Volatiles in glasses from the HSDP2 drill core
Caroline Seaman
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California 91125, USA
Sarah Bean Sherman and Michael O. Garcia
Department of Geology and Geophysics, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA
Michael B. Baker, Brian Balta, and Edward Stolper
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California 91125, USA
(ems@gps.caltech.edu)
[
1
]
H
2
O, CO
2
, S, Cl, and F concentrations are reported for 556 glasses from the submarine section of
the 1999 phase of HSDP drilling in Hilo, Hawaii, providing a high-resolution record of magmatic
volatiles over
200 kyr of a Hawaiian volcano’s lifetime. Glasses range from undegassed to having
lost significant volatiles at near-atmospheric pressure. Nearly all hyaloclastite glasses are degassed,
compatible with formation from subaerial lavas that fragmented on entering the ocean and were
transported by gravity flows down the volcano flank. Most pillows are undegassed, indicating
submarine eruption. The shallowest pillows and most massive lavas are degassed, suggesting formation
by subaerial flows that penetrated the shoreline and flowed some distance under water. Some pillow
rim glasses have H
2
O and S contents indicating degassing but elevated CO
2
contents that correlate
with depth in the core; these tend to be more fractionated and could have formed by mixing of
degassed, fractionated magmas with undegassed magmas during magma chamber overturn or by
resorption of rising CO
2
-rich bubbles by degassed magmas. Intrusive glasses are undegassed and have
CO
2
contents similar to adjacent pillows, indicating intrusion shallow in the volcanic edifice. Cl
correlates weakly with H
2
O and S, suggesting loss during low-pressure degassing, although most
samples appear contaminated by seawater-derived components. F behaves as an involatile incompatible
element. Fractionation trends were modeled using MELTS. Degassed glasses require fractionation at
p
H
2
O
5–10 bars. Undegassed low-SiO
2
glasses require fractionation at p
H
2
O
50 bars. Undegassed
and partially degassed high-SiO
2
glasses can be modeled by coupled crystallization and degassing.
Eruption depths of undegassed pillows can be calculated from their volatile contents assuming vapor
saturation. The amount of subsidence can be determined from the difference between this depth and
the sample’s depth in the core. Assuming subsidence at 2.5 mm/y, the amount of subsidence suggests
ages of
500 ka for samples from the lower 750 m of the core, consistent with radiometric ages. H
2
O
contents of undegassed low-SiO
2
HSDP2 glasses are systematically higher than those of high-SiO
2
glasses, and their H
2
O/K
2
O and H
2
O/Ce ratios are higher than typical tholeiitic pillow rim glasses
from Hawaiian volcanoes.
Components:
23,763 words, 14 figures, 3 tables
.
Keywords:
HSDP; Hawaii; volatiles; glasses; Mauna Kea.
Index Terms:
3640 Mineralogy and Petrology: Igneous petrology; 3670 Mineralogy and Petrology: Minor and trace element
composition; 8414 Volcanology: Eruption mechanisms.
Received
24 June 2003;
Revised
9 April 2004;
Accepted
12 May 2004;
Published
1 September 2004.
G
3
G
3
Geochemistry
Geophysics
Geosystems
Published by AGU and the Geochemical Society
AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES
Geochemistry
Geophysics
Geosystems
Article
Volume 5
, Number 9
1 September 2004
Q09G16, doi:10.1029/2003GC000596
ISSN: 1525-2027
Copyright 2004 by the American Geophysical Union
1 of 42
Seaman, C., S. B. Sherman, M. O. Garcia, M. B. Baker, B. Balta, and E. Stolper (2004), Volatiles in glasses from the HSDP2
drill core,
Geochem. Geophys. Geosyst.
,
5
, Q09G16, doi:10.1029/2003GC000596.
————————————
Theme:
Hawaii Scientific Drilling Project
Guest Editors:
Don DePaolo, Ed Stolper, and Don Thomas
1. Introduction
[
2
] The concentrations of volatile components
such as H
2
O, CO
2
,S,andthehalogensin
volcanic glasses can provide constraints on pet-
rogenesis and the nature of volcanic processes.
For example, H
2
O, S, and Cl contents of glasses
can distinguish between magmas erupted under
subaerial and shallow and deep submarine con-
ditions [e.g.,
Moore and Fabbi
, 1971;
Moore and
Schilling
, 1973;
Killingley and Muenow
, 1975;
Moore and Clague
, 1992;
Garcia and Davis
,
2001;
Davis et al.
, 2003], providing insight into
the eruptive history of individual samples. H
2
O
and CO
2
contents of glasses from submarine and
intrusive glasses can also provide quantitative
constraints on the pressure of eruption or
emplacement [e.g.,
Fine and Stolper
, 1985b;
Dixon
et al.
, 1991;
Newman et al.
, 2000;
Wallace
, 2002],
and the speciation of water can provide information
on thermal histories and the role of low-temperature
hydration in the formation of water-rich glasses
[
Zhang et al.
, 1991;
Dixon et al.
, 1995;
Zhang et
al.
, 1995;
Newman et al.
, 2000]. Also, perhaps most
importantly, for samples that retain volatile con-
tents inherited from melting and other processes in
the mantle, concentrations of H
2
O, CO
2
, etc. can
provide critical information on the role of volatiles
in petrogenesis and the nature of heterogeneity in
mantle sources of basaltic magmas [e.g.,
Michael
,
1988;
Stolper and Newman
, 1994;
Michael
, 1995;
Newman et al.
, 2000;
Dixon and Clague
, 2001;
Dixon et al.
, 2002;
Hauri
, 2002;
Michael and
Kamenetsky
, 2002;
Saal et al.
,2002;
Wallace
,
2002;
Davis et al.
, 2003].
[
3
] In this paper we report concentrations of
H
2
O, CO
2
, S, Cl, and F in a suite of 556 glasses
from the submarine section (i.e., at depths greater
than 1079 meters below sea level, or mbsl)
recovered by the 1999 phase of drilling at Hilo,
Hawaii of the Hawaii Scientific Drilling Project
(referred to hereafter as HSDP2, to distinguish it
from the pilot hole, or HSDP1, drilled in 1993).
These samples are presumed to represent output
from the Mauna Kea volcano [
Hawaii Scientific
Drilling Project
, 2001]. The potential of volatile
components for answering petrological, volcano-
logical, and geochemical questions of the sort
posed above is particularly great for the subma-
rine parts of the HSDP2 section, since fresh glass
is abundant [
Hawaii Scientific Drilling Project
,
2001;
Stolper et al.
, 2004]. Moreover, submarine
glasses from the HSDP2 core represent a detailed
sequence spanning
200 kyr, providing the oppor-
tunity to study the variation of volatiles and what
they can tell us about volcanic processes and petro-
genesis over a considerable fraction of the shield-
building stage of a Hawaiian volcano.
2. Sample Descriptions and Locations
[
4
] The 556 samples studied here were described
by
Stolper et al.
[2004], who report electron
microprobe analyses of 531 of these glasses.
The stratigraphy of the HSDP2 drill core is
summarized by
Hawaii Scientific Drilling Project
[2001]. The drill core penetrated subaerial Mauna
Loa lavas until reaching subaerial Mauna Kea
lavas at
245 mbsl. At
1079 mbsl, the core
passed into submarine Mauna Kea deposits
including hyaloclastites, pillow basalts, intrusive
units, and so-called ‘‘massive’’ basalts. During
core logging at the drill site, units were desig-
nated as ‘‘massive’’ if they could not be defin-
itively identified as either pillows or intrusive
units; thus this designation includes units with a
range of morphologies and characteristics. Nearly
all of the massive units in the core are from
1100–1800 mbsl, and all of those included in
this study are from this depth range.
[
5
] Samples were collected from glass fragments in
hyaloclastites and from pillow margins, intrusive
margins, contacts of massive basalts, and glassy
zones within massive basalts. There are actually
two independent data sets, one from the University
of Hawaii (UH) and one from Caltech, but in the
interests of a coherent presentation, we have
chosen to report them in a single publication. The
UH sample suite was collected on site during
drilling as part of the core logging procedure for
each core box. Samples in this suite span the
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
2of42
1100–3100 mbsl depth range; sampling was
particularly dense (roughly every 3 m) deeper than
1500 mbsl. The Caltech suite was collected at
Caltech during the summer of 2000 as a ‘‘reference
suite’’ for further glass analyses after all core logs
were completed and revised. In addition to regu-
larly spaced samples (roughly every 50 m),
preliminary major element chemistry of whole
rocks [
Rhodes and Vollinger
, 2004] and glasses
(from the UH sample suite) were used as guides for
additional sampling at Caltech to ensure coverage
of major geochemical trends and boundaries.
Additional details on the samples, their collection,
and their preparation are reported by
Stolper et al.
[2004].
3. Petrography
[
6
] Because of the large number of samples studied
in this project, we have limited our petrographic
descriptions primarily to reflected light microscopy
of the polished microprobe mounts of the Caltech
reference suite of glasses. Observations on pheno-
cryst and microphenocryst assemblages and on the
presence/absence of sulfides and vesicles are
reported in Table 1.
[
7
] The glass chips are brown both macroscopi-
cally and in thin section. Alteration rinds occur
both on the exterior surfaces of the glass chips and
less commonly along fractures within chips and on
interior surfaces of some vesicles. The extent of
alteration varies widely. Many samples display no
obvious alteration while other samples have rinds
that are up to 100
m
m thick. In all cases, these
alteration rinds were avoided during electron
microprobe analysis. Although we have not quan-
tified the abundance of altered glass in each sample
(since the chips in each probe mount were selected
based on their glassy appearance and are not
necessarily representative of the larger collection
of chips that make up each sample), visual inspec-
tion of the microprobe mounts suggests that the
extent of alteration roughly increases with increas-
ing depth. In addition, several intrusive glasses and
pillow glasses from near intrusive units have het-
erogeneous and elevated molecular water contents
(see section 5.2), suggesting some localization of
alteration processes.
[
8
] Olivine is the only phenocryst phase (defined
as grains >0.5 mm in longest dimension) in the
reference glasses. Microphenocrysts (0.05–0.5 mm
in longest dimension) include olivine, plagioclase,
augite, and more rarely spinel. Plagioclases in a
few of the samples have high aspect ratios (>10),
and on the basis of length would be classified as
phenocrysts, but given their elongate habit, we
have chosen to classify them as microphenocrysts.
One orthopyroxene grain (
0.5 mm in longest
dimension) partially rimmed by augite is present
in SR0907-2.8, a pillow basalt (unit 321) from
2789.5 mbsl. Orthopyroxene was also found
in SR0714-11.5, an intrusive (unit 263) from
1883.0 mbsl (Mark Kurz, personal communica-
tion). The rarity of orthopyroxene in our sample
suite and the fact that low-pressure MELTS calcu-
lations on selected glass compositions (see
section 7) do not predict orthopyroxene crystalliza-
tion suggests that orthopyroxene is not an important
low-pressure crystallizing phase in these samples.
Spinel inclusions are common in the olivine phe-
nocrysts and microphenocrysts. Sulfides were not
observed in any of the Caltech glasses. However,
four polished mounts of the UH glasses contain
sulfide blebs in the groundmass glass (SR0913-
11.80, SR0914-10.50, SR0916-7.40, SR0967-2.0),
and sulfide blebs are present in olivine-hosted melt
inclusions in another two UH samples (SR0969-
11.30, SR0965-1.40). Note that we only examined
11 of the UH samples for sulfides, chosen because
they were relatively rich in FeO* and S, and in all six
cases where sulfides were present, they are extreme-
ly rare, with only one or two blebs observed on the
polished surface of each sample.
[
9
] Vesicles are observed in all but five of the
polished Caltech glass samples (SR0694-4.9,
SR724-9.6, SR754-9.9, SR771-7.5, and SR0859-
1.0). Vesicles are
10–450
m
m in diameter; some
samples have vesicles spanning a wide size range,
while others have a narrow size distribution. We
measured the diameters of the five largest vesicles
in each sample (or fewer, if there were <5 vesicles
in the sample); the median of these measured
vesicle size for the degassed glasses is similar to
that for the partially degassed and undegassed
glasses (80 versus 90
m
m). Because of the small
surface areas of the mounted glasses, only approx-
imate modal abundances of vesicles (based on
visual estimates) are reported for the Caltech sam-
ples in Table 1; these estimates of vesicularity
range from zero to 5–10% (vesicle abundances
were not estimated for any of the UH samples).
Note that some of these estimates probably under-
estimate the actual vesicularity because vesicles
along the edge of a mounted chip were not includ-
ed in the volume estimate unless the vesicle
appeared to be more than 50% enclosed by glass.
All but one of the degassed samples have <5%
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
3of42
Table 1.
Phenocrysts and Microphenocrysts in Glasses from the Caltech Reference Suite
a
Sample Type Depth, mbsl Phenocrysts (>0.5 mm)
Microphenocrysts (0.05–0.5 mm)
Vesicles Volume Percent
Olivine Plagioclase Clinopyroxene Spinel
SR471-1.1 h
1119.8
ol [sp]
plag
cpx
2
SR485-0.9 h
1138.7
ol [sp]
ol [sp]
plag
cpx
2
SR495-0.9 h
1234.5
ol [sp]
ol [sp]
plag
cpx
2–5
SR508-8.6 h
1283.7
ol [sp] r plag
sp
2
SR523
h
1328.3
ol [sp]
plag
cpx
2
SR544-5.3 h
1391.2
ol [sp] vr plag
2–5
SR561-3.3 h
1436.9
ol [sp]
plag
2
SR595-6.4 h
1523.5
ol [sp]
plag
r cpx
2
SR646/647 h
1653.2
ol [sp]
plag
cpx
2
SR658-0.9 h
1685.8
ol [sp]
ol [sp] r plag
r cpx
2
SR675-8.8 h
1739.4
ol [sp]
plag
cpx
2
SR684-8.95 h
1766.5
ol
cpx
-
SR686-5.1 h
1771.5
ol
cpx
2–5
SR689-2.2 h
1779.4
ol [sp]
cpx
sp
2–5
SR694-4.9 m
1793
ol [sp]
ol [sp]
plag
cpx
2
SR696-8.4 h
1797.3
-
SR697-8.1 h
1800.1
ol [sp]
ol
r plag
cpx
2–5
SR707-11.4 h
1837.7
ol [sp]
plag
cpx
2
SR716-2.9 h
1890.7
ol
plag
cpx
2
SR724-9.6 h
1938.3
ol
plag
cpx
2
SR734-6.3 p
1987.8
ol
ol [sp]
plag
cpx
2
SR740-8.6 p
2003
ol [sp]
plag
r cpx
2
SR747-14.2 p
2045.7
ol [sp]
plag
r cpx
2
SR754-9.9 p
2087
ol [sp]
sp
2
SR763-14.9 p
2131.4
ol [sp] vr plag
2
SR770-2.2 h
2167.1
ol [sp]
plag
cpx
2
SR771-7.5 h
2175.3
ol
plag
cpx
2
SR774-0.6 h
2190.6
ol [sp]
ol [sp]
plag
cpx
2
SR780-20.8 p
2236
ol [sp]
plag
cpx
2
SR792-6.2 p
2285.1
ol [sp]
2
SR807-3.7 p
2340.5
ol [sp]
2
SR822-3.4 p
2382.5
ol [sp]
5–10
SR831-2.3 p
2438.4
ol [sp] r plag
2
SR837-21.1 p
2477.8
ol [sp]
2
SR839-5.5 h
2486.2
ol
plag
cpx
2
SR848-12.0 h
2540.4
ol [sp]
plag
cpx
5–10
SR855-0.6 h
2581.1
ol
plag
cpx
sp
2
SR858-4.1 i
2600.5
ol [sp] vr plag
b
5–10
SR859-0.8 i
2605.3
ol [sp]
5–10
SR859-1.0 p
2605.3
ol
plag
cpx
2
SR866-5.3 p
2639.6
ol [sp]
r cpx
2
SR883-0.3 p
2685.4
ol [sp] r plag
cpx
2
SR892-13.8 h
2735.7
ol [sp]
2
SR907-2.8 p
2789.4
ol
plag
cpx
2
SR916-7.2 p
2838.6
ol [sp]
plag
c
cpx
c
2–5
SR926-2.3 h
2888.3
ol [sp] vr plag
r cpx
2–5
SR933-4.9 p
2930.5
ol [sp]
ol [sp]
plag
cpx
2
SR944-11.3 i
2979.7
ol [sp]
ol [sp]
2–5
SR947-6.1 p
2987.3
ol [sp]
ol [sp]
plag
cpx
sp
2
SR949-8.2 i
2993.7
ol [sp]
ol [sp] vr plag
2
SR961-4.1 p
3037.5
ol [sp]
ol [sp]
sp
2–5
SR970-3.0 p
3078.1
ol [sp]
ol [sp]
vr cpx
2–5
SR970-6.2 i
3079
ol [sp]
2
SR975-0.5 p
3088.3
ol [sp]
2
SR975-2.6 p
3088.9
ol [sp]
ol [sp]
cpx
2
SR979-1.3 p
3095.2
ol [sp]
plag
c
cpx
c
2
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
4of42
vesicles and most have <2%; the only degassed
sample with more than 5% vesicles is SR0848-12.0
with 5–10%. Seven of the partially degassed and
undegassed samples have 2–10% vesicles, while
the remaining 15 samples have <2% vesicles. In
neither the degassed nor the undegassed and par-
tially degassed glasses does vesicle volume corre-
late with measured water content.
4. Analytical Methods
4.1. Infrared Spectroscopy
[
10
] Fourier transform infrared (FTIR) spectro-
scopy was performed on doubly polished glass
chips. Transmission spectra were collected with a
Nicolet Continuum Infrared Microscope connected
to a Nicolet 860 Magna series FTIR using a globar
source, a KBr beamsplitter, and a MCT/A detector.
Most spectra were collected with 128 scans. Glass
chips were 22–300
m
m thick; thickness for each
sample was based on the average of three measure-
ments taken on the same spot before each analysis;
1
s
on the thickness measurement was typically
less than 1
m
m. Glass density was assumed to be
2800 g/L [
Dixon et al.
, 1991].
[
11
] The concentration of total dissolved water
(i.e., dissolved as hydroxyl groups plus molec-
ular water) was calculated from the intensity of
the absorption band at
3550 cm
1
. The molar
absorptivity of this band was taken to be 63 ±
5 L/mol cm (see summary by
Newman et al.
[2000]) and assumed to be compositionally inde-
pendent over the range of compositions in this
study. The background for the 3550 band was
assumed to be linear and fixed by the spectrum at
3740 and
2510 cm
1
.
[
12
]
Dixon et al.
[1995] found that concentrations
of molecular water calculated using the 1630 cm
1
and 5200 cm
1
absorption bands are indistinguish-
able within error. Therefore, given the higher
sensitivity of the 1630 cm
1
band and the low
concentrations of water in most of our samples,
the 1630 cm
1
band was used to determine
molecular water contents. Except for three anoma-
lous intrusive samples with elevated molecular
water contents and detectable molecular CO
2
(see
section 5.2), all carbon dioxide in the HSDP2
glasses is dissolved as carbonate, manifested by a
doublet in the absorption spectrum with maxima at
1515 and 1435 cm
1
[
Fine and Stolper
, 1985b],
and the intensity of this doublet was used to
quantify dissolved CO
2
contents of the glasses.
To determine the intensities of these bands, we first
subtracted the spectrum of a volatile-poor pillow
rim glass (SR754-9.9) scaled to the same thickness
as the unknown. Each background-subtracted spec-
trum in the region 1800–1400 cm
1
was then
modeled as the sum of a 1630 cm
1
molecular
H
2
O band and a 1515–1435 cm
1
carbonate
doublet; the absorbances of the molecular H
2
O
and carbonate bands were the coefficients of these
bands in the best fit model spectrum. Concentra-
tions of molecular water were calculated using a
value of 25 ± 2 L/mol cm for the molar absorptivity
of the 1630 cm
1
band; this is the average (and 1
s
)
of values calculated for the HSDP2 glasses using
the equations of
Dixon et al.
[1995]. Carbonate
concentrations were calculated using a value of
355 ± 5 L/mol cm, based on values calculated for
the HSDP2 glasses using the results of
Dixon and
Pan
[1995]. The detection limit for CO
2
depends
upon sample thickness, the quality of the spectrum,
and the details of the background subtraction pro-
cedures, but it is typically 20–50 ppm; because of
the same factors, analytical precision can degrade to
several tens of percent at concentrations below 50–
100 ppm [
Fine and Stolper
, 1985b].
[
13
] Results of the infrared analyses are listed in
Table 1 of
Stolper et al.
[2004]. Total water
contents are reported for 172 samples; molecular
water contents are reported for 91 and carbon
dioxide contents for 77 of these samples. Multiple
spectra (2–5) were obtained for each sample;
analyses and uncertainties reported in Table 1 of
Stolper et al.
[2004] and shown in the figures in
this paper are the mean and 1
s
of the distribution
Notes to Table 1.
a
Note that the presence/absence of phases is based on examination of small glass chips; small surface areas precluded point counting. Immiscible
sulfide blebs were not observed in any of the CIT glass samples; sulfide blebs were observed in the matrix glass in UH samples SR0913-11.8,
SR0914-10.5, SR0916-7.4, SR0967-2.0; UH samples SR0969-11.3 and SR0965-1.4 had sulfide blebs in olivine-hosted melt inclusions. Vesicle
volume percent was estimated visually in three ranges:
2%, >2% and
5%, and >5% and
10%; dash means no vesicles observed. Abbreviations
are as follows: ol, olivine; plag, plagioclase; cpx, clinopyroxene; sp, spinel; [sp], spinel present as an inclusion; r, rare; vr, very rare; h, hyalo
clastite;
i, intrusive; m, massive; p, pillow.
b
Located in one small circular area at the edge of one chip.
c
Only present in one chip (out of the five in mount SR916-7.2 and out of the six in mount SR979-1.3); the composition of the glass in the
ol+plag+cpx-bearing chip in SR979-1.3 is more evolved than the glasses in the other chips in the same mount [see
Stolper et al.
, 2004, Table 1]; the
ol+plag+cpx-bearing chip in SR916-7.2 was not analyzed.
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
5of42
of these replicate analyses. The average uncertain-
ties for the FTIR determinations of water content,
molecular water content, and carbon dioxide dis-
solved as carbonate are 0.014 wt %, 0.09 wt %, and
8 ppm, respectively.
4.2. Ion Microprobe
[
14
] Because of the difficulties of making precise
FTIR measurements at the low-CO
2
contents of the
HSDP2 glasses (all are <120 ppm based on FTIR
measurements), we undertook a comparison
between FTIR and ion microprobe measurements
using our sample suite. Ion microprobe measure-
ments were made with a modified Cameca 3f ion
probe at the Lawrence Livermore National Labo-
ratory using a 2–3 nA beam of Cs
+
ions acceler-
ated at 10 kV; an electron flood gun was used to
prevent positive charging of the sample surfaces.
Sputtered ions were accelerated at a nominal volt-
age of 4.5 kV through a double-focusing mass
spectrometer and
12
C
+
ions were measured with
an electron multiplier. A zero energy offset was
used for all analyses, and the mass-resolving power
was 1100.
[
15
]Measured
12
C
+
/
30
Si
+
ratios in the HSDP2
glasses were converted to wt % CO
2
with a
calibration curve constructed using two natural
basaltic glasses (I. D. Hutcheon, personal commu-
nication, 2001) and the basaltic run products from
two moderate-pressure experiments [
Stolper and
Holloway
, 1988]. The CO
2
and SiO
2
concentra-
tions for the four standard glasses are in the ranges
29–550 ppm and 50.53–51.05 wt %. The stand-
ards were analyzed each day over a 6-day period
and the calibration curve consists of 45 analyses
with a correlation coefficient (R) of 0.9985. The
mean deviation between the 45 standard analyses
and the values predicted for these standards by the
calibration is 5.2% relative. Considering only the
three standards with the highest CO
2
contents
(152–550 ppm), the mean deviation drops to
2.6%; for the standard glass with the lowest CO
2
content, the mean deviation is
12%. Two Kilauea
basaltic glasses with 50.6 and 51.7 wt % SiO
2
(I. D.
Hutcheon, personal communication, 2001) were
also analyzed as secondary standards and the mean
ion probe CO
2
values (138 ± 7, 254 ± 7) overlap
with those determined by FTIR (117 ± 20, 278 ±
20) at the 1
s
level.
[
16
] In addition to the four primary and two
secondary glass standards discussed above, six
experimentally produced glasses quenched from
CO
2
-saturated liquids (containing up to
400 ppm
CO
2
), and 35 glass samples from the Caltech refer-
ence suite were analyzed for CO
2
by ion microprobe
(based on the calibration curve, four of the reference
glasses had CO
2
contents less than zero). The results
of these measurements are shown in Figure 1 and
tabulated by
Stolper et al.
[2004]. Although there is
an overall correlation between the two methods (R =
0.964), the best fit line does not pass through the
origin: i.e., for detectable concentrations <120 ppm
CO
2
(based on FTIR measurements), the concen-
trations determined by FTIR are systematically
lower than the ion probe determinations, suggesting
a problem with background corrections for one or
both techniques at low concentrations. Despite these
differences, both data sets show similar trends at the
level of interpretation that we present below. Given
this similarity and that CO
2
solubility models in the
literature have been calibrated using FTIR measure-
ments [
Dixon and Stolper
, 1995;
Dixon
, 1997],
that the CO
2
contents of most basaltic glasses in
the literature have been measured by FTIR
[e.g.,
Fine and Stolper
, 1985b;
Dixon et al.
, 1988,
1991;
Stolper and Newman
, 1994;
Dixon et al.
,
1997;
Newman et al.
, 2000;
Dixon and Clague
,
2001;
Wallace
, 2002;
Davis et al.
, 2003], and that
we have a larger number of FTIR measurements of
CO
2
contents than ion microprobe measurements,
we restrict our discussion of CO
2
contents in the
following sections to the FTIR measurements.
4.3. Electron Microprobe
[
17
] Major element and S concentrations of 531 of
the glasses were determined by electron micro-
probe analysis at the University of Hawaii or
Caltech; F and Cl concentrations were determined
at the University of Hawaii for 99 of these samples.
Analytical procedures and results are reported by
Stolper et al.
[2004].
5. Results
[
18
]H
2
O, S, CO
2
, Cl, and F contents of the HSDP2
glasses are listed in Table 1 of
Stolper et al.
[2004]
and shown as functions of depth in the drill core in
Figures 2a–2e and versus SiO
2
in Figure 3. Results
on the speciation of water in the HSDP2 glasses are
shown in Figure 4.
[
19
] All samples are from the submarine part of the
section (i.e., deeper than 1079 mbsl). Following
Stolper et al.
[2004], we have divided the samples
into a high-SiO
2
group (
50%; all concentrations
are in weight percent, unless otherwise indicated)
and a low-SiO
2
group (<50%), and they are differ-
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
6of42
entiated by color in all figures. Glasses from each
rock type (i.e., hyaloclastites, pillows, intrusive
units, massive lavas) are shown as distinctive
symbols in most figures. On the basis of the
distributions of the high- and low-SiO
2
glasses
with depth,
Stolper et al.
[2004] divided the
submarine part of the core into the four zones
shown in Figures 2a–2e.
[
20
] Glasses are referred to in this paper as either
‘‘undegassed’’ (H
2
O
0.45% and S
0.09%),
‘‘partially degassed’’ (0.45 > H
2
O
0.21%; 0.09 >
S
0.04%), or ‘‘degassed’’ (H
2
O < 0.21% and S <
0.04%); this classification is based largely on
concentrations in subaerial versus submarine
glasses [
Moore and Clague
,1992;
Garcia and
Davis
, 2001;
Sherman et al.
, 2002]. It is well
known that H
2
O and S degas significantly from
Hawaiian liquids only at relatively low total pres-
sures such as those pertaining to subaerial or
shallow submarine environments or to high-level
magma chambers connected to the atmosphere.
The point of this classification is thus that the
‘‘degassed’’ samples lost water and sulfur at pres-
sures approaching atmospheric, whereas the
‘‘undegassed’’ samples did not experience total
pressures of less than
40–50 bars for sufficient
time to vesiculate significantly [
Stolper et al.
,
2004].
[
21
] In this section, we describe the volatile con-
centrations, their relationship to rock type, and
their variations with chemical composition. In
section 6, we describe the variations in volatile
contents with depth in the core.
5.1. Water and Sulfur Contents
[
22
] Figures 3a and 3b show SiO
2
contents
versus H
2
O and S contents of the analyzed
glasses. Water contents are 0.06–3.85% and S
contents are 0.001–0.15%. The undegassed and
partially degassed glass samples define a low-
SiO
2
group (48.3–49.8% SiO
2
) and a high-SiO
2
group (50.5–52.0% SiO
2
) separated by a gap in
SiO
2
in which there are no undegassed or
partially degassed samples. Degassed samples
span the full range of SiO
2
contents, filling in
the gap in SiO
2
content defined by the unde-
gassed samples. There is also a relationship
between water content and MgO content, such
that the more fractionated samples (i.e., with
lower MgO contents) tend to be degassed.
Although there is not a one-to-one correlation,
the overall relationship is apparent in Figures 3a
and 3b, in which samples with
7% MgO
(shown by solid symbols) are concentrated
among the undegassed samples, and samples
with <7% MgO (shown by open symbols) are
concentrated among the degassed samples. This
relationship between degassing and fractionation
is developed in more detail below (see section
8.1). Finally, Figures 3a and 3b show that H
2
O
and S contents are bimodally distributed: i.e.,
there are few partially degassed glasses, a feature
also observed for S in submarine glasses from
other Hawaiian volcanoes [e.g.,
Clague et al.
,
2002;
Johnson et al.
, 2002;
Sherman et al.
,
2002;
Shinozaki et al.
, 2002;
Davis et al.
,
2003]. This reflects primarily the relatively nar-
row depth interval over which samples partially
degas; that is, only a small fraction of the depth
interval of the core corresponds to the pressures
at which submarine magmas would partially
degas, so most samples erupted either subaerially
(degassing fully) or deeper than
500 mbsl
(without degassing).
Figure 1.
Comparison of CO
2
concentrations in
basaltic glasses measured by FTIR and ion microprobe.
Open circles are HSDP2 glasses, solid circles are
experimentally produced glasses quenched from CO
2
-
saturated basaltic melts, and red diamonds are natural
Kilauea basaltic glasses; the two Kilauea glasses were
analyzed as secondary standards. Error bars represent 1
s
uncertainties based on replicate FTIR and ion probe
measurements and where not shown are smaller than the
size of the symbols. The weighted least squares fit
(dashed line) has the equation: FTIR CO
2
= 1.136(49)
(ion probe CO
2
)
30.742(6.196) and was determined
using the formulation of
Reed
[1992] that incorporates
errors on both the x and y coordinates. Parentheses
enclose 1
s
uncertainties on the slope and y intercept in
terms of the least units cited. The two samples with zero
CO
2
as determined by FTIR were not included in the fit.
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
7of42
[
23
] Although the distinction in terms of silica
content between the high- and low-silica groups
may at first appear arbitrary because there is a
continuous (though bimodal) distribution of silica
contents (and other major and minor elements
[
Stolper et al.
, 2004]), these groups are also dis-
tinguishable using several isotopic and incompati-
ble trace element ratios, so the formation of the
high- and low-SiO
2
magmatic groups must have
involved compositionally distinguishable mantle
sources. Moreover, the observations presented in
the preceding paragraph led
Stolper et al.
[2004] to
propose that there are two dominant, nonoverlap-
ping magma series for the HSDP2 rocks, not a
continuum, because among the undegassed glasses
there are unambiguously two distinct groups of
Figure 2.
(a) H
2
O, (b) S, (c) CO
2
, (d) Cl, and (e) F concentrations versus depth for HSDP2 glasses. Blue symbols
are samples with SiO
2
< 50% (‘‘low-SiO
2
’’); pink samples have SiO
2
50% (‘‘high-SiO
2
’’); green symbols are
glasses from the 2233–2280 mbsl excursion at the top of zone 3 [
Stolper et al.
, 2004]; cyan symbols are high-CaO-
K
2
O glasses from zone 1 [
Stolper et al.
, 2004]. Symbols with orange dots indicate samples designated as altered (see
text and Table 1 of
Stolper et al.
[2004]). Zone boundaries [
Stolper et al.
, 2004] are marked with solid lines, and
excursion boundaries (see text and
Stolper et al.
[2004]) are shown with dashed lines except when they coincide with
zone boundaries. Error bars in Figures 2a and 2c are 1
s
based on replicate analyses of each sample. Note that two
intrusive glasses (SR944-11.6, SR957-4.1) with >1.2% H
2
O are not shown in Figure 2a.
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
8of42
glasses. They further proposed that most glasses
with intermediate silica contents were produced by
mixing of the high-SiO
2
and low-SiO
2
magmas.
Finally, the observations that glasses with interme-
diate silica contents are always degassed and that
degassed glasses tend to be more fractionated
suggest that the site(s) of mixing are at high levels
in the volcanic edifice where pressures are low
enough for significant degassing to occur and that
mixing, degassing, and crystallization are connected
in the evolution of the Mauna Kea magmas sampled
in the HSDP2 core. The possible nature of this
connection is developed in section 8.1. Note that
there is a group of low-SiO
2
glasses identified by
Stolper et al.
[2004] with slightly elevated CaO,
Al
2
O
3
, and SiO
2
contents; these are all degassed and
appear to be the dominant low-SiO
2
mixing end
member in samples from the shallower parts of the
core.
[
24
] As shown in Figure 5a, H
2
OandSare
positively correlated. This is not surprising since
they are known to degas over similar relatively
low-pressure intervals [e.g.,
Dixon et al.
, 1991;
Moore and Clague
, 1992;
Garcia and Davis
, 2001;
Hauri
, 2002;
Davis et al.
, 2003]. In detail, however,
the high- and low-SiO
2
groups are distinguished in
Figure 5. The high-SiO
2
glasses form a continuous
linear array with roughly constant H
2
O/S ratio of
4–5, presumably representing a degassing trend.
The low-SiO
2
glasses are, however, mostly unde-
gassed, with a higher H
2
O/S ratio of
5–7.
Figures 3a and 5a show that primitive, undegassed
low-SiO
2
and high-SiO
2
glasses have distinguish-
able average H
2
O contents: excluding the intrusive
and pillow rim glasses suspected of alteration (see
section 5.2), the 70 undegassed low-SiO
2
glasses
with MgO
7% have 0.55–0.85% H
2
O (averag-
ing 0.67 ± 0.06% {1
s
}), whereas the 33 unde-
gassed high-SiO
2
glasses with MgO
7%
have 0.46–0.64% H
2
O (averaging 0.52 ± 0.05%
{1
s
}). The S contents of the two groups are not
as readily distinguished, with the undegassed,
MgO-rich, unaltered low-SiO
2
glasses having
0.91–0.119% S (averaging 0.104 ± 0.006%
{1
s
}) and the equivalent high-SiO
2
glasses
having 0.092–0.133% (averaging 0.114 ±
0.012% {1
s
}). However, as shown in Figure 6a,
the FeO* versus S relations of the HSDP 2glasses
suggest that the MgO-rich high-SiO
2
glasses are
indeed typically slightly richer in S than MgO-
rich low-SiO
2
glasses; moreover, petrographic
Figure 2.
(continued)
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
9of42
observations (Figure 6b) and comparison with
the observed conditions of sulfide saturation for
mid-ocean ridge basalts (Figure 6) suggest that
the most S-rich high-SiO
2
glasses are sulfide
saturated, whereas all low-SiO
2
glasses are
undersaturated with respect to an immiscible
sulfide liquid.
[
25
] In addition to the relationships between chem-
ical composition and H
2
O and S contents, there are
also patterns in volatile contents with respect to
rock type.
5.1.1. Hyaloclastites
[
26
] Nearly all of the hyaloclastite glasses are
degassed (i.e., 95% of 261 samples have 0.06–
0.16% H
2
O or 0.001–0.037% S). This suggests
that they represent subaerial or shallow submarine
eruptions that vitrified and fragmented on reaching
the shoreline and that their presence in the HSDP2
core reflects subsequent slumping of oversteepened
near-shore fragmental deposits [
Hawaii Scientific
Drilling Project
, 2001]. This interpretation is con-
sistent with the presence of highly vesicular lithic
clasts and the occurrence of charcoal (although this
Figure 3.
(a) H
2
O, (b) S, (c) Cl, and (d) F versus SiO
2
contents for HSDP2 glasses. Symbols are as in Figures 2a–
2e. Red horizontal lines divide undegassed, partially degassed, and degassed samples (see text and
Moore and Clague
[1992],
Garcia and Davis
[2001], and
Stolper et al.
[2004]). The dashed, black horizontal line in Figure 3c arbitrarily
distinguishes ‘‘contaminated’’ samples (>0.03% Cl). Note that one intrusive glass (SR957-4.1) with 3.85% H
2
O is not
shown in Figure 3a.
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
10 of 42
is rare) in the hyaloclastites [
Hawaii Scientific
Drilling Project
, 2001], both of which indicate
derivation from near-shore environments. A small
number of hyaloclastites (all from >2460 mbsl) have
elevated H
2
O contents (two samples from zone 3,
SR0837-21.0 and -21.1, both at 2477.9 mbsl,
0.55%; one sample from zone 4, SR0921-14.10,
2860.9 mbsl, 0.67%) and/or S contents (14 samples:
0.062–0.129%), perhaps indicating the presence of
deepwater hyaloclastites at depth in the core [e.g.,
Smith and Batiza
, 1989;
Maicher et al.
, 2000;
Head
and Wilson
, 2003].
[
27
]
Stolper et al.
[2004] emphasized the occur-
rence of high-CaO -K
2
O, low-SiO
2
hyaloclastite
glasses at
1765–1810 mbsl. Figure 2b shows that
these glasses are all degassed and that they are
lower in S than the high-SiO
2
glasses from around
this depth interval. We do not currently have an
explanation for the unusually low-S contents of the
low-SiO
2
glasses from this depth interval.
5.1.2. Massive Lavas
[
28
] Ten of the 11 glasses we analyzed from the
massive flows are degassed (0.07–0.12% H
2
O;
0.014–0.021% S). Given that most massive units
in the core are from shallower than 1800 mbsl in
the submarine section and most of those we ana-
lyzed are substantially degassed, we suggest that
most of the massive units represent subaerial flows
that penetrated the subaerial-submarine boundary
at the shoreline and flowed, perhaps in tubes, some
distance under water down the flank of the volcano
[
Moore et al.
, 1973;
Tribble
, 1991;
Garcia and
Davis
, 2001;
Hawaii Scientific Drilling Project
,
2001;
Davis et al.
, 2003].
[
29
] The single undegassed massive glass is sample
SR0508-8.00 at 1283.5 mbsl (0.46% H
2
O) from
the base of unit 191, a thick (>23 m), olivine-rich
(
30% phenocrysts) unit. There are no internal
Figure 4.
(a) Speciation of water in HSDP2 glasses
based on FTIR measurements. The black curve in all
panels is based on experiments on a MORB composi-
tion [
Dixon et al.
, 1995]. (a) HSDP2 glasses only.
Symbols and error bars are as in Figures 2a–2e, except
all samples (regardless of MgO content) are solid, and
symbols enclosed in red circles are glasses from within
1 m of an intrusion. Labeled samples with excess
molecular water are discussed in the text. (b) Same as
Figure 4a but at an expanded scale and other Hawaiian
glasses are shown for comparison [
Dixon et al.
, 1997;
Dixon and Clague
, 2001;
Davis et al.
, 2003]. Where
shown, error bars on data points from the literature are
from the original sources. (c) Same as Figure 4b but
with the scale expanded even further and with water-
rich back arc basin glasses shown for comparison
[
Stolper and Newman
, 1994;
Newman et al.
, 2000;
S. Newman, unpublished back-arc basin glass analyses,
2003].
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
11 of 42
Figure 5
Geochemistry
Geophysics
Geosystems
G
3
G
3
seaman et al.: volatiles from the hsdp2 drill core
10.1029/2003GC000596
12 of 42