(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Children's Library | Biodiversity Heritage Library | Additional Collections
Search: Advanced Search
Anonymous User (login or join us)
Upload
See other formats

Full text of "Annual report of the director of the Geophysical Laboratory"

■ 



CARNEGIE 
INSTITUTION 



Annual Report of the Director 
Geophysical Laboratory 

5251 BROAD BRANCH ROAD, NORTHWEST, WASHINGTON, D.C. 20015-1305 

1990-1991 



For the year July 1, 1990-June 30, 1991 

Issued December 1991 

Papers from the Geophysical Laboratory 

Carnegie Institution of Washington 

NO. 2250 



Digitized by the Internet Archive 

in 2012 with funding from 

LYRASIS Members and Sloan Foundation 



http://www.archive.org/details/annualreportofd199091carn 



Geophysical Laboratory 



Washington, District of Columbia 



Charles T. Prewitt 
Director 



Published by: Geophysical Laboratory 

5251 Broad Branch Rd., N.W. 
Washington, D.C., 20015-1305 
USA 



ISSN 0576-792X 
December 1991 



When used in bibliographic citations, The Annual Report should be cited as follows: 

Author, Title, Annu. Rep. Director Geophys. Lab., Carnegie Instn. Washington, 1990-1991, pagina- 
tion, 1991. 



GEOPHYSICAL LABORATORY 



Contents 



Introduction 1 

Igneous and Metamorphic Petrology — - 
Field Studies 3 

Global Convection and Hawaiian Upper Mantle 

Structures. T.Neil Irvine 3 

Megacrystalline Dunites and Peridotites: Hosts 

for Siberian Diamonds. N. P. Pokhilenko, D. 

G. Pearson , F. R. Boyd, 

andN.V.Sobolev 11 

Mantle Metasomatism: Evidence from a MARID 

- Harzburgite Compound Xenolith. 

F.R.Boyd 18 

Boron and Beryllium Concentrations in Subduc- 

tion-Related Metamorphic Rocks of the 

Catalina Schist: Implications for Subduction- 

Zone Recycling. Gray E. Bebout, Jeffrey G. 

Ryan, and William P. Leeman 23 

Laser Fluorination of Sulfide Minerals with F2 

Gas. D. Rumble, J. M. Palin, 

andT. C. Hoering 30 

Stable Isotope and Trace Element Indicators of 

Devolatilization History in Metashales and 

Metasandstones. Gray E. Bebout 34 

The Fa Content of Normative ol. 

Felix Chayes 40 

Igneous and Metamorphic Petrology — 
Experimental Studies 45 

Raman Spectra of High-Temperature Silicate 
Melts: Na20-Si02, K20-SiC>2, andLi 2 0-SiC>2 
Binary Compositions. John D. Frantz and 
Bjorn O. Mysen 45 

Peralkalinity and H2O Solubility Mechanisms in 
Silicate melts, Bjorn Mysen 53 

Partitioning of fluorine and chlorine between apa- 
tite and non-silicate fluids at high pressure and 
temperature. James Brenan 61 

Investigation of Fluid Immiscibility in the System 
H20-NaCl-CC>2 Using Mass Spectrometry and 
Microthermometry Techniques Applied to 
Synthetic Fluid Inclusions. 
Robert K. Popp,John D. Frantz, 
andThomas C. Hoering 68 



Akermanite-Gehlenite-Sodium Melilite Join at 



950 C and 5 kbar in the Presence of CO2 + 

H2O. H.G. Huckenholz, H.S. Yoder, Jr., T. 

Kunzmann, and W. Seiberl 75 

Merwinite Stability and High-Temperature Phase 

Relations in the Presence of CO2 +H2O. H. G. 

Huckenholz, H. S. Yoder, Jr., 

and W. Seiberl 81 

The System Mg2Si04-Fe2SiC>4 at Low Pressure. 

Hiroko Nagahara, Ikuo Kushiro, 

andBjorn O. Mysen 88 

Fe3+, Mg order-disorder in heated MgFe204: a 

powder xrd and 57pe mossbauer study. H. St. 

C. O'Neill, H. Annersten andD. Virgo ....93 

Crystallography - Mineral Physics 101 

Predicted High-Pressure Mineral Structures with 
Octahedral Silicon. Robert M. Hazen and Larry 
W. Finger 101 

Simultanous High P-T Diffraction Measurements 
of (Fe,Mg)Si03-Perovskite and (Fe,Mg)0 
Magnesiowiistite: Implications for Lower 
Mantle Composition. YingweiFei, Ho-Kwang 
Mao, Russell J. Hemley, and Jinfu Shu ..107 

High-Pressure Crystal Chemistry of Iron-Free 
Wadsleyite, p-Mg2SiC>4 Jinmin Zhang, Rob- 
ert M. Hazen, and Jaidong Ko 115 

Phase Transitions in Framework Minerals. 
David Palmer 120 

First-prirciples Studies of Elasticity and Post- 
Stishovite Phase Transitions in Si02. 
Ronald E. Cohen 126 

Molecular Dynamics Simulations of Melting of 
MgO at High Pressures. Zhaoxin Gong, 
Ronald E. Cohen, and Larry L. Boyer ... 129 

Glass Diffraction Measurements with Polychro- 
matic Synchrotron Radiation. Charles Meade 
and Russell J . Hemley 135 

X-Ray Diffraction of Solid Nitrogen-Helium Mix- 
tures. Willem L. Vos, Larry W. Finger, 
Russell J. Hemley, Ho-Kwang Mao, 
Jing Zhu Hu, Jin Fu Shu, Richard LeSar, 
Andre de Kuijper, 
and Jan A. Schouten 138 



CARNEGIE INSTITUTION 



Evidence for Orientational Ordering of Solid Deu- 
terium at High Pressures. Russell J. Hemley 
and Ho-Kwang Mao 141 

BlOGEOCHEMlSTRY 147 

Nitrogen Isotope Tracers of Atmospheric Deposi- 
tion in Coastal Shelf Waters off North Caro- 
lina. 
Marilyn L. Fogel and Hans W. Paerl 147 

Nitrogen Diagenesis in Anoxic Marine Sediments: 
Isotope Effects. David J. Velinsky, David J. 
Burdige, and Marilyn L. Fogel 154 

The Isotopic Ecology of Plants and Animals in 
Amboseli National Park, Kenya. PaulL. Koch, 
Anna K. Behrensmeyer, 
and Marilyn L. Fogel 163 



Rapid Racemization of Aspartic Acid in Mollusk 
and Ostrich Eggshells: A New Method for 
Dating on a Decadal Time Scale. Glenn A. 
Goodfriend, David W. von Endt, 
and P.E. Hare 172 

A Burning Question: Differences between Labo- 
ratory-Induced and Natural Diagenesis in Os- 
trich Eggshell Proteins. A S. Brooks, P.E. Hare, 
J.E. Kokis and K. Durana 176 

Publications 181 

Personnel 185 



GEOPHYSICAL LABORATORY 



Introduction 



Last year's introduction to the Annual 
Report described the co-location of the 
Geophysical Laboratory and the Depart- 
ment of Terrestrial Magnetism in our new 
and newly-renovated building complex on 
Broad Branch Road. This year the con- 
struction and renovation were completed 
and both departments are pursuing their 
normal research objectives with almost all 
of our equipment operating as it should. 
Moving and the relocation of laboratories 
has been disruptive for many staff mem- 
bers, but we all believe that the new envi- 
ronment and proximity to colleagues at 
DTM are worth the effort. 

The principal new research initiative 
for the Geophysical Laboratory this year is 
our participation in the new Center for 
High Pressure Research with the State 
University of New York at Stony Brook 
and Princeton University. This is one of the 
14 Centers established in 1991 through 
funding by the NSF Science and Technol- 
ogy Center Program. Depending on Con- 
gressional budgetary approvals, NSF in- 
tends to continue the Program for at least 
four more years, and it is possible that 
funding could extend for a total of eleven 
years. The funds supplied by NSF together 
with contributions from the three institu- 
tions and additional external grants will be 
used to support a variety of initiatives re- 
lated to high-pressure research. 

The Center is composed of staff, students, 
and laboratories at the Geophysical Labo- 
ratory, the Department of Earth and Space 
Sciences at Stony Brook, and the Depart- 



ment of Geological and Geophysical Sci- 
ences at Princeton. GL staff members 
involved with the Center are Francis Boyd, 
Ronald Cohen. Larry Finger, Robert Hazen, 
Ho-kwang Mao, Russell Hemley, Bjom 
Mysen, Charles Prewitt, and David Virgo. 
In addition to the above institutions, col- 
laboration is being developed between the 
Center and other laboratories in universi- 
ties, industry, and government. The princi- 
pal goal of the Center is to study fundamen- 
tal questions about the Earth's evolution, 
structure, and dynamic state, and about the 
nature of interiors of other planets. In addi- 
tion, we expect to generate extensive new 
information about material properties at 
high pressures and temperatures, and to 
synthesize new materials of interest to phys- 
ics, chemistry, and materials science as 
well as to the earth sciences. Experimental 
work will be complemented by theoretical 
computer simulations and by development 
of new equipment and techniques for high- 
pressure research, including larger-volume 
experimental apparatus. 

Systematic high-pressure work on ma- 
terials of geological interest has been a 
fundamental component of Geophysical 
Laboratory activities since 1904 and GL 
staff have played the major role in the 
development and utilization of many types 
of high-pressure apparatus, including pis- 
ton-cylinder, cold-seal, gas-media, and dia- 
mond-anvil devices. For example, in re- 
cent years Ho-kwang Mao and his col- 
leagues have led the development and ap- 
plication of the diamond-anvil cell for ex- 



CARNEGIE INSTITUTION 



perimental studies at high pressure. Static 
pressures of about 5.0 megabars — substan- 
tially greater than that at the Earth's cen- 
ter — have been attained, the important pres- 
sure-measuring scale using ruby fluores- 
cence has been extended to 1.8 megabars, 
and pressure-scale x-ray diffraction studies 
have been extended above 3 megabars. 
Techniques have been developed for heat- 
ing samples within the cell by laser and for 
studying them by means of a variety of 
spectroscopic techniques. The establish- 
ment of the Center will allow us to continue 
and extend this kind of innovation to greater 
extremes of pressure and temperature, larger 
sample volumes, and experiments on many 
different kinds of materials. 

In recent years, Mao, Russell Hemley, 
and colleagues have have concentrated 
much of their work on inert gases that 
crystallize into solid forms at high pres- 
sures. Their experiments with solid hydro- 
gen have exceeded 2.5 megabars, where 
they discovered new phase transitions and 
the first evidence of transformations into 
the metallic form. To support this research, 
a number of staff members and postdoctoral 
fellows have been active in developing and 
using x-ray and infrared beam lines at the 
National Synchrotron Light Source, 
Brookhaven National Laboratory, for ex- 
periments that could not be performed sat- 
isfactorily without the use of synchrotron 
radiation. In particular, the superconduct- 
ing wiggler beam line X17C provides a 
high-energy beam with very high intensity 
x-rays for probing tiny samples in dia- 
mond-anvil cells. The infrared beam line 
U4 provides high-intensity infrared radia- 
tion for spectroscopic measurements of 
samples in diamond-anvil cells, and is the 



only facility of its type anywhere in the 
World. 

Another new development this year is 
the installation of a Dilor micro-Raman 
system by John Frantz and Bj om My sen for 
examining the structures of silicate liquids 
at high temperature. Heretofore, most 
Raman studies of melts actually involved 
measurements on glasses quenched from 
high temperatures. Investigators were 
forced to assume that the glasses were 
representative of melts at temperature, but 
there were many doubts about the validity 
of this assumption. Now, Frantz and Mysen 
have made extensive recordings of Raman 
spectra on silicate melts at temperatures as 
high as 1600°C and it appears that they 
have opened up a new and exciting area of 
melt research. 

Douglas Rumble, Michael Palin, and 
Thomas Hoering have developed a method 
for laser fluorination of sulfide minerals 
with fluorine gas. This technique allows 
fast and precise in-situ micro-sampling on 
three of the four sulfur isotopes, 32 S, 33 S, 
and 34 S, and will provide information on 
mass transfer in sulfide system on a scale 
that was previously inaccessible. 

In addition to these initiatives related to 
a new NSF Center and to new instrumenta- 
tion, staff members, postdoctoral fellows 
and associates, and visiting investigators 
have been involved in a wide range of 
research, ranging from racemization dat- 
ing of ostrich shells to global convection of 
the upper mantle. Much of this research is 
described in brief summaries in this Annual 
Report, thus continuing the Geophysical 
Laboratory tradition of early communica- 
tion of results before full papers are pub- 
lished in scientific journals. 



GEOPHYSICAL LABORATORY 



Igneous and Metamorphic Petrology 

Field studies 



Global Convection and Hawaiian Upper 
Mantle Structure 



T. Neil Irvine 

This report gives further development 
of the global convection system proposed 
by Irvine (1989). In this system, upper 
mantle convection is stratified at both the 
400- and the 670-km seismic 
discontinuities, and the lower mantle fea- 
tures an orthogonal framework of six prin- 
cipal convection centers, or axes where 
upwelling occurs beneath Iceland, Hawaii, 
the Balleny Islands (near Antarctica), and 
the Okavango delta (in Botswana), and 
downwelling beneath Peru and the eastern 
edge of Vietnam. In last year's Report 
(Irvine, Annual Report 1989-1990, p. 3- 
1 1), the concept of an upper mantle vortex 
supercell was introduced for the Iceland 
center and explored on the basis of seismic 
data. Application of this concept is now 
extended to the Hawaiian center. 



Mantle Tomography and Hotspots 

When the above global-convection sys- 
tem was proposed, some support was cited 
from seismic tomography, but the 
tomographic maps then available portrayed 
only broad features and frequently seemed 
incompatible. More recent maps are not 



completely consistent either (e.g., see com- 
parisons made by Romano wicz, 1 99 1 ), but 
a set by Inoue et al. (1990) is especially 
interesting because the maps are unusually 
detailed. Much of the detail correlates 
meaningfully with surface geological fea- 
tures (particularly volcanic hotspots), so 
the data appear significant. Some features 
are notably compatible with the mantle 
convection relationships favored here. 1 

The Inoue et al. (1990) maps for nomi- 
nal depths of 478-629 km and 1203-1435 
km are illustrated in Fig. 1 , together with a 
map by Woodhouse and Dzie wonski ( 1 989) 
for 150 km. The relations for 1203-1435 
km warrant particular attention because ( 1 ) 
they are dominated by two major zones of 
low velocity, one under Hawaii, the other 
beneath Okavango, and (2) the extensions 
of these two zones in combination with 
several small, low-velocity anomalies, en- 
compass most of the world's hotspots. The 
two major anomalies are both prominent in 
several earlier tomographic maps (e.g., 
Giardini et al, 1987), but the velocity data 
in these cases were smoothed to low-order 
spherical harmonics, so the anomalies are 
only broadly delimited. The anomaly loca- 

1 In interpretations of mantle tomography here it 
is assumed simply that seismically slow regions 
are relatively warm and, thus, may represent zones 
of convective upwelling, whereas fast regions are 
cooler and, therefore, may reflect downwelling. 
Compositional and phase differences and seismic 
anisotropy are likely to be complicating factors 
but cannot be considered here. 



CARNEGIE INSTITUTION 




150 km 

Woodhouse and Dziewonski, 1989 
Model U84L85/SH Z =1-8 





+ + + + 
+ + + + 
+ + + + 
+ + + + 








% 


§ 





-3% 







3% 




478-629km 

Inoue et al.,1990 



::::: + 


+ + + + 
+ + + + 
+ + + + 
+ + + + 






'//// 
VK 


± 


± 


± 


X 


i 







2% 



GEOPHYSICAL LABORATORY 



tions on the Inoue et al. maps appear much 
better resolved. 

Neither the Iceland nor the Balleny cen- 
ter is seismically slow at 1203-1435 km, 
but Iceland is flanked by small low-veloc- 
ity anomalies on the northwest and south- 
east. Also, the Hawaii and Okavango 
anomalies are beltlike at this depth and tend 
to follow the great circle that includes Ice- 
land and Balleny. In combination, the two 
belts span almost half the Earth's circum- 



ference. (A rather similar anomaly arrange- 
ment is also indicated for 478-629 km in 
Fig. IB.) 

Several small anomalies in Fig. 1C do 
not correlate with hotspots, and a few 
hotspots are not associated with low ve- 
locities (notably Tristan da Cunha, Fernando 
de Norhana, and Martin Vas, all in a South 
Atlantic region where relatively high ve- 
locities are indicated); but at the present 
state of the science, it seems more signifi- 



1203-1435 km 
Inoue et al., 1990 




Volcanic hotspots 



FIG. 1. Global tomography maps for nominal depths of (A) 150 km, redrawn from Woodhouse and 
Dziewonski (1989) and (B and C) 478-629 km and 1203-1435 km, redrawn from Inoue etal. (1990). The 
maps also show the global convection framework of Irvine (1 989). 

The principal correlations between low-velocity anomalies and hotspots in Fig. 1C are as follows. The 
Hawaiian anomaly spreads southward beneath the Samoa, Marquesas, Tahiti, and Austral McDonald 
hotspots and comes close to Caroline (northeast of New Guinea). A strong small anomaly underlies the 
southeastern Australia hotspot, and weaker ones match the Galapagos and San Felix (south of Peru), and 
possibly Yellowstone. In the North Atlantic, a medium-sized anomaly is surrounded by the Azores, 
Canary Islands, Madeira, Cape Verde Islands, and New England hotspots; and in the South Atlantic, a 
strong small anomaly underlies the Cameroon hotspot, and a weak one matches St. Helena. Where the 
Okavango anomaly extends into northern Africa, it and a small satellite are associated with four 
continental hotspots; to the east and south, the Okavango anomaly encompasses the Comores, Reunion, 
Marion Island, Crozet Islands, and Kerguelen hotspots. 



CARNEGIE INSTITUTION 



cant that many hotspots are matched. A 
currently popular concept is that hotspots 
are initiated by mantle plumes originating 
in the D" seismic zone just above the core- 
mantle boundary, and then continue to be 
fed from this zone by thin, stemlike chan- 
nels of up welling (e.g., Olson, 1990). Im- 
pressive geoid and tomographic evidence 
for a source region at the core -mantle bound- 
ary has been available for some years (e.g., 
Chase, 1979; Crough and Jurdy, 1980; 
Woodhouse and Dziewonski, 1989), but 
the map in Fig. 1C contains what may well 
be the first discernible geophysical indica- 
tions of feeder stems at intermediate depths 
in the lower mantle. 



Stratified Mantle Convection 

A prominent feature of the three maps in 
Fig. 1 is that they are all very different; in 
fact, a general observation from seismic 
tomography is that the three mantle divi- 
sions delimited by the 400- and 670-km 
seismic discontinuities tend to exhibit con- 
trasting relations. Although rarely cited, 
this observation would seem a rather strong 
argument in favor of at least some stratified 
convection. 

In the mantle convection relationships 
proposed by Irvine (1989), it was assumed 
that the 400-km interface is everywhere 
mechanically coupled, and a combination 
of thermal and mechanical coupling rela- 
tionships was then devised for the 670-km 
interface, designed to account for the gen- 
eral tectonic features of the Earth's sur- 
face. An unorthodox feature of the result- 
ing arrangement is that, beneath the ob- 



served zones of upwelling along the mid- 
ocean ridges, there are zones of 
downwelling in the depth interval 400-670 
km. Some seismic evidence for this possi- 
bility was cited from the tomographic maps 
of Nataf et al. (1986), but it was acknowl- 
edged to be equivocal. The relationships of 
maps A and B of Fig. 1 suggest stronger 
evidence for this possibility, although this 
evidence too is not beyond question. In 
particular, the maps indicate that, while the 
regions beneath the mid-ocean ridges at 
150 km (map A) are generally seismically 
slow (as expected), those at 478-629 km 
(map B) commonly exhibit relatively high 
velocities. This contrast is especially con- 
spicuous along the East Pacific Rise, where 
it is most relevant to the Hawaiian convec- 
tion relationships. 

But, as explained by Inoue et al. ( 1 990), 
their map for 478-629 km does not have as 
high resolution as that for 1203-1425 km, 
and they specifically stated that the East 
Pacific Rise anomaly in the former is not 
reliable. This does not mean that the 
anomaly is invalid, however. Inoue et al. 
noted also that an increase of seismic ve- 
locity at 400 km beneath the East Pacific 
Rise had been observed by Suetsugu and 
Nakanishi ( 1 987) in a Rayleigh wave study; 
and maps by Dziewonski and Woodhouse 
(1987) of S-wave velocity variations at the 
670-km discontinuity similarly show high 
velocities beneath the Rise south of the 
Galapagos. Thus, the possibility of sub- 
ridge downwelling at 400-670 km, although 
still wanting of strong support, is still con- 
sidered viable. 



GEOPHYSICAL LABORATORY 



7 



Hawaiian Relationships and Supercell 
Structure 

It is well established that the islands and 
seamounts of the Hawaiian and Emperor 
volcanic chains become progressively older 
to the west and north from Hawaii. The 
widely accepted explanation is that the 
volcanoes formed in succession from a 
relatively fixed mantle hotspot, currently 
located beneath the Big Island, as the Pa- 
cific plate drifted first northward, then 
westward across it (cf. Clague and 
Dalrymple, 1987). The recent volcanism 
on Hawaii has been dominated by the tho- 
leiitic eruptions of Kilauea andMauna Loa, 
but the newest activity features eruptions 
of alkalic basalt from the submarine vol- 
cano Loihi on the south edge of the island. 
Along the older parts of the volcanic sys- 
tem, the seamounts at the Hawaiian-Em- 
peror bend are 42-43 Ma in age; those at the 
north end of the Emperor chain are 73-75 
Ma. 

In 1972, Jackson et al. pointed out that 
the Hawaiian-Emperor volcanoes tend to 
be paired, and portrayed them as being 
distributed along an en echelon (discon- 
tinuous) series of sigmoidal double lines. 
Recently, Garcia et al. (1990) identified a 
newly discovered submarine volcano 
(named Makuhona) just west of Hawaii to 
be the previously missing partner of the 
volcano Kohala on the northern peninsula 
of the island. With these observations as 
background, I have attempted to pair the 
volcanic structures of the entire system into 
a set of more continuous double lines, as 
shown in Fig. 2. En echelon overlap was 
required through the Midway Islands, but 



otherwise only two lines were necessary. 
Inasmuch as the only control was small- 
scale topography, the pairing is frequently 
conjectural, but in an overall count, per- 
haps 60 of some 70 possible pairs appear 
credible. Thus, given that there are prob- 
ably still other unrecognized eruption cen- 
ters of the Makuhona type, the more con- 
tinuous double lines seem realistic. The 
contention is that they reflect the continu- 
ous operation of the mantle convection 
supercell outlined in parts B and C of Fig. 
2 and explained below. 

Figure 2 also shows the ocean-floor 
troughs and arches that are associated with 
the volcanic chains. The main Hawaiian 
trough (here termed the "inner trough") 
encircles the south side of the Big Island 
and extends discontinuously westward on 
both sides of the island chain-, with widths 
locally exceeding 100 km. This trough has 
long been ascribed to loading of the ocean 
floor by the volcanic ridge (e.g., Moore, 
1987). Outboard from it are broad arches, 
or swells, 300-400 km wide. They are 
topographically prominent only along the 
younger half of the Hawaiian chain, where 
their relief exceeds 1 000 m and brings them 
to depths less than 5000 m (Fig. 2); but they 
can be readily identified throughout, even 
along the Emperor chain, by their distinc- 
tive positive gravity signatures (see Haxby, 
1987). Finally, fringing the swells around 
the younger part of the Hawaiian chain is a 
subtle "outer trough." This trough has 
special importance in the present context 
(see below). 

Figure 3 depicts an Hawaiian upper 
mantle convection supercell designed to 
account for the paired volcano chains and 



8 



CARNEGIE INSTITUTION 




FIG. 2. Maps of the Hawaiian and Emperor island and seamount chains in which the volcanic structures 
are paired into two lines (four lines through the Midway Islands). Based on the topography map of 
Mammerickx (1989), the ocean gravity map of Haxby (1987), and maps of earthquake epicenters and 
young volcanoes from Moore (1987) and Garcia etal. (1990). See text for discussion of the upper mantle 
convection supercell. 



GEOPHYSICAL LABORATORY 



WNW 



EAST PACIFIC RISE 



Mechanical coupling 




Tholeiitic shield volcano 
Early alkalic lavas 
nner Trough 
Outer trough 

KM 



PLAN VIEWS 
Vortex core flow 

© up 

down 



11111*11 in i 400 



FIG. 3. Schematic three-dimensional representation of the Hawaiian upper mantle vortex supercell. See 
text for description. 



related features. 2 A principal feature of the 
interpretation is the differential lateral flow 
of the upper mantle layers. As with the 
supercell proposed last year for Iceland, 
the vorticity in the supercell is ascribed to 
thermal tilting of this flow by heat rising 
from the underlying lower mantle axis of 
convective up welling (cf., Irvine, 1990, 

2 Garcia et al. (1990) noted, on the basis of an 
experimental study of mantle plume dynamics by 
Richards and Griffiths (1989), that a convection 
structure of this general type might account for the 
pairing of the Hawaiian volcanoes; and Sleep 
(1990) used a "stagnation streamline" model akin 
to part of the structure in Fig. 3 to estimate buoy- 
ancy flux values for the Hawaiian and other mantle 
plumes. 



Fig. 1). The differential flow also leads to 
the downwelling at 400-670 km under the 
East Pacific Rise, as discussed in relation to 
Fig. 2B. 

In the envisaged action of the supercell, 
a swath of the lower part of the lithosphere 
is stripped away (delaminated) and replaced 
by asthenospheric material upwelling from 
near the 400-km interface. The stripping 
process depresses the outer ocean-floor 
trough, and the buoyancy of the replace- 
ment material elevates the topographic 
swells. The swells eventually subside as 
the replacement material gradually cools 
and contracts, but through its increased 



10 



CARNEGIE INSTITUTION 



density, they continue to have their distinc- 
tive gravity signatures. Further postulates 
are (1) that the double -vortex circulation 
holds the supercell in place as a standing 
vortex against the laterally flowing mantle 
layers (see Irvine, 1990, Fig. 1), (2) that 
magma melting occurs mainly through the 
adiabatic rise of fertile peridotite along the 
two vortex axes of upwelling, and (3) that 
these two axes also deliver the magma to 
paired release points at the base of the 
lithosphere, from where it rises to the paired 
volcanoes at the surface. 

The supercell structure can also be used 
to rationalize much of the general history 
of individual Hawaiian volcanoes (for sum- 
mary, see Clague and Dalrymple, 1987). 
By concept, the magma from the main 
release points, having melted at relatively 
high pressures, is picritic tholeiite in com- 
position, and its eruption produces large 
shield volcanoes like Mauna Loa and 
Kilauea. But it can be postulated too that 
some alkalic magma should form in the 
surrounding, spreading parts of the vortex 
cells by differentiation of trapped intersti- 
tial melt at the more moderate pressures of 
this environment. This magma could be 
released both in advance of the tholeiitic 
shields, as at Loihi, and in arrears of them, 
as in the "alkalic post-shield stage" ob- 
served in volcanoes such as Mauna Kea 
and Hualalai. I suggest too that, through its 
buoyancy, this alkalic magma may tend to 
accumulate in substantial quantities at the 
tops of the inferred vortex axes of 
downwelling, from where it might be 
erupted at a considerably later stage in the 
history of a volcano if the volcano hap- 
pened to pass across such an axis by virtue 



of the plate motion. This kind of late 
eruption could correspond to the "alkalic 
rejuvenated stage" that is observed in most 
of the Hawaiian volcanoes between Maui 
and Necker Island. 

A fundamental general tenet in this 
analysis is that magma from the mantle is 
primarily released to the lithosphere from 
specific, critical flow points in the 
asthenospheric convection system. 



References 

Chase, C. G., Subduction, the geoid, and lower 
mantle convection: Nature, 282, 462-468, 
1979. 

Clague, D. A., and G. B. Dalrymple, The Hawai- 
ian-Emperor volcanic chain: Part I. Geologic 
evolution, U. S. Geol. Surv. Prof. Paper 1350, 
5-54, 1987. 

Crough, S. T , and D. M. Jurdy, Subducted litho- 
sphere, hotspots, and the geoid, Earth Planet. 
Sci. Letters, 48, 15-22. 1980. 

Dziewonski, A. M., andWoodhouse, J. H., Global 
images of the Earth's interior, Science, 236, 
37-48, 1987. 

Garcia, M. O., M. D. Kurz, and D. W. Muenow, 
Mahukona: The missing Hawaiian volcano, 
Geology, 18, 1111-1114, 1990. 

Giardini, D., L. Xiang-Dong, and D. H. 
Woodhouse, Three-dimensional structure of 
the Earth from splitting in free-oscillation spec- 
tra, Nature, 325, 405-41 1, 1987. 

Haxby, W. F., Map of the gravity field of the 
world's oceans, Natl. Oceanic Atmos. Adm. 
Rpt. MGG-3, 1987. 

Inoue, H., Y. Fukao, K. Tanabe, and Y. Ogata, 
Whole mantle P- wave travel time tomography, 
Phys. Earth Planet. Interiors, 59, 294-328, 
1990. 

Irvine, T. N., A global convection framework: 
concepts of symmetry, stratification and sys- 
tem in the Earth's dynamic structure, Econ. 
Geol, 84, 2059-21 14, 1989. 

Jackson, E. D., E. I. Silver, and G. B. Dalrymple, 
Hawaiian-Emperor chain and its relation to 
Cenozoic Circum-Pacific tectonics, Geol. Soc. 
Amer. Bull., 83, 601-618, 1972. 

Mammerickx, J., Bathymetry of the North Pacific 
Ocean, Geol. Soc. Amer., The Geology of 
North America, N, 1989. 



GEOPHYSICAL LABORATORY 



11 



Moore, J. G., Subsidence of the Hawaiian Ridge, 
U. S. Geol. Surv. Prof. Paper 1350, 85-100, 
1987. 

Nataf, H.-C, I. Nakanishi, and D. L. Anderson, 
Measurements of mantle wave velocities and 
inversion for lateral heterogeneity and anisot- 
ropy: III, Inversion, /. Geophys. Res., 91, 
7261-7308, 1986. 

Olson, P., Hot spots, swells and mantle plumes, in 
M. P. Ryan, Magma Transport and Storage, 
New York, J. Wiley & Sons, 33-51, 1990. 

Romano wicz, B., Seismic tomography of the 
Earth's mantle, Ann. Rev. Earth Planet. Sci., 
79,77-99, 1991. 

Sleep, N., Hotspots and mantle plumes: some 
phenomenology, /. Geophys. Res., 95, 6715- 
6736, 1990. 

Suetsugu,D.,andI.Nakashini, Three-dimensional, 
velocity map of the upper mantle beneath the 
Pacific Ocean as determined from Rayleigh 
wave dispersion, Phys. Earth Planet. Interi- 
ors, 47, 205-229, 1987. 

Woodhouse, J. H., and A. M. Dziewonski, Seis- 
mic modelling of the Earth's large scale three- 
dimensional structure, Phil. Trans. R. Soc. 
Lond., A 328, 291-308, 1989. 

Richards, M. A., and R. W. Griffiths, Thermal 
entrainment by depleted mantle plumes, Na- 
ture, 342, 900-902, 1989. 



Megacrystalline Dunites and 

Peridotites: 
Hosts for Siberian Diamonds 

N. P. Pokhilenko* , D. G. Pearson ** , 
F. R. Boyd, andN. V. Sobolev* 

Several investigations have identified 
xenoliths, consisting primarily of 
ultracoarse crystals of olivine, which ap- 
pear to be fragments of the principal host 
rocks of Siberian diamonds (Sobolev, 1974; 
Pokhilenko et ai, 1977; Soboley et al., 
1984). Twenty -three of these xenoliths 

Inst, of Mineralogy & Petrology, Siberian Branch 
of the USSR Academy of Sciences, Novosibirsk 
Dept. of Terrestrial Magnetism, Carnegie Instn. 
of Washington, Washington, D. C. 20015. 



contain diamonds, and many more contain 
garnets with a compositional range very 
similar to the range for garnets included in 
diamonds. These megacrystalline rocks 
have been found as xenoliths in the 
kimberlites of the Daldyn-Alakit region 
and in some pipes of the Upper Muna 
group. They are especially abundant in the 
xenolith-rich Udachnaya kimberlite, site 
of one of the world's richest diamond mines. 

The Siberian megacrystalline rocks dif- 
fer from diamondiferous peridotites from 
southern Africa, the only other region in 
which a number of such xenoliths have 
been found. African occurrences are pri- 
marily lherzolites and harzburgites having 
garnets and associated minerals with com- 
positional ranges that depart significantly 
from the majority of diamond inclusions 
(Viljoen etal., 1991; Boyd etal., in prepa- 
ration). Thus the principal host rocks for 
African diamonds have not yet been dis- 
covered as articulated xenoliths. 

The relative abundances and composi- 
tional ranges of the olivine and associated 
minerals indicate that most of the Siberian 
megacrystalline rocks are extremely re- 
fractor)'. They may be residues or cumu- 
lates of melting events in which a large 
proportion of melt was removed. Olivine 
forms more than 90% of most specimens. 
The primary modal proportions are diffi- 
cult to estimate because of disaggregation 
during eruption, but it is likely that these 
ultracoarse rocks were olivine-rich. The 
molar Mg/(Mg+Fe) for the olivine is pre- 
dominantly 0.92-0.95, and most of the gar- 
nets are strongly subcalcic and Cr-rich (Fig. 
4). Mineral assemblages of the 33 new 



12 



CARNEGIE INSTITUTION 



Table 1. Mineral assemblages in megacrystalline 
olivine-rich xenoliths from the Udachnaya 
kimberlite, U.S.S.R. 



Assemblage 



No. specimens 



olv + gar 

olv + gar + chr 

olv + gar + diam 

olv + gar + cpx 

olv + gar + opx 

olv + chr 

olv + opx + chr 

olv + gar + chr + diam 

olv + gar + opx + cpx + chr 



13 

8 

4 

2 

2 

1 

1 

1 

1 



specimens chosen for this investigation 
have a predominance of olivine + garnet 
(Table 1). Five are diamond-bearing. 



Isotope Results 

Preliminary neodymium isotope analy- 
sis of two hand-picked, acid-washed garnet 
separates (Uv 70/76 and Uv 49/76) reveal 
that the megacrystalline rocks, having ex- 
perienced an ancient trace element enrich- 
ment event, show similar Nd isotopic sig- 
natures to peridotite suite garnet inclusions 
in diamond from southern Africa analyzed 
by Richardson et al. (1984). The two 
samples record £Nd values of -28 to -33 at 
350 Ma (the time of pipe emplacement), 
and yield model ages relative to the Bulk 
Earth reservoir (CHUR) of 2.0 to 2.7 Ga. 
(Fig. 5). Although the latter age is late 
Archaean, it is substantially younger than 
the 3.3 Ga. model ages obtained for the 
diamond inclusions by Richardson et al. 
(1984). The younger model ages and the 
age-range obtained from the two samples 
from Udachnaya may be attributed to the 
more "open system" behavior of the small, 
coarse-grained megacrystalline samples 



o 

CO 

O 



12 

10 

5 8 

6 

4 
2 



— ■ r 

. o1 






— r— 


■-> 1 ■ — 


- 1 ■ 1 *• 


1 1 

o ° . 


a2 














• 3 














. B4 








o 


•^^^s* 


0^ 


- . 


• 


• 




□ _ 
° o* 


° %° 

o • % 

o 


■ 
• 

• 111 








• 


3 o° 

• ■ ■ 
c 

1 


_1 . 1 . 


I 



2 4 6 8 10 

Cr 2 3 ,wt% 



12 



14 



Fig. 4. A plot of CaO against O2O3 for garnets 
from megaciystalline xenoliths of different asso- 
ciations: 1 - without chromite; 2 - with chromite; 
3 - diamond-bearing without chromite; 4 - dia- 
mond-bearing with chromite. Data from this 
study and Sobolev et al. (1984). 

(compared with the armored diamond in- 
clusions), which could have allowed the 
xenolith garnets to be subjected to later 
interaction with a lighter rare-earth-enriched 
component. Further analyses of a suite of 
larger garnets from the megacrystalline 
rocks are being undertaken to resolve this 
problem. However, the results obtained 
indicate that the base of both the Kaapvaal 



0.512 




1000 2000 3000 4000 

Time, Ma 

Fig. 5. Neodymium isotope evolution diagram for 
two-garnet separates from the Udachnaya 
megacrystalline. Vertical arrows indicate model 
ages relative to the Bulk Earth or CHUR evolution 
curve. Diamond symbols represent the isotopic 
composition of the samples at the time of pipe 
emplacement 350 Ma ago. 



GEOPHYSICAL LABORATORY 



13 



and Siberian cratons experienced similar 
incompatible element enrichment events 
very early in their evolution. 



Petrography 

The megacrysts are friable and are not 
easily separated intact from hard kimberlite. 
The largest has a long dimension of 19 cm, 
but most specimens are ovoidal with di- 
mensions of the order of 5 cm. Olivine 
crystals range up to 10 cm, and in a few 
specimens smaller grains of olivine (6-9 
mm) with differing optic orientation are 
interspersed. Prominent parting, resem- 
bling cleavage, is characteristic. Rounded 
garnets are much finer-grained, 0.2-6 mm, 
and form up to 5 modal percent. The 
predominant low-Ca garnets are lilac -purple 
but those that are rich in Ca as well as Cr are 
dark purple or gray or even green. Alter- 
ation of garnet to kelyphite is variable; 
some grains with minor kelyphite are 
euhedral with a rhombic dodecahedral 
morphology complicated by uneven devel- 
opment of faces. The distribution of gar- 
nets is relatively regular in the larger xeno- 
liths. 

Enstatite is present in about 20% of the 
megacrystalline rocks, but clinopyroxene 
is much less common. Octahedra of chro- 
mite, 0.1-2 mm, form less than 1 modal 
percent, and some of them appear to show 
signs of resorption. 

Crystals of diamond in these rocks are 
characteristically clear, shaip-edged octa- 
hedra, some with spinel twins, and range up 
to 4 mm. Colored diamonds and those that 
are corroded or cracked are unusual. Small 



plates of graphite (0.2-1.5 mm) have also 
been observed both with and without dia- 
mond. The abundances of diamond and 
graphite are usually insignificant but one 
olivine-pyrope specimen having a volume 
of only 2.5 cm 3 contains four diamonds 
together with cavities from which two ad- 
ditional diamonds have broken away. 



Mineral Chemistry 

Most olivines in the megacrystalline 
rocks have Mg/(Mg+Fe) in the range 0.92 
-0.95, but two olivine-rich wehrlites in 
which the garnets have both high Ca and 
high Cr (Fig. 4) contain substantially more 
Fe-rich olivines, 0.87 - 0.89 (Fig. 6). Sur- 
prisingly, a variation in NiO from 0.32 to 
0.40 does not correlate with mg number. 
Values for CaO and Cr203 in the olivines 
do not exceed 0.06 wt %. 

Garnets in the megacrystalline xeno- 
liths can be assigned to three parageneses 
on the basis of their CaO and Cr203 con- 
tents: proportions for more than 200 speci- 
mens thus far studied (this Report and 
Sobolev et al., 1984) are harzburgite-dun- 
ite (80%), lherzolite (15%), and wehrlite 
(5%). Concentrations of FeO andTi02 are 
least in the harzburgite-dunite group (Tables 
2 and 3). There is a positive correlation 
between Ti02 and CaO in these garnets, 
apparently reflecting degree of depletion 
(Fig. 6B). The chromites are predomi- 
nantly rich in Cr203 (58-65 wt %) but 
spinels with Cr203 as low as 21.9 wt % 
have been found. A good correlation in Cr/ 
(Cr + Al) between the garnets and chro- 
mites (Fig. 6C) is evidence of equilibra- 



14 



CARNEGIE INSTITUTION 



a 
• i— < 

*—> 

■a 

>* 

CX 
• i— i 

CX 



t-l 
X> 



OS 

>> 

a 

c 
xi 
o 
a 

X3 



O 

OO 

.C 

4— > 

• r-< 

c 
t> 

X 

<— > 

J— > 

o 

■c 

ex 

D 

c 

00 

p 

S3 

a 

> 

• I— I 

a 

•*— > 
c 
<u 

oo 
<U 

ex 
u 

t-c 

oo 
t> 
oo 
j>. 

13 
c 
ed 

<u 

c 
• i—i 



CN 

a 

£0 
< 



no 

oo 

o" 
no 

I 

> 
!D 



NO 
oo 

CN 

NO 

> 
ID 



NO 
OO 

in 

CN 

i 
> 

!D 



NO 
oo 

NO~ 

CN 

I 

> 

ED 



bfl 



J3 

o 



X 

a, 
o 



53 



CX 

o 



J3 

bJQ 



X 

Oh 

O 



> 
O 



o 



ton 



X 

ex 
o 



incococoTtoouoioco© 

mO'— i'— i on © *— t^^OO 



CN Tf CN no r- no 



o o o 



oo 



en 



TfOOO-HO^tOOO 



r-» OO OO OO On Os c- 



© CO O 



oo oo 



-a -a 



© © On »0 *— i © ^h © c c 



^••oiovp-'^ooNior^ 



m d ^' ri -h o 



mo .r 



© t— no no co on <— no -g 
-4 -1 vd ^ °i "1 od "I ^. H 



lOOOO^OcoOO c 



^©©CNCN^hNOCN-CN 

^•OOOvO-HqO^CO 

Tj-ooor^o'mo c © 



^-*h^tOOO>^ 






oo 



CN Tf ^t 



<N 



CN 



NO r- O CN <N o o 



8rHlOOmHOOON 
< CO T-H Tt T— < ^ r- I © <— J 

loddd^dc^ddd 



en 

od 



On 



O CO On CO O NO 



O Tt 



O O CO 



Ttooor^-ou-iooo 



NO NO NO NO CN On 
'-1 P <*"* -J 00* "* 

o o r- ^o ^ © 



d c d 



coeoio^iocN©^ 



^t O 



oo 



oo Tt >o 



CN 



NO 



NO r^ O CN CN o o 



— ' On ^t NO r- <0 CN 



oo 



O^MCS 



m 



cn o ^ 



mOOOrJ-OcOOO CS 



hOOOhh ^_; 

^t © © © r^ © uo 



Pd^ 
o o o 



KP<0£SSozz 



oo 9, 



C BO 



d cs ^5 



oo 

r^ 

od 

ON 
ON 

d 
o 



in 

od 

On 



CN 



On 
CN 



On 
O 



On 

O* 

On 



On 
NO 



CO 

On 
On 



On 
CO 
On 



NO 
CN 

On 
On 



OO 



On 
On 



CO 



4—1 

o 
H 



CN ^f On 
ih d u-i 

r- co io 



NO 



OO 

OO 



CO 



CO 



oo 



CN NO OO 

"3* no »n 

On tJ- tj- 



CO NO iO 
O od iri 
OO '-h CN 



r .NO 



CO 

On 



CN 

CN 

On 



On 
CN 



"> NO "^ 



5 lO 

OO V) 



oo 



rn -: Tt 

On ° CN 



lO 

CN 

On 



r- On 

»o OO 



^ OO 



oo 
On 



CO 

On 



CN 

On 



3 CN 



£3^ 

+ 2 <- 

Sou 



GEOPHYSICAL LABORATORY 



15 



•i— i 

*-> 

M 

Oh 
• i— < 



X) 



a 

o 

a 

.13 

■4— > 

£ 
o 

<J3 

+- » 

•H 

c 

<D 

X 

<U 
■*-> 

•4— > 

O 

•c 

<u 
ex 

00 

C 



X> 

I 

C 

o 

B 



c 



o 
aj 

oo 



c 
< 



< 



ON 

oo 
O 

i 
> 



On 

oo 

r- 



> 

S3 



vo 

oo 

io 

VO 

I 

> 



vo 
oo 

^? 

On 
co 

i 

> 

S3 



vo 
oo 

vS 1 

I 

> 

S3 



00 



53 



53 

00 



00 



I— * 

o 



J3 



J3 
oo 



OX) 



a^tMCOIOMOrH 



-JP 
^t o 



vo 



oo 



On «t 



<N 



r-~ vo o <n co o 



""t m tJ- tJ- Tf \© ^ ro (N 



CN 



lOrHCO 



co 



OO 



Tj-ocN-^tr-ocNcNO 



On O 



>-Hi>(sa\M vo 
Ohh jo co 



^ooor-oino 



covor-»ocNcNO'— i 



eN 



ON 



CO On "tf ^ CO O 



TtO-HiONOONmO 



»— l QO'— iCIr-Hf—CO 

H'popcjHjq 
Tj-ooor^o>oo 



ON 
CO 



Tj-Q<— lOOVOOOVOCN 
Tt©— ivOUOOcNOO 



(N OO O O CO 



O vo 



CO 



vo 

CO 



^ooo«oo»oo 



CN CO «o OO CO CO 



^j O o o r- o 



CN 



^tooovoomo 



O <N m cn oo t-«- oo 



CN *-< VO 



o »o 



CN 



OOMnOhQhO 



cN^tcNooT-Hr^.r«-»o^HQ 

(vj O oo CN CO CO ^ <N © © 

rtO'-nr-VOOCN'-HOO 



t--<Nr'»r--vo, r +<— *on 



cn 

tJ- o 



oo 



CO CN 

\6 vd 



^ (sqq 

<N ">« o O 



VO 



i— i CN CO 

OON 



VO 



lO 

CO 



^tooovoo»oo 



m en 

OOnnO _ 

co H < <J pu ,§ S <J £ £ 



oo §o 



a oo 



o 



CN 



<—> w p © CO i— < J p CO © 

^t o o o r^ o m o © 2 



CN 



OO 

© 
o 



oo 

ON 

On 
On 



VO 



CN 

ON 

ON 
ON 



CO 



O « 



>o 



vO 
CO 



o Jr; 



CO 

oo 

ON 



VO 
CO 

On" 
On 



VO 

OO 

ON 
ON 



On 
On 

ON 



c3 

o 



oo 



VO 
CN 

ON 



CO 

oo 



oo 

CN 

ON 



VO 
OO 



VO 

CN 

On 



oo 
oo 



ON 



CO 

CO 

On 



CO 
ON 

iO 



oo 



CO 

oo 



vq 

CO 

ON 



+ 
00 



§ ^fc 



16 



CARNEGIE INSTITUTION 



0.42 

0.40 

0.38h 

0.36 

0.34 



0.32 



' r • i 

■ a. olivine 


— i 1 1 1 r 


■ 


- 


oo • 


- 


■ 


a • a 


• 


. 


o ■ 


■ 


- 


OOXD ■ • 


- 


o 


• <»«■ 


■ 


- 


*m m 


" 


o 


m m 

_J U . 1 L 


■ 



86 88 90 92 94 96 

mg# 



id 


■ b. garnet 


— i ■ — 

8 


10 




; 


8 




- 




• 


o 


6 


o 
o o 


o 




■ o- o 




4 


• °- 


- 




. °o * o 


. 




• o 




2 


"a • 

i.i. 


1 



0.0 0.1 0.2 

Ti0 2 , wt % 



0.3 0.4 



5U 


■ C. 


T ' 


1 •" 


1 ' 


40 








• 
■ 


30 








■ "■ . 


20 








• 


10 


" • 


• 
1 ■ 


• 
1 i 





20 40 60 80 

Cr/(Cr+AI), %, chr 



100 



Fig. 6. Compositional variations in minerals of the 
megacrystalline rocks. Symbols as in Fig. 4. a) a 
plot of NiO against Mg number for olivines of 
megacrystalline rocks, b) a plot of CaO against 
Ti02 for garnets from megacrystalline rocks, c) a 
plot of Cr/(Cr + Al) ratio for garnets against Cr/(Cr 
+ Al) ratio for chromites from megacrystalline 
rocks. 




700 



30 40 50 60 

Pressure, kbar 



70 



Fig. 7. Temperature-Pressure estimates for four 
xenoliths of megacrystalline rocks from 
Udachnaya pipe having coexisting pyrope and 
enstatite. Temperature was calculated with the 
olivine-garnet thermometer (O'Neill and Wood, 
1979) and pressure from the isopleths of 
MacGregor(1974). 

tion. Enstatites have low AI2O3 (Table 2), 
reflecting relatively low temperatures and 
high pressures of crystallization. 



Thermobawmetry 

The temperature of equilibration of a 
xenolith can be estimated from the parti- 
tion of Fe and Mg between olivine and 
garnet (O'Neill and Wood, 1979), but 
enstatite must also be present to obtain an 
estimate of the pressure or depth of equili- 
bration. Four of the megacrystalline rocks 
contain enstatite as well as olivine and 
garnet and<F-r estimates for them (Fig. 7) 
plot close to or within the diamond stabil- 
ity field. 

Puzzling features of the P-Tplot in Fig. 
7, however, are that the estimated points for 
megacrystalline rocks show a wide disper- 
sion in pressure and that they plot in a 
temperature range below the shield 
geotherm (40 mWm^) and below estimates 



GEOPHYSICAL LABORATORY 



17 



made for garnet lherzolite xenoliths from 
Udachnaya (Boyd, 1984). One of the 
megacrystalline rocks contains diopside as 
well as olivine and garnet (Uv-624/86, Table 
2), and its temperature can be estimated 
with the diopside solvus, giving 800°C, as 
well as with the Gar/Olv thermometer, 
which gives 840°C. The approximate agree- 
ment suggests that the discrepancy with the 
lherzolite data may not be a failure of 
thermobarometry. The dispersion in pres- 
sure for the megacrystalline rocks could be 
evidence of a wide range in depth of equili- 
bration. More data are clearly needed, 
however, before these questions can be 
properly addressed. 

Experimental studies of coexisting solid 
solutions of spinel and garnet show that the 
solubility of the knorringite component 
(Mg3Cr2Si30i2) in pyrope increases with 
increasing temperature and pressure 
(Malinovsky andDoroshev, 1975). Pyrope 
garnet that is rich in Cr can coexist with 



o 

03 
O 




4 6 8 10 

O2O3 ,Wt.% 



12 14 



Fig. 8. A plot of CaO against Q2O3 with isopleths 
of knorringite component in garnet estimated from 
experimental results (Malinovsky and Doroshev, 
1975). Fields for graphite and diamond are calcu- 
lated from these experimental data and the esti- 
mated temperature and pressure of the diamond- 
graphite transition in the Siberian mantle. 



chromite only at pressures within the dia- 
mond stability field (Kesson and Ringwood, 
1989). The experimental data can be used 
to estimate the compositions of garnets in 
equilibrium with Cr-spinel at P-T condi- 
tions corresponding to the diamond-graph- 
ite transition in the Siberian lithosphere 
(Fig. 8). The compositions of garnets in 
megacrystalline rocks that contain diamond 
and chromite are consistent with the field 
for diamond shown in Fig. 8 and with the 
knorringite isopleths based on experiment. 
Analyses for garnets in graphite + garnet + 
chromite assemblages (not plotted in Fig. 
8) all have less than 15% knorringite and 
are thus also consistent (Pokhilenko et al., 
1988). Thus, it may be possible to develop 
a useful barometer based on the Ca and Cr 
contents of the garnets for assemblages that 
include both pyrope and chromite. 



References 

Kesson, S. E., and A. E. Ringwood, Slab-mantle 
interactions 2. The formation of diamonds, 
Chem. Geol., 78, 97-118, 1989. 

Malinovsky, I. Yu., and A. M. Doroshev, Stability 
of garnet of pyrope-knorringite row in the field 
7=1000-1500°C and P=20-50 kb, Nauka. 
Novosibirsk, p. 23-31, 1975. 

O'Neill, H. St. C, and B. J. Wood, An experimen- 
tal study of Fe-Mg partitioning between garnet 
and olivine and its calibration as a 
geothermometer, Contrib. Mineral. Petrol. 70, 
59-70, 1979. 

Pokhilenko, N. P., A. S. Rodionov, T. M. Blinchik, 
and E. V. Malygina, Graphite-diamond phase 
transition and its significance for estimations of 
P-T conditions of equilibrium of ultrabasic 
xenoliths, Ext. Abstr. vol. Intern. Symposium 
on Composition and Processes of Deep-seated 
Zones of Continental Lithosphere, pp. 64-65, 
Novosibirsk, 1988. 

Pokhilenko, N. P., N. V. Sobolev, and Yu. G. 
Lavrent'ev, Xenoliths of diamondiferous ultra- 
mafic rocks from Yakutian kimberlites, 2nd 
Int. Kimb. Conf. Ext. Abstr. Vol. Santa Fe, 
1977. 



18 



CARNEGIE INSTITUTION 



Richardson, S. H., J. J. Gurney, A. J. Erlank, and 
J. W. Harris, Origin of diamonds in old en- 
riched mantle, Nature, 310, 198-202, 1984. 

Sobolev, N. V., Deep Seated Inclusions in 
Kimberlites and Problem of Upper Mantle 
Composition, p. 1-264 Nauka, Novosibirsk, 
1974 (English Translation), AGU, 1977. 

Sobolev, N. V., N. P. Pokhilenko, and E. S. 
Yefimova, Diamond-bearing peridotite xeno- 
liths in kimberlites and the problem of the 
origin of diamonds, Sov. Geol. Geophys., 25, 
62-76, 1984. 

Viljoen, K. S., D. H. Robinson, and P. M. Swash, 
Diamond and graphite peridotite xenoliths from 
the Roberts Victor Mine, 5th Intern. Kimb. 
Conf. Araxa, Brazil, Ext. Abstr. Vol., 1991. 



Mantle Metasomatism: Evidence from a 

MARIE) - Harzburgite Compound 

Xenolith 

F. R. Boyd 

Interpretation of metasomatized rocks 
in the Earth's crust depends critically on the 
study of outcrops. We are not fortunate, 
however, in having exposures of the deep 
portions of continental cratons which have 
been the sites of metasomatic events over 
the course of 3-4 billion years. In the 
absence of outcrops, the rare discoveries of 
xenoliths exhibiting contacts between rock 
types have particular importance. This is 
especially true if the rocks are fragments of 
an igneous intrusion and a metasomatized 
conduit wall. 

A xenolith containing a contact between 
a mica-rich igneous cumulate and a 
metasomatized peridotite was discovered 
and made available for study by staff of the 
Anglo Axnerican Research Laboratories. 
The xenolith was recovered from the com- 
bined coarse concentrate of the De Beers 
mines in Kimberley. 

Mica-rich xenoliths from the Kimberley 



pipes include varieties of both igneous and 
metasomatic origin. Those believed to be 
igneous have been designated by the acro- 
nym MARID, based on the names of the 
constituent primary minerals: mica, am- 
phibole, rutile, ilmenite, and diopside 
(Dawson and Smith, 1977). Metasomatized 
peridotites contain a variety of introduced 
phases that include phlogopite, potassic 
richterite and pargasite amphiboles, il- 
menite, rutile, and a number of exotic po- 
tassium and barium titanates (Erlank et al., 
1987). It has been a problem to understand 
the nature of the metasomatizing fluids and 
whether or not they are related to kimberlite. 
Is kimberlite the magma from which 
MARID rocks are derived and are MARID 
rocks the remnants of metasomatic sources? 
Only one xenolith containing a contact 
between a MARID rock and a peridotite 
has previously been described (Waters et 
al., 1989). This xenolith was interpreted as 
a fragment of an intrusive with attached 
wall rock and study of it has helped to 
establish the hypothesis that MARID rocks 
are remnants of sources of metasomatic 
fluids. Extensive post-metasomatic alter- 
ation of the peridotite to carbonate in this 
xenolith, however, has obscured the pri- 
mary mineralogy and the metasomatic im- 
print. 



Petrography 

The sawed fragment of the compound 
xenolith analyzed in the present study is 6 
cm in maximum dimension and 2 cm thick; 
the fragment is numbered FRB 1455. The 
MARID portion of the xenolith consists of 



GEOPHYSICAL LABORATORY 



19 



over 95 % phlogopite, predominantly in the 
form of tablets, 1-2 mm in length, aligned 
parallel to the contact with the peridotite. 
The coarse mica is strained, having undu- 
late extinction and bent cleavage. Finer- 
grained mica neoblasts are interspersed 
between the coarse tablets. The coarse 
mica is pale, but the neoblasts and mantles 
on the mica tablets have a darker, yellow- 
ish-brown color. 

Granules of diopside a few tenths of a 
mm in diameter are dispersed between the 
mica tablets and clustered with serpentine 
in lenticles. Elongate grains of rutile rang- 
ing up to a mm are less abundant. Small 
patches of calcite that may be primary are 
widely dispersed and enclose euhedral crys- 
tals of sphene, 0.1-0.2 mm. Sphene is an 
unusual mineral in mantle rocks, but 
Dawson and Smith (1977) have noted its 
occurrence as an accessory phase in a 
MARID xenolith from the Wesselton mine. 

The harzburgite consists primarily of 
coarse olivine, ranging up to 1 cm, with a 
minor proportion of smaller grains of 
enstatite. Olivine neoblasts locally form 
interstitial zones but are insufficiently abun- 
dant to envelope the primary grains. Coarse 
crystals of phlogopite, mantled and seamed 
with finer-grained mica, range up to 8 mm 
and have a color and degree of strain simi- 
lar to the mica in the adjacent MARID rock. 
The section analyzed contains several 1- 
mm grains of garnet enveloped in fine- 
grained mica, forming a replacement tex- 
ture like that cited by Erlank et al. (1987) 
and observed in Jagersfontein peridotites 
(Boyd and Mertzman, 1987). The 
harzburgite also contains irregularly- 
shaped, coarse blebs of ilmenite mantled 



by rutile which are up to a centimeter in 
maximum dimension. The contact between 
the peridotite and MARID is sharp; minor 
serpentine and ilmenite lenticles are local- 
ized in the peridotite adjacent to the con- 
tact. 



Mineral Chemistry 

The mica and accessory phases in the 
MARID portion of xenolith 1455 have 
compositions that are comparable to those 
previously reported for MARID rocks. The 
phlogopite and diopside have low Cr203 
(Table 4); the diopside is also relatively low 
in AI2O3. The mg number of the coarse 
phlogopite is 0.87, near the high end of the 
range reported by Dawson and Smith 
(1977). The diopside with an mg number 
of 0.86 is markedly more Fe-rich than the 
Cr-diopsides in common garnet lherzolites. 
The darker mica that forms neoblasts and 
mantles on the coarse tablets is enriched in 
Ti but not in total Fe. 

The olivine and enstatite in the associ- 
ated peridotite have extremely variable Mg/ 
(Mg+Fe). Cores of large olivine crystals 
are Mg-rich (0.94) but neoblasts and mar- 
gins on the coarse grains are as Fe-rich as 
0.88 (Tables 4, 5). These variations are 
irregular, differing widely from grain to 
grain; they are not systematic in reference 
to the MARID-peridotite contact. Enstatite 
grains have comparable enrichment in Fe 
and the secondary enstatite has higher Ca 
and Ti. Enstatite close to the MARID 
contact, however, is markedly depleted in 
Al (Table 5). The pale mica and darker 
mica in the harzburgite have compositions 



20 



CARNEGIE INSTITUTION 



Table 4. Compositions of primary minerals in MARID 
1455, Kimberley, RSA. 



harzburgite compound xenolith, FRB 







Harzburgite 






MARK) 






Olivine 


Enstatite 


Garnet 


Mica 


Diopside 


Rutile 


Sphene 


Si0 2 


41.5 


57.3 


41.4 


41.9 


54.3 


<0.03 


30.4 


Ti0 2 


<0.03 


0.10 


0.70 


1.77 


0.40 


97.2 


41.9 


AI2O3 


<0.03 


0.68 


20.5 


10.1 


0.50 


0.94 


0.55 


Cr 2 3 


0.04 


0.48 


3.26 


0.18 


0.48 


1.89 


0.03 


FeO 


5.74 


4.63 


8.20 


6.22 


5.11 


0.09 


0.79 


MnO 


0.07 


0.11 


0.39 


0.04 


0.13 


<0.03 


0.04 


MgO 


53.1 


36.4 


20.6 


23.5 


17.6 


0.09 


0.10 


CaO 


<0.03 


0.50 


5.51 


<0.03 


20.6 


0.09 


24.8 


Na20 


n.d. 


0.21 


0.11 


0.10 


0.94 


n.d. 


2.50 


K 2 


n.d. 


n.d. 


n.d. 


10.4 


<0.03 


n.d. 


0.07 


NiO 


0.36 


0.08 


<0.03 


0.15 


0.03 


0.03 


0.08 


total 


100.8 


100.5 


100.7 


94.4 


100.09 


100.3 


101.3 


Mg/(Mg+Fe) 


0.943 


0.934 


0.818 


0.871 


0.861 


- 


- 



Table 5. Compositions of secondary minerals occurring as discrete grains or mantles on 
primary grains in MARID-harzburgite compound xenolith, FRB 1455, Kimberley, RSA. 



Harzburgite 



MARID 





Olivine 


Enstatite 


Ilmenite 


Rutile 


Mica 


Mica 


Si0 2 


40.5 


56.4 


<0.03 


<0.03 


40.6 


40.0 


Ti0 2 


0.03 


0.30 


53.6 


94.0 


3.73 


3.64 


A1 2 3 


<0.03 


0.22 


0.81 


0.93 


11.1 


11.1 


Cr 2 3 


0.04 


0.10 


2.59 


2.97 


0.19 


0.23 


FeO 


11.5 


8.16 


30.4 


0.29 


5.29 


5.91 


MnO 


0.13 


0.13 


0.27 


<0.03 


0.08 


0.07 


MgO 


48.9 


33.9 


13.6 


0.10 


23.2 


22.7 


CaO 


0.06 


0.91 


0.04 


0.04 


<0.03 


<0.03 


Na 2 


n.d. 


0.14 


n.d. 


n.d. 


0.20 


0.17 


K 2 


n.d. 


n.d. 


n.d. 


n.d. 


10.2 


10.3 


NiO 


0.38 


0.07 


0.19 


<0.03 


0.12 


0.12 


total 


101.5 


100.3 


101.5 


98.3 


94.7 


94.2 


Mg/(Mg+Fe) 


0.884 


0.881 


0.443 


- 


0.887 


0.873 



very similar to counterparts in the MARID 
portion of the xenolith. Rutile in both the 
MARID and peridotite contains several 
percent of Cr 2 03, a distinguishing feature 
noted for MARID rutiles by Dawson and 
Smith (1977). 



Thermobarometry 

The occurrence of olivine and garnet 
together with enstatite makes it possible to 
estimate the temperature and depth of equili- 
bration of the harzburgite- MARID assem- 



GEOPHYSICAL LABORATORY 



21 



Table 6. Estimates of the temperature and depth of equilibration of the harzburgite- 
MARID xenolith, FRB 1455. 



Thermometer 


Barometer Temperature, 


°C 


Depth, km 




olv-gar-opx 








O'Neill-Wood (1979) 
O'Neill-Wood (1979) 


MacGregor(1974) 
Nickel-Green (1985) 

opx-gar-opx 


530 
560 




50 
70 


Harley (1984) 


MacGregor(1974) 


850 




115 








Discussion 



blage. Such estimates have not previously 
been possible for MARK) xenoliths be- 
cause they do not in themselves contain a 
necessary combination of phases. Enrich- 
ment of some of the olivine and enstatite in 
Fe creates an obvious uncertainty, but the 
most magnesian compositions are taken to 
be primary (Table 4). This assumption may 
not be correct for the enstatite, however, 
because it has a slightly lower mg number 
than the olivine, the reverse of the usual 
relationship. Combining the olivine-gar- 
net and the orthopyroxene-garnet thermom- 
eters with two Al-enstatite barometers pro- 
vides results that suggest a shallow mantle 
origin for this xenolith (Table 6), perhaps 
close to the upper limit of garnet peridotite 
stability. There is a discrepancy of about 
300°C, however, between temperatures cal- 
culated with the olivine-garnet and enstatite- 
gamet thermometers, and because of the 
temperature dependance of the barometer 
there is a corresponding discrepancy in 
depth. This is probably due to slight en- 
richment of the primary enstatite in Fe, and 
the olivine-gamet estimates are believed to 
be more reliable. 



The pronounced gradients in Fe/Mg in 
the olivine of xenolith 1455 may be evi- 
dence that the metasomatism occurred in 
association with the eruptions that pro- 
duced the Kimberley diatremes. Such gra- 
dients are unlikely to persist in the upper 
mantle for periods of many million years. 
The eruption of the six large diatremes at 
Kimberley, each with multiple units of 
kimberl'te, was a complex volcanic event, 
however, and the origin of the xenolith may 
have involved more than one body of 
magma. 

A second factor of importance is the 
relatively low ambient temperature and 
depth that are reflected by the composi- 
tions of the primary minerals in the 
harzburgite (Table 6). It appears that the 
metasomatic process occurred at shallow 
levels in the craton, far removed from the 
top of the asthenosphere and from the zone 
of kimberlite generation. In such circum- 
stances it is probable that the 
metasomatizing agent was either the 
MARID magma or the kimberlite itself, 
and possibly they are the same. A relation- 



22 



CARNEGIE INSTITUTION 



ship between MARID rocks and kimberlite 
has previously been proposed (Dawson 
and Smith, 1977). The only alternative 
source for the MARID rock that is known 
to exist are intrusions of mica-rich (Group 
II) kimberlite at nearby Loxtondal that have 
an age 40 My older than the Kimberley 
diatremes (E.M.W. Skinner, personal com- 
munication). 

Many of the peridotite and dunite xeno- 
liths from Kimberley have been subject to 
Fe-Ti metasomatism with introduction of 
ilmenite and development of Fe-enriched 
marginal zones on primary olivine and py- 
roxene. These effects are believed to have 
been caused by kimberlite magmatism 
(Boyd etal., 1983), and they have appeared 
to be distinct from the introduction of coarse, 
Mg-rich mica and less commonly pargasite 
that may have originated in events that long 
pre- dated kimberlite eruption. Crystalliza- 
tion of metasomatic K-richterite in perido- 
tite, however, has involved introduction of 
mica together with titaniferous phases 
(Erlankeftf/., 1987). The latter metasomites 
appear to be linked to MARID intrusions 
by similarities in mineralogy and by the 
evidence from xenolith 1455 and from the 
similar compound xenolith studied by 
Waters etal. (1989). Some of the variations 
in metasomatic imprint may be conse- 
quences of the varying state of kimberlite- 
related metasomatizing fluid and varying 
depth at which the reactions occurred. 

Compositional changes induced in the 
peridotite of xenolith 1455 are extremely 
irregular, and their magnitude does not 
vary systematically with distance from the 
MARID contact on a scale of several cen- 



timeters. Moreover, the irregular spatial 
variation does not seem to reflect the origi- 
nal presence of some other contact. Most 
likely, the agent of metasomatism was a 
hydrothermal fluid derived from nearby 
magma that penetrated the harzburgite ir- 
regularly along fractures and grain bound- 
aries. 

It is suggested that the MARID rock of 
xenolith 1455 is an aggregate of mica with 
minor diopside and rutile that was plated on 
the peridotite wall of a dike or other conduit 
through which kimberlite (?) magma was 
erupting. The magma from which MARID 
rocks have crystallized has been proposed 
to be lamproite (Waters, 1987). There are 
no lamproite s in the Kimberley area, how- 
ever, and kimberlite itself is an alternative 
possibility. The metasomatism could have 
been produced by the same volume of melt, 
but an earlier or later origin during the 
restricted period of kimberlite magmatism 
seems possible. The more darkly colored 
Ti-rich mantles on the mica in both the 
MARID and peridotite portions of the xe- 
nolith may be evidence of a change in 
composition of the metasomatic fluid or 
even a two-stage process. 



References 

Boyd, F. R., and S. A. Mertzman, Composition 
and structure of the Kaapvaal lithosphere, 
southern Africa, in Magmatic Processes: 
Physicochemical Principles, B. O. My sen, 
ed., Geochemical Society Special Publication, 
No. l,p. 13-24,1987. 

Boyd, F. R., R. A. Jones, and P. H. Nixon, Mantle 
metasomatism: the Kimberley dunites, 
Carnegie Instn. Washington Year Book, 82, 
330-336, 1983. 



GEOPHYSICAL LABORATORY 



23 



Dawson, J. B., and J. V. Smith, The MARID 
(mica-amphibole-rutile-ilmenite-diopside) 
suite of xenoliths in kimberlite, Geochim. 
Cosmochim. Acta, 41, 309-323, 1977. 

Erlank, A. J., F. G. Waters, C. J. Hawkes worth, S. 
E. Haggerty, H. L. Allsopp, R. S. Rickard, and 
M. Menzies, Evidence for mantle metasoma- 
tism in peridotite nodules from the Kimberley 
Pipes, South Africa, in Mantle Metasomatism, 
M. Menzies and C. J. Hawkesworth, eds., 
Academic Press, 1987. 

Harley, S. L., An experimental study of the parti- 
tioning of Fe and Mg between garnet and 
orthopyroxene, Contrib. Mineral. Petrol, 86, 
359-373, 1984. 

MacGregor, J. D., The system MgO-Al203-SiC>2: 
Solubility of AI2O3 in enstatite for spinel and 
garnet peridotite compositions, Amer. Min- 
eral., 59, 110-119, 1974. 

Nickel, K. G., and D. H. Green, Empirical 
geothermobarometry for garnet peridotites and 
implications for the nature of the lithosphere, 
kimberlite s and diamonds, Earth Planet. Sci. 
Lett., 73, 158-170, 1985. 

O'Neill, H. St. C., andB. J. Wood, An experimen- 
tal study of Fe-Mg partitioning between garnet 
and olivine and its calibration as a 
geothermometer, Contrib. Mineral. Petrol., 
70, 59-70, 1979. 

Waters, F. G., A suggested origin of MARID 
xenoliths in kimberlites by high pressure crys- 
tallization of an ultrapotassic rock such as 
lamproite, Contrib. Mineral. Petrol., 95, 523- 
533, 1987. 

Waters, F. G., A. J. Erlank, and L. R. M. Daniels, 
Contact relationships between MARID rock 
and metasomatized peridotite in a kimberlite 
xenolith, G eo chemical J ., 23, 11-17, 1989. 



Boron and Beryllium Concentrations in 

Subduction-Related Metamorphic Rocks 

of the catalina schist! implications for 

subduction-zone recycling 

Gray E. Bebout, Jeffrey G. Ryan* and 
William P. Leeman** 



Enrichments in B and 10 Be in arc volca- 
nic rocks have been interpreted to reflect 
slab additions to arc source regions via 
hydrous fluids or silicate liquids (Morris et 
ai, 1990; Ryan and Langmuir, 1991). The 
two-component mixing relations among B, 
10 Be, and 9 Be demonstrated by Morris et al. 
(1990) are believed to result from the addi- 
tion of a homogenized slab-derived compo- 
nent, probably a hydrous fluid which 
strongly fractionates B from Be. The con- 
centrations of B and Be in sediments and 
crustal materials believed to be subducted 
are reasonably well characterized. Exten- 
sive B and Be data sets exist for arc volcanic 
rocks (Teraef a/., 1986; Morris etal, 1990; 
Ryan and Langmuir, 1988, 1 991; Leeman et 
al., 1990); thus, the volcanic output of these 
elements can be estimated (see Ryan and 
Langmuir, 1991, for discussion of B). How- 
ever, relatively little is known about the 
effects of slab metamorphism on the concen- 
trations of 3, Be, and other trace elements in 
subducted rocks. These metamorphic pro- 
cesses may dictate the efficiency with which 

1 

these elements are transferred from surface 
reservoirs to arc source regions; Moran et 

* Department of Terrestrial Magnetism, Carnegie 
Institution of Washington 
** Keith -Wiess Geological Laboratories, Rice 
University, Houston, Texas 7725 1 



24 



CARNEGIE INSTITUTION 



al. (1991) discuss the consequences of B 
loss due to slab metamorphism for the 
mass -balance of B inputs to arc source 
regions. Also of particular interest are the 
relative capabilities of metamorphic 
de volatilization and migmatization pro- 
cesses to fractionate B and Be sufficiently 
to produce the wide ranges of B/Be ob- 
served in arc volcanic rocks (5 to 200; 
Morris etal, 1990). 

In this report, B and Be concentration 
data are presented for metamorphosed sedi- 
mentary and mafic rocks and veins and 
pegmatites from the Catalina Schist, ex- 
posed on Santa Catalina Island in southern 
California. Mineral reservoirs for these 
elements during devolatilization of 
metasedimentary and metamafic rocks are 
considered, and the extent to which B and 
Be were mobilized during progressive vola- 
tile loss and migmatization is discussed. 



Finally, the implications of these data for 
trace element fractionation among miner- 
als, H20-rich fluids, and silicate liquids 
deep in subduction zones are discussed, as 
well as the relevance of these observations 
for interpretation of B-Be concentrations 
in arc volcanic rocks. 

The Catalina Schist consists of three 
major metamorphic-tectonic units juxta- 
posed along low-angle faults (Piatt, 1975; 
Sorensen and Barton, 1987; Bebout and 
Barton, 1989). These units contain similar 
lithologies (metamorphosed sandstones, 
shales, and cherts; metabasaltic and 
metagabbroic rocks) and range in grade 
from lawsonite-albite to amphibolite. The 
Catalina rocks are well suited for examina- 
tion of the effects of metamorphism on 
trace element and stable isotope composi- 
tion over a wide range of metamorphic 
conditions (350°-750°C, 5-11 kbar). Evi- 



ocmm m 



Lawsonite-Albite 
Blueschist 



• o 



Greenschist/Epidote 
Amphibolite 



(KCD Amphibolite 



Seafloor Sediments 
50 - 200 ppm B 







40 



80 120 

Boron (ppm) 



160 



200 



Fig. 9. Boron content in whole-rock metasedimentary samples from the Catalina Schist. The large 
range in B concentration of the lawsonite-albite and blueschist grade rocks probably reflects 
variability in sedimentary protoliths (see B data for seafloor sediments in Moran et al., 1991). 
With increasing metamorphism, B content is decreased and becomes more uniform. 



GEOPHYSICAL LABORATORY 



25 



dence for fluid transport and associated 
mass transfer during metamorphism includes 
the occurrence of veins, reaction zones be- 
tween disparate lithologies, changes in bulk 
chemical composition, and changes in iso- 
topic composition (Bebout and Barton, 
1989). Stable isotope and petrologic stud- 
ies of the Catalina Schist have yielded evi- 
dence for progressive devolatilization and 
km-scale transport of H20-rich C-O-H-S- 
N fluid during metamorphism (Bebout and 
Barton, 1989; Bebout, 1991). Intheamphi- 
bolite unit, pegmatites represent high-P mass 
transfer via silicate liquids derived through 
vapor-saturated partial melting of sedimen- 
tary and mafic rocks (Sorensen and Barton, 
1987; Bebout and Barton, 1989). 



Boron and Beryllium Concentration Data 

Boron and Be concentration data for 96 
metasedimentary, metamafic, and meta-ul- 
tramafic rocks, mineral separates, veins and 
pegmatites were obtained by inductively 
coupled plasma (ICP) analytical techniques; 
samples were fused with Na2C03 flux. All 
chemistry was done in the laboratory of J. 
D. Morris and F. Tera at the Department of 
Terrestrial Magnetism. 

Boron concentrations and B/Be of 
metasedimentary rocks decrease progres- 
sively with increasing metamorphic grade 
(Figs. 9, 10); the decrease in B content is 
consistent with results for other metamor- 
phic suites (e.g., Shaw etal., 1988; Nabelek 
et al., 1990; Moran et ai, 1991). Lowest- 
grade lawsonite-albite rocks (inferred meta- 
morphic conditions of 350°-450°C, 5-8 kbar) 
range in B content from 12 to 181 ppm B 



with a mean of 73 ppm, whereas high-grade 
amphibolite equivalents (inferred metamor- 
phic conditions of 650°-750°C, 8-11 kbar) 
range from 5.4 to 19 ppm B with a mean of 
12.2 ppm. Beryllium contents in metasedi- 
mentary rocks of all grades range from 0.3 
to 1 .2 ppm. Thus, the decrease in B/Be from 
a mean of -72 for the lawsonite-albite grade 
metasedimentary rocks to a mean of -27 for 
amphibolite grade equivalents is attributable 
to loss of B (Fig. 10). Metamafic rocks 
contain from 3 to 20 ppm B (12 samples) 
and from 0.21 to 1.31 ppm Be, with the 
lower values for both elements occurring in 
the highest grade rocks. Metamafic B/Be 
shows no significant variation with increas- 
ing metamorphic grade and ranges from 3 to 
20 (Fig. 10). 

As measures of B and Be mobility in 
hydrous fluids and felsic silicate melts, B 
and Be concentrations were obtained for 
veins, which precipitated from the hydrous 
fluids, and for pegmatites, which reflect 
migmatization in the amphibolite unit. In 
blueschist metasedimentary exposures, 



1000F 



100 



CD 

CD 

m 



Lawsonite-Albite 
Blueschist —-*■"" *» ^* 



Greenschist 
& Epidote 
Atiphibolite 




10 100 

B (ppm) 



1000 



Fig. 10. B/Be vs. B content of metasedimentary 
rocks and metamafic rocks (shaded field). B and 
B/Be of metasedimentary rocks decrease with 
increasing grade. With the exception of one 
glaucophanic greenschist sample (60 ppm B, 1.98 
ppm Be, B/Be of 30), all of the measured 
metamafic rocks have B/Be less than 20. 



26 



CARNEGIE INSTITUTION 



i 1 — 

n = 10; 32.4 
±10.3 

Na-amphibole 

(n = 10; 33.4 ± 

11.5) 




Blueschist Metasediments 



Blueschist Veins 



OO Host-Rocks 



O Pegmatites 



O Metasedimentary 
Metamafic 







20 



40 



60 



B/Be 



80 



100 



Fig. 11. Comparisons of B/Be of veins and pegmatites with those of host-rocks from the 
blueschistand amphibolite units. 



nearly monomineralic sodic amphibole 
veins have B and Be contents and B/Be 
ranges similar to their host rocks (Fig. 11). 
One albite + graphite vein contains 4.6 ppm 
B and 0.13 ppm Be and also has B/Be 
(35.4) similar to that of the surrounding 
blueschist grade metasedimentary rocks 
(32.4 ±10). Pegmatites in the amphibolite 
unit show wide ranges in B content (6 to 26 
ppm) and Be content (0.6 to 4. 1 ppm) but 
have B/Be similar to or slightly lower than 
their mafic and sedimentary hosts (Fig. 11). 
Consideration of B and Be behavior 
during devolatilization and migmatization 
requires knowledge of B and Be mineral 
residency. Table 7 contains data for whole- 
rock and mineral separate samples from 
several high-grade metasedimentary rocks 
and pegmatites. These data demonstrate 
that B is concentrated in white micas, 
whereas Be appears to be somewhat more 
evenly distributed. For all but sample 7-3- 
23, which contains tourmaline, muscovite 



contains more B than the whole-rock 
sample; Be content of the muscovite is 
comparable to that of the whole-rock 
samples. Preferential enrichment of B in 
micas is consistent with the ion microprobe 
results of Domanik et al. ( 1 99 1 ) for samples 
of the Catalina Schist. With the exception 
of several occurrences in felsic pegmatites 
(amphibolite unit) and in greenschist-grade 
metacherts. tourmaline is not a significant 
host for B in the Catalina samples. 



Discussion 

The data presented in this report are 
consistent with removal of B and Be from 
metamorphosed sedimentary and mafic 
rocks by both H20-rich fluids and felsic 
silicate liquids. The B-Be signatures of 
these two "fluids" were apparently dra- 
matically different. The H20-rich fluids 



GEOPHYSICAL LABORATORY 

Table 7. Boron and beryllium concentrations of mineral separates (in ppm). 



27 



Whole-Rock 



Muscovite 



Feldspar 



Sample 



B 



Be 



B 



Be 



* not determined 

** tourmaline-bearing pegmatite from metasedimentary exposure 



B 



Be 







Metasedimentary Rocks 






6-3-41' 

(epidote amphibolite) 


32.0 


4.1 n.d.* n.d. 


4.2 


1.1 


8-1-3 
(amphibolite) 


19.0 


0.41 39.1 0.69 
Pegmatites - Amphibolite Unit 


n.d. 


n.d. 


6-3-24 
6-3-75 
7-3-23** 


37.0 
25.7 
970 


1.3 48.0 0.92 
4.1 60.6 4.0 
2.1 113 3.0 


n.d. 
11.1 
n.d. 


n.d. 
5.8 
n.d. 



liberated through progressive 
devolatilization are inferred to have had 
high B/Be; their removal resulted in de- 
crease of the B/Be of the residual rocks. To 
explain the similarity of the Na-amphibole 
vein and blueschist metasedimentary rock 
B and Be concentrations and B/Be, a model 
is suggested where veins precipitated from 
fluids which were previously equilibrated 
with respect to B and Be with host rocks (or 
with similar rocks upstream of current hosts). 
Stable-isotope data are consistent with this 
model; O, H, and C isotope compositions of 
vein minerals are in many cases similar to 
those of the same minerals in host rocks 
(Bebout and Barton, 1989). If this model is 
correct, the veins would have had mineral/ 
fluid partition coefficients for B and Be 
similar to the bulk-rock/fluid partition coef- 
ficients of the host rocks. If the pegmatites 
are directly representative of silicate liquid 
compositions produced during melting, the 
silicate liquids had B/Be similar to those of 



host rocks. The relatively low B/Be of the 
pegmatites may thus reflect earlier removal 
of B from the source rocks during 
devolatilization (Figs. 10, 1 1 ; cf. Leeman et 
al, 1991). 

Release of B and other fluid-mobile trace 
elements during devolatilization may be 
imagined as a result of discontinuous reac- 
tions involving breakdown of mineral hosts 
(e.g., see Nabelek et at., 1990), or by a 
process of continual partitioning from min- 
eral hosts into fluids, or by a combination of 
these processes. ThegoodfitoftheCatalina 
N concentration and isotope data with a 
Rayleigh distillation model (Bebout and 
Fogel, 1 991 ; Bebout, this Report) may have 
implications for the mechanisms of B loss 
during progressive devolatilization. Boron 
and N show correlated decreases in concen- 
tration with progressive volatile loss in the 
Catalina metasedimentary rocks (see 
Bebout, this Report), and both appear to be 
concentrated in white micas. During pro- 



28 



CARNEGIE INSTITUTION 



gressive de volatilization of the Catalina 
Schist, N isotopic composition evolved 
through incremental loss of N2 equilibrated 
with the rock N reservoir, which was largely 
dominated by the micas. Decreases in B 
concentration may also reflect progressive 
boron partitioning from B-rich minerals 
(primarily the micas) into H^O-rich fluids 
derived largely by chlorite-breakdown re- 
actions (see Bebout, 1991). Thus, the B 
loss may not simply reflect the breakdown 
of micas, which show no variation in com- 
bined modal abundance with grade. 

These results virtually eliminate a bulk 
sediment/slab mixing process (presumably 
mixing of a melt derived from heteroge- 
neous slab sources) as an explanation of B- 
Be systematics in high-B/Be volcanic arcs 
and strengthen the arguments of Morris et 
al. ( 1 990) for addition to arc magma sources 
of a fractionated (high-B/Be), slab-derived, 
hydrous-fluid component. The data pre- 
sented here predict that, before subducted 
mafic and sedimentary rocks reach depths 
of the Wadati-Benioff zone below arcs (80 
to 150 km), the B/Be of these rocks is likely 
to be decreased to <40. As is also suggested 
by Moran et al. (1991), the simple mixture 
of sediment and slab with B/Be reduced by 
devolatilization will not produce a mixing 
component with sufficiently high B/Be to 
explain the linear trends in the arc data (see 
Fig. 2 in Morris et al, 1990). 

Because fractionation of B and Be in the 
mantle wedge during melting processes is 
unlikely (Morris et al, 1990; Ryan and 
Langmuir, 1991), the varying impact of 
hydrous fluid or melt additions to arc source 
regions may explain some of the B/Be 
variability observed in arc volcanic rocks 



(see B-Be data for arcs in Morris et al., 
1990; Leeman et al, 1990; Ryan and 
Langmuir, 1991). Subduction zones with 
cooler inferred thermal structures on aver- 
age show higher B/Be (e.g., Aleutian and 
New Britain arcs with B/Be of 5-40 and 40- 
200, respectively), consistent with addi- 
tion of high B/Be hydrous fluids. Subduc- 
tion zones with hotter inferred thermal struc- 
tures produce volcanic rocks with rela- 
tively low B/Be (e.g., the Cascades and 
Woodlark Basin with B/Be of 2-10 and 4- 
15, respectively), perhaps consistent with 
addition of a melted component from a slab 
previously stripped of B by lower-tempera- 
ture devolatilization. Morris et al. (1990) 
and Ryan and Langmuir (1991) have docu- 
mented decreases in B/Be across individual 
arcs, in all cases varying from high front- 
arc values to low B/Be indistinguishable 
from MORB in back-arc regions. These 
decreases could presumably represent the 
diminishing effects of hydrous fluid addi- 
tion (contributing high-B/Be signatures) 
and/or the onset of melt dominated B-Be 
transfer (contributing low-B/Be signatures). 



Conclusions 

These results demonstrate that B and Be 
are significantly fractionated during meta- 
morphic devolatilization; this process pro- 
duces high-B/Be hydrous fluids and results 
in the dramatic reduction of the B/Be of the 
subducted rocks. In contrast, partial melt- 
ing, which does not strongly fractionate B 
and Be, may produce silicate melts with 
low B/Be inherited from previously 



GEOPHYSICAL LABORATORY 



29 



devolatilized source rocks. During 
devolatilization, loss of B and N occurred 
as the elements partitioned from B- and N- 
rich phases (e.g., micas) into H20-rich 
fluids produced primarily by chlorite- 
breakdown reactions. The B and Be con- 
centrations of metasedimentary and 
metamafic rocks, veins, and pegmatites of 
the Catalina Schist place constraints on 
models that invoke addition of these ele- 
ments to arc source regions by "fluids" 
(H20-rich solutions or silicate liquids). 
The varying impact of hydrous fluid or 
melt additions to arc source regions could 
in part explain the variability in B/Be of 
arc volcanic rocks, including cross-arc 
variability in some individual arcs. Boron 
and Be may serve as analogues for other 
incompatible elements (e.g., Cs and Zr, 
respectively) which appear to behave simi- 
larly in hydrous fluid-solid and melt-solid 
systems. 



References 



Bebout, G. E., Field-based evidence for 
devolatilization in subduction zones: impli- 
cations for arc magmatism, Science, 251,413- 
416, 1991. 

Bebout, G. E., and M. D. Barton, Fluid flow and 
metasomatism in a subduction zone hydro ther- 
mal system: Catalina Schist terrane, Califor- 
nia, Geology, 17, 876-980, 1989. 

Bebout, G. E., and M. L. Fogel, Nitrogen-isotope 
compositions of metasedimentary rocks in 
the Catalina Schist: implications for meta- 
morphic devolatilization history, Geochim. 
Cosmochim. Acta, in press, 1991. 

Domanik, K., R. L. Hervig, and S. M. Peacock, 
Beryllium and boron contents of subduction 
zone minerals : an ion microprobe study, EOS, 
72,293-294, 1991. 

Leeman, W. P., D. R. Smith, W. Hildreth, Z. 
Palacz, and N. Rogers, Compositional diver- 



sity of Late Cenozoic basalts in a transect 
across the southern Washington Cascades: 
Implications for subduction zone magmatism, 
Jour. Geophys.Res., 95, 19561-19582, 1990. 

Leeman, W. P., V. B. Sisson, and M. R. Reid, 
Boron geochemistry of the lower crust: Evi- 
dence from granulite terranes and deep crustal 
xenoliths, Geochim. Cosmochim. Acta, in press, 
1991. 

Moran, A. E., V. B. Sisson, and W. P. Leeman, 
Boron in progressively metamorphosed oce- 
anic crust and sediments: implications for com- 
positional variations in subducted oceanic slabs, 
Earth Planet. Sci. Lett., in press, 1991. 

Morris, J. D., W. P. Leeman, and F. Tera, The 
subducted component in island arc lavas: Con- 
straints from Be isotopes and B-Be systemat- 
ics, Nature, 344, 31-36, 1990. 

Nabelek, P. I., J. R. Denison, and M. D. Glascock, 
Behavior of boron during contact metamor- 
phism of calc-silicate rocks at Notch Peak, 
Utah, Amer. Mineral, 75, 874-880, 1990. 

Piatt, J. P., Metamorphic and deformational pro- 
cesses in the Franciscan Complex, California: 
Some insights from the Catalina Schist terrain, 
Geol. Soc. Amer. Bull, 86, 1337-1347, 1975. 

Ryan, J. G., and C. H. Langmuir, Beryllium sys- 
tematics in young volcanic rocks: Implications 
for 10 Be, Geochim. Cosmochim. Acta, 52, 237- 
244, 1988. 

Ryan, J. G., and C. H. Langmuir, The systematics 
of boron abundances in young volcanic rocks, 
Geochim. Cosmochim. Acta, in press, 1991. 

Shaw, D. M., M. G. Truscott, E. A. Gray, and T. A. 
Middleton, Boron and lithium in high-grade 
rocks and minerals from the Wawa- 
Kapuskasing region, Ontario, Can. Jour. Ear. 
Sci.,25, 1485-1502, 1988. 

Sorensen, S. S., and M. D. Barton, Metasomatism 
and partial melting in a subduction complex: 
Catalina Schist, southern California, Geology, 
15, 115-118, 1987. 

Tera, R, L. Brown, J. Morris, I. S. Sacks, J. Klein, 
and R. Middleton, Sediment incorporation in 
island arc magmas: Inferences from 10 Be, 
Geochim. Cosmochim. Acta, 50, 535-550, 1986. 



30 



CARNEGIE INSTITUTION 



Laser Fluorination of Sulfide Minerals 
with F2 Gas 

D. Rumble, J. M. Palin, and T. C. 
Hoering 

Excitement pervades the discipline of 
stable-isotope geochemistry at the pros- 
pects for scientific advances raised by the 
development of new, microanalytical and 
in situ sampling techniques. Plans can now 
be made, with reasonable chances of suc- 
cess, to measure fundamental indicators of 
mass transfer mechanisms in Earth's crust, 
i.e. submillimeter-scale gradients in isoto- 
pic compositions. Theoretical studies dem- 
onstrate it is the curvatures, slopes, and 
magnitudes of composition vs. distance 
profiles that provide definitive evidence on 
flux magnitude and direction and the rela- 
tive importance of diffusive and infiltrative 
mass transfer (Bickle and McKenzie, 1987; 
Blattner and Lassey, 1989; Baumgartner 
and Rumble, 1988; Bowman and Willett, 
1991). Experience shows that advancing 
the understanding of mass transfer in na- 
ture depends on the ability to resolve differ- 
ences in isotopic compositions over small 
distances (Bickle and Baker, 1990a,b). 
Isotopic microanalysis by laser-heating 



minerals immersed in a reactive atmosphere 
has been successfully demonstrated for sili- 
cate, oxide, and sulfide minerals. Sharp 
(1990) showed that laser heating of small 
amounts of powdered quartz, olivine, po- 
tassium feldspar, garnet, biotite, musco- 
vite, diopside, and magnetite in a BrFs 
atmosphere released O2 with 8^0 values 
that are in good agreement with the results 
of conventional BrF5 analyses. The in situ 
analysis of spots on the surfaces of 
uncrushed grains of quartz, plagioclase, 
olivine, and magnetite has also been car- 
ried out (Sharp, 1990; Schiffries and 
Rumble, Annual Report 1989-1990, p. 37- 
40; Elsenheimer et ai, 1990; Conrad and 
Chamberlain, 1991). In situ analysis for 
<53 4 S in the minerals pyrite, pyrrhotite, 
sphalerite, galena, and chalcopyrite has 
been accomplished by laser heating and 
combustion in an O2 atmosphere and 
(Kelley and Fallick, 1990; Crowe et al., 
1990; Crowe and Shanks, 1991). The di- 
ameter of analysis spots achieved in these 
studies is 100-300 Jim, with a precision of 
± 0.2%o. 

We are currently testing the efficacy of 
F2 gas in laser fluorination to analyze sul- 
fide, silicate, and oxide minerals for ^ 4 S 
and ^l 8 0. It is desirable to develop alterna- 



Table 8. Precision of laser fluorination using F2 gas in analyzing mineral powders 



Sample 



# 4 S 



533S 



#6 S 



a 



CDT +0.16 (±0.13) 17 

NBS-123 +17.62 (±0.15) 18 

E40 -26.72 (±0.3) 14 

M-pyrite +3.17 (±0.3) 33 



+0.03 


(±0.06) 


8 


-4.1 


(±6.1) 


9 


+8.98 


(±0.13) 


9 


+34 


(±11) 


9 


-13.82 


(±0.16) 


7 


-33.7 


(±4.5) 


4 


+1.58 


(±0.17) 


22 


+24 


(±74) 


11 



o~= Standard Deviation 

CDT: Canyon Diablo Troilite. NBS-123: Sphalerite. E40: Synthetic Ag2S prepared from 
pyrrhotite; three wildly errant values for <5^ 6 S have been dropped. M-pyrite: Pyritehedrons, 1 
cm in diameter. 



GEOPHYSICAL LABORATORY 



31 



tive fluorination reagents in order to opti- 
mize analysis conditions. Conventional 
analyses of silicates for <5 18 show that 
choosing BrFs or F2 may lead to incom- 
plete fluorination and inaccurate analytical 
results for certain minerals (Taylor and 
Epstein, 1962; Clayton and Mayeda, 1963). 
In the laser analysis of sulfides, it may 
prove that laser fluorination gives less frac- 
tionated analytical results than laser com- 
bustion with O2 (cf. Kelley and Fallick, 
1990; Crowe et ai, 1990). We report, 
below, verification of the feasibility of F2 
gas for micro analysis of 5^ 4 S, cf^S, and 
<53 6 S in sulfide minerals. 



Feasibility of Sulfide Laser Fluorination 
with F2 Gas 

The feasibility of using F2 gas as fluori- 
nating agent to produce SF6 for the analysis 
of 533s, #4S, and ^36s during laser heat- 
ing of sulfide minerals has been established 
by repeated analysis of aliquots of pow- 
dered troilite, pyrite, sphalerite, and syn- 



? 

v> 
« 

e 

CO 

00 



dSJ 


_ I 


1 1 1 1 1 1 


I I I | I I I I | I I I 


* 
* — 




- 


+ NBS-123 




10 





X CDT 


• 


_ 




— 


O E40-SO 2 


* 


— 




- 


O E40-BrF 5 


♦ 


— 





— 




.«* 






- 


-10 


** 


— 




• 


- 




* 
* 


- 


-20 




— 


™ 


— * 

-*r*\ 1 1 1 i 1 1 1 1 I 1 1 1 1 I 1 1 1 


Illl 1 



-30 



-20 



-10 

5 34 S, accepted 



10 



20 



Fig. 12. Plot of accepted vs. measured values of 
# 4 SforNBS-123 sphalerite, CDT-Canyon Diablo 
Troilite, and E40 - synthetic Ag 2 S prepared from 
pyrrhotite and analysed by conventional combus- 
tion (SO2) or conventional fluorination(BrF5). 

thetic Ag2S. The precision of analysis for 
troilite and sphalerite is ± 0.06-0. \5%c for 
S&S and # 4 S and± 6-1 Wcfor &£>S (Table 
8). The precision for pyrite and synthetic 
Ag 2 S is ± 0. 1 6-0 3%o for S^S and # 4 S and 
± 27-74%o for ^S (Table 8). The preci- 
sion for 5 34 S is comparable to that obtained 
by laser oxidation/combustion of sulfide 
minerals in an O2 atmosphere (Crowe et 
al., 1990). The scatter in the results for 
pyrite may be explained, at least in part, by 
the fact that the mineral was used for prac- 



Table 9. Accuracy of laser fluorination using F2 gas. Mineral powders. 







&*S 




Sample 


(measured) 


(accepted) 


(measured- accepted) 


CDT 

NBS-123 

E40 

E40* 


+0.16 
+17.62 
-26.72 
-26.72 


0.0 
+17.09 
-27.12 
-26.88 


+0.16 
+0.53 
+0.40 
+0.16 



CDT: Canyon Diablo Troilite; standard # 4 S value defined asO.O. NBS-123: Sphalerite: 
Value of +17.09 reported as average of results of intercomparsion of 1 1 laboratories. 
(International Atomic Energy Agency, 1986). E40: Synthetic Ag2S prepared from 
pyrrhotite. Accepted value analyzed by conventional combustion (Oliver et al., Annual 
Report 1989-1990, p. 30-33). E40': Synthetic Ag2S prepared from pyrrhotite. Accepted 
value analyzed by conventional BrFs fluorination (Oliver et al., Annual Report 1989- 
1990, p. 30-33). 



CARNEGIE INSTITUTION 



2.0 



_ I I I I | I I I I J...!...!...!...!... | ...!...j...j.. ..i...i...i....i....i. % i - 



1.5 - 



£ 1.0 ^ 

o 



W « IT 

S 0.5 



-0.5 



+ NBS-123 

X CDT 

O E40-SO 2 

O E40-BrFs (shift +2 along X-axis) 




o.o E- Q 



X 

N 

x 



- 1 i i i i i i i i i i i i i i i i i i i i i i i- 



-30 



-20 



-10 

6 34 S, accepted 



10 



20 



Fig. 13. Plot of accepted values of $ 4 S vs. the 
difference between measured and accepted values 
of^S. 

tice in learning how to use F2 gas. The 
scatter in results for synthetic Ag2S has not 
yet been fully explained but may be related 
to the relatively smaller amount of avail- 
able sulfur as constrained by stoichiom- 
etry. The accuracy of the results was evalu- 
ated by comparing 5^ 4 S of SF6 derived 
from F2 laser fluorination of CDT, NBS- 
123, and E 40 to "accepted" # 4 S values 
obtained by conventional combustion with 
O2 and fluorination with BrF5. The results 
show deviations of from +0.16 to +0.53%c 
in <P 4 S for F2 laser fluorination samples 
relative to accepted values (Figs. 12, 13, 
Table 9). 

The experimental apparatus used for 
F2 laser fluorination resembles, in general 
outline, that used by Sharp (1990) for BrF5 
fluorination of silicate minerals. There has 
been added, however, a fluorine gas gen- 
eration and delivery system, a heated KBr 
trap to dispose of excess F2, and an Inconel 
capacitance manometer to measure yields 
and to monitor the progress of reactions 
and the course of cryogenic transfers of 
condensable gases. Fluorine is generated 
by heating the compound K2NiF6 • KF in 
an evacuated nickel reservoir to tempera- 



tures of 290°-320°C (Asprey, 1976).* El- 
emental fluorine is incorporated at low 
temperatures (~250°C) and released at 
higher temperatures according to the re- 
versible reaction (Asprey, 1976) 

2(K2NiF6 • KF) = 2K3N1F6 + F2 (1) 
solid solid gas 

Use of the compound solves two problems 
encountered in fluorine chemistry. (1) F2 
gas can be obtained free from contamina- 
tion by N2 or O2. We outgas the compound 
at 100°C under vacuum prior to evolution 
of F2. (2) The hazard of handling large 
quantities of F2 gas is eliminated. We 
generate only the small amounts of F2 
needed for laser fluorination. Excess F2 
can be resorbed by the reversal of reaction 
(1) at lower temperature. 

The experimental procedure consists of 
generating F2 gas, expanding it into an 
evacuated sample chamber loaded with 
samples, and aiming and firing the laser. 
The reaction chamber is open to a liquid 
nitrogen cold trap to remove SF6 cryogeni- 
cally from the reaction site as soon as it is 
formed. After laser heating, excess F2 is 
removed by reaction with KBr. The prod- 
uct SF6 is held back in a cold trap to avoid 
mixing it with the bromine formed when F2 
reacts with KBr. The apparatus is not fully 
optimized for laser fluorination of sulfides 
because at this step of the procedure prod- 
uct SF6 must be removed in a Ni-metal 
container and carried to another vacuum 

1 We are indebted to J. O'Neil for alerting us to the 
existence of the K2NiF6 • KF compound, to G. P. 
Landis for furnishing unpublished information on 
its use, and to Ozark- Mahoning, Inc. for manufac- 
turing it. 



GEOPHYSICAL LABORATORY 



33 



line for purification. The purification con- 
sists of reacting SF6 with moist KOH to 
eliminate traces of F2 and HF followed by 
gas chromatography to remove trace con- 
taminants that give isobaric interferences 
with the ion beams of SF5 + in the mass 
spectrometer (Hoering, Annual Report 
1989-1990, p. 128-131). 

Experimental parameters, such as pres- 
sure of F2 gas and laser power, were varied 
systematically to establish optimal operat- 
ing conditions. Pressure of F2 gas in the 
reaction chamber was varied from 55 to 
175 torr. Fluorination proceeds readily at 
lower pressures and is preferred to econo- 
mize on consumption of the reagent. The 
20-watt, CO2 laser was operated in both 
continuous and pulsed mode. In the analy- 
sis of powders, minimum power was used 
in order to protect the fragile BaF2 win- 
dows from damage. For most samples, 
pulsed operation with a pulse spacing of 10 
milliseconds and pulse width of 10 milli- 
seconds at a laser power setting of 20% is 
adequate to achieve complete fluorination. 
An additional problem is encountered if the 
initial laser shot is set at high power: the 



10 
5 

-5 



_i i i i | i i i i | i i i i | i i i i | i i i i_ 

.4? 



j¥ 



<c*» 



<* 6 



■ r ^ • * * 



-10 - 

_ * 

.15 Eg- i i i 1 i 1 1 i 1 1 1 1 i i i 1 i i 1 1 1 1- 



►£• Sulfide Powders 



-30 -20 -10 



10 20 



5 34 S 



Fig. 14. Plot of $ 3 S vs. # 4 S showing measured 
vs. theoretical mass fractionation. 



powdered sample may be scattered around 
the sample chamber. The problem may be 
mitigated by starting the laser at low power 
and low pulse width and increasing incre- 
mentally until reaction is seen. It is diffi- 
cult to avoid some scattering of powdered 
samples. For this reason, yields of SF6 in 
relation to weighed amounts of samples are 
less than stoichiometric. 

Analysis for four sulfur isotopes, 32s, 
33S, 34 S, and 36s, is made simultaneously 
on aFinnigan-M AT '251 mass spectrometer 
with custom quadruple collector. The mea- 
surement of all the isotopes provides a 
useful test for the quality of the analysis 
(Fig. 14). 

The theory of isotope fractionation in 
chemical processes predicts that the ratio 
<534SA533s = 1.94 (±0.01) in samples such 
as terrestrial ores, where anomalous nucleo- 
synthetic effects are absent (Hulston and 
Thode, 1965). We obtained a value of 
1.950 (± 0.003) for the ratio (Fig. 14). 

The results reported above validate the 
use of F2 gas in laser fluorination of sul- 
fides. The laser fluorination of sulfides is 
far quicker than conventional fluorination: 
reactions are completed in seconds or min- 
utes rather than overnight. In comparison 
to laser combustion of sulfide minerals in 
O2 (Crowe et al. 1 990), laser fluorination is 
slower, because of the preparative gas chro- 
matography that is required to eliminate 
isobaric interferences in the mass spec- 
trometer. The advantages of the method 
are clear, however. One obtains precise 
data on three of the four isotopes 32 S, 33 S, 
and 34 S. The additional uncertainties in- 
troduced by oxygen isotope corrections in 



34 



CARNEGIE INSTITUTION 



mass spectrometry of SO2 are eliminated, 
for fluorine has only one stable isotope. 
Because SF6 is chemically inert, does not 
sorb onto the interior surfaces of vacuum 
lines, and is not sensitive to moisture, it is 
potentially a better working gas than SO2 
for mass spectrometry of very small samples 
(Rees, 1978). 



References 

Asprey, L. B., The preparation of very pure fluo- 
rine gas, J. Fluorine Chem., 7, 359-361, 1976. 

Baumgartner, L. P., and D. Rumble, Transport of 
stable isotopes. I. Development of a kinetic 
continuum theory for stable isotope transport, 
Contrib. Mineral. Petrol, 98, 417-430, 1988. 

Bickel, M., and D. McKenzie, The transport of 
heat and matter by fluids during metamor- 
phism, Contrib. Mineral. Petrol., 95, 384-392, 
1987. 

Bickel, J., and J. Baker, Advective-diffusive trans- 
port of isotopic fronts: an example from Naxos, 
Greece, Earth Planet. Sci. Lett., 97, 78-93, 
1990a. 

Migration of reaction and isotopic 

fronts in infiltration zones: Assessments of fluid 
flux in metamorphic terrains, Earth Planet. Sci. 
Lett., 98, 1-13, 1990b. 

Blattner, P., and K. R. Lassey, Stable-isotope 
exchange fronts, Damkohler numbers, and fluid 
torockratios, Chem. Geoi, 78, 381-392, 1989. 

Bowman, J. R, and S. D. Willett, Spatial patterns 
of oxygen isotope exchange during one dimen- 
sional fluid infiltration, Geophys. Res. Lett., 18, 
971-974, 1991. 

Clayton, R.N., and T. K. Mayeda, The use of BrF5 
in the extraction of O2 from oxides and silicates 
for isotopic analysis, Geochim. Cosmochim. 
Acta, 27, 43-52, 1963. 

Conrad, M. E., and C. P. Chamberlain, Laser- 
based analyses of small-scale variations in the 
oxygen isotope ratios of hydrothermal quartz, 
£05,72,292,1991. 

Crowe, D. E., J. W. Valley, and K. L. Baker, 
Micro-analysis of sulfur-isotope ratios and zo- 
nation by laser microprobe, Geochim. 
Cosmochim. Acta, 54, 2075-2092, 1990. 

Crowe, D. E., and W. C. Shanks, Laser micro- 
probe $ 4 S study of coexisting sulfide pairs: 
seeing through metamorphism, EOS, 72, 1991. 

Elsenheimer, D., J. W. Valley, and K. Baker, In- 
situ laser microprobe determinations of (5 18 0, 



GSAAbstractsw. Programs, 22, 160-161, 1990. 

Hulston, J. R., andH. G., Thode, Variations in the 
33 S, 34 S, and 36 S contents of meteorite and their 
relation to chemical and nuclear effects, /. 
Geophys. Res., 70, 3475-3484, 1965. 

Kelley, S. P.. and A. E. Fallick, High precision 
spatially resolved analysis of £* 4 S in sulfides 
using a laser extraction technique, Geochim. 
Cosmochim. Acta, 54, 883-888, 1990. 

Rees, C. E., Sulfur isotope measurements using 
SO2 and SF6, Geochim. Cosmochim. Acta, 42, 
383-389, 1978. 

Rumble, D., T. C. Hoering, and J. M. Palin, Mi- 
croanalysis for $ A S in sulfide minerals with 
laser fluorination, EOS, 72, 292, 1991. 

Sharp, Z. D., A laser-based microanalytical method 
for the in situ determination of oxygen isotope 
ratios of silicates and oxides, Geochim. 
Cosmochim. Acta, 545, 1353-1357, 1990. 

Taylor, H.P., and S. Epstein, Relationships be- 
tween 18 0/ 16 ratios in coexisting minerals of 
igneous and metamorphic rocks, Geol. Soc. 
Amer. Bull., 73, 461-480, 1962. 



Stable Isotope and Trace Element Indi- 
cators of Devolatilization History in 
Metashales and Metasandstones 

Gray E. Bebout 

Devolatilization of carbonate-poor 
metashales and metasandstones has the po- 
tential to release large amounts of H2O- 
rich C-O-H-S-N fluid (e.g., Walther and 
Orville, 1982). However, few studies have 
directly examined this process, in contrast 
with the many studies that have dealt with 
devolatilization/inf iltration in meta-carbon- 
ate systems (e.g., Valley, 1986). Because 
metashale and metasandstone make up a 
significant fraction of the continental crust, 
their devolatilization may be extremely 
important for crustal chemical, thermal, 
and rheological evolution. The 
metasedimentary rocks of the Catalina 
Schist (California) are well suited for a 



GEOPHYSICAL LABORATORY 



35 



study of devolatilization; rocks of similar 
bulk composition have been metamor- 
phosed under a wide range of metamor- 
phic conditions (350°-750°C, 5-11 kbar; 
Piatt, 1975; Sorensen and Barton, 1987; 
Bebout and Barton, 1989). Trace element 
and stable isotope compositions, mineral 
modes, and mineral compositions show 
distinct covariance with increasing meta- 
morphic grade. In this study, the integra- 
tion of petrologic and geochemical data 
results in a distillation model for progres- 
sive devolatilization. This model neces- 
sarily implies specific devolatilization 
mechanisms which have consequences for 
models of crustal chemical evolution, fluid 
transport dynamics, and metamorphic re- 
action kinetics. 

The three major metamorphic units of 
the Catalina Schist (lawsonite-albite/ 
blueschist, glaucophanic greenschist/epi- 
dote-amphibolite, and amphibolite) con- 
tain sedimentary, mafic, and ultramafic 
rocks underplated and metamorphosed 
during early Cretaceous subduction (Piatt, 
1975; Bebout, 1991a). Metasedimentary 
rocks from the three units show a similar 
range in lithology from metapelites to meta- 
graywackes; average grain size increases 
with increasing grade (several \im to sev- 
eral mm). Trends in H2O content with 
increasing metamorphic grade and pro- 
grade reaction histories inferred from min- 
eral modes indicate that devolatilization of 
the metasedimentary rocks of the Catalina 
Schist principally involved chlorite break- 
down reactions over the approximate tem- 
perature interval 400°-600°C (Bebout, 
1991a). H2O content, determined by H- 
isotope extraction techniques, decreases 



from 2-5.5 wt % in lowest-grade rocks to 
1.5-2.5 wt % in highest-grade, amphibolite- 
facies rocks (Bebout, 1991a). Chlorite 
breakdown reactions resulted in the produc- 
tion of muscovite-, biotite-, garnet-, and 
kyanite-bearing mineral assemblages 
through reactions of the following general 
types: 

2Phengite + Chlorite = Muscovite 

+ Biotite + Quartz + 4H2O (1) 

2Chlorite + 4Quartz 

= 3Garnet + 8H2O (2) 

3 Chlorite + 7 Muscovite + Quartz 

= Al2Si05(Kyanite) + 7Biotite 

+ 12H 2 (3) 

Chlorite + Muscovite = Biotite + 
Al2Si05(Kyanite) + Quartz + 8H2O (4) 

In addition to the decrease in H2O con- 
tent, devolatilization resulted in preferential 
loss of some trace elements and in shifts in 
the C and N isotope compositions of the 
rocks. The metasedimentary rocks show 
trends of decreasing N concentration and 
increasing whole-rock 8 5 N with increas- 
ing metamorphic grade (Bebout and Fogel, 
Annual Report 1989-1990, p. 19-26; Fig. 
15). The £ 13 C of carbonaceous matter in 
these rocks increases from values of from 
-26 to -24 %o in the lowest-grade rocks to 
values of from -21 to -19 %o in the highest- 
grade, amphibolite-facies rocks (Fig. 15). 
Concentrations of carbonaceous matter do 
not vary systematically with increasing grade 
and average -0.6 wt % for all grades. De- 
spite metamorphism at temperatures ex- 



36 



CARNEGIE INSTITUTION 



■18 



o 

CO 



-20 - 



§ -22 

Q. 



-24 



-26 



-28 




Lawsonite- • 
Albite & 
Blueschist 

JL 




Greenschist & 

Epidote 

Amphibolite 



j i_ 



j i_ 



5 15 N 



4 
Air 



Fig. 15. Whole-rock 5 15 N vs. <5 13 C of carbon- 
aceous matter for metasedimentary rocks of the 
Catalina Schist. Note parallel trends toward higher 
isotope values. 

ceeding 350°C, C/N of the lowest-grade 
metasedimentary rocks (5-20; mean -13) 
is similar to C/N of many unmetamorphosed 



sedimentary rocks, including those in trench 
and off-trench environments (cf. Patience 
et al., 1990). Higher-grade rocks have 
higher C/N that is attributable to preferen- 
tial N loss (for 6 greenschist and epidote 
amphibolite samples, range is from 5 to 
1 25; for four amphibolite samples, range is 
from 28 to 237). Boron concentration in 
the same rocks decreases with increasing 
metamorphic grade from an average of -73 
ppm in lawsonite-albite grade rocks to an 
average of -12 ppm in amphibolite -grade 
rocks (Fig. 16; Bebout et a/., this Report). 
Bebout and Fogel (Annual Report 1989- 
1990, p. 19-26) interpreted the N concen- 
tration and isotope data to be results of 
Rayleigh distillation behavior during pro- 
gressive devolatilization (see equation [4] 
in Valley, 1986). A remaining difficulty in 
the Rayleigh distillation calculations stems 
from variability in composition within an 
individual grade. This variability, which in 



1000 



800 



E 

Q. 



^ 600 



CD 

D) 

O 



400 - 



200 




Lawsonite-Albite & 
Blueschist (• ) 



Greenschist & Epidote 
Amphibolite (A ) 



Amphibolite (o ) 



100 

Boron (ppm) 



200 



Fig. 16. N and B concentrations of metasedimentary rocks of the Catalina Schist. 
See discussion of the significance of B loss for subduction zone recycling in 
Bebout et al. (this Report). 



GEOPHYSICAL LABORATORY 



37 



E 

CL 
Q. 

C 
CD 
D) 
O 



800 



600 



400 



200 




Mean F Calculated 

from Sample Pairs 

0.26 + 0.07 

(n = 12) 



Lawsonite-Albite & 
Blueschist 



F 0.4 - 




K>0 (weight %) 

Fig. 17. Relationship of N and K2O concentra- 
tions in metasedimentary rocks of the Catalina 
Schist. Tie lines connect samples with similar 
K2O content and demonstrate that high-grade 
samples contain 0.26 ± 0.07 of the N in low-grade 
samples with similar K content; calculated F is 
shown for several of the tie lines. The mean 8 N 
shift between low- and high-grade samples in 
these same pairs is 1.85%c (standard deviation of 

1.0%©). 

part represents protolith variability, compli- 
cates comparisons of trace element compo- 
sition and volatile content of low- and high- 
grade metamorphic equivalents. The 
protolith variability problem can be allevi- 
ated by normalization of the N data to K2O 
data for the same rocks. Because N is 
preferentially partitioned into micas (Honma 
and Itihara, 1981), and because the micas 
constitute the only significant K reservoir in 
the rocks, the concentrations of N correlate 
well with K2O content of the rocks (Fig. 
17). By comparing only samples with simi- 
lar K2O content (i.e., with similar modal 
abundance of micas), fluid-rock isotope frac- 
tionation factors can be more tightly con- 
strained, as shown in Fig. 18. 



Fluid - Rock 



Fig. 18. Demonstration of the interdependencies 
of 8 N shift due to Rayleigh distillation (labelled 
curves), fluid-rock N-isotope fractionation 
(A 15 Nfi u id-rqck; related to alpha, the fractionation 
factor, by 8 ^Ifiuid - 8 N roc k = 10 3 In alpha), and 
inferred N loss (F indicates the fraction of the 
original N remaining in the rock.) Indicated are 
estimates of N loss % based on differences in 
mean N content in the lawsonite-albite and am- 
phibolite facies rocks (0.22; dashed line) and 
based on normalizations by use of the K2O data 
indicates an F or ~0.26±0.07; see solid horizontal 
lines that indicate mean ± one standard devia- 
tion). Curves are for the mean of the 8 N shifts 
for the pairs with similar K2O content (± one 
standard deviation on lower and higher 8 N 
sides; 1.85 ± 1 .0 %p- solid curves), and the differ- 
ence in the mean 8 N of the lawsonite-albite and 
amphibolite samples (dashed curve labeled 2.4%o; 
see Bebout and Fogel, Annual Report 1989-1990, 
p. 19- 26 for data). Shaded region indicates range 
of fractionations (-1.511.0 %6) compatible with 
the Catalina data based on these data for N loss 
and 8 N shift. Vertical arrows indicate frac- 
tionations calculated by use of differences in 
mean N and 8 N of the low- and high-grade 
rocks (—1.6 %o) and by use of the K^O-normal- 
ized data (—1.4 %o). 

Values of A 15 Nfi u id-rock (<5 15 Nfi u id - 
<5 15 N r ock) of about -1 .5+1 .0%o inferred in 
Fig. 1 8 are similar to values of from -1 .0 to 
-4.0%o suggested by Haendel et ai (1986) 
based on isotope and concentration shifts 
in metamorphic suites, but are smaller 



38 



CARNEGIE INSTITUTION 



than ranges of from -3.0 to -4.0 %oand -6.0 
to -11.5 %o obtained by Kreulen et al 
(1982) and Bottrell et al (1988), respec- 
tively, for fluid inclusion-mineral/rock pairs 
in low to medium grade metasedimentary 
rocks. The range of fractionation factors 
indicated in this study is similar to the 
range calculated by Hanschmann (1981) 
for N2-NH4" 1 " exchange equilibrium at 400°- 
600°C (A 15 Nfiuid-rock of from -2.25 to -2.9 
%o for the temperature range of 400°- 
600°C). Speciation of N as N2 in the 
Catalina fluids is further suggested by the 
presence of small amounts of N2 in fluid 
inclusions from metasomatized eclogitic 
blocks in blueschist melange (<1 mol %; T. 
C. Hoering, personal communication, 1991; 
determined by quadrupole mass spectrom- 
etry). Furthermore, calculations of fluid C- 
O-H-N equilibria indicate that N2 is the 
dominant N species under most crustal 
metamorphic conditions (Duit etal, 1986; 
Ferry and Baumgartner, 1987; Bottrell et 
al., 1988). 

Carbon and O isotope data for the same 
rocks are compatible with the model of 
Rayleigh distillation derived from the N 
data. Progressive fractionation of 12 C from 
carbonaceous matter into CH4 in fluids by 
a Rayleigh distillation process could pre- 
sumably explain the observed shift in <5 13 C 
(Fig. 15; seeBebout, 1989). Because of the 
relatively large A 13 C of CH4-graphite ex- 
change at these temperatures (-3 to -7 %o\ 
see Bottinga, 1969), such a shift in £ 13 C 
could be achieved without a large decrease 
in whole -rock C concentration. Heating/ 
freezing experiments (Sorensen and Barton, 
1987) and quadrupole mass spectrometry 
indicate that CH4 is the dominant C species 



(<1-10 mol %) in fluid inclusions from 
metasomatized mafic blocks in blueschist 
and amphibolite unit melange. The unifor- 
mity of <5 18 of the metasedimentary rocks 
at all grades (quartz <5 18 of approximately 
+16 to +19 %o; Bebout, 1991a) indicates 
that the metasedimentary rocks did not 
pervasively reequilibrate with the H2O- 
rich fluids that produced veins and infil- 
trated more permeable portions of the 
Catalina Schist (Bebout, 1991a, b). This 
inferred relative impermeability of the 
metasedimentary rocks supports the as- 
sumption of closed- system behavior im- 
plicit in the Rayleigh distillation calcula- 
tions. 

The Rayleigh calculations assume that 
isotopic equilibrium is maintained between 
the fluid phase and the remaining mineral 
phases and among mineral phases during 
incremental loss of fluid. In the Catalina 
metasedimentary rocks, the mechanisms 
affording this continual reequilibration pre- 
sumably involved diffusive exchange and 
dissolution/reprecipitation during grain 
coarsening. The increase in white mica 
grain size, a trend in white mica chemistry 
{decrease in celadonite substitution, [(Mg, 
Fe 2+ ) + Si = 2A1]; Sorensen, 1986; Bebout, 
1989}, and increases in the grain size and 
degree of crystallinity of carbonaceous mat- 
ter (Bebout 1 989) presumably reflect these 
processes. For other rocks, Duit et al 
(1986) reported coupled decreases in mica 
and whole-rock N concentration with in- 
creasing metamorphic grade. Such de- 
creases are consistent with progressive par- 
titioning of N into fluids equilibrated with 
the remaining micas, as opposed to selec- 
tive release of N due to mica breakdown 



GEOPHYSICAL LABORATORY 



39 



(i.e., with the remaining mica retaining its 
original higher N content) . Similarly, in an 
ion microprobe study of the rocks ana- 
lyzed in this study, Domanik et al. (1991) 
report a general decrease in the B concen- 
tration of white micas with increasing meta- 
morphic grade. 

The proposed model requires that flu- 
ids released by devolatilization escape 
without appreciable infiltration by exter- 
nally-derived fluids significantly out of 
0-, C-, or N-isotopic equilibrium with the 
rocks. Such a scenario might arise as 
locally-derived fluids migrate within lay- 
ers of similar composition and at similar 
P-Tconditions toward fractures along pore 
fluid pressure gradients (cf. Etheridge et 
al., 1984). Veins, which represent the 
larger fractures, are abundant in the Catalina 
Schist, particularly at lower grades (Bebout 
and Barton, 1989). The fluids in this 
scenario would always be equilibrated with 
the rocks along their flow paths and would 
not leave a significant imprint on trace 
element and stable isotope compositions 
in the host rock. However, flow across 
layering (e.g., of mixed sandstone and 
shale sequences) would possibly homog- 
enize elemental and isotopic variations 
that resulted from local devolatilization 
history or protolith variability, depending 
on the magnitude of the variations, fluid- 
to-rock ratios, and fluid composition. The 
degree of homogenization would depend 
in part on fracture density, which would 
control the scale of the intergranular flow, 
and the relative permeabilities of the 
interlayered lithologies, which would dic- 
tate whether or not intergranular flow oc- 
curred across layering. To address further 



these issues, ongoing research involves 
analyses of trace element concentrations 
and stable isotope composition in finely 
interlayered, veined sequences of lawsonite- 
albite and blueschist grade metasandstone 
and metashale. Preliminary data indicate 
the preservation of -1.0 %o gradients in 
8 5 N within one finely interlayered (cm- 
scale) lawsonite-albite grade meta-sedimen- 
tary exposure. 

For carbonate-poor metamorphosed 
sandstones and pelitic rocks, N and C (and 
perhaps S) isotope systematics appear to be 
more useful measures of the extent of 
devolatilization than either the H or O sys- 
tems, owing to a more favorable mass- 
balance between the fluids and rocks. Be- 
cause of the large reservoir of H in high- 
X(H20) fluids and the low H/O of the rocks, 
trends in H isotope compositions due to 
fluid loss are probably modified more easily 
by interaction with the small amounts of 
fluid derived within local fluid generation/ 
extraction systems. The Catalina meta- 
sedimentary rocks show no obvious trend in 
SD with increasing metamorphic grade and 
range from -80 to -50 %oat all grades (Bebout, 
1989). Because of the large O reservoir in 
the rocks, Rayleigh distillation results in 
only minor change in whole-rock 5 °0 with 
progressive devolatilization; this change is 
probably unresolvable given the marked 
protolith variability in <5 18 0. Similar argu- 
ments were made by Oliver et al. (Annual 
Report 1989-1990, p. 30-33) for S-isotope 
systematics in sulfidic schists from the 
Waterville-Augusta area, Maine. Because 
of the low S content in the metamorphic 
fluids (2-5 mol %; Ferry, 1981), greenschist- 
grade metamorphic rocks retained S-iso- 



40 



CARNEGIE INSTITUTION 



tope signatures inherited from diagenetic 
and lower-grade metamorphic de- 
volatilization histories. These arguments 
are also analogous to those made for O and 
C isotope systematics during decarbon- 
ation of impure carbonates infiltrated by 
H20-rich, C-poor fluids. Because of the 
low C content of the infiltrating fluids, C- 
isotope composition of the rocks is re- 
garded as a better measure of the extent of 
decarbonation than the O-isotope system- 
atics, which are commonly controlled by 
infiltration (Rumble, 1982; VaUey, 1986). 
A distillation mechanism like that pro- 
posed for the Catalina N data may dictate 
the mobilities of other fluid-mobile trace 
elements (e.g., B, Cs, U; see Heier, 1973; 
Leeman et aL, 1990) residing in a range of 
mineral reservoirs (i.e., not only in micas, 
which are the predominant N and B reser- 
voir). The efficiency with which equilib- 
rium is maintained between fluids and min- 
erals during incremental loss may be dic- 
tated by the relative intracrystalline diffu- 
sion rates of the elements in their respective 
mineral reservoirs or by dissolution/ 
reprecipitation rates of the host minerals 
during fluid loss. Boron concentration, 
like N concentration and isotope composi- 
tion, may have been controlled by the bulk 
fluid loss and fluid-rock exchange history 
of the rocks. If so, B partitioned progres- 
sively from B-rich minerals (primarily the 
white micas) into H20-rich fluids derived 
largely by chlorite breakdown. Thus, as 
with N, the B loss need not be attributed to 
breakdown of host minerals such as white 
mica and biotite, which show no obvious 
decrease in their combined modal abun- 



dance with increasing metamorphic grade 
[see equations ( 1 -4)] . Continuous exchange 
reactions like those which stabilized micas 
during progressive metamorphism may also 
allow the continual reequilibration of fluid 
and rock stable isotope and trace element 
composition during progressive 
devolatilization and/or infiltration by ex- 
ternally-derived fluids. 



References 



Bebout, G. E., Geological and geochemical inves- 
tigations of fluid flow and mass transfer during 
subduction-zone metamorphism, Ph. D. Dis- 
sertation, University of California, Los Ange- 
les, 1989. 

Bebout, G. E., Field-based evidence for 
devolatilization in subduction zones: Implica- 
tions for arc magmatism, Science, 251, 413- 
416, 1991a. 

Bebout, G. E., Geometry and mechanisms of fluid 
flow at 15 to 45 kilometer depths in an early 
Cretaceous accretionary complex, Geophys. 
Res. Lett., 18, 923-926, 1991b. 

Bebout, G. E., and M. D. Barton, Fluid flow and 
metasomatism in a subduction zone hydrother- 
mal system: Catalina Schist terrane, Califor- 
nia, Geology, 17, 876-980, 1989. 

Bottinga, Y., Calculated fractionation factors for 
carbon and hydrogen isotope exchange in the 
system calcite-C02-graphite-methane-hydro- 
gen and water vapor. Geochim. Cosmochim. 
Acta 33, 49-64, 1969. 

Bottrell, S. H., L. P. Carr, and J. Dubessy, A 
nitrogen-rich metamorphic fluid and coexist- 
ing minerals in slates from North Wales, Min- 
eral. Mag., 52, 451-457, 1988. 

Domanik, K., R. L. Hervig, and S. M. Peacock, 
Beryllium and boron contents of subduction 
zone minerals: an ion microprobe study, EOS, 
72,293-294,1991. 

Duit, W., Jansen, J. B. H., Van Breemen, A., and 
A. Bos, Ammonium micas in metamorphic 
rocks as exemplified by Dome de L'Agout 
(France). Amer. J. Sci., 286, 702-732, 1986. 

Etheridge, M. A., Wall, V. J., Cox, S. R, and R. H. 
Vernon, High fluid pressures during regional 
metamorphism and deformation: Implications 
for mass transport and deformation mech- 
anisms,/ Geophys.Res.,89, 4344-4358, 1984. 



GEOPHYSICAL LABORATORY 



41 



Ferry, J. M., Petrology of graphitic sulfide-rich 
schists from south-central Maine: An example 
of desulfidation during prograde regional meta- 
morphism. Amer. Mineral., 66, 908-930, 198 1 . 

Ferry, J. M., andL. Baumgartner, Thermodynamic 
models of molecular fluids at the elevated pres- 
sures and temperatures of crustal metamor- 
phism, Rev. Mineral., 17, 323-365, 1987. 

Haendel, D., K. Muhle, H.-M. Nitzsche, G. Stiehl, 
and U. Wand, Isotopic variations of the fixed 
nitrogen in metamorphic rocks, Geochim. 
Cosmochim. Acta, 50, 749-758, 1986. 

Hanschmann, G., Berechnung von Isotop Effekten 
auf quantenchemischer Grundlage am Beispiel 
stickstoffhaltiger Molekule, Zfl-Mitt., 41, 19- 
39, 1981. 

Heier, K. S., Geochemistry of granulite facies 
rocks and problems of their origin, Phil. Trans. 
R. Soc. Lond. A., 273, 429-442, 1973. 

Honma, H., and Y. Itihara, Distribution of ammo- 
nium in minerals of metamorphic and granitic 
rocks, Geochim. Cosmochim. Acta, 45, 983- 
988, 1981. 

Kreulen, R., A. Van Breeman, and W. Duit, Nitro- 
gen and carbon isotopes in metamorphic fluids 
from the Dome de L'Agout, France, Proceed- 
ings of the Fifth International Conference on 
Geochronology, Cosmochronology, and Iso- 
tope Geology, p. 191, 1982. 

Leeman, W. P., A. E. Moran, and V. B. Sisson, 
Compositional variations accompanying meta^ 
morphism of subducted oceanic lithosphere: 
Implications for genesis of arc magmas and 
mantle replenishment, Abstr., Seventh ICOG 
Mtg., 58, 1990. 

Patience, R. L., C J. Clayton, A. T. Kearsley, S. J. 
Rowland, A. N. Bishop, A. W. G. Rees, K. B. 
Bibby, and A. C. Hopper, An integrated bio- 
chemical, geochemical, and sedimentological 
study of organic diagenesis in sediments from 
Leg 1 12, in Proceedings of the Ocean Drilling 
Program, Scientific Results, 112, E. Suess et 
al., eds., Ocean Drilling Program, College Sta- 
tion, Texas, 135-153, 1990. 

Piatt, J. P., Metamorphic and deformational pro- 
cesses in the Franciscan Complex, California: 
some insights from the Catalina Schist terrain, 
Geol. Soc. Amer. Bull, 86, 1337-1347, 1975. 

Rumble D., Stable isotope fractionation during 
metamorphic volatilization. Rev. Mineral. 10, 
327-353, 1982. 

Sorensen, S. S., Petrologic and geochemical com- 
parison of the blueschist and greenschist units 
of the Catalina Schist terrane, southern Califor- 
nia, Geol. Soc. Amer. Mem., 164, 59-75, 1986. 

Sorensen, S. S., and M. D. Barton, Metasomatism 
and partial melting in a subduction complex: 
Catalina Schist, southern California, Geology, 
15, 115-118, 1987. 



Valley, J. W., Stable isotope geochemistry of meta- 
morphic rocks, Rev. Mineral. 16, 445-489, 
1986. 

Walther, J. V., and P. M. Orville, Volatile produc- 
tion and transport in regional metamorphism, 
Contrib. Mineral. Petrol, 79, 252-257, 1982. 



The fa Content of Normative ol 
Felix Chayes 

Replacement of conventional "wet- 
way" analytical procedures by instrumented 
physical techniques is generating a large 
and rapidly expanding reservoir of rock 
analyses in which Fe is not partitioned by 
oxidation state. (Of the 14,722 analyses of 
igneous rocks included in the current ver- 
sion of the base IGBADAT, for instance, 
almost a quarter lack Fe partition.) Com- 
parison of analyses subject to this defi- 
ciency with older data, indeed, plotting 
them in many common petrographic varia- 
tion diagrams, requires some external 
nonanalytical assignment of Fe oxidation 
state. 

In the several procedures now in use 
(for a thorough review see Middlemost, 
1989), this adjustment consists of the ap- 
plication of a simple formula to convert the 
lone Fe or Fe-oxide value into two. With 
few exceptions the conversion is made 
without regard to the amounts of other 
components reported in the bulk analysis. 
And however important petrologic and 
mineralogical factors may have been in the 
development of these relations, they play 
no direct role in their application. In prac- 
tice, the preferred rule is simply applied by 
rote to any analysis that is to be used in a 
fashion requiring some estimate of the oxi- 



42 



CARNEGIE INSTITUTION 



dation state of Fe in the specimen in ques- 
tion. 

In olivine 1 bearing rocks this may lead 
to ol unrealistically rich in fa. Fortunately, 
loss of information about the oxidation 
state of Fe in bulk analyses of igneous 
rocks has been accompanied by a manifold 
increase in the amount of information about 
the chemical compositions of their con- 
stituent minerals. It has been pointed out 
(Chayes, Annual Report 1989-1990, p. 40- 
42) that whenever such information is in 
fact available for an analyzed specimen, it 
would be a simple matter to adjust the Fe 
oxidation ratio inversely, by using the fa 
content of modal olivine, or some multiple 
of it, as an upper limit on the normative 
ratio fa/ ol. What does one do, however, if, 
as is still true of many rocks and will always 
be true of some, no pertinent mineralogical 
information is available? 

Rather than fall back on a single rule for 
all rocks, one might then prefer an inverse 
adjustment based on the average/a/0/ con- 
tent found from more complete analyses of 
rocks similar to that in question. This note 
presents relevant summaries drawn from 
IGBADAT. Of the 14,722 specimen de- 
scriptions in the current version of that 
base, 11,294 are accompanied by "com- 
plete" bulk analyses, which sum to be- 
tween 95 and 105% and include determina- 
tions for both oxides of Fe. The cumulative 
frequency distribution offal ol in the 4,589 
of these that are 6>/-normative is shown by 
the upper line in Fig. 19. The lower line 
shows the same information for the 3,476 

1 In this note, "olivine" denotes the minerall, ol and 
fa the normative components computed from the 

analysis 




0.3 0.4 0.5 0.6 

Lower Class Mark 



Fig. 19. Cumulative frequency distributions of fa/ 
ol in analyses with summations in the range 95- 
105 and containing analytical determinations for 
both oxides of Fe. Data from IGB ADAT4 (4,589 
analyses in all, 3,476 with H2O <2%, class width 
= 0.05). 



which also contain less than 2% of H2O. 
Values of fa/ ol in excess of 0.5 comprise no 
more than 2.5% of the entries in either data 
set. In fact, in only 6% of all values 
included in either summary is fa/ol greater 
than 0.4, and in only 12% is it greater than 
0.35. 

The mean and standard error of this 
statistic in each of a number or rock types 
(really name groups) are shown in Table 
10. Despite the limited overall range of fa/ 
ol and its rather broad within group disper- 
sion, there appear to be marked differences 
between group means; in randomly drawn 
samples several of these differences would 
be considered statistically significant. In 
fact, with the exception of the tephrites and 
gabbros the groups seem to fall into three 
sets, with mean values of fa/ol in the 
ranges, respectively, 0.28-0.31, 0.20-0.21 
and 0.11-0.13. In Table 1 the members of 
each set are listed in order of decreasing 
size of the sample available for calculation 
of the mean and its error. 

As of this writing, calculations have 
been completed only for complete groups; 
from prior experience it is anticipated that 



GEOPHYSICAL LABORATORY 

Table 10. Average and standard error of fa/ ol, by rock type, data from IGBADAT4 



43 



Rock Type 


Number of Analyses 


falol 




All 


ol >0.5% 


mean 


std. error 


Tholeiite 


339 


202 


0.306 


0.005 


Dolerite 


359 


140 


0.285 


0.009 


Diorite 


358 


78 


0.310 


0.013 


Diabase 


200 


62 


0.298 


0.015 


Andesite 


794 


41 


0.281 


0.021 


Basalt 


2,532 


1,629 


0.206 


0.003 


Gabbro 


504 


342 


0.181 


0.007 


Basanite, hawaiite, mugearite 


384 


320 


0.197 


0.007 


Trachyandesite, benmoreiite, tristainite 


246 


158 


0.208 


0.012 


Phonolite 


319 


116 


0.208 


0.019 


Mafic plutonics, ultramafic nodules 


463 


384 


0.114 


0.005 


Ultramafic volcanic s & dikes 


313 


224 


0.128 


0.008 


Tephrite 


101 


94 


0.156 


0.011 



elimination of hydrated materials will shift 
means upward slightly, and since in all 
these groups high values of falol are either 
very rare or lacking, this may well reduce 
observed differences between means. It 
should also materially reduce within group 
dispersion, however, so that its effect on the 
significance of intergroup differences re- 
mains to be seen. 

The underlying situation, alas, is prob- 
ably not as simple as Table 10 may make it 
seem. Even ignoring the evident unruli- 
ness of the tephrites and gabbros, elaborate 
and sometimes rather elliptical rationaliza- 



tions about qualifications for membership 
in some of the groups cannot be altogether 
avoided in any organization of data based 
on the actual usage of rock names rather 
than on their a priori definitions. A more 
thorough discussion of the problems this 
raises is in preparation. 



References 



Middlemost, E.A.K., Iron oxidation, norms, and 
the classification of igneous rocks, Chem. 
GeoL, 77, 19-26, 1989. 



GEOPHYSICAL LABORATORY 



45 



Igneous and Metamorphic Petrology 
Experimental studies 



Raman Spectra of High-Temperature 

Silicate Melts: NA2O-S1O2, K2O-S1O2, 

and L12O-S1O2 Binary Compositions 

John D. Frantz and Bjorn O. My sen 

The structure of silicate liquids, deter- 
mined at high temperature, and relation- 
ships between structure and properties are 
centrally important to our understanding of 
natural magmatic processes. Principally 
from studies of quenched melts, for the 
compositional range of most natural mag- 
matic liquids, a simple equilibrium of the 
form; 

T 2 5 (2Q3) *=> T0 3 (Q 2 ) + TOiiQ 4 ), 1 (1) 

describes the principal elements of the struc- 
ture (e.g., Virgo et ai, 1980; Matson et al., 
1983; Stebbins, 1988), where T represents 
tetrahedrally coordinated cations. From 
spectroscopic data on quenched binary 
metal oxide-silica melts (glasses), it ap- 
pears that equation (1) shifts to the right 
with increasing Z/r 2 of the network-modi- 
fying metal cation. 



1 In this paper, stoichiometric units are used to 
describe the structural units. T represents tetrahe- 
drally coordinated cation(s) such as, for example, 
Si 4+ and Al-K In view of the frequent usage of Q- 
notations in the NMR literature, the equivalent 
notations are shown here for convenience. The 
superscript in the Q-notation refers to the number 
of bridging oxygens on the unit. 



Whether this relationship holds true in 
the molten state is not known. Studies 
using NMR (Stebbins, 1988; Brandriss and 
Stebbins, 1988) and Raman spectroscopy 
(Seifertertf/., 1981; My sen, 1990; Cooney 
and Sharma, 1990) indicate systematic 
structural changes with temperature. There- 
fore, even though the spectra of glasses and 
melts qualitatively resemble one another, 
at least for the limited compositions stud- 
ied, structural data from glasses may not be 
used to characterize relationships between 
melt properties and melt structure. In-situ, 
high-temperature spectra of melts are re- 
quired. Measurements of the Raman spec- 
tra of melts in their molten state to tempera- 
tures of above 1600°C are now possible. In 
the present study, the effect of temperature 
on the spectra of the Na20-Si02, K2O- 
Si02, and Li02-Si02 are investigated. 

The integration of a micro-heating stage 
with the focusing capability of the micro- 
Raman system is fundamental to the suc- 
cess of the proposed research. The heater/ 
thermocouple is fabricated from pieces of 
0.8-mm Pt and Pt9oRhiO wire which are 
welded and flattened to 500 jum thickness 
at the join. A 1-mm diameter hole is drilled 
through the Pt-PtooRh 10 junction (Fig. 20). 
In the heating stage, the Pt-PtooRhio ther- 
mocouple serves a dual function as both 
thermocouple and heater (Ohashi and 
Hadidiacos, Year Book 75, 828-834). The 
heater responds to applied power in a mat- 
ter of seconds. The temperature increments 



46 



CARNEGIE INSTITUTION 




Fig. 20. Microphotograph of heater with glass 
sample in place as indicated. 



reported here were accomplished within 
10-30 seconds. After a sample is melted 
into the hole it is held in place by surface 
tension during high temperature experi- 
ments. Sample thickness near the edge is 
about 500 Jim, and thicker in the center. 
Although the emf from the thermocouple 
yields an approximate temperature (uncer- 
tainties are introduced by electrical con- 
nections between dissimilar metals and al- 
loys), accurate temperature calibration is 
achieved by suspending a 0.04 mm Pt- 
Pt9oRhio thermocouple into the melt. The 
temperatures measured with this design are 
accurate to within 4 °C anywhere within the 
sample. The microheater was then placed 
on the microscope stage of a custom de- 
signed microscope port of a Dilor XY con- 
focal micro-Raman system equipped with 
an EG&G Model 1433-C cryogenic CCD 
detector. Small chips (~ 1 mg) of the glass 
were placed over the hole in the microheater 
(Fig. 20) and melted. The samples were 
excited with the 488-nm line of a Spectra 
Physics model 2025 Ar+ ion laser, operat- 



ing near 850 mW at the sample. 

The compositions 2 chosen for high-tem- 
perature spectroscopy were on binary M2O- 
Si02 joins, with M = Li, Na, andK. These 
three different metal cations were chosen 
so as to evaluate the influence of electronic 
properties of the alkali metal on the tem- 
perature-dependence of their structure. 
Relationships between bulk melt polymer- 
ization (NBO/Si) and the temperature de- 
pendence of the structure were addressed 
by changing the Si/O in each binary sys- 
tem. 

The widest range of metal oxide/silicon 
compositions were studied in the system 
Na20-Si02 (Fig. 21) as the problem of 
glass devitrification appeared to be mini- 
mal in this system. The high-frequency 
region (800-1300 cnr*) of the NS7, NS5, 
NS3, and NS2 glass spectra (marked 25 in 
Fig. 21a, b, c, d) are similar to those re- 
ported by Mysen et al. (1982). There is an 
approximately 1 00 cm - 1 wide (full width at 
half height) band centered near 1 100 cm -1 . 
A high-frequency shoulder near 1150 
cm - * can be discerned in the spectra of NS7 
and NS5 glasses. All spectra exhibit a dis- 
tinct band near 950 cm -1 , the intensity of 
which visually increases systematically 
with increasing Na/Si (increasing NBO/Si 
of the melt). The glass spectra have been 
deconvoluted previously (Mysen et al., 



2 These compositions are designated by the molar 
ratio of M2O to Si02, so that NS7, NS5, NS3, and 
NS2 are compositions Na207Si02, Na20»5Si02, 
Na 2 0*3Si02, and Na 2 0«2Si0 2 , and KS5, KS4, 
LS3, and LS2 are K 2 0»5Si0 2 , K 2 0»4Si0 2 , 
Li2O3Si02, Li20»2Si02, respectively. 



GEOPHYSICAL LABORATORY 



47 



6^ 
CO 

c 

CD 



200 



150 



100 V 



.'4 

r* •: # - v.- .;% 

fitist < 



NS7 

T| iq :1367°C 
T n :820°C 



200 




900 1000 1100 1200 

Wavenumber, cm" 



CO 

c 

CD 



B 



150 



100 



50 



*t 



.*** V 
WAS 



NS5 



T 



liq 



1255°C 
945°C 



% 







^■N. ^ ^y v*v .*>.>.% v s «*. ■-< y ■.* t 

•. •x : :-;.. : :; ; ::v ..;**■ V-. .- *>. • 




800 900 1000 1100 1200 

Wavenumber, cm" 1 



1300 



CO 

c 

CD 



200 - 



150 - 



NS3 

T„ q : 805°C 

T g : 445°C 




100 



800 900 1000 1100 1200 

Wavenumber, cm" 



1300 



co 

c 

CD 



200 - 



150 



100 



D 



VX ff£f*jJST 






NS2 

T, iq : 870'C 

T g : 490 °C 



l\%\... ...... w ,.. %v 1 1 55 




900 1000 1100 

Wavenumber, cm" 



Fig. 21. In-situ, high-temperature Raman spectra of glasses, supercooled melts, and melts on the join 
Na20-Si02 as a function of temperature (numbers on the right side of each spectrum represent 
temperature in °C). The liquidus, Tug, and glass transition, To, temperatures for each composition are 
indicated on the figures. The intensities are calculated relative to the greatest intensity within each 
spectrum. A - composition NS7 (Na20»7Si02, bulk melt NBO/Si=0.28); B - NS5 (Na20»5Si02, bulk 
melt NBO/Si=0.4); C - NS3 (Na 2 03Si0 2 , bulk melt NBO/Si=0.67); D - NS2 (Na 2 02Si0 2 , bulk melt 
NBO/Si=l). In each panel, the spectra from successively higher temperature are offset by 10% for 
clarity. 



48 



CARNEGIE INSTITUTION 



100 



80 



60 

'55 

c 

£ 

S 40 



20 



800 



NS2 



& 



Na 2 0-Si0 2 glass 
25°C 



[usa\ Jz 

; >NS5 N st. 
* .?.,*«*{«'. * 
* ,<NS7* 

"ft L_ 






c 
c 




900 



1000 



1100 



1200 



1300 



Wavenumber, cm" 



100 



80 



A Na 2 0-Si0 2 melts 
*\| 1144-1165°C 








• w 

NS2 • ; 


• 


60 


- #% . * 






.•; :* 


• .i 




*. »* 


• • 




• if 


■ • 






• ■ • 










. #• 


• •• 


40 


# : * -• 


9 \ 
• .i 




r : V7 


• . t 










.-. • 


•% 




« .„-. 


• > 




* NS3 i . 


• •:•• 


20 


- • >S /: 


*. ** 




; ; rHE&*$ •' 






.; f :J*v^- • 


t * 




.. ' f JT • • .' 


\ 'i 





•fr' ? *\ NS7 / 


vs. 



800 



900 



1000 



1100 



1200 



1300 



Wavenumber, cm" 



Fig. 22. Comparison of room temperature Raman spectra of glasses on the join Na20-Si02 (A) with 
spectra of the same materials at temperatures between 1144°C and 1165°C (B) for the sample 
compositions as shown in Fig. 21. The temperatures for all but the NS7 samples are above the liquidus 
temperature. For the NS7 sample, the temperature of spectra acquisition is above the glass transition 
temperature (see Fig. 21 for glass and liquidus temperatures of these samples). The intensities are 
calculated relative to the greatest intensity within each spectrum. 



1982) and the 950, 1100 and 1150 cm-1 
bands have been assigned to Si-0 stretch 
vibrations in specific structural units (see 
Virgo et al, 1980; Furukawa et ai, 1981; 
McMillan, 1984, for discussion of band 
assignments). The sharp band near 950 
cm -1 is assigned to Si-O" stretching in 
structural units with 2 nonbridging oxygens 
per silicon (Q 2 ). The main band centered 
near 1 100 cm -1 is due to Si-O" stretching in 
units with NBO/Si = 1 (1100 cm 1 band; 

3 For convenience, in this paper, we will refer to 
the 950, 1 100, and 1 150 cm" 1 bands as the Q 2 , iP 
and Q^ bands 



Q3). The high frequency shoulder on the 
1 100 cm - * band at approximately 1 1 50 cnr 
1 results from the presence of fully poly- 
merized units (Q 4 ). Thus, all the glass spec- 
tra from the Na20-Si02 system are consis- 
tent with the existence of structural units 
with their individual NBO/Si values = 2 
(Q 2 or Si03 2 " units), 1 (Q 3 or Si20s 2 " 
units) and (Q 4 or Si02 units) 3 . There is no 
evidence for structural units less polymer- 
ized than NBO/Si = 2 in these spectra, in 
accord with previous Raman and NMR 
data (e.g., Virgo et al. 1980; My sen et al, 
1982; Stebbins, 1987). Equation (1) can be 



49 



100 i_ 



80 



60 



CO 

c: 



40 



20 



:t» 



# 



NS5 




NS5 & KS5 
25°C 






%• 




/ 



'55 

c 

"E 



1100 



1200 



1300 



Wavenumber, cm" 



100 



80 



60 - 



B 



« # „ • 


NS5 & KS5 
1 370-1 380°C 


• • 

• 


T^ 5 :1255°C 
Tf 5 : 745°C 


1; 


T^ 5 : 940°C 


. 


Tg S5 : 540°C 





• 


"»• 






• 


» • 






6 * 


^t 




40 


v # 


'" • 






S 


, • 






^ * 


"• 






* 


" » 






.3= * 


\\ 






r. • 


o • 




20 


NS5 » w ; 


»J 






*4*.Vv .# 


*_v 












"e KS5 * 

o tft&ffW' 

i ■. •* i i 


\ 


v&f20»K 



800 900 1000 1100 1200 

Wavenumber, cm' 



1300 



Fig. 23. Comparison of room temperature (25°C - A) and high temperature (1370-1380°C - B) spectra 
of KS5 (K 2 05SiC>2) and NS5 (Na 2 0»5Si02) glasses (A) and melts (B). The bulk melt NBO/Si for both 
samples is 0.4. The intensities are calculated relative to the greatest intensity within each spectrum. 



used, therefore, to describe the anionic 
equilibria for these compositions. 

Increasing temperature has three effects 
on the Raman spectra. (1) The band near 
950 cm -1 (Q 2 band) relative to the band 
near 1100 cm-l increases in intensity as a 
function of increasing temperature (Fig. 
21 A through 2 ID). An increase of about 
60% from the 25° C spectrum to the high- 
est-temperature spectrum was determined 
for all four compositions (Fig. 22). A simi- 
lar increase in intensity with temperature 
was also noted by Seifert tt al. ( 1 98 1 ) in the 
high-temperature spectra of (Na2Si205)85- 
(Na2(NaAl)205)i5 melt. (2) The intense 
envelope centered near 1100 cm- 1 broad- 



ens with temperature (from slightly less 
than 100 cm -1 in spectra of room tempera- 
ture glass to more than 1 20- 1 30 cm - 1 in the 
spectra of melts and supercooled liquids. 
The -1150 cm -1 shoulder evident in the 
NS5 and NS7 room temperature spectra is 
not so clear. (3) The frequency of the -950 
cm -1 band as well as the envelope centered 
near 1100 cm -1 , shifts to slightly lower 
temperature. 

The relationships between temperature 
and electronic properties of the network- 
modifying metal cation (K, Na, and Li) are 
illustrated in Figs. 23 and 25. At room 
temperature (Fig. 23A), the spectrum of 
NS5 glass shows a more distinctive Q 2 



50 



CARNEGIE INSTITUTION 



1001- 



C 



NS3 & LS3 
25°C 




100 



80 



60 



1 40 



1300 



B 



/"A Nf 

. * • 1 1 

» : 1 • 



NS3 & LS3 
55-1165°C 



T^ S3 : 805°C 



LS3 / 



NS3 

g • 

LS3 

ii q : 



»: T 9 

1: 

\\ 

ft 



LS3 



445 °C 
1224°C 
725°C 



/ ns3 / ;\ 

Q EteJL ■ i i ^^i 



Wavenumber, cm 



800 900 1000 1100 

Wavenumber, cm 



1200 

-1 



1300 



100 1- 



80 - 



■ *— • 
CO 

c 

CD 



60 - 



40 - 



20; 



800 



pc 


i 


rV. NS2 & LS2 




9 
• 
r 
• 


:.: 25'C 

• • 


- 


/ 






7 
• 
• 


: »• 
•• 


- 


/ .' 


•• 

•• 
• 




,Als2,' : 


•• 
•• 
•• 


i 

9 


% * 1 


V 


— • 


v 


• 


« 


ft < 


> 

• 


:* 


\ j 


\ 


•7 


*NS2/ 


• 
9 


1 

I t 1 


1 


I I l 



100 



80 



60 



CO 

I 40 



20 



900 1000 1100 

Wavenumber, cm 



1200 
-1 



1300 



D 






NS2 & LS2 
1205-1 21 5°C 



LS2," J 

X / / 

* 1/ / 

? 'J 

: Vns2 



* 



ST 



7 

\ 



NS2 
Iiq • 

NS2. 

g : 

LS2. 
Iiq ■ 

LS2 



870°C 
490*C 
1033°C 
600°C 



y 



_ ^*^^ j 



800 900 1000 1100 



Wavenumber, cm 



1200 
-1 



1300 



Fig. 24. Comparison of Raman spectra of glasses and melts in the systems Li20-Si02 and 
Na20 for trisilicate (bulk melt NBO/Si = 0.67) (A - room temperature and B - 1 1 55- 1 1 65°C) 
and and disilicate (bulk melt NBO/Si = 1 .0) compositions (C - room temperature, D - 1 205- 
1215°C). Symbols: NS3 - Na 2 0«3Si0 2 , LS3 - Li 2 0»3Si0 2 , NS2 - Na 2 0«2Si02, LS2 - 
Li 2 0»2Si0 2 . Relevant liquidus and glass transition temperatures are indicated on the 
figures. The intensities are calculated relative to the greatest intensity within each spectrum. 



band than that of KS5 (in both glasses, the 
bulk melt NBO/Si = 0.4). When comparing 
the spectra of LS3 and NS3 (NBO/Si = 
0.667) and LS2 and NS2 (NBO/Si = 0.14) 
(Figs. 24A and 24C), the intensity of the Q 2 
band is again more pronounced in the spec- 



tra of glasses with the smallest metal cation 
(Li). This increased intensity of the 950 
cm -1 band would indicate enhanced abun- 
dance of Si03 2_ (or Q 2 ) structural units. 
High -temperature spectra of KS5 and 
KS4 melts reveal the same broadening of 



GEOPHYSICAL LABORATORY 



51 



250 i- 



200 - 



150 



m 



a) 

I 100 



50 




KS5 

T, iq :940°C 
T n : 536° C 











900 1000 1100 

Waven umber, cm 



1200 

-1 



1300 



250 - 



200 - 



KS4 

T liq : 772°C 
: 423°C 



o 



& 150 P 



05 



— 100 




800 900 1000 1100 

Wavenumber, cm 



1200 

1 



1300 



Fig. 25. Raman spectra of KS5 (K20»5SiC>2) glasses and melts (A) and KS4 (K20«4SiC>2) glasses and 
melts (B) as a function of temperature (°C) as indicated on the right side of each spectrum. The intensities 
are calculated relative to the greatest intensity within each spectrum, and spectra at successively higher 
temperatures are offset by 10 % for clarity. 



the 1100 cm -1 envelope in the KS5 and 
NS5 spectra (Fig. 25). The KS5 spectrum 
does not show evidence for a 950 cm -1 
band at any temperature studied, whereas 
that of NS5 does with its intensity 
increasing with temperature. In spectra of 
the KS4 composition, the 950 cm -1 band is 
present in the room temperature spectra as 
a very weak band or shoulder in the glass 
spectrum, and shows a distinctive intensity 
increase with increasing temperature (Fig. 
25B). Thus, it would appear that Si03 2_ 
structural units are no longer discernible 
(within the sensitivity of the spectroscopic 
technique) in glasses and melts on the K2O- 
Si02 join for compositions with bulk melt 
NBO/Si of 0.4. Increased temperature (at 
least to 1 380°C) does not change this obser- 
vation. In contrast, in the Na20-Si02 sys- 
tem, all structural units appear to be present 
at all temperatures with melts at least as 



polymerized as NS7 (NBO/Si = 0.28). 

The comparison of the glass and melt 
spectra of the tri- and di-silicates of Na and 
Li (Fig. 24) reveal (1) that the intensity of 
the Q 2 band in LS3 glass is greater by a 
factor of about 3 compared with that of 
NS3 (Fig. 24 A), whereas at high tempera- 
ture in the molten range, the difference has 
decreased to about a factor of 2 (Fig. 24B). 
These relative intensity changes can also 
been seen in the glass and melt spectra of 
LS2 and NS2 composition (Fig. 24C,D). 
(2) In the 1100 cm -1 envelope, the maxi- 
mum is at lower frequency in the spectra of 
the lithium samples than in the sodium 
samples and there might be a slight in- 
crease in frequency difference as the the 
glasses are transformed to melts. 

Previous deconvolutions of this high- 
frequency envelope (e.g., My sen et al., 
1982; Mysen, 1990) demonstrated that the 



52 



CARNEGIE INSTITUTION 



1150 cm- 1 band (Si02 or Q 4 ) on the high- 
frequency limb of the 1100 cm -1 band 
always increased when the intensity of the 
Q 2 band increased. These increases were 
accompanied by a concomitant decrease in 
the 1100cm- 1 (Si205 2 " or Q3) band inten- 
sity. Observations such as these, also con- 
sistent with interpretation of 29 Si NMR 
spectra (e.g., Stebbins, 1987), lead to the 
conclusion that in metal oxide silicate 
glasses whose anionic equilibrium can be 
described with equation ( 1 ), increased abun- 
dance of Q 2 (or Si03 2_ ) structural units is 
always accompanied with an increase in 
Q 4 (Si02) and a decrease in Ql (Si205 2 ) 
units. The spectra of glass, supercooled 
liquid, and liquid for Na20-Si02 and K2O- 
Si02 show visual evidence for a shift of 
equation (1) to the right with increasing 
temperature. In the absence of detailed 
statistical deconvolution, the evidence for 
the glasses and melts in the Li20-Si02 
glasses and melts is less obvious. It would 
appear, from comparison of NS3 with LS3 
and of NS2 with LS2 glass and melt spectra 
that increasing temperature has less effect 
on the spectra of Li-silicates than on those 
of the Na-silicates. Visually, the intensity 
near 950 cm -1 in the Li-silicate spectra 
does not change significantly with tem- 
perature, whereas those of NS3 and NS2 
do. The intensity difference between the 
two sets of spectra decreases, therefore, as 
the glasses are heated and eventually melted. 
In summary, high-quality Raman 
spectra of silicate melts can be recorded in- 
situ at magmatic temperatures and above 
with sample acquisition times on the order 
of one minute or less. From such spectra, it 
has been found that in binary alkali metal- 



silica systems in the bulk melt polymeriza- 
tion range between 0.28 and 1.0, glasses, 
supercooled melts, and melts in the tem- 
perature range 25-1475°C generally con- 
sist of coexisting Si032" (g 2 ), Si205 2 " 
(Q3), and Si02 (Q 4 ) structural units. In 
potassium-bearing systems, the upper NBO/ 
Si limit for Si03 2 ~ units probably is at 
NBO/Si between 0.4 and 0.5. No tempera- 
ture-dependence of this limit was observed. 
No additional units were identified within 
this temperature range. For compositions 
with the Z/r 2 of the alkali metal ranging 
from 2.8 to 0.6 (Li, Na, and K), equation (1) 
shifts to the right with increasing tempera- 
ture. The spectra probably indicate a com- 
positional dependence of the free energy 
for reaction (1). Qualitatively, the effect of 
temperature on equilibrium (1) decreases 
as the Zjr 2 of the metal cation increases 
(Li>Na>K) , but quantitative evaluations of 
its values have not been been carried out. 



References 



Brandriss, M. E., and J. F. Stebbins, Effects of 
temperature on the structures of silicate liq- 
uids: 29 Si NMR results, Geochim. Cosmochim. 
Acta, 52, 2659-2669, 1988. 

Cooney, T. F., and S. K.Sharma, High temperature 
Raman spectral study of Ge02 andRb4SigOi8 
crystals, glasses and melts, EOS, 71, 1672, 
1990. 

Furukawa, T., K. E. Fox, and W. B. White, Raman 
spectroscopic investigation of the structure of 
silicate glasses. HI. Raman intensities and 
structural units in sodium silicate glasses, J. 
Chem.Phys., 153, 3226-3237,1981. 

Matson, D. W., S. K. Sharma,and J. A. Philpotts, 
The structure of high-silica alkali-silicate 
glasses — A Raman spectroscopic investiga- 
tion, J. Non-Cryst. Solids, 58, 323-352, 1983. 

McMillan, P., A Raman spectroscopic study of 
glasses in the system CaO-MgO-Si02, Amer. 
Mineral., 69, 645-659, 1984. 



GEOPHYSICAL LABORATORY 



53 



Mysen, B. O., The role of aluminum in depoly- 
merized, peralkaline aluminosilicate melts in 
the systems Li20 - AI2O3 - Si02, Na20- 
Al203-Si02 and K 2 0-Al 2 03-Si02, Amer. 
Mineral., 75, 120-134, 1990. 

Mysen, B. O., D. Virgo, and F. A. Seifert, The 
structure of silicate melts: Implications for 
chemical and physical properties of natural 
magma, Rev. Geophys., 20, 353-383, 1882. 

Seifert, F. A., B. O. Mysen, and D. Virgo, Struc- 
tural similarity between glasses and melts rel- 
evant to penological processes, Geochim. 
Cosmochim. Acta, 45, 1879-1884, 1981. 

Stebbins, J. F., Effects of temperature and compo- 
sition on silicate glass structure and dynamics: 
Si-29 NMR results, /. Non-Cryst. Solids, 106, 
359-369, 1988. 

Stebbins, J. F., Identification of multiple structural 
species in silicate glasses by 29 Si NMR, Na- 
ture, 330, 465-467, 1987. 

Virgo, D., B. O. Mysen,and I. Kushiro, Anionic 
constitution of silicate melts quenched at 1 atm 
from Raman spectroscopy: Implications for 
the structure of igneous melts, Science, 208, 
1371-1373, 1980, 



Peralkalinity and H2O Solubility 
Mechanisms in Silicate melts 

Bjorn Mysen 

Relationships between properties of 
magmatic liquids and their water content 
have been examined since the pioneering 
work of Bowen (1928). Despite an exten- 
sive literature on the subject, the detailed 
nature of the interaction between H2O and 
the silicate melt structure remains unclear, 
and characterization of the relations be- 
tween the structure of hydrous silicate melts 
and their physical and chemical properties 
remains a topic of intense interest. 

The principal anionic equilibrium that 
describes the melt structure in magmatic 
liquids under anhydrous conditions is (Virgo 
etal, 1980; Mysen et al., 1982; Matsonef 
al, 1983; Stebbins, 1987) 



T2O5 2 - <^> TO3 2 - + IO2. 



(1) 



Dissolved H2O can interact not only with 
TO2. but with all the structural units in the 
melts. A study has been conducted, there- 
fore, under conditions of constant degree 
of bulk melt polymerization (NBO/T of 
anhydrous melts is 0.5) with Al/(A1+Si) 
and water content as the compositional 
variables. The NBO/T value corresponds to 
magma compositions intermediate between 
tholeiite and andesite (Mysen, 1988). The 
range in Al/(A1+Si) (0-0.3) covers that found 
in most natural magmatic liquids. 

Starting materials were mixtures of 
Na2C03+Al203 + Si02 on thetetrasilicate 
(Na2Si409)-tetra-aluminate 
[Na2(NaAl)409] join (Fig. 26). Exchange 




Fig. 26. Composition of anhydrous starting mate- 
rials superimposed on simplified liquidus phase 
relations in the system Na20-Al203-Si02 (from 
Osborn and Muan, 1960). 



54 



CARNEGIE INSTITUTION 



Anhydrous 




850 1075 1300 

Wavenumber, cm" 1 



- 100r 



7.5wt%H 2 




830 1065 1300 

Wavenumber, cm -1 



Fig. 27. Curve-fitted Raman spectra of composi- 
tions indicated. The V950 and vnoo bands are 
shaded for clarity. 



of Al 3 + for Si 4+ does not affect the bulk 
melt polymerization (NBOIT = 0.5) under 
anhydrous conditions because both Al 3+ 
and Si 4+ occupy tetrahedral coordination 
(Mysen, 1990). 

The glass starting materials were from 
the same batch of glasses used by Mysen 
(1990). About 20 mg of finely crushed 
glass together with distilled, deionized H2O 
was placed in sealed Pt containers for high- 
pressure synthesis. All water contents (<7.5 
wt %) were less than that needed to saturate 
the melts atthe 12kbarand 1400°Cusedfor 
sample preparation. These samples were 
subjected to 12 kbar at 1400°C in the solid- 
media, high-pressure apparatus (Boyd and 
England, 1960) for 90 min and tempera- 
ture-quenched at a quenching rate near 
100°C/s between the experimental tem- 
perature and ~500°C. The quenching rates 
were similar for all materials studied. The 
pressure uncertainty (as calibrated against 
the quartz <=> coesite and albite <=> jadeite + 
quartz transitions) is near ±1 kbar. The 
effect of pressure on the Pt-Pt9oRhio ther- 
mocouples is about ±10°C (Mao etal., Year 
Book 70, p. 281-287). With no pressure 
correction on the emf output from this 
thermocouple, the temperature is consid- 
ered accurate to ±10°C. 

Structural information was obtained 
from analysis of Raman spectra of the 
quenched melts. The spectra were recorded 
with an automated single-channel Raman 
spectrometer system with the frequency- 
doubled 532-nm line of an NdYAG laser 
operating at 1 W for sample excitation. 

The abundance of the structural units 
was determined from the Raman spectra, 
as described by Mysen {Annual Report 



GEOPHYSICAL LABORATORY 



55 



1988-1989, p. 47-54). From deconvoluted, 
high-frequency Raman spectral envelopes 
such as illustrated with Figure 27, the rela- 
tive intensities of the V950 [(Si, Al)-0" stretch 
band from TO3 2 - structural units] and vi 100 
bands (Si, Al)-0 stretch band from T2O5 2 - 
structural units] were used for this purpose. 
The distribution of Al among those units as 
well as the fully polymerized structural 
unit (7T)2.; V1150 and V1200 bands) was 
evaluated from frequency shifts of the indi- 
vidual bands as a function of H2O content 
and bulk melt Al/(A1+Si). 

There are systematic shifts in frequency 
of important Raman bands with increasing 
water content and with increasing bulk Al/ 
(Al+Si) (Fig. 28). The general descent of 
the vi 150 and V1200 frequencies with in- 
creasing H2O, also noted for hydrous 
NaA102-Si02 melts (Mysen and Virgo, 
1986), indicates that Al/(A1+Si) of the 
fully polymerized structural units (TO2.) in 
the melts is positively correlated with H2O 
concentration. It is notable, however, that 
even in anhydrous samples, the V1150 and 
vi 200 frequencies at given Al/( Al+Si) are 
lower than those observed in the spectra of 
fully polymerized (NBO/T = 0) NaA102- 
Si02 quenched melts with the same bulk 
Al/(A1+Si) (Fig. 28). If the frequency of 
these bands were used as a quantitative 
measure of Al/( Al+Si) in the 7X)2. units, 
[Al/Al+Si)]7T)2. in the anhydrous sodium 
aluminosilicate melts would be 0.32, 0.34, 
and 0.37, for NS4-A10 [bulk melt Al/ 
(Al+Si) = 0.1], NS4-A20 [bulk melt Al/ 
(Al+Si) = 0.2], and NS4-NA30 [bulk melt 
Al/( Al+Si) = 0.30], respectively. The er- 
rors in these numbers, however, are quite 
large due to the shallow slope of the fre- 



1300 



§ 1200 



900 




Anhydrous 



_» 1 i- 



0.0 0.1 0.2 0.3 0.4 0.5 0.6 
AI/(AI+Si) 



1300 



B 

> - » 



o 1200CT-- 



tf^- - -O 



o 

§ 1100, 
ex 

CD 

1000 




3 wt% H2O 



900 



0.1 0.2 .0.3 0.4 
AI/(AI+Si) 



1300 




1000 



900 



5wt%H 2 



0.0 0.1 0.2 0.3 0.4 
AI/(AI+Si) 



Fig. 28. Frequencies of selected (indicated on 
Figure) Raman bands as a function of bulk melt 
Al/(A1+Si) and H20.The bands shown are as- 
signed to (Si,Al)-0° stretch vibrations in fully 
polymerized structural units. The dashed line is 
from the system NaA102-Si02-H20 (data from 
Mysen and Virgo, 1986). 



56 



CARNEGIE INSTITUTION 



1.0 
0.8 



A 



NS4;AI/(AI+Si)=0; B r NS4-A10; Al/(AI+Si)=0. 




6 2 

wt% H 2 



8 



Fig. 29. Abundance of structural units (as shown on figure) as a function of H2O content 
and Al/(A1+Si). Open symbols are results from samples with D2O. 



quency vs. Al/(A1+Si) curves. The results 
are, nevertheless, consistent with a distinct 
preference of Al 3+ for the most polymer- 
ized among the coexisting structural units 
in the anhydrous melts, a conclusion con- 
sistent with other Raman and NMR data 
(e.g., Mysen etal, 1981; Engelhardt etal., 
1985; Kirkpatrick et al., 1986; Oestrike et 
al., 1987; Domine and Piriou, 1986; 
Merzbacher et al, 1990; Mysen, 1990). 
Thus, the Al/(A1+Si) of the TO2 structural 
units would exceed that of the bulk melt. 
The frequency reduction of the V1150 
and vi 200 bands with increasing H2O con- 
tent leads to the suggestion that the [Al/ 
Al+Si)]7U2 in hydrous melts is enhanced 



further. The effect of water on these Raman 
frequencies is less pronounced, however, 
in the peralkaline aluminosilicate melts 
studied here than in hydrous NaA102-Si02 
melts (Mysen and Virgo, 1986). If the fre- 
quency trajectories in spectra from anhy- 
drous Si02-NaA102 quenched melt were 
employed to calibrate the Al/(A1+Si) of the 
fully polymerized units in hydrous melts, 
the minimum frequency (near 1080 cm -1 ) 
corresponds to Al/(A1+Si) > 0.4 for for 
7T)2. units hydrous aluminosilicate melts. 
The influence of water on the abun- 
dance of structural units in the melts (Fig. 
29) can be inferred from the intensity varia- 
tions of the bands. At low water contents, 



GEOPHYSICAL LABORATORY 



57 



the AT2O5 in hydrous NS4 increases (the 
melts become depolymerized) with water 
content. This hydroxy lation mechanism 
may be expressed with the equation 



2Si0 2 + H 2 » SiO(OH) 2 , 



(2) 



which when considered with equation (1) 
probably results in an increased abundance 
of S12O5 2 - accompanied by a similar de- 
crease in Si03 2_ . If this were the only 
solubility mechanism, the rate of depoly- 
merization would be 0.0133 per mol % 
dissolved H2O. 

With high H2O contents there is evi- 
dence, however, that these melts undergo a 
slight polymerization (Fig. 29 A). Hydroxy - 
lation of network-modifying Na + can be 
described by the system of equations 



2Na+ + H 2 + Si0 3 2- 

<=> 2Na(OH) + Si0 2 , 

2Na+ + H 2 + Si20 5 2 " 

«> 2Na(OH) + 2Si02, 

and 

2Na+ + H 2 + 2Si0 3 2 " 

<=> 2Na(OH) + Si 2 5 2 -, 



(3) 



(4) 



(5) 



in which the network-modifying cation 
(Na + ) reacts with water to form Na(OH). 
Thus, even in this compositionally 
simple Na20-Si02-H20 system, the spec- 
tral data are consistent with water solubil- 
ity mechanisms that include three different 
types of OH-bonding [H-OH (molecular 
H2O), Na-OH (to form NaOH complexes), 
and Si-OH. 



The abundance trends of structural units 
in the aluminous samples have four fea- 
tures in common. (1) Whether anhydrous 
or water-bearing, the abundance of fully 
polymerized structural units is several tens 
of percent higher than in the absence of Al 
(Figs. 29B-D). (2) The abundance of TO2 
passes through a maximum with water con- 
tents between 1.5 wt %, and 3 wt %. (3) 
Both depolymerized ^Os 2 " (NBO/T= 1) 
andTC>3 2 (NBO/T=2) units generally coex- 
ist with fully polymerized TO2 (NBO/T=0). 
(4) There is a minimum in abundance of 
depolymerized units (T2O5 and TO3) at 
water contents corresponding to the maxi- 
mum TO2 concentration. 

The abundance patterns of the struc- 
tural units in Al-bearing samples indicate 
that at low H2O concentration (<1.5 wt % 
H2O) solution of water results in polymer- 
ization of the melts. The solubility mecha- 
nism consistent with polymerization is in- 
teraction between Na + and H2O along the 
lines of equation (3). 

The Al-bearing melts become depoly- 
merized with H2O contents > 1 .5-3 wt % as 
the abundance of TO3 and T2O5 units is 
positively correlated with H2O concentra- 
tion, whereas that of 702. is negatively 
correlated (Fig. 29B-D). Depolymeriza- 
tion via interaction of H2O with expulsion 
of tetrahedrally coordinated Al 3+ (because 
of reduction in Na+ charge -balance, or 
Al(OH)3 formation, or both) is a principal 
structural mechanism describing this be- 
havior. 

Although Al 3 + does not reside exclu- 
sively in fully polymerized anionic units 
[the frequencies of the V95oand vi 100 bands 
are weakly dependent on Al/(A1+Si), in 



58 



CARNEGIE INSTITUTION 



particular in the low water concentration 
ranges], available information (e.g, 
Engelhardt et al., 1985; Kirkpatrick et al., 
1986; Oestrike et al, 1987; Mysen, 1990) 
indicates a strong preference of aluminum 
for the most polymerized of the coexisting 
units at least in anhydrous melt systems. 
We will discuss, therefore, the solubility 
mechanism on the basis of all Al in such 
fully polymerized units and refer to them as 
NaA102. The principles derived from that 
discussion will not, however, be affected 
by some Al 3+ in more depolymerized struc- 
tural units. 

The depolymerization reactions of wa- 
ter-bearing alkali aluminosilicate melts rest 
on the premise that if Al 3 + in 4-fold coordi- 
nation in anhydrous melts forms Al(OH)3 
complexes after hydrolysis (with Al no 
longer in tetrahedral coordination), a frac- 
tion of the charge -balancing cation (Na + in 
the present system) equal to the proportion 
of Al3+ in these complexes becomes net- 
work-modifying. This mechanism could 
be described with the following set of 
equations: 

2NaA102 + 2Si02 + 3H 2 

<^> 2A1(0H) 3 + 2Na+ + Si20 5 2 ", (6) 

2NaA102 + Si02 + 3H20 

^> 2A1(0H) 3 + 2Na+ + Si0 3 2 ", (7) 

2NaA102 + Si20 5 2 " + 3H 2 

<=> 2A1(0H) 3 + 2Na+ + 2Si03 2 ". (8) 

An alternative or additional mechanism 
would operate if the charge -balancing cat- 
ion (Na+) interacts with dissolved H2O to 
form OH complexes. Then, an equivalent 



fraction of Al 3 + originally in tetrahedral 
coordination will become a network-modi- 
fying cation. 

2NaA102 + 3Si02 + H 2 

<=> 2Na(OH) + 2A13+ + Si0 3 2 ', (9) 

2NaA102 + 6Si02 + H2O 

<=> 2Na(OH) + 2A13+ + 3Si 2 5 2 -, (10) 

2NaA102 + 3Si205 2 " + H 2 

<=> 2Na(OH) + 2A13+ + 6Si03 2 ". (11) 

Manning et al. (1980) and Pichavant 
(1987) inferred from liquidus phase equi- 
libria in hydrous quartz-feldspar systems 
that this mechanism would explain their 
observations. Kohnera/. (1989) interpreted 
their NMR spectra of hydrous NaAlSi30s 
glass as consistent with NaOH complexing. 
In the case of Na-aluminosilicate melts, the 
latter mechanism (NaOH complexing) more 
efficiently depolymerizes the melt, be- 
cause for each charge-balancing Na + in 
anhydrous melts that forms the NaOH com- 
plex in the hydrous environment, three 
nonbridging oxygens can be stabilized with 
Al 3 +. For each tetrahedrally coordinated 
Al 3 + in anhydrous melts, transformed to 
Al(OH)3 in hydrous melts only one 
nonbridging oxygen can be stabilized with 
the Na + released in this process. 

These expressions are consistent with 
the observations in Figs. 29B-D, where the 
abundance of both T2O5 2 - and TO3 2 
increasesand that of TO2 decreases as the 
water concentration increases above 1.5 
wt%. Equation (5) (depolymerization) to- 
gether with equation (3) (polymerization) 
may be an adequate approximation to the 



GEOPHYSICAL LABORATORY 



59 



principal solution mechanism of H2O in 
highly aluminous, melts such as NS4-A30 
[Al/(A1+Si)=0.3]. 

Liquidus phase relations in the system 
Na20-Al203-Si02 in the compositional 
region near the Na2Si409-Na2(NaAl)409 
join can be employed to illuminate the 
relations between H2O activity and liquidus 
phase relations. Due to the lack of experi- 
mental data on necessary high-pressure 
liquidus phase relations, the 1 -bar informa- 
tion (Osborn and Muan, 1960) will be used 
for the purpose. It is recognized that by 
using the 1-bar data, significant errors are 
introduced, as it is well established that 
pressure by itself affects both liquidus tem- 
peratures and liquidus volumes in this (and 
other) systems (e.g., Boettcherera/., 1984). 
Nevertheless, it is informative to discuss 
the consequences of dissolved water only. 
Isopleths of 2.0 and 7.5 wt % were recast 
in terms of Na20, AI2O3, and Si02. The 



liquidus temperatures were corrected for 
water content by assuming ideal mixing of 
H2O in melts on the basis of the 8 -oxygen 
model of Burnham (1975) and by using the 
heat of fusion data for liquidus phases 
summarized by Richet and Bottinga ( 1 986) 
with the expression 



7(K) 



1 



-1— fail 



v melt\ 

A Hoo/ 



R 



(12) 



a o fusion 



where it was assumed that the heat of 
fusion of the relevant phases was tempera- 
ture independent. 

It is evident that with small amounts of 
water in melt solution (2 wt % H2O), the 
albite liquidus volume is greatly expanded 
at the expense of both quartz on its low-Al 
side and nepheline on its high- Al side (Fig. 
30). The small liquidus surface of crystal- 
line sodium disilicate (NS2; Na2Si20s) in 



1200 



1000 



O 

o 



£ 800 

■*— » 
co 

g. 600 

E 
a> 

•" 400 



200 



■ $\ 


— 1 

^ Ab 


1 T ■ 

""'Ab y 
NS2 


— , 1 1 1 

1 _ ^ 1 


^Ab 
1 


NS2 


• 



0.0 



0.1 



0.2 
AI/(AI+Si) 



0.3 



0.4 



Fig. 30. Calculated liquidus surfaces along the anhydrous (NBO/T = 0.5), 2 and 
7.5 wt % H20 isopleths, as a function of Al/(A1+Si). 



60 



CARNEGIE INSTITUTION 



the anhydrous system has disappeared. Ex- 
pansion of the albite liquidus volume at low 
PH 2 (-1.25 kbar) along the join 
NaAlSi308-Na2Si205 has also been docu- 
mented by experimental phase equilibrium 
results (Mustart, 1972). The liquidus sur- 
faces of fully polymerized albite and neph- 
eline and complete lack of depolymerized 
crystalline phases reflect, therefore, the 
tendency toward polymerization of the 
peralkaline aluminosilicate melts with small 
amounts of water in solution. 

With 7.5 wt % H2O in solution the 
liquidus temperatures are greatly depressed 
(Fig. 30). There is no quartz liquidus vol- 
ume, and crystalline NS2 (sodium disilicate) 
is an important phase from Al-free compo- 
sitions to Al/(A1+Si) near 0.15, where 
albite appears. The albite liquidus volume 
ranges from Al/(A1+Si) = 0.15 to about 
0.22, whereas with 2 wt % H2O the volume 
ranges from near 0.05 to about 0.26. This 
very significant expansion of the NS2 
liquidus volume at the expense of albite is 
a consequence of the depolymerization of 
the melt caused by the dissolved water. The 
activity of the nepheline component is in- 
creased relative to that of albite as reflected 
in the encroachment of the nepheline 
liquidus volume on that of albite. Even 
though the abundance of TO2 structural 
units has been lowered, most likely this 
expansion of nepheline relative to albite 
results from an enhancement of Al/(A1+Si) 
in the remaining fully polymerized struc- 
tural units. 



References 

Boettcher, A.L., Q, Guo, S. Bohlen, andB. Hanson, 
Melting of feldspar-bearing systems to high 
pressures and the structures of aluminosilicate 
liquids, Geolog, 12, 202-204, 1984. 
Bowen, N. L. The Evolution of the Igneous Rocks, 
332 pp., Princeton Univ. Press, Princeton, 1 928 
Boyd, F. R., and England J. L., Apparatus for 
phase equilibrium measurements at pressures 
up to 50 kilobars and temperatures up to 
1750°C,/. Geophys.Res., 65, 741-748, 1960. 
Burnham, C. W\, Thermodynamics of melting in 
experimental silicate-volatile systems, 
Geochim. Cosmochim. Acta, 39, 1077-1084, 
1975. 
Domine, R, and B. Piriou, Raman spectroscopic 
study of the Si02-Al203-K20 vitreous sys- 
tem: Distribution of silicon and second neigh- 
bors, Amer. Mineral, 71, 38-50, 1986. 
Engelhardt, G., M. Nofz, K. Forkel, F. G. 
Wishmann, M. Magi, A. Samoson, and E. 
Lippmaa, Structural studies of calcium alumi- 
nosilicate glasses by high resolution solid state 
29 Si and ^Al magic angle spinning nuclear 
magnetic resonance, Phys. Chem. Glasses, 26, 
157-165, 1985. 
Kirkpatrick, R. J., R. Oestrike, C. A. Weiss, K. A. 
Smith, and E. Oldfield High-resolution 27 A1 
and 29 Si NMR spectroscopy of glasses and 
crystals along the join CaMgSi206-CaAl2Si06, 
Amer. Mineral, 77,705-711,1986. 
Kohn, S. C., R. Dupree, and M. E. Smith, A 
Nuclear magnetic resonance study of the struc- 
ture of hydrous albite glasses, Geochim. 
Cosmochim. Acta, 53, 2925-2935, 1989. 
Manning, D. A. C, C. M. B. Hamilton C. M. B. 
Henderson, and M. J. Dempsey, The probable 
occurrence of interstitial Al in hydrous F- 
bearing and F-free aluminosilicate melts, 
Contr. Mineral. Petrol, 75, 257-262, 1980. 
Matson, D. W., S. K. Sharma, and J. A. Philpotts, 
The structure of high-silica alkali-silicate 
glasses — A Raman spectroscopic investiga- 
tion, /. Non-Cryst. Solids, 58, 323-352, 1983. 
Merzbacher, C, B. L.Sherriff, S. J. Hartman, and 
W. B. White, A high-resolution 29 Si and 27 A1 
NMR study of alkaline earth aluminosilicate 
glasses, /. Non-Cryst. Solids, 124, 194-206, 
1990. 
Mustart, D. A. Phase relations in the peralkaline 
portion of the systemNa20-Al203-Si02-H20 . 
Ph. D. thesis, Stanford University , 1972. 
Mysen, B. O., Effect of pressure, temperature, and 
bulk composition on the structure and species 
distribution in depolymerized alkali alumino- 
silicate melts and quenched melts, J. Geophys. 
Res., 95, 15733-15744, 1990. 



GEOPHYSICAL LABORATORY 



61 



My sen, B. O., Structure and Properties of Silicate 
Melts, 354 pp., Elsevier, Amsterdam, 1988. 
Mysen, B. O., D. Virgo, and F. A. Seifert, The 
structure of silicate melts: Implications for 
chemical and physical properties of natural 
magma, Rev. Geophys., 20, 353-383, 1982. 
Mysen, B. O., and D. Virgo, The structure of melts 
in the system Na20-CaO-Al 2 C>3-Si02-H20 
quenched from high temperature at high pres- 
sure. 2. Water in melts along the join NaAlC>2- 
Si02 and a comparison of solubility mecha- 
nisms of water and fluorine, Chem. GeoL, 57, 
333-358, 1986. 
Mysen, B. O., D. Virgo, and I. Kushiro, The 
structural role of aluminum in silicate melts — 
A Raman spectroscopic study at 1 atmosphere, 
Amer. Mineral, 66,678-701,1981. 
Oestrike, R., W.-H. Yang, R. J. Kirkpatrick, R. 
Hervig, A. Navrotsky, and B. Montez, High- 
resolution 23 Na, 27 A1 and 29 Si NMR spectros- 
copy of framework-aluminosilicate glasses, 
Geochim. Cosmochim. Acta, 51, 2199-2210, 
1987. 
Osborn, E. F., and A. Muan, Phase equilibrium 
diagrams for ceramists. Plate 4. The system 
Na20-Al203-Si02, Am. Ceram. Soc., Colum- 
bus Ohio, 1960. 
Pichavant, M., Effects of B and H2O on liquidus 
phase relations in the haplogranite system, 
Amer. Mineral, 72,1056-170,1987. 
Richet, P. , and Y. Bottinga, Thermochemical prop- 
erties of silicate glasses and liquids: A review, 
Rev. Geophys., 24, 1-26, 1986. 
Stebbins, J. F.,. Identification of multiple struc- 
tural species in silicate glasses by 29 Si NMR, 
Nature, 330, 465-467, 1987. 
Virgo, D., B. O. Mysen, and I. Kushiro, Anionic 
constitution of silicate melts quenched at 1 atm 
from Raman spectroscopy: Implications for 
the structure of igneous melts, Science, 208, 
1371-1373.,1980. 



Partitioning of Fluorine and Chlorine 

between Apatite and Non-Silicate Fluids 

at High Pressure and Temperature 



James Brenan 



mineral as a useful indicator of halogen 
activities in the geologic environment. Prior 
experimental work has focused on calibrat- 
ing apatite as a monitor of halogen activi- 
ties in the low pressure(f > )-temperature(7 T ) 
hydrothermal environment (Korzhinskiy, 
1981; Latil and Maury, 1977). Recently 
developed solution models for apatite, flu- 
ids and melts now provide a basis for theo- 
retical prediction of the halogen chemistry 
of apatite coexisting with these phases over 
a somewhat broader range of P and T 
(Candela, 1986;PiccoliandCandela, 1991; 
Tacker and Stormer, 1989; Zhu and 
Sverjensky, 1991). Although such work 
represents an important contribution to the 
interpretation of the halogen chemistry of 
natural apatite, its usefulness may be lim- 
ited to the overall low P-T range in which 
experiments were performed for both cali- 
bration and development of solution-model 
databases. In order to exploit apatite as a 
monitor of the halogen chemistry of fluids 
or melts in the high P-T environment, ex- 
perimental determination of apatite chem- 
istry at these conditions is a requisite. This 
report describes the results of experiments 
aimed at this goal with specific emphasis 
on measurements of the distribution of 
fluorine and chlorine between apatite and 
non-silicate fluids (H2O + dissolved salts ± 
CO2, carbonate melt) at P-T conditions 
appropriate to the lower crust and upper 
mantle. 



The presence of fluorine and chlorine as 
essential structural constituents of apatite, 
combined with the widespread occurrence 
of apatite amongst diverse parageneses, 
underscores the potential utility of this 



Experimental Technique 

Owing to the difficulties associated with 
analyzing fluids and low-viscosity melts 



62 



CARNEGIE INSTITUTION 



quenched from high P-T experiments, par- 
tition coefficients (D-values; wt % concen- 
tration in apatite/wt % concentration in 
fluid or melt) were determined by mass 
balance. The overall strategy was therefore 
to perform experiments by (1) encapsulat- 
ing known quantities of finely-powdered, 
natural apatite (previously well-character- 
ized in terms of fluorine and chlorine abun- 
dances) and fluid, or apatite and melt, in 
welded Pt containers, (2) equilibrating this 
mixture for 2-4 days at high pressure and 
temperature (950-1050°C, 1.0-2.0 GPa) 
with a solid-media, high-pressure appara- 
tus, then, (3) analyzing the run-product 
apatite for F and CI using an electron mi- 
croprobe. Fluids were added to these ex- 
periments as distilled H2O or as aqueous 
solutions of HC1 (1.8 and 5.3 wt %), NaCl 
(4, 15, and 25 wt %), Na2CC>3 (4 and 15 wt 
%) and NaOH (4 wt %). Water-carbon 
dioxide mixtures were produced by weigh- 
ing in distilled H2O with silver oxalate. 
The carbonate melt used in these experi- 
ments was a mixture of high-purity carbon- 
ates with the stoichiometry 80 wt % dolo- 
mite: 20 wt % Na2CCh. Water was added to 
melt-bearing experiments to yield 5-15 wt 
% in the melt. The apatite starting material 
consisted of hand-picked fragments (free 
of inclusions and visible alteration) of gem- 
quality crystals that originate at Durango, 
Mexico. Microprobe analysis of points on 
25 different apatite fragments yielded aver- 
age concentrations for F and CI of 3.57 (± 
0.13, lo) and 0.38 (± 0.06, la) wt %, 
respectively. A rastered beam was used to 
minimize mineral degradation during mi- 
croprobe analysis, and no loss of F or CI x- 
ray intensity was detected even for data 



acquisition times exceeding 4 minutes; stan- 
dard analytical conditions were a 45 nA 
sample current at 15 kV accelerating volt- 
age. 

Due to the low solubility of apatite in 
C02-H20-NaCl and dilute HC1 solutions 
(i.e.,<l wt%;Ayers and Watson, 1991), no 
correction for apatite dissolution was ap- 
plied to calculated partition coefficients 
involving these compositions. Based on 
the measurements of Ayers and Watson 
(1991), a solubility of 3 wt % apatite was 
used in partition coefficient calculations 
for 5.3 wt % HCl-bearing experiments and, 
combining the data of Baker and Wyllie 
(1990) with that of Watson (1980), a solu- 
bility of 12.5 wt % apatite was used to 
calculate D-values for experiments involv- 
ing carbonate melt. 



100 1 — ■ — 1 — ■ — 1 — ■ — 1 — ■ — ' — ■ — > — ■ — ' — ■ — ' — 

^ if- A - 2.2-2.6 Fluorme 

'=3 



.01 



.001 



*«* rwv 

ft TA A AA 

Chlorine 



m D 



03 n 




2 4 6 8 10 12 14 
Wt % CI 



Fig. 31. Apatite/fluid partition coefficients for 
fluorine (closed symbols) and chlorine (open sym- 
bols) as a function of total chlorine abundance. 
Data are from experiments at 2.0 GPa and 1050°C 
and pertain to H20-bearing fluids with and with- 
out added HC1 or NaCl (Circles, triangles and 
squares refer to experiments with H2O, H2O-HCI 
and H20-NaCl, respectively).) The numbered 
contours for the fluorine partitioning data refer to 
the total fluorine contents of these experiments 
(rev = reversal). 



GEOPHYSICAL LABORATORY 



63 



Results and Discussion 

Inspection of apatite run products re- 
vealed substantial grain-growth during the 
course of an experiment: samples starting 
out as <10 Jim, angular grains underwent 
coarsening to produce subhedral to euhedral 
material of from 15 jum to mpre than 100 
Jim grain size. The amount of grain growth 
was dependent on fluid composition and 
experiments containing HC1 or Na-carbon- 
ate solutions or carbonate melt produced 
the most abundant large grains. Multiple 
microprobe analyses across individual apa- 
tite grains from run-products showed no 
evidence of compositional zoning. 

Figures 31 and 32 portray D-values for 
fluorine and chlorine as functions of total 
chlorine and fluorine concentration mea- 
sured in experiments with H20-NaCl-HCl 
fluids at 1050 °C and 2.0 GPa. (Note that F 
concentrations were varied by adjusting 
the ratio of apatite to fluid; CI contents 
were varied by changing the CI concentra- 
tion in added solutions.) Average/) -values 
for CI from experiments containing H2O or 

100 



10 



0) 

a. 1 

(0 



.01 



.001 



Fluorine 


rev 

V 










*• 








m 




r 

Chlorine 


*» A ^- reV 

v 


1 
4& 




*HCI 


D O 

T 


H 


D 

-. 


2 NaCI 


, 


. 1 


: 


Oh 2 o 



1 2 

wt%F 



Fig. 32. Apatite/fluid partition coefficients for 
fluorine and chlorine as a function of total fluorine 
abundance (symbols as in Fig.31). Data are from 
the same experiments described in Fig. 31. 



HCl-solutions are -0.1 and average values 
obtained from aqueous NaCl-bearing ex- 
periments are -0.015; no dependence on 
either absolute F or CI concentration was 
seen for either of these values. Consistent 
with previous work (Korzhinskiy, 1981; 
Latil and Maury, 1977), D-values for F are 
well in excess of those measured for CI, and 
are generally similar at a given total CI 
concentration (for similar total F contents), 
regardless of the mode of CI addition. Par- 
tition coefficients for F exhibit no system- 
atic dependence on total CI concentration 
(Fig. 31), but values do systematically de- 
crease as a function of increasing F concen- 
tration (Fig. 32). Results of a reversal 
experiment in which apatite was first equili- 
brated in H2O, then in 1.8% HC1, gave 
good agreement with D-values measured 
in forward experiments. Average partition 
coefficients for F and CI in basic solutions 
(aqueous NaOH or Na2C03) are ~5 and 
<0.02, respectively. 

Figure 33 portrays fluorine and chlorine 
D-values as a function of pressure for ex- 
periments with aqueous HC1 or NaCI at 
1050°C. D- values for CI are invariant with 
pressure, regardless of fluid composition, 
as is the D for F in the HCl-bearing fluid. 
The F partition coefficient with aqueous 
NaCI at 1.0 GPa is, however, -10 times 
higher than values determined at 2.0 GPa 
for similar total F abundances. D-values 
for CI at 950°C and 2.0 GPa (Fig. 33) are 
identical to values measured at 1050°C 
whereas F partition coefficients are 4-10 
times higher. 

Fluorine and chlorine partition coeffi- 
cients generally decrease and increase, re- 
spectively, as a function of the mole frac- 



64 



CARNEGIE INSTITUTION 



100 



10 

p 

=3 

5= 1 
0) 

■ «— » 
TO 



Fluorine 



.01 



.001 




950°C 



HCI 




Chlorine 
A — 



950°C 



1.0 1.5 2.0 

Pressure (GPa) 



2.5 



Fig. 33. Apatite/fluid partition coefficients for 
fluorine (closed symbols) and chlorine (open sym- 
bols) as a function of pressure for experiments at 
950 and 1050°C (1050°C data unlabelled). Data 
are from experiments with 2.2-2.5 wt % total 
fluorine (variable chlorine contents) with added 
HCI (circles) or NaCl (triangles). 

Hon of CO2 [X(C02>] in H2O-CO2 mix- 
tures (Fig. 34). In terms of D-values for CI, 
this relation is greatly altered for experi- 
ments in which CI was added as solutions 
with >15 wt % NaCl (Fig. 35; F partition 
coefficients are relatively unaffected by 



100 



10 ■ 



■g 

"=3 



0) . 
♦J 

03 
Q. 

Si 



.01 





1 ■ 








> 
> 






• 


• 


• 


■ 


• 


fluorine 


r 


O 


chlorine 


* 


<D 


<D ; 


> 






) 








■ 




— 1__ 


■ 


■ 


■ 



0.0 



0.2 0.4 0.6 

Mol Frac C0 2 



0.8 



Fig. 34. Apatite/fluid partition coefficients for 
fluorine (closed symbols) and chlorine (open sym- 
bols) as a function of the mole fraction of CO2 for 
experiments involving CO2-H2O mixtures. Data 
are from experiments run at 1050°C and 2.0 GPa. 



Q. 

i - 1 

O 

SI 

O 
Q 



.01 



i 



H 2 



XC0 2 = 0.3-0.5 
5% HCI 



4% NaCl 



15%NaCI 
25% NaCl 



1 2 3 

Total CI (wt %) 

Fig. 35. Apatite/fluid partition coefficients for 
chlorine as a function of total chlorine concentra- 
tion. Data are from experiments involving CO2- 
H2O fluids with mole fractions of CO2 = 0.3-0.5. 
Data labels refer to experiments with no added 
chlorine (H2O; CI is from the apatite starting 
material) or in which CI was added as 5 wt % HCI 
or 4-25 wt % NaCl. 

these fluid composition variations). As 
illustrated in Fig. 35, similar D-values for 
CI (i.e., -0.25) were obtained from experi- 
ments in which CI was added as either the 
apatite starting material, 4 wt % NaCl, or 
5.3 wt % HCI. Experiments in which CI 
was added as 1 5 or 25 wt % NaCl, however, 
exhibit marked drops in the chlorine D- 
value (i.e., to -0.065 and -0.03, respec- 
tively). Inasmuch as CI partition coeffi- 
cients were found to be independent of total 
CI content for C02-free fluids, these results 
may be somewhat surprising. The overall 
reduction in the CI partition coefficient 
could, however, be accounted for if an 
H20-rich fluid evolved as a result of fluid 
unmixing in the runs with NaCl-rich com- 
positions. The lower chlorine D- value found 
for the experiment with 25 wt % NaCl, 
compared to that with 15 wt % NaCl, may 
therefore suggest a higher proportion of 
this H20-rich fluid in the former experi- 
ments. Figure 36 portrays fluid composi- 



GEOPHYSICAL LABORATORY 



65 



NaCI 



10 




Fig. 36. Bulk composition of fluids in the system 
C02-H20-NaCl (wt %) for experiments involving 
CO2-H2O fluids in which H26 was added as 4-25 
wt % NaCI solutions (see caption to Fig. 3 5 for 
more details). The tie-lines and phase field bound- 
ary are schematic but meant to be consistent with 
the observed partitioning relations (see text). 

tions for the experiments involving mixed 
CO2-H2O fluids with 4, 15 and 25 wt % 
NaCI plotted in the C02-H20-NaCl ter- 
nary system. Also shown in this figure is a 
topology for the two-phase field that is 
consistent with the above observations. 
Although no previous experimental mea- 
surements at high P and Thave been made 
with regard to the extent of immiscibility 
for C02-H20-NaCl fluids , results at low P- 
T conditions (Popp etal., this Report) indi- 
cate that the compositional range of the 
two-fluid field in this system can be exten- 
sive. Experiments employing synthetic 
fluid inclusions are now in progress in an 
attempt to confirm this interpretation of the 
partitioning data. 

Partition coefficients for fluorine and 
chlorine between apatite and carbonate melt 
(obtained at 1050°C, 2.0 GPa) as a function 
of total wt % chlorine are shown in Fig. 37. 
Fluorine D-values are lower than those 
measured for aqueous fluids (i.e., -1.5 vs 
>5, respectively), whereas chlorine parti- 
tion coefficients (-0.07) are similar to val- 
ues measured for experiments with H2O 
and HCl-bearing solutions. Partition coef- 



0) 

E 



o 



Q- 1 



.01 



Molten Carbonate (Dol80:NaCarb20) 
5-1 5 wt % H 2 



rev 



rev 



m 



m 



Fluorine 



Chlorine 



12 3 4 5 

Wt % CI 

Fig. 37. Apatite/carbonate melt partition coeffi- 
cients for fluorine (squares) and chlorine (circles) 
as a function of total chlorine concentration (ob- 
tained in experiments run at 2.0 GPa and 1050°C). 
Data pertain to experiments with 5-15 wt % H2O 
in the melt phase. 



ficients for both F and CI show little varia- 
tion with total CI content (all experiments 
had F abundances of -2.0-2.3 wt %). A 
reversal experiment that involved a two- 
hour preheating of the sample at 1300°C 
and 2.0 GPa (thus completely dissolving 
the apatite into the melt), then a slow, 
isobaric cooling to the final run condition 
of 1050°C (and holding there for 48 hours), 
yielded identical D-values as in forward 
experiments. 



Implications for the Fluorine 

and Chlorine Content of 

Upper-Mantle Fluids 

Testimony bearing on the action of flu- 
ids in the upper mantle may be found in 
certain suites of ultramafic xenoliths that 
contain evidence for mineral replacement 
by volatile-bearing phases (O'Reilly and 



66 



CARNEGIE INSTITUTION 



100 




meg aery sts 
A H 2 0-HCI 

A CO 3 Melt 

C-bearing (xenoliths) 
□ C0 2 -H 2 0-HCI 
■ C0 3 Melt 



1 10 

Wt%F 



Fig. 38. Calculated fluorine and chlorine contents 
of high P-T fluids based on the compositions of 
mantle-derived apatites. Fluid compositions were 
calculated using the partition coefficients mea- 
sured in this study for fluids capable of dissolving 
appreciable amounts of apatite (i.e., H2O-HCI (± 
CO2) fluids or carbonate melt). Apatite composi- 
tions from mantle xenolith parageneses were ob- 
tained from Wassetal. (1980), Smith etal. (1981) 
and Exley and Smith (1982). Carbon-bearing 
apatite compositions (xenolith paragenesis) are 
from O'Reilly and Griffin (1988) and apatite 
megacryst compositions were obtained from 
Hervig and Smith (1981) and Irving and Frey 
(1984). 



Griffin, 1988) or preserve textures indica- 
tive of a free vapor (Kovalenko etal., 1987) 
or both. Associated with these petrographic 
indications for the infiltration of fluids are 
elevated concentrations of elements that 
are typically present at only low levels in 
mantle rocks (e.g., Ba, Cs, Sr and the rare- 
earth elements). Analyses of the fluorine 
and chlorine content of apatites that occur 
as a minor phase in some such fluid-altered 
rocks, combined with the results presented 
here, may thus provide constraints on the 
halogen abundances of some mantle meta- 
somatic agents. 

Inasmuch as D-values for fluorine and 
chlorine were found to depend on fluid 
composition, accurate estimates of mantle 
fluid halogen contents will, therefore, be 
contingent on a judicious choice of parti- 



tion coefficients. Based on results of the 
apatite solubility experiments of Baker and 
WyUie(1990)andAyersandWatson(1991), 
the non-silicate fluids capable of transport- 
ing the most significant quantities of apa- 
tite are molten carbonate and HCl-bearing 
aqueous solutions. By applying the D- 
values measured for those compositions (at 
1050°C, 2.0 GPa) to analyses of apatites 
from mantle parageneses (i.e., present as 
interstitial grains in xenoliths or as 
megacrysts; see caption to Fig. 38 for data 
sources), fluorine and chlorine contents of 
fluids that may have coexisted with mantle 
apatites were calculated and are portrayed 
in Fig. 38. Two of the apatites analyzed by 
O'Reilly and Griffin (1988) contain car- 
bon abundances of -0.3 and -0.9 wt % and 
thus may preserve evidence for equilibra- 
tion with a carbon-bearing fluid; partition 
coefficients for a CO2-H2O-HCI fluid 
[X(C02) = 0.3-0.5] were therefore used in 
place of the aqueous HC1 values. 

Based on the observed P-T dependence 
of D-values found in this study, equilibra- 
tion of apatite with fluids at lower P-T 
conditions than that assumed for these cal- 
culations would not effect calculated CI 
abundances but calculated F contents might 
represent maximum values. Fluid compo- 
sitions determined in the above manner 
have minimum Cl/F ratios of > 1, F abun- 
dances <1 wt % and chlorine concentra- 
tions of from ~1 to -20 wt %. The overall 
Cl-rich nature of some calculated fluid 
compositions may suggest an important 
role for chlorine-bearing fluids as agents of 
mass transport in the upper-mantle. This 
result is in concert with the measurements 
of Brenan and Watson (1991), who ob- 



GEOPHYSICAL LABORATORY 



67 



served significantly greater partitioning of 
trace elements into aqueous chloride solu- 
tions (coexisting with olivine) relative to 
experiments involving pure H2O. In addi- 
tion, both Selverstone et al. (1990) and 
Philippot and Selverstone (1991) have docu- 
mented the occurrence of trace-element 
rich brines in Alpine eclogites, and thus 
chlorine-rich fluids may also play a role in 
trace element mobilization in other high- 
pressure environments. 



References 

Ayers, J. C. and E. B. Watson, Solubility of 
apatite, monazite, zircon, and rutile in 
supercritical fluids with implications for sub- 
duction zone geochemistry, Phil. Trans. R. 
Soc. Lond. A, in press, 1991 

Baker, M. B. and P. J. Wyllie, High-pressure 
solubility of apatite in carbonate-rich melt 
(abstr), EOS, 70, 1394, 1990. 

Brenan, J. M. and E. B. Watson, Partitioning of 
trace elements between olivine and aqueous 
fluids at high P-T conditions: Implications for 
the effect of fluid composition on trace ele- 
ment transport, Earth Planet. Sci. Lett., in 
press, 1991. 

Candela, P. A., Toward a thermodynamic model 
for the halogens in magmatic systems: an 
application to melt-vapor-apatite equilibria, 
Chem. Geoi, 57, 289-301, 1986. 

Exley, R. A. and J. V. Smith, The role of apatite in 
mantle enrichment processes and in the 
petrogensis of some alkali basalt suites, 
Geochim. Cosmochim. Acta, 46, 1375-1384, 
1982. 

Hervig, R. L. and J. V. Smith, Dolomite-apatite 
inclusion in chrome-diopside crystal, Bellsbank 
kimberlite, South Africa, Amer. Mineral., 66, 
346-349. 

Irving, A. J. and F. A. Frey, Trace element abun- 
dances in megacrysts and their host basalts: 
Constaints on partition coeffcients and 
megacryst genesis, Geochim. Cosmochim. 
Acta,48, 1201-1221, 1984. 



Korzhinskiy, M. A., Apatite solid solution as 
indicators of the fugacity of HC1° and HF° in 
hydrothermal fluids, Geochem. Int., 18, 44- 
60, 1981. 
Kovalenko, V. I., I. P. Solovova, I. D. Ryabchikov, 
D. A. Ionov, O. A. Bogatikov and V. B. 
Naumov, Fluidized C02-sulphide-silicate 
media as agents of mantle metasomatism and 
megacrysts formation: evidence from a large 
druse in a spinel-lhezolite xenolith, Phys. Earth 
Planet Int., 45,280-293,1987. 
Latil, C. and R. Maury, Contribution a l'etude des 
echanges d'ions OH*, CI" et F" et de leur 
fixation dans les apatites hydrothermales,5«//. 
Soc. Fran. Mineral. Cristall., 100, 246-250, 
1977. 
O'Reilly , S. Y. and W. L. Griffin, Mantle metaso- 
matism beneath western Victoria, Australia: 
1. Metasomatic processes in Cr-diopside 
lherzolites, Geochim. Cosmochim. Acta, 52, 
433-447, 1988. 
Philippot, P and J. Selverstone, Trace-element- 
rich brines in eclogite veins: implications for 
fluid composition and transport during sub- 
duction, Contrib. Mineral. Petrol., 106, 417- 
430, 1991. 
Piccoli, P. M. and P. A. Candela, The mathemati- 
cal modeling of the halogen composition of 
the mineral apatite during first and second 
boiling (abstr), EOS, 72, 312, 1991. 
Smith, J. V., J. S. Delaney, R. L. Hervig, and J. B. 
Dawson, Storage of F and CI in the upper 
mantle: geochemical implications, Lithos, 14, 
133-147, 1981. 
Tacker, R. C. and J. C. Stormer, A thermodynamic 
model for apatite solid solutions, applicable to 
high-temperature geologic problems, Amer. 
Mineral, 74, 877-888, 1989. 
Wass, S . Y. , P. Henderson and C. J. Elliott, Chemi- 
cal heterogeneity and metasomatism in the 
upper mantle: Evidence from rare earth and 
other elements in apatite-rich xenoliths in ba- 
saltic rocks from eastern Australia, Phil. Trans. 
R. Soc. Lond. A, 297, 333-346, 1980. 
W'atson, E. B., Apatite and phosphorous in mantle 
source regions: an experimental study of apa- 
tite/melt equilibria at pressures to 25 kbar, 
Earth Planet. Sci. Lett., 51, 322-335, 1980. 
Zhu, C. and D. A. Sverjensky, A set of consistent 
thermodynamic properties for fluorapatite, 
hydroxylapatite and chlorapatite (abstr), EOS, 
72, 145, 1991. 



68 



CARNEGIE INSTITUTION 



Investigation of Fluid Immiscibility in 

the System H2O-NACL-CO2 Using Mass 

Spectrometry and Microthermometry 

Techniques Applied to 

Synthetic Fluid Inclusions 

Robert K. Popp? John D. Frantz, and 
Thomas C. Hoering 

Popp and Frantz (1990) described the 
use of synthetic fluid inclusions to define 
the limits of fluid miscibility in the system 
H20-NaCl-C02. Inclusions were trapped 
in quartz prisms contained in fluids with 
sodium chloride and carbon dioxide con- 
tents up to 1 8 and 50 wt %, respectively, at 
temperatures of 500°, 600°, and 700°C and 
pressures of 1000, 2000, and 3000 bar. The 
presence of two different types of inclu- 
sions within a single sample was inter- 
preted as evidence of immiscibility for a 
particular fluid composition, temperature, 
and pressure. Samples that entrapped im- 
miscible fluids exhibited both (1) inclu- 
sions containing a salt crystal in addition to 
a vapor bubble and (2) inclusions without 
salt crystals, but with a much larger bubble 
than in the first case. The former type 
inclusion represents the quenched high- 
density (i.e. NaCl-rich, C02-poor) phase, 
whereas the latter type represent the 
quenched low-density (i.e. C02-rich, NaCl- 
poor) phase. Inclusions that formed in ex- 
periments in which the fluid was in the 
miscible field contained only a single type 
that did not contain NaCl crystals and had 
proportionally the same ratio of bubble to 



Department of Geology, Texas A&M Univer- 
sity, College Station, Texas 77843 



total inclusion volume. This report de- 
scribes the use of mass spectrometric tech- 
niques and microthermometric measure- 
ments to define more precisely the location 
of the solvus separating the one -phase and 
two-phase regions in T-P-X space, and to 
define the chemical compositions of the 
coexisting high-density and low-density 
fluid phases. 



Mass Spectrometry 

A Finnigan 4500 quadrupole mass spec- 
trometer was modified (Fig. 39) for the 




Fig. 39. Modification of Finnigan 4500 mass 
spectrometer. See text for details. 

analyses of individual and groups of fluid 
inclusions using the general technique of 
Barker and Smith (1986), as modified by 
Frantz et al. (1989). A crushed sample of 
the quartz prism (-20 mg) was placed in a 
silica tube (B) surrounded by a resistance 
heater (A). The tube, sealed at the outer 
end, was inserted into a flight tube (C) 
extending through the existing vacuum lock 
of the mass spectrometer. The tube is 
designed to stretch out the arrival time at 
the ion source of the pulses of released 
gases resulting from inclusion decrepita- 
tions. The flight tube connects with a modi- 
fied electron beam cup (G) and focuses 
most of the released gas molecules into the 
path of the electron-beam where they are 



GEOPHYSICAL LABORATORY 



69 



390 



Temperature, °C 

400 



Temperature, °C 



410 
— I 



408 



418 



428 



H,0* 



_K_»i , — lJ-a 



h ;5jJXm^^ 



— COfe 



*K~* 



jJv. 



— C02 



.wjs~ 




' I ■ 



J I 1_ 



J I I I 1_ 



J I I I I i_ 



21000 21500 22000 22500 

Scan Number 



i i i i i i i i i i i i i i i i ■'■■ 

23000 23500 24000 24500 25000 

Scan Number 



Fig. 40. Mass spectrograms showing H20+ and C02 + intensities for fluid-inclusion decrepitations from 
(a) a quartz sample equilibrated at 500°C, 1000 bar with a fluid of composition NaCl7.8-(C02)25.2- 
H2C>67.o(wt %) and (b) a quartz sample equilibrated at 500°C, 1000 bar with a fluid of composition 
NaCl9.4-H2O80.6-(CO2)i0.0 (wt %). The spectrogram in Fig. 40a demonstrates the existance of two types 
of inclusions having different CO2/H2O ratios; the spectrogram in Fig 40b, the existence of one type of 
inclusion. 



ionized. These ions are then accelerated 
and separated by the quadrupole mass filter 
(D) and detected by the ion multiplier de- 
tector (E). A cryogenic pump (F), chilled 
with liquid nitrogen, is used to reduce the 
water background at mass 18. Because the 
mixtures of low molecular weight gases 
analyzed in this study are simple compared 
to the organic compounds often analyzed 
with the instrument, the selected ion moni- 
toring mode of the INCOS 2300 data sys- 
tem of the mass spectrometer was used. 
Only the selected masses of interest were 
measured. A total scan time of 28 millisec- 
onds for two masses was utilized. 

The mass spectrometer was used to ana- 
lyze the ratio of CO2 and H2O in the fluid 
inclusions by monitoring channels corre- 
sponding to masses 18 (H2O 4 ") and 44 
(CO2 4 "). Approximately 45,000 scans were 
collected as the samples were heated from 
200 to 600 °C at 5 °C per minute. Decrepi- 
tation of the inclusions generally occurred 
between 300 and 573 °C (the latter being 
the approximate temperature of the oc-p 
transition for quartz). A spectrogram from 



the analysis of inclusions from an experi- 
ment at composition NaO7.8-CO225.2~ 
H2O67.O equilibrated at 500°C and 1000 
bar is shown in Fig. 40a. The 2000 scans 
shown in the figure correspond to a rise in 
the sample temperature from 390° to 410°C 
over a time period of approximately 240 
seconds. The peaks, which correspond to 
the decrepitations of individual and mul- 
tiple fluid inclusions, cover a range varying 
from 10 to 15 spectrometer scans. Indi- 
vidual peaks represent only nanogram 
amounts of H2O and CO2. The arrival 
times (scan number) and shapes of the CO2 
and H2O peaks are quite similar, though the 
sensitivity for water tends to be somewhat 
less than that of carbon dioxide due to the 
tendency of water molecules to absorb on 
the surface of the flight tube. Fluid immis- 
cibility clearly exists, as evidenced by the 
large variability in the area ratios of CO2 
and H2O. In the case of fluid of composi- 
tion NaCl9.4-CO2i0.0-H2O80.6 equilibrated 
at 500°C and 1000 bar (Fig. 40b), a single 
fluid phase was present at the experimental 



70 



CARNEGIE INSTITUTION 





O 





Mean = 0.406 
Std Dev = 0.039 










40 60 80 100 



Area %, C0 2 




40 60 80 100 



Area %, C0 2 



NaCI 




H 2 10 20 30 40 50 



CO- 



Fig. 41. Histograms of the frequency of the area % CO2 [areaco2(/areaco2+areaH20)] from mass 
spectrograms resulting from decrepitation of fluid inclusions All five samples were equilibrated with 
a series of compositions at 500°C, 2000 bar in the single-phase fluid compositional region, as shown 
in the ternary diagram in the lower right. 



run conditions, because the area ratios of 
CO2 and H2O are nearly identical for all the 
peaks. 

Five samples equilibrated at 500°C and 
2000 bar containing less than 5.2 wt % 
NaCI with varying CO2 contents were se- 
lected for calibration of the spectrometer. 



Based on the optical detection of only a 
single inclusion type in each sample, at was 
concluded that the five samples have trapped 
miscible fluids. For all five samples, areas 
under corresponding CO2 and H2O peaks 
resulting from decrepitations were com- 
puted, using the INCOS 2300 software 



GEOPHYSICAL LABORATORY 



71 




10 20 30 40 50 60 

wt % co 2 

Fig. 42. Calibration curve showing the relation 
between area % CO2 measured by mass spectrom- 
etry and wt % CO2 of the equilibrated fluid for the 
samples shown in Fig. 41. The squares represent 
the mean values of the measured area percents 
with the brackets indicating one standard devia- 
tion. 

standard with the Finnigan 4500 spectrom- 
eter. Histograms of area % CO2 measured 
for the peaks are shown for the five samples 
in Fig. 41. Fig. 42 shows the mean values 
for all five samples, with their correspond- 
ing standard deviations, plotted against the 
concentrations of CO2 (in wt %) initially 
added to the experimental charges. The 
second-order quadratic least squares fit of 
these data was then used as the calibration 
curve to define the wt% CO2 in inclusions 
grown in the two-phase region, for which 
C02-contents were unknown. 

Eight samples equilibrated at 500°C and 
1000 bar demonstrate the results obtained 
for inclusions grown in the immiscible 
field (Fig. 43). Histograms labelled 1 and 2 
are from samples that trapped miscible 
fluids so that their mean C02-contents cor- 
respond closely to the original CO2 con- 
tents of the fluid, denoted by the vertical 
dashed lines. The other six histograms, 
however, correspond to samples that trapped 
immiscible fluids. In histograms 3 through 
6, the intervals exhibiting the highest fre- 



quency are generally at much lower values 
of wt % CO2 than the initial bulk composi- 
tion, but some more C02-rich intervals 
contain smaller populations. The highly- 
populated intervals of low wt % CO2 rep- 
resent the concentration of cabon dioxide 
in the sodium chloride-rich, high-density 
fluid phase. In the case of histogram 7, for 
which the original bulk fluid composition 
lies extremely close to the carbon dioxide- 
rich limb of the solvus, the interval of 
highest frequency lies in the region of the 
original bulk composition, with less popu- 
lated intervals lying at lower CO2 concen- 
trations. In this case, the highly populated 
intervals correspond to the concentration 
of carbon dioxide in the C02-rich, low- 
density fluid phase. Histogram 8 
demonstates a case in which highly popu- 
lated intervals extend from the original 
bulk composition down to those represent- 
ing quite low concentrations of CO2. Be- 
cause the two immiscible phases are likely 
to be intimately intermixed rather than sepa- 
rated into a single high-density and a single 
low-density phase within the capsule 
(Ramboz et al., 1982; Zhang and Frantz, 
1989), many inclusions trap varying pro- 
portions of the two fluids; thus, a range of 
inclusions with properties intermediate 
between the two end members is com- 
monly observed in a single sample from the 
immiscible field. The less-populated inter- 
mediate intervals result from decrepita- 
tions of inclusions containing mixtures of 
the two fluids and from simultaneous de- 
crepitations of groups of vapor-rich and 
liquid-rich inclusions. 

The use of mass spectrometry measure- 
ments resulted in an expanded two-phase 



72 



CARNEGIE INSTITUTION 










20 40 60 80 100 

Wt%C0 2 




NaCI 
/ 




l ^£v 



20 40 60 80 100 

Wt%C0 2 



H 2 Q 10 20 30 40 50 C0 2 



Fig. 43. Histograms of the frequency of the values of wt % CO2 from mass spectrograms resulting from 
decrepitation of fluid inclusions grown at 500°C, 1000 bar. The values of area % CO2 were converted 
to wt% CO2 using the calibration curve shown in Fig. 42. 




NaCI + 
Solution 



20 30 40 50 60 

Wt % NaCI 

Fig. 44. Solubility diagram (in wt %) of NaCI in 
NaCl-H20 solutions as a function of temperature 
(data from Linke, 1958, 1965). 



field relative to that obtained from only the 
optical identification of the two inclusion 
types. That is, the mass spectrometry tech- 
nique has the increased precision neces- 



sary to identify inclusions of more than one 
composition in some samples where only 
one inclusion type has been identified op- 
tically. In addition, the technique permits 
the CO2 content of the high-density fluid to 
be determined. The concentration of CO2 
in the high-density fluid is taken to be that 
of the highest frequency intervals at rela- 
tively low weight % CO2 (Fig. 43). 



Microthermometry 

In order to determine the positions of 
tie-lines connecting the compositions of 
coexisting immiscible fluids, the concen- 
tration of NaCI in the high-density fluid 
inclusions was estimated from 



GEOPHYSICAL LABORATORY 



73 



500 °C 
2000 bar 




500 °C 
1000 bar 




H 2 10 20 30 40 50 



£ Q H 2 10 20 30 40 50 C0 2 



NaCI 



NaCI 



600 °C 

1000 bar __ f 

60 




H~0 10 20 30 40 50 



700 °C 

1 000 bar 60 




10 



CO H 2° 10 20 30 40 50 



CO. 



Fig. 45. Compositions of coexisting immiscible fluids. Each tie line represents 
the compositions of the coexisting fluid as determined for the sample denoted 
by the filled circle through which they pass. See text for further details. 



microthermometry measurements made 
using a heating stage supplied by Fluid 
Inc., Denver, Colorado, In the phase dia- 
gram shown in Fig. 44, the NaCI content of 
a given bulk composition is known if the 
temperature of the univariant boundary 
between the "Solution" field and the "NaCI 
+ Solution" field is known (i.e., if the 
temperature of NaCI melting is known). 
The presence of CO2 in the fluid inclusions 
might affect the temperatures obtained from 
the phase relations in the C02-free system. 
However, the results of the mass spectrom- 



etry measurements described above sug- 
gest that the bulk concentration of carbon 
dioxide in the high-density inclusions is 
relatively low. In addition, most of that 
carbon dioxide is contained in the vapor 
phase (i.e., the bubble) at the temperatures 
of NaCI melting. Therefore, only very 
small CO2 concentrations must be con- 
tained in the liquid phase, and thus the 
effect of CO2 on the temperature of NaCl- 
melting is considered insignificant. 

The melting temperatures of NaCI crys- 
tals in the high-density inclusions were 



74 



CARNEGIE INSTITUTION 



determined for selected samples equili- 
brated at 500°C, 1000 and 2000 bar; 600°C, 
1000 bar; and 700°C, 1000 bar. Arelatively 
large variation in the melting temperature 
was observed in each individual sample, 
with the largest frequency occurring at the 
highest temperatures. The lower-tempera- 
ture measurements are obtained from in- 
clusions that trapped mixtures of the two 
immiscible fluids, and therefore contain 
lower NaCl concentrations than inclusions 
that trapped only the high-density phase. 
With knowledge of NaCl contents, ob- 
tained from the highest NaCl melting tem- 
perature measured in a given sample, and 
knowledge of the CO2/H2O ratios mea- 
sured by mass spectrometry, the chemical 
composition of the high-density fluid phase 
is completely defined. Tie lines were lo- 
cated between the coexisting fluid phases 
at 500°Cand2000bar,500°Cand lOOObar, 
600°C and 1000 bar, and 700°C and 1000 
bar (Fig. 45). To construct the tie lines, a 
straight line was drawn from the composi- 
tion of the high-density fluid through the 
bulk composition to the carbon dioxide- 
rich limb of the solvus. At a given tempera- 
ture and pressure, the slopes of the tie lines 
systematically steepen as the NaCl-H20 
binary is approached. With increasing tem- 
perature, the tie-lines become increasingly 
steeper, as a result of greater partitioning of 
sodium chloride and reduced partitioning 
of carbon dioxide between the two immis- 
cible phases. 



Summary 
The use of mass spectrometry in 



conjunction with the more routine optical 
and microthermometric techniques pro- 
vides an improved basis on which to sepa- 
rate fluid inclusions that trapped one -phase 
fluids from those that trapped two-phase 
fluids. In addition, the technique provides 
a relatively straightforward method to de- 
fine tie lines between the compositions of 
the coexisting high-density and low-den- 
sity fluids. The techniques described here 
should be applicable to other systems of 
petrologic interest, such as those contain- 
ing the volatile species CH4, N2, and HC1. 

References 

Frantz, J.D., Y. Zhang, D. D. Hickmott, and T. C. 
Hoering, Hydrothermal reactions involving 
equilibrium between minerals and mixed 
volatiles. 1 . Techniques for experimentally load- 
ing and analyzing gases and their application to 
synthetic fluid inclusions, Chemical Geol., 76, 
57-70, 1989. 

Barker, C. and M. P. Smith, Mass spectrometric 
determination of gas in individual fluid inclu- 
sions in natural minerals. Anal. Chem., 58, 
1330-1333, 1986. 

Linke. W.F., Solubilities of Inorganic and Metal- 
Organic Compounds, 1. Van Norstrand, 
Princeton, N.J., 4th ed., 1958. 

Linke. W.F., Solubilities of Inorganic and Metal- 
Organic Compounds, 2. Am . Chem . Soc . , Wash- 
ington D.C, 4th ed., 1965. 

Popp, R.K. and J.D. Frantz, Fluid immiscibility in 
the system H20-NaCl-C02 as determined from 
synthetic fluid inclusions, Annu. Rep. Director 
Geophys. Lab., Carnegie Instn. Washington, 
1989-1990,43-47, 1990. 

Ramboz, C, M. Pichavant, and A. Weisbrod, 
Fluid immiscibility in natural processes: Use 
and misuse of fluid inclusion data, II. Interpre- 
tation of fluid inclusion data in terms of immis- 
cibility, Chemical Geol, 37, 29-48 r 1982. 

Zhang, Y. and J. D. Frantz, Experimental determi- 
nation of the compositional limits of immisci- 
bility in the system CaCl2-H20-NaCl at high 
temperatures and pressures using synthetic fluid 
inclusions, Chemical Geol., 74,289-308, 1989. 



GEOPHYSICAL LABORATORY 



75 



The Akermanite-Gehlenite-Sodium 

Melilite Join at 950°C and 5 kbar in the 

Presence of CO2 + H2O 

H.G. Huckenholz, H.S. Yoder, Jr., T. 
Kunzmann* and W.Seiberl* 

Melilites are primarily solid solutions 
between akermanite (Ca2MgSi20y), so- 
dium melilite (NaCaAlSi207), and 
gehlenite (Ca2Al2Si07). High-grade meta- 
morphism of impure limestones and dolo- 
mites favors the crystallization of members 
of the akermanite - gehlenite solid solution 
series, and the sodium melilite component 
is generally low. During the decarbonation 
process, CO2 and H2O play an active role 
and greatly influence the metamorphic as- 
semblages. On the other hand, igneous 
rocks, usually highly undersaturated and 
alkalic, are generally enriched in the so- 
dium melilite component and are close to 
the akermanite - sodium melilite join near 
ak 60 ge 1 o sm 30 0^ Goresy and Yoder, 1 974). 
A C02-rich fluid is probably involved in 
the upper mantle melting process, and is 
thought by some to be responsible for the 
formation of melilitite magma as well as 
nephelinite and kimberlite magmas. 

Experimental data on melilite-C02-H20 
are available for the endmember akermanite 
(Yoder, 1968, 1973, 1975; Huckenholz et 
al., this Report), gehlenite (Huckenholz 
and Yoder, 1974; Huckenholz, 1977; 
Hoschek, 1974) and for the akermanite - 
gehlenite solid solution series (Huckenholz 

Mineralogisch-Petrographisches Institut, 
Ludwig-Maximilians Universitat, D-8000 
Miinchen 2, Germany. 



et al., 1990). In order to elucidate the 
crystallization behavior of ternary melilites 
and their relationship to adjoining phases, 
an experimental study of the isothermal, 
isobaric section at 950°C and 5 kbar of the 
akermanite - gehlenite - sodium melilites 
was conducted in the presence of H2O and 
CO2 in both Washington and Munich. 



Experimental procedure 

Thirty crystalline melilite compositions, 
prepared by J.F. Schairer, from the 
akermanite - sodium melilite and the 
akermanite - gehlenite - sodium melilite 
joins (Schairer and Yoder, 1964; Schairer et 
al., 1967) were used in the experiments. 
Ten crystalline melilite compositions were 
also available from the akermanite - 
gehlenite join (Huckenholz et al., 1990) 
and (pure) sodium melilite from Yoder 's 
(1973) experimental study on that 
endmember composition. Water was added 
in excess to the samples for the melilite- 
H2O experiments, whereas oxalic acid 
dihydrate (H2C204»2H20) served as a 
source of CO2 + H2O. Oxalic acid dihydrate 
produces an initial X(C02) [CO2/ 
(H2O+CO2)] of 0.5. Higher X(C02) were 
obtained by adding calcite + oxalic acid 
dihydrate to the sample, and lower X(C02) 
were generated by a mixture of oxalic acid 
dihydrate + water. The weight ratio of 
oxalic acid dihydrate to sample was about 
1:2-3. When carbonation of sample took 
place, that is, formation of calcite and sea- 
polite, a final X(C02) of about 0.30 ± 0.03 



76 



CARNEGIE INSTITUTION 



Table 1 1. Liquid compositions 





Liquid + 


H20 






Liquid + 


CO2 + H2O 




Sm 100-1* 


Sm 100-2 


Sm8 


SmlOO 


Sm5 


SmlO 


Sm67 


Si0 2 


43.40 


42.57 


44.44 


48.01 


48.60 


50.03 


48.80 


AI2O3 


22.98 


22.08 


(23.00) 


21.30 


23.16 


20.27 


(21.70) 


Cat) 


0.03 


0.03 


0.50 


0.06 


0.51 


0.54 


0.50 


15.59 


15.65 


(15.60) 


9.30 


8.55 


9.18 


(9.10) 


Na20 


8.18 


9.25 


8.40 


11.49 


11.11 


10.02 


10.53 


totals 


90.18 


89.58 


(91.94) 


90.16 


91.93 


90.04 


(90.63) 


mel 


48.4 


53.1 


47.9 


39.6 


23.9 


29.1 


27.4 


(ak) 


- 


- 


6.6 


0.4 


3.6 


3.8 


3.4 


(ge) 


27.2 


19.5 


24.8 


7.4 


13.4 


10.7 


12.2 


(Sm) 


21.2 


33.6 


16.5 


31.8 


6.9 


14.6 


11.8 


jd 


23.2 


36.0 


26.8 


39.2 


55.5 


31.3 


44.4 


ab 


28.4 


10.9 


25.3 


21.2 


20.6 


39.6 


28.2 


X(CQ 2 ) V 


0.0 


0.0 


0.0 


0.72 


0.63 


0.54 


0.40 



Compositions of glasses from 950°C and 5 kbar experiments. SmlOO- 1, sodium melilite 
(starting from glass); Sm 100-2, sodium melilite (starting from crystalline sodium melilite); 
Sm8, akermanitel0-gehlenite20- sodium melilite70; Sm5, akermanite5-sodium melilite95; 
SmlO, akermanite 10- sodium melilite90; Sm67 } akermanite67- sodium melilite33. Molecu- 
lar proportions are (ak) Ca2MgSi207; (ge), Ca 2 Al 2 SiC>7; (Sm), NaCaAlSi 2 07; jd, 
NaAlSi 2 06; ab, NaAlSi308- ** Numbers in parentheses are estimates. 



Table 12. Compositions of crystalline materials 





Liquid + H 2 




Liquid 


+ CO2 + H 2 C 


> 






SmlOO 


Sm8 


SmlOO 


Sm5 


Sm64 


Sm67 


AklOO 




wo 


mel 


cc 


cc 


cc 


cc 


cc 


Si 


1.002 


1.716 


0.004 


0.019 


0.001 


0.001 


0.000 


Al 


0.002 


1.059 


0.001 


0.015 


0.000 


0.001 


0.000 


Mg 


0.001 


0.250 


0.000 


0.002 


0.003 


0.003 


0.002 


Ca 


0.995 


1.547 


0.992 


0.961 


0.996 


0.994 


0.997 


Na 


0.001 


0.432 


0.003 


0.001 
Liquid + C0 2 + 


0.001 
H 2 


0.001 


0.001 




Sm5 


SmlO 


Sm20 


Sm67 


Sm64 


Sm67 


AklOO 




wo 


wo 


wo 


wo 


cpx 


cpx 


cpx 


Si 


0.991 


0.991 


1.001 


0.998 


1.918 


1.925 


1.996 


Al 


0.003 


0.007 


0.002 


0.002 


0.226 


0.190 


0.003 


Mg 


0.008 


0.024 


0.011 


0.014 


0.848 


0.876 


0.997 


Ca 


0.996 


0.972 


0.984 


0.984 


0.979 


0.989 


1.002 


Na 


0.002 


0.005 


0.001 


0.001 


0.029 


0.020 


0.002 



Compositions of wollastonite, melilite, calcite, and clinopyroxene from 950°C and 5 kbar 
experiments. Compositions in cations per formula units. Bulk compositions as in Table 1 1 ; 
others are: Sm64, akermanite64-sodium melilite36; Sm20, akermanite 2 0- sodium meliliteso; 
aklOO, akermanite 100. 



GEOPHYSICAL LABORATORY 



77 



950"C-5 kbar 



Sm 
NaCaAISip 7 



L+wo+v 




Ge 
Ca 2 AI 2 Si0 7 



Ak 



Sm 
NaCaAISi^ 



L+cc+V 

+mel+cc+wo+V 



Ge 
CaAI^SJaG; 



L+mel+cc+cpx+wo+V 




Ca^AgS\p 7 



Fig. 46. Ak-Ge-Sm-H20 isothermal-isobaric sec- 
tion at 950°C and 5 kbar. Line A -Bis the limit of 
solid solution for ternary melilites found in natural 
rocks (El Goresy and Yoder, 1974). Abbrevia- 
tions are: V, fluid; L., liquid; mel, melilite; wo, 
wollastonite; gar, garnet; and cpx, clinopyroxene. 
Symbols refer to the phase assemblages as la- 
belled in the plane. Circle with dot is the melilite 
composition analyzed. 

in the subsolidus assemblages resulted. 

In hypersolidus assemblages, H2O par- 
titions between fluid and liquid. For the 
liquid, an X(C02) between 0.05 and 0. 1 is 
assumed. This assumption is derived from 
mass balance calculations conducted on a 
glass quenched from the sodium melilite 
endmember composition in which calcite 
is the only liquidus phase, and from CO2 + 
H2O solubility in albite melts (Kadik and 
Eggler, 1974). The CO2 partitions pre- 
dominantly between fluid and calcite + 
scapolite. The "final" X(C02) that results 
from X(C02)(fluid) + X(C02)(liquid) can 
be calculated, and is about 0.3 in most of 
the hypersolidus experiments. 

Experiments were carried out by means 
of two internally heated gas-media appara- 
tus in Washington and in Munich. The run 
duration was 24 hours in each case. 
Reversibility of experiments has not as yet 



Fig. 47. Ak-Ge-Sm-H20-CC>2 isothermal-iso- 
baric section at 950°C and 5 kbar at an X (CO2) of 
about 0.3. Abbreviations as Fig. 46; other is scap, 
scapolite. Symbols refer to the phase assemblages 
as labelled on the diagram. Square with dot is the 
phase assemblage at the beginning of melting. 
Line A - B is the position of the X(CC>2) versus 
composition plot of data for Fig. 48B. 



been demonstrated for this preliminary re- 
port. The run products were examined by 
optical methods, x-ray powder diffraction, 
and for 10 compositions, by microprobe 
analyses. Composition of glasses, calcite, 
wollastonite, clinopyroxene, and melilite 
are given in Tables 11 and 12. 



Experimental Results 

The experimental data obtained are 
presented in two isothermal-isobaric (950C 
- 5 kbar) sections of the ak (akermanite)-ge 
(gehlenite)-Sm (sodium melilite) join with 
H2O and with CO2 + H2O as fluids (Figs. 
46 and 47). The two ternary joins are each 
part of the cc (calcite)-di (diopside)-CaTs 
(Ca-Tschermak's component)-jd (jadeite) 
tetrahedron, which in turn is a four-compo- 
nent volume of the Na20-CaO-MgO- 



78 



CARNEGIE INSTITUTION 



950"C-5 kbar 



scap (mei) 4cc*cpx tsp.V 





B 


■ T T -1 1 1- 


1 




| 950'C-5 kbar | 


- 








cpx.cc 
+wo+V 

■ 


scap+cc* 
sp<?)+c<x+V 


scap+cc+cpx*sp(?)+V 




0*5 


* 




" 


mel+cc ^ 












\ 




\ mal+cpx+cc+V 




\ ■ 


■ 




- 


\ . 


md+V 












■ 





10 20 


30 


40 50 


60 


70 


80 


90 A 


/Sm15) 




mol% 








(Ak85/Sm15) 



Al203-Si02-C02-H20 system. Phases 
determined are fluid (V), liquid (L) calcite 
(cc,CaC03), wollastonite (wo, CaSi03), 
clinopyroxene [cpx, 

(Ca,Na)(Mg,Al)(Si,Al)20 6 ], garnet [gar, 
(Ca,Mg)3Al2Si30i2L scapolite [scap, 
Ca3(Ca,Na)iAl 5 -6Si6-7024(C0 3 )] and 
melilite [mel, (Ca,Na)2(Mg,Al) 
(Si,Al)207]. Compositions of calcite, 
clinopyroxene, and melilite plot in the cc- 
di-CaTs-jd system, whereas wollastonite, 
garnet, scapolite, and the liquid lie outside. 
Chemographical correlation of spatial phase 
assemblages from the multicomponent sys- 
tem Na20-CaO-MgO-Al203-Si02-C02- 
H2O, however, cannot always be appropri- 
ately illustrated. 

The isothermal-isobaric section of Fig. 
46 displays the phase assemblages encoun- 
tered in experiments with H2O as the fluid 
phase. The assemblages are outlined by 
dashed curves that are thought to be inter- 
faces of multiphase volumes intersecting 
the pseudo-ternary plane. The maximum 
extent of melilite solid solution, equivalent 



V?: 



950-C-5kbar 



, V+cryilaU+LxMd 



cpx 

scap 



± 

Lcfjd.c/yilals ± cak.«e -' 



Y ■ ' 1H9 *ec «cpx jx .3 - 

me)(1) .v 



o.ot 

SM 



70 



60 



SO 
mol% 



Fig. 48. Melilite composition versus AXCO2) .A. 
Melilite composition versus X(C02) plot of data 
along the join akermanite - gehlenite. Abbrevia- 
tions as in Figs. 46 and 47, others are mei, meionite; 
cor, corundum; sp, spinel. Symbols refer to the 
phase assemblages as labeled on the diagram. B. 
Melilite composition versus X(CC>2) plot of data 
along line A (Ak85/Sml5) -B (Ge 85/Sm 15) of 
Fig. 47. Melilite composition from above and 
below that line are projected. Abbreviations as 
Figs. 46 and 47; symbols refer to the phase assem- 
blages labelled. C. Melilite composition versus 
X(C02) plot of data along the join akermanite - 
sodium melilite. Abbreviations as in Figs. 46 and 
47. Mel (2) lies off the plane in the gehlenite - rich 
portion of Fig. 47. Vertical solid lines with dots 
are tie lines between fluid and liquid in composi- 
tions along akermanite - sodium melilite. Sym- 
bols in the subsolidus portion refer to phase assem- 
blages as labeled on the diagram. 



GEOPHYSICAL LABORATORY 



79 



to the H20-saturated solidus, is bound by 
akermanite58 sodium melilite42; 
akermanite26gehlenite3 1 sodium melilite43 
(microprobe data, Table 12; circle with 
dot); and by gehleniteyo sodium melilite3o 
compositions. The H20-saturated melilite 
solidus is close to the extent of melilite 
solid solution found for natural melilites by 
El Goresy and Yoder (1974). Solid phases 
in the hypersolidus portion of the plane 
coexist with a water-saturated liquid. 
Glasses quenched from sodium melilite ioo 
as well as akermaniteio gehlenite20 so- 
dium melilite70 were analyzed by electron 
microprobe. Their compositions (Table 
11) contain the melilite components (= ak + 
ge + sm) on the order of about 50 mol % but 
also jadeite (NaAlSi206) and albite 
(NaAlSi308). The total of constituents 
analyzed of the glasses run about 89-91%; 
the remainder is believed to be H2O dis- 
solved in the melt. The liquidus-phase 
wollastonite was also analyzed (Table 12), 
and is very close to wollastonite with minor 
solid solution, if any. No nepheline was 
detected in the run products, neither by 
optical, x-ray analysis, nor microprobe in- 
spection. 

The isothermal-isobaric melilite sec- 
tion at 950°C and 5 kbar with CO2 + H2O 
as a fluid is depicted in Fig. 48. For simplic- 
ity, the ternary plane is averaged forX(C02) 
of about 0.3. The bounding melilite solid 
solution on the CO2 + H20-saturated solidus 
was assumed to be akermanite20- 
gehlenite50-sodium melilite30. The appar- 
ent boundaries of the phase assemblages 
mapped in the plane tend to converge about 
this composition. 

Further indication of this special melilite 



composition can be deduced from the iso- 
thermal-isobaric, X(C02) vs. melilite com- 
position plot. Figure 48A displays such a 
relationship for akermanite-gehlenite. The 
solid solution is bound by decomposition 
of akermanite + CO2 to diopside + calcite 
at 0.12±0.02 and by gehlenite + CO2 to 
meionite + calcite + corundum at X(C02) 
0.30+0.02 (Huckenholz et al., 1990; 
Huckenholz and Seiberl, 1990), respec- 
tively. The breakdown assemblages of 
melilite + scapolite -1- calcite + corundum + 
spinel (?) in the gehlenite-rich portion and 
that of melilite + clinopyroxene + calcite in 
the akermanite-rich portion indicate a maxi- 
mum melilite solid solution of about 
akermanite 1 5 -gehlenite85, which appears 
to decompose at about X(C02) = 0.33 to 
scapolite + calcite + corundum + spinel (?). 

A C02-dependent compositional maxi- 
mum is also indicated when the relations 
are considered along line A -B on the 
akermanite-gehlenite-sodium melilite plane 
parallel to akermanite-gehlenite at about 
sodium melilite 15. In the section displayed 
in Fig. 48B, the melilite -1- CO2 stability has 
increased to about 0.45 AXCO2) . No stable 
melilite was found in the subsolidus assem- 
blage of scapolite + calcite + clinopyroxene 
+ spinel (?) at 0.5, and above the scapolite 
is a solid solution between meionite 
[Ca4Al6Si6024(C03)] and the (theoreti- 
cal) carbonate-marialite endmember 
[Na3CaAl 3 Si 9 024(C03)]. At 950°C and 5 
kbar its composition in (scapolite + calcite) 
- bearing assemblages is restricted to an 
equivalent an-contentof 0.75 (Huckenholz 
and Seiberl, 1990). 

Phase relations along the join 
akermanite-sodium melilite as a function 



80 



CARNEGIE INSTITUTION 



of X(C02) are shown in Fig.48C. The 
extent of solid solution toward sodium 
melilite is limited to akermanite70-sodium 
melilite30 and an X(C02) of 0. 1 2, which is 
at the breakdown of akermanite + CO2 to 
diopside + calcite. Melilites labelled mel 
(2) occur within the phase assemblages 
generated by the breakdown of akermanite 
+ CO2, but do not lie in the pseudo-binary 
akermanite-sodium melilite plane. Their 
composition is located in the gehlenite-rich 
portion of the akermanite-gehlenite-sodium 
melilite join as shown in Fig. 47, with 
X(C02) up to about 0.45. From the bound- 
ing melilite solid solution of the fluid- 
saturated solidus toward the sodium melilite 
endmember, hypersolidus phase assem- 
blages of V+mel+cc+cpx+scap+L, 
V+mel+cc+cpx+wo+L, V+mel+cc+wo+L, 
V+mel+cc+L and V+cc+L are traversed by 
the pseudo-binary join. Liquid with an 
assumed X(C02)(L) of 0.07 coexists with 
fluid having X(C02)(V) of 0.45 in the 
akermanite-rich portion of the join. To- 
ward the sodium melilite endmember com- 
position, the amount of the liquid increases 
thereby consuming increasing amounts of 
H2O, which in turn results in raising the 
X(C0 2 ) of the fluid. 

Quenched glasses from sodium 
melilite 100, akermanite5 -sodium melilite95, 
akermaniteio-sodium melilite^, and 
akermanite67-gehlenite33 bulk composition 
have been analyzed by microprobe (Table 
11). They exhibit (calculated molecular) jd 
(NaAlSi2C>6) and ab (NaAlSi308) compo- 
nents in addition to a large amount of 
ternary melilite (ak+ge+sm). In contrast, 
glasses quenched in the presence of a (pure) 



H2O fluid are lower in the jd and ab com- 
ponents but contain a ternary mel-compo- 
nent > 47%. Analyzed calcite (Table 12) 
contains minor amounts of MgO and Na20; 
wollastonite displays minor solid solution 
toward the sodium melilite (<1%) and 
akermanite (2-4%) components. 
Clinopyroxenes were analyzed from the 
akermanite64-sodium melilite36, 
akermanite67-sodium melilite33, and 
akermaniteioo bulk compositions. Ex- 
pressed as endmembers, they reduce to 
diopside87-CaTsio jadeite3, diopside89- 
CaTs9 jadeite2, and diopside99.7-CaTs<o.i 
jadeite<o.2, respectively. 



Reference 

El Goresy, A., and H. S. Yoder, Jr., Natural and 
synthetic melilite compositions. Carnegie 
Instn. Washington, Year Book, 73, 359-371, 
1974. 

Hoschek, G., Gehlenite stability in the system 
CaO-Al203-Si02-H20-C02- Contr. Min. 
Petr., 47, 245-254, 1974. 

Huckenholz, H. G., Gehlenite stability relations in 
the join Ca2Al2SiC>7 - H2O up to 10 kbar. 
NJb. Miner, Abh., 130, 169-186, 1977. 

Huckenholz, H. G., and H. S. Yoder, Jr., The 
gehlenite-H20 and \vollastonite-H2O systems. 
Carnegie Instn. Washington, Year Book, 73, 
440-443, 1974. 

Huckenholz, H.G., A. Wassermann, and K. T. 
Fehr, Stability and phase relations of gehlenite- 
akermanite solid solutions in the presence of a 
H20-C02-fluid. International Symposium of 
Experimental Mineralogy, Petrology and Geo- 
chemistry, Edinburgh, UK, p. 17, terra ab- 
stracts, 2, 1990. 

Huckenholz, H. G., and W. Seiberl, Stability and 
phase relations of carbonate scapolite solid 
solutions under the PT-regime of the deeper 
crust. Third International Symposium of Ex- 
perimental Mineralogy, Petrology and Geo- 
chemistry, Edinburgh, UK,. P. 17; terra ab- 
stracts 2 1990. 

Kadik, A. A., and D. H. Eggler,Melt-vapor rela- 
tions on the join NaAlSi308-H20-C02: 
Carnegie Instn. Washington, Year Book, 74, 
479-484, 1974. 



GEOPHYSICAL LABORATORY 



81 



Schairer, J. R, and H. S. Yoder, Jr., The join 
akermanite (Ca2MgSi207) - soda melilite 
(NaCaAlSi207) . Carnegie Instn. Washing- 
ton, Year Book, 63, 89-90, 1964. 

Schairer, J.F., H. S. Yoder, Jr., and C. E. Tilley, 
The high-termperature behavior of synthetic 
melilites in the join gehlenite-soda melilite- 
akermanite, Carnegie Instn. Washington, Year 
Book, 65, 217-226 1967. 

Yoder, H. S., Jr., Akermanite and related melilite- 
bearing assemblages. Carnegie Instn. Wash- 
ington, Year Book, 66, p47 1-477, 1968. 

Yoder, H. S. Jr., Melilite stability andparagenesis. 
Fortschr. Mineral, v.50, 140-173, 1973. 

Yoder, H.S., Jr., Relationship of melilite- bearing 
rocks to kimberlite: a preliminary report on the 
system akermanite-C02. Proc. Internat. 
kimberlite Conf., Cape Town. Phys. Chem. 
Earth, 9, 883-894, 1975. 



Merwinite Stability and 

High-Temperature Phase Relations 

in the Presence of CO2 + H2O. 

H. G. Huckenholz, H. S. Yoder, Jr., and 
W. Seiberl 



monticellite +melilite-bearing assemblages 
in high-temperature calc-silicate rocks oc- 
curring as inclusions in pyroxenites from 
the critical zone of the eastern Bushveld 
Complex. 

Phase equilibria studies on the join 
CaMgSi206-CaC03-C02 of the CaO- 
MgO-Si02-C02 system restrict the 
merwinite + C02 stability to high-tem- 
perature but low-pressure conditions. The 
reaction akermanite + calcite <=> merwinite 
+ CO2 (step 1 1 of the decarbonation series 
of Bowen, 1940) was studied by 
Shmulovich (1969) and Walter (1963a,b, 
1965) at low pressure but high tempera- 
ture. Merwinite + CO2 crystallizes from 
akermanite + calcite assemblages at tem- 
peratures of < 1065°C and pressures of < 
0.5 kbar. Merwinite + CO2, however, did 
not crystallize from diopside + 2 calcite 
assemblages between 950°C and 975°C at 
1 kbar (Yoder, 1975). 



Merwinite [Ca3Mg(Si04)2] was discov- 
ered and named by Larsen and Foshag 
(1921) from high-grade metamorphosed 
carbonaceous rocks at Crestmore near Riv- 
erside, California. At Crestmore (Burnham, 
1959; Walter, 1965) and other localities 
(e.g., Scawt Hill, Northern Ireland, Tilley, 
1929; Ardnamurchan, western Scotland, 
Agrell, 1965; Christmas Mountains, Big 
Bend region, Texas, Joesten, 1974), 
merwinite is mainly associated with cal- 
cite, spurrite, monticellite, melilite, and 
also with larnite. Recently, Wallmach et al. 
(1989) described merwinite from 

Mineralogisch-Petrographisches Institut, 
Ludwig-Maximilians Universitat, D-8000 
Munchen 2, Germany. 



Experimental Methods 

In order to elucidate the stability of 
merwinite in the presence of CO2 + H2O at 
pressures > 0.5 kbar, an experimental study 
of merwinite and its high-temperature phase 
relations was conducted. Stability and phase 
relations of merwinite with akermanite, 
diopside, calcite, liquid, and fluid were 
studied for pure CO2 in the 1-10 kbar 
pressure range at temperatures between 
900° and 1450°C. In addition, isobaric 
temperature versus CO2 + H2O relations 
were investigated at 1 and 3 kbar and at 
temperatures between 700° and 1 200°C. 

Mixtures of crystalline materials were 
used in all cases for the experiments. They 
consisted of: 



82 



CARNEGIE INSTITUTION 



03 
_Q 
J*: 

<D 

13 
CO 
CO 





12 
11 

ioh 

9 
8 

7 
6 
5 
4 
3 
2 
1 





800 



o /•/•/• •/«> 




900 



1000 1100 1200 

Temperature, °C 



1300 



1400 



Fig. 49. Pressure-temperature diagram for merwinite + CO2, akermanite + calcite, and diopside + calcite 
compositions. Heavy lines are univariant reaction curves; light lines are restricted reaction curves; short 
dashed curve is the compositional singularity for di + 2 cc = L; I\, I2, invariant points; Si, 52 singular 
points. Abbreviations for phases are: mer, merwinite (Ca3MgSi20g); ak, akermanite (Ca2MgSi2(>7); 
di, diopside (CaMgSi2C>6); cc, calcite (CaCC>3); L, Liquid. Symbols: Solid triangles (1) diopside + 
calcite = akermanite + CO2 (Shmulovich, 1969; Walter, 1963); solid triangles (2) akermanite + calcite 
= merwinite + CO2 (Shmulovich, 1969; Walter, 1963); open hexagons, di + cc from mixture B, C, and 
Y (for composition see text); open squares, akermanite + CO2 from mixture B, C, and Y as well as 
Yoder's data (1975); open diamonds, merwinite + CO2 from mixture B, C, and Y; solid diamonds, mer 
+ L + CO2 from mixtures B, C, and Y; solid squares, ak -1- L and ak + L + CO2 from mixture B, C, and 
Y; solid pentagon and solid circle are the 2 kbar run data on di + ak + CO2 = L (Yoder, 1975); solid 
hexagons, di + L and Di + L + CO2 from mixture C, Y, and Yoder's 1975 run data; open circles, Liquid 
on diopside + cc -1- CO2 compositions, mixture B, C, and Yoder's (1975) run data. 



(1). merwinite (crystallized at 1200°C, 1 
atm), mixture A; 

(2). akermanite (crystallized at 1050°C at 1 
atm) + calcite (Baker Chemical Com- 
pany, grade C.R.) equivalent (on a mo- 
lar basis) to merwinite + CO2; mixture 
B; 

(3). natural diopside (Twin Lakes, Califor- 
nia; Smith, 1 966) + 2 calcite equivalent 



to akermanite + calcite + CO2 or 
merwinite -1- 2 CO2, mixture C; and 

(4). natural 2 wollastonite (Willsboro, N. Y.) 
+ natural dolomite (Thornwood, N. Y.) 
equivalent to akermanite + calcite + 
CO2, or diopside + 2 calcite, or 
merwinite + 2 CO2, mixture Y. 
Other experiments were carried out on 

natural rock inclusions from the upper zone 



GEOPHYSICAL LABORATORY 



83 



of the eastern Bushveld Complex (locality 
Luipershoek, Joubert, 1976). Rocks con- 
taining akermanite + diopside (± 
monticellite), Ji, and monticellite + 
akermanite (± diopside), J6, were collected 
by H. G. Huckenholz (October, 1990). The 
synthetic assemblages merwinite + 
monticellite + akermanite60-gehlenite40 
(mixture E) and akermanite + monticellite 
+ calcite (mixture D) were also studied at 1 
and 3 kbar in the presence of CO2 + H2O. 
Experimental data were obtained by means 
of internally-heated, gas-media apparatus 
in both Washington, D. C, and Munich. 
Reversibility of the experiments was en- 
sured by the direction of reaction of the 
different crystalline mixtures listed above. 



Merwinite Relations with CO2 

Stability of merwinite -1- CO2 and 
merwinite phase relations with akermanite, 
diopside, calcite, liquid, and CO2 are dis- 
played in the temperature versus pressure 
plot in Fig. 49. With the data of Shmulovich 
(1969) at 0.5 kbar and below (see also 
Walter, 1963a,b), the reaction akermanite + 
calcite = merwinite + CO2 increases in 
temperature from 1015°C at 0.5 kbar up to 
1 1 82°C at 1 .3 ± 0.2 kbar, and results in the 
invariant assemblage (h) of merwinite + 
akermanite + diopside + calcite + liquid + 
CO2 in equilibrium. Four other univariant 
curves, 

[ak] merwinite + calcite + CO2 <=> liquid, 

[mer] akermanite + calcite + CO2 
<=> liquid, 

[cc] merwinite + CO2 

<=> akermanite -1- liquid, and 



[CO2] liquid + merwinite 

<=> akermanite + calcite, 

meet at that invariant point I2. The reaction 
[ak] was bracketed between 1150°C and 
1200°C at P = 1 kbar. Its position at about 
1180°C is fixed due to the (positive) slope 
of reaction [mer] and by the run at 1 200°C 
and 3 kbar in particular, which is just slightly 
above the solidus with a phase assemblage 
of akermanite + calcite + liquid. At 2 kbar 
and temperatures between 1200° and 
1450°C, merwinite + CO2 does not crystal- 
lize from akermanite + calcite nor from 
diopside + 2 calcite mixtures. Thus, the 
curves for the two univariant reactions of 
[cc] and [CO2] must pass between 1.3 and 
2 kbar. At reaction [cc], merwinite + CO2 
tie lines are interrupted by those of 
akermanite + liquid but with merwinite 
remaining in the C02-absent region of the 
merwinite + calcite + liquid and merwinite 
+ akermanite + liquid assemblages. 

Because of chemographic constraints, 
reaction [CO2] must occur on the high- 
pressure side of reaction [cc] and between 
1.3 and 2 kbar as well. In that pressure 
range, the melting curve of calcite (Yoder, 
1 973) intersects the reaction [CO2] at about 
1350°C. Calcite in the assemblage will 
melt and the restricted assemblage of 
merwinite + akermanite + liquid evolves 
from S\. At higher temperatures, the diop- 
side + akermanite + CO2 solidus and the 
diopside + CO2 solidus appear in the diop- 
side + akermanite + CO2 portion of the 
diopside -1- calcite + CO2 system. The 
temperature of the akermanite + diopside + 
CO2 solidus can be deduced from the run 
data of Yoder (1975) on diopside + calcite 
compositions obtained at 2 kbar. 
Akermanite + diopside + CO2 will melt 
slightly above 1400°C, and diopside + CO2 



84 



CARNEGIE INSTITUTION 



(Rosenhauer and Eggler, 1975) at 1415°C 
as well. 

In the temperature and pressure range 
studied, merwinite and diopside do not 
coexist in the presence of C02. They are 
separated by akermanite + calcite, or by 
akermanite + CO2 tie lines. Below 900°C 
there is only a very narrow CO2 pressure 
range of about 200 to 300 bar where 
akermanite + CO2 is formed from diopside 
+ calcite. With increasing temperature, 
akermanite + CO2 is stable up to about 5 
kbar. The reaction diopside + calcite = 
akermanite + CO2 (step 8 of Bowen's de- 
carbonation series) was studied at low pres- 
sures by Walter (1963) and up to 6 kbar by 
Yoder (1973), who exclusively used a crys- 
talline mixture of (natural) diopside -1- 1 
calcite. The slope of the reaction curve of 
Yoder (1973), drawn at the first appearance 
of akermanite, between 1 and 5.75 kbar is 
about 58°C/kbar. The reaction takes place 
through a range of temperatures 50°-70°C 
wide at a given pressure, presumably be- 
cause of possible solid solution of merwinite 
in akermanite, diopside in merwinite 
(Schairer et ai y 1967; Yoder, 1973), and 
minor substitution of Ca by Mg in calcite. 
The akermanite + diopside + calcite region 
was not observed in the present study when 
akermanite + calcite and diopside + 2 cal- 
cite compositions were used. The newly 
crystallized akermanite, however, contains 
tiny inclusions of (relict) diopside that are 
separated from calcite by the akermanite 
host. The diopside + calcite <=> akermanite 
+ CO2 reaction curve, now bracketed by 
means of akermanite + calcite and diopside 
+ 2 calcite compositions, has a revised 
slope of about 62°C/kbar. It terminates at 
the invariant point I\, which was found to 
be located at5.4±0.2kbar and 1215°±5°C 
with akermanite + diopside + calcite + 



liquid + CO2 in equilibrium. 

The invariant point I\ is the locus of four 
other reactions occurring clockwise: 

[CO2] akermanite + calcite + diopside 
<=> liquid, 

[ak] calcite + diopside 

<=> liquid + CO2, 

[cc] diopside + liquid 

<=> akermanite + CO2, and 

[di] akermanite -1- calcite + CO2 <=> liquid. 

Separation of reaction [C02] from reaction 
[ak] was not possible because the liquid 
field may be located very close to or even 
on the diopside + calcite join. Above the 
temperature of reaction [ak], diopside + 
liquid + C02 will reach the diopside + 2 
calcite composition (= compositional sin- 
gularity), and any further increase in tem- 
perature will move the liquid -1- C02 tie 
lines along diopside + calcite toward the 
join akermanite -1- CO2, that is, becoming 
compositionally equivalent to diopside + 1 
calcite. That particular composition is 
reached in S2 at 13 10°C and about 5.3 kbar 
located on reaction [cc], diopside + calcite 
= akermanite + CO2, from which the two 
restricted reactions diopside + calcite = 
liquid (higher pressure limb) and liquid = 
akermanite -1- CO2 (lower pressure limb) 
will occur. 



Merwinite Relations With CO2 and H2O 

Experiments on the stability of 
merwinite in the presence of C02 + H2O 
were conducted at 1 and 3 kbar (Figs. 50A 
and 50B). The merwinite + V stability field 



GEOPHYSICAL LABORATORY 



85 



1300 



1200 



1100 



O 

°- 1000 


| 900 
CD 

Q. 

E 

i® 800 



700 



600 



P=1 kbar 




(1)KUSHIRO&YODER,1964 

(2) WALTER, 1965 

(3)YODER,1973 

(4) HUCKENHOLZ etal., 1990 



0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 

XcOo 



1300 



_ . - 995' 




1195* 
1170' 



1090' 
1065' 



Fig. 50a and 50b. Isobaric temperature -X(<X>2) [X((X>2) = CO2ACO2 + H2O)] diagram for 1 kbar (Fig. 
50A) and 3 kbar (Fig. 50B). Abbreviations for phases are: mo, monticellite (CaMgSi04); per, periclase 
(MgO); fo, forsterite (Mg2Si04) spur, spurrite [Ca5Si208(C03)]; wo, wollastonite (CaSi03); geh, 
gehlenite (Ca2Al2Si07); gro, grossular (Ca3Al2Si30i2); cor, corundum (AI2O3); mei, meionite 
(Ca3 Al6Si6024*CaC03); V, fluid; other abbreviations and symbols as in Fig. 49. Letters on half-shaded 
symbols or symbols with dots refer to experiments on mixtures F, D, E, Jj, and J6- The solidus of these 
compositions (except for E) refers to ak + di + CO2 (above 1 1 50°C). Light curves and light dashed curves 
are decarbonation reactions (extrapolated) from Walter (1963a.b; 1965) h to/6 are not fully illustrated. 
The geh + V reaction is from Huckenholz et al. (1990). 



86 



CARNEGIE INSTITUTION 



decreases with increasing pressure when 
the C02 fluid is diluted with H2O. At 
AXCO2) = 1.00, akermanite + calcite = 
merwinite + V are restricted to a tempera- 
ture of 1 1 20°C at 1 kbar and 1 1 82°C at 1 .3 
kbar (isobaric invariant point I2). From I2 
the reaction akermanite + calcite <=> 
merwinite + V shifts to lower AXCO2) when 
the pressure increases (light dashed curves 
in Fig. 50A and 50B). At 810°C and at 1 
kbar the reaction spurrite -1- monticellite <=> 
merwinite + calcite (Walter, 1965), well 
displayed at Crestmore, California 
(Burnham, 1959), must intersect with 
akermanite + calcite <=> merwinite + V at 
about X(C02) = 0.10 in Fig. 50A The 
resulting invariant point labelled I5 be- 
comes the locus of the reaction of spurrite 
+ akermanite + monticellite <=> merwinite 
+ V and akermanite + calcite <=> spurrite + 
monticellite + V (not shown in Fig. 50A). 
With decreasing temperatures spurrite + 
akermanite + monticellite <=> merwinite + 
V intersects monticellite + wollastonite <=> 
akermanite at a pressure of 1 kbar and a 
temperature as low as 700°C and atX(C02) 
= 0.07 (isobaric invariant point Ie). 

Merwinite + calcite + V will melt at 1 
kbar over almost the entire range of X(C02) 
displaying merwinite + calcite + liquid as 
well as the merwinite -1- liquid + V assem- 
blage. Increasing pressure shifts the 
merwinite + calcite + V solidus toward 
lower X(C02) and lower temperatures as 
well. At 3 kbar, merwinite + calcite + V will 
melt atX(C02) < 0.15 and temperatures < 
1150°C. At X(C02) > 0.15, akermanite + 
calcite + V forms the subsolidus assem- 
blage, which melts between 1150° and 
1 200°C to akermanite + calcite + liquid and 
akermanite + liquid + V assemblages. The 
X(C02) of the liquid coexistent with fluid 



was not determined during the course of the 
present study. 

Melting experiments on silicate + H2O 
+ CO2 systems (Mysen, 1975; Rosenhauer 
and Eggler, 1975; Kadik and Eggler, 1975) 
demonstrate, however, the preference of 
H2O solubility over CO2 in silicate melts 
and, therefore, X(C02) of the liquid of 
about 0.05 was assumed for the tempera- 
ture and pressure conditions studied. The 
diopside + calcite = akermanite + CO2 
reaction was determined for 3 kbar at 
X(C02) of 1 .00, 0.50 and 0.05 (Huckenholz 
et ai, 1990; this study) and found to be 
1065°, 1010°, and 740°C, respectively. Cal- 
culations for X(C02) (Helgeson et ai, 
1978; Holloway; 1977) at 0.95, 0.10, and 
0.05 yielded temperatures of 1059°, 846°, 
and 803 °C. Experimental data for 1 kbar 
are available only for X(C02) at 1 .00 and 
950°C and for X(C02) at 0.5 and 895°C 
(Huckenholz etal., 1990; this study). They 
were calculated for X(C02) = 0.10 and 
0.20, and found to be about 735° and 770°C, 
respectively. 

The diopside + calcite <=> akermanite + 
V reaction is intersected by the fluid- and 
calcite-absent reaction of diopside + 
monticellite = forsterite + akermanite on 
the basis of the experimental data of Kushiro 
and Yoder (1964) and of Yoder (1973). The 
intersection will occur at 1 kbar at 880°C 
and about X(C02) = 0.45 and at 3 kbar at 
about 920°C and X(C02) = 0.20 ± 0.02, 
resulting in the isobaric invariant point 
labeled I4. The validity of the diopside + 
monticellite = forsterite + akermanite reac- 
tion is questioned by Helgeson etal. ( 1 978), 
who calculated reaction temperatures of 
about 1 300°C for 1 kbar and about 1 350°C 
for 3 kbar. It should be noted, however, that 
the anhydrous melting of diopside + 



GEOPHYSICAL LABORATORY 



87 



akermanite + forsterite assemblages will 
occur at 1357°C at 1 arm (Ricker and 
Osborn, 1954). A hydrous fluid phase 
involved in the melting will lower the 
solidus considerably, well below the tem- 
perature of the solid- solid reaction as cal- 
culated by Helgeson et al. (1978). The 
diopside + calcite <=> akermanite + CO2 
reaction will terminate at the CO2- and 
calcite-absent reaction of monticellite + 
wollastonite <=> akermanite (Yoder, 1973; 
Huckenholz, 1990; this study). The result- 
ing isobaric invariant point I4 for 1 kbar 
occurs at 700°C and atX(C02) = 0.1; for 3 
kbar, it is found to be at 720°C and at 
X(C0 2 ) = 0.05 ± 0.02. 



Discussion 

With the experimental data at hand, it 
has been demonstrated that merwinite + 
CO2 is stable over a wide range of tempera- 
ture but is restricted to pressures below 
about 1.5 kbar. Merwinite + V, however, 
may also form atX(C02) as low as 0.05 and 
at temperatures as low as 700°C from 
spurrite + monticellite + akermanite and 
from spurrite -1- monticellite + wollastonite 
assemblages in H20-rich aureoles of car- 
bonaceous rocks around basic intrusives. 
At pressures below 1 .3 kbar merwinite + 
calcite +V and merwinite + V assemblages 
reach the solidus between 1 1 50° and 1 200°C 
and will melt at any X(C02) up to 1 .0. At 
pressures above 1 .3 kbar the merwinite + V 
stability shifts toward lower X(C02) and 
akermanite + calcite + V assemblages reach 
the solidus and are subject to melting. At 3 
kbar, these same relationships occur at 
1 1 50°- 1 200°C and X(C02) > 0. 1 5 . 

High-temperature phase assemblages 
occur as rock inclusions in the critical zone 



of the Bushveld Complex: 

( 1 ) merwinite + monticellite + melilite 
(equivalent to mixture E), 

(2) calcite + periclase + monticellite, 
and 

(3) forsterite + periclase + 
monticellite. 

A different set of high-temperature phase 
assemblages occur as rock inclusions in the 
marginal zone: 

(4) akermanite + monticellite + cal- 
cite (equivalent to mixture D), 

(5) calcite + forsterite + monticellite, 

(6) akermanite + diopside + 
monticellite (equivalent to inclusions 
Jl and J6), and 

(7) diopside -1- forsterite + monticellite. 
According to Wallmach et al. (1989), tem- 
peratures as high as 1300° and 1200°C, 
respectively, for estimated pressures at 
about 1 kbar (0.6 to 1 . 1 kbar) and 2 kbar 
(1.1 to 2.4 kbar) are deduced. Experimen- 
tal results obtained on merwinite + V as- 
semblages (mixtures A, B, C, and Y), 
merwinite + monticellite + melilite assem- 
blages (mixture E), akermanite + 
monticellite + calcite assemblages (mix- 
ture D) as well as on the rocks (Ji and J6) 
containing akermanite + monticellite + di- 
opside are clearly above their solidi when 
treated at 1200°C and at pressures of 1 and 
3 kbar, respectively, with X(C02) between 
0.3 to 1.0. Thus, the deduced temperatures 
of Wallmach et al. (1989) appear to be 
excessive. 



References 

Agrell, S. O., Poly thermal metamorphism of lime- 
stones at Kilchoan, Ardnamurchan, Mineral. 
Mag., 34, (Tilley vol.) 1-15, 1965. 

Bowen, N. L., Progressive metamorphism of sili- 
ceous limestone and dolomite, J. Geoi, 48, 
225-274, 1940. 



88 



CARNEGIE INSTITUTION 



Burnham, C. W., Contact metamorphism of mag- 
nesian limestone at Crestmore, California, 
GeoL Soc, Amer. Bull, 70, 879-920, 1959. 
Helgeson, H. D., J. M. Delany, H. W. Nesbitt, and 
D. K. Bird, Summary and critique of the ther- 
modynamic properties of rock-forming min- 
erals, Amer. J. Sci., 278-A, 1-229, 1978. 
Holloway, J. R., Fugacity and activity of molecu- 
lar species in supercritical fluids, in Thermo- 
dynamics in Geology, D. Fraser, ed., 161-181, 
1977. 
Huckenholz, H. G., A. Wassermann, and K. T. 
Fehr, Stability and phase relations of gehlenite- 
akermanite solid solutions in the presence of a 
H2O-CO2 fluid, Third International Sympo- 
sium Experimental Mineralogy , Petrology, and 
Geochemistry, p. 81, Edinburgh, 1990. 
Joesten, R. C, Pseudomorphic replacement of 
melilite by idocrase in a zoned calc-silicate 
skarn, Christmas Mountains, Big Bend Re- 
gion, Texas, Amer. Mineral., 59, 694-699, 
1974. 
Joubert, J., Gemetamorfoseerde karbonaatins 
luitsels in die bosone van die 
Bosveldstollingskompleks, was van 
Roossenekel, Transvaal, M.Sc. thesis, Univ. 
Pretoria, Hillcrest, South Africa, 1976. 
Kadik, A. A., and D. H. Eggler, Melt- vapor rela- 
tions on the join NaAlSi30s-H20-C02, 
Carnegie Instn. Washington Year Book, 74, 
479-484, 1975. 
Kushiro, I., and H. S. Yoder, Jr., Stability field of 
akermanite, Carnegie Instn. Washington Year 
Book, 63, 84-86, 1964. 
Larsen, E. S., and W. F. Foshag, Merwinite, a new 
mineral from the contact zone at Crestmore, 
California, Amer. Mineral., 6, 143-148, 1921. 
Mysen, B. O., Stability of volatiles in silicate 
melts at high pressure and temperature, 
Carnegie Instn. Washington Year Book, 74, 
454-478, 1975. 
Ricker, R. W., and E. F. Osborn, Additional phase 
equilibrium data for the system CaO-Mg02- 
Si02, /. Amer. Ceram. Soc, 37, 133-139, 
1954. 
Rosenhauer, M., and D. H. Eggler, Solubility of 
H2O and CO2 in diopside melt, Carnegie 
Instn. Washington Year Book, 74, 474-479, 
1975. 
Schairer, J. F., H. S. Yoder, Jr., and C. E. Tilley, 
Behavior of synthetic melilites in the join 
gehlenite - soda melilite - akermanite, Carnegie 
Instn. Washington Year Book, 65, 217-226, 
1967. 
Shmulovich, K. I., Stability of merwinite in the 
system CaO-MgO-Si02-C02,D^/./4c^.5d, 
USSR, Earth Sci. Sect., 184, 125-127, 1969. 
Smith, J. V., X-ray emission microanalyses of 
rock-forming minerals, VI. Clinopyroxenes 
near the diopside-hedenbergite join, /. GeoL, 



74, 463-477, 1966. 

Tilley, C E., On larnite (calcium orthosilicate, a 
new mineral) and its associated minerals from 
the limestone xenoliths in the eastern Bushveld 
Complex, Canad. Mineral.,27, 509-523, 1989. 

Walter, L. S., Experimental studies on Bowen's 
decarbonation series, I: P-T univariant equi- 
libria of the "monticellite" and "akermanite" 
reactions, Amer. J. Sci., 261, 488-500, 1963a. 

Walter, L. S., Experimental studies on Bowen's 
decarbonation series II: P-T univariant equi- 
libria of the reaction: forsterite + calcite = 
monticellite + periclase + CO2, Amer. J. Sci., 
261, 173-179, 1963b. 

Walter, L. S., Experimental studies on Bowen's 
decarbonation series III: P-T univariant equi- 
libria of the reaction spurrite + monticellite = 
merwinite + calcite and analyses found at 
Crestmore, California, Amer. J. Sci., 263, 64- 
77, 1965. 

Yoder, H. S., Jr., Melilite stability and paragen- 
esis, Fortschr. Mineral, 50, 140-173, 1973. 

Yoder, H. S., Jr., Relationship of melilite- 
bearing rocks to kimberlite. A preliminary 
report on the system akermanite-C02, Phys. 
Chem. Earth, 9, 883-894, 1975. 



The System MG2S1O4-FE2S1O4 
at Low Pressure 

Hiroko Nagahara, Ikuo Kushiro* and 
Bjorn O. Mysen 

Gas -solid relationships are important 
when we consider condensation, evapora- 
tion, and fractionation of the solar nebula, 
especially in regard to bulk composition of 
the Earth and the terrestrial planets. Gas- 
solid relationships of minerals are different 
from those between liquid-gas, and ther- 
modynamic data are insufficient to con- 
struct gas-solid phase diagrams. Mysen 
and Kushiro (1988) and Kushiro and Mysen 
(1991) measured vapor pressures of MgO, 



* Geological Institute, University of Tokyo, Hongo, 
Tokyo 113, Japan 



GEOPHYSICAL LABORATORY 



89 



Si02, forsterite, and enstatite, and studied 
the phase relations of the MgO-Si02 sys- 
tem. In this study, we have measured the 
vapor pressure of fayalite, and based on 
those results along with those for forsterite, 
the gas-solid phase relationships of the 
olivine system as a function of the Fe/ 
(Mg+Fe) ratio are proposed. 

Vapor pressure measurements of fayalite 
were made using a Knudsen cell method 
similar to that described by Mysen and 
Kushiro (1988). The starting material is a 
single crystal of fayalite, about 2x2x5 mm 
in size, synthesized with the Czochralski- 
pulling method by H. Mori of the Univer- 
sity of Tokyo. The crystal was powdered 
(1-10 Jim), and several mg of the sample 
was placed in a molybdenum capsule with 
two 2-mm holes drilled in sides. The ex- 
periments were carried out in vacuum fur- 
naces in the University of Tokyo and the 
Geophysical Laboratory; the two furnace 
designs are nearly identical in size and 
design (Mysen and Kushiro, 1 988). Samples 
were heated at a rate of 15-20°/min from 
room temperature to experimental tem- 
peratures, which ranged from 1050°C to 
1 175°C. Run durations ranged from 4 days 
at 1175°C to 12 days at 1075°C. Total 
pressure of the vacuum chamber was 4.0 x 
10-7 torr (5.3 x 10" 1( > bar) to 6.0 x 10-7 torr 
(7.9x10-9 bar). 

Experiments in molybdenum capsules 
were at the oxygen fugacity (fo 2 ) of the 
M0-M0O2 buffer to ensure that fayalite is 
stable. The M0-M0O2 buffer is about 1.5 
orders of magnitude higher than the iron- 
wiistite (IW) buffer, and 1 to 1.5 orders of 
magnitude below the quartz-fayalite-mag- 
netite (QFM) buffer (Mysen and Kushiro, 



1988). Measured weight loss (1-14 %) was 
calibrated against vapor pressure by the 
equation 



P - 1 

1 m — 



m 



dw 



Ac dt 



v 



2kRT 
M 



(U 



where P m is the vapor pressure of a sub- 
stance, A is the area of the orifice of the 
capsule, c is the clausing factor, dw is the 
weight loss, dt is duration, M is the 
molecular weight of the effusing vapor, 7is 
the absolute temperature, and R is the gas 
constant (Paule and Margrave, 1967). The 
clausing factor for the cell has previously 
been determined for Cu and Ag by Mysen 
and Kushiro (1988). 

The residue of partial evaporation re- 
mained fayalite, indicating that fayalite 
evaporates congruently. Since forsterite 
evaporates congruently and intermediate 
olivine evaporates stoichiometrically 
(Mysen and Kushiro, 1988; Nagahara et 
al. y 1988), it is clear that olivine evaporates 
stoichiometrically regardless of the Fe/Mg 
ratio. Fayalite heated at 1175 °C melted, 
but fayalite heated at 1 160°C did not, sug- 
gesting that the melting point is about 
1 170°C and 2 x 10-8 bar. This is about 35°C 
lower than that at 1 bar (Bo wen and Schairer, 
1935). Lower melting temperatures rela- 
tive to melting temperature at 1 bar have 
also been found for Si02 and Mg2Si04 at 
low pressures (by 100° and 200°C, respec- 
tively) (Mysen and Kushiro, 1988). Mysen 
and Kushiro (1988) further demonstrated 
the presence of a three-phase region be- 
tween that of solid and gas and that of liquid 
and gas. This three phase region (forsterite 
+ liquid + vapor) implies that the system 



90 



CARNEGIE INSTITUTION 




1040 



1080 



1120 
T (<>C) 



1160 



1200 



Fig. 5 1 . Temperature and vapor pressure relation- 
ship of fayalite. Error bar represents weight mea- 
surement uncertainty. 



can not be described as a binary (MgO- 
S1O2). Therefore, phase relations where 
liquid is present will not be discussed in the 
present study. 

The experimentally determined tem- 
perature and vapor pressure (P v ) relation- 
ships for Fe2Si04 are summarized in Fig. 
51. The relationship is further shown in 



c 



-12 ■ 



-16 



-20 



-24 




1/TxlO 4 (1/K) 

Fig. 52. Arrhenius plot for vaporpressure of fayalite 
(this work) and forsterite, MgO, and Si02 (Mysen 
and Kushiro, 1988). Linear regression line from 
the fayalite data gives In / > v=(-608±60)/Rr + 
(273±9)/R and r = 0.9976. 



Fig. 52 together with the l/T vs. In P v of 
Mg2Si04, MgO, and Si02- The linear re- 
gression curve from the data points can be 
expressed in the equation for evaporation 



In P v = :Mv + ASv 
RT R 



(2) 



where A// v and AS V are enthalpy and en- 
tropy of evaporation, respectively. From 
the linear regression, A// v is 608 ± 60 (1 o) 
(kJ/mol) and AS V is 273 ± 9 (J/K-mol). 
These values are similar to those for 
forsterite (640 ± 36 and 210±54, respec- 
tively) at the same/02 (Mysen and Kushiro, 
1988). 

The gas-solid phase diagram was drawn 
by using the enthalpies and entropies for 
evaporation of forsterite and fayalite (Fig. 
53). In order to draw the phase diagram, 
two assumptions were made: (1) chemical 
equilibrium is achieved between crystals 
and gas in a Knudsen cell, and (2) both 
olivine and gas are ideal solutions. The 
assumption (1) can be valid. The residues 
are sintered homogeneous fayalite crystals 
regardless of experimental duration. Ac- 
cordingly, chemical compositions of coex- 
isting gas and solid have been uniform. The 
assumption (2), ideality of the olivine solid 
solution system, has been shown by many 
investigators (i. e., Wood and Kleppa, 1981). 
Gas can be treated as ideal. 

With the assumptions made above, mol 
fractions of the fayalite component in gas 
and solid at a given pressure are shown by 
the following equations 



GEOPHYSICAL LABORATORY 



91 



1800 



-" 1 • r 



^ 1600 

CD 

■*-> 
(0 

i_ 

0) 
Q. 

E 

£ 1400 



1200 



1800 



* 


1600 


<D 




k_ 




T 




■*-• 




CO 








O 




Q. 




F 




0) 


1400 


h- 





1200 



GAS 



OLIVINE s.s.+ GAS 



I0" 8 bar 



J I l_ 



■ I I 1_ 



20 40 60 80 100 



"> 1 " r 



T r 



10" 10 bar 



GAS 




_. 1 i_ 



1800 



^ 1600 



1200 



-i 1 «" 



GAS 




10 bar 
j i i k 



20 40 60 80 100 

Fe/(Mg+Fe)x100 



20 40 60 80 100 

Fe/(Mg+Fe)x100 



Fig. 53. Gas-solid phase diagrams of the olivine 
system at 10" 8 , 10A and 10 r ° bar. 



ln*- = 
x 



and 



'_ AH Fi 



\T ~fl 



In 



\-x_ AH Fo 
1 -x 



R 



L 

To 



(3) 



(4) 



where x and x y are the Fe/(Fe+Mg) ratios of 
gas and solid, respectively, and T and Tq 
are vaporus temperatures for fayalite and 
forsterite, respectively. 



The calculated vaporus and solidus 
curves are shown in Fig. 53. The conspicu- 
ous feature of the figure is that the binary 
loop is quite flat. With decreasing pressure, 
the vaporus temperatures for both forsterite 
and fayalite become lower and the loop 
becomes more flat. The figure sows that the 
compositional difference between coexist- 
ing solid and gas is extremely large com- 
pared to that between solid and liquid at L 
bar (Bo wen and Schairer, 1935). There is a 
very narrow temperature interval over 
which Fe -bearing olivine coexists with gas. 
Thus, olivine should have become forsterite 



92 



CARNEGIE INSTITUTION 



regardless of the primary composition when 
it was heated at subvaporus temperatures. 
Formation of forsterite would have been 
kinetically suppressed when equilibrium 
between gas and solid was not achieved. 
Hashimoto (1990) showed that forsterite 
evaporates very slowly at a free evapora- 
tion condition (at a rate one-tenth of that in 
equilibrium); that is, if the generated gas 
was removed from the system and did not 
equilibrate with the solid, formation of 
forsterite would be suppressed. Another 
possible factor preventing formation of 
forsterite is cation diffusion in solid 
forsterite, which depends on temperature, 
heating duration, and the grain size of oli- 
vine. Although the Mg-Fe inter-diffusion 
coefficient in olivine is largest among any 
other elemental diffusion coefficients in 
olivine and those known in any other sili- 
cate minerals (Freer, 1981), short heating 
and/or large grain size can be rate-limiting 
factors for formation of forsterite as partial 
evaporation residue. If these processes were 
not effective, olivine should have become 
forsterite quickly by heating at subvaporus 
temperatures. 

The phase diagram is not directly appli- 
cable to evaporation in the solar nebula; 
effects of other components, pressure, and 
foi should be evaluated. Other components, 
such as Al and Ca, would affect the abso- 
lute temperature of vaporus and solidus, 
but would not change significantly the shape 
of the diagram because of much smaller 
abundance in the solar nebula of these 
components compared with Si, Mg, and Fe 
(Anders and Ebihara, 1982). The pressure 
range in the present experiments is just 



applicable to the solar nebula. Elemental 
abundances of Mg and Si in the solar sys- 
tem are about 4 orders of magnitude smaller 
than that of H, and the total pressure at the 
midplane of about 3 A.U. has been gener- 
ally calculated to be between 10-3 and 
10-5 bar (Cameron, 1985; Morfill et al., 
1985). The approximate partial pressure 
for olivine component is thus 10 7 to 10 -9 
bar, which well agree with the pressure 
range in the present work. Oxygen fugacity 
condition of the present work is much more 
oxidizing than that estimated for the solar 
nebula, based on the elemental abundances 
of the solar system. However, oxygen fu- 
gacity as high as the present work has been 
proposed recently for the formation of vari- 
ous chondritic components in the solar 
nebula (Fegley and Palme, 1985; Palme 
andFegley, 1990; Weinbruclmtf/., 1990). 
The present result can be, thus, applied 
almost directly to the conditions for forma- 
tion of chondritic components in the solar 
nebula. 



References 

Anders, E., and M. Ebihara, Solar-system abun- 
dances of the elements. Geochim. Cosmochim. 
Acta, 46, 2363-2380, 1982. 

Bowen, N. L. and Schairer, J. F., The system 
MgO-FeO-Si02, Amer. Jour. Sci., ser. 5, 29, 
151-217, 1935. 

Cameron, A. G. W., Formation and evolution of 
the primitive solar nebula, in Protostar and 
Planets II, D. C. Black and M. S. Matthews, eds., 
Univ. Arizona Press, Tucson, Arizona, pp. 1073- 
1099, 1985. 

Fegley, B., Jr. and H. Palme, Evidence for oxidiz- 
ing conditions in the solar nebula from Mo and 
W depletion in refractory inclusions in carbon- 
aceous chondrites, Earth Planet. Sci. Lett., 72, 
311-326,1985. 



GEOPHYSICAL LABORATORY 



93 



Freer, R., Diffusion in silicate minerals and glasses: 
A data digest and guide to the literature. Contrib. 
Mineral Petrol, 76, 440-454, 1981. 
Hashimoto, A. Evaporation kinetics of forsterite 
and implications for the early solar nebula. 
Nature, 347, 53-55, 1990. 
Kushiro, I. and B. O. Mysen, in Progress in 
Metamorphic and Magmatic Petrology , L. L. 
Perchuked, Cambridge Univ. Press, Cambridge, 
England, 411-433, 1991. 
Morfill, G. E., W. Tscharnuter, and H. Volk, in 
Protostar and Planets II, D. C. Black and M. S. 
Matthews, eds., Univ. Arizona Press, Tucson, 
Arizona, pp. 493-533, 1985. 
Mysen, B. O. and I. Kushiro, Condensation, 
evaporation, melting, and crystallization in the 
primitive solar nebula: Experimental data in the 
system MgO-Si0 2 -H 2 to 1.0xl0- 9 barand 1870°C 
with variable oxygen fugacity, Amer. Miner., 
73, 1-19, 1988. 
Nagahara, H., I. Kushiro, B. O. Mysen, and H. 
Mori, Experimental vaporization and condensa- 
tion of olivine solid solution, Nature, 331, 516- 
518, 1988. 
Palme, H. and B. Fegley, Jr., High temperature 
condensation of iron-rich olivine in the solar 
nebula, Earth Planet. Sci. Lett., 101, 180-195, 
1990. 
Paule, R. C. and J, L, Margrave, Free-evaporation 
and effusion techniques, in The Characteriza- 
tion of High Temperature Vapors, J. L. Margraves 
ed., Wiley, New York, 130-151, 1967. 
Weinbruch, S., H. Palme, W. F. Muller, and A. El 
Goresy, FeO-rich rims and veins in Allende 
forsterite: Evidence for high temperature con- 
densation at oxidizing conditions, Meteoritics, 
25, 115-125, 1990. 
Wood, B. J. and O. J. Kleppa, Thermochemistry of 
forsterite-fayalite olivine solutions, Geochim. 
Cosmochim. Acta, 45, 529-534, 1981. 



Fe3 + , Mg Order-Disorder in Heated 

MGFE2O4: A Powder XRD 

and 57 Fe Mossbauer Study 

H. St. C. O'Neill; H. Annersten,** 
and D. Virgo 

The two extreme cation distributions in 
the spinel structure are the so-called nor- 
mal configuration A [Bii O4 and the inverse 
configuration B[AB]04 where the [ ] refer 
to the octahedrally coordinated cations and 
the remaining cations are in tetrahedral 
coordination. Disordered configurations of 
both these extreme arrangements can be 
represented as A\. x B X [A X #2-x]04. The 
degree of disorder can also be discussed in 
terms of the inversion parameter x, defined 
as the fraction of B cations occupying the 
tetrahedral sites. 

A general thermodynamic model of the 
cation distribution in spinels has been pro- 
posed by O'Neill and Navrotsky (1983, 
1984). The basic tenet of this model is that 
the equilibrium cation distribution is re- 
lated to the free energy of disordering in the 
following way, 



•RT\A—^ U^SL 

\(l-x)(2-x)j \ dx Itj,, 



N 



CD 



where AGd is the change in the non-con- 
figurational free energy of disordering. 

In their model, the free energy term 
AGd was shown to consist of an enthalpy 



*Bayerisches Geoinstitut, Universitat Bayreuth, 
Germany 

** Department of Mineralogy and Petrology, Insti- 
tute of Geology, University of Uppsala, Sweden 



94 



CARNEGIE INSTITUTION 



term, AHd which takes a quadratic depen- 
dence on x (AHd = OUC+ Pjc^). There was 
also a non configurational entropy term 
ASd- The non-linear nature of the enthalpy 
term implies that cation site preference will 
depend on the degree of inversion of the 
spinel into which it is substituting. Previ- 
ously it had been proposed (Navrotsky and 
Kleppa, 1967) that the site preference en- 
thalpies in spinels were independent of 
both temperature and the degree of order in 
the spinel structure. 

The experimental basis for the proposed 
model of non-linear enthalpy of disorder- 
ing in spinel solid solutions is limited 
(O'Neill and Navrotsky, 1983, Nell et aL, 
1989). Therefore, the present ordering and 
disordering experiments on MgFe204 have 
been specifically designed to test the above 
thermodynamic model. Magnesio-ferrite 
is an ideal composition in which to carry 
out such tests, since it shows a relatively 
large change in equilibrium cation distri- 
bution over a wide temperature range, it is 
stable in air, and the degree of inversion can 
be determined by several techniques such 
as x-ray refinement, and 57 Fe Mossbauer 
spectroscopy. Data in the literature are 
viewed as unsatisfactory, in view of the 
possibility either that the samples studied 
were not stoichiometric or that the cation 
site populations do not correspond to the 
annealing temperatures. 



Experimental 

MgFe204 was synthesized in air using 
a sodium tungstate flux. The oxide compo- 



nents consisting of 2 g MgO, 4 g Fe203, 20 
g Na2W04, and 2 g WO3 were melted in a 
Pt crucible at 1260 C and then cooled at 
6°C per hour to either 950°C (first batch) or 
900°C (second batch), at which tempera- 
tures the melt was crystallized for approxi- 
mately 12 hours before the final cooling to 
room temperature. The sodium tungstate 
flux was dissolved in warm water. Excess 
MgO was removed in dilute nitric acid. 
The MgFe204 run products consisted of 
euhedral, octahedra of reddish-brown 
spinel, -10 pxn in size. 

The acid cleaned MgFe204 was 
analysed by ICP and by two different elec- 
tron microprobe systems, one at the 
Bayerishes Geoinstitut and the other at the 
University of Uppsala. The MgO/Fe203 

values were respectively 1.014 ± 
0.019,1.014±0.009andl.028±0.015.All 

three sets of analyses indicate that MgFe204 
is stoichiometric within the analytical un- 
certainty (ie. within 2 standard deviations). 
Ideal stoichiometry is assumed in the fol- 
lowing discussion of the x-ray and 57 Fe 
Mossbauer results. 

Both disordering and ordering experi- 
ments were carried out in vertical quench 
furnaces in the range 400-1250°C. by heat- 
ing 40-100 mg of the starting material in 
platinum capsules welded at one end and 
crimped at the other end. Temperatures 
were controlled to ±0.5 °C, and run dura- 
tions were up to 50 days. The heated 
samples were quenched in water. Effective 
quench times calculated by extrapolation 
of the rate studies were of the order of a few 
milliseconds. 



GEOPHYSICAL LABORATORY 



95 



Lattice Parameter Measurements and 
Powder XRD Structural Refinements 

Lattice parameters were measured us- 
ing CoA'ai radiation on a STOE focusing 
diffractomer with a curved Ge monochro- 
mator. A position-sensitive detector of 8° 
20 width and 0.5° data interval was used to 
collect the diffraction data over the range 
40-120° 2£ NBSSi (0 = 5.43087) was used 
as an internal standard. Peak positions 
were determined by fitting the peak profile, 
and the observed 20 's were adjusted by 
using a linear correction based on the six 
main Si peaks in the two regions of interest. 
The spinel lattice parameter a was then 
determined by a weighted least squares 
refinement of the positions of the ten most 
intense spinel peaks. The mean internal 
estimate of one standard deviation of a is 
calculated to be ± 0.00015. An external 
estimate, based on determinations of a for 
samples equilibrated at 600, 650 and 700° C 
and heated for times that exceed that re- 
quired for equilibrium gives estimates in 
theuncertainty of0of±O.OOOll toO.00013. 

The XRD structural refinements were 
measured on data using the same 
diffractometer but with Mo/foci radiation 
and over the range in 20 from 7 to 80°. Data 
were collected from five consecutive scans 
and the scans were summed. Crystal struc- 
ture refinements were made using two con- 
trasting methods which are more fully dis- 
cussed by O'Neill et al. (1991a). Here, 
results for full profile Rietveld-type refine- 
ments using the program DBW (Wiles and 
Young, 1987) are considered (see also 
O'Neill et al, 1991b) The structure was 
refined in space group Fd3m with 8(a) 



andl6(d) cation sites and 32(e) oxygen 
sites all fully occupied. The refined quanti- 
ties were (1) a scale factor, (2) the Pearson 
VII profile shape parameters in which the 
component m could vary as m = a + b/20 , 
(3) an asymmetry parameter for peaks with 
2q < 34°, (4) a 20 zero parameter, (5) Peak 
halfwidth function of the form FWHM 2 = 
UtmO + VtmO + W, (6) the unit cell 
constant a (7) the oxygen positional param- 
eter m , (8) the inversion parameter, x, and 
(9) isotropic temperature factors for either 
cation sites or oxygen, or both. Refine- 
ments were also made for the cases where 
the raw data background was either in- 
cluded or subtracted. Bragg and total pat- 
tern residuals for the equilibrated samples 
were in the ranges 1 .97-3.63 and 2.89-4.57, 
respectively. There is no systematic corre- 
lation between these residuals and the de- 
gree of inversion. 



57Fe Mossbauer Spectroscopy 

Conventional transmission spectra were 
collected in a 512 multi-channel analyzer 
operated in conjunction with a constant 
accelerator electro-mechanical drive unit. 
57Co in Rh was used as a source. A calibra- 
tion spectrum for natural iron at room tem- 
perature was simultaneously recorded at 
the other end of the vibrator unit. The two 
mirror-symmetric spectra typically con- 
tained 0.6-0.8 x 10^ counts/channel, and 
the counts/channel for each half of the 
spectral data were averaged before analy- 
sis by a least-squares fitting program. 

The room temperature spectra of 
magnesio-ferrite (Neel point around 600 



96 



CARNEGIE INSTITUTION 



100.0 



o 99.0 



8 98.0 

jQ 

CO 

§ 97.0 

c 

o 

w ~~ ~ 

<i> 96.0 



95.0 - 




10 



•5 5 

Velocity, mm/s 



10 



o 
w 
-Q 

CO 



cz 
CO 
c 
o 

CD 
DC 



100.0 



o 99.0 



98.0 - 



E 97.0 - 



96.0 
95.0 




-10 




Velocity, mm/s 



10 



Fig.54. 57 Fe Mossbauer spectra of stoichiometric MgFe204 measured in an 
external field of 4.5 Tesla. The absorption patterns labeled A and B refer to Fe 3+ 
in the/4 and£-sites, respectively. Upper spectrum is MgFe204 heated at 1050°C, 
35 mins., Absorber temperature = 171 K. Lower spectrum is MgFe2C>4 heated 
500° C, 8 days, Absorber temperature = 189 K. 



K) indicates that there is almost complete 
overlap of the magnetically split patterns 
due to Fe 3+ in the tetrahedrally and octahe- 
drally coordinated sites. Accordingly, the 



spectral data were collected in the presence 
of an externally applied magnetic field 
using a superconducting magnet, cooled to 
4.2 K to produce a magnetic field of 4.5 



GEOPHYSICAL LABORATORY 



97 



Tesla. In these experiments, the external 
magnetic field is orientated parallel to the 
propagation of the gamma ray and thus the 
intensity of the transitions AMi =0 vanish; 
in addition, the magnetic hyperfine field at 
the tetrahedral site increases while that at 
the octahedral site decreases. Thus, each 
six-line pattern of the 298 K spectra of 
MgFe204 is reduced to four lines in the 
presence of the external field (cf Fig. 54). 
The absorbers were made by mixing 
powdered samples of equilibrated and 
quenched MgFe204 with -100 mg of 
thermo-setting plastic powder, (cf Virgo 
and Hafner, 1969). The absorber density 
was 7 mg Fe/cm 2 - The absorber was cooled 
in the cryostat to temperatures in the range 
1 2- 1 7 1 K; this procedure minimized differ- 
ences in the recoil-free fraction at the A and 
B sites. 

Least squares fitting of the spectral data, 
carried out using lines of Lorentzian shape, 
incorporated equal half- widths of each four- 
line pattern, and the intensity ratios were 
independent of whether the line intensities 
were unconstrained or whether the mag- 
netically split lines 1, 3, 4, and 6 were 
constrained in the ratio 3:1:1:3. It is evi- 
dent from Fig. 54 that the absorption due to 
Fe 3+ in the octahedrally coordinated sites 
is significant broader compared to that for 
Fe 3+ in the tetrahedrally coordinated sites. 
It is also evident from Fig. 54 that the half- 
width of the Z?-site absorption increases 
with increasing disorder. For the sample, 
heat treated at 1050°C, four hyperfine split 
patterns were refined in the fit of Fe 3+ (#- 
site) whereas only three such patterns were 
required for the remaining heat-treated and 
equilibrated samples in order to obtain sta- 



8.41 



< 
._- 8.40 

CD 
CD 

E 

cfl 

co 8.39 

Q. 

CD 
O 

'& 

CO 



8.38- 



8.37 



I I I I I I I I 
_ MgFe 2 4 


■ ■ ■ ■ 

■ 
■ 


— ■ — 
■ 


■ 


■ 


■ 


■ 


_ ■ _ 


i i i i i i i i 



400 



600 800 1000 1200 

Temperature of anneal, °C 



Fig. 55. Lattice parameter (a) measurements on 
heat-treated and quenched samples of stoichio- 
metric MgFe2C>4. 

tistically meaningful fits. For all values of 
x only a single hyperfine pattern was fitted 
to the absorption due to Fe 3+ in the tetrahe- 
dral site. 



Approach to Equilibrium 

The lattice parameters were used to 
monitor the approach to the equilibrium 
cation distribution in both disordering and 
ordering experiments. It can be shown 
from the results of this study that a varia- 
tion of ±0.0001 in a corresponds to a 
variation in x of about ±0.001. The cell 
constants for samples annealed and then 
quenched at 50° intervals in the range 450- 
1250°C are shown plotted in Fig. 55. Equi- 
librium is achieved on a time scale of about 
30 days at 450°C and less than five minutes 
at temperatures of 700°C and above. For 
equilibrium compositions in the range 550- 
700° C, equilibrium was verified by rever- 
sal experiments whereby the equilibrium 
values of x, at fixed temperature, were 
determined with starting material that was 



98 



CARNEGIE INSTITUTION 



both disordered and ordered with respect to 
the equilibrium value (O'Neil , 1991c). 

The data plotted in Fig . 5 5 show a smooth 
increase in a as a function of temperature, 
although there is only a small increase in a 
above 1050°C. Itisofconcernatthesehigh 
temperature that the rate of ordering is so 
fast in the high temperature experiments 
that the equilibrium distribution of Mg and 
Fe3+ is not quenched-in. There is also the 
possibility that there are deviations from 
the stoichoimetric composition at high tem- 
perature. While both those possibilities are 
discussed in more detail by O'Neill et al. 
(1991b), it is significant that the thermody- 
namic model to be discussed below and 
established from a fit to the cation distribu- 
tion on samples annealed at temperatures 
of less than 1000°C does, in fact, predict 
that there will only be comparatively small 
changes in ao at temperatures above 1 100°C. 



of the B -values with the degree of inver- 
sion. 

The values of x are shown plotted against 
the annealing temperature in Figure 56. As 
expected there is a smooth change in the 
cation distribution from nearly inverse at 
450° C to a more random configuration at 
high temperature. 

The spectra of MgFe204 in an applied 
field are similar to that reported by Sawatsky 
et al. (1969). Significantly, the absorption 
patterns due to Fe 3 + in the distinct crystal 
sites are well resolved (Fig. 54). The iso- 
mer shift values for Fe 3 + in tetrahedral 
versus octahedral coordination are 0.31- 
0.37mm/sec and 0.43-0.49 mm/sec, respec- 
tively. 

The broadening of the Z?-site absorption 
relative to the A -site absorption, noted pre- 
viously, can now be similarly explained as 
for other inverse spinels (Sawatsky et al., 



Cation Distributions in MgFe204: XRD 
and 57 Fe Mossbauer data. 

The structural refinements in which 
separate isotropic temperature factors were 
refined for each cation site and for oxygen 
gave the lowest values for RBragg and Rf 
although these statistical criteria were only 
slightly improved over a model wherein an 
average temperature factor for both cation 
sites and for oxygen are refined. The values 
for the temperature factors deceased in the 
order Btet < B ct <B oxygen, and the mean 
values are Btet = 0.304 ± 0.029 B ct = 
0.360 ± 0.023 and B 0X ygen = 0.507 
±0.03 1 A 2 . There is no systematic variation 



0.90 


I 

- 1% 


1 1 


1 1 


1 1 1 

S XRD 

o Mossbauer 


0.80 




O^i 




- 


n ir\ 


1 


1 1 


o 
1 1 


i i . T 



400 



600 



800 
Temperature, 



1000 



1200 



Fig. 56. A comparison of thermodynamic models 
for cation disordering in MgFe204 a) a two term 
model with no excess non-configurational en- 
tropy of disorder (dashed curve) and b) three term 
model, including an excess entropy term (unbro- 
ken curve). The XRD data are plotted with one 
standard deviation error bars; uncertainty in the 
Mossbauer data is 0.008 to 0.0 1 2 and is not shown 
for clarity. The three XRD data from 1 100 to 
1200°C (open symbols) were not included in the 
regressions, because of the possibility of a small 
oxygen deficiency in these samples. 



GEOPHYSICAL LABORATORY 



99 



1969). In the inverse MgFe204, each tetra- 
hedral site is surrounded by twelve nearest 
Fe 3+ neighbors. On the other hand, each 
octahedral site is surrounded by only six 
tetrahedral Fe 3 + neighbors. It is proposed 
that the broadening of the B-site absorption 
lines is due to the different number of 
probable distributions of Fe 3+ and Mg on 
the six nearest neighbor A -sites. Quite the 
opposite effect is proposed for the A-site 
Fe 3 +, because there is no apparent line 
broadening at this site. For MgFe204 with 
x ~ 0.66, the statistical probabilities that the 
B sites have distributions of 6 Fe, 5 Fe 1 
Mg, 4 Fe 2Mg and 3 Fe 3Mg in the next 
nearest neighbor A sites are 0.24, 0.39, 
0.26, and 0.09. For samples that are nearly 
inverse with x ~ 1.0, the corresponding 
probabilities are 0.78, 0.20, 0.02, 0.0. Thus, 
the increased halfwidth of the 5-site with 
increasing disorder (cf Fig. 54) is reason- 
ably explained in terms of additional hy- 
perfine fields due to the next-nearest-neigh- 
bor effect. 

The values of the inversion parameters 
calculated from the area ratios as deter- 
mined from the absorption spectra are com- 
pared with the values from the Rietveld 
refinements in Fig. 57. The mean differ- 
ence in the values of x is 0.0056 with a 
standard deviation of 0.0102. Thus, the 
5 1 Fe Mossbauer data are in excellent agree- 
ment with the XRD data, although it is 
noted that this agreement would become 
less satisfactory if fully ionized atom scat- 
tering factors were used in the structural 
refinement or if isotropic temperature fac- 
tors had not been separately refined for 
both cation sites, or if, in fact, only a single 
temperature factor was refined. 



Thermodynamic model. 

The equilibrium values of x in the tem- 
perature range 450-1050°C determined 
from the XRD Rietveld refinements and 
the 57pe Mossbauer data have been fitted 
to the expression 



-ln# = cc M S-Fe3+ + 2p;t, 



(2) 



where K is the distribution coefficient 
(Navrotsky and Kleppa, 1967). The values 
of x were weighted according to their stan- 
dard deviation, namely ±0.008 to ±0.012 



Comparison of Mossbauer and XRD 



0.90 



To 
a> 

jD 

<n 

o 0.80 



0.70 




1:1 line 



0.70 0.80 

x (Rietveld refinements) 



0.90 



Fig. 57. Comparison of the powder XRD determi- 
nations of x from the Rietveld refinements with the 
corresponding values determined from the 57 Fe 
Mossbauer spectra. Data are plotted with one 
standard deviation error bars. 



for Mossbauer data and ±0.004 for the 
XRD data. The results of the fit for this two- 
term model, shown plotted in Fig. 56, are 
a M g -Fe3+ = 26.6 ± 0.4, p = -21.7 ± 0.3 kJ/ 
mol and X^ v = 1.95. There is an improve- 
ment in the fit if an additional term repre- 
senting a nonconfigurational entropy of 
mixing is included. Equation (2) becomes 

-In K = a M g- p e 3+ - 7oMg-Fe3+ + 2 p;t , (3) 



100 



CARNEGIE INSTITUTION 



with a Mg-Fe3+ = 16>9 + 2.5 kJ/mol and 
a Mg-Fe3+ = .2.67 ± 1.52 andX^ = 1.10. It 
is of interest in the latter case that the values 
of x for the heat treated samples at 1100, 
1 150 andl250°C are in agreement with the 
predicted values using the three-term model. 
This latter result is significant, of course, in 
terms of whether the high temperature cat- 
ion distributions have been quenched at the 
respective annealing temperatures. 

In the formalism of the thermodynamic 
model proposed by O'Neill and Navrotsky 
[ 1983, 1984; equation (1)] the nonconfigu- 
rational entropy term inferred above is taken 
to refer to a vibrational energy contribution 
and/or the effect of short-range ordering 
(O'Neill and Navrotsky, 1983). In the 
latter case it is relevant that significant 
positional disorder of Fe 3+ and Mg on the 
B sub-lattice is inferred from the 57p e 
Mossbauer spectra (cf Fig. 54). 

In the literature, thermochemical data 
required to evaluate the proposed thermo- 
dynamic model are sparse. The enthalpy 
associated with the change in cation distri- 
bution in MgFe204 in the temperature range 
700-1200°C has been measured by trans- 
posed-temperature-drop calorimetry 
(Navrotsky, 1986; Table 3). The experi- 
mental value of -5.5 kJ does not agree with 
the value of — 1 kJ predicted from the 
present study [equations (2), (3)]. How- 
ever, it should be noted that the stoichiom- 
etry of the MgFe204 used in the thermo- 
chemical measurement was not specified. 
Further evaluation of the proposed model 
for cation disordering in spinels must there- 
fore await thermochemical data on well- 
characterized samples. 



References 



Navrotsky, A. and O. J. Kleppa, The thermody- 
namics of cation distributions in simple spinels. 
/. Inorg. Nucl. Chem., 29, 2701-2714, 1967. 

Navrotsky, A., Cation-distribution energetics and 
heats of mixing in MgFe204-MgAl204 } 
2nFe204-2nAl204 and NiAl204-2nAl204 
spinels: Study by high- temperature calorim- 
etry. Amer. Mineral, 71, 1160-1169, 1986. 

Nell J., B. J. Wood, T. O. Mason, High-tempera- 
ture cation distributions in Fe304-MgAl204- 
MgFe204-FeAl204 spinels from 
thermopower and conductivity measurements. 
Amer. Mineral, 74, 339-351, 1989. 

O'Neill, H. St. C, W. A. Dollase, and C. R. Ross 
II, Temperature dependence of the cation dis- 
tribution in nickel aluminate (NiAl204) spinel: 
a powder XRD study. Submitted to Physics 
and Chemistry of Minerals, 1991a. 

O'Neill, H St. C, H. Annersten, and D. Virgo D., 
The temperature dependence of the cation 
distribution in magnesioferrite (MgFe204) 
from powder XRD structural refinements and 
Mossbauer spectroscopy. Submitted to Amer. 
Mineral, 1991b. 

O'Neill, H. St. C. The rates of cation order- 
disorder in MgFe204 and MgAl204 spinels. 
In preparation, 1991c. 

O'Neill, H. St. C, and A. Navrotsky, Simple 
spinels: Crystallographic parameters, cation 
radii, lattice energies, and cation distribution. 
American Mineralogist, 68, 181-194, 1983. 

O'Neill, H. St. C, and A. Navrotsky, Cation 
distributions and thermodynamic properties 
of binary spinel solid soultions. Amer. Min- 
eral, 69, 733-753, 1984. 

Sawatsky, G.A., F. Van der Wande, and A. H. 
Morrish, Mossbauer study of several ferri- 
magnetic spinels. Physical Review, 187, 747- 
757, 1969. 

Virgo, D., S. S. Hafner, Fe^ + -Mg order-disorder 
in heated orthopyroxenes. Mineral Soc. Amer. 
Spec. Paper 2, 67 -87, 1969. 

Wiles, D.B., R. A. Young, A new computer pro- 
gram for Rietveld analysis of X-ray powder 
diffraction patterns. Journal App. Cry stall., 
14, 149-151, 1987. 



GEOPHYSICAL LABORATORY 



101 



Crystallography — Mineral Physics 



Predicted High-Pressure Mineral 

Structures 

with Octahedral Silicon 

Robert M. Hazen and Larry W. Finger 

Silicates are the most common minerals 
on the earth's surface, and they probably 
dominate throughout the Earth's mantle. 
Many hundreds of silicate structures have 
been determined and catalogued (e.g., 
Liebau, 1985), but only about 50 different 
structures account for the vast majority of 
all crustal silicates (Smyth and Bish, 1988). 
A common feature of all these low-pres- 
sure mineral structures is the presence of 
silicon cations exclusively in 4-coordina- 
tion [lYISi by oxide anions. 

High-pressure experiments demonstrate 
that all common crustal silicates undergo 
phase transitions to new structures with 6- 
coordinated silicon, [ VI ]Si, at pressures 
between 8 GPa (for pure Si02) to about 30 
GPa, which corresponds to the pressure at 
the top of the lower mantle. Mineral physi- 
cists identify silicon coordination number 
as a major crystal chemical difference be- 
tween the crust and lower mantle: silicon is 
virtually all four coordinated above about 
250 km, but is entirely six coordinated 
below 670 km. The earth's transition zone, 
on the other hand, is marked by the appear- 
ance of a group of high-pressure silicates 
with both [IVlSi and t VI lSi. The stability of 
these minerals is apparently confined to a 



rather narrow pressure range from approxi- 
mately 10-30 GPa. Within these limits, 
however, are silicate structures of remark- 
able complexity and great topological in- 
terest. 

The objectives of this review are to 
tabulate all known high-pressure silicate 
structures with six coordinated or mixed- 
coordinated silicon, to identify crystal 
chemical systematics among these struc- 
tures, and to predict additional high-pres- 
sure silicate structure types. 

There are only a dozen known high- 
pressure structures with Si06 polyhedra 
(Table 13). These silicates can be divided 
conveniently into two groups. Above about 
25 GPa, corresponding to the Earth's lower 
mantle, all silicates studied to date are 
observed to transform to one of seven dense 
structures, in which all Si is 6-coordinated. 
These structures — rutile, perovskite, il- 
menite, hollandite, calcium ferrite, 
pyrochlore, and K2NiF4 — are well known 
room-pressure topologies for transition 
metal oxides. In the high-pressure silicate 
isomorphs silicon occupies the octahedral 
transition metal site, while other cations 
may adopt six or greater coordination. 

At pressures between about 10 and 20 
GPa (in the Earth's transition zone) a sec- 
ond group of silicates forms with mixed 4- 
and 6-coordination. These phases include 
silica-rich modifications of the well known 
garnet, pyroxene, and wadeite structures, 
as well as complex new magnesium-bear- 



102 



CARNEGIE INSTITUTION 



Table 13. Compositions and calculated densities of high-pressure silicates with Si06 octahedra. 



Composition (Mineral Name) Structure Type 


p calc* 


References 






(g/cm 3 ) 






A. High-Pressure phases 


with Si06 


groups only 


SiC>2 (Stishovite) 


Rutile 


4.29 


Hill etal. (1983) 


CaSi03 


Cubic Perovskite 


4.25 


Mao etal. (1989) 


MgSi0 3 


Ortho Perovskite 


4.10 


Horiuchiertf/. (1987) 


MgSi0 3 


Ilmenite 


3.81 


Horiuchi eet al. (1982) 


ZnSi03 


Ilmenite 


5.25 


Ito and Matsui (1974) 


KAlSi 3 8 


Hollandite 


3.91 


Yamadaera/. (1984) 


BaAl 2 Si 2 08 


Hollandite 


5.3 


Reid and Ringwood (1969) 


CaAl 2 Si 2 08 


Hollandite 


3.9 


Madon era/. (1989) 


NaAlSi0 4 


Calcium Ferrite 


3.91 


Yamada etal. (1983) 


Sc 2 Si 2 O v 


Pyrochlore 


4.28 


Reid etal. (1977) 


In2 2 Si 2 7 


Pyrochlore 


6.34 


Reid etal. (1977) 


Ca 2 Si 2 7 


K 2 NiF 4 


3.56 


Liu (1978) 




B. High-Pressure phases with SiC>6 + 


Si04 groups 


MgSi0 3 (Majorite) 


Garnet 


3.51 


Angela al. (1989) 


MnSi0 3 


Garnet 


4.32 


Fujino etal. (1986) 


Na(Mgo.5Sio.5)Si 2 6 


Pyroxene 


3.28 


Angel etal. (1988) 


K 2 Si40 9 


Wadeite 


3.09 


Swanson and Prewitt (1983) 


Mgi4Si50 2 4 


Anhydrous Phase B 3.44 


Finger etal. (1991) 


Mg i2 SUOi 9 (OH) 2 


Phase B 


3.37 


Finger etal. (1991) 



*p calc = density calculated from unit-cell parameters at room pressure and temperature. 



ing phases designated "phase B" and "an- 
hydrous phase B." [A third group of room- 
pressure [VllSi silicates, detailed by Finger 
and Hazen (1991) but not considered here, 
includes silicon phosphates with relatively 
open framework structures.] 

The twelve high-pressure structures 
listed in Table 13 display a wide range of 
linkages between Si octahedra and other 
polyhedra. In stishovite, hollandite, and 
pyroxene a combination of edge- and cor- 
ner-sharing is observed, but in phase B and 
anhydrous phase B each Si octahedron 
shares all 12 edges with adjacent Mg octa- 
hedra. In pyrochlore, garnet, and wadeite 
the Si octahedra form part of a comer- 



linked framework, but additional cations in 
eight or greater coordination share edges 
and faces with the octahedra. Ilmenite pre- 
sents yet a different topology, with unusual 
face sharing between Mg and Si octahedra, 
as well as corner and edge sharing. Despite 
the differing polyhedral linkages, the size 
and shape of Si06 polyhedra are similar in 
all twelve high-pressure compounds. Poly- 
hedral volumes at room pressure, for ex- 
ample, vary by only about ± 4% from an 
average 7.67-A 3 value. All Si octahedra are 
close to regular (i.e., distortion indices are 
small) relative to the range observed for 
many divalent and trivalent cation octahe- 
dra. The observed tendency of silicon to 



GEOPHYSICAL LABORATORY 

Table 14. Predicted l VI ^Si structures that conform to more than one criterion. 



103 



Formula 


Structure 


1* 


2* 


3* 


4* 


5* 


(MgSi)0 2 (OH) 2 


Diaspore 


X 






X 




Ga^SiOg 


- 


X 






X 




Ga4Si7O20 


- 


X 






X 




MgioSi30i6 


Aerugite 




X 






X 


CaSi205 


Sphene 




X 


X 






MgSi(OH) 6 


Stottite/Gibbsite 




X 




X 




BaSU09 


Benitoite 




X 


X 






Fe2Si05 


Pseudobrookite 






X 


X 





* Criteria for predicting t VI lSi structures: 

1. Edge-sharing octahedral chains 

2. Germanate isomorphs 

3. Ti, Mn, and Fe oxides 

4. Substitution of (Mg + Si) for 2A1 

5. System Mg-Si-O-H 



adopt highly regular coordination leads us 
to conclude that silicon will not readily 
adopt exotic 5- or 7-coordination or highly 
distorted 4- or 6-coordination groups at 
high pressure in either crystalline or amor- 
phous condensed phases. 

Five systematic relations among the 
structures in Table 1 3 can be used to predict 
other possible high-pressure silicates (see 
also Table 14). Each of these criteria can be 
used to predict other potential [VI] Si phases. 







Fig. 58. Relationships among the rutile (A), 
IrSe2(B), ramsdellite(C), hollandite (D), 
psiomelane (E), and a hypothetical composite 
structure (F), after Bursill (1979). 



1. Three structure types (rutile, 
hollandite, calcium ferrite) are formed from 
edge-sharing chains of silicon octahedra. 
This relation points to other likely structure 
types, all of which incorporate edge-shar- 
ing octahedral chains linked to adjacent 
strips by corner sharing, as systematized by 
Wadsley (1964), Bursill and Hyde (1972), 
and Bursill and Hyde (1979). Rutile has 
single chains, leading to 1 x 1 square chan- 
nels, while hollandite and calcium ferrite 
have double chains, yielding larger chan- 
nels. Many similar octahedral chain struc- 
tures, such as ramsdellite (1x2) and 
psilomelane (2 x 3) are also known (Fig. 
58), and each of these could provide a 
topology suitable for silicon in 6-coordina- 
tion. 

2. Nine of the twelve known [VllSi 
high-pressure structure types were first 
synthesized as germinates at lower pres- 
sures. A systematic search of the Inorganic 
Crystal Structure Data Base (ICSD, FIZ 
Karlsruhe distributors) for germanates with 
[ vr lGe in systems containing the additional 
cations Na, K, Mg, Fe, Ca, Al, Ti, Si, and P 



104 



CARNEGIE INSTITUTION 



revealed 25 structure types, only nine of 
which have known silicate analogs (Finger 
andHazen, 1991). A number of these com- 
pounds, including Fe4Ge209, FesGe30i8, 
CaGe205,Ca2Ge70i6,Ca4Ge30io(H20), 
and K2BaGegOi8, are good prospects for 
high-pressure silicate analogs. 

3. All seven high-pressure t VI lSi struc- 
tures without tetrahedral Si are isomorphs 
of room-pressure oxides with trivalent or 
tetravalent transition metals (Ti, Mn, or Fe) 
in octahedral coordination. Structures of 
other binary oxides with octahedral tita- 
nium, manganese, or iron may also repre- 
sent possible topologies for mantle miner- 
als. Particularly relevant in this context are 
the structures of CaSi205 with the sphene 
structure, Fe2Si05 or Al2Si05 with the 
pseudobrookite structure, and CaSi409 with 
the benitoite structure. 

4. High-pressure ilmenite, garnet, and 
pyroxene forms of magnesium-bearing sili- 
cates are all related to room-pressure phases 
by the substitution of octahedral Mg and Si 
for a pair of aluminum cations. Similar 
substitutions might occur at high pressure 
in several other common rock-forming 
minerals, including kyanite, staurolite, 
pseudobrookite, lawsonite, cordierite, 
clinozoisite, gibbsite, and diaspore. Note 
that this substitution scheme will not work 
for many common aluminum-bearing min- 
erals with mixed 4- and 6-coordinated alu- 
minum. The substitution in muscovite 
[K[VI]Al 2 [ IV ](AlSi3)Oio(OH 2 )], for ex- 
ample, would yield the magnesian mica 
celadonite,K[VI](MgAl)[lV]Si 4 Oio(OH)2, 
in which all Si is tetrahedrally coordinated. 
Octahedral Al, thus, must constitute more 



than two-thirds of all aluminum to produce 
a [^1] Si phase by the substitution 2A1 — > 
(Mg + Si). 

5 . Finger and Pre witt ( 1 990) documented 
the close structural relations among a num- 
ber of hydrous and anhydrous magnesium 
silicates, and used those systematics to 
propose several as yet unobserved struc- 
tures, including high-pressure hydrous 
phases with octahedral silicon. They rec- 
ognized that several known phases, includ- 
ing chondrodite, humite, forsterite, phase 
B, and anhydrous phase B, are members of 
a large group of homologous magnesium 
silicates that can be represented by the 
general formula: 

m[M g4 « + 2[ ][V JSi2«08«(OH)4]Mg6 /2+ 4- 
2mod(n,2) [YI lSin+mod(n,2)OSn+4 

where mod{n,2) is the remainder when n is 
divided by 2. Finger and Prewitt (1990) 
examined cases where n = 1 ,2,3,4, °° and m 
= 1,2, oo. Structures with octahedral silicon 
result for all cases where m is not infinity. 
Of special interest is the proposed structure 
of "superhydrous phase B," a compound 
predicted by the logical progression from 
Mgi4Si5024 (anhydrous phase B) to 
Mgi2Si40i9(OH)2 (phase B) to 
MgioSi30i4(OH)4. Gasparik (1990) sug- 
gested that an as yet unanalyzed hydrous 
magnesium silicate synthesized at 1 8.6 GPa 
and 1600 °C possesses this structure, and 
further studies on that material are in 
progress. 

Several structure types appear to follow 
two of the five very different systematic 
trends. These structures, therefore, de- 



GEOPHYSICAL LABORATORY 



105 



mand further study. Of special interest to 
earth scientists are CaSi20s with the titanite 
structure, Fe2SiOs with the pseudobrookite 
structure, and Mg ioSi30 16 with the aerugite 
structure. Each of these phases, or their 
isomorphs with other cations replacing Ca, 
Mg, and Fe, might be represented in the 
Earth's mantle. In fact, Stebbins and 
Kanzaki (1991) mention the existence of 
titanite-type CaSi20s in some of their run 
products, though identification of this phase 
was provisional. 

Also worthy of further study are the 
proposed hydrous phases MgSi02(OH)2 
and MgSi(OH)6, which are isomorphs of 
diaspore and stottite, respectively. Such li- 
nen phases would be expected to occur 
only locally in the earth's deep interior, but 
their presence, integrated over the Earth's 
volume, could represent a major respository 
of water. 

Most common rock-forming cations, 
including Na, Mg, Fe, Ca, Mn, Al, Ti, and 
Si, are small enough to fit into the tetrahe- 
dral or octahedral interstices of a close- 
packed oxygen net. However, the presence 
of many other cations, including H, B, K, 
Rb, Pb, rare earths, and U, could disrupt the 
close-packed array and lead to other, as yet 
unrecognized, structure types. The gal- 
lium and barium silicates, Ga4SiOs, 
Ga4SiyO20, and BaSi409, are just three of 
the dozens of possible new [VlJSi struc- 
tures likely to be observed as high-pressure 
investigations extend beyond the traditional 
rock-forming elements. These structures 
are not likely to play a significant role in 
mantle mineralogy, but they will provide a 
more complete understanding of the crys- 
tal chemistry of octahedral silicon. 



Conclusions 

Is the earth's deep interior 
mineralogically simple? Are there only a 
few dominant structure types, or is there an 
unrecognized complexity in the crystal 
chemistry of octahedral silicon? There are 
hundreds of different crustal silicates with 
PYlSi, but only a dozen high-pressure [VI]Si 
structures have been produced. This dis- 
parity may reflect the relatively small num- 
ber of high-pressure studies, but it also 
arises, at least in part, from the nature of 
oxygen packing. Numerous crustal sili- 
cates, from the commonest minerals quartz 
and feldspar to the dozens of zeolites and 
other framework silicates, possess open, 
low-density topologies with correspond- 
ingly loose packing of oxygen. There are 
no obvious limits to the variety of silicates 
based on irregular oxygen packing. 

Volume constraints imposed by high 
pressure, however, favor structures with 
approximately close-packed oxygens. 
These restrictions on anion topology re- 
duce the number of possible cation con- 
figurations as well, and it is thus antici- 
pated that the number of different struc- 
tural topologies in the earth's deep interior 
will be much smaller than at the surface. 
Dense, close-packed, and for the most part 
high-symmetry structures, such as those 
represented by the seven known topologies 
with all t vr lSi, will predominate. Never- 
theless, within these restrictions there ex- 
ists opportunity for considerable structural 
diversity owing to three factors - reversible 
phase transitions, cation positional order- 
ing, and modularity, particularly based on 



106 



CARNEGIE INSTITUTION 



different close-packed layer stacking se- 
quences. This potential diversity is only 
hinted at by the known phases. 

Several of the known high-pressure 
types, including perovskite, K2NiF4, and 
pyrochlore, can adopt numerous structural 
variants based on slight changes in lattice 
distortions and cation distribution. The 
perovskite structure, in particular, can un- 
dergo dozens of phase transitions based on 
octahedral tilting, cation ordering, cation 
displacements, and anion defects (Megaw, 
1973; Hazen, 1988). We must study pro- 
posed mantle phases at the appropriate 
conditions of pressure and temperature to 
document the equilibrium structural varia- 
tions. 

Close packing of oxygen leads to modu- 
lar structures, with certain features (e.g., 
edge-sharing octahedral chains of rutile; 
the double chains of hollandite; the comer- 
sharing octahedral sheets of perovskite; the 
face-sharing topology of ilmenite) that can 
link together in many ways to form ordered 
superstructures of great complexity. Such 
complexity was recognized by Wadsley 
( 1 964) and Bursill and Hyde ( 1 979) in their 
descriptions of modular rutile-hollandite- 
Ga203 structures, and it is realized in the 
homologous series including phase B, an- 
hydrous phase B, and several other struc- 
tures. Phase B, for example, is based on 
oxygen close packing, yet it has 40 inde- 
pendent atoms in its asymmetric unit to 
yield one of the most complex ternary 
silicates yet described. Variations on the 
phase B structure could be based on chang- 
ing the relative number and position of the 
two different structural layers, by introduc- 



ing other types of layers, or by staggering 
layers to produce clino- and ortho-type 
structures as observed in other close -packed 
systems, for example, in the biopyriboles 
as describedbyThompson(1978) and Smith 
(1 982). The structure could be further com- 
plicated by element ordering among the 17 
different cation sites as Al, Fe Ti, Mn, and 
other elements enter the structure in a natu- 
ral environment . 

The study of octahedrally-coordinated 
silicon is still in its infancy, yet clear trends 
are beginning to emerge from the scattered 
data on diverse structures and composi- 
tions. It is now evident that while silicate 
perovskite may be the predominant phase 
in the Earth's lower mantle, a number of 
other dense silicate phases will compete for 
elements such as K, Ba, Ca, and Al. It 
appears that the earth's transition zone will 
display the varied mineralogy of mixed 
t VI lSi and PV] Si silicates, including some 
of the most complex structures known in 
the mineral kingdom. And it is certain that 
a detailed understanding of the mantle must 
await studies of these fascinating phases at 
temperatures and pressures appropriate to 
the Earth's dynamic interior. 



References 

Angel, R. J., T. Gasparik, N. L. Ross, L. W. Finger, 
C. T. Prewitt, and R. M. Hazen, A silica-rich 
sodium pyroxene phase with six-coordinated 
silicon, Nature, 335, 156-158, 1988. 

Angel, R. J., L. W. Finger, R. M. Hazen, M. 
Kanzaki, D. J. Weidner, and R. C. Liebermann, 
Structure and twinning of single-crystal 
MgSiC>3 garnet synthesized at 17 GPa and 
1800°C, Amer. Mineral, 74, 509-512, 1989. 

Bursill, L.A. and B. G. Hyde, Structural relation- 
ships between P-gallia, rutile, hollandite, 



GEOPHYSICAL LABORATORY 



107 



psilomelane, ramsdellite, and gallium titanite 
type structures, Acta Cryst., B35, 530-538, 
1979. 
Bursill, L. A., and B. G. Hyde, Rotation faults in 
crystals, Nature (London) Phys. Sci., 240, 122- 
124, 1972. 
Finger, L. W., and R. M. Hazen, Crystal chemistry 
of six-coordinated silicon: A key to under- 
standing the Earth ' s deep interior, Acta Cryst., 
A47, 1991, in press. 
Finger, L.W. and C. T. Prewitt, Predicted compo- 
sitions for high-density hydrous magnesium 
silicates, Geophys. Res. Lett., 16, 1395-1397, 
1990. 
Finger, L. W., R. M. Hazen, and C. T. Prewitt, 
Crystal structures of Mgi2Si40i9(OH)2 (Phase 
B) and Mgi4Si5024 (Phase AnhB), Amer. 
Mineral, 76, 1-7,1991. 
Fujino, K., H. Momoi, H. Sawamoto, and M. 
Kumazawa, Crystal structure and chemistry of 
MnSi03 tetragonal garnet, Amer. Mineral., ' 
71, 781-785, 1986. 
Gasparik, T. Phase relations in the transition zone, 

/. Geophys Res., 95, 15751-15769, 1990. 
Hazen, R. M., Perovskites, Sci. Amer., 74-8 1 , June 

1988. 
Hill R.J., M. D. Newton, and G. V. Gibbs, A 
crystal chemical study of stishovite, /. Sol. 
State Chem., 47, 185-200, 1983. 
Horiuchi, H., M. Hirano, E. Ito, and Y. Matsui, 
MgSiC>3 (ilmenite-type): Single crystal x-ray 
diffraction study, Amer. Mineral, 67, 788-793, 
1982. 
Horiuchi, H., I. Eiji, andD. J. Weidner, Perovskite- 
type MgSiC>3: Single-crystal x-ray diffraction 
study, Amer. Mineral, 72, 357-360, 1987. 
Ito, E., and Y. Matsui, High-pressure synthesis of 
ZnSiC>3 ilmenite, Phys. Earth Planet. Int., 9, 
344-352, 1974. 
Liebau, F., Structural Chemistry of Silicates, 

Springer, New York, 1985. 
Liu, L., High pressure Ca2SiC>4, the silicate 
K2NiF4-isotype with crystalchemical and geo- 
physical implications, Phys. Chem. Minerals, 
3, 291-299, 1978. 
Madon, M, J. Castex, and J. Peyronneau, A new 
aluminocalcic high-pressure phase as a pos- 
sible host of calcium and aluminum in the 
lower mantle, Nature, 342, 422-425, 1989. 
Mao, H.K., L. C. Chen, R. J. Hemley, A. P. 
Jephcoat, Y. Wu, and W. A. Bassett, Stability 
and equation of state of CaSi03-perovskite to 
134 GPdiJ.Geophys.Res.,94, 17,889-17,894, 
1989. 
Megaw, H. D., Crystal Structures: A Working 
Approach, Philadelphia, W. B. Saunders, 1973. 
Reid, A. F., and A. E. Ringwood, Six-coordinate 
silicon: High pressure strontium and barium 
aluminosilicates with the hollandite structure, 
J. Solid State Chem. 1, 6-9, 1969. 



Reid, A. F., C. Li, and A. E. Ringwood, High- 
pressure silicate pyrochlores, SC2S12O7 and 
In2Si207, Solid State Chem., 20, 219-226, 
1977. 

Smith, J. V., Geometrical and Structural Crystal- 
lography, Wiley, New York, 1982. 

Smyth, J. R., and D. L. Bish, Crystal Structures 
and Cation Sites of the Rock-Forming Miner- 
als, Allen and Unwin, Winchester, Massachu- 
setts, 1988. 

Stebbins, J. F., and M. Kanzaki, Local structure 
and chemical shifts for six-coordinated silicon 
in high-pressure mantle phases, Science, 251, 
294-298, 1991. 

Swanson, D.K., and C. T. Prewitt, The crystal 
structure of K2Si VI Si3 IV C>9, Amer. Mineral, 
65,581-585,1983. 

Thompson, J. B., Biopyriboles and polysomatic 
series, Amer. Mineral, 63, 239-249, 1978. 

Wadsley, A. D.,Non-Stoichiometric Compounds, 
L. Mandelcoin, ed., p. Ill, New York, Aca- 
demic Press, 1964. 

Yamada, H., Y. Matsui, andE. Ito, Crystal-chemi- 
cal characterization of NaAlSi04 with the 
CaFe204 structure, Mineral. Mag., 47, 177- 
181, 1983. 

Yamada, H., Y. Matsui, andE. Ito, Crystal-chemi- 
cal characterization of KAlSi3Us with the 
hollandite structure, Mineral. J. (Japan), 12, 
29-34, 1984. 



SlMULTANOUS HlGH P-^ 

Diffraction Measurements 

OF (Fe,Mg)Si03-PEROVSKITE 

and (Fe,Mg)0 Magnesiowustite: 

Implications for Lower 

Mantle Composition 

Yingwei Fei, Ho-Kwang Mao, Russell J. 
Hemley, and Jinfu Shu 

(Fe,Mg)Si03-perovskite and (Fe,Mg)0 
magnesiowustite are likely stable phases in 
the Earth's lower mantle. The thermal prop- 
erties of those phases are of critical impor- 
tance for constraining the composition of 
the lower mantle. In this paper, we report 
simultaneous high P-T synchrotron x-ray 
diffraction measurements of (Fe,Mg)Si03 



108 



CARNEGIE INSTITUTION 



>n Mw 



C/) 

c 

CD 



Mw200 



Mw220 



Au200 



Au111 

V 



'•'. Au311 

\ 

'• Au220 Mw311 :".* 



19.8 GPa, 310 K 



/ 




Energy, keV 



Fig. 59. Representative energy-dispersive x-ray diffraction spectra (20 = 15°) of magnesiowustite 
(Mg.6Fe. 4 )0 at three different P-T conditions. 



perovskite and (Fe,Mg)0 magnesiowustite. 
These data provided the first direct mea- 
surements of the effect of pressure on the 
thermal expansivity of these minerals. In 
the discussion we also examine the impli- 
cations of these results for the composition 
and mineralogy of the lower mantle. 



Experimental Methods 

The samples used in the experiments 
are (Feo.4Mgo.6)0 magnesiowustite, syn- 
thesized from mixtures of Fe2C>3 and MgO 



at a temperature of 1573 K and an oxygen 
fugacity (fen) of lO 10 - 8 bar (Rosenhauer et 
a/., 1976), and (Feo.iMgo.9)Si03 perovskite, 
synthesized from synthetic pyroxene by 
laser heating at 40 GPa in diamond-anvil 
cell (Mao etal, 1991). 

The experiments were carried out using 
a high-temperature diamond-anvil cell, 
made of inconel, and synchrotron radia- 
tion. A nickel alloy (Rene 41) gasket with 
thickness of 200 Jim was preindented to a 
pressure of 17 GPa. The powder sample 
was placed in the 250-|im-diameter sample 
chamber, filling less than one-third of the 



GEOPHYSICAL LABORATORY 



109 



volume. Gold foil and ruby grains were 
placed in the sample chamber as pressure 
calibrants at high temperature (Anderson 
et al. , 1 989) and at room temperature (Mao 
etal. , 1 986), respectively. The sample cham- 
ber was then filled with neon gas at 200 
MPa in a high-pressure gas-loading device, 
and sealed at a pressure of 2 GPa. The neon 
served as a quasi-hydrostatic pressure trans- 
mitting medium over the pressure range of 
measurement. 

The sample was heated with an external 
platinum-wire resistance heater (Mao et 
al, 1991). The heater was placed on the 
cylinder of the cell. Temperatures were 
measured with a chromel-alumel thermo- 
couple, while pressures were determined 
by measuring the lattice parameter of gold. 
During the experiment, pressure usually 
decreases with increasing temperature at 
the rate of about 5 GPa/100 K. 

Polychromatic (white) wiggler synchro- 
tron x-radiation at the National Synchro- 
tron Light Source, Brookhaven National 
Laboratory was used for the energy-disper- 
sive x-ray diffraction measurements. The 
diffraction data were collected with an in- 
trinsic germanium solid-state detector at a 
fixed 20 angle of 15°(± 0.005°). The energy 
was calibrated by using known energies of 
x-ray emission lines (Ka and Kp) of Mn, 
Cu, Rb, Mo, Ag, Ba, and Tb. The 20 angle 
was calibrated by collecting the diffraction 
pattern of platinum. The experimental con- 
ditions were optimized by considering the 
x-ray beam size, data collecting time, and 
slit size for the detector. With a 60-|im 
beam spot, a complete diffraction pattern 
of magnesiowustite with reasonable peak 
counts can be obtained in about five min- 



utes. Figure 59 shows three typical spectra 
of magnesiowustite with internal standard 
gold collected at different P-T conditions. 
All diffraction patterns show at least four 
sharp diffraction lines, 111, 200, 220, and 
311, of magnesiowustite. For theperovskite, 
the diffraction was carried out using mono- 
chromatic synchrotron x-ray and film meth- 
ods (Mao et al., 1991). These techniques 
were used because of the need for high 
resolution to resolve multiplets in the dif- 
fraction patterns arising from the or- 
thorhombic distortion of the perovskite. 



Results and Discussions 

Lattice parameters of the 
magnesiowustite were determined from 
diffraction lines 111, 200, 220, and 311 by 
using a peak-fitting program. For 
perovskite, diffraction peaks (mainly 020, 
112, and 200) were measured by both 
manual and computerized film-reading 
methods. The experimental results are plot- 
ted in Figures 60 and 61 . The uncertainties 
in pressure result from the measurements 
of lattice parameter of gold which was used 
as an internal high-pressure calibrant at 
high temperature. The error of ± 0.0015 A 
in the lattice parameter of gold, which is the 
typical measurement uncertainty in the ex- 
periments, corresponds to about ± 0.30 
GPa in pressure. 

To construct a P-V-T equation of state 
from a combination of 1-bar thermal ex- 
pansion data, 300-K compression data, and 
simultaneous high P-T volume data, we 
express pressure in a general form 

/W) = /W300K) + />*, (1) 



110 



CARNEGIE INSTITUTION 



o 
o 

CO 
CL 

P 
> 




~l 

10 15 

Pressure, GPa 



Fig. 60. Volume at high P-T relative to the 300 K 
isotherm. Experimental data are from Mao et al 
[1991] (crosses); Knittle et al, [1986] (solid 
squares); and Wang et al, [1991] (open squares). 
The lines are calculated isotherms. 



o 
o 

CO 

a." 
> 



2.5-1 



2.0_ 



1.5- 



1.0- 



0.0- 




"3 

S93"*- 



600 



0, „,, 4- _^«76 






565 

5*3 




*"1 «• 

323 -}-31Q 



500K 



300K 



1^ 

20 



—T 
25 



- 1 - 
30 



Pressure, GPa 



Fig. 61. Volume at high P-T relative to the 300 K 
isotherm. Experimental data are from Fei et al 
[1991b] (crosses); and Suzuki [1975] (solid 
squares). The lines are calculated isotherms. 

where P(V,300K) can be expressed by the 
standard third-order Birch-Murnaghan form 



P = h 

2 



4#-(#I-^-^fe 



(2) 



where Kto and K T q are the isothermal bulk 
modulus and its pressure derivative at room 
temperature, respectively. The thermal pres- 
sure can be modeled in two ways as dis- 
cussed below. 

By using the thermodynamic identity, 
the thermal pressure can be calculated by 
integrating the oK v i.e., 



/ > th = 



f 

/300K 



aKjiXT 



(3) 



where a is the thermal expansivity and K T 
is isothermal bulk modulus at T and V of 
interest. To parameterize the thermal pres- 
sure, we start with an assumption that (BK^ 
dT) p is independent of temperature, and 
obtain (dKflT), = -2 .7 '(±0.3) x 10 2 GPa/K 
for magnesiowiistite and (dK^dT^p = - 
6.3(±0.5) x 10 2 GPa/K for perovskite by 
fitting our P-V-T data to the thermal pres- 
sure model. A detailed discussion of these 
results is given in Mao et al ( 1 99 1 ) and Fei 
etal (1991b). 

The Anderson-Griineisen parameter S T 
is commonly used to measure the change in 
thermal expansivity at high P-T. It is de- 
fined by (Anderson, 1967) 



<5r = 



3lna 



_ 1 



l M T \ 



[dlnV It aK A dT 



(4) 



The parameters, 5 r for magnesiowiistite 
and perovskite derived from our data are 
listed in Table 15. They decrease with in- 
creasing temperature below the Debye tem- 
perature and approach a constant value at 
high temperature. The S T values for 
magnesiowiistite and perovskite are 4.3 
and 6.5, respectively, above the Debye 
temperatures. 



GEOPHYSICAL LABORATORY 



111 



Table 15. Temperature variation of some thermodynamic parameters for magnesiowiistite and 
perovskite 







Perovskite 








Magnesiowiistite 




T,K 


a(106) 


K T , GPa 


ocKt 


dr 


q 


a(106) tf r ,GPa 


aKj 


Sr 


<7 


300 


21.90 


260.9 


57.15 


11.02 


7.48 


31.32 


157.0 


49.16 


5.49 


1.97 


400 


28.94 


254.6 


73.68 


8.55 


5.28 


35.87 


154.3 


55.33 


4.88 


1.57 


500 


33.00 


248.3 


81.94 


7.69 


4.50 


38.47 


151.6 


58.31 


4.63 


1.40 


600 


35.90 


242.0 


86.86 


7.25 


4.10 


40.31 


148.9 


60.00 


4.50 


1.30 


700 


38.24 


235.7 


90.11 


6.99 


3.86 


41.79 


146.2 


61.07 


4.42 


1.24 


800 


40.28 


229.4 


92.40 


6.82 


3.70 


43.07 


143.5 


61.79 


4.37 


1.19 


900 


42.15 


223.1 


94.04 


6.70 


3.58 


44.25 


140.8 


62.28 


4.34 


1.15 


1000 


43.92 


216.8 


95.21 


6.62 


3.50 


45.35 


138.1 


62.60 


4.31 


1.13 


1100 


45.62 


210.5 


96.02 


6.56 


3.45 


46.41 


135.4 


62.81 


4.30 


1.11 


1200 


47.27 


204.2 


96.52 


6.53 


3.42 


47.43 


132.7 


62.92 


4.29 


1.09 


1300 


48.89 


197.9 


96.74 


6.51 


3.41 


48.44 


130.0 


62.94 


4.29 


1.09 


1400 


50.48 


191.6 


96.71 


6.51 


3.41 


49.43 


127.3 


62.90 


4.29 


1.08 


1500 


52.05 


185.3 


96.45 


6.53 


3.43 


50.40 


124.6 


62.78 


4.30 


1.08 


1600 


53.62 


179.0 


95.96 


6.57 


3.47 


51.37 


121.9 


62.60 


4.31 


1.09 


1700 


55.17 


172.7 


95.26 


6.61 


3.52 


52.33 


119.2 


62.36 


4.33 


1.10 



The thermal contribution can also be 
calculated by the Mie-Gruneisen relation, 

^ h = Qect, e D ) - £(3ook, e D )] 

where E(T,Q D ) is the harmonic internal en- 
ergy calculated from either a Debye model 
or a single Einstein oscillator model 
(Zharkov and Kalinin, 1971), which are 
equivalent in the high-temperature limit. 
The Debye temperature 6 D and Griineisen 
parameter /are considered to be functions 
of volume only: y= -dkiOJdlnV , with the 
volume dependence of y given by q =9lny 
/3lnK The model parameters, D , y, and q, 
can be obtained by fitting the experimental 
P-V-T data to the Mie-Griineisen equation 
of state. 

There is some uncertainty in the Debye 
temperature Q m for perovskite. As a result 
of the non-linear character of the fit, the 
parameters are correlated and there is a 
trade-off among the best-fit values. Figure 
61 illustrates the trade-off between y and q 
for values of B m ranging from 725 to 1025 



K for the silicate perovskite. The circles 
indicate the best fit to the experimental data 
when both y and q are simultaneously 
optimized. Notably, the q of 3 . 3 (at y = 1 .70 
and 6 m = 725 K) for perovskite obtained in 
this analysis is considerably higher than 
many other materials {e.g., q = 0-1 is com- 
monly found in shock- wave studies). How- 
ever, the high q value is consistent with 
high <5 7 -value derived from an independent 
method of analysis (Mao et al. , 1 99 1 ) using 
the thermodynamic relation 



q = 8 T + 1 



K T 



(6) 



when (dlnCydlnVOy, = 0, which is valid at 
high temperature. 



The trade-off between the values of K. 



TO 



and K n ' that fit the static compression data 
has been examined previously (Mao et al., 
1 99 1 ). That work showed that the assump- 
tion that K n * = 4 yields K w = 26 1 GPa. If a 
lower value for the bulk modulus is as- 
sumed, a higher value for K^ is required to 
be consistent with the static compression 



112 



CARNEGIE INSTITUTION 



data. For example, adopting K w = 247 
GPa, as obtained from Brillouin scattering 
measurements at zero pressure by Yeganeh- 
Haeri et al. (1989), requires K n * = 5.5. 
Because of the identity relating K^ and q 
[equation (6)], it is useful to consider the 
effect of a higher assumed value for A^' on 
the thermal properties. A fit to the experi- 
mental data with the assumption of K n ' = 
5.5 yields q = 1. 8 at y =1.70, with D0 = 725 
K (Fig. 62). The difference in densities 
calculated for perovskite at high P and T 
with K w ' varied over this range increases 
with pressure: for example, at 60 GPa K^ 
= 4 gives 0.8% higher density than that 
calculated with K^ = 5.5. 

The parameter values for calculating 
the P-V-T relations of perovskite and 
magnesiowiistite are summarized in Table 
16. Isotherms calculated with the Mie- 
Griineisen relation in the range of the x-ray 
diffraction measurements are shown in Fig- 
ures 62 and 63. For magnesiowiistite, both 
equations (3) and (5) both predict consis- 
tent P-V-T relations up to 140 GPa and 
3000 K(Fei etaL, 1991b). For perovskite, 
however, equation (3) predicts higher vol- 
ume at high P-T than equation (5) (about 
0.5% higher at 60 GPa and 2000 K). These 
differences can result in large uncertainties 
in the lower mantle composition. 




Fig. 62. The trade-off between y and q at various 
given Debye temperature D0 . The circles indi- 
cate the best fit to the experimental data when 
both y and q are simultaneously optimized. The 
solid lines are obtained when K w = 261 GPa and 
AYo = 4 are used. The dashed lines are obtained 
when K-ro - 247 GPa and K T q = 5.5 are used. (See 
text for discussion.) 



A detailed comparison of the densities 
calculated from P-V-T equations of state 
presented above and that determined by 
seismology (e.g., PREM, Dziewonski and 
Anderson, 1981) indicates that a large 
perovskite component relative to 
magnesiowusite with an upper mantle Fe/ 
Mg ratio (Ring wood, 1 975) matches PREM 
to within 1 % throughout the entire lower 



Table 16. Parameters of the thermal equations of state of perovskite and magnesiowiistite 



Parameters 



(Fe,Mg)SiC>3-perovskite 



(Fe,Mg)0-magnesiowustite 



Vo, cm3/mol 


24A6+0MX Fe 


11.25+1.02X Fe 


K m GPa 


261(4) 


160-7.5X Fe 


Kto 


4 


4 


(dK^dT) P , GPa/K 


-6.3(5) x lO- 2 


-2.7(5) x lO" 2 


(dKiftnv, GPa/K 


-2.5(3) x 10- 2 


-0.2(2) x lO" 2 


5r 


6.5(5) 


4.3(5) 


Oqo, K 


725(25) 


500(20) 


7o 


1.70(5) 


1.50(5) 


q 


3.3(5) 


1.1(5) 



GEOPHYSICAL LABORATORY 



113 



mantle (see Hemley et al., 1991). This 
implies a silica enrichment in the lower 
relative to the upper mantle. The conclu- 
sion is dependent on the geotherms and the 
thermoelastic parameters. Figure 63 shows 
the Fe/(Mg+Fe) and Si/Mg ratios for a best 
fit of density between the mineralogical 
model and PREM in the top portion of the 
lower mande (pressure range of 24-60 GPa). 
The composition trade-off between tem- 
perature and density is demonstrated by 
varying the adiabatic temperature at 670 
km from 1 800 K to 2000 K and its gradient 
from 7 K/GPa to 10 K/GPa. Abest fit of X 



Fe 



= 0.1 andX s .= 1.0 is obtained for the 2000- 
K adiabat. The X si value is not as well 
constrained by the density analysis. How- 
ever, the bulk modulus, the slope of density 
profile divided by density, provides a con- 
straint on Si/Mg. The minimum of X Si is 
always close to 1 (i.e., pure perovskite 
composition) for the equation of state of 
perovskite used in this study. The mini- 
mum shifts to more olivine-enriched 
composition if the volume dependence of 
the thermal expansivity of perovskite de- 
creases. 

Recently, we also showed that the cal- 
culated phase relations in the MgO-FeO- 
Si02 system under lower mantle condi- 
tions are sensitive to the volume depen- 
dence of the thermal expansion coefficient 
of component phases, with the value for the 
(Fe,Mg)Si03 perovskite being especially 
significant (Fei and Hemley, 1991). We 
find that <5rof at least 6 (q of 2.9) is required 
to match the observed phase boundary for 
Mg2Si04 (spinel) = MgSi03 (perovskite) 
+ MgO (periclase). A high value for 5j is 
also found to be consistent with the experi- 
mental phase equilibrium data (Ito and 
Takahashi, 1989) and element partitioning 



0.20 



0.15 



0.10 



0.05 - 



decreasing q 



q=2.3 
18O0K 

10K/GPa 9=3.3 
1800K 



(7=3.3 
2000K 
10K/GPa 




g=3.3 
10K/GPa 1800K 

7K/GPa 



0.00 



0.5 



0.6 



0.7 0.8 

Xsi 



0.9 1 .0 



Fig. 63. Misfits between the mineralogical model 
and seismic data for density as functions of Xp e and 
Xsi. The Fe/Mg and Si/Mg ratios for a best fit of 
density are calculated by varying the adiabatic 
temperature at 670 km from 1 800 K to 2000 K and 
its gradient from 7 K/GPa to 10 K/GPa. The effect 
of the volume dependence q for perovskite is also 
indicated by the arrow. 



data (Fei et al., 1991a) in the system 
Mg 2 Si0 4 -Fe2Si0 4 . The stability field of 
perovskite shifts to higher pressure and the 
Mg-rich region with decreasing &r value. 
A lower value of &r would give rise to two 
complete solid solutions between spinel 
and magnesiowustite and between 
perovskite and magnesiowustite, in contra- 
diction to experimental results in this sys- 
tem. Increasing XFe and temperature, and 
decreasing <5r, are shown to decrease sig- 
nificantly the stability field of perovskite. 
The primary conclusions of this study 
concern the top of the lower mantle (from 
670 - 1000 km depth). Comparison be- 
tween the extrapolated equation of state 
data and seismic results at greater depths 
indicates that the compositional arguments 
put forth here do apply to the bottom of the 
lower mantle; that is, the conclusion that a 
lower mantle assemblage dominated by 



114 



CARNEGIE INSTITUTION 



perovskite with an^Fe of -0.1 is consistent 
with the seismic data. The main uncertain- 
ties in this lower mantle composition model 
may rise from both temperature and pres- 
sure extrapolations in the thermal pressure 
models. The two models for perovskite 
used in the analyses result in 0.5% differ- 
ence in desity which accounts for about 2% 
difference in iron content by fitting to the 
PREM densities. Uncertainties in Kj can 
result in more significant changes in the 
composition model, especially to the Si/ 
Mg ratio argument. More accurate models 
of the composition and mineralogy in the 
deeper portions of the mantle would re- 
quire measurements at higher P-T condi- 
tions. 



References 

Anderson, O. L., Equation for thermal expansivity 
in planetary interiors, J. Geophys. Res., 72, 
3661, 1967. 

Anderson, O. L., D. G. Isaak, and S. Yamamoto, 
Anharmonicity and the equation of state for 
gold, /. Appl. Phys., 65, 1534-1543, 1989. 

Dziewonski, A. M., and D. L. Anderson, Prelimi- 
nary reference earth model, Phys . Earth Planet. 
Interiors, 25, 297-356,1981. 

Fei, Y., andR. J. Hemley, Stability of (Fe,Mg)Si0 3 - 
perovskite in the lower mantle, Geophys. Res. 
Lett., in press, 1991. 

Fei, Y., H. K. Mao, and B. O. Mysen, Experimen- 
tal determination of element partitioning and 
calculation of phase relations in the MgO- 
FeO-Si02 system at high pressure and high 
temperature, /. Geophys. Res., 96, 2157- 
2169, 1991a. 

Fei, Y., H. K. Mao, J. Shu, and J. Hu, P-V-T 
equation of state of magnesiowustite 
(Mgo6Feo4)0, Phys. Chem. Miner., submit- 
ted, 1991b. 



Hemley, R. J., Y. Fei, andH. K. Mao, Constraints 
on lowermantle composition from P-V-T mta- 
surements of (Fe,Mg)SiC>3-perovskite and 
(Fe,Mg)0, High Pressure Research in Min- 
eral Physics: Application to Earth and Plan- 
etary Science, edited by Y. Syono and M. H. 
Manghnani eds., submitted, 1991. 

Ito, E., and E. Takahashi, Postspinel transforma- 
tions in the system Mg 2 Si04-Fe 2 Si04 and 
some geophysical implications, /. Geophys. 
Res., 94, 10,637-10,646, 1989. 

Knittle, E., R. Jeanloz, and G. L. Smith, The 
thermal expansion of silicate perovskite and 
stratification of the Earth's mantle, Nature, 
319, 214-216, 1986. 

Mao, H. K., J. Xu, and P. M. Bell, Calibration of 
the ruby pressure gauge to 800 kbar under 
quasihydrostatic conditions,/. Geophys. Res., 
97,4673-4676,1986. 

Mao, H. K., R. J. Hemley, Y. Fei, J. F. Shu, L. C. 
Chen, A. P. Jephcoat, Y. Wu, and W. A. 
Bassett, Effect of pressure, temperature, and 
composition on lattice parameters and density 
of (Fe,Mg)SiC>3-perovskites to 30 GPa, J. 
Geophys. Res., in press, 1991. 

Ring wood, A. E., Composition and Petrology of 
the Earth's Mantle, pp. 618, McGraw-Hill, 
New York, 1975. 

Rosenhauer, M., H. K. Mao, and E. Woermann, 
Compressibility of magnesiowustite 
(Feo 4Mgo 6)0 to 264 kbar, Carnegie Inst. 
Washington Year Book , 75, 513-515, 1976. 

Suzuki, I., Thermal expansion of periclase and 
olivine and their anharmonic properties, J. 
Phys. Earth, 23, 145-159, 1975. 

Wang, Y., D. J. Weidner, R. C. Liebermann, X. 
Liu, J. Ko, M. T. Vaughan, Y. Zhao, A. 
Yeganeh-Haeri, and R. E. G. Pacalo, Phase 
transition and thermal expansion of MgSi03 
perovskite, Science, 251, 410-413, 1991. 

Yeganeh-Haeri, A., D. J. Weidner, and E. Ito, 
Elasticity of MgSiC>3 in the perovskite struc- 
ture, Science, 243, 787-789, 1989. 

Zharkov, V. N., and V. A. Kalinin, Equation of 
State for Solids at High Pressures and 
Temperatures, pp. 257, Consultants Bureau, 
New York, 1971. 



GEOPHYSICAL LABORATORY 



115 



High-Pressure Crystal Chemistry 

of Iron-Free wadsleyite, 

P-MG2S1O4 

Jinmin Zhang, Robert M. Hazen, and 
Jaidong Ko* 

Wadsleyite, P-(Mg,Fe)2Si04, may be 
the most abundant mineral in the upper 
mantle between 400 and 550 km, and the 
phase transformation of olivine to 
wadsleyite may explain the 400-km dis- 
continuity. Since this phase was first found 
to occur at high pressure, much work has 
been done to relate its crystal structure and 
properties to geophysically important prob- 
lems. For example, the 01 and even 02 
positions have been considered to be pos- 
sible sites for protonation; therefore, the 
wadsleyite phase may be a repository of 
water in the mantle (Smith, 1987; Downs, 
1989). 

The wadsleyite structure has three crys- 
tallographically non-equivalent octahedral 
sites, M7, Ml and M3. Wadsleyite in the 
mantle is believed to contain -10% Fe 2 +, 
which partitions in the order M3>M1>M2 
(Finger et al., in preparation). This prefer- 
ence should depend on the volume, the 
distortion and crystal field stabilization 
energy (CFSE) of each site. The character- 
istics of these cation sites in an iron-free 
phase at high pressure are important to 
understanding the distribution of Fe 2+ in 
wadsleyite in the mantle. In this paper, we 
report the result of structure refinements of 
P-Mg2Si04 at five pressures up to 4.84 



* Department of Earth and Space Science, State 
University of New York at Stony Brook 



GPa, with particular attention to the prop- 
erties of the octahedral sites. This study, 
together with work in progress on Fe-con- 
taining wadsleyites, will help to define the 
properties of this phase in the mantle. 

The single crystals used in this study 
were synthesized by Jaidong Ko at the 
High-Pressure Laboratory of the State Uni- 
versity of New York at Stony Brook. The 
sample was produced at a pressure of about 
16 GPa and a temperature of 1400°C. 

A subequant crystal 0.06 x 0.12 x 0.13 
mm in size was selected, and wasexamined 
optically and by x-ray diffraction at room 
conditions. Three intensity data sets were 
collected and several unit-cell parameter 
measurements were made on this crystal 
before it was crushed when trying to in- 
crease pressure after the run at 2.88 GPa. 
Another crystal 0.04 x 0.06 x 0.06 mm in 
size was used for higher pressure measure- 
ments. 

The selected crystal and a small piece of 
fluorite crystal were mounted in a dia- 
mond-anvil cell designed for single -crystal 
x-ray diffraction studies (Hazen and Fin- 
ger, 1982). The fluorite crystal was used as 
an internal standard of pressure, following 
the method of Hazen and Finger ( 1 98 1 ). An 
Inconel 750X gasket with 0.40 mm diam- 
eter hole was centered over one 0.60-mm 
diamond anvil; the crystal was then affixed 
to the anvil face inside the hole with a small 
dot of the alcohol insoluble fraction of 
vaseline petroleum jelly. A mixture of 4:1 
methanol rethanol was used as the hydro- 
static pressure medium. 

All x-ray measurements except for the 
and 1.16 GPa intensity data collections 
were performed with a Picker automated 



116 



CARNEGIE INSTITUTION 



Table 17. Crystallographic data for iron-free wadsleyite 









Pressure, GPa 




Paramter 

















1.16 


1.81 


2.88 


4.84 


crystal size, mm 


0.06x0.12x0.13 


. 


_ 


_ 


0.04x0.06x0.06 


IH> cm -1 


11.16 


11.25 


11.29 


11.35 


11.48 


Range of TO) 


0.92-0.94 


0.92-0.94 0.93-0.94 


0.93-0.94 


0.95-0.96 


RinP 


0.060 


0.069 


0.057 


0.057 


0.078 


Number of data 


307 


294 


303 


286 


239 


R J 3 ) all data 


0.056 


0.055 


0.056 


0.059 


0.060 


/?( 4 ) all data 


0.089 


0.094 


0.082 


0.072 


0.114 


Number F >2gf 


236 


216 


244 


242 


157 


Rw 


0.055 


0.054 


0.055 


0.057 


0.058 


R 


0.068 


0.069 


0.061 


0.057 


0.074 



T is transmission factor. R{ nt is residual for internal agreement of symmetry 
equivalent reflections. R w = (Luj(F - F c ) 2 /ZwF 2 f- 5 R = IXiF \ - IF C II/ZIIF II 



four-circle diffractometer with filtered Mo- 
Koc radiation (A = 0.7107 A). The intensity 
data at and 1.16 GPa were collected with 
a Huber four-circle diffractometer with 
graphite monochromated Mo-Koc radiation 
(X = 0.7093 A), but the cell parameters 
measured with the Picker diffractometer 
were used for the structure refinements for 
the purpose of consistency. 

Unit-cell parameters were measured 
with the procedure of King and Finger 
(1979), whereby several reflections are 
measured in eight equivalent orientations. 
The range of 20 for all reflections was 1 8- 
31° in order to avoid systematic errors that 
result from comparing angular data from 
different ranges (Swanson et al., 1985). 

Intensities were measured for all acces- 
sible reflections in a hemisphere of recipro- 
cal space with sin0/A < 0.7. Corrections 
were made for Lorentz and polarization 
effects, crystal absorption by the diamond 
and beryllium components of the pressure 



cell. Digitized step data were integrated by 
the method of Lehmann and Larsen ( 1 974). 
Refinement conditions, refined isotropic 
extinction coefficients, refined structural 
parameters, and isotropic temperature fac- 
tors are given in Tables 17 and 18. 

As observed by Hazen et al. ( 1 99 1 ), the 
c axis is the most compressible, with p c = 
2.32(4) x 1 0-3 GPa- l . The a and b axes have 
compressibilities of 1.72(14) x 10 -3 and 
1.71(3) x 10" 3 GPa- 1 , respectively, — al- 
most completely identical with the results 
of Hazen et al. (1991), who reported the 
value of 1.73(3)xl0 -3 for the a and b axes 
and 2.39(3) x 10" 3 for the c axis. The 
stiffness of a and b relative to c, according 
to Hazen et al. (1991), results from the 
pseudo-layering of the structure, with Mg- 
octahedral layers parallel to (001 ), and cross 
linking by Si207 tetrahedral pairs. 

The pressure-volume data were fitted 
with program VOLFIT to a Birch- 
Mumaghan equation of state with 4 as the 



GEOPHYSICAL LABORATORY 



117 



Table 18. Positional and equivalent isotropic thermal parameters. 









Pressure, GPa 




Parameter 

















1.16 


1.81 


2.88 


4.84 


Extn. 












Coef(10- 4 ) 


0.37(5) 


0.38(5) 


0.26(5) 


0.31(6) 


0.26(6) 


Ml, x,y,z 

















B 


0.62(10) 


0.66(11) 


0.55(8) 


0.45(7) 


0.54(18) 


M2,* 

















y 


1/4 


1/4 


1/4 


1/4 


1/4 


2 


0.9696(5) 


0.9694(6) 


0.9704(4) 


0.9697(4) 


0.9687(9) 


B 


0.52(8) 


0.52(8) 


0.43(6) 


0.49(6) 


0.47(13) 


M3,* 


1/4 


1/4 


1/4 


1/4 


1/4 


y 


0.1265(3) 


0.1264(3) 


0.1267(2) 


0.1267(2) 


0.1263(5) 


z 


1/4 


1/4 


1/4 


1/4 


1/4 


B 


0.91(6) 


0.91(7) 


0.70(5) 


0.70(5) 


0.58(10) 


Si, x 

















y 


0.1203(2) 


0.1203(2) 


0.1200(2) 


0.1199(1) 


0.1202(3) 


z 


0.6166(2) 


0.6165(3) 


0.6167(2) 


0.6168(2) 


0.6172(4) 


B 


0.55(5) 


0.54(5) 


0.41(4) 


0.37(4) 


0.36(7) 


Ol.z 

















y 


1/4 


1/4 


1/4 


1/4 


1/4 


z 


0.2173(10) 0.2181(10) 


0.2187(8) 


0.2204(7) 


0.2210(14) 


B 


0.9(2) 


0.8(2) 


0.5(1) 


0.6(1) 


0.2(2) 


02,x 

















y 


1/4 


1/4 


1/4 


1/4 


1/4 


z 


0.7158(9) 


0.7167(9) 


0.7159(8) 


0.7158(7) 


0.7199(14) 


B 


0.5(1) 


0.5(2) 


0.5(1) 


0.5(1) 


0.3(2) 


03,* 

















y 


0.9889(5) 


0.9895(5) 


0.9900(4) 


0.9899(4) 


0.9900(8) 


z 


0.2564(8) 


0.2557(8) 


0.2554(6) 


0.2553(5) 


0.2538(15) 


B 


0.9(1) 


0.6(1) 


0.54(9) 


0.40(8) 


0.8(2) 


04,* 


0.2605(10) 0.2597(12) 


0.2629(10) 


0.2608(9) 


0.258(2) 


y 


0.1226(3) 


0.1225(4) 


0.1218(4) 


0.1227(2) 


0.1211(7) 


z 


0.9923(4) 


0.9933(5) 


0.9933(4) 


0.9925(3) 


0.9936(7) 


B 


0.66(7) 


0.71(8) 


0.63(7) 


0.58(6) 


0.8(1) 



assumed pressure derivative (K 9 ). The bulk 
modulus is 162(6) GPa if Vo is not con- 
strained, comparable to the value 160(3) 
GPa reported by Hazen et al. (1991). 

The structure refinement of Fe-free 
wadsleyite by Finger et al. (in preparation) 
was used as the initial model in the present 
work except that isotropic temperature fac- 
tors were used in place of anisotropic ones 
in their study. The space group, Imma, was 
confirmed by the structure refinement. The 
scattering factor curves for Mg, Si, and O 



are those of neutral atoms in International 
Tables for X-ray Crystallography (1974). 

The RFINE6 program (Finger and 
Prince, 1975) was used for the structure 
refinements. The calculation converged 
quickly to the R values listed in Table 17. 
The extinction coefficients, atomic coordi- 
nates, and isotropic temperature factors are 
listed in Table 18. 

Polyhedral volumes were calculated 
using the program VOLCAL, written by 
Finger (in Hazen and Finger, 1982); qua- 



118 



CARNEGIE INSTITUTION 



Table 19. Polyhedral parameters for iron-free wadsleyite 




Pressure, GPa 
Parameter 


1.16 1.81 2.88 


4.84 



Ml, Volume 

QE* 

AV 

VS 

M2, Volume 

QE 

AV 

VS 

M3, Volume 

QE 

AV 

VS 

Si, Volume 

QE 

AV 

VS 

01, VS 

02, VS 

03, VS 

04, VS 



11.648(38) 
1.0053(10) 
15.83 
2.17 

11.880(38) 
1.0056(46) 
19.58 
2.09 

12.026(21) 
1.0070(18) 
23.11 
2.06 

2.276(10) 
1.0037(40) 
14.63 
3.84 

1.98 
2.01 
1.95 
2.04 



11.519(41) 
1.0045(11) 
13.62 
2.21 

11.779(42) 
1.0061(51) 
21.76 
2.11 

11.887(22) 
1.0065(20) 
21.74 
2.10 

2.294(11) 
1.0034(43) 
13.68 
3.79 

1.99 
2.02 
1.96 
2.04 



11.514(35) 
1.0049(10) 
15.52 
2.21 

11.959(35) 
1.0057(39) 
19.63 
2.07 

11.807(18) 
1.0061(15) 
20.10 
2.12 

2.254(9) 
1.0033(35) 
12.58 
3.90 

2.02 
2.03 
1.98 
2.08 



11.439(28) 
1.0046(8) 
14.43 
2.24 

11.768(30) 
1.0056(34) 
19.79 
2.12 

11.785(16) 
1.0058(13) 
18.79 
2.13 

2.254(8) 
1.0040(30) 
15.90 
3.89 

2.02 
2.05 
1.99 
2.09 



10.966(73) 
1.0044(21) 
13.35 
2.39 

11.554(70) 
1.0065(79) 
23.01 
2.18 

11.591(33) 
1.0057(31) 
18.72 
2.19 

2.298(20) 
1.0026(71) 
10.50 
3.79 

2.04 
2.07 
2.02 
2.09 



*QE and AV stand for the quadratic elongation and angle variance parameters, respec- 
tively, of Robinson et al. (1969). VS stands for the valence sum of Brown (1981). 



dratic elongations and angular variances 
were calculated according to the defini- 
tions of Robinson et al. (1971); the bond 
valence sums are after Brown (1981). The 
results are listed in Table 19. 

The Si04 tetrahedral volume does not 
decrease significantly with pressure, typi- 
cal of all high-pressure studies below 10 
GPa. Of the three Mg-0 octahedra, M3 is 
the largest and most distorted at room pres- 
sure (see Table 19); as pressure increases, it 
becomes smaller and perhaps more regular. 
Changes of the Ml and M2 octahedra are 
less well defined (Fig. 64). The Ml and M2 
octahedra become smaller with increasing 



pressure, but there is little change in the 
distortion. 

Least-square fit of volume-pressure data 
of the three polyhedra gives average 
compressibilities of 8.7(11), 4.3(11), and 
6.9(6) x 10' 3 GPa" 1 , respectively, forM7, 
M2, and M3 sites, corresponding to bulk 
moduli of 116(15), 234(59), and 145(12) 
GPa. 

Zhang et al. (1991) demonstrated that 
the unit-cell volume of an isostructural 
series is approximately linearly related to 
the bond length of a specific cation-anion 
bond as long as other bond lengths do not 
change. The compression of wadsleyite fits 



GEOPHYSICAL LABORATORY 



119 



11.80 



V= 11.677(30) -0.1 01 (13)P 

K = 116(15) GPa 
* 




i i i i — i i i — i i i i i i i i i i i i — i — i — i i i 



V=1 1.92(3) -0.051(13)P 
I K = 234(59) GPa 




11.60 

11.50 

£< 12.00 



i i i — i i i i i i i 



i i l i — i i i i — i— i- 



V= 11.998(16) -0.083(7)P 
K = 145(12) GPa 




2.0 3.0 4.0 

Pressure, GPa 



5.0 



Fig. 64. Variations of the volumes of the three M- 
O octahedra with pressure. 

into this specification because the Si-0 
bond length is virtually constant. Figure 
65 shows the variation of the unit-cell vol- 
ume with the average of the three M-0 
bond lengths; the linear relationship is a 
clear indication that the contraction of the 
unit cell is largely due to the shortening of 
the M-O bonds. This linear relationship is 
also true when Fe 2+ substitutes for Mg 2 + in 
the wadsleyite structure and thus increases 
the average M-0 bond length. In Fig. 66, 
using the data of Finger et al. (in prepara- 
tion), the unit-cell volume of wadsleyite 
with different Fe 2+ contents is plotted vs. 
the average M-0 bond length. Again it 




520 



2.045 2.055 2.065 2.075 2.085 

Average M-0 bond length, A 

Fig 65. Linear relationship between the unit-cell 
volume and the average of the three M-O bond 
lengths. 



co 
© 

E 

O 

> 

© 

o 

i 

■*± 

"c 

ID 











■ 


546 


- 




Fe25 


• S 


544 








• 


542 




Fe08v/ 


' Fe16 


■ 


540 




• yr 




• 


rqci 


x^» 


Te00 







2.078 2.081 2.084 2.087 2.090 2.093 

Average M-0 bond length, A 

Fig 66. Unit-cell volume as a function of the 
average M-0 bond length. The increases of the 
average M-O bond length and the unit-cell volume 
are due to the substitution of Fe 2+ for Mg 2+ . 



suggests that the increase or decrease of the 
unit-cell volume is due to the increase or 
decrease of the M-0 bond lengths. 
We reach several conclusions 
(1). Measurements of unit-cell param- 
eters at 10 pressures give the axial 
compressibilities of wadsleyite, which are 
1.72(14), 1.71(3) x 10-3, and 2.32(4) x 10- 
3 GPa -1 , for a, b y and c, respectively. As- 
suming a Birch- Murnaghan equation of 
state with K' = 4, the bulk modulus is 
162(6) GPa. 



120 



CARNEGIE INSTITUTION 



(2). The atomic coordinates, bond 
lengths, bond angles, and polyhedral vol- 
umes of the cation sites are documented. 
The bulk moduli of the three octahedra are 
116(15),234(59),andl45(12)GPaforM7, 
M2, and Mi, respectively. M3 is the octa- 
hedron for which the volume vs. pressure 
data points are the least scattered and is also 
the one which seems to become more regu- 
lar with increasing pressure. The other two 
octahedra, especially M2 , have bulk moduli 
with large errors, and no significant varia- 
tion in distortion is observed. 

(3). The unit-cell volume is linearly 
related to the average M-0 bond length, an 
indication that the decrease of the unit-cell 
volume is due to the shortening of the M-0 
bonds. 



References 

Brown, I. D., The bond-valence method: an em- 
pirical approach to chemical structure and 
bonding, in Structure and Bonding in Crys- 
tals, O'Keefe and Navrotsky, eds., 2, 1-30, 
Academic Press, Boston, 1981. 

Downs, J. W., Possible sites for protonation in (i- 
Mg2Si04 from an experimentally derived elec- 
trostatic potential, Amer. Mineral, 74, 1 124- 
1129,1989. 

Finger, L. W., andE. Prince, A system of Fortran 
IV computer programs for crystal structure 
computations, US National Bureau of Stan- 
dard Technical Note 854, Washington DC, 
1975. 

Hazen, R. M., and L. W. Finger, Calcium fluoride 
as an internal pressure standard in high pres- 
sure/high-temperature crystallography,/. Ap- 
plied Cry stallogr., 14, 234-236, 1981. 

Hazen, R. M., and Finger, L. W., Comparative 
Crystal Chemistry, Wiley, New York, 1982. 

Hazen, R. M., J. Zhang, and J. Ko, Effects of Fe/ 
Mg on the compressibility of synthetic 
wadsleyite: 0-(Mg \. x ^x)2^0a (jc<0.25), 
Phys. Chem. Minerals, 77, 416-419, 1991. 

International Tables of X-ray Crystallography, 
Vol. IV, Kynock Press, Birmingham, 1974. 



King, H. E., and L. W. Finger, Diffracted beam 
crystal centering and its application to high- 
pressure crystallography, /. Applied 
Crystallogr., 12, 374-378, 1979. 

Lehmann, M. S., and M. K. Larsen, A method for 
location of the peaks in step-scan-measured 
Bragg reflections, Acta Cry stallogr., A30, 580- 
584, 1974. 

Robinson, K., G. V. Gibbs, and P. H. Ribbe, 
Quadratic elongation: A quantitative measure 
of distortion in coordination polyhedra, Sci- 
ence, 772,567-570, 1971. 

Smith, J. R., p-Mg2SiC>4: a potential host for 
water in the mantle? Amer. Mineral., 72, 1051- 
1055, 1987. 

Swanson, D. K., D. J. Weidner, C. T. Prewitt, and 
J. J. Kandelin, Single-crystal compression of 
Y-Mg 2 Si0 4 , EOS, 66, 370, 1985. 

Zhang, J. M, D. N. Ye, and C. T. Prewitt, Rela- 
tionship between the unit-cell volumes and 
cation radii of isostructural compounds and 
the additivity of the molecular volumes of 
carbonates, Amer. Mineral., 76, 100-105, 1991. 



Phase Transitions in 
Framework Minerals 

David Palmer 

It is now clear that certain phase transi- 
tions previously dismissed as being "subtle" 
phenomena, may in fact produce dramatic 
anomalies in the physical properties of 
minerals and significantly modify their ther- 
modynamic behavior. These phase transi- 
tions, which typically involve displacive 
distortions of the crystal lattice or cation 
ordering effects, are common in most rock- 
forming minerals that exist in the Earth's 
crust. They are also expected to occur in 
mantle phases. 

Unlike heterogeneous reactions be- 
tween crystallographically unrelated phases 
(which are more usually studied by earth 
scientists), the influence of a displacive or 
order/disorder phase transition is not con- 



GEOPHYSICAL LABORATORY 



121 



fined to the phase boundaries of an equilib- 
rium system, but is significant at pressures 
and temperatures far below the transition 
point itself. As a consequence, stability 
relations between mineral assemblages may 
be perturbed throughout P-T space. 



Quantitative Analysis of 
Mineral Behavior 

Most of the breakthroughs in the study 
of phase transitions have come from solid- 
state physics, the original motivation hav- 
ing been to relate anomalies in physical and 
thermodynamic properties to changes at a 
crystal structural level. From a substantial 
body of work on simple crystal structures, 
came a number of "mean field" theories, all 
relating to the macroscopic behavior of 
crystals. These concern those phase transi- 
tions which involve the lowering of crystal 
symmetry, such that a relationship between 
the symmetries of "high" and "low" phases 
is maintained (i.e., a supergroup-subgroup 
relation). Most displacive and order/disor- 
der phase transitions fit into this category. 

Macroscopic theories based on ideas 
initially propounded by Landau (Landau 
and Lifshitz, 1980) have been used exten- 
sively in rationalizing the temperature de- 
pendence of crystal behavior. The funda- 
mental starting point for these theories is 
the concept of a macroscopic order param- 
eter, 2, which measures the progress of the 
phase transition, such that Q = in the high- 
symmetry phase, and 0<Q<1 in the low- 
symmetry phase. This might relate, for 
example, to the ordering of magnetic spins, 
positions of certain sets of cations such as 



Al and Si, or to a prevailing lattice distor- 
tion. Landau's original postulate was that 
the energy lowering due to the high - low 
transition, the excess free energy of the 
low-symmetry phase, could be represented 
as a power series in Q. From this potential, 
it is possible to derive the temperature 
dependence of the order parameter, in terms 
of a critical exponent, p. It also becomes 
possible to relate the excess thermody- 
namic properties — heat capacity, entropy, 
enthalpy etc. — to the order parameter and 
hence to the progress of the phase transi- 
tion. Further developments of Landau 
Theory allow the behavior of excess physi- 
cal properties to be related to the order 
parameter and the thermodynamic proper- 
ties. 

The Landau approach, by relating all 
excess physical, thermodynamic and struc- 
tural properties to the macroscopic order 
parameter considerably simplifies the de- 
scription of phase transition behavior and 
provides for a detailed understanding of 
more complex, mineralogical systems. 
Many minerals undergo more than one 
phase transition, which may lead to seem- 
ingly very complicated behavior: phase 
transitions in the same material are rarely 
independent of each other. In Landau 
Theory, these effects are predicted in terms 
of coupling between the various order 
parameters, according to strict symmetry 
rules. 

The aim of this particular study is to 
increase our understanding of phase-tran- 
sition-related mineral behavior, focusing 
on two less well studied areas, (1) high- 
temperature thermodynamic properties, and 
(2) high-pressure structural behavior. 



122 



CARNEGIE INSTITUTION 



Framework minerals provide the "model 
systems" for this work: these are the most 
abundant materials at the Earth's surface; 
their interconnected topologies ensure long 
correlation lengths throughout the crystal 
structure, and account for the many 
displacive and cation-ordering phase tran- 
sitions which occur with increasing tem- 
perature and pressure. In this context there- 
fore, "mean field" theories such as Landau 
Theory are particularly relevant. 



Order I Disorder Relations in Anorthite 

Feldspars dominate the mineralogy of 
the Earth's surface. They also show some 
of the most complex subsolidus behavior 
of all minerals, which may be attributed to 
the interplay between the effects of succes- 
sive displacive and cation-ordering phase 
transitions. Much of this behavior has now 
been quantified and rationalized within the 
scope of Landau Theory (Carpenter, 1988). 

The ordering of Al and Si at low tem- 
peratures is common in many minerals, but 
it has always been assumed that this pro- 
cess is largely configurational, and does 
not change the heat capacity. However, 
many studies of feldspars have shown that 
Al/Si ordering induces a lattice strain in the 
crystal, thereby coupling to any prevalent 
displacive distortion. It would be logical to 
suppose that such ordering would 
renormalize phonon frequencies, thereby 
altering the heat capacity. If this does turn 
out to be the case, then current attitudes to 
mineral energetics will have to be reevalu- 
ated. 



One is not predicting substantial AC P 
effects, and so testing this hypothesis re- 
quires sensitive experimental procedures. 
It would be desirable to compare the ener- 
getics of two (or more) samples from the 
same specimen, which differ only in the 
extent of their Al/Si ordering (measured by 
the order parameter Qod)- Changes in the 
Al/Si ordering require solid-state diffusion, 
which is an extremely sluggish process 
below 1500 K or so. In order to be able to 
prepare samples with different Qod in the 
laboratory, one needs a material with a very 
high equilibrium order/disorder phase tran- 
sition temperature (T c ). Samples can then 
be studied at lower temperatures without 
the risk of continued ordering. 

Anorthite, CaAl2Si20s, an end-mem- 
ber plagioclase feldspar, is a good candi- 
date for this study. Not only has the mineral 
been extensively studied, its transition be- 
havior is extremely well characterized (Car- 
penter 1988, 1991). On heating, there is a 
displacive phase transition from P\ to 7T at 
510 K. Continued heating induces pro- 
gressive disorder of Al and Si over the 
tetrahedral sites, and the symmetry ap- 
proaches CT, although anorthite melts (at 
-1800 K) before reaching the hypothetical 
/T - C\ phase transition temperature. The 
most ordered samples available have Qod 
~ 0.92, and this can be reduced, as required, 
by annealing at high temperatures. 

For this study on the effect of ordering 
on the molar heat capacity, a natural anor- 
thite was used (Qod = 0.92, as measured by 
x-ray diffraction). A portion of the material 
was annealed at 1723 K for 21 days to 
induce some Al/Si disorder, thereby reduc- 
ing Qod to 0.82. These samples were 



GEOPHYSICAL LABORATORY 



123 



provided by Michael Carpenter (Univer- 
sity of Cambridge). Differential scanning 
calorimetry (DSC) runs performed in Cam- 
bridge revealed no differences in the heat 
capacities of these samples at low tempera- 
tures (T< 900 K). However, this technique 
cannot be used at higher temperatures and 
so for a complete investigation it was de- 
cided to use transposed-temperature drop 
calorimetry at Princeton University, in col- 
laboration with Alexandra Navrotsky. This 
technique permits determination of heat 
capacity, albeit indirectly, by measuring 
the relative enthalpies at different tempera- 
tures. Precise determination of C p necessi- 
tates many measurements at closely spaced 
temperatures, but as a preliminary test to 
see whether Al/Si ordering does induce a 
ACp effect, it is sufficient to measure en- 
thalpies at a few temperatures, to see if AH 
between the samples varies as a function of 
temperature. 

Samples of anorthite were sealed in Pt 
foil and dropped into a receptacle within a 
"Setaram" calorimeter, set to the desired 
temperature. The enthalpy change of the 
sample from room temperature to the calo- 
rimeter temperature is then proportional to 
the energy required to restore the tempera- 
ture of the assembly. An alumina-filled 
capsule was used as calibration standard. 
Enthalpy measurements were repeated three 
to seven times per sample at each tempera- 
ture, to check reproducibility. The results 
are displayed in Table 20. 

It must be stressed that these are rela- 
tive enthalpies, that is, //r-//294K- The ac- 
tual enthalpy of ordering between the 
samples is ~ 8 kJmoH (Carpenter, pers. 
comm.). The increase in AH between the 



samples with increasing temperature is sig- 
nificant, and may well indicate a AC p ef- 
fect. Only at the highest temperature is 
there the possibility of some slight disor- 
dering within the calorimeter, though this 
seems unlikely because of the short mea- 
surement time (15 minutes, typically). 
Because the samples used were natural, 
albeit very pure, one cannot entirely dis- 
count other effects due to trace impurities. 
A second set of experiments, using syn- 
thetic anorthites is now underway in order 
to clarify this point. 

These measurements have shown that 
it is possible to measure enthalpy differ- 
ences between minerals on the order of a 
few joules for a 30-mg sample (3 kJmol -1 
for anorthite) at temperatures far beyond 
the reach of conventional calorimetric tech- 
niques. Although this method is not as 
precise as low-temperature DSC, further 
enhancements, such as more sensitive de- 
tector systems, are being developed to al- 
low increased resolution at higher tem- 
peratures. 



Table 20. Relative enthalpy of anorthite from 
973 K to 1673 K.* 



T[K] 


HtH 2 94K 

Qod=0M 


[kJmol 1 ] 

Qoo=om 


973 

1273 
1473 
1673 


194(4) 
288(3) 
349(3) 
429(6) 


195(3) 
290(4) 
369(3) 
444(8) 



* Errors are twice the standard deviation of 
the mean. 



124 



CARNEGIE INSTITUTION 



High-Pressure Studies of Phase 
Transitions in Minerals 



A New, High Pressure Phase Transition 
in Leucite 



It is possible to extend Landau Theory 
to the description of high-pressure phase 
transition behavior, using suitably chosen 
"model systems". The aim of this research 
is to concentrate on rock-forming minerals 
that exist within the Earth's crust, where 
moderate temperatures and pressures pre- 
vail. The study of the pressure dependence 
of such phases provides a logical step from 
previous, high- temperature studies of phase 
transition behavior. 

Feldspathoid minerals were selected for 
study. They are relatively abundant within 
the crust and, like the feldspars, have alu- 
minosilicate frameworks with cations in 
interstitial sites. Phase transitions involv- 
ing both displacive distortions and cation 
ordering are extremely common. 
Feldspathoids may be distinguished from 
feldspars by the presence of structural chan- 
nels, which provide an important reposito- 
ries for cations, water, organic molecules 
etc., and are pathways for ionic conduction 
(Palmer and Salje, 1990; Alpena ai, 1977). 
In this sense, feldspathoids may be com- 
pared to zeolite minerals. The presence of 
structural channels makes feldspathoids 
low-density minerals; thus feldspathoids 
might be expected to be very unstable at 
high pressures. However, high-tempera- 
ture studies of feldspathoids reveal a re- 
markable structural adaptability, and this 
may prolong the stability (or metastability) 
of such structures at much higher pressures 
than previously imagined. 



Leucite, KAIS12O6, is a feldspathoid 
mineral associated with SiC>2-poor, K-rich 
alkaline volcanics, which has been well- 
studied as a function of temperature. On 
cooling, leucite undergoes two phase tran- 
sitions, from a cubic phase Ia3d to an 
intermediate tetragonal phase I4\lacd at Tc 
= 938 K, described by order parameter Qi 
(E g symmetry); then to a low-7 tetragonal 
phase 14 1 la at 7c = 918 K, described by 
order parameter Qu with Tig symmetry 
(Palmer er al, 1989). There is a continuous 
volume decrease in the low-rphase, asso- 
ciated with collapse of the <1 1 1> structural 
channels; this is described by Qu . The first 
order parameter, Qj y describes a ferroelastic 
distortion of the unit cell (no volume 
change). Such a distortion — an acoustic 
shear mode — should show little or no 
pressure dependence, compared to the 
highly pressure dependent volume distor- 
tion. Increasing pressure, therefore, is ex- 
pected to modify the transition behavior by 
enhancing Qu relative to Q/. 

The pressure dependence of leucite has 
been followed up to 60 kbar, using Raman 
spectroscopy. A single crystal was placed 
within a Mao-Bell diamond-anvil cell, us- 
ing an organic liquid, "FC75" (C8F16O) as 
pressure medium. Pressure determination 
was by the ruby fluorescence method, with 
ruby spectra measured using profile refine- 
ment. 

Measurement of the Raman spectra was 
possible using a 0.6W He-Ne laser, with a 



GEOPHYSICAL LABORATORY 



125 



580 



570 



I 560 

| 550 

c 
® 

| 540 



530 



520 




■ increasing pressure 
□ decreasing pressure 



10 20 30 40 
Pressure, kbar 



50 



60 



Fig. 67. Pressure dependence of the Eg Raman 
mode forleucite. The large increase in frequency 
at P c , and a hysteresis on increasing and decreas- 
ing pressure, indicate the existence of a first-order 
phase transition. 



"Triplemate" detector and Princeton In- 
struments control system. The leucite 
Raman spectrum contains two intense 
peaks, and a number of much weaker modes. 
High-temperature Raman spectroscopy 
(Palmer et aL, 1990) showed that the in- 
tense modes relate to the two order param- 
eters, and may be assigned to T\g and Eg 
symmetries. 

For this study, the pressure dependence 
of the E g mode was followed, revealing the 
presence of a previously unknown phase 
transition at P = 23 kb on increasing pres- 
sure (Fig. 67). The similarity of the Raman 
spectra on either side of the phase transi- 
tion suggests that both "high" and "low" 
phases are related. The fact that the phase 
transition is totally reversible, together with 
its rapid kinetics, suggest that the mecha- 
nism is displacive. The existence of a 
hysteresis implies that the phase transition 
is first order in character. The existence of 



a supergroup/subgroup relation to this phase 
transition limits the choice of possible high- 
pressure phases. We hope to carry out x-ray 
work to refine the structure of the high 
pressure phase, and to fully characterize 
the phase transition behavior. 



References 

Alpen, U.U., H. Schulz, G. H. Tatat, and H. 
Boehm, One-dimensional cooperarive Li-dif- 
fusion in p-eucryptite. Solid State Commun., 
25,911-914, 1977. 

Carpenter, M. A., Thermochemistry of alumi- 
nium/silicon ordering in feldspar minerals, in 
Physical Properties and Thermodynamic 
Behaviour of Minerals, E.K.H. Salje, ed., 
D.Reidel, Dordrecht, pp. 265-323, 1988. 

Carpenter, M. A., Thermodynamics of phase tran- 
sitions in minerals: A macroscopic approach., 
in Stability of Minerals, G.D. Price, ed., Allen 
and Unwin, Boston (in press), 1991. 

Landau, L.D., and E. M. Lifshitz, Statistical 
Physics, Edition 3, part 1, Pergamon Press, 
Oxford, 1980. 

Palmer, D.C., and E. K. H. Salje, Phase transitions 
in leucite: dielectric properties and transition 
mechanism. Phys. Chem. Minerals, 17, 444- 
452, 1990. 

Palmer, D.C., E. K. H. Salje, and W. W. Schmahl, 
Phase Transitions in leucite: X-ray diffraction 
studies. Phys. Chem. Minerals, 16, 714-719, 
1989. 

Palmer, D.C., U. Bismayer, and E. K. H. Salje, 
Phase transitions in leucite: Order parameter 
behaviour and the Landau Potential deduced 
from Raman spectroscopy and birefringence 
studies. Phys. Chem. Minerals, 17, 259-265, 
1990. 

Salje, E., B. Kuscholke, B. Wruck, and H. Kroll, 
Thermodynamics of sodium feldspar II: ex- 
perimental results and numerical calculations. 
Phys. Chem. Minerals, 12, 99-1071, 1985. 



126 



CARNEGIE INSTITUTION 



First-principles Studies of Elasticity 

and Post-Stishovite Phase Transitions 

in S1O2* 

Ronald E. Cohen 

Stishovite is a candidate mineral for the 
Earth's transition zone and lower mantle, 
and is also the prototypical octahedrally 
coordinated silicate. It is an open ques- 
tion, however, whether stishovite remains 
the stable form of Si02 throughout the 
lowermantle, or if a new structure for Si02 
forms at the high pressure conditions of the 
lower mantle. This is an important ques- 
tion, because the presence or absence of 
stishovite depends on the chemical compo- 
sition of the lower mantle, particularly the 
Fe-Mg ratio (e.g. Fei and Hemley, 1991). 

Two recent studies suggestpossible phase 
transitions under mantle conditions. Park 
et al. (1988) predicted a phase transition 
from stishovite (rutile structure) to the 
pyrite structure (Pa3) at 60 GPa using self- 
consistent Linearized Augmented Plane 
Wave (LAPW) calculations. Tsuchidaand 
Yagi(1989) looked for this transition using 
in situ x-ray diffraction in the diamond 
anvil cell, and instead reported a phase 
transition in Si02 from stishovite to the 
CaCl2 structure between 80-100 GPa. A 
transition from stishovite to the CaCb struc- 
ture in the lower mantle would be very 
important in geophysical modeling of the 
Earth. This phase transition involves an 
elastic instability where c\\-c\2 becomes 



*The computations were performed on the Cray 2 
at the National Center for Supercomputing Appli- 
cations under the auspices of the National Science 
Foundation. 



unstable (Cohen, 1987; Hemley, 1987). 
Such a transition should be evident in seis- 
mological data (derived acoustic veloci- 
ties) if stishovite is present in any quantity 
in the deep Earth. No such features are 
observed in the lowermantle except for the 
anomalous D" zone at the base of the 
mantle. One must conclude either that 
little stishovite is present in the deep man- 
tle, or that the phase transition in Si02 
occurs in D" and seismic anomalies in D" 
reflect at least partially that transition. 

These questions are addressed here us- 
ing the Linearized Augmented Plane Wave 
method (Wei and Krakauer, 1985). The 
present calculations represent one of the 
most extensive studies of a single material 
using this technique — over 200 self-con- 
sistent calculations were performed to de- 
termine the phase relations, elasticity, and 
vibrational properties of Si02 in the 
stishovite, CaCl2, and Pa3 structures 
(Cohen, 1 99 1 a,b). These calculations make 
no assumptions about ionicity, bonding, or 
form of the electron distribution. The only 
inputs are the nuclear charges and posi- 
tions, and the output is a total energy for 
that nuclear configuration. The quantum 
mechanical problem is solved within the 
local density approximation (LDA) (Hedin 
and Lundqvist, 1 97 1 ) of density functional 
theory (Hohenberg and Kohn, 1964). The 
electrostatic and kinetic energy contribu- 
tions to the energy and potential are evalu- 
ated numerically, and can be converged to 
any necessary accuracy. In the LDA, the 
quantum-chemical contributions are mod- 
eled by assuming that the exchange and 
correlation contributions to the potential 
and energy can be obtained from the ex- 



GEOPHYSICAL LABORATORY 



127 



> 

ID 




O O * 



stishovite 



I i ■ i 'i 



30 



40 



50 



v(A J ) 



Fig. 68. Equation-of- state (energy versus volume 
at K) of stishovite, CaCl2, and Pa3. The transi- 
tion from stishovite to Pa3 is at about 160 GPa. 
CaCh becomes stable relative to stishovite at 45 
GPa. The initial energy difference between CaCl2 
and stishovite is very small relative to the energy 
changes with volume. 




50 75 

P (GPa) 



Fig. 69. The elastic constant c\ \-c\2 as a function 
of pressure. The solid line is a spline fit to the 
calculated points (circles). The plus is the Brillouin 
scattering data of Weidner et al. (1982). The 
dashed curve is for the high pressure CaCl2 phase. 
The phase transition is predicted to occur at 45 
GPa at K. 



change and correlation functionals for the 
uniform electron gas. During the last ten 
years, this approximation has been demon- 
strated to be quite accurate for metals, 
semiconductors, and insulators, with some 
exceptions primarily in magnetic crystals. 

Calculations were performed for Si02 in 
the stishovite (rutile, space group P4jJ 
mnm), CaCl2 {Pnnm), and pyrite (Pa3) 
structures as functions of volume. The 
internal structural parameters and lattice 
parameters were optimized at each vol- 
ume. Fitting the total energies to an equa- 
tion-of-state and to polynomials as func- 
tions of strain and phonon amplitudes gives 
the Alg and Big Raman modes as well as 
four independent elastic constants for 
stishovite as functions of pressure. The set 
of calculations also gives the phase transi- 
tion pressures from stishovite to the Pa3 
and CaCl2 structures. All of the present 
calculations are for static lattice energies, 
classically equivalent to a temperature of 
K. 

Figure 68 shows the energy versus vol- 
ume for the three phases. A phase transi- 
tion from stishovite to Pa3 is predicted at 
156 GPa. This is significantly higher in 
pressure than the value of 60 GPa obtained 
by Park et al. (1988) using LAPW. The 
reason for the difference is not clear, but 
extensive convergence tests of the present 
results indicate stability of about 1 GPa in 
the transition pressure with respect to the 
convergence parameters. 

Figure 69 shows the calculated elastic 
constant c\\-c\2 as a function of pressure. 
An elastic instability (c\\-c\2 vanishes) in 
stishovite is predicted at 45 GPa, at which 



128 



CARNEGIE INSTITUTION 



pressure stishovite would transform into 
the CaCl2 structure at zero temperature, 
Figure 68 shows that this phase transition 
has only a small effect on the energetics of 
Si02. It also has only a small effect, ini- 
tially, on the crystal structure. The phase 
transition is continuous, and the initial dis- 
tortions from the rutile structure are very 
small. However, the phase transition has 
enormous effects on the elastic properties 
of high pressure Si02, so it is crucial to 
consider whether this transition occurs un- 
der mantle conditions. 

The phase transition is likely to be very 
sensitive to temperature since the instability 
is driven by a Raman mode. Figure 70 
shows the Aig and Big Raman modes as 
functions of pressure. Agreement with the 
Raman data obtained by Hemley (1987) is 
excellent. At the CaCl2 phase transition 
the Big mode becomes an Ag mode and 
begins to increase in frequency with pres- 
sure, whereas as lower pressures the fre- 
quency decreases with increasing pressure. 
This is probably the most direct way of 
detecting the phase transition, since the 
distortions are very small at the transition 
and would be difficult to detect with x-rays. 
The phase transition is not a soft-mode 
transition in the traditional sense, since the 
Big frequency does not reach zero at the 
transition. The Big mode does drive the 
transition, since if the atoms were not al- 
lowed to distort along the Big eigenvector 
during a c\i-c\2 strain, c\\-c 12 would not 
become unstable. It is the coupling be- 
tween the strain and the phonon displace- 
ment that makes the energy decrease as a 
function of strain, and thus causes the phase 
transition. The Big Raman mode contrib- 




150 



Fig. 70. Pressure dependence of the Ai g and Big 
Raman frequencies for stishovite, and an Ag 
mode in CaCl2- The triangles and diamonds are 
experimental data from Hemley (1987). The 
dashed line is a prediction for the CaCl2 phase. 
Note that the "soft mode" does not go to zero, or 
even get very small at the phase transition. 



utes a factor to c\\-c\2 that is proportional 
to -1/co 2 (Miller and Axe, 1967), so that as 
the Big frequency decreases the elastic 
constant c\\-c yi is destabilized. Tempera- 
ture, however, would stabilize the higher 
symmetry rutile structure. At 150 GPa, the 
well depth (energy gain on the phase tran- 
sition) expressed as temperature is only 
1 275 K, so at mantle temperatures of 2000- 
3000 K the transition may not occur until 
much higher pressures than the 45 GPa 
calculated for zero temperature. The exact 
transition temperature can be calculated 
using molecular dynamics or Monte Carlo 
simulations using a potential model, or by 
fitting a simplified Hamiltonian to the total 
energy results and using statistical thermo- 
dynamics to evaluate the phase diagram 
(Rabe and Joannopoulos, 1987). Both ap- 
proaches require information about the 



GEOPHYSICAL LABORATORY 



129 



coupling between cells, as well as the total 
energies for zone center distortions that do 
not increase the unit cell size. A potential 
model is now being developed for high- 
pressure Si02 to investigate the thermal 
properties of stishovite and the tempera- 
ture dependence of the phase transition. It 
is quite possible that temperature will sig- 
nificantly increase the depth at which the 
transition occurs, and thus the transition 
may be partly responsible for the anoma- 
lous seismic properties of the so-called D" 
region. 



References 

Cohen, R. E., Calculation of elasticity and high 
pressure instabilities in corundum and 
stishovite with the potential induced breathing 
model, Geophys. Res. Lett., 14, 37-40, 1987. 

Cohen, R. E., Bonding and elasticity of stishovite 
Si02 at high pressure: Linearized augmented 
plane wave calculations, Amer. Mineral., 76, 
733-742, 1991a. 

Cohen, R. E., First-principles predictions of elas- 
ticity and phase transitions in high pressure 
Si02 and geophysical implications, in High 
Pressure Research in Mineral Physics: Appli- 
cation to Earth and Planetary Science (Pro- 
ceedings of U.S. -Japan Conference on High 
Pressure Geophysics, Ise, Japan, January, 
1991), M.H. Manghnani and Y. Syono, eds., 
in press, 1991b. 

Fei, Y. and R. J. Hemley, Stability of (Fe,Mg)Si03- 
perovskite in the lower mantle, Geophys. Res. 
Lett., in press, 1991. 

Hedin, L. and B. I. Lundqvist, Explicit local 
exchange-correlation potentials, /. Phys., C4, 
2064-2083, 1971. 

Hemley, R. J., Pressure dependence of Raman 
spectra of Si02 polymorphs: a-quartz, coesite, 
and stishovite. In High -Pressure Research in 
Mineral Physics, M.H. Manghnani and Y. 
Syono, eds., pp. 347-359. American Geo- 
physical Union, Washington, D.C., 1987 

Hohenberg, P., and W. Kohn, Inhomogeneous 
electron gas, P/ry.s./?ev.,7.?<5£, 864-871, 1964. 

Kohn, W. and L. J. Sham, Self-consistent equa- 
tions including exchange and correlation ef- 



fects, Phys. Rev., 140 A, 1133-1140, 1965. 

Miller, P. B. and F. D. Axe, Internal strain and 
Raman active vibrations in solids, Phys. Rev., 
163, 924-926, 1967. 

Park, K. T., K. Terakura, and Y. Matsui, Theoreti- 
cal evidence for a new ultra-high-pressure 
phase of Si02, Nature, 336, 670-672, 1988. 

Rabe, K. M., and J. D. Joannopoulos, Theory of 
the structural phase transition of GeTe, Phys. 
Rev. B, 36, 6631-6639, 1987. 

Tsuchida, Y. and T. Yagi, A new, post-stishovite 
high-pressure polymorph of silica, Nature, 
340, 217-220, 1989. 

Tsuneyuki, S . , M. Tsukada, H. Aoki, and Y. Matsui, 
First-principles interatomic potential of silica 
applied to molecular dynamics, Phys. Rev. 
Lett., 61, 869-872, 1988. 

Tsuneyuki, S., Y. Matsui, H. Aoki, andM. Tsukada, 
New pressure-induced structural transforma- 
tions in silica obtained by computer simula- 
tion, Nature, 339, 209-21 1, 1989. 

Wei, S. H., and H. Krakauer, Local density func- 
tional calculation of the pressure induced phase 
transition and metallization of BaSe and BaTe, 
Phys. Rev. Lett., 55, 1200-1203, 1985. 

Weidner, D. J., J. D. Bass, A. E. Ringwood, and W. 
Sinclair, The single-crystal elastic moduli of 
stishovite, /. Geophys. Res., 87, B4740-4746, 
1982. 



Molecular Dynamics Simulations of 
Melting of MgO at High Pressures* 

Zhaoxin Gong, Ronald E. Cohen, and 
Larry L. Boyer** 

We are developing a general-purpose 
molecular dynamics (MD) program for 
studying finite clusters of atoms using 
many-body potentials and long-range 
forces. This program will be used to study 
high-pressure and high-temperature phase 



* This work is supported by the Office of Naval 
Researchgrant#N00014-91-J-1227toREC. Com- 
putations were performed with the support of 
ONR at the NRL Connection Machine Facility 
and at the Pittsburgh Supercomputer Center under 
the auspices of the National Science Foundation. 
*Complex Systems Theory Branch, Naval Re- 
search Laboratory, Washington, D.C. 20375 



130 



CARNEGIE INSTITUTION 



transitions as well as ferroelectricity. 
Though many simulations have been per- 
formed on various systems of interest, rela- 
tively few have been performed on ionic 
systems using first-principles, non-empiri- 
cal potentials. Such potentials are expected 
to be more reliable outside the range of 
experimental data than empirical poten- 
tials, and furthermore can be constrained 
for arbitrary ionic configurations, not 
merely determined from bulk properties 
with atomic positions close to their equilib- 
rium sites. A cluster approach is important 
because periodic boundary conditions have 
been found to greatly inhibit phase transi- 
tions such as melting, especially in systems 
with long-range forces, because the peri- 
odic boundary conditions force the "liq- 
uid" state onto a periodic lattice. Also, in 
the case of ferroelectricity, electric fields 
and macroscopic polarization are by defi- 
nition "non-periodic," so that these phe- 
nomena can only be studied approximately 
using periodic boundary conditions. Finite 
clusters in free space were used in earlier 
MD studies of melting of NaF using a 
Gordon-Kim rigid ion potential (Boyer and 
Pawley, 1988; Boyer and Edwardson, 
1988). 

In order to use a cluster method, many 
atoms must be be included in order to 
minimize the effects of surfaces; however, 
with long-range forces the MD problem is 
an N 2 problem, where N is the number of 
atoms in the cluster, so computational effi- 
ciency is very important. We present re- 
sults here using a massively parallel com- 
puter, the Connection Machine CM-2, as 
well as results calculated on an IBM RS/ 
6000 superworkstation and a Cray YMR 



Here we report for the first time MD results 
of melting of MgO at high pressures using 
a many-body potential, the Potential In- 
duced Breathing (PIB) model. 

The PIB model is an ab initio model 
where no parameters are fitted to the ex- 
periment. This model has been very suc- 
cessful in predicting the thermodynamic 
and elastic properties of alkaline earth ox- 
ides, including the Cauchy violations (c\\ 
is not equal to cyi at zero pressure) (Boyer 
era/., 1985; Mehle/tf/., 1986; Cohen etaL, 
1987; Isaak et al., 1990). Monte Carlo 
(MC) and MD simulations for the PIB 
model are computationally demanding and 
complicated. So far only primitive MC 
results of equation of state at zero pressure 
using the PIB model have been reported for 
a sample of MgO with 64 atoms (Cowley et 
al„ 1990). 

Isaak et al. (1990) found that the PIB 
model predicts that the zero pressure iso- 
thermal elastic modulus C s (=c\i-C22) of 
MgO becomes negative at a temperature 
very close to the melting temperature of 
MgO. The Cy instability occurs before the 
bulk modulus instability discussed by Boyer 
(1985) with increasing temperature if the 
free energy, including the vibrational con- 
tribution, is calculated as a function of 
strain. To find out whether this instability 
is related to the melting, we investigate 
here whether the same PIB potential indi- 
cates a melting point close to the 
quasiharmonic elastic instability. Further- 
more, are there elastic instabilities that cor- 
respond to melting at high pressures? 

An important question of geophysical 
importance is the curvature of the melting 
curve at high pressures. At low pressures 



GEOPHYSICAL LABORATORY 



131 



the melt is generally (but not always) less 
dense than the crystalline form of a mate- 
rial, but the liquid is more compressible. 
As pressure is increased, the difference in 
densities of liquid and solid should de- 
crease, and the slope of the melting curve, 
dT/dP, should decrease with pressure. The 
melting curve may become horizontal or 
may reach a maximum and bend over. 
MgO, being a close-packed solid, is an 
interesting case because of the simple crys- 
tal structure and the probably simple struc- 
ture of the liquid. MgO is also an 
endmember of magnesiowustite (Mg,Fe)0, 
which is considered a likely mineral in the 
Earth's lower mantle. 

We used the same numerical PIB poten- 
tials as in Isaak et al. (1990), but we have 
fit them to a different form in order to 
increase accuracy and computational effi- 
ciency. One problem is that the Madelung 
(electrostatic) potentials on the oxygen ions 
on the surface of the cluster are lower in 
magnitude than in the bulk. The quantita- 
tive results presented here must be consid- 
ered preliminary, because the surface atom 
potentials often fell outside the range of the 
points at which the PEB potential was fit. 
We expect the changes to be small when a 
better potential is used, because the form 
we chose appears to extrapolate smoothly, 
and the great majority of atoms had poten- 
tials that fell within range of the fit. 

We tested the MD code by comparing 
the zero-pressure density of a free cluster 
with that from quasiharmonic lattice dy- 
namics (LD) calculations using the same 
model (Isaak etai, 1990). The density was 
calculated using the method of Boyer and 
Pawley (1988). At T= 1300 K, the average 



density for the innermost region of a cluster 
with 216 atoms differs from the LD result 
by about 7%. The difference decreases to 
about 6 % when the cluster size is increased 
to 512 atoms. We also found that the total 
energy is conserved to better than six sig- 
nificant figures over many thousands of 
time steps, giving confidence in the MD 
code. 

We performed a series of runs to find 
out the melting temperature at zero pres- 
sure. The starting configuration of 216 (= 
6x6x6) atoms in their perfect lattice 
positions was initially given a small amount 
of kinetic energy, equivalent to a tem- 
perature T = 300 K. The simulation was 
then allowed to proceed at constant energy 
for 30 ps (10,000 time steps); then the total 
energy of the cluster was increased by 
scaling up the velocities to give an increase 
in temperature of 400 K. This procedure 
was repeated several times. The results are 
drawn in Figures 71 and 72 as solid lines. 




0.55 



060 0.65 

E + 275 (Hartree) 



0.70 



Fig. 71. Plot of equilibrium temperature T versus 
the total energy E at P = GPa. The solid line 
corresponds to increasing energy while the dotted 
line corresponds to decreasing energy. 



132 



CARNEGIE INSTITUTION 




055 



060 



065 



0.70 



E + 275 (Hartree) 



Fig. 72. Zero pressure plot of equilibrium density 
versus the total energy, E, in a cubic region cen- 
tered at the center of mass with the cube side equal 
to the nearest neighbor distance of the perfect 
lattice. The solid line corresponds to increasing 
energy while the dotted line corresponds to de- 
creasing energy. 



At low temperatures, the temperature is 
almost a linear function of the energy E. 
Thus MgO is quite harmonic up to about 
half the melting temperature. The tempera- 
ture increases as the energy increases until 
T= 3815 K, where increasing energy leads 
to a temperature decrease to about 3045 K. 
(This run is three times longer than the 
other runs in this figure.) The temperature 
of the cluster then increases again as the 
energy is increased. The density of cluster, 
on the other hand, decreases slowly as the 
energy is increased until the temperature 
reaches about T= 38 1 5 K. The density then 
decreases by a much larger amount as the 
energy is increased by another increment 
and the temperature drops to 3045 K. This 
simultaneous larger decrease in tempera- 
ture and density indicates that the cluster 
melted and kinetic energy is transferred 
into potential energy (Boyer and Pawley, 



1988; Boyer and Edwardson, 1988). The 
dotted lines in these two figures correspond 
to the process of cooling. With decreasing 
energy, the cluster freezes to a crystal with 
defects due to the high cooling rate, which 
leads to hysteresis. The middle of the 
hysteresis loop is about 3100 K, which 
compares remarkably well with the experi- 
mental melting temperature of 3098 ± 20 K 
(Stull and Prophet, 1971) and the c\\ -c\ 2 
elastic instability (Isaak et al., 1990) using 
the same potential. 

In a second series of calculations, the 
cluster is confined in a box so that its 
pressure can be varied. It is natural to 
repeat procedures for a free cluster to find 
kinks in the T versus E curve, while the 
volume is kept fixed. At large volumes 
(low pressures) this procedure is satisfac- 
tory, but it becomes increasingly difficult 
to locate the kinks as they becomes less and 
less prominent as volume becomes smaller 
and smaller. At a kink, either a large differ- 
ence in temperature Tor a large difference 
in pressure P is observed. When the cluster 
melts, its temperature decreases while its 
pressure increases. For example, when the 
box size is set equal to 23.735 bohr, the 
drop in temperature is about 260 K, from 
about 5880 K to about 5620 K, and the 
increment in pressure is about 4.5 GPa, 
from about 7.7 GPa to about 12.2 GPa. 
However, for a box of cube side of 23.23 
bohr, we were not able to identify any kink 
with the drop in temperature larger than 70 
K and the increment in pressure larger than 
1.7 GPa. 

To determine the melting temperatures 
at high pressures, we first find the tempera- 
ture and energy at given pressure and given 



GEOPHYSICAL LABORATORY 



133 



* 8- 




0.55 0.60 0.65 0.70 

E + 275 (Hartree) 



0.75 



Fig. 73. Plot of equilibrium temperature T versus 
the total energy E at P = 10 GPa. The solid line 
corresponds to increasing volume while the dotted 
line corresponds to decreasing volume. 




P (GPa) 

Fig. 74. Plot of the melting temperature Tm versus 
the pressure P. The bars indicate the maxima and 
minima of the heating and cooling curves, respec- 
tively, around the melting points. 



volume, then we plot the temperature ver- 
sus energy at a given pressure (Fig. 73). 
The solid line represents the process of 
increasing volume, while the dotted line 
represents decreasing volume. The cube 
side increment in this figure is 0.505 bohr, 
and the starting box side is 20.2 bohr. We 
identify the kinks in the plot as associated 
with the melting. 

In Figure 74 we plot the melting tem- 
perature versus pressure. The melting tem- 
perature, because of the van der Waals loop 
typically present in our results, is taken to 
be the average temperature of two middle 
points in the heating and cooling curves. 
Although only three points have been cal- 
culated, it is obvious that there is signifi- 
cant curvature in the melting curve. Melt- 
ing temperatures are presented only to 20 
GPa because of an instability that occurred 
at high temperatures, due to the fitting 
range of the Watson sphere potential. At 
high temperatures the Watson potentials of 



many atoms lie outside the fitting range, 
which can be corrected by the development 
of better inter-atomic potentials. Further 
results will be on very high pressures and 
larger samples. 

Although the PIB model is more com- 
plicated than the rigid ion model, many 
calculations can utilize parallel processing. 
For example, checking for atoms that cross 
the box boundary can be done in parallel in 
all three directions for all the atoms in the 
cluster. The most time consuming part, the 
force calculation, can also be done in paral- 
lel. Many simulations were done on a CM- 
2 from Thinking Machine Inc. at the Naval 
Research Laboratory (NRL). On the CM- 
2, the row column difference (RCD) method 
by Boyer and Pawley (1988) is used when 
calculating pair interactions. The block 
sizes used are the largest possible, i.e, block 
sizes of N x N, which has been shown to be 
the most efficient way of calculating pair 
interactions for a cluster of 512 atoms 



134 



CARNEGIE INSTITUTION 



Table 21. Timing results (sec per time step) for MD simulations on the CM (one 8k 
sequencer), the CRAY-YMP (one processor), and the IBM RS/6000-320 power worksta- 
tion. 



Rigid Ion Model* 



PIB Model 







CM 


CM 


CRAY 


IBM 


CPU time 
Elapsed Time 


216 atoms 
512 atoms 
1000 atoms 

216 atoms 
512 atoms 
1000 atoms 


0.23 
0.76 

0.32 
1.44 


0.13 
0.31 
1.02 

0.23 
0.43 
1.22 


0.055 
0.29 
1.01 

0.082 
0.46 
2.16 


0.31 

2.45 
12.3 

0.32 
2.50 
12.5 



The timing results for rigid ion model are from Boyer and Edwardson (1988). 



(Boyer and Edwardson, 1988). For a clus- 
ter of 1000 atoms, Boyer and Edwardson 
(1988) were limited to a block size of 256 
x 256 by memory constraints. In the 
present hardware at NRL, we encountered 
no memory constraints for clusters of up to 
1000 atoms. Timings are given in Table 21 . 
The elapsed times are for an otherwise 
empty machine except for the Cray tim- 
ings, which were taken under a typical 
load. 

In summary, we have presented first 
results on melting of a oxide using a non- 
empirical many-body model. MD simula- 
tions were performed on finite clusters 
containing up to 1000 atoms on the mas- 
sively parallel CM-2. Significant curva- 
ture in the melting curve is predicted. This 
method will be applicable to study a large 
variety of phase transitions as functions of 
temperature and pressure. 



References 

Boyer, L. L., Theory of melting based on lattice 
instability, Phase Transitions, 5, 1-48, 1985. 

Boyer, L. L., M. J. Mehl, J. L. Feldman, J. R. 
Hardy, J. W. Flocken,, C. Y.andFong, Beyond 
the rigid ion approximation with spherically 



symmetric ions, Phys. Rev. Lett., 54, 1940- 
1943, 1985. 

Boyer, L. L., and P. J. Edwardson, Application of 
massively parallel machines to molecular dy- 
namics simulation of free clusters, Proceed- 
ings of the 2nd symposium on the frontiers of 
. massively parallel computation, Fairfax, Vir- 
ginia, USA, October 10-12, 1988. 

Boyer, L. L., andG. S. Pawley, Molecular dynam- 
ics of clusters of particles interacting with 
pairwise forces using a massively parallel com- 
puter, /. Comput. Phys., 78, 405-423, 1988. 

Cohen, R. E., L. L. Boyer, and M. J. Mehl, Lattice 
dynamics of the potential-induced breathing 
model: Phonon dispersion in the alkaline -earth 
oxides, Phys. Rev. B, 35, 5749-5760, 1987. 

Cowley, E. R., S. H. Liu, and G. K. Horton, Monte 
Carlo calculations of the equations of state of 
alkaline earth oxides, Ferroelectrics, 111, 33- 
42, 1990. 

Isaak, D. G., R. E. Cohen, and M. J. Mehl, Calcu- 
lated elastic and thermal properties of MgO at 
high pressures and temperatures, J. Geophys. 
Res., 95, 7055-7067, 1990. 

Mehl, M. J., R. J. Hemley, and L. L., Boyer, 
Potential-induced breathing model for the elas- 
tic moduli and high pressure behavior of the 
cubic alkaline-earth oxides, Phys. Rev. B, 33, 
8685-8696, 1986. 

Stull, D. R. and H. Prophet (Eds.), JANAF Ther- 
mochemical Tables, 2nd ed., Office of Stan- 
dardReference Data, NIST, Washington, D.C., 
1971. 



GEOPHYSICAL LABORATORY 



135 



Glass Diffraction Measurements with 
Polychromatic Synchrotron Radiation 

Charles Meade and Russell J. Hemley 

The structure of liquids and glasses at 
high pressures is an issue of great impor- 
tance in the Earth and Material Sciences. At 
present, however, little is known on this 
subject because of the difficulty of obtain- 
ing direct information about the structure 
of noncrystalline materials at high pres- 
sures. Spectroscopic studies (both optical 
and x-ray) have provided constraints on the 
vibrational properties and short-range or- 
der in glasses under pressure, but they 
typically provide only an indirect probe of 
structure. Perhaps the most direct means of 
determining glass and liquid structures at 
very high pressures (P > 10 GPa) is to 
obtain x-ray diffraction measurements from 
these materials in the diamond cell. Here, 
we investigate the use of high-energy poly- 
chromatic synchrotron radiation for this 
purpose. 

To obtain precise measurements of glass 
diffraction under these conditions, two prob- 
lems must be addressed. First, one has to 
know the intensity distribution of the x-ray 
source to interpret the x-ray patterns. Sec- 
ond, in the analysis of the measured dif- 
fraction spectra, one must be able to sub- 
tract the background diffracted intensity 
produced by the diamond anvils (both Bragg 
and Compton scattering) from the signal 
due to the amorphous material. 

In this report, we will address the first 
question, constraining the x-ray source spec- 
tra in glass diffraction experiments. To il- 
lustrate a new analysis that we have devel- 



oped for this problem, and to demonstrate 
the advantages of using high energy syn- 
chrotron radiation for these experiments, 
we have measured the x-ray diffraction 
from a small platelet of Si02 glass under 
ambient pressures and temperatures at the 
X-17C beamline of the National Synchro- 
tron Light Source, Brookhaven National 
Laboratory (B adding et a/., 1990; Mao et 
al. 1990). Because these measurements 
were made outside of the diamond cell on 
a known material (Mozzi and Warren, 1 969; 
Konnert and Karle, 1973), they provide an 
initial test of our ability to constrain the 
intensity distribution of the x-ray source in 
our experiments. 

Examples of the measured and known 
diffraction patterns for Si02 glass are shown 
in Figure 76. From this comparison, it is 
immediately evident that that the shape of 
the x-ray source distribution strongly influ- 
ences the observed x-ray measurements. 
To normalize these diffraction patterns, 
one can measure the intensity of the x-ray 
source; however, these measurements are 
difficult given the the detector geometry at 
X-17C. One could calculate the source 
distribution (e.g. Krinsky et al., 1983), 
though, this requires precise knowledge of 
the upstream filters and the critical energy 
of the superconducting wiggle r. Moreover, 
calculated spectra cannot account for par- 
tial absorption of the beam by upstream 
slits. 

Thus, we have adopted an alternative 
approach where we take advantage of the 
redundancy in our diffraction spectra ob- 
tained at several different scattering angles. 
Consider that the observed intensity spec- 
tra at particular diffraction angle 20/ (cor- 



136 



CARNEGIE INSTITUTION 




J tMJM T H"M J Tint "" J l t ti p I I I [ I I H |M t l|l 



B 




s.A- 1 



Fig. 76. (A) Diffraction pattern for Si02 glass at ambient pressures and temperatures measured from 
monochromatic x-ray source by Konnert and Karle (1973). The horizontal axis is the wavenumber 
S=47tsm0/X. (B) Observed intensities (corrected for sample and air absorption) from Si02 glass 
measured at two different scattering angles (26) with a polychromatic x-ray source at beamline X-17C 
of the National Synchrotron Light Source. The differences between these data and with the previously 
measured values are due to the distribution of x-ray intensities in the polychromatic synchrotron source. 

rected for air and sample x-ray absorption common scale we define the wavenumber 
and polarization of the diffracted beam) s = AKsin6lX. We can then write equation 
can be represented as (1 ) as 



/$ s (£) = [B(28i, E) + C(2Q, E)]f (E) ( i ) / obs W = [B(s) + C(j)]/ o (£,20) 



(2) 



where B and C are functions that respec- 
tively describe the coherent (Bragg) and 
incoherent (Compton) scattering from the 
sample and Io is the source 
spectrum. Whereas 5 and Care functions of 
2 and the energy of the incident beam (E), 
Io is a function of E only. In practice, we 
make measurements at several diffraction 
angles. Thus, to express all of the data on a 



Here, B, C, and I obs are functions of s only. 
In this representation, the source function 
depends on s and 20. With this transforma- 
tion, the measurements at different diffrac- 
tion angles can be collapsed down to the 
single function Iobs(s). Even though/? and 
C are not known a priori, one can use the 
knowledge that they are functions of s only 
to constrain the source function Iq. Specif i- 



Fig. 77. (A) Composite diffraction pattern for Si02 from measurements at seven angles ranging from 
2 = 6° to 45°. Below s = 1 , the data are extrapolated to zero intensity. (B) The same diffraction spectra 
as shown in Figure 76B. When the spectra are normalized with the correct source function, they are 
equivalent when plotted in terms of the wavenumber s. For clarity, the spectra are offset in the vertical 
direction. (C) X-ray source function for X-17C that is consistent with these measurements. The 
intensities drop off below 12 keV because of absorption in upstream beryllium and carbon filters. 



GEOPHYSICAL LABORATORY 



137 







1"" 


1 1 ii 


wh 


nrpiii|iiii|iin 


1 1 ■ 1 1 1 1 1 ( 1 


pm 


iTrrpTTT 


jm 


Mill 


















c 


"\ 




- 
















_ 


CO 

c 

CD 

•4— • 


















- 




u„J 


■ ml 


Mill 


jj,Liiii lui liin ' 


.ml.... 


I....I 


111 1I111 1 


7177 


TTTTT 



cally, we determine Io by requiring that 
I 0DS {s) must be the same for all diffraction 
angles. 

An example of a "composite" diffrac- 
tion pattern obtained in this way and the 
corresponding source function is shown in 
Figure 77. When compared to the previ- 
ously measured pattern, the advantages of 
using this approach is apparent. Because 
one can measure diffraction to high ener- 
gies (-60 keV) at high angles, diffraction 
spectra can be obtained to much higher 
values of s than in conventional laboratory 
measurements. In this study, we obtained 
data to s = 26 A" 1 . Measurements to 40 A" 
1 are possible. We expect that such mea- 
surements over a wide range of s values 
will allow strong constraints on glass struc- 
ture at high pressures. 



References 

Badding, J.V., H. K. Mao, J. Z. Hu, R. J. Hemley, 
C. Meade, and J. F. Shu, High pressure energy 
dispersive x-ray diffraction at X-17C, EOS, 
71, 1620, 1990. 

Konnert, J. H., and J. Karle, The computation of 
radial distribution functions for glassy materi- 
als, Acta. Cryst., A29, 702-710, 1973. 

Krinsky, S., M. L. Perlman, and R. E. Watson, 
Characteristics of synchrotron radiation and 
of its sources, in Handbook on Synchrotron 
Radiation, E. E. Koch, ed., North Holland, 
New York, pp. 65-171, 1983. 

Mao, H. K., J. Z. Shu, J. F. Shu, and R. J. Hemley, 
Ultrahigh-pressure experimentation above 300 
GPa, Bull Ame.r Phys. Soc, 36, 529, 1990. 

Mozzi, R. L., and B. E. Warren, The structure of 
vitreous silica, /. Appl. Cryst., 2, 162-172, 
1969. 



10 20 30 40 50 
Energy, keV 



60 70 



138 



CARNEGIE INSTITUTION 



X-ray Diffraction of Solid Nitrogen- 
Helium Mixtures* 

Willem L. Vos, Larry W. Finger, 
Russell J. Hemley, Ho-Kwang Mao, 

Jing Zhu Hu, Jin Fu Shu, 

Richard LeSar* Andre de Kuijper, * 

and Jan A. Schouten** 

The study of simple molecular mixtures 
in their solid phases at high pressures has 
recently become of increasing interest. This 
work is important for statistical and con- 
densed matter physics, specifically to in- 
vestigate the influence of size ratio, pack- 
ing, and van der Waals-like interactions on 
the stability of mixed structures. Such stud- 
ies also provide a starting point for for 
modeling the interiors of the outer planets 
of our solar system. 

One of the most studied mixtures in this 
respect is N2-He; helium is still a fluid 
under conditions where nitrogen has so- 
lidified (see Fig. 78). At room temperature, 
their freezing pressures differ by a factor of 
5: i.e., 24 vs. 120 kbar). By a combination 
of visual observations and quasi-isochoric 
p-T scans, Vos and Schouten (1990) found 
that around 10 mol% He is soluble in the 
orientationally ordered rhombohedral e - 
phase of nitrogen (Mills et ai, 1986); in 
contrast, the solubility of helium in the 
disordered (3 or 8 - phases (Cromer et ai, 

*This collaboration was supported by a NATO 
Collaborative Research Grant. 
* T-ll, Los Alamos National Laboratory, Los 
Alamos New Mexico 87545 

van der Waals-Zeeman Laboratorium, 
Universiteit van Amsterdam, 1018XE Amsterdam, 
The Netherlands. 



1981) is negligible. Due to this solubility, 
the stability of the e-phase was found to 
increase drastically with respect to the 8 - 
phase (see Fig. 78). The shift of the 8 - e 



100 



1 

J* 
<xi 

3 



CD 

k— 
CL 




250 350 450 550 

Temperature, K 

Fig. 78. PTphase diagrams of N2 (solid lines), He 
(short dashed line), and projected phase diagram 
of N2-He (dashed lines). The solid lines are in 
order of increasing pressure the P-fluid, 5-p and 
e-8 lines, the short dashed line is the melting line 
of helium and the dashed lines are the three - phase 
lines e-8 -F of the mixture. 



transition was subsequently confirmed by 
vibrational Raman spectroscopy 
(Scheerboom and Schouten, 1991). Here, 
we report preliminary investigations of the 
structure of the mixed phases (indicated by 
asterisks) by x-ray diffraction and a com- 
parison of the results with computer simu- 
lations. 

X-ray diffraction studies were performed 
at beamline X-17C of the National Syn- 
chrotron Light Source, at Brookhaven Na- 
tional Laboratory, using a single-crystal 
type diamond anvil cell (Mao and Bell, 
1980). The samples were prepared from 
high purity (99.999%) gases and loaded in 



GEOPHYSICAL LABORATORY 



139 



a pressure vessel at a pressure of about 2 
kbar. All experiments were performed at 
room temperature, and pressures were ob- 
tained from the linear ruby scale. Three 
experiments were performed: one on pure 
N2, to check the structure of the e-phase, 
since this had been obtained from powder 
diffraction (Mills et al., 1986, Olijnyk, 
1990), and two on the 8* - phase at compo- 
sitions of 5.0(2) and 10.0(2) mol% helium. 
The diffraction experiment on e-N2 was 
performed at 195 kbar since the transition 
from 8 to e takes place at 1 65 kbar (Olijnyk, 
1990). The sample showed clear colors 
under crossed polarizers, which distin- 
guishes it from the non-birefringent 8 - 
phase. The sample consisted of at least one 
large crystal and many small grains, since 
strong reflections were observed at distinct 
(X,oS) angles, while a powder-like pattern 
with preferred orientation was observed 
irrespective of the ix,(o) angles. This con- 
figuration was often encountered and was 
named "single-powder." The d-spacings of 
the powder pattern agree with the ones 
reported previously (Mills et aL, 1986; 
Olijnyk, 1990) and yield a unit cell of 
dimensions a= 7.63(4) A, c= 10.37(10) A, 
c/a= 1.36 with 24 molecules and space 

group R3c (Mills et ai, 1986). Ten reflec- 
tions were observed from the single crys- 
tal, that were also consistent with the R3c 
structure. Special attention was paid to 
overtones and to possible lower-order re- 
flections, since some computer simulation 
results yielded superstructures with unit 
cells containing up to 64 molecules (Nose 
and Klein, 1986, Belak era/., 1990). How- 
ever, no longer spacings were found and 
only a few weak overtones were observed, 



also consistent with the aforementioned 
structure. 

With 5 mol % He, experiments were 
performed at 126 and 144 kbar. The pres- 
ence of the e* - phase was checked by the 
observation of birefringence. The sample 
consisted again of a "single-powder." The 
spectrum of d-spacings consisted of lines 
from both 8* and £* - phases. This was 
confirmed by Raman scattering, which 
showed that the vi mode was clearly split, 
which is not the case in the pure phases. The 
spacings of the 8* - phase are the same as 
those of the 8 - phase at the same pressures 
(Olijnyk, 1990), which means that there is 
negligible solubility of helium in this phase. 
The other d-spacings could all be fitted to a 
hexagonal phase with a - 8.050(5) A, c - 
9.469(12) A, da = 1.176 at 126 kbar and 
7.959(5) A, 9.353(17) A, 1.175 respec- 
tively at 144 kbar. Since the intensities 
from the 8* - phase are large both in the x- 
ray diffraction and the Raman scattering 
experiments, this phase must be a signifi- 
cant portion of the volume. From mass- 
balance considerations, the composition of 
the e* - phase is clearly larger than 5 mol % 
He. 

The experiment on e* with 1 % He was 
performed at 92 kbar. Before the experi- 
ment was done, the sequence of transitions 
leading to this phase was verified, as well 
as the vibrational Raman spectrum. The 
sample consisted of several crystals, but no 
additional powder, probably due to the 
presence of a small amount of the helium- 
rich fluid phase. Strong reflections includ- 
ing series of overtones were obtained (see 
Fig. 79). From the measured d-spacings, a 
hexagonal unit cell with a - 8.258(2) A, c 



140 



CARNEGIE INSTITUTION 



100000 




' t t t I 



■ 



30 



50 



70 



Energy (keV) 



Fig. 79. Diffraction pattern of e* with 10 % helium 
taken at 26=9°. The fundamental reflection of the 
(101) class at 13.7 keV and 4 overtones at 27.3, 
41.1, 54.7 and 68.4 keV are indicated by the 
arrows. The peaks near 18 keV are escape peaks 
from the first overtone and the other peaks origi- 
nate from the gasket. 



= 9.747(5) A and c/a=1.180 was obtained, 
similar to the results at 5 mol% He. Four 
reflections could be attributed to one crys- 
tal and five reflections to a second one. 

The reflections of the e* - phase do not 
fulfill the condition -h+k+l=3n of rhombo- 
hedral structures. Therefore, we can rule 

out the R3c space group for this phase. 
However, the contents of the unit cell can 
be estimated on the basis of the available 
information. We assume the phase to be 
stoichiometric and calculate the free en- 
thalpy difference between this phase and 
the coexisting phases, using known p(V) 
isotherms of N2 (Olijnyk, 1 990) and He (Le 
Toullec et ai, 1990) and making assump- 
tions for the volumes of the mixture that are 
justified by theory and simulations. It then 
turns out that the e* - phase becomes stable 
when it contains at least 22 N2 molecules 
and 2 He atoms per unit cell. This also 
yields a volume difference with the coex- 



isting phases that is reasonable in view of 
the pressure jumps measured previously at 
transitions involving this phase (Vos and 
Schouten, 1990; Vos, 1991). Furthermore, 
this is also consistent with the fact that the 
Raman shifts of this phase are very similar 
to those of the 8 and e - phases at the same 
pressure (Scheerboom and Schouten, 1 99 1 ). 
Computer simulations were performed 
using a variable shape simulation cell 
(Rahman-Parinello method) to allow for 
crystal transformations. The best available 
potentials for N2 and He were used (for 
details see de Kuijper, 1991). At 300 K and 
pressures of 200 and 300 kbar, it was found 

that pure N2 remains in the R3c space 
group. This result was obtained with both 
hexagonal and rhombohedral simulation 
cells, which would have permitted the ap- 
pearance of the alternative structures that 
were observed earlier (Nose and Klein, 
1986; Belak et al, 1990) and therefore 
agrees with the experiment. For compari- 
son with pure N2, simulations were per- 
formed on N2-He, starting from the R3c 
space group. It turns out that this structure 
is maintained, indicating that it is at least 
metastable for the mixture. An interesting 
result is that the da drops considerably, 
from 1.30 in pure N2 to 1.17 at 10 mol % 
He. 

From the present x-ray experiments, it 
can be concluded that at room temperature 
there is a negligible solubility of He in the 
8 - phase of N2. Furthermore, the solubility 
in the e* - phase is clearly larger than 5 mol 
%, in agreement with previous experiments. 
However, it turns out that He stabilizes a 
phase with a different structure than pure e- 



GEOPHYSICAL LABORATORY 



141 



N2. This phase has a smaller c/a ratio than 
pure N2 at the same pressure, which is 
consistent with computer simulations. 

If the e* - phase turns out to be stoichio- 
metric, the behavior of N2-He documented 
here bears some similarity with that found 
for colloidal suspensions. There, it was 
recently found that mixtures with a size 
ratio of 0.61 (cf., roughly 0.6 for helium 
and nitrogen) form a stoichiometry with a 
structure that is not encountered in the pure 
components (Bartlett et al., 1990). 



Scheerboom, M. I. M., and J. A. Schouten, Detec- 
tion of the e-5 phase transition in N2 and the 
N2-He mixture by Raman spectroscopy: new 
evidence for the solubility of fluid He in solid 
N2,/. Phys.: Condens. Matter, in press, 1991. 

Vos, W. L., Phase equilibria in simple systems at 
high pressure, Ph. D. dissertation, Universiteit 
van Amsterdam, 1991. 

Vos, W. L., and J. A. Schouten, Solubility of fluid 
helium in solid nitrogen at high pressure, Phys. 
Rev. Lett., 64, 898-901, 1990. 



Evidence for Orientational Ordering of 
Solid Deuterium at High Pressures* 

Russell J. Hemley and Ho-Kwang Mao 



References 

Bartlett, P., R. H. Ottewill, and P. N. Pusey, 
Freezing of binary mixtures of colloidal hard 
spheres,/. Chem. Phys. ,93, 1299-1312, 1990. 

Belak, J., R. LeSar, and R. D. Etters, Calculated 
thermodynamic properties and phase transi- 
tions of solid N2 at temperatures 0<T<300 K 
and pressures 0<p<100 GPa, /. Chem. Phys., 
92,5430-5441,1990. 

Cromer, D. T., R. L. Mills, D. Schiferl, and L. A. 
Schwalbe, The structure of N2 at 49 kbar and 
299 K, Acta Cryst. B37, 8-11, 1981. 

de Kuijper, Jhr. A., Computer simulations of phase 
equilibria in molecular systems, Ph. D. disser- 
tation, Universiteit van Amsterdam 1991. 

Le Toullec, R., P. Loubeyre, and J. - P.Pinceaux, 
Refractive-index measurements of dense he- 
lium up to 16 GPa at T=298K: Analysis of its 
thermodynamic and electronic properties, 
Phys. Rev. B40, 2368-2378, 1990. 

Mao, H. K., and P. M. Bell, Design and operation 
of a diamond-window, high-pressure cell for 
the study of single-crystal samples loaded 
cryogenically, Carnegie Instn. Washington 
Year Book, 79, 409-411, 1980. 

Mills, R. L., D. T. Cromer, B. dinger, Structures 
and phase diagrams of N2 and CO to 1 3 GPa by 
x-ray diffraction, /. Chem. Phys., 84, 2837- 
2845, 1986. 

Nose, S., and M. L. Klein, Constant-temperature - 
constant-pressure molecular-dynamics calcu- 
lations for molecular solids: Application to 
solid nitrogen at high pressure, Phys. Rev. 
B33, 339-342, 1986. 

Olijnyk, H., High-pressure x-ray diffraction stud- 
ies on solid N2 up to 43.9 GPa, /. Chem. Phys., 
93, 8968-8972, 1990. 



At low pressures hydrogen forms an 
insulating molecular solid with the mol- 
ecules in states of complete rotational dis- 
order over a wide range of temperature. 
With increasing pressure, the rotational 
motion of the molecules is expected to 
become more restricted, ultimately leading 
to orientational ordering. Detailing the na- 
ture of possible ordering transitions is im- 
portant for understanding the mechanism 
of pressure-induced metallization, a transi- 
tion with important implications for both 
condensed-matter and planetary physics. 
Raman measurements of the high-fre- 
quency intramolecular vibrational mode 
(vibron) indicate that the molecular solid 
remains stable to at least -250 GPa but 
undergoes a phase transition at 150 GPa (at 
77 K; Hemley and Mao, 1988). The low- 
frequency vibrational spectrum provides 
information on the state of ordering in the 
solid and constraints on the crystal struc- 

* This work was supported by NSF (DMR-89 
12226 and EAR-8904080) and NASA (NAGW- 
1722). 



142 



CARNEGIE INSTITUTION 



ture at these pressures, which are beyond 
the range of current x-ray diffraction tech- 
niques (Mao etal., 1988). Previously, we 
reported measurements of the evolution of 
the low-frequency rotational bands and lat- 
tice phonon of hydrogen to 162 GPa at 77- 
295 K (Hemley et al. y 1990a). Over this 
pressure interval the rotational bands per- 
sist but broaden and the lattice phonon, 
which correlates with the Z?2g optical pho- 
non of the hexagonal-close packed struc- 
ture, shifts continuously. The continuity of 
the low-frequency bands as a function of 
pressure indicates that an underlying hex- 
agonal structure persists into the high-pres- 
sure phase above 1 50 GPa . 

Examination of the low-frequency spec- 
trum of deuterium is important for under- 
standing isotope effects on a variety of 
properties of hydrogen at high densities. 
Orientational ordering is energetically fa- 
vored in the heavier isotope as a result of its 
smaller rotational constant (#D2 = 29.9 
cm -1 versus #H2 = 59.3 cm -1 in the gas 
phase), which results in a stronger mixing 
of free molecule rotational states in con- 
densed phase. At low densities, changes in 
temperature result in variations in the rela- 
tive population of ortho and para species 
(even and odd / rotational states). How- 
ever, at high densities the single molecule 
ortho-para distinction breaks down as a 
result of mixing of rotational states, and 
this is expected to occur at lower densities 
in D2. Evidence for this is found in the 
differences in the pressure at which sym- 
metry breaking occurs in the 7=0 solids at 
very low temperatures (Silvera and 
Wijngaarden, 1981). Isotope effects are 
also observed in the pressure dependence 



Hydrogen 
93.8 GPa 




200 400 600 800 1000 

Wavenumber, cm" 1 



1200 



Fig. 80. Examples of low-frequency Raman spec- 
tra of hydrogen and deuterium at 77 K. The two 
broadened low-frequency bands observed in H2 
correlate with the So(0) and S\(0) rotational tran- 
sitions. The intense band observed in D2 at 250 
cm -1 is identified as a libra tional mode in an 
orientationally ordered structure. The optical pho- 
nons observed in both isotopes are indicated. 

of the molecular vibron; the shift is signifi- 
cantly stronger in hydrogen relative to deu- 
terium (see Hemley et al., 1991). It is 
therefore of interest to determine whether 
or not this difference is associated with 
structural differences between the two iso- 
topes. Finally, a distinct isotope effect is 
observed in the pressure of the low-tem- 
perature 150-GPa phase transition, which 
is >10 GPa higher in deuterium. Under- 
standing the origin of this effect is of inter- 
est because the transition appears to be 
associated with changes in electronic prop- 



GEOPHYSICAL LABORATORY 



143 



erties, such as metallization (Hemley etal., 
1990a; Hemley and Mao, 1990). 

Samples were loaded at room tempera- 
ture in a modified Mao-Bell diamond-cell 
with composite rhenium/T301 -stainless 
steel gaskets. Here we report the results of 
four separate experiments on deuterium 
carried out at pressures from 20 GPa to 
above 100 GPa and temperatures between 
77 K and 295 K. Raman spectra were 
measured using optical techniques de- 
scribed previously (Hemley and Mao, 1988; 
Hemley et aL, 1990a). Low-frequency 
Raman spectra show a strong, weakly pres- 
sure dependent, band at 240 cm -1 , together 
with a weaker peak at higher frequency 
which exhibits a large pressure dependence. 
A representative low-frequency Raman 
spectrum of deuterium at 99 GPa is com- 
pared with that measured for hydrogen at 
similar pressures in Fig. 80. On the basis of 
x-ray diffraction measurements (Mao et 
aL, 1988; Hemley et aL, 1990b) and the 
continuity of the spectra with increasing 
pressure, the higher frequency band is iden- 
tified as the E2g Raman-active phonon in 
the hexagonal-close packed structure, 
analogous to that found for hydrogen. 

The low-frequency spectrum of hydro- 
gen at 77 K below 100 GPa is characterized 
by two broadened rotational bands [So(0) 
and Si(0), corresponding to the AJ = 2, / = 
0-2 and AJ = 2, /= 1 -3 excitations in the free 
molecule]. The persistence of these bands 
to 100 GPa was interpreted as an indication 
of free (or nearly free) rotation of the mol- 
ecules. In contrast, the low-frequency band 
observed in D2 does not fit a rotational 
transition: i.e., the frequency of the band is 
240 cm~l, whereas the frequencies of the 



So(0) and Si(0) occur at 6£ = 180 cm" 1 and 
10# = 300 cm -1 . We interpret this band as 
indicative of a new high-pressure molecu- 
lar phase of deuterium. In view of the 
presence of the optical phonon, we suggest 
that the phase is an ordered (or partially 
ordered) form with a hexagonal close- 
packed structure and that the low-frequency 
band is associated with librational motion. 
The latter is consistent with its weak pres- 
sure dependence. 

Further evidence for ordering is found 
in the frequency shift of the optical phonon. 
The volume dependence of the optical pho- 
non frequency for the two isotopes is shown 
in Fig. 81. The D2 curve is parallel to that 
obtained by Wijngaarden et aL (1983) at 



E 
o 

i_r 

CD 

.Q 

E 

c 

CD 

> 

cd 



PHONON 
77 K 




200 _ 



Volume (cm 3 /mol) 

Fig. 81. Volume dependence of the optical pho- 
non for H2 and D2: squares, present work; circles, 
Wijngaarden etal. (1983). The volume was calcu- 
lated from the pressure using the experimental 
equation of state determined to 26.5 GPa at room 
temperature (Mao et aL, 1988; Hemley et aL, 
1990). The dotted line shows the frequencies 
expected for D2 on the basis of the measurements 
forH2. The data of Wijngaarden etal. (1983) have 
been corrected using the new equation of state. 



144 



CARNEGIE INSTITUTION 



much lower pressures, as noted previously 
for H2 (Hemley et al, 1990a). At low 
compressions, the frequencies of the modes 
differ by a factor of V2 as a result of the 
differences in masses [i.e., (mnz/^D2)^], 
as expected if the modes (assumed to be 
harmonic), crystal structure, and volume 
are identical for the two isotopes. It is 
evident, however, that this relationship does 
not hold at higher compressions: the D2 
curve is higher than that expected on the 
basis of the measurements for H2. This 
offset may arise from a modification of the 
crystal structure by ordering, which could 
affect the frequency of the optical phonon 
owing to changes in intermolecular inter- 
actions in the ordered state (even at con- 
stant volume). Alternatively, we note that 
orientational ordering should result in a 
more efficient packing of the molecules 
relative to the rotationally disordered state, 
and that the frequency of the phonon is a 
strong function of volume. Thus, a second 
possibility is that the offset indicates that 
the molar volume of deuterium (which is 
ordered) is lower than that of hydrogen 
(which appears not to be fully ordered) at 
the same pressures. This needs to be exam- 
ined by low-temperature x-ray diffraction. 
The available diffraction data provide some 
evidence for a lower volume in D2 even at 
room temperature (-2% at 30 GPa), so it is 
possible that effects of rotational ordering 
are present at room temperature (Fig. 81). 
It should be pointed out that if the volume 
difference at 77 K persists to higher pres- 
sure (>150 GPa), it may contribute to the 
isotope effect on the pressure of the high- 
pressure phase transition (i.e., the low- 
pressure phase is stabilized to higher pres- 



sures in the heavier isotope). 

The results may be compared with the 
measurements of Silvera and Wijngaarden 
(1981), who studied ortho-D2 at 5 K to 54 
GPa. They reported the observation of a 
broadened low-frequency band at 220-240 
cm - * above 28 GPa, which they interpreted 
as arising from an ordering-type transition 
(broken symmetry transition in the J = 
molecules). Since Silvera and Wijngaarden 
(1981) were unable to measure the optical 
phonon above the transition, they proposed 
that the high-pressure phase has the cubic 
Pa3 structure. The present observations of 
the optical phonon indicate that the struc- 
ture of the solid at 77 K (and above) is not 
Pa3 because the Raman-active excitations 
in this structure comprise only librational 
modes (phonon is inactive). The close simi- 
larity in the librational modes measured in 
the two studies strongly suggests that the 
transition observed at 5 K also takes place 
within the hep-type structure. This assign- 
ment is consistent with the results of recent 
theoretical calculations which indicate that 
structures based on hep are stable relative 
to cubic (e.g., Pa3) at high densities (Raynor, 
1987; Barbee etai, 1989; Ashcroft, 1991; 
Kaxirase/a/., 1991). 

Kaxiras et al. ( 1 99 1 ) have performed an 
extensive series of calculations of different 
molecular ordering schemes within hep. A 
new class of oriented hexagonal structures 
based on a herring-bone type configuration 
has been found to be energetically favored 
and to have larger band gaps than that of the 
structure assumed in previous work. We 
propose that the structure of the phase of D2 
observed here is closely related to these 
structures. This assignment is consistent 



GEOPHYSICAL LABORATORY 



145 



Deuterium 

130GPa 

77 K 




138 GPa 
295 K 




200 400 600 800 1000 

Wavenumber, cm -1 



1200 



Fig. 82. Raman spectrum of deuterium at 130-138 
GPa at 77 K and 295 K. 



with the experimental evidence for a band 
gap (insulating state) to high pressures 
(i.e., to at least -150 GPa), which is one of 
the key problems with previously proposed 
ordered hexagonal structures [see, Ashcroft 
(1991) and Kaxiras etal. (1991)]. Confir- 
mation of this structure should be possible 
by low-temperature single-crystal x-ray 
diffraction. Further work is also required to 
determine the P-T stability field of the 
phase as function of pressure, temperature, 
and ortho-para state (at lower pressures). 
The librational band weakens gradually 
with increasing pressure above 100 GPa, 
and diamond fluorescence tends to increase 
at these pressures. As a result, the phonon 
could not be measured above 100 GPa, 
although the stronger low-frequency band 
is readily apparent (Fig. 82). With increas- 
ing pressures above -140 GPa at 77 K the 



band appears to weaken somewhat but was 
observed through the high-pressure phase 
transition at -165 GPa, despite the marked 
discontinuity in the vibron frequency 
(Hemley etal., 1991). Hence, D2 is appar- 
ently ordered over the entire pressure inter- 
val of this study at 77 K, although the 
broadening of the band at room tempera- 
ture (Fig. 82) may indicate that the mol- 
ecules are disordered at higher tempera- 
tures. At still higher pressures, an increase 
in scattering intensity is observed in the 
vicinity of the low-frequency band, a de- 
tailed study of which will be presented 
elsewhere. 



References 

Ashcroft, N. W., Optical response near a band 
overlap: Application to dense hydrogen, in 
Molecular Systems under High Pressure, R. 
Pucci and G. Piccitto, eds., pp. 201-222, 
Elsevier, Amsterdam, 1991. 

Barbee, T. W., A. Garcia, M. L. Cohen, and J. L. 
Martins, Theory of high-pressure phases of 
hydrogen, Phys. Rev. Lett. 62, 1150-1153, 
1990. 

Hemley, R. J., and H. K. Mao, Phase transition in 
solid molecular hydrogen at ultrahigh pres- 
sures, Phys. Rev. Lett., 61, 857-860, 1988. 

Hemley , R. J., andH. K. Mao, Critical behavior in 
the hydrogen insulator-metal transition, Sci- 
ence, 249, 391-393, 1990. 

Hemley, R. J., H. K. Mao, and J. F. Shu, Low- 
frequency vibrational dynamics and structure 
of hydrogen at megabar pressures, Phys. Rev. 
Lett. 65, 2670-2673, 1990a. 

Hemley, R. J., H. K. Mao, L. W. Finger, A. P. 
Jephcoat, R. M. Hazen, and C. S. Zha, Equa- 
tions of state of solid hydrogen and deuterium 
from single-crystal x-ray diffraction to 26.5 
GPa, Phys. Rev. B 42, 6458-6470, 1990b. 

Hemley, R. J., H. K. Mao, and M. Hanfland, 
Spectroscopic investigations of the insulator- 
metal transition in solid hydrogen, in Molecu- 
lar Systems under High Pressure, R. Pucci 
and G. Piccitto, eds., pp. 223-243, Elsevier, 
Amsterdam, 1991. 

Kaxiras, E., J. Broughton, and R. J. Hemley, Onset 
of metallization and related transitions in solid 



146 



CARNEGIE INSTITUTION 



hydrogen, Phys. Rev. Lett., 67, 1138-1141, 
1991. 

Mao, H. K., A. P. Jephcoat, R. J. Hemley, L. W. 
Finger, C. S. Zha, R. M. Hazen, andD. E. Cox, 
Synchrotron x-ray diffraction measurements 
of single crystal hydrogen to 26.5 GPa, Sci- 
ence 239, 1131-1134,1988. 

Raynor, S., Novel ab initio self-consistent-field 
approach to molecular solids under pressure. 
II. Solid H2 under high pressure, /. Chem. 
Phys., 87, 2795-2799, 1987. 



Silvera, I. F., and R. J. Wijngaarden, New low- 
temperature phase of molecular deuterium at 
ultrahigh pressure, Phys. Rev. Lett., 47, 39-42, 
1981. 

Wijngaarden, R. J., V. V. Goldman, and I. F. 
Silvera, Pressure dependence of the optical 
phonon in solid hydrogen and deuterium up to 
230 kbar, Phys. Rev., B 27, 5084-5087, 1983. 



GEOPHYSICAL LABORATORY 



147 



BlOGEOCHEMISTRY 



Nitrogen Isotope Tracers of 
Atmospheric Deposition in Coastal 
Shelf Waters off North Carolina. 

Marilyn L. Fogel and Hans W. Paerl * 

Nitrogen plays a key role in regulating 
marine primary and secondary production 
both on regional and global scales (Ryther 
and Dunstan, 1971; Nixon et al, 1986). 
Nitrogen sources in aquatic ecosystems 
may be external ("new") or internal ("re- 
cycled"). The relative utilization of "new" 
and "recycled" N inputs by phytoplankton 
is important in determining levels of pri- 
mary and secondary production in coastal 
ecosystems. The need for a more detailed 
understanding of the dynamics of new vs. 
recycled production in coastal waters is 
pressing, because previously pristine seg- 
ments of the coastal oceans are now exhib- 
iting both incipient and advanced stages of 
eutrophication (Cosper etaL, 1987; Paerl, 
1988). Terrigenous point and nonpoint 
inputs have traditionally been identified as 
the most likely nutrient sources supporting 
new production in heavily impacted, eutro- 
phic estuaries, including the Chesapeake, 
San Francisco, Delaware, and Narraganset 
Bays (Boynton et al., 1982). The connec- 
tion between man's watershed activities 
and coastal eutrophication, however, ap- 



Institute of Marine Sciences, University of North 
Carolina 



pears more cryptic in many other places 
(e.g., South Atlantic Bight). Recently, ni- 
trogen deposition from rainfall has been 
shown to stimulate primary production in 
coastal waters adjacent to North Carolina 
(Paerl, 1985; Paerl et a/., 1990). 

Atmospheric deposition, as wet and 
dryfall, is an increasingly important source 
of biologically-usable nitrogen in estua- 
rine and coastal regions (Paerl, 1985; 
Legendre and Gosselin, 1989). Large East 
Coast estuaries and certain European seas 
currently receive about 20-50% of their 
combined nitrogen loading from atmo- 
spheric sources (Fisher etaL, 1988; Prado- 
Fiedler, 1990; Loye-Pilot et al, 1990). 
Nitrogen from atmospheric deposition (AD) 
may be a unique source in coastal waters, as 
direct surface water deposition may occur 
downstream of estuarine zones where much 
of the terrigenous nitrogen has been as- 
similated. Proper identification of differ- 
ent N sources and their fluxes are of prime 
concern in understanding and ultimately 
controlling recent eutrophication problems 
in coastal ecosystems. 

We have tested the possibility that AD 
represents both a unique and 
biogeochemically significant source of ni- 
trogen supporting new production in North 
Carolina's coastal Atlantic waters by trac- 
ing the fate of AD products into natural 
phytoplankton assemblages through the use 
of stable N isotope signatures. Nitrogen 
isotopes at the natural abundance level 
have been used extensively in tracing ei- 
ther biochemical processes or sources of 



148 



CARNEGIE INSTITUTION 



food in complex ecosystems (Owens, 1 987). 
A study by Showers et al. (1990) in the 
Neuse River of North Carolina demon- 
strated distinct nitrogen sources on the ba- 
sis of isotopic composition. The isotope 
ratio of the nitrate from sewage treatment 
or point sources was isotopically heavy 
((5 15 N =~12%o). Showers et al (1990) 
were able to distinguish a source of nitrate, 
enriched in 15 N due to intense agricultural 
activity, coming from nonpoint soil runoff 
(<5 15 N =7%o). 

The study site for this investigation 
was Bogue Sound and coastal waters di- 
rectly off Morehead City and Beaufort, 
North Carolina. In this region N sources 
from sewage treatment are almost nonex- 
istent. Therefore, remineralization and ag- 
ricultural runoff provide the primary sources 
of nitrogen to the phytoplankton in this 
ecosystem. We also expected that the <5 15 N 
particulate nitrogen would be influenced 
by recycled nitrogen, as 15 N-enriched am- 
monium (Cifuentes et al. , 1 989), and a 1 5 N- 
enriched nitrate from agricultural runoff, 
providing these were the only two sources 
of nitrogen for primary production. 

Nitrogen in wet or dry deposition is 
considerably more depleted in 15 N relative 
to recycled or agricultural inputs (Heaton, 
1986). Variability in the isotopic composi- 
tion of the nitrogen pools in acid deposition 
may be indicative of certain atmospheric N 
sources. For example, in industrial zones of 

Europe, the <5 15 N of dissolved ammonium 
has a mean value of -12 %c, whereas an 
average value for dissolved nitrate is -3 %c 
(Freyer, 1979). Atmospheric nitrate as 
NO x from industrial and automobile pollu- 
tion should have a 6 15 N near that of air (0 



%c). If, however, the source of NO x is from 
nitrification in soils, then owing to large 
isotope fractionation (Mariotti et al. ,1981), 

the 8 N of dissolved nitrate may be more 
negative. 

The concentration and isotopic compo- 
sition of N in certain rainfall events was 
determined. Collections of atmospheric 
wet and dry deposition were made on the 
roof of the Institute of Marine Sciences, a 
location free of potential sources of con- 
tamination (e.g., trees, powerlines, other 
buildings). Large polypropylene pans hav- 
ing splashproof walls were carefully acid- 
washed (0.01 N HC1), then rinsed three 
times with deionized water in order to 
remove any traces of nutrients. Pans were 
placed in elevated stands on the IMS roof- 
top, prior to precipitation events, to collect 
rainwater. Collectors were deployed just 
prior to, and removed immediately after, 
precipitation events in order to minimize 
contamination. 

Concentrations of dissolved inorganic 
nitrogen species were first analyzed at the 
University of North Carolina on a subsample 
using the methods of Strickland and Par- 
sons (1972). Rainwater samples were 
shipped frozen to the Geophysical Labora- 
tory and thawed just before isotope analy- 
sis. Initial trials with rainwater indicate 
that molecular sieve Zeolite W-85 adsorbed 
the ammonium from the rainwater directly 
without distillation. The <5 15 N of solutions 
containing a known NH + was within the 
standard error of the measurement (± 0.5 
%o) (Velinsky et al, 1989). After NH 4 + 
removal, some aliquots were freeze-dried. 
The residual material, containing the dis- 
solved nitrate and any dissolved organic 



GEOPHYSICAL LABORATORY 



149 




Atlantic Ocean 



Cape 

ookogt 



7 



65 km 



Offshore Sampling Site 



Fig. 83. Map of coastal North Carolina showing three sampling locations: Bogue Sound, Nearshore, and 
Offshore sites. 



150 



CARNEGIE INSTITUTION 



matter, was combusted for isotopic analy- 
sis. 

In bioassays and natural waters, the N 
isotopic ratio of the whole phytoplankton 
sample was measured. Samples were col- 
lected from three locations: a coastal 
nearshore site, Bogue Sound (near Beau- 
fort Inlet), and 90 km offshore (Fig. 83). 
These sites represent N-depleted, full-sa- 
linity Atlantic coastal and mesohaline es- 
tuarine waters. Hydrologically, Bogue 
Sound is a portion of the meso- to euhaline 
component of the Newport River Estuary. 
During incoming tides, however, Bogue 
Sound is a conduit for nearshore-coastal 
Atlantic Ocean water. Chronic N limita- 
tion characterizes both estuarine and coastal 
waters transitting Bogue Sound (Thayer, 
1978; Paerl, 1988). Water samples were 
filtered first through Nitex screening to 
remove zooplankton. Following initial 
screening, remaining phytoplankton were 
collected on Whatman GF/F filters (0.7 m 



Q 

Z 
in 



b- 














j_ 


NH 4 

NOx + DON 




■ 

A 


o- 




V A 








-5- 












10- 












15 i 
( 


i 
) 60 


i 
120 


■ 

180 


■ i • i 

240 30( 


■ i 
) 360 



Julian Days 

Fig. 84. Nitrogen isotopic composition of NH4+ 
and NO x plus dissolved organic N (DON) in 
continental rain events occurring on the North 
Carolina coast over a year's time from 1 January 
to 31 December (Julian Days). Samples were 
collected from April-May 1988 (n=2) and August 
1990- April 1991 (n = 6). Variations in <5 15 N most 
likely reflect differences in the sources of com- 
bined N to atmosphere (e.g., agricultural input vs. 
industrial input). 



nominal pore size). Particulate organic 
matter on glass fiber filters was ground 
with CuO and placed in preheated quartz 
tubes with copper metal. The quartz tubes 
were evacuated and sealed off, combusted 
batchwise at 910° C for 2 h, and cooled at a 
controlled rate. Replicate analysis of fil- 
ters plus organic material gave values with 
standard deviations of ± 0.5 %o. 

The <5 15 N of the NH. + from continental 

4 

rainfalls varied throughout the year (Fig. 
84), but had values considerably more 
negative than either recycled or agricul- 
tural inputs. When nitrogen is incorporated 
into phytoplankton during primary pro- 
duction, the isotopic compositions of algae 
will depend on (1) any biochemical frac- 
tionations that may occur and (2) the iso- 
tope ratios of the available nitrogen sources 
(Cifuentes et al, 1989). Accordingly, the 
<5 15 N of primary producers should shift 
with the addition of nitrogen from AD. In 
fact, when significant rainfall events oc- 
curred, the <5 15 N of particulate nitrogen 
decreased within a few days after the event 
(Fig. 85). 



20 



•z. 



15- 



10- 



5- 



A U 



:i 






t * 



r 



D Offshore 

/k Bogue Sound 

■ Nearshore 



30 60 90 120 150 180 210 240 270 
Fall Winter Spring 

Time (Days) 

Fig. 85. Nitrogen isotopic composition of particu- 
late material sampled from three coastal and estua- 
rine sites off North Carolina as a function of time. 
Day 1 = 1 August 1990. Arrows indicate the 
timing of a continental rainfall event. 



GEOPHYSICAL LABORATORY 



151 



^j" 


bU - 
50- 
40- 
30- 
20- 
10- 
0- 

c 


A 

• 










Mixing depth = 1 .5 m 


z 

Q 

|2 


• 




• 


• 


\ 

[PN] ( ji gN/L) 

V • Am •* • 




) 30 


i 
60 


90 




120 


150 180 210 240 



Time (Days) 



50 



40 



30- 



Q 20 

a 

o 

•" 10 



B 



Mixing Depth= 2 m 



[PN] ( W N/L) 



-i ■ r 



V I tSa fL 



30 60 90 120 150 180 210 240 

Time (Days) 



Mixing Depths 
■ 0.5m 
• 5m 




a 

o 



90 120 150 180 210 240 
Time (Days) 

mm Continental V~Z\ Mixed K/sA Oceanic 

Fig. 86. Total dissolved inorganic nitrogen (DIN) 
from atmospheric deposition that is mixed into 
surface layers of the North Carolina Coast. All 
rainfall events occurred from 1 August 1990 to 
April 1991 (Table 22). (A) Bogue Sound site. (B) 
Nearshore site. (C) Offshore site. Mixing depths 
are approximations estimated from the depth of 
the water column and the photic zone. The line 
indicates average ambient particulate N concen- 
trations measured periodically throughout the time 
period. Levels of DIN were generally < 1 |imole 
N/L. 



Table 22. Dissolved inorganic N (DIN) content of selected significant (more than 0.5 cm) 
rainfall events at the UNC Institute of Marine Sciences, Morehead City, North Carolina.* 



Date 


pH 


Origin [NOx] 


[NH/] 


Amount(cm) 


9 Aug 90 


3.58 


Continental 


659 


1115.5 


4.23 


10 Sep 90 


3.41 


Continental 


2213 


1416.4 


0.51 


24-26 Oct 90 


4.60 


Mixed 


295 


103.4 


3.98 


6 Nov 90 


5.74 


Mixed 


594.5 


128 


0.53 


30 Nov 90 


4.75 


Mostly Oceanic 


216.5 


48.7 


1.85 


4 Dec 90 


5.40 


Oceanic 


40.2 


77.9 


1.43 


9 Dec 90 


4.65 


Mostly Oceanic 


130.6 


44 


2.49 


21 Dec 90 


4.45 


Mixed 


280.4 


74.5 


1.19 


22 Dec 90 


4.81 


Oceanic 


60.1 


9.13 


0.77 


3-4 Jan 91 


4.17 


Mostly Continental 


434.4 


84.8 


0.44 


9 Jan 91 


4.79 


Mostly Oceanic 


75.2 


46.8 


4.66 


12 Jan 91 


4.79 


Mixed 


224 


122 


1.94 


16 Jan 91 


5.10 


Oceanic 


83.5 


27 


2.84 


20 Jan 91 


5.16 


Oceanic 


30.2 


43.9 


3.78 


25 Jan 91 


4.74 


Mixed 


118.8 


28.9 


2.27 


8 Feb 91 


4.89 


Oceanic 


44.6 


ND 


1.72 


5 Mar 91 


5.13 


Mostly Oceanic 


87.2 


76.2 


5.30 


14 Mar 91 


4.36 


Mostly Continental 


376.7 


275.6 


1.56 


30 Mar 91 


5.07 


Mostly Oceanic 


159.9 


187.3 


2.24 



* Weather fronts were documented from satellite imagery and dominant wind direction. 
DIN concentrations are in Lig N/ L. ND indicates non-detectable concentrations. 



152 



CARNEGIE INSTITUTION 



Table 23. Bioassay experiments to assay incorporation of atmospheric deposition into particulate 
nitrogen from coastal samples. 

Type of Date Final Rain Initial Final 5^N Rainwater Origin of Date of 

Bioassay dilution S^N PN PN S^N NH4 4 " last rain last rain 



Cubitainer 
Cubitainer 
Mesocosm 



13-Sep-90 
17-Oct-90 
27-Apr-91 



1% +4.2 

5% +10.8 
1.80% +16.3 



+4.4 ND* Continental 10-Sep-90 

+3.5 -9.5 Mixed 22-Sep-90 

+12.2 -1.5 Mixed 22-Apr-91 

+15.9 ND Oceanic 27-Apr-91 



Initial particulate nitrogen samples were collected from Bogue Sound, not determined. 



Rainfall events in the Beaufort, North 
Carolina, area originate from either conti- 
nental or oceanic sources, or are frequently 
a mixture of the two (Table 22). 

The amount of dissolved inorganic ni- 
trogen (DIN) and pH can usually be related 
to the origin of the rainfall event. Continen- 
tal storms had an average DIN concentra- 
tion of 1643 jig N/L (range = 519-3629 
jigN/L), whereas oceanic events contained 
as little as 150 jug N/L (range = 69-347 
jigN/L). To assess the relative importance 
of AD nitrogen to the ecosystem, the flux of 
N falling on the three coastal sites has been 
calculated for each rainfall event. Depend- 
ing on the mixing depth at each location 
and the amount of rainfall, rain from all 
three origins contributed significant 
amounts of nitrogen, especially at the off- 
shore site (Fig. 86). 

To determine direct uptake of nitrogen 
from AD, short and long-term bioassays 
were designed to examine impacts of AD 
on isotopic composition and primary pro- 
duction under natural irradiance and tem- 
perature conditions. Two independent bio- 
assays were employed. The first used rela- 
tively small volume Cubitainers (4L) incu- 
bated in situ for delineating short-term (1 
day-1 week) impacts (Paerl et a/., 1990). 
The second used previously constructed 



mesocosms (670 L), designed to evaluate 
longer-term chronic loading effects and to 
approximate the estuarine environments 
surrounding the Beaufort-Morehead City 
area. When a rainfall event was antici- 
pated, a control set of mesocosms was 
covered with transparent polyethylene, 
thereby excluding rain water. A second set 
of mesocosms remained uncovered, thereby 
allowing rain input. To a third set of 
mesocosms, rainfall from a previous con- 
tinental event was added. Added rainfall 
simulated dilutions commonly experienced 
in coastal waters (Table 23). 

An aliquot of 14 C-NaHC03 was 
added to each vessel in order to monitor 
photosynthetic 14 C02 assimilation as a 
measure of microalgal production. A par- 
allel 14 C-free set of vessels were deployed 
for stable isotope analyses. Initial samples 
for particulate <5 15 N analyses were taken at 
this time. After 4 days, samples were col- 
lected and filtered for 14 C02 assimilation 
and particulate <5 15 N. 

The first Cubitainer experiment fol- 
lowed directly after a continental rainfall 
that contained a significant amount of DIN 
(Table 23). Although there was some stimu- 
lation of primary production as measured 

with 14 C uptake, no isotopic changes re- 
sulted. In the second experiment, the added 
rainfall was from 9 Aug 1990 water, which 



GEOPHYSICAL LABORATORY 



153 



had a 5 15 N-NH, + of -9.5 %o. Prior to this 

4 

experiment, no significant rainfall had oc- 
curred in the area for 25 days. The addition 
of 5 % rainwater to the Cubitainer caused a 
7 %o decrease in the <5 15 N of the particulate 
N, indicative of the uptake of the isotopi- 
cally-light N from the acid rain. 

The mesocosm experiment was con- 
ducted following a period when rain fall 
events had been primarily of oceanic or 
mixed origin (see Table 22). Rain from a 
mixed event (22 Apr 91) was added to a 1 .8 
% dilution to certain mesocosms, while 
others received oceanic rainfall that oc- 
curred a day later (Table 23). Primary 
production was stimulated with the addi- 
tion of the 22 April rainwater, and 5 N of 
particulate N decreased by 4 %o. In con- 
trast, no stimulation of primary production 
as determined by 14 C uptake was mea- 
sured, relative to controls, with the oceanic 
event, and no change in the 5 N of particu- 
late N was detected. 

Previous results suggest that atmo- 
spheric N inputs, which may supply as 
much as 20 to 50 % of coastal ocean "new" 
nitrogen inputs, represent a significant 
source of N, that contribute to current rates 
of eutrophication. By demonstrating the 
actual incorporation of nitrogen from AD 
into primary producers with N isotope trac- 
ers, we conclude that the biological mea- 
surements of enhanced primary productiv- 
ity are due to the presence of this alterna- 
tive source of a limiting nutrient. Atpresent 
we do not know how primary productivity 
alterations in response to acid rain might 
shape short- or long-term production trends 
of estuarine and coastal ocean food chains. 
This problem is by nature a global one. The 



oxides of nitrogen responsible are largely 
generated in major urban and industrial 
centers located to the north and west of 
North Carolina, although their exact origin 
can only be speculated upon on the basis of 
concentration and location. Fertilizers and 
animal wastes associated with farming may 
also be important sources of NH + in this 
coastal area. Isotope tracers of atmospheric 
deposition may not only be used in tracing 
N into the food chain, but also may have 
some relevance to the origin of the atmo- 
spheric nitrogen. 



References 

Boynton, W. R., W. M. Kemp, and C. W. Keefe, 
A comparative analysis of nutrients and other 
factors influencing estuarine phytoplankton pro- 
duction, in Estuarine Comparisons, V. S. 
Kennedy, ed., pp. 69-90, Academic Press, NY, 
1982. 

Cifuentes, L. A., M. L. Fogel, J. R. Pennock, and 
J. H. Sharp, Biogeochemical factors that influ- 
ence the stable nitrogen isotopic ratio of dis- 
solved ammonium in the Delaware estuary, 
Geochim. Cosmochim. Acta, 53, 2713-2721, 
1989. 

Cosper, E. M, W. C. Dennison, E. J. Carpenter, V. 
M. Bncelj, J. G. Mitchell, S. H. Kuenstner, D. 
Colflesh, and M. Devey, Recurrent and persis- 
tent brown tide blooms perturb coastal marine 
ecosystem, Estuaries, 10, 284-290,. 1987. 

Fisher, D., J. Ceraso, T. Mathew, and M. 
Oppenheimer, Polluted Coastal Waters: The 
Role of Acid Rain. Environmental Defense 
Fund, New York, 1988. 

Freyer, H. D, Seasonal trends of NH 4 + and N0 3 ' 
nitrogen isotope composition in rain collected 
at Julich, Germany, Tellus, 30, 83-92, 1979. 

Heaton, T.H.E., Isotopic studies of nitrogen pollu- 
tion in the hydrosphere and atmosphere: A 
review, Chem. Geol, 59, 87-102, 1986. 

Legendre, L. O. and Gosselin, M. New production 
and export of organic matter to the deep ocean: 
consequences of some recent discoveries, 
Limnol. Oceanogr., 34, 1374-1380, 1989. 

Loye-PiJot, M. D., J.M. Martin, and J. Morelli, 
Atmospheric input of inorganic nitrogen to the 
western Mediterranean, Biogeochem., 9, 1 17- 
134, 1990. 



154 



CARNEGIE INSTITUTION 



Mariotti, A., J. C. Germon, P. Hubert, P. Kaiser, R. 
Letolle, A. Tardieux, and P. Tardieux, Experi- 
mental determination of nitrogen kinetic iso- 
tope fractionation, some principles; illustration 
for the denitrification and nitrification prin- 
ciples, Plant and Soil, 62, 413-430, 1981. 

Nixon, S. W., C. A. Oviatt, J. Frithsen, and B. 
Sullivan, Nutrients and the productivity of es- 
tuarine and coastal marine ecosystems. /. 
Limnol. Soc. So. Afr., 12, 43-71, 1986. 

Owens, N. J. P., Natural variations in 15 N in the 
marine environment, Adv. in Mar. Biol., 24, 
411-451,1987. 

Paerl, H. W., Enhancement of marine primary 
production by nitrogen-enriched acid rain, Na- 
ture, 316, 747-749, 1985. 

Paerl, H. W., Nuisance phytoplankton blooms in 
coastal, estuarine, and inland water, Limnol. 
Oceanogr., 33, 823-847, 1988. 

Paerl, H. W., J. Rudek, and M. A. Mallin, Stimula- 
tion of phytoplankton production in coastal 
waters by natural rainfall inputs: Nutritional 
and trophic implications, Mar. Biol,. 107, 247- 
254, 1990. 

Prado-Fiedler, R., Atmospheric input of inorganic 
nitrogen species to the Kiel Bight, Helgolander 
Meersuut, 44, 21-30, 1990. 

Ryther, J. H., and W. M. Dunstan, Nitrogen, 
phosphorus, and eutrophication in the coastal 
marine environment, Science, 171, 1008-1013, 
1971. 

Showers, W. J., D. M. Eisenstein, H. W. Paerl, and 
J. Rudek, Stable isotope tracers of nitrogen 
sources to the Neuse River, North Carolina, 
Water Resources Institute Report No.253, 1990. 

Strickland, J. D. H. and T. R. Parsons, A Practical 
Handbook of Seawater Analysis. Fish. Res. 
Board Can. Bull. 167, 1972. 

Thayer, G. W., Identity and regulation of nutrients 
limiting phytoplankton production in the shal- 
low estuaries near Beaufort, N.C., Oecologia, 
14, 75-92, 1978. 

Velinsky, D. J., L. A. Cifuentes, J. R. Pennock, J. 
H. Sharp, and M. L. Fogel, Determination of 
the isotopic composition of ammonium-nitro- 
gen at the natural abundance level from estua- 
rine waters, Marine Chemistry, 26, 351-361, 
1989. 



Nitrogen Diagenesis in Anoxic Marine 
Sediments: Isotope Effects 

David J. Velinsky, David J. Burdige* and 
Marilyn L. Fogel 

Due to the complexity of particulate 
nitrogen (PN) transformations, the stable 
isotopic composition of sedimentary nitro- 
gen has received little attention as an indi- 
cator of changes in the nitrogen biogeo- 
chemistry of the oceans. During diagenesis 
nitrogen can change oxidation state (-3 to 
+5) through a series of bacterially - medi- 
ated reactions (Klump and Martens, 1983). 
These transformations have a range of iso- 
tope effects, related mainly to kinetic dif- 
ferences in the reactivity between the light 
( 14 N) and heavy ( 15 N) isotopes (Wada et al. , 
1 975). As a result, diagenetic processes can 
affect the overall isotopic composition of 
dissolved and particulate nitrogen within 
the sediments and overlying waters. For 
example, if the extent of denitrification 
(N0 3 " — > N 2 ) in the water column and 
continental shelf sediments changed over 
time, the # 5 N of oceanic nitrogen could 
shift significantly due to the large isotope 
effect associated with denitrification (e = 
30-40 %c>; Wada et a/., 1975; Cline and 
Kaplan, 1975) and the extent of this process 
as a sink for combined nitrogen in the 
oceans (Christensen, 1987). The source 
and isotopic composition of nitrogen to the 
sediments can change depending on con- 
ditions in surface waters. Rau et al. (1987) 



* Department of Oceanography, Old Dominon 
University, Norfolk, Virginia 



GEOPHYSICAL LABORATORY 



155 



showed that the # 5 N of PN in organic 
carbon-rich Cretaceous marine sediments 
were isotopically light (-4.0 to +1.0 %6) 
compared to typical marine sediments (gen- 
erally >4 %6). They speculated that N 2 fixa- 
tion was the dominant source of organic 
nitrogen to these Cretaceous sediments. 

The type and isotopic composition of 
particulate nitrogen that is incorporated 
into the sediments depends on source varia- 
tions and isotope effects during diagenesis. 
For source variations to be evaluated, it is 
important to determine if any post-deposi- 
tional isotope effect occurs during early 
diagenesis. The purpose of this paper is to 
investigate possible isotope effects during 
diagenesis of PN in anoxic marine sedi- 
ments. Once diagenetic effects are deter- 
mined, a more accurate interpretation of 
nitrogen isotopes in ancient and present^ 
day marine sediments can be obtained. 




Framvaren Fjord 



Great Marsh 



▲ A 

Leaves, Framvaren Fjord 



-30 



-25 



-20 



-15 



5 13 C 



Fig. 87. The <5 13 C of organic carbon and <5 15 N of 
particulate nitrogen from sediments taken from 
the Chesapeake Bay, Framvaren Fjord and Great 
Marsh. Also included are isotope values from 
various plants around the Framvaren Fjord. 



Methods and Study Areas 

Concentrations and <5 15 N of pore water 
NH4 + and sedimentary PN were deter- 
mined from cores taken in three contrasting 
coastal marine environments: Framvaren 
Fjord, Norway (FF; June 1989), Great 
Marsh, Delaware (GM; June, 1988), and 
Chesapeake Bay (CB; June, 1988). 

The Framvaren Fjord is a permanently 
anoxic fjord with concentrations of dis- 
solved NH4 + and hydrogen sulfide (H2S) 
in the bottom waters (maximum depth 180 
m) of 2 mM NH4+ and 6 mM H2S, respec- 
tively. FF sediments are permanently 
anoxic. The sources of organic matter to the 
sediments are primarily from bacterial and 
phytoplankton production in the water col- 
umn along with some terrestrial inputs (e.g., 
leaves). 

The Great Marsh is a coastal salt marsh 
dominated by the short form of Spartina 
alterniflora. Marsh sediments in the upper 
12 cm (i.e., the active root zone) cycle 
seasonally betwenn oxidized and reducing 
conditions (Velinsky and Cutter, 1991). 
Sediments below -12 cm are permanently 
anoxic. Sources of organic matter to these 
sediments include Spartina production 
and upland runoff. 

Two sediment cores were obtained from 
the Chesapeake Bay, near 38°56' N, 
76°23'W, just south of the Chesapeake Bay 
Bridge near Annapolis, Maryland. The 
water column in this section of the bay 
(maximum depth 30 m) is seasonally anoxic 
or sub-oxic whereas the sediments are gen- 
erally permanently anoxic (San Diego- 
McGlone, 1991). Sources of organic mat- 
ter to the sediments of this portion of the 



156 



CARNEGIE INSTITUTION 








10 - 



20 - 



E 



30 - 



40 - 



50 



%PN and NH+ (mM) 
1.0 2.0 



5 15 N 



3.0 







8 



10 



1 1 1 1 1 1 1— 


—1 • 1 1 


C/N , 




- 


I \ 


\ nh; ■ 


PN A 




\ 


r 






i 1 


_i i 




10 15 

C/N (atomic) 



20 



Fig. 88. The depth distributions of particulate nitrogen (PN), dissolved ammonium (NH4 + ) and the 
organic carbon to particulate nitrogen ratio (C/N) along with the isotopic composition (5 15 N) of 
dissolved NH4 + and PN in Framvaren Fjord sediments. 



bay are primarily derived from phy toplank- 
ton production and to a lesser extent from 
land runoff. 

Box (CB) or gravity (FF and GM) cores 
were sectioned at specific intervals. The 
CB core was sectioned every 2 cm, and the 
FF core was sectioned at 5-cm intervals. 
The GM core was sectioned every 2.5 cm, 
and due to the small volume of pore fluids 
obtained and the low concentrations of 
dissolved NH4+ in the upper 15 cm, the 2.5- 
7.5 cm and the 10-15 cm sections were 



combined for isotope analysis. Pore fluids 
were separated from sediments by either 
centrif ugation (FF) or with Reeburgh ( 1 967 ) 
sediment squeezers (CB and GM), then 
filtered through Nuclepore 0.4 jiim filters. 
Both sediment and pore waters were stored 
frozen until sample preparation and analy- 
sis. 

The methods for the preparation and 
determination of the <5 15 N-NH4 + and <5 15 N- 
PN are described in Velinsky et al. (1989) 
and Cifuentes et al. (1988). The data are 



GEOPHYSICAL LABORATORY 

Table 24. Isotope discrimination between PN and NH4 + . 



157 



Location 


S 15 PN* 




Reference 


Frarnvaren Fjord 


3.3±0.9 


-0.210.8 


This Study 


Chesapeake Bay 

Core 41 
Core 42 


9.4±0.5 
9.8±1.1 


- 2.8 to - 0.7 
2.3 to - 0.4 


This Study 
This Study 


GreatMarsh,DE. 


5.2±0.8 


- 0.810.6 


This Study 


Santa Barabara Basin 

SOG 005 
GAS 24 
GAS25 
GAS 29 


7.111.2 
7.010.9 
7.310.3 
6.110.9 


+ 2.611.5 
+ 3.211.0 
+ 4.110.6 
+ 5.011.7 


Sweeney and Kaplan (1980) 



* Average ^l^PN 

**Average A = #1 5 NH4 + - 5 15 PN 



%PN and NH + (mM) 
1.0 2.0 3.0 





u 


* 


> p [ 


— I ' 1 r " 






C/N , 








10 


>v I / 


- 


E 


20 




\ NH 4 " 


CD 

Q 


30 




1 






PN 6 


► > 




40 


) ' 








[ 


3 • 






50 


i 


« 




10 15 20 

C/N (atomic) 

Fig. 89. Similar to Fig. 88, except for the Great Marsh. 



158 



CARNEGIE INSTITUTION 



reported in the standard S notation {i.e., 
<5 15 N = [(#sample/#standard) - 1] x 10 3 ; where 
R = 15n/ 14 N} and the ratios are reported 
against air (<5 15 N = 0). Precision of repli- 
cate samples for ammonium and PN isoto- 
pic analysis is approximately ± 0.5 %o and 
± 0.2 %o, respectively. Ammonium concen- 
trations were determined with the proce- 
dures described in Sharp et al. (1982) and 
solid phase PN concentrations were deter- 
mined with a Carlo-Erba ANA 1500 car- 
bon and nitrogen analyzer. 



Results and Discussion 

To illustrate the differences between 
sedimentary environments, the isotopic 
compositions of organic carbon (<5 13 C) and 
particulate nitrogen (<5 15 N) are plotted (Fig. 
87). Because of the predominance of 
Spartina aiterniflora (a C4 plant) in the 
GM, the <5 13 C of the sediments are greater 
than those in either the FF and CB sedi- 
ments. The carbon and nitrogen composi- 
tion of FF sediments reflect terrestrial and 



%PN 
0.3 0.4 



0.5 




15.0 




NH 4 (mM) 



Fig. 90a. Similar to Fig. 88, except for Chesapeake Bay, Core 41. 



GEOPHYSICAL LABORATORY 



159 



E 
o 



Q. 
CD 

Q 




5.0 



7.5 



8 15 N 

10.0 



12.5 



15.0 




NH^(mM) 



Fig. 90b. Similar to Fig. 90a, except for Chesapeake Bay, Core 42. 



water column-derived inputs, whereas the 
CB sediments are isotopically enriched in 
15 N and 13 C, indicating typical estuarine 
phytoplankton. 

In the FF sediments, the <5 15 N of both 
pore water NH4+ and sedimentary PN did 
not change significantly with depth (Fig. 
88). As such, the isotopic difference (A = 
5!5N-NH4 + - <5 15 N-PN) between the NH 4 + 
and PN is small (~ -0.2%o) with the NH4+ 
being 15n depleted (Table 24). These data 
indicate no expression of significant iso- 
tope effect during ammonification of PN in 
the FF sediments. 



The concentrations of PN decreased 
significantly (60%) in the GM sediments 
(Fig. 89) from the surface to approximately 
27 cm. Concurrently, the <5 15 N-PN in- 
creased very slightly in this interval with an 
overall average of 5.66 ± 0.46 %o (n = 8). 
Whereas pore water NH4+ concentrations 
increased with depth, the <5 15 N-NH4+ did 
not change and averaged 4.30 ± 0.41 %o 
(n=l). Similar to FF sediments, the <5 15 N- 
NH4+ was lighter than the <5 15 N-PN, indi- 
cating a small negative discrimination with 
depth (i.e., 1 4 N-NH4 + is preferentially re- 
leased compared to 15 N-NH4 + ; Table 24). 



160 



CARNEGIE INSTITUTION 



This discrimination is a result of a number 
of processes that could fractionate NH4+, 
including ion-exchange, deamination of 
amino acids (Macko and Estep, 1984), up- 
take by bacteria within the sediments and 
diffusion out of the sediments. The ob- 
served distribution of <5 15 N-NH4 + with 
depth most likely results from a combina- 
tion of these processes. 

The Chesapeake Bay cores exhibited a 
slightly different trend in <5 15 N-NH4+ with 
depth (Fig. 90a). In Core 41 , the concentra- 
tion of PN decreased with depth in the 



upper 10 cm and remains constant below 
10 cm, while the <5 15 N-PN did not change 
in the upper 10 cm. Conversely, as the 
concentration of NH4 4 " increased with 
depth, the <5 15 N-NH4+ also increased. The 
maximum increase in <5 15 N-NH4+ was in 
the same depth intervals as the decrease in 
PN and increase in dissolved NH4 + . The 
differences between Core 41 and 42 (Fig. 
90b) reflect the spatial heterogeneity of 
sediments in the bay. A sharp maximum in 
pore water NH4+ and <5 15 N occurs in the 4- 
6 cm interval and this was related to an 



PN 



PN 



R 



Nitrogen Cycle in Sediments 



NH 



NO, 



Water 



uuj 



2UUUUU 



i«p 



PN, 



PN 



R 



Burial 




K-2.B 



Sediment 



4 




no 3 


K 


N 
2 


k, 



ki 


flux of PN to the sediments 


k 4 


denitrification 


k 2 


am monification 


k 5 


nitrate diffusion 


k -2 


microbial uptake 


k 6 


ammonium diffusion 


k 3 


nitrification 


k 7 


burial 



Fig. 90. A conceptual model for the pathways of PN mineralization in marine sediments. Note that the 
PN is broken into two fractions: a labile phase (PNl) and a refractory phase (PNr). With depth and burial, 
the PNl fraction would decrease as would the rate of remineralization (k2; Burdige, 1991). 



GEOPHYSICAL LABORATORY 



161 



extensive shell layer found at this depth. 
This shell layer contained additional or- 
ganic material that enhanced sulfate reduc- 
tion causing the pore water sulfate concen- 
trations to decrease to undetectable levels 
at 4-6 cm versus 12-14 cm for Core 41. As 
a result, pore water concentrations of dis- 
solved NH4+ increased dramatically from 
the surface (1.3 mM) to 4-6 cm (6 mM). 
Below the NH4+ maximum, concentra- 
tions decreased to near constant levels, 
whereas the concentration of PN decreased 
to a minimum at approximately 1 8 cm (Fig. 
90b). The S^N of both NH 4 + and PN 
reflected the change in source of PN with 
distinctly different isotopic ratios of nitro- 
gen in the 4-6 cm interval. 

The absence of any significant isotope 
effect during PN remineralization may re- 
flect the variety of processes that are affect- 
ing PN in sediments. Each process could 
have an intrinsic isotope effect that is fully 
expressed in the overall solid phase PN 
distribution. Alternately, the lack of any 
observable shift in the <5 15 N-PN with depth 
may be due to a "mass" effect, in that only 
a small fraction of the PN is remineralized 
compared to the total PN. As such it would 
be difficult to see any discrimination unless 
it is very large. Pore water NH4+ should be 
a better indicator of mineralized material 
and, as such, should be a more accurate 
reflection of any diagenetic isotope effect 
on the PN. 

The distributions of A are similar for the 
Chesapeake Bay cores. In both cores, A in 
the surface section was approximately -2.5 
%o(i.e., the NH4 + is isotopically lighter than 
the PN) and increased with depth. This 
trend indicates a possible selective 



remineralization in the upper sediments of 
a more labile fraction of the PN with a 
lighter isotopic composition (PNl; Fig. 
91). This fraction of PN could represent 
"fresh" phytoplankton material that has 
reached the bottom sediments relatively 
unaltered. Montoya et al. (1990) showed 
that particulate material from the mainstem 
of Chesapeake Bay during the spring had 
nitrogen isotopic compositions of between 
6.2 and 10.5 %o. If this scenario is the case, 
the preferential degradation of this more 
labile organic matter (Westrich and Berner, 
1984; Burdige, 1991) would release dis- 
solved NH4+ into the pore waters with 
<5 15 N similar to that of the source material 
(Fig. 91). With depth less of this lighter 
material remains and eventually the <5l 5 N- 
NH4 + of the pore waters would reflect 
remineralization of the "background", more 
refractory, particulate nitrogen (PNr). Such 
results were shown by Sweeney and Kaplan 
(1980) for sediments taken from the Santa 
Barbara Basin (Table 24). They demon- 
strated that the <5 15 N of the pore water 
NH4 + reflected the degradation of a marine 
source of PN with a <5 i5 N of approximately 
10 %o. The range of <5 i5 N for total nitrogen 
in these sediments was thought to be de- 
rived from a mixture of a marine (i.e., 
phytoplankton at 10 %o) and terrestrial 
sources (i.e., sewage-derived at 2 %c). Dis- 
solved NH4+ in the pore waters was there- 
fore postulated to be derived mainly from 
the preferential degradation of the plank- 
tonic organic nitrogen. Although the 
data from Chesapeake Bay reflects the pos- 
sible degradation of a labile fraction of PN, 
the data from the Framvaren Fjord and 
Great Marsh did not reflect this distribu- 



162 



CARNEGIE INSTITUTION 



tion (Figs. 88 and 89; Table 24). It is pos- 
sible that in the Framvaren Fjord the major- 
ity of the degradation of PN occurred in the 
water column above the sediments. By the 
time this material reached the sediments 
any labile phase was already degraded. In 
the GM sediments, labile N could be re- 
leased in the upper 10-12 cm and either 
consumed by the Spartina or possibly 
released via diffusion to the adjacent creek 
waters 

In conclusion, only a small isotopic 
discrimination is expressed during diagen- 
esis (Table 24). Whereas, the bulk S l5 N- 
PN did not significantly change with depth, 
it appears that selective remineralization of 
a labile fraction of N may occur in certain 
environments. This observation indicates 
the <5 15 N of specific fractions of the PN 
could be used as a tracer of recently formed 
organic material. Downcore variations in 
<5 15 N of solid-phase nitrogen probably re- 
flect the material deposited to the sediment 
surface. It is not possible to tell however, if 
the isotopic composition of the PN formed 
in the water column is reaching the sedi- 
ment intact, only that the isotopic integrity 
of the bulk PN appears to remain unaf- 
fected. 



References 

Burdige, D. J., The kinetics of organic matter 
mineralization in anoxic marine sediments, 
Jour. Mar. Res., in press, 1991. 

Christensen, J. P., J. W. Murray, A. H. Devol, and 
L. A. Codispoti, Denitrification in continental 
shelf sediments has major impact on the oce- 
anic nitrogen budget, Global Biogeochem. 
Cycles, 1(2), 97-116, 1987. 

Cifuentes, L.A., J. H. Sharp, and M. L. Fogel, 
Stable carbon and nitrogen isotope biogeo- 



chemistry in the Delaware estuary, Limnol. 
Oceangr.,33, 1102-1115, 1988. 

Cline, J. D. and I. R. Kaplan, Isotopic fractionation 
of dissolved nitrate during denitrification in 
the eastern Tropical North Pacific, Mar. Chem., 
5,271-299, 1975. 

Klump, J. V., and C. S. Martens, Benthic nitrogen 
regeneration, in Nitrogen in the Marine Envi- 
ronment, Chapter 12, EJ. Carpenter and D.G. 
Capone, eds., Academic Press, New York, pp. 
411-458,1983. 

Macko, S. A., and M. L. F. Estep, Microbial 
alteration of stable nitrogen and carbon isoto- 
pic compositions of organic matter, Org. 
Geochem., 6, 787-790, 1984. 

Montoya, J. P., S. G. Horrigan, and J. J. McCarthy, 
Natural abundance of 15 N in particulate nitro- 
gen and zooplankton in the Chesapeake Bay, 
Mar. Ecol. Prog. Series, 65, 35-61, 1990. 

Rau, G. H., M. A. Arthur, and W. E. Dean, 15 N/ 14 N 
variations in Cretaceous Atlantic sedimentary 
sequences: implication for past changes in 
marine nitrogen biogeochemistry, Earth 
Planet. Sci. Let., 82, 269-279, 1987. 

Reeburgh, W. S., An improved interstitial water 
sampler, Limnol. Oceanogr., 12, 223-234, 
1967. 

San Diego-McGlone, M. L., Processes affecting 
the behavior of redox sensitive elements in 
Chesapeake Bay, Ph.D dissertation, Old Do- 
minion University, 1991. 

Sharp, J. H., C. H. Culberson, and T. M. Church, 
The chemistry of the Delaware estuary. Gen- 
eral Considerations, Limnol. Oceanogr., 27, 
1015-1028, 1982. 

Sweeney, R. E. and I. R. Kaplan, Natural abun- 
dances of 15 N as a source indicator for near- 
shore marine sedimentary and dissolved nitro- 
gen, Mar. Chem., 9, 81-94, 1980. 

Velinsky, D. J. and G. A. Cutter, Geochemistry of 
selenium in a coastal salt marsh, Geochim. 
Cosmochim.Acta,55, 179-191, 1991. 

Velinsky, D. J., J. R. Pennock, J. H. Sharp, L. A. 
Cifuentes and M. L. Fogel, Determination of 
the isotopic composition of ammonium- nitro- 
gen at the natural abundance level from estua- 
rine waters, Mar. Chem., 26, 351-361, 1989. 

Wada, E., T. Kadonaga and S. Matsuo, 15 N abun- 
dance in nitrogen of naturally occurring sub- 
stances and global assessment of denitrifica- 
tion from isotopic viewpoint, Geochem. Jour., 
9, 139-145, 1975. 

Westrich, J. T., and R. A. Berner, The role of 
sedimentary organic matter in bacterial sulfate 
reduction: The G model tested, Limnol. 
Oceangr., 29(2), 236-249, 1984. 



GEOPHYSICAL LABORATORY 



163 



The Isotopic Ecology Of Plants And 

Animals In Amboseli National Park, 

Kenya 

Paul L. Koch, Anna K. Behrensmeyer* 
and Marilyn L. Fogel 

Variations in the stable isotope ratios of 
carbon ("C/ 1 ^), nitrogen ("N/^N), and 
oxygen ("O/^O) provide information about 
the ecology, physiology, and habitats of 
living and extinct animals. For example, 
the ^ 3 C values of an animal's tissues are 
controlled by the isotopic composition of 
its diet, which, for herbivores, is related to 
the photo synthetic pathway of food plants 
(DeNiro and Epstein, 1978; Vogel, 1978). 
Although affected by dietary # 5 N values, 
N isotopes in animals vary with rainfall 
amounts among ecosystems and among 
trophic levels in an ecosystem (DeNiro and 
Epstein, 1981; Heaton et ai, 1986). 

We are investigating the isotopic 
ecology of plants and animals in Amboseli 
National Park, Kenya, for several reasons. 
First, investigations of floral and faunal 
isotopic composition in terrestrial ecosys- 
tems are uncommon (e.g., Ambrose and 
DeNiro, 1986; Sealy era/., 1987),andnone 
evaluates C, N, and O isotopes simulta- 
neously. Ecosystem studies test the gen- 
erality of relationships determined either in 
the laboratory or in comparisons of indi- 
viduals from different regions. Second, 
stable and radiogenic isotopes have been 
employed to identify sources of elephant 

*Department of Paleobiology, National Museum 
of Natural History, Smithonian Institution, Wash- 
ington, DC 



ivory and rhinoceros horn, in order to con- 
trol sales of poached versus legally hunted 
animals (van der Merwe etal., 1990; Vogel 
et ai, 1990). Ivory from various African 
parks can be distinguished by its N, C, and 
either Sr or Pb isotopic composition. How- 
ever, if the isotopic composition of el- 
ephants varies with time, because of habi- 
tat, diet, or climate change, isotopic identi- 
fication of source region may be unreli- 
able. Either the isotopic composition of a 
species must be constant through time 
within an ecosystem, or the secular trends 
must be minor when compared to differ- 
ences between populations. Finally, isoto- 
pic patterns in modern ecosystems can serve 
as analogs for interpretation of the fossil 
record. African faunas, with their diversity 
of large mammals, are excellent analogs of 
typical faunas before the Pleistocene ex- 
tinction. 



Study Area, Materials, and Methods 

Amboseli Park is located in southern 
Kenya (20°40'S,37°15'E; mean elevation, 
1140 m). Annually, temperature averages 
23°C and ranges from 15° to 31°C. Rain 
falls in two seasons and averages 300 cm/ 
year. However, the park is continuously 
supplied with spring water fed by melting 
snow on Mt. Kilimanjoro. Habitats in the 
park include grasslands, bushlands, 
swamps, seasonal lakes, and woodlands. 
Woodlands have retreated since the early 
1970s, perhaps due to overbrowsing by 
elephants or increased soil salinity. Tree 
loss has altered the abundances of herbi- 



164 



CARNEGIE INSTITUTION 



vores; there are more grazers (animals that 
eat grass) and fewer browsers (animals 
eating herbaceous and woody plants) and 
mixed feeders (D. Western, pers. comm.). 

Plant samples (mixtures of leaves and 
stems) were collected in September 1 990 at 
eight localities in the woodland, swamp, 
swamp edge, plains, and bush habitats. 
Faunal samples (tooth dentin or bone) were 
collected from carcasses throughout the 
park. Samples were collected from 1975 
through 1990, and were in different states 
of weathering. The minimum number of 
years since death can be estimated from 
weathering stage (Behrensmeyer, 1978). 
For carcasses in advanced weathering 
stages, however, determining actual time 
since death is difficult. 

Plants were air dried in the field, freeze- 
dried in the laboratory, and then lightly 
crushed. Bones and teeth were demineral- 
ized with EDTA or 0.1 N HC1 to isolate 
collagen (Tuross et ai, 1988), and then 
treated with chloroform/methanol solution 
to remove lipids. Plant and collagen samples 
were placed in preheated quartz tubes with 
CuO and Cu metal. Tubes were evacuated, 
sealed, combusted at 910°C for 2 h, then 
cooled at a controlled rate. Standard devia- 
tions for analysis of standards were ±0.2%c 
for #3C and S^N. 



I so topic Variation in Amboseli Plants 

The plants of Amboseli segregate into 
two populations isotopically (Table 25, Fig. 

92A). The 5 C values of grasses, which 
use C 4 photosynthesis, have a mean value 



iu ■ 


A 


E 


aquatic 




E 


succulent 


8 - 


ID 


shrub & tree 


w 






□ 


herb 


3 






■ 


grass 


.y e- 




- 


- 


"O 












c 












o 4- 








- r 




o 




„ 




n 


n 


2 - 




i 


_ 


. 


- X 


n - 




WW 


n 


"n 



6 13 C 



10 



8 
tn 
aJ 

| 6 

C 

o 4 

d 

z 



2- 



1 



^ :, 




□ elephant 

□ mixed feeder 

□ browser 
■ grazer 

Q carnivore 



fiii 



^ 00 O OJ ^t CD 



6 13 C 



Fig. 92. (A) Histogram of carbon isotope compo- 
sitions for Amboseli plants subdivided according 
to physiogamy. (B) Histogram of carbon isotope 
compositions for Amboseli mammals subdivided 
according to feeding type. 



± one standard deviation of -13.2 ± 0.9 %o. 
The herbaceous plants and woody plants 

employ CL photosynthesis and have mean 
values of -27.1 ± 1.9 %o and -27.7 ± 2.4 V 
respectively. This isotopic segregation 

between CL shrubs, trees and herbs and C 4 
grasses is expected in a hot, dry region 
(Tieszen and Boutton, 1988). Succulent 
herbaceous and woody plants exhibit a 



GEOPHYSICAL LABORATORY 



165 



Table 25. Isotopic data for Amboseli plants collected in 1990 



Species 


Family 


Habitat 


8^N 


5l3 C 




Grasses and Sedges 








Cynodon dactylon 


Graminae (m) 


swamp 


10.6 


-13.0 


Sporobolus consimilis 


Graminae (m) 


swamp edge 


8.9 


-13.4 


Sporobolus spicatus 


Graminae (m) 


swamp edge 


8.6 


-13.2 


Sporobolus kentrophyllum 


Graminae (m) 


swamp edge 


10.1 


-13.4 


Cynodon plectostachys 


Graminae (m) 


woodland 


9.4 


-14.2 


Sporobolus helvolus 


Graminae (m) 


bush 


8.8 


-15.1 


Sporobolus ioclades 


Graminae (m) 


bush 


11.9 


-13.0 


Chloris roxburghiana 


Graminae (m) 


bush 


8.7 


-13.4 


Chloris virgata 


Graminae (m) 


bush 


10.6 


-13.2 


Enneapogon cenchroides 


Gramiriae (m) 


bush 


9.4 


-13.3 


Cyperus immensus 


Cyperaceae (m) 


swamp 


4.4 


-11.2 


Cyperus laevigatus 


Cyperaceae (m) 
Submerged Aquatic Plants 


swamp 


7.8 


-12.2 


Ceratophyllum sp. 1 


Ceratophyllaceae 


swamp 


6.4 


-23.0 


Ceratophyllum sp. 2 


Ceratophyllaceae 
Herbs 


swamp 


9.3 


-19.7 


Pistia stratiotes 


Araceae (m) 


swamp 


11.5 


-29.0 


Solanum incanum 


Solanaceae 


swamp edge 


10.4 


-25.0 


Justicia odora 


Acanthaceae 


woodland 


13.3 


-27.6 


Diplictera albicauda 


n.d. 


woodland 


8.0 


-27.8 


Abutilon mauritanium 


Malvaceae 


plains 


11.0 


-29.7 


Pluchea ovalis 


Asteraceae 


plains 


8.1 


-30.5 


Cissampelos mucronata 


Menispermaceae 


plains 


7.5 


-27.7 


Commicarpus sp. 


Nyctaginaceae 


plains 


8.6 


-28.9 


Achyranthes aspera 


Amaranthaceae 


plains 


11.4 


-28.4 


Withania somnifera 


Solanaceae 


plains 


9.4 


-25.2 


Indigofera sp. 


Leguminosae 


bush 


10.4 


-25.0 


Duosperma eremophiloum 


Acanthaceae 


bush 


8.2 


-27.1 


Barleria spinisepala 


Acanthaceae 


bush 


11.2 


,24.3 




Shrubs. Trees, and Succulent Plants 






Trianthema ceratosepala 


Aizoaceae 


bush 


12.8 


-21.8 


Sansevieria sp. 


Agavaceae (m) 


bush 


12.6 


-14.5 


Euphorbia sp. 1 


Euphorbiaceae 


bush 


13.3 


-14.9 


Euphorbia sp. 2 


Euphorbiaceae 


bush 


13.9 


-14.0 


Sueda monoica 


Chenopodiaceae 


swamp edge 


13.6 


-13.3 



Continued on next page 



166 



CARNEGIE INSTITUTION 



Table 25. Continued 



Species 



Family 



Habitat 5 15 N <5 13 C 



Salvador a persica 
Maerua triphyllum 
Commiphora sp. 
Boscia angustifolia 
Acacia sp. 
Azima tetracantha 
Balanites glabra 
Phoenix reclinata 
Acacia tortilis 



Shrubs and Trees 








Salvadoraceae 


swamp edge 


5.3 


-25.4 


Capparidaceae 


bush 


8.8 


-29.3 


Burseraceae 


bush 


15.4 


-29.3 


Capparidaceae 


bush 


11.2 


-24.6 


Leguminosae 


bush 


10.2 


-23.7 


Salvadoraceae 


woodland 


7.6 


-26.8 


Balanitaceae 


woodland 


7.7 


-29.9 


Palmae (m) 


woodland 


7.1 


-29.4 


Leguminosae 


woodland 


8.9 


-29.7 



(m) indicates monocotyledons, all other plants are dicotyledons 



range of 8 C values and may use either C. 

or Crassulacean acid metabolism. Finally, 

i ^ 
submerged aquatic plants have 5 C values 

that can range between -12 and -33 %o, 
depending on the pathway of carbon up- 
take. Unlike terrestrial plants, which di- 
rectly incorporate atmospheric CCL, sub- 
merged plants may accumulate either dis- 
solved C0 2 or HC0 3 " (Raven, 1987). 
Amboseli aquatic plants have # 3 C values 
intermediate between C 3 and C. plants. 

The <5 15 N values of Amboseli plants 
form a unimodal distribution with a mean 
of 9.8 ± 2.4 %o (Table 25, Fig. 93 A). This 
mean value is slightly higher than that 
reported by Sealy et al. (1987) for plants 
from a region receiving 300 mm of rain. 

Plant <5 15 N values are not dependent on 
either location or habitat type, although 
plants from the bush habitat may be slightly 

15 N-enriched. Plant <5 15 N values are not 
influenced by physiogamy or photosyn- 
thetic pathway, with one exception. All 

Amboseli succulents are 15 N-enriched (1 3.2 



± 0.5 %c). These species are only distantly 
related to each other, and the cause of 15 N 
enrichment is unclear. 



Isotopic Variation in Amboseli Mammals 

The C isotope difference between C3 
and C4 plants provides a tool for tracing the 
diets of Amboseli mammals. Previous field 
studies demonstrate a consistent difference 
in <5 13 C values between diet and collagen 
of ~+5 %o (Vogel 1978; van der Merwe, 
1989). Amboseli mammals with pure graz- 
ing diets should have <5 13 C values of -8 to 
-9 %o. All Amboseli grazers (buffalo, spring 
hare, warthog, wildebeest, zebra) have col- 
lagen <5 13 C values in this range (Table 26, 
Fig. 92B). 

In contrast, Amboseli animals with pure 
browsing diets should have collagen <5 13 C 
values of -22 to -23 %o. None of the 
browsers (rninoceros, giraffe) have values 
this low, indicating a small but persistent 
fraction of grasses or succulents in their 



GEOPHYSICAL LABORATORY 



167 



Table 26. Isotopic data for Amboseli mammals, excluding elephants. 



Secies 


Common Name 


Year of 
death 


<5 15 N 


<5l3 C 




Carnivores. Insectivores. and Omn 


ivores 






Crocuta crocuta 


Spotted hyena 


*1984 


16.7 


-9.2 


Panthera leo 


Lion 


n.d. 


17.1 


-10.1 


Panthera leo 


Lion 


n.d. 


15.6 


-6.4 


A cinonyx jubatus 


Cheetah 


*1984 


17.2 


-15.7 


Canis adustus 


Jackal 


*1989 


15.4 


-11.2 


Canis adustus 


Jackal 


*1988 


13.8 


-9.9 


Canis adustus 


Jackal 


*1988 


14.9 


-14.8 


Otocyon megalotis 


Bat-eared fox 


*1971 


12.2 


-10.6 


Otocyon megalotis 


Bat-eared fox 


n.d. 


14.6 


-12.0 


Otocyon megalotis 


Bat-eared fox 


*1974 


14.2 


-15.7 


Ichneumia albicauda 


White-tailed mongoose 


n.d. 


17.4 


-8.3 


Orycteropus afer 


Aardvark 


n.d. 


9.7 


-12.1 


Papio cynocephalus 


Yellow baboon 

Grazers 


1989 


10.8 


-14.3 


Pedetes capensis 


Spring hare 


*1975 


9.8 


-9.0 


Equus burchelli 


Burchell's zebra 


1974 


9.8 


-8.6 


Equus burchelli 


Burchell's zebra 


*1984 


9.1 


-8.6 


Equus burchelli 


Burchell's zebra 


1990 


10.0 


-8.5 


Connochaetes taurinus 


White-bearded wildebeest 


1975 


12.4 


-8.0 


Connochaetes taurinus 


White-bearded wildebeest 


1974 


11.0 


-8.7 


Connochaetes taurinus 


White-bearded wildebeest 


*1988 


13.8 


-7.8 


Connochaetes taurinus 


White-bearded wildebeest 


*1989 


11.8 


-9.0 


Syncerus coffer 


Buffalo 


1968 


10.1 


-7.9 


Syncerus caffer 


Buffalo 


*1984 


10.8 


-8.0 


Phacochoerus aethiopicus 


Warthog 


1990 


10.8 


-8.8 


Phacochoerus aethiopicus 


Warthog 

Browsers 


*1989 


11.0 


-9.2 


Dicer os bicornis 


Black rhinoceros 


1961 


6.4 


-19.8 


Dicer os bicornis 


Black rhinoceros 


1974 


8.2 


-18.7 


Diceros bicornis 


Black rhinoceros 


*1984 


7.9 


-19.0 


Giraffa camelopardalis 


Giraffe 


*1989 


11.5 


-20.3 


Giraffa camelopardalis 


Giraffe 

Mixed Feeders 


*1986 


11.6 


-19.7 


Hystrix cristata 


Porcupine 


*1973 


9.9 


-15.6 


Hippopotamus amphibius 


Hippopotamus 


*1973 


8.9 


-10.0 


Hippopotamus amphibius 


Hippopotamus 


1990 


13.1 


-8.6 


Hippopotamus amphibius 


Hippopotamus 


*1989 


8.9 


-10.9 


Aepyceros melampus 


Impala 


1985 


12.4 


-14.0 


Aepyceros melampus 


Impala 


*1986 


11.7 


-13.6 


Aepyceros melampus 


Impala 


*1988 


13.1 


-15.8 


Gazella granti 


Grant's gazelle 


*1988 


10.5 


-16.0 


Gazella granti 


Grant's gazelle 


*1989 


10.2 


-15.6 


Gazella thomsoni 


Thomson's gazelle 


1990 


14.2 


-10.7 


Gazella thomsoni 


Thomson's gazelle 


*1986 


10.8 


-17.6 



* determined by weathering stage 



168 



20 



w 15 

CO 

■g 

"> 

■o 

c 




10 11 12 13 14 15 16 17 18 



8 15 N 



10 - 



5- 



□ elephant 

□ mixed feeder 

□ browser 
■ grazer 
H carnivore 



J=L 



3 4 5 6 7 




10 11 12 13 14 15 16 17 18 



CARNEGIE INSTITUTION 

Table 27. Isotopic data for Amboseli elephants. 



Secimen 



Year of 
death 



<5 15 N «5 13 C 



African Elephant: Loxodonta africana 



C75-3 


1974 


12.0 


-18.3 


C75-6 


*1973 


9.4 


-17.9 


E-l ) 


late 70s 


10.4 


-13.3 


E-3 ] 


late 70s 


10.3 


-13.9 


E-8 ] 


late 70s 


9.7 


-17.3 


E-ll ] 


late 70s 


10.7 


-13.2 


E-13 ] 


late 70s 


11.4 


-13.9 


E-15 ] 


late 70s 


11.9 


-13.7 


E-17 ] 


late 70s 


10.3 


-15.3 


E-19 ] 


late 70s 


10.1 


-13.4 


E-21 ] 


late 70s 


10.4 


-14.9 


E-23 ] 


late 70s 


10.4 


-11.9 


E-25 ] 


late 70s 


10.2 


-15.5 


E-27 ] 


late 70s 


10.4 


-15.5 


E-29 ] 


late 70s 


10.3 


-13.7 


E-30 ] 


late 70s 


10.7 


-11.9 


E-34 ] 


late 70s 


9.8 


-14.2 



All specimens with Year of death of late 70s were 
collected by Cynthia Moss and have known dates 
of death that we have not yet received. For Fig. 
94 A, these animals are plotted as deaths in 1978. 



Fig. 93. (A). Histogram of nitrogen isotope com- 
positions for Amboseli plants subdivided accord- 
ing to physiogamy. (B) Histogram of nitrogen 
isotope compositions for Amboseli mammals sub- 
divided according to feeding type. Note that the 
difference between the mean £ 15 N values of plants 
and animal collagen is only ~+l %o. 



diets. Animals known to eat a mixture of 
plants (elephant, Grant's and Thomson's 
gazelle, hippopotamus, impala, and porcu- 
pine) have <5 13 C values intermediate be- 
tween browsers and grazers (Tables 26 and 
27). 

The link between the <5 13 C of diet and 
collagen is more difficult to unravel in 
carnivores and omnivores, because these 
animals obtain carbon both from different 
tissues within a body (fat, muscle, skin) and 



from plants. All these sources may have 
different <5 13 C values. Generally, herbi- 
vore meat and carnivore collagen differ by 
~ +5%o, but the difference between herbi- 
vore collagen and carnivore collagen is 
+2%c (van der Merwe, 1989). Observa- 
tions of hunting carnivores suggest that 
hyena and lion consume chiefly C4-feed- 
ing herbivores (wildebeest and zebra), 
whereas cheetah eat mixed feeders 
(Thomson's and Grant's gazelle and im- 
pala). These observations are supported by 
#3C values (Table 26, Fig. 92B). The 
smaller carnivores (fox, jackal, mongoose) 
eat smaller animals from across the dietary 
spectrum and exhibit a spread of <5 13 C 
values. 



GEOPHYSICAL LABORATORY 



169 





IB - 
16- 


A 


a Carnivores 

• Browser & Mixed 














14- 






A # 


10 
to 


12- 
10- 


Carnivore A • 


4 


• 

• *" 




8- 


Browser & Mixed 




• 
• 




6- 


1 1 1 r 







1955 1960 1965 1970 1975 1980 1985 1990 1995 

Year of Death 



16 
14 
12 
10 H 

z 

£ 8 
to 

6- 

4- 
2- 



-f- 



4 

h T~'teHr H 



hh 



* 



-30 



-25 



I I '» I | ill l 



-20 
8 13 C 



-15 



-10 



A Rhinoceros 
• Elephant 
*±* Mean & std. dev. of 
African park elephants 



Fig. 94. (A) Secular variation in the nitrogen 
isotope composition of mammals. Grazers: Y= 
-68.25 +.0.04X r=0.24, slope is not significantly 
different from 0. Data and regression line are not 
plotted. Browsers: Y= -279.33 + 0.15X r=0.62, 
slope is significantly different than 0. Carnivores: 
Y= -253.66 + 0. 14X r=0.58, slope is significantly 
different than 0. The elephants listed as late 70s 
deaths are included on the figure, and given 1978 
as the year of death, but they were not used in the 
regression calculation. (B) Carbon and nitrogen 
isotopic composition of Amboseli elephants and 
rhinoceros. Also plotted are the means and stan- 
dard deviations for elephant ivory from 16 other 
African parks and preserves. Firm determination 
of temporal isotopic trends within species must 
await analysis of specimens with a greater range of 
known ages of death. However, the spread in the 
isotopic data from both species may result from 
coupled increases in <5 13 C and <5 15 N values with 
time. Although values from Amboseli elephants 
do not overlap with values from many other parks, 
they vary by amounts as great as those used to 
discriminate between other park populations. 



A fractionation of ~ +3%c between the 
<5 15 N value of plant food and herbivore 
collagen has been reported in previous 
studies (DeNiro and Epstein, 1981 ; Hare et 
ai, 1991). Given this fractionation, the 
average Amboseli herbivore should have a 
collagen S l 5 N value of 1 2- 1 3 %o. Although 
some animals have values in this range, 
most are more negative (Tables 26 and 27, 
Fig. 93B) Indeed, using the mean ^ 5 N of 
animals and a fractionation of +3 %o, we 
would expect dietary plants with values of 
l%o or less. Few plants within the park 
have isotopic values this low. 

There are several plausible explana- 
tions for this discrepancy. First, we ana- 



lyzed only plants collected in a dry season. 
In the wet season, plants may have lower 
<5 15 N values. Second, if the 5 15 N value of 
plants from Amboseli has increased re- 
cently, current vegetation may not be repre- 
sentative of the foods eaten by the sampled 
animals . Finally, the fractionation between 
diet and herbivore collagen may not be 
+3%o. Variability in this fractionation has 
been detected previously (Ambrose and 
DeNiro, 1986; Heatonef a/., 1986; Sealy et 
al., 1987), and attributed to differences in 
N metabolism between different herbivores. 
A fractionation of ~+\%o is observed be- 
tween current Amboseli plants and the 
sampled herbivores. 



170 



CARNEGIE INSTITUTION 



Differences in collagen <5 15 N between 
herbivores and carnivores are well studied 
and range from +3 to +6 %o (Schoeninger 
and DeNiro, 1984; Ambrose and DeNiro, 
1986; Sealy et al., 1987). Amboseli graz- 
ers, mixed feeders and browsers averaged 
10.9 ± 1.3 %o, 9.1 ± 2.3 %o 9 and 10.9 ± 1.3 
%o, respectively, whereas true Amboseli 
carnivores averaged 15.4 ± \.6%c. Conse- 
quently, there is a trophic level fraction- 
ation of ~+5 %o. The omnivorous yellow 
baboon has a lower <5 15 N value, suggesting 
a preponderance of plant foods in the diet. 
Finally, the aardvark, which consumes ants 
and termites, has a <5 15 N value within the 
herbivore range. However, insect chitin is 
known to be 15 N-depleted relative to di- 
etary plants (Schimmelmann,pers. comm.), 
which would lead to lower values in the 
collagen of insectivores relative to carni- 
vores. 



Secular Trends in the Isotopic Composi- 
tion of Amboseli Mammals 

The Amboseli ecosystem has changed 
dramatically since 1960 because of a loss 
of trees and the expansion of grassland. To 
document isotopic trends in park mammals, 
multiple samples of individual species from 
different time periods must be examined. 
Our data are currently insufficient for such 
a treatment, but trends within broad feed- 
ing categories may be examined. Collagen 
<5 15 N and <5 13 C values, and thus the diet of 
grazers, have remained constant from 1968 
to 1990 (Table 26). The &$N of browsers 
and mixed feeders has increased by a sta- 
tistically significant amount (Fig. 94A). 



This trend, however, is strongly influenced 
by a low value for a single old specimen. 
Although most elephant deaths are only 
roughly dated to the late 1 970s, the popula- 
tion seems to be trending towards higher 
#5n and 8&C values (Fig. 94B). The 
magnitude of this variation is significant 
when compared to the differences between 
populations from different parks. There is 
a suggestion of similar coupled increases 
for rhinoceros, but the sample is quite small. 
Finally, carnivores also increase in <5 15 N 
with time by amount similar to browsers 
(Fig. 94A). 

We hypothesize that as the park was 
stripped of trees, browsers and mixed feed- 
ers have been forced to consume more 
grass. Increased grass consumption is par- 
ticularly evident for elephants. However, 
grass is relatively nutrient-poor compared 
to browse. Eventually, the browsers and 
mixed feeders suffered nutritional stress. 
Nutritional stress may cause an animal to 
remetabolize previously deposited proteins, 
and can potentially produce an increase in 
collagen <5 15 N in bones equivalent to that 
generated by feeding at a higher trophic 
level (Tuross, pers. com.). The grazers 
thrived as the grasslands expanded and 
exhibit no isotopic changes. Carnivores 
eat both types of herbivores, and conse- 
quently exhibit intermediate isotopic trends. 



Conclusions 

The carbon and nitrogen isotopes in 
most plants from Amboseli National Park 
varied as expected, with a strong differen- 
tiation in 5 J 3 C between C3, C4, and aquatic 



GEOPHYSICAL LABORATORY 



171 



plants, and a high mean d^N value. The 
15 N enrichment of succulent plants was 
unexpected, and is currently unexplained. 
Differences in the <5 13 C of plants is re- 
flected in the collagen of the animals that 
consume these plants, and ultimately can 
be detected at higher trophic levels when 
these herbivores are preyed upon by carni- 
vores. The fractionations of C and N iso- 
topes between diet and collagen that we 
discovered match previous reports, with 
one exception. The fractionation of N 
between plant and herbivore collagen was 
much lower than expected. Finally, al- 
though our observations must be supported 
by more extensive sampling, there are sta- 
tistically significant secular trends in the 
<5 15 N of Amboseli browsers and mixed 
feeders and carnivores, whereas grazers 
are invariant. The substantial isotopic trends 
shown by Amboseli elephants may indi- 
cate that stable isotopes will be of limited 
utility in tracing the source of elephant 
ivory in changing habitats. 



References 

Ambrose, S. H., and M. J. DeNiro, The isotopic 
ecology of East African mammals, Oecologia, 
69, 395-406, 1986. 

Behrensmeyer, A. K., Taphonomic and ecologic 
information from bone weathering, Paleobio., 
4, 150-162, 1978. 

DeNiro, M. J., and S. Epstein, Influence of diet on 
the distribution of carbon isotopes in animals, 
Geochim. Cosmochim. Acta, 42, 495-506, 
1978. 

DeNiro, M. J., and S. Epstein, Influence of diet on 
the distribution of nitrogen isotopes in ani- 
mals, Geochim. Cosmochim. Acta, 45, 341- 
351, 1981. 

Hare, P. E., M. L. Fogel , T. W. Stafford, Jr., A. D. 
Mitchell, and T. C. Hoering, The isotopic 
composition of carbon and nitrogen in indi- 
vidual amino acids isolated from modern and 



fossil proteins, J. Archaeol. Sci, in press, 199 1 . 

Heaton, T. E., J. C. Vogel, G. von la Chevallerie, 
and G. Collett, Climatic influence on the iso- 
topic composition of bone nitrogen, Nature, 
322, 822-823, 1986. 

Raven, J. ., The application of mass spectrometry 
to biochemical and physiological studies, in 
The Biochemistry of Plants, Vol. 13, Aca- 
demic Press, Inc., New York, pp. 127-180, 
1987. 

Schoeninger, M. J., and M. J. DeNiro, Nitrogen 
and carbon isotopic composition of bone col- 
lagen from marine and terrestrial animals, 
Geochim. Cosmochim. Acta, 48, 625-639 
1984. 

Sealy, J.C., N. J. van der Merwe, J. A. Lee Thorp, 
and J. L. Lanham, Nitrogen isotopic ecology 
in southern Africa: implications for environ- 
mental and dietary tracing, Geochim. 
Cosmochim. Acta, 57,2707-2717, 1987. 

Tieszen, L. L., and T. W. Boutton, Stable carbon 
isotopes in terrestrial ecosystem research, in 
Stable Isotopes in Ecological Research, 
Rundel, P.W., J.R. Ehleringer, and K. A. Nagy , 
eds., Springer- Verlag, New York, pp. 167- 
195, 1988 

Tuross, N.,M. L. Fogel, andP. E. Hare, Variability 
in the preservation of the isotopic composition 
of collagen from fossil bone, Geochim. 
Cosmochim. Acta, 52, 929-935, 1988. 

van der Merwe, N. J., Natural variations in 13 C 
concentration and its effect on environmental 
reconstruction using 13 C/ 12 C ratios in animal 
bones, in The Chemistry of Prehistoric Human 
Bone, Price, T.D., ed., Cambridge Univ. Press, 
New York, pp. 105-125, 1989. 

van der Merwe, N. J., J. A. Lee-Thorp, J. F. 
Thackeray, A. Hall-Martin, F. J. Kruger, 
H.Coetzee, R. H. V. Bell, and M. Lindeque, 
Source-area determination of elephant ivory 
by isotopic analysis, Nature, 346, 744-746, 
1990. 

Vogel, J. C, Isotopic assessment of the dietary 
habits of ungulates, S. Afr. J. Sci., 74, 298-30 1 , 
1978. 

Vogel, J. C, B. Eglington, and J. M. Auret, Isotope 
fingerprints in elephant bone and ivory, Na- 
ture, 346, 747-749, 1990. 



172 



CARNEGIE INSTITUTION 



Rapid Racemization of Aspartic Acid 

in mollusk and ostrich eggshells : 

A New Method for Dating on 

a Decadal Time Scale 

Glenn A. Goodfriend, David W. von 
Endt* and RE. Hare 

Epimerization of L-isoleucine to D- 
alloisoleucine has been extensively ana- 
lyzed in mollusk shells as a means of deter- 
mining relative or absolute ages, primarily 
of Pleistocene samples. More recently, 
epimerization in Pleistocene ostrich egg- 
shells has been studied (Brooks et al. , 1 990). 
Racemization of a number of other amino 
acids has been studied in mollusk shells, 
leucine being the most widely studied (e.g., 
Wehmiller, 1984; reviewed in Goodfriend, 
1991). Until recently, research on aspartic 
acid racemization in mollusks had been 
limited to some Pleistocene marine samples 
from the West Coast of the U. S. 
(Kvenvolden et al., 1979). But recent stud- 
ies on aspartic acid racemization in desert 
land snails have turned up an interesting 
pattern: the initial rate of racemization is 
extremely high; the rate then slows down 
progressively with increasing time or D/L 
ratio. This pattern has been demonstrated 
in both a radiocarbon-dated series of Holo- 
cene shells (Goodfriend, 1 99 1 ) as well as in 
heating experiments of modern shells 
(Goodfriend and Meyer, 1991). 

The very rapid initial rate of racemiza- 
tion presents the possibility of using aspar- 



Conservation Analytical Laboratory, 
Smithsonian Institution, Washington, D.C. 20560 



tic acid racemization as a high-resolution 
dating method for young materials. This is 
of particular interest because radiocarbon 
generally cannot be used for dating of post- 
1650 A.D. samples (see radiocarbon cali- 
bration curve of Stuiver and Pearson ( 1 986) 
for this period). On the other hand, evalua- 
tion of the precision of aspartic acid racem- 
ization dating in this time range is difficult, 
because of the unavailability of radiocar- 
bon as an independent measure of age. For 
this reason, we turned to museum mollusk 
collections as a source of material of known 
age with which to evaluate the age predic- 
tive ability of D/L aspartic acid ratios. In 
addition, we examined aspartic acid race- 
mization in three samples of ostrich egg- 
shells, to see if the phenomenon of rapid 
initial racemization also occurs in this ma- 
terial. 



Materials and Methods 

Seven samples of the land snail 
Triodopsis multilineata from Iowa and 
Kansas were obtained from the collection 
of the U. S. National Museum of Natural 
History. The dates of collection of these 
samples range from 1881 to 1949. Most of 
these could be seen to have been collected 
alive because of the presence of some re- 
mains of the animals inside the shells. In 
addition, two modem samples (1990 and 
1991) of another Triodopsis species, T. 
tridentata, were collected in the Chevy 
Chase district of Washington, D. C. Samples 
of modem ostrich eggs hatched at the Front 
Royal, Virginia, breeding farm of the Na- 
tional Zoological Park in 1978 and at Dolly 



GEOPHYSICAL LABORATORY 



173 



Farms (Vicksburg, Mississippi) in 1990 
were obtained. A bead fashioned out of 
ostrich eggshell, excavated from a fort at 
Oudepost, South Africa (occupied ca. 1652- 
1660 A.D.; Schrire et al., 1990) was also 
obtained for analysis. (The authors are in- 
debted to R. Herschler and P. Greenhall of 
the U. S. National Museum of Natural 
History for providing the material of 
Triodopsis multilineata used in this study. 
Ostrich egg samples were kindly provided 
by A. Brooks, C. Schrire, J. Kokis, the 
National Zoological Park, and R. Shafer of 
Dolly Farms.) 

The periostracum (the outer organic 
layer, covering or partly covering the 
mollusk shells) was ground off the shell 
samples by abrasive-tipped bits using a 
motorized hand-held tool. Organic mate- 
rial on the surface of the ostrich egg pieces 
was similarly removed. Samples were then 
subjected to a short dip in dilute HC1, 
washed three times in distilled water, and 
dried under vacuum. Hydrolysis and 
derivatization of the amino acids to their N- 
trifluoracetyl isopropyl ester derivatives 
were carried out as described in Goodfriend 
(1991). The D/L amino acid ratios were 
analyzed by gas chromatography using a 
Hewlett-Packard model 5790. The values 
are reported as the ratio of the areas of the 
D and L peaks, as calculated by a Hewlett- 
Packard model 3 3 94 A integrator. For the 
Triodopsis samples, analyses were based 
on preparations made from small pieces of 
three individual shells (total weight: 13-19 
mg); a piece from a single shell was ana- 
lyzed in the case of the T. tridentata and 
ostrich egg samples. At least two analyses 
of each snail shell preparation were carried 



out and the mean of these is reported. The 
standard error of these sample means aver- 
aged 0.001 1. The ostrich eggshell samples 
were analyzed once. 



Results 

The Triodopsis shells show a progres- 
sive increase in the D/L aspartic acid ratio 
with increasing age (Fig. 95). The initial 
value of 0.041 to 0.044 represents either 
racemization induced by the preparation 
procedure or the occurrence of small 
amounts of D-aspartic acid in the modern 
shells. From this initial value, the D/L ratio 
increases up to 0.093 in the 110-year-old 
specimens (collected in 1881). Although 
the modem specimens are of a different 
species than the others, different mollusk 
species within the same genus generally 
show the same rates of epimerization (Lajoie 
et al., 1980), so it is expected that the 
modem T. tridentata values are represen- 
tative of those of modem T multilineata. A 
simple linear regression of the D/L ratios 
on the age yields the equation 



D/L = 0.000458 (age) + 0.00431, 

which indicates an average racemization 
rate of l%/22 yr. (In this and subsequent 
analyses, a single anomalous sample (D/L 
of 0. 12 for a 1949 sample) was left out; this 
high value may have been the result of 
heating of the shell to extract the bodies or 
to dry the shells after extraction of the 
bodies.) The correlation coefficient between 
the D/L ratio and age is 0.979. An estimate 



174 



CARNEGIE INSTITUTION 



0.10 



-i — i — i — i — i- 




■ T. multilineata 
• T. tridentata 



1990 1970 1950 1930 1910 1890 

year A. D. 

Fig.95. DA- aspartic acid ratios in shell samples of 
two species of land snails of the genus Triodopsis. 
The year of collection of the samples is indicated 
on the horizonal axis. 



of the error of an age predicted from the D/ 
L ratio of a specimen was calculated from 
the data as the square root of the mean 
square error of a regression of age on D/L 
ratio. A value of 9.5 yr was thus obtained 
and indicates that the year of collection of 
an undated shell sample can be estimated 
with approximately this degree of precision 
based on analysis of its D/L aspartic acid 
ratio. This estimate assumes that the scatter 
of the points about the regression line is 
uniform over the range of values analyzed, 
whereas it appears that the scatter is greater 
at higher ratios, which is as expected if the 
scatter is due to variation in the average 
racemization rate of different samples. Thus 
the error of age estimations would be lower 
than the calculated value for more recent 
ages and higher for older ages. 

The few results available for the ostrich 
egg samples (Table 28) also suggest a high 
aspartic acid racemization rate in this ma- 



terial. A net racemization of about 0.095 , or 
a rate of about 1% racemization per 35 
years, is indicated for the eggshell sample 
about 330 years old. 



Discussion 

The results confirm the occurrence of a 
very high rate of aspartic acid racemization 
in museum mollusk material, as expected 
based on earlier studies of desert land snails. 
Because the samples show a regular pattern 
of increasing D/L ratios with increasing 
age, aspartic acid racemization may be 
useful as a dating method for materials on 
a decadal time scale. For study of museum 
collections, this may have several applica- 
tions. In biogeographical studies, it is often 
of importance to know when a specimen or 
set of specimens was found at a particular 
location, since distributions may change 
over short time scales. Aspartic acid race- 
mization analysis could be used to deter- 
mine the approximate time of collection of 
undated samples in museum collections. It 
could also be used to determine if speci- 
mens were alive (or freshly dead) when 
collected, or whether they represent older, 
dead-collected material. Older records of 
distributions could be obtained from such 
material. The method may also be applied 



Table 28. D/L aspartic acid ratios in some ostrich 
eggshell samples. 



Year 



Source 



D/L 



1990 Dolly Farms 0.044 

1978 Natl. Zoological Park 0.057 

ca. 1652-1660 Oudepost, S. Afr. 0.129 



GEOPHYSICAL LABORATORY 



175 



to dating of recent deposits in nature and 
should provide good time resolution for the 
post- 1650 A.D. period not covered by ra- 
diocarbon. Such applications require an in 
situ rate calibration based on independentiy- 
dated material of the same species. An in 
situ calibration is required since different 
museum collections or different field sites 
will differ in their average temperatures. In 
museum collections, this calibration can be 
obtained from other specimens of known 
collection dates. In the field, radiocarbon 
dates from pre- 1650 A.D. samples or dat- 
ing by association with archeological arti- 
facts are the most likely sources of cali- 
bration dates. Possible problems with ap- 
plication of the method to museum mate- 
rials may arise if samples have been sub- 
jected to prolonged heating or high tem- 
peratures during processing or storage. 

Aspartic acid racemization has been 
applied previously to dating of human teeth 
(e.g., Helfman and Bada, 1976; Ohtani et 
al., 1988), where it shows a high rate of 
racemization (1% per 13 yr in dentine; 
Helfman and Bada, 1976). However, this 
high rate occurs at body temperature, or 
about 37°C. The projected rate at room 
temperature (about 2 1 °C) would be 14 times 
slower, or 1% per 180 yr (assuming an 
activation energy of 30 kcal/mol). Thus 
very rapid racemization of aspartic acid in 
young materials seems to be limited to 
biogenic carbonate materials, such as mol- 
lusk shells and bird eggshells; biogenic 
phosphates, as represented by teeth, show a 
considerably slower rate. One may expect 
the phenomenon of very rapid initial race- 
mization of aspartic acid to be found also in 
other biogenic carbonates, such as fora- 



miniferal tests and coral skeletons. 

Some possible modifications of ana- 
lytical procedures could lead to even higher 
temporal resolution. For example, it has 
been found that the free amino acid fraction 
in mollusks is always more highly 
epimerized than the total amino acid frac- 
tion that is obtained from hydrolysis (e.g., 
Miller and Hare, 1980). Although D/L en- 
antiomer ratios have never been measured 
in the free amino acid fraction, it might be 
expected that this would yield similar re- 
sults, i.e., that the DA- aspartic acid ratio 
may increase faster in the free amino acid 
fraction than in the total. It has been sug- 
gested that the rapid initial rate of aspartic 
acid racemization may actually be the re- 
sult of the racemization of asparagine rather 
than aspartic acid per se (Goodfriend, 1 99 1 ); 
asparagine is converted to aspartic acid 
during hydrolysis, so what is measured as 
"aspartic acid" is actually the sum of the 
aspartic acid and asparagine originally 
present in the sample. Development of 
methods for measuring the D/L ratio of 
asparagine may result in better time resolu- 
tion for dating applications. 



References 

Brooks, A. S., P. E. Hare, J. E. Kokis, G. H. Miller, 
R. D. Ernst, and F. Wendorf, Dating Pleisto- 
cene archeological sites by protein diagenesis 
in ostrich eggshell, Science, 248, 60-64, 1990. 

Goodfriend, G. A., Patterns of racemization and 
epimerization of amino acids in land snail 
shells over the course of the Holocene, 
Geochim. Cosmochim. Acta, 55, 293-302, 
1991. 

Goodfriend, G. A., and V. R. Meyer, A compara- 
tive study of amino acid racemization/ 
epimerization kinetics in fossil and modern 
mollusk shells, Geochim. Cosmochim. Acta, 
in press. 



176 



CARNEGIE INSTITUTION 



Helfman, P. M., and J. L. Bada, Aspartic acid 
racemisation in dentine as a measure of age- 
ing, Nature, 262, 279-281, 1976. 

Kvenvolden, K. A., D. . Blunt, andH. E. Clifton, 
Amino-acid racemization in Quaternary shell 
deposits at Willapa Bay, Washington, 
Geochim. Cosmochim. Acta, 43, 1505-1520, 
1979. 

Lajoie, K. R., J. F. Wehmiller, and G. L. Kennedy, 
Inter- and intrageneric trends in the apparent 
racemization kinetics of amino acids in Qua- 
ternary mollusks, mBiogeochemistry of Amino 
Acids, P. E. Hare, T. C Hoering, and K. King, 
Jr., eds., John Wiley and Sons, New York, pp. 
305-340, 1980. 

Miller, G. H. and P. E. Hare, Amino acid geochro- 
nology: integrity of the carbonate matrix and 
potential of molluscan fossils, mBiogeochem- 
istry of Amino Acids, P. E. Hare, T. C. Hoering, 
and K. King, Jr., eds., John Wiley and Sons, 
New York, pp. 415-443, 1980. 

Ohtani, S., S. Kato, H. Sugeno, H. Sugimoto, T. 
Marumo, M. Yamazaki, and K. Yamamoto, A 
study on the use of the amino-acid racemiza- 
tion method to estimate the ages of unidenti- 
fied cadavers from their teeth, Bull. Kanagawa 
Dental College, 16, 11-21, 1988. 

Schrire, C, J. Deetz, D. Lubinsky, and C. 
Poggenpoel, The chronology of Oudepost I, 
Cape, as inferred from an analysis of clay 
pipes, J. Archaeol. ScL, 17, 269-300, 1990. 

Stuiver, M. and G. W. Pearson, High-precision 
calibration of the radiocarbon time scale, AD 
1950-500 BC, Radiocarbon, 28, 805-838, 
1986. 

Wehmiller, J. F., Relative and absolute dating of 
Quaternary mollusks with amino acid racem- 
ization: evaluation, applications and questions, 
in Quaternary Dating Methods, W. C. 
Mahaney, ed., Elsevier, Amsterdam, pp. 171- 
193, 1984. 



A Burning Question: Differences 

between Laboratory-Induced 

and Natural Di agenesis in Ostrich 

Eggshell Proteins* 

A. S. Brooks, RE. Hare, J.E. Kokis, and 
K. Durana 

In earlier papers (Brooks et ai, 1990; 
Kokis et al., 1990), we demonstrated the 
utility of the D-alloisoleucine/L-isoleucine 
(A/I) ratio in ostrich eggshell for estimat- 
ing the age of archaeological specimens. 
Of the biogenic carbonates and phosphates 
tested so far, ostrich eggshell most nearly 
approximates a closed system with little 
loss of either water or protein breakdown 
products to the environment. Although 
ostrich eggshell is a common material in 
archaeological sites located in the arid and 
semiarid regions of the Old World, human 
activity and natural factors at these sites 
may produce anomalous epimerization ra- 
tios in two ways: by heating (camp fires, 
brush fires) and by stratigraphic mixing 
through ancient excavations (pits, burials, 
burrows, etc.) into underlying deposits 
which may also contain eggshell from hu- 
man occupation debris. For any archaeo- 
logical horizon, it is important to be able to 
distinguish between heating and strati- 
graphic admixture, especially of older ma- 
terials. Heating to certain higher tempera- 
tures may indicate the presence of human- 
controlled fire, whereas stratigraphic ad- 
mixture may call into question the interpre- 



This work is supported by NSF Grant BNS- 
9011657 



GEOPHYSICAL LABORATORY 



177 



tation of other materials at the site, e.g., 
human fossils. In addition, if temperature 
differentially affects two decomposition 
reactions because they have different acti- 
vation energies, we may be able to use 
these two reactions to determine simulta- 
neously both time and temperature. In this 
paper, we describe results from laboratory 
heating of ostrich eggshell fragments at 
controlled temperatures for varying peri- 
ods. Amino acid compositions of these 
heated samples were compared to amino 
acid compositions of archaeological 
samples from the last 80,000 years. 



Materials and Experimental Methods 

Heating experiments were conducted 
in a heated aluminum block with a Model 
71 A Temperature Controller (RFL Indus- 
tries, Inc., Boonton, New Jersey). Tem- 
perature readings were within ±0.2° C. 
Samples of eggshell fragments were 
weighed and dropped into pre -heated tubes 
placed in the aluminum block. Samples 
were then processed for free and total amino 
acids as described by Brooks et al. (1990). 

Three different sample series of ostrich 
eggshell fragments were heated in the labo- 
ratory, and two sets of archeological samples 
were analyzed. (Modem ostrich eggshell 
samples were provided by the National 
Zoological Park, Washington, D.C. and R. 
Shafer of Dolly Farms, Vicksburg, Missis- 
sippi. Archaeological eggshell samples 
were provided by O. Bar-Yosef, A.S. 
Brooks, J. Deacon, M. Mehlman, and W.E. 
Wendt.) 



Laboratory-Heated Samples: 

(1) A series heated dry at 300°C for 
incremental time periods of 15 min., 30 
min., 1 hr.,2hrs.,4hrs., 8hrs., 16hrs.,and 
32hrs. 

(2) A series heated dry for one hour 
each at temperatures of 160°C, 200°C, 
240°C, 280°C, 320°C, and 360°C. 

(3) A series heated for incremental 
time periods in water vapor at 157°C at 
times ranging from 0.5 to 256 hours. 
Archaeological Samples: 

(4) A stratified series of archaeological 
samples from a tropical zone site (7uGi, 
Botswana), in which the A/I ratio increases 
regularly with age and depth to above 1.0. 

(5) A group of pieces with anomalous 
A/I ratios from archaeological sites (7iGi, 
Boomplaas, Mumba Shelter, Qafzeh, 
Apollo 11). These pieces either had no 
significant A/I peaks or had ratios which 
were much higher than others from the 
same or underlying levels. 

For each series we measured the ratios 
of A/I peak areas subtracting 0.015 from 
each ratio to correct for laboratory-induced 
epimerization. We also measured peak ar- 
eas of aspartic acid (Asp), glutamic acid 
(Glu), glycine (Gly), alanine (Ala), and 
ammonia (NH3), as well as serine (Ser), 
threonine (Thr), and arginine (Arg). 



Results 

In the heating experiments, a sequence 
of changes in amino acid composition and 
concentrations occurs that can be described 
as a series of stages. 



178 



CARNEGIE INSTITUTION 



Stage 0: Modern unheated. Four major 
amino acids (Glu, Gly, Asp, Ala) occur in 
comparable amounts. Other amino acids, 
including Ser, Arg, Thr, and He, are also 
present at lower levels. No significant 
amounts of alloisoleucine are found. The 
level of NH3 relative to the four major 
amino acids is insignificant. 

Stage 1. Light heating (1 hour at 160°- 
200° C). The four major amino acids persist 
in comparable amounts, whereas Ser, Arg, 
and Thr diminish to trace levels. 
Alloisoleucine increases with length of heat- 
ing. The NH3 level is elevated. 

Stage 2. Moderate heating (1 hour at 
200°C-280°C). The four major amino ac- 
ids no longer remain at comparable levels; 
glutamic remains relatively constant while 
the others decrease. Serine, threonine, and 
arginine diminish to only trace levels or are 
completely absent. A/I values are not al- 
ways possible to determine due to the ap- 
pearance of interfering peaks. The level of 
NH3 steadily increases from 200°C to 280°C 
and over time. 

Stage 3. Strong heating (1 hour at 
300°C-360°C). The four major amino ac- 
ids diminish to only trace levels. Some Glu 
persists after other amino acids disappear. 
NH3 is the predominant peak. Some new 
peaks appear at this stage, which are tenta- 
tively identified as y-amino-butyric acid 
(GABA), and some amines, possibly 
methyl, ethyl, and propyl. Alloisoleucine 
and isoleucine levels are too low to calcu- 
late with confidence. Interestingly, there 
appears to be some synthesis of amino 
acids at these low levels. 



In the archaeological series the same 
four stages are observed. Samples that show 
good correlations with other age estimates, 
such as radiocarbon dating, exhibit only 
Stage or 1 patterns. Some of the anoma- 
lous samples from archaeological sites that 
do not correlate with the other age esti- 
mates show stage 2 or stage 3 patterns, with 
high NH3 levels and the presence of amines 
and probably GABA. Other anomalous ar- 
chaeological samples exhibit little change 
from stage or early stage 1 patterns; these 
show no evidence of heating and are pre- 
sumed to have derived from underlying 
levels by stratigraphic admixture. 



Discussion 

Differences were immediately apparent 
on inspection of the chromatograms for the 
archaeological series compared to the 
heated series. In the archaeological series, 
little significant decrease was noted in the 
four "stable" amino acids studied, even 
while A/I ratios increased to over 1 .0. In 
addition, there was little build-up of NH3. 
In the heated samples, in contrast, at tem- 
peratures as low as 200°C for 1 hour, or at 
157°C for 32 hours, there is a net increase 
in NH3 and a decrease in aspartic concen- 
tration relative to the more stable concen- 
tration of glutamic acid. The archaeologi- 
cal samples from 7cGi differ from all the 
pieces from series 2 heated at 200°C or 
above, all the pieces from series 1 (300°C), 
and all pieces from series 3 (157°C) heated 
32 hours or more in two respects: (1) heated 



GEOPHYSICAL LABORATORY 



179 



pieces exhibit a higher concentration of 
NH3 than of aspartic acid, and (2) no ar- 
chaeological piece has more than 50% as 
much NH3 as aspartic acid, except for the 
anomalous pieces. 

Of the anomalous archaeological pieces 
studied, three high A/I ratio pieces from 
7iGi were found in the top Later Stone Age 
horizons where A/I ratios normally ranged 
from 0.1 to 0.5 and radiocarbon calibra- 
tions suggested an age in the last 35,000 
years. Two of these pieces clearly fit with 
the earlier Middle Stone Age series and are 
presumably derived from below by human 
or animal disturbance. However, one of the 
three which provided an infinite radiocar- 
bon age (>40,000 yr B.R) must have been 
heated in antiquity, as it matched the amino 
acid composition of the strongly heated 
(early Stage 3) laboratory samples. Two 
anomalous pieces from the top levels at 
another site (Boomplaas) gave A/I ratios 
higher than the pieces from the bottom 
levels whose estimated age was 80,000 
B.R One of these pieces has been dated by 
TAMS 14 C to 5220 ± 70 yr B.R It exhibits 
an NH3-to-aspartic acid ratio greater than 
1.0, and is presumed to have been heated. 

Other anomalous archaeological pieces 
contained high NH3 concentrations but 
insufficient amounts of alloisoleucine or 



isoleucine to establish the A/I ratio. As 
mentioned above, experimental pieces sub- 
jected to stronger heating conditions (320°C 
for one hour, or 300° C for four or more 
hours) exhibited high NH3 and g-amino- 
butyric acid peaks as well as other very 
small peaks. A/I ratios in these strongly 
heated pieces could not be measured pre- 
cisely, but did not appear to progress much 
above 0.7. Examination of the small peaks 
suggests that at the higher temperatures in 
the heating experiments, serine as well as 
some other amino acids are being synthe- 
sized. Archaeological samples from Apollo 
11 in Namibia, Mumba Shelter in Tanza- 
nia, and Qafzeh Cave in Israel also con- 
form to this latter pattern and were almost 
certainly heated to relatively high tempera- 
tures in antiquity. 



References 

Brooks, A.S.,P.E. Hare, J. E. Kokis, G. H. Miller, 
R. D. Ernst, andF. Wendorf, Dating pleistocene 
archaeological sites by protein diagenesis in 
Ostrich eggshell, Science, 248, 60-64, 1990. 

Kokis, J. E., A. S. Brooks, andP. E. Hare, Chronol- 
ogy a nd aminostratigraphy of Middle and Late 
Stone Age sites from Sub-saharan Africa: A 
comparison of protein diagenesis and radio- 
carbon dating of ostrich eggshell, Geological 
Society of America Abstracts With Programs, 
22, A145-146, 1990, 



GEOPHYSICAL LABORATORY 



181 



Publications 

Reprints of the numbered publications listed below are available, except where noted, at no charge 
from the Librarian, Geophysical Laboratory, 5251 Broad Branch Road, N.W, Washington, D.C. 
20015-1305, U.S.A. Please give reprint number(s) when ordering. Youmay also request to be placed 
on the Laboratory's mailing list to receive periodic notifications of recent publications. 



Angel, R. J., N. L. Ross, L. W. Finger, and R. M. 
Hazen, Ba3CaCuSi60n : A new {1B,1s(i,oo)} 
{ 4 Si60i7} chain silicate, Acta Crystallogr. 
C46, 2028-2030, 1990 (G.L. Paper 2190). 

Angel, R. J., R. K. McMullen, and C. T. Prewitt, 
Substructure and superstructure of mullite by 
neutron diffraction, Am. Mineral., 76, 332- 
342, 1991 (G.L. Paper 2216). 

B adding, J. V., H. K. Mao, and R. J. Hemley, 
High-pressure synchrotron X-ray diffraction 
of Cs IV and Cs V, Solid State Commun., 77, 
801-805, 1991 (G.L. Paper 2208). 

Bebout, G. E., Field-based evidence for 
devolatilization in subduction zones: Implica- 
tions for arc magmatism, Science, 251, 413- 
416, 1991 (G.L. Paper 2206). 

Bebout, G. E., Geometry and mechanisms of fluid 
flow at 15 to 45 kilometer depths in an early 
cretaceous accretionary complex, Geophys. 
Res. Lett., 18, 923-926, 1991 (G.L. Paper 
2217). 

Chamberlain, C. P., J. M. Ferry, and D. Rumble, 
III, The effect of net-transfer reactions on the 
isotopic composition of minerals, Contrib. 
Mineral. Petrol., 105, 322-336, 1990 (G.L. 
Paper 2192; no reprints available for distribu- 
tion). 

Cifuentes, L. A., L. E. Schemel, and J. H. Sharp, 
Qualitative and numerical analysis of the ef- 
fects of river inflow variations on mixing 
patterns in estuaries, Estuarine Coastal Shelf 
Sci., 30, 41 1-427, 1990 (G.L. Paper 2198; no 
reprints available for distribution). 

Coffin, R. B., D. J. Velinsky, R. Devereux, W. A. 
Price, and L. A. Cifuentes, Stable carbon iso- 
tope analysis of nucleic acids to trace sources 
of dissolved substrates used by estuarine bac- 



teria, Appl. Environ. Microbiol., 56, 2012- 
2020, 1990 (G.L. Paper 2191). 

Cohen, R. E., Bonding and elasticity of stishovite 
Si02 at high pressure: linearized augmented 
plane wave calculations, Am. Mineral., 76, 
733-742, 1991 (G. L Paper 2220). 

Fei, Y., H. K. Mao, and B. O. My sen, Experimen- 
tal determination of element partitioning and 
calculation of phase relations in the MgO- 
FeO-SiC>2 system at high pressure and high 
temperature, /. Geophys. Res., 96, B2, 2157- 
2169, 1991 (G.L. Paper 2205). 

Finger, L. W., R. M. Hazen, and C. T. Prewitt, 
Crystal structures of Mgi2SUOi9(OH)2 (Phase 
B) and Mgi4Si5<I>24 (Phase AnhB), Amer. 
Mineral, 76, 1-7, 1991 (G.L. Paper 2219). 

Hare, P. E., M. L. Fogel, T. W. Stafford, Jr., A. D. 
Mitchell, and T. C. Hoering, The isotopic 
composition of carbon and nitrogen in indi- 
vidual amino acids isolated from modern and 
fossil proteins, /. Archaeol. Sci., 18, 277-292, 
1991 (G.L. Paper 2215). 

Hanfland, M., R. J. Hemley, and H. K. Mao, 
Optical absorption measurements of hydrogen 
at megabar pressures, Phys. Rev. B, 43, 8767- 
8770 1991 (G.L. Paper 2213). 

Hazen, R. M., Crystal structures of high-tempera- 
ture superconductors, in Physical Properties 
of High-Temperature Superconductors II, D. 
M. Ginsberg, ed., Chapter 3, pp. 121-198, 
World Scientific, New Jersey, 1990 (G.L. Pa- 
per 2158; no reprints available for distribu- 
tion). 

Hazen, R. M., and J. S. Trefil, Science Matters: 
Achieving Scientific Literacy, Doubleday, New 
York, 1991 (G.L. Paper 2195) (Available at 
your local bookstore or if you prefer direct 
from Doubleday) 



182 



CARNEGIE INSTITUTION 



Hazen, R. M., and J. S. Trefil, Achieving geologi- 
cal literacy, /. Geol. Educ, 39, 28-30, 1991 
(G.L. Paper 2196) 

Hazen, R. M, J. Zhang, and J. Ko, Effects of Fe/ 
Mg on the compressibility of synthetic 
wadsleyite:P-(Mg 1 . J ^ejt) 2 Si0 4 (x<Q.25),P/ry5. 
Chem. Minerals ,17, 416-419, 1990 (G.L. Pa- 
per 2197). 

Hemley, R. J., H. K. Mao, L. W. Finger, A. P. 
Jephcoat, R. M. Hazen, and C. S. Zha, Equa- 
tion of state of solid hydrogen and deuterium 
from single-crystal X-ray diffraction to 26.5 
GPa, Phys. Rev. B, 42, 6458-6470, 1990 (G.L. 
Paper 2180). 

Hemley, R. J., andH. K. Mao, Critical behavior in 
the hydrogen insultator-metal transition, Sci- 
ence, 249, 391-393, 1990 (G.L. Paper 2184). 

Hemley, R. J., H. K. Mao, and M. Hanfland, 
Spectroscopic investigations of the insulator- 
metal transition in solid hydrogen, in Molecu- 
lar Systems under High Pressure (Proceed- 
ings of the II Archimedes Workshop on Mo- 
lecular Solids under Pressure Catania, Italy, 
28-31 May 1990) R. Pucci, and G. Piccitto, 
eds., pp. 223-243, Elsevier, New York, 1991 
(G.L. Paper 2189). 

Hemley, R. J., H. K. Mao, and J. F. Shu, Low- 
frequency vibrational dynamics and structure 
of hydrogen at megabar pressures, Phys. Rev. 
Lett., 65, 2670-2673, 1990 (G.L. Paper 2201). 

Hemley, R. J., and J. D. Kubicki, Deep mantle 
melting, Nature, 349, 283-284, 1991 (G.L. 
Paper 2209). 

Hemley, R. J., M. Hanfland, andH. K. Mao, High- 
pressure dielectric measurements of solid hy- 
drogen to 170 GPa, Nature, 350, 488-491, 
1991 (G.L. Paper 2218). 

Kubicki, J. D., G. E. Muncill, and A. C. Lasaga, 
Chemical diffusion in melts on the 
CaMgSi206-CaAl2Si2C>8 join under high pres- 
sures, Geochim. Cosmochim. Acta, 54, 2709- 
2715, 1990 (G.L. Paper 2199). 

Kubicki, J. D., and A. C. Lasaga, Molecular dy- 
namics and diffusion in silicate melts, in Dif- 
fusion, Atomic Ordering, andMass Transport, 
J. Ganguly, ed., pp. 1-50, Advances in Physi- 
cal Geochemistry Series, Springer- Verlag, 
New York, 1990 (G.L. Paper 2202; no reprints 
available for distribution). 



Kubicki, J. D., and A. C. Lasaga, Molecular dy- 
namics simulations of pressure and tempera- 
ture effects on MgSiC>3 and Mg2Si04 melts 
and glasses, Phys. Chem. Minerals, 77, 661- 
673, 1991 (G.L. Paper 2224). 

Kudoh, Y., C. T. Prewitt, L. W. Finger,A. 
Darovskikh, and E. Ito, Effect of iron on the 
crystal structure of (Mg,Fe)SiC>3 perovskite, 
Geophys. Res. Lett. , 17, 1481-1484, 1990 
(G.L. Paper 2179). 

Kushiro, I., and B. O. Mysen, Experimental stud- 
ies of the system Mg2Si04-Si02-H2 at pres- 
sures 10"2-10"10 bar and temperatures to 

o 

1650 C: Application to condensation and 
vaporization processes in the primitive solar 
nebula, in Progress in Metamorphic andMag- 
matic Petrology, (D. S. Korzhinskiy Memo- 
rial Volume), L. L. Perchuk, ed., Chapter 16, 
pp. 4 1 1 -433 , Cambridge University Press, New 
York, 1991 (G.L. Paper 2214). 

Liu, X., and C. T. Prewitt, High- temperature dif- 
fraction study of LnCoC>3 perovskites: A 
high-order electronic phase transition. J. Phys. 
Chem. Solids, 52, 441-448, 1991 (G.L. Paper 
2204). 

Mao, H. K., R. J. Hemley, and M. Hanfland, 
Infrared reflectance measurements of the in- 
sulator-metal transition in solid hydrogen, 
Phys. Rev. Lett., 65, 484-487, 1990 (G.L. 
Paper 2186). 

Mao, H. K., Y. Wu, L. C. Chen, J. F. Shu, and A. 
P. Jephcoat, Static compression of iron to 300 
GPa and Feo.8Nio.2 alloy to 260 GPa: Implica- 
tions for composition of the core, /. Geophys. 
Res.,95, B13, 21,737-21,742, 1990 (G.L. Pa- 
per 2194). 

Mao, H. K., R. J. Hemley, Y. Fei, J. F. Shu, L. C 
Chen, A. P. Jephcoat, Y. Wu, and W. A. 
Bassett, Effect of pressure, temperature, and 
composition on lattice parameters and density 
of (Fe,Mg)Si03-perovskites to 30 GPa, /. 
Geophys. Res.,96,B5,S069-S019, 1991 (G.L. 
Paper 2212). 

Mao, H. K., and R. J. Hemley, Optical transitions 
in diamond at ultrahigh pressures, Nature, 
351, 721 724, 1991 (G.L. Paper 2222). 

Mysen, B. O., Volatiles in magmatic liquids, in 
Progress in Metamorphic and Magmatic Pe- 
trology (D. S. Korzhinskiy Memorial Vol- 
ume), L. L. Perchuk, ed., Chapter 17, pp. 435- 



GEOPHYSICAL LABORATORY 



183 



475, Cambridge University Press, New York, 
1991 (G.L. Paper 2172). " 

My sen, B. O., Relationships between silicate melt 
structure and petrologic processes, Earth-Sci- 
ence Reviews, 27, 281-365, 1990 (G.L. Paper 
2174). 

Mysen, B. O., Effect of pressure, temperature, and 
bulk composition on the structure and species 
distribution in depolymerized alkali alumino- 
silicate melts and quenched melts,/. Geophys. 
Res., 95, BIO, 15,733-15,744, 1990 (G.L. Pa- 
per 2181). 

Mysen, B. O., Interaction between water and melt 
in the system CaAl2-04-Si02-H20, Chem. 
Geol, 88, 223-243, 1990 (G.L. Paper 2188). 

Nagahara, H. , I. Kushiro, and B . O. Mysen, Vapor- 
ization and condensation experiments in the 
system olivine-hydrogen, in Dynamic Pro- 
cesses of Material Transport and Transforma- 
tion in the Interior of the Earth, S. Marumo, 
ed., pp. 473-490, Terra Pub., Tokyo, 1990 
(G.L. Paper 2225; no reprints available for 
distribution). 

Parise, J. B., Y. Wang, A. Yeganeh-Haeri, D. E. 
Cox and Y. Fei, Crystal structure and thermal 
expansion of (Mg,Fe)Si03 perovskite, 
Geophys. Res. Lett., 17,2089-2092,\990(G.L. 
Paper 2207; no reprints available for distribu- 
tion). 

Pickett, W. E., R. E. Cohen, and H. Krakauer, 
Lattice instabilities, isotope effect, and high- 
T c superconductivity in La2- x Ba x Cu04 )J P/ry 1 y. 
Rev. Lett., 67, 228-231, 1991. 

Ross, N. L., and R. M. Hazen, High-pressure 
crystal chemistry of MgSi03 perovskite, Phys. 
Chem. Minerals, 17, 228-237, 1990 (G.L. Pa- 
per 2176). 

Ross, N. L., J. F. Shu, R. M. Hazen, and T. 
Gasparik, High-pressure crystal chemistry of 
stishovite, Am. Mineral.,75, 739-747, 1990 
(G.L. Paper 2185). 

Ross, N. L., and K. Leinenweber, Single crystal 
structure refinement of high-pressure ZnGe03 
ilmenite, Z. Kristallogr., 191, 93-104, 1990 
(G.L. Paper 2203; no reprints available for 
distribution). 

Stafford, T. W., Jr., P. E. Hare, L. Currie, A. J. T 
Jull, and D. Donahue, Accuracy of North 
American human skeleton ages, Quaternary 
Research, 34, 111-120, 1990 (G.L. Paper 



2175). 

Stafford, T. W., Jr., P. E. Hare, L. Currie, A. J. T. 
Jull, andD. J. Donahue, Accelerator radiocar- 
bon dating at the molecular level, /. Archaeol. 
Sci., 18, 35-72, 1991 (G.L. Paper 2193). 

Stathoplos, L., and P. E. Hare, Amino acids in 
planktonic foraminifera: Are they phyloge- 
netically useful? in Origin, Evolution, and 
Modern Aspects ofBiomineralization in Plants 
and Animals, Proceedings of the Fifth Inter- 
national Symposium on Biomineralization, R. 
E.Crick,ed.,pp.329-338,PlenumPubl.Corp., 
New York, 1989 (G.L. Paper 2166; no reprints 
available for distribution). 

Velinsky, D. J., M. L. Fogel, J. F. Todd, and B. M. 
Tebo, Isotopic fractionation of dissolved am- 
monium at the oxygen-hydrogen sulfide inter- 
face in anoxic waters, Geophys. Res. Lett., 18, 
649-652, 1991 (G.L. Paper 2211). 

Yoder, H. S., Jr., Heat transfer during partial 
melting: An experimental study of a simple 
binary silicate system,/. Volcanol. Geotherm. 
Res. 43, 1-36, 1990 (G.L. Paper 2182). 

Zeitler, P. K., B. Barreiro, C. P. Chamberlain, and 
D. Rumble, IE, Ion-microprobe dating of zir- 
con from quartz- graphite veins at the Bristol, 
New Hampshire, metamorphic hot spot, Geol- 
ogy, 18, 626-629, 1990 (G.L. Paper 2187; no 
reprints available for distribution). 

Zhang, J., D. Ye, and C. T. Prewitt, Relationship 
between the unit-cell volumes and cation radii 
of isostructural compounds and the additivity 
of the molecular volumes of carbonates, Am. 
Mineral.,76, 100-105, 1991 (G.L. Paper 22 10). 

Zheng, Z Z., D. X. Gu, Y, Xin, D. O. Pederson, L. 
W. Finger, C G. Hadidiacos, andR. M. Hazen, 
A new 1212-type phase: Cr-substituted 
TlSr 2 CaCu207 with T c up to about 110 K, 
Modern Phys. Lett., 5, 635-642, 1991 (G. L. 
Paper 2223; no reprints available for distribu- 
tion). 



GEOPHYSICAL LABORATORY 



185 



Personnel 

July 1, 1990 to June 30, 1991 



Research Staff 

Charles T. Prewitt, Director 
Francis R. Boyd, Jr. 
Ronald E. Cohen 1 
Larry W. Finger 
Marilyn L. Fogel 
John D. Frantz 
P. Edgar Hare 
Robert M. Hazen 
Russell J. Hemley 
Thomas C. Hoering 
T. Neil Irvine 
Ho-Kwang Mao 
Bjorn O. My sen 
Douglas Rumble III 
David Virgo 
HattenS. Yoder, Jr. 

Postdoctoral Associates 

Zhaoxin Gong 2 
David Palmer 3 
Ellen K. Wright 4 

Research Associates 

Jingzhu Hu 
Jinfu Shu 

Postdoctoral Fellows 

John V. B adding 5 
Gray E. Bebout 
James Brenan 6 
Yingwei Fei 7 



Michael Hanfland 
David B. Joyce 8 
Paul L. Koch 9 
James D. Kubicki 10 
Charles Meade 11 
Craig M. Schiffries 12 
Hiroko Takahashi 13 
Willem L. Vos 14 
Jinmin Zhang 15 

Predoctoral Associate 

Julie Kokis 16 

Research Interns 

Craig Bates 17 
Jon Cramer 18 
Karen Durana 19 
Howard Lu 20 
Alistaire M. Moore 21 
Nicole Y. Morgan 22 

Supporting Staff 

Andrew J. Antoszyk, Shop Foreman 
Bobbie L. Brown, Instrument Maker 
Stephen D. Coley, Sr., Instrument Maker 
David J. George, Electronics Technician 
Christos G. Hadidiacos, 

Electronics Engineer 
Marjorie E. Imlay, Assistant to the Director 
Lavonne Lela, Librarian 23 
Yunye Luo, Library Technician 24 
Harvey J. Lutz, 

Technician/Mail Supervisor 



186 



CARNEGIE INSTITUTION 



Mary M. Moore, 

Word Processor Operator 

— Receptionist 
Lawrence B. Patrick, 

Maintenance Supervisor 25 
David Ratliff, Jr., 

Maintenance Technician 26 
Pedro J. Roa, 

Maintenance Technician 27 
Susan A. Schmidt, 

Coordinating Secretary 
John M. Straub, 

Business Manager 
Mark Vergnetti, 

Instrument Maker 28 
Stephanie Vogelpohl, 

Administrative Assistant 29 



J. Michael Palin, Yale University 
Nicolai P. Pokhilenko, Inst. Mineralogy & 

Petrology, Novosibirsk, USSR 
Robert Popp, Texas A. and M. 
Guoyin Shen, 

University of Uppsala, Sweden 
Bradley Tebo, 

Scripps Institution of Oceanography 
Noreen C. Tuross, Smithsonian Institution 
K. Vedam, Pennsylvania State University 
David von Endt, Smithsonian Institution 
YanWu, 

University of California, Berkeley 

Adjunct Senior Research Scientist 
Peter M. Bell 



Visiting Investigators 



Emeritus 



Rateb M. Abu-Eid, 

Kuwait Institute for Scientific Research 
Constance Bertka, 

Arizona State University 
Alison Brooks, 

George Washington University 
Robert T. Downs, 

Virginia Polytechnic Institute 

& State University 
Glenn A. Goodfriend, 

Weizmann Institute of Science, Israel 
Matthew Hoch, University of Delaware 
Hans G. Huckenholz, 

Munich University, Germnay 
Donald G. Isaak, 

Naval Research Laboratory 
James G. Kirklin, 

Johns Hopkins University 
Kevin Mandernack, 

Scripps Institution of Oceanography 



Hatten S. Yoder, Jr., Director Emeritus 
Felix Chayes, Petrologist Emeritus 



Appointed Sept. 1, 199U 

2 Appointed Jan. 28, 1991 

3 Appointed Oct. 1, 1990 

4 To June 30, 1991 

5 Accepted position as Assistant Professor, The 
Pennsylvania State University 

6 Appointed Nov. 15, 1990 

7 Accepted position as Associate Staff Member, 
Geophysical Laboratory 

8 To June 30, 1991 

9 Appointed Sept. 1, 1990 
°To December 30, 1990 

1 Appointed Gilbert Fellow July 1, 1990 
2 To Sept. 30, 1990 

3 Appointed March 1, 1991 

4 Appointed April 1, 1991 
5 To June 30, 1991 
Appointed July 1, 1990 
7 FromJune24, 1991 
8 FromMay 15, 1991 
9 FromJune 1, 1991 

20 From June 24, 1991 
21 From June 24, 1991 
22 From June 24, 1991 



GEOPHYSICAL LABORATORY 187 

23 Also associated with the Department of 
Terrestrial Magnetism (DTM) 

24 Also associated with DTM 

25 Also associated with DTM 

26 Also associated with DTM 

27 Also associated with DTM 

28 To Nov. 30, 1990 

29 AppointedJune3, 1991