Contrib Mineral Petrol (2017) 172:97
DOI 10.1007/s00410-017-1418-1
ORIGINAL PAPER
Lower crustal hydrothermal circulation at slow‑spreading ridges:
evidence from chlorine in Arctic and South Atlantic basalt glasses
and melt inclusions
Froukje M. van der Zwan1,6   · Colin W. Devey1   · Thor H. Hansteen1 ·
Renat R. Almeev2 · Nico Augustin1 · Matthias Frische1 · Karsten M. Haase3 ·
Ali Basaham4 · Jonathan E. Snow5 
Received: 29 March 2017 / Accepted: 25 October 2017
© Springer-Verlag GmbH Germany 2017
Abstract  Hydrothermal circulation at slow-spreading
ridges is important for cooling the newly formed lithosphere,
but the depth to which it occurs is uncertain. Magmas which
stagnate and partially crystallize during their rise from the
mantle provide a means to constrain the depth of circula-
tion because assimilation of hydrothermal fluids or hydro-
thermally altered country rock will raise their chlorine (Cl)
contents. Here we present Cl concentrations in combina-
tion with chemical thermobarometry data on glassy basaltic
rocks and melt inclusions from the Southern Mid-Atlantic
Ridge (SMAR; ~ 3 cm year−1 full spreading rate) and the
Gakkel Ridge (max. 1.5 cm year−1 full spreading rate) in
order to define the depth and extent of chlorine contami-
nation. Basaltic glasses show Cl-contents ranging from ca.
Communicated by Jochen Hoefs.
Electronic supplementary material  The online version of
this article (https://doi.org/10.1007/s00410-017-1418-1) contains
supplementary material, which is available to authorized users.
* Froukje M. van der Zwan
fzwan@geomar.de
1
Geomar Helmholtz Centre for Ocean Research Kiel, Kiel,
Germany
2
Leibniz Universität Hannover, Institute of Mineralogy,
Hannover, Germany
3
GeoZentrum Nordbayern, Universität Erlangen-Nürnberg,
Erlangen, Germany
4
Faculty of Marine Science, King Abdulaziz University,
Jeddah, Saudi Arabia
5
Department of Earth and Atmospheric Sciences, University
of Houston, Houston, USA
6
Present Address: Institute für Geowissenschaften,
Christian-Albrechts-Universität Kiel, Kiel, Germany
50–430 ppm and ca. 40–700 ppm for the SMAR and Gakkel
Ridge, respectively, whereas SMAR melt inclusions contain
between 20 and 460 ppm Cl. Compared to elements of simi-
lar mantle incompatibility (e.g. K, Nb), Cl-excess (Cl/Nb or
Cl/K higher than normal mantle values) of up to 250 ppm in
glasses and melt inclusions are found in 75% of the samples
from both ridges. Cl-excess is interpreted to indicate assimi-
lation of hydrothermal brines (as opposed to bulk altered
rock or seawater) based on the large range of Cl/K ratios in
samples showing a limited spread in H ­ 2O contents. Resorp-
tion and disequilibrium textures of olivine, plagioclase and
clinopyroxene phenocrysts and an abundance of xenocrysts
and gabbroic fragments in the SMAR lavas suggest multi-
ple generations of crystallization and assimilation of hydro-
thermally altered rocks that contain these brines. Calculated
pressures of last equilibration based on the major element
compositions of melts cannot provide reliable estimates of
the depths at which this crystallization/assimilation occurred
as the assimilation negates the assumption of crystallization
under equilibrium conditions implicit in such calculations.
Clinopyroxene–melt thermobarometry on rare clinopyroxene
phenocrysts present in the SMAR magmas yield lower crus-
tal crystallization/assimilation depths (10–13 km in the seg-
ment containing clinopyroxene). The Cl-excesses in SMAR
melt inclusions indicate that assimilation occurred before
crystallization, while also homogeneous Cl in melts from
Gakkel Ridge indicate Cl addition during magma chamber
processes. Combined, these observations imply that hydro-
thermal circulation reaches the lower crust at slow-spreading
ridges, and thereby promotes cooling of the lower crust. The
generally lower Cl-excess at slow-spreading ridges (com-
pared to fast-spreading ridges) is probably related to them
having few if any permanent magma chambers. Magmas
therefore do not fractionate as extensively in the crust, pro-
viding less heat for assimilation (on average, slow-spreading
13
Vol.:(0123456789)
 97   Page 2 of 23
ridge magmas have higher Mg#), and hydrothermal systems
are ephemeral, leading to lower total degrees of crustal alter-
ation and more variation in the amount of Cl contamina-
tion. Hydrothermal plumes and vent fields have samples in
close vicinity that display Cl-excess, mostly of > 25 ppm,
which thus can aid as a guide for the exploration of (active
or extinct) hydrothermal vent fields on the axis.
Keywords  Hydrothermal circulation · (Ultra)slow-
spreading ridges · Crystallization depths · Crustal
assimilation · MORB · Chlorine
Introduction
When new oceanic lithosphere is formed at a spreading
centre, heat is transferred upwards from the mantle either
by advection (magmatism) or tectonism. Although there is
fairly convincing evidence that hydrothermal circulation of
seawater cools the upper parts of this system (e.g. Stein and
Stein 1994), up to present no consensus on the maximum
depth of hydrothermal circulation in newly formed oceanic
crust (e.g. Mottl 2003) or what role hydrothermal circulation
plays in the deep cooling of ridges has been reached. Particu-
larly at (ultra)slow-spreading (< 1.2 cm year−1 full spreading
rate) and slow-spreading ridges (< 1.2 to < 5.5 cm year−1)
cooling of the lower lithosphere by hydrothermal circulation
may be relevant, as seafloor is partially produced by tectonic
processes (e.g. Dick et al. 2003; Cannat et al. 2006) and the
lack of permanent magma chambers and low crustal tem-
peratures (e.g. Harper 1985; Detrick et al. 1990) may allow
faulting to be particularly extensive and to penetrate deeper,
down to the lower crust, potentially providing pathways for
hydrothermal fluids (Harper 1985; Cannat et al. 1991; Mével
and Cannat 1991).
Direct evidence for deep circulation comes from studies
of oceanic drill sections (from crust formed at fast-spreading
ridges) and from ophiolites (e.g. Gregory and Taylor 1981;
Stakes and Vanko 1986; Nehlig and Juteau 1988; Lecuyer
and Reynard 1996; Kawahata et al. 2001; Coogan 2003;
Nicolas et al. 2003), both of which display high-temperature
alteration of the lower crust. Deep hydrothermal circulation
at fast-spreading ridges is further indicated by thermody-
namic modelling (Cherkaoui et al. 2003; Maclennan et al.
2005; Hasenclever et al. 2014) and the seismic structure of
the lower crust (Dunn et al. 2000). Evidence from slow-
spreading ridges include P-wave tomography underneath
the TAG hydrothermal field, which indicates hydrother-
mal circulation, related to a detachment fault, extending at
least 2 km, but potentially 3.5 km deep (Zhao et al. 2012).
Other direct information on the depth of penetration of
hydrothermal circulation at slow-spreading ridges is only
available from drill-cores that have a limited depth extent
13
Contrib Mineral Petrol (2017) 172:97
(e.g. < 1600 m in Alt and Bach 2006) or from tectonically
exposed parts on the lower crust (Gillis et al. 1993), which
have thus been exposed to ambient seawater for extended
periods of time.
Deep hydrothermal circulation could potentially be traced
by detecting it effects on the chlorine budget of erupted mag-
mas if those magmas assimilated the hydrothermal fluids
themselves or hydrothermally altered country rock at depth
(Michael and Schilling 1989; Michael and Cornell 1998;
Kent et al. 1999; Coogan et al. 2003; Gillis et al. 2003; Sun
et al. 2007; France et al. 2009; Kendrick et al. 2013; van der
Zwan et al. 2015). Assimilation of surrounding rocks by ris-
ing MORB magmas has been shown to take place in several
gabbroic complexes (e.g. Bédard et al. 2000 and references
therein; Godard et al. 2009; Drouin et al. 2009; Sanfilippo
et al. 2014; Fischer et al. 2016), but the depth and/or hydro-
thermal alteration state of the assimilated rocks is uncertain
so far.
Although work on fast-spreading ridges indicates addition
of Cl relative to trace elements of similar incompatibility
(e.g. K or Nb) in a reaction zone between the magma and
the overlying hydrothermal circulation cell at the roof of a
magma chamber (Coogan et al. 2003; Gillis et al. 2003; le
Roux et al. 2006; France et al. 2009; Kendrick et al. 2013),
previous studies on slow-spreading ridges (e.g. Michael
and Cornell 1998) had not found evidence of Cl addi-
tion to their magmas. Michael and Cornell (1998) argued
that because fast-spreading ridges generally have shallow
(100–300 MPa ≈ 3–10 km) axial magma chambers whereas
magmas from a typical slow-spreading ridge (e.g. not anom-
alously thick) generally yield high calculated crystallization
pressures (> 300 MPa = > 10 km), hydrothermal circulation
(at least at slow-spreading ridges) must occur at pressures
less than 300 MPa. More recent studies have now found
Cl addition in basalts of some slow-spreading ridges (e.g.
the Red Sea Rift) independent of their calculated crystal-
lization pressures (Jenner and O’Neill 2012; van der Zwan
et al. 2015) and advances in analytical methods mean that
the inherently lower Cl contents of the relatively less frac-
tionated slow-spreading ridge magmas have become less of
an analytical challenge.
To investigate how deep hydrothermal circulation reaches
at slower-spreading ridges, we performed high-precision Cl
measurements (van der Zwan et al. 2012) on basaltic glasses
and melt inclusions and combined this with barometric data
based on glass compositions and on glass and clinopyroxene
phenocryst compositions (cf. Ariskin and Barmina 2004;
Putirka 2008). As the composition and structure of the new
lithosphere at slower-spreading ridges is very heterogene-
ous, we selected two end-member ridges for our study: the
magmatically robust southern Mid-Atlantic Ridge (SMAR)
at 7–10°S (a slow-spreading ridge with ~  3  cm  year−1)
previously interpreted to show varying mantle sources
Contrib Mineral Petrol (2017) 172:97 Page 3 of 23  97

and crystallization pressures (0–800 MPa; Almeev et al.
2008; Hoernle et al. 2011) and the Arctic Gakkel Ridge
at 6°W–85°E (an ultraslow-spreading ridge with max.
1.5 cm year−1) which has a strongly tectonic character and
a very thin crust (Coakley and Cochran 1998; Dick et al.
2003; Michael et al. 2003). Note that at slow- and espe-
cially ultraslow spreading ridges classical “Penrose-type”
magmatic crust (e.g. Vine and Moores 1972) may not make
up the whole newly formed lithosphere (e.g. Cannat 1996)
and may even be totally absent (e.g. Sauter et al. 2013). In
this paper we therefore use the term “crust” to refer to the
wholly magmatic upper part of the lithosphere, if present.
The term “lithosphere” refers to the combination of crustal
and/or mantle rocks which make up the new plate.
Geological background
Southern Mid‑Atlantic Ridge
The southern Mid-Atlantic Ridge (SMAR) between 7° and
10°S, east of Ascension island, is a mature ridge with slow
spreading rates of 31–32  mm  year −1 full spreading rate
(DeMets et al. 2010; Fig. 1a). The area is bounded to the
north by the Ascension Fracture Zone and to the south by
the Bode Verde Fracture Zone. The region is divided into
four segments (A1–A4; Bruguier et al. 2003) that are sepa-
rated by overlapping spreading centres and show varying
morphologies. We chose to study samples from segments A1
and A2 as these samples display the largest range in calcu-
lated last crystallization pressures (200–900 MPa; Almeev
et  al. 2008). Segment A1 has a typical slow-spreading
ridge character with a well-developed deep rift valley of
> 2950 m, while segment A2 shows a shallower axial high
and is 2100–3000 m deep (Fig. 1a; Bruguier et al. 2003).
Segment A2 is associated with a thick seismically deter-
mined crust of ~ 11 km, compared to ~ 5 km under A1
(Minshull et al. 1998; Bruguier et al. 2003). Segment A1
displays microseismicity at 2–4 km depth and teleseismic-
ity (Grevemeyer et al. 2013), while segment A2 shows little
earthquake activity, thought to indicate a higher tempera-
ture gradient and therefore a shallower brittle/ductile transi-
tion in the crust (Fig. 1a; Devey et al. 2010). Presently, A2
appears to be in a period of reduced magmatic activity: the
youngest lava flows are tectonized and yield ages around
1000 years and no signs of active hydrothermalism have
been found (Devey et al. 2010; Haase et al. 2016). The sam-
ples used here are well studied in terms of their major and
trace element and isotope compositions and crystallization
pressures (Möller 2002; Almeev et al. 2008; Paulick et al.
2010; Hoernle et al. 2011). The geochemical studies show
mixing between a depleted unradiogenic MORB source
and radiogenic, trace element enriched HIMU-type sources
(Hoernle et al. 2011). The enriched samples display higher
­H2O contents and nearly isobaric calculated crystallization
pressures (100–300 MPa) while the more depleted samples
show lower H ­ 2O contents and a range of calculated last crys-
tallization pressures (200–900 MPa; Almeev et al. 2008).

Fig. 1  Overview of studied
areas on the a Southern Mid-
Atlantic Ridge (SMAR) and b
Gakkel Ridge. The SMAR study
area is subdivided in various
segments (after Bruguier et al.
2003; samples are from A1 and
A2). The Gakkel Ridge is seg-
mented into three zones (after
Michael et al. 2003): western
volcanic zone (WVZ), sparsely
magmatic zone (SMZ) and east-
ern volcanic zone (EVZ). White
dots indicate earthquake hypo-
centers with M ≥ 3 (from ISC,
1960 to present). Black stars
indicate hydrothermal activity
after German et al. (2002) and
Baker et al. (2004). Spreading
rates (after DeMets et al. 2010)
are indicated by arrows

13
 97   Page 4 of 23
Hydrothermal circulation in the studied area is known from
the Nibelungen hydrothermal field between A1 and A2 and
from two indications of plume activity at 7°57′S and 8°17′S
on A1 (Fig. 1a; Devey et al. 2005; Melchert et al. 2008).
Gakkel Ridge
The Gakkel Ridge (6°W–85°E) shows the lowest orthogonal
spreading rate on Earth at 11–13 mm year−1 full spread-
ing rate (DeMets et al. 2010) and lacks transform offsets
(Fig. 1b). The Gakkel Ridge has a very deep rift valley
(> 4600 m) and a thin seismically determined crustal thick-
ness (< 4 km; Coakley and Cochran 1998; Jokat et al. 2003).
The Gakkel Ridge can be subdivided into three different
magmato-tectonic units (Fig. 1b; Michael et al. 2003). The
Western Volcanic Zone (WVZ; 7°W–3°E) is a volcanic
area with many small cones on the ridge and displays a
larger extent of melting than the other two segments. In the
Sparsely Magmatic Zone (SMZ; 3°E–28°E) in contrast, the
seafloor consists mostly of peridotite with only rare basalts
or volcanic features. The Eastern Volcanic Zone (EVZ;
29°E–85°E) contains large (~ 30 km) widely spaced vol-
canic centres of which some show recent activity. Basalts
from the Gakkel Ridge are generally enriched in trace ele-
ments but with a large variability between the different
segments and also between samples (Michael et al. 2003).
Isotopically, two distinct sources were identified: samples
from the WVZ have a DUPAL-signature attributed to a
contribution from the sub-continental lithospheric mantle,
while samples from the EVZ have a signature similar to
Atlantic–Pacific MORBs (Goldstein et al. 2008) with the
boundary between these mantle domains lying in the SMZ.
Hydrothermal activity at Gakkel Ridge seems to be high
compared to other (ultra)slow-spreading ridges (Edmonds
et al. 2003; Michael et al. 2003). Although only two active
vent sites were confirmed by visual observations (Aurora,
85°E), 7–11 more vent sites were inferred by light scattering,
temperature and Mn concentration data in the water column,
from which we here consider the seven most certain ones
(Fig. 1b; Edmonds et al. 2003; Baker et al. 2004; Pontbriand
et al. 2012). Teleseismic earthquakes are present along the
whole ridge (Fig. 1b).
Analytical methods
Chlorine, major, trace, H­ 2O and C­ O2 measurements and
pressure calculations were performed on glassy basaltic
samples from collections of expeditions R/V Meteor 41/2
(SMAR; e.g. Möller 2002; Almeev et al. 2008; Hoernle et al.
2011) and AMORE 2001 (Gakkel Ridge; e.g. Michael et al.
2003; Goldstein et al. 2008). For each basaltic glass meas-
urement, three pieces of glass > 2 mm across were selected
13
Contrib Mineral Petrol (2017) 172:97
per rock under a binocular microscope based on their fresh
appearance and lack of visible vesicles, alteration rims or
minerals. The mounted glass was prepared, polished and
then cleaned with Milli-Q water following the procedures in
van der Zwan et al. (2012) to avoid Cl contamination. Opti-
cally clear SMAR Ol, Pl and Cpx minerals were prepared
following the same procedure. For selected SMAR samples
with elevated Cl/Nb glass contents, Ol and Pl minerals with
melt inclusions were mounted in UHU Hart glue mixed with
acetone and doubly polished sections (~ 60 µm thick) were
prepared. Melt inclusions were selected based on homoge-
neity and large sizes/thicknesses and with a good surface.
Major elements and Cl measurements were carried out
with a Jeol JXA-8200 ‘‘Superprobe’’ electron microprobe
at GEOMAR, Kiel, using an acceleration voltage of 15 kV.
Chlorine and potassium (K) were measured with a beam
current of 80 nA and a beam diameter of 10 µm using the
mapping technique described in van der Zwan et al. (2012,
2015). The Cl and K results are the average of 2–8 meas-
urements per sample and have an uncertainty of < 3 and
< 20 ppm for Cl and K, respectively (2SE). For the Cl and
K measurements of the melt inclusions, the method was
slightly adapted by a reduced beam current of 50 nA and a
mapping repetition of 14 times, in order to avoid a strong Cl
signal of the mounting material through the thin samples.
Internal glass standards (138-3, 147-2, 157-3; van der Zwan
et al. 2012) show a reproducibility of on average 3.2 ppm
(2SD) for this method. Nevertheless, an additional signal
from the mounting material is visible in the Cl data of the
melt inclusions and corrected for, but results in a higher
uncertainty of 19.2 ppm 2SD. This higher uncertainty does,
however, not influence the trends observed in the melt inclu-
sions, neither the main conclusions derived from Cl data of
the melt inclusions. The K measurements are not affected
by the mounting material. Any measurements that show dis-
crepancies (e.g. inhomogeneous Cl, mostly due to bubble
forming) were discarded.
Major elements of the glasses were measured with a defo-
cused spot of 5 µm and a beam current of 10 nA. Counting
times were 20/10 s (peak/background left and right of the
peak each 10 s) for all elements apart for 30/15 s for MnO.
­Na2O was measured first to avoid loss by heating. For cali-
bration and monitoring of data quality, we used natural ref-
erence samples from the Smithsonian Institute (Jarosewich
et al. 1980). The average was taken of nine analyses per
sample to assure homogeneity. Relative analytical preci-
sion is generally < 2.5%, but up to 5% for N
­ a2O and ~ 30%
for MnO and ­P2O5. Additionally, available (literature) data
for the samples were used; for the SMAR, data are utilized
from Möller (2002) and data measured with a CAMECA
SX-50 at GEOMAR and a JEOL JXA8900 Superprobe at the
Institut für Geowissenschaften, University of Kiel, partially
presented in Haase et al. (2016). Settings were the same as
Contrib Mineral Petrol (2017) 172:97 Page 5 of 23  97
above and accuracy and precision are better than 5% except
for MnO and ­P2O3. Additional major element data measured
on a Jeol JXA 8900RL electron probe microanalyzer at the
University of Mainz were used for Gakkel Ridge basalts.
Mineral point analyses and transects were measured with a
focused beam spot. Plagioclase and Cpx were measured with
a beam current of 20 nA and counting times of 20/10 s, apart
for ­K2O in Pl (40/20 s) and ­Cr2O3 and ­TiO2 in Cpx (30/15 s).
Olivine was measured with a beam current of 100 nA and
20/10 s counting times for S ­ iO2, ­Al2O3, MgO, FeO, 30/15 s
for MnO, ­Cr2O3, 60/30 s for CaO and 40/20 s for NiO.
For consistency, trace elements in all glass samples were
analysed by LA-ICP-MS at GEOMAR, Kiel, using a 193 nm
Excimer laser system (GeoLasPro, Coherent) coupled with
a double-focusing, high-resolution magnetic sector mass
spectrometer (AttoM, Nu-Instruments) under hot plasma
conditions [NAI > 10; ThO/Th ~ 0.03%; details in Fietzke
and Frische (2016)]. Spot analyses of samples were done
by 30 s ablation at a laser repetition rate of 10 Hz using
spot diameter of 90 µm and a fluence of 5 J cm−2. 50 s of
gas background data were collected prior to each ablation.
NISTSRM610 [30 s, 32 µm, 10 Hz, 5 J cm−2; Wise and
Watters 2012] was used for calibration. MPI-DING refer-
ence samples [ML3B-G and KL2-G; 30 s, 32–90 µm, 10 Hz,
5 J cm−2; Jochum et al. (2006)] were used to evaluate the
accuracy during the measurements. Relative analytical pre-
cision based on reproducibility (2SD) of the standard ref-
erence material used for calibration was about 3.5% (see
Online Resource 1). Data evaluation has been performed
applying the linear regression slope method (Fietzke et al.
2008). Ca and Si were used for internal standardization uti-
lizing data from EMP analyses.
Glass ­H2O and ­CO2 concentrations were determined
using a Fourier transformation infrared (FTIR) Bruker IFS88
spectrometer coupled with an IR Scope II microscope at
the Institute of Mineralogy, Leibniz University of Hanno-
ver. ­H2O and C­ O2 concentrations were measured on doubly
polished glass chips (~ 1 mm2 in size) that had a thickness
of 150–50 μm, measured with a digital micrometer Mitutoyo
[precision ± 2 μm (Behrens et al. 2009)] by three points
for every polished glass fragment. Absorption spectra in
the mid-infrared (MIR) range were collected using a spot
size of 100 × 100 μm (average of 50 scans) and using the
following operating conditions: globar light source, KBr
beam splitter, MCT (HgCdTe) detector, 4 cm−1 spectral
resolution, spectral range 13,000–0 cm−1. The H ­ 2O concen-
tration was measured at the peak that is attributed to the
OH stretch vibration (3550 cm−1) using a molar absorption
coefficient of 67 L cm−1 mol−1 (Stolper 1982). The C ­ O2 con-
centration was measured at the peak doublet with maxima at
1430 and 1510 cm−1 using molar absorption coefficient of
317 L cm−1 mol−1 (Shishkina et al. 2010). Glass densities
were calculated using the equation of Yamashita et al. (1997)
for basaltic compositions: ρ = 2819 − 20.8·H2O concen-
tration. The ­H2O and ­CO2 concentrations were calculated
based on the Lambert–Beer law using the peak height, which
was determined by reference to a straight tangential base
line. The average values of three measurements per sample
were used to calculate the H ­ 2O and C
­ O2 contents of the
glasses, with a standard deviation usually less than 0.02 and
0.003 wt%, respectively.
Pressure and temperature conditions at the point of last
crystallization equilibrium were calculated by two differ-
ent methods. Pressures for all samples were calculated with
the software COMAGMAT version 3.57 that is based on
empirical models calibrated with an experimental data-
base (Ariskin and Barmina 2004). Here we used inversed
modelling (see Almeev et al. 2008) of the major elements
and ­H2O concentrations of the glasses. The effects of small
amounts of ­H2O on crystallization were taken into account
by incorporating recent experimental calibration coeffi-
cients (Almeev et al. 2007a, b, 2012). The pressure models
of COMAGMAT have an uncertainty of ± 100 MPa and are
comparable with models used in Michael and Cornell (1998)
(Almeev et al. 2008). For the samples that contain Cpx, pres-
sures were also calculated using equilibrium exchange mod-
els between Cpx and the melt (Putirka et al. 1996; Putirka
2008), determined by analysing the average compositions
of Cpx-rims and their host glasses. Mid-ocean ridge basalts
have typically an ­Fe3+ content of 10–12% (Bézos and Hum-
ler 2005; Berry et al. 2017); for the calculations here we
applied a ­Fe3+ content of 12% (Bézos and Humler 2005),
but variations within these values do not affect the pres-
sure calculations. The standard error of estimate (SEE) of
the accuracy of the Cpx–melt model is ± 150 MPa (for an
extended evaluation of involved errors see Putirka 2008).
Results
Mineral textures and chemistry
Phenocrysts in the SMAR lavas are up to 2 mm in size and
mostly comprise olivine (Ol) and plagioclase (Pl), with
minor clinopyroxene (Cpx) observed in only five of the
samples. In samples that cover all different chemical com-
positions Ol and Pl show a range of textures from euhedral
crystals to anhedral grains displaying resorption at the min-
eral rims and occasionally sieve textures (Fig. 2a–e). A few
of the Ol grains display undulose extinction (Fig. 2f). Melt
inclusions are present in both Ol and Pl and can be up to
~ 200 µm in diameter; in the Pl they are frequently located
along growth zones, but also along healed cracks. Olivine
and Pl crystals often show clustering and intergrowth of
multiple minerals that probably represent cumulate frag-
ments (Fig. 2c–h).
13
 97   Page 6 of 23 Contrib Mineral Petrol (2017) 172:97

Fig. 2  Photomicrographs
of thin sections from SMAR
samples 149-1 (a, c–f, h), 142-7
(b) and 145-3 (g) in cross-
polarized light (a–d, f–h) and
plane-polarized light (e). a, b,
h Euhedral Ol occurs together
with highly dissolved and skel-
etal Pl and/or Ol. c–h Clusters
of multiple Pl (c), or Pl and Ol
(d–h) with complex intergrowth
structure of Pl and Ol (e, g) and
undulose extinction in Ol (f)
indicate xenocrystic fragments
that formed by in situ crustal
growth of minerals together and
subsequent assimilation of these
fragments

13
Contrib Mineral Petrol (2017) 172:97 Page 7 of 23  97

Fig. 3  Mineral composition data. a Transects over Ol grains from ▸

rim to core show unzoned Ol together with normally and slightly
reversely zoned Ol. b Fo-content of the Ol plotted against the Fo-con-
tent of Ol that would be in equilibrium with their host glasses (apply-
ing the equilibrium constant of 0.3  ±  0.03 from Roeder and Emslie
1970). Olivines in equilibrium fall onto the one-to-one line or within
the red field that represents the variation of the equilibrium constant.
c Transects over Cpx crystals show various zonation patterns from
limited zoned to complexly zoned crystals. d Plagioclase transects
display scatter showing oscillatory zoning, but display a generally
reversely zoned trend (147-2) and unzoned trend (143-2)

The forsterite (Fo) contents in Ol vary between 83.4 and

91.8%, with intra-sample variations up to 5.5% both between
different grains and within single Ol grains. Chemical pro-
files from core to rim over several selected Ol reveal (1)
crystals with a constant composition, (2) normally zoned
crystals, of which some may include diffusion profiles
(e.g. in 143-2 and 149-1) and (3) crystals with a slightly
reverse zonation (e.g. in 143-2, 147-2; Fig. 3a). Several Ol
rim compositions are too Fo-rich to be in equilibrium with
their host melts (Fig. 3b; using the equilibrium constant of
0.30 ± 0.03 from Roeder and Emslie (1970) and 12% F ­ e3+ in
the glasses (Bézos and Humler 2005) points to equilibrium
at Fo-contents of 79.4–89.6%) implying a xenocrystic origin
for these Ol. Also here applying a slightly different ­Fe3+
content does not affect this observation. Olivines in disequi-
librium are found for both incompatible element enriched
and depleted samples. The rare Cpx that occur in samples
147-2, 149-1, 149-4, 162-1 and 162-2 have a composition
of ­Wo36–41En51–55Fs6–10 and rim to core profiles display
zoning in some cases (Fig. 3c). Two profiles over Pl show
oscillatory zoning around a constant composition or slight
reversed-zoning (Fig. 3d).

Composition of the glasses

The major and trace elements together with the Cl data

are given in Online Resources 2 (SMAR) and 3 (Gak-
kel Ridge). The Cl concentration values range from 43
to 704 ppm for Gakkel Ridge and 52–429 ppm for the
SMAR (± < 3 ppm 2S.E.), although most values lie below
200 ppm (Fig. 4a). This range is similar to literature data
for other slow-spreading ridges (Michael and Cornell
1998; PetDB, Lehnert et  al. 2000; Jenner and O’Neill
2012). In the majority of both SMAR and Gakkel sam-
ples there is a negative correlation between Cl and MgO
(Fig. 4a). This trend is, however, poor (R2 = 0.22; n = 117)
and not parallel to a single fractionation model [(log(Cl8/
Cl) = 0.12·(MgO-8) from Weaver and Langmuir (1990),
assuming unfractionated Cl concentrations of 50, 100 or
150 ppm at 8 wt% MgO; Fig. 4a]. Several Cl-rich samples correlations are seen between Cl and several incompatible
from both SMAR and Gakkel Ridge lie well above the elements, e.g. K, Rb, Ba, Th, U, Nb, Ta, La, Ce, Pr, Nd
values that could be explained by fractionation. Positive and ­H 2O, with the best correlation existing between Cl

13
 97   Page 8 of 23 Contrib Mineral Petrol (2017) 172:97

Fig. 4  Chlorine composi-
tions of SMAR and Gakkel
Ridge basalt glasses compared
to major and trace element
compositions. The analytical
error is smaller than the symbol
sizes. a Cl vs. MgO compared
to fractionation trends for Cl
compositions of 50, 100 or
150 ppm at 8 wt% MgO (black
lines). Cl vs. Nb for b Gakkel
Ridge and c SMAR samples
together with lines of various
Cl/Nb ratios and compared to
samples from the Red Sea (van
der Zwan et al. 2015). The local
mantle Cl/Nb ratio (see “Dis-
cussion”) for the two locations
is highlighted. Note that many
samples lie above these mantle
values. d Chlorine concentra-
tions are related to radiogenic
isotope ratios (i.e. 87Sr/86Sr;
Hoernle et al. 2011). e ­H2O and
Ce contents are relatively well
correlated. f Cl/K and Cl/Nb are
correlated for each sample suite,
but display different trends
between the suites. g Variations
in Nb/K between the samples
suites are corresponding to Ce/
Pb

and Nb (R2 = 0.87; Fig. 4b, c), the element whose mantle
incompatibility is most similar to Cl (Sun et al. 2007).
Samples from the SMAR also show a correlation between
Cl and radiogenic Sr, Nd and Pb isotopes (isotope data
from Hoernle et  al. (2011); Fig.  4d; Table  1). No cor-
relation (R2 < 0.2) is observed between Cl and indicators
of degassing (e.g. C
­ O2/Nb, ­CO2/Ba, cf. Saal et al. 2002;
Michael and Graham 2015), melting degrees (e.g. ­Na8, cf.
Klein and Langmuir 1987) or continental input (e.g. Ce/
Pb, U/Nb, cf. Hofmann et al. 1986). A good correlation
between ­H2O and Ce (Fig. 4e) suggests that the magmas
have not lost or gained significant amounts of water since
leaving the melting zone (cf. Michael 1995).
To study Cl addition to the magmas, we examined the
ratios of Cl to elements of similar incompatibility; wher-
ever possible we used Nb (Cl/Nb) as Nb is slightly closer
in incompatibility to Cl than K (Sun et al. 2007) and is
more resistant to the effects of weathering and alteration
processes; for analytical reasons Cl/K is used for the melt
inclusions, and to compare our data to older studies, where
mostly only major elements were measured. Cl/K and Cl/
Nb are positively correlated for each study area although

13
Contrib Mineral Petrol (2017) 172:97 Page 9 of 23  97

Table 1  Pressure calculations Sample Cl/Nb Pressure Putirka et al. KD (Fe–Mg) Pressure COMAG- ∂ Pressure
based on the formulations for (1996) (MPa) MAT (MPa)
Cpx–melt equilibria compared
to COMAGMAT pressure 147-2_1 47.9 361 0.25 670 − 309
calculations
147-2_2 47.9 316 0.24 670 − 354
147-2_3 47.9 324 0.23 670 − 346
149-1_1a 32.8 − 13 0.12 5 − 18
149-1_2a 32.8 53 0.12 5 48
149-4 33.3 375 0.17 5 370
162-1_1 18.3 387 0.20 80 307
162-1_2 18.3 403 0.20 80 323
162-1_3 18.3 373 0.20 80 293
162-2_1 18.1 340 0.24 10 330
162-2_2 18.1 364 0.25 10 354
a
 The low/negative pressures together with the low KD (Fe–Mg) indicate disequilibrium and geologically
non-significant pressures for this sample

each area lies on a different subparallel trend (Fig.  4f).
These trends are consistent with source-derived variations
in Nb/K between the different sample suites as they also
show a grouping in their Ce/Pb and Nb/U contents (Fig. 4g).
SMAR samples reach a maximum Cl/K of 0.3 and Cl/Nb
of 64, higher than at most other slow-spreading ridges, but
lower than has been observed at fast-spreading ridges or in
the Red Sea (Fig. 5). The Gakkel Ridge samples, in contrast,
all lie within the previously defined slow-spreading ridge
field (Michael and Cornell 1998; Fig. 5). High Cl/K ratios
at SMAR mostly occur in trace element depleted samples
(e.g. samples with low K/Ti, Fig. 5a), consistent with the
negative relation observed between Cl/Nb and incompat-
ible trace elements (e.g. Nb, K, Rb, REE; Fig. 5c). Samples
from Gakkel Ridge show a range of Cl/Nb values across
the whole range of trace element concentrations (Fig. 5b).
High Cl/Nb is observed in both segments of the SMAR and
in the WVZ and SMZ of the Gakkel Ridge. The EVZ of the
Gakkel Ridge in contrast, displays comparatively few high
Cl/Nb samples.
High‑precision chlorine measurements on SMAR melt
inclusions
High-precision Cl and K concentrations were measured in
melt inclusions in both Ol and Pl crystals in selected high Cl/
Nb (28–48) samples from the SMAR (137-3, 143-1, 143-2,
147-2, 147-3, 149-1, 149-4 and 151-3; Online Resource 4).
The SMAR melt inclusions display a generally positive trend
between Cl concentrations (20–168 ppm in Ol, 30–234 ppm
in Pl) and K (67–2256 ppm), but with more scatter towards
higher Cl-values than observed in the bulk glasses (Fig. 6a).
This scatter is not concentration-dependent, and is consider-
ably larger than analytical error. Cl/K ratios in the inclusions
(0.08–0.56) extend mainly to higher values than observed
in the host glasses (0.12–0.19). For individual samples,
Cl, K and Cl/K of the melt inclusions vary to both higher
and lower values than their host glass (stars in Fig. 6b). An
exception are the trace element enriched samples from sta-
tion 149, which display a larger variation of the melt inclu-
sions in Cl/K, but both Cl and K are considerably lower than
the host glass and overlap with the more depleted samples
(Fig. 6c). The highest Cl/K values in the melt inclusions are
observed in 143-1 and 143-2 that also have the highest Cl/K
of their host glasses (0.19).
Pressures of equilibrium crystallization
Previous work (Almeev et al. 2008) had used the experi-
mentally calibrated COMAGMAT program to determine the
pressures at which the SMAR magmas were last in equilib-
rium with the three-phase assemblage Ol–Pl–Cpx. Applying
COMAGMAT with improved experimental calibration coef-
ficients for the effect of water (see method section) to our
samples yields pressures of last equilibrium crystallization
between 5 and 700 MPa for the SMAR and between 100 and
840 MPa for the Gakkel Ridge (Online Resources 2, 3). The
average pressures are 357 and 426 MPa, respectively, similar
to values found by Michael and Cornell (1998) for slow-
spreading ridges which did not display high Cl/K. SMAR
segment A1 has slightly higher calculated crystallization
pressures than the shallower segment A2 with on average
385 and 322 MPa, respectively. At Gakkel Ridge, the high-
est average pressures are found in the Eastern Volcanic Zone
(538 MPa), while the Western Volcanic Zone and Sparsely
Magmatic Zone display lower pressures (340 and 375 MPa,
respectively).
For the rare samples that contain Cpx (SMAR 147-2,
149-1, 149-4, 162-1, 162-2), we also calculated pressures
(given in Table  1) using the formulations for Cpx–melt

13
 97   Page 10 of 23
Fig. 5  Indicators of hydrothermal Cl addition (Cl/Nb, Cl/K) in basal-
tic glass plotted against major and trace element data. The analytical
error is smaller than the symbol sizes. a Cl/K ratios reach higher val-
ues than previously observed for average slow-spreading ridges, but
are not as high as observed for Red Sea Rift basalts or fast-spread-
ing ridges. High Cl/K (>  0.1) is not observed for samples with K/
Ti  >  0.22. Numerous b Gakkel Ridge and c SMAR samples have
Cl/Nb above the Cl/Nb mantle ratio. Samples from Gakkel Ridge
show high Cl/Nb over the complete range of Nb; at the SMAR this is
mainly for samples with low Nb
equilibrium of Putirka et  al. (1996). Melt–phenocryst
equilibrium was assessed by comparing the predicted and
observed values for the Cpx components (cf. Putirka 2008)
13
Contrib Mineral Petrol (2017) 172:97
Fig. 6  Melt inclusion Cl and K data. a Melt inclusions in Ol (dia-
monds) and Pl (circles) show a similar trend in Cl and K as the
SMAR glasses, but with larger scatter. There is no systematic dif-
ference between melt inclusions in Ol or Pl. b Cl/K vs. K for melt
inclusions (filled symbols) together with their host glasses (stars; col-
ours distinguish samples). Melt inclusions in minerals hosted in trace
element depleted glasses display both lower and higher Cl/K and K
than their host glasses. Note that the most K-depleted inclusions show
extremely high Cl/K ratios, exceeding the range in their host glasses.
c Melt inclusions from trace element-enriched (high K) glass sample
149-1 have lower Cl and K than their host and overlap with concen-
trations of melt inclusions in trace element depleted samples (b)
Contrib Mineral Petrol (2017) 172:97 Page 11 of 23  97
and by comparing the observed and predicted values for KD
(Fe–Mg)cpx–liq (Online Resource 5; Table 1). In terms of
the Cpx components, the calculated and observed composi-
tions were within 5% and indicate equilibrium in all sam-
ples. Samples 147-2 and 162-2 also show equilibrium in
KD (Fe–Mg)cpx–liq. Samples 149-4 and 162-1 display minor
[KD(Fe–Mg)cpx–liq = 0.17–0.20] and sample 149-1 major  et al. 2006; le Roux et al. 2006), while values derived from
[KD(Fe–Mg)cpx–liq = 0.12] disequilibrium in Fe and Mg of
the Cpx with the glass, indicating some late-stage modi-
fication of the glass for these fast-diffusing elements. As
the pressure calculations are dependent on variations in the
Cpx components, which involve slower diffusing elements,
this minor disequilibrium in Fe–Mg for samples 149-4 and
162-1 has no significant effect on the resulting pressure cal-
culations. The disequilibrium effect is significant, however,
for sample 149-1 so we have excluded it from the subse-
quent discussion. The four samples in (close) equilibrium
yield pressures between 300 and 400 MPa, which are not
consistent with the pressures calculated by COMAGMAT
(Table 1). This will be discussed further below.
Discussion
Chlorine contents and Cl/Nb values of the mantle
and magma
In order to identify Cl addition to the magmas from external
processes, it is necessary to determine the possible range in
local mantle Cl contents (and hence its probable concentra-
tions in mantle-derived magmas). The correlation between
Cl and many incompatible trace elements (particularly those
that have a compatibility close to Cl, for example Nb) and
to Sr-, Nd- and Pb-radiogenic isotopes in the case of the
SMAR (Fig.  4b–d) indicates that much of the variation
in absolute Cl concentrations between the samples can be
explained by source processes. Mantle source variations at
the SMAR were described by, e.g. Almeev et al. (2008) and
Hoernle et al. (2011) and were attributed to mixing between
a depleted mantle source (with low Cl, Nb and K concentra-
tions) and several enriched (HIMU) mantle sources (with
higher Cl, Nb and K concentrations), while variations in
mantle source compositions at the Gakkel Ridge were identi-
fied by, e.g. Michael et al. (2003) and Goldstein et al. (2008).
We see from Fig. 4f, g that each individual axial locality has
its own inherent Nb/K signature and, by inference, Cl/Nb
and Cl/K mantle signatures. If we can define these mantle
ratios for the individual regions, then Cl/Nb and/or Cl/K
above these values should indicate Cl addition.
Cl/Nb (and Cl/K) mantle ratios are not well constrained
and cover a large range in the literature: Palme and O’Neill
(2003) suggest the highest Cl/Nb of 51, McDonough and
Sun (1995) propose 25.8, le Roux et al. (2006) favour 14,
Salters and Stracke (2004) choose 6.9 and Saal et al. (2002)
give a ratio of 3. However, Palme and O’Neill (2003) cal-
culate a relatively high Cl mantle value (30 ppm) compared
to values that were derived from basaltic melts and melt
inclusions, which are all lower than 7 ppm (Ryabchikov
2001; Saal et al. 2002; Salters and Stracke 2004; Kovalenko
mantle minerals are even lower < 0.38 ppm (Urann et al.
2017). If we assume mantle concentrations of 0.3–0.6 ppm
for Nb (Hofmann 1988; McDonough and Sun 1995; Palme
and O’Neill 2003) and a Cl concentration of < 7 ppm, then
mantle Cl/Nb ratios should lie at least below the value of
25.8 proposed by McDonough and Sun (1995). To deter-
mine in more detail the local Cl/Nb in our study areas we
can use the variations we see in our samples. Assuming (a)
no fractionation of Cl/Nb during melting (which can be
excluded due to their similar distribution coefficients) and
(b) no degassing of Cl from the magma prior to eruption
(see below) then interactions of the magma with its environ-
ment should always lead to Cl/Nb increases. In this case, the
magmas with the lowest Cl/Nb represent the least contami-
nated samples and their Cl/Nb should be closest to the local
(upper limit) mantle Cl/Nb ratio. The evidence for a mix of
sources at both locations may imply that the Cl/Nb and Cl/K
mantle ratios are not constant within an area, which may be
particularly the case for the SMAR ridge samples as HIMU
sources may have higher Cl/K values (Stroncik and Haase
2004; Kendrick et al. 2017). However, Fig. 5b, c, shows a
very consistent lower Cl/Nb ratio in our samples over a large
range of absolute Nb contents at both locations. At Gakkel
Ridge these values span the full range of compositions and
a constant Cl/Nb mantle value is confirmed by the fact that
we see, for example, no correlations between Cl/Nb and any
trace element ratios indicating variations in continental input
(e.g. Ce/Pb, U/Nb, cf. Hofmann et al. 1986) at the Gakkel
Ridge.
In the SMAR samples we see a negative correlation
between the lowest Cl/Nb values and trace element con-
centrations as all depleted SMAR basalts (Nb < 4.5) show
higher Cl/Nb ratios (significantly above analytical error, see
“Analytical methods”), while SMAR samples with low Cl/
Nb and Nb > 4.5 ppm fall on a line with Cl/Nb of 20 ± 2
(Fig. 5b). We cannot rule out that this results from mixing
of a magma from a low Nb—high Cl/Nb (~ 50) source with
magma from a trace element enriched, low Cl/Nb source,
but this appears unlikely for the following reasons: (a) we
see a range of Cl/Nb values for similar trace element (Nb)
concentrations that cannot be explained by mantle varia-
tions; (b) we see no positive correlation between Cl/Nb and
other indicators of mantle enrichment (e.g. Zr/Hf; Nb/Zr);
(c) depleted mantle sources have been shown to have lower
Cl/K (and by implication, Cl/Nb) than enriched, particularly
HIMU, mantle sources (Michael and Cornell 1998; Stroncik
13
 97   Page 12 of 23 Contrib Mineral Petrol (2017) 172:97

and Haase 2004; Kendrick et al. 2017). Therefore, if the
depleted mantle source (low Nb samples) has a different
Cl/Nb ratio, this is expected to be lower than the Cl/Nb of
20 ± 2 indicated by the HIMU source samples (high Nb
values). Instead, the higher Cl/Nb for all SMAR samples
with Nb < 4.5 ppm, more likely can be explained by their
inherently higher sensitivity to Cl contamination due to their
low magmatic Cl (the effect of adding a constant amount of
excess Cl to magmas with different Nb concentrations are
shown as dashed lines on Fig. 5b, c; note that the smooth-
ness of the curves observed are an artefact of plotting the
ratio of an element against the same element). We thus take
these values to indicate the local mantle Cl/Nb ratios (ca.
20 ± 2 at SMAR, 18.5 ± 1.5 at Gakkel). All samples with
Cl/Nb above these local mantle Cl/Nb ratios are taken to
show Cl-excess (defined as Cl content above the expected
Cl in the melts based on their Nb content and the local Cl/
Nb mantle ratio).
Using the correlation regression shown in Fig. 4f, Cl/Nb
mantle ratios of 18.5 ± 1.5 and 20 ± 2 for the Gakkel Ridge
and SMAR, respectively, correspond to Cl/K mantle ratios
of 0.040 and 0.086 (Fig. 6b, c), which is consistent with
typical Cl/K values of 0.02–0.04 and 0.04–0.08 determined
by Stroncik and Haase (2004) for EM (Gakkel Ridge) and
HIMU (SMAR) suites.
Using these Cl/Nb mantle ratios, we can calculate Cl-
excess in ppm for each sample based on its Nb or K content
(Fig. 7a, b). In contrast to the absolute Cl/Nb ratios, Cl-
excess is independent of the trace element (e.g. Nb) con-
centrations of the samples, and shows only the addition of
Cl to magmas derived from any particular mantle source.
Cl-excess is up to 250 ppm, and ~ 75% of the measured sam-
ples from both the Gakkel Ridge and SMAR show Cl addi-
tion (Online Resources 2, 3) over a range of compositions,
spreading rates and both comparatively thick and thin crust
(Fig. 7). The amount of Cl-excess is similar for the SMAR
and Gakkel Ridge, apart from samples from the EVZ that
show significantly less Cl-excess (Fig. 7a). The SMAR melt
inclusions show the same extent of Cl-excess as the glasses
(Fig. 7b). Both ridges show considerably lower Cl-excess
than basalts from the slow-spreading Red Sea (< 1400 ppm;
Fig. 7a; cf. van der Zwan et al. 2015).

Fig. 7  Cl-excess (ppm) calcu-

lated by the Cl concentrations
minus the magmatic Cl based
on Cl/Nb mantle values plotted
against major and trace element
data. a Cl-excess is present in
the glasses over various Nb
concentrations and is up to
250 ppm. b The melt inclusions
show the same of Cl-excess as
the glasses. c Cl-excess is not
related to MgO (fractionation),
d ­Na8 (melt degree), e 87Sr/86Sr
(source variation) or f ­CO2
(degassing)

13
Contrib Mineral Petrol (2017) 172:97 Page 13 of 23  97
Chlorine excess and magmatic, seafloor
and hydrothermal processes at SMAR and Gakkel
Ridge
Magmatic processes such as fractional crystallization and
partial melting as well as source variations influence the
absolute Cl contents of the melts, but do not affect the Cl/
Nb ratios and hence cannot cause the observed Cl-excess.
This is additionally confirmed by the lack of a correlation
between Cl-excess and indicators of fractionation (MgO)
or degree of melting (­ Na8) and source variations (87Sr/86Sr;
Fig. 7c–e). Degassing of the samples is indicated by the
­CO2/Nb of the samples of mostly < 200 and the C ­ O2/Ba
of < 60 that are significantly lower than values of 239 or
283 for C­ O2/Nb and 105 for C­ O2/Ba for undegassed magma
derived by Saal et  al. (2002) and Michael and Graham
(2015). Degassing of ­CO2 can also be seen in a plot of C ­ O2
against ­H2O compared to isobars of equilibration eruption
pressures where we see degassing trends for ­CO2 towards the
eruption pressures of the samples (Online Resource 6). ­H2O
shows, however, no degassing trends and is undersaturated
compared to the isobars, implying no ­H2O degassing took
place, consistent with the good ­H2O/Ce correlation. Also
variations in Cl due to Cl removal by degassing can be ruled
out as there is no correlation between C ­ O2/Nb or ­CO2/Ba
and Cl-excess (Fig. 7f).
As melt inclusions display the same Cl-excess as the
glasses (Figs. 5a, 6, 7a, b) and Cl-excess is homogeneous
across multiple glass chips of the same sample and also not
correlated to indicators of surficial alteration e.g. U, Ba (e.g.
Alt et al. 1986; Alt and Teagle 2003; Schramm et al. 2005;
Augustin et al. 2008), Cl addition by surficial processes (sea-
floor alteration or syn-eruptive magma–seawater interaction,
e.g. Hart et al. 1974; Soule et al. 2006) can be ruled out
as playing a significant role in generating the Cl-excess we
observe. Instead, the source of the Cl-enrichment must lie
deeper in the magmatic system, before eruption but during
crystal growth and magma homogenization.
The assimilation of hydrothermally altered oceanic crust
was shown to lead to Cl-excess in glasses at other locations
(cf. Bédard 1991; Bédard and Hébert 1996). Furthermore,
the global lack of Cpx in most MORB magma despite a
positive correlation between Ca and Mg has been explained
by assimilation of (lower) crustal rocks (e.g. Kvassnes and
Grove 2008; Lissenberg and Dick 2008; Godard et al. 2009;
Drouin et al. 2009) and melt–rock reactions (syntexis; Béd-
ard et al. 2000 and references therein; Kelemen et al. 1995).
The petrology of our SMAR samples shows that they have a
complex history of melt/rock interactions, with phenocryst
rim/glass disequilibrium (Fig. 3b), complex or reverse zon-
ing and dissolution features (Fig. 2) all suggesting magma
mixing and xenocryst inheritance. The presence of depleted
melt inclusions in phenocrysts from enriched magmas
(Fig. 6c) further underlines their xenocrystic origin and the
late addition of the enriched component to the system before
eruption (Fig. 6c). The occurrence of rock fragments identi-
fied by deformation textures in the SMAR samples (Figs. 2,
3) suggests that the magmas assimilated oceanic crust or
partially solidified crystal mushes. Evidence for assimila-
tion of country rocks at Gakkel comes from U-series data
(Elkins et al. 2014).
To determine whether the Cl-excess in our glasses comes
from assimilation of country rock during ascent we can use
the ­H2O/Cl ratios of the samples, as this ratio is different in
seawater, altered rock and hydrothermal brines. When we
plot ­H2O/Cl against K/Cl (following Kendrick et al. 2013;
Fig. 8), we see that our samples follow a mixing line from
mantle values to a contaminant with low ­H2O/Cl (< 9). Both
hydrothermally altered lithosphere and seawater have high
­H2O/Cl ratios (> 16 and ~ 50; see Ito et al. 1983; Bach
et al. 2003; Sano et al. 2008; Barnes and Cisneros 2012)
and would introduce H ­ 2O to the magma. The good cor-
relation between ­H2O and Ce (maximum 0.22 wt% varia-
tion for a given Ce content per sample suite; Fig. 4e), the
undersaturation of the magmas in H ­ 2O (Online Resource
6), and H ­ 2O/Ce values of < 350, all indicate no significant
­H2O addition (see also Michael 1995; le Roux et al. 2006;
Wanless et al. 2010, 2011; van der Zwan et al. 2015). The
Cl-excess we observe cannot, therefore, be explained by the
bulk addition of either seawater or bulk altered lithosphere.
A Cl-rich phase (that adds up to 250 ppm Cl) with low H ­ 2O/
Cl can be introduced by assimilation of 0.25–0.08% fluids
with NaCl contents of 16 to > 50%, respectively (brines, e.g.
Michael and Schilling 1989; Kendrick et al. 2013; van der
Zwan et al. 2015; Fig. 8). Such fluids are generated, particu-
larly at high temperatures and pressures, by phase separa-
tion of hydrothermal fluids (Bischoff and Rosenbauer 1987;
Fig. 8  The relation of the K/Cl and the H­ 2O/Cl contents of the sam-
ples compared to their mantle values and potential contaminants
(seawater, altered oceanic crust, brines). Salinity of the brines with
10–50 wt% salts and mantle data are after Kendrick et al. (2017), but
with the local K/Cl mantle values as derived in this paper. Altered
oceanic crust data are from Ito et al. (1983), Bach et al. (2003), Sano
et al. (2008), Barnes and Cisneros (2012). See text for discussion
13
 97   Page 14 of 23
Fournier 1987; Berndt and Seyfried 1990). Hydrothermal
brines could potentially be added directly to the magmas
if fluids intrude through the conductive cracking front of
the magmas (e.g. Kendrick et al. 2013); the petrographic
evidence for assimilation of country rocks in both sample
sets argues, however, against such a simple incorporation
of fluids. Assimilation of hydrothermally altered rocks with
brines present in the pore-spaces (Coogan et al. 2003) or as
fluid inclusions in hydrothermally formed minerals, would
lead to Cl-excess but little H
­ 2O addition. This mechanism for
Cl-contamination has also been suggested for samples at the
boundary between the magmatic and hydrothermal systems
from IODP Site U1256 (fast-spreading EPR), where Cl was
found to be present in fluid inclusions or pore fluids (Zhang
et al. 2017) and stoping of altered crust was thought to be the
dominant Cl enrichment process (Fischer et al. 2016). Alter-
natively, Cl could be added by partial melting (anatexis) of
the hydrothermally altered lithosphere and particularly by
the breakdown of Cl-rich hydrothermally formed minerals
such as amphiboles (Barnes and Cisneros 2012); these min-
erals melt preferentially due to a lower solidus than the fresh
rock, and partition Cl much more strongly than ­H2O into the
melt, producing a low ­H2O/Cl contaminant (Michael and
Schilling 1989; Kent et al. 1999; France et al. 2010, 2014;
Wanless et al. 2010, 2011). This process has been proposed
for some basaltic to dacite lavas on the EPR based on their
major and trace element data (France et al. 2010, 2014; Wan-
less et al. 2010, 2011) although it was excluded for the rocks
from IODP Site U1256 which apparently reacted under dry
conditions without the influence of amphibole (Fischer et al.
2016; Erdmann et al. 2017). Since there is no correlation in
our samples between Cl-excess and any of the other meas-
ured elements (Fig. 7; Online Resources 2, 3) that could be
indicative for any of these processes, we cannot distinguish,
with our data, between the assimilation of altered country
rock containing interstitial brines or brine in fluid inclusions
and partial melt of (potentially amphibole-bearing) altered
lithosphere. We therefore group all these potential deep Cl
contamination processes under the term ‘assimilation of
hydrothermally altered lithosphere’.
Crystallization depths and the validity of the pressure
calculations
The discrepancies between the pressure of last equilibrium
calculated using COMAGMAT (0–840  MPa) and those
derived from the Putirka et al. (1996) Cpx–melt equilib-
rium model (300–400 MPa) indicate that at least one type
of pressure estimate is inaccurate. The COMAGMAT pres-
sures are representative for melt-based pressure calculations
(see Online Resource 7 for discussion), but the assumption,
inherent in all phase-based pressure calculations, of a sim-
ple crystallization history and equilibrium at the Ol–Pl–Cpx
13
Contrib Mineral Petrol (2017) 172:97
cotectic as the defining final event in the magmatic evolution
is probably incorrect. Petrographically and geochemically
our samples show signs of mineral–melt chemical disequi-
librium (Fig. 3), assimilation of lithic xenoliths (Fig. 2)
and magma mixing (Figs. 3, 6). Moreover, the lack of Cpx
phenocrysts in most of the samples is inconsistent with the
assumption that the melts are at a multiply-saturated cotectic
with Ol, Pl and Cpx.
Melt–rock reactions in the lower lithosphere may signifi-
cantly change the compositions of both the cumulates (e.g.
Meyer et al. 1989; Bédard 1991; Bédard and Hébert 1996;
Bédard et al. 2000; Coogan et al. 2000; Dick et al. 2002;
Ridley et al. 2006; Grimes et al. 2008) and the melt. As
we have no precise information on the original melt and
contaminant compositions, the extent of assimilation can-
not be determined straightforwardly by backward-modelling
although Kvassnes and Grove (2008) and Lissenberg and
Dick (2008) have shown, using experiments and forward
modelling, that assimilation of various types of oceanic
crust has a significant effect on the composition of the melts
and their ­Al2O3–CaO–Si–MgO relations. They concluded
that any method that uses the composition of a melt (glass)
alone to determine, e.g. crystallization pressure or temper-
ature from a melt can be erroneous. Lissenberg and Dick
(2008) further showed that calculated pressures increase
with increasing assimilation of lower crustal troctolite (up
to 25%) and that the full range of crystallization pressures
observed in MORB (100–800 MPa) can be achieved by reac-
tion of melts with the lower crust. This is consistent with the
basalt data from the SMAR, where many of the samples that
display high COMAGMAT pressures (> 300 MPa; Table 1)
show Cl-excess implying they have assimilated altered crust.
As all samples with Cl-excess likely have modified major
element compositions due to assimilation, pressure calcula-
tions based on major elements cannot be used to determine
the depth of Cl addition.
Pressure calculations based on Cpx–melt element
exchange (Putirka 2008) can give a better control on the
crystallization pressures, as the degree of equilibrium
between glass and Cpx can be assessed and, by using the
Cpx rim compositions, only equilibrium in the last phase of
crystallization is required, thus the pressures are not affected
by earlier magma modifications. A statistically robust
assessment of the relationship between Cl-excess and crys-
tallization pressures is hindered by the lack of Cpx in most
(MORB) samples, but our samples with Cpx in equilibrium
give geologically relevant depths.
The depth of assimilation of hydrothermally altered
lithosphere
The SMAR Cpx–melt pressures (after Putirka et al. 1996) all
lie in a relatively restricted range (316–403 ± 150 MPa ≈ 
Contrib Mineral Petrol (2017) 172:97 Page 15 of 23  97
10–13 ± 5 km; Table 1). For a crustal thickness of ~ 11 km
on segment A2 (Bruguier et al. 2003) where these samples
originate, these pressures imply that the last episode of crys-
tallization occurred at lower crustal depths or close to the
Moho. The Cl-enriched melt inclusions in Ol and Pl imply
an increase of Cl in the magma prior to crystallization and
inclusion formation. The complete equilibrium observed for
samples 147-2 and 162-2 between the Cpx phenocrysts and
melt show no substantial modification of the melt subse-
quent to Cpx crystallization. Thus Cpx crystallized together
or after the Ol and Pl that host the Cl-contaminated melt
inclusions. This places a strong limit on the minimum depth
of hydrothermal circulation in this region of ~ 10 km. We
note that not all samples that display lower crustal Cpx crys-
tallization pressures display Cl-excess, indicating that not all
assimilated crust is altered or that assimilation of crust is not
the only method of magma cooling at depth.
Also the investigated samples that show a slight modifica-
tion of the host melt after crystallization of the Cpx of the
faster diffusing elements by minor disequilibrium in Fe–Mg
between the Cpx and glass, but with Cpx components that
are in equilibrium, host melt inclusions with Cl-excess indi-
cating deep assimilation of brines (Table 1). Disequilibria
between melts and several Ol rims imply assimilation or
mixing processes shortly before eruption, as most Ol in
disequilibrium do not show any evidence of diffusive re-
equilibration that would have led to the formation of obvious
diffusion profiles within a few days at magmatic tempera-
tures (cf. Klügel 1998; Shaw and Klügel 2002). Even the one
observed diffusion profile with a diffusion length of < 50 µm
(sample 143-2; Fig. 3a), indicates a short residence time
of the Ol in the melt of maximum a few weeks (cf. Shaw
and Klügel 2002). The lack of diffusion profiles in the Ol
(Rutherford 2008) shows that the rising of magma was rela-
tively fast (days to a few weeks) and therefore Cl addition by
assimilation of crust is less likely to happen during ascent
in the upper crust. This is in agreement with what would be
expected at slow-spreading ridges, where the cooler upper
crust (Reid and Jackson 1981; Dick et al. 2003; Montési
and Behn 2007) requires that magma rises fast in order to
not crystallize.
Clinopyroxene phenocrysts were only found at SMAR
A2, where the anomalously thick crust probably favours
the early crystallization of clinopyroxene. At slow-spread-
ing ridges with typical or lower crustal thickness without
clinopyroxene phenocryst (SMAR segment A1 and Gak-
kel Ridge) crystallization and assimilation depths cannot
be directly determined. Nevertheless, the ranges in Cl, K,
Cl/K and Cl-excess observed in melt inclusions in indi-
vidual samples also from SMAR segment A1 imply that
they were formed from both different mantle melts and
melts which had experienced varying degrees of assimi-
lation before extensive crystallization (Figs. 6, 7b). As
many Ol grains are xenocrystic, they must have crystal-
lized from earlier batches of melt that already assimilated
Cl as shown by the Cl-excess of the melt inclusions in
the Ol. Hence, the various amounts of Cl-excess observed
in the glass and melt inclusions from one rock probably
represent multiple generations of hydrothermal alteration,
assimilation of crust and crystallization, rather than pro-
cesses occurring all to the same magma batch. Production
of the crust by multiple cycles of crystallization was also
suggested by Grimes et al. (2008) and Lissenberg et al.
(2009), who found a range of ages in gabbros from the
Mid-Atlantic Ridge.
We see very similar degrees of Cl-excess in all regions
studied and the homogeneous distribution of Cl in the bulk-
rock glasses from both the SMAR and Gakkel Ridge shows
extensive homogenization after assimilation, which requires
magma reservoir processes. Therefore, also at Gakkel Ridge,
Cl cannot simply have been added during ascent, imply-
ing that hydrothermal circulation must reach the levels of
magma pockets at depth. At the ultraslow-spreading Gakkel
Ridge, the thin lithosphere (Reid and Jackson 1981; Dick
et al. 2003; Montési and Behn 2007) will be inherently more
strongly cooled by conduction, which promotes deep crys-
tallization and so magma reservoirs are expected to form
deeper there (Rubin and Sinton 2007; Morgan and Chen
1993). Both deep pooling of melt (Shaw et al. 2010) as well
as melt–rock interactions at depths < 15 km (Elkins et al.
2014) have been suggested at Gakkel Ridge. This evidence
and the consistent level of Cl-excess between the ridges
implies that deep hydrothermal circulation is also occur-
ring there. Crystallization and crustal assimilation in magma
bodies in the lower lithosphere at (ultra)slow spreading
ridges is consistent with the evidence of gabbroic samples
and melt–rock reactions in the lower crust and mantle found
at the South-West Indian Ridge and the MAR 30°N (e.g.
Dick et al. 2000; Expedition Scientific Party 2005).
Deep circulation of water can be facilitated by deep faults-
pathways. Data for teleseismic earthquakes (1960 to present;
International Seismological Centre 2011) record events at
depths of > 10 km for both study areas. Although the depth
estimate of these earthquakes is not very precise (Engdahl
et al. 1998), the > 3 magnitude of these earthquakes (Fig. 1)
indicates that the faults causing them must be large and thus
have the potential to extent to larger depths. Also, micro-
seismicity on SMAR segment A1 down to 2–4 km within
the 5-km-thick crust, indicates that faults extend to lower
crustal depths (Grevemeyer et al. 2013). Deep faults, down
to the lower crust, were found in gabbros at slow-spreading
ridges, e.g. by Harper (1985) Cannat et al. (1991) and Mével
and Cannat (1991), while Bach et al. (2004) showed that
also serpentinites that may be present in the lithosphere of
ultraslow-spreading ridges can break and alter at depth. In
addition, Hasenclaver et al. (2014) showed that the physical
13
 97   Page 16 of 23 Contrib Mineral Petrol (2017) 172:97

process of deep hydrothermal circulation can be modelled
at greater depths.
In conclusion, hydrothermal circulation is not restricted
to the upper crust and the assimilation depth is therefore
not a limiting factor on Cl-addition to rising magmas as
suggested by Michael and Cornell (1998). The difference
between our conclusion and those of previous work is prob-
ably related to analytical precision, which needs to be high
to distinguish between small, but significant variations in
Cl/Nb and to identify appropriate Cl/Nb mantle values and
Cl-excess in slow-spreading-ridge magmas.
Chlorine excess in lavas and sites of hydrothermal
activity
To study the spatial relationship between Cl-excess in lavas
and hydrothermal vent sites, we compared the sampling
locations of basalts showing Cl-excess with those of known
active hydrothermal vent sites and indirect indications for
hydrothermal activity (e.g. plume signatures; Fig. 9). At both
the SMAR and Gakkel Ridge there are distinct areas with
multiple lavas containing Cl-excess. These areas are most
apparent on the SMAR (Fig. 9a), where, within 10 km of a
known hydrothermal site or plume, samples with Cl-excess
are always seen, which are also some of the highest val-
ues observed along the ridge. High Cl-excess is also seen
at the axial high of segment A2 (8°57′S) and the relatively
shallow area in the north of A1 (7°51′S; Fig. 9a). Axial
topographic highs and domes are volcanic structures with a
higher average potential to host hydrothermal systems due
to a higher magmatic activity, shown by detailed studies
on the relation between hydrothermal occurrences and rift
geomorphology (Fouquet 1997; Anderson et al. 2017). The
axial high of segment A2 was earlier interpreted to be a
remnant of past high magmatic activity (Devey et al. 2010)
and there are currently no signs of hydrothermal activity
(Devey et al. 2010; Haase et al. 2016). Therefore, we take
the high Cl-excess in basalts at the axial highs of A2 and in
the north of A1 to indicate that alteration of the crust took
place there previously by ancient, now inactive hydrothermal
systems, which makes these areas good targets for the search
for extinct hydrothermal deposits.
At Gakkel Ridge, samples collected within 10 km of the
Aurora hydrothermal field at 6°20′W and plume sites at
1°45′W and 37°E display high Cl-excess (> 25 ppm; Fig. 9b)
consistent with the observations at the SMAR. Plume sites
at 2°10′E, 7°30′E and 43°10′E display samples with high
Cl-excess within the larger vicinity of 25  km, which is
likely related to the highly uncertain locations for the origin
of most of the hydrothermal plumes (Baker et al. 2004).
The only exception among our samples are basalts that do
not exhibit Cl-excess collected < 10 km from the present-
day hydrothermal active site at 85°E (Fig. 8b; Baker et al.
2004; Pontbriand et al. 2012). But in detail, these sampled
glasses come from visibly older lava flows than the young
glassy basalts, believed to have erupted in 1999 that the pre-
sent hydrothermal venting was linked to (Pontbriand et al.
2012). We therefore suggest that the 85°E system is a young

Fig. 9  Locations of high
Cl-excess samples (diamonds)
and samples without Cl-excess
(circles) and hydrothermal sites
at the a SMAR and b Gakkel
Ridge. Note in the vicinity of a
known hydrothermal site there
are always samples taken with
elevated Cl-excess (> 25 ppm)
with the exception of 85°E at
Gakkel Ridge, where independ-
ent evidence suggests that the
hydrothermal activity is prob-
ably younger than the magmas
sampled (for discussion see
text). The axial high at SMAR
segment A2 (8°57′S) yielded
Cl-excess samples but lacks any
recent hydrothermal activity,
which suggests that the high Cl-
excess results from former, now
inactive hydrothermal activity

13
Contrib Mineral Petrol (2017) 172:97 Page 17 of 23  97
hydrothermal system, that was established after the eruption
of the here measured samples.
In summary, Cl-excess in erupted lavas always indicate
that they have interacted with altered crust implying active
or former hydrothermal circulation in the vicinity; the large
amount of samples with Cl-excess show that hydrothermal
alteration is widespread. Basalts with the highest Cl-excess
(> 25 ppm) are related to places that display active or fossil
hydrothermal venting in the vicinity and as such represents
a guide that can aid to identify MOR segments with hydro-
thermal deposits.
Differences in Cl contamination between slow‑
and fast‑spreading ridges
Our data show that Cl-contamination by assimilation of
hydrothermally altered lithosphere is not restricted to
Fig. 10  Chlorine concentrations at different ridge types. a Frequency
plots of Cl-contents and b Cl/K for slow-, medium- and fast-spread-
ing ridges. Slow-spreading ridges show on average lower Cl-contents
and Cl/K than medium- and fast-spreading ridges, nevertheless 23.9%
of the slow-spreading MORB show Cl/K above a upper mantle limit
fast-spreading ridges or ridges with a thick crust, contrary
to the conclusions of Michael and Cornell (1998), but also
occurs in several ultraslow- and slow-spreading ridges
that show a range of morphologies and thicknesses. Nev-
ertheless, differences in the amounts of Cl-contamination
between fast- and slow-spreading ridges are clearly evident.
Following Stroncik and Niedermann (2016) we assembled
Cl and Cl/K data from the PetDB database (Lehnert et al.
2000) and used them to assess degrees of Cl-excess globally.
Absolute MORB Cl-concentrations at slow-spreading ridges
are overall considerably lower (only 14% has > 200 ppm
Cl) than at fast- or medium-spreading ridges (36 and 31%
has > 200 ppm Cl; Fig. 10a). Using an upper value of 0.09
for mantle Cl/K implies that at least 25% of the samples at
slow-spreading ridges show Cl-excess (Fig. 10b), consider-
ably less than observed at fast- and medium-spreading ridges
(69 and 79%). The maximum Cl-excess on slow-spreading
value of 0.09 (dashed line). c Comparing samples with definite Cl-
excess with hydrothermal sites on the most studied, slow-spreading
Norther Mid-Atlantic Ridge (20°–65°N) shows that Cl-excess is pre-
sent along the whole ridge and that where data are present hydrother-
mal sites are mostly associated with Cl-excess samples close by
13
 97   Page 18 of 23
ridges in the PetDB database overlaps with our study
(< 550 ppm, but mostly < 250 ppm) and is clearly lower than
at fast- and medium-spreading ridges (mostly < 1000 ppm).
The fact that only ca. 25% of slow-spreading magmas in
PetDB are classified as showing Cl-excess compared to
our 75% estimate for Gakkel and SMAR shows that using
a global mantle Cl/K of 0.09 underestimates the samples
with Cl-excess at slow-spreading ridges and emphasizes the
need for both high-precision Cl determinations and good
estimates of local mantle Cl/K values in these inherently Cl-
poor magmas. Cl-excess in the PetDB data is not restricted
to specific sections of slow-spreading ridges, e.g. with thick
crust as suggested by Michael and Cornell (1998), as for
example can be seen from the well-studied slow-spreading
Northern Mid-Atlantic Ridge (Fig. 10c).
Hence, Cl addition by hydrothermal contamination is
present on all types of ridges, but shows a difference in the
amount of Cl-excess in basalts from the two end-member
type of ridges. While Cl-excess at fast-spreading ridges is
high and present in most samples (e.g. Michael and Cornell
1998; Saal et al. 2002; le Roux et al. 2006; Kendrick et al.
2013), Cl-excess at slow-spreading ridges is more variable
and significantly lower (Fig. 5a; Michael and Cornell 1998;
van der Zwan et al. 2015; this paper). The lower absolute
intensity of Cl-contamination in slow-spreading magmas
implies either a lower degree of assimilation and/or a lower
degree of lithospheric alteration. The less differentiated
nature of magmas from slow- compared to fast-spreading
ridges (higher average Mg#; Rubin and Sinton 2007) sug-
gest that the magmas at slow-spreading ridges on average
underwent less fractional crystallization and therefore
have lost less heat during passage through the lithosphere.
Whether that results in less extensive hydrothermal circula-
tion and alteration or less assimilation of country rock, the
net effect will be a lower average Cl addition into the slow-
spread magmas. A decrease of hydrothermal alteration with
depth (e.g. Dick et al. 2000; Coogan 2003) may also play a
role in the lower Cl-excesses at slow-spreading ridges. The
lower crustal crystallization depths shown here and deep
assimilation of less altered rock will result in comparatively
minor Cl addition compared to shallower magma chambers
with chronic hydrothermal systems at fast-spreading ridges
(< 3 km; le Roux et al. 2006; Kendrick et al. 2013) that
assimilated more strongly altered crust. If less cooled by
hydrothermal circulation, this deeper lithosphere needs to
conductively loose its heat, consistent with thermodynamic
models and inferred temperature gradients that indicate more
effective conductive cooling at slow-spreading ridges (Reid
and Jackson 1981; Niu and Hekinian 1997; Montési and
Behn 2007). Slow-spreading ridges display a much higher
compositional variety with heterogeneous amounts of altera-
tion (e.g. Dick et al. 2000) and magma bodies are probably
small and ephemeral, which implies that both assimilation
13
Contrib Mineral Petrol (2017) 172:97
and alteration processes are more localized, explaining
that not all assimilated rock is altered at slow-spreading
ridges leading to more variable amounts of Cl-excess in the
samples. The large permanent magma lenses beneath fast-
spreading ridges provide the opportunity for more extensive
mixing of melts and assimilation of the crust over a larger
scale, leading to a common chemical signature of assimila-
tion of hydrothermally altered crust (cf. Saal et al. 2002;
Kendrick et al. 2013).
Global implications of hydrothermal Cl contamination
The depth of hydrothermal circulation is important for the
cooling of the lithosphere. With our data, we can demon-
strate that there is a cooling mechanism by circulation of
hydrothermal fluids down to the lower crust that can facili-
tate deep crystallization at slow-spreading ridges. This cir-
culation most likely happens at all low spreading rates. This
means crust is not merely formed at shallow levels but that
lower crustal and lithospheric mantle magma chamber pro-
cesses may be common at slow-spreading ridges and that
oceanic crust can form in situ at all depths (cf. Kelemen
et al. 1997; Kelemen and Aharonov 1998; Lissenberg et al.
2004; Grimes et al. 2008).
Hydrothermal circulation throughout the newly formed
lithosphere can increase the Cl budget and to a smaller
extent the ­H2O budget of an altered oceanic slab and thus
the potential of Cl (and ­H2O) to subduct, although we cannot
determine with the present data the Cl content and distribu-
tion in the altered plate and so cannot give a realistic esti-
mate of its total Cl content. This has important consequences
for global water and halogen budgets (see e.g. Kendrick et al.
2017). Although most other elements occur in too low con-
centrations in hydrothermal fluids to significantly change the
composition of the deeper lithosphere, H- and S-isotopes, B
and noble gasses may be significantly changed (Kent et al.
1999; Gillis et al. 2003; Kendrick et al. 2013; Labidi et al.
2014; Stroncik and Niedermann 2016). Any study of those
elements and isotopes should take into account the effect of
hydrothermal circulation and assimilation of hydrothermally
altered rocks. In addition, melts that show Cl-excess at slow-
spreading ridges likely interacted with the lithosphere and
care should be taken when using their major element data
for e.g. pressure or melt degree calculations.
Conclusion
Novel high-precision Cl measurements of basaltic glasses
and melt inclusions in Ol and Pl from the slow- and ultra-
slow-spreading Southern Mid-Atlantic Ridge and Gakkel
Ridge show up to 75% of these samples to have been affected
by Cl-contamination, raising their Cl/Nb and Cl/K ratios
Contrib Mineral Petrol (2017) 172:97 Page 19 of 23  97
above mantle values, which cannot be explained by mag-
matic processes such as degassing, melting, fractionation
or source variations, or by seafloor processes. This contami-
nation most likely occurs by the addition of hydrothermal
brines to the magmas during the assimilation of hydrother-
mally altered country rock. Cl-contamination is shown to
be a common process at (ultra-) slow-spreading ridges, in
contrast to conclusions drawn from early studies based on
lower precision Cl data. The compositional effects of crus-
tal assimilation lead to disequilibrium and phase-equilib-
ria-based estimates of magma fractionation depths, such as
those provided by COMAGMAT, to be erroneous. Pressures
calculated from Cpx–melt pairs are more reliable as equi-
librium can be assessed and yield Cl-contamination depths
(and thus minimum depths to which hydrothermal circula-
tion penetrates) of at least 10 km for SMAR segment A2.
This is near the base of the lower crust in this studied region.
Melt inclusions with Cl-excess show assimilation of hydro-
thermal crust in magma chambers before crystallization at
depth in magma reservoirs. Deep cooling by hydrothermal
circulation and crustal assimilation is consistent with inde-
pendent earlier evidence for lower crustal crystallization and
melt–rock reactions and for extensive and deep faulting at
slow-spreading ridges.
All MORB basalts that display Cl-excess were at some
point in time affected by hydrothermal fluids, while the high-
est observed values here are associated with hydrothermal
vent fields. Although globally both fast- and slow-spread-
ing ridges display a similar proportion of samples with Cl-
excess, slow-spreading ridges show both lower absolute Cl
concentrations and maximum degrees of Cl-excess. This
may result from either less alteration of the crust or less
interaction of the magmas with the crust at slow-spreading
ridges. The generally less fractionated nature of slow-spread-
ing magmas provides support for the latter explanation—
the former explanation would require a larger proportion
of conductive cooling of the lower crust as spreading rates
decrease. The detection of the effects of deep-seated hydro-
thermal circulation in slow-spreading crust has implications
for the volatile and halogen budget of the oceanic plates and
this needs to be taken into account when assessing the Cl,
­H2O, H-, S-, B- and noble gas isotopic signatures of global
geochemical cycles.
Acknowledgements  We are very grateful to Mario Thöner for the
extensive technical assistance at the EMP and to Dagmar Rau for the
technical assistance at the LA-ICP-MS. Further, we like to thank Jan
Fietzke (all GEOMAR) for the help with the modification of the Cl
measurement method for the melt inclusions. The suggestion of three
anonymous reviewers and editorial handling by Jochen Hoefs was
greatly appreciated. We acknowledge generous financial support from
the Jeddah Transect Project between King Abdulaziz University and
Helmholtz-Center for Ocean Research GEOMAR that was funded by
King Abdulaziz University (KAU) Jeddah, Saudi Arabia, under Grant
no. (T-065/430).
References
Almeev RR, Holtz F, Koepke J, Ariskin AA (2007a) The effect of
minor H2O content on crystallisation in MORB: experiments,
model, applications. In: Goldschmidt 2007 abstracts Geochim
Cosmochim Acta vol 71 (15S1). p A15
Almeev RR, Holtz F, Koepke J, Parat F, Botcharnikov RE (2007b) The
effect of H
­ 2O on olivine crystallization in MORB: experimental
calibration at 200 MPa. Am Miner 92(4):670–674
Almeev R, Holtz F, Koepke J, Haase K, Devey C (2008) Depths of
partial crystallization of ­H2O-bearing MORB: phase equilibria
simulations of basalts at the MAR near Ascension Island (7–11
S). J Petrol 49(1):25–45
Almeev RR, Holtz F, Koepke J, Parat F (2012) Experimental calibra-
tion of the effect of ­H2O on plagioclase crystallization in basaltic
melt at 200 MPa. Am Miner 97(7):1234–1240
Alt JC, Bach W (2006) Oxygen isotope composition of a section of
lower oceanic crust, ODP hole 735B. Geochem Geophys Geosyst
7(12):Q12008. https://doi.org/10.1029/2006GC001385
Alt JC, Teagle DA (2003) Hydrothermal alteration of upper oceanic
crust formed at a fast-spreading ridge: mineral, chemical, and iso-
topic evidence from ODP Site 801. Chem Geol 201(3):191–211
Alt JC, Honnorez J, Laverne C, Emmermann R (1986) Hydrothermal
alteration of a 1 km section through the upper oceanic crust,
Deep Sea Drilling Project Hole 504B: mineralogy, chemistry and
evolution of seawater–basalt interactions. J Geophys Res Solid
Earth 91(B10):10309–10335
Anderson MO, Chadwick WW, Hannington MD, Merle SG, Resing
JA, Baker ET, Butterfield DA, Walker SL, Augustin N (2017)
Geological interpretation of volcanism and segmentation of the
Mariana back-are spreading center between 12.7°N and 18.3°N.
Geochem Geophys Geosyst 18(6):2240–2274
Ariskin AA, Barmina GS (2004) COMAGMAT: development of a
magma crystallization model and its petrological applications.
Geochem Int 42(1):S1–S157
Augustin N, Lackschewitz K, Kuhn T, Devey CW (2008) Mineral-
ogical and chemical mass changes in mafic and ultramafic rocks
from the Logatchev hydrothermal field (MAR 15 N). Mar Geol
256(1):18–29
Bach W, Peucker-Ehrenbrink B, Hart SR, Blusztajn JS (2003) Geo-
chemistry of hydrothermally altered oceanic crust: DSDP/ODP
Hole 504B–Implications for seawater-crust exchange budgets and
Sr- and Pb-isotopic evolution of the mantle. Geochem Geophys
Geosyst 4(3). https://doi.org/10.1029/2002GC000419
Bach W, Garrido CJ, Paulick H, Harvey J, Rosner M (2004) Sea-
water–peridotite interactions: First insights from ODP Leg
209, MAR 15 N. Geochem Geophys Geosyst 5(9). https://doi.
org/10.1029/2004GC000744
Baker ET, Edmonds HN, Michael PJ, Bach W, Dick HJB, Snow JE,
Walker SL, Banerjee NR, Langmuir CH (2004) Hydrother-
mal venting in magma deserts: the ultraslow-spreading Gak-
kel and Southwest Indian ridges. Geochem Geophys Geosyst
5(8):Q08002. https://doi.org/10.1029/2004GC000712
Barnes JD, Cisneros M (2012) Mineralogical control on the chlorine
isotope composition of altered oceanic crust. Chem Geol 326–
327:51–60. https://doi.org/10.1016/j.chemgeo.2012.07.022
Bédard JH (1991) Cumulate recycling and crustal evolution in the Bay
of Islands ophiolite. J Geol 2:225–249
Bédard JH, Hébert R (1996) The lower crust of the Bay of Islands
ophiolite, Canada: petrology, mineralogy, and the importance of
syntexis in magmatic differentiation in ophiolites and at ocean
ridges. J Geophys Res Solid Earth 101(B11):25105–25124.
https://doi.org/10.1029/96JB01343
Bédard JH, Hebert R, Berclaz A, Varfalvy V (2000) Syntexis and the
genesis of lower oceanic crust. In: Dilek Y (ed) Ophiolites and
13
 97   Page 20 of 23
oceanic crust: new insights from field studies and the Ocean
Drilling Program. Geological Society of America, pp 105–120
Behrens H, Misiti V, Freda C, Vetere F, Botcharnikov RE, Scarlato
P (2009) Solubility of ­H2O and ­CO2 in ultrapotassic melts at
1200 and 1250 °C and pressure from 50 to 500 MPa. Am Miner
94(1):105–120
Berndt ME, Seyfried WE Jr (1990) Boron, bromine, and other trace
elements as clues to the fate of chlorine in mid-ocean ridge vent
fluids. Geochim Cosmochim Acta 54(8):2235–2245. https://doi.
org/10.1016/0016-7037(90)90048-P
Berry AJ, O’Neill HStC, Rowe MC, Moselmans JFW, Rivard C (2017)
The oxidation state of iron in basaltic glasses. In: Goldschmidt
2017 abstracts. Nr 327
Bézos A, Humler E (2005) The F ­ e3+/ΣFe ratios of MORB glasses and
their implications for mantle melting. Geochim Cosmochim Acta
69(3):711–725
Bischoff JL, Rosenbauer RJ (1987) Phase separation in seafloor geo-
thermal systems; an experimental study of the effects on metal
transport. Am J Sci 287(10):953–978
Bruguier N, Minshull T, Brozena J (2003) Morphology and tectonics
of the Mid-Atlantic Ridge, 7°–12°S. J Geophys Res Solid Earth
108(B2)
Cannat M (1996) How thick is the magmatic crust at slow spreading
oceanic ridges? J Geophys Res Solid Earth 101(B2):2847–2857
Cannat M, Mével C, Stakes D (1991) Stretching of the deep crust at
the slow-spreading Southwest Indian ridge. Tectonophysics
190(1):73–94. https://doi.org/10.1016/0040-1951(91)90355-V
Cannat M, Sauter D, Mendel V, Ruellan E, Okino K, Escartin J, Com-
bier V, Baala M (2006) Modes of seafloor generation at a melt-
poor ultraslow-spreading ridge. Geology 34(7):605–608
Cherkaoui AS, Wilcock WS, Dunn RA, Toomey DR (2003) A numeri-
cal model of hydrothermal cooling and crustal accretion at a
fast spreading mid-ocean ridge. Geochem Geophys Geosyst
4(9):8616
Coakley BJ, Cochran JR (1998) Gravity evidence of very thin crust
at the Gakkel Ridge (Arctic Ocean). Earth Planet Sci Lett
162(1):81–95
Coogan LA (2003) Contaminating the lower crust in the Oman ophi-
olite. Geology 31(12):1065–1068. https://doi.org/10.1130/
g20129.1
Coogan LA, Saunders AD, Kempton PD, Norry MJ (2000) Evidence
from oceanic gabbros for porous melt migration within a crystal
mush beneath the Mid-Atlantic Ridge. Geochem Geophys Geo-
syst 1(9):1044. https://doi.org/10.1029/2000GC000072
Coogan LA, Mitchell NC, O’Hara MJ (2003) Roof assimilation at
fast spreading ridges: an investigation combining geophysical,
geochemical, and field evidence. J Geophys Res Solid Earth
108(B1):ECV2-1–ECV2-14
DeMets C, Gordon RG, Argus DF (2010) Geologically current plate
motions. Geophys J Int 181(1):1–80
Detrick RS, Mutter JC, Buhl P, Kim II (1990) No evidence from mul-
tichannel reflection data for a crustal magma chamber in the
MARK area on the Mid-Atlantic Ridge. Nature 347(6288):61–64
Devey CW, Lackschewitz KS, Baker E (2005) Hydrothermal and
volcanic activity found on the Southern Mid-Atlantic Ridge.
EOS Trans Am Geophys Union 86(22):209–212. https://doi.
org/10.1029/2005EO220001
Devey CW, German C, Haase K, Lackschewitz K, Melchert B, Con-
nelly D (2010) The relationships between volcanism, tectonism,
and hydrothermal activity on the southern equatorial Mid-Atlan-
tic Ridge. In: Rona PA, Devey CW, Dyment J, Murton BJ (eds)
Diversity of hydrothermal systems on slow spreading ocean
ridges, pp 133–152
Dick HJB, Natland JH, Alt JC, Bach W, Bideau D, Gee JS, Haggas S,
Hertogen JGH, Hirth G, Holm PM, Ildefonse B, Iturrino GJ, John
BE, Kelley DS, Kikawa E, Kingdon A, LeRoux PJ, Maeda J,
13
Contrib Mineral Petrol (2017) 172:97
Meyer PS, Miller DJ, Naslund HR, Niu Y-L, Robinson PT, Snow
J, Stephen RA, Trimby PW, Worm H-U, Yoshinobu A (2000) A
long in situ section of the lower ocean crust: results of ODP Leg
176 drilling at the Southwest Indian ridge. Earth Planet Sci Lett
179(1):31–51. https://doi.org/10.1016/S0012-821X(00)00102-3
Dick H, Ozawa K, Meyer P, Niu Y, Robinson P, Constantin M, Hebert
R, Maeda J, Natland J, Hirth G (2002) Primary silicate min-
eral chemistry of a 1.5-km section of very slow spreading lower
ocean crust: ODP Hole 735B, Southwest Indian ridge. Proc
Ocean Drill Program Sci Results 176:1–61
Dick HJB, Lin J, Schouten H (2003) An ultraslow-spreading class of
ocean ridge. Nature 426(6965):405–412
Drouin M, Godard M, Ildefonse B, Bruguier O, Garrido CJ (2009)
Geochemical and petrographic evidence for magmatic impregna-
tion in the oceanic lithosphere at Atlantis Massif, Mid-Atlantic
Ridge (IODP Hole U1309D, 30°N). Chem Geol 264(1–4):71–88.
https://doi.org/10.1016/j.chemgeo.2009.02.013
Dunn RA, Toomey DR, Solomon SC (2000) Three-dimensional seis-
mic structure and physical properties of the crust and shallow
mantle beneath the East Pacific Rise at 9°30′N. J Geophys Res
Solid Earth 105(B10):23537–23555
Edmonds HN, Michael PJ, Baker ET, Connelly DP, Snow JE, Lang-
muir CH, Dick HJB, Muhe R, German CR, Graham DW
(2003) Discovery of abundant hydrothermal venting on the
ultraslow-spreading Gakkel Ridge in the Arctic Ocean. Nature
421(6920):252–256. http://www.nature.com/nature/journal/
v421/n6920/suppinfo/nature01351_S1.html
Elkins L, Sims K, Prytulak J, Blichert-Toft J, Elliott T, Blusztajn J,
Fretzdorff S, Reagan M, Haase K, Humphris S (2014) Melt gen-
eration beneath Arctic ridges: implications from U decay series
disequilibria in the Mohns, Knipovich, and Gakkel Ridges. Geo-
chim Cosmochim Acta 127:140–170
Engdahl ER, van der Hilst R, Buland R (1998) Global teleseismic
earthquake relocation with improved travel times and procedures
for depth determination. Bull Seismol Soc Am 88(3):722–743
Erdmann M, France L, Fischer LA, Deloule E, Koepke J (2017) Trace
elements in anatectic products at the roof of mid-ocean ridge
magma chambers: an experimental study. Chem Geol 456:43–57.
https://doi.org/10.1016/j.chemgeo.2017.03.004
Expedition Scientific Party (2005) Oceanic core complex formation,
Atlantis Massif, Mid-Atlantic Ridge: drilling into the footwall
and hanging wall of a tectonic exposure of deep, young oceanic
lithosphere to study deformation, alteration, and melt genera-
tion. IODP Prelim Rep. https://doi.org/10.2204/iodp.pr.305.2005
Fietzke J, Frische M (2016) Experimental evaluation of elemental
behavior during LA-ICP-MS: influences of plasma conditions
and limits of plasma robustness. J Anal At Spectrom. 31(1):234–
244. https://doi.org/10.1039/C5JA00253B
Fietzke J, Liebetrau V, Günther D, Gürs K, Hametner K, Zumholz K,
Hansteen T, Eisenhauer A (2008) An alternative data acquisition
and evaluation strategy for improved isotope ratio precision using
LA-MC-ICP-MS applied to stable and radiogenic strontium iso-
topes in carbonates. J Anal At Spectrom 23(7):955–961
Fischer LA, Erdmann M, France L, Wolff PE, Deloule E, Zhang C,
Godard M, Koepke J (2016) Trace element evidence for anatexis
at oceanic magma chamber roofs and the role of partial melts
for contamination of fresh MORB. Lithos 260:1–8. https://doi.
org/10.1016/j.lithos.2016.05.001
Fouquet Y (1997) Where are the large hydrothermal sulphide depos-
its in the oceans? Philos Trans R Soc Math Phys Eng Sci
355(1723):427–441
Fournier R (1987) Conceptual models of brine evolution in
magmatic-hydrothermal systems. US Geol Surv Prof Pap
1350(2):1487–1506
France L, Ildefonse B, Koepke J (2009) Interactions between magma
and hydrothermal system in Oman ophiolite and in IODP Hole
Contrib Mineral Petrol (2017) 172:97 Page 21 of 23  97
1256D: Fossilization of a dynamic melt lens at fast spread-
ing ridges. Geochem Geophys Geosyst 10(10).https://doi.
org/10.1029/2009GC002652
France L, Koepke J, Ildefonse B, Cichy SB, Deschamps F (2010)
Hydrous partial melting in the sheeted dike complex at fast
spreading ridges: experimental and natural observations. Con-
trib Miner Petrol 160(5):683–704
France L, Koepke J, MacLeod CJ, Ildefonse B, Godard M, Deloule E
(2014) Contamination of MORB by anatexis of magma cham-
ber roof rocks: constraints from a geochemical study of experi-
mental melts and associated residues. Lithos 202–203:120–
137. https://doi.org/10.1016/j.lithos.2014.05.018
German C, Connelly D, Evans A, Parson L (2002) Hydrothermal
activity on the southern Mid-Atlantic Ridge. In: AGU Fall
Meeting Abstracts, vol 1. p 1361
Gillis KM, Thompson G, Kelley DS (1993) A view of the lower
crustal component of hydrothermal systems at the Mid-Atlan-
tic Ridge. J Geophys Res Solid Earth 98(B11):19597–19619.
https://doi.org/10.1029/93JB01717
Gillis KM, Coogan LA, Chaussidon M (2003) Volatile element (B,
Cl, F) behaviour in the roof of an axial magma chamber from
the East Pacific Rise. Earth Planet Sci Lett 213(3–4):447–462.
https://doi.org/10.1016/s0012-821x(03)00346-7
Godard M, Awaji S, Hansen H, Hellebrand E, Brunelli D, Johnson K,
Yamasaki T, Maeda J, Abratis M, Christie D, Kato Y, Mariet
C, Rosner M (2009) Geochemistry of a long in situ section of
intrusive slow-spread oceanic lithosphere: results from IODP
Site U1309 (Atlantis Massif, 30°N Mid-Atlantic-ridge). Earth
Planet Sci Lett 279(1–2):110–122. https://doi.org/10.1016/j.
epsl.2008.12.034
Goldstein SL, Soffer G, Langmuir CH, Lehnert KA, Graham
DW, Michael PJ (2008) Origin of a `Southern Hemisphere’
geochemical signature in the Arctic upper mantle. Nature
453(7191):89–93. http://www.nature.com/nature/journal/v453/
n7191/suppinfo/nature06919_S1.html
Gregory RT, Taylor HP (1981) An oxygen isotope profile in a section
of Cretaceous oceanic crust, Samail Ophiolite, Oman: evidence
for δ18O buffering of the oceans by deep (> 5 km) seawater-
hydrothermal circulation at mid-ocean ridges. J Geophys Res
Solid Earth 86(B4):2737–2755
Grevemeyer I, Reston TJ, Moeller S (2013) Microseismicity of the
Mid-Atlantic Ridge at 7°S–8°15′S and at the Logatchev Massif
oceanic core complex at 14°40′N–14°50′N. Geochem Geophys
Geosyst 14(9):3532–3554
Grimes CB, John BE, Cheadle MJ, Wooden JL (2008) Protracted
construction of gabbroic crust at a slow spreading ridge: con-
straints from 206Pb/238U zircon ages from Atlantis Massif and
IODP Hole U1309D (30 N, MAR). Geochem Geophys Geosyst
9(8). https://doi.org/10.1029/2008GC002063
Haase K, Brandl PA, Devey CW, Hauff F, Melchert B, Garbe-Schön-
berg D, Kokfelt T, Paulick H (2016) Compositional variation
and 226Ra-230Th model ages of axial lavas from the south-
ern Mid-Atlantic Ridge, 8°48′S. Geochem Geophys Geosyst
17(1):199–218
Harper GD (1985) Tectonics of slow spreading mid-ocean ridges
and consequences of a variable depth to the brittle/ductile
transition. Tectonics 4(4):395–409. https://doi.org/10.1029/
TC004i004p00395
Hart S, Erlank A, Kable E (1974) Sea floor basalt alteration:
some chemical and Sr isotopic effects. Contrib Miner Petrol
44(3):219–230
Hasenclever J, Theissen-Krah S, Rüpke LH, Morgan JP, Iyer K,
Petersen S, Devey CW (2014) Hybrid on-axis plus ridge-per-
pendicular circulation reconciles hydrothermal flow observa-
tions at fast spreading ridges. Nature 508:508–512
Hoernle K, Hauff F, Kokfelt TF, Haase K, Garbe-Schönberg D, Werner
R (2011) On- and off-axis chemical heterogeneities along the
South Atlantic Mid-Ocean-ridge (5–11°S): shallow or deep recy-
cling of ocean crust and/or intraplate volcanism? Earth Planet Sci
Lett 306(1–2):86–97. https://doi.org/10.1016/j.epsl.2011.03.032
Hofmann AW (1988) Chemical differentiation of the Earth: the rela-
tionship between mantle, continental crust, and oceanic crust.
Earth Planet Sci Lett 90(3):297–314
Hofmann AW, Jochum KP, Seufert M, White WM (1986) Nb
and Pb in oceanic basalts: new constraints on mantle evo-
lution. Earth Planet Sci Lett 79(1–2):33–45. https://doi.
org/10.1016/0012-821X(86)90038-5
International Seismological Centre (2011) On-line bulletin. Interna-
tional Seismological Centre. http://www.isc.ac.uk. Accessed 19
Dec 2012
Ito E, Harris DM, Anderson AT Jr (1983) Alteration of oce-
anic crust and geologic cycling of chlorine and water.
Geochim Cosmochim Acta 47(9):1613–1624. https://doi.
org/10.1016/0016-7037(83)90188-6
Jarosewich E, Nelen JA, Norberg JA (1980) Reference samples for elec-
tron microprobe analysis. Geostand Newsl 4(1):43–47. https://
doi.org/10.1111/j.1751-908X.1980.tb00273.x
Jenner FE, O’Neill HSC (2012) Analysis of 60 elements in 616 ocean
floor basaltic glasses. Geochem Geophys Geosyst 13(2):Q02005
Jochum KP, Stoll B, Herwig K, Willbold M, Hofmann AW, Amini M,
Aarburg S, Abouchami W, Hellebrand E, Mocek B (2006) MPI-
DING reference glasses for in situ microanalysis: new reference
values for element concentrations and isotope ratios. Geochem
Geophys Geosyst 7(2). https://doi.org/10.1029/2005GC001060
Jokat W, Ritzmann O, Schmidt-Aursch MC, Drachev S, Gauger S,
Snow J (2003) Geophysical evidence for reduced melt produc-
tion on the Arctic ultraslow Gakkel mid-ocean ridge. Nature
423(6943):962–965. http://www.nature.com/nature/journal/
v423/n6943/suppinfo/nature01706_S1.html
Kawahata H, Nohara M, Ishizuka H, Hasebe S, Chiba H (2001) Sr
isotope geochemistry and hydrothermal alteration of the Oman
ophiolite. J Geophys Res Solid Earth 106(B6):11083–11099.
https://doi.org/10.1029/2000JB900456
Kelemen PB, Aharonov E (1998) Periodic formation of magma frac-
tures and generation of layered gabbros in the lower crust beneath
oceanic spreading ridges. In: Buck WR, Delaney PT, Karson
JA, Lagabrielle Y (eds) Faulting and magmatism at Mid-Ocean
ridges. American Geophysical Union, pp 267–289
Kelemen PB, Shimizu N, Salters VJM (1995) Extraction of mid-ocean-
ridge basalt from the upwelling mantle by focused flow of melt
in dunite channels. Nature 375(6534):747–753
Kelemen PB, Koga K, Shimizu N (1997) Geochemistry of gabbro
sills in the crust-mantle transition zone of the Oman ophiolite:
implications for the origin of the oceanic lower crust. Earth
Planet Sci Lett 146(3–4):475–488. https://doi.org/10.1016/
S0012-821X(96)00235-X
Kendrick MA, Arculus R, Burnard P, Honda M (2013) Quantify-
ing brine assimilation by submarine magmas: examples from
the Galápagos Spreading Centre and Lau Basin. Geochim
Cosmochim Acta 123:150–165. https://doi.org/10.1016/j.
gca.2013.09.012
Kendrick MA, Hemond C, Kamenetsky VS, Danyushevsky L, Devey
CW, Rodemann T, Jackson MG, Perfit MR (2017) Seawater
cycled throughout Earth’s mantle in partially serpentinized
lithosphere. Nat Geosci 10:222–228. https://doi.org/10.1038/
NGEO2902
Kent AJR, Norman MD, Hutcheon ID, Stolper EM (1999) Assimilation
of seawater-derived components in an oceanic volcano: evidence
from matrix glasses and glass inclusions from Loihi seamount,
Hawaii. Chem Geol 156(1–4):299–319. https://doi.org/10.1016/
S0009-2541(98)00188-0
13
 97   Page 22 of 23
Klein EM, Langmuir CH (1987) Global correlations of ocean ridge
basalt chemistry with axial depth and crustal thickness. J Geo-
phys Res Solid Earth 92(B8):8089–8115. https://doi.org/10.1029/
JB092iB08p08089
Klügel A (1998) Reactions between mantle xenoliths and host
magma beneath La Palma (Canary Islands): constraints on
magma ascent rates and crustal reservoirs. Contrib Miner Petrol
131(2–3):237–257
Kovalenko VI, Naumov VB, Girnis AV, Dorofeeva VA, Yarmolyuk
VV (2006) Estimation of the average contents of ­H2O, Cl, F,
and S in the depleted mantle on the basis of the compositions
of melt inclusions and quenched glasses of mid-ocean ridge
basalts. Geochem Int 44(3):209–231. https://doi.org/10.1134/
s0016702906030013
Kvassnes AS, Grove T (2008) How partial melts of mafic lower crust
affect ascending magmas at oceanic ridges. Contrib Miner Petrol
156(1):49–71. https://doi.org/10.1007/s00410-007-0273-x
Labidi J, Cartigny P, Hamelin C, Moreira M, Dosso L (2014) Sulfur
isotope budget (32S, 33S, 34S and 36S) in Pacific-Antarctic ridge
basalts: a record of mantle source heterogeneity and hydrother-
mal sulfide assimilation. Geochim Cosmochim Acta 133:47–67.
https://doi.org/10.1016/j.gca.2014.02.023
le Roux PJ, Shirey SB, Hauri EH, Perfit MR, Bender JF (2006) The
effects of variable sources, processes and contaminants on the
composition of northern EPR MORB (8–10°N and 12–14°N):
evidence from volatiles (­ H2O, ­CO2, S) and halogens (F, Cl). Earth
Planet Sci Lett 251(3–4):209–231. https://doi.org/10.1016/j.
epsl.2006.09.012
Lecuyer C, Reynard B (1996) High-temperature alteration of oceanic
gabbros by seawater (Hess Deep, Ocean Drilling Program Leg
147): evidence from oxygen isotopes and elemental fluxes. J
Geophys Res Solid Earth 101(B7):15883–15897. https://doi.
org/10.1029/96JB00950
Lehnert K, Su Y, Langmuir C, Sarbas B, Nohl U (2000) A global
geochemical database structure for rocks. Geochem Geophys
Geosyst 1(5):1012. https://doi.org/10.1029/1999GC000026
Lissenberg CJ, Dick HJB (2008) Melt–rock reaction in the lower
oceanic crust and its implications for the genesis of mid-ocean
ridge basalt. Earth Planet Sci Lett 271(1–4):311–325. https://doi.
org/10.1016/j.epsl.2008.04.023
Lissenberg CJ, Bédard JH, van Staal CR (2004) The structure and geo-
chemistry of the gabbro zone of the Annieopsquotch ophiolite,
newfoundland: implications for lower crustal accretion at spread-
ing ridges. Earth Planet Sci Lett 229(1–2):105–123. https://doi.
org/10.1016/j.epsl.2004.10.029
Lissenberg CJ, Rioux M, Shimizu N, Bowring SA, Mével C (2009) Zir-
con dating of oceanic crustal accretion. Science 323(5917):1048–
1050. https://doi.org/10.1126/science.1167330
Maclennan J, Hulme T, Singh SC (2005) Cooling of the lower oceanic
crust. Geology 33(5):357–366. https://doi.org/10.1130/g21207.1
McDonough WF, Sun SS (1995) The composition of the
Ear th. Chem Geol 120(3–4):223–253. https://doi.
org/10.1016/0009-2541(94)00140-4
Melchert B, Devey CW, German C, Lackschewitz K, Seifert R, Wal-
ter M, Mertens C, Yoerger D, Baker E, Paulick H (2008) First
evidence for high-temperature off-axis venting of deep crustal/
mantle heat: the Nibelungen hydrothermal field, southern Mid-
Atlantic Ridge. Earth Planet Sci Lett 275(1):61–69
Mével C, Cannat M (1991) Lithospheric stretching and hydrothermal
processes in oceanic gabbros from slow-spreading ridges. In:
Peters TJ, Nicolas A, Coleman R (eds) Ophiolite genesis and
evolution of the oceanic lithosphere. Springer, pp 293–312
Meyer P, Dick HB, Thompson G (1989) Cumulate gabbros from the
Southwest Indian ridge, 54°S–7°16′E: implications for mag-
matic processes at a slow spreading ridge. Contrib Miner Petrol
103(1):44–63. https://doi.org/10.1007/BF00371364
13
Contrib Mineral Petrol (2017) 172:97
Michael P (1995) Regionally distinctive sources of depleted
MORB: evidence from trace elements and H2O.
Earth Planet Sci Lett 131(3–4):301–320. https://doi.
org/10.1016/0012-821X(95)00023-6
Michael PJ, Cornell WC (1998) Influence of spreading rate and
magma supply on crystallization and assimilation beneath
mid-ocean ridges: evidence from chlorine and major ele-
ment chemistry of mid-ocean ridge basalts. J Geophys Res
103(B8):18325–18356. https://doi.org/10.1029/98jb00791
Michael PJ, Graham DW (2015) The behavior and concentration of
­CO 2 in the suboceanic mantle: inferences from undegassed
ocean ridge and ocean island basalts. Lithos 236–237:338–351.
https://doi.org/10.1016/j.lithos.2015.08.020
Morgan JP, Chen YJ (1993) The genesis of oceanic crust: Magma
injection, hydrothermal circulation, and crustal flow. J Geophys
Res Solid Earth 98(B4):6283–6297
Michael PJ, Schilling J-G (1989) Chlorine in mid-ocean ridge mag-
mas: evidence for assimilation of seawater-influenced compo-
nents. Geochim Cosmochim Acta 53(12):3131–3143. https://
doi.org/10.1016/0016-7037(89)90094-x
Michael P, Langmuir C, Dick H, Snow J, Goldstein S, Graham D,
Lehnert K, Kurras G, Jokat W, Mühe R (2003) Magmatic and
amagmatic seafloor generation at the ultraslow-spreading Gak-
kel Ridge, Arctic Ocean. Nature 423(6943):956–961
Minshull T, Bruguier N, Brozena J (1998) ridge-plume interactions
or mantle heterogeneity near Ascension Island? Geology
26(2):115–118
Wise SA, Watters RL (2012) Certificate of Analysis, Standard Refer-
ence Material 610. National Institute of Standards and Tech-
nology. http://www.nist.gov/srm
Möller H (2002) Magma Genesis and Mantle Source at the Mid-
Atlantic Ridge East of Ascension Island. Dissertation at Chris-
tian-Albrechts-Univeristät zu Kiel
Montési LG, Behn MD (2007) Mantle flow and melting underneath
oblique and ultraslow Mid-Ocean ridges. Geophys Res Lett
34(24). https://doi.org/10.1029/2007GL031067
Mottl M (2003) Partitioning of energy and mass fluxes between
mid-ocean ridge axes and flanks at high and low temperature.
In: Halbach P, Tunnicliffe V, Hein JR (eds) Energy and mass
transfer in marine hydrothermal systems. Dahlem University
Press, Berlin, pp 271–286
Nehlig P, Juteau T (1988) Flow porosities, permeabilities and
preliminary data on fluid inclusions and fossil thermal
gradients in the crustal sequence of the Sumail ophiolite
(Oman). Tectonophysics 151(1–4):199–221. https://doi.
org/10.1016/0040-1951(88)90246-6
Nicolas A, Mainprice D, Boudier F (2003) High-temperature sea-
water circulation throughout crust of oceanic ridges: a model
derived from the Oman ophiolites. J Geophys Res Solid Earth
108(B8):2371. https://doi.org/10.1029/2002JB002094
Niu Y, Hekinian R (1997) Spreading-rate dependence of the extent
of mantle melting beneath ocean ridges. Nature 385:326–329
Palme H, O’Neill HSC (2003) Cosmochemical estimates of mantle
composition. In: Heinrich DH, Karl KT (eds) Treatise on geo-
chemistry, vol 2. Pergamon, Oxford, pp 1–38
Paulick H, Münker C, Schuth S (2010) The influence of small-scale
mantle heterogeneities on Mid-Ocean Ridge volcanism: evi-
dence from the southern Mid-Atlantic Ridge (7 30′S to 11 30′S)
and Ascension Island. Earth Planet Sci Lett 296(3–4):299–310.
https://doi.org/10.1016/j.epsl.2010.05.009
Pontbriand CW, Soule SA, Sohn RA, Humphris SE, Kunz C, Singh
H, Nakamura Ki, Jakobsson M, Shank T (2012) Effusive
and explosive volcanism on the ultraslow-spreading Gakkel
Ridge, 85°E. Geochem Geophys Geosyst 13(10). https://doi.
org/10.1029/2012GC004187
Contrib Mineral Petrol (2017) 172:97 Page 23 of 23  97
Putirka KD (2008) Thermometers and barometers for volcanic systems.
Rev Miner Geochem 69(1):61–120
Putirka K, Johnson M, Kinzler R, Longhi J, Walker D (1996) Thermo-
barometry of mafic igneous rocks based on clinopyroxene-liquid
equilibria, 0–30 kbar. Contrib Miner Petrol 123(1):92–108
Reid I, Jackson H (1981) Oceanic spreading rate and crustal thickness.
Mar Geophys Res 5(2):165–172
Ridley WI, Perfit MR, Smith MC, Fornari DJ (2006) Magmatic pro-
cesses in developing oceanic crust revealed in a cumulate xeno-
lith collected at the East Pacific Rise, 9°50′N. Geochem Geophys
Geosyst 7(12):Q12O04. https://doi.org/10.1029/2006GC001316
Roeder P, Emslie R (1970) Olivine-liquid equilibrium. Contrib Miner
Petrol 29(4):275–289
Rubin KH, Sinton JM (2007) Inferences on mid-ocean ridge ther-
mal and magmatic structure from MORB compositions. Earth
Planet Sci Lett 260(1–2):257–276. https://doi.org/10.1016/j.
epsl.2007.05.035
Rutherford MJ (2008) Magma ascent rates. Rev Miner Geochem
69(1):241–271
Ryabchikov ID (2001) Deep geospheres and ore genesis. Geol Rudn
Mestorozhd 43:195–207
Saal AE, Hauri EH, Langmuir CH, Perfit MR (2002) Vapour under
saturation in primitive mid-ocean-ridge basalt and the volatile
content of Earth’s upper mantle. Nature 419(6906):451–455.
http://www.nature.com/nature/journal/v419/n6906/suppinfo/
nature01073_S1.html
Salters VJM, Stracke A (2004) Composition of the depleted man-
tle. Geochem Geophys Geosyst 5(5):Q05B07 https://doi.
org/10.1029/2003gc000597
Sanfilippo A, Tribuzio R, Tiepolo M (2014) Mantle–crust interactions
in the oceanic lithosphere: constraints from minor and trace
elements in olivine. Geochim Cosmochim Acta 141:423–439.
https://doi.org/10.1016/j.gca.2014.06.012
Sano T, Miyoshi M, Ingle S, Banerjee NR, Ishimoto M, Fukuoka
T (2008) Boron and chlorine contents of upper oceanic crust:
Basement samples from IODP Hole 1256D. Geochem Geophys
Geosyst 9(12):Q12O15. https://doi.org/10.1029/2008GC002182
Sauter D, Cannat M, Rouméjon S, Andreani M, Birot D, Bronner A,
Brunelli D, Carlut J, Delacour A, Guyader V (2013) Continu-
ous exhumation of mantle-derived rocks at the Southwest Indian
Ridge for 11 million years. Nat Geosci 6(4):314
Schramm B, Devey CW, Gillis KM, Lackschewitz K (2005) Quanti-
tative assessment of chemical and mineralogical changes due
to progressive low-temperature alteration of East Pacific Rise
basalts from 0 to 9 Ma. Chem Geol 218(3–4):281–313. https://
doi.org/10.1016/j.chemgeo.2005.01.011
Shaw C, Klügel A (2002) The pressure and temperature conditions
and timing of glass formation in mantle-derived xenoliths from
Baarley, West Eifel, Germany: the case for amphibole break-
down, lava infiltration and mineral-melt reaction. Miner Petrol
74(2–4):163–187
Shaw AM, Behn MD, Humphris SE, Sohn RA, Gregg PM (2010) Deep
pooling of low degree melts and volatile fluxes at the 85 E seg-
ment of the Gakkel Ridge: evidence from olivine-hosted melt
inclusions and glasses. Earth Planet Sci Lett 289(3):311–322
Shishkina T, Botcharnikov R, Holtz F, Almeev R, Portnyagin MV
(2010) Solubility of ­H2O- and ­CO2-bearing fluids in tholeiitic
basalts at pressures up to 500 MPa. Chem Geol 277(1):115–125
Soule SA, Fornari DJ, Perfit MR, Ridley WI, Reed MH, Cann JR
(2006) Incorporation of seawater into mid-ocean ridge lava flows
during emplacement. Earth Planet Sci Lett 252(3–4):289–307.
https://doi.org/10.1016/j.epsl.2006.09.043
Stakes D, Vanko DA (1986) Multistage hydrothermal altera-
tion of gabbroic rocks from the failed Mathematician
ridge. Earth Planet Sci Lett 79(1–2):75–92. https://doi.
org/10.1016/0012-821X(86)90042-7
Stein CA, Stein S (1994) Constraints on hydrothermal heat flux through
the oceanic lithosphere from global heat flow. J Geophys Res
Solid Earth 99(B2):3081–3095
Stolper E (1982) The speciation of water in silicate melts. Geochim
Cosmochim Acta 46(12):2609–2620
Stroncik NA, Haase KM (2004) Chlorine in oceanic intraplate basalts:
constraints on mantle sources and recycling processes. Geology
32(11):945–948. https://doi.org/10.1130/g21027.1
Stroncik NA, Niedermann S (2016) Atmospheric contamination of the
primary Ne and Ar signal in mid-ocean ridge basalts and its
implications for ocean crust formation. Geochim Cosmochim
Acta 172:306–321. https://doi.org/10.1016/j.gca.2015.09.016
Sun WD, Binns RA, Fan AC, Kamenetsky VS, Wysoczanski R, Wei
GJ, Hu YH, Arculus RJ (2007) Chlorine in submarine volcanic
glasses from the eastern manus basin. Geochim Cosmochim Acta
71(6):1542–1552. https://doi.org/10.1016/j.gca.2006.12.003
Urann BM, Le Roux V, Hammond K, Marschall HR, Lee C-TA, Mon-
teleone BD (2017) Fluorine and chlorine in mantle minerals and
the halogen budget of the Earth’s mantle. Contrib Miner Petrol
172(7):51. https://doi.org/10.1007/s00410-017-1368-7
van der Zwan FM, Fietzke J, Devey CW (2012) Precise measurement of
low (< 100 ppm) chlorine concentrations in submarine basaltic
glass by electron microprobe. J Anal At Spectrom 27:1966–1974
van der Zwan FM, Devey CW, Augustin N, Almeev RR, Bantan RA,
Basaham A (2015) Hydrothermal activity at the ultraslow- to
slow-spreading Red Sea Rift traced by chlorine in basalt. Chem
Geol 405:63–81. https://doi.org/10.1016/j.chemgeo.2015.04.001
Vine F, Moores E (1972) A model for the gross structure, petrology,
and magnetic properties of oceanic crust. Geol Soc Am Mem
132:195–206
Wanless V, Perfit M, Ridley W, Klein E (2010) Dacite petrogenesis
on mid-ocean ridges: evidence for oceanic crustal melting and
assimilation. J Petrol 51(12):2377–2410
Wanless V, Perfit M, Ridley W, Wallace P, Grimes C, Klein E (2011)
Volatile abundances and oxygen isotopes in basaltic to dacitic
lavas on mid-ocean ridges: the role of assimilation at spreading
centers. Chem Geol 287(1):54–65
Weaver SJ, Langmuir CH (1990) Calculation of phase equilibrium in
mineral-melt systems. Comput Geosci 16(1):1–19
Yamashita S, Kitamura T, Kusakabe M (1997) Infrared spectroscopy
of hydrous glasses of arc magma compositions. Geochem J Jpn
31:169–174
Zhang C, Wang L-X, Marks MAW, France L, Koepke J (2017) Vola-
tiles ­(CO2, S, F, Cl, Br) in the dike-gabbro transition zone at
IODP Hole 1256D: magmatic imprint versus hydrothermal influ-
ence at fast-spreading mid-ocean ridge. Chem Geol 459:43–60.
https://doi.org/10.1016/j.chemgeo.2017.04.002
Zhao M, Canales JP, Sohn RA (2012) Three-dimensional seis-
mic structure of a Mid-Atlantic Ridge segment characterized
by active detachment faulting (Trans-Atlantic Geotraverse,
25°55′N–26°20′N). Geochem Geophys Geosyst 13(11). https://
doi.org/10.1029/2012GC004454
13