The geochemical components that distinguish Loa- and Kea ...

Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

ScienceDirect

Geochimica et Cosmochimica Acta 185 (2016) 160–181

The geochemical components that distinguish Loa- andKea-trend Hawaiian shield lavas

Frederick A. Frey a,⇑, Shichun Huang b, Guangping Xu c, Klaus P. Jochum d

aMassachusetts Institute of Technology, Cambridge, MA 02139, United StatesbUniversity of Nevada Las Vegas, Las Vegas, NV 89154, United StatescSchlumberger – Doll Research, Cambridge, MA 02139, United States

dClimate Geochemistry Department, Max Planck Institute for Chemistry, D-55020 Mainz, Germany

Received 20 July 2015; accepted in revised form 5 April 2016; available online 30 April 2016

Abstract

Recent (<5 Ma) Hawaiian volcanoes define two sub-parallel spatial trends, Loa and Kea. Despite the short distance(�30 km) between adjacent volcanoes on these trends, most of the Loa-trend shield lavas are geochemically distinct from mostof the Kea-trend shield lavas. These geochemical differences arise from small amounts of the LOA component in the source ofLoa-trend shield lavas. This component is most prominent in the uppermost shield lavas of Koolau, Lanai and Kahoolawevolcanoes. Correlations between abundance ratios of incompatible elements and isotopic ratios of Sr, Nd, Hf and Pb inHawaiian shield lavas indicate that the LOA component consists of three geochemically distinct materials formed by diverseprocesses. A gabbroic adcumulate (i.e. no trapped melt) with abundant cumulus plagioclase is responsible for the high Sr/Nd,La/Th and La/Nb in Loa-trend shield lavas relative to Kea-trend shield lavas. Also it has relatively low 206Pb/204Pb and high208Pb/204Pb at a given 206Pb/204Pb, consistent with the low U/Pb and Th/Pb that are characteristic of plagioclase; these dis-tinctive Pb isotope ratios require a long-time interval, �3 Ga, to develop. This material is most abundant in the uppermostshield lavas of Koolau volcano. Possible origins of adcumulate gabbros with abundant cumulus plagioclase are the loweroceanic and continental crust. A second material in the LOA component is distinctive because it is offset from the linear trendof 176Hf/177Hf versus 143Nd/144Nd, known as the terrestrial array, to high 176Hf/177Hf at low 143Nd/144Nd. This offset requiresan ancient material with high Lu/Hf. It is equally abundant in the shield lavas at Koolau, Lanai and Kahoolawe volcanoes.Possible origins of this material are ancient pelagic sediment or ancient depleted lithosphere. A third material in the LOAcomponent is characterized by relatively high 87Sr/86Sr, but the Rb/Sr of this material is too low to explain the high 87Sr/86Srin 4.5 Ga. A relatively recent (<1 Ga) event must have lowered the Rb/Sr. This material is most abundant in the shield lavas ofKahoolawe volcano.� 2016 Elsevier Ltd. All rights reserved.

Keywords: Mantle plume; Hawaiian volcanism; Mantle heterogeneity

1. INTRODUCTION

From 1838 to 1842 a US Navy operation, the USExploring Expedition, was charged with mapping andcharting little-known regions of the Pacific and Southern

http://dx.doi.org/10.1016/j.gca.2016.04.010

0016-7037/� 2016 Elsevier Ltd. All rights reserved.

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

Oceans. From September 1840 to April 1841 the expeditionwas in Hawaii. The young geologist, J. D. Dana, one ofnine on board scientists, made many significant observa-tions of Hawaiian volcanoes, including recognition of a sys-tematic age progression to the northwest, now described asa hotspot track, and a spatial arrangement of volcanoesthat are now known as the Loa- and Kea-trends (Dana,1849; Appleman and Dana, 1987). In order to understand


mailto:[email protected]


http://crossmark.crossref.org/dialog/?doi=10.1016/j.gca.2016.04.010&domain=pdf

F.A. Frey et al. /Geochimica et Cosmochimica Acta 185 (2016) 160–181 161

the volcanism that occurs along a hotspot track it is neces-sary to understand (1) the physical processes that createddiscrete islands forming sub-parallel trends (Hieronymusand Bercovici, 1999), and (2) the processes that determinethe geochemical characteristics of magmas erupted alongthe hotspot track. In this research we use geochemical dif-ferences between Loa- and Kea-trend shield lavas to under-stand the processes that created geochemical differencesbetween volcanoes forming Loa-trend shields from MaunaLoa to Koolau and Kea-trend shields from Kilauea to EastMolokai (Fig. 1).

The archetype Loa and Kea volcanoes, Mauna Loa andKilauea, are in the subaerial shield stage of rapid growthand samples of unaltered lava flows are easily collected;consequently there are abundant compositional data forMauna Loa and Kilauea shield lavas. Powers (1955) foundthat at a given MgO content, the abundance of SiO2 is �2%lower in Kilauea lavas than in Mauna Loa lavas. Heinferred that these differences reflect distinct magmabatches whose compositions reflect different sources foreach volcano. Wright (1971) also compared the composi-tion of shield lavas erupted at Mauna Loa and Kilauea vol-canoes. He found that lavas from both volcanoes have thesame range of K2O/P2O5 and concluded that uniformity ofthis ratio of incompatible elements suggests a commonsource for Mauna Loa and Kilauea lavas. ConsequentlyWright (1971) concluded that geochemical differencesbetween these adjacent and simultaneously active volcanoes

Loa Trend

Kea Trend

Mauna Kea

Mauna Loa

Kohala

Haleakala

East Molokai

Koolau

200 km

158° 156°22°

20°

18°Loihi

West MauiWest M

olokai

LanaiKahoolawe

Mahukona

HualalaiKilauea

Molokai FractureZone (MFZ)

Penguin B

ank

Waianae

Fig. 1. Locations of Hawaiian volcanoes showing the subparallelKea (blue) and Loa (red) trends. Abouchami et al. (2005) suggestedthat the different geochemical characteristics of shield lavas eruptedalong these trends ended at the Molokai Fracture Zone. The Keatrend clearly ends at East Molokai Volcano. However thedistinctive geochemical features of the LOA component found inshield lavas from Kahoolawe and Lanai are also present in theuppermost shield lavas of Koolau. The youngest Koolau lavas, theMakapuu-stage defined by Haskins and Garcia (2004), are themost extreme manifestation of the LOA component; consequentlythe Loa-trend has commonly been extended to include Koolauvolcano, indicated by dashed red line, and even to Kauai further tothe northwest (e.g., Weis et al., 2011). (For interpretation of thereferences to colour in this figure legend, the reader is referred tothe web version of this article.)

were caused by differences in process rather than differencesin source compositions.

Do different sources or different processes cause the geo-chemical differences between Loa- and Kea-trend shieldlavas? Tatsumoto (1978) answered this question when hefound that lavas from each volcano on the Island of Hawaiihave distinctive Pb isotope ratios. In particular, he empha-sized that lavas erupted at Kea-trend volcanoes have higher206Pb/204Pb than lavas erupted at Loa-trend volcanoes, aresult that provides strong evidence for geochemically dis-tinct sources for Loa-and Kea-trend volcanoes. The useful-ness of Pb isotopic ratios to distinguish Loa- and Kea-trendvolcanoes was strengthened by Abouchami et al. (2005)who found that 208Pb*/206Pb* is a better discriminantbetween Loa- and Kea-trend lavas than 206Pb/204Pb.208Pb*/206Pb* = (208Pb/204Pb – 29.475)/(206Pb/204Pb –9.307) is a measurement of the radiogenic ingrowth of208Pb and 206Pb since earth formation.

Abouchami et al. (2005) used Pb isotope data for Loa-and Kea-trend volcanoes to infer an asymmetrically zonedplume (Fig. 2). The zonation in this model differs from theprevious concentrically zoned models (Hauri et al., 1996;Kurz et al., 1996; Lassiter et al., 1996). However, Pb isotopedata for lavas from East Molokai-West Molokai-PenguinBank volcanoes that define a transect perpendicular to thestrike of the Loa and Kea trends (Fig. 1) show that it isnot possible to use Pb isotope ratios to distinguish betweenLoa- and Kea-trend lavas in these older shields (Xu et al.,2014). For example, the West Molokai shield on the Loa-trend (Fig. 1) includes lavas with both Loa- and Kea-typegeochemical characteristics (Xu et al., 2007). Additionalcomplexity in using 208Pb*/206Pb* as a discriminant betweenLoa- and Kea-trend lavas is that the Kea-trend volcanoes,Kilauea, Mauna Kea and Haleakala have some basalts with208Pb*/206Pb* ratios like Loa-trend lavas (Eisele et al., 2003;Marske et al., 2007, 2008, respectively).

The geochemical characteristics of shield lavas at twopairs of adjacent volcanoes, Koolau and Waianae volca-noes on Oahu and Mauna Loa and Kilauea volcanoes onHawaii (Fig. 1) were compared by Frey and Rhodes(1993). They showed that for each pair at a given MgO con-tent there are differences in abundance of SiO2, CaO, K2Oand TiO2. The development of new analytical techniques,such as instrumental neutron activation analysis (INAA),isotope dilution (ID) using mass spectrometers and mostrecently inductively coupled plasma mass spectrometry(ICP-MS), has led to acquisition of precise and accurateconcentration data for incompatible trace elements inHawaiian shield lavas. These data show that at a givenMgO content, Kilauea lavas have higher abundances ofincompatible elements than Mauna Loa lavas, and differentincompatible element ratios, such as Th/La, Zr/Nb, Sr/Nband K/Nb, than Mauna Loa lavas (Fig. 3; Pietruszka et al.,2013). Frey and Rhodes (1993) also showed that the incom-patible element ratios, such as Sr/Nb and Zr/Nb, are corre-lated with 87Sr/86Sr and 143Nd/144Nd. In each Loa- andKea-trend pair, the Loa-trend lavas have higher 87Sr/86Srand lower 143Nd/144Nd than the shield lavas in the adjacentKea-trend volcano. However, there are poor correlationsbetween the parent/daughter ratio, Rb/Sr, and 87Sr/86Sr

par�ally ordered

Kilauea

Mauna Kea

Kohala

Loihi

MaunaLoa

Mahukona

Fig. 2. Geochemical zonation models for the Hawaiian plume. Filled circles showing location of the 5 volcanoes on the Island of Hawaii andthe submarine Loihi Volcano located southeast of the island. The two spatial trends are shown with Kea-trend volcanoes indicated by bluecircles and line and Loa-trend volcanoes as green circles and red dashed line. The large circles indicate the area of the currently activevolcanoes. Loihi, the youngest, is a submarine volcano; Mauna Loa, the oldest active volcano, is northwest of Loihi; and Kilauea, theyoungest Kea volcano, is east of Mauna Loa. Left Panel shows the bilateral asymmetrically zoned model that is based on high precision Pbisotopic data of Abouchami et al. (2005). Note the abrupt boundary between Kea and Loa trend geochemical characteristics. Right Panel isbased on incompatible element abundance ratios in melt inclusions included in olivine phenocrysts from Koolau and Haleakala basalt (Renet al., 2005) and on recently acquired isotopic data for more volcanoes, including the transect from East Molokai to West Molokai to PenguinBank (Fig. 1 and Xu et al., 2014). These additional data show that there are lavas with Kea geochemical features erupted at Loa volcanoes andvice versa. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

162 F.A. Frey et al. /Geochimica et Cosmochimica Acta 185 (2016) 160–181

and between Sm/Nd and 143Nd/144Nd (e.g., Roden et al.,1994).

Frey and Rhodes (1993) discussed constraints on thestructure of the Hawaiian plume arising from the differ-ences between adjacent volcanoes in major and trace ele-ment abundances and in Sr and Nd isotope ratios. Theyproposed alternative physical characteristics of the Hawai-ian plume that could lead to the geochemical differencesbetween the Loa- and Kea-trend volcanoes. For example,tilting of an ascending plume caused by interaction withthe overlying oceanic lithosphere, varying proportions ofgeochemical heterogeneities distributed within the plumeor a plume consisting of separate diapirs for each volcano.Also the source of Hawaiian shield lavas may have varyingproportions of peridotite and pyroxenite. Hauri (1996) con-cluded that the composition of lavas forming Loa-trendshields contain a SiO2-rich melt (dacite) derived from apyroxenite in the source of Loa-trend shield lavas. InHawaiian shield lavas isotope ratios of Sr, Nd and Pb arecorrelated with abundance of SiO2, TiO2 and CaO, afteradjustment to a common MgO content. Jackson et al.(2012) evaluated in more detail the correlations betweenmajor element compositions and radiogenic isotopic ratiosin Hawaiian shield lavas. They suggested that lavas withrelatively low SiO2 and radiogenic Pb isotope ratios (Keageochemical characteristics) are derived from a source ofmixed pyroxenite/eclogite and peridotite whereas the lavaswith relatively high SiO2 and unradiogenic Pb isotope ratios(Loa geochemical characteristics) are derived from pyrox-enite/eclogite. The pros and cons for derivation of Loa-trend shield lavas from a pyroxenite source are discussedby Hauri (1996), Herzberg (2006), Sobolev et al. (2005),

Jackson et al. (2012), Rhodes et al. (2012), Herzberg et al.(2014) and Huang and Humayun (2016).

Jackson et al. (2012) also proposed that the variableSiO2 content of Kea- and Loa-trend shield lavas reflectsthe extent of slab dehydration during subduction; in partic-ular the high SiO2 content of the LOA component reflectsthe absence of dehydration. Variable extents of dehydrationof a subducted oceanic lithosphere are a variable that hasbeen used to explain intershield differences in abundancesof fluid mobile elements such as Ba and Sr (Fig. 7 ofPietruszka et al., 2013).

2. DATA

Large data sets of accurate abundances of highly incom-patible elements in basaltic rocks have only recently becomeavailable as ICP-MS emerged as the analytical method ofchoice. This paper uses abundance data for the incompati-ble elements Ba, Th, Nb, La, Sr, Nd, Zr, Sm, Eu, Dy andLu in Hawaiian shield lavas, Kilauea (82 samples), MaunaKea (99 samples), Mauna Loa (36 samples), Lanai (40 sam-ples), Kahoolawe (29 samples) and Koolau (34 Makapuu-stage samples and 90 samples recovered by the Koolau Sci-entific Drilling Project (KSDP)). Isotopic ratios of Sr, Nd,Hf and Pb for a subset of these samples were also used.These data are compiled in Appendix A. Our approach isto utilize isotopic ratios and ratios of incompatible elementabundances to understand igneous processes. It is necessaryto be aware of element mobility during post-eruption alter-ation (Appendix B; Clague et al., 2016 and Huang et al.,2016), contamination during sample preparation (AppendixB) and inter-laboratory differences (Appendix C).

0

5

10

15

20

25

Ba Th Nb La Ce Pb Nd Sr Sm Zr Hf Eu Ti Tb Dy Ho Y Er Tm Yb Lu

Bas

alt /

Prim

itive

Man

tle

Kilauea (7.35% MgO)

Mauna Loa (6.64% MgO)MORB (7.58%MgO)

Increasing Incompatibility

Fig. 3. Incompatible element abundances in shield basalt fromKilauea and Mauna Loa normalized to primitive mantle(McDonough and Sun, 1995). These samples are glasses, KL2-Gfrom Kilauea and ML3B-G from Mauna Loa, that are used asreference samples because their compositions, including abundanceof incompatible elements, are well known (Jochum et al., 2006).Relative to an average Mid-Ocean Ridge Basalt (MORB) (Galeet al., 2013), the Hawaiian shield basalts are enriched in incom-patible elements ranging from Dy to Ba but are depleted in theheavy rare earth elements (HREE) from Er to Lu. For this paperthe important observations are the geochemical differences betweenKea-trend (Kilauea) and Loa-trend (Mauna Loa) basalt. Specifi-cally, as shown by bold green lines are: (Ba/Th)PM > 1 with similarratios in Kea and Loa basalt; (La/Nb)PM > 1 in Loa basalt and <1in Kea basalt; (Sr/Nd)PM > 1 in both Loa and Kea basalt but Loabasalt > Kea basalt. (For interpretation of the references to colourin this figure legend, the reader is referred to the web version of thisarticle.)


3. DISCUSSION

3.1. Overview

Our goals are: (1) to establish the geochemical differ-ences between the shield lavas erupted on the Kea- andLoa-trends; (2) to evaluate the geochemical heterogeneityof the LOA component and its variable abundance in thesource of lavas forming different shields; and (3) to presenthypotheses for the origin and processes that created theLOA component.

We use abundance ratios of incompatible elements toevaluate the mass balance equation:

Mauna Loa shield lava ¼ Kilauea shield lava

þ the LOA component:

This mass balance equation is used to determine theabundance ratios of incompatible elements and radiogenicisotope ratios of the LOA component that must be addedto a Kilauea basalt to make a Mauna Loa basalt. Thisequation is used because relative to Kea-trend lavas, Loa-trend lavas have unusual geochemical characteristics (e.g.,Frey and Rhodes, 1993; Pietruszka et al., 2013). Kilauea

shield lavas are assumed to be derived from a peridotitesource. The shield lava compositions used are Mauna Loaand Kilauea samples in Fig. 3 (Appendix D).

The first step is to determine the differences in incompat-ible element abundance ratios between Loa- and Kea-trendshield lavas. Such differences have previously been recog-nized (e.g., Frey et al., 1994; Hemond et al., 1994;Hofmann and Jochum, 1996; Huang and Frey, 2003,2005; Pietruszka et al., 2013). However the recent increasein high quality abundance data for highly incompatible ele-ments such as Th, Nb and La, combined with long drillcores into the shields of Koolau and Mauna Kea volcanoesled us to look for trends that were not previouslyrecognized.

3.2. Abundance ratios of selected elements

We first identify the abundance ratios that distinguishLoa- and Kea-trend lavas. A plot of incompatible elementabundances in lavas from Loa- and Kea-trend volcanoesnormalized to primitive mantle (PM) shows important fea-tures (Fig. 3). For example, (Ba/Th)PM > 1 for both Loa-and Kea-trend lavas whereas (Ba/Th)PM < 1 for MORB;(Nb/La)PM < 1 in both MORB and Loa-trend shield lavasbut >1 in Kea-trend lavas; (Th/La)PM < 1 in MORB andboth Loa- and Kea-trend lavas; (Sr/Nd)PM < 1 for MORBbut both Loa- and Kea-trend lavas have (Sr/Nd)PM > 1with Loa-trend shield lavas having the highest values;(Zr/Nb)PM >1 in MORB but <1 in Loa- and Kea-trendlavas with Loa lavas having higher Zr/Nb than Kea lavas.Also MORB have (Dy/Lu)PM �1 whereas (Dy/Lu)PM > 1in both Loa- and Kea-trend shield lavas (subscript PM indi-cates the ratio normalized to the Primitive Mantle ratio andMORB is Mid-Ocean Ridge Basalt).

3.2.1. Sr and Nd

During igneous processes that do not involve feldspar,Sr and Nd are similar in incompatibility; that is, the ratioof solid/melt partition coefficients, DSr/DNd, is close to 1.Because DSr

Plag/Melt > 1 and DNdPlag/Melt < 1 (e.g., Drake and

Weill, 1975; Bindeman and Davis, 2000; Bedard, 2006),the simplest process for explaining Sr enrichment, i.e.,(Sr/Nd)PM >1, is formation of an adcumulate rock rich incumulus plagioclase. It is necessary that the cumulate rockbe an adcumulate, that is without trapped melt, becausesuch melt would overwhelm the geochemical signature ofplagioclase. If this adcumulate is metamorphosed to pyrox-enite/eclogite in a closed system and then partially meltedto >50%, the melt will inherit the high Sr/Nd of the sourcerock (Appendix E)

A large variation in MgO content, �5–25% (Fig. 4b), istypical of Hawaiian shield lavas and is caused by olivineremoval or addition. However, olivine does not affect theSr/Nd ratio because olivine/melt partition coefficients forSr and Nd are very low (Bedard, 2005). Combined geo-chemical and petrographic studies of the liquid line of des-cent for Hawaiian tholeiitic basalt find that plagioclasesaturation is not reached until the MgO content of the melthas decreased to about 7% MgO (e.g., Wright, 1971). Frac-tionation of plagioclase from melts with <7% MgO lowers

Fig. 4. Panels a–e: Sr/Nd versus Sr/Nb, MgO, 143Nd/144Nd, 206Pb/204Pb and Ba/Th in Loa and Kea shield lavas. Primitive mantle (PM)element ratios are indicated by arrows. Because Sr in shield lavas can be mobile during alteration, only samples with K2O/P2O5 in the intervalfrom 1.3 to 2.2 are plotted in Figs. 4, 7–9 and 12 (see Appendix B; Huang and Frey, 2003; Xu et al., 2007; Huang et al., 2016). Shield basaltsamples from Kahoolawe are not plotted because this filter eliminates most of the samples. In this and subsequent figures the geochemicaldifferences between Mauna Loa lavas and the two Kea shields (Kilauea and Mauna Kea) in our figures are modest. Lavas from the Koolau,Lanai and presumably Kahoolawe shields define the extreme geochemical characteristics of Loa-trend lavas. Koolau lavas are divided into theyoungest shield lavas that are exposed at Makapuu Point and older samples recovered in a 679 m drill core obtained by the Koolau ScientificDrilling Project (KSDP). Panel a: Sr/Nb and Sr/Nd are highly correlated, and range by factors of 3.3 and 2.4, respectively. The superimposedblack and red squares indicate mean MORB (Gale et al., 2013) and old Pacific MORB (Fekiacova et al., 2007). Panel b: Sr/Nd vs. MgOshowing that at >8% MgO, Sr/Nd does not vary systematically with MgO content. In contrast, two Mauna Loa samples with low, <7% MgOcontent, have the lowest Sr/Nd and Sr/Nb because they have fractionated plagioclase which lowers Sr content. The near vertical trend forLanai samples with 6–7% MgO is also likely to be a result of plagioclase fractionation. Panels c and d: Sr/Nd vs 143Nd/144Nd and 206Pb/204Pbshowing that, except for the two Mauna Loa samples with low MgO contents, in each panel the fields for individual shields define an inversetrend. Panel e: Sr/Nd vs Ba/Th. This trend is surprisingly poor suggesting that different processes controlled these ratios. However the twoMauna Loa lavas with low MgO have low Sr/Nd and Ba/Th, probably reflecting plagioclase fractionation. The gray field showing nocorrelation is for tholeiitic basalt from the Galapagos archipelago (Saal et al., 2007). (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)



the Sr/Nd ratio significantly. In Fig. 4b examples of plagio-clase fractionation are the near vertical trend of Lanai lavasand the two Mauna Loa lavas with Sr/Nd <10. Scatter inthe Sr/Nd vs MgO trend for a specific shield trend canreflect mixing of low-MgO magmas that are plagioclase sat-urated with high-MgO olivine-saturated magmas (Fig. 12of Rhodes and Vollinger, 2004).

Although Sr enrichment can occur as result of processesoccurring in the crust, the correlations of Sr/Nd with143Nd/144Nd and 206Pb/204Pb (Fig. 4c, d) indicate that rela-tive enrichment of Sr in Loa-trend shield lavas is a sourcecharacteristic. Fig. 4a–d is an example of the trends definedby the intershield differences between Loa- and Kea-trendshield lavas. In this paper our focus is on understandingthe processes that created geochemical differences betweenshields. In all figures that plot ratios of incompatible ele-ments and isotopic ratios, the fields for lavas from thetwo Kea-trend volcanoes Kilauea and Mauna Kea overlap.The Loa-trend lavas from the Makapuu stage of Koolauand Lanai volcanoes are the most extreme; lavas cored bythe Koolau Scientific Drilling Project (KSDP) are olderthan the Makapuu stage lavas and together with MaunaLoa lavas they fill the gap between the Kea-trend lavasand Loa extreme lavas (Fig. 4). Lavas from Loa-trendKahoolawe volcano are not included in plots involvingincompatible element abundances because most Kahoolawelavas are highly altered; however isotopic ratios of Kahoo-lawe lavas are in Figs. 12 and 13 because the effects of alter-ation were minimized by acid leaching prior to analysis.

The largest Sr enrichments, (Sr/Nd)PM = 3.1, have beenfound in a small number of melt inclusions in olivine phe-nocrysts of Mauna Loa lavas by Sobolev et al. (2000)who proposed that Sr enrichment in melt inclusions in oli-vine phenocrysts of some Mauna Loa shield lavas could beexplained by a source composed of gabbros that containcumulus plagioclase with Sr/Nd of �200 that was sub-ducted and incorporated into the Mauna Loa mantlesource. Plagioclase and clinopyroxene have very differentmineral/melt partition coefficients that can be used to iden-tify a petrogenetic role for these two minerals (Fig. 5a).Gabbroic rocks with cumulus plagioclase and clinopyrox-ene are common in the lower oceanic crust (e.g., Meyeret al., 1989; Elthon et al., 1992; Hart et al., 1999; Natlandand Dick, 2001; Coogan et al., 2001). Analyses of coexistingcumulus plagioclase and clinopyroxene in these gabbrosfrom the oceanic crust reflect their mineral/melt partitioncoefficients; that is relative to clinopyroxene, plagioclase isenriched in Ba, La, Pb, Sr and Eu and depleted in Nband Zr (Fig. 5a, b). Also the abundances of incompatibleelements in coexisting cumulus plagioclase and clinopyrox-ene in gabbroic cumulates from the lower oceanic crust areconsistent with forming in equilibrium with a MORB melt(Fig. 5c, d).

Increasing pressure as the oceanic plate subducts con-verts the gabbroic cumulate to eclogite. If this metamor-phism occurs in a closed system, the eclogite retains theSr enrichment resulting from cumulus plagioclase. If thiseclogite is then incorporated into the mantle source ofLoa-trend basalt, the basalt will have a high Sr contentinherited from the cumulus plagioclase that is no longer

present in the eclogite. Sobolev et al. (2000) described thisSr enrichment as resulting from ‘‘Ghost Plagioclase”.

An alternative hypothesis for the relative enrichment ofSr in ocean island basalt is assimilation of plagioclase in thegabbroic lower crust of islands (Danyushevsky et al., 2003;Saal et al., 2007; Peterson et al., 2014). Sobolev et al. (2000)evaluated this alternative for the melt inclusions in olivinephenocrysts of Mauna Loa lava. They argued that themajor element contents of the melt inclusions were not sat-urated with plagioclase, a necessity for the shallow-levelassimilation model. In a study of basalt erupted in theGalapagos Islands, Saal et al. (2007) found that basalt inthe eastern part of the archipelago have the trace elementsignature of plagioclase, e.g., high Sr/Nd. They concludedthat direct assimilation of plagioclase-rich lower crust isunlikely because this process requires ‘‘an unreasonablylarge proportion of digested plagioclase-rich gabbro” (seeSupplementary material Fig. 3 of Saal et al. (2007)); theyfavored a more complex crustal process that involves diffu-sive control during reaction of percolating basaltic magmawith plagioclase-rich cumulates in the lower crust underly-ing the volcano.

A strong argument against the Sr enrichment in theLOA component arising from interaction of hotspotderived magmas with the lithosphere beneath Hawaiianvolcanoes is that the Sr/Nd is inversely correlated with143Nd/144Nd (Fig. 4c); that is, the high Sr/Nd of the LOAcomponent has relatively low 143Nd/144Nd. This result isinconsistent with the high 143Nd/144Nd typical of MORBrelated lithosphere (Fig. 4c).

Lead isotope ratios provide a way to assess the timing ofthe process that created the geochemical characteristics ofGhost Plagioclase (Peterson et al., 2014). In plagioclasePb is a moderately incompatible but U and Th are highlyincompatible (Fig. 5a; Leeman, 1979; Bindeman et al.,1998; Bindeman and Davis, 2000; Norman et al., 2005;Bedard, 2006; Bedard, 2007). Clinopyroxene also hasDPb > DU and DTh (Fig 5a; Pertermann et al., 2004). Sincethe major phases of a plagioclase- and clinopyroxene-richcumulate have low U/Pb, little ingrowth of 206Pb and207Pb will occur. If ascending basaltic magma interacts withcumulates in the lower oceanic crust of islands, this lowercrust will be too young to have developed retarded206Pb/204Pb and 207Pb/204Pb. In contrast, a long time isrequired for recycling of eclogite derived from aplagioclase-rich cumulate in the lower oceanic crust intothe mantle where it was incorporated into a mantle plume,such as the Hawaiian hotspot. If the gabbro to eclogitetransition occurred within a closed system, the eclogite willhave retarded 206Pb/204Pb and 207Pb/204Pb ratios reflectingthe low U/Pb of the gabbroic cumulate. Melting of thiseclogite in a hotspot environment could explain the low206Pb/204Pb (Fig. 6a), and 207Pb/204Pb ratios of the LOAComponent (Fig. 8c of Xu et al., 2014).

What age of the recycled gabbro is consistent with themeasured Pb isotope ratios of the LOA component? Toanswer this question the first step is to take the Pb isotopiccompositions of average modern MORB and calculate thetemporal trend of Pb isotope ratios in the MORB mantlesource backwards in time in 0.5 Ga increments (Fig. 6b)

Fig. 5. Panel a – Clinopyroxene/melt partition coefficients define a positive slope with small anomalies (such as for Pb). Plagioclase/meltpartition coefficients are very distinctive in having a negative slope, indicated by red solid and dashed line, from light REE (La, Ce and Nd) toheavy REE (Er to Lu), strong peaks for Ba, Sr and Eu, a moderate peak at Pb and valleys at Th, Nb, Zr and Hf (data from Appendix G).Panel b: Plagioclase/Clinopyroxene abundance ratios of incompatible elements in gabbroic cumulates recovered by drilling on the AtlantisMassif on the Mid-Atlantic Ridge (Drouin et al., 2009). Note that relative to clinopyroxene, plagioclase has peaks in Ba, Pb, Sr and Eu andvalleys for Nb and Zr. Panels c and d: Primitive mantle-normalized abundances of incompatible elements in clinopyroxene and plagioclasefrom gabbroic adcumulates recovered by ODP drilling on the Atlantis Massif (Drouin et al., 2009). In both panels the abundance patterns incumulus clinopyroxene and plagioclase are similar to clinopyroxene and plagioclase in equilibrium with MORB (the thick black lines) usingthe partition coefficients in Appendix G. Panel e: Garnet/melt and clinopyroxene/melt partition coefficients (experimental sample A343 ofPertermann et al., 2004). Both minerals show the expected positive slope indicating that their mineral/melt partition coefficients decrease asthe elements become more incompatible. Important differences between these minerals are that garnet has a much steeper slope from Dy to Luand pronounced minimums at Sr and Th. The high DZr/DNb (>10) in both minerals indicates that Zr/Nb is affected by variable extents ofmelting garnet peridotite. Nevertheless Zr/Nb in oceanic island basalt, including Hawaiian lavas, is usually correlated with 143Nd/144Nd;consequently, Zr/Nb differences in the sources of Hawaiian shield lavas are not completely masked by the effects of processes. There is verygood agreement between different experimental studies of clinopyroxene and garnet; e.g., for garnet see Figs. 2 and 5 in Pertermann et al.(2004). For clinopyroxene/melt partition coefficients note the similarity of experimental results of Pertermann et al. (2004) with clinopyroxenephenocryst/melt ratios in a Kilauea basalt (Norman et al., 2005). (For interpretation of the references to colour in this figure legend, the readeris referred to the web version of this article.)



using the average U/Pb and Th/U ratios of MORB (Galeet al., 2013). Then we choose the average of eight leastaltered gabbros from a 500 meter gabbroic section of drillcore from Ocean Drilling Project Hole 735B near theSouthwest Indian Ridge (Table 2 of Hart et al., 1999). Thisaverage has the geochemical characteristics of cumulateplagioclase, such as high Sr/Nd (Fig. 9 of Hart et al.,1999). However, plagioclase has very low partition coeffi-cients for Th and U, much lower than clinopyroxene(Table 3 of Blundy and Wood, 2003). Therefore clinopyrox-ene with its high U/Pb, Th/Pb and Th/U will control theevolution of lead isotopic ratios. Evidence that cumulusclinopyroxene in the gabbros at Hole 735B controlled the

evolution of Pb isotope ratios is that the Th/U in unalteredgabbro is 3.75 (Table 2 of Hart et al., 1999) leads to a ratioof (cumulus cpx in gabbro = 3.75)/(MORB average 3.09)equal to 1.21. This value is within the experimentally deter-mined range of 0.8–1.3 for (Th/U)cpx/(Th/U)melt (Fig. 1 ofBlundy and Wood, 2003).

We assume that this gabbro formed at various times inthe past in increments of 0.5 Ga from 0.5 to 3.5 Ga. Usingthe average Th/U and U/Pb of the gabbros, we calculatedthe evolution of the 206Pb/204Pb and 208Pb*/206Pb* to pre-sent day. These are the blue and black lines in Fig. 6c.The intersection of these lines with the red regression linedefined by Pb ratios for Loa- and Kea-trend shield lavasshows that to create the Pb isotopic characteristics of theLOA component the average age of unaltered gabbro isslightly greater than 3 Ga. For altered gabbro with lowerTh/U of 1.78, reflecting a higher U content caused by lowtemperature alteration, the calculated formation age isslightly greater than 3.5 Ga.

It is a surprising result that both the altered gabbro andunaltered gabbro yield similar model Pb ages for the LOA

Fig. 6. (a) 206Pb/204Pb versus 208Pb*/206Pb* for Hawaiian shieldlavas. Data are from the compilation of Huang et al. (2011a), withadditional data for Mauna Loa and Lanai (Appendix A). Thechange from Kea- to Loa-trend shields is a decrease 206Pb/204Pband increase in 208Pb*/206Pb*. Estimates of upper continental crust(UCC) and lower continental crust (LCC) are from Zartman andHaines (1988). Clearly UCC is not suitable for the LOA compo-nent. In contrast, LCC has the low 206Pb/204Pb and high 208Pb*/206-

Pb* that are characteristics of the LOA component.(208Pb*/206Pb* = (208Pb/204Pb – 29.475)/(206Pb/204Pb) –9.307) is ameasure of the radiogenic ingrowth of 208Pb and 206Pb since earthformation). (b) 206Pb/204Pb versus 208Pb*/206Pb* for MORB mantlesource from present day to 3.5 Ga. We start with the averagemodern MORB values of 206Pb/204Pb = 18.412,207Pb/204Pb = 15.515, 208Pb/204Pb = 38.100, U/Pb = 0.179, andTh/Pb = 0.551 (Gale et al., 2013). Assuming that melting at mid-ocean ridge does not fractionate U/Pb or Th/Pb, we calculate thePb isotopic compositions of MORB source backwards in time insteps of 0.5 Ga. (c) A plagioclase-rich adcumulate gabbro formedfrom ancient MORB will have U/Pb and Th/Pb ratios determinedby its cumulus phases and their proportions. The Pb isotopicevolution of this gabbro will deviate from that of MORB source.We calculated forward in time the present-day Pb isotopicsignature of the gabbro using the unaltered and altered gabbrocompositions from Tables 2 and 7 of Hart et al. (1999). The labeledages are the ages of the gabbro formation, and unless specified, theincremental step is 0.5 Ga. The LOA component is at theintersection of the regressed Hawaiian data (red line) with ancient,3–3.5 Ga, gabbros indicated by the curved trends (black and blue).The model ages of ancient gabbros are not very sensitive to the U/Pb and Th/Pb ratios of the gabbros. This is because 206Pb/204Pbdecreases and 208Pb*/206Pb* increases with the age of MORBmantle source, and ancient MORB mantle source has lower206Pb/204Pb and higher 208Pb*/206Pb*. As long as the gabbros havesufficiently low (U, Th)/Pb ratios, they preserve their ancient Pbisotopic signatures, i.e., low 206Pb/204Pb and high 208Pb*/206Pb*,which are not sensitive to the Th/U ratios of the gabbros. (Forinterpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

3

7

10

12

15

17

20

10 20 30 40 50

Zr/N

b

Sr/Nb

7

10

12

15

17

20

22

0.5125 0.5127 0.5129 0.5131

Zr/N

b

143Nd/144Nd

PM

b

PM

a Mauna KeaKilauea

Mauna Loa

KSDPMakapuu

Koolau:

LOA Trend

KEA Trend

Lanai

Fig. 7. Zr/Nb versus 143Nd/144Nd and Sr/Nb: Arrows labeled PMindicate the primitive mantle ratio of McDonough and Sun (1995).Panel a: The negative Zr/Nb versus 143Nd/144Nd correlation showsthat Zr/Nb is a source characteristic; however, each shield showsconsiderable scatter that may reflect the effects of process on Zr/Nb, such as variable extents of partial melting. In fact most Panelsin Figs. 4, 7 and 8 show a broad field for each shield. We infer thatthere are processes occurring during the growth of each shield thatintroduce scatter into the inter-shield trends. The inter-shield trendsdefined by the average ratio in each shield are the trends that we useto define the LOA component. Panel b: shows that Zr/Nb is highestin Loa shields lavas and that Zr/Nb is positively correlated with Sr/Nb.

a

b

c

PM

PM (0.12)

Mauna KeaKilauea

Mauna Loa

KSDPMakapuu

Koolau:

LOA Trend

KEA Trend

Lanai

PM (0.12)

Fig. 8. (a) La/Nb versus Sr/Nb; (b) Th/Nb versus Sr/Nb and (c)Th/La versus Zr/Nb. The well defined intershield trends in Panels aand c show that Loa-trend lavas have high Sr/Nb and La/Nb butlow Th/La as expected for cumulus plagioclase (Fig. 5b). Incontrast Th/Nb, a ratio not controlled by plagioclase, is notcorrelated with Sr/Nb (Panel b).


component, despite their very different Th/U ratios (1.78 vs.3.75, respectively). The explanation is that radiogenicingrowth of 208Pb is slower than that of 206Pb in the MORBsource reservoir; consequently, its 206Pb/204Pb increases and208Pb*/206Pb* decreases with time (Fig. 6b). That is, ancientPb is characterized by low 206Pb/204Pb and high 208Pb*/206-

Pb*; this Pb isotopic signature is characteristic of the LOAcomponent (Fig. 6a). The ancient gabbros, altered or unal-tered, formed as adcumulates rich in cumulus plagioclasethey are characterized by high Sr/Nd and low U/Pb andTh/Pb ratios; consequently, they preserve their ancient Pbisotopic signatures. There is no evidence that the U/Pband Th/Pb of the cumulus plagioclase and clinopyroxenewere increased by the massive loss of Pb during dehydra-tion of a subducted plate (Kelley et al., 2005).

3.2.2. Zr and Nb

Because Zr and Nb are minimally affected by alteration,and their abundance can be accurately determined by X-ray

fluorescence, Zr/Nb has been used extensively to providepetrogenetic constraints for basalt. They are high-field-strength elements whose extent of incompatibility differssubstantially (Fig. 3). Both clinopyroxene and garnet canchange the Zr/Nb ratio because for each phase, DZr/DNb

is >1 (Fig. 5e); consequently, Zr/Nb varies with the extentof melting of peridotite. Nevertheless this ratio can be usedto evaluate geochemical heterogeneity in the mantle sourcesof basalt. In particular Hawaiian shield lavas define inversetrends for Zr/Nb versus 206Pb/204Pb (Fig. 11 of Chen et al.,1996) and for Zr/Nb versus 143Nd/144Nd (Fig. 7a).

The Zr/Nb ratio is also highly correlated with Sr/Nb,thereby indicating that these ratios reflect differencesbetween the sources of Loa- and Kea-trend lavas (Fig. 7b).

Fig. 9. La/Nb and Th/La, ratios are correlated with 143Nd/144Nd and 206Pb/204Pb. Therefore these ratios of highly incompatible elements arerelated to the sources of Hawaiian shield lavas.


The high ratios in the Loa-trend lavas are consistent withthe LOA component forming as an adcumulate gabbrocomposed of cumulus plagioclase and clinopyroxene. Suchrocks are common in the oceanic crust (e.g., Meyer et al.,1989; Elthon et al., 1992; Hart et al., 1999; Natland andDick, 2001; Coogan et al., 2001). Moreover plagioclaseand clinopyroxene in gabbros cored from the Atlantis Mas-sif on the Mid Atlantic Ridge (Drouin et al., 2009) haveincompatible element ratios consistent with equilibriumcrystallization from a MORB melt (Fig. 5c, d). Increasingpressure during subduction converts the gabbro to eclogite.If metamorphism occurred in a closed system, the eclogiteretains the Sr enrichment caused by cumulus plagioclaseand the relatively high Zr/Nb caused by clinopyroxene.

3.2.3. Th, Nb and La

We focus on Th, Nb and La because these elements arehighly incompatible during the petrogenesis of tholeiiticbasalt, and they are relatively immobile during post-magmatic alteration of Hawaiian shield lavas (e.g., Huanget al., 2016). Because few processes can change the ratiosof highly incompatible elements, it is surprising that La/Nb, Th/La and Th/Nb ratios in Hawaiian shield lavas varyby factors of 1.9, 1.8 and 1.6, respectively (Figs. 8 and 9).

Ratios of La/Nb and Th/La, but not Th/Nb, are correlatedwith Zr/Nb and Sr/Nb and with radiogenic isotope ratios(Figs. 8 and 9). What process created these correlations?Equilibrium partial melting changes ratios only if the extentof melting is less than the (bulk solid)/melt partition coeffi-cients (Appendix F). The low extent of melting required,<5%, to change La/Nb, Th/La and Th/Nb ratios is toolow to generate tholeiitic basalt (e,g., Green andRingwood, 1968).

In contrast, residues created by fractional melting willhave very different incompatible element ratios than thesource and these ratios are very sensitive to extent of melt-ing (e.g., Johnson et al., 1990; Appendix H). However thefractional melting model is not realistic because it requiresthat solid-state diffusion is able to maintain a homogeneousresidue as fractional melts are segregating instantaneouslyfrom the residue. More realistic melting models, such asdynamic melting that assume a critical proportion of meltis not removed from the residue, i.e., the residue is porous,are much less effective in changing abundance ratios ofincompatible elements.

High Sr/Nb, La/Nb and La/Th are characteristic of pla-gioclase (Fig. 5a and Appendix I). For La the plagioclasepartition coefficient at an appropriate An/Ab ratio is


�0.07 (Appendix G). The sensitivity of La/Nb and Th/Lato extent of melting of a pyroxenite/eclogite is shown inAppendix E. A simple interpretation of the trends inFig. 8a and c is mixing of Kea-trend melt with melt derivedby high, >50%, extents of melting of pyroxenite/eclogitethat formed as the plagioclase-rich adcumulate enteredthe mantle. If the system was closed during this reaction,the pyroxenite/eclogite inherited the incompatible elementsin the plagioclase-rich adcumulate. Melts formed by high(>50%) extents of melting of the pyroxenite/eclogite arebasaltic, and they would have the characteristic ratios ofcumulus plagioclase, i.e. high Sr/Nb, La/Nb and low Th/La (Appendix E).

Both Loa and Kea components have (Th/Nb)PM and(Th/La)PM < 1 (Fig. 8b, c; Hofmann and Jochum, 1996);therefore the sources of all Hawaiian shield lavas are rela-tively depleted in Th. Depletion in Th is consistent withthe high Ba/Th in both Loa- and Kea-trend lavas (Fig. 4e;Hofmann and Jochum, 1996). Plagioclase has high DBa/DTh (Fig. 5a; Bindeman and Davis, 2000; Tepley et al.,2010). High Ba/Th has been attributed to accumulationof plagioclase and recycling of cumulate plagioclase-richgabbros into the source of Hawaiian lavas (Hofmann andJochum, 1996; Yang et al., 2003) and to reaction of basalticmagma with plagioclase-rich cumulates in the oceanic litho-sphere (Saal et al., 2007). However Ba/Th is poorly corre-lated with Sr/Nd (Fig. 4e; Yang et al., 2003) possiblybecause of Ba mobility during alteration process (Huanget al., 2016).

3.3. The Origin of the LOA Component: The Role of Lower

Oceanic Crust

Lavas from different shields define strong correlationsbetween La/Nb, Th/La, Sr/Nb and Zr/Nb (Fig. 8). In aplot of plagioclase/(basaltic melt) partition coefficient ver-sus atomic number the REE define a negative slope, i.e.,LREE are more compatible than HREE, and there areprominent peaks at Ba, Pb, Sr and Eu and valleys at Th,Nb, Zr and Hf (Fig. 5a). Appendix I shows that cumulusplagioclase is the explanation for the high Sr/Nd, Sr/Nb,La/Nb and low Th/La that characterize the LOA compo-nent. Adcumulate gabbros with cumulus plagioclase andclinopyroxene are common in the oceanic lower crust(e.g., Natland and Dick, 2001). However the low143Nd/144Nd of Loa-trend shield lavas relative to MORBis not explained by involving only the oceanic lower crust(Fig. 4c).

3.4. The Origin of the LOA component: an alternative

approach to evaluating the role of lower oceanic crust

Pietruszka et al. (2013) evaluated the source of geochem-ical heterogeneity in the Hawaiian plume by assuming thatthe heterogeneity reflects processes at ancient subductionzones where oceanic crust was introduced into the uppermantle and subsequently incorporated into the Hawaiianhotspot. Instead of using the intershield differences in ratiosof highly incompatible elements, which they described asenigmatic, Pietruszka et al. (2013) constrained the source

of Hawaiian shield lavas by calculating a mass balanceamong partial melts of ambient Hawaiian mantle peridotiteand two components of subducted oceanic crust, variablydehydrated upper crust of altered MORB and lower crustof unaltered gabbro. For parental Makapuu-stage Koolaushield lavas, they concluded that the lava source was dom-inantly peridotite that included 8–10% altered MORB and7–8% lower oceanic crust.

Previously we argued that the high La/Nb of the Loa-component reflects a deficiency in Nb caused by a smallamount of sediment in the Loa-component (Jacksonet al., 1999; Huang and Frey, 2003, 2005). This conclusionis consistent with Li, Ca, Nd, Hf and Os isotopic data forMakapuu-stage lavas in Koolau volcano (Lassiter andHauri, 1998; Blichert-Toft et al., 1999; Huang et al.,2011b; Harrison et al., 2015). However we now favor theinterpretation that the high Sr/Nd and La/Nb that distin-guish the LOA component were created by cumulusplagioclase.

Pietruszka et al. (2013) proposed a very different model;specifically that the high La/Nb of the LOA componentindicates that rutile was not present in recycled oceaniccrust that led to Loa-trend lavas. They inferred that theabsence of rutile indicates that these slabs were not stronglydehydrated during subduction. We point out that Fig. 3c ofPietruszka et al. (2013) shows that the component with highSr/Nd and La/Nb in their mass balance model is ‘‘freshlower crustal gabbro”. We suggest that both high Sr/Ndand La/Nb in their gabbroic lower oceanic crust reflectcumulus plagioclase. Therefore their inference of less dehy-dration of the subducted oceanic crust for Loa-trend shieldsis not required.

Another difficulty with the less dehydration inference isthat Hauri (2002) in a study of the volatile content ofolivine-hosted melt inclusions concluded that the subductedslab component contributing to Koolau lavas ‘‘was effi-ciently devolatilized during ancient subduction”. This evi-dence for extensive dehydration is strengthened by therecent discovery that melt inclusions in olivine phenocrystsof Loa-trend magmas have lower H2O, F and Cl contentsthan those in Kea magmas (Marske et al., 2014).

3.5. The origin of the LOA component: the role of granulites

in the lower continental crust

Our preceding discussion of incompatible element ratiosled to the conclusion that the LOA component is rich inplagioclase. Although plagioclase-rich rocks are abundantin the lower oceanic crust (e.g., Natland and Dick, 2001),they also occur in the lower crust of arcs and in the lowercontinental crust as granulites (Rudnick and Fountain,1995; Rudnick and Gao, 2014). Could the LOA componenthave originated in a continental environment? McKenzieand O’Nions (1983) proposed that the continental litho-sphere mantle could become detached and incorporatedinto the convecting mantle. Stern and Scholl (2010) sug-gested that there are three processes, subduction, subduc-tion erosion and lower crust foundering (delamination)that transfer lower continental crust into the mantle. Con-sistent with deep recycling of continental crust, there are


several examples of ocean island basalt and MORB that areinterpreted to have had a source component derived fromcontinental lithosphere (e.g., Kamenetsky et al., 2001,2009; O’Reilly et al., 2009; Regelous et al., 2009).

3.6. The Origin of the LOA Component: the role of cumulate

rocks in the lower crust of arcs?

Volcanic arcs are a tectonic setting where mafic cumu-lates in the lower crust may delaminate and founder intothe convecting mantle. Jagoutz and Schmidt (2013) esti-mated that worldwide the mass of delaminated lower crustin arcs is 1/3 to 1/2 of subducted oceanic crust. In order toachieve sufficient density for delamination, Jagoutz andSchmidt (2013) emphasized the important role of garnet-bearing cumulates. They calculated delaminate for theKohistan arc; it has high ratios of Ba/Th, La/Nb, Sr/Nd,Sm/Zr and Pb/Ce and relative depletion in Zr and Nb.These are characteristics of plagioclase/melt partition coef-ficients (Fig. 5a). However, the high abundance of heavyREE in the calculated delaminate is not consistent with pla-gioclase as the only aluminous cumulus phase. Since HREEare compatible in garnet, the cumulate in the Kohistan arcmay have contained both plagioclase and garnet.

3.7. Where did the Loa Component form? Can incompatible

element ratios constrain the tectonic setting?

Rocks that formed as plagioclase-rich adcumulatesoccur in oceanic and continental lower crust and in deepsections of volcanic arcs. Examples are:

(1) The lower oceanic crust has been sampled by drillcores into gabbroic sections exposed on the seafloor.The 1508 m section recovered by the Ocean DrillingProgram in Hole 735B near the Southwest IndianRidge has been especially useful. For example,Natland and Dick (2001) emphasized that many ofthese gabbros formed as cumulates and commonlyas adcumulates with very little trapped melt. How-ever Loa-trend shield lavas with high Sr/Nd (>19)have 143Nd/144Nd ratios less than 0.5128, distinctlylower than the mean value for normal MORB of0.5131 (Fig. 4c). Although this result seems to pre-clude a relationship to MORB we calculated a massbalance (Fig. 11a) for an average lower oceanic crust(Hart et al., 1999).

(2) The lower continental crust is sampled by xenoliths inexplosive eruptions of mafic magmas. For example atLashaine, a tuff cone in Tanzania, there are xenolithsof garnet-clinopyroxene granulites. These xenolithswere originated as plagioclase-rich cumulates in thelower continental crust (LCC) at depths within thestability field of plagioclase; subsequently garnetformed during subsolidus metamorphic reactions(Jones et al., 1983; Mansur et al., 2014). Rudnickand Gao (2014) recognized the problems with usingxenoliths to sample the lower crust and they favoreda LCC composition that is based on seismic velocitiesof lower crust coupled with ‘‘typical” compositions of

lower crust lithologies (Rudnick and Fountain, 1995).In our subsequent mass balance calculations, we useestimates of LCC composition based on granulitexenoliths and a composition based on seismic veloci-ties in the lower continental crust (Fig. 11a, b, e, h).

(3) The lower crust of volcanic arcs is sampled as xeno-liths in arc magmas. It includes 1–2 km ofhornblende- and garnet-granulite that are inferredto have formed as cumulates. Garnet was probablynot a cumulus phase and probably formed duringsubsolidus metamorphic reaction (Jagoutz andSchmidt, 2013).

We used two simple end-member models (Fig. 10) toevaluate the suitability of lower crust in the three differenttectonic areas to satisfy the mass balance equation:

Mauna Loa primary melt ¼ X% Loa componentþ ð1� X%ÞKilauea primary melt:

In this equation X is unknown; the incompatible ele-ment contents of primary melt compositions are inferredfrom Mauna Loa and Kilauea glass compositions(Appendix D) and the LOA component is the composi-tion of plagioclase-rich lower crust in the three differenttectonic settings. We chose this simple mass balancebecause Loa-trend lavas have several geochemical charac-teristics indicating that their source contained recycledcrust. For example, among Hawaiian shield lavas, theyhave: the highest d18O (Eiler et al., 1996). Koolau(Makapuu-stage) lavas also have the highest 187Os/186Os(Fig. 2 of Lassiter and Hauri, 1998) and the most anoma-lous (Sr/Nd)PM and (La/Nb)PM (1.28 and 1.38, respec-tively) in contrast to near PM ratios (1.04 and 0.89,respectively) for Kilauea lavas. Clearly the Loa-trendlavas have more unusual compositions than Kea-trendlavas. The simplest model is addition of an exotic compo-nent to Kilauea magma to make a Loa-trend magma.The goal of these calculations is to determine if theincompatible element content of plagioclase-rich gabbroicadcumulate in the lower crust in different tectonic settingsor estimates of lower crust composition can satisfy such amass balance for incompatible elements; if they do sosuch a model for Loa-trend lavas is plausible.

The first model is mixing of two melts that are createdas mantle containing peridotite and eclogite (garnetpyroxenite) ascend in an upwelling plume (Fig. 10a).The eclogite (garnet pyroxenite) melts to a large extent,>60%, before the peridotite reaches its solidus (e.g.,Hirschmann and Stolper, 1996; Pertermann andHirschmann, 2003). In this model it is assumed that themelt derived from the eclogite (garnet pyroxenite) is insu-lated enabling ascent to a shallow magma accumulationzone where it mixes with the peridotite-derived melt(Pertermann and Hirschmann, 2003; Kogiso et al.,2004; Liang et al., 2010). Fig. 11a shows that a mixedmelt consisting of 65% of Kilauea primary magma and35% of the melt derived by large extent of melting theeclogite (garnet pyroxenite) (see Appendix E for details)matches the incompatible element content of Mauna

Fig. 10. Alternatives for partial melting of an ascending diapir composed of garnet peridotite and pyroxenite (lower panels in Case A andCase B). In Case A (left side) because the pyroxenite has a lower solidus temperature than the peridotite, the pyroxenite begins to melt deeper,~150 km, than the peridotite; in middle panels melts of pyroxenite are indicated by yellow rims on pyroxenite minerals. Partial melting ofperidotite occurs at shallower depth, as indicated by red rims on peridotite minerals. At this depth the pyroxenite is melted to a high extent,>50%. In this alternative the partial melts of pyroxenite and garnet peridotite are formed at different depths and mix within crustal magmachambers. In Case B, pyroxenite begins to melt deeper, ~150 km, than the peridotite. Then such melts wet both pyroxenite minerals andperidotite minerals, as indicated by the yellow lines (melts) linking pyroxenite and peridotite in the middle panel. If melt is in equilibrium withall minerals, this forces the garnet and clinopyroxene in peridotite have the same minerals compositions as those in pyroxenite. This effect isindicated by the same colors for garnet and clinopyroxene in both pyroxenite and peridotite in the upper panel, and the same color for meltswetting both pyroxenite and peridotite minerals. At this stage, it is equivalent to partial melting of a peridotite enriched by a pyroxenite. Theeffect of this alternative can also be interpreted using a slightly modified version of the Sobolev et al. (2005) model: pyroxenite begins to meltdeeper. By the time peridotite begins to melt, pyroxenite melts to ~100%. At this stage, melt of pyroxenite reacts with peridotite to form anenriched peridotite. Then partial melts of this enriched peridotite contributes to a Hawaiian volcano. Calculation details are in Appendix J.


Loa primary magma within 10% for 17 of the 19 ele-ments; the largest discrepancy �20% is for Th and Nb(Fig. 11f; note that Panels f and j use a linear verticalaxis to compare the calculated results with the MaunaLoa composition). We view 10% as a satisfactory matchbecause our choice of end-member LOA componentsfrom diverse localities is surely not the LOA componentin Hawaiian Loa-trend basalt. Note that the altered gab-bro is not a suitable mixing component because thealtered gabbro is similar to the Mauna Loa compositionfor the elements from Eu to Lu but is very deficient Ba,Rb and Th (Fig. 11a).

In contrast our choices of incompatible element abun-dances of the lower continental crust, two granulite xeno-

liths and an estimate that used seismic velocities, did notproduce a reasonable mass balance (Fig. 11f).

A third possibility for the LOA component is delami-nates of the lower crust in arcs. Estimated compositionsof delaminates (Table 4 of Jagoutz and Schmidt, 2013)mixed with Kilauea basalt in a 35/65 proportion providesa fit to the mass balance equation comparable to that ofthe unaltered gabbro from the oceanic crust (Fig. 11d).

The second model is solid–solid mixing whereby peri-dotite and pyroxenite equilibrate. This is an end-membermelt–rock reaction model whereby total melts of recycledeclogite (garnet pyroxenite) react with olivine in the peri-dotite to form a homogenous solid (Fig. 10b) that can bemelted to different extents (Yaxley and Green, 1998;

Fig. 11. Results of the Mass Balance calculation: Mauna Loa Basalt = Kilauea basalt + LOA component. Panels a–e show results for Case Ain Fig. 10 with various choices of the LOA component. Panel f shows the fit of the model calculation to the data. The best fits result fromunaltered oceanic gabbro adcumulate or Kohistan lower crust as the LOA component. Panels g, h, i show results for Case B in Fig. 10 withvarious choices of the LOA component. Panel j shows the fit of the model calculation to the data is similar for each choice for the Loacomposition.


Appendix J; Fig 11g, h, i). This calculation involves amixing ratio for the two solids and an extent of meltingof the homogenous solid that fits most of the trace ele-ment abundances within 10%; exceptions are Th andNb which deviate by 15–25% (Fig. 11j).

In summary, plagioclase-rich cumulates transformed toeclogite (garnet pyroxenite) from all three tectonic locationscan create melts with abundances of incompatible elementsthat are similar to the LOA component. Therefore it is notpossible to use incompatible element abundance ratios toassociate the LOA component with a particular tectoniclocation.

4. USING ISOTOPIC RATIOS OF O, SR, ND, HF, PB

AND OS TO IDENTIFY THE SOURCE AND

TECTONIC ENVIRONMENT OF THE LOA

COMPONENT?

4.1. Oxygen

Eiler et al. (1996) found inter-shield differences in d18Oof olivine phenocrysts, relatively low in Kea-trend lavasand high in Loa-trend lavas. Among Hawaiian shieldbasalts the highest d18O are in Makapuu-stage lavaserupted during late growth of the Koolau shield; specifi-

Fig 11. (continued)


cally, olivine from seven samples ranges in d18O from+5.71 to +5.98 and olivine from three Lanai samplesranges from 5.39 to 5.61 (Eiler et al., 1996). In contrast,olivines in Kea-trend lavas have d18O < 5.2 (Eiler et al.,1996) and an average of historical Kilauea lavas is 5.11± 0.07 (Garcia et al., 2008). Most importantly, d18O iscorrelated with isotopic ratios such as 143Nd/144Nd and206Pb/204Pb, the major element ratio CaO/Al2O3 andincompatible element ratios such as Sr/Nb (Wang et al.,2010). The important inference is that the high d18O ofthe Makapuu lavas is a source characteristic rather thanone controlled by a process such as recent interaction ofan ascending magma with plagioclase-rich lithosphere(Saal et al., 2007; Peterson et al., 2014). The inferredd18O = +6.4 of coexisting Makapuu melt is well abovethat of MORB and similar to the d18O = +6.5 ± 1 of gab-bro in the lower km of oceanic crust core recovered atODP Hole 735B (Alt and Bach, 2006).

Granulite xenoliths from the lower continental crust alsohave high oxygen isotopic ratios. Fowler and Harmon(1990) found that 86 granulites from 7 locations have aver-age d18O = +7.5. If partial melts of granulites are mixedwith melts derived from peridotite with d18O = +5.7, agranulite component could also explain the high d18O of+6.4 in Koolau (Makapuu stage) lavas. The major conclu-sion is that oxygen isotope ratios of altered lower oceaniccrust and continental crust are similar and not useful inconstraining the source of the LOAcomponent.

4.2. Osmium

Lassiter and Hauri (1998) determined that Makapuustage lavas of the Koolau shield have high 187Os/188O. Theyconcluded that high Os ratios accompanied by high d18Oindicate that the source of Loa-trend lavas contained recy-cled oceanic crust including up to 5% pelagic sediments. Aswe noted earlier a role for only recycled normal MORB isnot consistent with the low 143Nd/144Nd that accompaniesthe characteristic high Sr/Nd of Loa-trend shield basalt(Fig. 4c). The presence of pelagic sediment in the LOAcomponent may explain why the LOA component haslow 143Nd/144Nd relative to MORB (Fig. 4c).

4.3. Lead

Previously we used the Pb isotopic ratios that are char-acteristic of the LOA component, i.e. relatively low206Pb/204Pb and high 208Pb*/206Pb*, to conclude thatancient oceanic plagioclase-rich adcumulates formed fromMORB magma could explain the Pb isotopic characteristicsof the LOA component (Fig. 6a, b).

Lustrino (2005) emphasized that lower continentalcrust incorporated into the convecting mantle also resultsin distinctive isotopic signatures such as the low206Pb/204Pb and high 187Os/188Os and unradiogenic Nd.Lower continental crust in arcs can also generate the Pbisotopic characteristic of the LOA component because


when Pb is lost from subducting slabs and incorporatedinto the source of arc lavas (Kelley et al., 2005), thedelaminates in arc settings have very low U/Pb and Th/Pb; consequently, they retain their initial Pb isotopicratios (Jagoutz and Schmidt, 2013; Huang et al., 2014).In summary the distinctive Pb isotopic characteristics ofLoa-trend lavas can develop in each of the three tectonicsettings.

4.4. Strontium and Neodymium

Hawaiian shield lavas, like most ocean island basaltdefine an inverse trend of 143Nd/144Nd vs. 87Sr/86Sr(Fig. 12a). At a given eNd, shield lavas increase in 87Sr/86Srfrom Koolau (Makapuu) to Lanai to Kahoolawe(Fig. 12a). There is no obvious correlation of 87Sr/86Sr withRb/Sr. In particular, even ancient oceanic crust of 3.5 Gawill not reach the lower range of 87Sr/86Sr in Loa-trend

-5

5

15

25

35

45

0.699 0.701 0.703 0.705

Nd

87Sr/86Sr

10

14

18

22

26

0.00

Sr/N

d

0.00

0.02

0.04

0.06

0.08

0.10

0

Rb/Sr

-2

0

2

4

6

8

10

0.7030 0.7035 0.7040 0.7045 0.7050

Nd

87Sr/86Sr

a

c

MeanMORB

Fig. 12. Panel a: eNd versus 87Sr/86Sr for Hawaiian shield lavas. As inplotted. The important result is that lavas from the three Loa-trend volcasame eNd (as indicated by the three dashed lines). Panel b shows Sr/Nd veDelta 87Sr/86Sr is the vertical distance from the regression line defined byKoolau and the three Kahoolawe samples form a positive trend conplagioclase diminishes as 87Sr/86Sr increases. However the Lanai samplesSr/Nd. The offset of Lanai data from this study (blue ‘‘ + ”, Appendix TaTable A10) also suggests that may be an inner laboratory difference (Apphave < 7% MgO. If these samples fractionated plagioclase, Sr/Nd in highshows eNd versus 87Sr/86Sr for Hawaiian shield lavas at an expanded sc0.0016–0.0021, (Hart et al., 1999) to explain the 87Sr/86Sr of Loa-trend latime. The black square) indicates present day MORB (Gale et al., 2013). Pto reach 87Sr/86Sr = 0.7045 at present day. Dashes on line indicate incremeRb/Sr (vertical axis) in an ancient gabbro that has present-day 87Sr/86Sr oaxis). Even for 4.5 Ga the required Rb/Sr in ancient gabbro is much high(For interpretation of the references to colour in this figure legend, the r

basalts (Fig. 12c, d). The simplest explanation is that inaddition to cumulus plagioclase, the Loa componentincludes a recent and variable addition of a high 87Sr/86Srcomponent such as carbonate (Huang et al., 2009, 2011b),or seawater altered lithospheric mantle (Sobolev et al.,2011). Another possibility is loss of incipient, volatile-richpartial melt with high Rb/Sr as melting in the hotspot com-menced (Roden et al., 1994).

4.5. Hafnium and Neodymium

Most oceanic island basalts have highly correlated176Hf/177Hf and 143Nd/144Nd ratios because their parent/-daughter ratios, Sm/Nd and Lu/Hf, are usually highly corre-lated. However because Lu is compatible in garnet (Fig. 5e),a petrogenetic role for garnet will decouple Hf and Nd iso-tope ratios. Jagoutz and Schmidt (2013) concluded thatgarnet-bearing delaminated lower crust in arcs would lead

Mauna KeaKilauea

Mauna Loa

KSDPMakapuu

Koolau:

LOA Trend

KEA Trend

KahoolaweLanai

005 0.00005 0.00015 0.00025 0.00035delta 87Sr/86Sr

2 4Age (Ga)

d

b

++

Figs. 4, 7 and 8 only samples with K2O/P2O5 from 1.3 to 2.2 arenoes Koolau, Lanai and Kahoolawe have different 87Sr/86Sr at thersus the offset in 87Sr/86Sr from the trend of Koolau lavas in Panel a.Koolau lavas in Panel a. The samples from the Makapuu stage of

sistent with our interpretation that the influence of the cumulusare not consistent with this trend because they have high values ofble A9) and from Gaffney et al. (2005) with green ‘‘ + ” (Appendixendix C). However most of the Lanai samples (blue ‘‘ + ” symbols)MgO Lanai lavas exceeds that of Koolau (Makapuu) lavas. Panel cale showing that the Rb/Sr ratios of oceanic gabbros are too low,vas. Green and black lines show the path of MORB backwards inanel d: Age versus Rb/Sr showing the Rb/Sr ratio and time requirednts of 0.5 Ga. This plot shows the relationship between the requiredf 0.7045 (LOA end member value) and its formation age (horizontaler than that in MORB (0.020) or oceanic gabbros (0.0016–0.0021).eader is referred to the web version of this article.)

-20-10010203040506070

-10 -6 -2 2 6 10

Hf

Nd

2

4

6

8

10

12

14

-2 0 2 4 6 8

Hf

Nd

LCC

Ancientdepleted

lithosphere

5%

15%

Loa component

a bMauna KeaKilauea

Mauna Loa

KSDPMakapuu

Koolau:

LOA Trend

KEA Trend

KahoolaweLanai

Fig. 13. Panel a: eNd versus eHf for Hawaiian shield lavas. Data are from the compilation of Huang et al. (2011a) and Appendix A.Terrestrial array is from Vervoort et al. (2011). eNd and eHf describe the 143Nd/144Nd and 176Hf/177Hf differences of a particular sample withrespect to the chondritic reservoir in parts per 10,000, with (143Nd/144Nd)CHUR = 0.512638 and (176Hf/177Hf)CHUR = 0.282785. Hawaiianshield lavas define an eNd versus eHf trend shallower than the terrestrial array. This deviation has been attributed to pelagic sediments(Blichert-Toft et al., 1999). As an alternative explanation for the offset to high eHf, Salters et al. (2011) proposed that ancient peridotiterepresented by xenoliths at Salt Lake Crater, Oahu was a component in Loa-trend shield lavas. In Panel b: peridotite xenolith (SCL326-4 witheHf �65, Bizimis et al., 2004) is mixed with lower continental crust (LCC) on the terrestrial array (granulite 89–726, Mansur et al., 2014), TheLCC percentage is shown next to the mixing line. Sample SCL326-4 is ancient recycled lithosphere assumed to have 20% clinopyroxene, andother phases that contain negligible Nd and Hf. This mixing line trends out of the terrestrial array, and the LOA component can be explainedby a 95–5 mixture of LCC and ancient depleted lithosphere.


to oceanic islandbasaltic lavas offset from themantle array tohigh 176Hf/177Hf at a given 143Nd/144Nd. Although garnet isa commonmineral in granulites anddelaminated lower crust,petrographic observations commonly conclude that garnetformed as a metamorphic mineral (e.g., Mansur et al.,2014). If garnet formed by solid state reaction in a closed sys-tem, ancient arc delaminates are unlikely to deviate from thehighly correlated 176Hf/177Hf-143Nd/144Nd terrestrial array.The Loa-trend lavas from Lanai, Kahoolawe and Koolau(Makapuu stage) deviate from the mantle array to high eHfat the low endof the array (Fig. 13).Garnetwould cause devi-ation from the array at the high end of the array (Bizimiset al., 2004). The offset of Loa-trend lavas at the low end ofthe array has been interpreted to indicate that the LOA com-ponent includes a pelagic sediment component (Blichert-Toftet al., 1999), an inference that is consistent with the low143Nd/144Nd and high d18O and 187Os/186Os of Makapuustage Koolau lavas (Lassiter and Hauri, 1998).

However, Salters et al. (2006) concluded that the incom-patible element and radiogenic isotope ratios in Loa-trendlavas cannot be explained by adding only ancient pelagicsediments to the source. The slope of the Kahoolawe, Lanaiand Koolau lavas in eHf versus eNd (Fig. 13) is the shallow-est observed at any island (Salters et al., 2006, 2011). Salterset al. (2006) proposed that the low eNd-eHf end of theHawaiian trend that deviates from the terrestrial array isbest explained if the LOA component contained ancientdepleted oceanic lithosphere, and possibly pelagic sediment(Fig. 13) in addition to plagioclase-rich adcumulates.

5. SUMMARY

5.1. Spatial distribution of LOA component

Volcanoes forming the Hawaiian Islands define two sub-parallel spatial trends known as the Loa-trend and Kea-trend. The Kea-trend volcanoes clearly end at East Molo-

kai Volcano whereas the Loa-trend extends to Koolau Vol-cano and even perhaps to Kauai (Fig. 1, Weis et al., 2011;Garcia et al., 2015). Most lavas erupted at Loa-trend volca-noes are geochemically different from those erupted at Kea-trend volcanoes. These differences reflect the geochemicalcharacteristics of the LOA component that is most abun-dant in the uppermost shield lavas of Kahoolawe, Lanaiand Koolau volcanoes. Curiously at West Molokai, locatedbetween Koolau and Lanai (Fig. 1), only 1 of the 39 shieldsamples analyzed has a Loa-trend geochemical signature,such as high Sr/Nd (Xu et al., 2007).

5.2. What is the LOA component?

It is important to recognize the heterogeneity of the LOAcomponent that created the distinctive geochemical charac-teristics of Loa-trend shield lavas. The most obvious geo-chemical signature of the LOA component is that ofancient gabbroic adcumulate rich in cumulus plagioclase;such material is recognized by high Sr/Nd, La/Nb, low143Nd/144Nd and high Sr/Nd, high 208Pb*/206Pb* and low206Pb/204Pb (Figs. 4, 6, 9 and 12b); it is most prominent inthe uppermost shield lavas of Koolau volcano. There is alsoa relatively enriched material with high 87Sr/86Sr that is notsupported by its Rb/Sr; it is most prominent in Kahoolawevolcano (Fig. 12a). A third geochemically distinct materialis inferred from the offset of Koolau, Lanai and Kahoolaweshield lavas from the eNd-eHf terrestrial array to high176Hf/177Hf at low 143Nd/144Nd (Fig. 13). This material hasbeen interpreted modified pelagic sediment and recycledancient depleted lithosphere (Salters et al., 2006).

5.3. Is the LOA component dominantly recycled oceanic

lower crust?

Ratios of Sr/Nd and La/Nb in Loa-trend lavas exceedestimates of primitive mantle and indicate that a gabbroic


adcumulate with cumulus plagioclase is a major part of theLoa-component. Such rocks are abundant in the oceaniccrust and their incompatible element content satisfies themass balance equation Mauna Loa basalt = Kilaueabasalt + LOA component.

Also the Pb isotope ratios of cumulate gabbros in theoceanic crust, i.e., low 206Pb/204Pb and high 208Pb*/206Pb*,are consistent with the LOA component as unaltered gab-bro formed from MORB as a plagioclase-rich adcumulateand aged for >3 Ga.

There are, however, geochemical characteristics of theLOA component that are not consistent with derivationfrom only igneous oceanic crust. Most obvious is theinverse trend between Sr/Nd and 143Nd/144Nd (Fig. 4c) thatshows the high Sr/Nd material has lower 143Nd/144Nd thanmost MORB. Another component such as pelagic sedimentis required to explain the low 143Nd/144Nd. Also from eNd+2 to �2, 87Sr/86Sr increases from Koolau (Makapuu) toLanai to Kahoolawe (Fig. 12a); this increase in 87Sr/86Sris not related to normal MORB because the average Rb/Sr = 0.0016 of oceanic gabbro does not lead to the high87Sr/86Sr of the Loa-trend lava (�0.7045) in 4.5 Ga(Fig. 12c,d).

Relative to the mantle array in 176Hf/177Hf versus143Nd/144Nd, Loa-trend lavas define a shallower slope thatreflects an ancient high Lu/Hf material that evolved to high176Hf/177Hf at low ratios of 143Nd/144Nd; this material wasinitially inferred to be ancient pelagic sediment consistentwith Koolau (Makapuu stage) lavas having high d18Oand 187Os/186Os and low 143Nd/144Nd. An alternative inter-pretation for this component is ancient depleted oceaniclithosphere (Fig. 13; Salters et al., 2006).

5.4. Is the LOA component recycled lower continental crust

(LCC) or as delaminated lower crust beneath volcanic arcs?

Samples of lower crust beneath continents and arcsinclude rocks that formed as adcumulates with abundantcumulus plagioclase. These rocks commonly occur as xeno-liths in explosive volcanic eruptions on continents andrarely as exhumed basement beneath volcanic arcs such asKohistan in Pakistan. The ratios of incompatible elementscan be explained by the mass balance equation MaunaLoa basalt = Kilauea basalt + LOA component. However,some choices for the composition of LCC, such as xenolithsfrom Tanzania and the estimate for LCC composition byRudnick and Fountain (1995) cannot be explained by mix-ing of two melts model presented in Fig. 10a, and evaluatedin Fig. 11f.

Relative to MORB, continental rocks have lower143Nd/144Nd and higher 87Sr/86Sr. Therefore unlike igneousoceanic lower crust they do not have a problem in explain-ing the Nd and Sr ratios in the LOA component (Fig. 12).Also the Pb isotopic ratios estimated for LCC can explainthe trend of Loa-trend lavas to low 206Pb/204Pb and high208Pb*/206Pb* (Fig. 6). However, as with igneous oceaniclower crust, a process is needed to explain that Kahoolaweshield lavas have high 87Sr/86Sr at a given eNd (Fig. 12a).

5.5. Important role of cumulus plagioclase in creating the

LOA component

Although we cannot make a compelling argument for aspecific tectonic setting for origin of the LOA component,we conclude that regardless of tectonic setting there is acommonality in the sequence of processes that created thegeochemical signature of cumulus plagioclase in the LOAcomponent. They are: (a) ancient formation in the lowercrust of gabbroic adcumulates with abundant cumulus pla-gioclase, that is, no melt was trapped; (b) recrystallizationto eclogite-facies mineralogy in a closed system, i.e., therewas no compositional change accompanying metamor-phism; (c) incorporation into the convecting mantle via sub-duction, foundering or delamination; (d) addition of acomponent with a long-term high Lu/Hf ratio, such as resi-dues from ancient melting processes; or pelagic sediment (e)during ascent of the buoyant Hawaiian plume melts formedby low extents of partial melting of peridotite and highextents of melting of pyroxenite/eclogite are mixed(Fig. 10a). Alternatively, the mineralogically heterogeneousperidotite and pyroxenite are homogenized, by melt-rockreaction, described as solid–solid mixing. This solid par-tially melts as the buoyant Hawaiian hotspot ascends(Fig. 10b).

Interestingly, Loa-trend Hawaiian volcanoes are not theonly locality where cumulate gabbros rich in cumulus pla-gioclase are a source component for basalt. Adcumulaterocks have also been proposed as a source for depleted Ice-landic basalt (Chauvel and Hemond, 2000), for lavaserupted at the easternmost volcanoes in the GalapagosArchipelago (Saal et al., 2007; Peterson et al., 2014) andfor basalt erupted in Sardinia derived from plagioclase-rich cumulates formed in oceanic plateaus (Gasperiniet al., 2000) or recycled lower continental crust (Lustrinoet al., 2000).

ACKNOWLEDGMENTS

We thank A. Saal, A. Hofmann, K. Purtirka, R. Rudnick andespecially J. Natland for discussion and the geochemists who pub-lished the data that we use in this paper, M. Norman for providingthe data for 21 Mauna Loa basalts from Jason Cruise and PiscesDive 184 in TablA5, and A. Hofmann, W. Abouchami and J.Blichert-Toft for isotopic data for Lanai basalts in Table A9 ofAppendix. J. Bedard provided the calculated partition coefficientsin Appendix G. The reviewers (anonymous and A. Pietruszka)and Guest Editor M. Garcia made numerous comments thatimproved the paper. GX thanks D. McCormick for useful com-ments. Fred thanks M. Garcia for insisting that I present a talkat the 2014 annual meeting that led to this GCA issue. SH acknowl-edges support from NSF EAR-1524387.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.gca.2016.04.010.




REFERENCES

Abouchami W., Hofmann A. W., Galer S. J. G. and Frey F. A.(2005) Lead isotopes reveal bilateral asymmetry and verticalcontinuity in the Hawaiian mantle plume. Nature 434, 851–856.

Alt J. C. and Bach W. (2006) Oxygen isotope composition of asection of lower oceanic crust, ODP Hole 735B. Geochem.

Geophys. Geosyst. 7. http://dx.doi.org/10.1029/2006GC001385.Appleman D. E. and Dana J. D. (1987) The origins of Hawaiian

Volcanology: The U.S. exploring expedition in Hawaii, 1840-41. In Volcanism in Hawaii, Chapter 60. Department of MineralSciences, Smithsonian Institution, Washington, D.C..

Bedard J. H. (2005) Partitioning coefficients between olivine andsilicate melts. Lithos 83, 394–419.

Bedard J. H. (2006) Trace element partitioning in plagioclasefeldspar. Geochim. Cosmochim. Acta 70, 3717–3742.

Bedard J. H. (2007) Trace element partitioning coefficients betweensilicate melts and orthopyroxene: Parameterizations of Dvariations. Chem. Geol. 244, 263–303.

Bindeman I. N. and Davis A. M. (2000) Trace element partitioningbetween plagioclase and melt: Investigation of dopant influenceonpartition behavior.Geochim.Cosmochim.Acta 64, 2863–2878.

Bindeman I. N., Davis A. M. and Drake M. J. (1998) Ionmicroprobe study of plagioclase-basalt partition experiments atnatural concentration levels of trace elements. Geochim. Cos-

mochim. Acta 62(7), 1175–1193.Bizimis M., Sen G. and Salters V. J. M. (2004) Hf-Nd isotope

decoupling in the oceanic lithosphere: Constraints from spinelperidotites from Oahu, Hawaii. Earth Planet. Sci. Lett. 217, 43–58.

Blichert-Toft J., Frey F. A. and Albarede F. (1999) Hf isotopeevidence for pelagic sediments in the source of Hawaiianbasalts. Science 285, 879–882.

Blundy J. and Wood B. (2003) Mineral-melt Partitioning ofUranium, Thorium and Their Daughters. In Chapter 3 Reviews

in Mineralogy & Geochemistry, 52 (eds. B. Bourdon, G. T. M.Henderson, C. Lundstrom and S. P. Turner).

Chauvel C. and Hemond C. (2000) Melting of a complete section ofrecycled oceanic crust: Trace element and Pb isotopic evidencefrom Iceland. Geochem. Geophys. Geosyst.. http://dx.doi.org/10.1029/1999GC000002.

Chen C.-Y., Frey F. A., Rhodes J. M. and Easton R. M. (1996)Temporal geochemical evolution of Kilauea volcano: Compar-ison of Hilina and Puna basalt. In Earth Processes: Reading the

Isotopic Code, 95 (eds. A. Basu and S. Hart), pp. 161–181.Geophys. Mono.

Clague D. A., Frey F. A., Garcia M. O., Huang S., McWilliams M.and Beeson M. H. (2016) Compositional heterogeneity of theSugarloaf Melilite Nephelinite Flow, Honolulu Volcanics,Hawaii. Geochim. Cosmochim. Acta 185, 251–277, Special Issueof Geochimica et Cosmochimica Acta on ‘‘Magmas and theirsources: A tribute to Fred Frey”.

Coogan L. A., MacLeod C. J., Dick H. J. B., Edwards S. J.,Kvassnes A., Natland J. H., Robinson P. T., Thompson G. andO’Hara M. J. (2001) Whole-rock geochemistry of gabbros fromthe Southwest Indian Ridge: constraints on geochemicalfractionations between the upper and lower oceanic crust andmagma chamber processes at (very) slow-spreading ridges.Chem. Geol. 178, 1–22.

Dana J. D. (1849), p. 756. Geology United States ExploringExpedition Philadelphia. C. Sherman.

Danyushevsky L. V., Perfit M. R., Eggins S. and Falloon T. J.(2003) Crustal origin for coupled ‘ultra-depleted’ and ‘plagio-clase’ signatures in MORB olivine-hosted melt inclusions:evidence from the Siqueiros Transform Fault, East PacificRise. Contrib. Mineral. Petrol. 144, 619–637.

Drake M. J. and Weill D. F. (1975) Partition of Sr, Ba, Ca, Y,Eu2+, Eu3+, and other REE between plagioclase feldspar andmagmatic liquid: an experimental study. Geochim. Cosmochim.

Acta 39, 689–712.Drouin M., Godard M., Ildefonse B., Bruguier O. and Garrido C.

J. (2009) Geochemical and petrographic evidence for magmaticimpregnation in the oceanic lithosphere at Atlantis Massif,Mid-Atlantic Ridge (IODP Hole U1309D, 30�N). Chem. Geol.

264, 71–88.Eiler J. M., Farley K. A., Valley J. W., Hauri E., Craig H., Hart S.

and Stolper E. M. (1996) Oxygen isotope variations in oceanisland basalt phenocrysts. Geochim. Cosmochim. Acta 61, 2281–2293.

Eisele J., Abouchami W., Galer S. J. G. and Hofmann A. W. (2003)The 320 kyr Pb isotope evolution of Mauna Kea lavas recordedin the HSDP-2 drill core. Geochem. Geophys. Geosyst.. http://dx.doi.org/10.1029/2002GC000339.

Elthon D., Stewart M. and Kent Ross. D. (1992) Compositionaltrends of minerals in oceanic cumulates. J. Geophys. Res. 97,15189–15199.

Fekiacova Z., Abouchami W., Galer S. J. G., Garcia M. O. andHofmann A. W. (2007) Origin and temporal evolution ofKo’olau Volcano, Hawai’i: Inferences from isotope data on theKo’olau Scientific Drilling Project (KSDP), the HonoluluVolcanics and ODP Site 843. Earth Planet. Sci. Lett. 261, 65–83.

Fowler M. B. and Harmon R. S. (1990) The oxygen isotopecomposition of lower crustal granulite xenoliths. In Granulites

and Crustal Evolution (eds. D. Vielzeuf and Ph. Vidal). KluwerAcademic Publishers, pp. 493–506.

Frey F. A. and Rhodes J. M. (1993) Inter-shield geochemicaldifferences among Hawaiian volcanoes: implications for sourcecompositions, melting process and magma ascent paths. Philos.Trans. R. Soc. London A 342, 121–136.

Frey F. A., Garcia M. O. and Roden M. F. (1994) Geochemicalcharacteristics of Koolau volcano: implications of intershieldgeochemical differences among Hawaiian volcanoes. Geochim.

Cosmochim. Acta 518, 1441–1462.Gaffney A. M., Nelson B. K. and Blichert-Toft J. (2005) Melting in

the Hawaiian plume at 1–2 Ma as recorded at Maui Nui: Therole of eclogite, peridotite, and source mixing. Geochem.

Geophys. Geosyst 6, 10.Gale A., Dalton C. A., Langmuir C. H., Su Y. and Schilling J. G.

(2013) The mean composition of ocean ridge basalts. Geochem.

Geophys. Geosyst, 14.Garcia M. O., Ito E. and Eiler J. (2008) Oxygen isotope evidence

for chemical interaction of kilauea historical magmas withbasement rocks. J. Petrology 49(4), 757–769.

Garcia M. O., Smith J. R., Tree J. P., Weis D., Harrison L. andJicha B. R. (2015) Petrology, geochemistry, and ages of lavasfrom Northwest Hawaiian Ridge volcanoes. In The Origin,

Evolution and Environmental Impact of Oceanic Large Igneous

Provinces (eds. C. R. Neal, W. W. Sager, T. Sano and E. Erba).Geological Society of America Special Paper 511.

Gasperini D., Blichert-Toft J., Bosch D., Del Moro A., Macera P.,Telouk P. and Albarede F. (2000) Evidence from Sardinianbasalt geochemistry for recycling of plume heads into theEarth’s mantle. Nature 408, 701–704.

Green T. H. and Ringwood A. E. (1968) Genesis of the calcalkalineigneous rock suite. Contrib. Mineral. Petr. 18, 105–162.

Harrison L., Weis D., Hanano D. and Barnes E. (2015) Lithiumisotopic signature of Hawaiian basalts. Hawaii an volcanoes,

From Source to Surface, Geophysical Monograph 208, first ed,pp. 79–104.

Hart S. R., Blusztajn J., Dick H. J. B., Meyer P. S. andMuehlenbachs K. (1999) The fingerprint of seawater circulation

http://refhub.elsevier.com/S0016-7037(16)30169-7/h0005



http://dx.doi.org/10.1029/2006GC001385






























http://dx.doi.org/10.1029/1999GC000002

http://dx.doi.org/10.1029/1999GC000002










































http://dx.doi.org/10.1029/2002GC000339

http://dx.doi.org/10.1029/2002GC000339



















































I a 500-eter section of ocean crust gabbros. Geochim. Cos-

mochim. Acta 63(23/24), 4059–4080.Haskins E. H. and Garcia M. O. (2004) Scientific drilling reveals

geochemical heterogeneity within the Ko’olau shield, Hawai’i.Contrib. Mineral. Petrol. 147, 162–188.

Hauri E. H. (1996) Major-element variability in the Hawaiianmantle plume. Nature 382, 415–419.

Hauri E. H. (2002) SIMS investigations of volatiles in silicateglasses 2: abundances and isotopes in Hawaiian melt inclusions.Chem. Geol. 183, 115–141.

Hauri E. H., Lassiter J. C. and DePaolo D. J. (1996) Osmiumisotope systematics of drilled lavas fromMauna Loa, Hawaii. J.Geophys. Res. 101(B5), 11793–11806.

Hemond C., Hofmann A. W., Heusser G., Condomines M.,Raczek I. and Rhodes J. M. (1994) U-Th-Ra systematics inKilauea and Mauna Loa basalts, Hawaii. Chem. Geol. 116,163–180.

Herzberg C. (2006) Petrology and thermal structure of theHawaiian plume from Mauna Kea volcano. Nature 444, 605–609.

Herzberg C., Cabral R. A., Jackson M. G., Vidito C., Day J. M. D.and Hauri E. H. (2014) Phantom Archean crust in Mangaihotspot lavas and the meaning of heterogeneous mantle. EarthPlanet. Sci. Lett. 396, 97–106.

Hieronymus C. F. and Bercovici D. (1999) Discrete alternatinghotspot islands formed by interaction of magma transport andlithospheric flexure. Nature 397, 604–607.

Hirschmann M. M. and Stolper E. M. (1996) A possible role forgarnet pyroxenite in the origin of the ‘‘garnet signature” inMORB. Contrib. Mineral. Petrol. 124, 185–208.

Hofmann A. W. and Jochum K. P. (1996) Source characteristicsderived from very incompatible trace elements in Mauna Loaand Mauna Kea basalts. Hawaii Scientific Drilling Project. J.Geophys. Res. Solid Earth 101, 11831–11839.

Huang S. and Frey F. A. (2003) Trace element abundances ofMauna Kea basalt from phase 2 of the Hawaiian ScientificDrilling Project: petrogenetic implications of correlations withmajor element content and isotopic ratios. Geochem. Geophys.

Geosyst. 4(6), 8711, 10;1029/2004GC000757.Huang S. and Frey F. A. (2005) Recycled oceanic crust in the

Hawaiian Plume: evidence from temporal geochemical varia-tions within the Koolau Shield. Contrib Mineral. Petrol. 149,3556–3575.

Huang S. and Humayun M. (2016) Petrogenesis of High-CaO lavasfrom Mauna Kea, Hawaii: Constraints from trace elementabundances. Geochim. Cosmochim. Acta.

Huang S., Abouchami W., Blichert-Toft J., Clague D. A., CousensB. L., Frey F. A. and Humayun M. (2009) Ancient carbonatesedimentary signature in the Hawaiian plume: Evidence fromMahukona volcano Hawaii. Geochem. Geophys. Geosyst 10,Q08002.

Huang S., Hall P. S. and Jackson M. G. (2011a) Geochemicalzoning of volcanic chains associated with Pacific hotspots. Nat.

Geosci. 4, 874–878. http://dx.doi.org/10.1038/NGEO1263.Huang S., Farkas J. and Jacobsen S. B. (2011b) Stable calcium

isotopic compositions of Hawaiian shield lavas: Evidence forrecycling of ancient marine carbonates into the mantle.Geochim. Cosmochim. Acta. 75, 4987–4997. http://dx.doi.org/10.1016/j.gca.2011.06.010.

Huang S., Lee C.-T. A. and Yin Q.-Z. (2014) Missing lead and high3He/4He in ancient sulfides associated with continental crustformation. Sci. Rep. 4, 5314. http://dx.doi.org/10.1038/srep05314.

Huang S., Vollinger M. J., Frey F. A., Rhodes M. J. and Zhang Q.(2016) Compositional variation within thick (>10 m) flow units

of Mauna Kea Volcano Cored by the Hawaiian ScientificDrilling Project. Geochim. Cosmochim. Acta 185, 182–197.

Jackson M. C., Frey F. A., Garcia M. O. and Wilmoth R. A.(1999) Geology and geochemistry of basaltic lava flows anddikes from the Trans-Koolau tunnel, Oahu, Hawaii. Bull.

Volcanol. 60, 381–401.Jackson M. G., Weis D. and Huang S. (2012) Major element

variations in Hawaiian shield lavas: Source features andperspectives from global ocean island basalt (OIB) systematics.Geochem. Geophys. Geosyst. 13. http://dx.doi.org/10.1029/2012GC004268.

Jagoutz O. and Schmidt M. W. (2013) The composition of thefoundered complement to the continental crust and a re-evaluation of fluxes in arcs. Earth Planet. Sci. Lett. 371–372,177–190.

Jochum K. P. et al. (2006) MPI-DING reference glasses for in situmicroanalysis: new reference values for element concentrationsand isotopic ratios. Geochem. Geophys. Geosyst. 7, 11.1029/2005GC001060, Q02008.

Johnson K. T. M., Dick H. J. B. and Shimizu N. (1990) Melting inthe oceanic upper mantle; an ion microprobe study of diopsidesin abyssal peridotites. J. Geophys. Res. 95, 2661–2678.

Jones A. P., Smith J. V., Dawson J. B. and Hansen E. C. (1983)Metamorphism, partial melting, and K-Metasomatism ofgarnet-scapolite-kyanite granulite xenoliths from Lashane,Tanzania. J. Geol. 91, 143–166.

Kamenetsky V. S., Maas R., Sushchevskaya N. M., Norman M.D., Cartwright I. and Peyve A. A. (2001) Remnants ofGondwanan continental lithosphere in oceanic upper mantle:Evidence from the South Atlantic Ridge. Geology 3, 243–246.

Kamenetsky V. S., Kamenetsky M. B., Sobolev A. V., Golovin A.V., Sharygin V. V., Pokhilenko N. P. and Sobolev N. V. (2009)Can pyroxenes be liquidus minerals in the kimberlite magma?Lithos 112, 213–222.

Kelley K. A., Plant T., Farr L., Ludden J. and Staudigel H. (2005)Subduction cycling of U, Th, and Pb. Earth Planet. Sci. Lett.

234, 369–383.Kogiso T., Hirschmann M. M. and Reiners P. W. (2004) Length

scales of mantle heterogeneities and their relationship to oceanisland basalt geochemistry. Geochim. Cosmochim. Acta 68(2),345–360.

Kurz M. D., Kenna T. C., Lassiter J. C. and DePaolo D. J. (1996)Helium isotopic evolution of Mauna Kea volcano: First resultsfrom the 1-km drill core. J. Geophys. Res. 101(B5), 11781–11791.

Lassiter J. C., DePaolo D. J. and Tatsumoto M. (1996) Isotopicevolution of Mauna Kea volcano: Results from the initial phaseof the Hawaii Scientific Drilling Project. J. Geophys. Res. 101,11769–11780.

Lassiter J. C. and Hauri E. H. (1998) Osmium-isotope variations inHawaiian lavas: evidence for recycled oceanic lithosphere in theHawaiian plume. Earth Planet. Sci. Lett. 164, 483–496.

Leeman W. P. (1979) Partitioning of Pb between volcanic glass andcoexisting sanidine and plagioclase feldspars. Geochim. Cos-

mochim. Acta 41, 171–175.Liang Y., Schiemenz A., Hesse M. A., Parmentier E. M. and

Hesthaven J. S. (2010) High-porosity channels for melt migra-tion in the mantle: Top is the dunite and bottom is theharzburgite and lherzolite. Geophys. Res. Lett. 37, L15306.

Lustrino M. (2005) How the delaminating and detachment of lowercrust can influence basaltic magmatism. Earth Sci. Rev. 72, 21–38.

Lustrino M., Melluso L. and Morro V. (2000) The role of lowercontinental crust and lithospheric mantle in the genesis of Plio-Pleistocene volcanic rocks from Sardinia (Italy). Earth Planet.

Sci. Lett. 180, 259–270.





















































http://dx.doi.org/10.1038/NGEO1263



http://dx.doi.org/10.1038/srep05314

http://dx.doi.org/10.1038/srep05314









http://dx.doi.org/10.1029/2012GC004268

http://dx.doi.org/10.1029/2012GC004268
























































Mansur A. T., Manya S., Timpa S. and Rudnick R. L. (2014)Granulite-facies xenoliths in rift basalts of northern Tanzania:Age, composition and origin of Archean lower crust. J. Petrol.55(7), 1243–1286.

Marske J., Pietruszka A. J., Weis D., Garcia M. O. and Rhodes J.M. (2007) Rapid passage of a small-scale mantle heterogeneitythrough the melting regions of Kilauea and Mauna Loavolcanoes. Earth Planet. Sci. Lett. 259, 34–50.

Marske J., Garcia M. O., Pietruszka A. J., Rhodes J. M. andNorman M. D. (2008) Geochemical variations during Kilauea’sPu’u ‘O’o eruption reveal a fine-scale mixture of mantleheterogeneities within the Hawaiian Plume. J. Petrol. 49,1297–1318.

Marske J., Hauri E., Trusdell F., Garcia M. O. and Pietruszka A.(2014) From Purgatory to Paradise: The volatile life of Hawaiian

Magmas Abstract V33C–4893 presented at 2014 Fall MeetingAGU, San Francisco, CA, 15–19, December.

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

McKenzie D. and O’Nions R. K. (1983) Mantle reservoirs andocean island basalts. Nature 301, 229–231.

Meyer P. S., Dick H. J. B. and Thompson G. (1989) Cumulategabbros from the Southwest Indian Ridge, 54�S-7�16’E: impli-cations for magmatic processes at a slow spreading ridge.Contrib. Mineral. Petrol. 103, 44–63.

Natland J. H. and Dick H. J. B. (2001) Formation of the lowerocean crust and the crystallization of gabbroic cumulates at avery slowly spreading ridge. J. Volcanol. and Geotherm. Res. 11,191–233.

Norman M., Garcia M. O. and Pietruszka A. J. (2005) Trace-element distribution coefficients for pyroxenes, plagioclase, andolivine in evolved tholeiites from the 1955 eruption of KilaueaVolcano, Hawaii, and petogenesis of differentiated rift-zonelavas. Am. Mineral. 90, 888–899.

O’Reilly S. Y., Zhang M., Griffin W. L., Begg G. and Hronsky J.(2009) Ultra-deep continental roots and their oceanic remnants:A solution to the geochemical ‘‘mantle reservoir” problem?Lithos 2115, 1043–1054.

Pertermann M. and Hirschmann M. M. (2003) Partial meltingexperiments on a MORB-like pyroxenite between 2 and 3 GPa:Constraints on the presence of pyroxenite in basalt sourceregions from solidus location and melting rate. J. Geophys. Res.108(B2), 2125. http://dx.doi.org/10.1029/2000JB000118.

Pertermann M., Hirschmann M. M., Hametner K., Gunter D. andSchmidt M. (2004) Experimental determination of traceelement partitioning between garnet and silica-rich liquidduring anhydrous partial melting of MORB-like eclogite.Geochem. Geophys. Geosyst. 5, 5, 1029/2003GC000638.

Peterson M. E., Saal A. E., Nakamura E., Kitagawa H., Kurz M.and Koleszar A. M. (2014) Origin of the ‘Ghost Plagiocalse’signature in Galapagos melt inclusions: New evidence from Pbisotopes. J. Petrol. 55(11), 2193–2216.

Pietruszka A. J., Norman M. D., Garcia M. O., Marske J. P. andBurns D. H. (2013) Chemical heterogeneity in the Hawaiianmantle plume from the alteration and dehydration of recycledoceanic crust. Earth Planet. Sci. Lett. 361, 298–309.

Powers H. A. (1955) Composition and origin of basaltic magmaof the Hawaiian Island. Cosmochim. Geochimica Acta 7, 77–107.

Regelous M., Niu Y., Abouchami W. and Castillo P. R. (2009)Shallow origin for South Atlantic Dupal Anomaly from lowercontinental crust: Geochemical evidence from the Mid-AtlanticRidge at 26�S. Lithos 112, 57–72.

Ren Z.-Y., Ingle S., Takahashi E., Hirano N. and Hirata T. (2005)The chemical structure of the Hawaiian mantle plume. Nature

436, 837–840.

Rhodes J. M. and Vollinger M. (2004) Composition of basalticlavas sampled by phase-2 of the Hawaii Scientific DrillingProject: Geochemical stratigraphy and magma types. Geochem.

Geophys. Geosyst. 5. http://dx.doi.org/10.1029/2002GC000434,Q03G13.

Rhodes J. M., Huang S., Frey F. A., Pringle M. and Xu G. (2012)Compositional diversity of Mauna Kea shield lavas recoveredby the Hawaii Scientific Drilling Project: Inferences on sourcelithology, magma supply, and the role of multiple volcanoes.Geochem. Geophys. Geosyst.. http://dx.doi.org/10.1029/2011GC003812.

Roden M. F., Trull T., Hart S. R. and Frey F. A. (1994) New He,Nd, Pb, and Sr isotopic constraints on the constitution of theHawaiian plume – results from Koolau Volcano. Oahu,Hawaii, USA. Geochim. Cosmochim. Acta 58, 1431–1440.

Rudnick R. L. and Fountain D. M. (1995) Nature and compositionof the continental crust: A lower crustal perspective. Rev.

Geophys. 33(3), 267–309.Rudnick R. and Gao S. (2014) Composition of the continental

crust. Treat. Geochem. 3, 1–51.Saal A. E., Kurz M. D., Hart S. R., Blusztajn J. S., Blichert-Toft J.,

Liang Y. and Geist D. J. (2007) The role of lithospheric gabbroson the composition of Galapagos lavas. Earth Planet. Sci. Lett.

257, 391–406.Salters V. J. M., Blichert-Toft J., Fekiacova Z., Sachi-Kocher A.

and Bizimis M. (2006) Isotope and trace element evidence fordepleted lithosphere in the source of enriched Koolau basalts.Contrib Mineral. Petrol. 151(3), 297–312.

Salters V. J. M., Mallick S., Hart S. R., Langmuir C. E. and StrackeA. (2011) Domains of depleted mantle: New evidence fromHafnium andNeodymium isotopes.Geochem. Geophys. Geosyst.

12(8), Q08001. http://dx.doi.org/10.1029/2011GC003617.Sobolev A. V., Hofmann A. W. and Nikogosian J. K. (2000)

Recycled oceanic crust observed in ‘ghost plagioclase’ withinthe source of Mauna Loa lavas. Nature 404, 986–990.

Sobolev A. V., Hofmann A. W., Sobolev S. V. and Nikogosian I.K. (2005) An olivine-free mantle source of Hawaiian shieldbasalts. Nature 434, 590–597.

Sobolev A. V., Hofmann A. W., Jochum K. P., Kuzmin D. V. andStoll B. (2011) A young source for the Hawaiian plume. Nature

476, 434–437.Stern R. J. and Scholl D. W. (2010) Yin and Yang of continental

crust creation and destruction by plate tectonic processes. Int.Geol. Rev. 52(1), 1–31.

Tatsumoto M. (1978) Isotopic composition of lead in oceanicbasalt and its implication to mantle evolution. Earth Planet.

Sci. Lett. 38, 63–87.Tepley F. J., Lundstrom C. C., McDonough W. F. and Thompson

A. (2010) Trace element partitioning between high-An plagio-clase and basaltic to basaltic andesite melt at 1 atmospherepressure. Lithos 118, 82–94.

Vervoort J. D., Plank T. and Prytulak J. (2011) The Hf-Nd isotopiccomposition of marine sediments. Geochim. Cosmochim. Acta

75, 5903–5926.Wang Z., Eiler J. M., Asimow P. D., Garcia M. O. and Takahashi

E. (2010) Oxygen isotope constraints on the structure andevolution of the Hawaiian plume. Am. J. Sci. 310, 683–720.

Weis D., Garica M. O., Rhodes J. M., Jellinek M. and Scoates J. S.(2011) Role of the deep mantle in generating the compositionalasymmetryof theHawaiianmantle plume.Nat.Geosci.4, 831–838.

Wright T. L. (1971) Chemistry of Kilauea and Mauna Loa lava inspace and time. In Geological Survey Professional Paper, 735,

Washington, D.C., p. 40.Xu G., Frey F. A., Clague D. A., Abouchami W., Blichert-Toft J.,

Cousens B. and Weisler M. (2007) Geochemical characteristicsof West Molokai shield- and postshield-stage lavas: Constraints








































http://dx.doi.org/10.1029/2000JB000118
























http://dx.doi.org/10.1029/2002GC000434

http://dx.doi.org/10.1029/2011GC003812

http://dx.doi.org/10.1029/2011GC003812


















http://dx.doi.org/10.1029/2011GC003617

































on Hawaiian plume models. Geochem. Geophys. Geosyst. 8.http://dx.doi.org/10.1029/2006GC001554.

Xu G., Huang S., Frey F. A., Blichert-Toft J., Abouchami W.,Clague D. A., Cousens B., Moore J. and Beeson M. H. (2014)The distribution of geochemical heterogeneities in the source ofHawaiian shield lavas as revealed by a transect across the strikeof the Loa and Kea spatial trends: East Molokai to WestMolokai to Penguin Bank. Geochim. Cosmochim. Acta 132,214–237.

Yang H.-J., Frey F. A. and Clague D. A. (2003) Constraints on theSource Components of Lavas Forming the Hawaiian NorthArch and Honolulu Volcanics. J. Petrol. 44(4), 603–627.

Yaxley G. M. and Green D. H. (1998) Reactions between eclogiteand peridotite: mantle refertilisation by subduction of oceaniccrust. Schweiz Mineral. Petogr. MITT 78, 243–255.

Zartman R. E. and Haines S. M. (1988) The plumbotectonic modelfor Pb isotopic systematics among major terrestrial reservoirs –A case for bi-directional transport. Geochim. Cosmochim. Acta

52, 1327–1339.

Associate editor: Michael Garcia

http://dx.doi.org/10.1029/2006GC001554


















The geochemical components that distinguish Loa- and Kea ...

Documents

Transcript of The geochemical components that distinguish Loa- and Kea ...