xfen
2, F
, W
SA
Keywords:
Obsidian
ra?
ow
AN
n g
ic i
lass
e s
We conclude that the most likely reducing agent is hydrogen, produced by magnetite crystallization within
obsidians dramatically capture redox disequilibrium on the micoscale and
hydrous ?uid liberation and late-stage crystallization to the redox signature of
es in m
drogen
the con
rred to
Chemical Geology 268 (2009) 272?280
Contents lists available at ScienceDirect
Chemical
l se(e.g., Geschwind and Rutherford, 1995) and melt stiffening (Hess and
Dingwell, 1996), as magma degasses. The relationship between water
and the ratio of ferric to ferrous iron in silicatemelts remains an active
area of research. In H2-buffered systems, the dissociation of water
(e.g., H2O?H2+1/2O2, Mueller, 1971) controls oxygen fugacity
(fO2); thus aH2O controls Fe3+/Fe2+. The effect of water as a chemical
component at ?xed P, T, and fO2 is more nuanced. The study of (Baker
and Rutherford, 1996) suggested that the Fe3+/Fe2+ ratio increased
with the addition of water; however, as discussed by Wilke et al.
(2005) and Gaillard et al. (2001), the study of Baker and Rutherford
Most natural magmas crystallize and vesiculate during their
ascent towards and emplacement at the Earth's surface. Because
these phase transformations redistribute and ultimately release
volatile components from the system, they can drive oxidation?
reduction (redox) reactions that will alter the speciation of iron in
the magma. Crystallization of silicate minerals that preferentially
incorporate ferrous iron in their structures, such as olivine and
pyroxene, result in a relatively oxidized melt residuum. The manner
by which crystallization of non-ferrous minerals such as quartz and
feldspar affects the residual melt redox state is currently unknown;was not at constant oxygen fugacity (fO2). E
rhyolite (Moore et al., 1995) and basalt (Bo
indicate that water, as a chemical component
P, and oxygen fugacity (fO2), has no effect on
? Corresponding author. Tel.: +1 202 633 1810; fax:
E-mail address: jonathan.castro@cnrs-orleans.fr (J.M
0009-2541/$ ? see front matter ? 2009 Elsevier B.V. A
doi:10.1016/j.chemgeo.2009.09.006collectively as ?water,?
mineral liquidus curves
sulting in crystallization
(e.g., Mathez, 1984; Burgisser and Scaillet, 2007) or deciphering
the oxidation states of magma source regions (e.g., Carmichael,
1991).dissolved in magma govern the positions of
andmelt viscosities; both rise dramatically, reXANES
1. Introduction
The chemical and physical process
the concentration and reaction of hy
the silicate melt. It is well known that
H2O and OH? groups, hereafter refeagma are in?uenced by
-bearing components in
centrations of molecular
contrast, other experimental studies indicate that as the bulk water
content in increases so too does the Fe3+/Fe2+ in the melt (Gaillard
et al., 2001; Gaillard et al., 2003b). The relationship between water
and iron redox state in natural magmas is thus an important factor
to consider when reconstructing redox history of degassing magmaxperimental studies on
tcharnikov et al., 2005)
in a melt held at ?xed T,
the Fe3+/Fe2+ ratio. By
however, becau
redistribute wat
2009), thereby
(Gaillard et al., 2
The release
outward diffusio
to oxidize lavas (
Candela, 1986;H
+1 202 357 2476.
. Castro).
ll rights reserved.? 2009 Elsevier B.V. All rights reserved.
FTIR glassy lavas.Spherulite
Oxidation?reduction
the spherulite. The Kra?a
highlight the importance ofSpherulite crystallization induces Fe-redo
Jonathan M. Castro a,?, Elizabeth Cottrell b, Hugh Tuf
a ISTO, UMR 6113 Universit? d'Orl?ans-CNRS, 1a rue de la F?rollerie, 45071 Orl?ans cedex
b Department of Mineral Sciences, Smithsonian Institution, 10th and Constitution Ave. NW
c Department of Environmental Science, Lancaster University, LA1 4YQ, UK
d Graduate School of Oceanography, University of Rhode Island, Narragansett, RI 02882, U
a b s t r a c ta r t i c l e i n f o
Article history:
Received 9 March 2009
Received in revised form 2 September 2009
Accepted 7 September 2009
Editor: D.B. Dingwell
Rhyolitic obsidians from K
liberated during crystal gr
synchrotron ?-FTIR and ?-X
distinct colorless and brow
groups and depleted in ferr
between brown and clear g
front that originated from th
j ourna l homepage: www.eredistribution in silicic melt
c, Amelia V. Logan b, Katherine A. Kelley c,d
rance
ashington, DC 20560, USA
a volcano, Iceland, record the interaction between mobile hydrous species
th and the reduction of ferric iron in the silicate melt. We performed
ES measurements along a transect extending from a spherulite into optically
lass zones. Measurements show that the colorless glass is enriched in OH
ron, while the brown glass shows the opposite relationship. The color shift
is sharp, suggesting that the colorless glass zone was produced by a redox
pherulite margin and moved through surrounding melt during crystallization.
Geology
v ie r.com/ locate /chemgeose these phases are anhydrous, their growth must
er in the melt (Castro et al., 2008; Watkins et al.,
in?uencing the chemical environment of iron
003c).
of H2 from magma, either through its continuous
n or by liberation of H2 gas in bubbles is widely thought
e.g., Sato and Wright, 1966; Sato, 1978; Mathez, 1984;
olloway, 2004). The diffusive transport of hydrogenout
of themelt (m) and into bubbles (v) displaces the following equilibrium
to the right: 2FeO(m)+H2O(m)=H2(v)+Fe2O3(m). Crystallization of
magnetite may similarly alter the oxidation state of basaltic and silicic
systems by the ?auto-oxidation? reaction as de?ned by Holloway
(2004):
3Fe2+ O
?m? + H2O?m? = H2?fluid? + Fe
2+ O?Fe3+2 O3?magnetite? ?1?
The net effect of this reaction on system redox will be determined
by the degree to which H2 gas exits the system. If the hydrogen
escapes the magmatic system completely (e.g. to bubbles or the
atmosphere), the system is left relatively oxidized. If the hydrogen is
retained within the system, however, it may generate relatively
reducing ?uids adjacent to the zone of crystallization as proposed by
Holloway (2004).
Here we describe adjacent zones of oxidation and reduction on a
scale of only hundreds of microns in Kra?a obsidians. Our observa-
tions and measurements link spherulitic plagioclase, quartz, and
magnetite crystallization to reduction of ferric iron in the rhyolitic
melt (Fig. 1). We show that reduced, colorless glass rims jacketing
spherulites could have been produced by the expulsion of molecular
water from growing spherulites, followed by crystallization of
magnetite and concomitant production of hydrogen. The result is a
boundary layer of reduced melt that grows with time. Given the
considerable sizes (>1 m; Smith et al., 2001) and high volume
proportions of spherulites in many rhyolitic lavas and ignimbrites
(~90 vol.%; Stevenson et al., 1994; Tuffen and Castro, 2009; Tuffen
et al., in review) the interplay of crystallization and microscale
adjustment of oxidation states has important implications for the
post-emplacement chemical evolution of spherulite-rich lavas.
2. Geological background
Spherulites are radiating, often concentrically arranged crystalline
aggregates set in a glassy matrix (Fig. 1). They occur in obsidian
domes, vitrophyric ash-?ow tuffs (e.g., Smith et al. 2001), large-
volume rhyolite ?ows such as those at Yellowstone (e.g., Wright,
1915), and in shallow volcanic conduits (e.g., Stasiuk et al., 1993;
Tuffen and Castro, 2009). Spherulites nucleate and grow in response
to large undercoolings (>200 ?C) rapidly imposed on the magma by
its degassing and quenching (e.g., Swanson et al., 1989). As dictated
by the thermal pro?le of amagma body (Manley, 1992; Tuffen et al., in
review), spherulitic obsidian develops in spatially restricted zones
(e.g., Manley and Fink, 1987; Stevenson et al. 1994), comprising a
transitional facies that separates the rapidly quenched, outermost
vitrophyric rhyolite from a devitri?ed microcrystalline core.
Anomalously high volatile contents exist within and just above the
spherulitic zones in lava domes (e.g., Westrich et al. 1988). Castro
et al. (2008) have recently shown that the OH? concentrations in
glass around spherulites are elevated above the background level.
ario
py)
bac
d alo
273J.M. Castro et al. / Chemical Geology 268 (2009) 272?280Fig. 1. A)Photomicrographof spherulites (roundandelliptical, black) inobsidian (brown).V
Fourier Transform Infrared Spectroscopy; XANES= X-ray absorption near-edge spectrosco
swaths are tracks left by a laser ablation ICPMS (these data are not discussed here). B) A
clinopyroxene andmagnetite. C) AnOH? concentration pro?le determined by FTIR collecte
the colorless?brown glass boundary. (For interpretation of the references to colour in this ?guusanalyticalpoints and traverses are shown(emp=electronprobemicroanalyzer; FTIR=
. Black diamonds indicate approximate locations of ?-XANESmeasurements. Broader dark
kscattered electron image showing typical internal spherulite texture. Bright phases are
ng the lower right traverse in frame A. The vertical grey line is the approximate location of
re legend, the reader is referred to the web version of this article.)
274 J.M. Castro et al. / Chemical Geology 268 (2009) 272?280They interpreted these OH? concentration gradients to re?ect the
combined advection and diffusion of H2O away from the growing
spherulites and numerically modeled these processes in order to
estimate crystallization timescales. While their work con?rms that
spherulite crystallization drives volatile enrichment in silicic glass, it
neither identi?es the speci?c form of hydrous species ejected from the
growing spherulite, nor constrains why an iron redox shift, mani-
fested as a sharp color difference in the obsidian matrix (Fig. 1), is
superimposed on the OH? concentration gradient in the glass. In this
paper, we use microscopic chemical and textural evidence collected
from the same sample studied by Castro et al. (2008) to demonstrate
that regions of elevated OH? are linked to zones of iron reduction in
the glass surrounding spherulites. This in turn suggests that spherulite
growth may cause changes in the Fe-valence state in rhyolite melt.
3. Analytical and experimental techniques
All analyses were made on a decimeter-sized obsidian sample
collected from the Hrafntinnuhryggur ridge system on Kra?a
volcano, Iceland (Tuffen and Castro, 2009). Spherulite mineralogy
was determined by 1) microscopic observation, 2) sample magne-
tism to identify Fe-oxides as magnetite, and 3) compositional data
from energy dispersive spectra (EDS) collected on a FEI NOVA
nanoSEM600 FEG Variable Pressure Scanning Electron Microscope at
the Smithsonian Institution National Museum of Natural History,
operated at 7?12 keV, 5 mm working distance and beam current
ranging from 0.5 to 1 nA.
Major element glass compositions were analyzed using a JEOL JXA-
8900R electron microprobe (EPMA) running software with ZAF cor-
rections at the Smithsonian National Museum of Natural History.
Analyses were performed with an acceleration voltage of 15 keV, a
10 ?mbeam, and a 10 nAbeamcurrent. Standardizationwas performed
on the following natural mineral standards: Quartz (Si), Anorthite (Ca),
Bytownite (Al), Microcline (K), Albite (Na), and Hornblende (Fe, Mg). A
natural rhyolitic glass (VG568, Obsidian Cliffs, Yellowstone, USA) of
known major element composition was periodically analyzed to check
for instrument drift.
Determination of ferrous iron content was via wet chemical analysis
of a powdered (grain size~180 ?m) obsidian aliquot chipped from the
same mass of sample on which all other measurements and observa-
tions were made. We attempted to separate out spherulite fragments
from the glass powder. However, some glass-encrusted spherulitesmay
have been over looked. Furthermore, it was not possible to separate the
thin, clear glass halos from the pervasive brown glass. Thus, these
measurements provide an average FeO concentration for the bulk
brown glass with very minor dilution by the clear glass fraction. A total
of nine (9) analyses were performed following the technique of Peck
(1964) with some minor modi?cations in order to minimize oxidation
during sample digestion. Digestion began in 8 ml of ?uoroboric acid
(HBF4) for 30 min in the ultrasonic bath. We then added 5 ml HF and
2 ml extra of HBF4 to the solution and completed the digestion under
heat for approximately 10 min. Three ferrous iron determinations were
also made on the U.S.G.S. Glass Mountain Rhyolite standard RGM-1,
which yielded a mean value of 1.19 wt.% FeO?0.03 (1?). The nominal
value for RGM-1 is 1.18 wt.% FeO.
H2O concentrations were determined by synchrotron-FTIR at the
Advanced Light Source, Lawrence Berkeley National Laboratory.
Measurements were made along traverses oriented perpendicular to
the spherulite?glass boundaries on a Thermo Nicolet Magna 760 FTIR
spectrometer interfaced with a NicPlan IR microscope (at beamline
1.4.3). The IRbeamhas a diffraction-limiteddiameter of about 3 ?m. The
uncertainty in spot position is ?2 ?m. Transmittance spectra were
obtained over the mid-IR (1400?4000 cm?1) to the near-IR (3700?
6500 cm?1) regions with MCT detectors, KBr beam-splitters, and the
synchrotron light source. 128 scans were used to obtain each spectrum
and these spectrawere corrected by subtracting a background spectrumcollected every hour. We determined OH? concentrations from the
intensity of the broad 3570 cm?1 absorption band, utilizing an
absorption coef?cient of 100 Lmolcm?1 (Newman et al. 1986). We
estimate the analytical uncertainty of OH? concentration to be?10% of
the measured value.
The oxidation state of Fe in the glass was determined at the micro-
scale using Fe K-edge X-ray Absorption Near Edge Structure (?-XANES)
spectroscopy. The area-weighted average energy of the two pre-edge
peaks, or centroid, shifts to higher energy as the ratio of ferric to ferrous
iron increases, allowing quanti?cation of Fe3+/?Fe in silicate glasses
(e.g. Berry et al., 2003, Wilke et al., 2005, Cottrell et al., 2009).
Commensurate with this, the intensity of the peak corresponding to
Fe3+ (at higher energy) grows proportionately larger relative to the
intensity of the peak corresponding to Fe2+ (at lower energy); thus the
ratio of peak intensities can also be used to quantify oxidation state
(Wilke et al., 2005, Cottrell et al., 2009).
Spectra were collected in ?uorescence mode using a 9 element Ge
array detector and a silicon channel-cut (311) monochromator at
station X26A (bending magnet) at the National Synchrotron Light
Source (NSLS), Brookhaven National Lab. The spot size on the sample
was 9?5 ?m. Spectra were recorded from 7020 to 7220 eV with a
0.1 eV step over the pre-edge from 7106 to 7118 eV at 5 s dwell. The
pre-edge was deconvolved from the background absorption edge by
simultaneously ?tting the background with a damped harmonic
oscillator function plus a line constrained to have a positive slope and
the pre-edge features with two Gaussian peaks. The oxidation state of
the glass was quanti?ed using the empirical calibrations in Cottrell
et al. (2009) based on pre-edge peak intensity ratios measured on a
series of 16 basalt reference glasses with Fe3+/?Fe ratios (0.088?
0.601) independently determined by M?ssbauer spectroscopy and 7
rhyolitic glasses with Fe3+/?Fe ratios (0.238?0.806) independently
determined by wet chemistry (Moore et al., 1995).
We performed a heating experiment on Kra?a obsidian in order to
induce spherulite crystallization and determine if optical and
chemical changes in neighboring melt could result from this
crystallization. Two cubes (2 cm-on-edge) of obsidian were subjected
to temperatures above the rheological glass transition (~690 ?C) in a
Lindberg Blue horizontal tube furnace with an internal Cromel?
Alumel (CR?AL) thermocouple positioned in the middle of the
chamber. We also used an external inconel-sheathed CR?AL thermo-
couple connected to a digital display to monitor the temperature of
the cube at a position close to the center of the cube. The cubes were
placed on a ceramic plate at the middle of the chamber, directly above
the internal thermocouple. Target temperatures were 770 ?C and
870 ?C. Only the 770 ?C experiment yielded useful results as the
sample heated to 870 ?C vesiculated making comparisons with the
natural unvesiculated sample impossible.
The temperature path comprising the heating, dwell, and cooling
stages was established by running a temperature calibration exper-
iment using an equal-sized ?dummy? cube of obsidian with a 3 mm
diameter hole drilled to the center of the cube to accommodate the
external thermocouple. This procedure allowed us to know the
approximate temperature?time path of the center of the cube during
heat treatment. We did not drill the cubes on which textural and
analytical observations were made because of the possibility that the
sample might outgas into the thermocouple hole, thereby affecting
the viscosity of the melt and hence the reaction kinetics at the dwell
temperature. Heat treatment consisted of a ramp to the target
temperature (770 ?C) at a rate of about 65 ?Cmin?1. We observed a
temperature overshoot of about 10 ?C above 770 ?C that took about
5 min to relax. Once the temperature had stabilized, we left the
sample in the furnace for 90 min and then rapidly removed the cube
in order to quench it with a cold blast of compressed air. We estimated
the cooling time on the dummy cube by leaving the external
thermocouple in the hole while cooling the sample. It took about
3 min for the sample to cool from 770 ?C to 25 ?C.
Following the experiment, we extracted a thin (~225 ?m), doubly
polished wafer from the center of the cube and examined the glass
and spherulites near the center of that wafer. We also measured a
water concentration pro?le (via FTIR) along a traverse extending from
one of the spherulites in the wafer center. Due to a lack of obsidian
starting material, we could not repeat the experiment at 770 ?C.
4. Results
4.1. Analytical measurements
Fig. 1A is a photomicrograph of a natural spherulitic obsidianwafer
that is the subject of the measurements discussed in the following
paragraphs. The spherulites in this sample are composed of
plagioclase and quartz (~95 vol.%), and minor amounts of clinopyr-
oxene (2?3 vol.%) and magnetite (<0.8 vol.%; Castro et al., 2008). All
of the spherulites are enclosed in haloes of colorless to light brown
glass, which separate them from the pervasive dark brown matrix
glass. The transitions between brown and colorless glass tend to be
sharp around the larger spherulites (radii >100 ?m) whereas the
boundaries are more diffuse around the smaller spherulites.
Fig. 1B shows a representative OH? concentration pro?le mea-
iron,with the darker brown glass havingmore ferric iron than the light-
colored glass (Gaillard et al. 2002; Donald et al., 2006; Moriizumi et al.,
2008). This hypothesis was con?rmed with ?-XANES measurements.
Spectra were collected on the spherulite shown in Fig. 1A both within
the clear glass rim, approximately 20 ?m from the spherulite?glass
border, and in the far-?eld brown matrix glass where the OH?
concentration ?attens out, about 330 ?m from the spherulite edge.
Fig. 3 shows the raw pre-edge spectra, model components, and total
model ?ts to the spectra. The spectrum taken in the distal brown glass
(spot 2) displays a proportionately larger Fe3+ peak, indicative of a
greater contribution from ferric iron. This can be seen evenmore clearly
in Fig. 3C when the baseline-subtracted spectra are superimposed.
Quantitatively, the centroid (area-weighted average of the two pre-
edgepeaks) shifts to higher energyby0.18 eVmoving fromthe clear rim
to the distal brown glass. Consistentwith this, the ratio of pre-edge peak
intensities (i.e. I(Fe3+)/I(Fe2+)I(Fe3+)/[I(Fe2+)+I(Fe3+)]) shift to
re?ect a greater contribution from ferric iron in the distal brown glass
than in the halo, consistent with a relative change in the ratio of Fe3+/
?Fe of 0.06. The ?XANES results clearly indicate a difference in the
relative Fe-oxidation state of the two glass regions, with the clear glass
halos demonstrably reduced relative to the brown matrix glass. The
absolute ferric iron content determined for the distal brown glass (Fe3+/
(0.04
(0.04
)/[I
.%. F
0.5%
ol.
275J.M. Castro et al. / Chemical Geology 268 (2009) 272?280sured by Castro et al. (2008). The area under the pro?le is proportional
to the amount of OH? groups in the silicate glass surrounding the
spherulite. Castro et al. showed that the OH? concentration increases
with the spherulite size, and typically matches the amount of water
that would be ejected during complete crystallization of anhydrous
minerals from a melt volume equal to that of the spherulite. Differ-
ences between the measured and predicted OH? contents show that
some spherulites retained water during their growth, consistent with
the presence of glass and microvesicles (Castro et al., 2008).
Table 1 shows representative major element glass compositions
measured within and across the differently colored glass regions. These
data show that the composition of glass in the colorless halo, the
transitional zone, and the distal brown glass surrounding the spherulite
in Fig. 1A are indistinguishable within analytical error. Most impor-
tantly, there is no discernable change in the total iron content as the
spherulite?glass boundary is approached; i.e., glass color is not a
function of the total iron content but rather the ratio of ferric to ferrous
iron. Fig. 2 is a graph showing the major element compositions of glass
along a line traverse emanating from the spherulite shown in Fig. 1A.
The color difference between the halo and distal glass in the natural
sample was suspected to re?ect the proportions of ferric and ferrous
Table 1
Glass and spherulite composition in Hrafntinnuhryggur obsidian.
Glass composition
SiO2 TiO2 Al2O3 FeOtot MnO
EPMA
Brown (n=7) 75.3 (0.68)a 0.24 (0.01) 12.9 (0.06) 3.22 (0.15) 0.11
Trans (n=1) 76.2 0.24 12.7 3.18 0.06
Colorless (n=8) 74.9 (0.76) 0.24 (0.03) 12.7 (0.08) 3.30 (0.20) 0.08
?-XANES I(Fe3+)/I(Fe2+) I(Fe3+
Clear halo 0.57 0.36
Brown matrix 0.76 0.43
Fe3+ converted to Fe2+ in rim: 9.1?10?8 mol
blog fO2-brown glass ?14.4
blog fO2-clear glass ?13.7
Wet chemistry: FeO of brown matrix (n=7): 2.31 ?0.02 wt.%. Fe2O3 implied: 1.04 wt
Log fO2 of brown glass due to exposure to fH2=0.02 bar: ?14.6.
Spherulite characteristics
Mineralogy (vol.%): plagioclase (45%); quartz (45%); clinopyroxene (5%); magnetite (<
Colorless glass rim width: 140 ?m (+10 ?m).
mass OH? in clear glass rim: 0.00052 mg moles OH? in the colorless rim: 3.0?10?8m
a Values in parentheses represent 1 s.d. about the mean value.
b fO2 calculations according to Kress and Carmichael (1991).?Fe=0.23?0.04) compares favorablywith, but ismore reduced than,
thewet chemical determination of Fe3+/?Fe=0.29?0.02,whereas
Fe in the clear brown glass is more reduced (Fe3+/?Fe=0.17?
0.04).
Fig. 4 shows measurements of the widths of several colorless glass
haloes versus their corresponding spherulite radii. Estimated uncer-
tainties in rim widths, stemming from the diffuse nature of the glass
color boundaries, are shown as vertical error bars. The nonlinear
regression to the data is a power law function relating rim width to
spherulite radius. Also shown in Fig. 4 are the predicted reduction rim
widths for each spherulite based on calculations using empirical
relationship of Gaillard et al. (2003a), which relates reduction front
position or width to hydrogen fugacity, temperature (800 ?C) and
time. These model calculations will be discussed further below.
4.2. Experimental results
The optical character and OH? concentration of the heated
obsidian are shown in Figs. 5 and 6. To aid comparisons with the
natural state, we also include images of unheated spherulite?rim
combinations collected on the spherulite shown in Fig. 1. The
MgO CaO Na2O K2O OH? Total
) 0.08 (0.01) 1.64 (0.05) 3.70 (0.20) 3.02 (0.14) 0.13 (0.02) 100.3
0.09 1.56 3.85 2.86 0.18 100.9
) 0.08 (0.01) 1.64 (0.04) 3.79 (0.14) 3.03 (0.14) 0.19 (0.01) 99.9
(Fe2+)+I(Fe3+)] Centroid Fe3+/?Fe FeO Fe2O3
7112.22 0.165 (0.04) 2.70 0.59
7112.40 0.225 (0.04) 2.52 0.79
e3+/?Fe: 0.29.
); interstitial glass (<5%).
276 J.M. Castro et al. / Chemical Geology 268 (2009) 272?280transmitted-light photomicrographs were taken at the same illumi-
nation and focal depth, and the imaged wafers were about the same
thickness.
The salient features and changes in the heated sample are: 1) thin,
~10 ?m wide veneers of plagioclase (identi?ed with EDS on an SEM)
jacketing all spherulites in the sample, 2) thin brown fringes composed
of glass, plagioclase, and magnetite crystals located just inboard of the
plagioclase veneers, 3) the colorless glass rims became brighter, and on
their outer periphery, thin, light brown glass zones developed that
appear to have propagated toward the spherulites, 4) circumferential
cracks formed parallel to the colorless?brown glass boundaries, in
some cases directly on them, and 5) along many of these cracks, small,
~10?30 ?m, ?bubble-trains? grew, comprising groups of spherical
(Fig. 5C) vesicles.
The most apparent changes that occurred during the heating
experiment were crystallization and brightening of the colorless
haloes (Fig. 5A,B). That the crystallization was ?new? is indicated by
the fact that none of the unheated spherulites have a jacket of feldspar
around them. The crystallization rate implied by the amount of new
plagioclase (~1.7?10?9m s?1; i.e., width of plagioclase veneer
divided by the experiment time) is similar to the average value
(~1.9?10?9m s?1) determined by Castro et al. (2008) from their
diffusion models at 800 ?C.
The circumferential fractures and small vesicle trains resulted from
processes occurring, respectively, below and above the rheological
glass transition (Tg). At the given heating rate, the sample reached Tg
(~700 ?C; Castro et al., 2008) in ~10 min. During this short time the
Fig. 2. Major element compositions of glass leading up to a spherulite margin (at position=
grey bars demarcate the approximate (?10 ?m) position of the colorless?brown glass bouglass may have fractured due to its greater thermal expansion
compared to the mineral phases in the spherulite. If this were the
case, then fractures formed in tension as the expanding glass pulled
away from the spherulite. Once the fractures formed, and the glass
passed through Tg, bubble nucleation occurred preferentially along the
fractures, which channeled volatile components to the growing
bubbles. The cracks apparently served as conduits for degassing of
the volatile-enriched boundary layers. An FTIR pro?le (Fig. 6)
measured across one of these fractures indicates that the dissolved
OH? content decreases in the vicinity of a fracture, suggesting
degassing into that fracture.
5. Mechanism of Fe-reduction during crystallization
It is clear from the geometry of the colorless glass haloes, speci?cally
their proportional increase in width with spherulite size (Fig. 4), and
their mimicking of the spherulite shapes, that the colorless glass rims
resulted from the growth of the spherulites. The ?-XANES spectra
indicate that some (~5?7% absolute or 26% relative) of the ferric iron in
the glass adjacent to the spherulites has been reduced to ferrous iron
(Fig. 3). In this section we address how these redox changes could have
occurred in light of the analytical and experimental evidence.
We ?rst consider the possibility that the heterogeneous ferric?
ferrous iron distribution in the glass is related to the cooling history.
M?trich et al. (2006) showed that the ferric?ferrous ratio in peralka-
line rhyolitic glass inclusions appeared to increase upon slow cooling
over several hundred ?C in high temperature ?-XANES experiments.
0 ?m). All analyses were collected with an electron probe microanalyzer. The vertical
ndary.
277J.M. Castro et al. / Chemical Geology 268 (2009) 272?280They attributed the apparent oxidation in XANES spectra upon cooling
to a change in the coordination environment of iron; high temper-
ature favors tetrahedrally coordinated ferrous iron while lower
temperatures stabilize octahedrally coordinated ferric iron. M?trich
et al. conclude that the XANES spectra of slowly cooled samples will
appearmore oxidized due to this coordination shift. To account for the
reduced halos in the spherulitic obsidian, the reduced halos would
have had to cool much more rapidly than the more oxidized distal
brown glass, which is physically unlikely because latent heat liberated
from the growing spherulites would have slowed cooling of the melt
Fig. 3. Raw ?-XANES spectra of the pre-edge region, model components, and model ?ts
for two points within the glassy matrix depicted in Fig. 1A. A) Point 1 inside the
colorless glass halo and (B) point 2 in the distal brown glass. C) Model ?ts of the spectra
in A and B baseline-subtracted and superimposed. Approximate positions of analysis
points are shown on the inset photomicrograph. These data show the relative changes
in the ratio of Fe2+ to Fe3+ in the colorless and brownmatrix glass. Iron in the glass near
the spherulite (point 1) is signi?cantly reduced compared to that in the far-?eld brown
glass (point 2).adjacent to the spherulite (Tuffen et al., in review). Moreover, the
color change in the glass supports a real difference in the oxidation
state of iron, not a coordination change. For these reasons, this
mechanism was probably not important for the formation of the color-
less rims.
We next consider the possibility that the reduced-Fe signature in
the rims arose from the late-stage crystallization of magnetite. The
Fig. 4. Colorless glass rim width versus the apparent spherulite radius, measured on
photomicrographs of the Kra?a obsidian (solid dots). Error in rim width, shown as
vertical bars, ranges from about 10 to 20%, and originates from the diffuse nature of the
colorless?brown glass boundaries. Regression curve is a power law function of the
form: y=2x0.69; R2=0.97. Solid triangles are the predicted rim widths based on the
experimentally constrained hydrogen?iron reaction rate constant of Gaillard et al.
(2003b), and the assumption that the rims developed during the interval of time that
the spherulites grew, which in turn, is taken from the growth rate data of Castro et al.
(2008).distribution of magnetite within spherulites, mainly as radial
aggregates sandwiched between larger domains of plagioclase and
quartz (Fig. 1), suggests that magnetite grains formed in the latest
stage of crystallization, from an interstitial melt that would have been
enriched in water, further stabilizing magnetite (Sisson and Grove,
1993). Magnetite crystallization could act to reduce the melt adjacent
to the spherulite via two potential mechanisms. First, hydrogen
produced by magnetite crystallization in the presence of water (i.e.,
?auto-oxidation,? Reaction (1)) would necessarily diffuse out of the
spherulite, and would have subsequently reduced ferric iron in the
neighboring melt. Second, magnetite crystallization alone will reduce
iron in the glass residuum by virtue of its higher ferric/ferrous ratio
(just as olivine crystallization would raise the ferric/ferrous ratio). We
show below that either mechanism, or both in concert, could have
generated the reduced halos.
?Auto-oxidation? as described by Holloway (2004) may proceed in
hydrous silicates when magnetite with a higher ferric/ferrous ratio
than the melt from which it is crystallizing becomes stable. Pure end-
member magnetite has a Fe3+/?Fe of about 0.67, compared to the
value of 0.22?0.29 in the brown glass; therefore magnetite precipi-
tation could have resulted in ?auto-oxidation? in this hydrous lava
from Kra?a. In this scenario, magnetite precipitation proceeds
through the consumption of water, generating one mole of H2 for
every mole of magnetite crystallized (Reaction (1)). If H2 leaves the
system completely through degassing along fractures, the remaining
material is left relatively oxidized. As H2 diffuses through a melt,
however, it necessarily results in a reduction front. We believe that
these spherulites capture this disequilibrium state.
Simple mass balance arguments con?rm the plausibility of this
scenario. The modal proportion of magnetite in the spherulites,
278 J.M. Castro et al. / Chemical Geology 268 (2009) 272?280determined by BSE image analysis on 10 different spherulites, is about
0.3 vol.% (?0.12). The amount of Fe3+ now residing in the magnetite
grains within the large spherulite pictured in Fig. 1A is about 0.003 mg
Fe3+ (5.4?10?08 moles of Fe3+), and because these magnetite grains
Fig. 5. Photomicrographs (500?; 200 ?m f.o.v.) collected on: (A) a natural spherulite?glass
the natural appearance of the spherulite margin (at top) and colorless and brown glass as vie
typical appearance of the spherulite and colorless and brown glass after heating a similar piec
the thin veneer (~10 ?m) of plagioclase coating the spherulite margin (upper right) and the r
center of the colorless halo with a string of small vesicles on its tip. Frame C shows another vi
The vesicles here formed along a fracture in the colorless glass zone. Frame D is a backscatt
microlite rich fringe (a), plagioclase veneer (b), and the colorless and Fe-reduced matrix gla
referred to the web version of this article.)
Fig. 6. An OH? concentration pro?le measured along a traverse extending from a spherul
photomicrograph. The vertical dashed and solid grey lines on the pro?le bound the light bro
shown in the photomicrograph. Note the coincidence between the depression in the OH?contain some Ti (Castro et al., 2008), this estimate of Fe3+ in
magnetite is a maximum. If all of the Fe3+ residing in the magnetite
were created by oxidation of Fe2+ in the liquid, magnetite
crystallization would produce a maximum of 2.7?10?8 moles of H2
combination, and (B?D) an experimentally heated spherulitic obsidian. Frame A shows
wed in transmitted, plane polarized light. The wafer is 197 ?m thick. Frame B shows the
e of obsidian to 770 ?C for approximately 90 min. This wafer is about 202 ?m thick. Note
elative brightening of the colorless glass halo. Note also the crack running roughly in the
ew of the incipient crystallization and vesiculation that occurred during heat treatment.
ered electron image of the three zones developed at the spherulite margin: the brown
ss (c). (For interpretation of the references to colour in this ?gure legend, the reader is
ite in the heated obsidian sample. The position of traverse is shown on the subjacent
wn glass zone that may have formed as a result of degassing of hydrogen into the crack
concentration pro?le and the position of the crack at ~75?80 ?m.
279J.M. Castro et al. / Chemical Geology 268 (2009) 272?280via Reaction (1). This hydrogen could, in turn, reduce 5.4?10?8
moles of ferric iron (i.e. 0.003 mg) in the adjacent melt (and produce
water) according to:
H2?s? + Fe
3+
2 O3?m=g? = 2Fe
2+ O
?m=g? + H2O?m=g? ?2?
Subscripts above refer to: (s)?spherulite and (m/g)?melt/glass. If
all of the H2 were transferred to the 100 ?m halo jacketing the
spherulite in Fig. 1A, it would shift the percentage of Fe3+ from ~22%
(measured via XANES) to 9%. We observe Fe3+ in the halo equal to
16% (Table 1), indicating that ?auto-oxidation? within the spherulite
could actually be responsible for iron reduction in the halo. This
calculation is a maximum because it assumes that 100% of the ferric
iron in the magnetite had to be converted from FeO (i.e. Reaction (1)).
We observe, however, that the brown glass contains about 22% ferric
iron. Assuming that Reaction (1) proceeded at 78% ef?ciency (i.e. 22%
of the ferric iron incorporated into the magnetite was already ferric),
we still ?nd that the percentage of Fe3+would drop to 11%.Within the
error of the XANES measurements, the uncertainty of the reaction
ef?ciency and mass balance, and even taking into account that the
magnetite is Ti-bearing, this scenario remains plausible.
The auto-oxidation scenario is also corroborated by the high
concentration of water in the halos. We observe the number of moles
of ?H2-equivalent? water currently residing in the halo (i.e. one half
the moles of OH?, calculated from the OH? concentration of 0.19%
measured by FTIR) equal to 5?10?8. This is ~45% higher than the
number of moles of H2 expected from magnetite crystallization via
Reaction (1). We therefore observe more ?water? (hydroxyl and
molecular) in the glass surrounding these spherulites than produced
by magnetite crystallization. This ?excess? water likely re?ects water
extruded as an incompatible component from minerals growing
within the spherulite (e.g., Castro et al., 2008; Watkins et al., 2009).
A second mass balance argument can be made that magnetite
precipitation alone (i.e. not relying on the presence of water) could
have caused the reduction halos. Magnetite incorporates iron in a ratio
of 2Fe3+:1Fe2+. Only 0.4% magnetite crystallization (as a percent of the
volume of glass nowoccupied by the reduced halos) is required to cause
the ferric iron to decrease from ~22 to the observed value of ~16%. If we
assume that all the magnetite in the spherulites utilized components
from the volume of glass now occupied by the reduced halos, we would
expect 0.7% crystallization (which would reduce the ferric iron to 11%).
Ifweassume that all of themagnetite in the spherulites crystallized from
the combined volumes of glass now comprising the spherulites plus the
reduced halos, we would expect 0.3% crystallization (which would
reduce the ferric iron to 18%). These two values therefore bracket the
maximum and minimum extents of magnetite crystallization, and
either extreme provides a reasonable mechanism by which the glass in
the halos could have been reduced. Crystallization of 0.4% magnetite
from the halo would simultaneously cause the total iron concentration
(FeO) in the halo to fall from 3.2 to 2.8%. This is close to within the error
of themicroprobemeasurements. Nevertheless, nodecrease or gradient
in the iron concentration is observed in the halos relative to the distal
brown glass. This observation, combined with the demonstrable
presence ofwater in the system and the likelihood of the auto-oxidation
reaction proceeding, leads us to favor the auto-oxidation scenario as a
means of generating the halos, but neither can be ruled out and in fact
both could have contributed.
6. Experimentally induced Fe-reduction and spherulite growth
Gaillard et al. (2003a,c) simulated the process of hydrogen-?ux
iron reduction by exposing natural Fe-bearing rhyolitic melt and glass
cylinders to reducing atmospheres of hydrogen and hydrogen?argon
gas mixtures. Interestingly, their experiments produced many of the
features we observe around the natural spherulites: 1) a sharp changein glass color bounding a zone of reduced Fe3+/Fetotal in hydrous glass,
2) sigmoidal OH? concentration pro?les emanating from the sample
edge and attributable to hydrogen incorporation in the melt followed
by diffusion of molecular water along the concentration gradient,
and 3) an offset between the reaction front position and the point of
elevated OH? in the glass.
Gaillard et al. (2003a,c) proposed that because hydrogen is very
reactive with iron in the silicate melt, the reaction front progress in
the melt is governed by the solubility and diffusion of hydrogen in the
melt. The effective diffusion rate of H2, in turn, was limited by the fH2
of their experiments. Gaillard et al. observed that the rate of progress
of the reduction front (?m/h) was several orders of magnitude slower
than the expected hydrogen diffusivity (?m/s) in the melt.
Gaillard et al. (2003a) performed time series experiments in
order to characterize the rate of advancement of the reduction front
with time. They observed a square root of time dependence of the
front position and, based on linear relationships between the square
of the reduction front position and run duration, they extracted
reduction rate constants, K=?2/t, where ? is the reaction front
position, and t is time, for their experiments at 800 ?C and a range of
hydrogen fugacities (0.02?50 bars). They proposed that the reduc-
tion rate was limited by hydrogen incorporation, which in turn, is a
function of the fugacity-dependent H2 solubility and diffusivity in
the melt.
It is possible that the spherulites behaved in a similarmanner, that is,
they acted as hydrogen point sources to the neighboringmelt. Although
theboundary conditions are slightly different, e.g., the spherulites are an
internal, as opposed to external hydrogen source, the empirical kinetic
data of Gaillard et al. (2003c) can be used to assess whether the widths
of the natural reduction rims are compatible with the timescales of
spherulite growth. In other words, we assume that the reduction front
startedmoving at the onset of spherulite growth and stoppedwhen the
growth ceased. By this simple scenario, the rim widths depend on the
growth timescale, and therefore the duration that the expelled
hydrogen had to react with the rhyolite melt according to the reaction
rate constant K (Gaillard et al., 2003a).
Castro et al. (2008) estimated average spherulite growth rates by
modeling the combined advection and diffusion of water away from the
growing spherulites and ?tting model pro?les to the natural OH?
concentration pro?les. Even though their model did not account for
hydrogen incorporation in themelt and its unknowneffect on thekinetics
of H2O diffusion, their growth rates match those determined experimen-
tally on compositionally similar melts (e.g., Baker and Freda, 2001), and
are probably accurate to an order ofmagnitude. According to their results
at 800 ?C the spherulites grew at an average rate of ~10?9m/s (Castro
et al., 2008), which is similar to the rate determined in our spherulite
growth experiments at 770 ?C.
Results of rim-width calculations are shown in Fig. 4. Individual
spherulite growth timescales were determined by dividing the
average growth rate (800 ?C; Castro et al., 2008) by their spherulite
radii. Using the individual growth timescales (t), we calculated rim
widths (?) using the relation ?=(K?t)1/2. The best match between the
calculated and natural rim widths was attained with the K value
derived from the lowest H2 fugacity (fH2=0.02 bar) in the experi-
ments of Gaillard et al. (2003a). We found a very large mismatch at
their next highest fH2 (~0.25 bar H2) and attribute this to the nearly
one order of magnitude increase in K at fH2=0.25 bar, which
translates to comparatively rapid rim growth relative to the spherulite
growth timescales.
The agreement between the calculated and real Fe-reduction rim
widths is good, suggesting that the natural rims could have developed
under conditions of relatively low hydrogen fugacity (~0.02 bar
according to Gaillard et al., 2003a). This match supports our
hypothesis that it is molecular hydrogen that extrudes from the
growing spherulite, thereby ?uxing the ferric iron in the neighboring
melt and causing the propagation of a redox front.
7. Conclusion
Our analysis indicates that the spherulitic growth of anhydrous
Donald, S.B., Swink, A.M., Schreiber, H.D., 2006. High-iron ferric glass. Journal of Non-
crystalline Solids 352, 539?543.
Gaillard, F., Scalliet, B., Pichavant, M., Beny, J.-M., 2001. The effect of water and fO2 on
the ferric?ferrous ratio of silicic melts. Chemical Geology 174, 255?273.
Gaillard, F., Scalliet, B., Pichavant, M., 2002. Kinetics of iron oxidation?reduction in
hydrous silicic melts. American Mineralogist 87, 829?837.
280 J.M. Castro et al. / Chemical Geology 268 (2009) 272?280neighboring melt or glass around the growing crystals. Fe-redox
reactions are driven by the liberation of hydrogen, and we propose
that it is hydrogen produced as a product of magnetite crystallization.
The result is an Fe-reduction reaction that propagates through the glass
or melt with time. That spherulite crystallization can cause reduction of
ferric iron in the silicate melt suggests that this phenomenon could
prevent oxidation of silicic lavas during their emplacement despite
extensive crystallization. For example, Carmichael (1991) indicates that
voluminous post-caldera rhyolite lavas at Yellowstone National Park,
USA underwent little to no change in redox state compared to earlier-
erupted ash ?ows. Carmichael (1991) used the apparent lack of
oxidation of these lavas to support his hypothesis that ?silicic magmas
have redox states that re?ect their source regions rather than H2 loss.?
He further explained that the low bulk H2O content of the post-caldera
lavas could have suppressed the activity of the H2O asmost of thewater
would be speciated as OH groups (Stolper, 1982).
Spherulite crystallization is widespread in the interiors of the
Yellowstone rhyolites (Wright, 1915; Colony and Howard, 1934), with
some ?ows containing well over 50 vol.% spherulites (e.g., in the Nez
Pierce ?ow). We suggest that the lack of variation in redox state may
have been enabled by spherulite crystallization and local hydrogen
solute rejection, but retention of generated hydrogen within the ?ows.
Thus the conclusion of Carmichael (1991) that ?silicic magmas with
small amounts of iron and large amounts of water do not have their
redox states reset? upon eruption may be a consequence of the
offsetting effects of crystallization and glass reduction.
Acknowledgements
We thank M.C. Martin for assistance with the SFTIR measurements
and T. Gooding with sample preparation. Antonio Lanzirotti provided
critical assistance at X26A, NSLS, Brookhaven National Lab. We thank
F. Gaillard and M. Rutherford for valuable discussions. Finally, we
appreciate the help of Landsvirkjun and the staff at Kra?a power
station during our sampling campaign in Iceland.
References
Baker, D.R., Freda, C., 2001. Eutectic crystallization in the undercooled orthoclase?
quartz?H2O system: experiments and simulations. European Journal of Mineralogy
13, 453?466.
Baker, L.L., Rutherford, M.J., 1996. The effect of dissolved water on the oxidation state of
silicic melts. Geochimica et Cosmochimica Acta 60, 2179?2187.
Berry, A.J., O'Neill, H.S., Jayasuriya, K.D., Campbell, S.J., Foran, G.J., 2003. XANES
calibrations for the oxidation state of iron in a silicate glass. American Mineralogist
88, 967?977.
Botcharnikov, R.E., Koepke, J., Holtz, F., McCammon, C., Wilke, M., 2005. The effect of
water activity on the oxidation and structural state of Fe in a ferro-basaltic melt.
Geochimica Cosmochimica Acta 69, 5071?5085.
Burgisser, A., Scaillet, B., 2007. Redox evolution of a degassing magma rising to the
surface. Nature 445, 194?197. doi:10.1038/nature05509.
Candela, P.P., 1986. The evolution of aqueous vapor from silicate melts: effect on
oxygen fugacity. Geochimica Cosmochimica Acta 50, 1205?1211.
Carmichael, I.S.E., 1991. The redox states of basic and silicic magmas: a re?ection of
their source regions ?. Contributions to Mineralogy and Petrology 106, 129?141.
Castro, J.M., Beck, P., Tuffen, H., Nichols, A.R.L., Martin, M., 2008. Timescales of spherulite
crystallization in obsidian inferred from water concentration pro?les. American
Mineralogist 93, 1816?1822.
Colony, R., Howard, A.D., 1934. Observations on spherulites. American Mineralogist 19,
515?524.
Cottrell, E., Kelley, K.A., Lanzirotti, A.T., Fischer, R., 2009. High-precision determination
of iron oxidation state in silicate glasses using XANES. Chemical Geology.
doi:10.1016/j.chemgeo.2009.08.008.Gaillard, F., Schmidt, B., Mackwell, S., McCammon, C., 2003a. Rate of hydrogen?iron
redox exchange in silicate melts and glasses. Geochimica et Cosmochimica Acta 67,
2427?2441.
Gaillard, F., Pichavant, M., Scaillet, B., 2003b. Experimental determination of activities of
FeO and Fe2O3 components in hydrous silicic melts under oxidizing conditions.
Geochimica et Cosmochimica Acta 67, 4389?4409.
Gaillard, F., Pichavant, M., Mackwell, S., Champallier, R., Scaillet, B., McCammon, C.,
2003c. Chemical transfer during redox exchanges between H2 and Fe-bearing
silicate melts. American Mineralogist 88, 308?315.
Geschwind, C.H., Rutherford, M.J., 1995. Crystallization of microlites during magma
ascent: the ?uidmechanics of 1980?1986 eruptions at Mount St. Helens. Bulletin of
Volcanology 57, 356?370.
Hess, K.U., Dingwell, D.B., 1996. Viscosities of hydrous leucogranitic melts: a non-
Arrhenian model. American Mineralogist 81, 1297?1300.
Holloway, J.R., 2004. Redox reactions in sea?oor basalts: possible insights into silicic
hydrothermal systems. Chemical Geology 210, 225?230.
Kress, V., Carmichael, I.S.E., 1991. The compressibility of silicate liquids containing
Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on
their redox states. Contributions to Mineralogy and Petrology 108, 82?92.
Manley, C.R., 1992. Extended cooling and viscous ?ow of large, hot rhyolite lavas:
implications of numerical modeling results. Journal of Volcanology and Geothermal
Research 53, 27?46.
Manley, C.R., Fink, J.H., 1987. Internal textures of rhyolite ?ows as revealed by research
drilling. Geology 15, 549?552.
Mathez, E.A., 1984. In?uence of degassing on oxidation states of basaltic magmas.
Nature 310, 371?374.
M?trich,N., Susini, J., Foy, E., Farges, F.,Massare,D., Sylla, L., Lequien, S., Bonnin-Mosbah,M.,
2006. Redox state of iron in peralkaline rhyolitic glass/melt: X-ray absorption micro-
spectroscopy experiments at high temperature. Chemical Geology 231, 350?363.
Moore, G., Righter, K., Carmichael, I.S.E., 1995. The effect of dissolved water on the
oxidation state of iron in natural silicate liquids. Contributions to Mineralogy and
Petrology 120, 170?179.
Moriizumi, M., Nakashima, S., Okumura, S., Yamanoi, Y., 2008. Color-change processes
of a plinian pumice and experimental constraints on color-change kinetics in air of
an obsidian. Bulletin of Volcanology. doi:10.1007/s00445-0080202-5.
Mueller, R.F., 1971. Oxidative capacity of magmatic components. American Journal of
Science 270, 236?243.
Newman, S., Stolper, E.M., Epstein, S., 1986. Measurement of water in rhyolitic glasses:
calibrationofan infraredspectroscopic technique.AmericanMineralogist71, 1527?1541.
Peck, L.C., 1964. Systematic analysis of silicates: U.S.G.S. Bulletin 1170, 80.
Sato, M., Wright, T.L., 1966. Oxygen fugacities directly measured in magmatic gases.
Science 153, 1103?1105.
Sato, M., 1978. Oxygen fugacity of basaltic magmas and the role of gas-forming elements.
Geophysical Research Letters 5, 447?449.
Sisson, T.W., Grove, T.L., 1993. Experimental investigations of the role of H2O in calc-
alkaline differentiation and subduction zonemagmatism. Contributions toMineralogy
and Petrology 113, 143?166.
Smith, R.K., Tremallo, R.L., Lofgren, G.E., 2001. Growth of megaspherulites in a rhyolitic
vitrophyre. American Mineralogist 86, 589?600.
Stasiuk, M.V., Barclay, J., Carroll, M.R., Jaupart, C., Ratte, J.C., Sparks, R.S.J., Tait, S.R., 1993.
Degassing duringmagmaascent in theMuleCreek vent (USA). Bulletinof Volcanology
58, 117?130.
Stevenson, R.J., Briggs, R.M., Hodder, A.P.W., 1994. Physical volcanology and
emplacement history of the Ben Lomond rhyolite lava ?ow, Taupo Volcanic Centre,
New Zealand. New Zealand Journal of Geology and Geophysics 37, 345?358.
Stolper, E.S., 1982. The speciation of water in silicate melts. Geochimica et Cosmochimica
Acta 46, 2609?2620.
Swanson, S.E., Naney,M.T.,Westrich, H.R., Eichelberger, J.C., 1989. Crystallizationhistoryof
Obsidian Dome, Inyo Domes, California. Bulletin of Volcanology 51, 161?176.
Tuffen, H., Castro, J.M., 2009. The emplacement of an obsidian dyke through thin ice:
Hrafntinnuhryggur, Kra?a, Iceland. Journal of Volcanology and Geothermal
Research. doi:10.1016/j.jvolgeores.2008.10.021.
Tuffen, H., Castro, J.M., Wilson, L., in review. The thermal effects of spherulite
crystallization in rhyolitic lava. Earth and Planetary Science Letters.
Watkins, J., Magma, M., Huber, C., Martin, M., 2009. Diffusion-controlled spherulite
growth in obsidian inferred from H2O concentration pro?les. Contributions to
Mineralogy and Petrology 157, 163?172.
Westrich, H.R., Stockman, H.W., Eichelberger, J.C., 1988. Degassing of rhyolitic magma
during ascent and emplacement. Journal of Geophysical Research 93, 6503?6511.
Wilke,M., Partzsch,G.M., Bernhardt, R., Lattard,D., 2005.Determinationof the ironoxidation
state in basaltic glasses using XANES at the K-edge. Chemical Geology 220, 143?161.
Wright, F.E., 1915. Obsidian from Hrafntinnuhryggur, Iceland: its lithophysae and
surface markings. Bulletin of the Geological Society of America 26, 255?286.phases can cause signi?cant changes in Fe-oxidation state in the