Full text of DOI:10.1093/petrology/42.9.1643

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
PAGES 1643–1683
2001
Petrology and Geochemistry of the
Lamongan Volcanic Field, East Java,
Indonesia: Primitive Sunda Arc Magmas
in an Extensional Tectonic Setting?
S. A. CARN∗ AND D. M. PYLE
DEPARTMENT OF EARTH SCIENCES, CAMBRIDGE UNIVERSITY, DOWNING STREET, CAMBRIDGE CB2 3EQ, UK
RECEIVED SEPTEMBER 30, 1999; REVISED TYPESCRIPT ACCEPTED FEBRUARY 26, 2001
New geochemical data are presented from prehistoric and historical multifarious process, involving possible inputs from the
eruptive products of the Lamongan volcanic field (LVF), East Java; subducted oceanic lithosphere, subducted oceanic sedi-
2
a region of the Sunda arc covering >260 km and containing ments and fluids, the asthenospheric and lithospheric
>
90 eruptive vents plus the historically active Lamongan volcano. portions of the mantle wedge above the subduction zone,
LVF lavas include medium-K basalts and basaltic andesites from and the arc crust (e.g. Tatsumi & Eggins, 1995). Following
historical eruptions of Lamongan and prehistoric eruptions in the melt generation, processes such as crystal fractionation,
eastern LVF, along with a high-K suite represented by prehistoric accumulation and crustal assimilation during transit to
deposits in the western LVF. Although lacking some of the char- the surface may obscure the true nature of the magma
acteristics of truly primary basalts, the least evolved lavas identified source region. The effects of these processes on the
in the LVF have some of the lowest SiO contents (>43 wt % compositions of erupted lavas must therefore be removed
2
SiO ) yet reported in Sunda arc volcanic rocks. Mass balance before models of arc magma genesis can be tested.
2
considerations indicate that two chemically distinct LVF magmas
For this reason it is desirable to sample magmas that
may be parental to suites currently being erupted from the neighbouring have suffered minimal modification since segregation
volcanoes, Semeru and Bromo. Lamongan ’s historical lavas can be from their source region. The compositions of the primary
related to the medium-K andesitic products of Semeru by fractional or primitive magmas from different arcs provide a window
crystallization, despite the former ’s location at the same distance on the diverse mechanisms of, and the components
from the trench as Bromo, a high-K volcano. Extensional tectonics, involved in, melt production at convergent margins,
possibly related to arc segmentation in the region of the LVF, creating and as such are essential to our understanding of the
conditions that promote the rapid ascent of parental magmas, is subduction zone environment.
probably responsible for this and several other features of the complex.
Many studies have highlighted the correlation between
extensional tectonics in volcanic arcs and the eruption
of lavas with primitive chemical characteristics (e.g. Luhr
&
Carmichael, 1981; Knittel & Oles, 1995; Bacon et al.,
KEY WORDS: geochemistry; primitive magmas; Sunda arc; volcanic field;
extensional tectonics
1
997; Smith et al., 1997). In such settings, the development
of thinned and fractured crust allows rising magma to
erupt at the surface in a relatively unmodified state,
rather than ponding at the crust–mantle boundary. The
rapid passage to the surface of individual magma batches
inhibits the formation of long-lived crustal magma cham-
INTRODUCTION
The generation and subsequent evolution of magma in bers, and invariably leads to the formation of monogenetic
subduction zone settings is widely acknowledged to be a vent fields comprising tens to thousands of individual
∗
Corresponding author. Present address: JCET, University of Mary-
land, Baltimore, MD 21250, USA. E-mail: Simon@jcet.umbc.edu
 Oxford University Press 2001

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
eruptive centres (such as cinder cones and maars) in these Along much of the arc Indian Ocean crust is subducting
environments [e.g. the Michoac a´ n–Guanajuato Volcanic beneath the Eurasian plate at a rate of >6 cm/yr (Ham-
Field, Mexico (Luhr, 1997) and the Macolod Corridor, ilton, 1979), with the downgoing slab probably continuous
Philippines (Knittel & Oles, 1995)].
from the surface to the lower mantle beneath Java
The Sunda arc in Indonesia is a mature island arc (Puspito & Shimazaki, 1995; Widiyantoro & van der
with several structural features indicative of an active Hilst, 1996, 1997). This continuous slab is compatible
extensional regime at certain points along its length, such with a regional subduction history that dates back to the
as developing back-arc basins (e.g. the Madura Basin, Permian (Katili, 1975).
East Java; Hamilton, 1979; Fig. 1). Furthermore, the
The Indian Ocean crust ages eastwards along the arc
present seismic strain field beneath the volcanic arc on (Fig. 1), being around 130–135 Ma old south of Java
Java is indicative of extension oriented at a high angle (Widiyantoro & Hilst, 1996). In the eastern Sunda arc,
to the arc, a feature common to many convergent margins the incursion of continental crust and crustal fragments
(
Apperson, 1991). However, there have been few docu- on the Indo-Australian plate complicates the system (Fig.
mented studies of monogenetic vent fields on the Sunda 1). East of Flores the progressive docking of Australian
arc, which, as discussed above, are also symptomatic of continental lithosphere onto the Banda arc since the
intra-arc extension.
Neogene (e.g. Rangin et al., 1990) has led to a cessation
This paper presents the results of a petrological and of volcanism at the arc front. This arc–continent collision
geochemical investigation of the Lamongan volcanic field is also believed to be responsible for incipient back-arc
(
LVF) in East Java, which encompasses the historically thrusting (from Bali eastwards; Fig. 1) and extensional
active Lamongan volcano along with numerous cinder features on the arc (e.g. Silver et al., 1983; Charlton,
cones and maars (Carn, 2000). The surrounding region 1991).
contains at least five recently active volcanoes (Fig. 1).
The Sunda arc has existed in roughly its current
The LVF sample suite covers most of the exposed vents configuration for around 5 m.y., although an active vol-
and lava flows, and allows the characterization of the canic arc has probably persisted in the region since at
most primitive magmas in the complex. Modelling of
fractional crystallization is used to place these lavas in a
local and regional context. The data are also used to
make preliminary judgements on the mechanisms of
magma genesis in this part of the Sunda arc and on
the possible effects of the regional tectonic regime on
volcanism in the area.
least the Eocene (Katili, 1975; Hamilton, 1979; Rangin
et al., 1990). The Quaternary volcanoes generally overlie
Late Miocene–Pliocene volcanic rocks and sediments
(Soeria-Atmadja et al., 1994). The changing nature of the
arc crust, from continental in Sumatra to transitional in
Java and oceanic in Flores, gives rise to along-arc variation
in the style and composition of volcanism (Whitford et
al., 1979). Silicic volcanism and large calderas (e.g. Lake
Toba) on Sumatra give way to mafic–intermediate vol-
canism on Java, where the crust is of near-continental
thickness but comprises m e´ langes and mafic–intermediate
plutons. From Bali to Sumbawa (Fig. 1) a mature oceanic
island arc has developed (Hamilton, 1988). Across-arc
OVERVIEW OF LAMONGAN
VOLCANO AND THE LVF
Tectonic and geological setting
Lamongan volcano (8·00°S, 113·34°E; altitude 1651 m) geochemical variations, primarily in K and Sr/ Sr,
lies on the Sunda arc between the massifs of Tengger– have also been documented (Whitford & Nicholls, 1976;
Semeru and Iyang–Argapura, around 200 km above the Whitford et al., 1979).
Wadati–Benioff Zone (Fig. 1). Its location corresponds
87
86
to one of the narrowest sections of the island of Java,
and graben structures in the vicinity of the LVF suggest
local extension (Carn, 1999). The region has been little
Lamongan volcano and the LVF
studied, and prior petrological studies are only cursory Lamongan volcano is the only historically active vent in
2
(
Mulyana, 1989). The occurrence of three seismic crises the LVF, a region covering some 260 km occupied by
in the area during the 20th century, each entailing numerous prehistoric eruptive centres including around
the formation of east–west- and ENE–WSW-oriented 61 cinder cones and 29 maars (Fig. 2). Two prehistoric
tensional fractures on Lamongan’s western flanks vents, Gunung Tarub and Gunung Tjupu, complete the
(
Matahelumual, 1990), led to the establishment of an central edifice (Lamongan–Tarub–Tjupu or LTT) of
observatory to monitor the volcano.
which Lamongan forms the southwestern segment (Figs
The Sunda arc extends from the Andaman Islands 2 and 3). Carn (2000) presented an analysis of the physical
north of Sumatra to the island of Alor in the Banda Sea aspects of the LVF, and his compilation of volumetric
over a distance of >3000 km, and incorporates 76% of data for the cinder cones, maars and lava flows of the
Indonesia’s Holocene volcanoes (Simkin & Siebert, 1994). complex is given in Table 1. The number of vents in the
1644

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1645

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
cones (Fig. 2) indicating a sustained eruption of magma
following maar formation.
The sample suite collected from the LVF is rep-
resentative of most of the eruptive units and monogenetic
vents in the complex (Figs 2 and 3). Samples were
categorized as: (1) historical lavas (erupted from La-
mongan volcano, 1799–1898); (2) cinder or spatter cones
Table 1: Summary of
juvenile magma volume
estimates for LVF cones,
lava flows and maars;
from Carn (2000)
[
including lava flows clearly associated with cones
3
Sample size
Volume (km )
(mapped in Fig. 2) and cones within maars]; (3) prehistoric
lava flows (flows not covered by the two preceding
categories); (4) maar deposits (Table 2). No particular
genetic relationships between samples in each group
are implied by this classification. Samples were further
subdivided on the basis of vent morphology and location,
as detailed in a later section. The majority of LVF
lavas are basalts and basaltic andesites, with subordinate
picrobasalts, basanites, trachybasalts and trachyandesites
Cinder or spatter cones
Maars
36
22
11
1
1·50
0·21
Prehistoric lava flows
Lamongan–Tarub–Tjupu
Lamongan eruptions (1799–1898)
Total
0·11
10·14
0·049
10
>12
(
Table 2).
Samples from cinder or spatter cones were invariably
LVF is exceptional for an Indonesian volcano, exceeding scoriae or spindle bombs found close to the surface.
the 35 scoria cones reported from Gunung Slamet in Dense vegetation in the SE of the region precluded
central Java (Vukadinovic & Sutawidjaja, 1995), although detailed exploration, but there is little evidence for young
it is somewhat fewer than that observed in larger cinder vents in this area (Bronto et al., 1986). Maar deposits
cone fields elsewhere (e.g. Mexico; Luhr, 1997). Vent were poorly preserved and juvenile material was difficult
distributions in the LVF suggest a strong regional tectonic to identify. The provenance of maar samples is hence
influence on prehistoric volcanism in the field, which largely unknown; some material may have been reworked
may have involved fissure-style eruptions (Carn, 1999, by the disruption of subsurface lava flows. For com-
2000).
parison, several new analyses from volcanoes adjacent to
Lamongan’s first recorded historical activity occurred the LVF (Semeru and Bromo) are also presented (Fig.
in 1799. This heralded >100 years of unrest during 1; Table 2).
which the volcano erupted up to 15 lava flows, the last
in 1898 (Matahelumual, 1990; Said, 1992; Simkin &
Siebert, 1994). The steady-state eruptive rate at La-
mongan from 1843 to 1898 was >0·03 m /s, and the
3
ANALYTICAL TECHNIQUES
age of the volcano has been estimated at 13–40 ka, Whole-rock samples were analysed for major and trace
assuming a constant time-averaged eruptive rate over its elements by X-ray fluorescence (XRF) at the Open
lifetime (Carn, 2000). With the exception of seismic University, Milton Keynes, on an ARL 8420+ wave-
crises on the volcano’s flanks in 1925, 1978, 1985 and length-dispersive XRF spectrometer. Fresh cores of lavas
1988–1989 (Matahelumual, 1990), Lamongan has been were washed in deionized water and dried before crushing
dormant since 1898.
in an agate grinder. Pressed powder pellets for trace
element determination were prepared in Cambridge, and
fused glass discs for major element analyses were prepared
in Milton Keynes. Data quality was monitored by running
in-house igneous rock standards with each sample batch.
Trace element detection limits are given in Table 2.
Mineral analyses were performed on carbon-coated pol-
FIELD OBSERVATIONS AND
SAMPLING
The LVF is situated >8° south of the equator and much ished slides using a Cameca SX-50 electron microprobe
of the field is well vegetated. All the observed lava flows in Cambridge. Analytical conditions were a 10 kV ac-
in the LVF (historical and prehistoric) are of a’a, and celerating voltage, beam currents of 4–15 nA and a beam
are several metres thick and up to 2–3 km long (Fig. 3). diameter of 5 ꢀm; all analyses were energy dispersive.
Cross-sections of individual prehistoric flows, >10–30 m Precision, monitored using various standards, is generally
thick, were found in outlying stream sections and in the within 1–2% for major elements. To determine Sr isotopic
vertical walls of maar craters although they often proved ratios, >150 mg aliquots of rock powder were dissolved
difficult to trace beyond these isolated exposures. Several in HF–HNO and the resulting solution was evaporated
3
of the dry maars were observed to contain small cinder until dry. After further dissolution in HNO the solution
3
1646

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
Fig. 2. Map of the LVF showing the monogenetic vents and lava flows, and the distribution of samples used in this study. Cinder cones are
classified as either ‘young’ (KVM series) or ‘old’ (KVT series) according to the morphological criteria of Carn (2000). Dashed box indicates
the location of Fig. 3. Contour interval is 100 m, except for supplementary contours on some cinder cones. Contours on the east edge of the
map mark the western flanks of the Iyang–Argapura massif (Fig. 1). Derived partly from Bronto et al. (1986).
was evaporated again; the samples were then redissolved variable amounts of glass (often devitrified). The pet-
in HCl. After centrifuging, samples were loaded onto rography of the lavas is generally rather uniform through-
cation-exchange resin and Sr was eluted using HCl. The out the series. Modal analyses (point-counted) for 10
dried residue was loaded in HCl onto an outgassed historical lavas are listed in Table 3.
Ta filament and analysed on a VG Sector 54 mass
spectrometer in Cambridge. Ratios were corrected for accounting for up to 70% of the modal mineralogy
Subhedral Plag is the predominant phase in all samples,
8
6
88
mass fractionation by normalizing to Sr/ Sr = 0·1194. (Table 3). It typically occurs as complex aggregates or
Repeat analyses of NBS 987 gave a mean value of glomerocrysts (up to 3–4 mm square; associated with Cpx
0
·710263.
and Mag), and often exhibits fine-scale concentric zoning,
sieve texture and resorption features (e.g. embayed rims),
and zoned inclusions suggesting a complex history. Mi-
crophenocrysts often show similar features. Inclusions of
pyroxene and Mag are often observed in Plag. Ground-
mass alignments of Plag are rarely observed and are never
well developed. Phenocrysts range up to 3–4 mm in size.
Cpx also occurs in aggregates (with Plag, Fe–Ti oxides
PETROGRAPHY
Historical lavas
Historical lavas from Lamongan volcano typically contain
phenocrysts of plagioclase (Plag), clinopyroxene (Cpx), and Ol ), although less frequently than Plag. It is typically
olivine (Ol ) and Ti-magnetite (Mag) in a crystal-rich, subhedral, concentrically zoned, and occasionally displays
seriate-textured groundmass with similar mineralogy and twin lamellae; phenocrysts are up to >2 mm in size.
1647

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Fig. 3. Map of the flanks of Lamongan volcano showing the lava flow field emplaced during 19th-century activity (from Carn, 2000). The
provenance of samples used in this study is indicated. Eruption dates are reasonably well constrained from historical records (e.g. Matahelumual,
1
990), although these are less reliable for the older flows.
Embayed and sieve-textured Cpx (and rare skeletal grains)
Also observed in several samples (e.g. LAM18981,
is also present in some samples. Inclusions of Plag are LAM18831, LAM1070) are clots or nodules containing
rare, but zones of Fe–Ti oxide inclusions are frequent; Plag, Cpx, Mag and occasional Ol, up to 0·8 mm in
a common texture observed in the historical lavas (and size. The pyroxene in these clots is generally unzoned
in other samples; see below) is the mantling of Cpx and often shows a subophitic or ophitic texture,
phenocrysts by small Fe–Ti oxide crystals. Low-Ca pyr- distinguishing them from the ubiquitous glomerocrysts
oxene is present only as a minor phase in the more that have similar constituents. Several nodules show
evolved lavas (e.g. LAM1561). It occurs as resorbed disequilibrium features such as sieve-texture and re-
crystal cores or as rims enveloping Cpx phenocrysts.
sorbed boundaries. These nodules are interpreted as
Ol is typically less abundant than Plag and Cpx in the cumulate material derived from the edges of an evolving
historical lavas (Table 3). It occurs as phenocrysts (up to magma chamber.
1
mm across) and microphenocrysts (generally <0·5 mm
across) that are frequently embayed, and it is also present lavas (LAM18982, LAM18832; Table 3), suggesting that
in the groundmass of several samples. a flotation mechanism (e.g. Campbell et al., 1978) may
Plag glomerocrysts are less common in the younger
1648

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1649

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
1650

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1651

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
1652

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1653

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
1654

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1655

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
1656

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1657

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Table 3: Modal proportions (vol. %, vesicle-free) of glomerocryst (Gl), phenocryst (Ph) and groundmass
Gm) phases in selected LVF rocks (point-counted using 700-1000 points per sample)
(
Sample
Plag %
Gl
Olivine %
Pyroxene %
Oxides Glass
Others Crystals
%
%
%
%
Ph
Gm
Total
Ph
Gm
Total
Ph
Gm
Total
LAM18982
LAM18832
LAM18641
LAM1070
LAM1261
LAM1562
LAM1564
LAM2962
LAM7070
LAM1762
ALA1161
GCK2960
GKL256
9·3
9·9
22·2
17·6
15·8
19·4
11·3
17·4
12·6
14·4
18·0
21·3
8·6
38·5
27·8
32·7
22·5
32·0
22·9
26·1
27·1
25·0
22·4
28·0
25·5
23·8
70·0
55·4
68·5
66·4
63·5
63·1
62·5
58·5
66·1
69·1
36·6
61·9
55·1
3·1
2·8
2·0
6·0
6·5
5·3
3·9
4·4
8·5
5·0
1·1
6·7
8·1
5·0
3·9
2·4
0·9
0·8
0·8
3·7
1·9
2·5
0·5
3·6
0·7
2·4
8·1
6·6
4·5
6·9
7·3
6·1
7·7
6·3
11·0
5·5
4·7
7·5
10·6
8·1
0
8·1
4·7
11·9
13·8
9·3
8·1
7·3
6·8
5·8
9·4
9·8
7·8
7·7
8·7
13·7
0
1·9
2·5
0·2
0
100
4·7
11·5
2·1
3·7
1·4
2·2
5·8
0·6
3·5
4·8
6·5
8·0
0
13·8
5·8
86·2
20·0
24·5
20·2
22·9
23·9
17·1
23·1
25·4
0
0
11·5
12·3
13·6
12·1
9·2
94·2
93·7
91·6
88·1
84·7
88·3
98·0
99·7
65·4
98·9
99·6
10·2
9·9
6·3
8·4
0
10·6
7·0
11·9
15·3
11·7
2·0
0
0
8·1
14·0
11·2
17·3
16·3
20·9
20·1
0
10·6
13·8
11·5
14·4
12·1
0
0·3
0
34·6
1·1
0
14·1
9·7
22·4
21·5
0
0·4
0
have controlled their distribution in the chamber and and microphenocrysts (<1·5 mm long) are usually em-
biased their occurrence to earlier-erupted flows.
bayed and occasionally skeletal. Plag is abundant and
sometimes occurs as glomerocrysts, but less commonly
than observed in the historical samples (Table 3), and
disequilibrium textures in Plag are also present, but less
widespread in the KVM samples. Plag phenocrysts are
typically a maximum of >1·5 mm in length. Fe–Ti
oxides are abundant, particularly in the groundmass;
phenocrysts are up to 2 mm in size. Cumulate clots of
Plag, Cpx and Fe–Ti oxides, similar to those discerned in
Cinder or spatter cones
Samples from the LVF cinder or spatter cones display a
wide range of petrographic features. Following Carn
(
2000), the cones are subdivided morphologically into a
young series (centuries old) and an old series (millennia
old; Fig. 2). The former is referred to as the kerucut the historical lavas, are also evident in the KVM lava
volkanik muda (KVM) series and the latter as the kerucut flow samples. There is a general trend of decreasing
volkanik tua (KVT) series. Cinder cones observed within phenocryst dimensions with distance from the central
maars belong to the KVM series. The KVT cones are (LTT) edifice.
generally situated in the western part of the LVF (Fig.
Other KVM samples are basaltic scoriae or small
2). The modal mineralogy of three KVM samples is bombs, and hence contain more groundmass glass than
given in Table 3.
the associated lava flows (e.g. ALA1161, Table 3). The
The KVM lavas contain phenocrysts of Plag, Cpx, scoriae have the same mineralogy as the lavas (with the
Fe–Ti oxides and Ol. Very rare crystals of oxy-hornblende exception of the two samples that contain minor oxy-
are also present in two scoria samples (ALA1161 and hornblende), although phenocrysts are generally smaller
GCK2961). Many of the KVM cones are spatter ramparts in the former; typically <1 mm with rare Cpx up to
associated with flank lava flows (Fig. 2), and samples >5 mm and Plag up to 3–4 mm in length. Other textural
from these flows (e.g. GCK2960, GKL256) are similar features are similar, such as the mantling of Cpx phe-
in appearance and modal mineralogy to the historical nocrysts by Fe–Ti oxide inclusions, the presence of cu-
lavas (Table 3). Large Cpx phenocrysts (up to >7 mm mulate nodules (maximum 5 mm diameter; largely Cpx
long, often euhedral) are common in the KVM lava and Fe–Ti oxides), and some resorptional features in Ol,
flows, typically with inclusions of Fe–Ti oxides, often Cpx and Plag crystals. Glomerocrysts are typically absent
concentrated at crystal rims (as observed in the historical or rare and disequilibrium textures are less common than
lavas). Some Cpx shows resorption features (e.g. em- in the historical samples.
bayments, eroded rims, sieve texture) and exsolution
KVT samples are petrographically diverse but some
lamellae are common. Ol phenocrysts (up to 6 mm long) spatial patterns can be discerned. Cones in the western
1658

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
LVF, west of the town of Klakah (Fig. 2), are particularly LAO306, LAO1207, TAR2760; Figs 2 and 3). Their
distinct and represent some of the oldest vents in the petrography exhibits many similarities with the historical
complex (Carn, 2000). Samples from these cones are samples from Lamongan.
generally glassy and contain Plag, Ol and minor Cpx in
order of decreasing abundance. Plag typically exists as
microlites (normally aligned) or microphenocrysts, often
zoned and occasionally showing disequilibrium textures
in larger crystals, which may have fresh mantles. Glo-
merocrysts of Plag up to 3–4 mm in diameter are also
MINERALOGY
The major phenocryst phases in LVF lavas are Ol (Fig. 4),
Cpx (Fig. 4), Plag (Fig. 5) and Fe–Ti oxides. Representative
mineral analyses are presented in Tables 4–6, and a
summary of observed mineral compositions is given in
Table 7. The majority of Fe–Ti oxide crystals analysed
observed in some samples, and individual phenocrysts
range up to >3 mm in length. Ol occurs as micropheno-
crysts (up to >1·5 mm in size), which are frequently
fall on the magnetite–ulv o¨ spinel solid solution (titano-
embayed or skeletal. Cpx appears rare in these samples
magnetite), with TiO contents as high as 17 wt % but
typically around 7–10 wt %.
2
and is generally restricted to the groundmass.
Other KVT samples form localized groups with dis-
tinctive petrography, although they are generally akin to
the KVM lavas. Textural similarities are observed be-
tween samples from collinear vents (e.g. GMI15072 and Olivine
GKN3060; both highly Plag-phyric).
Forsterite (Fo) contents of Ol phenocrysts and micro-
phenocrysts in LVF lavas range from Fo29 to Fo , with
79
considerable variation often observed on the scale of
individual samples (Table 7). This variation is generally
greater in the historical lavas. Many crystals display
normal zoning (Table 4). The more Fo-rich crystals are
generally found in cinder cone samples, and Fo contents
are typically highest in glomerocrysts and crystals found
in cumulate clots or nodules, with microphenocrysts
typically more iron rich. Disequilibrium textures in Ol
are seen in a range of compositions (Table 4). CaO
contents rarely exceed >0·3 wt %, which is a typical
value for basaltic olivines ( Jurewicz & Watson, 1988)
and higher than that found in mantle olivines (<0·1
wt %; e.g. Foden, 1983).
Prehistoric lavas
Prehistoric lava flows (PLFs) are distributed throughout
the LVF although they are rarely well exposed. Samples
from outlying flows (PLF1; e.g. TAR1361, KAP2062,
RBI1162; Fig. 2) are typically Plag-phyric with a glassy
matrix. Cpx and Ol are present as microphenocrysts
(
>1 mm) although Ol is often unstable. Plag crystals
are usually flow-aligned and show rare disequilibrium
textures. Some clots (>1 mm diameter) of Cpx, Fe–Ti
oxides and Plag are also present, and the groundmass is
often rich in Fe–Ti oxides.
The other PLF samples form two distinct groups. One
set (PLF2) crops out around much of the flank region of
the LTT edifice, usually in the walls of maar craters as
thick flows (e.g. RWG116, RJN278, RAG1962,
RLG1262; Figs 2 and 3); these lavas are highly por-
phyritic, containing phenocrysts of Ol, Plag and Cpx in a
There is widespread evidence for Ol accumulation in
LVF lavas. Ol phenocrysts and microphenocrysts are
typically not in equilibrium with their host rock, with Fo
contents generally being too low for their respective
bulk-rock mg-number; this is particularly evident in the
historical lavas and PLFs (Fig. 6). This may be due to
coarse groundmass (same mineralogy, plus Fe–Ti oxides). elevation of the measured Fe O /FeO (wt %) ratio by
2
3
Ol is typically anhedral and unstable (skeletal in places), post-eruption alteration and oxidation, which would
occurring as microphenocrysts and phenocrysts up to result in artificially high mg-numbers. However, most
>
3 mm in length and occasionally in nodules (with lavas were not visibly altered, with the exception of some
variable amounts of Fe–Ti oxides, Cpx and Plag) several PLF2 samples that show iddingsite rims on Ol phenocrysts
millimetres in diameter. Plag often occurs as glomerocrysts (e.g. RAG1962), which may explain the more extensive
(
3–4 mm in size), and frequently displays sieve-texture disequilibrium in these samples (Fig. 6). Assuming an
and fine-scale compositional zoning, the latter sometimes Fe O /FeO ratio of 0·2 still places the majority of his-
2
3
mantling the former. Cpx phenocrysts (3–6 mm) and torical samples and PLFs in disequilibrium, although the
glomerocrysts (up to 1 cm long) are less abundant, some- cinder cones (largely KVM) plot close to the high-
times euhedral, but generally unstable. Some contain pressure equilibrium field (Fig. 6). Cinder cone Ol also
inclusions of Ol, Plag and Fe–Ti oxides (concentrated at shows the narrowest range of Fo contents, particularly
the edges of the host crystal).
sample ALA1161 (KVM), whereas the PLF2 samples
The other group (PLF3) crops out at a slightly higher (RWG116, RAG1962; Fig. 6) show the clearest evidence
stratigraphic level in flows with seldom pristine mor- for Ol accumulation. Mean Fo contents in Ol are 63%
phologies, largely on the flanks of Lamongan (e.g. (historical lavas), 74% (cinder cones) and 62% (PLFs).
1659

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Table 4: Representative olivine analyses
Cinder cones
PLFs
Historic lava flows
Sample: GCK1071
GKL-
56
GGI2661
GOO13071
LAO296
RA-1
LAM- LAM18981
1561
LAM1070
2
Xtal:
Phxt
Phxt
Rim
Mphxt Phxt
Phxt
Rim
Phxt
Core
Emb
Phxt
Rim
Emb
Phxt
Core
St
Phxt
Glxt
Mphxt Mphxt Mphxt
Mphxt Mphxt
Core/rim: Core
Texture:
Core
Res
Core
Rim
Res
Core
Core
Core
Rim
Core
St
Rim
SiO
2
38·84 38·72 38·42 38·55 37·86
18·95 20·60 22·45 20·97 23·70
36·59 36·80
31·14 29·41
37·88 36·55 38·81
25·34 32·24 21·02
35·81 37·41 36·34
35·98 25·65 31·27
35·75 36·00
35·74 35·30
FeO
MnO
MgO
CaO
Total
0·36
41·08 39·90 38·43 39·92 36·98
0·22 0·20 0·23 0·23 0·30
99·46 99·78 99·98 99·96 99·35
0·36
0·46
0·29
0·51
0·56
30·32 31·85
0·28 0·29
98·88 98·89
0·53
0·53 0·91
36·09 29·85 39·43
0·26 0·29 0·14
100·1 99·86 99·80
0·39
0·92
26·81 35·29 30·74
0·20 0·26 0·29
99·72 99·21 99·32
0·61
0·69
0·88
27·42 27·80
0·24 0·23
100·04 100·13
0·80
Fo
Fa
79·4
20·6
77·5
22·5
75·3
24·7
77·2
22·8
73·5
26·5
63·4
36·6
65·9
34·1
71·7 62·3
28·3 37·7
77·0
23·0
57·1
42·9
71·0
29·0
63·7
36·3
57·8
42·2
58·4
41·6
Oxides in wt %. Glxt, Phxt, Mphxt, glomerocryst, phenocryst, microphenocryst, respectively. Textures: Res, resorbed; Emb,
embayed; St, sieve texture. Fo, Fa, molar percentages of forsterite and fayalite, respectively.
Table 5: Representative pyroxene analyses
Cinder cones
PLFs
RA-1
Historic lava flows
Sample:
Xtal:
GCK1071 GKL256
GGI2661 ALA1161
LAM1561
LAM18981 LAM2961
Glxt
Mphxt
Core
Mphxt
Rim
Mphxt
Core
Mphxt
Core
Phxt
Core
Phxt
Rim
Glxt
Core
St
Glxt
Rim
Phxt
Core
Mphxt
Core
Core/rim: Core
Texture:
SiO
TiO
Al
Cr
FeO
2
46·96
1·12
47·66
1·27
48·14
1·28
45·61
1·36
48·96
0·99
48·62
0·63
48·10
0·94
51·53
51·48
50·48
0·64
52·37
0·32
2
0·37
2·83
0·46
2·34
2
O
3
7·14
6·98
5·47
8·63
4·86
6·70
7·00
2·93
3·23
2
O
3
0·02
0·06
0·09
0·06
0·07
0·47
0·26
0·08
0·09
0·07
0·09
7·90
7·61
9·62
8·45
7·77
6·40
6·72
20·46
0·73
12·71
0·62
9·18
22·45
0·84
MnO
MgO
CaO
0·11
0·14
0·28
0·10
0·13
0·15
0·09
0·28
12·40
23·46
0·39
12·77
23·16
0·35
12·28
22·30
0·41
12·16
22·96
0·45
14·03
22·87
0·35
13·95
23·02
0·26
13·31
22·76
0·31
22·24
1·78
15·07
16·96
0·48
14·15
21·02
0·42
14·06
5·44
Na
2
O
0·13
0·81
Total
99·51
100·00
99·87
99·79
100·03
100·21
99·49
100·2
100·2
99·16
99·61
Mg
Fe
36·8
13·2
50·0
37·9
12·7
49·4
36·4
16·0
47·5
36·4
14·2
49·4
40·3
12·5
47·2
40·9
10·5
48·5
39·8
11·3
48·9
63·5
32·8
3·7
43·8
20·7
35·4
41·1
15·0
43·9
46·0
41·2
12·8
Ca
Oxides in wt %. Abbreviations as in Table 4. Mg, Fe, Ca, molar percentages of enstatite, ferrosilite and wollastonite,
respectively.
1660

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
1661

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Table 7: Summary of mineral compositions in LVF samples
Sample
Olivine
Fo)
n
Zoning
Pyroxene
n
Zoning
Plag
(An)
n
Zoning
(
(En, Fs, Wo)
Historical lava flows
LAM18981
LAM18831
LAM18642
LAM1070
LAM2961
LAM1762
LAM1207
LAM1563
LAM1561
LAM1761
LAM6071
LAM1763
LAM7070
74–29
69–46
65–44
65–52
66–41
59–46
69–55
64–42
63–48
72–36
68–59
73–36
68–37
22
17
26
26
23
6
N
36–42, 12–19, 43–48
—
24
—
20
13
11
—
2
N
95–60
91–50
94–34
92–40
90–26
86–52
89–43
93–24
87–26
93–24
89–23
91–28
88–29
67
45
50
64
66
24
40
56
58
96
65
66
58
N, R, O
N, O
N, R, O
N, O
N
N
—
N
N
38–54, 19–37, 9–43
39–55, 23–35, 10–38
39–46, 23–41, 13–38
—
N
N
N
N
—
—
N
—
—
—
—
—
—
N
N, O
N, O
N
17
24
20
37
9
37–42, 34–35, 24–28
15–43, 17–45, 40–46
34–65, 13–32, 3–52
20–43, 17–29, 41–51
38–42, 16–57, 5–41
38–43, 14–21, 41–42
36–42, 19–25, 39
16
21
40
9
—
—
—
—
N
N, O
N, R, O
N, O
N
29
29
29
13
N
N, R, O
Cinder cones
GKK1107
GCK1071
ALA116
78–64
79–40
78–74
70–60
77–61
77–32
23
12
11
31
44
23
—
N
24–41, 12–31, 45–47
37–41, 13–15, 48–50
10–47, 21–56, 32–34
50, 25, 25
4
4
—
96–59
94–31
96–71
79–39
93–58
95–29
41
13
26
87
46
72
N
—
—
N
—
—
—
N
33
1
N
GOO13071
GGI2661
—
N
34–41, 12–15, 47–51
20–39, 13–16, 47–65
31
49
N
N
GKL256
N, R
N, O
Prehistoric lava flows
RWG116
RAG1962
LAO296
68–51
39
22
24
N
N
N
38–43, 17–21, 36–44
39–42, 11–15, 43–49
34–41, 16–20, 43–46
10
16
13
—
93–28
95–43
94–31
43
52
61
N, O
N, O
N, O
78–58
72–47
—
N, R
n, number of analyses. Zoning: N, normal; R, reverse; O, oscillatory.
Pyroxene
more Fe rich. CaO contents in excess of 23 wt % (salite)
are not uncommon in samples from cinder cones and
PLFs (Table 5). Chromian pyroxene is absent, with
The vast majority of pyroxenes are typical augites (Fig.
4), with minor low-Ca pyroxene observed only in the
Cr O typically <0·1 wt %.
more evolved historical lavas, and rare ferroaugite. Where
present, low-Ca pyroxene is occasionally mantled by Cpx
and often unstable. As mentioned above, inclusions of
Mag are very common towards the rims of pyroxene
2
3
Pyroxenes from cinder cone and PLF samples are
distinguished by their high Al contents, often in excess
of 5 wt %, and also by high TiO (>1 wt % or more;
2
phenocrysts. This may be a reaction texture resulting Table 5). Al O concentrations in clinopyroxene from
2
3
from the breakdown of unstable Fe-rich pyroxene, or an subcontinental and suboceanic peridotites range from 2
oxidation effect. Pyroxenes can contain high con- to 8 wt %, with decreasing Al associated with increasing
centrations of structurally bound OH (up to 110 ppm proportions of coexisting melt (Seyler & Bonatti, 1994).
H O), particularly mantle pyroxenes (e.g. Skogby, 1994). Similar augites and Ti-rich aluminous salites to those
2
Such OH may provide fuel for oxidation– found in the LVF have been reported in alkaline (potassic)
dehydroxylation reactions that occur during the for- basalts from Ringgit–Beser, an extinct volcanic complex
2
+
3+
mation of oxy-hornblende and convert Fe to Fe . situated on the north coast of East Java (Fig. 1; Edwards
Phenocryst rims and groundmass crystals are typically et al., 1994), and in calc-alkaline basalts from behind the
1662

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
Fig. 4. Pyroxene and olivine phenocryst compositions, plotted on the pyroxene quadrilateral and Fo–Fa tie-line respectively, for LVF historical
lavas, cinder or spatter cones and prehistoric lavas. Tie-lines in the pyroxene quadrilateral join coexisting CPX and OPX phenocrysts at 5 kbar,
based on the thermometry of Lindsley (1983). Ticks are in increments of 10%.
volcanic front along the Sangihe arc north of Sulawesi
(Morrice & Gill, 1986).
Plagioclase
Plag exhibits a wide compositional range in many samples,
with a maximum anorthite (An) content of >96 mol %
(
Table 7; Fig. 5). The more An-rich compositions are
found in glomerocrysts and as phenocryst cores (Table
6
), and are more common in cinder cone samples.
Highly calcic plagioclase is common in high-Al arc
basalts, with An content dependent upon the H O content,
2
crystallization pressure, Al O content and CaO/Na O
2
3
2
value of the melt (e.g. Panjasawatwong et al., 1995). Melt
CaO/Na O exerts the strongest control on plagioclase
2
An content, with values of CaO/Na O > 8 required to
2
produce highly calcic (An ) plagioclase as observed in
95
Fig. 5. Plagioclase phenocryst compositions plotted on the An–Ab–Or
ternary, for LVF historical lavas, cinder or spatter cones and prehistoric
lavas. Ticks are in 10% increments.
the LVF samples. The highest CaO/Na O ratios in the
2
LVF suite (>6–7) occur in the KVM lavas from the
1663

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
2+
Fig. 6. Forsterite (Fo) contents of olivines vs mg-number [100Mg/(Mg + Fe )] of host basalts. mg-number is calculated using the Fe-oxide
ratios determined by titration (Table 2). Data shown are for historical lavas (Χ, Β), cinder cones (Α, Μ) and PLFs (Φ, Ε); symbols represent
phenocrysts and microphenocrysts, respectively. Historical lavas are labelled with eruption dates (where known); cinder cones and PLFs are
2 3
labelled with sample names. Dotted lines show where the data would plot if mg-numbers were recalculated assuming Fe O /FeO = 0·2. The
shallow-pressure (SP) and high-pressure (HP) equilibrium fields for basaltic magma (Roeder & Emslie, 1970; Ulmer, 1989) are indicated.
northern and western flanks of LTT (Fig. 2), which also The relatively high oxide content of some samples may
contain An-rich Plag (Table 7), but these are probably indicate widespread oxidation and breakdown of am-
not true liquid compositions. High values (CaO/Na O phibole in LVF magmas, which is only rarely preserved.
2
>
7) also occur in the PLF2 samples, although these are
highly phyric.
Disequilibrium textures (embayed crystals, sieve tex-
ture, etc.) are common in Plag (particularly in the historical GEOCHEMISTRY
lavas and the PLF2–PLF3 samples) but are not restricted
to a particular compositional range. Zoning of several
Overview
Whole-rock major and trace element XRF data for the
types (normal, reverse, oscillatory) is prevalent in Plag,
especially in the historical lavas (Table 7). Sieve-textured
or cellular Plag can form through resorption (e.g. as a
result of decreasing P, increasing PH2O, or mixing with
hotter magma) or rapid crystal growth under vapour-
saturated conditions or after mixing with cooler magma
LVF suite are reported in Table 2. In Fig. 7, the LVF
data are compared with existing analyses from other
active and extinct Indonesian volcanoes. The least
evolved LVF basalts (>43 wt % SiO ) represent some
of the most SiO -poor lavas reported from the archipelago
2
2
to date, considerably lower than the mean SiO of 55·5
2
(
e.g. Tsuchiyama, 1985; Morrice & Gill, 1986). These
wt %. Although many of the LVF lavas have higher
MgO contents than the relatively low Indonesian average
of 4·5 wt % (calculated from the data in Fig. 7b), they
also have rather low MgO at a given value of SiO2
in comparison with other Indonesian volcanoes. K O
contents in LVF samples are lower than the mean of 2·5
processes can be distinguished using textural evidence,
because resorption leads to cellular zones with diffuse
inner boundaries, whereas rapid growth results in zones
with sharp borders, often associated with oscillatory zon-
ing. The latter appear to dominate in the LVF lavas,
although the cellular cores of Plag are often more diffuse.
2
wt % for the data in Fig. 7a.
Variations in K O content broadly divide the LVF
2
suite into two series, a medium-K series and a high-K
calc-alkaline series (Fig. 8a). In Fig. 8, the LVF lavas are
Amphibole
The occurrence of amphibole in the LVF lavas is re- plotted along with data from Semeru and Bromo (Fig.
stricted to rare phenocrysts of oxy-hornblende in two 1) and from Iliboleng and Lewotolo, which lie east of
KVM samples. These crystals were not analysed by Flores and north of the arc–continent collision zone (Fig.
electron microprobe. They have narrow rims of fine- 1). The bulk of the LVF samples, including the historical
grained oxides, with dark brown pleochroic interiors. lavas, most of the KVM series and the PLF2 and PLF3
1664

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
(
including the oldest historical flows), trachybasalts and
basaltic trachyandesites.
As is commonly observed in subduction-related vol-
canic rock classification (e.g. Vukadinovic & Sutawidjaja,
1995), the use of different parameters to categorize the
LVF lavas gives slightly contradictory results. On an
AFM diagram (Fig. 9) the LVF samples straddle the
tholeiitic–calc-alkaline boundary with a definite tholeiitic
tendency, showing marked Fe enrichment with respect
to the Semeru and Bromo–Tengger lavas, particularly
at intermediate compositions. This is contrary to the
K O plot (Fig. 8a), but as Vukadinovic & Sutawidjaja
2
(
1995) pointed out, the two methods of discrimination
probably relate to mutually exclusive aspects of the lava
suite, namely its source (K O) and its crystallization
2
history (AFM). The LVF samples are thus relatively and
variably K rich (probably inherited from the source) and
also show Fe enrichment (probably related to magma
chamber processes).
Major and trace element variations
Harker diagrams of major element variations with SiO2
in the LVF suite are shown in Fig. 10. Both MgO and
CaO behave compatibly throughout the entire series,
indicating the influence of Ol and/or Cpx removal. A
group of three samples with >10–11 wt % MgO (PLF2
and maar deposits) have accumulated significant Ol; the
other lavas have MgO <7·5 wt %. The KVM series
show the highest MgO values, and the early historical
flows (e.g. the 1821 flow) and more evolved KVT rocks
Fig. 7. (a) K
2
O vs SiO
2
data for LVF lavas from this study (Ε) and
from Mulyana (1989) (Χ), plotted with 678 existing analyses (Β) from
active and extinct volcanoes of Java, Sumatra, the Lesser Sunda Islands,
the Banda Arc, Sulawesi, Maluku, Sangihe and Halmahera. Data the lowest (>2·7 wt %). CaO also peaks in the most
sources: Whitford (1975), Nicholls & Whitford (1976, 1983), Whitford mafic KVM samples at 13·4 wt %.
et al. (1979, 1981), Foden & Varne (1980), Hutchison (1982), Foden
(
Levels of Fe O ∗ and TiO show similar patterns to
2
3
2
1983), Self et al. (1984), Wheller & Varne (1986), Camus et al. (1987),
Wheller et al. (1987), Stolz et al. (1988, 1990), Van Bergen et al. (1989),
Varekamp et al. (1989), Vukadinovic & Nicholls (1989), Leterrier et al. become more scattered, possibly reflecting the in-
MgO and CaO until >51 wt % SiO when the data
2
(
1
1990), Wahyudin (1991), Gerbe et al. (1992), Edwards et al. (1993,
994), Van Gerven & Pichler (1995), Vukadinovic (1995), Vukadinovic
volvement of two distinct magma series (most clearly
defined in the TiO diagram). Elevated Fe O ∗ (>15–16
2
2
3
&
Sutawidjaja (1995), Mandeville et al. (1996), Self & King (1996) and
Chesner (1998). (b) MgO vs SiO
2
for LVF lavas from this study wt %) and TiO (up to >2 wt %) contents are seen in
2
and Mulyana (1989), plotted with 374 existing analyses from other the PLF1 and some KVT samples, and in several maar
Indonesian volcanoes. Symbols and data sources as in (a).
deposits; all these samples also show relatively high K O
2
and P O (Fig. 10). The latter are both incompatible in
2
5
the LVF suite and there is no evidence for appreciable
samples, fall in the medium-K field and form an array apatite fractionation, although there is a slight decrease
that continues to higher SiO through the Semeru lavas. in P O with increasing SiO in the high-K group.
2
2
5
2
Crystal accumulation in the PLF2 series may have dis-
placed them towards transitional tholeiitic compositions Al O ). There is considerable scatter in Al O contents,
The LVF lavas are highly aluminous (>14·5–20 wt %
2
3
2
3
(
Fig. 8a). The KVT series, PLF1 samples and the maar probably as a result of variations in Plag abundance, as
deposits generally lie in the high-K field, along with the the most Plag-rich sample is also the most aluminous
rocks from the Bromo–Tengger complex and the eastern (GKN3060; 20·2 wt % Al O ). Despite this, Al O appears
2
3
2
3
Sunda arc (Iliboleng, Lewotolo). Similar patterns emerge to increase initially in the LVF lavas, until an inflexion
from a total alkali–silica plot (Fig. 8b), wherein the at 50–51 wt % SiO after which there is a general decline.
2
majority of LVF lavas are classified as typical basalts with This may also reflect multiple magma series, or possibly
subordinate picrobasalts (KVM series), basaltic andesites the onset of significant Plag fractionation.
1665

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Fig. 8. (a) K
and maar deposits (Ο)], Semeru (Η), Bromo–Tengger (Φ), Iliboleng (Β) and Lewotolo (Α). Data sources—LVF: this study; Semeru: Wahyudin
1991) and this study; Bromo: Van Gerven & Pichler (1995) and this study; Iliboleng and Lewotolo: S. A. Carn (unpublished data, 1996). Field
2 2
O vs SiO plot, after Peccerillo & Taylor (1976). Data are shown for the LVF [historical lavas (Χ), cinder cones (Μ), PLFs (Ε)
(
boundaries are from Rickwood (1989). (b) Total alkali–silica (TAS) diagram, after Le Bas et al. (1986). Data and symbols as in (a). (See text for
discussion.)
Variations in trace element concentrations with SiO2 show considerable scatter, as for Al O , suggesting the
2
3
are shown in Fig. 11. Rb contents show similar patterns influence of Plag accumulation (e.g. Vukadinovic, 1993).
to K O, with the historical lavas, PLF2, PLF3 and most The ‘low-SiO , high-Rb’ samples mentioned above also
2
2
of the KVM samples forming a linear array with lower have slightly elevated Sr, suggesting that these magmas
Rb (>6–18 ppm), and a more diffuse group with higher may have suffered similar contamination.
Rb (>21–77 ppm). The latter group comprises a small
Ba contents show a wide range (>200–1000 ppm) and
number of KVM and KVT samples at low SiO and also define a linear array with relatively low con-
2
relatively high Rb that are closely related in space centrations increasing with SiO , and a more enriched
2
(
samples with prefixes GCK, GKG, GMA and GKN, cluster at >50 wt % SiO . Concentrations in several PLF
2
situated NE of Klakah; Fig. 2), and a cluster at higher samples displaced below the linear array (Fig. 11) may
SiO (>50 wt %) consisting of the PLF1 series, the KVT have been diluted through the accumulation of ferro-
2
lavas from the western LVF and many of the maar magnesian phenocrysts. Similar patterns are displayed
deposits (Fig. 11).
by Zr, Y and Nb, with the divergent trend or enriched
The relatively enriched, more evolved group of samples group at >50 wt % SiO particularly evident. Zr contents
2
is also evident in the Ba, Zr, Nb, Y and (to a lesser are rather depleted in the low-SiO samples (<50 ppm)
2
extent) Sr plots (Fig. 11). Sr contents (>300–600 ppm) but increase to >75 ppm at 54–55 wt % SiO in the
2
1666

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
Fig. 9. AFM diagram [where A is Na
and Bromo–Tengger (Α). Data sources as in Fig. 8. Boundaries between the tholeiitic (T) and calc-alkaline (CA) fields from Kuno (1968) and
Irvine & Baragar (1971) are shown.
2
O + K O, F is FeOtot (total Fe as FeO) and M is MgO] showing lavas from the LVF (Φ), Semeru (Ε)
2
most evolved historical lavas, and attain up to 226 ppm a prominent trough at Nb, which is characteristic of
in some PLF1 and KVT rocks. Y varies from >15 to subduction zone magmas. This depletion of Nb relative
2
7 ppm with increasing SiO in the linear array and up to the large-ion lithophile elements (LILE; e.g. Rb, Ba,
2
to >58 ppm in the enriched group. Niobium contents K) is attributed primarily to two processes: (1) the addition
are also relatively low in the LVF lavas, with many of an LILE-enriched, Nb-poor fluid component to the
samples close to detection limits (Table 2) and the highest mantle wedge; (2) the preferential retention of Nb in
contents (10–11 ppm) recorded in a KVT sample and amphibole relative to other phases present in the mantle
maar deposit. In the enriched group of samples, Zr, Y, source (e.g. Borg et al., 1997). Similar processes are
Nb and possibly Rb and Sr all appear to decrease with inferred to explain the general depletion of the high field
increasing SiO (Fig. 11).
strength elements (HFSE) Zr, Ti and Y with respect to
Both Sc (>15–60 ppm) and V (>100–670 ppm) LILE in arc magmas (e.g. Pearce & Peate, 1995).
contents exhibit a steady decline with increasing SiO2, Inter-element trends for the LVF samples generally
2
although there is also evidence for an enriched group of resemble a typical primary island-arc basalt (IAB) profile,
samples at >50 wt % SiO , as observed in the in- but with more enriched values of most elements, although
2
compatible element plots (Fig. 11). This compatibility of this is of course dependent on the IAB values used.
Sc and V in the LVF lavas attests to the influence of Historical samples form the most homogeneous group,
Cpx removal throughout the suite, with other deviations with notable enrichments in Rb, Ba, K, Pb, P, Ti and
(
particularly in Sc) probably attributable to varying modal Y compared with IAB, and levels of Ba and Pb narrowly
proportions of the mineral. exceeding OIB (Fig. 12a). Ba enrichments in Javanese
Ni (<60 ppm) and Cr (<160 ppm) contents are both volcanic rocks are probably related to the input of Ba-
low in the LVF lavas, peaking in PLF2 samples that have rich oceanic sediments into the subduction zone (e.g.
visibly accumulated Ol (Fig. 11). Low Cr values imply Plank & Langmuir, 1993).
minimal contamination by chromian spinel (e.g. in Ol )
The cinder cone lavas display varying levels of en-
or by Cr-rich pyroxene. The group of KVT and PLF1 richment, although some of the variation will be due to
basalts that are relatively enriched in incompatible ele- crystal fractionation (Fig. 12b). KVT samples from the
ments also have slightly elevated Ni (>30–40 ppm) but western LVF are typically most enriched, up to 10×
the values are still low. The historical lavas and some primary IAB for Rb, and with relatively high values
KVM samples are particularly depleted in Ni and Cr, (compared with IAB) for all other elements except Sr
even where Sc and V are elevated, which suggests that (indicating Plag fractionation), exceeding or approaching
even the least evolved lavas have undergone some Ol ocean-island basalt (OIB) values in most cases. The KVM
fractionation. This is surprising given their low SiO2 rocks and the cinder cones NE of Klakah (Fig. 2) generally
contents, and another possibility is that low Ni and Cr represent the least enriched (i.e. most IAB-like) samples
were inherited from the source.
in the LVF. The latter have an interesting profile in Fig.
Figure 12 shows N-MORB normalized trace element 12b, showing significant Rb enrichment relative to IAB
abundances in LVF lavas. All of the samples show without a corresponding Ba anomaly. They are also
1667

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Fig. 10. Major element variation (vs SiO
cinder or spatter cones within maars (Μ) are now also distinguished. Fe
2
) in the LVF suite, showing historical lavas (Β), cinder cones (Α), PLFs (Φ) and maar deposits (Η);
∗, total Fe as Fe
O
2 3
2 3
O .
depleted in Nb, P, Zr and, to a lesser extent, Sr, although Ti and Y with respect to typical IAB and consists of the
enriched in Ti and Y, compared with IAB. PLF1 series; these samples also show a small negative Sr
The PLF data display a similar range to the cinder anomaly similar to the KVT cones in the western LVF.
cone samples, although there are two discrete groups The less enriched group comprises the PLF2 and PLF3
rather than a continuum of compositions (Fig. 12c). The series, with the former exhibiting slight depletions in Nb,
upper group is highly enriched in Rb, Ba, K, Pb, P, Zr, P and Zr, and enrichments in Ti and Y, relative to IAB
1668

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
2
Fig. 11. Trace element variation (vs SiO ) in LVF lavas. Symbols as in Fig. 10.
(
Fig. 12c). The only element that appears to be relatively subduction signature of Nb and HFSE depletion and
constant throughout the suite is Sr, which is typically at LILE enrichment, with the cinder cone and PLF samples
or close to typical IAB abundance. showing the greatest variability. Several of the KVM and
In summary, all of the LVF samples are enriched in PLF2 samples show low abundances of Nb, P and Zr
Rb, Ba, K, Pb, Ti and Y to some degree relative to a relative to typical primary IAB and N-MORB, but with-
typical primary IAB. Most of the lavas carry the typical out corresponding depletions in Ti and Y, and some cones
1669

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
Whitford, 1975). This is consistent with the general
87
86
decline in Sr/ Sr from West Java to Bali (Fig. 1).
87
86
The Sr/ Sr ratios do not correlate with SiO or Rb/
2
Sr, although the highest ratio occurs in a KVM sample
with anomalously high Rb and relatively high Rb/Sr
(
GCK2961; Table 2), which suggests source enrichment
in this region. The lowest ratios occur in the historical
lavas.
Principal component analysis
To clarify the complex patterns presented on binary
diagrams by the large number of samples and elements
analysed, the major and trace element XRF data from the
LVF suite have been subjected to principal component
analysis (PCA). PCA locates the directions of maximum
variance in a multivariate dataset by finding the eigen-
values and eigenvectors of the data correlation matrix,
and selects the combination of the initial variables re-
sponsible for this variance (e.g. Le Maitre, 1968; Al-
bar e` de, 1995). In this analysis the historical lavas, cinder
cones and PLFs have been treated as separate groups.
The first three principal components (PCs) of the
historical lava data, which represent 85% of the total
variance, are depicted in Fig. 13. The horizontal spread
of these data (PC1) is generally related to the eruption
age of the lavas, with PC2 probably picking out subtle
variations in phenocryst content.
At opposite extremes of the PC1 axis, we find the
oldest (1821) and youngest (1898) historical samples. The
former correlate with the incompatible elements (Rb, Zr,
Ba, etc.) and relatively evolved components (SiO , K O,
2
2
Fig. 12. N-MORB-normalized trace element diagrams for (a) historical Na O; e.g. alkali feldspar), whereas the latter correlate
2
lavas, (b) cinder cones (the cinder cone cluster NE of Klakah is
distinguished by dashed lines), and (c) PLFs. Samples with Nb or Pb
with a ‘pyroxene and Fe–Ti oxide’ component (i.e. MgO,
Fe O , TiO , CaO, V, Sc; Fig. 13). This probably implies
2
3
2
below or close to detection limits (Table 2) are excluded. Typical
ocean-island basalt (OIB), primary island-arc basalt (IAB) and E- that the magma chamber feeding the 19th-century erup-
MORB trends are shown for reference. Normalizing values, OIB and tions at Lamongan was stratified, with the relatively
E-MORB compositions from Sun & McDonough (1989); IAB is sample
evolved basalts at the top of the chamber erupted first
ID16 from Nye & Reid (1986), a primary arc basalt from the Aleutian
(1821) and the more mafic magmas at the chamber base,
Islands.
enriched in dense ferromagnesian phenocrysts, erupted
last (1898) and at the lowest altitude (Fig. 3).
appear to preserve localized geochemical heterogeneities
Lavas erupted in 1864 cluster at the centre of the plot
such as enrichment in Rb relative to Ba. Abundances of (Fig. 13), i.e. closest to the ‘average’ composition of
Rb, P and Zr show the greatest, and Sr the smallest, historical Lamongan basalts. Significantly, Al O , Ni, Cr
2
3
range. A similar consistency of Sr abundance is seen in and Nb appear to exert little influence on the chemistry
lavas from Gunung Slamet in central Java (e.g. Vu- of erupted lavas, except possibly through minor variations
kadinovic & Sutawidjaja, 1995).
in Plag content, suggesting that the abundance of these
variables (i.e. high Al O , low Ni, Cr, Nb) was controlled
2
3
primarily by the ‘source’ or in this case the magma
supplied to the chamber.
A corollary of using PCA is that lavas of similar overall
Sr isotopes
The limited number of Sr isotope ratios available occupy composition, and by inference of similar age, will cluster
a fairly narrow range (0·7042–0·70445; Table 2), with together in PC space. This is of use at Lamongan, where
the lowest values close to the Sunda arc minimum (e.g. there is poor age control on several samples. The plot
1670

CARN AND PYLE
LAMONGAN VOLCANIC FIELD, EAST JAVA
Fig. 13. Principal component analysis (PCA) of major and trace
elements in 19th-century lavas from Lamongan volcano, showing the
first three principal components (PCs). Β, positions of the initial
variable axes in PC space (see inset for key); Χ, historical lava samples
Fig. 14. PCA of major and trace elements in LVF cinder cone samples,
showing the first three PCs, plotted as in Fig. 13. Μ, cinder cone
samples. The first two components account for 77·5% of the total
variance. In the lower plot, group a comprises the KVT samples from
the western LVF; group b comprises two K-rich (shoshonitic) samples.
In the upper plot, the following groups are demarcated: ꢁ, most of the
cones to the north and south of LTT; ꢂ, the cones in the ‘transition
zone’ west of LTT; ꢃ, the KVT cones in the western LVF. (See text
for discussion.)
(
several samples are labelled with eruption ages). Units on both axes
are standard deviations. The first two components account for >77%
of the total variance.
of PC3 against PC1 (Fig. 13) splits the samples into a
‘Y’-shaped array, with the 1821 samples grouping with
sample LAM1207 on the lower ‘arm’. This latter sample from the western LVF (west of the roughly north–south
was thus probably erupted at a similar time to the former, lineament defined by the three lake-filled maars im-
but from a different vent (Fig. 3). Similarly, samples mediately east of Klakah; Fig. 2) and two transitional
LAM1261 (1847 flow) and LAM1563 (1821 flow?) also high-K–shoshonitic samples (GTE288 and GAP1070).
have similar characteristics (e.g. relatively high Nb con- The KVM series and KVT samples from the central
tents) and hence may in fact be part of the same flow and eastern LVF form the other limb, with the younger
field (probably 1847; Fig. 3).
cones plotting towards the left-hand side (Fig. 14). There
The cinder cone samples form a broadly L-shaped is no appreciable clustering of samples at the ‘apex’,
array in the plot of the first two PCs (Fig. 14), and the indicating that the LVF cones constitute two distinct
first three PCs account for >87% of the variance. This magma series with little compositional overlap between
L-shaped pattern is often observed for suites of lavas them. However, several of the lavas that do plot in the
from individual volcanoes, with picritic and differentiated region of the apex (GRG14071/2, GCK2960,
flows occupying separate ‘arms’ and common com- GKN3060) are derived from cones situated close to the
positions clustering at the apex (e.g. Albar e` de et al., 1997). boundary between the compositionally distinct cones of
In Fig. 14, one limb is composed of the KVT cones the western LVF and the rest of the complex (Fig. 2).
1671

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 9
SEPTEMBER 2001
The western LVF samples are predominantly con-
trolled by K O, P O and the incompatible elements,
2
2
5
whereas the spread of compositions in the other lavas
correlates with a CaO–MgO–Fe O –Sc–V–Co com-
2
3
ponent (Cpx or possibly amphibole) and a SiO2–
Na O–Al O –Sr component (Plag), the former
2
2
3
predominating in the KVM samples (Fig. 14). As ob-
served in the historical lavas, Ni and Cr are not strongly
correlated with the ferromagnesian component and plot
at the centre of the diagram, suggesting minimal influence
of mantle phases (Ni-rich Ol, Cr-rich pyroxene) on the
major compositional trends and reflecting the overall low
Ni and Cr contents of the suite. These elements exert
more influence in the PC3 plot (Fig. 14), which probably
reflects minor variations as a result of Ol accumulation
in some magmas. Also in the PC3 diagram, the Al O3
2
axis plots towards the pyroxene component, indicating
the significance of Al-rich pyroxene in these lavas.
The spatial variation of magma composition in the
LVF is also subtly manifested in the PC3 plot (Fig. 14),
wherein three distinct geographical groups of cinder
cones can be recognized. One group collects most of the
cones on the northern and southern flanks of the LTT
edifice (ꢁ; Fig. 14), which show a strong correlation with
Fe O and TiO . Another assemblage (ꢂ) contains the
2
3
2
samples from a narrow region to the west of LTT and
east of the three lake-filled maars (GMA5070, GCK2961/
2960/1071, GKG136, GKL256, GRG14071/2,
GKN3060; Fig. 2). This acts as a ‘transition zone’ between
the previous group and the third group (ꢃ), which contains
the KVT samples from the western LVF. This last group
seems to correlate with Sr and the incompatible elements,
and possibly Ni and Cr.
Fig. 15. PCA of major and trace elements in LVF prehistoric lava
flows, showing the first three PCs, plotted as in Fig. 13. Ε, the PLF
samples. The first two components account for 81·3% of the total
variance. PLF1 and PLF2/3 samples are clearly separated in these
The PLF samples also define an L-shaped array in PC
space (Fig. 15), with the PLF1 group clearly distinguished. diagrams. (See text for discussion.)
These lavas correlate with the axes for K O, P O and
2
2
5
the incompatible elements in the PC1–PC2 plot, whereas
the PLF2/3 trend is controlled by a ferromagnesian
component (Ol, Cpx) and Plag (SiO , Al O , Na O, Sr).
Thus the PLF1 samples, which come from localities in
the western LVF (Fig. 2), have similar overall char-
acteristics to the KVT cones in the same region of the
field.
Contrary to the historical lavas and cinder cones, the
PLFs show a more robust correlation between Ni, Cr
and MgO. This indicates that the Ni and Cr budget is
primarily controlled by the accumulation of phenocrysts
rich in compatible elements (e.g. Ol ) in the PLFs, and
by the source in the historical lavas and cinder cones.
To summarize, many latent geochemical features of the
LVF lavas are revealed by PCA. Spatial heterogeneities in
magma composition are particularly apparent, with a
Zr, K, P, ± Nb are interrelated in all samples, implying
that these elements behaved incompatibly during magma
evolution. Compatibility of Sr in Plag is indicated by the
general correlation of Sr with the Plag component on
plots of the first two PCs. Similarly, the general lack of
correspondence between Ni, Cr and the ferromagnesian
phases (except in the PLFs) attests to the low con-
centrations of these elements in the LVF suite (Fig. 11),
which are probably a feature of the source magmas.
2
2
3
2
FRACTIONAL CRYSTALLIZATION IN
THE LVF LAVAS
Major elements
clear distinction evident between lavas from the eastern The results of PCA suggest that fractional crystallization
and western parts of the LVF, and transitional samples of Ol, Cpx, Plag and Fe–Ti oxides may be responsible for
with intermediate compositions. Contents of Rb, Ba, Y, much of the compositional variation within individual
1672