Fractionation of Nb and Ta from Zr and Hf at

JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
PAGES 221–232
2001
Fractionation of Nb and Ta from Zr and Hf
at Mantle Depths: the Role of Titanian
Pargasite and Kaersutite
M. TIEPOLO1∗, P. BOTTAZZI2, S. F. FOLEY3, R. OBERTI2,
R. VANNUCCI1,2 AND A. ZANETTI2
1
DIPARTIMENTO DI SCIENZE DELLA TERRA, UNIVERSITÀ DI PAVIA, VIA FERRATA 1, I-27100 PAVIA, ITALY
2
CNR-CENTRO DI STUDIO PER LA CRISTALLOCHIMICA E LA CRISTALLOGRAFIA (CSCC), VIA FERRATA 1,
I-27100 PAVIA, ITALY
3
MINERALOGISH-PETROLOGISCHES INSTITUT, UNIVERSITÄT GÖTTINGEN, GOLDSCHMIDTSTRASSE 1,
37077 GÖTTINGEN, GERMANY
RECEIVED NOVEMBER 28, 1999; REVISED TYPESCRIPT ACCEPTED JUNE 26, 2000
Selective enrichment or depletion in either Zr and Hf (HFSE4+)
or Nb and Ta (HFSE5+) is a feature commonly observed in many
mantle-derived melts and amphiboles occurring as either disseminated
minerals in mantle xenoliths and peridotite massifs or in vein
assemblages cutting these rocks. The fractionation of Nb from Zr
seen in natural mantle amphiboles suggests that their incorporation
is governed by different crystal-chemical mechanisms. An extensive
set of new partitioning experiments between pargasite–kaersutite and
melt under upper-mantle conditions shows that HFSE incorporation
and fractionation depends on amphibole major-element composition
and the presence or absence of dehydrogenation. Multiple regression
analysis shows that Amph/LDNb/Zr is strongly dependent on the mgnumber of the amphibole as a result of a combination of amphibole
and melt structure effects, so that the following generalizations
apply: (1) high-mg-number amphiboles crystallized from unmodified
mantle melts more easily incorporate Zr relative to Nb leading to
an increase of the Nb/Zr ratio in the residual melt; (2) low-mgnumber amphiboles, such as those found in veins cutting peridotites,
may strongly deplete the residual melt in Nb and cause very low
Nb/Zr in residual melts. Implications and applications to mantle
environments are discussed.
INTRODUCTION
KEY WORDS: trace elements; high field strength elements; partition coefficients; amphibole; upper mantle
Because of their similar crystal-chemical properties (i.e.
high charge and ionic radius), high field strength elements
(HFSE: Ti, Zr, Hf, Nb and Ta) are considered to behave
as a coherent group of substituents in rock-forming silicate
minerals, and thus are not expected to fractionate greatly
during partial melting and crystallization processes (Sun
& McDonough, 1989). Nevertheless, fractionation of
HFSE4+ from HFSE5+ is a common feature in many
primitive mantle melts such as island-arc volcanics and
intra-plate alkaline melts, suggesting either the presence
of a solid phase with S/LDNb,Ta significantly different from
S/L
DZr,Hf, or of two distinct phases controlling these element
pairs. Experimental investigations of HFSE decoupling
have been so far focused on Ti minerals (rutile and
ilmenite), because of their high compatibility for HFSE
and their consequent capability of altering HFSE signatures of equilibrium melts. Rutile has been shown to
have partition coefficients for Ta and Nb exceeding that
for Zr by one to two orders of magnitude, with Zr a
further one to two orders higher than other trace elements
( Jenner et al., 1993; Foley et al., 2000). Rutile could thus
cause significant decoupling between Nb and Zr, and
has the potential to produce the pattern typical of islandarc basalts (Foley & Wheller, 1990; Foley et al., 2000), in
which Nb and Ta show much stronger deviation from
∗Corresponding author. Telephone: +39 0382 505867 Fax: +39
0382 505890. E-mail: [email protected]
 Oxford University Press 2001
JOURNAL OF PETROLOGY
VOLUME 42
the behaviour of other trace elements than do Zr and
Hf (Briqueu et al., 1984).
However, rutile and other Ti-bearing minerals are
rarely observed in mantle assemblages; rutile is not expected to be stable in mantle peridotites as a result of
reaction with olivine to form orthopyroxene and ilmenite,
whereas ilmenite has to be more abundant than typically
seen in xenoliths if it is to explain the HFSE anomalies
(mainly Nb and Ta) observed in island-arc volcanic rocks
(Ayers, 1998). Among other mineral candidates, spinel
has very similar S/LDHFSE, which assume high partition
coefficients only in Ti-rich spinel compositions not found
in mantle peridotites (Horn et al., 1994; Nielsen & Beard,
2000), whereas clinopyroxene and olivine cannot significantly alter the HFSE signature of the equilibrium
liquids, because of their very low S/LDHFSE.
Volatile-bearing minerals such as amphibole and phlogopite have been recently recognized as important repositories for HFSE (Brenan et al., 1995; Foley et al.,
1995; Ionov & Hofmann, 1995; LaTourrette et al., 1995;
Vannucci et al., 1995; Ionov et al., 1997), although their
retention behaviour toward HFSE has not yet been
investigated in detail. Amphibole has comparable (or
slightly lower) S/LDTi and significantly higher S/LDZr,Hf and
S/L
DNb,Ta than phlogopite, and may, therefore, play a
leading role not only in hosting HFSE but also in their
fractionation.
Insight into site preferences and crystal-chemical mechanisms of incorporation for REE3+, HFSE4+ and HFSE5+
in pargasite–kaersutite has been recently provided (Bottazzi et al., 1999; Oberti et al., 2000; Tiepolo et al.,
2000) by an extensive project combining the results of
partitioning experiments and structure refinements (site
dimensions and site populations) of the experimentally
synthesized amphiboles with the lattice-site elastic-strain
theory. The aim of this paper is to highlight the differing
behaviour of Nb (Ta) and Zr (Hf ), and to discuss the
factors that can fractionate HFSE4+ from HFSE5+ in
pargasite and kaersutite under mantle conditions.
HFSE DISTRIBUTION IN
AMPHIBOLES FROM UPPERMANTLE ASSEMBLAGES
An increasing number of data on the HFSE incorporation
in amphibole have been made available recently because
of two main advances in analytical techniques: the small
number of separate grains now required for bulk analysis
by inductively coupled plasma mass spectrometry (ICPMS) (Ionov & Hofmann, 1995; Ionov et al., 1997) and
the wide distribution of high-sensitivity microanalytical
techniques such as secondary ion mass spectrometry
(SIMS) and laser ablation microprobe ICP-MS (LAMICP-MS) (Witt-Eickschen & Harte, 1994; Vannucci et
NUMBER 1
JANUARY 2001
Fig. 1. Nb and Ti enrichment with respect to Zr in mantle amphibole
available in the literature (see text for references). Φ, amphiboles from
peridotite massifs; Χ, amphibole from xenoliths. The line represents
primordial mantle (Hofmann, 1988).
al., 1995; Vaselli et al., 1995; Zanetti et al., 1995; Chazot
et al., 1996). Figure 1 shows the relative behaviour of Zr
vs Nb and Zr vs Ti contents observed in amphiboles
(Zanetti et al., 1996; Ionov et al., 1997; Mazzucchelli et
al., 1999) from peridotite massifs and mantle xenoliths.
No clear relation is observed in the two plots. However,
a comparison with the reference lines for Primordial
Mantle (Zr/Nb = 15·7 and Ti/Zr = 112; Hofmann,
1988) shows a general enrichment in Nb relative to Zr
and a frequent enrichment in Ti relative to Zr. As the
enrichment in Nb and Ti is widespread and not restrained
to peculiar P, T, X conditions, we conclude that it must
be governed by a crystal-chemical mechanism depending
on other intensive factors. In this respect, it is worth
noting that the presence of partial dehydrogenation (i.e.
of the oxy-component at the O3 site, O3O2−) is ubiquitous
in pargasites and kaersutites, occurring either in disseminated or in vein amphiboles in mantle assemblages
and functional to the stability of amphibole phase at high
T and low f H2. [For a detailed discussion of dehydrogenation in amphibole, see Oberti et al. (2000).]
Figure 2 shows the relative abundances of a series of
geochemically relevant elements in coexisting amphiboles
222
TIEPOLO et al.
FRACTIONATION OF HFSE4+ FROM HFSE5+
EXPERIMENTALLY DETERMINED
AMPH/L
D HFSE
Fig. 2. Trace element incorporation in coexisting amphibole and
clinopyroxene pairs from the literature [Ionov et al. (1997) and references
therein].
and clinopyroxenes from mantle assemblages available in
the literature [Ionov et al. (1997) and references therein].
Although these two-mineral D values relevant for subsolidus conditions cannot be considered totally analogous
to Amph/CpxD values inferred from experimentally determined partition coefficients (Green, 1995), similar values have been also obtained for amphibole–clinopyroxene
pairs obtained from our high-pressure experiments (Tiepolo et al., 2000).
As a whole, the ranges of D values observed for both
natural and synthetic samples are in agreement with
present knowledge on mechanisms of trace-element incorporation in amphibole and clinopyroxene based on
simple crystal-chemical considerations about ionic radius
and charge. Amph/CpxDREE are in the range of 1–2, Amph/
Cpx
DSr in the range of 1–3, with an average value of 2·6
(Ionov et al., 1997). This can be explained by incorporation of such elements according to the same
crystal-chemical mechanisms in the two phases, that is,
into the analogous eight-fold co-ordinated M2 site in
Cpx and M4 site in Amph, which have multiplicity 1:2
per formula unit (p.f.u.), and with the higher compliance
for large cations of the amphibole structure. Amph/CpxDZr
is frequently close to one and never exceeds two, in
agreement with incorporation into the analogous octahedral M1 site in Cpx and M2 site in amphibole, again
with multiplicity 1:2 p.f.u. (Oberti et al., 2000). Amph/CpxD
values for Nb and Ti show a far larger variation (up to
about 10 and 100, respectively). This is in agreement
with their incorporation in amphibole according to one
or more mechanisms not available for clinopyroxenes,
as will be discussed in a following section.
Titanian pargasites and kaersutites were synthesized,
using a piston-cylinder apparatus, at the MineralogischPetrologisches Institut, University of Göttingen, from two
bulk-rock compositions, an alkali–olivine basalt (472213a, Wedepohl, 1983) and a basanite (WR13-141,
Wörner et al., 1989) at 1·4 GPa pressure and temperature
from 850 to 1070°C. The natural rock powders and
synthetic mixtures of similar major element composition
were doped with Nb, Ta, Zr and Hf in the range
70–700 ppm for each element. The synthetic starting
compositions were also varied in different runs along
the compositional vectors K2O/(Na2O + K2O), MgO/
(MgO + FeO) and TiO2/(TiO2 + SiO2) to cover the
maximum possible range of natural amphiboles and
exchange mechanisms within them. Details of the synthesis and equilibrium conditions have been given in the
studies by Bottazzi et al. (1999), Tiepolo et al. (1999) and
Oberti et al. (2000), to which the reader is referred for
more information on microanalytical techniques, X-ray
analysis and structure refinement. Amphibole–glass pairs
were characterized for major elements by electron probe
(EMP), and trace and volatile (H, F, and Cl) elements
were quantified by ion probe (SIMS) in Pavia. Singlecrystal structure refinement (SREF) was performed on
at least one crystal from each experimental charge. The
amount of H2O (wt %) was thus independently estimated
by SIMS analyses, and the amount of O3O2− was checked
by comparison of structural variations and site populations obtained by SREF [see Tiepolo et al. (1999) and
Oberti et al. (2000) for details].
The experimental run temperatures were approached
slowly from higher temperatures to facilitate the growth
of large, homogeneous amphibole crystals. Run products
show a degree of crystallization lower than 50%; amphibole is the dominant phase, the last in the crystallization sequence after olivine and clinopyroxene. No
significant major-element zoning was found by EMP
investigations of both glasses and crystals. Their homogeneity and the regular crystal morphology are consistent
with equilibrium conditions.
Amphiboles are titanian pargasites and kaersutites according to the nomenclature scheme of Leake et al. (1997);
the oxy-component ranges from 0·61 to 1·08 atoms per
formula unit (a.p.f.u.), and is 0·24 a.p.f.u. in the F-rich
amphibole sample 16 (Tiepolo et al., 2000). Glasses range
in composition from basalt through trachyandesite to
trachyte, with mg-number in the range 0·08–1 and analysed H2O in the range 2·4–6·1 wt %. Amphibole and
glass composition strongly reflects that of the starting
materials. Slight changes in composition (mainly mgnumber and SiO2) were also observed in experimental
products derived from the same starting material (e.g. the
223
JOURNAL OF PETROLOGY
VOLUME 42
Table 1: Sample codes and relevant parameters
of melt (Xnf/X) and amphibole (mg-number)
Sample
Code
Teq (°C)
Xnf /Xmelt
1
A-N-melt∗
1015
0·55
0·75
A-N-synth
1015
0·62
0·52
3
A-N-melt
1015
0·54
0·74
4
A-K-1·00
1015
0·59
0·62
5
A-K-0·81
1015
0·61
0·63
6
A-K-0·71
1015
0·62
0·57
7
A-M-0·45
950
0·63
0·36
8
B-N-melt∗
1015
0·65
0·55
9
B-T-0·89
975
0·66
0·49
10
B-T-0·89
1015
0·65
0·53
11
B-T-0·89
1035
0·63
0·56
12
B-T-0·89
1055
0·61
0·60
13
B-T-0·94
1015
0·59
0·55
14
B-T-0·94
1035
0·68
0·52
15
B-T-0·94
1055
0·62
0·58
16
B-T-0·94
1075
0·50
0·87
17
B-T-0·97
975
0·65
0·43
18
B-T-0·97
1015
0·62
0·50
19
B-M-0·45
1045
0·59
0·50
20
B-M-0·75
1050
0·59
0·76
21
B-M-0·90
1050
0·62
0·89
22
B-M-1·00
1070
0·59
1·00
23
B-K-1·00
1070
0·58
0·57
24
B-K-0·50
1030
0·58
0·59
25
B-K-0·81
1030
0·55
0·65
JANUARY 2001
Table 2: Major element composition of three
glasses and amphiboles selected for their
representative behaviour towards mg-numberAmph
and SiO2glass
mg-numberAmph
2
NUMBER 1
Sample:
16
4
4
17
17
Glass
Amph
Glass
Amph
Glass
SiO2
43·44
41·53
39·36
50·40
38·70
TiO2
3·86
5·39
5·33
2·07
2·94
0·44
Al2O3
12·35
13·22
14·80
18·62
15·16
19·42
—
—
53·20
Cr2O3
0·01
FeOT
4·82
MnO
0·01
MgO
18·17
7·98
11·67
2·25
8·14
CaO
11·32
9·54
8·84
4·88
9·57
2·76
Na2O
2·62
2·99
3·73
4·05
2·96
5·15
K2O
1·55
1·38
0·03
H2O
1·06
3·65
1·01
—
F
—
8·01
—
0·01
12·88
0·01
—
9·92
—
0·06
19·06
0·02
1·36
1·27
3·42
1·23
3·03
1·66
—
0·03
—
0·06
−0·06
—
–0·01
—
–0·03
—
Total
100·81
93·68
97·68
92·24
99·08
98·02
Si
6·23
5·96
5·93
Al
1·77
2·04
2·07
T
8·00
8·00
8·00
Al
0·31
0·60
0·66
0·27
0·56
—
Ti
0·42
0·61
0·34
Mg
3·75
2·47
1·86
Fe2+
0·51
1·05
1·58
M(1,2,3)
5·00
5·00
Mg
0·12
0·16
The code is a combination of: (1) the composition of the
starting material (A, olivine alkali basalt 472213a; B, basanite
WR13-141); (2) the vector along which the composition was
varied [N, natural composition; K, K2O/(K2O + Na2O); M,
MgO/(MgO + FeOT); T, SiO2/(SiO2 + TiO2)]; (3) the value of
the A/(A + B) ratio between the two oxides that were varied
(melt indicates rock powder; ∗not doped).
9·25
—
–O=F
Fe3++Cr
T-series of samples) at different equilibrium temperature,
and this is mainly related to the different proportions of
solid and residual liquid in the system. Sample codes,
synthesis conditions and compositional characteristics
relevant to the present discussion are shown in Table 1.
In Table 2 representative major-element compositions
of amphibole and equilibrium glasses are reported; for
the complete dataset and geometric variables the reader
is referred to tables 2–4 of Tiepolo et al. (2000).
Table 3 lists the Amph/LDHFSE calculated from amphiboles
and glass at equilibrium in our experimental charges.
The relative variation of Amph/LDNb and Amph/LDZr as a
function of Xnf/X, the sum of the molar fractions of
network-forming cations (i.e. Si and the fraction of Al
16
Amph
5·00
—
Fe2++Mn
0·07
0·31
0·30
Ca
1·74
1·43
1·57
Na
0·07
0·10
0·13
M4
2·00
2·00
2·00
Na
0·65
0·99
0·75
K
0·28
0·01
0·25
A
0·93
1·00
1·00
OH
1·01
1·04
1·26
F
0·75
0·02
0·04
O
0·24
0·94
0·70
X
2·00
2·00
2·00
mg-no.
0·87
0·64
0·62
—
0·29
0·43
0·21
The whole dataset has been reported by Tiepolo et al. (2000).
that can be balanced by Na and K; Nielsen, 1990)
normalized to the molar fraction of total number of
cations in the melt is shown in Fig. 3. Xnf/X is a simplified
expression of the degree of polymerization of the melt.
224
TIEPOLO et al.
FRACTIONATION OF HFSE4+ FROM HFSE5+
It differs from the widely quoted NBO/T in that the
contributions of Fe3+ and H-speciation are neglected.
Amph/L
DNb increases with increasing Xnf/X by a factor of
12 whereas Amph/LDZr increases by a factor of four. Nb
and Zr behave most differently in samples 8 and 17,
where SiO2 contents are high and TiO2 contents in the
melt are <1%. In Ti-poor samples, Amph/LDNb values
clearly deviate from the trend defined by the other
samples and even exceed unity, providing evidence that
amphibole may be a repository for Nb. In contrast,
Amph/L
DZr values slightly increase with Ti depletion in the
melt, so that Zr does not become compatible, at least
in our dataset. In the two peculiar samples crystallized
from Ti-poor melts, the maximum increase observed for
Amph/L
DNb and Amph/LDZr is 26 and six times, respectively.
The insets of Fig. 3 show variations in Amph/LDZr and
in Amph/LDNb as a function of Amph/LDHf and Amph/LDTa,
respectively; Amph/LD values for Zr and Hf are linearly
related with a slope of 1·7, whereas those for Nb and
Ta are not linearly related. Amph/LDNb/Ta values are a
function of major-element composition of the amphibole
and can be predicted according to Tiepolo et al. (2000).
Thus, the following discussion is focused on Nb and Zr
as indicator elements for the behaviour of all HFSE.
Although the variation of temperature is coupled with
a change in melt and crystal composition, its effect on
Amph/L
D can be assumed to be negligible relative to that
of crystal-chemistry in our experimental charges as a
result of the restricted temperature range (<100°C in the
systems with the same starting composition).
DISCUSSION
HFSE site preference in amphibole
Titanium can partition among the three octahedral M1,
M2 and M3 sites and the tetrahedral T2 site in amphiboles. Detailed crystal-chemical work has shown that
its incorporation is ruled by the following exchange
vectors:
(1) M1Ti4+ O3O2−2 M1(Mg, Fe)2+–1 O3OH−–2;
(2) M2,M3Ti4+ T1,T2Al3+2 M2,M3(Mg, Fe)2+–1 T1,T2Si4+–2;
(3) T2Ti4+ T2Si4+–1;
whereby mechanism (3) is active only in richterite [see
Tiepolo et al. (1999) and Oberti et al. (2000) for more
details]. At upper-mantle conditions, partial dehydrogenation seems to be a requisite for amphibole
stability, and the major proportion of titanium (65–80%
depending on f H2 and T conditions) is involved in
mechanism (1). The occurrence of M3Ti is restricted to
Fe-depleted systems at high T, in which Al has a strong
preference for tetrahedral co-ordination, but is negligible
in most amphiboles. The crystal-chemical evidence allowing accurate determination of Ti partitioning has
been discussed by Tiepolo et al. (1999).
To infer the site preference for Zr4+ and Hf 4+ at the
trace-element level, Oberti et al. (2000) compared the
aggregate ionic radius at the various sites measured by
SREF (obtained by subtracting the ionic radius of [4]O2−,
0·138 nm, from the mean bond lengths) with their
optimum ionic radius calculated from partition coefficients for homovalent cations entering the same structural site according to the elastic-strain model of Blundy
& Wood (1994). Oberti et al. concluded that Zr4+ and
Hf 4+ have the same site preference as the fraction of
Ti4+ not related to dehydrogenation and occur at M2.
The ordering of Zr at M2 in amphibole and at M1 in
clinopyroxene as a consequence of the same crystalchemical mechanism, namely the temperature-dependent
Al substitution for Si, explains well the observation that
Amph/Cpx
DZr values do not exceed two (Fig. 2). Recent
work on Zr-rich arfvedsonites from Greenland (R. Oberti,
unpublished data, 1999) confirms this site assignment.
Nb and Ta are the only R5+ trace elements available,
and thus we cannot decipher their site preference by
comparing mean bond lengths with site characteristics
derived from the elastic-strain theory. However, the
Amph/Cpx
DNb values in Fig. 2 suggest that their incorporation into amphibole is ruled by a crystal-chemical
mechanism not available in clinopyroxene. Tiepolo et
al. (2000) showed that Amph/LDNb,Ta values are strongly
correlated with Amph/LDTiM1, and poorly correlated
with Amph/LDTiM2,M3, and that Amph/LDNb/Ta values correlate
with the dimension of the M1 site. Both the oxy component and Amph/LDNb increase by a factor of three in two
synthetic amphiboles in equilibrium with melts with
nearly the same SiO2 content (numbers 16 and 25). This
evidence suggests that Nb and Ta are incorporated into
the M1 site, and contribute together with M1Ti to the
electroneutrality of the O3 site in which O2− and not
OH− occurs. The peculiar crystal-chemical mechanism
that governs the incorporation of Nb and Ta into the
amphibole structure makes their partition coefficients
very sensitive to compositional changes in the melt,
especially the TiO2 content. The lack of OH groups
stabilizes the amphibole structure at upper-mantle conditions. When the contents of high charge cations in the
melt are low, their Amph/LD must increase strongly to allow
amphibole crystallization. Consequently, Nb may even
become compatible when the TiO2 content and the f H2
in the melt are low (Fig. 3).
Observation and prediction of decoupling
between DNb( Ta) and DZr(Hf) in amphibole
In our experimental dataset, Amph/LDNb/Zr values vary by
about one order of magnitude (from 0·2 to 1·8), thus
confirming that amphibole can decouple Nb from
Zr during mantle processes. The range of variation for
225
JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
JANUARY 2001
Table 3: Experimentally determined
Amph/L
Nb
Ta
Zr
Hf
Nb/Zr
1
0·138 ± 0·002
—
0·188 ± 0·012
—
0·734 ± 0·009
2
0·515 ± 0·049
0·395 ± 0·027
0·374 ± 0·004
0·667 ± 0·048
1·377 ± 0·068
3
0·139 ± 0·024
0·115 ± 0·019
0·153 ± 0·025
0·238 ± 0·039
0·913 ± 0·031
4
0·363 ± 0·025
0·320 ± 0·023
0·476 ± 0·108
0·792 ± 0·215
0·764 ± 0·085
5
0·344 ± 0·038
0·315 ± 0·023
0·451 ± 0·016
0·757 ± 0·020
0·762 ± 0·032
6
0·447 ± 0·015
0·349 ± 0·005
0·418 ± 0·016
0·636 ± 0·023
1·070 ± 0·024
7
0·708 ± 0·071
0·483 ± 0·068
0·419 ± 0·022
0·679 ± 0·052
1·690 ± 0·125
8
1·135
—
0·553 ± 0·146
—
2·054 ± 0·299
Sample
DHFSE values for the samples of this work
9
0·703 ± 0·133
0·490 ± 0·083
0·636 ± 0·188
1·022 ± 0·336
1·106 ± 0·255
10
0·525 ± 0·094
0·419 ± 0·098
0·520 ± 0·127
0·857 ± 0·300
1·009 ± 0·160
11
0·386 ± 0·031
0·336 ± 0·021
0·436 ± 0·048
0·715 ± 0·144
0·887 ± 0·051
12
0·403 ± 0·008
0·365 ± 0·022
0·405 ± 0·013
0·670 ± 0·040
0·994 ± 0·015
13
0·440 ± 0·023
0·357 ± 0·032
0·370 ± 0·036
0·593 ± 0·041
1·189 ± 0·051
14
0·605 ± 0·021
0·409 ± 0·008
0·584 ± 0·049
0·900 ± 0·084
1·035 ± 0·055
15
0·391 ± 0·083
0·326 ± 0·082
0·353 ± 0·094
0·553 ± 0·125
1·108 ± 0·138
16
0·062 ± 0·003
0·062 ± 0·003
0·148 ± 0·028
0·258 ± 0·008
0·421 ± 0·012
17
1·628 ± 0·127
0·999 ± 0·109
0·887 ± 0·098
1·554 ± 0·155
1·837 ± 0·295
18
0·691 ± 0·049
0·543 ± 0·052
0·568 ± 0·048
0·920 ± 0·103
1·216 ± 0·084
19
0·468 ± 0·083
0·361 ± 0·033
0·374 ± 0·042
0·621 ± 0·152
1·251 ± 0·116
20
0·266 ± 0·016
0·283 ± 0·028
0·338 ± 0·010
0·559 ± 0·011
0·788 ± 0·014
21
0·174 ± 0·027
0·231 ± 0·035
0·514 ± 0·076
0·947 ± 0·853
0·339 ± 0·0277
22
0·128 ± 0·022
0·180 ± 0·027
0·439 ± 0·111
0·773 ± 0·139
0·291 ± 0·033
23
0·377 ± 0·081
0·339 ± 0·015
0·473 ± 0·019
0·819 ± 0·022
0·796 ± 0·066
24
0·289 ± 0·017
0·248 ± 0·002
0·359 ± 0·084
0·626 ± 0·243
0·804 ± 0·069
25
0·181 ± 0·013
0·166 ± 0·013
0·281 ± 0·006
0·509 ± 0·017
0·642 ± 0·009
Amph/L
DZr,Hf are calculated from a larger dataset from the same experimental charges from which selected crystals were
separated for SREF investigations (Oberti et al., 2000); Amph/LDNb,Ta are from Tiepolo et al. (2000); 1 standard deviations are
calculated considering at least two different points within each charge.
Amph/L
DNb is twice that of Amph/LDZr, suggesting that the
main factor affecting Amph/LDNb/Zr is incorporation of Nb
in amphibole.
Amph/L
DNb/Zr is expected to be a function of the composition and structure of both amphibole and melt as
Nb and Zr have different ionic charge (5+ and 4+,
respectively) and ionic radius [0·064 nm for [6]Nb and
0·072 nm for [6]Zr; although the radius of Nb may be
effectively increased in distorted sites (Shannon, 1976)].
It has been shown that the compatibility of trace elements
in the solid phase in equilibrium with highly polymerized
melt varies as a function of their ionic charge (Z ) and
ionic radius (r) ratio (Ryerson & Hess, 1978). Z/r is 7·8
for [6]Nb and 5·5 for [6]Zr, so that the compatibility of
Nb is expected to increase with increasing degree of melt
polymerization at a faster rate than for Zr. This is
confirmed in Fig. 4 by the positive (although highly
scattered) correlation of ln( Amph/LDNb/Zr) vs Xnf/X. Nevertheless, the scatter in the correlation implies that another
effect, namely the control of the crystal structure, plays
a role in fractionation of the Amph/LDNb/Zr ratio. Noticeably,
the two deviating samples (16 and 21) have the highest
mg-number, and thus the smallest octahedral sites. As a
result, the increase of Amph/LDNb/Zr expected from compositional effects (i.e. melt polymerization) is counterbalanced by the lower capability of high-Mg amphibole
to incorporate high amounts of Nb.
The effects of the crystal structure can be understood
by considering that element incorporation depends on
the availability not only of structural sites of suitable
dimensions in the crystal, but also of appropriate
crystal-chemical mechanisms allowing for local charge
balance. To decipher the role played by the crystal
structure, multiple-regression analysis was carried out
with ln( Amph/LDNb/Zr) as the dependent variable and all
the refined mean bond lengths as independent variables.
A combination of mean bond lengths of all the structural
sites was found to explain 92% of the ln( Amph/LDNb/Zr)
226
TIEPOLO et al.
FRACTIONATION OF HFSE4+ FROM HFSE5+
Fig. 5. The effect of the mg-numberAmph on the relative compatibility
of Nb and Zr in amphiboles.
Fig. 3. The effects of the degree of melt polymerization (Xnf/X ) on
the incorporation of Nb and Zr in amphibole. The relative dependences
of DNb vs DTa and DZr vs DHf are shown in the insets.
For the formulation of a predictive expression, which
may be widely applied in petrogenetic studies, it is better
to avoid the need of SREF analyses. A simple and yet
accurate tool of potentially widespread use can be derived
by expressing the relevant site dimensions in terms of
compositional factors that can be straightforwardly calculated from an electron microprobe analysis.
In upper-mantle environments similar to those of this
experimental dataset, dehydrogenation can be assumed
to be nearly constant, and <M1–O> can be considered
to be a direct function of the mg-number and of the total
Ti content of the amphibole; in fact, Mg and Fe2+ are
nearly equally distributed between the M1 and M3 sites
at high T (>800°C), and most Ti occurs at M1. The
other compositional factors affecting the relevant site
dimensions are the cummingtonite component [ M4(Fe2+
+ Mg)] and the K content at the A sites ( AK); the best
multiple regression equation with ln( Amph/LDNb/Zr) as the
dependent variable is
ln(Amph/LDNb/Zr)=−2·47 mg-no.Amph+
1·12M4(Fe2++Mg)Amph+1·19AK+0·81
(1)
(R2 = 0·93). Given the difficulty of calculating the cummingtonite component in absence of an accurate estimate
of the dehydrogenation, we provide also the dependence
of ln( Amph/LDNb/Zr) on the mg-number:
ln(Amph/LDNb/Zr) =−2·76 mg-numberAmph + 1·57 (2)
Fig. 4. The relation of melt structure to the relative compatibility of
Nb and Zr in amphiboles.
variability, whereby <M1–O> is by far the most important factor, explaining >64% of the variability.
(R2 = 0·86), which is an oversimplified model that may
confidently be applied (Fig. 5).
Although the crystal structure plays a leading role in
fractionating Nb and Zr, complete discrimination between the effects of amphibole and melt structures is
prevented by the correlation between their compositions.
It is expected that amphibole crystallization will cause
the residual melt to evolve towards high silica and alkali
227
JOURNAL OF PETROLOGY
VOLUME 42
contents and towards low mg-number. The mg-number
of the starting material was independently varied in our
experimental runs, and it roughly correlates with Xnf/X.
The strong correlation between the mg-number of the
melt and that of the amphibole shows that this variable
includes both melt and crystal effects.
Implications and applications: Nb/Zr
fractionation in upper mantle
Two major generalizations can be made from the foregoing discussion, which have important implications for
mantle studies at constant pressure and in a narrow range
of temperature:
(1) Nb and Ta can be incorporated in amphiboles in
amounts far greater than inferred from both the HFSE4+
behaviour and from general considerations based on
their ionic radii and charge. In particular, Nb becomes
compatible, whereas Zr remains incompatible, in amphibole crystallized from Ti-poor systems.
(2) Amph/LDNb/Zr in our dataset is not constant, varying
from values lower to those higher than unity as a function
of both amphibole crystal-chemistry and degree of polymerization of the melt. Multiple regression analysis
showed that the ability of amphibole to vary the Nb/Zr
ratio is mainly controlled by its mg-number value. Zr is
incorporated more easily than Nb in high-mg-number
amphiboles, with Amph/LDNb/Zr values down to 0·2. Nb
incorporation is strongly favoured in low-mg-number
amphiboles, with Amph/LDNb/Zr values up to 1·8.
Pargasite and kaersutite are stable up to 1150°C in
the 2·5–3·0 GPa range under water-undersaturated conditions (Niida & Green, 1999), so that the results of this
work have important applications for melt–rock reactions
in the upper mantle. At first sight, it could be argued
that amphibole in equilibrium with mantle peridotite is
not capable of generating low Nb/Zr ratios in coexisting
melts because of its high mg-number as a result of the
buffering of mg-number by other silicate minerals, particularly olivine. Amphiboles from peridotite massifs and
mantle xenoliths have mg-number broadly ranging from
0·80 to 0·90 (Ionov & Hofmann, 1995; Vannucci et al.,
1995; Woodland et al., 1996; Ionov et al., 1997), in
agreement with evidence from both subsolidus and nearsolidus experiments (mg-numberAmph from 0·86 to 0·90)
on mid-ocean ridge basalt (MORB) pyrolite (Wallace &
Green, 1991; Niida & Green, 1999). Amphibole crystallized from unmodified mantle melts (mg-number =
0·65–0·75) in equilibrium with olivine should have mgnumber >0·85–0·90 according to the KD( Fe/Mg) for amphibole–melt obtained from our experiments (>0·29).
Thus, Amph/LDZr should be higher than Amph/LDNb (both
being <1), and Nb and Zr contents as well as Nb/Zr
ratio in the residual melt should increase.
NUMBER 1
JANUARY 2001
However, mg-number values of the initial melt can be
significantly modified during its infiltration and migration
through peridotite because of crystallization and modification accompanying dissolution processes related to
Fe–Mg exchange with ambient peridotite. Two situations
can be envisaged: (1) migration of melts in magma
conduits and crystallization of amphibole in veins; (2)
reactive pervasive porous melt flow.
In the former case, a limited exchange with the host
peridotite is expected, but closed-system conditions are
approached only after a long length-scale migration along
a vein. On a restricted scale, the minerals crystallized at
vein edges may represent liquidus crystals from a melt
that is no longer seen close by, as it moved on during
the process of flow crystallization (Irving, 1980). As crystallization in the vein proceeds and the melt flows along
the vein, the composition of the melt is enriched in SiO2
and alkali, and its degree of polymerization increases.
Later amphiboles therefore will have lower mg-number,
and can incorporate increasing amounts of Nb. As the
initial amphiboles are expected to be rich in TiO2 to
allow partial dehydrogenation, the melt will evolve to
Ti-poorer compositions, and Amph/LDNb will exceed unity.
This process may explain the high Nb/Zr values frequently observed in vein amphiboles (Ionov & Hofmann,
1995; Wulff-Pedersen et al., 1999).
In the case of reactive porous flow, amphibole mgnumber will remain high in conditions of low porosity
because the interstitial melt is buffered at high mg-number
value by the surrounding peridotite. Lower mg-number
values may occur in the case of high melt/rock ratios
(Xu et al., 1998), a situation that could be considered
transitional to melt migration in veins. However, hydrous
basaltic melts with either ocean island basalt (OIB) or
MORB signatures migrating in oceanic and lithospheric
upper mantle have mg-number values too high to allow
significant fractionation of Nb from Zr by the amphibole.
An example of the effects on Nb/Zr variation induced
in newly formed amphibole and coexisting melt during
reaction between an infiltrating melt and the host peridotite is shown in Fig. 6. Under the very simplified
assumption of assimilation–fractional crystallization
(AFC) modelling (DePaolo, 1981), dissolution and mineral-forming reactions are simulated by pyroxene consumption and amphibole crystallization. Three different
sets of Amph/LD values were used to model the evolution
of the melt and the solid phase at different mg-number
values. In the case of intermediate mg-number, the Nb/
Zr value in the melt (Fig. 6a) and in the amphibole (Fig.
6b) is unaffected. The higher incompatibility of Nb
relative to that of Zr for high-mg-number amphibole
strongly increases Nb/Zr (up to 2·6 times) in the residual
melt towards high values. In the case of low-mg-number
amphiboles the residual melt is quickly depleted in Nb
as crystallization proceeds. At residual melt fractions of
228
TIEPOLO et al.
FRACTIONATION OF HFSE4+ FROM HFSE5+
Fig. 6. AFC modelling of peridotite dissolution and amphibole-forming
reaction produced by migration of melts through a slightly depleted
lherzolite. To illustrate the fractionation of the Nb/Zr ratio caused by
amphibole, the initial melt is set to have Nb/Zr = 1. The assimilation
rate (Ma/Mc) has been supposed equal to 0·2, whereas the consumed
assemblage consists of orthopyroxene and clinopyroxene (in proportions
of 0·8 and 0·2, respectively), as is expected in the case of migration of
OIB-like melts. The concentration of Nb and Zr in orthopyroxene and
clinopyroxene are from Rampone et al. (1995); Nb in orthopyroxene
is arbitrarily assumed to be 0·01 ppm. Amphibole has been assumed
to be the only crystallizing phase and three different sets of Amph/LD [i.e.
low mg-number (sample 17); medium mg-number (sample 12); high mgnumber (sample 21)] were used to model the evolution of the melt and
the solid phase under different mg-number conditions.
0·5, Nb/Zr in the melt is half of the initial value and it
is 10 times lower at the end of the process. The Nb/Zr
ratio in the amphibole shows the opposite behaviour; at
the beginning of the crystallization it is about two times
that of the melt, and quickly drops to 0·2 as crystallization
proceeds.
Our experimental dataset does not contain results for
high-mg-number and Ti-poor systems; however, it is
reasonable that pargasite and kaersutite crystallized under
these conditions, which require almost constant dehydrogenation during progressive crystallization, will incorporate increasing amounts of Nb and Ta to allow
local electroneutrality at the O3 site. If this hypothesis is
correct, then open-system melt migration by reactive
porous flow may also lead to crystallization of amphibole
significantly enriched in Nb and Ta relative to Zr and
Hf, provided that the interstitial melt is progressively
depleted in Ti by dissolution and mineral-forming reactions. The occurrence of amphiboles with high-mgnumber value and high Nb/Zr ratios disseminated
through mantle peridotites (Chazot et al., 1996; Laurora
et al., 1999; Litasov et al., 2000) strongly supports this
interpretation.
Silica- and alkali-rich melts occur as inclusions and
vein glasses in mantle xenoliths. Whatever the origin of
the xenolith glasses (i.e. melts produced by in situ melting
of mantle peridotite, breakdown of hydrous phases, reaction between peridotite and infiltrating basaltic melts),
there is a general consensus that they can act as effective
agents of mantle metasomatism (Schiano et al., 1994;
Zinngrebe & Foley, 1995; Wulff-Pedersen et al., 1999,
and references therein). Moreover, experimental evidence
has been recently provided that glasses with SiO2 contents
up to 55 wt % may represent melts in, or near, equilibrium
with mantle mineralogy in the upper reaches of the
upper mantle (Baker et al., 1995; Draper & Green, 1997;
Robinson et al., 1998; Hirschmann et al., 1999) and may,
therefore, be able to circulate through mantle peridotite.
Xenolith glasses indicate melts with a wide range of
compositions, with mg-number ranging from 0·25 to 0·90
with the peak around 0·60 (Draper & Green, 1997).
Injection of these relatively low-mg-number metasomatic
agents may be a suitable mechanism to crystallize pargasitic and kaersutitic amphiboles, which are able to
incorporate Nb to a much greater extent than Zr as the
system becomes progressively Ti depleted.
In subduction-related environments, the presence of
amphiboles with high Nb/Zr values is documented by
some xenolith occurrences. Xenoliths from sub-arc
mantle in Kamchatka (Kepezhinskas et al., 1996) contain
amphiboles with mg-number = 0·69 and NbN/ZrN values
of >2·5. Sp-facies mantle xenoliths from Neogene Southern Patagonian Plateau (Gobernator Gregores, Santa
Cruz, Argentina), which is in a back-arc position with
respect to the Chile trench, contain amphibole with NbN/
ZrN values up to 91, which Laurora et al. (2001) showed
to be formed during infiltration of slab-derived melts in
the mantle wedge. Alternatively, the crystallization of
high-Nb amphiboles in sub-arc mantle may be related
to the infiltration of a fluid phase in the mantle wedge
as suggested by Ionov & Hofmann (1995). Although
characterized by distinct water and silica contents relative
to a silicate melt, silica-rich aqueous fluids released from
the slab can be regarded as analogues of water-rich silicic
melts in that they should lead to the crystallization of
low-mg-number amphiboles capable of fixing Nb in the
overlying mantle-wedge peridotites.
229
JOURNAL OF PETROLOGY
VOLUME 42
CONCLUSIONS
In titanian pargasites and kaersutites stable at uppermantle conditions, HFSE are partitioned among three
independent octahedral sites depending on the endmember composition and the abundance of the oxycomponent. In this respect, they represent a special case
relative to other incompatible elements, which display
amphibole–melt partitioning behaviour more coherent
with that of clinopyroxene–melt. Nb and Ta enter the
M1 site, co-operating with Ti in providing local electroneutrality when H is lacking, whereas Zr and Hf have
the same site preference (M2) as the fraction of Ti4+ that
balances for the presence of tetrahedral Al. In Ti-depleted
systems, this decoupling results in an abrupt increase of
Amph/L
DNb relative to Amph/LDZr. Selective uptake of HFSE5+
by amphibole in veined mantle or during melt migration
by reactive porous flow is thus probable and provides a
simple explanation for the occurrence of both disseminated and vein amphiboles selectively enriched in
Nb relative to Zr in several mantle occurrences (Ionov
& Hofmann, 1995; Litasov et al., 2000).
The new partitioning data reveal that the accurate
prediction of the behaviour of HFSE requires a reasonable estimate of the water and titanium activity in the
melt. Amph/LD values of elements that are involved in
charge compensation of the oxy-component, namely Ti,
Nb and Ta, cannot be assumed to be constant during
the crystallization of hydrous phases from the interstitial
melt. Former crystallization of amphibole implies sink–
source effects for both H and Ti, which in turn either
affect Nb and Ta partitioning in amphibole crystallized
later or prevent the formation of new amphibole beyond
the reaction zone.
This mechanism may play a significant role in controlling the HFSE signatures of melts migrating through
the upper mantle, either lithospheric or oceanic, as well
as of magmas generated at convergent margins.
ACKNOWLEDGEMENTS
This paper greatly benefited from constructive comments
by two anonymous reviewers. Funding from this work
was provided by the Consiglio Nazionale delle Ricerche
to the CSCC laboratories, by the Ministero della Università e della Ricerca Scientifica e Tecnologica (project
‘Transformations in subducted materials and mass transfer to the mantle wedge’) to Riccardo Vannucci, and by
the Deutsche Forschungsgemeinschaft (Grant Fo 181/9)
to Steve Foley.
REFERENCES
Ayers, J. C. (1998). Trace element modeling of aqueous fluid–
peridotite interaction in the mantle wedge of subduction zones.
Contributions to Mineralogy and Petrology 132, 390–404.
230
NUMBER 1
JANUARY 2001
Baker, M. B., Hirshmann, M. M., Ghiorso, M. S. & Stolper, E.
M. (1995). Composition of near-solidus peridotite melts from
experiments and thermodynamic calculations. Nature 375, 308–311.
Blundy, J. & Wood, B. (1994). Prediction of crystal–melt partition
coefficients from elastic moduli. Nature 372, 452–454.
Bottazzi, P., Tiepolo, M., Vannucci, R., Zanetti, A., Brumm, R.,
Foley, S. F. & Oberti, R. (1999). Distinct local configurations
for heavy and light REE in amphibole and the prediction of
Amph/L
DREE. Contributions to Mineralogy and Petrology 137, 36–45.
Brenan, J. M., Shaw, H. F., Ryerson, F. J. & Phinney, D. L. (1995).
Experimental determination of trace-element partitioning between
pargasite and a synthetic hydrous andesitic melt. Earth and Planetary
Science Letters 135, 1–11.
Briqueu, L., Bougault, H. & Joron, J.-L. (1984). Quantification of
Nb, Ta, Ti and V anomalies in magmas associated with subduction
zones: petrogenetic implications. Earth and Planetary Science Letters
68, 297–308.
Chazot, G., Menzies, M. A. & Harte, B. (1996). Determination of
partition coefficients between apatite, clinopyroxene, amphibole,
and melt in natural spinel lherzolites from Yemen: implications
for wet melting of the lithospheric mantle. Geochimica et Cosmochimica
Acta 60, 423–437.
DePaolo, D. J. (1981). Trace element and isotopic effects of combined
wallrock assimilation and fractional crystallization. Earth and
Planetary Science Letters 53, 189–202.
Draper, D. S. & Green, T. H. (1997). P–T phase relations of silicic,
alkaline, aluminous mantle-xenolith glasses under anhydrous and
C–O–H fluid-saturated conditions. Journal of Petrology 38, 1187–
1224.
Foley, S. F. & Wheller, G. E. (1990). Parallels in the origin of the
geochemical signature of island arc volcanics and continental
potassic igneous rocks: the role of residual titanates. Chemical
Geology 85, 1–18.
Foley, S. F., Jenner, G. A., Konzett, J. & Sweeney, R. J. (1995).
Trace element partitioning in natural phlogopite- and K-richteritebearing xenoliths from southern African kimberlites. Sixth International Kimberlite Conference; Extended Abstracts, pp. 164–166.
Foley, S. F., Barth, M. G. & Jenner, G. A. (2000). Rutile/melt
partition coefficients for trace elements and an assessment of the
influence of rutile on the trace element characteristics of subduction
zone magmas. Geochimica et Cosmochimica Acta 64, 933–938.
Green, T. H. (1995). Experimental versus natural two mineral
partition coefficients—a ‘high teach’ controversy. International Geology
Review 37, 851–865.
Hirschmann, M. M., Ghiorso, M. S. & Stolper, E. M. (1999).
Calculation of peridotite partial melting from thermodynamic
models of minerals and melts. II. Isobaric variations in melts
near the solidus and owing to variable source composition. Journal
of Petrology 40, 297–313.
Hofmann, A. W. (1988). Chemical differentiation of the Earth: the
relationship between mantle, continental crust, and oceanic crust.
Earth and Planetary Science Letters 90, 297–314.
Horn, I., Foley, S. F., Jackson, S. E. & Jenner, G. A. (1994).
Experimentally determined partitioning of high field strength- and
selected transition elements between spinel and basaltic melt.
Chemical Geology 117, 193–218.
Ionov, D. A. & Hofmann, A. W. (1995). Nb–Ta-rich mantle
amphiboles and micas: implications for subduction-related metasomatic trace element fractionations. Earth and Planetary Science
Letters 131, 341–356.
Ionov, D. A., Griffin, W. L. & O’Reilly, S. Y. (1997). Volatilebearing minerals and lithophile trace elements in the upper
mantle. Chemical Geology 141, 153–184.
TIEPOLO et al.
FRACTIONATION OF HFSE4+ FROM HFSE5+
Irving, A. J. (1980). Petrology and geochemistry of composite
ultramafic xenoliths in alkalic basalts and implications for magmatic
processes within the mantle. American Journal of Science 280A,
389–426.
Jenner, G. A., Foley, S. F., Jackson, S. E., Green, T. H., Fryer,
B. J. & Longerich, H. P. (1993). Determination of partition
coefficients for trace elements in high pressure–temperature
experimental run products by laser ablation microprobe-inductively
coupled plasma-mass spectrometry (LAM-ICP-MS). Geochimica et
Cosmochimica Acta 58, 5099–5103.
Kepezhinskas, P., Defant, M. J. & Drummond, M. S. (1996).
Progressive enrichment of island arc mantle by melt–peridotite
interaction inferred from Kamchatka xenoliths. Geochimica et
Cosmochimica Acta 60, 1217–1229.
LaTourrette, T. Z., Hervig, R. L. & Holloway, J. R. (1995). Trace
element partitioning between amphibole, phlogopite, and basanite
melt. Earth and Planetary Science Letters 135, 13–30.
Laurora, A., Rivalenti, G., Mazzucchelli, M., Bottazzi, P., Barbieri,
M. A., Cingolani, C. & Vannucci, R. (1999). Carbonated peridotite
xenoliths from the mantle wedge: the Patagonia case. Ofioliti 24,
123–124.
Laurora, A., Mazzucchelli, M., Rivalenti, G., Vannucci, R., Zanetti,
A., Barbieri, M. A. & Cingolani, C. A. (2001). Metasomatism
and melting in carbonated peridotite xenoliths from the mantle
wedge: the Gobernador Gregores case (Southern Patagonia).
Journal of Petrology 42, 69–87.
Leake, B. E., Woolley, A. R., Arps, C. E. S., Birch, W. D., Gilbert,
M. C., Grice, J. D., Hawthorne, F. C., Kato, A., Kisch, H. J.,
Krivovichev, V. G., Linthout, K., Laird, J., Mandarino, J.,
Maresch, W. V., Nickel, E. H., Tock, N. M. S., Schumacher, J.
C., Smith, D. C., Stephenson, N. C. N., Ungaretti, L., Whittaker,
E. J. W. & Youzhi, G. (1997). Nomenclature of amphiboles:
report of the Subcommittee on amphiboles of the International
Mineralogical Association Commission on new minerals and
mineral names. American Mineralogist 82, 1019–1037.
Litasov, K. D., Foley, S. F. & Litasov, Y. D. (2000). Magmatic
modification and metasomatism of the subcontinental mantle
beneath the Vitim volcanic field (East Siberia): evidence from
trace element data on pyroxenite and peridotite xenoliths from
Miocene picrobasalt. Lithos (in press).
Mazzucchelli, M., Rivalenti, G., Zanetti, A., Vannucci, R. &
Cavazzini, G. (1999). Origin and significance of late noritic dikes
in the Baldissero peridotite massif (Ivrea–Verbano zone). Ofioliti
24, 129–130.
Nielsen, R. L. (1990). Simulation of igneous differentiation processes.
In: Nichols, J. & Russell, J. K. (eds) Modern Methods of Igneous
Petrology. Mineralogical Society of America, Reviews in Mineralogy 23,
65–105.
Nielsen, R. L. & Beard, J. S. (2000) Magnetite–melt HFSE
partitioning. Chemical Geology 164, 21–34.
Niida, K. & Green, D. H. (1999). Stability and chemical composition
of pargasitic amphibole in MORB pyrolite under upper mantle
conditions. Contributions to Mineralogy and Petrology 135, 18–40.
Oberti, R., Vannucci, R., Zanetti, A., Tiepolo, M. & Brumm, R.
C. (2000). A crystal-chemical re-evaluation of amphibole/melt
and amphibole/clinopyroxene DTi in petrogenetic studies. American
Mineralogist 85, 407–419.
Rampone, E., Hofmann, A. W., Piccardo, G. B., Vannucci, R.,
Bottazzi, P. & Ottolini, L. (1995). Petrology, mineral and isotope
geochemistry of the External Liguride peridotites (Northern
Apennines, Italy). Journal of Petrology 36, 81–105.
Robinson, J. A. C., Wood, B. J. & Blundy, J. D. (1998). The
beginning of melting of fertile and depleted peridotite at 1·5
GPa. Earth and Planetary Science Letters 155, 97–111.
Ryerson, F. J. & Hess, P. C. (1978). Implications of liquid–liquid
distribution coefficients to mineral–liquid partitioning. Geochimica
et Cosmochimica Acta 42, 921–932.
Schiano, P., Clochiatti, R., Shimizu, N., Weiss, D. & Mattielli, N.
(1994). Cogenetic silica-rich and carbonate-rich melts trapped in
mantle minerals in Kerguelen ultramafic xenoliths: implications
for metasomatism in the oceanic upper mantle. Earth and Planetary
Science Letters 123, 167–178.
Shannon, R. D. (1976). Revised effective ionic radii and systematic
studies of interatomic distances in halides and chalcogenides. Acta
Crystallographica 32A, 751–767.
Sun, S.-S. & McDonough, W. F. (1989) Chemical and isotopic
systematics of oceanic basalts: implications for mantle composition
and processes. In: Saunders, A. D. & Norry, M. J. (eds) Magmatism
in the Ocean Basins, Geological Society, London, Special Publications 42,
313–345.
Tiepolo, M., Zanetti, A. & Oberti, R. (1999). Detection, crystalchemical mechanism and petrological implications of [6]Ti4+
partitioning in pargasite and kaersutite. European Journal of Mineralogy
11, 345–354.
Tiepolo, M., Vannucci, R., Oberti, R., Foley, S. F., Bottazzi, P. &
Zanetti, A. (2000). Nb and Ta incorporation and fractionation
in titanian pargasite and kaersutite: crystal-chemical constraints
and implications for natural systems. Earth and Planetary Science
Letters 176, 185–201.
Vannucci, R., Piccardo, G. B., Rivalenti, G., Zanetti, A., Rampone,
E., Ottolini, L., Oberti, R., Mazzucchelli, M. & Bottazzi, P.
(1995). Origin of LREE-depleted amphiboles in the subcontinental
mantle. Geochimica et Cosmochimica Acta 59, 1763–1771.
Vaselli, O., Downes, H., Thirlwall, M., Dobosi, G., Coradossi, N.,
Seghedi, I., Szakacs, A. & Vannucci, R. (1995). Ultramafic
xenoliths in Plio-Pleistocene alkali basalts from the Eastern
Transylvanian Basin: depleted mantle enriched by vein metasomatism. Journal of Petrology 36, 23–53.
Wallace, M. E. & Green, D. H. (1991). The effect of bulk rock
composition on the stability of amphibole in the upper mantle:
implications for solidus positions and mantle metasomatism.
Mineralogy and Petrology 44, 1–19.
Wedepohl, K. H. (1983). Die chemische Zusammensetzung der
basaltischen Gesteine der nordlichen Hessichen Senke und ihrer
Umgebung. Geologisches Jahrbuch Hessen 111, 261–302.
Witt-Eickschen, G. & Harte, B. (1994). Distribution of trace
elements between amphibole and clinopyroxene from mantle
peridotites of the Eifel (Western Germany). Chemical Geology 117,
235–250.
Woodland, A. B., Kornprobst, J., McPherson, E., Bodinier, J. L. &
Menzies, M. A. (1996). Metasomatic interactions in the lithospheric
mantle; petrologic evidence from the Lherz Massif, French
Pyrenees. Chemical Geology 134, 83–112.
Wörner, G., Viereck, L., Hertogen, J. & Niephaus, H. (1989). The
Mt. Melbourne volcanic field (Victoria Land, Antarctica) II.
Geochemistry and magma genesis. Geologisches Jahrbuch 38,
395–433.
Wulff-Pedersen, E., Neumann, E.-R., Vannucci, R., Bottazzi, P. &
Ottolini, L. (1999). Silicic melts produced by reaction between
peridotite and infiltrating basaltic melts: ion probe data on glasses
and minerals in veined xenoliths from La Palma, Canary Islands.
Contributions to Mineralogy and Petrology 137, 59–82.
Xu, Y.-G., Menzies, M. A., Bodinier, J.-L., Bedini, R. M., Vroon,
P. & Mercier, J.-C. C. (1998). Melt percolation and reaction atop
a plume: evidence from the poikiloblastic peridotite xenoliths from
Borée (Massif Central, France). Contributions to Mineralogy and
Petrology 132, 65–84.
231
JOURNAL OF PETROLOGY
VOLUME 42
Zanetti, A., Vannucci, R., Oberti, R. & Dobosi, G. (1995). Traceelement compositions and crystal-chemistry of mantle amphiboles
from the Carpatho-Pannonian Region. Acta Vulcanologica 7,
265–276.
Zanetti, A., Vannucci, R., Bottazzi, P., Oberti, R. & Ottolini, L.
(1996). Infiltration metasomatism at Lherz as monitored by
232
NUMBER 1
JANUARY 2001
systematic ion-microprobe investigations close to a hornblendite
vein. Chemical Geology 134, 113–133.
Zinngrebe, E. & Foley, S. F. (1995). Metasomatism in mantle
xenoliths from Gees, West Eifel, Germany: evidence for the
genesis of calc-alkaline glasses and metasomatic Ca-enrichment.
Contributions to Mineralogy and Petrology 122, 79–96.