You are on page 1of 20

Journal of Volcanology and Geothermal Research 197 (2010) 259278

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j vo l g e o r e s

A melt inclusion study of the Toba Tuffs, Sumatra, Indonesia


Craig A. Chesner a,, James F. Luhr b
a
b

Department of Geology/Geography, Eastern Illinois University, Charleston, IL 61920, United States


Department of Mineral Sciences, Smithsonian Institution, Washington, DC, 20013, United States

a r t i c l e

i n f o

Article history:
Received 26 November 2008
Accepted 9 June 2010
Available online 18 June 2010
Keywords:
Toba Tuffs
melt inclusions
geochemistry
dissolved volatiles
aerosols

a b s t r a c t
The Toba Caldera in Northern Sumatra, Indonesia is the site of Earth's largest Quaternary volcanic eruption.
This eruption, dated at 74 ka, produced the 2800 km3 Youngest Toba Tuff (YTT) and ash-fall. Quartz-bearing
silicic pumices ranging from 68 to 77 wt.% SiO2 indicate that the YTT magma was zoned compositionally and
mineralogically. Prior to the YTT eruption, two other silicic quartz-bearing tuffs known as the Middle Toba
Tuff (MTT) and the Oldest Toba Tuff (OTT) were erupted from Toba at 0.501 Ma and 0.840 Ma respectively.
Although the volatile contents of the Toba magmas are poorly constrained, aerosols generated by the YTT
eruption are generally thought to have caused a global volcanic winter. An evaluation of the pre-eruptive
dissolved gas contents of the YTT magma is fundamental to understanding the global effects of this eruption.
We used melt inclusions in quartz crystals to determine the dissolved H2O, CO2, S, Cl, and F contents of the
YTT, MTT, and OTT magmas. Quartz crystals selected from pumice blocks and welded tuffs that spanned the
compositional ranges of the three units were chosen for study. Major and trace element analyses of melt
inclusions were also conducted to characterize melt evolution and provide context to the volatile data. Melt
inclusions from the YTT, MTT, and OTT are rhyolitic in composition (7377 wt.% SiO2) and have overlapping,
indistinguishable major element trends. Inclusions from the large eruptions (YTT and OTT) have similar
geochemical relationships where the most evolved melt inclusions occur in the least silicic bulk rock samples
and are more evolved than their matrix glasses. In contrast, the least evolved melt inclusions occur in the
most silicic samples and either overlap with, or are less evolved than their matrix glasses. These patterns can
be explained if quartz crystals in the least silicic magmas were inherited from more evolved melt, while
those in the most silicic melt formed in-situ. Several lines of evidence suggest that crystal settling of earlyformed quartz crystals in long-lived magma bodies was an important process in the YTT and OTT magmas.
Trace element data from the YTT melt inclusions have wide ranges (Rb = 250830 ppm, Ba = b 10485 ppm)
and mostly support these models; however, incompatible trace elements are enriched 23 over their
matrix glass compositions in all samples. In the smaller MTT, melt inclusions overlapped with the
compositional range of the bulk rock samples and are consistent with in-situ crystallization. FTIR analyses
indicate that the YTT, MTT, and OTT melts contained about 4.05.5, 2.05.5, and 2.05.5 wt.% H2O
respectively. CO2 contents in all units are generally b 100 ppm and Cl is b 2000 ppm. Water gradients occur in
all 3 tuffs, Cl gradients are evident only in the YTT, and the least silicic samples from all units had no
detectable CO2. Sulfur contents in melt inclusions are low in all samples (b 32 ppm) and overlap with matrix
glass S contents. Applying broad degassing constraints, about 1014 g of H2SO4 aerosols were likely loaded into
the atmosphere during the YTT eruption, 2 orders of magnitude less than previous petrologic estimates.
Considerably more Cl and F, about 1015 g of each, may have been part of the YTT aerosol cloud.
2010 Elsevier B.V. All rights reserved.

1. Introduction
The Toba Caldera Complex in northern Sumatra (Fig. 1) was the site
of 4 caldera-forming eruptions in the past 1.2 m.y. (Chesner and Rose,
1991). Tuffs known as the Haranggoal Dacite Tuff (HDT), Oldest Toba
Tuff (OTT), Middle Toba Tuff (MTT), and Youngest Toba Tuff (YTT)
erupted at 1.2, 0.840, 0.501, and 0.074 Ma respectively (Nishimura

Corresponding author. Tel.: +1 217 581 6323.


E-mail address: cachesner@eiu.edu (C.A. Chesner).
0377-0273/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jvolgeores.2010.06.001

et al., 1977; Diehl et al., 1987; Chesner et al., 1991). Only welded tuffs
are preserved from the three earlier eruptions and their estimated
volumes are HDT = 35 km3, OTT = 500 km3, and MTT = 60 km3
(Chesner and Rose, 1991; Knight et al., 1986). The culminating event
engulfed all previous calderas and was Earth's largest Quaternary
volcanic eruption (Smith and Bailey, 1968), producing 2800 km3 of
non-welded to densely welded ignimbrite (YTT) and co-ignimbrite
ash-fall (Rose and Chesner, 1987). Ignimbrite deposits cover 20,000
30,000 km2 in northern Sumatra and extend to the Indian Ocean and
the Straits of Malacca (Van Bemmelen, 1949; Aldiss and Ghazali,
1984). Ash deposits from the eruption have been found in Malaysia,

260

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Fig. 1. Location map for the Toba Caldera Complex (modied after Chesner, 1998). Sample sites for rock samples used in this study are indicated by Y (YTT), M (MTT), O (OTT), and
P (PYLD). Sample numbers are written next to the code letters.

India, the Indian Ocean, the Arabian Sea, and the South China Sea
(Ninkovich et al., 1978; Ninkovich, 1979; Stauffer et al., 1980; Rose and
Chesner, 1987; Dehn et al., 1991; Pattan et al., 1999; Acharyya and
Basu, 1993; Shane et al., 1995; Westgate et al., 1998; Bhring et al.,
2000; Lee et al., 2004; Raj, 2008). A sulfate rich layer in Greenland ice
core GISP2, dated at 71 5 ka, has been attributed to the YTT eruption
(Zielinski et al., 1996) suggesting global distribution of the YTT aerosol
cloud. Aerosols from the eruption are generally thought to have caused
a global volcanic winter (Rampino and Self, 1992) that lasted
approximately 6 yrs (Zielinski et al., 1996; Bekki et al., 1996) and
may have contributed to enhanced cooling for 200 yrs (Zielinski et al.,
1996). The harsh conditions from ash fallout in southeast Asia and
subsequent global cooling led Rampino and Ambrose (2000) to
suggest that the eruption decimated the human population.
It has been shown that the mass of sulfur emitted during eruptions
can be an important contributor to brief global cooling (Rampino et al.,
1988). The SO2 and H2S emitted during eruptions combine quickly
with OH to form H2SO4 aerosols, which can absorb and scatter
incoming solar radiation. Volcanic aerosols that form in the troposphere have short residence times of weeks or months. However, those
that are injected into the stratosphere from large-volume eruptions
can have residence times of well over a year (Turco et al., 1982). A

recent generic super-eruption climate simulation by Jones et al. (2005)


using a stratospheric sulfate loading 100 times larger than the 1991 Mt.
Pinatubo levels, suggests that a 10 C global temperature drop is likely,
and cooling can persist for over a decade. Climate model simulations
specic to the YTT eruption (Robock et al., 2009) and using 33 to 900
times the Pinatubo SO2 stratospheric injection, predict global cooling
of 817 C and climate recovery within a few decades. Thus, the mass
of sulfurous gases emitted during an eruption is a crucial parameter
when modeling its climatic effects. Rose and Chesner (1990) used
phase equilibria of sulde blebs in YTT oxides (Whitney, 1984;
Toulmin and Barton, 1964) and other data to calculate that approximately 3 1015 g of H2S (1016 g H2SO4) were emitted during the YTT
eruption. Data from the Greenland ice cores suggest a similar amount of
H2SO4 (0.74.4 1015 g) was present in the atmosphere during the 6 yrs
following the eruption (Zielinski et al., 1996). Rampino and Self (1993)
have calculated that the 74 ka Toba eruption aerosol layer, consisting of
H2SO4 and ne dust, may have caused a global temperature decrease of
up to 5 C. These calculations are based upon estimates of increased
optical depths that plausibly resulted from the Toba H2SO4 aerosol
cloud. Scaillet et al. (1998) however, have argued that due to the low
solubility and diffusivity of S in low temperature silicic melts, S yields
during the YTT eruption may have been 23 orders of magnitude lower

Table 1
Representative analyses of YTT and PYLD melt inclusions and their whole rock and matrix glasses. All analyses have been normalized to 100% on a volatile free basis, and all Fe is reported as FeO. Melt inclusion (MI) and matrix glass (MG) major
elements were determined by microprobe, and whole rock (WR) major elements by XRF. Melt inclusion trace elements are by LA-ICP-MS, bulk rock and matrix glass trace elements were analyzed by ICP-MS. Water is by FTIR and Cl by
microprobe. Samples are arranged from left to right in order of decreasing SiO2 contents of their bulk rocks. PYLD samples have TT prexes.
Sample

5B3WR 5B3-3 MI 5B3 MG 55A1 WR


76.76
0.06
13.29
1.09
0.08
0.18
0.67
2.93
4.92
0.02

76.35
0.07
12.77
0.97
0.10
0.07
0.78
3.65
5.21
0.02
4.37
0.15

77.31
0.07
12.37
0.92
0.13
0.05
0.71
3.41
5.01
0.02

75.96
0.18
12.98
1.55
0.07
0.30
1.20
2.83
4.91
0.01

2.9
271.6
343.1
32
20.35
44.15
54.5
90
95.32
17.74
20.14
12.14
14.46
136
88.73
30.44
26.34
61.2
55.42
7.1
25.18
22.3
5.99
5.57
0.35
0.22
5.83
5.39
1.08
0.99
7.15
1.55
1.46
4.69
5.05
0.77
5.28
6.25
0.86
1.01
3.95
3.99
2.57
2.95
23.58
47.71
41.64
49.42
7.02
7.89

2.5
335.8
7
59.87
89
23.44
17.27
17
22.28
49.19
6.19
22.65
6.25
0.14
6.59
1.26
8.3
1.83
5.64
0.98
6.98
1.16
4.38
3.08
24.28
42.44
8.2

3.4
242.8
85
33.27
121
15.57
10.23
440
30.04
57.75
6.48
22.27
4.83
0.57
4.46
0.82
5.38
1.17
3.41
0.57
3.86
0.63
4.36
1.84
30.97
31.31
4.96

55A1-3-MI 55A1 MG 23A4 WR 23A4-3-MI


75.78
0.11
13.04
1.03
0.06
0.07
0.85
3.69
5.32
0.05
4.88
0.13
428.96
56.27
57.7
125.77
22.66
17.19
421.98
40.23
81.73
30.64
5.96
0.47
6.15
1.09
1.73
5.32
6.23
1.01
4.46
2.88
56.21
54.26
8.69

77.10
0.08
12.43
0.93
0.10
0.06
0.80
3.39
5.08
0.02

74.98
0.19
13.30
1.86
0.07
0.44
1.62
2.97
4.54
0.02

2.3
288.4
22
44.38
74
17.31
12.94
87
23.81
50.59
5.88
20.47
4.99
0.22
5.01
0.92
6.06
1.34
4.12
0.71
5.07
0.84
3.37
2.13
37.93
33.9
5.8

5.5
200.3
134
31.41
157
14.99
7.52
554
45.31
81.17
9.99
34.04
6.92
0.95
5.51
0.95
5.79
1.17
3.35
0.53
3.53
0.56
5
1.5
27.66
28.02
4.17

76.13
0.06
12.76
1.13
0.06
0.05
0.61
3.67
5.52
0.01
4.98
0.16

23A4 MG 12A1 WR 12A1- 3-MI 12A1 MG 21A4 WR 21A4-3MI 21A4 MG TT2 WR TT2-3- MI TT2 MG TT7 WR TT7-3- MI TT7 MG
77.01
0.09
12.52
0.92
0.12
0.07
0.87
3.17
5.23
0.01

71.99
0.27
15.92
2.50
0.08
0.48
2.08
2.80
3.85
0.03

2
247.6
39
35.43
76
14.54
10.06
260
29.48
55.44
7.14
24.48
5.42
0.38
4.76
0.83
5.17
1.1
3.34
0.57
4.05
0.67
3.17
1.7
35.28
28.63
4.73

5.7
146
153
24.58
159
14.09
5.35
727
44.01
77.7
8.28
27.17
5.01
0.93
4.11
0.7
4.38
0.89
2.6
0.41
2.73
0.45
4.97
1.35
27.75
26.1
3.01

76.58
0.07
12.38
1.01
0.08
0.06
0.73
3.77
5.31
0.01
4.75
0.14

76.88
0.09
12.56
0.91
0.09
0.07
0.95
3.21
5.23
0.01

70.18
0.41
15.67
3.30
0.08
0.76
2.86
3.10
3.57
0.06

1.8
209.8
50
30.32
84
14.05
7.98
473
31.82
59.32
6.78
22.46
4.55
0.43
4.05
0.7
4.34
0.93
2.86
0.49
3.47
0.58
3.26
1.58
36.89
28.01
3.96

10
119
246
34.73
209
13.22
4.29
881
54.51
88.65
10.96
38.22
7.18
1.47
6.36
1.03
6.32
1.28
3.53
0.52
3.33
0.52
5.82
0.96
23.66
18.05
2.66

76.51
0.09
12.23
0.97
0.21
0.05
0.81
4.32
4.78
0.03
4.40
0.13

76.76
0.13
12.69
1.06
0.09
0.10
1.08
3.14
4.94
0.02

75.87
0.12
13.13
1.26
0.07
0.18
1.14
3.40
4.82
0.01

1.9
177
79
30.38
97
13.63
5.9
821
57.64
96.78
10.82
34.68
5.81
0.67
4.89
0.75
4.5
0.93
2.77
0.45
3.12
0.52
3.31
1.26
38.71
25.29
3.37

2.9
233.2
69
34.28
96
14.67
9.6
281
32.93
63.44
6.84
22.87
4.74
0.47
4.45
0.83
5.46
1.17
3.54
0.59
3.98
0.66
3.7
1.94
34.27
33.92
5.23

76.96
0.09
12.32
0.97
0.07
0.06
0.73
3.78
4.99
0.03
4.77
0.15
472.57
17.31
63.89
121.68
27.87
22.58
80.27
36.48
81
32.4
6.9
0.2
7.45
1.27
2.07
6.8
8.23
1.3
5.22
4.34
66.38
67
11.99

77.40
0.08
12.26
0.89
0.16
0.05
0.71
3.40
5.03
0.01

69.78
0.40
15.37
3.23
0.08
1.01
3.22
3.46
3.34
0.11

2=.5
300.6
16
49.07
80
18.69
14.03
57
23.77
50.96
6.02
21.12
5.39
0.18
5.49
1.02
6.72
1.47
4.62
0.8
5.71
0.95
3.75
2.34
42.83
37.22
6.52

6.7
135.2
210
24.63
179
12.35
4.63
900
49.73
89.09
9.22
30.48
5.37
1.17
4.48
0.72
4.4
0.89
2.52
0.38
2.45
0.4
5.1
0.95
21.6
19.92
2.24

76.33
0.06
12.55
1.02
0.06
0.08
0.89
3.70
5.30
0.00
4.68
0.13
257.69
28.74
38.76
77.84
14.75
10.99
199.28
24.63
52.56
19.77
4.61
0.31
4.68
0.78
1.18
3.72
4.35
0.66
2.95
1.95
38.16
32.66
5.58

76.65
0.10
12.71
1.04
0.10
0.07
0.98
3.15
5.18
0.02

2.2
201.4
72
27.48
82
13.44
7.38
759
38.34
71
7.45
23.95
4.43
0.52
3.86
0.64
3.92
0.83
2.58
0.43
3.03
0.51
2.96
1.32
33.65
23.91
3.3

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

SiO2
TiO2
Al2O3
FeO
MnO
MgO
CaO
Na2O
K2O
P2O5
H2O
Cl
Sc
Rb
Sr
Y
Zr
Nb
Cs
Ba
La
Ce
Pr
Nd
Sm
Eu
Gd
Tb
Dy
Ho
Er
Tm
Yb
Lu
Hf
Ta
Pb
Th
U

261

262

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

than previous estimates. Oppenheimer (2002) concurs and suggests


that global cooling associated with the YTT eruption may have only been
about 1 C. Thus, a rened evaluation of the pre-eruptive dissolved gas
contents of the YTT magma, especially S, is fundamental to understanding the global effect of this eruption.
One of the most widely used methods to determine pre-eruptive
volatiles in magmas is to analyze melt inclusions trapped in
phenocrysts (Newman et al., 1986; Skirius et al., 1990; Lowenstern,
2003). Quenched melt inclusions can retain their dissolved volatiles,
especially if they remain sealed after entrapment. Quartz is one of the
best minerals for inclusion studies because its simple composition
allows minimal chemical exchange with the included melt, it has no
planar weaknesses that promote leakage, and it is transparent. Quartz
is abundant in the YTT, MTT, and OTT and was thus chosen as the host
mineral for this melt inclusion study. We used melt inclusions in
quartz crystals to determine the dissolved H2O, CO2, S, Cl, and F
contents of the YTT magma. In an effort to characterize the evolution
of the Toba magma system we also studied melt inclusions from the
MTT, OTT, and post YTT lava domes (PYLD) that erupted shortly after
the YTT eruption. Major and trace element analyses were also
conducted to characterize melt evolution and provide context to the
volatile data.
2. Toba composition and mineralogy overview
The quartz-bearing Toba magmas were compositionally and
mineralogically zoned (Chesner, 1998). Bulk rock YTT pumices range
from 68 to 77 wt.% SiO2, welded samples from the MTT contain 72
76 wt.% SiO2, OTT rocks have 6974 wt.% SiO2, and PYLD contain 70
76 wt.% SiO2. Their phenocryst mineralogy consists of quartz,
plagioclase, sanidine, biotite, and amphibole, with accessory orthopyroxene, magnetite, allanite, zircon, and ilmenite. Fayalite has been
noted in the YTT and MTT. The less silicic YTT pumice samples (68
73 wt.% SiO2) represent 25% of the YTT pumice suite, have the highest
crystal contents (3040 wt.%), contain little or no sanidine, and feature
shattered quartz phenocrysts with euhedral outlines up to 2 cm in
diameter. Pumices with 7377 wt.% SiO2 comprise the remaining 75%
of the suite, have lower crystal contents (1225 wt.%), and contain
smaller crystals. Crystallinity and crystal size relationships for the MTT
and OTT magmas cannot readily be determined from welded tuff
samples. Irontitanium oxide temperatures are 701780 C, 743
751 C and 704759 C for the YTT, MTT, and OTT respectively. Oxygen
fugacities generally range between 16 and 14 log units, with the
lowest values in the most silicic magmas. Chesner (1998) attributed
the compositional zonation of the Toba tuffs to extensive crystal
fractionation from a rhyodacitic source. Although Beddoe-Stephens
et al. (1983) have studied melt inclusions in Toba rocks, their work was
done prior to recognition that multiple tuffs erupted from the caldera
complex, and without the benets of Fourier Transform Infrared
Spectrometer (FTIR) analysis. A small subset of YTT melt inclusions
were analyzed by FTIR in a pilot study and indicated H2O contents of
4.95.5 wt.% and CO2 concentrations of 0200 ppm (Newman and
Chesner, 1989).
3. Sample and melt inclusion selection
Although melt inclusions are common in Toba quartz, not all melt
inclusions are suitable for study. In rocks that remain hot for an
extended period of time after eruption, such as those that weld or are
in the interior of non-welded ignimbrite, melt inclusions commonly
devitrify (Skirius et al., 1990). This can result in a change from
compositional homogeneity to heterogeneous domains that no longer
represent the pre-eruptive melt composition. Slow cooling can also
result in differential shrinkage of the melt and the host crystal, forming
a shrinkage bubble prior to solidication of the glass (Roedder, 1984).
Conceivably, volatiles could diffuse from the melt into the shrinkage

bubble, reducing the dissolved volatile content of the melt. Tait (1992),
however, has shown that volatiles remain dissolved in the melt unless
the crystal cracks. Consequently, shrinkage bubbles are expected to
have minimal effect on dissolved volatile contents of inclusion glass,
although post-quenching volatile diffusion into the bubble is possible.
During the violent process of explosive eruption, crystals can crack and
intersect inclusions, or over-pressurized large inclusions can burst,
fracturing their host crystals (Anderson et al., 1989; Tait, 1992). When
crystals crack, volatiles can exsolve from the melt and form multiple or
variably sized vapor bubbles. Thus, the ideal inclusions to study are
colorless with minimal signs of devitrication, vapor bubble free, have
no visible cracks, and to accommodate sample preparation and FTIR
analysis, are N100 m in their maximum dimension.
Quartz crystal fragments and occasional complete crystals were
hand-picked from several disaggregated samples of YTT, MTT, OTT,
and PYLD that span the compositional range of each unit. These
separates were then examined in immersion oil to characterize their
melt inclusions and identify the best samples for further study. Several
YTT pumices and MTT and OTT welded tuff samples were excluded
from study due to signicant devitrication of inclusion glass. A total of
5 YTT samples (Table 1) were selected for study and consist of pumices
from each of the three varieties described by Chesner and Rose (1991)
and Chesner (1998) as follows: 1) Type A Low-SiO2 (Type A-LS)
coarse (quartz phenocrysts up to 2 cm), crystal-rich (3040 wt.%)
pumices with abundant biotite and 6873% SiO2 (samples 21A4 and
12A1), 2) Type A High-SiO2 (Type A-HS) ner-grained, moderate
crystal content (1225 wt.%) pumices with much less biotite and 73
76% SiO2 (samples 23A4 and 55A1) and, 3) Type B-rare crystal poor
pumice (sample 5B3) with sparse biotite and the most evolved
composition of all Toba rocks (N76% SiO2). These pumice blocks were
collected from widely spaced exposures around the caldera, mostly
from the outow sheet near the caldera rim (Fig. 1). Three welded tuff
samples from the MTT (Table 2) representing a composite stratigraphic section were studied. From bottom to top, these samples
consist of a black basal vitrophyre collected near Haranggoal (sample
99), a light gray eutaxitic tuff with glassy amme (sample 7A)
collected up-section from the vitrophyre near Haranggoal, and a white
eutaxitic tuff with gray amme (sample 66) collected at the top of the
MTT section near Sipisupisu (Fig. 1). Based upon bulk rock samples
collected from the Haranggoal section, Chesner (1998) interpreted the
MTT as a reverse compositionally zoned tuff. Stratigraphic relations
among the 3 densely welded OTT samples are not as well constrained
and consist of 2 samples collected from the Uluan resurgent dome
(samples 77 and 25A), thought to represent later phases of the OTT
eruption, and a sample of welded outow collected in the deeply
incised Asahan River canyon (sample 85), presumable emplaced
during the earlier part of the eruption (Table 2, Fig. 1). The 2 PYLD
samples (Table 1) were collected from the Tuk-Tuk peninsula on
Samosir Island, the main resurgent dome associated with the YTT
eruption. Tuk-Tuk consists of several post YTT lava domes that were
emplaced along 3 parallel faults. One sample was selected from the
easternmost fault (TT-7) while the other came from the westernmost
fault (TT-2). These lava domes are white to light gray mineralogical
and chemical equivalents of the YTT. Furthermore their 40Ar/39Ar ages
are indistinguishable from the YTT (McIntosh, personal communication 1999).
A minimum of 6 inclusions that approached the ideal characteristics (outlined above) as closely as possible were then chosen from
quartz separates of each sample. In order to study inclusions located
near the rims and cores of intact crystals, one large (1 cm) quartz
phenocryst was selected from thick-sections of each YTT pumice
sample. Virtually all Toba melt inclusions, regardless of their location
within the host crystal or their size, are hexagonal dipyramidal
negative crystal shapes (Fig. 2). Exceptions consist of hourglass
inclusions (Anderson, 1991) with or without a connection to a crystal
edge, large irregular ovoid inclusions in the interiors of a few euhedral

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

263

Table 2
Representative analyses of MTT and OTT melt inclusions and their whole rock and matrix glasses. All analyses have been normalized to 100% on a volatile free basis, and all Fe is
reported as FeO. Melt inclusion (MI) and matrix glass (MG) major elements were determined by microprobe, and whole rock (WR) major elements by XRF. Water is by FTIR and Cl by
microprobe. Samples are arranged from left to right in order of decreasing SiO2 contents of their bulk rocks. MTT samples are 6699, OTT samples are 8577.
Sample

66 WR

66-1 MI

7A WR

7A-1 MI

99 WR

99 3AMI

99 MG

85 WR

85-1 MI

27 MG

25A WR

25A1 MI

77 WR

77-2 MI

SiO2
TiO2
Al2O3
FeO
MnO
MgO
CaO
Na2O
K2O
P2O5
H2O
Cl

77.93
0.12
13.47
0.68
0.01
0.18
0.58
2.56
4.44
0.02

76.56
0.04
12.44
1.14
0.11
0.66
3.90
5.15

4.36
0.15

76.12
0.16
13.15
1.65
0.05
0.24
1.24
3.19
4.18
0.02

75.03
0.05
13.40
1.25
0.01
0.01
0.64
3.67
5.94

5.03
0.16

74.42
0.26
14.14
1.70
0.04
0.41
1.72
2.87
4.42
0.02

74.50
0.03
13.44
1.30
0.10
0.02
0.73
4.47
5.40

3.25
0.16

77.19
0.06
12.57
0.75
0.04
0.01
0.73
3.12
5.51
0.02

73.19
0.22
13.88
2.23
0.06
0.47
1.82
3.48
4.58
0.06

74.75
0.09
13.59
1.23
0.10

0.60
3.72
5.91

4.32
0.21

77.54
0.06
12.38
0.66
0.08
0.01
0.70
2.51
6.05
0.01

71.29
0.37
14.34
3.14
0.06
0.78
2.68
3.41
3.82
0.10

76.76
0.02
12.64
1.04
0.08

0.72
3.53
5.21
0.01
4.66
0.14

69.19
0.51
14.76
3.98
0.08
0.94
3.44
3.43
3.51
0.16

76.98
0.03
12.37
1.15
0.07
0.01
0.63
3.26
5.49
0.02
2.60
0.19

crystals from the Type B YTT sample (5B3), and a few prismatic
shaped negative crystal inclusions in a large quartz phenocryst from
YTT pumice 55A1. Negative crystal shapes are thought to develop
from primary ovoid inclusions as a result of post entrapment
dissolution and recrystallization processes (Chaigneau et al., 1980;
Beddoe-Stephens et al., 1983; Manley, 1996). Consequently, ovoid
inclusions were trapped shortly before eruption, while negative
crystal shapes are recrystallized inclusions. The preponderance of
negative crystal inclusions in the Toba samples suggests that the large
quartz crystal fragments and single phenocrysts chosen for study are
from quartz crystals that resided in the magma for some time prior to
eruption. Hexagonal dipyramidal shape can be attributed to recrystallization at temperatures greater than the inversion temperature of 573 C. YTT and PYLD inclusions chosen for study were
typically 100200 m in size, while those from the MTT and OTT were
100150 m. Most larger inclusions are intersected by fractures and
contain variably sized vapor bubbles indicative of leakage. In addition
to colorless inclusions from each sample, a few tan ones from a YTT
Type A-HS sample (23A4) were also studied. Most inclusions in PYLD
sample TT-7 are light tan in color. Inclusions from 3 YTT samples and
one post YTT lava dome were vapor free, while all welded tuff
samples, 2 YTT pumice samples, and one PYLD sample yielded
inclusions with tiny vapor bubbles. These bubbles ranged in size from
about 10% of the inclusion diameter in the YTT and PYLD samples to
about 20% in the MTT and OTT samples. Inclusions with identical
vapor bubble to inclusion volume ratios were chosen from these
samples to ensure that the vapor bubbles were due to shrinkage and
not leakage from cracking of the host crystal. Generally, inclusions
containing tiny vapor bubbles also showed slight evidence of
devitrication (evenly dispersed incipient spherulitic microlites)
and contain a small colorless daughter crystal attached to the
bubble/inclusion walls. Inclusions from the welded tuffs, typically
exhibit micro-cracks extending short distances from the apices of the
inclusions, sub-parallel to the c-axis. Skirius et al. (1990) observed
similar cracks in some melt inclusions from Bishop Tuff quartz and
suggest that they formed during the 1% volume shrinkage of quartz
during the inversion (Ghiorso et al., 1979; Roedder, 1984).
4. Preparation and analysis of melt inclusions
Each host crystal was ground until a window was opened into
the melt inclusion. The glass exposed in the window was then
polished in preparation for electron microprobe analysis. Melt
inclusion glass was analyzed for 10 major elements (SiO2, TiO2,
Al2O3, FeO, MnO, MgO, CaO, Na2O, K2O, and P2O5) as well as SO3 and
Cl, using the JEOL 8900 Super Probe at the Department of Mineral
Sciences, Smithsonian Institution. Typically, 3 spots, 20 or 30 m in
diameter were analyzed from each inclusion, and an average

composition determined. Spot to spot analyses within the same


inclusions exhibited little compositional variation. A total of 152 melt
inclusions (78 YTT: 58 in quartz fragments and 20 in large single
crystals, 29 MTT, 31 OTT, 14 post YTT lava domes) was analyzed
(Tables 1 and 2). Matrix glass analyses were done on pumiceous glass
in polished sections from the YTT and PYLD samples, and hand-picked
vitric domains from MTT and OTT vitrophyre samples. The vitric
domains had conspicuous perlitic cracks, and contained tiny microlites that were generally avoided during analysis. An average matrix
glass composition was determined for each sample from a minimum
of 10, 20 m spot analyses (Tables 1 and 2).
After microprobe analyses, the analyzed surfaces were repolished
and then wafers were prepared for FTIR analyses. The host crystals were
ground to expose a window into the opposite side of the inclusion, and
then polished. During this process, special efforts to retain maximum
thickness were made; nished wafers were 30160 m thick.
Inclusion thicknesses were measured accurately with a petrographic
microscope by orienting each inclusion wafer vertically, and using the
calibrated ocular scale. Inclusions in large wafers were also measured
with a digital micrometer. Using a microscope equipped with a Bio Rad
(Digilab) Excaliber FTS 3000 FTIR unit, each inclusion was then analyzed
for H2O and CO2 by collecting near-IR (30008000 cm 1) and mid-IR
(4004000 cm 1) spectra on two overlapping areas of about 1600 m2.
FTIR data was collected on 103 separate melt inclusions (54 YTT: 39 in
quartz fragments and 15 in large single crystals, 20 MTT, 20 OTT, 12 post
YTT lava dome). Absolute peak heights from the 5230 cm 1 and
4520 cm 1 H2O and 2350 cm 1 CO2 absorption peaks on the spectra, in
conjunction with thickness and glass densities, were used to calculate
the wt.% H2O and ppm CO2 of each melt inclusion following the methods
of Zhang et al. (1997), Blank et al. (1993), and Newman et al. (1986). If
the data quality and results were similar for the 2 overlapping areas
analyzed in each inclusion, they were averaged, otherwise, the poorest
analysis was rejected.
Wafers from 4 YTT and 2 PYLD samples were then mounted on an
aluminum block and taken to the Plasma Laboratory in the
Department of Geological Sciences, University of Maryland. Laser
ablation inductively coupled plasma mass spectrometer (LA-ICP-MS)
analyses were made on 36 melt inclusions (22 YTT, and 14 PYLD).
These analyses (Table 1) provided data on 24 trace elements (Ni, Rb,
Sr, Y, Zr, Nb, Cs, Ba, La, Ce, Nd, Sm, Eu, Gd, Tb, Ho, Er, Yb, Lu, Hf, Ta, Pb,
Th, and U). In order to provide context to this data, ICP-MS trace
element analyses were done on bulk rock and glass separates from the
5 YTT pumices and 2 PYLD samples at Washington State University
(Table 1). Major element XRF bulk rock analyses for all samples are
from Chesner (1988).
A small subset (13) of YTT inclusions and 1 matrix glass sample were
analyzed by ion-probe for H2O, CO2, S, Cl, and F at the Department
of Terrestrial Magnetism, Carnegie Institution.

264

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

5. Major and trace element geochemistry


Melt inclusions analyzed from all ignimbrites and PYLDs are
rhyolitic in composition and have similar SiO2 ranges (YTT 7477, MTT
7377, OTT 7477, and PYLD 7677 wt.% SiO2). About 98% of their
anhydrous total consists of SiO2, Al2O3, K2O, and Na2O; the remaining

2% is mostly FeO and CaO. The entire suite has overlapping major
element trends and individual units cannot be distinguished by their
melt inclusion compositions. In this study, we refer to elements whose
concentrations diminish relative to increasing SiO2 contents within a
particular data group (bulk rock, melt inclusion, or matrix glass) as
compatible elements. Those elements that increase relative to SiO2

Fig. 2. Examples of melt inclusions in quartz from the YTT, MTT, OTT, and PYLD. In each unit and the PYLD, samples are arranged from top to bottom in order of decreasing SiO2. Rare
irregular shaped and tan colored inclusions are shown in samples 5B3 and 23A4 respectively. One of the single large quartz phenocrysts used in core-to-rim analyses is also shown
(sample 21A4). Microprobe and FTIR analysis windows can be seen in the high magnication views of MTT and OTT inclusions. OTT samples 85 and 25A have daughter crystals
between the vapor bubble and inclusion wall.

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Fig. 2 (continued).

265

266

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Fig. 3. A. Compositional variation diagrams for the YTT samples, showing melt inclusion, whole rock, and matrix glass analyses for the 5 YTT samples that were studied (w = whole
rock analysis, mg = matrix glass). B. Matrix glass and melt inclusion analyses for PYLD samples and a limited set of YTT samples showing contrasting trends. Legend: YTT samples are
yellow = 5B3, pink = 55A1, purple = 23A4, green = 12A1, red = 21A4. PYLD samples are blue = TT-2, black = TT-7. Same color open symbols are matrix glasses, whole rock analyses
are indicated by same color different symbols that are lled.

within a data group on variation diagrams are referred to as


incompatible. Thus, bulk rock, melt inclusions, and glass matrix analyses
with higher SiO2 contents and incompatible elements are assumed to be
more evolved than those with lower concentrations.
5.1. YTT and PYLD
Melt inclusions from the Type A-LS, Type B, and PYLD samples
contain 7677 wt.% SiO2, while those from the Type A-HS samples

have 74.576 wt.% SiO2 (Fig. 3A). Variations in Al2O3 (1.5 wt.%), K2O
(1 wt.%) and Na2O (1 wt.%) are apparent on Harker diagrams, while
smaller variations in CaO and TiO2 occur as well. Concentrations of
FeO and MgO vary little across the suite. Melt inclusions from the Type
A-HS samples show a greater range in major element compositions
but generally are not as evolved as inclusions from the Type A-LS
samples. Inclusions from the Type B sample have narrow major
element chemical ranges, similar to those from Type A-LS samples.
The two Type A-LS pumices and one of the PYLD samples (TT-7)

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

have melt inclusions that are far more silicic than their bulk rock
samples, while the Type A-HS, Type B pumices, and the other PYLD
sample (TT-2) either overlap with, or are less silicic than their bulk
rock counterparts. Although K2O increases with SiO2 content in the
whole rock suite, it decreases with respect to SiO2 in the melt
inclusion and glass matrix samples. Trace element concentrations of
the inclusion subset that was analyzed by LA-ICP-MS (Fig. 4) have
considerable variation (Ba b 10485 ppm, Sr b 1058 ppm, Rb 250
830 ppm). The only trace elements in the melt inclusion suite that show

267

compatible element trends are Sr, Ba, and Eu. Glass matrix analyses have
narrow major element compositional ranges (SiO2 = 76.577.5 wt.%),
but signicant ranges in their trace elements (Ba 17881 ppm, Sr 11
79 ppm, Rb 177385 ppm). Major and trace element trends in the glass
matrix data are parallel or sub-parallel to whole rock trends (except
K2O) and are consistent with fractionation of the Toba mineral
assemblage. A few elements that strongly partition into accessory
minerals such as zircon and allanite (Zr, La, Ce, and Nd) show compatible
matrix glass trends and incompatible melt inclusion trends (Fig. 5).

Fig. 4. Trace element plots of a subset of the YTT samples (A) and PYLD samples (B) that were analyzed by LA-ICP-MS. Whole rock and matrix glass analyses are included (w = whole
rock analysis, mg = matrix glass). Legend is the same as in Fig. 3.

268

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

than their matrix glasses with respect to major elements and compatible
trace elements (lower concentrations). This relationship, exemplied by
CaO, TiO2, Sr, Ba, and Eu, indicates that the host quartz crystals did not
crystallize from their surrounding matrix, but must have been inherited
from a more evolved melt. This concept is supported further by a subset
of melt inclusions whose distance from a crystal face has been
established. Melt inclusions from the Type A-LS magma show clear
trends in compatible major elements (TiO2 and CaO) with lower
concentrations in the core and higher concentrations near the rim
(Fig. 6), consistent with settling into less silicic melt. Type A-HS
inclusion core to rim trends are inconclusive, while crystals from Type B
magma (sample 5B3) clearly show opposite trends, with inclusions
containing lower compatible element concentrations near their rims,
consistent with in-situ crystallization. Sample 5B3 is also the only
pumice that contains some euhedral quartz crystals with ovoid shaped
inclusions. Such inclusion morphology has been interpreted to
represent inclusions that were captured shortly before eruption, unlike
those with negative crystal shapes associated with longer magma
residence times (Beddoe-Stephens et al., 1983). Crystals that resided in
the melt for short durations are more likely to remain in-situ, than those
that existed in the melt for long periods of time, assuming similar size
and sinking rates. Additional evidence for the Type A-LS quartz crystals
originating in a more evolved melt is the similarity between their
inclusion major element chemistry and those of inclusions from sample
5B3 that are presumed to have formed in the most evolved roof melts.
Crystal settling, magma mixing, and inux of new melts, have been
invoked to explain similar melt inclusion trends in the Bishop Tuff
(Wallace et al., 1999; Anderson et al., 2000; Peppard et al., 2001;
Hildreth and Wilson, 2007), Bandelier Tuff (Dunbar and Hervig, 1992),
Southwest Nevada Volcanic Field (Vogel and Aines, 1996) and Tarawera
Volcano (Shane et al., 2008) respectively.
Although major element and compatible trace element data are
suggestive of in-situ quartz crystallization in the most evolved pumice
samples and inheritance of quartz crystals in the least silicic rocks,
incompatible trace element data offers conicting evidence. In all
samples, incompatible trace elements (Rb, Y, Sm, Nb, Cs, Gd, Tb, Ho, Er,
Yb, Lu, Hf, Ta, Pb, Th, and U) are higher in the melt inclusions than in
their matrix glasses as shown by selected plots in Fig. 5. Incompatible
trace elements are typically enriched in inclusions by 23 and are
enriched as much as 6 in tan colored inclusions from sample 23A4.
Such enrichments of incompatible trace elements in the melt
inclusions should be expected in the Type A-LS pumices and TT-7
based on the concept of derivation from a more evolved melt, but is
inconsistent with the compatible major and trace element data for the
Type A-HS and Type B inclusions. In these samples, if indeed they
formed as a result of in-situ crystallization, their incompatible trace
element concentrations should be higher in their matrix glasses than
in their melt inclusions.

5.2. MTT
Fig. 5. Trace element plots showing contrasting whole rock, matrix glass, and melt
inclusion trends in the Zr and Ce plots. Only matrix glass and melt inclusions are plotted
in the Nb and U plots. Legend is the same as in Fig. 3.

Comparing melt inclusions to their matrix glasses results in patterns


that collectively cannot be explained by crystal-liquid fractionation.
Melt inclusions from the most silicic magma (Type A-HS, Type B, and TT2) have major element and compatible trace element concentrations
that are less evolved (higher concentrations) or overlap with their
matrix glasses. In-situ growth of quartz phenocrysts during crystalliquid fractionation of the Toba mineral assemblage (quartz, plagioclase,
sanidine, biotite, amphibole, and accessories) can reasonably explain
these chemical trends. In contrast, samples from the less silicic magma
(Type A-LS and TT-7) contain melt inclusions that are more evolved

Melt inclusions from the MTT welded tuffs (Fig. 7) were either less
silicic than their bulk rock samples (2 most silicic samples, 66 and 7A),
or overlapped (least silicic sample, 99). Unlike the YTT, none of the
bulk rock analyses were less evolved than melt inclusions from their
quartz phenocrysts. In the MTT, the most evolved sample (66) contains
the most evolved melt inclusions while the least evolved sample (99)
has the least evolved melt inclusions. The glass matrix composition of
the least evolved sample is signicantly more evolved than its melt
inclusions. Because of extensive devitrication, matrix glass analyses
were not possible for the more evolved samples. Although no trace
element data is available for the MTT inclusions, major element
geochemistry of the bulk rock, glass matrix, and melt inclusions
suggest that in-situ crystallization was the dominant process during
capture of the MTT melt inclusions.

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

269

Fig. 7. Major element compositional variation diagrams of the MTT inclusions, whole
rock samples, and matrix glasses (w = whole rock analysis, mg = matrix glass). Legend:
MTT samples are red = 66, blue = 7A, green = 99. Same color open symbols are whole
rock counterparts, and green X is matrix glass from sample 99.

Fig. 6. Rim-to-core analyses for inclusions that were analyzed in the large single crystals
from the YTT. A few analyses are also from quartz fragments where rims could be
identied (solid symbol 5B3 and 23A4 analyses). Legend: Colors are the same as
previous gures, but open symbols represent inclusion in large single crystals, whereas
closed symbols are inclusions from fragmented crystals.

5.3. OTT
All of the OTT melt inclusions are more silicic than their bulk rock
samples (Fig. 8). Many of the same compositional relationships noted
in the YTT pumices are also observed in the OTT. The least evolved
whole rock samples contain the most evolved melt inclusions (sample
77), the most evolved bulk rock samples contain the least evolved melt

270

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

inclusions, the similarities between the chemical patterns of the YTT


lead us to conclude that quartz crystals in the least silicic sample were
inherited from more evolved melt while those in the more silicic rock
are representative of in-situ crystallization.
6. Volatile contents
6.1. YTT

Fig. 8. Major element compositional variation diagrams of the OTT inclusions, whole
rock samples, and matrix glasses (w = whole rock analysis, mg = matrix glass). Legend:
OTT samples are purple = 85, yellow = 25 A, gray = 85. Same color open symbols are
whole rock samples, and the red + is vitrophyric matrix glass from sample 27.

FTIR analyses of individual inclusions typically range from 3.3 to


6.4 wt.% H2O, although most have between 4.0 and 5.5 wt.% H2O
(Fig. 9A). Ion-probe H2O values for a limited set of samples overlaps
this range. Although considerable overlap exists within the suite of
inclusions, the mean H2O contents of inclusions from the least silicic
Type A pumice (21A4) and the most silicic Type A pumice (55A1) are
respectively lower (4.1 wt.%) and higher (5.1 wt.%) than the rest of the
sample suite (4.6 wt.%). Thus, based on sample means, a 1% gradient in
H2O is apparent within the Type A pumices. The large quartz
phenocrysts prepared to study core to rim inclusions, indicate lower
H2O contents in the rims of the 3 least silicic samples; data from the
other 2 samples was inconclusive.
Evaluation of the mid-IR FTIR spectra indicated that many inclusion
wafers produced no 2350 cm 1 peaks at all. Spectra that displayed the
CO2 peak were typically collected from thicker wafers of Type A highSiO2 and Type B samples. FTIR CO2 contents range from 10 to 175 ppm,
mostly overlap in the 1075 ppm range, are similar to ion-probe
analyses (1060 ppm), and had overlapping ranges in most samples
(Fig. 9B). The only inclusions from Type A low-SiO2 samples that had
measurable CO2 were in the large intact phenocryst that was studied
from sample 12A1; sample 21A4 had no inclusions with detectable
CO2. Rim-to-core analyses of inclusions in pumices 12A1 and 23A4
indicate inclusions with the lowest CO2 contents occur closer to the
crystals' rims (Fig. 6).
Microprobe analyses of SO3 across the YTT suite result in S contents
that average about 16 ppm with maximum values of about 32 ppm.
These concentrations are near microprobe detection limits, however,
they are consistent with the small group of ion-probe analyses that
indicate S contents of 620 ppm for the YTT melt inclusions. Melt
inclusion and glass matrix SO3 data overlap and no variation is evident
among the sample suite or within the large quartz phenocrysts.
Chlorine contents of melt inclusions analyzed by microprobe range
from 1100 to 1800 ppm (Fig. 9C) and are consistent with the small
group of samples analyzed by ion-probe that have 11502000 ppm Cl.
Melt inclusions from the Type A-HS and Type B pumices generally
contain the most Cl; average Cl contents systematically increase from
about 1250 to 1625 ppm with increasing SiO2 content. In some
samples the matrix glasses contain considerably less Cl than the melt
inclusions, suggesting Cl degassing up to about 550 ppm; other
samples showed overlapping matrix and inclusion Cl contents
(Fig. 10). The large quartz phenocrysts from the Type A-LS pumices
(21A4, and 12A1) appear to have melt inclusions with lower Cl
contents near their rims vs their cores, whereas inclusions in more
silicic pumices (5B3 and 23A4) have higher Cl in their rims. Melt
inclusions from the ion-probe subset of samples have F contents
between 500900 ppm; one glass matrix contained about 200 ppm F
(Fig. 9D).
6.2. PYLD

inclusions (sample 85), matrix glass from the most evolved sample is
more evolved than its melt inclusions, and K2O increases with SiO2
content in bulk rock samples but decreases with respect to SiO2 in the
melt inclusions. Since all of the OTT rock samples are devitried,
matrix glass data of a basal vitrophyre sample that did not have
suitable melt inclusions (sample 27), was used as a proxy for the most
silicic sample. Even though we have no trace element data for the OTT

Water contents determined by FTIR of inclusions from the two


PYLD samples, overlap, range between 4.1 and 4.9 wt.% H2O, and show
much less variation than inclusions analyzed from the YTT samples
(Fig. 11A). Even though the bulk rock compositions of these samples
are similar to the compositional range of the YTT pumice samples (TT7 = 70% SiO2, TT-2 = 76% SiO2), there is no correlation between their
water contents and bulk rock chemistry. However, like the YTT

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

271

Fig. 9. Histograms of volatile contents determined on YTT melt inclusions. H2O (A) and CO2 (B) were determined by FTIR analyses, Cl (C) by electron microprobe, and F (D) by ionprobe. Lighter shade bars indicate melt inclusions from the subset of large single crystals. Samples are arranged from top to bottom by decreasing SiO2 contents. Same color,
horizontally aligned histograms are from the same sample. Histogram absence represents no detectable CO2 and no F analyses.

samples, the least silicic sample (TT-7) has no detectable CO2, while
the most silicic sample (TT-2) has 2585 ppm CO2 determined by FTIR
(Fig. 11B). SO3 contents determined by microprobe are near detection
limits, and indicate maximum S contents of 21 ppm with an average of
13 ppm. Microprobe analyses of Cl give an overall range of about

13001500 ppm (Fig. 11C). Inclusions in both samples average about


1400 ppm and show no correlation with bulk rock chemistry. Chlorine
contents of matrix glass from the most silicic sample overlap with melt
inclusion concentrations; however, the least silicic sample (TT-7)
suggests degassing of about 300 ppm Cl.

272

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Fig. 10. Cl vs CaO plot of selected samples from YTT, PYLD, MTT, and OTT that have
matrix glass Cl contents signicantly lower than their melt inclusions. Legend is the
same as Figs. 3, 7 and 8.

rock samples from both units. In the MTT, average H2O contents increase
from 3.5 wt.% in the least silicic rock to 5.1 and 4.3 wt.% in the most
evolved rocks. A water gradient is also evident in the OTT, where
average H2O contents vary from 2.4, 3.8, and 4.2 wt.% with increasing
SiO2 contents of their bulk rock samples.
FTIR analyses of melt inclusions from 2 of the MTT samples (7A and
66) have a range of about 2090 ppm CO2; one inclusion from 7A
contained about 250 ppm CO2. The most evolved bulk rock sample
(66) has inclusions with low CO2 contents (23 ppm or less), while
those in 7A average 62 ppm CO2. Absence of the 2350 cm 1 CO2 peak
on the FTIR spectra from sample 99 inclusions, may indicate that they,
like the least evolved bulk rock YTT and PYLD samples, contain the
least CO2. Alternatively, their undetectable CO2 could be due to
limitations in the FTIR analyses of smaller inclusions. Therefore, CO2
gradients in the MTT are indeterminable. Melt inclusions from the OTT
contain about 20125 ppm CO2, one inclusion from sample 85
contained about 280 ppm. In the OTT, inclusions with the highest
CO2 contents were from the most evolved bulk rock samples, which
also had the least silicic glass composition.
Sulfur contents measured in both the MTT and OTT inclusions by
microprobe average about 11 ppm S with maximum concentrations of
about 20 ppm S. As with the YTT, S contents cannot be correlated with
chemistry, and matrix glass S analyses overlap with the melt inclusions.
Chlorine contents of the melt inclusions range from 1200 to
1800 ppm in the MTT and 10002000 ppm in the OTT. Unlike the YTT,
there is no apparent correlation of Cl content with bulk rock chemistry
in the MTT or OTT. Considerable Cl degassing (about 450 ppm in the
MTT and 600 in the OTT) is suggested by matrix glass analyses with
distinctly lower Cl contents (Fig. 10).
7. Discussion

6.3. OTT and MTT


7.1. Magmatic evolution as suggested by melt inclusions
Water contents determined by FTIR in both the MTT and OTT
inclusions range from about 2.05.5 wt.% H2O (Figs. 12 and 13). Melt
inclusions with the lowest water contents are found in the least evolved

The melt inclusion data presented in this study provide important


constraints on the Toba magmatic system. Chesner (1998) attributed

Fig. 11. Histograms of volatile contents determined on PYLD melt inclusions. H2O (A) and CO2 (B) were determined by FTIR analyses, and Cl (C) by electron microprobe. Samples are
arranged from top to bottom by decreasing SiO2 contents. Same color, horizontally aligned histograms are from the same sample. Histogram absence represents no detectable CO2.

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

273

Fig. 12. Histograms of volatile contents determined on MTT melt inclusions. H2O (A) and CO2 (B) were determined by FTIR analyses, and Cl (C) by electron microprobe. Samples are
arranged from top to bottom by decreasing SiO2 contents. Same color, horizontally aligned histograms are from the same sample. Histogram absence represents no detectable CO2.

the compositional zonation of the Toba magmas to crystal fractionation of mostly quartz, plagioclase, sanidine, biotite, amphibole, and
lesser amounts of magnetite, ilmenite, allanite, and zircon from
parental magma similar to the least evolved Type A-LS bulk rock
compositions. The most evolved rock compositions could be produced
by about 40% fractional crystallization, and an additional 10%
fractionation was required to produce the matrix glass compositions
of the Type A-HS pumices. Melt inclusions indicate that the YTT melts
ranged in composition from only about 74.577 wt.% SiO2, not from
6877 wt.% SiO2 like the bulk rocks. Thus, instead of 68% SiO2 melts
evolving to higher SiO2 contents by crystal fractionation as previously
suggested (Chesner, 1998), high-SiO2 parental melts are required.
These high-SiO2 melts (N74.5 wt.% SiO2) may have stratied by
density and evolved further by in-situ fractionation, while magma
compositions could evolve by removal of crystals due to wallrock
crystallization or crystal settling to form the Type A-HS and Type B
magmas. Generation of the bulk rock compositions of the Type A-LS
pumices cannot be done by crystal fractionation from the high-SiO2
melts, but instead requires addition of a mineral assemblage whose
overall composition is far less silicic than the high-SiO2 parental melt.
Mixing calculations and petrographic observations (Chesner, 1998)
indicate that the Type A-LS pumices consist of about 3040 wt.%
crystals with up to 23% plagioclase, 8% quartz, 4% biotite, 4%
amphibole, and 1% FeTi oxides. Given the small overall range in
YTT melt composition recorded by melt inclusions and matrix glasses
(74.577.5 wt.% SiO2), such an assemblage cannot be explained solely

by in-situ fractionation. Thus, a signicant portion of the high crystal


content observed in these samples may be the result of crystals
settling and accumulating from higher levels of the magma body. We
believe that an open system, large volume, long-lived silicic magma
system with complex evolutionary dynamics similar to those
described by Anderson et al. (2000), Hildreth and Wilson (2007),
and Bachmann and Bergantz (2008) could supply the necessary
crystals. The resultant crystal mush zone would then consist mostly of
crystals that were inherited from more evolved melt, and inclusion
glass would be expected to be more silicic than the surrounding
matrix, as well as being far more evolved than the bulk rock
composition. Major and trace element, and core to rim melt inclusion
data are consistent with such a history for the Type A-LS samples.
Furthermore, these crystals are likely to be the oldest and largest
crystals in the magma body at the time of eruption. In contrast,
crystals in the Type A-HS and Type B pumices that erupted from the
upper parts of the chamber are more likely to contain melt inclusions
that are similar to, or less evolved than their matrix glasses. Major
elements and compatible trace elements of inclusions from the Type
A-HS and Type B pumices support this model, as well as core and rim
inclusion analyses from sample 5B3.
Enrichment of incompatible trace elements in the Type A-HS
inclusions over their matrix glasses cannot be easily explained by insitu crystallization. Because Type B pumice samples are rare, represent
the most evolved rocks in the system, have the smallest quartz crystals,
contain some quartz with irregular shaped inclusions, and have ample

274

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Fig. 13. Histograms of volatile contents determined on OTT melt inclusions. H2O (A) and CO2 (B) were determined by FTIR analyses, and Cl (C) by electron microprobe. Samples are
arranged from top to bottom by decreasing SiO2 contents. Same color, horizontally aligned histograms are from the same sample.

evidence of in-situ crystallization, their incompatible trace element


trends were expected to offer insights into the proposed model.
However, they too have inclusions with higher incompatible trace
element concentrations than their matrix glasses. Spandler et al. (2007)
have shown that REE concentrations in olivine melt inclusions
equilibrate with their surrounding matrix in b100 yrs. Thus, anomalous
melt inclusion compositions require both entrapment and crystal
settling/magma mixing to occur shortly before eruption. Although
Type B quartz and their inclusions have morphological evidence for
short pre-eruption residence times, Type A-HS quartz crystals and their
inclusions do not. Furthermore, major element and compatible trace
element data for both groups are consistent with in-situ crystallization.
Consequently, we do not believe that their ndings can be applied to
Toba quartz. Alternatively, incompatible trace element concentrations
could be elevated in a boundary layer around growing crystals. Bacon
(1989) has shown that if crystals grow faster than elements can diffuse
to them, then signicant gradients in incompatible elements can
develop immediately adjacent to growing phenocrysts. Growth rates
of Toba quartz, however, are presumed to be quite slow due to their
euhedral shapes (Watt et al., 1997). Boundary layer processes were
investigated by Lu et al. (1995) in quartz and sanidine from the Bishop
Tuff and thought to be negligible for inclusions greater than 50 m in
diameter. Thomas et al. (2003) analyzed trace elements in melt
inclusions from YTT zircon, allanite, plagioclase, and quartz crystals
and concluded that boundary layer processes did not affect the
compositions of melt inclusions that ranged in size from 20 to 100 m.
If diffusional boundary processes are insignicant, then complex magma

mixing is required to explain the incompatible trace elements. However,


we maintain that our model of crystal accumulation and in-situ
crystallization is consistent with the majority of the physical and
chemical data, and the possibility that diffusional boundary effects have
inuenced incompatible trace element compositions in quartz-hosted
melt inclusions at Toba should not be dismissed.
If the models presented above to explain the major and trace
element geochemistry of the YTT melt inclusions are valid, then
volatile data should be consistent with them as well. Dissolved H2O
and Cl contents in melt inclusions are the lowest in the Type A-LS
samples that presumably crystallized near the roof of the chamber and
settled to lower levels. At rst glance, this is inconsistent with the
model; however, considering the expected longevity of these large
crystals, they may be capturing roof melt before it evolved to higher
water and Cl contents. Indeed the melt inclusions from core to rim
indicate that the melt captured near the rims has less H2O and Cl than
melt captured earlier, indicating possible descent of the crystals into
lower portions of the magma. Crystals that formed more-or-less insitu in the upper parts of the magma (Type A-HS and Type B) have the
highest H2O and Cl contents, and may thus be younger than the
crystals that sank to deeper levels. Applying the same argument to
CO2 however is problematic. Our data indicates that the least silicic
Type A-LS sample (21A4) has undetectable CO2, and CO2 concentration
is invariant in the other samples. Core to rim analyses of single crystals
in other Type A samples suggest a decrease in CO2 contents towards
their rims. Settling crystals would be expected to have the least CO2 in
their earliest formed inclusions and higher CO2 contents toward

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

crystal rims. We believe the absence of CO2 in 21A4 is real because the
least silicic PYLD (TT-7) also has undetectable CO2. Low CO2 contents
could be consistent with crystallization near the roof of the chamber;
however, core to rim trends in single crystals cannot easily be accounted
for by crystal settling. Using dissolved H2O and CO2 contents from melt
inclusions, and crystallization temperatures from FeTi oxides (Chesner,
1998), gas saturation pressures (Newman et al., 1986) of about 1.1
1.4 kb were determined for inclusions in the Type A-HS and Type B
pumices, and the most evolved PYLD sample (TT-2). In a H2OCO2 gas
saturation plot (Fig. 14), the overlapping trends of these inclusions,
parallel to the pressure contours, suggests similar crystallization
pressures for the four samples that were plotted. This is consistent
with our interpretation that quartz from these samples crystallized insitu near the roof of the magma chamber. Although the Type A-LS
inclusions fall within the 1.11.4 kb gas saturation range as well, the
limited number of inclusions with measurable CO2 precludes further
interpretation.
Settling of quartz and sanidine crystals in the Bishop Tuff magma
chamber has been invoked by Peppard et al. (2001) and Anderson et al.
(2000) based on compositions of rim and core melt inclusions and
cathodoluminescent zoning. In order for settling to be a viable
mechanism, crystals must exist in the magma chamber for signicant
periods of time. Based on isotopic evidence, quartz crystals were
thought to have resided in the Bishop Tuff magma from 100 ky to as
long as 1.2 my prior to eruption (Christensen and DePaolo, 1993; Van
den Bogaard and Schirnick, 1995; Christensen and Halliday, 1996).
Using melt evolution and crystal settling arguments, Anderson et al.
(2000) concur that quartz phenocrysts record an evolutionary history
of 100 ky or more. Hildreth and Wilson (2007) conclude that zircon,
the earliest crystallizing mineral in the Bishop Tuff magma, provides
the most reliable age limits for the Bishop magma, about 100160 kyr.
Several lines of evidence suggest that the quartz phenocrysts in the
YTT also record a signicant duration of crystal-liquid evolution at
Toba. Black (personal communication, 1995) used U-series disequilibrium to date quartz crystals in the YTT. Quartz from two Type A-HS
pumice samples have ages close to the eruption age (74 ka), while
quartz from a Type A-LS sample has an age 40 ky older. This data
suggests that the large quartz in the least evolved magma have longer
magmatic residence times than the quartz in the most evolved samples,
and is consistent with the melt inclusion data suggesting that these
quartz were inherited from more evolved magma, possibly through the
process of crystal settling. Furthermore, Vazquez and Reid (2004) used
U/Th disequilibrium techniques to determine that allanite phenocrysts
resided in the Toba melts for 150 ky. The presence of faceted (negative
crystal) inclusions in all quartz crystals from Type A pumices suggests

Fig. 14. Dissolved H2O and CO2 contents for melt inclusions from Type A-HS and Type B
YTT pumices, and PYLD sample TT-2. Pressure contours determined at 725 C using
VolatileCalc 1.1 (Newman and Lowenstern, 2002). Legend is the same as in Fig. 3.

275

that these crystals resided in the magma chamber long enough for
solutionreprecipitation processes to transform irregular shaped inclusions into negative crystal shapes (Manley, 1996); inclusion composition has been shown to be unaffected by this maturation process (Lukacs
et al., 2002; Manley, 1996). Perhaps the most compelling argument for
mature quartz phenocrysts at Toba is their size. YTT pumices contain
euhedral quartz crystals up to 2 cm in diameter. The largest and most
abundant quartz crystals occur in the Type A-LS samples, although the
Type A-HS and Type B pumices can also contain unusually large quartz
crystals. These large quartz crystals are highly shattered and crumble
out of the pumice when exposed at the surface of a pumice block.
Although YTT quartz has a continuum of sizes from unusually large
phenocrysts to those that are only a few mm in diameter, it is likely that
many of the quartz fragments chosen for this study originate from these
large phenocrysts. If we use Anderson et al.'s (2000) estimated average
growth rate for Bishop Tuff quartz of 1 mm/100 ky, then Toba's quartz
could be up to 2 my old. Considering that no quartz-bearing rocks are
known at Toba prior to the OTT eruption 840 ky ago, it seems unlikely
for the quartz crystals to be much older than 1 my. An argument could
also be made that the large YTT quartz crystals are antecrysts derived
from the older MTT and/or OTT magma bodies or their crystallization
rinds. The euhedral nature of the quartz phenocrysts is inconsistent with
an antecryst origin. Large clots of subhedral plagioclase, quartz, biotite
in the Type A-LS samples could be interpreted as antecrysts, however,
considering the quartz crystal morphology, perhaps they originated by
synneusis. Clearly, the crystallization history of Toba quartz is long and
complex and has major implications on the evolution of the Toba
magmas. Our data suggests that the quartz crystals erupted within a
single eruptive cycle are likely to have vastly different ages, and
some have moved signicantly within the magma after their initial
crystallization. Thus, a better understanding of the age and growth of
Toba quartz is now required, and further discussions of Toba magmatic
evolution based upon melt inclusions in quartz must await the results of
a cathodoluminescence study of Toba quartz.
The nearly identical major element chemistry and volatile contents
of melt inclusions from the YTT, MTT, OTT, and PYLD indicates that the
Toba system is a long-lived magmatic system in which similar
processes have operated for up to 1 m.y. Melt inclusions from both
the YTT and OTT indicate that these large magma bodies evolved over
long periods of time, allowing early-formed quartz crystals to settle into
deeper levels of the magma. Other, presumably younger quartz crystals
found in the more evolved rocks that erupted from the upper levels of
the magma bodies, formed dominantly by in-situ crystallization. Melt
inclusions in the smaller reversely zoned MTT record only in-situ
crystallization. Geochemical patterns and the paucity of large melt
inclusions in the MTT vitrophyre compared to the later erupted tuffs,
suggests that large quartz crystals did not settle and accumulate in the
deeper portions of the smaller MTT magma chamber, and imply a
shorter residence time for the MTT magma prior to eruption. Much of
the data presented here is consistent with the rhyolite mush model
presented by Hildreth and Wilson (2007) at Long Valley caldera, and
the compositionally zoned magma chamber dynamics reviewed by
Bachmann and Bergantz (2008). Applying their models to Toba, the
eruption of the compositionally zoned quartz-bearing rhyolite tuffs in
the past 840 ky, represents the surcial expression of a compositionally
monotonous, large granitoid pluton, that has episodically provided
nearly identical composition melts to sub-volcanic magma bodies.
These water-rich, crystal poor melts then evolved for at least 150 ky
mostly by crystal-liquid fractionation. Near the roof of the chamber
magma became more silicic as the oldest and largest crystals (quartz
and possibly plagioclase, biotite, and amphibole) eventually settled to
lower levels. With time, accumulation of early-formed crystals in the
deeper portions of the chamber led to less silicic bulk magma
compositions and ultimately a compositional zoned magma body
whose bulk rock compositions were both more and less evolved than
the original melt.

276

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

7.2. Sulfur yields


Because S contents of melt inclusions overlap with those in matrix
glasses, it was not possible to calculate S degassing from any of the Toba
melts as hoped. Assuming that the low, near detection limit S
concentration in the melt inclusions is a reasonable bound for preeruptive S contents, broad constraints of S degassing can be made. If we
assume that all dissolved S in the melt inclusions was lost from the
magmatic melt during eruption, a maximum of 36 1014 g H2SO4
(range based on average and maximum S contents) aerosols could have
been produced from the 2800 km3 of YTT magma that erupted. Given
the low rates of S diffusion in silicic melts (Baker and Rutherford, 1996),
Scaillet et al. (1998) suggest that at most, 25% of the S dissolved in the
melt could exsolve. In this case, only 0.81.5 1014 g of H2SO4 might
have been added to the atmosphere during the YTT eruption. This mass
represents about 25 the total H2SO4 (3 1013 g H2SO4) generated
during the 1991 Pinatubo eruption (Bluth et al., 1994), and is similar to
the H2SO4 (0.93 1014 g) amounts that followed the 1815 Tambora
eruption (Devine et al., 1984; Rampino et al., 1988; Scaillet et al., 1998).
The observed temperature drop in the northern hemisphere in the year
following the Tambora eruption was 0.7 C (Stothers, 1984). Thus,
although the S melt inclusion data is not of high quality, it provides
constraints on S degassing, and supports Scaillet et al. (1998) and
Oppenheimer's (2002) assertions that previous estimates of signicant
global temperature perturbations (based on S emissions) following the
YTT eruptions were excessive. This new S yield estimate is 2 orders of
magnitude less than that determined for the YTT based on S fugacities
calculated from pyrrhotite compositions (Rose and Chesner, 1990).
Considering that pyrrhotite occurs as rounded blebs in YTT magnetite,
ilmenite, orthopyroxene, amphibole, allanite, and zircon phenocrysts
(Chesner, 1998), an origin as included immiscible sulde melt from
differentiated boundary layers is likely (Bacon, 1989). In this case, the S
fugacities of the silicate melt in such boundary layers would not be
representative of the overall YTT melt. Furthermore, inclusion of
pyrrhotite in early forming phases, and not in quartz and feldspars,
may be indicative of S fugacities of boundary layer melts captured
during the earliest stages of YTT magma evolution, instead of
immediately prior to eruption. Thus, we believe that constraints on
dissolved S in the YTT melts presented in this study supersede those
previously derived from pyrrhotite. The likelihood of a separate preeruption S-rich uid at Toba has been dismissed by Scaillet et al. (1998)
due to the low redox state (close to NNO, Chesner, 1998) of the YTT
magma. They also determined that any S stored in a uid phase would
be one order of magnitude less than that available from the melt, and
therefore have a negligible effect on the total S budget of the eruption.
Similar low total S yields for reduced, cool, silicic magma have been
reported for the small 1362 rfajkull eruption in Iceland (Sharma
et al., 2008), as well as for other large rhyolite eruptions such as Taupo
(Palais and Sigurdsson, 1989) and the Bishop Tuff (Anderson et al.,
1989).
Using similar arguments and calculations to those made for the
YTT, S emissions from the 60 km3 MTT and 500 km3 OTT can be made
as well. Assuming complete degassing, a total of 58 1012 g H2SO4
aerosols could have been generated from the small MTT eruption and
a total of 47 1013 g of H2SO4 from the much larger OTT eruption.
If degassing of 25% of this total occurred during the eruptions, then
12 1012 and 12 1013 g of H2SO4 would have been produced during
the eruptions respectively. Thus, the MTT sulfur emissions may have
been similar to those estimated petrologically from the 1883 Krakatau
eruption (8 1012 g H2SO4; Mandeville et al., 1996), while those from
the OTT were more similar to the 1991 Pinatubo eruption.
7.3. Chlorine and uorine yields
Microprobe analyses of Cl show that in some samples, the matrix
glasses had similar Cl contents to the melt inclusions, indicating that Cl

was not signicantly degassed. Matrix glass from other samples had
considerably less Cl than associated melt inclusions (up to 550 ppm),
allowing Cl degassing calculations. Assuming that the difference
between melt inclusion and matrix glass Cl contents represents Cl that
was lost during eruption, and using an average Cl loss of 200 ppm for the
5 YTT samples, a total of 1015 g of Cl may have been emitted to the
atmosphere. Although a Cl spike should be expected in the GISP2 ice
cores, Yang et al. (1996) report diminished Cl contents in the Toba
sulfate horizon. Using the same arguments, about 450 ppm and
600 ppm Cl was lost from the MTT and OTT magmas respectively
during their eruptions. Considering their volumes, a total of 6 1013 g
and 7 1014 g of Cl may have entered the atmosphere at 0.50 and
0.84 ma. Similar calculations can be done to estimate F degassing. Using
the average YTT melt inclusion F content of 675 ppm, and F loss of
450 ppm, then about 3 1015 g could have been lost to the atmosphere
during the YTT eruption. Although signicant quantities of Cl and F can
be removed from volcanic aerosols while they are still in the
troposphere, some portion of these volatiles may have reached the
stratosphere (Symonds et al., 1988). In light of the effects of these
halogens on ozone (Tie and Brasseur, 1995), the implications of
considerable Cl and F (an order of magnitude greater than S) associated
with the YTT event should be considered.
8. Conclusions
Melt inclusions in quartz crystals from the YTT indicate that the
parental melts for the Toba magmas were high-silica rhyolites, mostly
containing 74.577 wt.% SiO2. Compositional zonation of the preeruptive magma body from 68 to 77 wt.% SiO2 developed by crystal
fractionation near the roof of the chamber and crystal accumulation
near the base of the chamber. Melt inclusions have major and trace
element compositions that indicate the least evolved magma
contained quartz that had crystallized from more evolved melts. In
contrast, quartz crystals in the most evolved rocks have melt inclusion
compositions that are consistent with in-situ crystallization. Crystal
settling is the preferred mechanism of transport for the quartz crystals
based upon their extraordinary size (up to 2 cm), their longevity in
the magma chamber (at least 150 ky), and core to rim variation in
melt inclusion chemistry. Thus, in addition to compositional and
mineralogical zonation, the large YTT magma chamber had a gradient
in crystal ages with the oldest crystals near the oor of the chamber
and youngest crystals near the roof.
The YTT melts contained about 4.05.5 wt.% H2O and 20175 ppm
CO2 indicating gas saturation pressures of 1.11.4 kb. Water and
possibly CO2 gradients are evident with higher contents in inclusions
from the most silicic rocks. Low dissolved S contents (632 ppm) in
melt inclusions indicate a maximum of 36 1014 g H2SO4 aerosols
could have been generated during the YTT eruption, and based upon
the low S diffusivities of rhyolite melts, 0.81.5 1014 g H2SO4 is more
likely. This amount of H2SO4, which is far less than previously
speculated, is similar to the mass produced by the 1815 eruption of
Tambora. Dissolved Cl and F contents in the YTT melts were about
11001800 ppm and 500900 ppm respectively. The highest Cl
contents were in inclusions from the most evolved rocks and about
1015 g of both Cl and F could have been injected into the atmosphere
during the YTT eruption.
The much smaller MTT magma was dominated by in-situ
crystallization in which melts contained 2.05.5 wt.% H2O, b90 ppm
CO2, b20 ppm S, and 12001800 Cl. Only water shows a gradient with
higher contents in the most silicic samples. Processes similar to those
of the YTT magma chamber are suggested by melt inclusions in the
OTT. In-situ crystallization dominated the most silicic magmas, while
the least silicic magma appears to have inherited quartz crystals that
had crystallized from more silicic melt. Dissolved volatiles consist of
2.05.5 wt.% H2O, b125 ppm CO2, b20 ppm S, and 10002000 ppm Cl.
Like the YTT, gradients in water and CO2 are apparent. Similarities in

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

melt inclusion major element chemistry and dissolved volatiles


among the YTT, MTT, OTT, and PYLD strongly suggest that the Toba
magma system consists of a single granitoid pluton that episodically
supplies sub-volcanic magma chambers with water-rich, high-SiO2
rhyolite. Similar differentiation processes involving signicant
amounts of crystal settling over long durations of time have resulted
in the compositional variation of the large volume YTT and OTT.
Acknowledgements
This report is dedicated to the memory of co-author James F. Luhr. It
has been an honor and a privilege to have been taught the art of melt
inclusion preparation and analysis by such a legendary petrologist.
May his enthusiasm, professionalism, patience, and guidance be
carried on by all who have had the opportunity to know and work
with him. The Department of Mineral Science, National Museum
of Natural History, Smithsonian Institution and the Eastern Illinois
University Honors College supported this research. Electron microprobe and FTIR analyses were done at the Smithsonian Institution.
Amelia Logan assisted with the microprobe analyses and Tim Gooding
helped prepare the samples. Assistance with the MTT and OTT samples
was provided by Rachel Ens. LA-ICP-MS analyses were performed at
the University of Maryland's Plasma Mass Spectrometry Laboratory
under the direction of Bill McDonough. Ion-probe data were collected
in the Department of Terrestrial Magnetism at the Carnegie Institution.
The GeoAnalytical Laboratory at Washington State University analyzed the whole rock and glass matrix samples by ICP-MS. Steve Self
and Tom Vogel provided constructive reviews of this manuscript.
References
Acharyya, S.K., Basu, P.K., 1993. Toba ash on the Indian subcontinent and its implications
for correlation of late Pleistocene alluvium. Quaternary Research 40, 1019.
Aldiss, D.T., Ghazali, S.A., 1984. The regional geology and evolution of the Toba volcanotectonic depression, Indonesia. Journal of the Geological Society, London 141,
487500.
Anderson, A.T., 1991. Hourglass inclusions: theory and application to the Bishop
rhyolitic tuff. American Mineralogist 76, 530547.
Anderson, A.T., Newman, S., Williams, S.N., Druitt, T.H., Skirius, C., Stolper, E., 1989. H2O,
CO2, Cl, and gas in plinian and ash-ow Bishop rhyolite. Geology 17, 221225.
Anderson Jr., A.T., Davis, A.M., Lu, F., 2000. Evolution of Bishop Tuff rhyolitic magma based
on melt and magnetite inclusions and zoned phenocrysts. Journal of Petrology 41,
449473.
Bachmann, O., Bergantz, G.W., 2008. Deciphering magma chamber dynamics from
styles of compositional zoning in large silicic ash ow sheets. Reviews in
Mineralogy and Geochemistry, Minerals, Inclusions and Volcanic Processes, 69,
pp. 651674.
Bacon, C.R., 1989. Crystallization of accessory phases in magmas by local saturation
adjacent to phenocrysts. Geochimica et Cosmochimica Acta 53, 10551066.
Baker, L.L., Rutherford, M.J., 1996. Sulfur diffusion in rhyolitic melts. Contributions to
Mineralogy and Petrology 123, 335344.
Beddoe-Stephens, B., Aspden, J.A., Shepherd, T.J., 1983. Glass inclusions and melt
compositions of the Toba Tuffs, Northern Sumatra. Contributions to Mineralogy and
Petrology 83, 278287.
Bekki, S., Pyle, J.A., Zhong, W., Toumi, R., Haigh, J.D., Pyle, D.M., 1996. The role of
microphysical and chemical processes in prolonging the climate forcing of the Toba
eruption. Geophysical Research Letters 23, 26692672.
Blank, J.G., Stolper, E.M., Carroll, M.R., 1993. Solubilities of carbon dioxide in rhyolitic
melt at 850 C and 750 bars. Earth and Planetary Science Letters 119, 2736.
Bluth, G.J.S., Doiron, S.D., Krueger, A.J., Walter, L.S., Schnetzer, C.C., 1994. Global tracking
of the SO2 clouds from the June, 1991 Mount Pinatubo eruptions. Geophysical
Research Letters 21, 28332836.
Bhring, C., Sarnthein, M., Leg 184 Shipboard Scientic Party, 2000. Toba ash layers in
the South China Sea: evidence of contrasting wind directions during eruption ca.
74 ka. Geology 28, 275278.
Chaigneau, M., Massare, D., Clocchiatti, R., 1980. Contribution a l'etude des inclusions
vitreuses et des elements volatils contenus dans les phenocristaux de quartz de
roches volcaniques acides. Bulletin Volcanolgique 43, 233240.
Chesner, C.A., 1988. The Toba Tuff and Caldera Complex, Sumatra, Indonesia: Insights
into Magma Bodies and Eruptions. Ph.D. Thesis, Michigan Technological University,
Houghton.
Chesner, C.A., 1998. Petrogenesis of the Toba Tuffs, Sumatra, Indonesia. Journal of
Petrology 39, 397438.
Chesner, C.A., Rose, W.I., 1991. Stratigraphy of the Toba Tuffs and evolution of the Toba
Caldera Complex, Sumatra, Indonesia. Bulletin of Volcanology 53, 343356.

277

Chesner, C.A., Rose, W.I., Deino, A., Drake, R., 1991. Eruptive history of Earth's largest
Quaternary caldera (Toba, Indonesia) claried. Geology 19, 200203.
Christensen, J.N., DePaolo, D.J., 1993. Timescale of large volume silicic magma systems:
Sr isotopic systematics of phenocrysts and glass from the Bishop Tuff, Long Valley,
California. Contributions to Mineralogy and Petrology 113, 100114.
Christensen, J.N., Halliday, A.N., 1996. RbSr ages and Nd isotopic compositions of melt
inclusions from the Bishop Tuff and the generation of silicic magma. Earth and
Planetary Science Letters 144, 547561.
Dehn, J., Farrell, J.W., Schmincke, H.-U., 1991. Neogene tephrochronology from Site 758
on northern Ninetyeast Ridge: Indonesian arc volcanism of the past 5 Ma.
Proceedings Ocean Drilling Program, Scientic Results 121, 273295.
Devine, J.D., Sigurdsson, H., Davis, A.N., 1984. Estimates of sulfur and chlorine yield to
the atmosphere from volcanic eruptions and potential climate effects. Journal of
Geophysical Research 89, 63096325.
Diehl, J.F., Onstott, T.C., Chesner, C.A., Knight, M.D., 1987. No short reversals of Brunhes
age recorded in the Toba tuffs, north Sumatra, Indonesia. Geophysical Research
Letters 14, 753756.
Dunbar, N.W., Hervig, R.L., 1992. Volatile and trace element composition of melt
inclusions from the Lower Bandelier Tuff: implications for magma chamber
processes and eruptive style. Journal of Geophysical Research 97, 1515115170.
Ghiorso, M.S., Carmichael, I.S.E., Moret, L.K., 1979. Inverted high-temperature quartz.
Unit cell parameters and properties of the inversion. Contributions to
Mineralogy and Petrology 68, 307323.
Hildreth, W., Wilson, C.J.N., 2007. Compositional zoning of the Bishop Tuff. Journal of
Petrology 48, 951999.
Jones, G.S., Gregory, J.M., Stott, P.A., Tett, S.F.B., Thorpe, R.B., 2005. An AOGCM
simulation of the climate response to a volcanic super-eruption. Climate Dynamics
25, 725738.
Knight, M.D., Walker, G.L., Ellwood, B.B., Diehl, J.F., 1986. Stratigraphy, paleomagnetism,
and magnetic fabric of the Toba tuffs: constraints on the sources and eruptive
styles. Journal of Geophysical Research 91, 355382.
Lee, M.Y., Chen, C.H., Wei, K.Y., Iizuka, Y., Carey, S., 2004. First Toba supereruption
revival. Geology 32, 6164.
Lowenstern, J.B., 2003. Melt inclusions come of age: volatiles, volcanoes, and Sorby's
Legacy. In: De Vino, B., Bodnar, R.J. (Eds.), Melt Inclusions in Volcanic Systems;
Methods, Applications, and Problems, 5, pp. 121.
Lu, F., Anderson, A.T., Davis, A.M., 1995. Diffusional gradients at the crystal/melt
interface and their effect on the composition of melt inclusions. Journal of Geology
103, 591597.
Lukacs, R., Czuppon, G., Harangi, S., Szabo, C., Ntaos, T., Koller, F., 2002. Silicate melt
inclusions in ignimbrites, Bukkalja Volcanic Field, northern Hungary texture and
geochemistry. Acta Geologica Hungarica 45 (4), 341358.
Mandeville, C.W., Carey, S., Sigurdsson, H., 1996. Magma mixing, fractional crystallization and volatile degassing during the 1883 eruption of Krakatau volcano,
Indonesia. Journal of Volcanology and Geothermal Research 74, 243274.
Manley, C.R., 1996. Morphology and maturation of melt inclusions in quartz phenocrysts
from the Badlands rhyolite lava ow, southwestern Idaho. American Mineralogist 81,
158168.
Newman, S., Chesner, C.A., 1989. Volatile compositions of glass inclusions from the 75
Ka Toba Tuff, Sumatra. GSA Abstracts with Programs 21, 271.
Newman, S., Lowenstern, J.B., 2002. VolatileCalc: a silicate meltH2OCO2 solution
model written in Visual Basic for Excel. Computers & Geosciences 28 (5), 597604.
Newman, S., Epstein, S., Stolper, E., 1986. Measurement of water in rhyolitic glasses:
calibration of an infrared spectroscopic technique. American Mineralogist 71, 15271541.
Ninkovich, D., 1979. Distribution, age and chemical composition of tephra layers in
deep-sea sediments off western Indonesia. Journal of Volcanology and Geothermal
Research 5, 6786.
Ninkovich, D., Shackleton, N.J., Abdel-Monem, A.A., Obradovich, J.D., Izett, G., 1978. KAr
age of the late Pleistocene eruption of Toba, north Sumatra. Nature 276, 574577.
Nishimura, S., Abe, E., Yokoyama, T., Wirasantosa, S., Dharma, 1977. Danau Tobathe
outline of Lake Toba, North Sumatra, Indonesia. Paleolimnology Lake Biwa Japan
Pleistocene, 5, pp. 313332.
Oppenheimer, C., 2002. Limited global change due to largest known Quaternary
eruption, Toba 74 kyr BP? Quaternary Science Review 21, 15931609.
Palais, J.M., Sigurdsson, H., 1989. Petrologic evidence of volatile emissions from major historic
and pre-historic volcanic eruptions. In: Berger, A., Dickinson, R.E., Kidson, J.W. (Eds.),
Understanding Climate Change: Geophysical Monograph, pp. 3153.
Pattan, J.N., Shane, P., Banakar, V.K., 1999. New occurrence of youngest Toba tuff in
abyssal sediments of the central Indian Basin. Marine Geology 155, 243248.
Peppard, B.T., Steele, I.M., Davis, A.M., Wallace, P.J., Anderson, A.T., 2001. Zoned quartz
phenocrysts from the rhyolitic Bishop Tuff. American Mineralogist 86, 10341052.
Raj, R., 2008. Occurrence of volcanic ash in the Quaternary alluvial deposits, lower
Narmada basin, western India. Journal of Earth Systems Science 117, 18.
Rampino, M.R., Ambrose, S.H., 2000. Volcanic winter in the Garden of Eden: the Toba
super-eruption and the Late Pleistocene human population crash. In: McCoy, F.W.,
Heiken, G. (Eds.), Volcanic Hazards and Disasters in Human Antiquity: Geological
Society of America Special Paper, 345, pp. 7182.
Rampino, M.R., Self, S., 1992. Volcanic winter and accelerated glaciation following the
Toba super-eruption. Nature 359, 5052.
Rampino, M.R., Self, S., 1993. Climate-volcanism feedback and the Toba eruption of 74,
000 yr ago. Quaternary Research 40, 269280.
Rampino, M.R., Self, S., Stothers, R.B., 1988. Volcanic winters. Annual Reviews in Earth
and Planetary Science 16, 7399.
Robock, A., Ammann, C.M., Oman, L., Shindell, D., Levis, S., Stenchikov, G., 2009. Did the
Toba volcanic eruption of 74 ka B.P. produce widespread glaciation? Journal of
Geophysical Research 114, D1010710.1029/2008JD011652.

278

C.A. Chesner, J.F. Luhr / Journal of Volcanology and Geothermal Research 197 (2010) 259278

Roedder, E., 1984. Fluid inclusions. Mineralogical Society of America Reviews in


Mineralogy, 12. 644 p.
Rose, W.I., Chesner, C.A., 1987. Dispersal of ash in the great Toba eruption, 75 ka.
Geology 15, 913917.
Rose, W.I., Chesner, C.A., 1990. Worldwide dispersal of ash and gases from Earth's largest
known eruption: Toba, Sumatra 75 Ka. Global and Planetary Change 3, 269275.
Scaillet, B., Clemente, B., Evans, B.W., Pichavant, M., 1998. Redox control of sulfur
degassing in silicic magmas. Journal of Geophysical Research 103, 2393723949.
Shane, P., Westgate, J., Williams, M., Korisettar, R., 1995. New geochemical evidence for
the youngest Toba tuff in India. Quaternary Research 44, 200204.
Shane, P., Smith, V.C., Nairn, I., 2008. Millenial timescale resolution of rhyolite magma
recharge at Tarawera volcano: insights from quartz chemistry and melt inclusions.
Contributions to Mineralogy and Petrology 156, 397411.
Sharma, K., Self, S., Blake, S., Thordarson, T., Larsen, G., 2008. The AD 1362 rfajkull
eruption, S.E. Iceland: physical volcanology and volatile release. Journal of
Volcanology and Geothermal Research 178, 719739.
Skirius, C.M., Peterson, J.W., Anderson Jr., A.T., 1990. Homogenizing rhyolitic glass
inclusions from the Bishop Tuff. American Mineralogist 75, 13811398.
Smith, R.L., Bailey, R.A., 1968. Resurgent cauldrons. In: Coats, R.R., Hay, R.L., Anderson, C.A.
(Eds.), Studies in Volcanology: Geological Society of America, Memoir, 116, pp. 613662.
Spandler, C., O'Neill, H., St, C., Kamenetsky, V.S., 2007. Survival times of anomalous melt
inclusions from element diffusion in olivine and chromite. Nature 447, 303306.
Stauffer, P.H., Nishimura, S., Batchelor, B.C., 1980. Volcanic ash in Malaya from a
catastrophic eruption of Toba, Sumatra, 30,000 years ago. In: Nishimura, S. (Ed.),
Physical Geology of Indonesian Island Arcs. Kyoto University.
Stothers, R.B., 1984. The great Tambora eruption and its aftermath. Science 224,
11911198.
Symonds, R.B., Rose, W.I., Reed, M.H., 1988. Contribution of Cl- and F-bearing gases to
the atmosphere by volcanoes. Nature 334, 415418.
Tait, S., 1992. Selective preservation of melt inclusions in igneous phenocrysts.
American Mineralogist 77, 146155.
Thomas, J.B., Bodnar, R.J., Shimizu, N., Chesner, C.A., 2003. Melt inclusions in zircon. In:
Hanchar, J.M., Hoskin, P.W.O. (Eds.), Reviews in Mineralogy and Geochemistry,
Zircon, 53, pp. 6387.
Tie, X.X., Brasseur, G., 1995. The response of stratospheric ozone to volcanic eruptions:
sensitivity to atmospheric chlorine loading. Geophysical Research Letters 22,
30353038.

Toulmin, P., Barton Jr., P.B., 1964. A thermodynamic study of pyrite and pyrrhotite.
Geochimica Cosmochimica Acta 28, 641671.
Turco, R.P., Whitten, R.C., Toon, O.B., 1982. Stratospheric aerosols: observation and
theory. Reviews in Geophysics and Space Physics 20, 233279.
Van Bemmelen, R.W., 1949. The geology of Indonesia. In: Nijhoff, Martinus (Ed.),
General Geology of Indonesia and Adjacent Archipelagos 1A. Government Printing
Ofce, The Hague. 732 pp.
Van den Bogaard, P., Schirnick, C., 1995. 40Ar/39Ar laser probe ages of Bishop Tuff quartz
phenocrysts substantiate long-lived silicic magma chamber at Long Valley, United
States. Geology 23, 759762.
Vazquez, J.A., Reid, M.R., 2004. Probing the accumulation history of the voluminous
Toba magma. Science 305, 991994.
Vogel, T.A., Aines, R., 1996. Melt inclusions from chemically zoned ash ow sheets from the
Southwest Nevada Volcanic Field. Journal of Geophysical Research 101, 55915610.
Wallace, P.J., Anderson Jr., A.T., Davis, A.M., 1999. Gradients in H2O, CO2, and exsolved gas in
a large-volume silicic magma system: interpreting the record preserved in melt
inclusions from the Bishop Tuff. Journal of Geophysical Research 104, 2009720122.
Watt, G.R., Wright, P., Galloway, S., McLean, C., 1997. Cathodoluminescence and trace
element zoning in quartz phenocrysts and xenocrysts. Geochimica Cosmochimica
Acta 61, 43374348.
Westgate, J.A., Shane, P.A.R., Pearce, N.J.G., Perkins, W.T., Korisettar, R., Chesner, C.A.,
Williams, M.A.J., Acharyya, S.K., 1998. All Toba tephra occurrences across
peninsular India belong to the 75 ka eruption. Quaternary Research 50, 107112.
Whitney, J.A., 1984. Fugacities of sulfurous gases in pyrrhotite-bearing silicic magmas.
American Mineralogist 69, 6978.
Yang, Q., Mayewski, P.A., Zielinski, G.A., Twickler, M., 1996. Depletion of atmospheric
nitrate and chloride as a consequence of the Toba volcanic eruption. Geophysical
Research Letters 23, 25132516.
Zhang, Y., Belcher, R., Ihinger, P.D., Wang, L., Xu, Z., Newman, S., 1997. New calibration
of infrared measurement of dissolved water in rhyolitic glasses. Geochimica et
Cosmochimica Acta 61, 30893100.
Zielinski, G.A., Mayewski, P.A., Meeker, L.D., Whitlow, S., Twickler, M.S., Taylor, K., 1996.
Potential atmospheric impact of the Toba mega-eruption 71, 000 yr ago. Geophysical
Research Letters 23, 837840.

You might also like