You are on page 1of 23

Journal of Petrology, 2022, 63, 1–23

https://doi.org/10.1093/petrology/egac044
Original Manuscript

Making Andesites and the Continental Crust: Mind the


Step When Wet
Olivier Reubi* and Othmar Müntener

Institut des Sciences de la Terre, Université de Lausanne, Géopolis, CH-1015 Lausanne, Switzerland

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


*Corresponding author. E-mail: olivier.reubi@unil.ch

Abstract
Andesites are iconic of subduction zone magmatism. Yet intermediate magmas (57–66 wt % SiO2 ) are less abundant than generally
thought in arc settings. A comparison of experimental hydrous liquid lines of descent, melt inclusions and bulk-rock compositions
demonstrates the importance of polybaric crystallization–differentiation in producing the compositional range and dictating the
relative abundance of arc melts, but also highlights the preponderant role of mixing (sensu lato) in producing andesitic magmas. Based
on their P2 O5 contents, at least 74% of the arc magmas with around 64 wt % SiO2 are inferred to be mixing products. In addition to
their surprisingly low abundance, andesitic melt inclusions are characterized by relatively low H2 O, Al2 O3 , ± Na2 O contents compared
to the ranges measured in mafic and silicic melt inclusions. These compositional characteristics suggest that there is a sweet spot
for the production of andesitic melts delimited by the low-pressure stability limit of amphibole (<150 MPa) and the adiabatic ascent
path of mafic melts, but that this low-pressure differentiation pathway plays a minor role in the production of silicic arc magmas that
principally form along high-pressure hydrous liquid line of descents (>700 MPa) before decompression. The compositional bimodality
recorded by the melt inclusions and in well-preserved intra-oceanic arc crustal sections is a fundamental characteristic of differen-
tiation in transcrustal arc magmatic systems, with important consequences for the chemical evolution of the continental crust. We
propose that the overall bimodality shown by arc melts does not relate to a compositional gap in the differentiation mechanisms but
results from a combination of (1) the disparity in volume of differentiated magmas produced by low and high-pressure crystallization–
differentiation and (2) the strong nonlinearity of the high-pressure liquid lines of descent in composition–temperature–crystallinity
space related to crystallization of amphibole-rich assemblages. In this context, the compositional characteristics shared by andesitic
magmas and the continental crust principally depict the central role of mixing and mass balance processes in producing andesitic
compositions. The step in differentiation efficiency encountered by hydrous magmas entering the amphibole stability field at high
pressure plays an important role in defining the silicic component involved in these scenarios.

Keywords: continental crust, crystallization–differentiation, arc magma, andesite

INTRODUCTION crystallization–differentiation of mantle-derived basaltic


Intermediate to silicic magmas produced in subduction melts (Bowen, 1928; Sisson & Grove, 1993; Grove et al.,
zones are pivotal in the Earth’s evolution, since magma 2003); (2) partial melting of mafic lower crustal rocks
addition in the crust beneath volcanic arcs is thought (Rapp & Watson, 1995); (3) partial melting of pelitic
to be an important contributor to continental crust to granitic crust accompanied by variable amounts of
growth (Taylor & White, 1965; Gill, 1981). In this context, hybridization with mantle-derived melts (Hildreth &
the differentiation processes that produce intermediate Moorbath, 1988; Clemens et al., 2011); (4) direct partial
to silicic magmas from mantle-derived basaltic melts melting of hydrous peridotite to produce magnesian
are essential to produce a continental crust with an andesites (Kushiro, 1975; Kelemen, 1986; Grove et al.,
andesitic bulk composition (Taylor & McLennan, 1995; 2003) and (5) slab melting to account for andesites
Rudnick & Gao, 2003; Jagoutz & Kelemen, 2015). In and dacites with high Sr/Y ratios (Kay, 1978). Although
addition, intermediate to silicic arc magmas produced all these processes certainly contribute the chemical
among the largest and most destructive eruptions diversity of arc magmas, crystallization–differentiation
on Earth (Newhall & Self, 1982) and are associated or partial melting of mantle-derived hydrous basaltic
with significant economic resources (Richards, 2011). sources are emerging as the predominant processes
Therefore, understanding the mechanisms controlling producing differentiated arc magmas (Grove et al., 1997;
their production and evolution is an important task in Sisson et al., 2005; Blatter et al., 2013; Jagoutz, 2014;
geosciences. Nandedkar et al., 2014; Müntener & Ulmer, 2018; Ulmer
A number of processes have been proposed for the et al., 2018). Investigations of the relatively complete
production of differentiated arc magmas, including (1) crustal sections of the accreted Kohistan (Pakistan) and

Received: August 6, 2021. Revised: May 2, 2022. Accepted: May 6, 2022


© The Author(s) 2022. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com
2 | Journal of Petrology, 2022, Vol. 63, No. 6

Talkeetna (South Central Alaska) island arcs provide a is of prime importance to understand the magmatic
strong support for crystallization–differentiation as the systems feeding hazardous arc stratovolcanoes, the
prime process to produce differentiated arc magmas. formation of arc batholiths and ultimately the chemical
Here, the observation that cumulate rocks exposed in priming of the continental crust.
the lower crust of these sections are complementary Here we take advantage of the availability of extensive
to the differentiation trends formed by the mid- to melt inclusion and whole-rock compositional data, as
upper-crustal plutonic and volcanic rocks is a key well as experimental petrology data designed to under-
argument (Greene et al., 2006; Rioux et al., 2007; Jagoutz stand differentiation processes in magmatic arcs to fur-
et al., 2011). Additional arguments are provided by ther investigate the mode of production of intermedi-
experimental petrology data showing that the cumu- ate and silicic arc magmas. By combining the datasets,
lates derived from high-pressure (0.7–1 GPa) hydrous we document the principal differentiation pathways fol-
fractional crystallization experiments closely reproduce lowed by arc magmas in transcrustal magmatic systems,

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


the cumulate sequences exposed in the roots of island we evaluate their relative importance, as well as that of
arcs and produce rhyolitic liquids close to natural rock mixing in producing andesites, and we gauge the con-
compositions (Müntener & Ulmer, 2018). sequences for the evolution of the continental crust. In
Although fractional crystallization of hydrous melts exploiting the significantly larger datasets now available,
can generate the entire compositional range observed in we also shed new light on the debated importance of the
arc magmas, i.e. basalt to rhyolite (Nandedkar et al., 2014; dearth of andesitic melt previously identified by Reubi &
Ulmer et al., 2018), the prime mechanism responsible for Blundy (2009) in the overall differentiation scheme of arc
the abundance of intermediate andesitic magmas, which magmas.
are iconic of arc volcanism, is nevertheless controversial.
These magmas are generally porphyritic and typically
show petrographic characteristics of magma mixing METHODS
or mingling between silicic and mafic magmas (e.g. Data compilations
Eichelberger, 1978; Gill, 1981; Kent et al., 2010). As a The compilation of published melt inclusion (MI) data in
result, andesitic bulk-rock compositions often represent arc magmas (n = 8973) was extracted from the GEOROC
mixtures of liquids and crystals, which are frequently not global MI compilation (https://georoc.eu). Only the
co-genetic. How close these compositions are to the com- analyses corresponding to volcanism along the main arc
position of true magmatic liquids and what is the overall axes were considered here. Available analyses show that
abundance of true intermediate liquids in the volcanic MIs contain ≤10 wt % H2 O. Consequently, the analyses
and plutonic records are open to debate. On the one hand, with totals <90 or > 101.5 wt % (excluding H2 O contents
it is suggested that andesitic bulk rock compositions that are often not given) were removed from the dataset.
essentially correspond to the compositions of melts Whole-rock compositions (n = 39 636) were obtained from
forming the liquid line of descent (LLD) generated by the GEOROC pre-compiled file available for each arc.
crystallization–differentiation from hydrous basalts (e.g. Plutonic whole-rock compositions for relatively well-
Lee & Bachmann, 2014; Keller et al., 2015). In this view, preserved Mesozoic plutonic arc sections were obtained
magma mixing and mingling are secondary processes from the GEOROC and EarthChem Portal (http://portal.
paramount for petrographic and volcanologic concerns earthchem.org) databases. Analyses with totals <98.5
(e.g. Sparks et al., 1977; Cashman & Blundy, 2013), but or > 101.5 wt % were rejected. All data are provided
not necessarily fundamental to the genesis of andesites. as Supplementary data and are normalized to 100%
Accordingly, the abundance of intermediate magmas anhydrous. To limit the effect of regional sampling
may reflect preferential extraction of these liquids bias and to enable a comparison of MI and volcanic
from crystallizing magmatic systems and the andesitic whole-rock compositional distributions (Fig. 1), bootstrap
composition of the bulk continental crust may reflect a subsampling of both data sets was performed to obtain
primary magmatic feature. On the other hand, mixing an equal number of samples per 100 km along the arc
and mingling are regarded as the predominant processes strike (23 and 12 samples per 100 km for whole-rock and
producing the continuous range of compositions shown MIs, respectively). The curves shown in Fig. 1a and 1c
by arc magmas, hence masking a relative paucity of are the average relative sample distributions for 1000
melts with intermediate compositions (Anderson, 1976; subsampling iterations, and the gray areas represent ±
Eichelberger, 1978; Reubi & Blundy, 2009; Tatsumi & two standard deviation (SD) of the relative proportions.
Suzuki, 2009). In this case, the composition and abun- The compositional distribution of the full dataset and
dance of intermediate magmas depict essentially the the averaged subsets are, nevertheless, relatively similar.
mixing dynamics in magmatic systems and the relative The principal difference being that the amplitude of
paucity of intermediate melts is the prime characteristic silicic peaks around 75–78 wt % SiO2 for the MIs is
of the differentiation mechanisms. As a consequence, reduced in the subsets average. For plutonic rocks, the
the andesitic composition of the continental crust may compositional distribution is strongly biased by the
not be a primary magmatic feature. Establishing the extent of exposure and the depth of the exposed sections,
relative contributions of these contrasting scenarios limiting comparison between terranes. Consequently,
Journal of Petrology, 2022, Vol. 63, No. 6 | 3

we used the full dataset available for each section bulk continental crust
to establish the relative compositional distribution, (a)
Volcanic bulk-rock All arcs
although it is clear that these distributions are not 0.18
Oceanic arcs
representative of the complete arc crustal section.

Relative sample abundance


Continental arcs
MIs studies often focus on olivine-hosted inclusions 0.14
(42% of the dataset), thus creating an additional sam-
pling bias. To alleviate this issue, we performed bootstrap
0.10
subsampling to obtain a number of MIs per type of
host phase proportional to the overall modal abundance
of each phase in a complete fractional crystallization 0.06
sequence from a primary hydrous basalt to a rhyolite.
To this purpose, we used the total modal proportions 0.02

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


reported by Nandedkar et al. (2014) for fractional crys-
tallization experiments at 700 MPa from 1150 to 700 ◦ C. (b) Kohistan-Ladakh
Plutonic bulk-rock
Compared to the geographical subsampling, this signif- 0.08 Talkeetna

Relative sample abundance


Andean
icantly reduces the amplitude of the mafic peak and Sierra Nevada
enhances the silicic peak (Fig. 1c). Post-entrapment crys- & Peninsular Range
0.06
tallization of the host mineral is known to be an issue
in the case of MIs. The composition of MIs in olivine are
typically reported after correction for post-entrapment 0.04
crystallization, while data for MIs in other phases are
reported as raw compositions. This may create artifact
trends in the MI record and hinder comparison with bulk- 0.02

rock data. The principal chemical trends identified in this


work for intermediate and silicic compositions are, how-
ever, observed for MIs independently of their host phase, (c) All arcs melt inclusions
Volcanic Glass
while addition of few percent of the host phase to gauge 0.18 Izu Bonin matrix glass
melt incl. proportional
the effect of eventual post-entrapment crystallization is
Relative sample abundance

to host phase
sufficient to yield distinct chemical trends (Supplemen- 0.14
tary Figure S1). This indicates that post-entrapment crys-
tallization is typically limited to a few percent (<5%) and
0.10
has a restricted influence on the major trends identified
in Figs 2, 4 and 5. In addition, experiments have shown
that diffusive H2 O loss from MIs may be a substantial 0.06

issue (Gaetani et al., 2012). To which extent natural MIs


populations are affected by H2 O loss is more difficult to 0.02
assess, particularly in global compilations. This issue is
addressed in the relevant section of the discussion. 40 45 50 55 60 65 70 75 80 85
SiO2 wt%
The full volcanic bulk-rock dataset was used to con-
struct the contours in the compositional plots (Fig. 2). For Fig. 1. Relative sample abundance as a function of silica content for (a)
MIs, data from the Taupo rhyolite ignimbrite were omit- arc volcanic rocks, (b) arc plutonic rocks and (c) arc melt inclusions and
matrix glass. For the volcanic rocks, the continental arcs include the
ted in Fig. 2 as the large number of analyses (n = 1060) Aegean, eastern Aleutian, Andean, Mexican, Cascades and Kamchatka
within a tight range of rhyolitic compositions swamp arcs. The oceanic arcs comprise the central and western Aleutian,
the patterns for less evolved compositions. In Fig. 3, only Bismarck, Izu-Bonin, Kermadec, Kurile, Lesser Antilles, Mariana, New
Hebrides, Solomon and Tonga arcs. The Banda, Central American,
the compositions of plutonic rocks within the range of Honshu, Kurile, Luzon and Sunda with less well-defined crustal settings
experimental melts and volcanic bulk-rocks were plotted were included in the ‘All arcs’ dataset only. All data are from the
to keep the same scales. Hence, cumulate compositions GEOROC database. Melt inclusions data for all arcs are taken from the
GEOROC global melt inclusion database. For the volcanic rocks (a) and
strongly enriched in compatible elements are not repre- melt inclusions (c), the relative sample abundances are based on
sented, although they form a significant portion of the random subsampling to obtain an equal number of samples per 100 km
data set. along the arc strike (23 and 12 samples per 100 km for whole-rock and
melt inclusions, respectively). For the MIs, random subsampling was
High-pressure experiments aimed at reproducing the also performed to obtain a number of MI per host phase proportional to
composition of intermediate to silicic arc magmas were the total modal proportion of the host in fractional crystallization
compiled from Baker et al. (1994), Grove et al. (1997), experiments from 1150 to 700 ◦ C (experimental data from Nandedkar et
al., 2014). The curves shown in a and c are the average distributions for
Blatter et al. (2013), Nandedkar et al. (2014) and Ulmer 1000 subsampling iterations, the gray areas represent ± two SDs of the
et al. (2018). These experiments cover a range of pri- relative proportions. Bin size is 2 wt % SiO2 . Plutonic rock data are
mary and derivative hydrous arc-magmas, of pressures extracted from the GEOROC and EarthChem databases. The full dataset
for each batholith was used to calculate the probability density function
from 1.0 to 0.1 GPa and comprise both equilibrium and in (b). All data are normalized to 100% anhydrous and are provided as
fractional crystallization experiments at f O2 conditions Supplementary data.
4 | Journal of Petrology, 2022, Vol. 63, No. 6

Melt inclusions Bulk rock volcanic


80
X = 160 X = 250
(a) (b)
75

SiO2 wt% 70

65

60

55

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


50

45

(c) (d)
CaO wt%

X = 400 X = 550

(e) (f)
FeO wt%

X = 300 X = 400

MgO wt% MgO wt%


X
Nandedkar et al. (2014), FC, 700 MPa Blatter et al. (2013, EC, 400 MPa
X/2 Ulmer et al. (2018), Mg-basalt, FC, 1.0 GPa Grove et al. (1997), EC, 200 MPa
Ulmer et al. (2018), b-a, FC, 1.0 GPa Grove et al. (1997), EC, 100 MPa
0 bulk crust upper crust lower crust (Rudnick & Gao, 2003)

Fig. 2. Comparison of chemical variations in melt inclusions, volcanic bulk-rocks and experimental melts. Melt inclusions are from the full arc melt
inclusion GEOROC compilation excluding data from the Taupo volcanic zone. The volcanic bulk rock dataset comprises the GEOROC compilations for
all the arcs listed in Fig. 1. Color contours represent the number of sample per cell. Grids of 60 × 60 and a 40 × 40 cells were used respectively for the
volcanic rocks and the melt inclusions. X is the maximum number of samples per cell in each bivariate diagram. The liquid lines of descent are based
on experimental studies of hydrous basalts at the pressure given with the references. FC = fractional crystallization experiments, EC = equilibrium
crystallization experiments. The white dashed line represents an arbitrarily chosen mixing line for visual reference. Stars are the compositional
estimates for the lower, upper and bulk crust from Rudnick & Gao (2003). All data are normalized to 100% anhydrous.
Journal of Petrology, 2022, Vol. 63, No. 6 | 5

Melt inclusions Bulk rock volcanic

X = 60 X = 170
(i) (j)

Al2O3 wt%

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


X = 60 X = 150
(g) (h)
Na2O wt%

X = 45 X = 250
(k) (l)
P2O5 wt%

SiO2 wt% SiO2 wt%


Fig. 2. Continued

relevant for arc magma genesis. They were selected to considered here. Discussion of the relevance of these
cover a range of pressures reproducing crystallization at experiments to understand fractionation processes in
lower- to upper-crustal conditions and to cover a range the roots of magmatic arcs and comparison of the exper-
of temperature sufficient to produce a continuous LLD imentally produced liquids and cumulates with natural
from mafic to silicic compositions. As the MIs provide plutonic rocks were previously presented in Müntener
ample evidence that mafic arc melts contain significant & Ulmer (2018). We emphasize that the experimental
amounts of H2 O (2–6 wt %) at the onset of degassing, only studies were not designed to simulate LLD’s completely
experiments using hydrous starting compositions were as they were done at constant pressure.
6 | Journal of Petrology, 2022, Vol. 63, No. 6

(a) (b)
X = 50 X = 20

MgO wt%

Al2O3 wt%

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


SiO2 wt% SiO2 wt%

(c) (d)
X = 30 X = 60
P2O5 wt%

FeO

SiO2 wt% MgO wt%

X
Talkeetna Kohistan

X/2
Mid-crustal plutons Upper crust
Lower-crustal rocks Lower crust
0 Bulk crust

Fig. 3. Comparison of chemical variations in plutonic bulk-rocks and experimental melts. Color contours as in Fig. 2 for grids of 60 × 60 cells. Colored
curves are experimental liquid lines of descents as in Fig. 2. The white dashed line is the same reference mixing line as in Fig. 2. Bulk-rock data as in
Fig. 1. Stars are the compositional estimates for the lower, upper and bulk crust from Rudnick & Gao (2003). The average composition for the
mid-crustal plutons and lower crustal gabbroic rocks from the Talkeetna and Kohistan arc crustal section are also shown (data from Jagoutz &
Kelemen, 2015).

Analytical errors for major elements are small rela- Terminology


tive to the compositional ranges investigated and are, The intermediate peak observed in the compositional
consequently, omitted. In the case of the minor oxide distribution of volcanic arc rocks (Fig. 1a) straddles the
P2 O5 , the errors are not negligible and could impact andesite and part of the dacite fields defined by LeBas
the comparison between the bulk-rock data (XRF anal- et al. (1992) (57–63 wt % SiO2 for andesite). This peak
yses) and MI data (electron microprobe). In our quan- also encompasses the compositional estimates for the
titative estimates of the proportion of mixed magmas bulk continental crust, which range from 57.1 to 65.2 wt
based on P2 O5 , we took into consideration the calibration % SiO2 (Rudnick & Gao, 2003, and reference therein,
root mean square error (RMSE) and analytical uncer- Hacker et al., 2011). Here we focus on the petrogenesis of
tainties (SD) for P2 O5 obtained by measuring geologi- magma forming this intermediate peak and its potential
cal reference materials (rock powders and glasses) on relationship to the composition of the continental crust.
a PANalytical AxiosmAX x-ray fluorescence spectrometry We thus use indiscriminately the terms andesite and
(XRF) (RMSE = 0.004 and SD = 0.004) and a JEOL JXA-8530F intermediate to describe rocks with 57–66 wt % SiO2 . The
microprobe (RMSE = 0.011 and SD = 0.01) at the University term silicic and mafic are used respectively to describe
of Lausanne. rocks with >66 and < 57 wt % SiO2 . The term magma
Journal of Petrology, 2022, Vol. 63, No. 6 | 7

mixing unless otherwise specified is used in the geo- Andean batholiths essentially show a mirror image of
chemical sense to describe the mathematical compo- the above distributions, with a prominent broad silicic
sition resulting from mixing between silicic and mafic peak, which regionally comprises two shoulders, and a
end-members, irrespective of the actual physical process subdued mafic peak (Fig. 1b). The relative proportion of
that may include magma mixing sensu stricto, magma preserved mafic and silicic magmas in continental and
mingling or partial assimilation of crystal cumulates. oceanic sections may in part be an artifact of the expo-
sure depth, with deep mafic sections exposed in oceanic
arcs while mid to upper crustal plutons predominate
RESULTS in continental setting. Although the amount and geo-
On the occurrence of intermediate magmas graphic spread of the dataset does not allow for robust
A full understanding of the mode of production of differ- resampling to compare the plutonic sections, let alone
entiated magmas requires knowledge of the volumetric approximate the volumetric abundances, it is striking

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


proportion of magma produced per compositional range that the intermediate peak seen in the volcanic record
from the scales of a single magmatic system to the global and corresponding to the bulk continental crust compo-
volume. This information is, however, rarely available sition estimates is essentially absent from the plutonic
and only limited to the level of exposure. The relative record. Continental batholiths show instead a significant
compositional abundance of analyzed arc magmas nev- amount of magmas in the range 65–70 wt % SiO2 . Inter-
ertheless provides first order indications on their respec- mediate magmas around 60 wt % SiO2 can nevertheless
tive occurrence and hints about the processes at play. be locally important as in the central Aleutian plutons
Intermediate magmas in the range 57–66 wt % (Kay et al., 2019), the Peninsular Range (Todd et al., 2003)
SiO2 represent a significant proportion of the arc and in the Alpine plutons (Müntener et al., 2021).
volcanic rock samples, even though mafic samples with Volcanic glasses preserved in volcanic rocks as MIs
48–54 wt % SiO2 predominate (Fig. 1a). Intermediate trapped in phenocrysts provide important information
compositions are not distributed evenly among the on the composition of liquid during differentiation. In
arcs. They predominate in continental arcs, largely the case of petrologically complex magmas, they rep-
due to the common occurrence of large andesitic resent the only unequivocal record of the composition
stratovolcanoes (e.g. Hildreth, 2007), but are less com- of crystallizing liquids. As previously noted by Reubi &
mon in oceanic arcs (Fig. 1a). Some oceanic arcs, for Blundy (2009), the MIs show a bimodal distribution with
instance the Kermadec arc, even show a marked major peaks around 50–55 and 70–77 wt % SiO2 , and a
bimodality with a dearth of intermediate compositions marked paucity of intermediate compositions (Fig. 1C).
(e.g. Smith et al., 2006). Intermediate magmas can Resampling strategies to limit the effect of regional sam-
nevertheless be relatively frequent locally in island pling bias (equal number of MIs per 100 km of arc) or to
arcs, for example Montserrat in the Lesser Antilles compensate for preferential investigation in a particular
(e.g. Zellmer et al., 2003). A first order comparison host phase (number of MIs proportional to the total
between the number of published data and eruptive modal proportion of the host phase in a fractional crys-
volume estimates per compositional range is only tallization sequence from basalt to rhyolite) both yield a
available for the Cascades (Kent, 2014). It shows a good trough at intermediate compositions (Fig. 1C). This con-
degree of consistency for the major volcanic edifices, firms the dearth of intermediate MIs previously observed
but indicates that the smaller dispersed volcanic edifices by Reubi & Blundy (2009) using a significantly smaller
and cinder cones, which are predominantly mafic, are melt inclusion dataset. The paucity of MIs with interme-
underrepresented in terms of equivalent number of diate compositions, despite the abundance of volcanic
samples. This suggests that the andesitic peak may not rocks within this compositional range, largely reflects
translate the volumetric importance of andesites but the fact that magmas from andesitic stratovolcanoes
rather preferential sampling of big active andesitic stra- generally contain MIs dacitic to rhyolitic in composition
tovolcanoes. A bias difficult to correct for in the case of (Reubi & Blundy, 2009; Kent et al., 2010). Volcanic glasses
island arcs where small dispersed edifices are rarely doc- (pumices, glass shards, groundmass glasses, and pillow
umented. This would require more elaborate resampling rims) show the same bimodal distribution as the MIs
strategies accounting for the geographic particularities (Kent, 2014). This distribution is also observed at the
of individual arc segments in the case of continental arcs. arc scale for glasses from distal tephra fallout from the
The intermediate peak observed in the volcanic record Izu Bonin arc (Straub, 2003) (Fig. 1C). Collectively, this
is not observed in relatively well-preserved plutonic arc demonstrates that the paucity of andesitic melt is not an
sections (Fig. 1b). Plutonic rocks from accreted island artifact caused by sampling or analytical strategies of MIs
arcs such as the Kohistan-Ladakh and Talkeetna sections investigations.
(Greene et al., 2006; Jagoutz et al., 2009; Jagoutz & Schmidt, Andesitic magmas turn out to be less ubiquitous than
2012) show bimodal distributions with a major peak in generally thought in arc settings. Overall, the plutonic
the range 45–55 wt SiO2 , a smaller peak between 68– record indicates that there is a subjacent bimodality
75 wt % SiO2 and a trough between 60 and 65 wt % SiO2 . with a dearth of intermediate compositions. A pattern
The continental Sierra Nevada–Peninsular Ranges and akin to the indubitable bimodality in melt composition
8 | Journal of Petrology, 2022, Vol. 63, No. 6

depicted by the MIs and volcanic glasses, although the at a given SiO2 content. Na2 O in mafic MIs essentially
glass silicic peak is offset to higher silica contents. These follow the experimental LLD, although the 0.7–1.0 GPa
observations are based on the relative abundance of experiments used a starting composition with low Na2 O
samples and further research is necessary to ascertain contents (Fig. 2g). The silicic MIs show a trend of net
the volumetric proportions. Considering the degree of decrease in Na2 O not reproduced by the experimental
consistency between the plutonic and MI records that LLDs. P2 O5 in all experimental melts increases up to 64–
are subject to independent sampling bias, and in view 70 wt % SiO2 before dropping sharply, a trend also shown
of the volumetric predominance of the plutonic record by the intermediate and differentiated MIs (Fig. 2k). The
over the volcanic record, it is nevertheless evident range in P2 O5 shown by the dacitic to low-SiO2 rhyolite
that the andesitic trough reflects subdued volumes of MIs reach higher values than most experimental LLDs,
intermediate magmas in mid- to upper-crustal magmatic which in all likelihood is related to the higher variability
systems. It is worth pointing out that there is no in P2 O5 in natural basaltic melts compared to the starting

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


compositional gap as such. The observed andesite compositions of the experiments (Fig. 2k).
trough is, however, at odds with simple crystallization– The volcanic bulk-rock compositions match relatively
differentiation models, which predict a progressive well the LLDs and MIs for mafic and silicic composi-
volumetric decrease from mafic to silicic magmas. The tions, but diverge for intermediate compositions. The
presence of the andesitic and silicic peaks in the volcanic intermediate bulk-rock population is slightly displaced to
and plutonic records, respectively, also raises questions higher SiO2 and lower CaO and FeO at a given MgO con-
as the opposite may be expected if progressive differenti- tent compared to the LLDs and MIs (Fig. 2b, d and f). In
ation of melts ascending through the crust produced the intermediate magmas, Al2 O3 contents are in most cases
silicic magmas. This warrants further examination of the significantly lower than the high pressure LLDs and only
mode of production of intermediate melts and magmas reproduced by equilibrium crystallization experiments at
in order to understand the differentiation mechanisms low pressure (<200 MPa) and water saturated conditions
in arcs. (Fig. 2j). As with the MIs, a decrease in Na2 O is observed
for silicic volcanic rocks (Fig. 2h). However, silicic rocks
Experimental melts, MIs and bulk-rock with Na2 O > 5 wt % are scarce, when these compositions
compositions are recurrent in silicic MIs. P2 O5 contents in intermediate
The LLDs produced experimentally from hydrous basaltic magmas are mostly lower than in the experimental melts
melts (Grove et al., 1997; Sisson et al., 2005; Blatter et and MIs at a given SiO2 content (Fig. 2l). If the major ele-
al., 2013; Nandedkar et al., 2014; Ulmer et al., 2018), ment arrays may equally be produced by differentiation–
as well as the arc MI and bulk-rock compositions are crystallization and mixing, the P2 O5 contents in inter-
presented in Fig. 2. Overall, the MIs show a good match mediate magmas suggest a strong influence of magma
to the experimental liquids for SiO2 , MgO, CaO, FeO and mixing.
P2 O5 , although the compositional variability of mafic The plutonic bulk-rock compositions from exposed arc
MIs is higher than accounted for by the experiments. sections are shown in Fig. 3 along with the experimental
Intermediate and silicic MIs follow the hydrous LLDs pro- LLDs. The trends for the plutonic rocks are concordant
duced at high pressure (0.7–1.0 GPa) for most elements with the experimental LLDs to about 57 wt % SiO2 and
and generally reproduce the LLDs curvature for the above 67 wt % SiO2 . In between, the plutonic rocks form
compatible elements MgO, CaO and FeO (Fig. 2a, c and e). arrays that cut under or above the more curved LLDs.
In detail, CaO in intermediate melts from 1 GPa fractional As for volcanic rocks, the 100 to 200 MPa water saturated
crystallization experiments (Ulmer et al., 2018) are on LLDs can partly account for the low Al2 O3 of intermediate
the high side of the MI range. The intermediate melts magmas (Fig. 3b), but are difficult to reconcile with the
produced by equilibrium crystallization at 400 MPa low P2 O5 in the majority of these rocks (Fig. 3c). As
(Blatter et al., 2013) have FeO contents on the low side pointed out by Keller et al. (2015), above 50 wt % SiO2
of the MI range and produce a trend that cannot fit the the chemical trends for volcanic and plutonic rocks are
differentiated MIs. Fractional crystallization experiments identical although the compositional distributions differ
performed at 700 MPa reproduce well the MI trends, (Fig. 1). We stress that cumulates strongly enriched in
except for FeO in differentiated melts, but this is due to compatible elements were not included in Fig. 3. These
slight iron loss not corrected for some of the experiments cumulates form a significant portion of the relatively
(Nandedkar et al., 2014). The principal difference between well-preserved Kohistan-Ladakh and Talkeetna lower
the experimental LLDs and MIs are observed for Al2 O3 crust sections (Greene et al., 2006; Jagoutz et al., 2011).
and Na2 O (Fig. 2g and i). The majority of mafic MIs show
trends of marked Al2 O3 decrease from 50 to 60 wt %
SiO2 and from 6 to 4 wt % MgO. In intermediate MIs, the DISCUSSION
Al2 O3 contents are mostly lower than in experimental Isobaric versus polybaric liquid lines of descents
melts and high Al2 O3 compositions are rare. Some of The chemical similarities between the experimen-
the silicic MIs have high Al2 O3 contents matching the tal LLDs and the MIs demonstrate that crystalliza-
high pressure LLDs, but the majority show lower Al2 O3 tion–differentiation of hydrous basalts can effectively
Journal of Petrology, 2022, Vol. 63, No. 6 | 9

< 60 wt% SiO2 60-66 wt% SiO2 > 66 wt% SiO2 simulated in experiments vs polybaric decompression-
4000
induced crystallization recorded by the MIs, and of the
strong control this exerts on the stability of plagioclase
in hydrous melts. Diffusive loss may hamper the
500
3000 MP reliability of H2 O contents measured in MIs and may
aR
hyo
lite produce or contribute to the steep trends in Fig. 5b.
CO2 (ppm)

50
0M
The steep negative correlations shown by the same
Pa MIs in the Al2 O3 (and Na2 O) vs SiO2 diagram (Fig. 5a)
300 R Ba
.
2000 sa
lt support, nevertheless, the proposal that these trends
essentially record decompression induced crystallization
200 R of mainly plagioclase. In essence, the MIs complement
.
the experimental data by tracing the decompression-
1000 200 B.

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


induced crystallization trends followed by the melts as
100 MPa they ascend through the upper crust, hence providing
a more exhaustive view of the potential crystallization
path followed by the magmas. Figure 5 illustrates this
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 complementarity by showing the MI evolution trends
H2O (wt%) for several mafic and intermediate volcanoes. These
Fig. 4. H2 O vs CO2 for melt inclusions from arc magmas. Vapor
examples illustrate several important points regarding
saturation isobars shown respectively for generic basaltic (with 49 wt % the production of intermediate and silicic magmas:
SiO2 ) (B) and rhyolitic (R) melts and 100, 200 and 500 MPa pressures (1) Decompression-induced crystallization of hydrous
calculated using Newman & Lowenstern (2002). Melt inclusions data are
taken from the GEOROC global melt inclusion database. Mafic and silicic
mafic melts to very low pressure produces residual melts
melt inclusions show broad spectrum of vapor saturation pressure, that rarely extends beyond 60 wt % SiO2 . Furthermore,
while the majority of intermediate melt inclusions record pressures these residual melts are characterized by low Al2 O3 (and
<200 MPa. The low CO2 contents in the intermediate melt inclusions
indicate that their low H2 O contents are not the result of diffusive
volatiles) and are evidently not parental to the silicic MIs
H2 O-loss. (Figs 2i and 5a), nor to the andesitic bulk rock compo-
sitions that are characterized by higher Al2 O3 contents
(Fig. 2j).
(2) The least differentiated silicic MIs (≥66 wt % SiO2 )
reproduce the compositional range of melts in arcs from individual volcanoes have distinctly higher Al2 O3
and are in all likelihood the predominant processes and Na2 O wt % than the vast majority of bulk-rocks
controlling differentiation. There are, however, important within the same SiO2 range (Fig. 2). Their composition
distinctions between the conditions recorded by the matches the melts produced experimentally by frac-
MIs and the experimental LLDs. While the experiments tional crystallization of hydrous melts at 0.7 to 1 GPa
simulate isobaric fractional or equilibrium crystalliza- (Figs 2 and 5a), suggesting that the differentiation step
tion at pressure up to 1 GPa, the volatile elements that produced the silicic melts occurred at mid- to lower-
measured in MIs indicate polybaric crystallization crustal pressure.
in volatile saturated conditions at pressures below (3) The compositions of the silicic melts at the onset
500 MPa (Fig. 4) (Wallace, 2005; Metrich & Wallace, of the degassing-induced trends cannot be produced by
2008). Furthermore, the MIs from individual volcanoes fractional crystallization from compositions similar to
often show steep negative correlations between H2 O intermediate bulk-rocks, nor to the intermediate MIs that
and SiO2 (Fig. 5b) that are taken as indicating that are characterized by lower and decreasing Al2 O3 and
decompression is principally driving crystallization at Na2 O contents (and H2 O for the MIs) (Figs 2i, 2g and 6a).
moderate to low pressure (Sisson & Layne, 1993; Blundy (4) Al2 O3 and Na2 O diminishes markedly during
& Cashman, 2005). One remarkable aspect of the isobaric decompression-induced crystallization of silicic melts,
experimental LLDs is the correlation between the Al2 O3 hence producing a MI population shifted from the high
content in intermediate melts and the pressure of pressure LLDs. The trends shown by the silicic MIs extend
crystallization, in spite of the variability in parental to lower Al2 O3 and Na2 O (and higher SiO2 ) contents than
melt compositions (Figs 2 and 5). The overall pattern the bulk-rock compositions (Figs 2 and 5), implying that
indicates that earlier saturation of plagioclase with decompression yields significantly greater amounts of
decreasing H2 O contents, and therefore pressure in the crystallization than required to produce the bulk-rock
case of vapor-saturated melts, controls the melt Al2 O3 compositions by crystallization–differentiation from the
content during crystallization (Blundy & Cashman, 2001; high pressure LLDs. Considering that the quartz-feldspar
Grove et al., 2003; Blatter et al., 2013; Müntener & Ulmer, cotectic in the haplogranite system is shifted to higher
2018). The individual MI trends in Fig. 5a imply that SiO2 at low P (Holtz et al., 1992; Blundy & Cashman, 2001),
the poor fit between the experimental LLDs and MIs crystallization at the end of the decompression trends
for Al2 O3 , and Na2 O in silicic inclusions, (Fig. 2g and i) also explain the shift to higher SiO2 shown by the most
are an expression of the contrasted isobaric conditions differentiated MIs (Figs 1 and 5).
10 | Journal of Petrology, 2022, Vol. 63, No. 6

.
at
e rs
und
O
H2
Colima Llaima Villarrica Soufriere Hills sat.
isobaric H2O

de O sa
H2
co
Cerro Negro St Helens Pinatubo

mp t.
23.0

res
a) b)

si
on
10.0
21.0
Pa

Pa
1G

M
a

0
MP

70
19.0 8.0 300
400
200
Al203 (wt%)

H2O (wt%)
17.0 100 Pa
1 GPa

6.0 200 M

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


700 MPa

15.0 0.1
400
4.0 100 MPa
13.0
50 MPa
2.0
11.0
10 MPa

9.0 0.0
45 50 55 60 65 70 75 80 85 45 50 55 60 65 70 75 80 85
SiO2 (wt%) SiO2 (wt%)

Fig. 5. Al2 O3 and H2 O contents vs SiO2 content of melt inclusions from several arc volcanoes. Curves in (a) are experimental LLDs as in Fig. 2.
Numbers in italic indicate the pressure of the experiments in MPa. Dashed lines in (b) are isopleths of H2 O contents extrapolated from calculated
values for basaltic (with 49 wt % SiO2 ) and rhyolitic compositions using the solubility model of Newman & Lowenstern (2002). Continuous curves in (b)
are the H2 O contents measured in melts from the H2 O-undersatured experiments shown in (a). H2 O contents in the H2 O-saturated experimental
melts follow the isopleths at the corresponding pressure. Inset, schematic trends anticipated for different crystallization conditions. Melt inclusion
data are from Colima (Reubi et al., 2013); Llaima (Bouvet de Maisonneuve et al., 2012), Villarrica (Pioli et al., 2015), Soufriere Hills (Humphreys et al., 2010;
Williamson et al., 2010; Mann et al., 2013; Cassidy et al., 2015), Cerro Negro (Roggensack et al., 1997; Roggensack, 2001; Sadofsky et al., 2008; Portnyagin et
al., 2014), Mount St. Helens (Blundy & Cashman, 2005) and Mt. Pinatubo (de Hoog et al., 2004; Borisova et al., 2005; Di Muro et al., 2008). References for
the experimental data are given in Figs 2 and 3. Although diffusive loss of H2 O from the MIs is likely to contribute to the steep trends shown in (b), the
combination with the negative correlations in (a), indicates that the trends for individual volcanoes primarily result from
decompression-crystallization of plagioclase-rich assemblages in H2 O-saturated melts. All reported data are normalized to 100% anhydrous.

The crystallization–differentiation pathway to that they formed at this or lower H2 O-partial pressures
andesite if vapor-saturated (Ptot < 200 MPa, Fig. 4). Alternatively,
Andesitic MIs, when found in appreciable number at these melts could result from crystallization at higher
individual arc volcanoes, show a significant composi- pressure of vapor-undersaturated damp basaltic melts
tional spread (Fig. 6). At these volcanoes, the mafic and with ∼1–2 wt % H2 O. In the case of the volcanoes
intermediate MIs with <60 wt % SiO2 have variable Al2 O3 highlighted in Fig. 6, the presence of mafic MIs with 4
contents that extend to high values, when intermediate to 5 wt % H2 O suggest, however, that this is unlikely. In
MIs with >60 wt % SiO2 are characterized by relatively view of the compositional spread, it is difficult to estab-
low Al2 O3 and H2 O contents (and in most case low Na2 O) lish unequivocally the differentiation pathways that
compared to more differentiated inclusions (Fig. 6). It produced the intermediate MIs (Fig. 6). Decompression-
is noteworthy that the global MI dataset comprises induced crystallization of hydrous mafic melts fol-
surprisingly few intermediate compositions with high lowed by near isobaric crystallization at pressures
H2 O, CO2 , Na2 O and Al2 O3 contents (Figs 2, 4 and 6). between 100 and 10 MPa is one plausible pathway.
As discussed above, low H2 O may be a consequence Decompression-induced crystallization from more dif-
of diffusive H2 O loss from the MIs, but the low CO2 ferentiated (< 60 wt % SiO2 ) high-Al2 O3 hydrous melts
rather indicate low entrapment pressures. This is also is another possibility, although none of the examples in
consistent with the relatively low Na2 O and Al2 O3 Fig. 5 show a continuous trend as observed in Fig. 4. In
that require extensive plagioclase crystallization. The any case, the low Al2 O3 , H2 O, ± Na2 O contents in MIs
predominant compositions among MIs in the range with 60–66 wt % SiO2 indicate that the intermediate
60 to 65 wt % SiO2 imply that they formed under a melts formed at relatively low pressure after extensive
limited range of H2 O contents and/or pressure conditions plagioclase crystallization and along LLDs distinct from
compared to more mafic and silicic MIs (Fig. 6). H2 O- the high-pressure LLDs that produced the silicic melts
saturated equilibrium crystallization experiments at shown in Fig. 5.
100 MPa (Grove et al., 1997) form an upper bound to the The vast majority of andesitic volcanic and plutonic
majority of intermediate MIs (Figs 2 and 6), indicating rocks essentially follow the trajectories of the 100 MPa
Journal of Petrology, 2022, Vol. 63, No. 6 | 11

t.
sa
d er
un
Emmons lake Tungurahua Hudson All melt inclusions O
H2
Gorely Volcanoes from Fig. 4 sat.
isobaric H2O

de O sa
H2
co
23.0 10.0

mp t.
b)

Pa
a)

res

0M
9.0

70
si
on
21.0
8.0

Pa
1G
19.0 400 7.0
200
6.0 Pa
Al203 (wt%)

200 M

H2O (wt%)
17.0
100
1 GPa

5.0

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


0.1

15.0 400 100 MPa

700
4.0

MP
a
13.0 3.0 50 MPa

2.0 25 MPa
11.0
10 MPa
1.0

9.0 0.0
45 50 55 60 65 70 75 80 85 45 50 55 60 65 70 75 80 85
SiO2 (wt%) SiO2 (wt%)

Fig. 6. Al2 O3 and H2 O contents vs SiO2 content of melt inclusions from several arc volcanoes with significant populations of melt inclusions
intermediate in composition. Curved LLDs and dashed isopleths as in Fig. 5. Melt inclusion data are from Emmons Lake (Zimmer et al., 2010),
Tungurahua (Myers et al., 2014), Hudson (Kratzmann et al., 2009), Gorely (Tolstykh et al., 2012). The composition of melt inclusions with >60 wt % SiO2
indicate that for the majority they formed at pressures ≤100 MPa and that isobaric crystallization at shallow pressures may be significant in these
compositions. All reported data are normalized to 100% anhydrous.

vapor-saturated experiments and have the 200 MPa errors for P2 O5 indicate that in the range 58–62 wt
melt compositions as an upper bound in Al2 O3 vs % SiO2 , 47% of the volcanic and 44% of the plutonic
SiO2 and MgO diagrams (Figs 2j and 3b). However, as rocks have lower P2 O5 contents than the melts obtained
discussed above, the dominant bulk-rock trajectories experimentally at these SiO2 contents (Fig. 7a). In the
in P2 O5 vs SiO2 (and to a lesser extent FeO or CaO vs range 62–66 wt % SiO2 (roughly corresponding the apex
MgO) diagrams do not follow the LLDs and rather match of the bell-shaped curves), the proportions increase to
linear mixing arrays that may be difficult to distinguish 74% for the volcanic and 85% for the plutonic rocks
from the relatively straight 100 MPa LLD’s in Al2 O3 (Fig. 7b). This is consistent with the relatively linear
vs SiO2 and MgO diagrams (Fig. 2). The low-pressure bulk-rock trajectories in major elements Harker dia-
differentiation conditions recorded by intermediate MIs grams at intermediate compositions (Figs 2 and 3) and
are likely to account only for a limited proportion of emphasizes the prevalence of mixing in producing
andesitic magmas, consistent with the low proportion of intermediate compositions. If these proportions of
intermediate MIs (Fig. 1c). This issue is assessed further mixed magmas are subtracted from the compositional
in the next paragraph. distribution (Fig. 1a), the intermediate peak in the
volcanic record effectively disappears, leaving a trough
On the relative importance of magma mixing around 64 wt % SiO2 as for the MIs and plutonic rocks
and crystallization–differentiation in making (Fig. 1). The inferred proportions of mixed intermediate
intermediate magmas magmas are in fact minimum estimates because (1)
Lee & Bachmann (2014) highlighted the potential of P2 O5 the mafic mixing end-member may have higher P2 O5
to distinguish between fractional crystallization and than the experimental melts (e.g. Pinatubo in Fig. 8); (2)
magma mixing in arc magmas. In damp to hydrous melts, the trends recorded by the MIs indicate that in many
P2 O5 is incompatible until apatite saturation, at which cases differentiation of hydrous magmas yields higher
point it decreases significantly forming bell-shaped P2 O5 than the experimental LLDs (Figs 2k and 7) and (3)
differentiation trajectories as shown by the experimental assimilation or cannibalization of P2 O5 –rich cumulates of
LLDs (Fig. 2k and l). The intermediate and silicic MIs earlier magmatic pulses can produce magmas with P2 O5
follow these trajectories, consistent with crystalliza- high enough to mimic the composition of crystallization-
tion–differentiation trends (Fig. 2k). The intermediate derived melts. The first two points are illustrated by
bulk-rocks follow distinct trajectories cutting straight the good match between the bulk-rock P2 O5 distribution
under the bell-shaped LLDs, unequivocally indicating and the Monte Carlo mixing simulations using random
the predominance of mixing trajectories (Fig. 2l). Monte selections of mafic and silicic melt inclusions as end-
Carlo simulations taking into account the analytical members (Fig. 7).
12 | Journal of Petrology, 2022, Vol. 63, No. 6

Volcanic WR Plutonic WR Melt inclusions rand. mixing

58-62 wt% SiO2 experimental melts 62-66 wt% SiO2


Relative sample abundance experimental melts

relative proportion with relative proportion with


P205 < than in exp. melts: P205 < than in exp. melts:

47% volcanic WR 74% volcanic WR


44% plutonic WR 85% plutonic WR
20% melt inclusions 24% melt inclusions

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


P2O5 (wt%) P2O5 (wt%)

Fig. 7. Relative sample abundance as a function of P2 O5 content (anhydrous) for arc volcanic and plutonic rocks, melt inclusions, experimental melts
and Monte Carlo simulation of mixing. Bin size is 0.02 wt % P2 O5. The curves were obtained by performing Monte Carlo uncertainty propagation on
10 000 randomly selected samples from each population. Typical calibration errors (RMSE) and analytical uncertainties respectively for XRF analyses
(bulk-rock data, RMSE = 0.004 and SD = 0.004) and electron microprobe analyses (melt inclusions and experimental glasses, RMSE = 0.011 and SD = 0.01)
were taken into account in the simulations. Mixing simulations (random mixing curve) were performed by 5000 random selections of a mafic melts
melt inclusion with <58 wt % SiO2 and a melt inclusions with >68 wt % SiO2 , and applying a random mixing proportion in the range 80–20%. P2 O5
contents in 0.1 to 1.0 GPa experimental melts from Grove et al. (1997), Grove et al. (2003), Nandedkar et al. (2014) and Ulmer et al. (2018) were used to
obtain the experimental melts curves. Experimental melts from Blatter et al. (2013) were omitted from these calculations in regard of the relatively
high P2 O5 content in the starting composition. The MIs nevertheless indicate that basaltic melts with high P2 O5 contents are present in arcs (Fig. 2k).
The proportions given in the panels are the relative proportions of bulk-rock and MI compositions with P2 O5 contents lower than in the experimental
melts (gray area). They correspond to the proportions of samples most likely to be produced by magma mixing (sensu lato). These are minima as
demonstrated by the random mixing simulation curve that covers the bulk-rock populations, hence demonstrating that the bulk-rock P2 O5 ranges are
potentially fully achievable through mixing of documented melt compositions. In addition, mingling of P2 O5 -rich cumulates is not taken into account,
although it would yield a wider spread in the mixing curve. Unlike the volcanic and plutonic whole rock compositions, the MI compositions generally
do fall within the range of experimental liquids, indicating that they are likely to represent LLD melts.

As may be anticipated from Fig. 2j, intermediate significant amounts of intermediate melts in the lower-
bulk-rocks with P2 O5 contents high enough to be to mid-crust. This concurs with the observation that
compatible with the experimental LLDs and to represent plutonic rocks inferred to be dry or damp may represent
‘true’ crystallization-derived melts have Al2 O3 content a significant portion of accreted lower crustal arc section
typically in the range 15–17 wt %. These compositions but rarely exceed 57 wt % SiO2 (Jagoutz et al., 2011).
are best accounted for by vapor-saturated crystallization
at pressures between 200 and 50 MPa (but mostly The prevalent mixing end-members
<100 MPa). Hence, we conclude that low-pressure The compositions of the mixing end-members provide
crystallization–differentiation yield at most ∼55% of the important information on the components involved. The
magmas in the range 58–62 wt % SiO2 and 25% of the trends shown by the bulk-rock compositions in Figs 2
magmas between 62 and 66 wt % SiO2 , while mixing and 3 imply that the predominant mixing end-members
accounts for the rest. Intermediate volcanic rocks with have respectively around 52–56 wt % SiO2 and 4 to 6 wt
P2 O5 contents equivalent to experimental melts and the % MgO, and 68–72 wt % SiO2 and < 1 wt % MgO. Such
MIs are found in continental and oceanic arcs, examples compositions are abundant in the melt and magma
include Anathan volcano in the Mariana arc (Wade et records and are close to experimentally produced melts
al., 2005) or Mocho volcano in Chile (Rawson et al., 2015). (Figs 1–3). Mixing or mingling between basaltic andesites
These compositions are, however, surprisingly rare in the and high-SiO2 dacites or low-SiO2 rhyolites that can
plutonic record (Figs 3c and 7). This may nevertheless be be related by crystallization–differentiation or partial
expected if volatile-saturated crystallization at relatively melting is apparently the norm. In detail, there must be
low pressures is the prime pathway to produce inter- significant variability in the composition of the mixing
mediate melts, as the plutonic record is volumetrically end-members to account for the spread in intermediate
dominated by magmas that formed at higher pressure. compositions.
The low proportion of ‘true’ andesitic melts in the The compositions of MIs, bulk-rocks and mafic
plutonic record further suggests that crystallization– enclaves from the same rocks are presented for several
differentiation from damp mafic melts does not produce intermediate stratovolcanoes in Fig. 8. Despite the
Journal of Petrology, 2022, Vol. 63, No. 6 | 13

22.0 10.0
(a) a (b)
GP
0
1. 9.0
a. FC, 700 MPa
, b-
20.0 FC
FC
8.0
,M
g-b
asa
EC, 400 lt, 1 7.0
.0 G
18.0 Pa

6.0
Al2O3 wt%

MgO wt%
*
16.0 5.0

4.0
* EC, 200 MPa
14.0

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


majority of intermediate 3.0
volcanic rocks
2.0
12.0
1.0

10.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 45 50 55 60 65 70 75 80
MgO wt% SiO2 wt%

0.8
(c) Soufrière Hills 1995-2010 WR

Soufrière Hills 1995-2010 melt inclusion


0.7
Soufrière Hills 1995-2010 mafic enclave

0.6 Soufrière Hills 1995-2010 mafic enclave average

Colima 1998-2010 WR
0.5
P2O5 wt%

Colima 1998-2010 melt inclusion

0.4 Colima 1998-2010 crystal clots average

Mount St. Helens 1980-1986 WR


0.3
Mount St. Helens 1980-1986 melt inclusion

Mount St. Helens 1980-1983 gabbroic inclusion


0.2
Mount St. Helens 1980-1983 gabbroic inclusion av.

0.1 Pinatubo 1991 WR

Pinatubo 1991 melt inclusion


0.0
45 50 55 60 65 70 75 80 Pinatubo 1991 mafic inclusions
SiO2 wt%
Pinatubo 1991 mafic inclusions av.

Fig. 8. Mixing trends linking the melt inclusions, bulk-rock and mafic enclaves in intermediate magmas at several arc volcanoes. Data are from Volcan
de Colima 1998–2010 eruption (Reubi & Blundy, 2008; Reubi et al., 2013); Soufriere Hills 1995–2010 eruption (Humphreys et al., 2010; Mann et al., 2013;
Plail et al., 2018), Mount St. Helens 1980–1986 eruption (Melson, 1983; Heliker, 1995; Blundy & Cashman, 2005) and Mt. Pinatubo 1991 eruption (Pallister
et al., 1992; de Hoog et al., 2004; Borisova et al., 2005). Gray fields represent the region of high density in bulk-rock compositions (red to green colors in
Fig. 2). Dashed curves are experimental LLDs as in Fig. 2. The colored mixing lines are visual fits connecting the mafic enclave, bulk-rock and melt
inclusion compositions for each case study. These mixing trends illustrate the variability in the end-member compositions and petrological
characteristics (see main text) involved in the production of intermediate magmas. All reported data are normalized to 100% anhydrous.

variability in the composition of the mafic enclaves and mafic mixing end-members. It is likely that the mixing
silicic MIs for each volcano, the bulk-rock trends can components often comprise several crystal populations,
be reasonably fitted by mixing the mafic enclaves with a variety of crystal cargo and/or several magma batches
melts slightly more evolved than the least differentiated mixed in successive events (e.g. Cashman & Blundy,
MIs. We note that the mixing lines are visual fits and do 2013). The examples shown in Fig. 8 correspond to
not exactly correspond to a regression through the bulk- two distinct mixing scenarios. In the case of Colima
rock compositions. In reality, the mixing trends are not and Mount St. Helens, the mafic enclaves are gabbroic
perfectly linear, either because there are more than two fragments from the plutonic roots of the volcano that
components involved and/or because using the average were entrained and partly resorbed by the ascending
composition of the mafic enclaves is too restrictive silicic melts represented by the silicic MIs (Heliker,
to represent the variability in composition of the 1995; Blundy & Cashman, 2005; Reubi & Blundy, 2008).
14 | Journal of Petrology, 2022, Vol. 63, No. 6

bulk magma differenaon melt differenaon only magma mixing

(A) amphibole step (C)

2
mixing

px
SiO
1200 plutonic

abundance
relave
wt
55
amp
t
gr ol
1000
c d
23
(B)
800 a
21
amphibole step
2

60
Pressure (MPa)

40
70 wt SiO

60 wt SiO2

19

Al203 wt%
600

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


70
plg 8 wt% H2O 20
17 mixi
a ng
400 H2O saturaon 80
15
d 6 c
b * d
13 80
?
200
4
c 11
b
2
0 * 9 * Andesite sweet spot
800 900 1000 1100 1200 1300 45 50 55 60 65 70 75 80 85
Temperature (°C) SiO2 wt%

Fig. 9. Diagrams showing the principal differentiation pathways envisaged for arc magmas. (A) Pressure–temperature diagram showing the
crystallization and ascent paths for hydrous melts. The phase relations for hydrous basaltic to basaltic andesite melts are based on experimental data
from Baker et al. (1994), Blatter et al. (2013), Grove et al. (1997), Grove et al. (2003), Moore & Carmichael (1998), Müntener et al. (2001), Nandedkar et al.
(2014), Sisson et al. (2005) and Ulmer et al. (2018). The black dashed lines contour the SiO2 wt % of the residual melts in these experiments. Gray dashed
lines represent H2 O saturation pressures for basaltic (1100 ◦ C) and rhyolitic (850 ◦ C) compositions (after Newman & Lowenstern, 2002). The blue
arrows illustrate representative differentiation pathways based on the melt inclusion and experimental data. a Differentiation at lower crustal
pressures. Decompression-induced crystallization trends followed respectively by the mafic b and silicic melts d. The light blue arrow c represents a
crystallization path combining decompression followed by isobaric cooling producing the intermediate melts. The yellow and orange fields in (A) and
(B) are based on the melt inclusions compositions and H2 O contents. The gray fields represent the step in the differentiation process caused by the
crystallization of amphibole (see main text for details). In (B), the lower crustal differentiation path a is an averaged composition based on the 0.7 and
1.0 GPa LLDs from Nandedkar et al., (2014) and Ulmer et al. (2018). The differentiation path b, c and d are based on melt inclusion data (Figs 5 and 6).
The italic numbers along the differentiation paths a and d (to 70) are the total crystal vol% taken from Nandedkar et al. (2014) experiments and
calculated using the K2 O content of the melt inclusions from Mount St. Helens (Blundy & Cashman, 2005) for d (70 to 80). The mixing array is based on
the compositional array of volcanic bulk-rocks. (C) Relative abundances of magma inferred to result from the four differentiation paths. Filled curves
represent the inferred deconvoluted volcanic distribution based on Fig. 1a, while the orange curve shows the silicic peak for the plutonic record
inferred from Figs 1b and 10. The overall relative abundances are based on the relative sample abundances shown in Fig. 1. The proportion of
intermediate magmas resulting from mixing and differentiation are based on the P2 O5 contents shown in Fig. 6.

For Soufriere Hills and Pinatubo, the mafic enclaves important dynamic distinctions between the volcanic
are quenched mafic magmas that intruded into the and plutonic realms.
differentiated magmas and likely triggered the eruption
(Pallister et al., 1992; Murphy et al., 2000). In the first case, The step in the hydrous liquid lines of descents
partial assimilation and dissemination of mafic crystals The bimodal distributions observed in Fig. 1 for the
from the gabbroic clots and cumulates form most of the MIs and the oceanic arc plutonic sections are often
mafic component (Reubi & Blundy, 2008). In the second referred to as a compositional gap (or Daly gap). This
case, mafic melt droplets and mafic microcrystals from terminology is, however, misleading because when taken
disaggregated mafic inclusions compose most of the individually all major elements in the MI record show a
mafic end-member (Humphreys et al., 2009). Although continuous range of composition (Fig. 2). Furthermore,
not exhaustive and certainly not fully accounting for considering only a single chemical element (generally
the complexity of the mixing processes, the range of SiO2 ) yields a partial picture. In compositional terms,
mixing end-members illustrated in Fig. 8 provide a the MI silicic peak is associated with shifts to higher
good fit to the variability in the global volcanic trends Al2 O3 , H2 O and Na2 O contents (Figs 2 and 6), indicating
(Figs 3b and 8). that the relative proportion of andesitic and silicic
Interestingly, mixing in plutonic systems does not yield melts translate the variability in the volume of melt
an andesitic peak like in the volcanic bulk-rock record sampled at a given degree of differentiation along
but rather a dacitic shoulder on the silicic peak (Figs 1 low and high pressure hydrous LLDs, respectively. The
and 3). The causes and consequences of this difference principal differentiation paths envisaged based on the
will require further investigations, as it may translate melt inclusion compositions and relative abundances, as
Journal of Petrology, 2022, Vol. 63, No. 6 | 15

bulk continental crust


23

relative sample abundance


0 Volcanic Volcanic

Al203 wt%
19

Kohistan - Ladakh Batholith


15
10 gabbros to granites

melt dominated
11

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


23
2-4 kbar
relative sample abundance
20 2-8 kbar
4-8 kbar

Al203 wt%
19
Depth (km)

15
30 mixing array
gabbros - diorite
amphibolites

+ cumulates
- granite

11
melts
Southern Plutonic Complex

23
relative sample abundance

40 8-12kbar 8-12kbar
ultramafic cumulates

amphibole step
cumulate dominated

Al203 wt%
19

15
50 mixing array

11

40 50 60 70 80 40 50 60 70 80
SiO2 wt% SiO2 wt%
Fig. 10. Crustal section of the Kohistan-Ladakh arc. Pressure estimates are based on amphibole barometry. The differentiation–crystallization paths as
in Fig. 9. Lithologic section after Jagoutz (2014), the geochemical data, relative abundances and amphibole barometry for the plutonic section are from
Jagoutz (2014) and our new data. Volcanic rocks data are from the GEOROC Kohistan-Ladakh compilation. All data are provided as Supplementary data.

well as the experimental data are summarized in Fig. 9. 2021). The plutonic and MI records concur, however,
In essence, crystallization–differentiation of hydrous in indicating that the efficiency of melt production
basaltic melts at pressure and temperature conditions and/or extraction varies non-systematically along high-
encountered throughout crustal arc sections produces pressure hydrous LLDs. The melts issued from the
a continuum of melt compositions, but limited volumes lower crust are predominantly either basaltic to basaltic
of differentiated melts (>65 wt % SiO2 ) are produced at andesite (<57 wt % SiO2 ) or dacitic (>65 wt % SiO2 ) to
pressures less than 200 MPa, whereas vast volumes of rhyolitic in composition (Fig. 9). The crustal section of
silicic melts are produced at lower crustal pressures. the Kohistan-Ladakh arc provides an example of this
The cumulate rocks present at the base of exposed (Fig. 10), although we acknowledge that the number of
crustal sections and the voluminous silicic plutons samples with amphibole barometry constraints on the
observed from mid-crustal pressures upward are strong pressure of crystallization is limited in regard to the
supports in favor of the volumetric importance of geographic extent of the section. Magmas escaping from
differentiation in the deep crust (Müntener et al., 2001; the ultramafic to mafic cumulate-dominated roots of
Greene et al., 2006; Jagoutz et al., 2011; Jagoutz & the arc to form the mid- to upper-crustal batholith are
Klein, 2018; Müntener & Ulmer, 2018; Klein & Jagoutz, bimodal with a dearth of andesites (Fig. 10). Furthermore,
16 | Journal of Petrology, 2022, Vol. 63, No. 6

high-Al andesitic melts compatible with the high- melt in equilibrium with amphibole
pressure hydrous LLDs are absent at these depths. On the
melt inclusion
other hand, the volcanic record shows a compositional (a)
distribution typical of oceanic arcs with a secondary 30

andesitic peak and a minor dacitic to rhyolitic peak. 1.


0
This implies that production of andesites by magma 700 GP 0
MP a
mixing and low-pressure crystallization–differentiation a
@
in sub-volcanic magmatic systems, combined with
#
preferential emplacement of silicic magmas in mid- to

Al2O3 wt%
upper-crustal batholith are key in producing the relative *

1.0 GPa
sample abundance observed for oceanic arc volcanics
(Figs 1a and 10).

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


Additional support for the prevalence of differentia-
tion at high pressure is provided by the composition of
amphiboles in intermediate to silicic volcanic rocks. For @ 400 MPa
instance, Al-rich hornblende preserved in the core of # 200 MPa
amphibole phenocrysts from the Pinatubo and Mount St
* 100 MPa
Helens magmas indicate that the silicic melts feeding
these volcanoes initially formed at high pressure (0.7–
1 GPa) and relatively low temperature (950 ◦ C) (Prouteau SiO2 wt%
& Scaillet, 2003; Blatter et al., 2017;Humphreys et al., 2019;
Wanke et al., 2019). Furthermore, the melts inferred to (b) 40
be in equilibrium with hornblendes yield a continuous
range from low SiO2 -andesite for the cores to dacite for
0
the rims (Humphreys et al., 2019; Wanke et al., 2019),
hence demonstrating that crystallization–differentiation
and not partial melting of basaltic sources produced the
silicic melts. The amphibole trace elements suggest, nev-
Al2O3 wt%

ertheless, that scavenging of partially remelted intrusive


material occurred at deep crustal levels (Humphreys et
al., 2019). The importance of the amphibole record is
further illustrated for the volcanic rocks of the high Cas-
cades arc (Fig. 11), by comparing the composition of MIs,
bulk-rocks and melts in equilibrium with the amphiboles
calculated using the empirical formulation of Zhang et al.
(2017). Despite the significant errors associated with this
approach, the results highlight the offset between the
composition of the crystallizing melts and the majority of
intermediate bulk rock compositions. This demonstrates
that amphiboles trace the evolution of the intermediate MgO wt%
melts differentiating in the deeper section of the crust,
Fig. 11. Comparison of chemical variations in volcanic bulk-rocks, melt
a record rare in the MI and bulk-rock data. The amphi-
inclusions and calculated melts in equilibrium with amphiboles from
bole record is a strong indication that the paucity of the High-Cascades arc. The bulk-rocks, melt inclusions and amphibole
intermediate melts is a consequence of crystallization compositions are from the GEOROC compilations. The composition of
the melts in equilibrium with the amphiboles were calculated using the
of amphibole-rich assemblages and that it represents
formulation of Zhang et al. (2017). The black crosses show the error
a local drop in melt extraction efficiency along high- (2 S.E) reported by these authors for the calibration. Color contours
pressure LLDs, not a compositional gap in the production represent the number of bulk-rock sample per cell in a 60 × 60 grid.
Colored curves are experimental liquid lines of descent as in Fig. 2. The
mechanism (Fig. 9). The 0.7–1 GPa experimental data
complementarity between the melts calculated from the amphibole
indicate that plagioclase, ± orthopyroxene, ± clinopy- compositions and the melt inclusions, as well as the offset between the
roxene, ±garnet crystallize along with amphibole. High- bulk-rock and amphibole melts compositions suggest that
crystallization of amphibole at mid- to lower-crustal pressures
pressure crystals recording crystallization through inter-
contributes in producing the compositional step observed in the melt
mediate compositions are, consequently, also expected inclusions.
for these minerals in arc magmas, although distinguish-
ing them from low-pressure crystals requires careful
analysis and examination. (Grove & Donnelly-Nolan, 1986; Reubi & Blundy, 2009); (2)
A local drop in melt extraction efficiency along the increments in the efficiency of melt-crystal separation
high-pressure LLDs may be caused by (1) the strong during crystallization (Dufek & Bachmann, 2010) and
nonlinearity of LLDs in composition-temperature space (3) thermal dwelling along the peritectic reaction
Journal of Petrology, 2022, Vol. 63, No. 6 | 17

opx + cpx + melt = hbl (Blatter et al., 2017). The last two Considering the above arguments, we conclude that
hypotheses require near equilibrium crystallization over fractional crystallization of amphibole-rich assemblages
a significant temperature interval to take place. For the is key in producing both the voluminous silicic intrusions
second mechanism, theoretical calculations indicate observed in mid- to upper-crustal batholiths and the
that crystal-liquid separation is most efficient within paucity in intermediate melts issued from the lower
a crystallinity window of 50–70 vol% crystals, with an crust (Figs 1b and 10). The combination of large com-
optimum around 60 vol% (Dufek & Bachmann, 2010). positional changes over relatively small temperature
According to equilibrium crystallization experiments, a and crystallization intervals produces a step in the
first extraction optimum is expected in the range 53– differentiation efficiency that significantly increases
57 wt % SiO2 , which could fit the mafic peak in Fig. 1a. through intermediate compositions, which effectively
Subsequent crystallization of the extracted mafic melts reduces the probability to extract residual melts with
to ∼60 vol% crystals yields a second extraction optimum these compositions. Equilibrium crystallization may,

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


around 62 wt % SiO2 , right in the trough shown by the MIs however, become important in high-crystallization
and plutonic rocks. This mechanism is therefore unlikely temperature intervals. Incremental efficiency in melt-
to control the paucity of intermediate melts extracted crystal separation and consumption of andesitic melt
from the high-pressure LLD. The third mechanism occurs to produce amphibole in peritectic reactions may then
near the amphibole high-temperature stability limit contribute in producing the step. Additional petrolog-
(Blatter et al., 2017). According to 700 and 900 MPa ical and geochemical investigations focusing on melt
equilibrium crystallization experiments (Blatter et al., extraction from lower crustal cumulates are necessary
2013), thermal dwelling and resulting enhanced melt to further address this issue. We note also that the
extraction probability should yield basaltic to basaltic- size of the step appears to be variable with the least
andesite residual melt compositions (Fig. 12). This may differentiated silicic MIs at individual volcanoes ranging
account for the mafic peak in Fig. 1a, but requires an from 65 to 71 wt % SiO2 (Fig. 5). Spatial and temporal
additional process to account for the low extractability changes in the thermal state of lower crustal magmatic
of intermediate melts. Fractional and equilibrium systems will inevitably control the temperature at which
crystallization experiments at 700 MPa show a slope in the silicic melts are extracted (Annen, 2009) and are
composition-temperature space that shallows markedly additional parameters likely to control the size of the
between ∼52 and ∼ 68 wt % SiO2 (Fig. 12a) in response to step.
the appearance of amphibole, high-An plagioclase and
Fe–Ti oxides in the crystallizing assemblage (Blatter et al., The andesite sweet spot
2013; Nandedkar et al., 2014). These large compositional One of the standing question is why the intermediate
changes over relatively small temperature intervals melts are principally extracted from relatively low-
reduce the probability of extracting intermediate melts pressure LLDs (<200 MPa) (Figs 2, 6 and 9). Insights may
relative to compositions with steeper sections of the be gained by looking at the composition–temperature–
liquids and larger crystallization intervals. This is melt fraction relationships for the experimental LLDs
certainly paramount in causing the local drop in (Fig. 12).
the abundance of melts extracted from the lower Vapor-saturated equilibrium crystallization experi-
crust. ments at pressure < 200 MPa show relatively constant
In melt fraction-composition space, equilibrium slopes in composition–temperature and composition–
and fractional crystallization experiments show dis- crystallinity space (Fig. 12). Following the arguments
tinct behavior. The equilibrium crystallization LLDs presented above, a step appreciably diminishing the
have a steep slope up to 59 wt % SiO2 and produce likelihood to extract intermediate melts is not expected
silicic compositions only at very low melt fraction at these pressures. At very low pressure (0.1 MPa LLD)
(Fig. 12b). This is a consequence of the peritectic and in the case of decompression induced crystallization
reaction cpx + opx + melt = amphibole. Consumption of of basaltic melts to very low pressure, the liquidus is
the andesitic melt to produce amphibole reduces the shifted to higher temperature (Figs 9 and 12a) and the
melt fraction reaching rhyolitic compositions despite crystallizing assemblage dominated by plagioclase and
crystallization of low SiO2 assemblages. Amphibole- characterized by the late appearance of Ti-magnetite
bearing fractional crystallization experiments show a results in moderate SiO2 increase and rapid decrease
distinct behavior. Suppression of the peritectic reaction in melt fraction (Fig. 12b). Both effects combine to limit
results in marked SiO2 increase associated with subdued greatly the differentiation potential of this crystallization
drop in melt fraction due to crystallization of amphibole- path. Considering the existence of the amphibole
rich SiO2 poor residues irrespective of the coexisting step, the hydrous mafic melts differentiating in the
liquid that becomes increasingly SiO2 -rich (Müntener lower crust will reach ∼58 wt % SiO2 at most (Fig. 9).
& Ulmer, 2018). Fractional crystallization experiments at Although they will temporarily crystallize plagioclase
700 MPa yield a shallow slope from 57 to 77 and appear to upon ascending near adiabatically, these melts can
be the most efficient LLD to produce significant volumes differentiate efficiently only if they stall and cool at
of high-SiO2 melts (Fig 12). pressures less than the stability of amphibole (the
18 | Journal of Petrology, 2022, Vol. 63, No. 6

a) EC, 900 MPa b)


1200 100
FC, 700 MPa
EC, 700 MPa 90
1150
EC, 200 MPa
80
EC, 0.1 MPa
1100
70

melt fraction wt%


Temperature (°C)

1050
60
amph in
1000 50

amph in 40 amph in
950

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


30
900
20
850
10

800 0
48 50 52 54 56 58 60 62 64 66 68 70 72 48 50 52 54 56 58 60 62 64 66 68 70 72
SiO2 wt% SiO2 wt%
Fig. 12. Silica enrichment of experimental melts as a function of temperature (a) and melt fraction (c). FC = fractional crystallization experiments,
EC = equilibrium crystallization experiments. 900 MPa data from Blatter et al. (2013), 700 MPa data from Nandedkar et al. (2014) (FC) and Blatter et al.
(2013) (EC), 200 MPa data from Grove et al. (1997) and Grove et al. (2003), 0.1 MPa data from Baker et al. (1994). High-pressure crystallization results in
marked SiO2 enrichments over small temperatures intervals below 1050 ◦ C. At lower pressures (≤200 MPa) silicate enrichment as a function of
temperature is more constant (a). The curves in (b) illustrate the more efficient production of silicic melts by fractional crystallization at
high-pressure, when crystallization of hydrous basaltic andesite melts at 100 MPa yield intermediate melts at the highest melt fraction. The shaded
area indicates the optimal crystallinity window for crystal-liquid separation [50–70 vol% crystals (Dufek & Bachmann, 2010)]. The black arrows show
the composition of the melts at the separation optimum (60 vol%), when the dashed lines illustrate the compositional range for each separation
window. The second shaded area (9–25 vol%) represents a second step of crystallization for melts issued from the first crystallinity window.

small SiO2 differential between plagioclase and these Trans-crustal degassing, crystallization and
melts dampens SiO2 increase during decompression). differentiation
This inferred crystallization–differentiation pathway One consequence of crystallization–differentiation from
to andesite corroborates the proposal of Carmichael hydrous basalts at high pressure is that the silicic melts
(2002) that two-pyroxene andesites are produced from can reach high H2 O contents. Starting from mafic melt
hydrous melts decompressing at temperatures above with H2 O contents similar to the MI values, fractional
the amphibole field and subsequently crystallizing crystallization in the lower crust will produce rhyolitic
upon reaching the vapor-saturated liquidus. Our results melts with up to 13 wt % H2 O, a value significantly
reveal, however, the limited contribution of this pathway higher than recorded by any silicic MI (≤ 9 wt % H2 O)
in the overall differentiation scheme of arc mag- (Fig 6b). It is plausible that the silicic MIs do not record
mas. the H2 O contents of the most hydrous melts when they
We conclude that there is a sweet spot in andesite attain vapor-saturation due to the shape of the crys-
production delimited by the low-pressure stability limit tallinity isolines for hydrous silicic melts in P–T space.
of amphibole (<150 MPa) and the adiabatic ascent path of Dacitic to rhyolitic melts ascending along the adiabat at
mafic melts (Fig. 9). Based on the P2 O5 -SiO2 systematics, temperatures below the stability of amphibole, i.e ∼ 920
this differentiation pathway is assumed to produce at ◦
C, do not crystallize appreciably before reaching about
most 50% of the volcanic rocks in the range 58–62 wt 300 MPa (Fig. 13). The H2 O-CO2 contents of the MIs effec-
% SiO2 , but yield limited amounts of more differen- tively indicate that they mostly form (or are sealed) from
tiated magmas (Fig. 7). A situation related to the fact this point onward as crystallization and undercooling
that the melt fraction diminishes rapidly over extended increase (Figs 4 and 5). H2 O-rich silicic melts reaching
crystallization intervals and intermediate compositions vapor-saturation above this pressure may degas signifi-
are reached at lower melt fraction, limiting the potential cant amounts of H2 O before effective sealing of the MIs.
productivity of these LLDs (Fig. 12). Fractional crystal- In the case of the examples shown in Fig. 13, 3–4 wt
lization may alleviate these limitations. However, equi- % H2 O could be missing from the MI record. The H2 O
librium and fractional crystallization experiments under fluxes released by arc magmas estimated on the basis
identical conditions outside the amphibole stability field of MI data may consequently be significantly underes-
show similar melt fraction vs composition relationship to timated, particularly at mid-crustal pressures. We note
low melt fractions (Villiger et al., 2004; Müntener & Ulmer, that the H2 O contents in mafic arc melts inferred from
2018). the mafic MIs may themselves be underestimates, as
Journal of Petrology, 2022, Vol. 63, No. 6 | 19

between the least differentiated and the main popula-


melt extracted from mid to lower crust
tion of MI were calculated using their K2 O contents and
H2O saturation
the compositions of the experimental phase assemblage
highest melt inclusion H2O obtained experimentally at the relevant conditions. The
main melt inclusion population amounts obtained, ∼20 vol%, fit the amount of crystal-
1200 68 wt% SiO2 lization expected on the basis of the experimental data

@ 10wt% liquidus
72 wt% SiO2 10 wt% H2O (Fig. 12) and also accounts for the offset in Al2 O3 and

H2O
11 wt% H2O Na2 O between the initial high-pressure LLDs and the

ersat.
predominant bulk-rock compositions of silicic magmas

amph
1000
pl (Figs 9b and 10). We conclude that silicic arc magmas typ-

H2O-und
px ically undergo 20% fractional crystallization as they are
800 ascending through the crust (but up to 60% crystalliza-
Pressure (MPa)

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


tion). If the whole differentiation pathway from parental
64 wt% SiO2 basalts is considered (i.e. high-pressure + decompression
60% melt

8 wt% H2O
600 induced crystallization–differentiation), this represents
80%

90%

only ∼12% of the total amount of fractional crystalliza-


tion (Fig. 9b). Although modest in the overall percentage,
10 wt% H2O
400
crystallization of the plagioclase dominated assemblages
Pi px
taking place during decompression down to 300 MPa
8
essentially erases the high pressure chemical signature.
MSH 6 Notably, this drives the silicic magmas to moderate pera-
200 78
71 luminosity, if not to metaaluminosity. Mixing these differ-
4
sol Col entiated magmas with basaltic andesite magmas in sub-
H2O
idu 66 sa t. volcanic magmatic systems accounts for the metaalumi-
s
0
700 800 900 1000 1100 1200 nosity of arc andesites as previously suggested by Blatter
Temperature (°C) et al. (2013).

Fig. 13. Pressure–temperature diagram showing the melt fraction for Implications for the formation of continental
dacitic melts with ∼10 wt % H2 O initial. Phase relations and melt
fractions are based on the experimental data from Holtz et al. (2005) crust
Prouteau & Scaillet (2003), Rutherford et al. (1985) and Scaillet & Evans The andesitic composition of the continental crust it
(1999). Water saturated granite solidus from Holtz et al. (1992). Dashed
lines represent H2 O saturation pressures calculated for rhyolitic
is often implicitly taken as translating the abundance
compositions using the solubility model of Newman & Lowenstern of andesitic magma produced in arc setting. The com-
(2002). The white and blue arrows illustrate the ascent paths inferred for positional distributions of arc plutonic rocks, but also
three eruptions: PI = Pinatubo 1991 (Borisova et al., 2005; Holtz et al.,
2005), MSH = Mount St Helens 1980 (Blundy et al., 2006), Col = Colima
volcanic rocks in oceanic arcs, challenge this assumption
1998–2005 (Reubi et al., 2013). White circles are potential starting and imply that, if modern arcs are taken as proxies for the
conditions matching the compositions of the least differentiated melt production of the continental crust, its bulk composition
inclusions in each system. The starting conditions are melt
compositions extracted from the 1 GPa LLD from Ulmer et al. (2018) and
has to result from of a mass balance between at least
the 700 MPa LLD from Nandedkar et al. (2014). White triangles mark the two reservoirs (Fig. 1). Estimates for the composition of
pressure of water-saturation and the stars indicate the highest pH2 O the lower and upper crust of Rudnick & Gao (2003) fit
recorded by the melt inclusions. The white rectangles delineate H2 O
contents of the melt inclusion main populations and the inferred
the required chemical characteristics of these reservoirs
temperatures of crystallization based on phase equilibria or (Fig. 2) and are evident entities in this scenario. Modern
experimental data given in the above references. Italic numbers next to arc data also indicate that derivation of the continental
the white rectangles are the SiO2 contents of the melt inclusions on the
least differentiated end of the main populations. Dacitic melts extracted
crust from high-Mg andesites is an unlikely scenario, as
from the lower crust and ascending along an adiabat crystallize these compositions are rare in the bulk-rock record and
extensively only when they reach 100 to 300 MPa. At the conditions essentially absent from the MI record (Fig. 2). The shared
depicted, significant amounts of H2 O may degas before the melts start
crystallizing extensively and the melt inclusions are formed.
compositional characteristics of the continental crust
and andesitic magmas, notably their low P2 O5 contents
(Fig 2l), should consequently be taken as an indication of
the importance of magma mixing in the petrogenesis of
volatile-rich mafic melts will reach vapor saturation at andesites, not as translating the volumetric importance
pressures greater than those at which they intersect the of andesites. As a corollary, the bulk crust composition
vapor-saturated liquidus and before formation of the MI should not be used as an argument to support primary
(Fig. 9a). petrogenetic processes for its formation.
Although their maximum H2 O contents may not be Magmas matching closely the bulk continental crust
trustworthy, the composition of the silicic MIs suggest composition are rare (Figs 2 and 3), implying that the
that they record efficiently the crystallization path of reservoirs involved in the crustal mass balance are
melts ascending from the lower crust (Fig. 5). For the not equivalent to the predominant magma mixing
volcanoes shown in Fig. 13, the amount of crystallization end-members. If the upper crust can reasonably be
20 | Journal of Petrology, 2022, Vol. 63, No. 6

assumed to be initially formed by silicic plutons with related to crystallization of amphibole-rich assemblages.
minor amounts of mafic magmas, commonly expressed Fractional crystallization of amphibole-rich SiO2 poor
as mafic enclaves and volcanics, and is therefore residues at high pressure efficiently yield voluminous
compositionally relatively close to arc batholiths, the silicic magmas but apparently produces a step in the
nature of the lower crust is more open (Fig. 3). Mafic extraction probability of residual melts that limits the
compositions similar to Rudnick & Gao (2003) lower amount of intermediate compositions escaping from
crust estimate are common in the volcanic record and the lower crust. Mid- to upper-crustal magmatic and
along the experimental LLDs (Fig. 2). The estimated plutonic systems are consequently bimodal. There is
compositions for the lower crust in accreted arcs are, a sweet spot for the production of andesitic melts by
however, systematically depleted in highly incompatible crystallization–differentiation of the ascending mafic
trace elements compared with the estimate for the melts delimited by the low-pressure stability limit of
continental lower crust (Jagoutz et al., 2011; Kelemen amphibole (<150 MPa) and the adiabatic ascent path

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


& Behn, 2016) and they have noticeably lower SiO2 and of mafic melts. The productivity of these low-pressure
higher Al2 O3 contents (Fig. 3). The depleted composition differentiation pathways is volumetrically too minor
of the arc lower crust is expected if it comprises to fill the high-pressure andesite trough. In addition,
significant amount of cumulates complementary to these low-pressure trends play a minor role in the
differentiation of the silicic arc magmas from hydrous production of silicic arc magmas that principally form
basalts. The less depleted nature of the lower continental along high-pressure hydrous LLDs (>700 MPa) before
crust suggests either that it does not comprise significant decompression. Decompression-induced crystallization
volumes of cumulates or that addition of less depleted of silicic melts ascending through the crust typically
mafic magmas subsequent to extraction of the silicic represents a limited fraction of the total amount of
magmas forming the upper crust played an important differentiation (∼12%), but it strongly overprints the
role in its chemical evolution. Addition of damp melts chemical signature of high-pressure crystallization–
issued from decompression melting of the mantle wedge differentiation.
beneath the arc or relamination of subducted buoyant In this context, the andesitic composition of the bulk
arc magmatic rocks have been proposed to account for continental crust cannot result from primary magmatic
the less depleted composition of the lower crust (Jagoutz processes but must be a mass balance between the com-
et al., 2011; Kelemen & Behn, 2016). In any case, removal positions of the upper and lower crust. The geochemical
of dense cumulates complementary to the silicic arc similarity between andesitic magmas and juvenile crust
magmas is necessary to obtain an andesitic bulk crust. in arcs should accordingly be taken as telltales of the
Independently of the exact nature of the lower crust, importance of mixing in the production of andesite mag-
the amphibole step encountered by hydrous magmas mas, although the mafic components involved are not
differentiating at high pressure is fundamental in identical. The amphibole step encountered by hydrous
producing the ‘mixed’ andesitic composition of the bulk magmas differentiating at high pressure is paramount in
continental crust, as it defines the silicic composition of controlling not only the magmatic dynamics of andesitic
the batholiths and therefore the upper crust. stratovolcanoes but also in defining the chemical evolu-
tion of the continental upper crust.

CONCLUSIONS
Andesites are emblematic of arc volcanism, yet they FUNDING
represent a limited proportion of the magmas erupted at This work was supported by the University of Lausanne.
oceanic arcs, of the plutons in crustal arc sections and O.M. acknowledges the Swiss National Science Founda-
of arc MIs. The comparison of experimental melt, MI and tion for support (grant no. 200021-162666).
bulk-rock compositions indicates that at least half of the
volcanic and plutonic rocks with around 60 wt % SiO2
are mixing (sensu lato) products, while this proportion DATA AVAILABILITY
exceed three quarters around 64 wt % SiO2. The chemi- The data underlying this article are available in its
cal bimodality shown by arc magmas is a fundamental online supplementary material and were obtained
characteristic of magmatic differentiation in arc crust from GEOROC (https://georoc.eu) and EarthChem Portal
with important consequences for the evolution of the (http://portal.earthchem.org).
continental crust.
The overall bimodality shown by arc melts does not
relate to a compositional gap in the differentiation mech- ACKNOWLEDGMENTS
anisms but result from a combination of (1) the disparity Constructive reviews from Dawnika Blatter and Lena
in the amount of differentiated magmas produced by low Melekhova and the useful editorial comments from
and high pressure crystallization–differentiation and (2) Madeleine Humphreys are gratefully acknowledged.
the strong nonlinearity of the high pressure liquid lines of Gabriel Buret and Jean-Luc Epard are thanked for sharing
descent in composition–temperature–crystallinity space data on the Ladakh Batholith.
Journal of Petrology, 2022, Vol. 63, No. 6 | 21

SUPPLEMENTARY DATA Eichelberger, J. C. (1978). Andesitic volcanism and crustal evolution.


Nature 275, 21–27.
Supplementary data are available at Journal of Petrology
Gaetani, G. A., O’Leary, J. A., Shimizu, N., Bucholz, C. E. & Newville,
online.
M. (2012). Rapid reequilibration of H2 O and oxygen fugacity in
olivine-hosted melt inclusions. Geology 40, 915–918.
Gill, J. B. (1981) Orogenic Andesites and Plate Tectonics. Berlin: Springer.
References
Greene, A. R., DeBari, S. M., Kelemen, P. B., Blusztajn, J. & Clift, P.
Anderson, A. T. (1976). Magma mixing—petrological process and D. (2006). A detailed geochemical study of island arc crust: the
volcanological tool. Journal of Volcanology and Geothermal Research Talkeetna arc section, south-Central Alaska. Journal of Petrology
1, 3–33. 47, 1051–1093.
Annen, C. (2009). From plutons to magma chambers: thermal con- Grove, T. L. & Donnelly-Nolan, J. M. (1986). The evolution of young
straints on the accumulation of eruptible silicic magma in the silicic lavas at medicine Lake volcano, California—implications
upper crust. Earth and Planetary Science Letters 284, 409–416. for the origin of compositional gaps in calc-alkaline series lavas.

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


Baker, M. B., Grove, T. L. & Price, R. C. (1994). Primitive basalts and Contributions to Mineralogy and Petrology 92, 281–302.
andesites from the Mt. Shasta region, N. California: products of Grove, T. L., DonnellyNolan, J. M. & Housh, T. (1997). Magmatic pro-
varying melt fraction and water content. Contributions to Mineral- cesses that generated the rhyolite of Glass Mountain, medicine
ogy and Petrology(2) 118, 111–129. Lake volcano, N California. Contributions to Mineralogy and Petrology
Blatter, D. L., Sisson, T. W. & Ben Hankins, W. (2013). Crystallization 127, 205–223.
of oxidized, moderately hydrous arc basalt at mid- to lower- Grove, T. L., Elkins-Tanton, L. T., Parman, S. W., Chatterjee, N., Mün-
crustal pressures: implications for andesite genesis. Contributions tener, O. & Gaetani, G. A. (2003). Fractional crystallization and
to Mineralogy and Petrology 166, 861–886. mantle-melting controls on calc-alkaline differentiation trends.
Blatter, D. L., Sisson, T. W. & Ben Hankins, W. (2017). Voluminous arc Contributions to Mineralogy and Petrology 145, 515–533.
dacites as amphibole reaction-boundary liquids. Contributions to Hacker, B. R., Kelemen, P. B. & Behn, M. D. (2011). Differentiation of
Mineralogy and Petrology 172, 27. the continental crust by relamination. Earth and Planetary Science
Blundy, J. & Cashman, K. (2001). Ascent-driven crystallisation of Letters 307, 501–516.
dacite magmas at mount St Helens, 1980-1986. Contributions to Heliker, C. (1995). Inclusions in Mount St. Helens dacite erupted
Mineralogy and Petrology 140, 631–650. from 1980 through 1983. Journal of Volcanology and Geothermal
Blundy, J. & Cashman, K. (2005). Rapid decompression-driven crys- Research(1–4) 66, 115–135.
tallization recorded by melt inclusions from Mount St. Helens Hildreth, W. & Moorbath, S. (1988). Crustal contributions to arc Mag-
volcano. Geology(10) 33, 793–796. matism in the Andes of Central Chile. Contributions to Mineralogy
Blundy, J., Cashman, K. & Humphreys, M. (2006). Magma heating and Petrology 98, 455–489.
by decompression-driven crystallization beneath andesite volca- Hildreth, W. (2007). Quaternary magmatism in the Cascades. Geologic
noes. Nature 443, 76–80. perspectives. USGS Professional Paper 1744.
Borisova, A. Y., Pichavant, M., Beny, J. M., Rouer, O. & Pronost, J. (2005). Holtz, F., Pichavant, M., Barbey, P. & Johannes, W. (1992). Effect of H2O
Constraints on dacite magma degassing and regime of the June on liquidus phase-relations in the haplogranite system at 2 and
15, 1991, climactic eruption of Mount Pinatubo (Philippines): new 5 kbar. American Mineralogist 77, 1223–1241.
data on melt and crystal inclusions in quartz. Journal of Volcanology Holtz, F., Sato, H., Lewis, J., Behrens, H. & Nakada, S. (2005). Exper-
and Geothermal Research 145, 35–67. imental petrology of the 1991-1995 Unzen dacite, Japan. Part I:
Bouvet de Maisonneuve, C., Dungan, M. A., Bachmann, O. & Burgisser, phase relations, phase composition and pre-eruptive conditions.
A. (2012). Insights into shallow magma storage and crystallization Journal of Petrology 46, 319–337.
at Volcán Llaima (Andean southern volcanic zone, Chile). Journal de Hoog, J. C. M., Hattori, K. H. & Hoblitt, R. P. (2004). Oxidized sulfur-
of Volcanology and Geothermal Research 211–212, 76–91. rich mafic magma at Mount Pinatubo, Philippines. Contributions to
Bowen, N. L. (1928) The evolution of the igneous rocks. Princeton Univer- Mineralogy and Petrology 146, 750–761.
sity Press, Princeton. Humphreys, M. C. S., Christopher, T. & Hards, V. (2009). Microlite
Carmichael, I. S. E. (2002). The andesite aqueduct: perspectives on transfer by disaggregation of mafic inclusions following magma
the evolution of intermediate magmatism in west-central (105– mixing at Soufrière Hills volcano, Montserrat. Contributions to
99◦ W) Mexico. Contributions to Mineralogy and Petrology(6) 143, Mineralogy and Petrology(5) 157, 609–624.
641–663. Humphreys, M. C. S., Edmonds, M., Christopher, T. & Hards, V. (2010).
Cashman, K. & Blundy, J. (2013). Petrological cannibalism: the chem- Magma hybridisation and diffusive exchange recorded in hetero-
ical and textural consequences of incremental magma body geneous glasses from Soufriere Hills volcano, Montserrat. Geo-
growth. Contributions to Mineralogy and Petrology 166, 703–729. physical Research Letters 37, https://doi.org/10.1029/2009GL041926.
Cassidy, M., Edmonds, M., Watt, S. F. L., Palmer, M. R. & Gernon, T. M. Humphreys, M. C. S., Cooper, G. F., Zhang, J., Loewen, M., Kent, A.
(2015). Origin of basalts by hybridization in andesite-dominated J. R., Macpherson, C. G. & Davidson, J. P. (2019). Unravelling the
arcs. Journal of Petrology 56, 325–346. complexity of magma plumbing at Mount St. Helens: a new
Clemens, J. D., Stevens, G. & Farina, F. (2011). The enigmatic sources trace element partitioning scheme for amphibole. Contributions
of I-type granites: the peritectic connexion. Lithos 126, 174–181. to Mineralogy and Petrology 174, 9.
Di Muro, A., Pallister, J., Villemant, B., Newhall, C., Semet, M., Martinez, Jagoutz, O. (2014). Arc crustal differentiation mechanisms. Earth and
M. & Mariet, C. (2008). Pre-1991 sulfur transfer between mafic Planetary Science Letters 396, 267–277.
injections and dacite magma in the Mt. Pinatubo reservoir. Journal Jagoutz, O. & Kelemen, P. B. (2015). Role of arc processes in the for-
of Volcanology and Geothermal Research 175, 517–540. mation of continental crust. Annual Review of Earth and Planetary
Dufek, J. & Bachmann, O. (2010). Quantum magmatism: magmatic Sciences 43, 363–404.
compositional gaps generated by melt-crystal dynamics. Geology Jagoutz, O. & Klein, B. (2018). On the importance of crystallization–
38, 687–690. differentiation for the generation of SiO2 -rich melts and the
22 | Journal of Petrology, 2022, Vol. 63, No. 6

compositional build-up of arc (and continental) crust. American Metrich, N. & Wallace, P. J. (2008) Volatile Abundances in Basaltic
Journal of Science 318, 29–63. Magmas and Their Degassing Paths Tracked by Melt Inclusions.
Jagoutz, O. & Schmidt, M. W. (2012). The formation and bulk com- In: (Putirka K. D. & Tepley F. J. (eds)) Minerals, Inclusions and Vol-
position of modern juvenile continental crust: the Kohistan arc. canic Processes, Mineralogical Society of America and Geochemical
Chemical Geology 298, 79–96. Society, Reviews in Mineralogy and Geochemistry, vol. 69, pp.
Jagoutz, O. E., Burg, J. P., Hussain, S., Dawood, H., Pettke, T., Iizuka, T. 363–402.
& Maruyama, S. (2009). Construction of the granitoid crust of an Moore, G. & Carmichael, I. S. E. (1998). The hydrous phase equi-
island arc part I: geochronological and geochemical constraints libria (to 3 kbar) of an andesite and basaltic andesite from
from the plutonic Kohistan (NW Pakistan). Contributions to Miner- western Mexico: constraints on water content and conditions of
alogy and Petrology 158, 739–755. phenocryst growth. Contributions to Mineralogy and Petrology 130,
Jagoutz, O., Müntener, O., Schmidt, M. & Burg, J. (2011). The roles of 304–319.
flux- and decompression melting and their respective fraction- Müntener, O. & Ulmer, P. (2018). Arc crust formation and differenti-
ation lines for continental crust formation: evidence from the ation constrained by experimental petrology. American Journal of

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


Kohistan arc. Earth and Planetary Science Letters(1–2) 303, 25–36. Science 318, 64–89.
Kay, R. W. (1978). Aleutian magnesian andesites—melts from sub- Müntener, O., Kelemen, P. B. & Grove, T. L. (2001). The role of H2 O
ducted Pacific Ocean crust. Journal of Volcanology and Geothermal during crystallization of primitive arc magmas under upper-
Research 4, 117–132. most mantle conditions and genesis of igneous pyroxenites: an
Kay, S. M., Jicha, B. R., Citron, G. L., Kay, R. W., Tibbetts, A. K. & Rivera, experimental study. Contributions to Mineralogy and Petrology 141,
T. A. (2019). The calc-alkaline Hidden Bay and Kagalaska plutons 643–658.
and the construction of the central Aleutian oceanic arc crust. Müntener, O., Ulmer, P. & Blundy, J. D. (2021). Superhydrous arc
Journal of Petrology 60, 393–439. magmas in the Alpine context. Elements 17, 35–40.
Kelemen, P. B. (1986). Assimilation of ultramafic rock in subduction- Murphy, M. D., Sparks, R. S. J., Barclay, J., Carroll, M. R. & Brewer, T. S.
related magmatic arcs. Journal of Geology 94, 829–843. (2000). Remobilization of andesite magma by intrusion of mafic
Kelemen, P. B. & Behn, M. D. (2016). Formation of lower continental magma at the Soufriere Hills volcano, Montserrat, West Indies.
crust by relamination of buoyant arc lavas and plutons. Nature Journal of Petrology 41, 21–42.
Geoscience 9, 197–205. Myers, M. L., Geist, D. J., Rowe, M. C., Harpp, K. S., Wallace, P. J. &
Keller, C. B., Schoene, B., Barboni, M., Samperton, K. M. & Husson, J. Dufek, J. (2014). Replenishment of volatile-rich mafic magma into
M. (2015). Volcanic-plutonic parity and the differentiation of the a degassed chamber drives mixing and eruption of Tungurahua
continental crust. Nature(7560) 523, 301–307. volcano. Bulletin of Volcanology 76.
Kent, A. J. R. (2014) Preferential eruption of andesitic magmas: impli- Nandedkar, R. H., Ulmer, P. & Müntener, O. (2014). Fractional crystal-
cations for volcanic magma fluxes at convergent margins. In: lization of primitive, hydrous arc magmas: an experimental study
(Gomez-Tuena A., Straub S. M. & Zellmer G. F. (eds)) Orogenic at 0.7 GPa. Contributions to Mineralogy and Petrology 167, 1015.
andesites and crustal growth, Vol. 385. Geological Society, London, Newhall, C. G. & Self, S. (1982). The volcanic explosivity index (VEI) an
pp.257–280. estimate of explosive magnitude for historical volcanism. Journal
Kent, A. J. R., Darr, C., Koleszar, A. M., Salisbury, M. J. & Cooper, of Geophysical Research Oceans(C2) 87, 1231–1238.
K. M. (2010). Preferential eruption of andesitic magmas through Newman, S. & Lowenstern, J. B. (2002). VOLATILECALC: a silicate
recharge filtering. Nature Geoscience 3, 631–636. melt-H2 O-CO2 solution model written in visual basic for excel.
Klein, B. Z. & Jagoutz, O. (2021). Construction of a trans-crustal Computers & Geosciences 28, 597–604.
magma system: building the Bear Valley intrusive suite, southern Pallister, J. S., Hoblitt, R. P. & Reyes, A. G. (1992). A basalt trigger
Sierra Nevada, California. Earth and Planetary Science Letters 553, for the 1991 eruptions of Pinatubo volcano? Nature(6368) 356,
116624–116615. 426–428.
Kratzmann, D. J., Carey, S., Scasso, R. & Naranjo, J. A. (2009). Compo- Pioli, L., Scalisi, L., Costantini, L., Di Muro, A., Bonadonna, C. & Clavero,
sitional variations and magma mixing in the 1991 eruptions of J. (2015). Explosive style, magma degassing and evolution in the
Hudson volcano, Chile. Bulletin of Volcanology 71, 419–439. Chaimilla eruption, Villarrica volcano, southern Andes. Bulletin of
Kushiro, I. (1975). On the nature of silicate melt and its significance Volcanology 77.
in magma genesis; regularities in the shift of the liquidus bound- Plail, M., Edmonds, M., Woods, A. W., Barclay, J., Humphreys, M. C. S.,
aries involving olivine, pyroxene, and silica minerals. American Herd, R. A. & Christopher, T. (2018). Mafic enclaves record syn-
Journal of Science(4) 275, 411–431. eruptive basalt intrusion and mixing. Earth and Planetary Science
LeBas, M. J., Lemaitre, R. W. & Woolley, A. R. (1992). The construction Letters 484, 30–40.
of the total alkali-silica chemical classification of volcanic rocks. Portnyagin, M. V., Hoernle, K. & Mironov, N. L. (2014). Contrasting
Mineralogy and Petrology 46, 1–22. compositional trends of rocks and olivine-hosted melt inclusions
Lee, C. T. A. & Bachmann, O. (2014). How important is the role of from Cerro Negro volcano (Central America): implications for
crystal fractionation in making intermediate magmas? Insights decompression-driven fractionation of hydrous magmas. Interna-
from Zr and P systematics. Earth and Planetary Science Letters 393, tional Journal of Earth Sciences 103, 1963–1982.
266–274. Prouteau, G. & Scaillet, B. (2003). Experimental constraints on
Mann, C. P., Wallace, P. J. & Stix, J. (2013). Phenocryst-hosted melt the origin of the 1991 Pinatubo dacite. Journal of Petrology 44,
inclusions record stalling of magma during ascent in the con- 2203–2241.
duit and upper magma reservoir prior to vulcanian explosions, Rapp, R. P. & Watson, E. B. (1995). Dehydration melting of Metabasalt
Soufriere Hills volcano, Montserrat, West Indies. Bulletin of Vol- at 8-32 kbar: implications for continental growth and crust-
canology 75, 687. mantle recycling. Journal of Petrology(4) 36, 891–931.
Melson, W. G. (1983). Monitoring the 1980-1982 eruptions of Mount Rawson, H., Naranjo, J. A., Smith, V. C., Fontijn, K., Pyle, D. M., Mather,
St. Helens—compositions and abundances of glass. Science 221, T. A. & Moreno, H. (2015). The frequency and magnitude of
1387–1391. post-glacial explosive eruptions at Volcán Mocho-Choshuenco,
Journal of Petrology, 2022, Vol. 63, No. 6 | 23

southern Chile. Journal of Volcanology and Geothermal Research 299, Straub, S. M. (2003). The evolution of the Izu Bonin-Mariana volcanic
103–129. arcs (NW Pacific) in terms of major element chemistry. Geochem-
Reubi, O. & Blundy, J. (2008). Assimilation of plutonic roots, forma- istry Geophysics Geosystems 4, 1018.
tion of high-K exotic melt inclusions and genesis of andesitic Tatsumi, Y. & Suzuki, T. (2009). Tholeiitic vs Calc-alkalic differ-
magmas at Volcán de Colima, Mexico. Journal of Petrology 49, entiation and evolution of arc crust: constraints from melting
2221–2243. experiments on a basalt from the Izu-Bonin-Mariana arc. Journal
Reubi, O. & Blundy, J. (2009). A dearth of intermediate melts at of Petrology 50, 1575–1603.
subduction zone volcanoes and the petrogenesis of arc andesites. Taylor, S. R. & McLennan, S. M. (1995). The geochemical evolution of
Nature 461, 1269–1273. the continental-crust. Reviews of Geophysics(2) 33, 241–265.
Reubi, O., Blundy, J. & Varley, N. R. (2013). Volatiles contents, degassing Taylor, S. R. & White, A. J. R. (1965). Geochemistry of Andesites and
and crystallisation of intermediate magmas at Volcan de Colima, the growth of continents. Nature(5007) 208, 271–273.
Mexico, inferred from melt inclusions. Contributions to Mineralogy Todd, V. R., Shaw, S. E. & Hammarstrom, J. M. (2003). Cretaceous
and Petrology 165, 1087–1106. plutons of the peninsular Ranages batholith, San Diego and

Downloaded from https://academic.oup.com/petrology/article/63/6/egac044/6586529 by guest on 22 May 2023


Richards, J. P. (2011). Magmatic to hydrothermal metal fluxes westernmost Imperial counties California; intrusion across a late
in convergent and collided margins. Ore Geology Reviews Jurassic continental margin. Geological Society of America Special
40, 1–26. Paper(5007) 374, 185–235.
Rioux, M., Hacker, B., Mattinson, J., Kelemen, P., Blusztajn, J. & Gehrels, Tolstykh, M. L., Naumov, V. B., Gavrilenko, M. G., Ozerov, A. Y. &
G. (2007). Magmatic development of an intra-oceanic arc: high- Kononkova, N. N. (2012). Chemical composition, volatile compo-
precision U-Pb zircon and whole-rock isotopic analyses from the nents, and trace elements in the melts of the Gorely volcanic cen-
accreted Talkeetna arc, south-Central Alaska. Geological Society of ter, southern Kamchatka: evidence from inclusions in minerals.
America Bulletin 119, 1168–1184. Geochemistry International 50, 522–550.
Roggensack, K. (2001). Unraveling the 1974 eruption of Fuego volcano Ulmer, P., Kaegi, R. & Müntener, O. (2018). Experimentally derived
(Guatemala) with small crystals and their young melt inclusions. intermediate to silica-rich arc magmas by fractional and equilib-
Geology(10) 29, 911–914. rium crystallization at 1.0 GPa: an evaluation of phase relation-
Roggensack, K., Hervig, R. L., McKnight, S. B. & Williams, S. N. (1997). ships, compositions, liquid lines of descent and oxygen fugacity.
Explosive basaltic volcanism from Cerro Negro volcano: influence Journal of Petrology 59, 11–58.
of volatiles on eruptive style. Science 277, 1639–1642. Villiger, S., Ulmer, P., Müntener, O. & Thompson, A. B. (2004). The
Rudnick, R. L. & Gao, S. (2003) Composition of the continental crust. liquid line of descent of anhydrous, mantle-derived, tholeiitic
In: (Rudnick R. L. (ed)) Treatise on Geochemistry, Elsevier-Pergamon, liquids by fractional and equilibrium crystallization—an exper-
Oxford, Vol. 3, pp.1–64. imental study at 1 center dot 0 GPa. Journal of Petrology 45,
Rutherford, M. J., Sigurdsson, H., Carey, S. & Davis, A. M. (1985). 2369–2388.
The May 18, 1980, eruption of Mount St. Helens: 1. Melt com- Wade, J. A., Plank, T., Stern, R. J., Tollstrup, D. L., Gill, J. B., O’Leary, J. C.,
position and experimental phase equilibria. Journal of Geophysical Eiler, J. M., Moore, R. B., Woodhead, J. D., Trusdell, F., Fischer, T. P. &
Research(B4) 90, 2929–2947. Hilton, D. R. (2005). The May 2003 eruption of Anatahan volcano,
Sadofsky, S. J., Portnyagin, M., Hoernle, K. & van den Bogaard, P. (2008). Mariana Islands: geochemical evolution of a silicic island-arc vol-
Subduction cycling of volatiles and trace elements through the cano. Journal of Volcanology and Geothermal Research 146, 139–170.
central American volcanic arc: evidence from melt inclusions. Wallace, P. J. (2005). Volatiles in subduction zone magmas: concentra-
Contributions to Mineralogy and Petrology 155, 433–456. tions and fluxes based on melt inclusion and volcanic gas data.
Scaillet, B. & Evans, B. W. (1999). The 15 June 1991 eruption of Journal of Volcanology and Geothermal Research 140, 217–240.
Mount Pinatubo. I. Phase equilibria and pre-eruption P-T-Fo2- Wanke, M., Karakas, O. & Bachmann, O. (2019). The genesis of arc
Fh2O conditions of the dacite magma. Journal of Petrology(3) 40, dacites: the case of Mount St. Helens, WA. Contributions to Miner-
381–411. alogy and Petrology 174, 10.
Sisson, T. W. & Grove, T. L. (1993). Experimental investigations of Williamson, B. J., Di Muro, A., Horwell, C. J., Spieler, O. & Llewellin, E.
the role of H2O in calc-alkaline differentiation and subduc- W. (2010). Injection of vesicular magma into an andesitic dome at
tion zone magmatism. Contributions to Mineralogy and Petrology the effusive-explosive transition. Earth and Planetary Science Letters
113, 143–166. 295, 83–90.
Sisson, T. W. & Layne, G. D. (1993). H2O in basalt and basaltic andesite Zellmer, G. F., Hawkesworth, C. J., Sparks, R. S. J., Thomas, L. E.,
glass inclusions from four subduction-related volcanoes. Earth Harford, C. L., Brewer, T. S. & Loughlin, S. C. (2003). Geochemi-
and Planetary Science Letters(3–4) 117, 619–635. cal evolution of the Soufrière Hills volcano, Montserrat, Lesser
Sisson, T. W., Ratajeski, K., Hankins, W. B. & Glazner, A. F. (2005). Antilles volcanic arc. Journal of Petrology 44, 1349–1374.
Voluminous granitic magmas from common basaltic sources. Zhang, J., Humphreys, M. C. S., Cooper, G. F., Davidson, J. P. &
Contributions to Mineralogy and Petrology 148, 635–661. Macpherson, C. G. (2017). Magma mush chemistry at subduction
Smith, I. E. M., Worthington, T. J., Price, R. C., Stewart, R. B. & Maas, zones, revealed by new melt major element inversion from calcic
R. (2006). Petrogenesis of dacite in an oceanic subduction envi- amphiboles. American Mineralogist 102, 1353–1367.
ronment: Raoul Island, Kermadec arc. Journal of Volcanology and Zimmer, M. M., Plank, T., Hauri, E. H., Yogodzinski, G. M., Stelling, P.,
Geothermal Research 156, 252–265. Larsen, J., Singer, B., Jicha, B., Mandeville, C. & Nye, C. J. (2010). The
Sparks, R. S. J., Sigurdsson, H. & Wilson, L. (1977). Magma mixing: a role of water in generating the calc-alkaline trend: new volatile
mechanism for triggering acid explosive eruptions. Nature(5609) data for Aleutian magmas and a New Tholeiitic index. Journal of
267, 315–318. Petrology 51, 2411–2444.

You might also like