Geochemistry of rare high-Nb basalt lavas: Are they derived from a mantle wedge metasomatised by slab melts

Page 1

Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Geochemistry of rare high-Nb basalt lavas: Are they derivedfrom a mantle wedge metasomatised by slab melts?

Alan R. Hastie a,⇑, Simon F. Mitchell b, Andrew C. Kerr c, Matthew J. Minifie c,Ian L. Millar d

a School of Geography, Geology and the Environment, Kingston University, Penrhyn Road, Kingston upon Thames, Surrey KT1 2EE, UKb Department of Geography and Geology, University of the West Indies, Mona, Kingston 7, Jamaica

c School of Earth and Ocean Sciences, Cardiff University, Main Building, Park Place, Cardiff CF10 3YE, UKd NERC Isotope Geoscience Laboratories, Keyworth, Nottingham NG12 5GG, UK

Received 24 January 2011; accepted in revised form 13 June 2011; available online 23 June 2011

Abstract

Compositionally, high-Nb basalts are similar to HIMU (high U/Pb) ocean island basalts, continental alkaline basalts andalkaline lavas formed above slab windows. Tertiary alkaline basaltic lavas from eastern Jamaica, West Indies, known as theHalberstadt Volcanic Formation have compositions similar to high-Nb basalts (Nb > 20 ppm). The Halberstadt high-Nb bas-alts are divided into two compositional sub-groups where Group 1 lavas have more enriched incompatible element concen-trations relative to Group 2. Both groups are derived from isotopically different spinel peridotite mantle source regions, whichboth require garnet and amphibole as metasomatic residual phases. The Halberstadt geochemistry demonstrates that the lavascannot be derived by partial melting of lower crustal ultramafic complexes, metasomatised mantle lithosphere, subductingslabs, continental crust, mantle plume source regions or an upper mantle source region composed of enriched and depletedcomponents. Instead, their composition, particularly the negative Ce anomalies, the high Th/Nb ratios and the similar isoto-pic ratios to nearby adakite lavas, suggests that the Halberstadt magmas are derived from a compositionally variable spinelperidotite source region(s) metasomatised by slab melts that precipitated garnet, amphibole, apatite and zircon. It is suggestedthat high-Nb basalts may be classified as a distinct rock type with Nb > 20 ppm, intraplate alkaline basalt compositions, butthat are generated in subduction zones by magmatic processes distinct from those that generate other intraplate lavas.� 2011 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

In most modern subduction zones dehydration reactionsin the downgoing oceanic slab release aqueous fluid, whichenters the overlying mantle wedge, lowers the mantlesolidus and triggers partial melting to form continentaland island arc magmas (e.g. McCulloch and Gamble,1991; Pearce and Peate, 1995; Elliott, 2003; Tatsumi,2003). However, in several present-day convergent margins[e.g. Aleutian arc (Kay, 1978; Yogodzinski et al., 1995) andCentral American arc (Defant et al., 1992)] a suite of lavas

0016-7037/$ - see front matter � 2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.gca.2011.06.018

⇑ Corresponding author.E-mail address: [email protected] (A.R. Hastie).

with compositions similar to trondhjemite, tonalite, grano-diorite/dacites (TTG/Ds) is identified and have been termedadakites by Defant and Drummond (1990).

Defant et al. (1992) defined adakites as having SiO2

> 56 wt%, Al2O3 > 15 wt%, MgO < 6 wt%, low Y and hea-vy rare Earth elements (HREE) (Y and Yb < 18 and1.9 ppm, respectively), high Sr (>400 ppm) and low highfield strength elements (HFSE). Adakites are derived fromthe partial melting of mafic source material (either a sub-ducting slab or lower crust) which has been transformed intoeither amphibolite, garnet amphibolite or eclogite (e.g.Rapp et al., 1991, 1999; Atherton and Petford, 1993;Drummond et al., 1996; Smithies et al., 2003; Martinet al., 2005; Macpherson et al., 2006).

Page 2

Table 1Major and trace element compositions of the Halberstadt lavas analysed by ICP-OES and ICP-MS at Cardiff University using methodologiesdescribed in McDonald and Viljoen (2006). Major elements are re-calculated to total 100% anhydrous values (excluding LOI). The averageNEB and HNB values are from Castillo (2008) and the NEAB samples VAL55 and 8710 (that are classified as HNBs with our geochemicaldefinition) are from Kepezhinskas et al. (1996). JB-1a certified values are taken from McDonald and Viljoen (2006) and Govindaraju (1994).The JB-1a international standard data are based on 10 analyses carried out within the Jamaica analytical run. r.s.d. = relative standarddeviation. Detection limits are based on blank analyses.

AHHB01 AHHB02 AHHB03 AHHB04 AHHB05 AHHB07 AHHB08

Major elements (wt%)

SiO2 54.63 52.77 57.99 56.47 50.21 55.04 53.19TiO2 2.39 2.79 2.39 2.23 3.13 2.71 3.06Al2O3 14.24 16.13 14.68 13.37 15.45 16.25 15.23Fe2O3 9.91 12.63 10.20 8.53 10.55 9.42 8.43MnO 0.12 0.06 0.04 0.08 0.10 0.04 0.06MgO 8.43 11.17 8.71 8.50 8.78 8.38 7.30CaO 6.84 1.37 3.30 6.54 6.14 2.60 5.96Na2O 2.88 3.05 2.30 2.78 3.38 3.64 4.18K2O 0.07 0.03 0.05 0.03 0.11 0.09 0.10P2O5 0.30 0.34 0.41 0.36 0.73 0.33 0.73Total 99.81 100.33 100.07 98.87 98.58 98.50 98.24(SiO2)8.0 56.48 55.46 59.92 58.33 52.17 56.87 54.68

Trace elements (ppm)

Sc 20.7 23.5 21.5 18.7 18.1 24.2 18.0V 199 229 200 184 195 229 196Cr 113 125 127 116 165 169 168Co 33.1 38.4 30.2 29.6 35.5 34.1 27.2Ni 131 97 96 106 123 136 107Ga 16.2 18.9 16.8 15.0 19.3 16.4 17.3Rb 0.3 0.1 0.2 0.1 0.4 0.7 0.5Sr 177 146 97 110 161 167 171Y 20.1 17.9 23.5 22.3 19.9 26.5 21.3Zr 100.9 116.8 104.1 97.3 168.8 117.3 147.0Nb 28.82 33.36 29.50 27.56 45.97 33.69 43.36Ba 89 39 79 382 124 134 153La 17.91 11.76 13.87 13.54 27.61 19.31 24.11Ce 33.03 22.79 27.67 26.73 63.87 38.51 55.72Pr 4.28 3.08 3.84 3.65 8.93 5.09 7.74Nd 17.84 13.54 17.07 15.82 36.73 21.24 30.59Sm 4.40 3.79 4.75 4.32 7.71 5.35 6.21Eu 1.51 1.18 1.63 1.52 2.47 1.84 1.73Gd 4.32 3.72 4.91 4.43 6.39 5.41 5.02Tb 0.66 0.60 0.78 0.71 0.82 0.85 0.70Dy 3.89 3.58 4.48 4.13 4.13 4.99 3.82Ho 0.66 0.63 0.78 0.71 0.65 0.87 0.66Er 1.72 1.66 2.02 1.90 1.63 2.29 1.80Tm 0.25 0.26 0.28 0.27 0.22 0.32 0.27Yb 1.47 1.50 1.65 1.60 1.28 1.91 1.66Lu 0.22 0.21 0.26 0.23 0.18 0.26 0.25Hf 2.70 3.22 2.80 2.58 4.01 3.15 3.35Ta 1.68 1.99 1.70 1.56 2.42 1.88 2.37Th 2.30 2.73 2.32 2.14 2.60 2.53 2.52U 0.61 0.74 0.57 0.52 0.72 0.72 0.69

AHHB09 AHWG08 AHWG09 AHWG10 NEB average HNB average

Major elements (wt%)

SiO2 50.53 51.21 51.50 46.34 50.46 47.50TiO2 3.13 3.05 3.18 2.83 1.70 2.49Al2O3 14.97 16.73 17.29 15.46 14.70 15.58Fe2O3 8.60 9.08 8.72 10.51 11.96 11.90MnO 0.11 0.14 0.16 0.23 0.15 0.17MgO 7.90 4.21 4.06 5.98 7.39 8.09CaO 8.17 9.06 8.55 12.01 8.17 8.92Na2O 3.51 4.64 4.65 3.62 3.59 3.53K2O 0.11 0.62 0.80 0.77 0.88 1.64

5050 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Page 3

Table 1 (continued)

AHHB09 AHWG08 AHWG09 AHWG10 NEB average HNB average

P2O5 0.79 0.77 0.77 0.66 0.31 0.58Total 97.82 99.51 99.68 98.42(SiO2)8.0 52.21 51.75 51.99 47.42

Trace elements (ppm)

Sc 17.8 17.0 18.1 15.5 19.6 23.9V 199 204 200 78 169 202Cr 229 62 75 88 221 220Co 28.0 21.0 21.2 25.0Ni 116 70 69 43 151 145Ga 17.4 20.9 20.3 17.3Rb 1.0 2.7 4.0 4.9 11.8 30.3Sr 216 489 488 224 415 568Y 22.4 22.2 22.3 19.4 20.2 29.8Zr 144.5 153.5 161.0 90.6 131 237Nb 41.83 47.21 48.62 35.56 19.10 47.70Ba 131 823 957 514 265 378La 34.71 35.97 35.38 28.56 15.80 32.20Ce 75.75 77.74 77.21 65.87 31.50 64.50Pr 10.27 10.42 10.28 8.86Nd 40.27 41.36 40.80 35.55 18.80 31.70Sm 7.85 8.32 8.23 7.32 4.71 6.88Eu 2.31 2.52 2.54 2.27 1.59 2.21Gd 6.57 6.59 6.57 5.91 4.76 7.50Tb 0.86 0.88 0.88 0.78Dy 4.30 4.37 4.46 3.90Ho 0.70 0.70 0.71 0.61Er 1.74 1.78 1.78 1.52Tm 0.24 0.25 0.24 0.21Yb 1.35 1.44 1.42 1.27 1.53 2.43Lu 0.20 0.21 0.21 0.21Hf 3.25 3.39 3.62 1.87Ta 2.34 2.73 2.80 2.31Th 2.51 3.37 3.48 2.79 1.78 4.14U 1.09 0.88 0.89 0.70

VAL55 8710 JB-1a certified values Average JB-1a value for this analysis r.s.d. Detection limits

Major elements (wt%)

SiO2 48.08 47.10 52.16 52.80 0.86 0.0119TiO2 2.03 2.41 1.30 1.28 2.56 0.0002Al2O3 15.89 16.00 14.51 14.71 1.45 0.0055Fe2O3 10.20 10.90 9.10 8.95 1.65 0.0044MnO 0.16 0.17 0.15 0.15 3.61 0.0194MgO 8.51 7.74 7.75 7.94 1.24 0.0004CaO 9.67 10.30 9.23 9.52 2.07 0.0029Na2O 3.01 3.46 2.74 2.58 6.88 0.0029K2O 1.96 2.08 1.42 1.37 6.96 0.0169P2O5 0.50 0.58 0.26 0.26 2.84 0.0044Total(SiO2)8.0

Trace elements (ppm)

Sc 25.5 21.3V 281 244 206 196 2.25 0.07Cr 339 350 415 423 4.26 0.21Co 39.5 38.1 3.31 0.03Ni 136 115 140 139 7.23 0.34Ga 18.0 17.9 1.89 0.022Rb 22 26 14 14 91.24 0.031Sr 667 768 443 466 9.32 0.29Y 22 22 24.0 24.0 1.58 0.02Zr 162 167 146.0 137.5 4.86 0.05Nb 32.10 48.30 27 28 5.46 0.09

(continued on next page)

Derivation of Jamaican high-Nb basalts 5051

Page 4

Table 1 (continued)

VAL55 8710 JB-1a certified values Average JB-1a value for this analysis r.s.d. Detection limits

Ba 448 430 497 502 2.18 0.41La 25.00 29.40 38.1 37.6 3.65 0.011Ce 49.30 56.60 66.1 66.4 1.88 0.006Pr 7.3 7.2 2.14 0.003Nd 23.10 30.10 25.5 25.8 1.52 0.006Sm 4.60 5.70 5.07 5.12 3.53 0.005Eu 1.58 1.97 1.47 1.49 1.44 0.002Gd 4.60 6.10 4.54 4.60 3.31 0.028Tb 0.69 0.68 3.30 0.009Dy 3.90 4.60 4.19 4.06 1.67 0.003Ho 0.72 0.74 2.65 0.001Er 2.18 2.13 2.59 0.003Tm 0.31 0.31 2.65 0.001Yb 2.00 2.20 2.1 2.07 2.38 0.003Lu 0.30 0.34 0.32 0.31 6.44 0.004Hf 3.70 4.90 3.48 3.36 5.12 0.002Ta 1.6 1.6 2.41 0.001Th 3.00 4.40 8.8 8.8 3.28 0.002U 0.90 1.20 1.6 1.6 9.15 0.004

5052 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Adakites are commonly associated with high-Mg ande-sites, which represent partial melts of mantle wedge mate-rial that has been variably metasomatised by slab melts(e.g. Kay, 1978; Yogodzinski et al., 1994, 1995; Rappet al., 1999; Smithies et al., 2003; Martin et al., 2005). Inaddition, rare high-Nb (>20 ppm) basalts (HNB) are asso-ciated with adakite successions; however, their petrogenesisremains controversial (Reagan and Gill, 1989; Defant et al.,1992; Sajona et al., 1996; Polat and Kerrich, 2001; Castilloet al., 2007; Castillo, 2008). They are not to be confusedwith Nb-enriched basalts (NEB) or Nb-enriched arc basalts(NEAB) that are also associated with adakite sequences(Sajona et al., 1993, 1996; Kepezhinskas et al., 1996;Wyman et al., 2000; Polat and Kerrich, 2001; EscuderViruete et al., 2007; Castillo, 2008).

HNB, NEB and NEAB have not always been clearly de-fined in the literature and somewhat confusingly, the termshave been used interchangeably between different studies(e.g. Reagan and Gill, 1989; Kepezhinskas et al., 1995,1996; Aguillon-Robles et al., 2001; Polat and Kerrich,2001; Escuder Viruete et al., 2007). Nevertheless, Sajonaet al. (1996) and Castillo et al. (2007) distinguish NEBsfrom HNBs in the Sulu arc based on the HNBs being moreenriched in large ion lithophile elements (LILEs), light rareEarth elements (LREEs), Nb and Ta (Table 1).

For the purpose of this study, NEABs are considered tobe analogous to NEBs. LILE- and LREE-enriched NEBshave weakly negative or positive primitive mantle-norma-lised Nb and Ta anomalies and have high-Nb concentrations(�5–20 ppm) (Table 1) (e.g. Sajona et al., 1993, 1996;Wyman et al., 2000; Castillo et al., 2007; Castillo, 2008). Theyare thought by many to be derived from partially meltingmantle source regions that have been variably metasoma-tised by slab-related melts and fluids (e.g. Sajona et al.,1993, 1996; Wyman et al., 2000; Benoit et al., 2002; Smithieset al., 2003; Martin et al., 2005; Pallares et al., 2007).

Conversely, HNBs are more enriched in LILEs, LREEsand HFSEs (Nb > 20 ppm) than NEBs, although they can

still have either slightly negative or positive Nb and Taanomalies on multi-element normalised diagrams (Table 1)(e.g. Reagan and Gill, 1989; Defant et al., 1992; Sajonaet al., 1996; Castillo et al., 2002, 2007; Castillo, 2008).HNBs can have trace element compositions similar toHIMU (high U/Pb) ocean island basalt (OIB) (e.g. Palaczand Saunders, 1986; Weaver, 1991; Chauvel et al., 1992;Castillo, 2008), continental alkaline basalts (Fitton andDunlop, 1985) and alkaline basalts formed above slab win-dows (e.g. Hole et al., 1991; Aguillon-Robles et al., 2001).

Reagan and Gill (1989) proposed that HNBs in CostaRica (2–20 Ma) were derived from complex mixing pro-cesses involving OIB-type small degree partial melts andlarge degree MORB-like melts. In addition, Castillo et al.(2007) and Castillo (2008) also argued that HNBs (andNEBs) in the Sulu arc and Baja California, Mexico are de-rived from the partial melting of upper mantle composed ofenriched OIB-like and depleted N-MORB-type mantlecomponents.

In contrast, Kepezhinskas et al. (1995, 1996) describeNEBs (their NEABs) from the Kamchatka arc that wouldbe classified as HNBs with our geochemical definition[and other definitions (Table 1) (e.g. Sajona et al., 1996;Castillo et al., 2007; Castillo, 2008)]. Kepezhinskas et al.(1995, 1996) clearly demonstrate, based on mineralogyand geochemistry, that the Kamchatka HNBs cannot bederived from an OIB-type source region or formed from amantle-derived magma that has been contaminated by con-tinental material. By studying related mantle xenoliths theyshowed that the HNBs are formed by partial melting ofdepleted mid-ocean ridge (MOR)-type mantle (DMM) thathas been metasomatised by ascending slab melts. Similarly,Defant et al. (1992) also suggested that HNBs in the CentralAmerican arc could be derived from adakitic metasomatisedperidotite residue in the mantle wedge. It is clear, therefore,that the generation of NEBs and HNBs is controversial.

In this paper we report the geochemistry of someTertiary alkaline basaltic lavas from eastern Jamaica, West

Page 5

Derivation of Jamaican high-Nb basalts 5053

Indies, known as the Halberstadt Volcanic Formation.These lavas have similar trace element characteristics toHIMU-type OIB, continental alkaline basalts and HNBsamples. We will use this data to shed light on the petrogen-esis of HNBs, i.e. did they form by slab-melt interactions ina mantle wedge (e.g. Kepezhinskas et al., 1995, 1996;Sajona et al., 1996) or by mixing and/or melting of enrichedand depleted upper mantle (e.g. Reagan and Gill, 1989;Castillo et al., 2007; Castillo, 2008)? We will also assessthe compositional and tectonic variability between HNBand similar rock types to determine if HNBs merit beingclassified as a distinct rock type.

Fig. 1. (a) Map of the Caribbean region, (b) location of tectonic blockgeological map of the Wagwater Belt (modified from Jackson and SmHalberstadt and Newcastle Volcanic Formations in eastern Jamaica.

2. GEOLOGICAL BACKGROUND

2.1. The Caribbean plate

The Caribbean plate has northern and southern bound-aries marked by dextral and sinistral strike slip motions,respectively (Fig. 1a). The eastern and western margins ofthe plate are active subduction zones with the Cocos Platesubducting under the Central American arc to the westand the Atlantic oceanic crust subducting beneath the Les-ser Antilles to the east (Fig. 1a) (e.g. Pindell and Barrett,1990).

s and volcano-sedimentary Cretaceous inliers in Jamaica and (c)ith, 1978; Hastie et al., 2010b), which shows the location of the

Page 6

5054 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Most of the Caribbean plate consists of a 8–20 km thickLate Cretaceous oceanic plateau that formed in the Pacific(e.g. Edgar et al., 1971; Mauffret and Leroy, 1997; Kerret al., 2003), possibly derived from the initial plume headphase of the Galapagos hotspot (e.g. Geldmacher et al.,2003; Thompson et al., 2003; Hastie and Kerr, 2010). Theoceanic plateau erupted onto the Farallon Plate at �89–94 Ma and was transported northeastwards to collide withthe northwestern margin of South America and an intra-American oceanic island arc (the Great Arc of theCaribbean) by �80 Ma (Burke, 1988; Kerr et al., 2003).The plateau crust was too thick, hot and buoyant tosubduct, thus, the southern portion of the oceanic plateaucollided and accreted to Colombia and Ecuador in the LateCretaceous (e.g. Kerr et al., 2002a,b). When the northernportion of the oceanic plateau subsequently collided withthe Great Arc of the Caribbean, it initiated subductionpolarity reversal whereby the proto-Caribbean crust begansubducting in a southwesterly direction beneath the oceanicplateau (e.g. Duncan and Hargraves, 1984; Burke, 1988;Kerr et al., 1996, 1999; Thompson et al., 2003). Throughoutthe Cenozoic, the plateau and the Great Arc were tectoni-cally emplaced between the westward moving North andSouth American continents to form the Caribbean plate(Duncan and Hargraves, 1984; Burke, 1988; Kerr et al.,2003; Hastie and Kerr, 2010).

2.2. Jamaican geology – the Tertiary Wagwater basin

Jamaica is situated on the northwestern Caribbean platemargin (Fig. 1a) (Robinson et al., 1972; Jackson and Smith,1978) and can be divided into three structural blocks. To

Fig. 2. Location map for the Halbersta

the west lies the Hanover Block that is separated from thecentral Clarendon Block by a graben structure known asthe Montpelier–Newmarket Belt (Robinson et al., 1972;Draper, 1986). In the east the Clarendon Block is separatedfrom the Blue Mountain Block by the Wagwater Belt(Fig. 1b) (Robinson et al., 1972; Jackson and Smith, 1978).

The Wagwater Belt contains a thick (at least 7 km) suc-cession of Palaeocene to Early Eocene (�65–45 Ma) clasticsedimentary and volcanic rocks that accumulated in theWagwater rift basin (Mann and Burke, 1990). The lower5.6 km consists of coarse-grained terrigenous, dominantlyterrestrial strata (Wagwater Formation), overlain by some1.2 km of upward-fining marine rocks (Richmond Forma-tion). During the last 12 Ma, the basin has been invertedto form the Wagwater Belt, which has steep dips towardsthe east. The Wagwater Formation consists of conglomer-ates with subordinate sandstones, limestones and evapor-ites. The Halberstadt Volcanics are exposed in thesoutheastern part of the belt and are represented by a suc-cession of pillow basalts which, based on vesicle sizes, wereextruded in water depths estimated at about 300 m (Fig. 1c)(Jackson, 1987). The Newcastle Volcanics are intercalatedthrough the middle and upper part of the Wagwater For-mation and may have been extruded from three differentvolcanic centres (Jackson and Smith, 1978, 1979; Jackson,1987).

Hastie et al. (2010a,b) studied the petrogenesis of theNewcastle Volcanic Formation and demonstrated that thelavas are adakitic and are derived from the partial meltingof mafic, underthrust Caribbean oceanic plateau material at52.74 ± 0.34 Ma as the Caribbean plate was being em-placed between the American continents.

dt samples collected in this study.

Page 7

Tab

le2

Sr–

Nd

–Hf

iso

top

ed

ata

for

the

Hal

ber

stad

tla

vas

(HL

)an

dn

ewS

rd

ata

for

the

New

cast

lela

vas

(NL

)(m

,m

easu

red

valu

es;

i,in

itia

lva

lues

).T

he

cho

nd

rite

valu

esu

sed

toca

lcu

late

the

epsi

lon

Hf

and

Nd

valu

esar

e0.

2827

3an

d0.

5125

6,re

spec

tive

ly.

Nd

and

Hf

iso

top

ed

ata

for

the

New

cast

lela

vas

are

also

pre

sen

ted

inH

asti

eet

al.

(201

0b).

Init

ial

and

epsi

lon

valu

esca

lcu

late

du

sin

gth

eav

erag

e40A

r/39A

rp

late

auag

eo

f52

.74

Ma

fro

mth

eN

ewca

stle

Vo

lcan

icF

orm

atio

n(H

asti

eet

al.,

2010

b).

(87S

r/86S

r)m

(87S

r/86S

r)i

176L

u/1

77H

f(1

76H

f/177H

f)m

(176H

f/177H

)ie H

f(i)

147S

m/1

44N

d(1

43N

d/1

44N

d)m

(143N

d/1

44N

d)i

e Nd(i

)

AH

HB

03H

L0.

7048

50.

7048

40.

0134

0.28

307

0.28

306

11.3

70.

168

0.51

298

0.51

292

6.80

AH

HB

05H

L0.

7045

30.

7045

20.

0064

0.28

306

0.28

306

11.2

90.

127

0.51

293

0.51

288

6.13

AH

HB

07H

L0.

7046

00.

7045

90.

0116

0.28

308

0.28

307

11.5

60.

152

0.51

297

0.51

292

6.77

AH

WG

08H

L0.

7044

60.

7044

50.

0088

0.28

306

0.28

305

11.1

70.

122

0.51

294

0.51

290

6.40

AH

WG

12N

L0.

7043

10.

7041

80.

0056

0.28

309

0.28

308

12.2

50.

124

0.51

294

0.51

289

6.32

AH

WG

19N

L0.

7048

10.

7046

80.

0064

0.28

309

0.28

308

12.0

50.

113

0.51

292

0.51

288

6.06

AH

WG

21N

L0.

7048

50.

7047

90.

0058

0.28

309

0.28

309

12.3

60.

125

0.51

293

0.51

289

6.18

AH

WG

32N

L0.

7048

20.

7047

60.

0046

0.28

310

0.28

309

12.5

00.

104

0.51

297

0.51

293

7.03

AH

WG

33N

L0.

7045

20.

7044

80.

0047

0.28

309

0.28

309

12.2

80.

108

0.51

296

0.51

292

6.80

Derivation of Jamaican high-Nb basalts 5055

3. SAMPLE LOCATIONS AND PETROGRAPHY

Samples AHHB01–09 were collected along the road sec-tion from Newstead to Bito (Fig. 2) and are greenish-grey,porphyritic amygdoidal/vesicular pillow lavas. AHWG08–10 were collected in a network of quarries located to thesouth of Bito. These latter lavas can be relatively freshand form massive, porphyritic amygdoidal/vesicular basal-tic flows which have a dark grey colour. They largely lackthe pillow structures seen in AHHB01–09.

In thin section, all the lavas are composed of plagioclasefeldspar phenocrysts set in a groundmass of plagioclase andopaque minerals. The feldspar phenocrysts are partially re-placed by clay minerals, chlorite and sericite. The amyg-dales range in size from <1 to �5 mm and are composedof quartz, calcite and chlorite.

4. GEOCHEMICAL RESULTS

4.1. Analytical techniques

Major and trace element data were analysed using a JYHoriba Ultima 2 inductively coupled plasma optical emis-sion spectrometer (ICP-OES) and a Thermo X7 seriesinductively coupled plasma mass spectrometer (ICP-MS)at Cardiff University, UK (Table 1). A full description ofthe analytical procedures and equipment at Cardiff Univer-sity can be found in McDonald and Viljoen (2006). Multi-ple analyses of international reference materials JB-1a,BIR-1, W2, JA-2, MRG-1 and JG-3 and four in-housestandards ensured the accuracy and precision of the analy-ses. Most elements did not deviate more than 5% from stan-dard values and have first relative standard deviationsbelow 5% (see Table 1 for JB-1a data).

Sr, Nd and Hf isotope compositions were analysed atthe NERC Isotope Geoscience Laboratories, Nottingham,UK. Determinations of Sr and Nd isotopes followed theprocedures of Kempton (1995) and Royse et al. (1998). Srwas loaded on single Re filaments using a TaO activator,and run using static multicollection on a Finnigan Tritonmass spectrometer. Nd was run as the metal species usingdouble Re–Ta filaments on a Finnigan Triton mass spec-trometer (Table 2). For Hf isotope analysis, the sampleswere prepared following the procedures in Nowell andParrish (2001) and Munker et al. (2001), and run on aNu-Plasma multicollector ICP-MS.

4.2. Element mobility

Previous geochemical studies on altered Cretaceousigneous rocks in the Caribbean have demonstrated thatthe majority of the LILE can be variably mobilised by arange of weathering, hydrothermal and metamorphic pro-cesses (Revillon et al., 2002; Thompson et al., 2003; EscuderViruete et al., 2007; Hastie et al., 2007; Hastie, 2009; Hastieet al., 2009).

Conversely, Th, the HFSEs, the REEs and the transitionelements (e.g. Co, V, Cr, Ni and Sc) are considered to berelatively immobile during sub-solidus alteration and low-temperature metamorphism (up to greenschist-grade) (e.g.

Page 8

Fig. 3. Variation diagrams for a range of incompatible elements plotted against Nb. If the elements are (1) incompatible during partialmelting and fractional crystallisation and are (2) immobile, the data should form linear near-diagonal (1:1) vectors on log–log plots.

5056 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Winchester and Floyd, 1976; Pearce, 1982, 1996). If animmobile element is plotted on the horizontal axis of a var-iation diagram other elements can be plotted on the verticalaxes. If the two elements are moderately-highly incompati-ble and immobile, and the samples are co-genetic, the datashould give trends with slopes close to unity (Cann, 1970).

Representative variation diagrams for the Halberstadtlavas are shown in Fig. 3 whereby an incompatible elementis plotted against immobile Nb on the abscissa (e.g. Cann,1970; Hill et al., 2000; Kurtz et al., 2000). The bivariate plot

for Ba (Fig. 3a) shows a large degree of scatter with little orno evidence of the expected, pre-alteration linear trendwithin the Halberstadt lavas. Other LILEs, such as Naand K, similarly do not show a linear trend which is inter-preted to be the result of post-eruptive alteration of therocks by hydrothermal and weathering processes.

In contrast, Sr, Th, Zr, La and Sm show much smallerdegrees of scatter and form linear trends with moderatelyhigh correlation coefficients, indicating that these elementsare relatively immobile (Fig. 3b–f). These diagrams also

Page 9

Derivation of Jamaican high-Nb basalts 5057

indicate that the Halberstadt lavas can be separated into twogroups: Group 1 has higher concentrations of Sr, Th, Zr, La,Sm and Nb relative to Group 2 (Fig. 3b–f). These twogroups are also clearly identified in the Er and Yb diagrams;however, Group 1 has lower abundances of the HREEscompared to Group 2. Nonetheless, the HREE should beas immobile as the LREE and HFSE, therefore the low con-centrations in the Group 1 samples are probably related to amagmatic process and are not due to sub-solidus alteration.

4.3. Major elements and classification

The lavas have high loss on ignition values (LOI), lar-gely because of the calcite in the amygdales, and so the ma-

Fig. 4. (a) Zr/TiO2–Nb/Y and (b) Th–Co discrimination diagrams (Pcalc-alkaline; H-K and SHO, high-K calc-alkaline and shoshonite; B, basa� indicates that latites and trachytes also fall in the D/R fields.

jor elements are re-calculated on an anhydrous basis. TheSiO2 and MgO concentrations of the Halberstadt lavasrange between 46.3–58.0 and 4.1–11.2 wt%, respectively(Table 1). Although the lavas are obviously basaltic in nat-ure, the mobility of elements of low ionic potential limitsthe use of standard classification diagrams, such as the totalalkali silica and the K2O–SiO2 diagrams (Peccerillo andTaylor, 1976; Le Bas et al., 1986, 1992). Alternatively,the lavas can be classified on the updated (Pearce, 1996)immobile trace element Zr/TiO2–Nb/Y diagram ofWinchester and Floyd (1976) and the Th–Co discriminationdiagram of Hastie et al. (2007). These diagrams confirmthat, with the exception of AHHB03, the Halberstadt lavasare alkaline basalts and basaltic andesites (Fig. 4). Relative

earce, 1996; Hastie et al., 2007). IAT, Island arc tholeiite; CA,lt; BA/A, basaltic–andesite and andesite; D/R*, dacite and rhyolite.

Page 10

5058 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

to N-MORB, the lavas also have high TiO2 contents rang-ing from 2.2 to 3.2 wt%.

4.4. Trace elements

Primitive mantle-normalised multi-element and chon-drite-normalised REE diagrams are shown in Fig. 5. TheHalberstadt lavas have positive Nb and Ta anomalies, havemore enriched incompatible element concentrations thanN-MORB and NEBs and have fractionated REE patternssimilar to other HNBs (Fig. 5). Group 1 lavas have higherLREE and HFSE and lower Y and HREE normalised con-centrations than the Group 2 samples. The HNBs also haveslight negative Ce, Zr and Hf anomalies and a positive Tianomaly. Interestingly, positive Ti anomalies are also pres-ent in the NEBs of the Sulu arc (Sajona et al., 1993). Theprimitive mantle-normalised multi-element patterns of theHalberstadt lavas are similar to HIMU-type OIB, continen-tal alkali basalts and basalts formed in slab window envi-ronments (e.g. Fitton and Dunlop, 1985; Hole et al.,1991; Weaver, 1991).

Fig. 5. (a) Primitive mantle-normalised multi-element plot and (b) chonvalues from McDonough and Sun (1995). Dark grey fields are HNBs frofields are NEBs from Castillo et al. (2007).

4.5. Sr–Nd–Hf radiogenic isotopes

An adakite from the Newcastle Volcanic Formationyielded an average 40Ar/39Ar plateau age of 52.74 ±0.34 Ma (Hastie et al., 2010b). Therefore, radiogenic iso-tope data for Halberstadt and Newcastle lavas have beenage-corrected to 52.74 Ma. The Halberstadt lavas have(87Sr/86Sr)i = 0.7044–0.7048, eNd(i) = +6.13 to +6.80 andeHf(i) = +11.17 to +11.56 (Fig. 6 and Table 2). The New-castle lavas have (87Sr/86Sr)i = 0.7042–0.7048. The Ndand Hf isotope systems are resistant to alteration processes(e.g. White and Patchett, 1984); hence, their ratios shouldrepresent the primary composition of the Halberstadtand Newcastle lavas. Fig. 3 shows evidence that theHalberstadt LILEs have been mobilised by sub-solidusalteration processes. However, the Sr variation diagramshows that Sr, when plotted against Nb, forms a pre-alter-ation magmatic linear trend with low/moderate degrees ofscatter. In addition, the samples were leached in 6 M HClprior to analysis in order to minimise the effects of anyalteration.

drite-normalised REE plot of the Halberstadt lavas. Normalisingm Kepezhinskas et al. (1996) and Castillo et al. (2007). Light grey

Page 11

Fig. 6. (a) (87Sr/86Sr)i–eNd(i) and (b) eNd(i)–eHf(i) diagrams for theHalberstadt and Newcastle lavas (modified from Hastie et al.(2008, 2009), respectively).

Fig. 7. (a) (SiO2)8.0–Nb and (b) (SiO2)8.0–Zr diagrams showingevidence of a partially melting trend in the Halberstadt data.Symbols as in Fig. 3.

Derivation of Jamaican high-Nb basalts 5059

Consequently, the evidence suggests that the Halbers-tadt initial 87Sr/86Sr ratios represent primary magmacompositions. They also correspond to the primary compo-sition of the Newcastle lavas because the Sr in the adakiteshas also not been mobilised by post-eruptive alteration pro-cesses (Hastie et al., 2010b). Fig. 6 shows that the Halbers-tadt and Newcastle lavas have near identical Sr and Ndisotope signatures and similar eNd(i) and eHf(i) values.

5. DISCUSSION

5.1. Halberstadt compositional variability: source

heterogeneity and/or variable partial melting?

Halberstadt Group 1 lavas are more enriched in themost incompatible elements (e.g. Th, Nb, La, Sm) andslightly more depleted in the HREEs relative to Group 2 la-vas (Table 1 and Figs. 3–5). This chemical difference in theHalberstadt rocks may be due to (1) both groups being de-rived from compositionally different source regions and/or

(2) variably small degrees of partial melting in generatingthe different Groups. The samples are basalts and basalticandesites with mostly >5 wt% MgO; thus, large degrees offractional crystallisation cannot generate the compositionaldifferences in both Groups. Table 2 shows that the(87Sr/86Sr)i, eNd(i) and eHf(i) radiogenic isotopic composi-tion of the Group 1 and Group 2 basalts are slightly differ-ent, with Group 1 having more enriched compositions.Age-corrected radiogenic isotope ratios are not modifiedby partial melting processes, therefore, the different valuessuggest that Group 1 and Group 2 basalts are likely to bederived from compositionally different source regions.

Partial melting of peridotite can result in variable majorelement melt compositions due to differing source composi-tions, melt fractions and pressures (e.g. Takahashi andKushiro, 1983; Klein and Langmuir, 1987, 1989; Hiroseand Kushiro, 1993; Walter, 1998). Consequently, if bothHalberstadt groups are assumed to be formed at a similarpressure, the SiO2 content can be re-calculated to 8 wt%MgO to take into account low degree fractional crystallisa-tion (Klein and Langmuir, 1987). When (SiO2)8.0 is plottedagainst Nb and Zr (Fig. 7) the data have a negative slopewith Group 1 lavas having higher Nb and Zr contentsbut lower (SiO2)8.0 values than Group 2. Klein andLangmuir (1987, 1989) and Walter (1998) show that, with

Page 12

5060 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

constant pressure, at low melt fractions the concentrationof SiO2 is also low and increases with higher degrees of par-tial melting (up to �40–50%). Therefore, Fig. 7a and b sug-gests that Group 1 major and trace element compositionscan similarly be derived by smaller degrees of partial melt-ing relative to Group 2 lavas.

The low HREE and Y contents and wide range ofLREE/HREE ratios (e.g. La/Yb = 7.8–25.8) with relativelyconstant Sc/Yb ratios (10.8–15.6), suggests that residualgarnet may have been in the source region for all the Hal-berstadt lavas. Residual garnet in the Halberstadt sourceregion also explains the scattered negative trends in theEr and Yb variation diagrams in Fig. 3.

Therefore, Group 1 and 2 lavas are derived from isoto-pically distinct peridotite mantle source regions, whichboth contain garnet as a residual phase. Group 1 lavaswere derived from a more enriched source region thatmay have imparted higher concentrations of incompatibleelements to Group 1 magmas relative to Group 2. On theother hand, the major and trace element differences be-tween the two groups can also be partly explained by gen-erating the Group 1 magmas by smaller degrees of partialmelting.

5.2. Identifying the Halberstadt source region

5.2.1. A lower ultramafic crustal source?

The lowermost crust of an arc is characterised by ultra-mafic complexes (Alaskan-type complexes) (e.g. Debariet al., 1987; Andrew et al., 1995). Therefore, could theseultramafic complexes undergo partial melting to generatealkaline basalts with compositions similar to the JamaicanHNB? Medard et al. (2006) have experimentally determinedthe composition of magmas generated from melting theselower crustal ultramafic rocks at 0.5–1.0 GPa.

The conditions in the lower arc crust (�1.0 GPa) enablearc magmas to undergo fractional crystallisation and gener-ate cumulates of olivine, clinopyroxene, amphibole, magne-tite and possibly plagioclase (Medard et al., 2006). As aconsequence, Medard et al. (2006) used a starting mineral-ogy composed of olivine, clinopyroxene and amphibole.The melts generated from these experiments, at comparablesilica contents to the HNBs, are compositionally distinctfrom the Halberstadt lavas. Additionally, such a source re-gion would be too shallow for garnet to form (e.g. Kinzler,1997) and, as has been noted above, garnet is required inthe source in order to generate the fractionated REE pat-terns in the Jamaican HNBs. Therefore, it is unlikely thatthe HNBs are derived from partially melting a lower crustalultramafic complex.

5.2.2. Assimilation of upper continental crust?

Contamination of mantle-derived magmas with conti-nental crust or island arc crust can be ruled out as thiswould strongly increase the LILE/HFSE and LREE/HFSEratios of the HNB magmas and would impart negative Nb–Ta anomalies on normalised multi-element diagrams(Fig. 5) (e.g. Kerr et al., 2000). Furthermore there is noevidence of continental crust beneath Jamaica, which islargely made up of Cretaceous–Tertiary carbonate and

island arc igneous rocks overlying an assumed late Jurassic–Cretaceous altered oceanic basement (Robinson et al.,1972; Draper, 1986). This also rules out a deep continentallithosphere source for the Halberstadt magmas.

5.2.3. Depleted MORB mantle (DMM) source region?

MORBs are compositionally more depleted than theHalberstadt lavas and are considered to be derived by�5–20% fractional partial melting of ascending astheno-sphere below a mid-ocean ridge (Kinzler and Grove,1992a,b; Kinzler, 1997; Schwab and Johnston, 2001). How-ever, MORBs can be compositionally variable (e.g.Workman and Hart, 2005); therefore, can the Halberstadtbasalts be generated from a DMM source? Experimentalpartial melts (±fractional crystallisation), with comparablesilica abundances to the HNBs, generated from intermedi-ate/depleted plagioclase and spinel peridotites at 0.4–2.5 GPa by Kinzler and Grove (1992b, 1993) and Schwaband Johnston (2001) fail to produce liquids with similar ma-jor element compositions to the Halberstadt alkalinebasalts.

Moreover, the more enriched Sr, Nd and Hf radiogenicisotope systematics in the Jamaican HNBs shows that theyare not derived from low degree partial melting of a DMMsource region (Fig. 6) (e.g. Nowell et al., 1998; Workmanand Hart, 2005). Similarly, the enriched-DMM (E-DMM)source of Workman and Hart (2005) cannot be a viablesource region because it also has differing radiogenic iso-tope ratios relative to the Jamaican HNBs.

5.2.4. Melting a subducting slab?

Partial melting of a subducting slab is generally regardedto be the main way of generating siliceous melts with adak-itic compositions (e.g. Defant et al., 1992; Drummondet al., 1996). Nevertheless, basaltic compositions have alsobeen generated from mafic protoliths during partial meltingexperiments (e.g. Rapp et al., 1991). A palaeo-subductionzone in California (the Catalina Schist) was described bySorensen (1988) and Sorensen and Grossman (1989).High-Nb mafic migmatites associated with the schist areconsidered to be derived by partial melting of an amphibo-lite or eclogite residue (a subducting slab melt?). However,the major and trace element geochemistry of the experimen-tally derived basalts of Rapp et al. (1991) and theCalifornian migmatites are not analogous to HNBs, whichhave predominantly higher Na2O and Ni abundances andlower Al2O3, Sc, Zr, Y, LREE, MREE and HREE concen-trations relative to the high-Nb mafic migmatites [compareHNBs in Table 1 of Kepezhinskas et al. (1996) and migma-tites in Table 1 of Sorensen and Grossman (1989)]. Conse-quently, it is highly unlikely that the Halberstadt HNBmagmas are generated via a slab melting process.

5.2.5. An OIB source region?

Many geochemists consider that the majority of OIBmagmas are derived from partial melting fertile lower man-tle peridotite that has risen through the upper mantle inhot, buoyant mantle plumes (e.g. Woodhead, 1996;Hofmann, 1997; Farnetani and Samuel, 2005; Campbell,2007). However, other studies consider OIB magmas to

Page 13

Derivation of Jamaican high-Nb basalts 5061

be derived from lithospheric mantle source regions thathave been contaminated by carbonatite/plume derivedmelts (e.g. Nakamura and Tatsumoto, 1988; Dupuy et al.,1992; Chauvel et al., 1997). Although a discussion of thepotential source regions of OIB is beyond the scope of thisstudy we will briefly address carbonatite metasomatism as itmay represent a mechanism to form the Jamaican HNBsource region.

Carbonatite contamination of a peridotite source regionwould produce high and fractionated Zr/Hf ratios of up to100 in any resultant melt (Woodhead, 1996; Chauvel et al.,1997), which is in contrast to the average Zr/Hf ratios forMORB, continental flood basalts, arc rocks and someOIB lavas (�33–49) (Dupuy et al., 1992; Chazot et al.,1996; Woodhead, 1996). The Jamaican HNBs have Zr/Hfratios from 36.2 to 48.5, which are lower than that expectedif they were derived from a mantle source metasomatised bycarbonatite melts.

In addition, carbonatite melts and the sources theymetasomatise, have very high concentrations of the LREEs(�100) and negative Ti anomalies on chondrite/primitivemantle-normalised multi-element diagrams (Nakamuraand Tatsumoto, 1988; Hauri et al., 1993; Woodhead,1996; Chauvel et al., 1997). The lower LREE concentra-tions and the lack of negative Ti anomalies in the Halbers-tadt lavas in Fig. 6 also argues against them being derivedfrom a carbonatite metasomatised source region; the Jamai-can HNBs actually have positive Ti anomalies.

Regardless of the source of OIBs, anhydrous partialmelt experiments on fertile spinel peridotite KR4003 (anOIB source?) from 3 to 7 GPa by Walter (1998) formed par-tial melts that are unlike any Jamaican HNB at similar sil-ica values. This result is comparable to other anhydrousfertile spinel peridotite partial melt experiments performedat slightly shallower depths (0.5–3.5 GPa) that can alsonot generate liquids with major element compositions sim-ilar to the Jamaican HNBs at similar concentrations of SiO2

(e.g. Takahashi and Kushiro, 1983; Hirose and Kushiro,1993).

Dasgupta et al. (2007) performed partial melting exper-iments on CO2-rich fertile peridotite [CO2-rich KLB-1 ofTakahashi (1986)] at 3 GPa to generate melts with majorelement compositions similar to some OIB-like alkalinebasalts. Additionally, Kawamoto and Holloway (1997) car-ried out H2O-saturated partial melt experiments on KLB-1up to 11 GPa to generate melts with compositions similar tokimberlites. However, like the anhydrous experiments, themelts generated in these deep H2O- and CO2-saturated peri-dotite experiments have compositions distinct from theHalberstadt lavas.

The HNBs cannot be derived from a EMI- or EMII-typeOIB end-member mantle source region because they havedifferent radiogenic isotope ratios (Fig. 6). In addition, theirincompatible trace element abundances and ratios are dis-similar (e.g. Zindler and Hart, 1986; Salters and Hart,1991; Weaver, 1991; Hart, 1994; Woodhead, 1996; Saltersand White, 1998). For example, the average Th/Nb ratioof the Jamaican HNBs is 0.072 whereas the ratios forEMI- and EMII-type OIB are 0.10–0.12 and 0.11–0.16,respectively (e.g. Weaver, 1991; Woodhead, 1996). Also, be-

cause the Jamaican HNBs have been generated by small de-grees of partial melting, their average Th/Nb ratiorepresents a maximum value. At higher degrees of partialmelting the Th/Nb ratio would further decrease. Therefore,incompatible trace element ratios and radiogenic isotopevalues rule out EMI and EMII end-members as potentialsources.

The Jamaican HNBs and HIMU-OIB also have differ-ent radiogenic isotope compositions (Fig. 6). However,incompatible trace element ratios of the Halberstadt basaltsand HIMU-OIB can be similar. Weaver (1991) and Wood-head (1996) show that the Th/Nb and the La/Nb ratios ofHIMU-OIB are 0.07–0.10 and 0.64–0.77, respectively.These values are analogous to the average Jamaican HNBratios of 0.072 and 0.62, respectively. Nevertheless,although the incompatible trace elements between HIMU-OIB and the HNBs can be alike, the more enriched Hfradiogenic isotopes for the HIMU-OIB indicates that theHNBs are not derived from a HIMU source region.

However, could the Jamaican HNBs represent a mixturebetween the different OIB end-member components? Addi-tionally, Castillo et al. (2007) showed that the elemental andradiogenic isotope compositions of NEBs and HNBs in theSulu island arc overlap, and thus, both rock types are de-rived from chemically similar mantle source regions, com-posed of an N-MORB-like component and an enrichedmantle component with OIB-like trace element concentra-tions. Castillo (2008) also provided geochemical evidenceto suggest that HNBs and NEBs in Baja California, Mexicoare derived from partially melting Pacific asthenophericmantle that is composed of enriched and depleted compo-nents. Nonetheless, when considering a “true” OIB compo-nent or the depleted and enriched mantle model of Castilloet al. (2007) and Castillo (2008) for the formation of theJamaican HNBs the regional evolution of the Caribbeanplate and the composition of previous Jamaican volcanicrocks must be considered.

There is little evidence for a hotspot in the Caribbean re-gion from 120 Ma to the present-day (e.g. Kerr et al., 2003;Pindell and Kennan, 2009); thus, there should be no “true”

OIB-like enriched mantle beneath the island arc crust of Ja-maica at �53 Ma (age of Halberstadt lavas). The Carib-bean oceanic plateau is tectonically close to, or isunderthrusting Jamaica at this time (Hastie et al., 2010a),but previous geochemical studies on primary magmas ofthe Caribbean oceanic plateau show that the lavas areformed by �20–30% partial melting of a depleted and/orenriched spinel/garnet peridotite source (e.g. Kerr et al.,1996, 2002a,b; Hauff et al., 1997; Revillon et al., 2000;Hastie and Kerr, 2010). As a result, the mantle plumeresidue is highly depleted and refractory (e.g. Fitton andGodard, 2004) and cannot be the source of any enrichedOIB-type mantle. Finally, island arc lavas studied onJamaica from the Hauterivian/lower Barremian (ca.136.4–122 Ma) to the Tertiary are derived from mantlecomponents compositionally similar to N-MORB andshow no evidence for an enriched OIB-like upper mantlecomponent, similar to that described by Castillo et al.(2007) in the Sulu arc and Castillo (2008) in Mexico (Hastieet al., 2007, 2009; Hastie, 2009).

Page 14

Fig. 8. Th/Yb–Ta/Yb discrimination diagram of Pearce (1982).Halberstadt lavas and HNBs from Castillo et al. (2007) are seenplotting just above the MORB array, similar to back-arc lavasfrom Jamaica.

5062 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Taking all these things into consideration it is thereforeunlikely that the Halberstadt lavas are derived from a“true” OIB-type source region or a source region composedof a mixture of a depleted N-MORB and an enriched man-tle component.

5.2.6. Slab-melt metasomatised mantle wedge

It has been noted in several studies (e.g. Defant et al.,1992; Sajona et al., 1993, 1996) that NEBs and HNBs areassociated with adakites and the subduction of young oce-anic crust. Therefore, it seems likely that HNBs may begenetically linked to adakite formation.

Sajona et al. (1996) calculate the composition of the pri-mary magmas of the Sulu arc NEBs, which have composi-tions similar to HNBs. Their results demonstrate that themantle source region(s) of the Sulu arc NEBs probably con-tained garnet, amphibole, clinopyroxene and Fe–Ti oxides,but no rutile. The results are similar to experimental workon the reaction of slab-derived acidic (TTG/D and adakitic)melts with depleted and fertile spinel peridotite (e.g. Carrolland Wyllie, 1989; Johnston and Wyllie, 1989; Sen andDunn, 1994a; Rapp et al., 1999; Prouteau et al., 2001).

A slab melt migrating through a mantle wedge will frac-tionate, hybridise and metasomatise the peridotite wherebythe original mineralogy of the peridotite (olivine, orthopy-roxene, clinopyroxene and spinel) is broken down and re-placed by precipitating Nb and Ti-enriched pargasiticamphibole, garnet, phlogopite, Na-rich clinopyroxene andFe-enriched orthopyroxene (also Na-plagioclase at lowpressures) (e.g. Carroll and Wyllie, 1989; Johnston andWyllie, 1989; Adam et al., 1993; Sen and Dunn, 1994a;Kepezhinskas et al., 1995; Rapp et al., 1999; Prouteauet al., 2001). If the melt:rock ratio is high enough the resul-tant hybridised melt, which will have lower SiO2 and higherMgO, Ni and Cr contents after it has reacted with the peri-dotite, can ascend and erupt as an adakite (e.g. Carroll andWyllie, 1989; Johnston and Wyllie, 1989; Rapp et al., 1999).If the melt:rock ratio is lower, the slab melt(s) are consumedand the metasomatised peridotite can subsequently par-tially melt to produce adakitic and/or high-Mg andesitemagmas depending on the exact melt:rock ratio (Sajonaet al., 1993, 1996; Rapp et al., 1999).

The model of Sajona et al. (1993, 1996), (which is alsosimilar to one proposed by Kepezhinskas et al., 1995,1996), proposes that the first slab-derived melts are pro-duced at a depth of 75–85 km. On ascent, they react withthe mantle wedge to produce the aforementioned mineralassemblage in a metasomatised spinel peridotite. The meta-somatic minerals will subsequently scavenge trace elements[amphibole would take Ti, Nb and Ta (e.g. Sorensen andGrossman, 1989)] from continually ascending slab-meltsuntil the unmelted metasomatised peridotite mantle sourceregion is subsequently convected to a sufficient depthwhereby amphibole breaks down thus promoting partialmelting of the wedge and generation of NEBs.

We propose a similar model for generating the JamaicanHNBs. The negative Ce anomalies on the N-MORB-normalised multi-element plot in Fig. 5 suggest that theHNBs may be contaminated with a slab-related pelagic sed-imentary component. Comparable negative Ce anomalies

are present in island arc lavas (e.g. Hole et al., 1984;McCulloch and Gamble, 1991; Elliott, 2003) and have beenquantified by using an N-MORB normalised Ce/Ce* ratioe.g. Ce/(((La � Pr)/2) + Pr). The Jamaican HNBs have(Ce/Ce*)n–mn ratios from 0.85 to 0.97, which suggests thatat least some of the lavas display a slab-related componentin their genesis.

Also, the Jamaican HNBs, and HNB samples from Cas-tillo et al. (2007), plot just above the MORB-OIB array inthe Th/Yb–Ta/Yb plot of Pearce (1982) (Fig. 8). This indi-cates that the HNB rocks have higher Th/Ta ratios than“normal” mantle derived oceanic basalts. The relativeenrichment of Th in the HNBs may be derived from (1) in-put of a subduction slab-fluid/melt into the source region or(2) crustal contamination of the ascending HNB magmas(e.g. Pearce and Peate, 1995). Nevertheless, the higher Th/Nb ratios for the Halberstadt lavas, together with the lowCe/Ce* values and similar Sr–Nd–Hf radiogenic isotope ra-tios to the Newcastle adakites, suggests that the Halbers-tadt source region was contaminated with a subductioncomponent compositionally similar to the Newcastleadakites.

Hastie et al. (2010a) have shown that the majority ofNewcastle adakites did not hybridise significantly withany peridotite component. This information, along withother geochemical signatures in the adakites, was used topropose that the adakites were generated by partial meltingof an underthrusting (or subducting) portion of the Carib-bean oceanic plateau beneath Jamaica at �53 Ma. Thethickness and buoyancy of the oceanic plateau meant thatit underthrust at an extremely shallow angle thus prevent-ing the formation of an extensive asthenospheric mantlewedge above it. This allowed the Jamaican adakites to as-cend from their source region without hybridising with peri-dotite material.

Hastie et al. (2010a) suggested that the Newcastle andHalberstadt lavas are isotopically distinct; however, thenew data presented in this paper show that they havecomparable (87Sr/86Sr)i and eNd(i) values and very similar

Page 15

Derivation of Jamaican high-Nb basalts 5063

eHf(i) values. These near identical values imply that the Hal-berstadt lavas are from source regions with similar radio-genic isotope compositions to the Newcastle adakites(Fig. 6).

Partial melting a basaltic protolith, at 3.2–3.8 GPa, withno added peridotite material, will form a magma containing<1.4 wt% MgO (Rapp et al., 1999). Conversely, by adding10% depleted or 12% enriched peridotite to the basalticstarting material and partially melting it, the reaction pro-duces a hybridised liquid with a greater “peridotite” com-ponent, e.g. 2.4–2.8 wt% MgO (Rapp et al., 1999). Withincreasing peridotite:basalt ratios the MgO compositionof the resultant liquid continues to increase (Rapp et al.,1999).

Many of the Jamaican adakites have 61.4 wt% MgOand, thus, probably did not substantially interact with aperidotite source rock (Hastie et al., 2010b). However,two samples (AHWG18 and AHWG19) have �2.0 wt%MgO. Although this MgO concentration is still low, it doessuggest that these adakites may have had limited interac-tion with peridotite material. Interestingly, models of gener-ating adakitic magmas by flat subduction of overthickenedoceanic crust by Gutscher et al. (2000) have a thin astheno-spheric mantle wedge remaining between the thick subduct-ing oceanic crust and the overlying lithosphere.

Therefore, although many of the Jamaican adakite mag-mas did not hybridise with a peridotite source, it is possiblethat some of the adakite magmas did undergo limitedhybridisation with a mantle source. Consequently, thenew data in this paper suggests that a small (laterally dis-continuous?) mantle wedge may have been present abovethe underthrusting oceanic plateau (Gutscher et al., 2000)and that as some of the Jamaican adakite magmas rapidlyascended melt:rock reactions took place to metasomatisethe peridotite. The adakitic slab melts, possibly with associ-ated slab-derived aqueous fluids, also imparted the negativeCe anomaly onto the HNB mantle source region. This en-riched, and compositionally variable, metasomatised peri-dotite underwent variable small degrees of partial meltingto generate the Halberstadt HNB magmas.

In contrast to the study of Rapp et al. (1999), theJamaican-type adakites are derived from lower pressures(�1.0–1.6 GPa) because both plagioclase and garnet are re-quired as residual phases to explain the composition of theNewcastle adakites (Hastie et al., 2010a,b). As a conse-quence, if the HNBs are derived from a metasomatisedsource, it has to have formed at pressures <1.6 GPa. Thiswill reduce the likelihood that phlogopite is present in thesource region, as it is usually stable at pressures >1.5 GPain metasomatised peridotite mantle (e.g. Adam et al.,1993; LaTourrette et al., 1995).

Within upper mantle anhydrous peridotite the garnet–spinel transition, although variable due to compositionaldifferences, is commonly proposed to be in the region�2.5–3.0 GPa (e.g. Kinzler, 1997; Walter, 1998). Con-versely, partial melting experiments of mixtures of tonaliteand peridotite at 1.5 GPa and 5–10 wt% H2O by Carrolland Wyllie (1989) have shown that, at a range of tempera-tures, the tonalite–peridotite partial melt is in equilibriumwith crystallising garnet, clinopyroxene, orthopyroxene

and pargasitic amphibole. Similarly, Sen and Dunn(1994a,b) partially melted mixtures of amphibolite (meta-morphosed subducting slab) and spinel peridotite at1.5 GPa and 950–1025 �C. Their experiments generated apartial melt of the amphibolite that was in equilibrium withgarnet, clinopyroxene, Ti-rich amphibole, plagioclase andrutile.

Therefore, even at low mantle pressures of 1.5 GPa, thereaction of an ascending slab-melt with the overlying peri-dotite mantle wedge should precipitate garnet and amphi-bole (Carroll and Wyllie, 1989; Sen and Dunn, 1994a;Prouteau et al., 2001). Residual garnet in the source regionof the Jamaican HNBs (e.g. Carroll and Wyllie, 1989) canexplain the fractionated HREEs in the Halberstadt rocksand their liquid lines of descent in Fig. 3. Unfortunately,the strong garnet signature prevents the identification of apossible “U-shaped” REE pattern in Fig. 5, which wouldbe formed by any residual amphibole.

However, it is possible that amphibole was in the mantlesource region and, if so, it would preferentially scavengeNb, Ta and Ti from further ascending acidic slab melts be-cause of these elements’ higher amphibole/acid melt parti-tion coefficients compared to Th, LREE and the MREE(e.g. Kepezhinskas et al., 1996; Sajona et al., 1996; Kleinet al., 1997). Kepezhinskas et al. (1995) also report Ti-richamphiboles in mantle xenoliths metasomatised by slabmelts in the Kamchatka arc. Fusion of this garnet amphi-bole peridotite source will generate a melt enriched inNb–Ta–Ti and this may explain the positive Nb, Ta andTi anomalies in Fig. 6.

6. A SIMPLE PETROGENETIC MODEL OF

HALBERSTADT FORMATION

6.1. Background on the Newcastle adakites

In order to model the generation of the Halberstadt la-vas from a theoretical metasomatised mantle source regionthe approximate composition of the adakitic melts that areresponsible for the metasomatism, must first be determined.Hastie et al. (2010b) used major and trace element andradiogenic isotope ratios to model the generation of theNewcastle adakites from a garnet amphibolitised oceanicplateau protolith.

The geochemistry of the Newcastle rocks can be ex-plained if the lavas were generated from a source regionthat was composed of amphibole, plagioclase and garnetwith minor amounts of quartz, zircon and apatite (Hastieet al., 2010b). This mineralogy is stable from �1.0 to1.6 GPa and would intersect the amphibole dehydrationmelting solidus at 850–900 �C (e.g. Peacock et al., 1994;Martin et al., 2005; Hastie et al., 2010a). Wolf and Wyllie(1994) show that a garnet amphibolite at �900 �C and1 GPa is composed of 55% amphibole, 25% plagioclase,8% garnet and 12% pyroxene. During similar experiments,Sen and Dunn (1994b) showed that 2.3% quartz is presentin a metamorphosed mafic protolith at 900 �C and 1.5 GPaand that pyroxene does not form until higher temperaturesare attained.

Page 16

5064 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Consequently, Hastie et al. (2010b) combined the twoexperimental results by replacing 2.3% of the pyroxenefrom Wolf and Wyllie (1994) with quartz and 0.1% zirconand 0.1% apatite to form a theoretical Newcastle adakitesource region. Using the Caribbean oceanic plateau as thestarting composition and by calculating the melt modesfrom changing phase proportions in the experiments inWolf and Wyllie (1994) and Sen and Dunn (1994b), Hastieet al. (2010b) were able to model partial melting of thissource to generate liquid compositions very similar to theNewcastle adakites.

However, the identification of the Halberstadt lavas asHNBs derived from a Newcastle adakite metasomatisedmantle source requires the original modelling of the New-castle magmas to be modified. In order to generate a Nb-rich Halberstadt source region, the Newcastle magmas arerequired to add large concentrations of Nb (and otherincompatible elements) into that source. It has been arguedthat Nb enrichment can be largely explained by amphiboleprecipitation (e.g. Kepezhinskas et al., 1996; Sajona et al.,1996). Unfortunately, most experimental studies show thatthe partition coefficient of Nb in amphibole is low andamphibole may not represent an ideal residual mineral toconcentrate Nb (LaTourrette et al., 1995; Klein et al.,1997; Adam and Green, 2006, 2010).

However, Carroll and Wyllie (1989) and Sen and Dunn(1994a) modelled slab-melt metasomatism at 1.5 GPa and950–1025 �C and showed that pargasitic amphibole is pro-duced in the reaction zone. Importantly, Tiepolo et al.(2000) determined that Nb partition coefficients in pargasit-ic amphiboles at 1.4 GPa and 950–1030 �C with basaniticand alkali-basaltic melts are varied, but can be moderatelyhigh (up to 1.63). These partition coefficients are generatedin equilibrium with basic melts and are not strictly applica-ble to amphibole in equilibrium with a TTG/D-like melt.Nevertheless, partition coefficients should increase withmore acidic compositions (e.g. LaTourrette et al., 1995),so if Nb partition coefficients can be high with a basic meltwhy are the experimental partition coefficients lower inTTG/D experiments (e.g. Klein et al., 1997)? To answer thisquestion is beyond the scope of this study, but the study ofTiepolo et al. (2000) suggests that precipitating pargasiticamphibole from a slab melt may be an effective mechanismof enriching the mantle in Nb. Consequently, the higherpargasitic partition coefficient values of Tiepolo et al.(2000) will be incorporated into the partial melt models ofboth the Newcastle and Halberstadt magmas.

6.2. The modified Newcastle source region

All the Newcastle magma partial melting variables inHastie et al. (2010b) remain unchanged apart from thechoice of partition coefficients. Partition coefficients arenow taken from Bedard (2006) except for Nb, Ta and Lain the pargasitic amphibole, which are from Tiepolo et al.(2000).

The Bedard (2006) dataset is used because (1) modellingthe generation of both Newcastle and Halberstadt magmasintroduces many more variables and we wish to simplify themodels so that there is less scope for modelling error, (2) the

use of one dataset removes the experimental and analyticalvariation between different studies/laboratories (e.g. Kleinand Langmuir, 1989; LaTourrette et al., 1995; Kinzler,1997; Walter, 1998), (3) the partition coefficients of Bedard(2006) are for temperatures of 850 �C in equilibrium withTTG/D melts, which are the conditions required for New-castle magma formation and (4) in addition to the TTG/D partition coefficients, Bedard (2006) also presents a com-plimentary set of updated peridotite partition coefficientsthat can be used in the modelling of the Halberstadt mantlesource; thus, supporting points (1) and (2). Partition coeffi-cients for Nb, Ta and La in the source amphibole are aver-age pargasitic values from Tiepolo et al. (2000) (0.6, 0.5 and0.2, respectively).

Hastie et al. (2010b) presented a Gd/Yb–Nb/Sm dia-gram to attempt to explain the compositional variation inthe Newcastle adakites. This diagram seemed to indicatethat by varying the amounts of garnet and amphibole inthe source region the composition of the Newcastle lavascan be explained. Nonetheless, the partial melt trends werenot an ideal fit; however, the new partial melt trend inFig. 9a now plots very close to the Newcastle adakite dataand if the garnet and amphibole starting composition isvaried from 8% and 55% to 4% and 59% respectively thepartial melt curve passes through the adakite data.

Hastie et al. (2010a) show a Nb/Yb–Yb diagram to illus-trate that the trace element composition of the Newcastleadakites can be explained with a partial melt curve fromthe fusion of a Caribbean oceanic plateau metabasic proto-lith as opposed to a N-MORB protolith. These results areconfirmed with modified partial melt curves in Fig. 9b usingthe new partition coefficients and garnet and amphibolemineral modes. Furthermore, Fig. 9c shows that with themodified model parameters the N-MORB normalised mul-ti-element signature of the Newcastle adakites can again bevery closely replicated. This confirms that the modified the-oretical garnet amphibolite source region of Hastie et al.(2010b) in this study represents a viable source protolithfor the Newcastle adakites.

6.3. The Halberstadt lavas

Mineral and melt modes for the peridotite mineralphases are taken from the spinel peridotite data of Johnson(1998). This spinel peridotite is then modified by introduc-ing variable percentages of garnet and amphibole that haveprecipitated from ascending slab-melts at the expense ofolivine, clinopyroxene, orthopyroxene and spinel (e.g.Rapp et al., 1999). To a first approximation this proceduresimulates the consumption of the original peridotite phasesand the precipitation of the metasomatic phases during mel-t:rock interaction. In reality, however, other minerals arelikely to precipitate, e.g. clinopyroxene and orthopyroxene(Carroll and Wyllie, 1989; Johnston and Wyllie, 1989;Rapp et al., 1999). Nevertheless, in order to simplify themodel, these minerals are not considered.

However, Hastie et al. (2010b) have shown that theNewcastle lavas are saturated in zircon and apatite. As aconsequence, it is highly likely that if the Newcastle mag-mas metasomatised the Halberstadt source region zircon

Page 17

Fig. 9. (a) Gd/Yb–Nb/Sm trace element diagram from Hastie et al. (2010b) showing updated partial melt models of possible source regionsfor the Newcastle lavas. Numbers on the curves represent percentage of partial melting. (b) Updated Nb/Yb–Yb diagram from Hastie et al.(2010a) illustrating that the Newcastle lavas plot on, or close to, the Caribbean oceanic plateau partial melting trend. (c) N-MORB normalisedmulti-element diagram showing the results of non-modal batch partial melting of a theoretical Newcastle source region from Wolf and Wyllie(1994) and Sen and Dunn (1994b) with an initial mineralogy of 59% amphibole, 25% plagioclase, 9.5% clinopyroxene, 4% garnet, 2.3% quartz,0.1% zircon and 0.1% apatite. Grey field represents the Newcastle adakite data from Hastie et al. (2010b) and the numbers represent data for1–30% non-modal batch partial melting. The starting composition is data from the Caribbean oceanic plateau (Hastie et al., 2008). Partitioncoefficients are taken from Tiepolo et al. (2000) and Bedard (2006). The melting mode of the garnet amphibolite is taken from melt reactions inSen and Dunn (1994b) and Wolf and Wyllie (1994) and includes quartz (0.08) + plagioclase (0.20) + amphibolite (0.72) = clinopyroxene(�0.4) + garnet (�0.1) + melt (�0.5). Normalising values from Sun and McDonough (1989).

Derivation of Jamaican high-Nb basalts 5065

and apatite would have precipitated and, therefore, we haveincluded these two minerals in the theoretical source region.Natural peridotite metasomatised by slab melts in Kam-chatka show evidence of apatite accessory mineral growth(Kepezhinskas et al., 1995). Apatite is also crystallised insub-liquidus mafic melt experimental studies at the requiredP–T conditions (e.g. Adam et al., 2007). Furthermore, zir-con can be a common accessory phase in metasomatisedperidotite (e.g. Konzett et al., 1998); although, recent stud-ies suggest that it is present in very small volumes <0.1 vol%(e.g. Kerr et al., 2010).

Fig. 9c indicates that much of the Newcastle data can beexplained by 5–30% partial melting of the modified sourceregion of Hastie et al. (2010b). We assume that the theoret-ical 10% adakite melt composition represents the averagecomposition of the primary adakite magmas as they ascendthrough the mantle (e.g. Sen and Dunn, 1994b). This pri-mary magma is then fractionated using variable propor-

tions of amphibole and garnet along with 0.1% zirconand 0.1% apatite precipitation (similar to other studies,e.g. Carroll and Wyllie, 1989; Sen and Dunn, 1994a; Adamet al., 2007; Kerr et al., 2010).

To be compatible with the Newcastle adakite modelling,the same partition coefficients from Tiepolo et al. (2000)and Bedard (2006) are used in the fractional crystallisationmodels. The composition of residues formed by 2.5–30%fractional crystallisation of this 10% adakite melt are addedto the spinel peridotite starting composition to form thetheoretical source region of the Jamaican HNBs. To becompatible with recent intraplate studies (e.g. Wymanet al., 2000; Adam and Green, 2006, 2010), the spinel peri-dotite starting composition is that of E-DMM of Workmanand Hart (2005). These “metasomatised” peridotite sourceregions are then partially melted. Partition coefficients forall of the elements in the spinel peridotite phases are takenfrom Bedard (2006). The garnet and amphibole partition

Page 18

5066 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

coefficients for a peridotite source are taken from LaTour-rette et al. (1995), Adam et al. (2007) and Adam and Green(2010).

The results of these models are shown in Fig. 10 as Zr/Yb–Nb diagrams. These elements are chosen because oftheir conservative nature in subduction environments.Thus, the data represents the geochemistry of the mantlewedge and any slab-melt component and eliminates anycompositional interference from aqueous slab-fluids. Sec-ondly, Zr and Yb will indicate how garnet and zircon willaffect the geochemistry of the magmas and Nb concentra-tion varies with degree of partial melting and enrichmentof the source due to amphibole precipitation.

Fig. 10. Zr/Yb–Nb diagrams showing the Halberstadt data plotted with ssource regions. In (a)–(c) the Nb partition coefficient for amphibole inpartition coefficient has been changed to 1.63. The first two numbers in tfractionated and subsequently partially melted and the third number reoriginal spinel peridotite mineralogy: Garnet–Amphibole–Material Addfractionated adakite assemblage composed of �50% garnet and �50%assemblage. In reality the assemblage also contains 0.1% zircon and 0.1%clarity. Numbers on the curves represent percentage of partial melting.

Fig. 10a shows the partial melting trends for a spinelperidotite whose original mineralogy has been replaced bya 2.5%, 5%, 10% or 15% fractionally crystallised Newcastleadakite component comprising 49.9% garnet, 49.9% amphi-bole, 0.1% zircon and 0.1% apatite. This diagram indicatesthat all but three of the Halberstadt lavas can be generatedby 0.1–1.0% partial melting of a source contaminated by2.5–10% adakite. Geochemical studies have suggested that,even at melt percentages as low as 0.1%, melt pathways candevelop to enable liquid to separate from the source residueand ascend (Salters and Hart, 1989; Johnson et al., 1990).Fig. 10b shows partial melt trends for a mantle sourcemetasomatised by 2.5–10% of an adakite component repre-

everal partial melt trends from theoretical metasomatised peridotitethe fractionating metasomatic phase is 0.6 and in (d)–(f) the Nbhe partial melt labels refer to the volume of amphibole and garnet

presents the volume of fractionated material that has replaced theed. For example, the 50–50–10 partial melt curve refers to a 10%

amphibole being used to replace 10% of the spinel peridotiteapatite, but we have overlooked this in the labels in the interests of

Page 19

Derivation of Jamaican high-Nb basalts 5067

sented by a fractionated assemblage of 74.9% garnet, 24.9%amphibole, 0.1% zircon and 0.1% apatite. Again, partialmelt modelling of these garnet-rich source regions can gen-erate all but the three most enriched samples from a 0.1–1.0% partial melt of a �2.5–10% metasomatised mantlesource region. Fig. 10c shows the partial melting trends ofspinel peridotite source regions that are replaced by a2.5%, 5%, 10%, 15% or 20% amphibole-rich assemblagecomprising 24.9% garnet, 74.9% amphibole, 0.1% zirconand 0.1% apatite. This suggests that, apart from the en-riched samples, the HNBs can be formed by 0.1–1.0% par-tial melting of source contaminated by 2.5–15% of thisassemblage.

From these models most of the Halberstadt samples canbe generated with variable small degrees of partial melting.However, instead of using the averaged pargasitic valuesfrom Tiepolo et al. (2000), models were formed using theNb partition coefficient from Sample 7 in Tiepolo et al.(2000), which is run at 1.4 GPa and 950 �C and is thusthe closest match to the P–T conditions at the site of possi-ble peridotite metasomatism. In these models, all variablesare the same as the modelling in Fig. 10a–c apart from theNb partition coefficient during amphibole precipitation inthe mantle that is now changed from 0.6 to 0.71. The resultsare very similar to Fig. 10a–c for most of the samples; how-ever, the higher partition coefficient does produce a slightlymore enriched mantle source and the amphibole-richassemblage can now nearly generate all of the HNB sam-ples from �0.1% to 1.0% partial melting.

Subsequently, the Nb partition coefficient value fromSample 7 in Tiepolo et al. (2000) (0.71) is replaced withthe highest pargasitic Nb partition coefficient obtained inthe Tiepolo experiments (Sample 17, KD = 1.63)(Fig. 10d–f). Fig. 10d shows the partial melting trends fora source contaminated by 49.9% garnet, 49.9% amphibole,0.1% zircon and 0.1% apatite. The Halberstadt lavas can begenerated by �1.0–2.5% partial melting of a source metaso-matised by 5–20% adakite. Partial melting of a contami-nated garnet-rich source can form the Halberstadt lavasfrom 0.1% to 1.0% partial melting of a �2.5–10% metaso-matised mantle source region. The amphibole-rich sourcein Fig. 10f can generate the Halberstadt rocks by �2.5–5.0% partial melting of source contaminated by 10–30%of this assemblage.

These simplified models show that the composition ofthe Halberstadt lavas can be theoretically generated froma mantle wedge metasomatised by continually ascendingslab-melts. It is emphasised, however, that there is notone particular compositional source region that is more via-ble than another. The models show that the Halberstadtsource region may be highly variable, both mineralogicallyand geochemically, and that a number of metasomatisedmantle compositions can explain the derivation of the Hal-berstadt HNBs. It is interesting that the more incompatibleelement-enriched Group 1 HNBs lie on partial meltingtrends of source regions contaminated with larger volumesof the slab-melt component; thus explaining their more en-riched nature.

It must also be stressed that there are still many uncer-tainties in these models that cannot be rectified without a

detailed knowledge of the exact composition (including vol-atiles), fO2, temperature and pressure of the Halberstadtsource region(s) (e.g. Johnston and Wyllie, 1989; Adamet al., 2007). In addition, the choice of partition coefficientsaffects the results of the modelling (e.g. Dalpe and Baker,1994; Johnson, 1998). For instance, if mantle peridotiteamphibole and clinopyroxene partition coefficients fromChazot et al. (1996) are used, the resulting models suggestthat the HNB magmas may be generated from slightly high-er degrees of partial melting (up to 10%) (Electronic Annex:Fig. 1). Nevertheless, within the constraints of the availabledata all of these models provide a plausible explanation forthe formation the Jamaican HNBs by the partial melting ofmantle peridotite metasomatised by slab melts.

7. HNB A DISTINCT ROCK TYPE?

HNBs are compositionally similar to some mantleplume-derived OIB lavas and other intraplate alkaline lavasthat are not related to mantle plume activity (e.g. Wood-head, 1996; Castillo, 2008; Adam and Green, 2010). How-ever, it has been demonstrated in this study, and others(e.g. Sajona et al., 1996; Castillo et al., 2007), that unlikeOIB lavas, HNBs are not generated from a pure mantleplume source or metasomatism of the mantle by carbona-tite/mantle plume melts. HNBs are also not intraplate lavaslike OIB and other alkaline lavas (Adam and Green, 2010);they are only found in subduction zone environments (e.g.Defant et al., 1992). Thus, it may be argued that HNBscould be considered a different rock type from OIB andother intraplate alkaline basalts, but it is difficult to classifythem based on a single petrogenetic process because theyare derived from mantle source regions formed from either(1) mixtures of enriched OIB-like and depleted MORB-typeupper mantle (Reagan and Gill, 1989; Castillo et al., 2007;Castillo, 2008) or (2) metasomatism of a mantle wedge byslab melts (Sajona et al., 1993, 1996; Kepezhinskas et al.,1996).

Additionally, we agree with the findings of Castillo et al.(2007), and consider that NEBs are also formed from thelatter two processes that give rise to HNBs. However, rela-tive to the HNB source regions, the NEBs are derived froma source composed of either (1) volumetrically less OIB-likeenriched mantle or (2) a metasomatised mantle that hasscavenged less incompatible elements from ascending slabmelts.

Consequently, we suggest that HNBs could be classifiedas a distinct rock type with Nb > 20 ppm, intraplate alka-line basalt compositions, but that are generated in subduc-tion zones by magmatic processes distinct from those thatgenerate other intraplate lavas.

8. CONCLUSIONS

The Halberstadt HNBs are divided into two composi-tional sub-groups that are derived from isotopically differ-ent spinel peridotite mantle source regions, which bothcontain garnet and amphibole as residual phases. The Hal-berstadt geochemistry demonstrates that the lavas cannotbe derived by partial melting of lower crustal ultramafic

Page 20

5068 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

complexes, metasomatised mantle lithosphere, subductingslabs, continental crust or mantle plume source regions.Instead, their composition, particularly the negative Ceanomalies, the high Th/Nb ratios and the similar isotopicratios to the Newcastle adakite lavas, suggests that theHalberstadt magmas are derived from a spinel peridotitesource region metasomatised by slab melts that precipitatedgarnet, amphibole, apatite and zircon.

These conclusions support the findings of Hastie et al.(2010a,b) whereby, as the Caribbean oceanic plateau wasbeing tectonically emplaced in between the two Americancontinents, the northern portion of the plateau underthrustJamaica. As it did so, the plateau partially melted to formthe Newcastle magmas that ascended and metasomatiseda thin (laterally discontinuous?) mantle wedge. The modi-fied mantle wedge peridotite was subsequently convectedto a sufficient depth where it partially melted to generatethe Halberstadt magmas.

ACKNOWLEDGEMENTS

Alan Hastie acknowledges NERC Ph.D. Studentship NER/S/A/2003/11215. The authors thank Prof. Andrew Saunders forconfirming the reliability of our data and Iain Neill and JulianPearce for fruitful geochemical discussions. We are grateful toCarib Cement (Jamaica) for allowing access to their quarries.Paterno Castillo, Ali Polat and an anonymous reviewer arethanked for their comments that improved the manuscript.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.gca.2011.06.018.

REFERENCES

Adam J., Green T. H. and Sie S. H. (1993) Proton microprobedetermined partitioning of Rb, Sr, Ba, Y, Zr, Nb and Tabetween experimentally produced amphiboles and silicate meltswith variable F content. Chem. Geol. 109, 29–49.

Adam J. and Green T. H. (2006) Trace element partitioningbetween mica- and amphibole-bearing garnet lherzolite andhydrous basanitic melt: 1. Experimental results and the inves-tigation of controls on partitioning behaviour. Contrib. Min-

eral. Petrol. 152, 1–17.

Adam J., Oberti R., Camara F. and Green T. H. (2007) An electronmicroprobe, LAM-ICP-MS and single-crystal X-ray structurerefinement study of the effects of pressure, melt–H2O concen-tration and fO2 on experimentally produced basaltic amphi-boles. Eur. J. Mineral. 19, 641–655.

Adam J. and Green T. H. (2010) Trace element partitioningbetween mica- and amphibole-bearing garnet lherzolite andhydrous basanitic melt: 2. Tasmanian Cainozoic basalts and theorigins of intraplate basaltic magmas. Contrib. Mineral. Petrol.

doi:10.1007/s00410-010-0570-7.Aguillon-Robles A., Calmus T., Benoit M., Bellon H., Maury R.

C., Cotton J., Bourgois J. and Michard F. (2001) Late Mioceneadakites and Nb-enriched basalts from Vizcaino Peninsula,Mexico: indicators of East Pacific Rise subduction belowsouthern Baja California? Geology 29, 531–534.

Atherton M. P. and Petford N. (1993) Generation of sodium-richmagmas from newly underplated basaltic crust. Nature 362,

144–146.

Andrew A. S., Hensen B. J., Dunlop A. C. and Agnew P. D. (1995)Oxygen and hydrogen isotope evidence for the origin ofplatinum-group element mineralization in Alaska-type intru-sions at Fifield, Australia. Econ. Geol. 90, 1831–1840.

Bedard J. H. (2006) A catalytic delamination-driven model forcoupled genesis of Archaean crust and sub-continental litho-spheric mantle. Geochim. Cosmochim. Acta 70, 1188–1214.

Benoit M., Aguillon-Robles A., Calmus T., Maury R. C., BellonH., Cotton J., Bourgois J. and Michard F. (2002) Geochemicaldiversity of Late Miocene volcanism in southern Baja Califor-nia, Mexico: implication of mantle and crustal sources duringthe opening of an asthenospheric window. J. Geol. 110, 627–

648.

Burke K. (1988) Tectonic evolution of the Caribbean. Annu. Rev.

Earth Planet. Sci. 16, 201–230.

Campbell I. H. (2007) Testing the plume theory. Chem. Geol. 241,

153–176.

Carroll M. R. and Wyllie P. J. (1989) Experimental phase relationsin the system tonalite–peridotite–H2O at 15 kb: implications forassimilation and differentiation processes near the crust–mantleboundary. J. Petrol. 30, 1351–1382.

Castillo P. R., Solidum R. U. and Punongbayan R. S. (2002) Originof high field strength element enrichment in the Sulu Arc,southern Philippines, revisited. Geology 30, 707–710.

Castillo P. R., Rigby S. J. and Solidum R. U. (2007) Origin of highfield strength element enrichment in volcanic arcs: geochemicalevidence from the Sulu Arc, southern Philippines. Lithos 97,

271–288.

Castillo P. R. (2008) Origin of the adakite–high-Nb basaltassociation and its implications for postsubduction magmatismin Baja California, Mexico. Geol. Soc. Am. Bull. 120, 451–462.

Chauvel C., Hofmann A. W. and Vidal P. (1992) HIMU-EM: theFrench Polynesian connection. Earth Planet. Sci. Lett. 110, 99–

119.

Chauvel C., McDonough W., Guille G., Maury R. and Duncan R.(1997) Contrasting old and young volcanism in Rurutu Island,Austral chain. Chem. Geol. 139, 125–143.

Chazot G., Menzies M. A. and Harte B. (1996) Determination ofpartition coefficients between apatite, clinopyroxene, amphi-bole, and melt in natural spinel lherzolites from Yemen:implications for wet melting of the lithospheric mantle.Geochim. Cosmochim. Acta 60, 423–437.

Cann J. R. (1970) Rb, Sr, Y, Zr and Nb in some ocean floorbasaltic rocks. Earth Planet. Sci. Lett. 10, 7–11.

Dalpe C. and Baker D. R. (1994) Partition coefficients for rare-earth elements between calcic amphibole and Ti-rich basaniticglass at 1.5 GPa, 1100 �C. Mineral. Mag. 58A, 207–208.

Dasgupta R., Hirschmann M. M. and Smith N. D. (2007) Partialmelting experiments of peridotite + CO2 at 3 GPa and genesisof alkalic ocean island basalts. J. Petrol. 48, 2093–2124.

Debari S., Kay S. M. and Kay R. W. (1987) Ultramafic xenolithsfrom Adagdak volcano, Adak, Aleutian Islands, Alaska:deformed igneous cumulates from the MOHO of an islandarc. J. Geol. 95, 329–341.

Defant M. J. and Drummond M. S. (1990) Derivation of somemodern arc magmas by melting of young subducted litho-sphere. Nature 347, 662–665.

Defant M. J., Jackson T. E., Drummond M. S., De Boer J. Z.,Bellon H., Feigenson M. D., Maury R. C. and Stewart R. H.(1992) The geochemistry of young volcanism throughoutwestern Panama and southeastern Costa Rica: an overview. J.

Geol. Soc. Lond. 149, 569–579.

Page 21

Derivation of Jamaican high-Nb basalts 5069

Draper G. (1986) Blueschists and associated rocks in easternJamaica and their significance for Cretaceous plate-margindevelopment in the northern Caribbean. Geol. Soc. Am. Bull.

97, 48–60.

Drummond M. S., Defant M. J. and Kepezhinskas P. K. (1996)Petrogenesis of slab-derived trondhjemite–tonalite–dacite/adakite magmas. Trans. Soc. Edinb.: Earth Sci. 87, 205–215.

Duncan R. A. and Hargraves R. B. (1984) Plate tectonic evolutionof the Caribbean region in the mantle reference frame. In The

Caribbean–South America Plate Boundary and Regional Tec-

tonics, vol. 162 (eds. W. E. Bonini, R. B. Hargraves and R.Shagam). Geol. Soc. Am. Mem., pp. 81–93.

Dupuy C., Liotard J. M. and Dostal J. (1992) Zr/Hf fractionationin intraplate basaltic rocks: carbonate metasomatism in themantle source. Geochim. Cosmochim. Acta 56, 2417–2423.

Edgar N. T., Ewing J. I. and Hennion J. (1971) Seismic refractionand reflection in the Caribbean Sea. Am. Assoc. Petrol. Geol. 55,

833–870.

Elliott T. (2003) Tracers of the Slab. Inside the subduction Factory.Geophys. Monogr. 138, 23–45.

Escuder Viruete J., Contreras F., Stein G., Urien P., Joubert M.,Perez-Estaun A., Friedman R. and Ullrich T. (2007) Magmaticrelationships and ages between adakites, magnesian andesitesand Nb-enriched basalt-andesites from Hispaniola: record of amajor change in the Caribbean island arc magma sources.Lithos 99, 151–177.

Farnetani C. G. and Samuel H. (2005) Beyond the thermal plumeparadigm. Geophys. Res. Lett. 32. doi:10.1029/2005GL022360.

Fitton J. G. and Dunlop H. M. (1985) The Cameroon line, WestAfrica, and its bearing on the origin of oceanic and continentalalkali basalt. Earth Planet. Sci. Lett. 72, 23–38.

Fitton J. G. and Godard M. (2004) Origin and evolution ofmagmas on the OJP. In Origin and Evolution of the OJP, vol.229 (eds. J. G. Fitton, J. J. Mahoney, P. J. Wallace and A. D.Saunders). Geol. Soc. Lond. Spec. Publ., pp. 151–178.

Geldmacher J., Hanan B. B., Blichert-Toft J., Harpp K., HoernleK., Hauff F., Werner R. and Kerr A. C. (2003) Hafniumisotopic variations in volcanic rocks from the Caribbean LargeIgneous Province and Galapagos hot spot tracks. Geochem.

Geophys. Geosyst. 4, Paper No. 2002GC000477.

Govindaraju K. (1994) Compilation of working values and samplesdescription for 383 geostandards. Geostand. Newsl. 18, 1–158.

Gutscher M. A., Maury R., Eissen J. P. and Bourdon E. (2000)Can slab melting be caused by flat subduction? Geology 28,

535–538.

Hart S. R. (1994) Mantle plumes and other mail from the mantle.Mineral. Mag. 58A, 382–383.

Hastie A. R., Kerr A. C., Pearce J. A. and Mitchell S. F. (2007)Classification of altered volcanic island arc rocks using immo-bile trace elements: development of the Th–Co discriminationdiagram. J. Petrol. 48, 2341–2357.

Hastie A. R., Kerr A. C., Mitchell S. F. and Miller I. (2008)Geochemistry and petrogenesis of Cretaceous oceanic plateaulavas in eastern Jamaica. Lithos 101, 323–343.

Hastie A. R. (2009) Is the Cretaceous primitive island arc (PIA)series in the circum-Caribbean region geochemically analogousto the modern island arc tholeiite (IAT) series? In Geology of the

Area Between North and South America, with Focus on the

Origin of the Caribbean Plate, vol. 328 (eds. K. H. James, M. A.Lorente and J. Pindell). Geol. Soc. Lon. Spec. Pub., pp. 397–409.

Hastie A. R., Kerr A. C., Mitchell S. F. and Millar I. (2009)Geochemistry and tectonomagmatic significance of lowerCretaceous island arc lavas from the Devils RacecourseFormation, eastern Jamaica. In Geology of the Area between

North and South America, with Focus on the Origin of the

Caribbean Plate, vol. 328 (eds. K. H. James, M. A. Lorente andJ. Pindell). Geol. Soc. Lond. Spec. Publ., pp. 337–359.

Hastie A. R. and Kerr A. C. (2010) Mantle plume or slab window?Physical and geochemical constraints on the origin of theCaribbean oceanic plateau. Earth Sci. Rev. 98, 283–293.

Hastie A. R., Kerr A. C., Mitchell S. F., Pearce J. A., McDonald I.,Millar I. and Wolstencroft M. (2010a) Do Cenozoic analoguessupport a plate tectonic origin for the Earth’s earliest conti-nental crust? Geology 38, 495–498.

Hastie A. R., Kerr A. C., Mitchell S. F., Pearce J. A., McDonald I.,Millar I., Barfod D. and Mark D. F. (2010b) Geochemistry andpetrogenesis of rhyodacite lavas in eastern Jamaica: a newadakite subgroup analogous to early Archaean continentalcrust? Chem. Geol. 276, 344–359.

Hauri E. H., Shimizu N., Dieu J. J. and Hart S. R. (1993) Evidencefor hotspot-related carbonatite metasomatism in the oceanicupper mantle. Nature 365, 221–227.

Hauff F., Hoernle K., Schmincke H.-U. and Werner R. (1997) Amid Cretaceous origin for the Galapagos hotspot: volcanolog-ical, petrological and geochemical evidence from Costa Ricanoceanic crustal segments. Geol. Rundsch. 86, 141–155.

Hill I. G., Worden R. H. and Meighan I. G. (2000) Yttrium: theimmobility–mobility transition during basaltic weathering.Geology 28, 923–926.

Hirose K. and Kushiro I. (1993) Partial melting of dry peridotitesat high pressures: determination of compositions of meltssegregated from peridotite using aggregates of diamond. Earth

Planet. Sci. Lett. 114, 477–489.

Hofmann A. W. (1997) Mantle geochemistry: the message fromoceanic volcanism. Nature 385, 219–229.

Hole M. J., Saunders A. D., Marriner G. F. and Tarney J. (1984)Subduction of pelagic sediments: implications for the origin ofCe-anomalous basalts from the Mariana Islands. J. Geol. Soc.

Lond. 141, 453–472.

Hole M. J., Rogers G., Saunders A. D. and Storey M. (1991)Relation between alkalic volcanism and slab-window forma-tion. Geology 19, 657–660.

Jackson T. A. and Smith T. E. (1978) Metasomatism in theTertiary volcanics of the Wagwater Belt, Jamaica. Geol.

Mijnbouw. 57, 213–220.

Jackson T. A. and Smith T. E. (1979) The petrochemistry of somemafic volcanics, Jamaica, West Indies. In Transactions of the

Fourth Latin American Geological Conference, Port-of-Spain,Trinidad, pp. 387–396.

Jackson T. A. (1987) The petrology of Jamaican Cretaceous andTertiary volcanic rocks and their tectonic significance. InProceedings of a Workshop on the Status of Jamaican Geology

(ed. R. Ahmad). J. Geol. Soc. Jamaica, pp. 107–119.Johnson K. T. M., Dick H. J. B. and Shimizu N. (1990)

Melting in the oceanic upper mantle: an ion microprobe studyof diopside in abyssal peridotites. J. Geophys. Res. 95, 2661–

2678.

Johnson K. T. M. (1998) Experimental determination of partitioncoefficients for rare earth and high-field-strength elementsbetween clinopyroxene, garnet, and basaltic melt at highpressures. Contrib. Mineral. Petrol. 133, 60–68.

Johnston A. D. and Wyllie P. J. (1989) The system tonalite–peridotite–H2O at 30 kbar, with applications to hybridization insubduction zone magmatism. Contrib. Mineral. Petrol. 102,

257–264.

Kawamoto T. and Holloway J. R. (1997) Melting temperature andpartial melt chemistry of H2O-saturated mantle peridotite to 11gigapascals. Science 276, 240–243.

Kay R. W. (1978) Aleutian magnesian andesites: melts fromsubducted Pacific Ocean crust. J. Volcanol. Geoth. Res. 4, 117–

132.

Page 22

5070 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Kempton P. D. (1995) Common Pb chemical procedures for silicaterocks and minerals, methods of data correction and anassessment of data quality at the NERC Isotope GeosciencesLaboratory, NIGL Report Series, 78.

Kepezhinskas P., Defant M. J. and Drummond M. S. (1995) Nametasomatism in the island-arc mantle by slab melt–peridotiteinteraction: evidence from mantle xenoliths in the NorthKamchatka Arc. J. Petrol. 36, 1505–1527.

Kepezhinskas P., Defant M. J. and Drummond M. S. (1996)Progressive enrichment of island arc mantle by melt–peridotiteinteraction inferred from Kamchatka xenoliths. Geochim.

Cosmochim. Acta 60, 1217–1229.

Kerr A. C., Tarney J., Marriner G. F., Nivia A., Saunders A. D.and Klaver G. T. (1996) The geochemistry and tectonic settingof late Cretaceous Caribbean and Colombian volcanism. J. S.

Am. Earth Sci. 9, 111–120.

Kerr A. C., Iturralde-Vinent M. A., Saunders A. D., Babbs T. L.and Tarney J. (1999) A new plate tectonic model of theCaribbean: implications from a geochemical reconnaissance ofCuban Mesozoic volcanic rocks. Geol. Soc. Am. Bull. 111,

1581–1599.

Kerr A. C., White R. V. and Saunders A. D. (2000) LIP reading:recognizing oceanic plateaux in the geological record. J. Petrol.

41, 1041–1056.

Kerr A. C., Aspden J. A., Tarney J. and Pilatasig L. F. (2002a) Thenature and provenance of accreted oceanic terranes in westernEcuador: geochemical and tectonic constraints. J. Geol. Soc.

Lond. 159, 577–594.

Kerr A. C., Tarney J., Kempton P. D., Spadea P., Nivia A.,Marriner G. F. and Duncan R. A. (2002b) Pervasive mantleplume head heterogeneity: evidence from the late CretaceousCaribbean–Colombian oceanic plateau. J. Geophys. Res. Solid

Earth 107(B7), Paper No. 2001JB000790.

Kerr A. C., White R. V., Thompson P. M. E., Tarney J. andSaunders A. D. (2003) No oceanic plateau–no Caribbean plate?The seminal role of an oceanic plateau in Caribbean plateevolution. In The Circum Gulf of Mexico and Caribbean:

Hydrocarbon Habitats Basin Formation and Plate Tectonics, vol.79 (eds. C. Bartolini, R. T. Buffler and J. Blickwede). Am.

Assoc. Petrol. Geol. Mem., pp. 126–268.Kerr A. C., Khan M., Mahoney J. J., Nicholson K. N. and Hall C.

M. (2010) Late Cretaceous alkaline sills of the south Tethyansuture zone, Pakistan: initial melts of the Reunion hotspot?Lithos 117, 161–171.

Kinzler R. J. and Grove T. L. (1992a) Primary magmas of mid-ocean ridge basalts: 1. Experiments and methods. J. Geophys.

Res. 97, 6885–6906.

Kinzler R. J. and Grove T. L. (1992b) Primary magmas of mid-ocean ridge basalts: 2. Applications. J. Geophys. Res. 97, 6907–

6926.

Kinzler R. J. and Grove T. L. (1993) Corrections and furtherdiscussion of the primary magmas of mid-ocean ridge basalts, 1and 2. J. Geophys. Res. 98, 22339–22347.

Kinzler R. J. (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition: application tomid-ocean ridge basalt petrogenesis. J. Geophys. Res. 102, 853–

874.

Klein E. M. and Langmuir C. H. (1987) Global correlations ofoceanic ridge basalt chemistry with axial depth and crustalthickness. J. Geophys. Res. 92, 8089–8115.

Klein E. M. and Langmuir C. H. (1989) Local versus globalvariations in oceanic ridge basalt composition: a reply. J.

Geophys. Res. 94, 4241–4252.

Klein M., Stosch H.-G. and Seck H. A. (1997) Partitioning of highfield-strength and rare-earth elements between amphibole and

quartz–dioritic to tonalitic melts: an experimental study. Chem.

Geol. 138, 257–271.

Konzett J., Armstrong R. A., Sweeny R. J. and Compston W.(1998) The timing of MARID metasomatism in the Kaapvaalmantle: an ion probe study of zircons from MARID xenoliths.Earth Planet. Sci. Lett. 160, 133–145.

Kurtz A. C., Derry L. A., Chadwick O. A. and Alfano M. J. (2000)Refractory element mobility in volcanic soils. Geology 28, 683–

686.

LaTourrette T., Hervig R. L. and Holloway J. R. (1995) Traceelement partitioning between amphibole, phlogopite, andbasanite melt. Earth Planet. Sci. Lett. 135, 13–30.

Le Bas M. J., Le Maitre R. W., Steckeisen A. and Zanettin B.(1986) A chemical classification of volcanic rocks based on thetotal alkali-silica diagram. J. Pet. 27, 745–750.

Le Bas M. J., Le Maitre R. W. and Woolley A. R. (1992) Theconstuction of the total alkali-silica chemical classification ofvolcanic rocks. Mineral. Petrol. 46, 1–22.

Macpherson C. G., Dreher S. T. and Thirlwall M. F. (2006)Adakites without slab melting: high pressure differentiation ofisland arc magma, Mindanao, the Philippines. Earth Planet.

Sci. Lett. 243, 581–593.

Mann P. and Burke K. (1990) Transverse intra-arc rifting: Palaeo-gene Wagwater Belt, Jamaica. Mar. Petrol. Geol. 7, 410–412.

Martin H., Smithies R. H., Rapp R., Moyen J.-F. and ChampionD. (2005) An overview of adakite, tonalite–trondhjemite–granodiorite (TTG), and sanukitoid: relationships and someimplications for crustal evolution. Lithos 79, 1–24.

Mauffret A. and Leroy S. (1997) Seismic stratigraphy and structureof the Caribbean igneous province. Tectonophysics 283, 61–104.

McCulloch M. T. and Gamble J. A. (1991) Geochemical andgeodynamical constraints on subduction zone magmatism.Earth Planet. Sci. Lett. 102, 358–374.

McDonald I. and Viljoen K. S. (2006) Platinum-group elementgeochemistry of mantle eclogites: a reconnaissance study ofxenoliths from the Orapa kimberlite, Botswana. Appl. Earth

Sci. (Trans. Inst. Min. Metall. B) 115, 81–93.

McDonough W. F. and Sun S.-S. (1995) The composition of theEarth. Chem. Geol. 120, 223–253.

Medard E., Schmidt M. W., Schiano P. and Ottolini L. (2006)Melting of amphibole-bearing wehrlites: an experimental studyon the origin of ultra-calcic nepheline-normative melts. J.

Petrol. 47, 481–504.

Munker C., Weyer S., Scherer E. and Mezger K. (2001) Separationof high field strength elements (Nb, Ta, Zr, Hf) and Lu fromrock samples for MC–ICPMS measurements. Geochem. Geo-

phys. Geosyst. 2, Paper No. 2001GC000183.

Nakamura Y. and Tatsumoto M. (1988) Pb, Nd, and Sr isotopicevidence for a multicomponent source for rocks of Cook–Austral Islands and heterogeneities of mantle plumes. Geochim.

Cosmochim. Acta 52, 2909–2924.

Nowell G. M. and Parrish R. R. (2001) Simultaneous acquisition ofisotope compositions and parent/daughter ratios by non-isotope dilution solution-mode Plasma Ionisation Multi-collector Mass Spectrometry (PIMMS). In Plasma Source Mass

Spectrometry – The New Millennium (eds. G. Holland andS. D. Tanner). Royal Society of Chemistry, Cambridge, pp.

298–310.

Nowell G. M., Kempton P. D. and Noble S. R. (1998) Highprecision Hf isotope measurements of MORB and OIB bythermal ionisation mass spectrometry: insights into the depletedmantle. Chem. Geol. 149, 211–233.

Palacz Z. A. and Saunders A. D. (1986) Coupled trace element andisotope enrichment in the Cook–Austral–Samoa islands, south-west Pacific. Earth Planet. Sci. Lett. 79, 270–280.

Page 23

Derivation of Jamaican high-Nb basalts 5071

Pallares C., Maury R. C., Bellon H., Royer J.-Y., Calmus T.,Aguillon-Robles A., Cotton J., Benoit M., Michard F. andBourgois J. (2007) Slab-tearing following ridge-trench collision:evidence from Miocene volcanism in Baja California, Mexico.J. Volcanol. Geoth. Res. 161, 95–117.

Peacock S. M., Rushmer T. and Thompson A. B. (1994) Partialmelting of subducting oceanic crust. Earth Planet. Sci. Lett.

121, 227–244.

Pearce J. A. (1982) Trace element characteristics of lavas fromdestructive plate boundaries. In Andesites (ed. R. S. Thorpe).John Wiley and Sons, Chichester, pp. 525–547.

Pearce J. A. (1996). A Users Guide to Basalt DiscriminationDiagrams. In Trace Element Geochemistry of Volcanic Rocks:

Applications for Massive Sulphide Exploration, vol. 12 (ed D. A.Wyman). Geol. Ass. Canada Short Course Notes, pp. 79�113.

Pearce J. A. and Peate D. W. (1995) Tectonic implications of thecomposition of volcanic arc magmas. Annu. Rev. Earth Planet.

Sci. Lett. 23, 251–285.

Pearce J. A. and Peate D. W. (1995) Tectonic implications of thecomposition of volcanic arc magmas. Annu. Rev. Earth Planet.

Sci. Lett. 23, 251–285.

Peccerillo R. and Taylor S. R. (1976) Geochemistry of Eocene calc-alkaline volcanic rocks from the Kastamonu area, northernTurkey. Contrib. Mineral. Petrol. 58, 63–81.

Pindell J. L. and Barrett S. F. (1990) Geological evolution of theCaribbean region: a plate tectonic perspective. In The Geology

of North America, vol. H (eds. G. Dengo, J. E. Case). Geol. Soc.

Am., Boulder, Colorado, The Caribbean Region, pp. 405–432.Pindell J. L. and Kennan L. (2009) Tectonic evolution of the Gulf

of Mexico, Caribbean and northern South America in themantle reference frame: an update. In Geology of the Area

Between North and South America, with Focus on the Origin of

the Caribbean Plate, vol. 328 (eds. K. H. James, M. A. Lorenteand J. Pindell). Geol. Soc. Lond. Spec. Publ., pp. 1–55.

Polat A. and Kerrich R. (2001) Magnesian andesites, Nb-enrichedbasalt-andesites, and adakites from late Archean 2.7 Ga Wawagreenstone belts, Superior Province, Canada: implications forlate Archean subduction zone petrogenetic processes. Contrib.

Mineral. Petrol. 141, 36–52.

Prouteau G., Scaillet B., Pichavant M. and Maury R. (2001)Evidence for mantle metasomatism by hydrous silicic meltsderived from subducted oceanic crust. Nature 410, 197–200.

Rapp R. P., Watson E. B. and Miller C. F. (1991) Partial melting ofamphibolite/eclogite and the origin of Archean trondhjemitesand tonalites. Precambrian Res. 51, 1–25.

Rapp R. P., Shimizu N., Norman M. D. and Applegate G. S.(1999) Reaction between slab-derived melts and peridotite inthe mantle wedge: experimental constraints at 3.8 GPa. Chem.

Geol. 160, 335–356.

Reagan M. K. and Gill J. B. (1989) Coexisting calcalkaline andhigh-niobium basalts from Turrialba volcano, Costa Rica:implications for residual titanites in arc magma sources. J.

Geophys. Res. 94, 4619–4633.

Revillon S., Arndt N. T., Chauvel C. and Hallot E. (2000)Geochemical study of ultramafic volcanic and plutonic rocksfrom Gorgona Island, Colombia: the plumbing system of anoceanic plateau. J. Petrol. 41, 1127–1153.

Revillon S., Chauvel C., Arndt N. T., Pik R., Martineau F.,Fourcade S. and Marty B. (2002) Heterogeneity of theCaribbean plateau mantle source: heterogeneity of the Carib-bean plateau mantle source: Sr, O and He isotopic composi-tions of olivine and clinopyroxene from Gorgona Island. Earth

Planet. Sci. Lett. 205, 91–106.

Robinson E., Lewis J. F. and Cant R. V. (1972) Field guide toaspects of the geology of Jamaica. In International Field

Institute Guidebook to the Caribbean Island Arc System

1970. American Geological Institute, Washington, DC, pp. 1–45, 1–4.

Royse K. R., Kempton P. D. and Darbyshire F. D. (1998)Procedure for the analysis of rubidium–strontium and samar-ium–neodymium isotopes at the NERC Isotope GeosciencesLaboratory, NIGL Report Series, 121.

Sajona F. G., Maury R. C., Bellon H., Cotton J., Defant M. J. andPubellier M. (1993) Initiation of subduction and the generationof slab melts in western and eastern Mindanao, Philippines.Geology 21, 1007–1010.

Sajona F. G., Maury R. C., Bellon H., Cotton J. and Defant M. J.(1996) High field strength element enrichment of Pliocene–Pleistocene island arc basalts, Zamboanga Peninsula, WesternMindanao (Philippines). J. Petrol. 37, 693–726.

Salters V. J. M. and Hart S. R. (1989) The Hf paradox and the roleof garnet in the source of mid-ocean ridge basalts. Nature 342,

420–422.

Salters V. J. M. and Hart S. R. (1991) The mantle sources of oceanridges, islands and arcs: the Hf-isotope connection. Earth

Planet. Sci. Lett. 104, 364–380.

Salters V. J. M. and White W. M. (1998) Hf isotope constraints onmantle evolution. Chem. Geol. 145, 447–460.

Sen C. and Dunn T. (1994a) Experimental modal metasomatism ofa spinel lherzolite and the production of amphibole-bearingperidotite. Contrib. Mineral. Petrol. 119, 422–432.

Sen C. and Dunn T. (1994b) Dehydration melting of a basalticcomposition amphibolite at 1.5 and 2.0 GPa: implications forthe origin of adakites. Contrib. Mineral. Petrol. 117, 394–409.

Schwab B. E. and Johnston A. D. (2001) Melting systematics ofmodally variable, compositionally intermediate peridotites andthe effects of mineral fertility. J. Petrol. 42, 1789–1811.

Smithies R. H., Champion D. C. and Cassidy K. F. (2003)Formation of Earth’s early Archaean continental crust.Precambrian Res. 127, 89–101.

Sorensen S. S. (1988) Petrology of amphibolite-facies mafic andultramafic rocks from the Catalina Schist, southern California:metasomatism and migmatization in a subduction zone meta-morphic setting. J. Metamorph. Geol. 6, 405–435.

Sorensen S. S. and Grossman J. N. (1989) Enrichment of traceelements in garnet amphibolites from a paleo-subduction zone:Catalina Schist, southern California. Geochim. Cosmochim.

Acta 53, 3155–3177.

Sun S.-S. and McDonough W. F. (1989) Chemical and isotopesystematics of oceanic basalts: implications for mantle compo-sition and processes. Magmatism in the Ocean Basins. Geol.

Soc. Lon. Spec. Pub. 42, 313–345.

Takahashi E. and Kushiro I. (1983) Melting of a dry peridotite athigh pressures and basalt magma genesis. Am. Mineral. 68, 859–

879.

Takahashi E. (1986) Melting of a dry peridotite KLB-1 up to14 GPa: implications on the origin of peridotitic upper mantle.J. Geophys. Res. 91, 9367–9382.

Tatsumi Y. (2003) Some constraints on arc magma genesis. Insidethe subduction factory. Geophys. Monogr. 138, 277–292.

Thompson P. M. E., Kempton P. D., White R. V., Kerr A. C.,Tarney J., Saunders A. D. and Fitton J. G. (2003) Hf–Ndisotope constraints on the origin of the Cretaceous Caribbeanplateau and its relationship to the Galapagos plume. Earth

Planet. Sci. Lett. 217, 59–75.

Tiepolo M., Vannucci R., Oberti R., Foley S., Bottazzi P. andZanetti A. (2000) Nb and Ta incorporation and fractionation intitanian pargasite and kaersutite: crystal-chemical constraintsand implications for natural systems. Earth Planet. Sci. Lett.

176, 185–201.

Walter M. J. (1998) Melting of garnet peridotite and the origin ofkomatiite and depleted lithosphere. J. Petrol. 39, 29–60.

Page 24

5072 A.R. Hastie et al. / Geochimica et Cosmochimica Acta 75 (2011) 5049–5072

Weaver B. L. (1991) Trace element evidence for the origin ofocean–island basalts. Geology 19, 123–126.

White W. M. and Patchett P. J. (1984) Hf–Nd–Sr isotopes andincompatible element abundances in island arcs: implicationsfor magma origins and crust–mantle evolution. Earth Planet.

Sci. Lett. 67, 167–185.

Winchester J. A. and Floyd P. A. (1976) Geochemical magma typediscrimination: application to altered and metamorphosed basicigneous rocks. Earth Planet. Sci. Lett. 28, 459–469.

Wolf M. B. and Wyllie P. J. (1994) Dehydration-melting ofamphibolite at 10 kbar: the effects of temperature and time.Contrib. Mineral. Petrol. 115, 369–383.

Woodhead J. D. (1996) Extreme HIMU in an oceanic setting: thegeochemistry of Mangaia Island (Polynesia), and temporalevolution of the Cook–Austral hotspot. J. Volcanol. Geoth. Res.

72, 1–19.

Workman R. K. and Hart S. R. (2005) Major and trace elementcomposition of the depleted MORB mantle (DMM). Earth

Planet. Sci. Lett. 231, 53–72.

Wyman D. A., Ayer J. and Devaney J. R. (2000) Niobium-enrichedbasalts from the Wabigoon subprovince, Canada: evidence foradakitic metasomatism above an Archean subduction zone.Earth Planet. Sci. Lett. 179, 21–30.

Yogodzinski G. M., Volynets O. N., Koloskov A. V. andSeliverstov N. I. (1994) Magnesian andesites and the subduc-tion component in a strongly calcalkaline series at Piip volcano,far western Aleutians. J. Petrol. 34, 163–204.

Yogodzinski G. M., Kay R. W., Volynets O. N., Koloskov A. V.and Kay S. M. (1995) Magnesian andesite in the westernAleutian Komandorsky region: implications for slab meltingand processes in the mantle wedge. Geol. Soc. Am. Bull. 107,

505–519.

Zindler A. and Hart S. (1986) Chemical geodynamics. Annu. Rev.

Earth Planet. Sci. 14, 493–571.

Associate editor: Steve Shirey