Mineral Constraints on the Source Lithologies at Fogo, Cape Verde

Examensarbete vid Institutionen för geovetenskaper

Degree Project at the Department of Earth Sciences

ISSN 1650-6553 Nr 433

Mineral Constraints on the Source Lithologies at Fogo, Cape Verde

Geokemiska ledtrådar till de aktiva mantel- komponenterna på Fogo, Kap Verde

Elin Rydeblad

INSTITUTIONEN FÖR GEOVETENSKAPER

D E P A R T M E N T O F E A R T H S C I E N C E S

Examensarbete vid Institutionen för geovetenskaper

Degree Project at the Department of Earth Sciences

ISSN 1650-6553 Nr 433

Mineral Constraints on the Source Lithologies at Fogo, Cape Verde

Geokemiska ledtrådar till de aktiva mantel- komponenterna på Fogo, Kap Verde

Elin Rydeblad

ISSN 1650–6553

Copyright © Elin Rydeblad

Published at Department of Earth Sciences, Uppsala University (www.geo.uu.se), Uppsala, 2018

Abstract

Mineral Constraints on the Source Lithologies at Fogo, Cape Verde Elin Rydeblad

Variations in major, minor, and trace elements compositions and ratios, as well as isotope ratios are all useful tools in studying the composition of the Earth’s mantle, and heterogeneities present therein. Since the mantle itself doesn’t easily lend itself to study, ocean island basalt (OIBs) are commonly used as a proxy due to compositional differences combined with the range of origination depth, a combination that allows them to represent the heterogeneity of the mantle, sampling everything from the core mantle boundary to the old or recent additions of recycled oceanic crust. Fogo, being one of the most active volcanoes in the world, continuously samples the interior of our planet, and as such is a prime location for studies of mantle geochemistry.

This study aims to determine the origin of the mantle lithologies present at Fogo. The study is a continuation and extension of the studies conducted by Barker et al. (2014) and Magnusson (2016).

This study utilises major, minor, and trace element geochemistry in clinopyroxene and olivine phenocrysts, as well as Ni-isotopes from whole rock samples. Using the relative values of Ni, Mn, and trace elements and their ratios in olivine and clinopyroxene phenocrysts we aim to further unravel the mechanics of the creation of ocean islands and provide additional constraints regarding the mechanics of the formation of heterogeneities in the Earth’s mantle. This study will focus on Ni* and Mn* in olivine phenocrysts, trace element composition and ratios of olivine phenocrysts and clinopyroxene phenocrysts, and Ni-isotope data.

This study found evidence for both pyroxenite, carbonatite, and carbonated eclogite source lithologies at Fogo. A correlation between La/Sm and δ

⁶⁰

Ni was also found, indicating a control on the δ

⁶⁰

Ni by source pyroxenite.

This study suggests a carbonated eclogite origin for the lithologies present at Fogo, which would have hosted the majority of the olivine phenocrysts. The phenocrysts then resided within a separated carbonatite melt fraction that either contaminated or metasomatized a pyroxenite melt where the clinopyroxene phenocrysts nucleated. The melt then evolved to an alkali basalt melt through melt-rock reactions, principally via the dissolution of orthopyroxenes and concomitant precipitation of clinopyroxene and olivine (Zhang, Chen, Jackson & Hofmann 2017).

Keywords: olivine, clinopyroxene, carbonatite, Cape Verde, Eclogite, source lithologies

Degree Project E in Earth Science, 1GV085, 45 credits Supervisor: Abigail Barker

Department of Earth Sciences, Uppsala University, Villavägen 16, SE-752 36 Uppsala (www.geo.uu.se) ISSN 1650–6553, Examensarbete vid Institutionen för geovetenskaper, No. 433, 2018

The whole document is available at www.diva-portal.org

Populärvetenskaplig sammanfattning

Geokemiska ledtrådar till de aktiva mantelkomponenterna på Fogo, Kap Verde Elin Rydeblad

Sammansättningen av jordens inre är någonting som länge har fascinerat geologer, geokemister, och geofysiker, och många vetenskapliga undersökningar har utförts för att försöka utröna hur våran planet är uppbyggd och hur den fungerar. Undersökningar av mantelns kemiska sammansättning och de olika mantelkomponenterna utförs ofta vid vulkaniskt aktiva öar, då de olika typer av basalt som finns på dessa öar ofta behåller ett tydligt avtryck av dess ursprungliga mantelkomponent. Fogo, en av öarna i den sydvästra delen av ögruppen Kap Verdes skärgård, är en av världens i dagsläget mest aktiva vulkaner och därmed en optimal utgångspunkt för att studera jordens inre.

Den här studien ämnar utröna vilka mantelkomponenter som är närvarande på Fogo. Studien är en vidareutveckling av två tidigare studier, en som studerade mantelkomponenter på närliggande öar och en som studerade förändringen i spårämnen och syreisotoper över tid i samma prover som används i denna studie.

I den här studien används förekomsten av grundämnen, spårämnen, och sällsynta jordartsmetaller i silikatmineralerna olivin och pyroxen för att ta reda på de bidragande mantelkomponenterna. Skillnaden mellan två olika nickel isotoper,

⁵⁸

Ni och

⁶⁰

Ni (uttryckt som δ

⁶⁰

Ni), kommer också att undersökas för att utöka insikten om detta nya tämligen outforskade isotopiska system, och koppla samman dess isotopiska sammansättning med de analyserade syreisotoperna från tidigare studier, och om möjligt även mineralsammansättning.

I denna studie har kan vi påvisa att pyroxenit, karbonatit, och karbonatrik eklogit alla har varit inblandade i bildandet av Fogo. Bevis fanns även för att mängden pyroxenit i en smälta kontrollerar sammansättningen av δ

⁶⁰

Ni.

Den här studien föreslår, baserat på mineralsammansättningen av olivin och pyroxen, karbonat-och granat-rik källa som ursprunglig smälta. Det var i denna smälta som olivinkristallerna växte. Dessa kristaller följde sedan med en karbonatitsmälta när den bröt sig fri. Denna karbonatitsmälta har sedan antingen kemiskt förändrat, eller blandat sig med, den pyroxenitsmälta som pyroxenkristallerna har växt i. Denna blandade smälta har slutgiltige utvecklats till en alkalisk basaltisk smälta genom interaktioner mellan smältan och de omkringliggande mantelkomponenterna.

Nyckelord: olivin, pyroxen, karbonatit, Kap Verde, eklogit, mantelkomponenter

Examensarbete E i geovetenskap, 1GV085, 45 hp Handledare: Abigail Barker

Institutionen för geovetenskaper, Uppsala universitet, Villavägen 16, 752 36 Uppsala (www.geo.uu.se) ISSN 1650–6553, Examensarbete vid Institutionen för geovetenskaper, Nr 433, 2018

Hela publikationen finns tillgänglig på www.diva-portal.org

Table of Contents

1. Introduction ... 1

2. Background ... 3

2.1 Geological Background ... 3

2.1.1 Tectonic Setting ... 3

2.1.2 Fogo ... 4

2.1.3 Previous Work ... 4

2.2 Source Lithologies and Mantle Reservoirs... 5

2.2.1 Mantle Reservoirs ... 5

2.2.2 Source Lithologies ... 6

2.3 Nickel Isotope Geochemistry ... 8

3. Methods ... 11

3.1 Electron Microprobe Analysis ... 11

3.2 LA-ICP-MS Trace Element Analysis ... 11

3.3 Nickel Isotope Clean Laboratory procedures and MC-ICPMS ... 12

3.3.1 Nickel Isotope Clean Laboratory procedures ... 12

3.3.2 MC-ICPMS ... 13

4. Results ... 15

4.1 Petrography ... 15

4.2 Mineral Chemistry ... 16

4.2.1 Olivine ... 16

... 20

4.2.2 Clinopyroxene ... 21

4.3 Trace element composition of minerals ... 23

4.4 Nickel Isotope Geochemistry ... 27

5. Discussion ... 29

5.1 Magmatic processes ... 29

5.1.1 Fractional Crystallisation... 29

5.1.2 Olivine-liquid thermometry and variations in Ni* and Mn* ... 31

5.1.3 Variations in Mg# ... 33

5.2 Source lithologies ... 34

6. Conclusions ... 41

7. Acknowledgments ... 42

8. References ... 43

Appendix A: Mineral Chemistry ... 43

Major and Minor Elements ... 49

Olivine ... 49

Clinopyroxene ... 55

Mn*, Mn, Ni*, Ni, and Fo. ... 87

Mg# and Wo-En-Fs ... 92

Appendix B: Trace Element Composition of Minerals ... 107

Olivine ... 107

Clinopyroxene ... 110

Trace element ratios ... 113

Olivine ... 113

Clinopyroxene ... 113

Partition coefficients... 114

Appendix C: Nickel Isotope Geochemistry ... 115

1

1. Introduction

Determining the heterogeneities of the Earth’s mantle is an important tool when studying the evolution of our planet, as it can be used to unravel its original composition and subsequent evolution. The main volcanic lithologies that are found on volcanic islands are commonly called ocean island basalts, or OIBs. Although OIBs are mafic rocks they tend to show much greater chemical and isotopic variation than other types of mafic rocks, such as mid-ocean ridge basalts (MORBs) (Hofmann, Albrecht W.

1988).

OIBs are typically sourced from hotspot volcanism and are commonly thought to be the surface expression of an upwelling mantle plume. These mantle plumes are thought to rise from different levels within the mantle, possibly even from the core-mantle boundary, and as such may provide a window in to the mantle at several different levels (Winter, 2013). They are distinct from the MORBs in both trace elements and isotopic ratios, ranging from significantly more enriched in some ratios to more depleted in others, and their compositional differences combined with the range of origination depth are why they are thought to represent the heterogeneity of the mantle, sampling everything from the core mantle boundary to the old or recent additions of recycled continental and oceanic crust (Hofmann, Albrecht W. 1988; Hofmann, A. W. 2003). Due to this, OIBs are commonly used as a proxy for the composition of mantle geochemistry (Barker et al., 2014). OIBs are also more suitable since ocean islands commonly sit on oceanic crust which is much thinner than continental crust, leading to less crustal contamination and less modification of the rising melt (Zindler & Hart 1986).

The great variation in chemistry and isotopic ratios has been attributed to a variation in mantle sources. These mantle sources are inferred by studying differences in major and trace element geochemistry, as well as radiogenic isotopes. While there is a host of different types of mantle sources, the most common contributors to OIBs are high U/Pb ratio sources (HIMU), depleted mantle (DMM), and two different enriched mantle sources (EMI and EMII) (Hofmann, Albrecht W. 1988).

There are several mantle reservoirs present within the Cape Verde archipelago (Gerlach et al., 1988;

Hoernle et al., 2002; Barker et al., 2014). A HIMU, or HIMU-like component is present throughout the archipelago, coupled with an EMI component in the southern part, and a DMM component in the northern part. The DMM end-member has been determined to be derived from a peridotite melt, while the EMI and HIMU component indicate an origin that is a mixture between eclogite, pyroxenite, and peridotite. The modelled HIMU component has been shown to have a highly variable parental composition, ranging from a peridotite-pyroxenite mixture to one containing over 75% eclogite melt (Barker et al., 2014).

To further study these heterogeneities, we turn to one of the most active volcanoes in the world, the

island of Fogo in the southern archipelago of Cape Verde (Gerlach et al. 1988). Pico de Fogo is

composed of mainly high-alkali silica undersaturated rocks, such as basanites, tephrites, and

nephelinites, as well as carbonatites and phonolites (Gerlach et al. 1988), and shows contribution from

2

both the HIMU and EMI source lithologies with the contributions from the EMI component increasing in modern eruptions (Magnusson 2016; Mata et al. 2017).

This study aims to determine the origin of the mantle lithologies present at Fogo. The study is a continuation and extension of the studies conducted by Barker et al. (2014) and Magnusson (2016). The goals of this study are to explore the source lithologies and processes involved in the creation of the Fogo volcano.

This study utilises major, minor, and trace element geochemistry in clinopyroxene and olivine phenocrysts, as well as Ni-isotopes from whole rock samples. The study will primarily focus on the absolute and relative values of Mn, Ni, Fo in olivine to trace the fingerprints of peridotite, pyroxenite, and eclogite in the Fogo samples, using methods detailed in Sobolev et al., (2005, 2007), and Barker et al., (2014). Using the relative values of these elements and their ratios in olivine and clinopyroxene phenocrysts we aim to further unravel the mechanics of the creation of ocean islands and provide us with additional constraints regarding the mechanics of the formation of heterogeneities in the Earth’s mantle. The previous studies listed above have determined the Ni and Mn content in olivine of these different lithologies, as well as the Fo-content, and at the same time introduced the Mn/FeO and Ni*MgO/FeO ratios in olivine which allow us to see through the fractional crystallisation and other magmatic processes to find the fingerprint of the original mantle lithology (Sobolev et al., 2005, 2007;

Barker et al.,2014). The ratios of trace elements Ba/Th and La/Nb will also be utilised, as they are some of the few bulk partitioning ratios to show different behaviours between pyroxenite and peridotite and thus highly useful tracers of source lithologies (Stracke & Bourdon 2009).

As nickel is a non-traditional stable isotopic system, the data produced by this study will provide

additional information regarding this novel stable isotopic system. Only one previous study made by

Gall et al (2017) has looked in to the δ

⁶⁰

Ni of natural mantle rocks, in which they found an exciting

relationship between δ

⁶⁰

Ni and δ

⁵⁷

Fe (Gall, Williams, Halliday & Kerr 2017). This study will allow us

to investigate the relationship between δ

⁶⁰

Ni, high-precision geochemical data, and the δ

¹⁸

O obtained

by Magnusson (2016), and to compare them to the data obtained by Gall et al (2013, 2017) thus

providing additional constraints on the isotopic system, as well as on the variation of nickel isotopic

signatures in the mantle.

3

2. Background

2.1 Geological Background

2.1.1 Tectonic Setting

The Cape Verde archipelago is situated in the Atlantic Ocean, approximately 500 kilometres west of the African coast. The archipelago is composed of ten larger active and inactive volcanic islands, as well as a cluster of small volcanic islands and seamounts (Pim, Peirce, Watts, Grevemeyer & Krabbenhoeft 2008). The Cape Verde islands are commonly subdivided in to the North and the South archipelago, with Fogo being a part of the Southern division (Foeken, Day & Stuart 2009).

The islands are emplaced on the Cape Verde swell, a bathymetric anomaly which is 2.2 kilometres high and 1200 kilometres wide (Ali, Watts & Hill 2003), and currently widely considered the largest bathymetric anomaly in the world’s oceans (Carracedo et al. 2015). Magnetic anomaly studies have suggested that the swell beneath Cape Verde is composed of oceanic crust (Ali et al., 2003) with an age of approximately 125-150 Ma (Pim et al., 2008). The islands themselves range in age from approximately 8 Ma to 20 Ma, with the younger islands located to the West, and the older to the East (Pim et al., 2008).

The swell, and the emplacement of the island themselves, are commonly accepted to be associated with hotspot volcanism, related to the Cape Verde mantle plume (Pim et al., 2008). Hotspot volcanism, also commonly referred to as “oceanic intraplate volcanism”, mainly occurs within lithospheric plates, as the rising plume melts. The typical “sea chain” appearance of volcanic islands associated with hotspots are due to the movement of the lithospheric plate over a static mantle plume (Courtillot, Davaille, Besse &

Stock 2003; Yoshida 2014).

The Cape Verde hotspot has been linked to the proximal Canary hotspot. Patriat and Labails (2006) argued that the current treatment of the two volcanic-island archipelagos as expressions of two distinct mantle plumes may be erroneous. They base this on the similarities between the islands, the timing of volcanism and deformation on the islands, as well as events such as subsidence and uplift have been found to be close to contemporaneous, as well as the presence of a continuous oceanic ridge between the two (Patriat & Labails 2006).

The work of Montelli et al., (2004) also indicate a cogenetic relationship between the Cape Verde and Canary island hotspots, as well as the Azores hotspot. These three plumes have been shown to be a part of a larger plume complex with the Azores and Canary plumes merging at approximately 1450 kilometres, and the Cape Verde plume at >1900 kilometres (Montelli et al. 2004). Interestingly, the Canary and Cape Verde plumes show opposite

³

He/

⁴

He values. Cape Verde has a

³

He/

⁴

He ratio higher than the accepted MORB value of 8, and the Canary hotspot has a lower

³

He/

⁴

He ratio (~5-7) (Farley &

Neroda 1998; Courtillot et al. 2003), something that would indicate a distinct difference in the source of

the contributing mantle reservoirs.

4 2.1.2 Fogo

The island of Fogo rises to a height of approximately six kilometres over the ocean floor, making it the second highest island in the Atlantic Ocean (Ali et al., 2003). The highest point of the island, Pico do Fogo, reaches a peak height of just under three kilometres above sea level (Worsley 2015).

Pico do Fogo — Spanish for “peak of fire”— is a stratovolcano located in the caldera field of Chã das Caldeiras, which was created by the collapse of the ancient Monte Amarelo shield volcano (Carracedo et al. 2015). This collapse has been dated to roughly 63-123 ka through cosmogenic

³

He exposure ages (Worsley 2015).

The rock types present on Fogo are silica undersaturated, highly alkaline, mafic to ultramafic volcanic rocks (Hoernle et al. 2002; Carracedo et al. 2015), and Pico do Fogo is composed almost entirely of basaltic lapilli (Carracedo et al. 2015).

The samples used in this study cover eruptions from the late 1700’s to the most recent eruption of 2014-2015, as well as one pre-historic, undated sample. The samples have been collected from the Chã das Caldeiras collapse scar, with no samples heralding from the Monte Amarelo (Magnusson 2016).

The nature of the samples as well as the analysis conducted are detailed in table 1 below.

Table 1: Table detailing the samples used in this study. The table shows the sample name/number, rock type, year of eruption, and analyses conducted. Rock type and eruption year from Magnusson (2016).

Sample Rock type Eruption Year EPMA LA-ICP-MS Ni-isotopes

CVF01 Tephrite 2014 Yes Yes Yes

CVF02 Tephrite 1995 CPX only CPX only Yes

CVF03 Tephrite 1951 No No Yes

CVF04 Basanite 1857 Yes Yes Yes

CVF05 Basanite Pre-historic Yes Yes Yes

CVF06 Basanite 1799 Yes Yes Yes

CVF07 Basanite 1847 Yes Yes Yes

CVF08 Tephrite 1816 Yes Yes Yes

CVF09 Basanite 1852 Yes Yes Yes

CVF10 Basanite 1951 Yes Yes Yes

CVF11 Tephrite 1816 Yes Yes Yes

CVF13 Basanite/Tephrite 2014 No No Yes

CVF14 Basanite/Tephrite 2014 No No Yes

2.1.3 Previous Work

The Cape Verde archipelago has been investigated in several previous studies, including, but no limited

to: Furnes and Stillman, (1987) Holm et al., (2006, 2008), Barker et al., (2009, 2010, 2014) Hildner,

5

Klügel and Hauff, (2011) and Carracedo et al., (2015). The full body of work is too extensive to be discussed here, however, a few recent works are directly relevant to the research conducted in this study.

The study conducted by Barker et al., (2014) aimed to determine and quantify the different magma reservoirs involved in the heterogenous lithologies at Cape Verde. They studied trace elements in olivine and found that the cores of the olivine crystal differed from the global average as they showed low Mn/FeO and Ni*FeO/MnO values, rather than the negative array that is usually seen. They also found that this could not be explained by fractional crystallisation (Barker, A. K., Holm & Troll 2014).

Magnusson (2016) studied samples from modern eruptions at Fogo, covering the years 1799 to 2015.

He conducted the first study of

¹⁸

O isotopes in the area, as well as major and trace element whole rock analysis of the same samples. He also found evidence for a pyroxenite-rich magma source as a dominant contributor, based on the ratio of La/Nb as well as high TiO

2

values (Magnusson, 2016). Additionally, the magma under Cape Verde was found to have undergone both assimilation and crystal fractionation, and that there have been several episodes of magma mixing. Variations in trace elements are present on geologically very short timescales, with clear differences in lithologies in the eruptions covered in this study. This was interpreted as being due to differences in mantle sources, with a variation between pyroxenite and peridotite facies. The variation in

¹⁸

O-values was also interpreted to be due to a mantle heterogeneity since the values were incompatible with crustal assimilation. He also showed that some trace element ratios and the oxygen isotope values are increasing with time, something that indicates an increased influence from EMI in modern eruptions (Magnusson, 2016).

2.2 Source Lithologies and Mantle Reservoirs

2.2.1 Mantle Reservoirs

There are several mantle reservoirs that may contribute to ocean island volcanism, all characterised by depletion or enrichment in different isotopic and trace element ratios. The most commonly used trace elements are incompatible trace elements, which are enriched in OIBs relative to MORBs (Hofmann, Albrecht W. & White 1982). Given the absolute value of any given trace elements may fractionate through several magmatic processes, such as fractional crystallisation, the ratios of incompatible elements are more commonly employed to distinguish between mantle sources. Commonly used trace element ratios are La/Nb, Ba/Th, La/Sm, and Zr/Hf (Hofmann, A. W. 1997; Stracke & Bourdon 2009).

Nb/U and Ce/Pb ratios may also be used to differentiate between continental crustal sources and mantle

sources, but the trace element ratios are too uniform to be to differentiate between different basaltic

sources (47±10 and 25±5, respectively) (Hofmann, A. W., Jochum, Seufert & White 1986). Isotopic

systems are commonly used in conjunction with trace element geochemistry as mantle melting and

magmatic processes can cause fractionation patterns in trace elements that obscure the original mantle

reservoir or source lithology.

6

Radiogenic systems are a useful tool in geochemistry as they do not fractionate in the way trace elements do during partial melting or fractional crystallisation, and as such commonly provide a better indication of the source reservoir (Hofmann, A. W. 2003).

Five main mantle end-members have been distinguished based on differences in the radiogenic isotope ratios of Sr, Nd, Pb, He, and Os: HIMU, EMI, EMII, FOZO, and DMM (Hofmann, A. W. 2003).

These end-members and their interactions will be discussed below, although the main focus will be on HIMU, EMI, and DMM as these are the mantle reservoirs hypothesised to be the main contributors of the Cape Verde archipelago. The acronyms we use for the mantle end-members today were coined by Zindler and Hart (1986) and refer to the defining characteristics of each. Enriched mantle I (EMI) and II (EMII) refer to two different end-members that are enriched in

⁸⁷

Sr/

⁸⁶

Sr, with near-primordial (~0.705) and continental crust (>0.722) values, respectively (Zindler & Hart 1986). Both EM-reservoirs have similarly low

¹⁴⁴

Nd/

¹⁴³

Nd, although the Pb isotopic ratios differ, with EMI having very low

²⁰⁶

Pb/

²⁰⁴

Pb and high

²⁰⁸

Pb/

²⁰⁶

Pb (Hofmann, A. W. 2003). This has been attributed to several different factors, such as assimilation and recycling of subducted continental or oceanic crust and sediments, or the assimilation of delaminated subcontinental lithosphere, while the high

⁸⁷

Sr/

⁸⁶

Sr of EMI is attributed to the assimilation of a crustal or sedimentary component (Hofmann, A. W. 2003).

HIMU (high-μ) is thought to be the mantle end-member responsible for the enrichment of radiogenic lead. HIMU has the highest radiogenic lead ratios, but is comparatively depleted in other radiogenic isotopes, approaching depleted MORB values. This requires a mantle source with incredibly high U/Pb and Th/Pb ratios, and comparatively low parental Rb/Sr and Sm/Nd. The high radiogenic Pb of HIMU has been suggested to be due to either the metasomatism of oceanic lithosphere by partial melts with high U/Pb and Th/Pb, or lead and alkali-loss of recycled oceanic crust during subduction-related alteration (Hofmann, A. W. 2003).

DMM, or depleted MORB mantle, may also be referred to as DM, for depleted mantle. This mantle reservoir is characterised by high

¹⁴³

Nd/

¹⁴⁴

Nd (and ɛ-Nd values), and low

⁸⁷

Sr/

⁸⁶

Sr,

¹⁸⁷

Os/

¹⁸⁶

Os, and Pb- isotope ratios. This end member is commonly presumed to be the source of the depleted N-type MORB (Zindler & Hart 1986). DMM has a standard mantle

³

He/

⁴

He with an average of approximately 8±1R

A

. This value is remarkably stable, and OIBs show a much large range of

³

He/

⁴

He ratios, spaning from 5 to 43 R

A

(Zindler & Hart 1986). This indicates that a source with higher

³

He/

⁴

He ratios must be involved in the genesis of OIBs. FOZO, or “focal zone”, has been suggested to be the contributor of the primitive He values. This component is thought to be representative of the composition of the primitive lower mantle, and to be present as a mixing component in all plumes originating in the deep mantle (Hofmann, A. W. 2003).

2.2.2 Source Lithologies

The three main lithologies present in the Earth’s mantle are peridotite, pyroxenite, and eclogite.

7

Peridotite is the most common lithology in the upper part of the Earth’s mantle and is generally composed of mainly olivine and pyroxene, with varying modal proportions of other ultramafic minerals such as garnet and spinel, depending on temperature and pressure. Peridotite is subdivided in to four groups based on the modal proportions of its constituent minerals: Dunite, Wehrlite, Harzburgite, and Lherzolite. Dunite has typically ≥90% olivine, with the modal proportion of olivine falling as the other subgroups incorporate clino- and orthopyroxenes and garnet (Winter, 2013).

Eclogite lithologies are a product of recycled subducted crust, commonly produced in subduction zones (Sobolev et al. 2007). They are commonly composed of eclogite-facies minerals, such as omphacite and garnet (Zheng & Chen 2017). Due to the lower solidus temperature of eclogites, they start melting deeper in the mantle where they react with the ambient peridotite to form pyroxenite (Sobolev et al., 2007; Barker et al., 2014). Pyroxenite is a metasomatized pyroxene-rich mantle lithology, mainly composed of different pyroxene minerals, such as enstatite or augite (Winter, 2013), although they commonly contain garnet and inherited olivine as well (Sobolev et al. 2007). Eclogite has been thought to be the cause of the recycled crustal signature (high-μ,

⁸⁷

Sr/

⁸⁶

Sr, Ba, Th, LREE, Pd, and low Ti, Sr, Hf, and

¹⁴³

Nd/

¹⁴⁴

Nd) which is commonly present at OIBs (Dasgupta et al., 2004).

To distinguish between pyroxenite and peridotite when reconstructing a parental magma, the proportions and absolute values of trace elements Ni and Mn, along with the forsterite content (Fo) of olivine are commonly used. This is due to the fact that olivine, especially high-Mg olivine, commonly occurs as the earliest mineral to crystallise in a melt, and thus has the highest probability of retaining the “foot print” of the parental magma itself (Sobolev et al. 2005). Several different workers, such as the experimental work performed by Pertermann and Hirschmann, (2002), as well as the work by Sobolev et al., 2005, 2007 have modelled the end member compositions of FeO, MgO, Cao, Mn, and Ni in eclogite, peridotite, and pyroxenite in olivine and whole rock compositions, and as such they are now easily distinguished.

Melts derived from the average peridotite end-member have low to moderate Ni content (≤560 ppm), and high Mn (≥1400 ppm), FeO, CaO and MgO. The melt derived from the average modelled pyroxenite end-member have high Ni content (≥1000 ppm), moderate Mn (≤920 ppm) and CaO, and low to moderate MgO and FeO. The melt derived from experimental eclogite melts have very low Ni contents (≤70 ppm), roughly 14 times lower than pyroxenites and 8 times lower than peridotites, and a MgO content that is roughly 12-13 times lower than both the other lithologies. FeO and CaO content is comparable to pyroxenites, and the Mn content is approximately 675 ppm (Pertermann & Hirschmann 2002; Sobolev et al. 2005, 2007).

The key to the dispersal of these elements lies in their constituent minerals. Nickel, for example, is

highly compatible in olivine. This means that a peridotitic melt, which has a high modal proportion of

olivine, will have a lower proportion of Ni in the melt in comparison to pyroxenite, since a lot of the

nickel will have been sequestered by the olivine (Sobolev et al. 2007). Recently, it has been argued that

8

the coefficients of Mn and Ni depend on two different parameters. Matzen et al., (2017) argue the weight percent of MgO in the liquid controls the partition coefficient of Mn in olivine, with the compatibility decreasing with increasing MgO. Nickel, on the other hand, was shown to be mostly independent on the MgO content of the melt but rather depended on temperature (Matzen, Wood, Baker

& Stolper 2017). Given the higher MgO contents of the average peridotite melt relative to a pyroxenite one, this could perhaps be another factor in the differing enrichment of trace elements. Trace element ratios are a helpful tool in resolving the relative contributions of pyroxenite and peridotite. Ba/Th, La/Sm, and La/Nb are especially useful due to differences in partitioning behaviour and controls, and may as such, when applied together, be fairly robust measures of source lithology. Their relative ratios are well established, for example a Ba/Th ratios of ≤62 indicates a pyroxenite source, a La/Sm ratio of

≥0.4 indicates a peridotite source, and a La/Nb ratio of 1.6 and above indicates a peridotite source (Stracke & Bourdon 2009).

Another lithology that is worth discussing given the scope of this project are carbonatites.

Carbonatites are formed from the low-degree partial melting of carbonate rich mantle lithologies, and are known for their unusually high concentration (1000s of ppm) of Sr and REEs, as well as their significant variation in content of HFSEs (Chakhmouradian 2006; Bell & Simonetti 2010). There is still ongoing debate in regard to the petrogenesis of these melts, and several different parental melts have been suggested such as a deep plume origin, crustal recycling, metasomatized HFSE rich mantle, sub- lithospheric origin, and complex combination of all four. The trace element, radiogenic, and stable isotope constraints of carbonatite petrogenesis are severe, and it is incredibly complex to find a model that satisfies all requirements (Bell & Simonetti 2010). However, Zhang et al., (2017), showed that carbonated silicate melts, which are widely considered closely related to carbonatites, can exist in the shallow mantle. They also showed that carbonated silicate melts may evolve to alkali basalts through reactions with the lithospheric mantle (Zhang et al. 2017). This may open up additional pathways into resolving the genesis of carbonatites.

2.3 Nickel Isotope Geochemistry

Nickel is a trace element, belonging to the first row of transition metals on the periodic table. Nickel has five stable isotopes and 11 radiogenic isotopes, although only one is suitable for radiometric dating (Elliot and Steele, 2017). The five stable isotopes,

⁵⁸

Ni,

⁶⁰

Ni,

⁶¹

Ni,

⁶²

Ni, and

⁶⁴

Ni, are cosmochemically abundant (Elliot and Steele, 2017). Nickel may exist in several different oxidation states, ranging from 0 to +IV, although natural samples are present in +II only. More than 99% of the Earths nickel reservoir is present in the upper mantle. In mantle rocks, nickel is preferentially sequestered in to olivine, as it is a highly compatible element under mantle melting conditions and has a KD

_liq^ol

of as high as 47 (Adam

& Green 2006). Nickel is also sequestered in to clinopyroxene, spinel, and garnet, although in lesser

amounts (Gall, Williams, Siebert & Halliday 2012).

9

Table 2: Table showing the atomic fraction (with 2 SE) and nuclide mass of the five stable isotopes of nickel.

Data from Elliot (2017).

58

Ni

⁶⁰

Ni

⁶¹

Ni

⁶²

Ni

⁶⁴

Ni

Atomic Fraction

0.687076886 0.26223146 0.01139894 0.03634528 0.00925546

2 SE

5.92*10^-5 5.14*10^-5 4.33*10^-5 1.14*10^-5 5.99*10^-5

Nuclide Mass

57.93534241 59.93078589 60.93105557 61.92834537 63.92796682

Table 3: Table showing the atomic ratios (with 2 SE) of the five stable isotopes of nickel.

Data from Elliot (2017).

60

Ni /

⁵⁸

Ni

⁶¹

Ni/

⁵⁸

Ni

⁶²

Ni/

⁵⁸

Ni

⁶⁴

Ni/

⁵⁸

Ni Atomic ratio

0.385198965 0.016744215 0.053388576 0.013595598

2 SE

8.27*10^-5 6.52*10^-6 1.74*10^-5 8.88*10^-6

Nickel does not fractionate through kinetic or equilibrium fractionation, something which is common with other transition metals (Gueguen, Rouxel, Ponzevera, Bekker & Fouquet 2013). Nickel fractionation may occur, however, due to changes in speciation or coordination (Gall et al. 2012). This means that there is less change in the isotopic composition of nickel as the magma evolves relative to other stable isotopic systems, giving us a better footprint of the magmas initial composition (Gueguen et al., 2013). Nickel is also not redox sensitive under the conditions present during mantle melting (Gall et al. 2017).

The most commonly used isotopic ratio in the nickel system is

⁶⁰

Ni /

⁵⁸

Ni, which is expressed as δ

⁶⁰

Ni.

This ratio reflects, the relationship between the

⁶⁰

Ni /

⁵⁸

Ni of the sample versus the

⁶⁰

Ni /

⁵⁸

Ni of a reference standard, commonly NIST-SRM986, a pure nickel metal, as shown by the equation below:

δ

⁶⁰

Ni = [

^{60𝑁𝑖/58𝑁𝑖𝑠𝑎𝑚𝑝𝑙𝑒}

60𝑁𝑖/58𝑁𝑖_{𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑}

− 1] ∗ 1000

This type of notation is referred to as the δ notation and common in other stable isotopic systems, such as oxygen and carbon. It is expressed as parts per mil, or ‰ (Gall et al. 2017).

Comparatively few studies exist on nickel in high-temperature systems, relative to other stable isotope

systems (Gueguen et al., 2013, 2017). The studies that do exist are mainly focused on cosmochemical

applications or ore geochemistry, and only one of the studies, Gall et al., (2017), has focused on natural

samples rather than standards or reference materials (Gall et al. 2012). This is likely due to the difficulty

in purifying nickel from natural samples. This difficulty stems from the similarity in the behaviour of

nickel compared to other elements that are very common in geological samples, such as Mg and Ca, and

is further exacerbated by the fact that the major elements are present on a percentage scale while nickel

10

is commonly on present on a ppm level, and thus are thousands of times more abundant (Gall et al.

2012).

Gall et al., (2017) is the first study to date to investigate the nickel isotope systematics of mantle rocks. They analysed several different types of mantle and volcanic rocks, such as komatiites, peridotites, and mantle xenoliths, as well as mineral separates of olivine, garnet, and orthopyroxenes and clinopyroxene (Gall et al. 2017). They found that the mantle is likely to be broadly homogeneous in the distribution of nickel isotopes, with a δ

⁶⁰

Ni of 0.23 ± 0.06‰. They also found that δ

⁶⁰

Ni is not strongly affected by either serpentinization or metasomatism (Gall et al. 2017).

Additionally, they observed a strong correlation between the modal proportions of clinopyroxene and

the δ

⁶⁰

Ni of their samples, with the more clinopyroxene-rich samples being by far the isotopically

heaviest, regardless of rock type. This indicates that the isotopic composition of natural samples is

controlled by the modal proportion of clinopyroxene, despite the majority of the nickel being present in

the olivine (Gall et al. 2017). They also found a correlation between δ

⁶⁰

Ni and δ

⁵⁷

Fe, where the samples

with heaviest δ

⁶⁰

Ni also have the heaviest δ

⁵⁷

Fe, meaning that the Fe and Ni shows similar behaviour in

mantle lithologies (Gall et al. 2017). The partitioning of iron isotopes has in the past been partially

attributed to changes in redox state, but this is harder to defend given the correlation of δ

⁶⁰

Ni and δ

⁵⁷

Fe

since nickel is not redox sensitive at these conditions. This means that the correlation must reflect

another process capable of altering both isotopic systems. A likely candidate is heterogeneity within

mantle reservoirs or source lithologies, as iron isotopes have been shown to differ between different

source lithologies (Williams & Bizimis 2014; Gall et al. 2017).

11

3. Methods

3.1 Electron Microprobe Analysis

The olivine and clinopyroxene phenocrysts present in sample CVF01, CVF02 and CVF04-11 were analysed using the JEOL JXA-8530F Hyperprobe FEG-EMPA at the Department of Earth Sciences, Uppsala University.

A high-precision measurement program was used to measure the olivine phenocrysts, utilizing an accelerating voltage of 15 keV and beam current of 200 nA, with a 1-micron beam diameter. 60 seconds were spent on peak analysis and 30 seconds on background. The high beam current was chosen to gain better analytical precision and detection limits, which enabled the analysis of the trace elements Ni and Mn, as these elements have a detection limit of ≤135 ppm, 45 ppm higher than the major elements which have a detection limit of ≤90 ppm (Barker et al., 2015). The uncertainties for these analyses are <3.6%

s.d. for Ni, Mn, and Ca, and <3% s.d. for the major elements based on analyses of reference materials (Nobre-Silva et al., in prep).

Each crystal was analysed in three separate spots centred in the core of the phenocrysts, and the final measurement used is the calculated average of these three analyses. Effort was made to analyse 30 crystals per sample whenever possible in order to acquire a suitable amount of data. The core was chosen as the sample point as olivine is the first mineral to crystallise in a melt as it ascends, and thus records the Mn and Ni contents most similar to its source lithology (Barker et al., 2014).

The clinopyroxene phenocrysts were analysed using a standard measurement protocol, employing the same voltage as the olivine protocol, but with a lower beam current of 10 nA. The clinopyroxene phenocrysts were analysed based on the placement of zonation. Clinopyroxene composition evolves with the melt, therefore the zonation present is likely to record processes that have changed the composition of the melt prior to eruption, and thus are important clues in to the evolution of the melt.

Analytical precision for this analytical protocol has been determined based on measurements made on Smithsonian Institute mineral standards and showed that elements with a concentration of 10 wt% or above have uncertainties of ≤1.5% s.d., while elements in a wt% range of 5-10 and 2-5 wt% have uncertainties of ≤2.2 and ≤4.5% s.d. respectively. Elements with a concentration of 0.2 to 1.5 wt% have uncertainties of ≤10% s.d. (Barker et al., 2015).

3.2 LA-ICP-MS Trace Element Analysis

Thin sections of samples CVF01, CVF02, and CVF04-11 were analysed for trace element

compositions using a 193 nm Resonetics ArF excimer laser coupled to a Thermo Element XR ICP-MS

system at the Institute of Geochemistry and Petrology, ETH Zürich. The analyses were performed by

Dr. Ben Ellis, using a similar protocol to the one used in Szymanowski et al., (2015).

12

Laser spot size was 43 microns with NIST612 used as the primary standard and GSD-1G as secondary standard (data in supplementary materials). The output energy of the laser was maintained at approximately ~3.5 J cm

-2

for all rounds of measurement. The resulting raw data was reduced using the MATLAB based program SILLS (Guillong et al., 2008) using the average SiO

2

from microprobe measurements as internal standard.

For all trace elements present at least two times the limit of detection (LOD), the precision is approximately 5% 2σ. The precision for trace elements present at LOD is slightly lower. The typical LOD for all trace elements measured are 10

^-2

to 10

^-4

. All trace elements measured in the

clinopyroxene phenocrysts are present at concentrations well above the LOD. In the olivine

phenocrysts, Sr, Y, Zr, Nb, Sm, Dy, Ho, Er, Tm, Yb, Lu, Pb, Th, and U are present at concentrations of 1.5 times LOD or better, while Rb, Cs, Ba, La, Ce, Pr, Nd, Eu, Gd, Tb, Hf, Ta are present at LOD.

The location of clinopyroxene and olivine phenocrysts were mapped using scans of the thin sections, and the spots used for analysis of clinopyroxene were chosen using backscattered electron imagery on the JEOL JXA-8530F Hyperprobe at CEMPEG, Uppsala University. This was done in an effort to retrieve trace element data from the same crystals and spots as previously analysed for major and minor elements.

3.3 Nickel Isotope Clean Laboratory procedures and MC-ICPMS

All samples (CVF01-14) have been crushed and milled using an agate mill at the Department of Earth Sciences in preparation for isotopic analysis of mass-dependent Ni isotope variations. All Ni-isotope analyses were performed at the Department of Earth Sciences, Oxford University.

3.3.1 Nickel Isotope Clean Laboratory procedures

The method of Ni purification used in this study involves the same protocols as used in Gall et al., (2012). Based on trace element data, sufficient sample was weighed to have 2μg of sample Ni. The samples were then optimally spiked by weight with 5μg a

⁶¹

Ni and

⁶²

Ni double-spike before digestion.

This was done in order to correct for instrumental mass-fractionation and ensure a high-precision isotopic measurement (Gall et al. 2012). For each run approximately 10% of each sample, or 700 nanograms of Ni, were analysed. In addition to the samples of this study, two USGS rock standards, BIR-1a and BHVO-2, and a procedural blank were processed through the Ni- purification protocol of Gall et al., (2012).

The dimethylglyoxime (DMG), ammonium citrate, and sodium hydroxide solutions used in the Ni

purification were prepared from crystalline commercial reagents. Ethanol, acetone, and ammonia

solutions used were reagent grade solutions, and the dichloromethane used for solvent extraction was of

trace element grade. The HCl, HNO

3

, and HF used were all purified in-house, using sub-boiling

13

distillation. All water added during chemistry was 18.2M-grade Milli-Q water from a Milli-Q Element water purification system (Gall et al. 2012).

The samples were dissolved in Teflon beakers using a 1:3 mixture of concentrated HF-HNO

3

, dried and redissolved in 6M HCl to check for complete dissolution. They were then dried prior to preparation for purification. The purification involves a three-column ion exchange chromatography protocol. The first column is designed to separate Ni from alkaline and alkaline earth elements, such as Na, K, Mg, and Ca using a combination of NH

4

OH and di-ammonium citrate ((NH

4

)

2

C

6

H

6

O

7

). The second column uses an oxalic acid-HCl step, designed to remove the remaining Mg, Ca, Na, NH4+, Al, Cr and Ti, followed by an HCl-acetone mixture which removes Mn, Cu, Zn, Sn, and Cd. The nickel is then eluted from this column using Ni-DMG complex in HCl-acetone media, dried down and treated with concentrated HNO

3

to break down the DMG. The samples are then redissolved in HCl-H

2

O

2

and loaded on a final clean-up anion exchange column, which separates Fe from Ni using HCl-H

2

O

2

mixtures.

These steps yield close to 100% Ni for high-Ni samples, and a slightly lower yield of 85-95% for low- Ni samples (Gall et al. 2012).

3.3.2 MC-ICPMS

The isotopic ratios of

⁶⁰

Ni/

⁵⁸

Ni,

⁶¹

Ni/

⁵⁸

Ni, and

⁶²

Ni/

⁵⁸

Ni were analysed using a Nu Plasma-HR multi- collector inductively coupled mass spectrometer. The NIST SRM 986 was used as a reference standard for standard-sample bracketing during analysis.

For each measurement four nickel isotopes (

⁶²

Ni,

⁶¹

Ni,

⁶⁰

Ni, and

⁵⁸

Ni) were measured simultaneously, together with

⁵⁷

Fe. Measurements were carried out using pseudo high-resolution (Weyer and Schwieters, 2003) in order to accurately and precisely correct for polyatomic interferences on both

⁵⁷

Fe (which is used for correction of

⁵⁸

Fe on

⁵⁸

Ni) and

⁵⁸

Ni (J Barling, 2018, personal communication, April 6). This is necessary for accurate measurement of 

^60/58

Ni

NIST

.

The samples were introduced to the plasma using a micro-concentric PFA nebuliser coupled with a DSN-100 desolvator. For each measurement all isotopes were measured simultaneously, using a measurement sequence of 40 cycles with 10 second integration of the ion beam intensity. Typical instrument settings include a RF power of 1300 W, and an acceleration voltage of 5.84 kV. Sample uptake time is 60 seconds, with a background measurement time of 20 seconds (Gall et al. 2012).

All nickel isotope data in this study is reported as δ

⁶⁰

Ni (in ‰), which is the ratio of

⁶⁰

Ni/

⁵⁸

Ni relative to the same ratio in the nickel isotope standard NIST SRM 986. This ratio is calculated using the following equation:

δ

⁶⁰

Ni = [

^{60𝑁𝑖/58𝑁𝑖𝑠𝑎𝑚𝑝𝑙𝑒}

60𝑁𝑖/58𝑁𝑖_{𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑}

− 1] ∗ 1000

The accuracy and external reproducibility of Ni-isotopic measurements produced using this method

are assessed by means of two USGS standards, BIR-1a and BHVO-2. The mean measured δ

⁶⁰

Ni of BIR-

14

1a is 0.181 ± 0.047‰ (2σ, n=3) within error are the long-term Oxford average of 0.147 ± 0.062‰ (2σ, n=22; N.J. Saunders pers. comm.). The measured δ

⁶⁰

Ni of a single analysis of BHVO-2 is 0.064‰ is within error of the long-term Oxford average of 0.028 ± 0.065‰ (2σ, n=39; Gall et al., 2012, 2017, N.J.

Saunders pers. comm.). In addition to these standards, two in-house standards (Sudbury and Kambala) are used to assess the efficiency of the pseudo high-resolution interference correction. The δ

⁶⁰

Ni for Sudbury and Kambala measured in this study are 0.749 ± 0.058‰ (2σ, n=4) and -0.904 ± 0.047‰ (2σ;

n=5), respectively. These are within error of the long-term averages for these standards (N.J. Saunders

pers. comm.). External reproducibility of the analysed samples CVF01-11, 13, and 14 is comparable to

that of the USGS reference samples.

15

4. Results

4.1 Petrography

The twelve samples (CVF01 to 12) were studied to identify the locations of olivine and clinopyroxene phenocrysts. The samples are composed of olivine and clinopyroxene phenocrysts, as well as opaque minerals. The samples all have a porphyritic texture and microcrystalline to glassy groundmass.

The groundmass is composed of olivine, clinopyroxene, and opaques, with the occasional plagioclase and feldspathoids. Grains of amphibole xenocrysts showing intense reaction rims are present in CVF04, CVF08, CVF10, and CVF12 (figure 1).

Figure 1. Images taken with the transmitted light microscope showing representative grains of amphibole xenocrysts (A and B), clinopyroxene (C and D), and olivine (E and F). The images also show the range of crystallinity present in the groundmass.

16

The samples show great variation between the modal proportions of olivine and clinopyroxene, as well as the modal proportion of groundmass versus phenocrysts.

Clinopyroxene is, however, the dominant mineral in all thin sections, regardless of the amount of olivine phenocrysts. CVF07, CVF08, and CVF09 are the most olivine rich with 6 to 10% of olivine, while CVF03 and CVF10 contains the least amount of olivine phenocrysts with roughly 1-3% olivine (Magnusson 2016). CVF02 contains exclusively clinopyroxene phenocrysts, constituting roughly 30%

of the sample, with olivine present only within the microcrystalline groundmass.

The clinopyroxene phenocrysts show several different zonation patterns, including oscillatory and hourglass zonation, as well as simple twinning. The clinopyroxene crystals also commonly occur in clusters (figure 1C), and intergrowths between clinopyroxene grains and olivine crystals are common.

The olivine crystals, however, very rarely show any zonation, although inclusions of opaques and vesicles are common (figure 1E, F). The size of the olivine phenocrysts varies from a few μm in size, microphenocrysts, up to several millimetres. The largest phenocrysts are found in CVF07 and CVF09.

4.2 Mineral Chemistry

All data collected and calculated is presented in appendix A.

4.2.1 Olivine

The forsterite content (Fo) of all the olivine phenocrysts have been calculated, and the data is presented in figure 2. The majority of the olivine phenocrysts have a forsterite content of Fo

79

to Fo

82

. 15% of the samples have a forsterite content exceeding Fo

82

, and ~9% of the samples have a forsterite content below Fo

79

.

Figure 2. Histogram showing the forsterite content and distribution of all measured olivine phenocrysts.

17

In figure 3 we can see nickel and manganese (in ppm) plotted against forsterite content and compared with high-precision olivine data from worldwide (Sobolev et al., 2017) and other Cape Verde samples (Barker et al., 2014). The figures show that the samples of this study lie on the lower end of the global trend regarding forsterite (Fo

72

to Fo

83

versus Fo

79

to Fo

95

, respectively) and nickel content (295 ppm to 1330 ppm versus 812 ppm to ~4500 ppm, respectively), with moderate to high manganese (1610 ppm to 3300 ppm versus 2330 ppm to 720 ppm, respectively). One sample (CVF06) has a very low nickel content (46 ppm - 132 ppm) and is separated from the general trend. This outlier is visible in the manganese as well, although the low values are not as severe (1610 ppm to 2680 ppm).

In an attempt to remove any magmatic processes, such as fractional crystallisation, from the results all nickel and manganese data was transformed in to Ni* and Mn*. This was done using the equations Ni*MgO/FeO and Mn/FeO, respectively, based on the methods presented in Sobolev et al., (2005, 2007).

Figure 3. Diagram showing the samples of this study plotted together with high-precision data from published studies. The data from Barker et al., (2014) are olivine phenocrysts from other areas of the Cape Verde archipelago (Santiago, San Antão), and the data from Sobolev et al., (2007) represents the global array. The samples of this study have been divided according to their Mn* and Ni* values.

18

The filtering was not entirely successful (figure 4), and the correlation between Ni* and Mn* and forsterite content is still present, although it is less pronounced.

In figure 3 and 4 we can also see that the samples with the highest Mn and Mn* (based on the division made in figure 5) are the samples that shows the closest similarity to the samples from Barker et al., (2014) on the Ni and Ni* diagrams, and also tend to be the samples with the highest, and thus most primitive, forsterite content. The same trend is not present in the Mn and Mn* diagrams, but the samples can rather be said to be more or less clustered together, with the high Mn and low Ni and Mn samples showing the same values as the global array, while the samples that are present on the global array falls on a high-Mn low-Fo trend and thus forms the end tail of the global array.

In an attempt to quantify the contributions of the source lithologies present in the different Fogo eruptions, Ni* was plotted versus Mn* (figure 5).

Pyroxenite, peridotite, and eclogite end-members have been labelled based on values that were experimentally determined by Sobolev et al., (2005, 2007) (pyroxenite and peridotite) and Pertermann and Hirschmann (2002) (eclogite). The pyroxenite end-member has a Ni* value of approximately 1000,

Figure 4. Diagram showing the samples of this study plotted together with high-precision data from published studies.

The data from Barker et al., (2014) are olivine phenocrysts from other areas of the Cape Verde archipelago (Santiago, San Antão). The samples of this study have been divided according to their Mn* and Ni* values.

19

and a Mn* value of approximately 90. The peridotite end member has a Ni* value of approximately 350

and a Mn* value of approximately 130. The eclogite end-member has a Ni* of approximately 400,

comparable to the peridotite end-member, but the Mn* value of approximately 80 is lower than the other

two end-members. Values representing the global array has been added for reference, along with a

pyroxenite-peridotite mixing line. In figure 5 we can see that the samples are divided in to three clear

groups, one that is composed solely of sample CVF06 and has very low Ni* (38 to 54) and Mn* (91 to

107), one group that is on or close to the global array with a Ni* range of 210 to 506 and a Mn* range

of 91 - 131, and one group that has Ni* values close to or lower than that of the global array (165 to

411), but moderate to high Mn* values (131 to 179).

20

Figure 5. Diagram showing the samples of this study plotted together with high-precision olivine data from published studies (light grey diamonds) and data from Barker et al., (2014) (dark grey outlined diamonds). Data for the global array was provided by Abigail Barker (PhD). The samples of this study have been divided according to their Mn* and Ni* values.

21 4.2.2 Clinopyroxene

The proportions of wollastonite, enstatite, and ferrosilite, have been calculated for all measurements, and are presented on the ternary diagram in figure 6A. This figure shows that all samples plot within a fairly narrow Ca range, with the exception of a few outliers, and show a larger variation in the relative Fe and Mg content. The samples mainly classify as diopside, with a few samples classifying as augite and one as hedenbergite.

The magnesium number (Mg#) of all the clinopyroxene measurements have also been calculated, and the data is presented in figure 6B. The total range in Mg# is fairly large, but this is due to isolated outliers rather than a spread in crystal composition between samples. The vast majority, 70%, of the measured clinopyroxene phenocrysts have a Mg# between 73 and 79. 11% have a Mg# between 79 and 85, and 19% have a Mg# below 73.

Figure 6. A ternary diagram (A) showing the proportions of wollastonite, enstatite, and ferrosilite in the measured samples, and a histogram showing the magnesium number (Mg#) and distribution of all measured clinopyroxene phenocrysts (B).

B

A

22

The composition of clinopyroxene phenocrysts evolves with the melt, a process that is likely to be reflected in compositional changes of the mineral grains. These compositional changes are visible as zonation. To further investigate this, the composition of clinopyroxene phenocrysts were measured along lines that were placed to cut through the zonation of the grains. These Mg# of the measurements taken along these lines were then calculated and plotted against the distance of the lines (figure 7).

The zonation profiles presented in the diagrams in figure 7 show that the differences in Mg# are highly varied between different samples. Some profiles, such as the phenocrysts measured in CVF05 (figure 7A) and CVF01 (figure 7B) show relatively smooth profiles, with an initial increase in Mg#, followed by a one or two increases and subsequent decreases in Mg#. Both profiles show, however, quite a large difference in the change in overall Mg#, with the phenocryst analysed in CVF05 showing an overall change from 79 to 71, and the phenocryst analysed in CVF01 shows a change from 76.5 to 79.5.

The zonation of some of the other phenocrysts profiles show a much more erratic pattern of zonation, such as CVF11 (figure 7C) and CVF07 (figure 7D). The profile shown in figure 7C shows an initial increase in Mg# from 77 to 79, followed by a small decrease, another small increase, and then a larger decrease from 79 to approximately 75.5.

The trend continues slowly downwards, punctuated by several small increases in Mg#, before a drastic increase from Mg# from approximately 74 to 81. The profile then shows a moderate decrease from 81 to 77 followed by an increase of almost comparable size (77 to 80.5) over 10-20 μm. The end of the profile shows a decrease, followed by an increase, and then a downward slope. The profile shown in figure 7D is the most erratic one, and also has the largest overall change in Mg#, from ~64 to ~78.

Figure 7. Diagram showing the change in Mg# over distance from four representative clinopyroxene phenocrysts. Two of the diagrams show an expected zonation profile (A, B), and two of the phenocrysts show a highly erratic zonation profile (C, D). The distance (in μm) is from the rim (0 μm) towards the core of the phenocryst. Grey X’s indicate a measurement point. The phenocrysts shown above are from CVF05 (A), CVF01 (B), CVF11 (C), and CVF07 (D).

23

The profile shows five initial increase and decrease peaks, who display an overall downward trend from a high Mg# of 74 to 78 towards a lower Mg# of 64. Halfway through the profile the trend changes to an increase in overall Mg# with one initial small peak of approximately 68, followed by a plateau at a higher Mg# of 76 before an additional increase to 79. The profile then shows two more decreases and subsequent increases.

4.3 Trace element composition of minerals

All data collected and calculated is presented in appendix B. All samples except CVF03 and CVF12 were analysed for phenocryst trace element composition. Clinopyroxene phenocrysts were analysed in samples CVF01, 02, and 04 - 11, and olivine phenocrysts were analysed in samples CVF01 and 04 - 11.

The trace mineral composition of the parental melts was calculated from the resulting mineral data using partition coefficients (K

Dmin-liq

) from Adam and Green (2006). The partition coefficients were chosen as they are basanite or alkali basalt values, and as such are most appropriate for the host rock values of the samples that were analysed. Figures 8 and 9 show the calculated melt compositions for each sample plotted as a full spidergram and REE-diagram, respectively, along with reference samples from the Cape Verde archipelago (from Escrig et al., (2005)) and Cape Verde carbonatites (Hoernle et al. 2002).

The carbonatites plotted are the mean of the full set of samples, divided in to their respective subtype, magnesio-carbonatites, and calcio-carbonatites.

Figure 8. Full spidergram after McDonough and Sun (1995). The samples in this study are plotted in purple (olivine phenocrysts) and dark green (clinopyroxene phenocrysts). The carbonatites are plotted based on their respective subtype, with calcio-carbonatites shown in turquoise and the magnesio-carbonatites shown in bright green. The silicate rocks from Escrig et al., (2005) are plotted in light grey and the whole rock values from Magnusson (2016) are plotted in dark grey.

Carbonatite data from Hoernle et al., (2002) and silicate reference data from Escrig et al., (2005).

24

In figure 8, a full spidergram after McDonough and Sun (1995), we can see that the samples of this study have similarities with both the silicate reference samples and the carbonatites, and also differences and similarities between the olivine phenocrysts and the clinopyroxene phenocrysts.

The olivine phenocrysts are generally depleted relative to clinopyroxene phenocrysts, and to a lesser extent the silicate rocks, with the exception of CVF01 and CVF07. These two samples are generally comparable to the carbonatites and the clinopyroxene phenocrysts, and in some cases CVF01 exceeds both (Sm, Sr, Ce). The remaining olivine phenocrysts are depleted relative to the silicate rocks in Ba, Ce, Sr, Nd, and the HREEs, but otherwise show comparable values.

The clinopyroxene phenocrysts, however, are in general more enriched than the silicate rocks, and in some cases also exceed the carbonatites (Cs, Rb, Ta, Zr). The clinopyroxene phenocrysts are comparable to the silicate rocks in U, Pb, Sr, and Y, but are otherwise enriched. They are enriched relative to the olivine phenocrysts in all elements except U and Pb. They are generally comparable to the carbonatite samples and are only depleted relative to the carbonatites in regard to Sr and the HREEs (with the exception of Yb), although they are moderately to highly enriched in Nb, Ta, and Zr.

A REE diagram was plotted to further investigate the relationship between the samples of this study, Cape Verde carbonatites, and Cape Verde silicate rocks (figure 9).

In figure 9, we can see that the clinopyroxene phenocrysts have a clear affinity for the carbonatite rocks, especially the calcio-carbonatites which they overlap in all but the HREEs where they are marginally lower. All REEs are enriched relative to the silicate rocks and the pattern is mainly smooth

Figure 9. REE diagram after McDonough and Sun (1995). The samples in this study are plotted in purple (olivine phenocrysts) and dark green (clinopyroxene phenocrysts). The carbonatites are plotted based on their respective subtype, with calcio-carbonatites shown in turquoise and the magnesio-carbonatites shown in bright green. The silicate rocks from Escrig et al., (2005) are plotted in light grey and the whole rock values from Magnusson (2016) are plotted in dark grey.

Carbonatite data from Hoernle et al., (2002) and silicate reference data from Escrig et al., (2005).

25

except for Yb, which is slightly enriched. The olivine phenocrysts are more chaotic and are mildly depleted in LREEs, with the exception of La and Sm where they are comparable to the silicate rocks, and highly depleted in the HREEs. Two samples, CVF07 and CVF01, divert from the general trend and are moderately to highly enriched (respectively) in the LREEs, with CVF01 even surpassing the carbonatites in La and Ce. The differences and similarities between the calculated melt values of the clinopyroxene phenocrysts, olivine phenocrysts, and the whole rock samples from Magnusson (2016) are readily visible when plotting the trace elements versus Y (figure 10).

The enriched nature of CVF01 is especially visible in Nb, Ba, La, and Th, but it is not as much of an outlier in Sm. The clinopyroxene phenocrysts and the silicate rocks are roughly comparable in Y-values with the silicate rocks ranging from approximately 25 to 35, and the clinopyroxene phenocrysts from approximately 32 to 51. The olivine phenocrysts show much lower values and range from approximately 4.5 to 8. No trend or correlation can be seen between Y and any of the trace elements in the olivine phenocrysts, but a positive trend is visible in the clinopyroxene.

A trend between the silicate rocks and the clinopyroxene phenocrysts may be inferred for Th, but for Nb, Ba, La, and Sm the silicate rocks have trace element values comparable to the olivine phenocryst, but higher Y-values. Nb, Sm, and La appear to trace variations from the olivine phenocrysts to the clinopyroxene phenocrysts.

Figure 10 (continued on next page). Plots showing selected trace elements (y-axis) versus Y (x-axis).

The plots follow the same colour scheme as figure 8 and 9, and as such clinopyroxene phenocrysts are plotted as green triangles, olivine phenocrysts as purple triangles, and the whole rock trace element data from Magnusson (2016) are plotted as grey squares.