Contributions to Mineralogy and Petrology (2021) 176:13
https://doi.org/10.1007/s00410-020-01764-3
ORIGINAL PAPER
Chemical and textural diversity of Kameni (Greece) dacites: role
of vesiculation in juvenile and mature basal crystal masses
Michael D. Higgins1
· Anouk Debecq2 · Jacqueline Vander Auwera3 · Paraskevi Nomikou4
Received: 25 August 2020 / Accepted: 12 December 2020 / Published online: 28 January 2021
© The Author(s), under exclusive licence to Springer-Verlag GmbH, DE part of Springer Nature 2021
Abstract
Dacite lavas erupted from Kameni Islands volcanic centre (Greece) during the last 2000 years have a limited range in chemical
composition (SiO2 = 64.0–68.5%) which contrasts with their wide range in plagioclase abundance (3–22%) and crystal size
distributions. Most plagioclase crystals have simple zoning and occur independently or in loose clusters with finer-grained
cores. We propose that magmatic diversity was produced by the interaction between crystals that formed at the base of a
magma reservoir and bubbles produced by injection and vesiculation of more mafic magma. Two end-member situations can
be identified: in juvenile systems, the basal crystal mass is loosely connected and readily disrupted by bubble formation. The
crystal–bubble couples accumulate at the top of the reservoir, from where they can enter the sub-volcanic plumbing system
to produce high-crystal content, chemically unevolved magmas. In the mature system, the crystal mass is well connected so
bubbles displace the evolved, interstitial magma and liberate only a smaller number of crystals from the crystal mass. This
process produces chemically evolved magmas, with lower crystal contents. The oldest lavas seem to have been produced
from mature systems, whereas the youngest eruptions were of lavas produced from juvenile systems. This progression may
reflect an overall reduction in repose times during the last 2000 years.
Keywords Dacite · Enclaves · Bubbles · Texture · CSD · Santorini
Introduction
An important challenge in the study of volcanos is to understand how the underlying magmatic system has evolved with
time, and how it will behave in the future. For recent eruptions or intrusive events, we can use geological observation,
cultural information, and geophysical methods to determine
Communicated by Timothy L. Grove.
Supplementary Information The online version contains
supplementary material available at https://doi.org/10.1007/s0041
0-020-01764-3.
* Michael D. Higgins
mhiggins@uqac.ca
1
Sciences Appliquées, Université du Québec à Chicoutimi,
Québec G7J 4M2, Canada
2
Université Libre de Bruxelles, Bruxelles, Belgium
3
Université de Liège, Liège, Belgium
4
National and Kapodistrian University of Athens, Athens,
Greece
the timing and movements of magma (Pyle 2017). Aspects
of current magma distribution can be measured using seismic tomography (Hooft et al. 2019; McVey et al. 2019).
However, many older eruptions can only be understood by
studying lava and ash produced by the eruptions.
An ideal volcano for such a study would be easily accessible, would erupt frequently with well-described and timed
volcanic events, and volcanic products would be well preserved. The Kameni Islands volcanic centre has many of
these characteristics and has hence been the object of many
studies to be described later. Most work to date has examined
the chemical composition of the lavas and enclaves, with a
much smaller number of studies on the texture (microstructure) of the rocks. However, few have used the combination
of geochemical and textural work that has proved to be so
powerful in the understanding of other active volcanos (e.g.
Higgins et al. 2015).
Here we want to address the following problems: how was
the textural and chemical diversity of the Kameni dacites
produced and what were the mechanisms in the underlying magmatic system that produced this diversity; can the
eruptions of the Kameni Islands during the last 2000 years
13
Vol.:(0123456789)
13
Page 2 of 20
be considered as a single, stable phase, or must the history
be divided up into several separate volcanic cycles? This
leads us to consider if we are now near the end of an eruptive cycle and how this may inform us about the nature of
future eruptions.
We would like to mention that although magma reservoirs
are discussed in the text and figures we acknowledge that the
magmatic ‘plumbing system’ under the Kameni volcanos
is much more complex than this, as shown by the recent
seismic unrest and tomographic models (Hooft et al. 2019;
McVey et al. 2019; Parks et al. 2012).
The Kameni islands
Eruptive history
Santorini volcano (also known as Thira or Thera) was
formed at least 650,000 years ago and has had numerous
major Plinian eruptions (Druitt et al. 2019b). The most
recent was the ‘Minoan’ eruption in ~ 1600 BCE which
produced a caldera that forms the central bay between
Thira and Thirasia islands (Druitt et al. 2019a). Since
that time, eruptions have been confined to the central part
of the caldera. We do not know when extrusive activity
restarted, but the earliest subaerial post-Minoan eruption
was in 197 BCE and produced the island of Hiera (Theodorakopoulou et al. 2020). It was probably situated 4 km
southwest of Fira town and reduced to a bank before being
Fig. 1 The Santorini Island group includes Thira (Thera), Thirasia as
well as the Kameni islands whose lava flows are studied here. Highresolution bathymetry/topography of the Kameni group shows the
13
Contributions to Mineralogy and Petrology (2021) 176:13
buried by the 1866–70 eruption (Fig. 1). The next subaerial eruption was in 46–47 CE and produced the core of
Palea Kameni Island (PK). The eruptive products extend
underwater to the south and west. The 726 CE subaerial
eruption was the most violent in the history of the islands
(VEI = 4) and caused considerable damage (Vougioukalakis and Fytikas 2005). It produced large quantities of
pumice which floated as far as the Dardanelles and may
have been used to justify the imposition of the Iconoclasm
by the Byzantine Emperor Leo III. Although knowledge of
underwater eruptions is sparse, there may have been a hiatus for over 800 years until 1570 CE when subaerial activity restarted 3 km to the northeast. Although the current
outcrop of this flow is small the flow appears to continue
underwater for 2 km to the north. Large effusive eruptions
continued to occur at decreasing intervals building Nea
Kameni Island (NK). The volume of the subaerial flows
diminished in the twentieth century and the last event in
1950 CE was very small.
Recent high-resolution bathymetric observations have
considerably expanded our view of the subaerial flows,
showing them to be significantly larger than previously
estimated. In addition, the studies have revealed at least
three submarine flows unrelated to the subaerial flows: NK
North, NK East, and Drakon (Fig. 1; Nomikou et al. 2014).
The eruption dates of these flows are partly constrained by
historic lead-line bathymetry (Watts et al. 2015) and their
relationships to adjacent flows: NK North erupted sometime
underwater continuity of subaerial flows and the presence of some
previously unknown flows. Source: (Nomikou et al. 2014)
Page 3 of 20 13
Contributions to Mineralogy and Petrology (2021) 176:13
between 1848 and 1925; Drakon erupted after 1848, and NK
East erupted before 1848.
Seismic unrest in 2011–2012 was related to magma
movements beneath the edifice (Newman et al. 2012; Parks
et al. 2012, 2015). Earthquake loci showed that the magma
rose along the southern caldera fault, essentially beneath the
Kameni Islands, and stalled at a depth of about 4 km (Parks
et al. 2015). This depth accords with seismic tomography
models (Hooft et al. 2019; McVey et al. 2019). The magma
then appeared to feed a sill beneath the caldera a few 100 m
north of Nea Kameni, close to the centre of the northern part
of the bay. Inflation indicated that about 1–2.107 m3 of new
magma was emplaced during the event (Parks et al. 2015).
Helium isotopes suggest that the event was triggered by the
injection of primitive magma (Rizzo et al. 2015).
The eruptive history of the currently accessible parts of
the Kameni Island volcanic system appears to comprise
two or three distinct phases. The first phase produced Palea
Kameni island and culminated in the major explosive eruption of 726 CE. There was a hiatus for 800 years before
activity resumed 3 km to the northeast and built Nea Kameni
island. Chemical and textural data support this division into
PK and NK phases, and also a less well-defined phase that
comprises the latest eruptions (1939–1950). It is not clear
if the 197 BCE eruption was part of the PK phase as we do
not know its exact position or composition.
Relevant earlier work
The Kameni islands’ lava flows, and their enclaves, have
been the subject of several geochemical and textural studies
over the last 30 years, commonly as part of a wider study on
the whole Santorini volcanic system. The subaerial Kameni
volcanic ensemble has many advantages: it is accessible, the
rocks are mostly fresh, and all the flows are well dated. Samples of some of the underwater phases of the Kameni volcanic ensemble have recently become available (Nomikou
et al. 2014).
The earliest study of the geochemistry of Santorini volcano was by Nicholls (1971), who examined a few Kameni
samples. However, analyses mostly used uncorrected XRF
and hence cannot be compared with more recent data. The
first study using modern XRF methods was that of Huijsmans (1985). Some of his data was published in (Barton and
Huijsmans 1986). Most of the analyses were of dacite samples and they concluded that despite a range in composition
between individual samples, the mean chemical composition
of flows changed little during the last 2000 years. Crystal
size distributions (CSD) of plagioclase were determined in
10 samples of dacite by Higgins (1996b). These samples
make up some of the material reanalysed in this study. Francalanci et al. (1998) analysed a large number of samples of
dacite and enclaves but only published mean compositional
values for each flow and a few representative analyses of
enclaves. Zellmer et al. (2000) analysed dacites and enclaves
as part of an isotopic study. Martin et al. (2006) included
two analyses of dacite as part of a study of the enclaves.
The overall conclusion of these studies is that the chemical
composition of the dacites is relatively uniform, as compared to products of the Santorini volcano as a whole, but
that there is enough geochemical variation to be studied by
modern analytical techniques. However, textural variations
are more pronounced than geochemical variations and the
two together can help us understand the evolution of this
magma system.
The enclaves in the dacite lavas have been studied as
part of larger studies and independently. Most studies have
looked at the overall chemical composition of the enclaves
and concluded that their chemical composition is much more
variable than that of the dacites (Barton and Huijsmans
1986; Francalanci et al. 1998; Martin 2005; Martin et al.
2006; Zellmer et al. 2000). The enclaves’ textures in post1925 flows were studied by Martin et al. (2006) who classified the enclaves into three groups of micrographic enclaves
(A-1, A-2, B) and one group of cumulate enclaves. The latter
is made of plagioclase crystals up to 10 mm long, frequently
intergrown with olivine: they crystallised from mafic magma
and been coarsened by equilibration at high temperatures.
There are no data on the overall chemical composition of
the cumulates. Finally, it should be noted that the enclaves
are vesicular with a much higher bubble content than their
surrounding magma (Martin 2005; Martin et al. 2006).
Different populations of plagioclase crystals have been
recognised in the Kameni lavas, as in most andesites and
dacites elsewhere. Plagioclase populations in enclaves have
not been considered. Barton and Huijsmans (1986) identified ‘phenocrysts’ and ‘xenocrysts’ based on their composition and zoning. Their phenocrysts are euhedral, up to
1 mm long, with brown glass inclusions and compositions
of An55–An42. The xenocrysts are generally larger, with
resorbed anorthitic cores (An90–An71), and are considered to
be derived from disaggregated enclaves. StamatelopoulouSeymour et al. (1990) studied the composition of a small
number of crystals in much detail. They produced a similar
classification but divided the first class of Barton and Huijsmans (1986) into inherited crystals and those that crystallised in situ. The abundance of these crystal populations was
not discussed in either study.
Methods
Sampling
Samples were taken from the surface of the lava flows
on Palea Kameni and Nea Kameni islands during two
13
13
Page 4 of 20
campaigns in 1990 (Higgins 1996b) and 2017 (see supplementary table 1). All samples were fresh and unaltered. Dacite lava samples were selected to exclude visible enclaves,
and enclaves were sampled avoiding contamination by dacite. We did not sample material from the scoria cones. The
flow dates were determined from existing studies and field
observations (Druitt et al. 1999).
The surface samples were complemented by samples
taken underwater using an ROV (Sigurdsson et al. 2006).
Most lava samples were fresh, but scoria samples appeared
to have been altered by long exposure to seawater. We used
the flow association and nomenclature of Nomikou et al.
(2014). All the underwater lava samples were assigned to the
NK North flow. Scoria samples were taken from the underwater surface of the 46–7 and 726 eruptions.
Chemistry
The samples were coarsely crushed with a hammer and
ground in an agate planetary mill to a fine powder. Major
elements were analysed by X-ray fluorescence at the Université de Liège (Belgium) using a Thermofisher PERFORM’X
X-ray fluorescence spectrometer with a Rh tube. These elements were measured on fused glass discs that were prepared with lithium tetra- and meta-borate and 0.35 g of rock
powder previously dried at 1000 C for two hours. Some
trace elements were also measured by X-ray fluorescence
on pressed powdered pellets. The instrument was calibrated
using 40–60 international standards for major and trace elements, respectively. Iron was expressed as Fe2O3. Reproducibility was estimated to be better than 1% for most major
elements and 3% for minor elements. Most trace elements
Fig. 2 Dacite lava (NK-21) with
a microgranular mafic enclave
in, (a) linear cross-polarised
image and (b) cathodoluminescence image. Variation in the
cathodoluminescence colour of
crystals is due to compositional
differences and variable orientations of the crystal lattice with
respect to the direction of electron illumination: hence zoning
and twinning are sometimes
visible. Cathodoluminescence
colours: Green = plagioclase;
yellow = apatite; black = pyroxenes, magnetite and vesicles.
The enclave is of type A-1
according to the classification of
Martin (2006)
13
Contributions to Mineralogy and Petrology (2021) 176:13
were analysed by ICP-MS following alkali fusion using a
Thermo Scientific X-Series 2 with a collision cell (MRAC
Tervuren, Belgium). Reproducibility was estimated to be
better than 3%. Chemical data are presented in electronic
appendix table S1.
Petrography
Plagioclase textures were examined in polished thin sections using conventional transmitted light microscopy and
cathodoluminescence (Fig. 2). The intensity and colour of
the light emitted during cathodoluminescence are dominantly controlled by trace chemical composition (Pagel et al.
2000). The most important activator in plagioclase is Mn2+,
but Cu2+, Fe3+, Ti4+, Ce3+ and Eu2+ may also be important
(Mora and Ramseyer 1992). Another factor is lattice orientation, hence, twinning is commonly visible.
In this study, cathodoluminescence colour was used not
only to help quantify plagioclase textures but also to identify
the abundance of different petrographic classes of plagioclase (Götze et al. 2012; Higgins et al. 2015). The cathodoluminescence images were obtained using a CITL CL8200
cold-cathode instrument mounted on a regular petrographic
microscope. Multiple images were mosaicked to give composite images that typically covered half a regular thin section. The texture of the plagioclase crystal population was
quantified from the composite cathodoluminescence image.
The cathodoluminescence images were thresholded for plagioclase and the binary image was edited manually to separate touching crystals and remove artefacts. The resulting
binary image was quantified using the software package
Page 5 of 20 13
Contributions to Mineralogy and Petrology (2021) 176:13
ImageJ (Rasband 2010). The smallest crystal that could be
consistently measured was 0.05 mm long.
The intersection data were converted to CSDs using the
stereological program CSDCorrections version 1.6 (Higgins
2000). This method needs an estimate of the crystal shape.
Overall aspect ratios were determined from the mode of
intersection width/intersection length as well as the shape
of [010] sections which are parallel to the tablet faces. The
roundness value was determined by minimising errors in the
intersection width/intersection length and total phase abundance from the sum of the intersection areas and the CSD
(Higgins 2006). An overall shape of 1:3:4 and a roundness of
0.5 were used for all samples. It should be noted that plagioclase CSDs for some of our Kameni samples were originally
published in Higgins (1996b) but were not converted using
stereologically correct methods.
The complete CSD analyses described above were complemented by determinations of the total plagioclase content
in other samples. Thin sections were illuminated using circular cross-polarised light (Higgins 2010a) and the images
thresholded for plagioclase abundance.
Our results and analytically compatible earlier
analyses
Outcrop geology
The overall characteristics of the Kameni lavas were determined in the field. All lavas are phyric with evident plagioclase crystals distributed homogeneously throughout
the rock on the scale of an outcrop. Almost all outcrops
of lava contain mafic enclaves, but their abundance is very
variable within and between flows. In a typical outcrop of
several square metres, we observed one 3 cm enclave per
square meter, giving an abundance of ~ 0.07%. Parts of some
flows have greater concentrations, especially the 46–47 CE
flow on Palea Kameni, which has up to 5% enclaves. Most
enclaves are rounded and microgranular, except in the 1950
flow in which coarse-grained plagioclase-olivine cumulates
dominate.
Petrography and CL mineralogy
The petrography of the dacite lavas is much more variable
than their geochemistry (see later). The lavas have a finegrained groundmass, with crystals of plagioclase (2.4–22%)
up to 3 mm long. Clinopyroxene is much less abundant
(0.1–1.7%), followed by orthopyroxene (0.3–0.7%) and
magnetite (0.5–1.2%) (This study; Barton and Huijsmans
1986).The dacites contain several populations of plagioclase
crystals, as is common in many intermediate volcanic rocks.
Most plagioclase crystals are euhedral or subhedral with little zoning (Fig. 3a–d) and a uniform pale-green CL colour
(Fig. 2b). Pale spots within the crystals in the CL images are
melt inclusions, apatite crystals or artefacts due to charging (Fig. 2b). Rare plagioclase crystals have complex interior zones, with zoning and melt inclusions. These crystals
mostly occur as larger, independent crystals (Fig. 3e, f). The
abundance of these complex crystals is less than 2% of most
lava samples. They are considered to be xenocrysts and will
not be discussed further.
In the dacites, some of the plagioclase, pyroxene, and
oxide crystals occur in clusters of up to about 15 crystals,
commonly with an interstitial pale brown glass that is largely
crystal-free (Fig. 3a, b). In the interior of the clusters the
crystals are frequently small and interlocking, suggesting
that initially plagioclase, pyroxene, and oxide nucleated
nearby and were fused by initial crystal growth (Fig. 3c, d).
Towards the exterior, the crystals have euhedral forms indicating that further growth of both phases occurred simultaneously into the silicate liquid. The larger exterior crystals
strongly resemble independent crystals elsewhere in the section in shape, zoning and CL colour. The crystal clusters do
not resemble the enclaves, which have a different texture
(Martin et al. 2006).
Other important phases are orthopyroxene, clinopyroxene, oxides, and apatite, which also occur within the clusters
as well as independent crystals. As for the plagioclase, there
does not seem to be a great contrast between the forms of the
independent crystals and those in the clusters. Minor apatite
crystals are generally small and most are enclosed within
pyroxene crystals.
The petrography of the enclaves has been described in
detail in other studies and we do not want to repeat it (Martin
2005; Martin et al. 2006). All that we can add here is the plagioclase composition, as revealed by cathodoluminescence.
This phase has a distinctive cathodoluminescence pattern,
with a dark green interior and a pale green rim (Fig. 2) that
contrasts with the pale green uniform colour of the dacite
plagioclase. We do not see any crystals with similar CL colours in the dacite magma away from the enclaves.
Overall chemical variations
The freshness and lack of alteration of samples were examined using the amount of loss-on-ignition (LOI). All subaerial samples have low LOI and are considered fresh and
unaltered (appendix table 1). However, five of the samples
taken underwater have high LOI (5–25%). Three samples
are scoria and two are lava. If the analyses are recalculated
on a volatile-free basis then the scoria samples (EN 419–91,
-92, -93) do not resemble any of the volcanic rocks of the
Kameni islands: they may have been modified by precipitation of calcite and dolomite in the vesicles.
Many new samples have been analysed in this study, but
it behoves us to combine our data with earlier studies where
13
13
Page 6 of 20
Fig. 3 Thin sections in non-polarised (PL) and linearly cross-polarised illumination (XP). a, b Cluster of plagioclase, pyroxene and
magnetite crystals. The groundmass in the interstices is glassy, with
fewer microlites than away from the cluster centre. c, d Cluster of plagioclase, pyroxene and magnetite crystals. Plagioclase and pyroxene
crystals are narrow near the centre of the cluster, which is presumed
13
Contributions to Mineralogy and Petrology (2021) 176:13
to be near the initial nucleation site. There are vesicles close to the
crystals which are probably in contact away from the plane of the section. e, f A euhedral plagioclase crystal with a complex core and a
wide, simple rim. Small pyroxene and magnetite crystals were incorporated during the growth of the rim
Page 7 of 20 13
Contributions to Mineralogy and Petrology (2021) 176:13
possible. Most major element analyses of dacites from earlier studies fall in a similarly restricted field to our analyses.
However, there are significant differences, particularly in
Na, which can be difficult to analyse by XRF due to the low
energy of the characteristic X-rays and may also be affected
by diffusion within the glass beads. There are also problems with the reproducibility of some trace element analyses, probably caused by inadequate use of standards. This
is significant for the dacite analyses given the small range
in compositions. For this reason, we are unable to combine
early datasets and our discussion will be restricted to our
data alone. In contrast, the wide range in enclave compositions, and the relatively small number analysed in this study,
makes it essential to combine our data with that of other
studies.
Dacite samples have been divided up into three groups:
the earliest lavas from 46 to 726 CE, which are now exposed
only on Palea Kameni (first phase; PK); the post-paroxysmal
lavas from 1570 to 1925 CE (second phase; Nea Kameni);
and the most recent eruptions from 1939 to 1950 CE (Nea
Kameni). This last group is defined on the basis of its contrasting textural and chemical compositions, to be documented later.
The Total-Alkalis-Silica diagram is commonly used for
the classification of volcanic rocks and their components (Le
Maitre et al. 2002). All the lavas fall in a restricted group
in the dacite field with silica contents from 64.0 to 68.5%
(Fig. 4a, b). There is a strong correlation between silica
and total alkalis. The first phase (PK) dacites have generally higher SiO2 and alkali contents than the second phase
dacites, but there is not a clear separation. The third phase
dacites have the lowest silica contents. The enclaves have a
much wider range than the lava compositions and fall in the
andesite, basaltic andesite, and basalt fields. Most enclaves
fall along with a relatively narrow trend with just two outliers which are from the first phase eruptions.
The affinity of the underwater lava samples can be determined using the TAS diagram. One lava sample taken close
to the other NK North samples (EN 419–58) has an anhydrous composition similar to that of the other NK North
samples (Fig. 4). The anhydrous composition of the other
lava sample (EN 419–94) did not resemble those of the
Kameni Islands. It was from a ridge SW of the Kameni
islands and is hence not part of the same magmatic system.
All five samples were not considered further in this study.
Magnesium contents can inform us on the role of mafic
phases. MgO and SiO2 are strongly correlated for the dacite
lavas but with a small range in compositions (Fig. 4c). The
enclaves have a much larger range in compositions, but the
correlation is not so strong, especially for low SiO2 samples. The division of enclave compositions into two groups
for low silica contents is probably an artefact of the small
numbers of analyses.
Phosphorus abundances can inform us on the role of apatite. P2O5 and SiO2 are strongly correlated for the second and
third phase dacites (Fig. 4d). However, the first phase dacites
fall off this trend with lower P2O5 contents. These elements
are also strongly correlated in most of the enclaves, except
for some enclaves also from the first phase.
Zr is an incompatible element in the Kameni dacites as
they do not contain zircon. There is a very strong correlation between SiO2 and Zr for the dacites (Fig. 5a). Ba is
also weakly correlated with Zr (Fig. 5b). In both diagrams,
there is an overall trend to lower SiO2, Zr, and Ba with the
progression of the eruptions.
Incompatible element ratios are unchanged by fractional crystallisation of major minerals and can be used for
examining compositional variations in the source materials (Fig. 5c). The dacites show a slight difference in Th/
Ta between samples from the three phases. However, the
enclaves show a considerable variation in both Th/Ta and
Zr/Nb (Fig. 5c).
Rare-earth elements (REE) are very useful for understanding the role of plagioclase in magmatic systems due
to the anomalous behaviour of europium with respect to its
neighbouring REE. The easiest way to express REE variations is in terms of Eu/Eu* (Eu anomaly) versus LaN/LuN,
which reflects the overall slope of the chondrite normalised
REE spectrum (Fig. 5d). All of the dacite samples have a
small range in Eu/Eu* values (0.67–0.78) which are negatively correlated with LaN/LuN. There is a general trend to
smaller Eu anomalies and shallower REE patterns with the
progression of the eruptions. Enclaves have a much wider
range of values with a negative correlation between Eu/
Eu* and LaN/LuN. Almost all enclaves have Eu deficits (Eu/
Eu* < 1).
Secular chemical variations
Secular variation in the overall composition of dacite lavas
can be determined because the age of most flows is well
known. The only exception analysed here is the NK North
flow: its age is between 1848 and 1925 but has been arbitrarily placed at 1900 CE in Fig. 6a. The compositional
variation shown by dacites and enclaves in our new data
mirrors that of earlier studies (Huijsmans 1985; Martin
2005; Zellmer et al. 2000) and hence the datasets can be
considered together. SiO2 has been chosen as a key parameter for characterising the magmas in each flow group, as
the uncertainty in its determination is much less than the
variation in the data. The simplest way of illustrating this
data is with a linear graph of time versus composition
(Fig. 6a). The overall trend is clear: medium SiO2 contents
in the earliest flows rise in the 726 flows and then descends
slightly with much higher dispersion in the 1570–1950
eruptions. Although this graph preserves a linear time
13
Contributions to Mineralogy and Petrology (2021) 176:13
Page 8 of 20
(a)
12
10
K2O+Na2O (%)
7.4
Enclaves - Zellmer
Enclaves - Huijmanns
Enclaves - Marn
Enclaves - This study
46-726 Dacites -This study
1570-1925 Dacites - This study
1939-50 Dacites - This study
Underwater scoria - This study
8
(b)
Trachyte
92
7.2
Trachyandesite
K2O+Na2O (%)
13
93
91
58
6
7.0
6.8
4
Dacite
Andesite
6.6
Basaltic
Andesite
2
Basalt
Error bars at
centre of plot
Error bars at
centre of plot
6.4
0
45
50
55
60
65
64
70
65
66
SiO2 (%)
67
68
69
SiO2 (%)
(c)
(d)
8
0.2
P2O5 (%)
MgO (%)
6
4
0.1
2
Error bars at
centre of plot
Error bars at
centre of plot
0
0
45
50
55
60
65
70
SiO2 (%)
45
50
55
60
65
70
SiO2 (%)
Fig. 4 Major element compositional data for dacites (this study) and
micrographic enclaves (this study; Huijsmans 1985; Martin 2005;
Zellmer et al. 2000). Analyses of the underwater scoria samples have
been recalculated on a volatile-free basis. a Total-alkalis-silica diagram (Le Maitre et al. 2002). Underwater sample numbers are abbre-
viated: 91 = EN 419–91. They fall away from the main trend and are
hence assumed to have been severely altered. They will not be considered in further discussions. b Detail of TAS diagram for dacites
analysed in this study. c MgO versus SiO2. d P2O5 versus SiO2
component, variations in recent data are not very clearly
shown. Hence, we have plotted SiO2 content for each eruption (Fig. 6b). The overall pattern is similar for the older
flows, but this figure reveals better the variations in the
younger eruptions. Eruptions from 1570 to 1940 have a
similar, relatively wide range in SiO2 composition. However, lower SiO2 samples become more common with time
and dominate the 1950 eruption.
Enclave SiO2 contents have a wide range in values but
show groupings for certain eruptions, although we must
acknowledge the relatively small number of enclaves analysed from older events (Fig. 6c).
Plagioclase abundance variations
13
Plagioclase abundance in dacites was determined from the
sum of the area of crystal outlines and varies from 2.4 to
22%. The earliest flows have about 4–7% plagioclase crystals, which descends to its lowest value in the 726 CE eruption products (Fig. 7a). After that event, plagioclase abundance generally increases, reaching the highest values in the
1950 eruption. The plagioclase content is also related to the
SiO2 content of the magmas: most low SiO2 magmas have
medium to high plagioclase contents, whereas more silicic
magmas tend to have lower plagioclase contents (Fig. 7b).
Page 9 of 20 13
Contributions to Mineralogy and Petrology (2021) 176:13
(a) 70
(b)
450
400
Dacites
Dacites
65
350
Ba (ppm)
300
SiO2 (%)
60
Enclaves
250
200
55
Enclaves
150
46-726 Dacites
1570-1925 Dacites
1939-1950 Dacites
Enclaves
50
100
Analytical
error
50
Analytical error
45
0
0
50
100
150
200
0
250
50
100
150
200
250
Zr (ppm)
Zr (ppm)
(c)
(d)
1.0
Dacites
15
Enclaves
Eu/Eu*
Th/Ta
0.9
Enclaves
0.8
10
0.7
Analytical
error
5
5
10
15
Zr/Nb
Dacites
Analytical
error
0.6
1.5
2.0
2.5
3.0
3.5
LaN/LuN
Fig. 5 Trace element variations in enclave and dacite data from this
study. a SiO2 versus Zr. b Ba versus Zr. c Incompatible element
ratios. d Europium anomalies versus REE normalised slope. Analyti-
cal error estimates do not account for error correlations and are hence
probably overestimates in panels c and d
There is very little variation in Al2O3 content with plagioclase abundance, but the analytical error is large as compared to the overall variation (Fig. 7c). Although there is
much dispersion, plagioclase content is broadly correlated
with Eu/Eu* values, with a general increase with eruption
date (Fig. 7d).
by the smallest crystal outline that could be measured
(0.04 mm) and the accumulation of correction errors in for
the smallest size bins (Higgins 2000). Hence, the CSD is
undefined below 0.1 mm.
There is strong theoretical and experimental evidence that
CSDs in many simple systems are straight on a semi-logarithmic diagram (Cashman and Marsh 1988; Marsh 1988).
Curved CSDs can be produced in many situations (Higgins
2006), but the most popular involve mixing of two crystals
populations (Higgins 1996b) or crystallisation under two
different cooling regimes (e.g. Fornaciai et al. 2015). Such
CSD variations
All dacite CSDs are strongly curved, concave up (Fig. 8a,
electronic appendix 1). The lower size limit is controlled
13
13
Contributions to Mineralogy and Petrology (2021) 176:13
Page 10 of 20
SiO2 (%)
(a) Lava composition by date
68
67
66
65
Zellmer
Martin
64
0
69
Huijsmanns
This study
200
Analytical
error
400
600
(b) Lava composition by eruption
800
1000
Year
1200
1400
1600
1800
2000
(c) Enclave composition by eruption
Huijsmanns
Zellmer
Martin
This study
68
SiO2 (%)
60
SiO2 (%)
67
55
66
50
65
Huijsmanns
Zellmer
Analytical
error
45
64
46-7
726
1570-3 1707-10 1866-70
Eruption
NK 1925-8 1939-41 1950
North
Analytical
error
Francalanci
This study
46-7
726
1570-3 1707-10 1866-70
1925-8 1939-41 1950
Eruption
Fig. 6 a Variation of dacite SiO2 contents with eruption date. b Variation of dacite SiO2 contents for different eruptions. c Variation of enclave
SiO2 contents for all the eruptions (no data from NK North; this study; Huijsmans 1985; Martin 2005; Zellmer et al. 2000)
curved CSDs can be modelled by summing two straight
CSDs (Fig. 8B) and fitting the curve by trial and error to
the measured CSD (Higgins 1996a). One population has a
steep slope with many small crystals and is termed microcryst, whereas the other population has a shallow slope and
abundant large crystals and is termed macrocryst.
The slope of a CSD has units of 1/length, hence it is easier to characterise the distribution using the characteristic
length (CL) which equals − 1/slope of the fitted line and
has the units of length. For a straight CSD, it is equal to the
mean crystal size. The curvature of the CSD can then be
easily expressed on a graph of plagioclase abundance versus
characteristic lengths of the two components. It is then much
easier to compare curved CSDs from different samples using
this approach than simply to plot all the CSDs together on
the same diagram (Fig. 9).
The range of CL values of the microcryst population is
almost constant for all samples except those from the last
two eruptions (Fig. 9a). Similarly, the range in CL of the
macrocryst population is great for all samples except those
from the last two eruptions (Fig. 9a).
13
It is of interest to determine the relationship between the
CLs of the two model crystal populations and the repose
time between eruptions (Fig. 9b). This has been calculated
for the subaerial eruptions only, as we do not know the exact
age of all underwater eruptions. For the 46–7 CE eruption,
the predecessor is assumed to be the 197 BCE eruption,
whose products are no longer visible, but there may have
been other eruptions that never reached the surface, in which
case the repose time would be shorter.
The microcryst populations have small CLs for the short
repose times of the latest eruptions (1939–41, 1950), and
the range of CLs is greater for longer repose times of the
first and second eruptive phases. The macrocryst populations
have rather similar CLs for all repose times. However, the
four samples with greater CLs all have repose times greater
than 150 yrs.
The abundance of plagioclase in the two model populations can be compared with their chemical composition,
as was done in Fig. 7 for the total crystal population.
There is an overall negative correlation between silica
content and macrocryst plagioclase abundance (Fig. 9c)
Contributions to Mineralogy and Petrology (2021) 176:13
25
13
Page 11 of 20
25
(a)
1939 - 1950 CE
(b)
1570 - 1925 CE
46-726 CE
20
Plagioclase (vol %)
Plagioclase (vol %)
20
15
10
5
15
10
Closed system
crystallisaton
5
0
0
46-7
726
1570-3 1707-10 1866-70
Eruption
25
NK 1925-8 1939-41
North
1950
64
25
(c)
Crystal
addition
Analytical
error
65
66
67
SiO2 (wt %)
Crystal
fractionation
68
69
(d)
Analytical
error
20
Plagioclase (vol %)
Plagioclase (vol %)
20
15
10
5
Overall temporal trend
15
10
5
Analytical
error
0
1 5. 0
15.5
Al2O3 (%)
16.0
0
0.68
0.70
0.72
0.74
Eu/Eu*
0.76
0.78
0.80
Fig. 7 Plagioclase total volumetric abundances were determined from
textural data and have errors smaller than the symbol. a Plagioclase
abundance for different eruptions. b, c Plagioclase abundance versus
SiO2 and Al2O3 content. Vectors show variations for addition, crystallisation or fractionation of 5% plagioclase An50. d Plagioclase abundance versus Eu/Eu*
as well as a secular variation in both these parameters.
This correlation is not significant for the microcryst
population abundances (Fig. 9d). Similarly, there is no
significant correlation between alumina and macrocryst
abundance (Fig. 9e). However, there is a correlation
between europium anomaly and macrocryst abundance
(Fig. 9f).
13
Contributions to Mineralogy and Petrology (2021) 176:13
Page 12 of 20
10
10
(a) Representative CSDs
MH-TH-06 (726)
NK26 (1950)
NK08 (1866)
MH-TH-20 (1925)
8
6
(b) Modelling of a CSD
Microcryst
Population
8
-4
Population density
- (mm )
Fig. 8 a Representative plagioclase CSDs from the Kameni
dacites. Complete CSD data
is in electronic appendix 1.
b Kameni CSDs have been
modelled by the addition of two
straight CSDs (here for sample
NK25), a population of smaller
crystals termed microcrysts and
a population of larger crystals
termed macrocrysts, following
Higgins (1996a)
-4
Population density
- (mm )
13
4
2
0
-2
-4
6
4
2
Data
Sum of populations
0
-2
-4
-6
Macrocryst
Population
-6
0
0.5
1
Size (mm)
Discussion
1.5
2
0
0.5
1
Size (mm)
1.5
2
the production of magmatic diversity in the Kameni volcanic
system.
Existing models for the Kameni system
Origins of magmatic diversity
Many of the existing studies of the Kameni volcanic
rocks have tried to constrain the properties of the magmatic plumbing system that fed the Kameni eruptions.
The earliest models envisaged that plagioclase and other
phases crystallised throughout the magma reservoir and
descended by gravity towards the lower part of the reservoir, creating a stratified reservoir (Fig. 10) (Barton and
Huijsmans 1986). Eruptions during the last 2000 years
sampled successively lower levels of this reservoir, with
increasing plagioclase contents. Such a model requires
that the temperature remain constant and hence that the
magma reservoir must have been large, or more likely,
that it was frequently reheated by the influx of new mafic
magma. However, the main weakness of this model is that
such additions of heat would necessarily produce convection, which would destroy the stratification. In addition,
the high viscosity of the dacite magma would inhibit the
gravitational transfer of crystals.
Other models have been proposed in subsequent studies. Higgins (1996b) proposed that there were two magma
reservoirs. Eruptions were rooted in the upper one which
was topped up from a deeper reservoir before each event.
Francalanci et al. (1998) and Petrone et al. (2013) proposed
that magmatic diversity of the dacite and the enclaves were
produced by mingling and mixing between the dacite and
mafic magmas that are injected into the dacite. The work of
Martin et al. (2006) was mostly concerned with the origin
of the enclaves and not of the dacites themselves. All these
models were based on the study of either chemistry or texture: in this study, we use these complementary datasets to
modify existing models and propose a new mechanism for
13
Enclaves
Enclaves are a very minor, but widespread component of the
Kameni lavas and it is important to determine their role, if
any, in the development of magmatic diversity. The broad
collinearity between the dispersion of dacite compositions
and those of the enclaves has suggested to some authors that
magmatic diversity was produced by mixing between primordial dacite and mafic magmas (Francalanci et al. 1998).
Such a model would be expected to produce a continuous
spectrum of compositions, from the most mafic enclave to
the most silicic dacite. We have combined data from all
available sources and show in Fig. 4 that this is clearly not
the case and there is a conspicuous gap between the dacite
and enclave compositions. The enclave compositions do not
lie on a linear trend, hence at least three end-member components are necessary for an origin by mixing.
Petrone et al. (2013) suggested that there was a field and
petrographic evidence for disaggregation (‘crumbling’) of
enclaves in the Kameni lavas. Our cathodoluminescence
images are a sensitive test for such a process. As we mentioned before, plagioclase crystals in enclaves are generally
small and have a distinctive cathodoluminescence pattern,
with a dark green interior and a pale green rim (Fig. 2).
There are no crystals with this pattern in the dacite. Indeed,
there is a distinct paucity of crystals of this size in the dacite.
Hence, if the enclaves have been disaggregated, then the
crystals must have been completely resorbed. Enclaves from
different eruptions have, in some cases, distinctive compositions that contrast with that of earlier or later eruptions
Contributions to Mineralogy and Petrology (2021) 176:13
(Fig. 6c). This suggests that enclaves have a relatively short
life in the magma reservoir as is suggested by experimental
studies (Ruprecht et al. 2020). All these arguments indicate
that although enclaves are an important indicator of mafic
input to the base of the magma reservoir, they have not influenced the composition of the dacites in a significant way.
Even if the enclaves did not modify significantly the
composition of the dacites, the mafic magma that produced
them may have had another important role in the magmatic
development of the system. Most of the enclaves are vesicular, showing that vapour saturation was achieved by cooling
after the mixing event. At this time the crystal network was
sufficiently strong that the enclave could not expand, so most
of the exsolved gas must have escaped into the surrounding
magma, where some bubbles may have become attached
crystals in the mush (Pleše et al. 2018). This idea will be
developed later.
Crystal clusters in the dacites
The degree of clustering of crystals in volcanic rocks is very
variable, but its petrological significance has rarely been
explored. In the Kameni dacites, many crystals of plagioclase, pyroxene, and magnetite are clustered and there does
not appear to be any clear mineralogical differences between
clustered and independent crystals (Fig. 3a, b). Crystals
within the cores of the larger clusters are generally smaller
than crystals in the exterior of the cluster and independent
crystals. In some cases, narrow crystals in the core widen out
in their more peripheral parts (Fig. 3c, d). This suggests that
the crystals initially nucleated and grew near to each other,
in an environment where all major phases could nucleate
and grow. The simplest explanation is that crystallisation
occurred in a peripheral region, perhaps at the base of a
magma reservoir, where the density of the clusters would
reduce the likelihood of dispersion by simple magma convection. However, it is clear that clusters have been disrupted
and dispersed, indicating the need for special events, such
as the input of new mafic magma into a semi-consolidated
crystal mass.
The preservation of glass between the crystals in the
interior of the clusters suggests that the clusters have been
cooled more rapidly than the surrounding magma. One
possible mechanism is by attachment of bubbles from the
enclaves to crystals in the basal part of the reservoir. The
resultant buoyance of some crystal–bubble couples would
have disrupted the mush and enabled rapid uplift of crystal
clusters. We will return to this model below.
Nature of the crystal cargo
We can now discuss the origin of the remarkable variability in the abundance of crystals in the dacites and if it can
Page 13 of 20
13
inform us about magmatic processes active in the Kameni
volcanic system. Crystals in lavas can be phenocrysts, which
grew in place from their surrounding magma, antecrysts,
which grew in a distant, but the magmatically connected
environment, or xenocrysts, which come from a separate
system. Isolated xenocrysts can be recognised by their complex zoning (Fig. 3e, f), but they are not abundant and will
not be considered further. Sub-volcanic magmatic systems
are generally complex, with transfer and mixing between
different magmatic reservoirs in each of which crystallisation can occur and here the problem is to choose the simplest system that can account for the observed chemical and
textural variations.
We will first consider the possibility that the whole
crystal cargo (microcrysts and macrocrysts) is dominantly
antecrysts, incorporated in the magma by mixing of a crystal-rich material with a sparsely-phyric magma. The most
abundant phase in the dacites is plagioclase, hence here we
model the effects for plagioclase of composition An50—the
inclusion of other phases seen in the clusters does not make
a significant difference to the modelling (Barton and Huijsmans 1986). The addition of plagioclase to a single magma
produces a simple diagonal variation in a plot of plagioclase
abundance versus SiO2 (Fig. 7b). Our data show a band of
variations, which necessitates multiple original magma compositions. Most of the dispersion can be accounted for by
the addition of 4–20% plagioclase to magmas with original
SiO2 contents of 66–68%. However, such a model is unable
to explain the variation in Al2O3 with plagioclase abundance
(Fig. 7c), hence the simple, pure antecryst model is rejected.
We will discuss later the possibility that only the larger crystals (macrocrysts) were added.
Growth of plagioclase phenocrysts in the Kameni dacites is very probable, but the high viscosity of the dacite
magma and the small contrast in density makes it unlikely
that significant gravity-driven fractionation could occur by
compaction or settling. Involvement of other, denser phases,
as suggested by the occurrence of polymineralic clusters in
the dacite lavas, would give a greater density, but probably
not enough to make a significant difference. In this case,
the overall composition of the magma would be unchanged
by crystallisation. This model would produce a vertical dispersion on the plot of plagioclase abundance versus chemical composition (Fig. 7b–d). However, there are significant
chemical variations and hence this simple model must be
rejected. One possible modification would be crystallisation
from multiple original magma compositions. In this model,
the earliest lavas would have formed from the most evolved
magmas (highest SiO2 and lowest Al2O3) and have shown
the least amount of crystallisation. Later lavas would have
formed from progressively less evolved magmas and crystallised more, culminating in the high-crystallinity 1939–1950
magmas. The observed Eu/Eu* values (Eu anomaly) are
13
13
Contributions to Mineralogy and Petrology (2021) 176:13
Page 14 of 20
(b) Repose times
(a) Eruptions
1950
800
1939-41
700
1925-8
Repose time (yr)
600
NK
North
1866-70
1707-10
500
400
300
1570-3
200
726
100
46-7
Microcryst population
0
0.1
Macrocryst population
0
0.2
0.3
0.4
Characteristic Length (mm)
(c) Macrocrysts
1570-1928 CE
8
1939-1950 CE
7
6
Crystal
addition
5
Closed system
crystallisaton
4
Plagioclase Microcrysts (vol %)
Plagioclase Macrocrysts (vol %)
0.2
0.3
0.4
Characteristic Length (mm)
0.5
10
46-726 CE
9
9
8
7
6
5
4
3
3
2
2
1
64
1
Crystal
fractionation
Analytical
error
0
65
66
67
SiO2 (%)
68
64
69
(f)
Analytical
error
8
7
6
5
4
3
66
67
SiO2 (%)
68
69
10
8
7
6
5
4
3
2
2
1
1
0
15.0
65
9
Plagioclase Macrocrysts (vol %)
9
Analytical
error
0
(e) 10
13
0.1
(d) Microcrysts
10
Plagioclase Macrocrysts (vol %)
0
0.5
15.5
Al2O3 (%)
16.0
0
0.68
Analytical
error
0.70
0.72
0.74
Eu/Eu*
0.76
0.78
0.80
Contributions to Mineralogy and Petrology (2021) 176:13
◂Fig. 9 a CLs of microcryst and macrocryst model plagioclase popu-
lations versus eruption. b CLs of the two populations versus repose
time. c Silica versus macrocryst model abundance. The vectors are
for plagioclase of composition An50. d Silica versus microcryst model
abundance. e Alumina versus macrocryst model abundance f Europium anomaly versus macrocryst model abundance
consistent with such a model: the earliest, most fractionated magmas have the smallest values (Fig. 7d). Hence, a
model is preferred in which the crystal cargo is essentially
dominated by phenocrysts that grew within the main magma
reservoir. However, we cannot rule out the possibility that
some crystals are antecrysts that were added from connected
reservoirs.
Textural constraints on magmatic processes
The curved plagioclase CSDs observed in the Kameni dacites were first interpreted as evidence of mixing of two different magmas (Higgins 1996b). However, the uniformity
of the plagioclase macrocryst populations and the lack of
chemical evidence for simple mixing weakens this model.
Hence, we favour here an origin for the curved CSDs by
crystallisation under two different pressure and temperature regimes, represented by the characteristics of the two
straight CSDs used in the modelling.
The microcryst population is dominated by the smallest
crystals and represents a growth event immediately before
the eruption. The steep slope indicates that growth occurred
at significant undercooling, presumably in a shallow magmatic reservoir. The characteristic length of this population
tends to be smaller for the latest eruptions (Fig. 9a). There
does not appear to be a correlation between characteristic
length and repose time (Fig. 9b).
The macrocryst population reflects an earlier phase of
plagioclase growth. Most samples have similar characteristic lengths for different flows (Fig. 9a), except for four
samples with much greater characteristic lengths. These all
have lower plagioclase contents and most erupted during the
PK phase. Characteristic length is broadly correlated with
repose time, except for two samples from the 1570 CE flow
which have the longest repose times (Fig. 9b).
CSDs with greater characteristic lengths can be produced
by textural coarsening, in which small crystals dissolve and
large crystals grow (Higgins 2010b). Coarsening develops
when a magma is maintained for some time close to the
mineral saturation temperature and may be accelerated if
the temperature oscillates around this value (e.g. Mills and
Glazner 2013). The infrequency of eruptions during the PK
phase may have led to a stable environment that permitted
coarsening. The lack of small, recently nucleated, crystals in
the glassy parts of the clusters (Fig. 3a, b) suggests that the
clusters must have coarsened, presumably at the base of a
Page 15 of 20
13
deep magma reservoir. These crystal-free glasses were then
preserved by rapid uplift.
Characteristic crystal growth times can be determined if
the growth rate of plagioclase can be established (Marsh
1988). Cashman (1993) established a correlation between
growth rate and cooling time for plagioclase in mafic
magmas and a similar relationship is generally assumed
for more felsic magmas. For the microcryst population a
cooling time of 3 years, which accords with seismic data
from the 2011–12 magma injection event, would suggest a
growth rate of about 10–9 mm/s. Such a value would indicate
characteristic growth times of 1.4–3.5 years for the microcryst population. The shortest times are associated with the
1939–41 and 1950 flows, which have the shortest repose
times and highest plagioclase contents. The macrocryst population may have formed in a deeper reservoir and hence the
growth rate may have been lower: Higgins (1996b) used a
growth rate of 10–10 mm/s, which would give characteristic
crystal growth times of 60–160 years. This broadly accords
with repose times of 150–250 years for three of the four
coarsened populations (Fig. 9b).
Pertinent temporal constraints on magmatic processes
in the Kameni systems have been determined using other
methods. Ra/Th ratios in Kameni rocks suggest that crystallisation of plagioclase occurred less than about 1000 years
before eruption (Zellmer et al. 2000). Trace element profiles
in plagioclase suggest a residence time of 100–450 years
(Zellmer et al. 1999). In contrast, Martin et al. (2008) used
trace element profiles in an olivine crystal from a cumulate
enclave in the 1925–8 lavas to conclude that mafic magma
was added only a few months before the eruption. These data
will be used to constraint a magmatic model.
The total macrocryst population can be compared to various chemical parameters as was done earlier for the whole
crystal population (Figs. 7, 9). As before, it is not possible
to generate the observed dispersion only by the addition of
plagioclase to a single liquid composition. However, here
the model fits better on the graph of Al2O3 versus macrocryst abundance (Fig. 9e) than on the graph of SiO2 versus
macrocryst abundance (Fig. 9c). Again, the data support a
model of closed-system crystallisation from liquids of variable composition.
Magmatic development by bubble‑mediated liquid
and crystal movements
The chemical and textural data presented above suggest that
Kameni magmas were formed by the closed-system crystallisation of variable amounts of plagioclase, and lesser
amounts of other phases, from a range of dacite liquids,
but how was this achieved? The presence of crystal clusters
shows that much of the plagioclase macrocrysts formed in
an environment with a high crystal content. The density of
13
13
Contributions to Mineralogy and Petrology (2021) 176:13
Page 16 of 20
(a) Stratified magma chamber
(b) Withdrawal from top
(c) Continued withdrawal
Lava dome
Lava dome
Surface
Low crystallinity magma
High crystallinity magma
Scale change
Scale change
Scale change
Fig. 10 Schematic model of a static, stratified magma reservoir following the work of Barton and Huijsmans (1986). a A stratified
magma reservoir is created by crystallisation of plagioclase and other
phases followed by accumulation in the lower part of the reservoir. b
The first magma to be withdrawn is from the upper, low crystallinity
part of the reservoir. c The magma reservoir is emptied and successively higher crystallinity magmas are withdrawn
plagioclase is close to that of the magma, hence this zone
cannot have formed by gravity-driven accumulation: instead
crystallisation must have occurred in situ, probably largely
at the base of the reservoir. The degree of connectedness
of the individual crystals was probably very variable. This
high-crystallinity material was subsequently disaggregated
into clusters and individual crystals and incorporated into
the dacite liquid. Simple recharge of the reservoir with mafic
magma cannot lift enclaves, otherwise we would see abundant enclaves in all dacites, which is not the case. Hence, it
is proposed that the buoyant forces necessary to disrupt the
basal crystal mass and lift the material to the top of the reservoir were produced by the physical association of crystals
with vapour bubbles. Such a process has been proposed for
the origin of crystal-bearing vesicle pipes (e.g. Helz 1987),
but not for crystals and crystal clusters in a magma reservoir.
Interactions between bubbles and crystals at the base
of a magma reservoir have been explored in several recent
papers. Mungall (2015) modelled mathematically the problem of non-wetting bubble migration through a basal accumulation of crystals and concluded that bubbles would be
trapped in a cumulate. Boudreau (2016) used an analogue
experiment to show exactly the opposite—that non-wetting
bubbles can escape from cumulates easily. However, both
studies may not be completely applicable to natural systems
as Pleše et al. (2019; 2018) showed that bubbles nucleate
and grow on plagioclase and pyroxene crystals and hence
that these surfaces are wetted. If the bubble–crystal couples
stay together long enough then they will rise buoyantly in
the reservoir through the dacite magma. There, there may
be sufficient time for the bubbles to detach, which is why
we do not see many bubbles on the crystals now (Pleše et al.
2019; 2018). This model can be used to construct a model
for magmatic development in the Kameni system.
We propose that magmatic development started in a
deep dynamic magma reservoir, or a nexus of magma storage areas filled with low-crystallinity dacite magma, where
two different scenarios may have developed, here termed
juvenile and mature (Fig. 11). Paradoxically, the juvenile
scenario appears to be applicable to the later magmas, and
the mature model to the earliest magmas.
The juvenile model is invoked for low to high crystallinity, less chemically evolved magmas mostly produced after
the 726 CE paroxysmal event. Crystallisation occurred at the
base of the reservoir, producing a loose mass of crystals that
were cemented in places by further growth (Fig. 11a). Mafic
magma was injected into the base of the reservoir where it
was cooled by the dacite and became saturated with vapour.
Vapour pockets developed in the semi-crystalline enclaves
and grew until bubbles were forced out of the enclaves into
the dacite magma. Here, some of the bubbles attached to
silicate crystals. Buoyance of the crystal–bubble couples disrupted the crystal mush. The couples then rose rapidly and
accumulated at the top of the reservoir, where the interstitial liquid in the crystal clusters solidified as glass. Thermal
convective movements induced by the mafic input may have
contributed to the uplift process. Subsequent eruptions may
have sampled both the high crystallinity magmas at the top
of the reservoir or the lower crystallinity magma beneath
this layer.
In the mature model, crystals of plagioclase, pyroxene
and oxides, nucleated and grew at the base of the reservoir until the crystals impinge on each other forming a rigid
network (Fig. 11b). This crystal network may have formed
13
Contributions to Mineralogy and Petrology (2021) 176:13
(a) Juvenile system
Lava dome
Page 17 of 20
(b) Mature system
High crystallinity magma
Upper reservoir
13
Lava dome
Low crystallinity magma
Upper reservoir
Scale change
Scale change
Evolved magma
Primitive magma
Loose crystal mush
Mafic magma injected into base of lower reservoir
Connected
crystal network
Mafic magma injected into base of lower reservoir
Fig. 11 Schematic dynamic magma reservoir models for producing magmas of variable chemical composition and crystallinity.
The lower and upper components are not all at the same scale and
are shown for a single, simplified magma storage and transfer system. In both the juvenile (a) and mature models (b) the process starts
with the creation of a reservoir filled with dacite magma. Cooling
and crystallisation produce a crystal mush at the base of the reservoir. This mush is initially made of loose crystals (a) but with time
growth will fill in interstitial space to make a semi-rigid connected
network (b). In both models mafic magma is injected into the base
of the reservoir, causing vesiculation. In the juvenile system (a),
the loose mush is disrupted by bubbles out-gassed from the mafic
magma. Some bubbles will adhere to the crystals in the dacite and
make crystal-bubble couples that rise in the magma reservoir. Bubbles may detach when the couples are stored on the roof. Accumulation of crystals near the roof will produce a high-crystallinity magma
that is relatively primitive as the original liquid composition is almost
unchanged. For the mature system (b), the crystal network acts as a
rigid filter and the bubbles will force out evolved magma that was
originally between the crystals into the magma reservoir. Minor parts
of the mush may be mobilised in the same way as in the juvenile system and rise to the top of the reservoir, forming a chemically evolved
low-crystallinity magma. In both juvenile and mature systems, the
magma in the upper part of the reservoir does not necessarily erupt
directly but instead may feed an upper chamber from where it may be
stored or erupted
because the host rocks were relatively cool at the start of the
eruptive cycle, allowing for more growth. However, CSDs
suggest that some crystal-bearing regions were coarsened,
which would have maintained channels for the subsequent
escape of interstitial magma (Higgins 2006, p. 67). The
interstitial silicate liquid was isolated and evolved towards
more silicic compositions by fractional crystallisation. As
in the juvenile system, injection of mafic magma into the
base of the reservoir produced enclaves with vapour pockets.
Again, bubbles were transferred to the dacite crystal network
where they attached to the silicate phases. However, buoyant
forces could not overcome the strength of the network, so the
vapour pocket instead displaced the interstitial fractionated
liquid which mixed with the dacite liquid in the reservoir.
This evolved liquid may have picked up some loose crystals
from the top of the pile that rose to the top of the reservoir
to produce a low-crystallinity, chemically evolved magma.
The presence of gabbroic cumulate enclaves in the 1950
CE flow suggests that bubble-mediated crystal floatation
may have removed all crystals that formed from the dacite
magma and excavated down to the underlying mafic unit.
Here, slow cooling would have coarsened plagioclase and
olivine that originally crystallised from the mafic magma.
Bubbles could have wafted these materials into the magma
forming what we see now as gabbroic enclaves.
In both these models, magma may have been withdrawn
from the top of the lower reservoir directly or transferred
to a sub-volcanic reservoir. The latter seems to have been
the case for the 2011–2012 CE intrusive event (Newman
et al. 2012; Parks et al. 2012). The trigger for such a transfer
may have been mafic magma injection in the lower reservoir
(Rizzo et al. 2015).
Chemical and textural data can be used to constrain the
timing of the magmatic models. The macrocryst population
characteristic growth times, Ra/Th ratios (Zellmer et al.
2000) and plagioclase trace element profiles (Zellmer et al.
1999) all suggest that plagioclase grew in the lower magma
reservoir in less than 1000 years (Fig. 11). The microcryst
population characteristic growth times suggest storage in the
sub-volcanic upper reservoir for up to a few years, which fits
with seismic evidence from the 2011–12 CE event (Newman
et al. 2012). The olivine trace element profile indicates a
short transition time for an enclave between the base of the
lower reservoir and the surface (Martin et al. 2008). Perhaps
this reflects different movements of enclaves and plagioclase
macrocrysts in the lower reservoir. It may also confirm that
13
13
Page 18 of 20
storage times in the upper reservoir are short-less than a
few months.
Conclusions
Contributions to Mineralogy and Petrology (2021) 176:13
for their help with access to the islands, particularly Mr Thanasis who
landed us in difficult spots. Comments of anonymous reviewers considerably improved the manuscript.
Funding MDH-Natural Sciences and Engineering Research Council
(Canada); JVA-Fonds de la Recherche Scientifique (Belgium).
Data availability Data will be provided in a data repository.
The Kameni volcanic centre has produced dacitic magmas
that show a limited range in chemical composition, but a
huge range in textures, clearly expressed in the abundance of
plagioclase crystals. We propose that the observed chemical
and textural diversity was produced by interactions between
basal crystal masses and bubbles. Two extreme situations
can be identified: Juvenile and mature systems.
In a juvenile system, the crystallisation of plagioclase and
other phases at the base of a lower magma reservoir creates
a loose network of crystals. In some areas, growth will link
crystals to form small connected clumps. Injection of mafic
magma into the base of the reservoir produces enclaves that
are saturated in vapour. Expansion of the gas pockets in the
enclaves will lead to the escape of bubbles into the dacite
and their attachment to silicate crystals. Buoyant forces then
disrupt the loose crystal mass and bubble–crystal couples
rise towards the top of the reservoir where they accumulate
to be tapped periodically to feed eruptions. This process
produces a less chemically evolved magma, with a crystal
content that can range to high values.
In the mature system, more protracted crystallisation of
plagioclase and other phases at the base of a magma reservoir creates a strongly-connected network of crystals. Again,
vapour pockets are produced in enclaves of mafic magma
and bubbles escape to adhere to the silicate phases in the
crystal network of the reservoir. However, in this model, the
bubbles displace the evolved interstitial liquid which mixes
into the body of the magma reservoir. The ejected magma
may also pick up a small number of bubble–crystal couples
from the top of the mush zone which rise in the reservoir.
This process produces a crystal-poor but chemically evolved
magma at the top of the reservoir.
In both situations, there may have been periodic withdrawal of magma from the top of the reservoir to a high-level
staging volume, from which the magma may have erupted
periodically to produce the diverse magmatic compositions
observed in the Kameni lavas.
The juvenile system was more important for the more
recent eruptions, whereas the mature system seems to have
dominated the oldest eruptions. This may reflect the generally decreasing repose times of the system, which allowed
for the development of an interconnected mush zone at the
base of the magma reservoir.
Acknowledgements We would like to thank Steve Carey for supplying
the underwater samples. Pantelia Sorotou gave us tremendous help on
the islands. We would like to thank the Santorini Union of Boatman
13
Compliance with ethical standards
Conflict of interest None.
References
Barton M, Huijsmans JPP (1986) Post-caldera dacites from the Santorini volcanic complex, Aegean sea, Greece: an example of the
eruption of lavas of near-constant composition over a 2,200 year
period. Contrib Miner Petrol 94:472–495. https://doi.org/10.1007/
BF00376340
Boudreau A (2016) Bubble migration in a compacting crystal-liquid
mush. Contrib Miner Petrol 171(4):32. https://doi.org/10.1007/
s00410-016-1237-9
Cashman KV (1993) Relationship between plagioclase crystallisation
and cooling rate in basaltic melts. Contrib Miner Petrol 113:126–
142. https://doi.org/10.1007/BF00320836
Cashman KV, Marsh BD (1988) Crystal size distribution (CSD) in
rocks and the kinetics and dynamics of crystallisation II. Makaopuhi lava lake. Contrib Miner Petrol 99:292–305. https://doi.
org/10.1007/BF00375363
Druitt TH, Edwards L, Mellors RM, Pyle D, Sparks RSJ, Lanphere M,
Davies M, Barreirio B (1999a) Santorini volcano. In: Memoir vol.
19. Geological Society, London, p 176
Druitt TH, McCoy FW, Vougioukalakis GE (2019a) The Late Bronze
Age eruption of Santorini volcano and Its impact on the ancient
Mediterranean world. Elements 15(3):185–190. https ://doi.
org/10.2138/gselements.15.3.185
Druitt TH, Pyle DM, Mather TA (2019b) Santorini volcano and
its plumbing system. Elements 15(3):177–184. https ://doi.
org/10.2138/gselements.15.3.177
Fornaciai A, Perinelli C, Armienti P, Favalli M (2015) Crystal size
distributions of plagioclase in lavas from the July–August 2001
Mount Etna eruption. Bull Volcanol 77(8):1–15. https ://doi.
org/10.1007/s00445-015-0953-8
Francalanci L, Vougioukalakis G, Eleftheriadis G, Pinarelli L, Petrone
C, Manetti P, Christofides G (1998) Petrographic, chemical and
isotope variations in the intracaldera post-Minoan rocks of the
Santorini volcanic field, Greece. In: Casale R (ed) Proceedings of
the second workshop, Santorini, Greece, pp 175–186
Götze J, Schertl H-P, Neuser RD, Kempe U, Hanchar JM (2012) Optical microscope-cathodoluminescence (OM–CL) imaging as a
powerful tool to reveal internal textures of minerals. Mineral Petrol 107(3):373–392. https://doi.org/10.1007/s00710-012-0256-0
Helz RT (1987) Differentiation behavior of Kilauea Iki lava lake,
Kilauea Volcano, Hawaii; an overview of past and current work.
Magmatic processes; physicochemical principles; a volume in
honor of Hatten S Yoder, Jr, vol 1. Geochemical Society, Dayton,
Ohio, pp 241–258
Higgins MD (1996a) Crystal size distributions and other quantitative textural measurements in lavas and tuff from Mt Taranaki
(Egmont volcano), New Zealand. Bull Volcanol 58:194–204
Contributions to Mineralogy and Petrology (2021) 176:13
Higgins MD (1996b) Magma dynamics beneath Kameni volcano,
Thera, Greece, as revealed by crystal size and shape measurements. J Volcanol Geoth Res 70(1–2):37–48. https ://doi.
org/10.1016/0377-0273(95)00045-3
Higgins MD (2000) Measurement of crystal size distributions. Am
Mineral 85(9):1105–1116. https://doi.org/10.2138/am-2000-8-901
Higgins MD (2006) Quantitative textural measurements in igneous
and metamorphic petrology. Cambridge University Press, Cambridge, UK
Higgins MD (2010a) Imaging birefringent minerals without extinction
using circularly polarized light. Can Mineral 48(1):231–235. https
://doi.org/10.3749/canmin.48.1.231
Higgins MD (2010b) Textural coarsening in igneous rocks. Intern
Geol Rev 53(3–4):354–376. https ://doi.org/10.1080/00206
814.2010.496177
Higgins MD, Voos S, Vander Auwera J (2015) Magmatic processes
under Quizapu volcano, Chile, identified from geochemical and
textural studies. Contrib Miner Petrol 170(5–6):1–16. https://doi.
org/10.1007/s00410-015-1209-5
Hooft EEE, Heath BA, Toomey DR, Paulatto M, Papazachos CB,
Nomikou P, Morgan JV, Warner MR (2019) Seismic imaging of
Santorini: subsurface constraints on caldera collapse and presentday magma recharge. Earth Planet Sci Lett 514:48–61. https://doi.
org/10.1016/j.epsl.2019.02.033
Huijsmans JPP (1985) Calc-alkaline lavas from the volcanic complex
of Santorini, Aegean Sea, Greece: a petrological, geochemical and
stratigraphic study. Geologica Ultraiectina 41:1–316
Le Maitre RW, Streckeisen A, Zanettin B, Le Bas MJ, Bonin B, Bateman P, Bellieni G, Dudek A, Efremova S, Keller J, Lameyre J,
Sabine PA, Schmid R, Sorensen H, Woolley AR (2002) Igneous
rocks : a classification and glossary of terms: recommendations
of the International Union of Geological Sciences subcommission on the systematics of igneous rocks. Cambridge University
Press, Cambridge
Marsh BD (1988) Crystal size distribution (CSD) in rocks and the
kinetics and dynamics of crystallization I. Theory. Contrib Miner
Pet 99:277–291. https://doi.org/10.1007/BF00375362
Martin VM (2005) Geochemical and textural analysis of mafic enclaves
from Nea Kameni, Santorini, Greece. PhD Thesis, University of
Cambridge
Martin VM, Holness MB, Pyle DM (2006) Textural analysis of magmatic enclaves from the Kameni Islands, Santorini, Greece. J
Volcanol Geoth Res 154(1–2):89–102. https://doi.org/10.1016/j.
jvolgeores.2005.09.021
Martin VM, Morgan DJ, Jerram DA, Caddick MJ, Prior DJ, Davidson JP (2008) Bang! month-scale eruption triggering at Santorini
volcano. Science 321(5893):1178. https://doi.org/10.1126/scien
ce.1159584
McVey BG, Hooft EEE, Heath BA, Toomey DR, Paulatto M, Morgan
JV, Nomikou P, Papazachos CB (2019) Magma accumulation
beneath Santorini volcano, Greece, from P-wave tomography.
Geology 48(3):231–235. https://doi.org/10.1130/g47127.1
Mills RD, Glazner AF (2013) Experimental study on the effects of
temperature cycling on coarsening of plagioclase and olivine in
an alkali basalt. Contrib Miner Petrol 166(1):97–111. https://doi.
org/10.1007/s00410-013-0867-4
Mora CI, Ramseyer K (1992) Cathodoluminescence of coexisting plagioclases, boehls butte anorthosite-Cl activators and fluid-flow
paths. Am Mineral 77(11–12):1258–1265
Mungall JE (2015) Physical Controls of Nucleation, Growth and
Migration of Vapor Bubbles in Partially Molten Cumulates. In:
Charlier B, Namur O, Latypov R, Tegner C (eds) Layered Intrusions. Springer, Dordrecht, pp 331–377
Newman AV, Stiros S, Feng L, Psimoulis S, Moschas S, Saltogianni V,
Jiang Y, Papazachos C, Karaginni E, Vamvakaris D (2012) Recent
Page 19 of 20
13
geodetic unrest at Santorini Caldera, Greece. Geophys Res Lett.
https://doi.org/10.1029/2012gl051286
Nicholls IA (1971) Petrology of Santorini Volcano, Cyclades. Greece
J Petrol 12(1):67–119. https://doi.org/10.1093/petrology/12.1.67
Nomikou P, Parks MM, Papanikolaou D, Pyle DM, Mather TA,
Carey S, Watts AB, Paulatto M, Kalnins ML, Livanos I, Bejelou K, Simou E, Perros I (2014) The emergence and growth of
a submarine volcano: The Kameni islands, Santorini (Greece).
GeoResJ 1-2(Supplement C):8-18 doi:https://doi.org/10.1016/j.
grj.2014.02.002
Pagel M, Barbin V, Blanc P, Ohnenstetter D (2000) Cathodoluminescence in geosciences. Springer, Berlin
Parks MM, Biggs J, England P, Mather TA, Nomikou P, Palamartchouk
K, Papanikolaou X, Paradissis D, Parsons B, Pyle DM, Raptakis
C, Zacharis V (2012) Evolution of Santorini volcano dominated
by episodic and rapid fluxes of melt from depth. Nat Geosci
5(10):749–754. https://doi.org/10.1038/ngeo1562
Parks MM, Moore JDP, Papanikolaou X, Biggs J, Mather TA, Pyle
DM, Raptakis C, Paradissis D, Hooper A, Parsons B, Nomikou P
(2015) From quiescence to unrest: 20 years of satellite geodetic
measurements at Santorini volcano, Greece. J Geophys Res Solid
Earth 120(2):1309–1328. https://doi.org/10.1002/2014JB011540
Petrone C, Francalanci L, Vougioukalakis G (2013) Mixing, mingling
and enclave crumbling in the post-Minoan dacitic magmas of
Santorini volcano, Greece. In: Goldschmidt 2013 conference
proceedings
Pleše P, Higgins MD, Baker DR, Kudrna Prašek M (2019) Nucleation
and growth of bubbles on plagioclase crystals during experimental
decompression degassing of andesitic melts. J Volcanol Geoth Res
388:106679. https://doi.org/10.1016/j.jvolgeores.2019.106679
Pleše P, Higgins MD, Mancini L, Lanzafame G, Brun F, Fife JL, Casselman J, Baker DR (2018) Dynamic observations of vesiculation reveal the role of silicate crystals in bubble nucleation and
growth in andesitic magmas. Lithos 296–299:532–546. https://
doi.org/10.1016/j.lithos.2017.11.024
Pyle DM (2017) What Can We Learn from Records of Past Eruptions to Better Prepare for the Future? In: Fearnley CJ, Bird DK,
Haynes K, McGuire WJ, Jolly G (eds) Observing the volcano
world. Springer, Heidelberg, pp 1–18
Rasband WS (2010) ImageJ. U. S. National Institutes of Health,
Bethesda, Maryland, USA http://rsb.info.nih.gov/ij/
Rizzo AL, Barberi F, Carapezza ML, Di Piazza A, Francalanci L,
Sortino F, D’Alessandro W (2015) New mafic magma refilling
a quiescent volcano: evidence from He-Ne-Ar isotopes during
the 2011–2012 unrest at Santorini, Greece. Geochem Geophys
Geosyst 16(3):798–814. https://doi.org/10.1002/2014gc005653
Ruprecht P, Simon AC, Fiege A (2020) The Survival of Mafic Magmatic Enclaves and the Timing of Magma Recharge. Geophysical
Research Letters 47(14):e2020GL087186 doi:https://doi.org/https
://doi.org/10.1029/2020GL087186
Sigurdsson H, Carey S, Alexandri M, Vougioukalakis G, Croff K,
Roman C, Sakellariou D, Anagnostou C, Rousakis G, Ioakim
C, Goguo A, Ballas D, Misaridis T, Nomikou P (2006) Marine
investigations of Greece’s Santorini volcanic field. Eos, Transactions American Geophysical Union 87(34):337–342. https://doi.
org/10.1029/2006eo340001
Stamatelopoulou-Seymour K, Vlassopoulos D, Pearce TH, Rice C
(1990) The record of magma chamber processes in plagioclase
phenocrysts at Thera volcano, Aegean volcanic arc, Greece. Contrib Miner Petrol 104:73–84. https://doi.org/10.1007/BF00310647
Theodorakopoulou K, Kyriakopoulos K, Athanassas CD, Galanopoulos
E, Economou G, Maniatis Y, Godelitsas A, Dotsika E, Mavridis F,
Darlas A (2020) First Speleothem Evidence of the Hiera Eruption
(197 BC), Santorini, Greece. Environmental Archaeology:1-13
doi:https://doi.org/10.1080/14614103.2020.1755196
13
13
Page 20 of 20
Vougioukalakis GE, Fytikas M (2005) Volcanic hazards in the Aegean
area, relative risk evaluation, monitoring and present state of the
active volcanic centers. In: Fytikas M, Vougioukalakis GE (eds)
The South Aegean active volcanic arc-present knowledge and
future perspectives, milos conferences, vol 7. Elsevier, Greece,
pp 161–183
Watts AB, Nomikou P, Moore JDP, Parks MM, Alexandri M (2015)
Historical bathymetric charts and the evolution of Santorini submarine volcano. Greece Geochem Geophys Geosyst 16(3):847–
869. https://doi.org/10.1002/2014gc005679
Zellmer G, Turner S, Hawkesworth C (2000) Timescales of destructive
plate margin magmatism; new insights from Santorini, Aegean
volcanic arc. Earth Planet Sci Lett 174(3–4):265–281. https://doi.
org/10.1016/S0012-821X(99)00266-6
13
Contributions to Mineralogy and Petrology (2021) 176:13
Zellmer GF, Blake S, Vance D, Hawkesworth C, Turner S (1999)
Plagioclase residence times at two island arc volcanoes (Kameni
Islands, Santorini, and Soufriere, St. Vincent) determined by Sr
diffusion systematics. Contributions to Mineralogy and Petrology
136(4):345-357 doi:https://doi.org/10.1007/s004100050543
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.