Magmatic and volcanic processing of
volatile and chalcophile elements

Olivia Rhiannon Hogg

Department of Earth Sciences

University of Cambridge

This thesis is submitted for the degree of

Doctor of Philosophy

Queens’ College

September 2024





Declaration

This thesis is the result of my own work and includes nothing which is the outcome of work
done in collaboration except as declared in the preface and specified in the text. It is not substan-
tially the same as any work that has already been submitted, or is being concurrently submitted,
for any degree, diploma or other qualification at the University of Cambridge or any other Uni-
versity or similar institution except as declared in the preface and specified in the text. It does
not exceed the prescribed word limit for the relevant Degree Committee.

Olivia R Hogg

September 2024

ii





Acknowledgements

I am deeply grateful to Dr Mary Benton and Prof Jon Blundy for encouraging me to pursue a
PhD. It has well and truly been the right path and a rewarding challenge.

Thank you, Marie, for your steadfast support over the past few years. Thank you for always
being around to call whenever inspiration struck or confidence wavered, and for giving me the
freedom to explore new topics (and countries) far beyond the intended scope of this PhD. I have
learnt so much from your supervision and guidance, and best of all, the journey has been so
much fun. I am truly grateful. Fran, thank you for being a sounding board and providing a
fresh perspective to help untangle the chaos of trace element data. Nick, I want to thank you
especially. From the first day I stepped into Cambridge, you welcomed me with open arms.
You have not only been a great friend and colleague, but an excellent mentor. Whether its chats
about code, copper, or coffee, you are always there to listen to me harp on!

There are many people to thank who contributed to this thesis. Cees Jan de Hoog, for your
assistance with SIMS; Barbara Kunz, for your help and our insightful conversations about LA-
ICP-MS; and Iris Buisman, for handling my impromptu emails requesting another session on
the probe or SEM to hunt for those elusive sulfides! I also want to express my gratitude to
Paula Antoshechkina for your incredible help and support with using alphaMELTS; over the
last two years it has been invaluable for helping me (and so many others) learn and navigate this
powerful tool for igneous petrologists.

It takes an army to build a village and the mez to compile a PhD. Thank you to so many people
who have shared wisdom, food and laughs of the past four years – especially Hassan, Jack,
Norb, Sarah Shi, Lizzy, Sara, Charlie, Dee, Kit, and Devesh. To Sylvia and Tatti, the first faces
I see each day – thank you for always having the kettle on early and brightening up my day. I
will greatly miss our mornings together! Carrie and Ayesha – where do I even begin? You’ve
survived four years of listening to me talk endlessly about brines and volcanoes, often without
much context or notice. Your support and office armchair made all the difference, and will be
utterly missed. Some days research can feel like one step forward and ten steps back, but you
guys knew how to lighten the mood when it felt heavy. Library Sarah, thank you for introducing

iv



me to ’when taken’ and woollen swimming togs, for lack of a better term, they’ve changed my
life. And to friends beyond the four walls of Downing Site – Shauna, Georgia, Serena, Maria,
Alice, Ashley, and Beibs – thank you for helping me forget about magma, metals, and melt
inclusions on weekends, and for keeping me sane while starting a PhD during a pandemic.

Sometimes work stayed at the office, but more often than not it followed me home. Sonny
and my family, thank you for your endless patience and advice, and for nodding along to my
constant stream of ideas. You’ve been my biggest cheerleaders, and I truly appreciate you.

v



I dedicate this thesis to the Hoggs. You said do what you love, and it worked!

“Live in the now”
- Garth Algar, Waynes World

“The best mountains have no peak”
- Alex Hormozi

vi





Publications

Publications arising from this thesis

Chapter 4: Hogg, O.R., Edmonds, M. and Blundy, J., 2023. Water-rich magmas optimise
volcanic chalcophile element outgassing fluxes. Earth and Planetary Science Letters, 611,
p.118153.

Chapter 5: Hogg, O.R., Edmonds, M., Wieser, P.E., Gleeson, M., Jenner, F. and Blundy, J.,
(in review). Sulfide resorption by water-rich melt yields copper-rich magmatic fluids. Nature
Communications.

Chapter 6 is being prepared for submission.

Co-author publications using work presented in this thesis

Edmonds, M., Mason, E. and Hogg, O., 2022. Volcanic outgassing of volatile trace metals.
Annual Review of Earth and Planetary Sciences, 50 (1), pp.79-98.

viii



Supplementary Data

Some of the data presented in this thesis is supplied as an Electronic Appendix. There are
three folders – Electronic Appendix A, B, and D – containing multiple excel files with several
spreadsheets. Each file contains a ‘readme’ sheet which describes the data and any abbreviations
included in each excel file. Excel files are named according to the following structure: Appendix
Letter – Number of file – Brief description (e.g., ‘B1 – Models and Data Compilations.xlsx’).
When applicable, references are made to the Electronic Appendix and corresponding data table
number.

ix



Abstract

Chalcophile elements possess significant economic and environmental value. The transition
to an electric economy has led to an unprecedented rise in the demand for critical metals, in-
cluding several chalcophile elements such as copper, selenium, and silver. Copper in particular
is essential – it is a key component in wind and solar technologies, as well as energy storage
systems, all of which are central to the energy transition. However, with existing ore grades de-
clining and newer resources harder to find, meeting the growing global demand for metals poses
a significant challenge to current supply capacity. Over 70% of global copper is derived from
porphyry copper deposits (PCDs) associated with hydrous oxidised calc-alkaline arc magmas
at convergent margins. It is not uncommon for these deposits to generate other critical element
byproducts like selenium, tellurium, and bismuth, which in the current economic climate also
face surging demands. Despite widespread recognition of the role that magmas play in the for-
mation of hydrothermal ore deposits, there is still no consensus on the relative importance of
the various magmatic processes involved.

An important and undisputed step in the process is the generation of a hydrothermal saline
magmatic fluid with a proclivity to carry high quantities of metals. These fluids exsolve at depth
from fractionating magmas and are subsequently conveyed to sites of mineralisation. The con-
ditions that optimise the masses and concentrations of copper and other chalcophile elements
partitioning in to these fluids are not fully understood. By definition, chalcophile elements have
a strong affinity for sulfur, such that in magmatic systems they fractionate into precipitating
sulfide phases. Several chalcophile elements are also highly volatile, therefore will also parti-
tion into exsolving magmatic fluids. Degassing and sulfide saturation occur ubiquitously during
magma evolution, yet the impact that these processes exert on the abundance and distribution
of copper in magmatic systems, and more specifically in exsolving magmatic fluids, remain
unclear.

Exsolved magmatic fluids are not only important within the crust, but also play a critical
role in surface environments: these fluids may advect to the surface and manifest as volcanic
gas plumes, which emit vast quantities of chalcophile elements into the atmosphere. These ele-
ments exist primarily as aerosols or particulate matter that eventually settle out of the plume and

x



into the surface environments. Over a narrow interval, these elements transition from serving as
essential nutrients to becoming toxic pollutants, highlighting their significant environmental im-
plications. Metal assemblages in volcanic gas plumes vary systematically with tectonic setting.
Arcs tend to be more enriched in lead, thallium, and bismuth compared to hotspots. How-
ever, despite their distinct metal fingerprints, the concentrations and mass fluxes of outgassing
metals can differ by several orders of magnitude even within individual arcs. What controls
the distribution and abundance of volatile and chalcophile elements in magmatic and vol-
canic systems? I explore this fundamental question by combining natural geochemical datasets
with numerical models of degassing and sulfide saturation during fractional crystallisation and
decompression. The commonality that emerges across the work contributing to this thesis is
the importance of magma water concentrations on the fate of volatile chalcophile elements in
magmatic and volcanic environments.

xi



Contents

1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The behaviour of chalcophile elements . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Introduction to controls on Cu systematics in magmatic systems . . . . . . . . 7
1.4 Key research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Thesis aims and structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Methods 11
2.1 Geological setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.2 Acquisition procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.3 Data quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.4 Secondary Ion Mass Spectrometry (SIMS) . . . . . . . . . . . . . . . . . . . . 17
2.4.1 Acquisition procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2 Calibration and data quality . . . . . . . . . . . . . . . . . . . . . . . . 18

2.5 Electron Probe Micro-Analyser (EPMA) . . . . . . . . . . . . . . . . . . . . . 20
2.5.1 Analytical procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.2 Data quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5.3 Oxide spot analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.6 Laser Ablation Inductively Coupled Mass Spectrometry (LA-ICP-MS) . . . . . 26
2.6.1 Acquisition parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.6.2 Data processing and quality . . . . . . . . . . . . . . . . . . . . . . . . 28

2.7 Post-entrapment crystallisation corrections . . . . . . . . . . . . . . . . . . . . 29

3 Modelling crystallisation, degassing and sulfide saturation in magmas 33
3.1 Degassing of volatile chalcophile elements during fractional crystallisation and

decompression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.1 Partitioning behaviours of chlorine and groups of chalcophile elements . 34
3.1.2 Modelling isobaric crystallisation and second boiling . . . . . . . . . . 35
3.1.3 Modelling decompression and first boiling . . . . . . . . . . . . . . . . 38

3.2 Degassing and sulfide saturation of chalcophile elements during fractional crys-
tallisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Modelling fractional crystallisation and degassing of H2O and CO2 . . 42
3.2.3 Modelling degassing of sulfur, chlorine and chalcophile elements . . . . 42
3.2.4 Modelling sulfide saturation . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.5 Sulfide-silicate melt partitioning of chalcophile elements . . . . . . . . 49

xii



Contents

3.2.6 Modelling sulfide assimilation by hydrous magmas . . . . . . . . . . . 57
3.3 Degassing and sulfide saturation of chalcophile elements during decompression 59

3.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.2 Modelling degassing and crystallisation during ascent . . . . . . . . . . 59
3.3.3 Modelling degassing of sulfur, chlorine and chalcophile elements at Yasur 60
3.3.4 Modelling sulfide saturation and sulfide-silicate melt partitioning at Yasur 61
3.3.5 Modelling isobaric crystallisation at Yasur . . . . . . . . . . . . . . . . 63

4 Water-rich magmas optimise volcanic chalcophile element outgassing fluxes 65
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2 Global ore-forming fluid and volcanic gas datasets . . . . . . . . . . . . . . . . 68
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.4.1 Models of degassing of trace metals during second boiling and crystalli-
sation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.4.2 Models of decompression degassing of trace metals . . . . . . . . . . . 75
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

4.5.1 Comparison of models with data from global volcanoes . . . . . . . . . 79
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

5 Sulfide resorption by water-rich melt yields copper-rich magmatic fluids 82
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Modeling chalcophile element behaviour in arc magmas . . . . . . . . . . . . . 85
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Melt interaction with sulfide cumulates generates copper-rich fluids . . . . . . . 88
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6 Decompression crystallisation forms copper-rich melt signatures in arc magmas 91
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.2 Sample preparation and methods . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6.3.1 Petrography and host geochemistry . . . . . . . . . . . . . . . . . . . . 102
6.3.2 Melt inclusion geochemistry: major elements . . . . . . . . . . . . . . 103
6.3.3 Melt inclusion geochemistry: volatile elements . . . . . . . . . . . . . 106
6.3.4 Melt inclusion geochemistry: Cu and Ag . . . . . . . . . . . . . . . . . 108

6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.4.1 Cu diffusion into melt inclusions through phenocryst hosts . . . . . . . 110
6.4.2 Isobaric crystallisation and sulfide saturation cannot reproduce Yasur

melt inclusion data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.4.3 Decompression crystallisation and degassing generates Cu-rich melts . 115
6.4.4 Sulfur degassing maintains the melt at sulfide saturation during decom-

pression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

7 Conclusions and future work 120
7.1 Watery magmas for chalcophile element-rich magmatic and volcanic volatile

phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2 An underappreciated role of degassing in magmatic Cu systematics . . . . . . . 121
7.3 The origins of Cu-enriched magmas . . . . . . . . . . . . . . . . . . . . . . . 123

xiii



List of Figures

1.1 Selected chalcophile element fluid-melt and sulfide-melt partition coefficients . 4
1.2 Schematic to emphasise the distinction between masses and concentrations for

fluid-melt partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1 Bathymetric map of the Vanuatu Arc . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Photographs of epoxy mounts and scoria samples . . . . . . . . . . . . . . . . 14
2.3 CO2 density (g/cm3) versus diad splitting (∆, cm−1) . . . . . . . . . . . . . . . 17
2.4 Calibration curve for CO2 and H2O measurements acquired during SIMS analysis 19
2.5 Comparing uncorrected MgO concentrations acquired by EPMA to those de-

rived from SIMS measurements on melt inclusions. . . . . . . . . . . . . . . . 19
2.6 Fe-Ti oxide compositions derived by stoichiometry and charge balance . . . . . 25
2.7 Ag-line scan signal in an olivine-hosted melt inclusion . . . . . . . . . . . . . 28
2.8 Rhodes diagram for assessment of olivine-melt equilibrium . . . . . . . . . . . 31

3.1 Comparison of the fluid-melt partitioning behaviours of theoretical trace elements 38
3.2 Harker diagrams showing the impact of magma water concentration on frac-

tional crystallisation paths over a range of pressures . . . . . . . . . . . . . . . 43
3.3 Relative volatility and sulfide affinity of Cu and Ag with respect to the silicate

melt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Fractional crystallisation models run for different initial S and Cl concentrations 47
3.5 Changes to melt SCSS2−, SCAS and S6+/Stotal during fractional crystallisation

in buffered models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.6 Evaluation of different chalcophile element sulfide liquid-silicate melt partition-

ing parametrisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7 Comparison of different published DCu

SL/melt and DAg
SL/melt parametrisations and

experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.8 The effect of magma water concentration on Cu and S masses and concentra-

tions in fractionating melts and their exsolving fluids . . . . . . . . . . . . . . 58
3.9 Fluid-melt partition coefficients of S, Cl and Cu modelled during decompression 61
3.10 Modelled changes in melt SCSStot and FeOt with MgO and pressure . . . . . . 63

4.1 The composition of volcanic gas/aerosol plumes compared with the composi-
tion of deeper exsolved fluids, represented by fluid inclusions hosted by quartz . 71

4.2 Mass fluxes (kg/day) of trace metals carried via gas and aerosol plumes at three
arc volcanoes and one hotspot. . . . . . . . . . . . . . . . . . . . . . . . . . . 73

4.3 The behaviour of Cl and trace metals during second boiling . . . . . . . . . . . 76
4.4 The behaviour of Cl and trace metals during first boiling . . . . . . . . . . . . 78
4.5 Modelled trace metal mass fluxes compared to natural volcanic gas data from

Etna and Yasur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5.1 Global whole rock volcanic arc rock compositions showing Cu and FeOt . . . . 84

xiv



List of Figures

5.2 Modelling the effect of magma water concentration and pressure on the mass
distribution of Cu in the melt-fluid-sulfide system . . . . . . . . . . . . . . . . 87

5.3 The effect of sulfide resorption on Cu mass and concentration in melts and fluids 90

6.1 Global arc melt inclusion (MI) and volcanic whole rock (WR) FeOt, and Cu
concentrations categorised by magma composition . . . . . . . . . . . . . . . . 93

6.2 Transmitted and reflected light images of an olivine-hosted melt inclusion and
BSE map of a scoria thin section from Yasur . . . . . . . . . . . . . . . . . . . 96

6.3 Composition of melt inclusions, matrix glasses and adjacent host phenocryst in
samples from Yasur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6.4 Core and rim compositions of olivine, clinopyroxene and plagioclase phenocrysts
measured by EPMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6.5 Major element composition of melt inclusions, whole rocks and matrix glasses
from Yasur volcano and the Vanuatu Arc . . . . . . . . . . . . . . . . . . . . . 106

6.6 Volatile concentrations (H2O, CO2, S and Cl) in melt inclusions and matrix
glasses from Yasur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

6.7 Insights to Cu, S and Cl systematics in Yasur magmas from melt inclusions and
whole rocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

6.8 Assessing the impact of melt inclusion post-entrapment Cu diffusion on global
datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

6.9 Modelled fractional crystallisation trajectories for Yasur magmas . . . . . . . . 114
6.10 Modelled decompression of an oxidised hydrous basalt generates high Cu con-

centrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

B.1 P-T-X evolution of intermediate-low salinity magmatic hydrothermal fluids . . 152
B.2 The behaviour of Cl and trace metals during second boiling with initial melt

concentrations of 20 ppm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
B.3 The behaviour of Cl and trace metals during second boiling with initial melt

concentrations of 100 ppm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
B.4 The behaviour of Cl and trace metals during first boiling, with initial metal

concentrations of 20 ppm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
B.5 The behaviour of Cl and trace metals during first boiling, with initial metal

concentrations of 100 ppm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
B.6 Modelled trace metal mass fluxes compared to natural data from Etna and Yasur

volcano . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
C.1 Relative fluid-melt and sulfide-melt partition coefficients for selected chalcophile

elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
C.2 Model results for melt ratios of Cu/Ag, Cu/Re, Cu/Au, Cu/Se, assuming sulfide

is present only as SL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
C.3 Model results for melt ratios of Cu/Ag, Cu/Re, Cu/Au, Cu/Se, assuming sulfide

transitions from SL to MSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
C.4 Modelled melt Cu/Ag ratios for initial S concentrations of 1000 and 2000 ppm,

for systems of differing magma water concentrations . . . . . . . . . . . . . . 161
D.1 EDS map for inspection of sulfide phases in Yasur scoria samples . . . . . . . . 162
D.2 Inspection of PEC in plagioclase- and clinopyroxene-hosted melt inclusions

from Yasur, this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
D.3 Selected major element, Cl, S and Cu concentrations for melt inclusion and

whole rock data along the Vanuatu Arc . . . . . . . . . . . . . . . . . . . . . . 164

xv



List of Figures

D.4 Selected trace element ratios for melt inclusion and whole rock data along the
Vanuatu Arc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

D.5 Selected trace element concentrations analysed in melt inclusions from Yasur in
this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

D.6 Selected chalcophile element concentrations analysed in melt inclusions from
Yasur in this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

D.7 Comparing the global distribution of CA and TH whole rocks and melt inclusions168
D.8 Cooks Distance plot to assess the influence of individual data points on an

ANOVA test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
D.9 Tukey-HSD results from an ANOVA test on groups of phenocryst-hosted melt

inclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

xvi



List of Tables

2.1 Summary of the mean concentration, precision, and accuracy for different
standards and elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Beam conditions for spot analyses of phenocryst hosts, melt inclusions and
oxides measured over three analytical sessions. . . . . . . . . . . . . . . . . 21

2.3 Precision and accuracy estimates for measurements in session 1 . . . . . . . 23
2.4 Precision and accuracy estimates for measurements in session 2 . . . . . . . 24
2.5 Precision and accuracy estimates for measurements in session 3 . . . . . . . 25
2.6 LA-ICP-MS analytical conditions for spot and line scan analyses . . . . . . 26

3.1 Input parameters of the five magmatic systems, A, B, C, D, and E used to
model the development of an MVP . . . . . . . . . . . . . . . . . . . . . . 34

3.2 Input parameters of the five magmatic systems, A*, B*, C*, D*, and E* used
to model the development of an MVP during decompression . . . . . . . . . 39

3.3 Average DCl
MVP/melt of trachybasalts (modelled as a melt fractionated to F=0.5) 40

3.4 Starting parameters used for fractional crystallisation models that are dis-
cussed in Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.1 Major, minor, volatile and trace element concentrations of melt inclusions
from Yasur. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

A3.1.1 Acquisition conditions for EPMA spot analyses on plagioclase measured in
Session 1. All run with a 15 kV and 10 nA defocused beam with 5 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

A3.1.2 Acquisition conditions for EPMA spot analyses on clinopyroxene measured
in Session 1. All run with a 15 kV and 20 nA defocused beam with 5 µm
spot size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

A3.1.3 Acquisition conditions for EPMA spot analyses on olivine measured in Ses-
sion 1. All run with a 15 kV and 40 nA defocused beam with 1 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

A3.1.4 Acquisition conditions for EPMA spot analyses on glass measured in Ses-
sion 1. All run with a 15 kV and 10 nA defocused beam with 10 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

A3.1.5 Acquisition conditions for EPMA spot analyses on plagioclase measured in
Session 2. All run with a 15 kV and 10 nA defocused beam with 5 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

A3.1.6 Acquisition conditions for EPMA spot analyses on clinopyroxene measured
in Session 2. All run with a 15 kV and 20 nA defocused beam with 5 µm
spot size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

A3.1.7 Acquisition conditions for EPMA spot analyses on olivine measured in Ses-
sion 2. All run with a 15 kV and 40 nA defocused beam with 1 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

xvii



List of Tables

A3.1.8 Acquisition conditions for EPMA spot analyses on glasses measured in Ses-
sion 2. All run with a 15 kV and 20 nA defocused beam with 5 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

A3.1.9 Acquisition conditions for EPMA spot analyses on oxides measured in Ses-
sion 3. All run with a 15 kV and 20 nA defocused beam with 1 µm spot
size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

A3.2.1 Detection limits for materials measured in Session 1. . . . . . . . . . . . . . 149
A3.2.2 Detection limits for materials measured in Session 2. . . . . . . . . . . . . . 150
A3.2.3 Detection limits for materials measured in Session 3. . . . . . . . . . . . . . 151
D.1 Mean Cu concentrations (ppm) for tholeiitic (TH) and calc-alkaline (CA)

compositions of whole rocks and melt inclusions. . . . . . . . . . . . . . . 168
D.2 Statistical test results for global arc whole rock (TH, CA) and melt inclusion

(TH, CA) datasets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
D.3 Results of the ANOVA test comparing mean Cu concentrations in arc melt

inclusions (N=372) across different phenocryst hosts. . . . . . . . . . . . . 169

xviii



1
Introduction

1.1 Motivation

The significance of chalcophile elements are twofold: they possess economic and environmen-
tal value, and serve as effective tracers of magmatic and volcanic processes. The transition to a
low-carbon future relies heavily on electricity generation via renewable sources, which has cre-
ated and will continue to generate an unprecedented demand for critical metals. These metals,
defined by their economic importance and risk to supply, include several chalcophile elements
such as copper (Cu), silver (Ag), and selenium (Se). Previous research has extensively inves-
tigated the metallogenesis of critical metal deposits and the processing of associated minerals.
However, the fundamental prerequisites for their formation are still debated (Richards, 2011;
Park et al., 2021; Kouzmanov and Pokrovski, 2012; Rezeau and Jagoutz, 2020). Chalcophile
elements fractionate and become enriched in magmatic and hydrothermal environments by par-
titioning between the silicate melt, sulfide, and fluid reservoirs (Reekie et al., 2019; Jenner
et al., 2015). Partitioning is dependent on several factors that change during the evolution of the
magmatic-hydrothermal system (Kiseeva and Wood, 2015; Zajacz et al., 2008).

The work presented in this thesis focuses on one particular chalcophile element: Cu. Copper
is an indispensable component of wind and solar technologies and energy storage. These char-
acteristics that are central to the energy transition attributed to its high electrical conductivity, its
ductility, and its relatively high efficiency and recyclability (Seck et al., 2020; Gu et al., 2024).
To help alleviate the demand on individual metals, some critical metals can be substituted for

1



1.1. Motivation

those with similar properties, e.g., lithium (Li) for sodium (Na) (Nayak et al., 2018). However,
no green substitute exists for Cu, and with existing ore grades declining and newer resources
harder to find, this poses a significant challenge to surmounting the supply capacity limitations.

Over 70% of global Cu supply is sourced from porphyry deposits associated with silicic
volatile-rich oxidised calc-alkaline arc magmas at convergent margins (Blundy et al., 2015).
These deposits are often found associated with other critical element byproducts, like Se, Te,
and Bi, which also face surging demands in the current economic climate (McNulty et al.,
2022). While multiple factors contribute to the formation of porphyry copper deposits, such as
tectonic regime (Loucks, 2021), fluid focusing (Park et al., 2021), long-lived thermal sustain-
ability (Chiaradia and Caricchi, 2017; Chelle-Michou et al., 2017) and precipitation efficiency
(Chiaradia, 2020), an important step is the generation of a hydrothermal metal-enriched saline
fluid (Hedenquist and Lowenstern, 1994; Sillitoe, 2010; Richards, 2015). Aqueous magmatic
fluids exsolve from magmas fractionating at depth in the crust and convey metals to areas of
mineralisation and precipitation (Sillitoe, 2010; Lee and Tang, 2020; Richards, 2021). But the
role that precursor magmatic sources and crustal processes institute on the composition and
abundance of these fluids is not well constrained (Richards, 2015; Audétat and Simon, 2012).
Of particular debate is the relative importance of sulfide saturation and degassing on the fate of
Cu and other chalcophile elements, as both work to deplete melts in Cu (Jenner et al., 2015).

The value of chalcophile elements transcends their mere association with ore deposits. Vol-
canoes emit huge plumes of gases and aerosols into the atmosphere, comprised of water (H2O),
carbon dioxide (CO2), sulfur (S), halogens (Cl and F) and a suite of volatile chalcophile ele-
ments (Cu, Pb, Zn, Bi, Tl, As and Se), at rates comparable to the industrial emissions of entire
countries (Ilyinskaya et al., 2021; Nriagu, 1989). Eventually, these elements settle out of the
plume and deposit into surface environments (Dupont et al., 2010), and transition from critical
nutrients to toxic pollutants within a very narrow range, which thereby poses substantial en-
vironmental and biological risk (Floor and Román-Ross, 2012). These volcanic gases derive
from exsolved magmatic fluids in the crust, which are crucial for ore formation in arc settings,
underscoring the integral role volcanoes play in the global cycling of chalcophile elements.
Plume-derived metal assemblages vary systematically across tectonic settings; arcs are partic-
ularly enriched in volatile elements that speciate with chloride (Pb, Bi, Tl), while rifts often
have high concentrations of elements that form sulfides and hydrides, or exist as free atoms
(Te and Se) (Edmonds et al., 2018; Zelenski et al., 2021). The concentration and mass flux
of chalcophile elements varies several orders of magnitude, yet the underlying controls remain
unclear. What influences the composition and abundance of chalcophile elements in fluid and
gaseous reservoirs?

2



Chapter 1. Introduction

1.2 The behaviour of chalcophile elements

Chalcophile elements (including Au, Ag, Bi, Cu, Pb, Se, Re, and Zn) were originally defined by
Goldschmidt (1937), as those with an affinity for sulfur. In magmatic systems, this behaviour is
demonstrated upon fractionation of a dense sulfide phase. The amount of sulfide that can dis-
solve in the melt is described by the sulfide solubility (SCSS, see Chapter 3.3 and 5 for details)
and depends on several factors, including melt FeO (O’Neill, 2021; Jugo et al., 2010; Smythe
et al., 2017), oxidation state (Jugo et al., 2010; Nash et al., 2019; O’Neill and Mavrogenes,
2022) and pressure (Matjuschkin et al., 2016; Cox et al., 2019). In sulfide-saturated conditions,
chalcophile elements partition from the silicate melt into dense sulfides, e.g., CuFeS2, which
may become sequestered in the crust (Jenner et al., 2010). The proportion of these elements
partitioning out of the silicate melt is linked to their affinity for the sulfide phase (Reekie et al.,
2019), which can be quantified using Nernst style partition coefficients that expresses the con-
centration ratio of an element (X) between two coexisting phases at equilibrium:

DX
sulfide/melt = CX

sulfide/CX
melt (1.1)

where DX
sulfide/melt is the partition coefficient, CX

sulfide is concentration of X in the sulfide and
CX

melt is the concentration of X in the silicate melt. DX
sulfide/melt > 1 represents an element with

a preference for the sulfide phase, hence by definition, describes most chalcophile elements
(Figure 1.1). Sulfides in arc settings initially take the form of sulfide liquid (SL) (Nadeau et al.,
2010) but transform to monosulfide solid solution (MSS) at lower temperatures (Chang et al.,
2018; Keith et al., 2017; Li and Audétat, 2013). This transition itself may influence the affinity
of these elements for the sulfide phase (Figure 1.1). Some elements (Au, Ag, Cu) are highly
chalcophile (with partition coefficients between SL-silicate melt (DSL/melt) and MSS-silicate
melt (DMSS/melt) reaching > 104), thus theoretically, will become more depleted in the melt
compared to weakly chalcophile elements like Pb and Zn (DSL/melt an DMSS/melt up to 102)
(Li and Audétat, 2015; Li et al., 2021) (Figure 1.1). Our knowledge of chalcophile element
partitioning between the various forms of sulfides and silicate melts derives from extensive
work on natural and experimental samples (Li et al., 2021; Li and Audétat, 2015; Kiseeva and
Wood, 2013; Brenan, 2015; Patten et al., 2013). However, behaviours of chalcophile elements,
such as Cu and Ag, during sulfide saturation in an evolving magmatic system are far better
constrained than for other chalcophile elements, such as Se and Te.

3



1.2. The behaviour of chalcophile elements

Figure 1.1: Selected chalcophile element partition coefficients for DSL/melt (pink), DMSS/melt

(red), and Dfluid/melt (blue). Ranges in partition coefficients are taken from natural and exper-
imental data (Zajacz et al., 2008; Zelenski et al., 2021; Li and Audétat, 2015; Li et al., 2021;
Kiseeva and Wood, 2013; Brenan, 2015; Tattitch and Blundy, 2017b; Pokrovski et al., 2013).
The relative magnitude of fluid/silicate melt and sulfide/silicate melt partition coefficients across
chalcophile elements modeled here should be considered rather than their absolute values. Note
that Cu and Ag have almost identical DSL/melt, but DMSS/melt differs by an order of magnitude,
with Cu being more compatible in crystalline sulfide than Ag (Li and Audétat, 2015; Jenner
et al., 2010). Partition coefficients cover several orders of magnitude that reflect the changing
conditions of the magmatic system. For DSL/melt and DMSS/melt, this can reflect melt tempera-
ture and composition, oxidation state, and pressure. Dfluid/melt is also affected by melt composi-
tion, pressure, and temperature (Zajacz et al., 2008; Zelenski et al., 2021; Tattitch and Blundy,
2017b). For Cl-speciating elements like Cu, Pb, and Bi, Dfluid/melt increases as fluid salinity
increases (Zajacz et al., 2008). The limited range in partition coefficients plotted for Se reflects
the limited data available. Experiments show DCu

fluid/melt reach up to 300 (Pokrovski et al., 2013),
but these values relate to hypersaline brines rather than single-phase fluids, for which DCu

fluid/melt

is typically < 100 (Zajacz et al., 2008; Audetat et al., 2008).

Many chalcophile elements, despite being defined by their S-loving tendencies, are also
relatively volatile (Pokrovski et al., 2013; Mason et al., 2024; Zelenski et al., 2021) (Figure 1.1);
therefore, in volatile-saturated magmas will partition into exsolving aqueous fluids. Predicting
the fate of these elements in magmatic systems necessitates an understanding of their dual
behaviours in order to trace key magmatic processes, such as sulfide saturation and degassing,
which occur ubiquitously in evolving magmatic systems (Jenner et al., 2015). As described
previously, the relative proportion of an element partitioning into the fluid is governed by its
affinity for the fluid phase relative to the silicate melt, which is represented by a fluid-melt
partition coefficient:

4



Chapter 1. Introduction

DX
fluid/melt = CX

fluid/CX
melt (1.2)

where DX
fluid/melt is the partition coefficient that expresses the concentration ratio of an el-

ement in the fluid (CX
fluid), versus the silicate melt (CX

melt). For a given system, the onset of
degassing depletes the melt more extensively in Se, Cu and Bi (with DX

fluid/melt on the order of
101 to 102) relative to Pb and Zn (with DX

fluid/melt between 10−1 and 101). However, there are
knowledge gaps in our understanding of how elements like Se behave during magma degassing
(Zelenski et al., 2021). Some chalcophile elements (e.g. Cu, Ag, Re, Pb, Zn) are primarily
Cl-complexing, so their partitioning behaviour between the melt and fluid phases can be linked
to Cl where DX

fluid/melt (or fluid affinity) increases as fluid salinity increases (Figure 1.1) (Iveson
et al., 2019; Webster et al., 2018; Zajacz et al., 2008; Dolejš and Zajacz, 2018). Chlorine par-
titions most strongly into the fluids exsolving from evolved magmas at high pressures (Tattitch
et al., 2021). Some chalcophile elements (e.g. Se and Au) speciate with OH− and S ligands
(Zajacz et al., 2013, 2008); how these elements behave at different pressures, temperatures, melt
compositions and oxidation states are less well constrained (Zajacz et al., 2012; Zelenski et al.,
2021). Figure 1.1 shows that some chalcophile elements (Pb, Zn, Bi) display similar affinities
for both sulfides and fluids with respect to the melt (Edmonds et al., 2018), whereas others (Cu,
Ag, Se) show a strong preference for the sulfide phase. To a first order, this suggests that mag-
mas depleted in Cu, Ag and Se largely reflect the onset of sulfide saturation, whereas for Pb,
Zn, and Bi, evidence of degassing and sulfide saturation in the rock record are convolved. Can
these features of chalcophile element behaviour explain global chalcophile element systemat-
ics in volcanic rocks and gases, or do other factors beyond simple geochemical affinity exert a
greater influence?

Partition coefficients, such as those introduced in equations 1.1 and 1.2, describe the ratio
of concentrations of a particular element in two reservoirs (e.g. sulfide and melt) but do not
indicate the total mass of the element contained within these reservoirs directly. Although the
concentration of ligands, particularly Cl, in magmas and fluids influence the partitioning of
certain chalcophile elements, the fluids are dominated by H2O (+ CO2) (Audétat and Edmonds,
2020). Exsolved magmatic fluids are generated during crystallisation and decompression where
the mass of exsolved fluids increases as the concentration of dissolved volatiles exceeds the
solubility of the melt during second and first boiling, respectively (Newman and Lowenstern,
2002). Magmas with initially higher concentrations of water will start exsolving a volatile
phase at relatively higher melt fractions, from a higher mass of melt (Rezeau and Jagoutz,
2020), thereby partitioning potentially higher masses of chalcophile elements into exsolving
fluids (Figure 1.2). For a chalcophile element with a fixed DX

fluid/melt , magmas exsolving greater

5



1.2. The behaviour of chalcophile elements

masses of fluids will partition higher masses (but equal concentrations) of metals into the fluid
reservoir, compared to more water-poor systems, as demonstrated in Figure 1.2. Therefore, not
only is fluid-melt and sulfide-melt partitioning an important control on the abundance and dis-
tribution of chalcophile elements in magmatic systems, but it also plays a key role in mediating
the absolute masses of these elements entering the fluid and sulfide reservoirs (Figure 1.2).

This principle also holds for the precipitation of sulfides. For Cu, Ag and Au, which are
highly chalcophile and just moderately volatile, the mass transfer of these elements between the
melt-fluid-sulfide system during fractionation or decompression of a sulfide-volatile-saturated
melt, will also depend on the mass of these reservoirs. Understanding the conditions that max-
imise elemental masses and concentrations in magmatic reservoirs requires consideration of the
impact of simultaneous degassing and sulfide saturation on the fate of moderately to highly
volatile chalcophile elements, such as Cu, in magmatic systems.

Figure 1.2: Schematic to emphasise the distinction between masses and concentrations. Box
sizes represent absolute masses of the reservoirs. Two systems of identical mass but one with
high magma H2O concentration and the other with lower. Both systems have the same dissolved
concentration of a metal (Mx) in the melt which has a fixed DX

fluid/melt in both systems. As
degassing initiates and a fluid reservoir is established, the proportion and concentration of Mx

partitioning into the fluid phase is governed by DX
fluid/melt, hence is identical in both systems.

However, the absolute mass of the element in the fluid phase differs and is ultimately controlled
by the concentration of H2O in the system.

6



Chapter 1. Introduction

1.3 Introduction to controls on Cu systematics in magmatic
systems

Global arc volcanic rock data generally show that Cu behaves incompatibly in tholeiitic (TH)
magmas reaching up to 300 ppm in intermediate compositions and compatibly in calc-alkaline
(CA) magmas. Some research suggests the differences in Cu behaviour are controlled primarily
by the thickness of the overriding crust, which implicates the fractionation trends and oxidation
state of evolving magmas (Chiaradia, 2014). Thick crusts promote high pressure fractionation
of phases like garnet and amphibole, which in turn, depletes concentrations of FeO in the melt
(Lee and Tang, 2020). Others suggest that dissolved magma water concentrations have a princi-
pal control on magma fractionation sequences (Sisson and Grove, 1993; Richards, 2015; Barber
et al., 2021), with more hydrous magmas promoting higher temperature fractionation of Fe-rich
phases, like amphibole, resulting in Fe-poor calc-alkaline magmas. Regardless, a drop in melt
FeO endorses sulfide saturation by lowering the sulfide solubility of the melt (SCSS2−), result-
ing in the generation of Cu-poor calc-alkaline magmas (Chiaradia, 2014). However, sustained
Fe-enrichment in tholeiites is able to suppress sulfide saturation and form Cu-rich magmas
(Chiaradia, 2014; Lee et al., 2012; Richards, 2015).

Concentrations of water dissolved in primitive arc magmas can vary from <1 to >6 wt%
(Klein et al., 2023). It is widely recognised that sulfide saturation prior to degassing sequesters
Cu into sulfides, preventing Cu from entering later exsolving fluids (Jenner et al., 2015; Cox
et al., 2019). Water-rich magmas will become volatile saturated at higher melt fractions and ex-
solve greater masses of fluids relative to more water-poor counterparts (Klein et al., 2023) and,
moreover, may initiate degassing prior to sulfide saturation. Sulfur is highly volatile, and there-
fore will partition strongly into exsolving fluids. This potentially impacts the ability of the melt
to become (or remain) sulfide saturated, and in turn, influences the abundance and distribution
of Cu in the melt-fluid-sulfide system (Jenner et al., 2015). These processes typically operate
in tandem, making it difficult to determine their relative importance for forming Cu-rich fluids:
can degassing of large masses of fluid (Figure 1.2) outweigh the impact of high sulfide-melt
partition coefficients (Figure 1.1) on chalcophile elements abundances in magmatic reservoirs?
Several models describing Cu systematics in response to sulfide saturation exist (Chiaradia,
2014; Lee et al., 2012; Wieser et al., 2020) but no frameworks to date incorporate the effect of
concomitant degassing.

The aforementioned published models each highlight an enigma: porphyry Cu deposits are
generally associated with Cu-poor volcanic rocks which confronts an ongoing theme in por-
phyry research space, of whether fractionation of Cu-rich dense sulfides are detrimental to the

7



1.4. Key research questions

formation of metal-rich endowments, and more specifically, the generation of a metal-rich fluid.
In the last few decades models have shown that these systems may rely on sulfide formation as
an essential pre-concentration step in deeper crustal reservoirs (Chiaradia, 2014), and that sub-
sequent events such as the infiltration of oxidised hydrous magmas or aqueous fluids may serve
as a mechanism to destabilise these sulfides and transport their Cu and S inventories to sites of
porphyry formation, in the upper crust (Wilkinson 2013; Lee et al. 2012; Lee and Tang 2020;
Heinrich and Connolly 2022, Hogg et al., in review). However, this view on how metal-rich
fluids may form in calc-alkaline magmas is not universally accepted. Other prevailing models
de-emphasise the role that sulfides play, instead highlighting the importance of fluids exsolv-
ing from water-rich oxidised magmas despite early sulfide saturation (Loucks, 2014). It has
been suggested that high pressure (> 500 MPa) fractionation of water-rich magmas are essen-
tial prerequisites for the formation of porphyries by generating an early forming, hypersaline
fluid with the ability to carry large proportions of metals (Loucks, 2021, 2014). Through mod-
elling, others have demonstrated, that deep fractionation is optimal for accumulating high melt
Cu concentrations through volatile-undersaturated fractionation, but argue that later ascent and
degassing of these magmas to lower pressures (<400 MPa) is critical for liberating high masses
of Cu and volatile chalcophile elements into high mass fluid reservoirs (Chiaradia and Caricchi,
2017; Chiaradia, 2020; Chelle-Michou et al., 2017; Richards, 2015).

1.4 Key research questions

Salient research has been done in pursuit of understanding how precursor magmatism aids min-
eralised porphyry development (Rezeau and Jagoutz, 2020; Chiaradia, 2014; Blundy et al.,
2015; Lee et al., 2012; Richards, 2018, 2011; Loucks, 2014; Chiaradia and Caricchi, 2017;
Chelle-Michou et al., 2017; Du and Audétat, 2020), and while these models acknowledge the
role that degassing and sulfide saturation play, the relative importance of these processes re-
mains contentious. Moreover, while these processes account for most global data, there are
subsets of uniquely Cu-rich (>500 ppm) magmas that cannot be explained by degassing and
sulfide saturation during simple fractionation, raising questions about their origins (Deng et al.,
2022; Iveson et al., 2022).

Establishing a comprehensive understanding of the mechanisms controlling the abundance
and distribution of chalcophile elements in magmatic and volcanic systems is foundational to
advancing our knowledge of critical metal deposit formation and the contextualisation of vol-
canic gas plumes. This understanding is paramount for propelling future research and theo-
retical development in these fields. In this thesis, I address the existing knowledge gap by
integrating detailed numerical models with globally synthesised datasets. In detail, I pose the

8



Chapter 1. Introduction

following questions:

1. What conditions (pressure, melt and fluid composition, differentiation processes) opti-
mise the concentration and mass of Cu and chalcophile elements in exsolved magmatic
fluids?

2. What is the relative importance of degassing and sulfide saturation for the formation of
Cu-rich fluids?

3. Are there systematic differences in the distribution of Cu in magmatic reservoirs in dif-
ferent arc environments?

4. Can natural datasets such as melt inclusion, whole rock and volcanic gas compositions
be used to fingerprint these mechanisms?

9



1.5. Thesis aims and structure

1.5 Thesis aims and structure

This thesis consists of three studies designed to explore the principal controls on chalcophile
element abundance and distribution in magmatic and volcanic systems. Chapter 4 examines a
suite of volatile chalcophile elements, while Chapters 5 and 6 provide a detailed analysis of Cu.
This thesis takes the following structure:

Chapter 2 details sample preparation techniques, analytical methodologies and data collec-
tion procedures that were used to analyse the samples that are presented in Chapter 6.

Chapter 3 describes the construction of models presented in Chapters 4, 5 and 6. The first
model simulates degassing and partitioning of selected chalcophile elements during sulfur-free
fractional crystallisation and decompression (Chapter 4). The second and third models replicate
the three-way partitioning of chalcophile elements between melt-fluid-sulfide reservoirs during
fractional crystallisation (Chapter 5), and decompression (Chapter 6).

Chapter 4 establishes the control of Cl versus water concentrations for the composition and
mass flux of volcanic outgassing plumes and crustal magmatic fluids. I identify a decoupling in
the conditions required for optimising chalcophile elements mass fluxes and concentrations in
exsolved magmatic fluids and volcanic gases. This chapter is published as Hogg et al., (2023)
in Earth and Planetary Science Letters.

Chapter 5 combines global natural volcanic datasets with numerical models that simulate
concomitant degassing and sulfide saturation during fractional crystallisation of magmas. The
goal was to investigate the principal controls on the abundance and distribution of Cu in the
crust. I showcase the importance of intensive parameters such as magmatic water concentra-
tions, and magmatic processes like sulfide resorption, for generating Cu-rich fluids. This chapter
is currently in review in the journal of Nature Communications.

Chapter 6 investigates the origins of uniquely Cu-enriched signatures in arc magmas. Global
arc volcanic whole rock data show a trend in Cu and FeO with MgO that is otherwise absent
from the global arc melt inclusion record. I present melt inclusion data from Yasur volcano,
Vanuatu Arc which accommodates some of the highest Cu concentrations recorded by melt in-
clusions globally. Using a combination of geochemical observations and modelling (applying
the framework constructed for Chapter 5) I showcase the potential of degassing and crystallisa-
tion during decompression to facilitate the generation of Cu-rich melts by effectively offsetting
sulfide saturation.

Chapter 7 summarises the conclusions of this thesis and proposes areas of future research.

10



2
Methods

2.1 Geological setting

Throughout this thesis I use Yasur volcano as a case study. The samples presented in Chapter
6 were generated during fountaining eruptions of Mount Yasur, Vanuatu Arc in 2016 (Figure
2.1). Mount Yasur sits on Tanna Island, which belongs to the southernmost segment of the
Vanuatu volcanic arc, in the southwest Pacific (Figure 2.1). The arc is generated from eastward
subduction of the Australian plate beneath the Pacific plate (Figure 2.1) (Allard et al., 2016).
Volcanism along the arc (including Yasur) produces dominantly tholeiitic basalts to basaltic-
andesites that are considered to be water-poor relative to the global range in arc magmatic
water contents (Klein et al., 2023; Cottrell et al., 2021). Multiple studies show that magma
compositions at Yasur have been homogenous for millennia and point towards a long-lived and
singular feeding system (Métrich et al. 2011 and references therein).

Yasur is an open-vent, persistently degassing volcano characterised by frequent strombo-
lian and vulcanian eruptions, although is subject to occasionally more intense eruptions (sub-
Plinian) (Métrich et al., 2011; Woitischek et al., 2020). SO2 fluxes range from 400 to 700 tons
per day which equates to complete degassing of ∼ 0.05 km3 per year of unerupted magma. This
is approximately 50 times the volume of magma erupted at the surface at Yasur (Edmonds et al.,
2022). Only one published study of volcanic gas and aerosol chalcophile element composition
data exists for Yasur (Mandon et al., 2019), as showcased in Chapter 4.

11



2.1. Geological setting

Due to their low abundances in magmas, several chalcophile elements are difficult to analyse
with accuracy in whole rocks, minerals and melt inclusions (Jenner, 2017). At Yasur, Cu con-
centrations in whole rocks have been published by Dupuy et al. (1982), but no published records
exist for melt inclusions. In this thesis I present the first Cu (and Ag) data for melt inclusions
from Yasur, which preserve some of the highest Cu concentrations (500 – 600 ppm) recorded
by the global melt inclusion dataset. In Chapter 6, the composition of phenocryst-hosted melt
inclusions is used to understand the key processes controlling Cu systemics at Yasur and use
these principles to contextualise the global arc Cu array.

Figure 2.1: Bathymetric map showing the Vanuatu Arc in the SW Pacific in a) Volcanic centres
along the arc are coloured according to commonly referenced groups. Solid black line traces
the arc front. Dashed lines represent slab dip contours below the arc (data downloaded from
Hayes et al., 2018). Tanna Island sits in the southernmost block (red square). Inset map of
Tanna Island in b) – red star highlighting scoria sample localities at (19°31.807N,169°27.005E)
in Mount Yasur crater. Map constructed in QGIS.

12



Chapter 2. Methods

2.2 Sample preparation

Samples were collected by L. Melekhova and J. Blundy on August 27th, 2016. Sample prepa-
ration was completed by myself, at the University of Cambridge. Samples of freshly quenched
(0 Ma in age) plagioclase-rich basaltic scoria (Figure 2.2a) from Yasur volcano, Tanna Island
(T1, T2 and T3), were erupted from fire fountains reaching up to 100 m high. Samples were
jaw crushed and sieved into different size fractions (<500, 500–710, 710–1000, <1000 µm)
(Figure 2.2b-c), with the jaw crusher being cleaned between each sample change. Using a
binocular microscope, bulk tephra were hand-picked for olivine, plagioclase and clinopyroxene
phenocrysts of a range of sizes that included glassy, enclosed, bubble-bearing and non-bubble
bearing melt inclusions that were >25 µm. The 710-500 µm and >710 µm fractions contained
the largest proportion of euhedral phenocrysts. Once picked, phenocrysts were mounted indi-
vidually on glass slides using CrystalBondT M and polished using progressively finer grit papers
(9, 6 and 3 µm) until at least one melt inclusion was <30 µm from the surface. Samples were
placed in a sonic bath before switching to different polishing grades in order to minimise risk
of contamination.

Samples hosting vapour-bubbles underwent Raman spectroscopy so that CO2 densities could
be measured in the bubble. Vapour bubble-bearing melt inclusions were extremely rare in the
population of phenocrysts analysed in these Yasur samples. To prepare samples for SIMS anal-
ysis, all melt inclusion-bearing phenocrysts were removed from individually mounted glass
slides by dissolving CrystalBondT M in acetone and subsequently re-mounting crystals in epoxy
(Bueller Epothin2) rounds (Figure 2.2d-e). These mounts were then left to set for 24 hours in an
evac chamber. Epoxy rounds were polished using a Saphir 520 polishing rig with 3 to 1/4 µm
diamond paste until melt inclusions were exposed at the surface (observed using transmitted
and reflected light images). A thin uniform gold coat was then applied to each epoxy round
at the University of Edinburgh prior to SIMS analyses. Once SIMS analyses were complete,
the gold coat was removed by polishing each epoxy stub automatically on the Saphir rig with
1 µm diamond paste. A subsequent carbon coat was added in preparation for Electron Probe
Micro-Analyser (EPMA) and Laser Ablation Inductively Coupled Plasma Mass Spectrometry
(LA-ICP-MS).

13



2.2. Sample preparation

Figure 2.2: Photographs showing examples of a) Yasur scoria collected in 2016, b) petri dish of
crushed Yasur scoria, c) clinopyroxene phenocrysts that were picked for melt inclusion inspec-
tion, d) examples of epoxy moulds used to make the sample mounts and e) an epoxy mount of
phenocrysts hosting melt inclusions that would be coated in gold or carbon for SIMS or EPMA
analysis, respectively.

14



Chapter 2. Methods

2.3 Raman Spectroscopy

2.3.1 Background

The purpose of Raman spectroscopy is to estimate CO2 concentrations in the vapour bubbles
of melt inclusions (Moore et al., 2015). Bubbles commonly develop in melt inclusions in re-
sponse to a series of post-entrapment processes that induce volatile exsolution: cooling causing
differential thermal contraction of the host and melt inclusion (Rose-Koga et al., 2021); post-
entrapment crystallisation on the walls of the melt inclusion reducing the internal pressure and
lowering the solubility of volatiles in the melt (Roedder, 1984; Steele-MacInnis et al., 2011); or
by decompression-induced diffusive H+ loss (Hartley et al., 2014; Gaetani et al., 2012). These
bubbles can host substantial proportions (> 90%) of volatiles, such as CO2 (Hartley et al., 2014)
that were originally dissolved in the melt inclusion, hence failure to reconcile the concentration
of CO2 trapped in bubbles can lead to significant underestimations of magma storage pressures
and the timing of fluid exsolution (Venugopal et al., 2020; Moore et al., 2015; Hartley et al.,
2014). Raman spectra of vapour bubble-bearing melt inclusions were collected using a JY
Horiba LabRam HR (800 mm) Raman spectrometer at the University of Cambridge. LabSpec
software was used to calibrate and acquire spectra. CO2 spectra consist of two peaks, Fermi
Diads, which are situated at 1285 and 1388 cm−1 at 1 bar. The distance between Fermi Diads
(diad splitting) is proportional to the density of CO2 in the vapour bubble, so by converting
the distance between the two peaks, with the aid of a densimeter (Lamadrid et al., 2017), CO2

density can be calculated and subsequently converted to concentration.

2.3.2 Acquisition procedure

Calibrations were performed at the beginning of each analytical session with the LabSpec 6
autocalibration function. A 100 mW, 532.02 nm, argon laser with filter set to 100% and focused
on a x50 LWD objective, was used. Spectra were collected on a CCD detector with confocal
hole diameter of 300 µm, 1800 mm–1 gratings, a slit width of 100 µm and a spectral window
of 1250 cm–1. Initial diad hunts had acquisition times of 30 s and three accumulations that
were averaged, and final quantifications had acquisition times of 90 s and four accumulations
that were averaged. Each vapour bubble analysis was taken three times at progressively deeper
depths in the bubble. CO2 peaks were fitted for each Raman vapour bubble analysis using a
Gaussian peak fitting method developed by Wieser et al. (2021) where the distance between

15



2.3. Raman Spectroscopy

each diad was calculated and converted to density:

pCO2 (g/cm3) = 0.3217∆ ±0.026(cm−1) −32.995±2.7 (2.1)

Concentrations of CO2 in the vapour bubbles were subsequently calculating according to
the method of Tucker et al. (2019):

CVB
CO2

= 106 ×
pCO2 ·VVB

pmelt ·Vmelt
(2.2)

Ctotal
CO2

=CVB
CO2

+Cmelt
CO2

(2.3)

where VVB is the volume of the vapour bubble (cm3); Vmelt is the volume of the melt
inclusion (cm3); pCO2 and pmelt are the densities of vapour bubble CO2 and melt in g/cm3; CVB

CO2

is the concentration of CO2 in the vapour bubble (ppm) from conversion; CVB
CO2

is concentration
dissolved in the melt, acquired from SIMS analysis; and total concentration Ctotal

CO2
(ppm) in the

melt inclusion overall. Vapour bubbles were identified in 54 melt inclusions from Yasur scoria
(T1, T2 and T3) but only 6 returned quantifiable Fermi Diads (Electronic Appendix A1, Table
A1.2).

2.3.3 Data quality

Conversion to density is reliant upon parametrisation to a densimeter, however the use of non-
instrument specific densimeters can significantly overestimate calculated CO2 densities (Wieser
et al., 2021; Taracsák et al., 2019; Venugopal et al., 2020). An instrument specific densimeter
(developed by Wieser et al. 2021) was used to avoid systematic uncertainties in the absolute
CO2 densities that propagate from the use of generalised or non-standardised densimeters (see
Wieser et al. 2021; DeVitre et al. 2021). The standard deviation of three repeated measurements
taken within each melt inclusion-vapour bubble were used to assess the precision of analyses
(Figure 2.3).

A more significant source of error on the measurement of CO2 in vapour bubbles comes
from calculation of bubble volume fraction, under the assumption that vapour bubbles and the
ellipsoidal axes of melt inclusions were spherical (Tucker et al., 2019). The dimensions of

16



Chapter 2. Methods

the two observable axes were measured photomicrographically using the Zeiss Scope software.
The unobservable axis of each melt inclusion was assumed to be the arithmetic mean of the two
known axes (Tucker et al., 2019) and is associated with errors ranging from –48 to 37%.

Figure 2.3: CO2 density (g/cm3) versus diad splitting (∆, cm−1) of the 6 vapour bubbles in melt
inclusions that were quantified from Yasur. Errors bars on each data point represent the preci-
sion of the data, taken as the standard deviation calculated from three repeated measurements
of a single vapour bubble.

2.4 Secondary Ion Mass Spectrometry (SIMS)

2.4.1 Acquisition procedure

Volatile (H2O, CO2) concentrations in melt inclusions discussed in Chapter 6 were determined
by SIMS (Cameca IMS-7f GEO) at the NERC Ion Microprobe Facility, University of Edin-
burgh. Decisions regarding the instrument set up and analysis conditions were made by Dr.
Cees-Jan de Hoog. All SIMS analyses were performed prior to EPMA to minimise the risk of
volatile migration under the electron beam and to avoid contamination of CO2 measurements
with the carbon coat that is required for EPMA. Epoxy stubs were gold coated and left to outgas
in the ion microprobe airlock at ca. 10−7 mbar for 12 hours.

Procedures followed those described by Wieser et al. (2020), which included a 16O− pri-
mary ion beam (5 nA) accelerated to 13 kV and focused onto the sample surface to create a 15
µm spot size, as this is optimal for analyzing the concentrations of volatile elements. A rela-

17



2.4. Secondary Ion Mass Spectrometry (SIMS)

tively high mass resolution of 1200 was used to separate 24Mg2+ and 12C peaks. An energy
offset of 50 V with a 50 V energy window was applied to reduce volatile backgrounds and
matrix effects. Isotopes measured were 1H+, 24Mg2+, 12C+, 26Mg+, and 30Si+. Prior to each
analysis, the area was pre-sputtered for 180 s using a 15 µm beam raster to remove any H and
C contamination from the surfaces of the epoxy-mounted grains. At the start of each analysis,
beam positions were centered, and 30Si peak positions were verified. At the beginning of each
session, calibrations were completed, and peak positions for isotopes were reaffirmed.

SIMS volatile measurements are sensitive to instrument background concentration varia-
tions - CO2 is most problematic because it has a low positive ion yield and typically low con-
centrations in natural samples. In addition to working under ultrahigh vacuum (ca. 10−9 mbar),
the first 2 out of 10 cycles of each 12C and 1H analysis were eliminated to minimise the effects
of surface contamination. Backgrounds of CO2 and H2O were checked on a sputtered olivine
crystal, assumed to be anhydrous. Each melt inclusion was examined under reflected light to
ensure the SIMS beam was entirely within the area of the melt inclusion.

2.4.2 Calibration and data quality

Raw counts per second (with background subtracted) of 12C were converted to concentrations
using a calibration curve of CO2 concentration (ppm) vs. 12C/30Si × SiO2 (wt%) for the cal-
ibration standards measured (Figure 2.4a). To calculate H2O concentrations, 1H/30Si was not
corrected for SiO2 wt% (Figure 2.4b) as it has been shown that matrix effects for H/Si correlate
with the SiO2 content of the samples, such that samples with different SiO2 content will plot
along a single calibration curve. The concentrations of SiO2 (wt%) used to convert counts to
concentrations are measured in sample melt inclusions determined by EPMA and range from
51.4 to 61.0 wt%. I measured 26Mg and compared MgO concentrations of melt inclusions ac-
quired with SIMS during H2O and CO2 runs to concentrations derived from EPMA (Figure 2.5)
to identify analyses that breached the melt inclusion and hit the host phenocryst. Samples that
breached the melt inclusion and hit the host phenocryst were discarded from further analysis.

Internal precision refers to the 1σ uncertainty for each individual analysis based on the
standard error of the mean of 8 cycles for each measurement. For H2O, these uncertainties range
from 0.15 to 1.82%, whereas for CO2, due to the relatively low 12C count rate, uncertainties are
larger and range from 3.1 to 19.8% (Electronic Appendix A2, Table A2.1). External precision,
or reproducibility, of the data is measured from repeated measurements on a homogeneous
secondary standard (where M5 is the secondary standard). Precision on the secondary standard

18



Chapter 2. Methods

Figure 2.4: Calibration curve for CO2 (a) and H2O (b) across each session that was used to
convert count rates of 1H and 12C to concentrations. Symbols refer to different days (sessions).
R2 and standard error (SE) are shown. Equation of the calibration line and the error on the slope
are plotted. Symbols differ depending on the day that analyses were measured.

Figure 2.5: Comparing uncorrected MgO concentrations acquired by EPMA to those derived
from SIMS measurements on olivine-hosted (ol), plagioclase-hosted (plag) and clinopyroxene-
hosted (cpx) melt inclusions.

19



2.5. Electron Probe Micro-Analyser (EPMA)

is taken as the standard deviation of repeated measurements divided by the mean and gives
uncertainties of 1.81% for H2O and 0.87% for CO2 (Table 2.1, Electronic Appendix A2, Table
A2.2). Reproducibility of the data is calculated as the slope of the calibration line divided by
the standard error of the slope. Accuracy of these measurements was calculated as the mean of
the secondary standard divided by the laboratory accepted concentration of the standard (Table
2.1). In Chapter 6, I combine the internal and external errors and the reproducibility of the slope
to give the total error on concentration measurements of CO2 (4.09 – 11.35%) and H2O (2.38
– 2.44%) (Electronic Appendix A2, Tables A2.1 and A2.3) — both represent an uncertainty of
1σ .

Element Standard N Mean concen-
tration

1 σ Precision (%) Accuracy (%)

H2O M5 3 0.78 wt% 0.02 1.81 122.35
CO2 M5 3 1137.82 ppm 9.89 0.87 114.93
H2O M40 3 3.00 wt% 0.024 0.80 97.75
CO2 M40 3 2086.80 ppm 35.97 1.72 95.59
H2O M36 1 104.83
CO2 M36 1 98.79
H2O M10 1 107.28
CO2 M10 1 153.97
H2O 519-4-1 1 89.30
CO2 519-4-1 1 144.52

Table 2.1: Summary of the mean concentration, precision, and accuracy for different standards
and elements. Standard M5 was the chosen secondary standard for analysis. Empty cells where
only one measurement on the standard was taken, hence no mean, precision or standard devia-
tion (σ ) could be calculated.

2.5 Electron Probe Micro-Analyser (EPMA)

2.5.1 Analytical procedure

The geochemistry of matrix glasses, melt inclusions and phenocryst hosts that are discussed in
Chapter 6 were measured on the electron microprobe (Cameca SX100 EMP) at the University
of Cambridge in two sessions in November 2022 and February 2023. The composition of Fe-Ti
oxides in scoria samples were analysed on the electron microprobe (JEOL, JXA-iHP 200F Hy-
per Probe) in a third session in April 2024. These analyses were acquired in collaboration with

20



Chapter 2. Methods

Dr. Iris Buisman, who helped with the instrument set up, analysis of standard materials and
decisions regarding suitable beam conditions for each analytical routine. Epoxy stubs were car-
bon coated with uniform thickness of 15 nm prior to analysis and then loaded into the electron
microprobe, under vacuum. Melt inclusions hosted in phenocrysts of olivine, clinopyroxene
and plagioclase were identified with the electron beam off to avoid damage to hydrous silicate
glasses by irradiation of the beam. Analyses were conducted using WDS on five spectrometers.
The standards, crystals and count times for olivine, plagioclase and clinopyroxene host spot
analyses can be found in Electronic Appendix A3, and Tables A3.1.1 through to A3.2.3 in this
thesis. Each element was calibrated using a mineral standard at the start of each analytical ses-
sion. The peak count times for elements varied depending on the phase that they were analysed
in (Electronic Appendix A3, Tables A3.1.1 through to A3.2.3).

Alkali element X-ray lines drop in intensity during EPMA therefore deciding upon a suitable
beam current and exposure time is highly important. Using a dual routine, the beam was set to
15 kV and currents and counting times varied according to the material under inspection (Table
2.2). In the first session Na, Al, P, Ca, K, Ti, Si, Mg, Fe and Mn were analysed using a 10
µm defocused beam and 10 nA for glasses and melt inclusions; 20 nA and 5 µm defocused
beam for clinopyroxene; and 40 nA and 1 µm defocused beam for olivine; and 10 nA and 5
µm defocused beam for plagioclase crystals. In the second session melt inclusions and matrix
glasses were re-analysed for Cl and S using a defocused beam of 5 µm and current of 20 nA.The
beam conditions for phenocryst hosts of the analysed melt inclusions were also updated. In the
third session only oxides were analysed, at 15 kV using a 20 nA and 1 µm defocused beam. To
minimise alkali migration, Na and K were measured first followed by remaining major, minor
and volatile elements. Space permitting, a minimum of two repeat analyses per host, melt
inclusion and matrix glass were taken.

Table 2.2: Beam conditions for spot analyses of phenocryst hosts (olivine, plagioclase, clinopy-
roxene) and melt inclusions (glass) measured over two analytical sessions, and oxides mea-
sureds in the third and final session.

Material Session Beam voltage (kV) Beam current (nA) Beam spot (µm)
Plagioclase 1 15 10 5
Plagioclase 2 15 10 5
Olivine 1 15 40 1
Olivine 2 15 40 1
Clinopyroxene 1 15 20 5
Clinopyroxene 2 15 20 5
Glass 1 15 10 10
Glass 2 15 20 5
Oxide 3 15 20 1

21



2.5. Electron Probe Micro-Analyser (EPMA)

2.5.2 Data quality

Tables A3.2.1, A3.2.2, A3.2.3 details the mean and 3σ of the detection limits on each mate-
rial analysed, per session. Data reported with concentrations below detection limits and with
totals < 96 or >102 wt% were discarded from further evaluation. Precision and accuracy of
the EPMA analyses reported in this study were calculated from repeated measurements of sec-
ondary standards for each of the analysed materials (olivine, plagiolcase, clinopyroxene, glass
- including melt inclusions and matrix glasses, oxide), per session (Table 2.3 – 2.5). The large
discrepancy in the accuracy of S and Cl concentrations measured in the glass secondary stan-
dard, VG2 1, relative to laboratory accepted values, likely reflects the low concentrations of
these elements in the standards themselves and perhaps exposure to heterogeneity within the
standard. Assessing the accuracy of EPMA values for these elements is difficult due to the lack
of synthetic and experimental glass standards available for S and Cl. Precision on the VG2 1
glass standard for S (10.87%) and Cl (5.82%) are however reasonable, therefore these analyses
were accepted.

22



Chapter 2. Methods

Ta
bl

e
2.

3:
Pr

ec
is

io
n

an
d

ac
cu

ra
cy

es
tim

at
es

fo
rm

ea
su

re
m

en
ts

in
se

ss
io

n
1

au
g

16
49

05
au

g
12

21
42

sa
n

ca
rl

os
ol

la
b

11
59

00
A

99
ba

sa
lt

cl
in

op
yr

ox
en

e
cl

in
op

yr
ox

en
e

ol
iv

in
e

pl
ag

io
cl

as
e

gl
as

s
E

le
m

en
t

Pr
ec

is
io

n
%

A
cc

ur
ac

y
%

Pr
ec

is
io

n
%

A
cc

ur
ac

y
%

Pr
ec

is
io

n
%

A
cc

ur
ac

y
%

Pr
ec

is
io

n
A

cc
ur

ac
y

%
Pr

ec
is

io
n

%
A

cc
ur

ac
y

%
N

a 2
O

4.
52

10
0.

92
3.

43
10

5.
02

1.
80

11
4.

24
4.

53
10

0.
76

K
2O

22
0.

95
48

.2
7

14
0.

99
9.

33
73

.4
9

4.
10

10
3.

86
Si

O
2

0.
73

10
0.

48
0.

53
99

.2
0

0.
56

99
.7

6
0.

50
99

.9
0

0.
62

98
.7

4
C

aO
0.

97
98

.7
3

1.
08

10
1.

31
1.

40
96

.5
7

1.
15

96
.5

5
Ti

O
2

1.
29

96
.0

6
0.

88
11

7.
06

11
.7

0
81

.7
1

0.
75

99
.9

6
Fe

O
1.

50
13

5.
05

1.
45

94
.3

2
0.

85
95

.0
7

6.
19

92
.5

7
1.

53
10

0.
55

M
nO

20
.9

6
10

0.
24

13
.1

2
10

7.
26

13
.0

6
95

.5
8

10
7.

30
17

7.
40

10
.1

9
13

8.
92

C
r 2

O
3

4.
58

10
2.

46
13

.1
3

A
l 2

O
3

0.
64

93
.2

3
0.

56
10

4.
72

0.
62

95
.1

3
0.

80
95

.5
3

M
gO

0.
71

99
.3

8
0.

85
97

.2
8

0.
24

10
0.

50
4.

69
82

.6
6

1.
03

98
.9

4
P2

O
5

4.
90

11
3.

65

23



2.5. Electron Probe Micro-Analyser (EPMA)

Ta
bl

e
2.

4:
Pr

ec
is

io
n

an
d

ac
cu

ra
cy

es
tim

at
es

fo
rE

PM
A

m
ea

su
re

m
en

ts
in

se
ss

io
n

2

V
G

2
1

sa
n

ca
rl

os
ol

au
g

12
21

42
la

b
11

59
00

gl
as

s
ol

iv
in

e
cl

in
op

yr
ox

en
e

pl
ag

io
cl

as
e

E
le

m
en

t
Pr

ec
is

io
n

%
A

cc
ur

ac
y

%
Pr

ec
is

io
n

%
A

cc
ur

ac
y

%
Pr

ec
is

io
n

%
A

cc
ur

ac
y

%
Pr

ec
is

io
n

%
A

cc
ur

ac
y

%
N

a 2
O

1.
61

10
3.

34
1.

52
10

5.
63

2.
61

10
8.

27
K

2O
6.

45
10

7.
42

32
9.

42
10

.9
3

66
.6

3
Si

O
2

0.
51

99
.3

3
0.

50
10

0.
07

0.
61

99
.9

0
0.

97
10

0.
65

C
aO

0.
55

97
.5

5
12

.7
6

0.
56

10
1.

31
1.

30
10

0.
06

Ti
O

2
0.

95
10

2.
48

33
.3

0
0.

78
11

7.
81

17
.8

6
77

.6
5

Fe
O

0.
81

99
.7

4
1.

35
94

.8
7

1.
31

95
.1

8
6.

37
95

.1
0

M
nO

13
.5

1
93

.0
1

9.
86

89
.4

8
19

.5
4

10
1.

19
19

8.
89

75
.5

3
C

r 2
O

3
95

.7
9

45
.2

8
14

.1
6

A
l 2

O
3

0.
26

95
.9

9
6.

37
0.

22
10

4.
79

1.
17

97
.8

6
M

gO
0.

80
99

.6
7

0.
23

99
.6

5
0.

76
97

.3
5

4.
78

83
.7

7
P 2

O
5

8.
74

10
7.

39
SO

2
5.

82
20

7.
18

C
l

10
.8

7
42

.7
7

N
iO

5.
07

97
.0

0

24



Chapter 2. Methods

Cr-Augite 164905
oxide

Element Precision % Accuracy %
Na2O 2.27 99.25
SiO2 0.09 101.06
CaO 0.09 100.40
TiO2 1.13 95.36
FeO 0.61 98.11
MnO 2.24 98.33
Cr2O3 2.75 98.59
Al2O3 0.31 91.57
MgO 0.49 99.73

Table 2.5: Precision and accuracy estimates for measurements in session 3

2.5.3 Oxide spot analyses

The oxide compositions that are mentioned in Chapter 6 were calculated using concentrations
from EPMA measurements (Electronic Appendix A3, Table A3.8). Oxide stoichiometry was
calculated by assuming all Fe is present as Fe2+ and using the deficit charge balance to deter-
mine the molar proportion of Fe3+. The compositions of Fe-Ti oxides analysed in scoria from
Yasur plotted in the magnetite field of the ternary in Figure 2.6 and can be found in Electronic
Appendix A3, Table A3.9.

Figure 2.6: Fe-Ti oxide compositions represented in a ternary with TiO2, Fe2O3 and FeO end
members. Compositions calculated from stoichiometry and charge balance.

25



2.6. Laser Ablation Inductively Coupled Mass Spectrometry (LA-ICP-MS)

2.6 Laser Ablation Inductively Coupled Mass Spectrometry
(LA-ICP-MS)

2.6.1 Acquisition parameters

LA-ICP-MS was used to determine trace element abundances in melt inclusions using methods
described by (Jenner and O’Neill, 2012). I used a Photon Machine G2 193 nm excimer laser
system, equipped with a HelEx 2-volume cell coupled to an Agilent 8800 ICP-QQQ-MS at
the School of Environment, Earth and Ecosystem Sciences at The Open University. Decisions
regarding the instrument set up, conditions for spot and line scan analyses were made by Dr.
Barbara Kunz.

Two analytical routines were used under consideration of the analytical conditions needed
to maximise counting times of low abundance elements 1) trace elements by spot and 2) Ag
by line scan. Samples were placed in the sample chamber under a He atmosphere and with a
laser beam that ranged in diameter (25, 30, 40 and 50 µm) focused on the melt inclusion surface
with a fluency of 3.63 J/cm2. For sample aerosol transport 0.91 L/min He gas was used, which
was subsequently mixed with 0.77 L/min Ar downstream before injection into the plasma for
ionisation. Table 2.6 provides conditions for spot analyses and line scans.

Analysis
type

Background
/gas blank
(s)

Repetition
Rate (Hz)

Scan speed
(µm/s)

Integration
time (s)

Ar flow
rate
(L/min)

Spot (trace
elements)

30 10 N/A 0.606 0.77

Line scans
(Ag)

30 50 15 0.267 0.70

Table 2.6: Analytical conditions for spot and line scan (Ag only) analyses.

Trace elements were measured using 10 Hz repetition rate and spot sizes varied (25, 30,
40, 50 µm) depending on the size of the target melt inclusion – note that smaller spot sizes
have lower resolution but quicker processing time. The overall integration time (0.606 s) i.e.
time taken to complete one cycle of all masses from lightest (Li) to heaviest (U) element, was
optimised to keep the overall time at a minimum while maximising individual dwell times for
low abundance elements (Table 2.6).

26



Chapter 2. Methods

Samples were pre-ablated prior to analysis to avoid surface contamination. Signal acqui-
sition included a gas blank for each spot of 30 s, followed by signal for 30 s and finally a 50
s washout period. NIST-SRM 612 was used as a primary standard and BCR-2G was used as
the secondary standard for trace elements and the primary standard and major elements. EPMA
analyses were used as a secondary data quality check for major elements. A standard bracketing
approach was taken for spot analyses two spots per standard were measured in the following
order: 2 × NIST-SRM 612, 3 × SCOl, 2 × BCR, 4 × melt inclusion, 2 × BCR, 2 × NIST-SRM
612. The San Carlos Olivine was analysed after NIST-SRM 612 to help wash out the signal,
i.e. ensure that elements (e.g. Sb, W) with high concentrations in NIST-SRM 612 relative to
the melting inclusions were brought to background levels.

A separate analytical routine was used for Ag; analyses were acquired using line scans, at
50 Hz repetition rate and at scan speeds of 15 µm/s. In addition to the He-Ar-sample gas mix,
5 ml/min of N2 was introduced via a y-piece to enhance ionisation. Integration time for the Ag
analysis was 0.267 s (Table 2.6) to maximise integration cycles for short signal durations. The
length of each line scan varied depending on the size of the melt inclusion. The background
and washout time was the same as for the spot analysis.

The procedure for line scans is ablating across the entire length of the target melt inclusion,
starting at the host in order to maximise the length of the processed signal. Due to the nature
of acquisition, melt inclusion signals are susceptible to contamination through mixing of the
host in relation to ablation of inclusions through to host. Contamination can be identified by
manually inspecting an element signal that has high concentrations in the host and relatively
lower concentrations in the glass. Line scans were only conducted in olivine hosts, therefore
careful attention was paid to the Ni signal - focusing particularly on where the Ni signal was a
flat plateau in glasses (Figure 2.7). Furthermore, Ni concentrations obtained for melt inclusions
during the Ag line scans were later on compared with Ni concentrations obtained for spots, as
an additional quality check to choose only the glass proportion of the signal.

27



2.6. Laser Ablation Inductively Coupled Mass Spectrometry (LA-ICP-MS)

Figure 2.7: Example of an Ag-line scan in an olivine-hosted melt inclusion. Following back-
ground, the laser is turned on and the Ni signal rises in the host olivine. Ca increases and Ni
drops as the signal from the host olivine crystal and melt inclusion mix. The selected signal
(∼4 s) is taken as the flat Ni signal which represents the melt inclusion.

The accuracy of Ag data is challenged by interferences that arise from molecular or elemen-
tal species that share almost identical mass-to-charge ratios as the targeted isotope. For 107Ag,
mass interferences with 91Zr 16O+ occur and are corrected by calculating the ZrO production
based on in-house standards with known Ag concentrations (Jenner et al., 2015). The standard
bracketing order for the Ag routine was modified as follows: 2 × NIST-SRM 612, 3 × SCOl,
39-1, 41-1, 49-1, 51-1 (which are all in-house standards for ZrO correction), 4 × melt inclusion
spots, 2 X NIST-SRM 612.

2.6.2 Data processing and quality

Data processing and assessment of data quality was done in collaboration with Dr. Barbara
Kunz. Raw data were processed using Iolite v3.71 (Paton et al., 2011) where signal intensity
(cps) was converted to concentration (ppm). First, a background correction is applied where the
background is subtracted from the signal. Individual signals of each melt inclusion are inspected
for signal quality to exclude i.e. contamination or mixing. Standardisation is then performed
using external standards (NIST-SRM 612) and internal standards (29Si or 43Ca depending on
which element provides the most accurate results for the secondary standards). Spot diameters
of 30 and 25 µm were internally calibrated using 43Ca; spots with diameters of 40 and 50 µm
and line scans were internally calibrated against 29Si. Internal precision is taken as the stan-
dard error on internal standards (see Electronic Appendix A4, Table A4.1 for all trace element

28



Chapter 2. Methods

concentration measurements and their internal precision).

Accuracy is calculated as the average concentration of repeated measurements on the sec-
ondary standard (BCR-2G) relative to in-house long term averages and published values from
Jenner and O’Neill (2012). External precision is taken as the relative standard deviation of re-
peated measurements of BCR-2G divided by the mean of the secondary standard. Precision and
accuracy were calculated for each spot size (Electronic Appendix A4, Tables A4.2 to A4.8).
The accuracy and precision of repeated measurements of secondary standard, BCR-2G, for
chalcophile elements were generally better than 10%. Poorer accuracy was obtained for Bi and
Tl (±40%) and Sb (±20) likely due to heterogeneity of the concentrations within the standards
themselves (Jenner and O’Neill, 2012); and higher precision was obtained for Sb (20%) and Tl
(26%). For Rare Earth Elements (REEs) (La to Lu), accuracy was within ±5% and precision
up to 13%. Smaller spot sizes (25 µm) show marginally poorer accuracy and precision for REE
analyses. Accuracy on the remaining lithophile elements are ±5%; Be was more variable (-20
to 5%), as was Ta and W (-15%). Precision is generally ∼7% besides Be (10-20%), W (5-20%)
and Cr (5-15%). No systematic differences in the precision or accuracy of chalcophile element
and REE measurements relating to differences in spot size were observed. Errors reported for
Cu and Ag concentrations in Chapter 6 are propagated by combining the internal and exter-
nal errors (ppm). Refer to Electronic Appendix A4, Tables A4.9 and A4.10 for the complete
dataset.

2.7 Post-entrapment crystallisation corrections

Providing that melt inclusions remain chemically and physically isolated from the external en-
vironment, they may faithfully record the composition of the melt at the time of entrapment
(Rose-Koga et al., 2021; Gaetani and Watson, 2002; Danyushevsky and Plechov, 2011; Danyu-
shevsky et al., 2000). However, from their formation at depth to eruption at the surface, pristine
preservation is unlikely and almost all melt inclusions undergo post-entrapment modifications
in response to changes in pressure, temperature, melt composition and redox state (Rose-Koga
et al., 2021; Steele-MacInnis et al., 2011; Wieser et al., 2021; Hartley et al., 2017; Moore and
Bodnar, 2019). Post-entrapment crystallisation (PEC) is common in response to pre-eruptive
magma cooling and diffusive H+ loss (Rose-Koga et al., 2021) . In olivine-hosted melt in-
clusions, crystallisation of an olivine rim causes inclusions to deplete more in Mg than Fe,
according to the equilibrium distribution coefficient (Roeder and Emslie, 1970):

Kd = [XFe/XMg]ol/[XFe/XMg]melt (2.4)

29



2.7. Post-entrapment crystallisation corrections

where Kd is the equilibrium distribution coefficient of Fe and Mg between the inclusion and
olivine host; XFe and XMg are the molar fractions of Fe and Mg in olivine (ol) and the melt. With
more cooling and crystallisation, the olivine rim becomes progressively richer in Fe and poorer
in Mg.

PEC can be corrected numerically (Danyushevsky et al., 2000) to restore the original com-
position of the inclusion. This is done by adding small increments of equilibrium olivine (i.e.,
the crystallising rim composition) back into the measured melt inclusion. This is repeated, with
the Kd being recalculated after each olivine addition until the Fe-Mg Kd reaches the equilib-
rium value with the olivine host. The composition of the host olivine is used to determine the
equilibrium melt composition based on the coupled diffusion of Fe and Mg in olivine (Roeder
and Emslie, 1970). Since [XFe/XMg]ol in equation 2.4 is taken as the olivine host, it is important
that the host olivine is not zoned and for the spot analysis of the host to be taken sufficiently far
from the melt inclusion, such that it does not interfere with the crystallised rim. Having reached
equilibrium, the mean mass of olivine added can be calculated and the PEC correction is then
applied.

PEC corrections are further complicated by diffusive re-equilibrium of the inclusion with the
host crystal and/or external melt which results in differential fractionation of elements, which
are dependent on their relative diffusivities (Danyushevsky et al., 2000; Rose-Koga et al., 2021;
Gaetani and Watson, 2002). If cooling is sufficiently slow, the Fe-rich olivine rim that crys-
tallised in the inclusion can then equilibrate with the more Mg-rich olivine host. Fe diffusing
out of the rim into the host olivine causes inclusions to equilibrate by means of melt Fe-loss
(Danyushevsky et al., 2000), hence inclusions no longer reflect magmatic variation.

An initial assessment of olivine-melt equilibrium was done using a Rhodes diagram in Fig-
ure 2.8 which shows the minimum and maximum equilibrium fields for Roeder and Emslie
(1970) and Matzen et al. (2011), using an open source Python tool, Thermobar (Wieser et al.,
2022). Petrolog3 was used to reconstruct the original trapped composition of melt inclusions
by correcting for PEC and Fe-loss (Danyushevsky and Plechov, 2011). The ’Reverse Crystalli-
sation’ and ‘Reconstruct MI composition’ option was chosen and requires users to define the
composition of the inclusion, host olivine (Fo#) content (taken adjacent to the target inclusion),
initial FeO∗ and fO2. Since the composition of the groundmass glass should approximate that
of the residual melt inclusion at the time of eruption (Danyushevsky et al., 2000), I fixed initial
FeO∗ to 8.88 wt% which represents the average matrix glass composition from Yasur scoria
samples. I used the mineral-melt equilibrium model of Toplis (2005), and Fe oxidation state
model of Kress and Carmichael (1991).

The program then simulates addition of equilibrium olivine back into the measured melt

30



Chapter 2. Methods

Figure 2.8: Olivine-liquid equilibrium curves were modelled in Thermobar to assess whether
a) melt inclusions (OHMIs) were in equilibrium with their olivine host, and if b) matrix glasses
were in equilibrium with adjacent olivines. Uncorrected olivine-hosted melt inclusions (unPEC-
corrected) are Mg-depleted relative to the melt composition in equilibrium with host olivine
(according to Roeder and Emslie 1970). PEC-corrected olivine-hosted melt inclusion (PEC-
corrected OHMIs) compositions were computed using Petrolog3 and required addition of less
than 5% olivine to restore equilibrium.

inclusion until the computed melt inclusion composition reaches equilibrium with the host Fo#
(supplied by user). The host is held at the liquidus olivine temperature of the inclusion until
equilibrium is restored (Danyushevsky and Plechov, 2011). Once equilibrium is reached, the
programmed FeO of the melt inclusion is compared to the user specified FeO∗ – if FeO∗ > FeO
programmed, Petrolog3 will simulate a temperature increase whilst simultaneously maintaining
host-inclusion equilibrium. The process is continued until the FeO content of the inclusion
matches the FeO∗ value specified. The mean mass of olivine added is calculated and the PEC
correction is applied. Olivine-hosted melt inclusions presented in Chapter 6 (26 out of 44 melt
inclusions) are estimated to have < 5% olivine crystallisation. This correction is applied to Cu,
Ag, Cl and S concentrations (which are assumed to be incompatible with respect to crystallising
silicate phases) that are presented in Chapter 6.

PEC corrections for clinopyroxene- and plagioclase-hosted melt inclusions are more diffi-
cult due to their complex mineralogy (Neave et al., 2017; Barber et al., 2021; Adams et al.,
2021), but also due to a lack of streamlined ‘user-friendly’ software that is available for olivine.
Barber et al. (2021) applied a manual assimilation approach using RhyoliteMELTS (Ghiorso

31



2.7. Post-entrapment crystallisation corrections

and Gualda, 2015) to correct for crystallisation in clinopyroxene-hosted melt inclusions. Neave
et al. (2017) take a similar approach to restore equilibrium melt compositions in plagioclase-
hosted melt inclusions where equilibrium plagioclase compositions were calculated with the
model of Namur et al. (2012). An initial assessment of PEC in plagioclase- and clinopyroxene-
hosted melt inclusions in my samples was done by comparing the MgO, CaO and Al2O3 con-
centrations at a given Mg# to those of existing Yasur melt inclusions and whole rocks (Nielsen
2011, Figure D.2). For a given Mg#, all melt inclusions align with the trends defined by previ-
ously published Yasur whole rocks and melt inclusions, therefore I decided not to perform PEC
corrections on these data (for more discussion see Chapter 6 and Appendix D, Figure D.2).

32



3
Modelling crystallisation, degassing and sul-
fide saturation in magmas

3.1 Degassing of volatile chalcophile elements during frac-
tional crystallisation and decompression

The modelling described in the following section has been published as “Water-rich mag-

mas optimise volcanic chalcophile element outgassing fluxes”, in Earth and Planetary Science

Letters. Co-authors include, M. Edmonds and J. Blundy. The results of these models are pre-

sented in Chapter 4.

For a closed magma system comprised of melt and a magmatic volatile phase (MVP) in
equilibrium, the initial magmatic H2O concentration determines the mass fraction of the ex-
solved MVP at any particular pressure or degree of crystallisation, according to the relationship:

MX
MVP = MX

Melt ×G×DX
MVP/melt (3.1)

where MX
MVP is the total mass of an element X in the MVP, MX

Melt is the total mass of X in
the melt, DX

MVP/melt is the fluid-melt partition coefficient of X, and G is the mass ratio of MVP

33



3.1. Degassing of volatile chalcophile elements during fractional crystallisation and decompression

to melt in the system which, for a fixed pressure and degree of fractionation, increases with
magma water content. Note that equation 3.1 explicitly accounts for the mass of metals (the
quantity of interest here) rather than the concentration of metal in the fluid. MVP development
during isobaric crystallisation and decompression in sulfur-free conditions is modelled using
five hypothetical magmatic systems to encapsulate the range in the natural data (Table 3.1). I use
MagmaSat (Ghiorso and Gualda, 2015) to model the solubility of H2O and CO2 under different
pressure, temperatures using the five cases in Table 3.1. Initial melt volatile contents are further
modified by crystallisation during magma ascent, which is modelled using RhyoliteMELTS
(Ghiorso and Gualda, 2015).

System Bulk H2O, wt% Bulk CO2, wt% Bulk Cl, wt%

A 1 0.2 1

B 1 0.2 0.5

C 3.4 1.5 0.5

D 3.4 1.5 0.05

E 3.4 1.5 1

Table 3.1: Input parameters of the five magmatic systems, A, B, C, D, and E used to model the
development of an MVP. Starting compositions are basaltic and differ only in their initial bulk
water (H2O wt%) and chlorine (Cl wt%) contents in ranges typical of arc magmatic settings
(Wallace, 2005).

3.1.1 Partitioning behaviours of chlorine and groups of chalcophile ele-
ments

I model chlorine and fluid-mobile chalcophile elements partitioning into the exsolving MVP.
The solubility behavior of Cl is complex and varies with melt composition (Métrich and Ruther-
ford, 1998; Signorelli and Carroll, 2002), fluid composition (Botcharnikov et al., 2004; Webster
et al., 2009), temperature, oxygen fugacity and pressure (Botcharnikov et al., 2004). Some stud-
ies have postulated an inverse relationship between DCl

MVP/melt and pressure, i.e. that DCl
MVP/melt

decreases with increased pressure (Alletti et al., 2009; Shinohara, 2009); this is explained by
the large and negative pressure dependence of NaCl partitioning into a melt and the HCl–NaCl
exchange reaction between a silicate melt and an aqueous fluid, which favours HCl in aqueous
fluids at lower pressures (Shinohara, 2009). These pressure dependencies cause chlorine to ap-
pear as HCl in low pressure (∼0.1 MPa) volcanic gases and NaCl in high pressure (∼50 MPa)

34



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

fluids. More work is required, however, to fully understand the implications of Cl speciation on
fluid-melt partitioning (Shinohara, 2009; Thomas and Wood, 2021).

Chlorine complexes with some chalcophile elements and therefore directly influences their
partitioning behaviour between melt and fluid phases (Iveson et al., 2019; Webster et al., 2018;
Zajacz et al., 2008). However, some chalcophile elements speciate with S and OH– ligands
instead, and are consequently independent of Cl. I assume that limited-to-no sulfide saturation
occurs during fractional crystallisation or decompression. To replicate the different behaviours
of groups of selected chalcophile elements I define three hypothetical metals X, Y and Z, dis-
tinguished by their partition coefficients: X (moderately volatile and Cl-complexing, e.g., Pb),
Y (highly volatile and Cl-complexing, e.g., Bi) and Z (highly volatile and non-Cl-complexing,
e.g., Se). I assume that all metals begin with the same concentration (10 ppm) in the primitive
melts in the first set of models, so that the effect of varying the volatile contents can be extracted
from the results. The effects of increasing the initial melt concentrations of metals to 20 and
100 ppm, were also investigated (Chapter 4).

3.1.2 Modelling isobaric crystallisation and second boiling

I used MagmaSat (Ghiorso and Gualda, 2015) to model the equilibrium volatile composition
(H2O and CO2) of the magma and coexisting MVP during isobaric fractionation at pressures
of 80, 160, 240 and 400 MPa, representing the case where magma is stored in the crust and
undergoes equilibrium crystallisation driving second boiling. Within the isobaric crystallisa-
tion models, I assume that the exsolved volatile component (the MVP) remains in equilibrium
with the evolving magma. The melt fraction (F) decreases from 1 to 0.1 in the model as the
magma crystallises; the bulk volatile content (in the melt and exsolved fluid together) increases
correspondingly:

Xvolatile
bulk =

Xvolatile
i

Dvolatile
solid/melt(1−F)+F

(3.2)

which for a mineral/melt partition coefficient of e.g., element X, Dvolatile
solid/melt, of zero (assum-

ing volatiles are not taken up in any crystal phases during crystallisation), this simplifies to:

35



3.1. Degassing of volatile chalcophile elements during fractional crystallisation and decompression

Xvolatile
bulk =

Xvolatile
i

Fi
(3.3)

where Xvolatile
bulk , in wt%, is the initial bulk magma concentration of volatile species X (from

Table 3.1); Fi is the melt fraction at each step (modelled from 1 to 0.1); and Xvolatile
bulk wt%, is

the bulk concentration of volatile species X in the melt-MVP system. MagmaSat (Ghiorso and
Gualda, 2015) was used to calculate the melt and exsolved MVP compositions in terms of H2O
and CO2 at each crystallisation step, using melt compositions shown in Electronic Appendix B,
Table B.1. The fraction of volatiles remaining in the melt at each step, Vmelt, is given by:

Vmelt =
XH2O

melt +XCO2
melt

XH2O
bulk +XCO2

bulk

(3.4)

where Vmelt, is the mass fraction of volatiles left dissolved in the melt XH2O
melt +XCO2

melt relative
to the entire system XH2O

bulk +XCO2
bulk and is calculated at each melt fraction step for four different

pressures. The fraction of volatiles in the MVP is therefore:

VMV P = 1−Vmelt (3.5)

The total mass of fluid generated in each step is:

MMV P = Mbulk ×
(

XMV P

100

)
(3.6)

The concentrations of chlorine in both the MVP and melt are calculated using partition
coefficients which vary with pressure and with melt fraction (Tattitch et al., 2021). As the
melt evolves towards rhyolitic compositions, I model fluid-melt partitioning of chlorine using
the synthesis of experimental data of (Tattitch et al., 2021). DCl

MVP/melt is low (< 10) for basaltic
compositions and decreases as pressure decreases (Tattitch et al., 2021) (see section 3.1.1). I use
three initial Cl bulk contents (Table 3.1) and calculate the Cl content of the melt-MVP system

36



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

XCl
bulk at each step during isobaric fractionation using equation 3.2. Melt Cl concentrations were

calculated as:

XCl
melt =

XCl
bulk

DCl
MV P/melt +1

(3.7)

and used to model chlorine partitioning into the MVP:

XCl
MV P = XCl

bulk −XCl
melt (3.8)

where XCl
melt and XCl

MVP are the concentrations (wt%) of Cl in the residual melt fraction and
associated MVP; DCl

MVP/melt is the fluid-melt partition coefficient. Fluid salinity (SA) and molar
chlorinity (ML) are given by:

SA = XCl
MV P ×

mNaCl

mCl
(3.9)

ML =
XCl

MV P ×104

mCl
103

(3.10)

where XCl
MVP is the concentration (wt%) of Cl in the fluid reservoir; mNaCl is the molecular

mass of NaCl (58.44 g) and mCl is the atomic mass of Cl (35.453 g). The mass of Cl in the
MVP per unit mass of melt is given by:

MCl
MV P = MMV P ×XCl

MV P (3.11)

where MCl
MVP is the total mass (kg) of Cl in the fluid per kg of magma; MMVP is the total

mass of volatiles in the MVP, and XCl
MVP is the concentration (wt%) of Cl in the MVP. I model

37



3.1. Degassing of volatile chalcophile elements during fractional crystallisation and decompression

the bulk, melt, and fluid concentrations of elements X, Y, and Z using equations 3.7-3.11. The
fluid-melt partition coefficient of X (DX

MVP/melt = 6 × ML) and Y (DY
MVP/melt = 10 × ML) are

scaled empirically to the molar chlorinity of the fluid (after Zajacz et al. 2008), generating a
range comparable to that observed experimentally for Pb and Bi (Zajacz et al., 2008), as shown
in Figure 3.1. DZ

MVP/melt was set to 50, independent of the molar chlorinity of the fluid (as
observed for e.g., Se, As; Audétat 2019; Zajacz et al. 2008).

Figure 3.1: Comparison of the partitioning behaviour of theoretical trace elements. Partition
coefficients used in my models describing the partitioning of theoretical trace metals X, Y and
Z between the MVP and the silicate melt, plotted against a) MVP salinity where the fluid-
melt partition coefficient of elements X (DX

MVP/melt) and Y (DY
MVP/melt) are proportional to the

salinity of the MVP, and DZ
MVP/melt is independent of MVP salinity; b) melt fraction (1 is a basalt

and 0.2 is a rhyolite), where DX
MVP/melt and DY

MVP/melt increase whereas DZ
MVP/melt remains the

same with decreasing melt fraction; and c) pressure, where DX
MVP/melt and DY

MVP/melt drop with
pressure whereas DZ

MVP/melt remains constant.

3.1.3 Modelling decompression and first boiling

Magmas exsolve volatiles during decompression (‘first boiling’) owing to the dependence of
H2O and CO2 solubility on pressure (e.g., Newman and Lowenstern 2002). As magmas ascend
to the surface an exsolved MVP is generated, into which chlorine and other volatile species,
including some metals and metalloids, partition. Degassing that occurs during magma ascent
preceding and accompanying eruption, accelerates magma up the conduit by lowering bulk
magma density. I assume here that magmas have no exsolved MVP prior to the onset of decom-
pression from different starting pressures in the model (effectively assuming that all magmas are

38



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

System
Pressure
(MPa)

Bulk magma
H2O (wt%)

Bulk magma
CO2 (wt%)

Bulk magma
Cl (wt%)

A*

400 4.92 0.32 0.06

240 4.01 0.12 0.06

160 3.29 0.06 0.08

80 2.26 0.02 0.17

B*

400 4.92 0.32 0.01

240 4.01 0.12 0.01

160 3.29 0.06 0.01

80 2.26 0.02 0.02

C*

400 1.99 0.25 0.06

240 1.96 0.13 0.06

160 1.93 0.07 0.08

80 1.82 0.02 0.17

D*

400 1.99 0.25 0.12

240 1.96 0.13 0.12

160 1.93 0.07 0.17

80 1.82 0.02 0.33

E*

400 4.92 0.32 0.12

240 4.01 0.12 0.12

160 3.29 0.06 0.17

80 2.26 0.02 0.33

Table 3.2: Starting bulk volatile contents for modelled decompressed magmas. Concentrations
were calculated in MagmaSat using the melt compositions given in Electronic Appendix B, Ta-
ble B.1. A*- E* relate to systems described in Table 3.1 but have been isobarically fractionated
to 50% and lost volatiles and chlorine via open system degassing, prior to decompression.

at the point of saturation just prior to ascent). These models therefore represent an end member
of degassing behaviour. I initialise the model using a range of melt compositions encompass-
ing the concentrations of dissolved volatiles derived from 50% crystallisation of the primitive
magmas (as seen in Electronic Appendix B, Tables B.1 and B.2).

This extent of crystallisation will generate the trachybasalt magmas typically erupted from
Yasur, for example, from the primitive basalts at depth (Métrich et al., 2011, 2004) and typical
Etna hawaiites and mugearites from alkali basalts (Métrich and Rutherford, 1998). The model
describes chemical evolution of the MVP during decompression from the starting pressure to
0.1 MPa via a single batch closed system step. I have chosen closed system degassing because
the relatively small size of the exsolved vapour bubbles and rapid ascent of hydrous magmas

39



3.1. Degassing of volatile chalcophile elements during fractional crystallisation and decompression

means that relatively little fluid-melt segregation can occur on ascent timescales. The starting
bulk volatile compositions for the two cases considered are shown in Table 3.2: a relatively
water-rich melt and a water-poor melt. Saturated water and CO2 concentrations in melts at the
starting conditions were calculated using MagmaSat (Ghiorso and Gualda, 2015).

In MagmaSat, bulk volatile contents (Table 3.2) and primitive basaltic compositions from
Etna (Métrich et al. 2004, sample 36,10) and Yasur (Métrich et al. 2011, TAN23-18) were used
as proxy wet and dry starting melts for the decompression model. I assumed no crystallisation
during decompression, therefore melt fraction is assumed constant although I acknowledge that
decompression-induced-crystallisation will occur during ascent (Lipman et al., 1985). The total
fluid concentration (XMVP wt%; equation 3.3), the volatile fraction remaining in the melt (Vmelt;
equation 3.4), the volatile fraction exsolved to the MVP (VMVP; equation 3.4), and the total mass
of fluid (MMVP; equation 3.6) were calculated. Results are shown in Electronic Appendix B,
Table B.2.

The concentration of Cl in the decompressed melt (XCl
melt) and fluid (XCl

MVP), salinity (SA),
molar chlorinity (ML), and mass yield (MCl

MVP) were calculated using equations 3.7-11. The
starting bulk concentration of Cl, (XCl

bulk) used for melts in the decompression model was taken
as the melt concentration (XCl

melt) in the isobaric melt after 50% crystallisation, i.e., melt fraction
(Fi) of 0.5 (Table 3.2). For simplicity, the starting bulk concentrations of X (Xbulk), Y (Ybulk),
and Z (Zbulk) were set to 10 ppm across each modelled system and starting pressure. For each
case, an average DCl

MVP/melt was used to describe partitioning of chlorine through each pressure
interval during decompression to the surface, given as 0.1 MPa (Table 3.3; Electronic Appendix
B, Tables B.2 and B.3) using the parametrisations of (Tattitch et al., 2021) to replicate the bulk
fluid compositions after closed system decompressional degassing. DX

MVP/melt and DY
MVP/melt

were again scaled against the molar chlorinity of the fluid (equation 3.9).

Pressure (MPa) Average DCl
MVP/melt

80 2.9

160 3.9

240 5.1

400 5.9

Table 3.3: Average DCl
MVP/melt of trachybasalts (modelled as a melt fractionated to F=0.5) calcu-

lated for each pressure interval during decompression (Electronic Appendix B, Table B.3). For
the range in modelled fluid salinities, increasing bulk magmatic chlorine content had negligible
influence on the overall DCl

MVP/melt therefore it was modelled to be the same for all systems (A*-
E*) at a given, despite changes to a system’s bulk Cl wt%.

40



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

3.2 Degassing and sulfide saturation of chalcophile elements
during fractional crystallisation

The modelling described in the following section has been submitted as part of the manuscript,

“Sulfide resorption by water-rich melts yields copper-rich magmatic fluids”, in the journal of

Nature Communications, and is currently in review. Co-authors include, M. Edmonds, P. Wieser,

M. Gleeson, F. Jenner and J. Blundy. The results of these models are presented in Chapter 5.

3.2.1 Background

The behaviour of chalcophile elements, such as Cu, Ag, Au, Re, Se, in magma are influenced
by processes of sulfide saturation and degassing during fractionation. Therefore assumptions
and methods described in section 3.1, for sulfur-free systems, are unable to faithfully model
the behaviour and fate of these elements in evolving magmatic systems. Understanding the
relative effects of these processes may be key to understanding whether these elements become
sequestered in crustal cumulates, or transported via hydrous fluids to upper crustal reservoirs,
which may lead to the formation of porphyry ore deposits.

Here I formulate a model that combines crystallisation and sulfide saturation, but also incor-
porates the degassing of water, sulfur and chlorine, which is a critical and up to now, unexplored
part of the process. This approach allows a detailed examination into the impact of magmatic
water and sulfide formation on the metal-carrying capacity of the co-existing magmatic fluids.
The following terms are referred to frequently throughout this section and Chapter 5:

• SCSS2−: sulfur content of the silicate melt at sulfide saturation

• S6+/Stotal: ratio of sulphate concentration (S6+) to total sulfur concentration (Stotal) in the
melt

• SCAS: sulphate content of the silicate melt at sulphate saturation

• Fe3+/Fetotal: concentration ratio of Fe3+ in the melt to total Fe (Fetotal)

• DCu
SL/melt, DAg

SL/melt: sulfide liquid-silicate melt partition coefficient for copper (Cu) and
silver (Ag)

• DCu
sulf/melt, DAg

sulf/melt: sulfide phase-silicate melt partition coefficient for copper (Cu) and
silver (Ag) (for modelling sulfide as sulfide liquid and then monosulfide solid solution)

41



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

3.2.2 Modelling fractional crystallisation and degassing of H2O and CO2

I use RhyoliteMELTS v1.2.0 (Ghiorso and Gualda, 2015) run through an open-source Python
package, PetThermoTools, (v.0.2.1) (Wieser, 2023) to model the composition of the melt during
fractional crystallisation and degassing from 1300◦C to 800◦C for a range of magma water con-
tents (0 to 6 wt%), storage pressures (50 MPa to 400 MPa) and oxygen fugacities (∆QFM+1 to
+1.4) (Figure 3.2). Recognising that natural arc magmatic systems are not strictly buffered,
unbuffered models yielded log fO2 values ranging from –5 to –25, which approximates to
∆QFM+1 to -10 (Frost, 1991). However, these values do not encapsulate the range recorded
by arc magmas globally (Cottrell et al., 2021) and so I opted to buffer models, resulting in a
range in fO2 that aligns closely with the observed natural data (Figure 5.1). RhyoliteMELTS
normalises major element oxides and volatiles to 100 wt%, hence initial H2O concentrations
(wt%) will vary slightly from those originally input to PetThermoTools. For the most water-
rich cases (3-6 wt% H2O), a magmatic volatile phase already exists prior to crystallisation, at
the onset of the model (Figure 3.2h).

3.2.3 Modelling degassing of sulfur, chlorine and chalcophile elements

As a hydrous exsolved fluid phase forms during crystallisation sulfur, chlorine and chalcophile
elements partition into it, with the ratio of concentrations in each phase described by a fluid-
melt partition coefficient. Chalcophile elements are assumed to be incompatible with respect to
the crystallising silicate phases. Some chalcophile elements (e.g., Cu, Ag) are Cl-complexing
and their partitioning behaviour between the melt and fluid phases can be linked to chlorine de-
gassing (Iveson et al., 2019; Webster et al., 2018; Zajacz et al., 2008), with partition coefficients
increasing as fluid salinity increases (Figure 3.3a). Fluid-melt partition coefficients for Cl are
taken from Tattitch et al. (2021) and are lowest for basalts (1-3) and the highest for rhyolites
(25-65). Fluid-melt partition coefficients for S are lowest in basalts (1-10) and highest in rhyo-
lites (15-200), but in contrast to Cl, increase with decreasing pressure (Berlo et al., 2014; Fiege
et al., 2015; Grondahl and Zajacz, 2022; Masotta et al., 2016; Zajacz et al., 2012). Sulfur fluid-
melt partitioning is highly dependent on fO2 (Keppler, 2010). Final results represent models
run through RhyoliteMELTS at ∆QFM +1.2, and therefore use oxidised values for DS

fluid/melt.

42



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

Figure 3.2: Harker diagrams demonstrating the effect that changing magma water contents has
on fractional crystallisation paths over a range of pressures (shaded area represents pressures
from 100 to 400 MPa). Modelled system presented here is buffered to ∆QFM+1.2.

43



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

Figure 3.3: Relative (a) volatility and (b) sulfide affinity of Cu and Ag with respect to the sili-
cate melt. Dfluid/melt for Cl-speciated chalcophile elements, Cu (Tattitch and Blundy, 2017a) and
Ag (Zajacz et al., 2008) increase as the salinity of the fluid increases during magma differenti-
ation to lower MgO. Dashed and solid lines differentiate pressures. Bold lines in b) represent
Dsulfide/melt where SL-silicate melt partition coefficients (Kiseeva and Wood, 2013; Li et al.,
2021; Li and Audétat, 2015; Patten et al., 2013) are modelled up to FeOt 3.6 wt% and transition
to MSS-silicate melt partition coefficients (Li et al., 2021; Li and Audétat, 2015) below this
threshold. Faint lines in b) show contrasting modelled Dsulfide/melt for models that assume only
SL is present.

44



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

The mass of chalcophile elements transported by fluids are controlled by the mass of the
fluid reservoir, which is a function of magma water contents, pressure and the extent of crys-
tallisation (Hogg et al., 2023). Under the assumption that mass balance is maintained, the mass
fraction of fluid forming in a given system and the mass fraction of a given chalcophile element
(X) partitioning into the fluid can be described as:

MX
fluid = MX

total · (
DX

fluid/melt
Gfluid

)/(
DX

fluid/melt
Gfluid

+1) (3.12)

where MX
fluid and MX

total are the total masses of X in the fluid and total system respectively and
DX

fluid/melt is the fluid-melt partition coefficient of X. Gfluid is the ratio of melt mass to fluid mass
in the system which, for a fixed pressure and degree of fractionation, increases with magma wa-
ter content. I report the results for models that best fit the global arc copper array in Chapter 5
(Figure 5.1a) where initial magma S and Cl concentrations were set to 0.1 wt% (Muth and Wal-
lace, 2022; Wallace, 2005). Although models run at different concentrations of S and Cl (0.05
– 0.2 wt%) changed the absolute concentration and masses of chalcophile elements distributed
among the melt and fluid, the relative relationships remain unchanged (Figure 3.4). The mass
of exsolved fluid (MX

fluid) is calculated from RhyoliteMELTS where the fluid is comprised only
of H2O. The total concentration and mass of S and Cl in the entire system are as follows:

CX
total =

CX
initial
Fi

(3.13)

MX
total = M f luid +Mmelt ×

(
CX

total
106

)
(3.14)

where CX
total ppm is the total concentration of X (S or Cl) in the system; CX

initial ppm is the
initial total magma concentration; and Fi is the melt fraction at each step representing the ratio
of Mmelt to the initial Mmelt ; and MX

total is the total mass of X (S or Cl) in the system in grams.
The mass of Cl and S in the exsolving fluid can be found using equation 3.12, and the mass of
Cl and S in the melt residual (CX

melt) as:

45



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

MX
melt = MX

total −MX
f luid (3.15)

The concentration of Cl and S in ppm in the fluid and residual melt are therefore:

CX
f luid = (106 ×MX

f luid)/M f luid (3.16)

CX
melt = (106 ×MX

melt)/Mmelt (3.17)

The molar chlorinity (ML) and salinity (SA) of the exsolving fluid are given by:

ML = 10−6 ×
CCl

f luid

mCl/103 (3.18)

SA = 10−4 ×CCl
f luid ×mNaCl/mCl (3.19)

where CCl
fluid is the concentration (ppm) of Cl in the fluid; mNaCl is the molecular mass

of NaCl (58.44 g) and mCl is the atomic mass of Cl (35.453 g). The total, melt and fluid
concentrations and masses of chalcophile elements are modelled using equations 3.12 to 3.17.
Parental magma Cu concentrations were fixed at 75 ppm to reflect average global arc primitive
basalt compositions (Richards, 2015). Initial magma Cu/Ag ratio is within the range of primitive
MORB values (Jenner et al., 2010); initial concentrations of Ag (0.03 ppm) were based on these
assumptions.

46



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

Figure 3.4: Models run at ∆QFM +1.2 and for different initial S and Cl concentrations, a) S
= 1000 ppm, Cl = 500 ppm; b) S = 1500 ppm, Cl = 1000 ppm. Relationship with changing
magmatic water contents remains the same as in Figure 5.1. Shaded regions represent the range
in concentrations modelled for a given system between 100 and 400 MPa.

3.2.4 Modelling sulfide saturation

The relatively oxidised nature of arcs means that a higher proportion of dissolved S is present
as sulphate (S6+) compared to MORB and OIBs (Cottrell et al., 2021). A variety of models
are available for estimating S speciation in silicate melts (Jugo et al., 2010; Nash et al., 2019;
O’Neill and Mavrogenes, 2022). Nash et al. (2019) calculates S6+/Stotal with reference to melt
Fe3+/Fetotal in basalt-dacitic melts at 1 atm. Water in the melt lowers liquidus temperatures,
which in turn affects melt Fe3+/Fetotal, making S6+/Stotal ratios modeled this way highly depen-
dent on temperature and water contents. O’Neill and Mavrogenes (2022) introduce a parameter
to describe the sulfide and sulfate capacity of a melt that can be used to calculate the S6+/Stotal of
the melt. S6+/Stotal parameters modeled using O’Neill and Mavrogenes (2022) showed signif-
icant variation with magma water contents. Jugo et al. (2010) calculate S6+/Stotal based on the
oxidation state of the melt, where ∆QFM is calculated relative to a redox buffer (Frost, 1991).
Minor misalignment between the QFM buffer position in RhyoliteMELTS and that of Jugo
et al. (2010) based on Frost (1991) does result in slight changes to S6+/Stotal through fractional
crystallisation, but these did not impact the overall model results.

The SCSS2− correlates positively with melt fO2 (Jugo et al., 2010; Nash et al., 2019), FeO
(Jugo et al., 2010; Nash et al., 2019; O’Neill, 2021; Smythe et al., 2017) and negatively with
pressure (Matjuschkin et al., 2016). The SCSS2− increases with fO2 at high pressure (Cox

47



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

et al., 2019; Matjuschkin et al., 2016) as demonstrated by the ability of arc magmas to dissolve
up to 1.5 wt% S (mostly present as S6+) at pressures of ∼1 GPa (Wallace and Edmonds, 2011)
in the absence of sulfide formation. The effect of dissolved H2O on the SCSS2− is relatively
understudied and presently remains debated (Fortin et al., 2015; Li and Zhang, 2022; O’Neill,
2021). I use the model of O’Neill (2021) where the SCSS2− decreases from ∼1600 ppm in
basalts to >400 ppm in rhyolitic melts (Figure 3.5a). For simplicity, to model SCSS2− I use
a constant value of 0.6 for the ratio of Fe/(Fe+Ni+Cu) in the sulfide but acknowledge there is
likely a range, although poorly constrained over different pressures and temperatures (Wieser
et al., 2020; Ding and Dasgupta, 2018).

Final models were buffered at ∆QFM+1.2 – within the range reported for average arc mag-
mas (Cottrell et al., 2021) – to control the effect that fO2 has on the SCSS2−. Experimental work
on anhydrite solubility shows a positive relationship with increasing temperature (and pressure)
(Jugo et al., 2010; Zajacz and Tsay, 2019). Evidence of sulfate melt saturation in oxidised arc
dacites and andesites from Yanacocha, Peru highlight its stability under magmatic conditions
(Chambefort et al., 2008; Hutchinson et al., 2020). Sulfate is also detected as a phase included
in magmatic apatite phenocrysts from Qulong porphyry, Gangdese belt, Tibet (Xia et al., 2023).
I model S6+ solubility using parametrisations by Zajacz and Tsay (2019) (Figure 3.5b). I used
an open-source Python package, PySulfSat (v.1.0.3) (Wieser and Gleeson, 2023) to model the
S6+/Stotal, SCSS2− and SCAS of the melt through fractional crystallisation at different pressures
and magma water contents.

Figure 3.5: Modelling changes in a) SCSS2− ; b) SCAS and c) S6+/Stotal during fractional
crystallisation of magmas with S 1000 ppm; Cl 1000 ppm; Cu 75 ppm; and different initial
dissolved water concentrations (0 to 6 wt%). Models are run over different pressures, between
100 and 400 MPa.

48



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

3.2.5 Sulfide-silicate melt partitioning of chalcophile elements

Several parametrisations have been developed to calculate Cu and other chalcophile elements’
sulfide liquid-silicate melt partition coefficients, (Kiseeva and Wood, 2015, 2013; Li and Audétat,
2015). Kiseeva and Wood (2013) (hereafter called KW2013) determined a simple correlation
between sulfide liquid-silicate melt partitioning behaviour of chalcophile elements (e.g., Cu and
Ag), and melt FeO content:

logDX
SL/melt =−m log(FeO)+ c (3.20)

where DX
SL/melt is the sulfide liquid-silicate melt partition coefficient for a chalcophile ele-

ment, X; FeO is the concentration of FeO in the melt (wt%); and m and c are experimentally-
derived constants. Subsequent studies by Li and Audétat (2015) (LiAud2015) and Kiseeva and
Wood (2015) (KW2015) expand on the relationships derived by KW2013 and show temperature
and oxygen fugacity to be important constraints on the partitioning behaviour of chalcophile el-
ements between sulfide liquid and silicate melt. In Figure 3.6, modelling of DCu

SL/melt and DAg
SL/melt

using equations of KW2013, KW2015 and LiAud2015 are shown alongside the experimentally-
derived DCu

SL/melt and DAg
SL/melt reported in these studies. When applied to my model, equations

from KW2015 and LiAud2015 yield very high DCu
SL/melt values, surpassing most experimental

data reported in these studies (Figure 3.6a). A similar relationship is observed for modelled
DAg

SL/melt using KW2015 (Figure 3.6b). By contrast, sulfide liquid-silicate melt partition coef-
ficients modelled using KW2013 are smaller and overlap with experimental data across these
studies (Figure 3.6a-b).

To further evaluate the performance of DX
SL/melt parametrisations, ratios of DCu

SL/melt relative
to DAg

SL/melt are plotted for both modelled outputs and published experimental data (Figure 3.6c).
Ratios of DCu

SL/melt / DAg
SL/melt should remain consistent with geochemical evidence for limited

fractionation between Cu and Ag during fractionation of sulfide liquids (Figure 3.6c) (Jenner,
2017; Jenner et al., 2010). Modelled ratios using KW2015 and KW2013 align well with ex-
perimental data and remain constant through fractional crystallisation (Figure 3.6c). However,
results using LiAud2015 incur significant fractionation associated with the large differences in
DCu

SL/melt and DAg
SL/melt (Figure 3.6c). I therefore model DCu

SL/melt and DAg
SL/melt using the parametri-

sation by KW2013, as this best matches the experimental data and maintains coherence with
literature (e.g., Jenner 2017) by exhibiting no significant fractionation in the presence of SL.

49



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

Figure 3.6: Evaluation of different chalcophile element sulfide liquid-silicate melt partitioning
parametrisations. Lines represent modelled DCu

SL/melt (a) and DAg
SL/melt (b) and the ratio of these

SL - melt partition coefficients (c) using equations by Kiseeva and Wood (2013) (KW2013,
pink); Kiseeva and Wood (2015) (KW2015, orange); and Li and Audétat (2015) (LiAud2015,
turquoise). Different symbols represent individual sulfide liquid-silicate melt experimental data
points from KW2013, KW2015 and LiAud2015 and are referenced by the same colours as
corresponding model lines. Use of KW2015 or LiAud2015 to model DCu

SL/melt and DAg
SL/melt return

extremely high values that increase exponentially as melt FeO decreases. DCu
SL/melt modelled

using LiAud2015 become more than an order of magnitude greater than DAg
SL/melt modelled

using LiAud2015 (in c), which would result in strong fractionation of melt Cu/Ag ratios, in the
absence of MSS.

50



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

Sulfides take the form of sulfide liquid (SL) in arc melts at high temperature (Nadeau et al.,
2010) but may transform to monosulfide solid solution (MSS) at lower temperatures (Chang
et al., 2018; Keith et al., 2017; Li and Audétat, 2015). Experimental work define this transition at
around 3.6 wt% FeOt or ∼1000°C (Li and Audétat, 2015) (Figure 3.7a). Since the affinity of Ag
for the sulfide phase drops relative to Cu in the presence of MSS (Li and Audétat, 2015) (Figure
3.7b), models assuming SL to be the only sulfide phase do not produce Cu/Ag fractionation
during crystallisation (Figure 3.7b), which is contrary to the trends displayed by arc magmas
(Cox et al., 2020; Jenner, 2017; Jenner et al., 2010). I therefore assume all sulfide in the melt to
be present as SL up to 3.6 wt% FeOt, or as MSS below 3.6 wt% FeOt (Figure 3.7c), although I
acknowledge that natural arc systems likely have mixed proportions of both phases (Chiaradia,
2020; Chiaradia and Caricchi, 2017). The resultant DCu

sulf/melt and DAg
sulf/melt modelled with this

approach replicate the experimental data well (Figure 3.7c). The concentration of S, as S2− in
ppm, in the melt is given by:

CS2−
melt =CS

melt ×

(
1− S6+

Stotal

)
(3.21)

where CS2−
melt is concentration of sulfide in the melt; CS

melt is the total concentration of S, ppm,
in the melt; and S6+/Stotal is proportion of S6+ in melt as predicted by speciation models (Jugo
et al., 2010). Where melt CS2−

melt is greater than the SCSS2− calculated, the concentration and
mass of S2− in the fractionating sulfide can be calculated as:

CS2−
sulf =CS2−

melt −SCSS2− (3.22)

MS2−
sulf = Mmelt +MH2O

fluid ×

(
CS2−

sulf
106

)
(3.23)

where CS2−
sulf is the concentration of S2− in ppm, in the sulfide – note this is either sulfide

liquid (SL) or monosulfide solid solution (MSS) depending on if the melt is above or be-
low FeOt 3.6 wt% respectively; and MS2−

sulf is the mass of S2−, in grams, in the sulfide. The
concentration of sulfide remaining in the melt CS2−

sat melt in ppm, is equal to the SCSS2− or for

51



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

sulfide-undersaturated melts, is equal to CS2−
melt. The concentration of sulphate in the melt can be

calculated as follows:

CS6+

melt =CS
melt ×

S6+

Stotal
(3.24)

where CS6+

melt is the concentration, in ppm, of S6+ in the melt. If CS6+

melt is greater than the SCAS
calculated for given melt composition. The concentration and mass of S6+ in the crystallising
anhydrite can be calculated as:

CS6+

anhyd =CS6+

melt −SCAS (3.25)

MS6+

anhyd = Mmelt +MH2O
fluid ×

CS6+

anhyd

106

 (3.26)

where CS6+

anhyd is the concentration of S6+ in ppm, in anhydrite; and MS6+

anhyd is the mass of

S6+, in grams, in anhydrite. The concentration of S6+ remaining in the melt CS6+

sat melt in ppm, is
therefore equal to the SCAS or for sulphate-undersaturated melts, is equal to CS6+

melt. The sum of
total S concentration and mass of S in the melt after accounting for fluid and sulfide saturation
is:

CStotal

sat-melt =CS6+

sat-melt +CS2−
sat-melt (3.27)

MStotal

sat-melt =CStotal

sat-melt ×
Mmelt

106 (3.28)

where CStotal

sat-melt and MStotal

sat-melt are the total concentration (in ppm) and mass (in grams) of

52



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

S remaining in the melt residual after both degassing and sulfide saturation has occurred. I
assume S constitutes approximately one third of the total mass of the sulfide. The mass of a
chalcophile element, X, sequestered by a sulfide phase (MX

sulf) can be described as in equation
3.12 by substituting Gfluid for Gsulfide (the mass fraction of melt to sulfide in the system) and
DX

fluid/melt for DX
sulf/melt where DX

sulf/melt is (the sulfide-silicate melt partition coefficient of X); and
the concentration (CX

sulf) in ppm is:

CX
sulf =

106 ×MX
sulf

3×MS2−
sulf

(3.29)

where I assume MS2−
sulf is ∼third of the total mass of the sulfide (e.g., CuFeS2). The mass and

concentration of X remaining in the melt post-degassing and sulfide saturation is:

MX
sat-melt = MX

melt −MX
sulf (3.30)

CX
sat-melt = 106 ×

MX
sat-melt

Mmelt
(3.31)

where MX
sat-melt and CX

sat-melt represent the mass (in grams) and concentration (in ppm) of a
chalcophile element (X) remaining in the melt. The mass fraction of fluid, melt and solid at any
stage of differentiation can then be calculated as:

Xfluid =
Mfluid

Mfluid
initial +Mmelt

initial +Msolid
initial

(3.32)

replacing Mfluid with Msolid and Mmelt to find Xsolid and Xmelt respectively. I assume the
fluid is pure H2O, since only minor proportions are comprised of S and Cl. The mass fraction
of sulfide was:

53



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

Xsulf =
3×MS2−

sulf

Mfluid
initial +Mmelt

initial +Msolid
initial

(3.33)

where Xsulf is the mass fraction of sulfide forming at each step. Hence, Xtotal was calculated
as:

Xtotal = Xfluid +Xmelt +Xsolid +Xsulf (3.34)

Normalised mass fractions were then calculated, and referred to as Xnorm
fluid , Xnorm

melt etc. To
maintain mass balance, I re-calculate the final concentration of S and chalcophile elements in
the melt, fluid and sulfide as:

MX
initial =CX

initial ×Xnorm
total (3.35)

CX
melt =

MX
initial

Xnorm
melt +(Xnorm

fluid ·DX
fluid/melt)+(Xnorm

sulf ·DX
sulf/melt)

(3.36)

CX
fluid =CX

melt ×DX
fluid/melt (3.37)

CX
sulf =CX

melt ×DX
sulf/melt (3.38)

where MX
initial is initial mass (which must remain constant for mass balance to be maintained)

of S and chalcophile elements (in grams); CX
melt, CX

fluid, CX
sulf are the re-calculated equilibrium

concentrations of S and chalcophile elements (in wt%). The final mass of S and chalcophile
elements in the melt, fluid and sulfide is therefore:

54



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

MX
melt =CX

melt ×Xnorm
melt (3.39)

where MX
melt is the equilibrated mass of an element in the melt (in grams). MX

fluid and MX
sulf

are calculated by substituting CX
melt and Xnorm

melt for fluid and sulfide counterparts (see equation
3.32 – 3.38). CS2-

melt was then recalculated as in equation 3.21. Where CS2-

melt was still greater than
the SCSS2−, equations 3.22-38 were iterated until CS2-

melt approached the SCSS2−. This threshold
was set to be within 0.1% of the SCSS2−. MS2-

sulf (equation 3.23) formed in each iteration was
accumulated such that:

MS2−
acc sulf = MS2−

sulf (i)+MS2−
sulf (i−1) (3.40)

where MS2-

acc sulf is the accumulated mass of sulfide formed for each iteration of the model.
MS2-

acc sulf for each iteration then substitutes for MS2-

sulf in equation 3.33. This model was con-
structed using open source Python code, available on my GitHub (https://github.com/
oliviahogg/CuRichFluid).

55

https://github.com/oliviahogg/CuRichFluid
https://github.com/oliviahogg/CuRichFluid


3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

Figure 3.7: Comparison of different published DCu
SL/melt and DAg

SL/melt parametrisations and ex-
perimental data. a) coloured lines are for fractional crystallisation models run for different
magma water contents and pressures, overlain experimental data from Li and Audétat (2015).
Larger symbols highlight best fit or most relevant experiments to my modelled crystallisation
paths. Circles represent LiAud2015 experiments containing SL and stars represent LiAud2015
experiments that were MSS-bearing. b) and c) Data for Cu and Ag are plotted in green and
orange respectively; symbol shape and size for experimental data from LiAud2015, are same
as in a); squares are experimental data from KW2013 and crosses from KW2015; lines in b)
represent the modelled DCu

SL/melt and DAg
SL/melt using KW2013 and do not replicate experimentally

derived DAg
SL/melt; lines in c) represent modelled DCu

sulf/melt and DAg
sulf/melt using the SL-silicate melt

partition coefficients of Cu and Ag from KW2013 up to 3.6 wt% FeO, thereafter, MSS-silicate
melt partition coefficients of Cu and Ag from LiAud2015 are used. By transitioning from SL to
MSS at 3.6 wt% FeOt, modelled DCu

sulf/melt and DAg
sulf/melt in c) fit the experimental data well.

56



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

3.2.6 Modelling sulfide assimilation by hydrous magmas

Sulfur-undersaturated magmas ascending through crustal mushes have the potential to assim-
ilate pre-existing sulfides, and thereby raise the S and Cu concentration of the melt. I model
this process by adding 0.1 g increments of sulfide to mafic to intermediate (Fi between 1 and
0.6) sulfide-undersaturated melts that initially had 1000 ppm S and 75 ppm Cu, fractionating
at shallow pressures (100 MPa) until the melt became sulfide-saturated, to explore the effect
this has on the evolving fluid composition. The mass of Cu and S added to the melt per 0.1 g
increment of sulfide (here denoted as CuFeS2) was calculated as:

MX
CuFeS2

= mX ×mol (3.41)

where MX
CuFeS2

is the mass (in grams) of an element, either Cu or S respectively in CuFeS2;
mX is atomic mass of an element, either Cu (63.55 g) or S (32.07 g); and mol are the moles of
S or Cu in 0.1 g of CuFeS2. Hence the stoichiometric S concentration of CuFeS2 is 34 wt%.
This mass was added to the initial mass of S and Cu in the system (MX

total; equation 3.14) and
the model was performed again and iterated until the final melt sulfide concentration (CS2-

melt;
equation 3.21) reached the SCSS2−.

57



3.2. Degassing and sulfide saturation of chalcophile elements during fractional crystallisation

Figure 3.8: The effect of magma water concentration on Cu and S: melt concentrations (a, b);
melt mass (c,d); fluid mass (e,f, note logarithmic scale); sulfide mass (g,h). The mass fraction of
Cu and S partitioned in to sulfides versus fluids are shown in (i,j). Water-rich systems (3 and 6
wt% H2O) generate Cu-poor magmas due to coincident degassing and sulfide saturation. Water-
poor (0, 1 and 0.1 wt% H2O) systems generate Cu-enriched magmas where sulfide saturation
is delayed and minimal degassing occurs. Dry (0 wt% H2O) system is highlighted in red and
water-bearing systems become progressively darker in blue.

58



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

3.3 Degassing and sulfide saturation of chalcophile elements
during decompression

3.3.1 Background

The results of modelling presented in this section are discussed in Chapter 6. I apply the same
three-way partitioning method as in section 3.2 but repurpose this for magmas decompressing
isothermally from a given storage depth in the crust (200 MPa to the surface, 0.5 MPa). I
apply this model to the magmatic system of Yasur volcano, Vanuatu Arc, to help describe the
mechanisms involved in shaping the composition of silicate melt inclusions analysed in this
study.

3.3.2 Modelling degassing and crystallisation during ascent

I use RhyoliteMELTS v1.2.0 (Ghiorso and Gualda, 2015) via alphaMELTS (Smith and Asi-
mow, 2005), run through terminal, to model the composition of the melt during isothermal
decompression-induced crystallisation and degassing. AlphaMELTS requires the following in-
puts: initial melt composition, pressure, temperature and oxidation state. I run two sets of mod-
els; one starting with an initial melt composition close to the suspected parental basalt for Yasur
(Tuk3, Dupuy et al. 1982) and one starting with a basaltic trachyandesite (Tan2318, Métrich
et al. 2011) which covers the range in composition of the erupted products at Yasur (Métrich
et al., 2011). Hereafter, I will refer to the decompression models run with these compositions
as Tuk3 or Tan2318. All models were run starting with initial H2O and CO2 concentrations of
1.2 wt% and 0.1 wt%, respectively (Métrich et al., 2011; Hogg et al., 2023). Reported are the
model outputs that best fit the major, minor and volatile element and Cu concentrations of Yasur
magmas. Initial magmatic Cl and S concentrations were set to 300 ppm and 500 ppm, respec-
tively, such that modelled decompression trajectories overlapped the concentrations recorded
by Yasur melt inclusion data from this study.

A range of melt oxygen fugacities were trialled and evaluated based on the fit of major,
minor, volatile elements and mineral-melt equilibrium compositions predicted by models with
the natural data. Final models are run at ∆QFM since these best fit the data. Models were
buffered for reasons explained in section 3.2.2. Whether models started from 400, 200 or 100
MPa observed minor differences in the overall major element and Cu concentration trajectories
for both Tuk3 or Tan2318 models. A launch pressure of 200 MPa was decided, in line with the

59



3.3. Degassing and sulfide saturation of chalcophile elements during decompression

depth range suspected for the dominant magma storage reservoirs under Yasur (Métrich et al.,
2011). Models for both Tuk3 and Tan2318 were run isothermally at temperatures of 1075◦C
and 1050◦C) consistent with what has previously been reported for Yasur (Métrich et al., 2011).

Published Cu data available for Yasur are rare and consist only of whole rock data (Dupuy
et al. 1982; Deng et al. 2022 and references therein). I take the sample Ta93 (Dupuy et al.,
1982) with the highest concentration of MgO to represent a composition that is similar to that
of the primary or parental basalt feeding Yasur - this enables us to tentatively constrain initial Cu
concentrations to ∼145 ppm. No Ag concentrations in either whole rock or melt inclusions have
been reported for Yasur, therefore the global Cu/Ag array (∼3000–4000) is used to constrain
the Ag concentration that were input to Yasur models (42 ppb).

3.3.3 Modelling degassing of sulfur, chlorine and chalcophile elements at
Yasur

The behaviour of S and Cl during decompression differs from that during fractional crystalli-
sation. The solubility of S decreases during decompression, with magmas exsolving fluids at
the shallowest pressures partitioning the greatest proportion of their S inventory compared to
greater pressures (Keppler, 2010; Ding et al., 2023; Zajacz et al., 2012) (Figure 3.9). All equa-
tions for calculating the mass and concentration of chalcophile elements in exsolving fluids
and evolving melt remain unchanged from Section 3.2. Only the fluid-melt partition coefficient
parametrisations for S, Cl and Cu described in section 3.2.3, are modified for decompression
(Figure 3.9). As in section 3.2.3, chalcophile elements (Cu and Ag) and S are assumed to be
incompatible with respect to the crystallising silicate phases.

I use values of DS
fluid/melt reported by Ding et al. (2023) that were modelled in SulfurX, an

open source Python model of sulfur degassing during ascent. DS
fluid/melt is highly sensitive to

the oxidation state of the melt (Zajacz et al. 2012; Keppler 2010; section 3.2.3). Ding et al.
(2023) reports the range in DS

fluid/melt modelled for basaltic andesites during decompression
for a relatively anhydrous reduced system (Mauna Kea) and a hydrous and relatively oxidised
system (Fuego). Since Yasur volcano generates basaltic magmas that are moderately oxidised,
and to avoid ‘over-modelling’, I simply extrapolate DS

fluid/melt values to fit the range of these two
endmember cases, where at 200 MPa for a basaltic andesite DS

fluid/melt is 100 and increases up
to 1100 near the surface (Figure 3.9).

During decompression, the affinity of Cl for the fluid phase decreases (Webster et al., 1999;
Tattitch et al., 2021). In section 3.1 (Table 3.3; Electronic Appendix B, Tables B.2 and B.3),

60



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

Figure 3.9: Fluid-melt partition coefficients (Dfluid/melt) of S, Cl and Cu modelled during de-
compression. Values on the left hand axis, in purple, are for sulfur. Values on the right hand
axis, in black, are for Cl and Cu. DCl

fluid/melt decreases with decompression. DCu
fluid/melt has a posi-

tive with relationship with chlorine therefore mirrors its behaviour with decompression, falling
from 32 to 20. DS

fluid/melt increases during decompression with values starting at 15 (200 MPa)
and rising to 1100 at near surface conditions.

DCl
fluid/melt parametrisations of Tattitch et al. (2021) were used to model bulk fluid compositions

during closed system decompressional degassing. However by employing these DCl
fluid/melt val-

ues (from ∼20 at depth, to <5 near the surface), models were unable to replicate the degassing
trend which is marked by decreasing melt Cl concentrations with decreasing MgO wt%, cap-
tured by existing Yasur melt inclusions (Métrich et al., 2011). To fit the decompression models
to the melt inclusion data presented in Chapter 6, I kept the relationship between DCl

fluid/melt and
pressure the same, but increased DCl

fluid/melt values by a factor of ∼4 – 5 bigger than those ob-
served in experiments. DCl

fluid/melt started at 70 at 200 MPa and decreased to 60 at 0.5 MPa, in
each run (Figure 3.9). Fluid salinity and molinity were calculated as in section 3.2.3. Since
DCu

fluid/melt is a function of fluid salinity, its parametrisations remain unchanged as the effect of
decompression is already encompassed by use of the equation 3.19.

3.3.4 Modelling sulfide saturation and sulfide-silicate melt partitioning at
Yasur

Methods for modelling sulfide saturation are as described in section 3.2.4, where PySulfSat
(v.1.0.3) (Wieser and Gleeson, 2023) is used to model the S6+/Stotal, SCSS2− and SCAS of

61



3.3. Degassing and sulfide saturation of chalcophile elements during decompression

the melt through decompression at different temperatures (1050◦C and 1075◦C) and starting
magma compositions (Tuk3 and Tan2318). Melt inclusion datasets report the total S concen-
tration (ppm) of the melt, therefore a correction must be applied to the SCSS2− output by these
models in order to correctly calculate the total sulfur solubility (SCSStot) of the melt:

SCSStot =
SCSS2−

1− (S6+/Stotal)
(3.42)

In section 3.2.4, several parametrisations for modelling the S6+/Stotal of the melt were in-
troduced (Nash et al., 2019; O’Neill and Mavrogenes, 2022; Jugo et al., 2010). The model of
Nash et al. (2019) is discarded due to its temperature dependency (see section 3.2.4). Using
parametrisations of O’Neill and Mavrogenes (2022) (OM22) or Jugo et al. (2010) (J2010) to
model the S6+/Stotal of the melt, made negligible differences to the overall SCSStot output by
my models (Figure 3.10), therefore for consistency with the modelling presented in section 3.2,
I continue to use the equations by Jugo et al. (2010) to model S6+/Stotal.

The SCSS2− was modelled using the parametrisation of O’Neill (2021) as in section 3.2.4.
The resultant SCSStot are shown in Figure 3.10; SCSStot generally increases with decompres-
sion up to a certain pressure before decreasing towards the surface (Figure 3.10b). For a given
melt composition, the SCSS2− correlates positively with melt fO2 (Jugo et al., 2010; Nash et al.,
2019), FeOt (Jugo et al., 2010; Nash et al., 2019; O’Neill, 2021; Smythe et al., 2017) but nega-
tively with pressure (Matjuschkin et al., 2016). These models show that during decompression,
crystallisation causes melt compositions to change (specifically FeOt), which competes with
the effect of falling pressure on the SCSStot of the melt (Figure 3.10c). Due to the limited
availability of data, the SCSS2− is modelled under the assumption that the composition of the
fractionating sulfide remains constant, with Fe/(Fe+Ni+Cu) = 0.6 (Wieser et al., 2020; Ding
and Dasgupta, 2018). The SCAS is implemented within the model using parametrisations by
Zajacz and Tsay (2019).

I apply the same parametrisations for DCu
sulf/melt that were modelled in section 3.2.5 where

all sulfide in the melt is assumed to be present in the form of sulfide liquid (SL) up to 3.6 wt%
FeOt, and there after as monosulfide solid solution (MSS) (Figure 3.7c).

62



Chapter 3. Modelling crystallisation, degassing and sulfide saturation in magmas

Figure 3.10: All models presented here are run at ∆QFM and 1075 to 1050◦C. Plots a) and
b) show the evolution of SCSStot with MgO wt% and pressure, respectively. In c) the correla-
tion between melt FeOt and SCSStot is demonstrated with the Tuk3 model run at 1075◦C and
1050◦C. Both Tuk3 and Tan2318 use the model of O’Neill (2021) to calculate the SCSS2− and I
show that calculating S6+/Stotal from either O’Neill and Mavrogenes (2022) or Jugo et al. (2010)
has almost no effect on the overall SCSStot calculated. Green lines in c) show the concentration
of FeOt during decompression and its positive relationship with the SCSStot of the melt (in red
and blue).

3.3.5 Modelling isobaric crystallisation at Yasur

The isobaric crystallisation models presented in Chapter 6 apply the same methods and parametri-
sations that are introduced in section 3.2. I use a primary basaltic Yasur sample, Tuk3, for the
initial magma composition input to fractional crystallisation models presented in Chapter 6.
The starting magma CO2 and Cl concentrations are the same as those input to decompression
models using Tuk3. Initial magma water concentrations are varied from 0.5 to 1.0 wt% H2O
in line with the range reported for basaltic melt inclusions from Yasur (Métrich et al., 2011).

63



3.3. Degassing and sulfide saturation of chalcophile elements during decompression

Models were run at 200 MPa to be consistent with pressure range estimated from melt inclu-
sions (Métrich et al., 2011) and seismic data (Bani et al., 2013). Additional models were run
between 400 and 100 MPa but made no significant difference to the conclusions drawn from
this modelling. Three subsets of models were run to determine if sulfide saturation could be
avoided during fractional crystallisation, and whether this could reconstruct the Cu signatures
of the melt inclusions discussed in Chapter 5. The differences in these models are highlighted
in Table 3.4.

Model # Si ppm ∆QFM Comment Colour in Figure
1 500 0 conditions most similar to Yasur red

2 500 +1.5 oxidised model green

3 200 0 low S model pink

Table 3.4: Different starting parameters used for fractional crystallisation models that are dis-
cussed in Chapter 6. All models were run at 200 MPa and with H2O concentrations of 0.5
and 1.0 wt%. Model #1 has geochemical conditions most similar to those observed at Yasur;
Model #2 increases the oxidation state of the melt and leaves all other geochemical parameters
as original values; Model #3 lowers the initial concentration of S (ppm) in the melt, all other
geochemical parameters remain the same.

64



4
Water-rich magmas optimise volcanic chal-
cophile element outgassing fluxes

This chapter is an adaptation of the work published as: “Water-rich magmas optimise volcanic
chalcophile element outgassing fluxes.”, Hogg, O.R., Edmonds, M. and Blundy, J., 2023. Earth

and Planetary Science Letters, 611, p.118153. https://doi.org/10.1016/j.epsl.2023.

118153.

Co-author contributions: In this manuscript, O. R. Hogg was responsible for conceptuali-

sation, methodology, investigation, visualisation, writing and editing the final publication. M.

Edmonds supervised this work and contributed to the conceptualisation, methodology, writing

and editing of the manuscript prior to publication. J. Blundy supervised this work and con-

tributed to editing prior to final publication. Electronic Appendix B is relevant to this work.

Abstract

Magmatic-hydrothermal fluids transport chalcophile metals to the atmosphere as volcanic
gases; and to the crust, where they may play a role in the formation of ore deposits. Global
volcanic gas datasets show considerable variability in the flux and composition of metals out-
gassed between volcanoes, but the controls on this variability are unclear. Magmatic chloride
is a key ligand for metal transport but magmatic water dominates the exsolved fluid reservoir

65

https://doi.org/10.1016/j.epsl.2023.118153
https://doi.org/10.1016/j.epsl.2023.118153


4.1. Introduction

into which metals partition during crystallisation and decompression. Here I develop models
simulating decompression-driven degassing (‘first boiling’) and isobaric crystallisation-driven
degassing (‘second boiling’) of magmas to show that while moderate concentrations of chlorine
are essential for metal partitioning into the magmatic fluids, magmatic water contents have the
greatest potential to control the mass yield of metals carried by exsolving fluids. My models
explain why water-rich magmatic-volcanic systems like Mount Etna (Italy) deliver the largest
mass fluxes (per unit of degassing magma) of metals to the atmosphere, whereas relatively dry
magmatic-volcanic systems like Yasur (Vanuatu) deliver the smaller mass fluxes. These re-
sults establish the important role of magmatic water in generating large reservoirs of metal-rich
aqueous fluids in the shallow crust.

4.1 Introduction

Volcanoes outgas volatile trace chalcophile metals and metalloids, such as lead (Pb), copper
(Cu), arsenic (As) and bismuth (Bi), into the atmosphere as gases and particulates at rates
comparable in magnitude to the industrial budget of entire countries (Ilyinskaya et al., 2021;
Nriagu, 1989). Volcanoes therefore constitute an integral part of the global geochemical cycle
for volatile trace metals, which involves them being transported to the crust by melt, partitioned
into an exsolved magmatic volatile phase, then discharged to the atmosphere during volcanism.
Deposition to the surface environment, as toxic pollutants or as bio-reactive molecules essential
for life (Dupont et al., 2010), is followed by weathering or biological uptake, and eventual
removal to the oceans, where they may be incorporated into sediment, returned to the mantle via
subduction, and recycled via slab devolatilisation (Debret et al., 2014; Zheng, 2019). Volatile
trace metals typically speciate as chlorides, oxides or sulfates in the magmatic exsolved volatile
phase and in volcanic gases (Mason et al., 2021; Pokrovski et al., 2013; Zelenski et al., 2014)
and then rapidly condense into the aerosol phase after emission into the atmosphere (Allen
et al., 2002; Symonds et al., 1987). In volcanic gas and aerosol plumes volatile trace metals
are typically enriched over the parent silicate melt by 10 – 105 times relative to refractory, non-
volatile elements such as Mg (Edmonds et al., 2018; Mather et al., 2012; Zelenski et al., 2021).
Gas-melt partition coefficients may be calculated for each element based on their abundance in
gases relative to melts (as melt inclusions or submarine glasses, prior to degassing), providing
a measure of their volatility (Zelenski et al., 2021). The chalcophile metals and metalloids such
as Te, Se, Tl, Bi, Cd and As are relatively abundant in volcanic gases (Zelenski et al., 2021),
with gas-melt partition coefficients of >>1.

It has been observed that gases and aerosols emitted from arc volcanoes have specific as-
semblages of trace metals that are distinct from those outgassed by hotspot volcanoes (Edmonds

66



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

et al., 2018; Zelenski et al., 2021). Arc-related volcanic gases are rich in elements such as Pb,
Bi, Tl, Pl, Cd and Sb, and relatively poor in elements such as Se and Te. These compositional
differences in volcanic gas plumes are linked to both the slab flux of fluid-mobile elements (Cox
et al., 2019) and to the high magmatic chlorine contents of arc magmas leading to enrichment
in the exsolved volatile phase of those metals that speciate with chloride (Edmonds et al., 2018;
Mandon et al., 2019; Zelenski et al., 2014).

Trace volatile metals partition into an exsolved magmatic volatile phase (MVP) that forms
during magma storage and transport in the crust through first boiling (decompression-induced
degassing) and second boiling (degassing induced by isobaric crystallisation) (Audétat et al.,
1998; Candela, 1997). The MVP comprises principally H2O and CO2 with lower concentra-
tions of chlorine, fluorine and sulfur and a range of trace species, including metals and met-
alloids. The role of chlorine in influencing metal partitioning behaviour (Zajacz et al., 2008;
Webster et al., 2018; Grondahl and Zajacz, 2022) into crustal MVPs (Audétat and Zhang, 2019;
Tattitch et al., 2021; Webster et al., 1999) and volcanic gases (Mandon et al., 2020, 2019; Ze-
lenski et al., 2021) has been elucidated previously: chloride is an important ligand for many
chalcophile metals, e.g., Zn, Pb and Cu, although the latter may have an equally high affinity
for sulfur species (Iveson et al., 2019; Mandon et al., 2020; Zajacz et al., 2017, 2008). Experi-
ments and thermodynamical modelling suggest that partitioning of Cl-complexing elements is
strongly influenced by the salinity of the MVP (Audetat et al., 2008; Tattitch et al., 2021; Web-
ster et al., 2015; Zajacz et al., 2017, 2008; Dolejš and Zajacz, 2018). However, the partitioning
of elements such as As and Se, which do not speciate with chloride, is likely independent of
salinity (Zajacz et al., 2008). Owing to the strong partitioning of chlorine into the exsolved
fluid phase during crystallisation during sub-volcanic magma storage, especially at elevated
pressures, Cl-rich magmas typical of those found in arcs will exsolve single-phase fluids with a
bulk salinity between 5 – 15 wt% NaCl (Audétat and Edmonds, 2020). During decompression
and cooling, these magmas may exsolve a brine which is significantly more saline, reaching up
to 70 wt% NaCl for those formed during shallow phase separation (Lecumberri-Sanchez and
Bodnar, 2018). Chlorine concentrations in silicate melts at the point of fluid exsolution corre-
late with the efficiency of Pb, Zn and Cu (and other Cl-complexing elements) sequestration into
the MVP (Audétat, 2019; Hsu et al., 2019; Tattitch et al., 2021).

As well as chlorine, H2O plays a key role in providing the reservoir into which metals par-
tition. Magmatic H2O has long been linked to the efficiency of metal extraction from magmas
linked to porphyry ore systems (Chiaradia, 2020; Lee and Tang, 2020; Loucks, 2014; Rezeau
and Jagoutz, 2020; Richards, 2018). Global geochemical compilations of arc whole-rock data
suggest high primary trace metal concentrations are of subordinate importance for the forma-
tion of large porphyry copper deposits (PCDs –Audetat et al. 2008; Audétat 2019; Barber et al.
2021; Chiaradia and Caricchi 2017; Chiaradia 2022); instead, it has been suggested that ele-

67



4.2. Global ore-forming fluid and volcanic gas datasets

vated magmatic H2O in subduction-related parental basalts may be an important precursor to
PCD formation (Chiaradia, 2020; Rezeau and Jagoutz, 2020; Richards, 2018; Loucks, 2014).
High Sr/Y ratios typical of porphyry-forming magmas are linked to amphibole (± garnet) frac-
tionation and plagioclase suppression, which are promoted by high dissolved H2O contents
consistent with extensive differentiation of parental arc basalt (Chiaradia, 2014; Lee and Tang,
2020). Water-rich magmas saturate in an exsolved MVP at higher melt fractions compared to
water-poor magmas, hence producing considerably greater fluid reservoir masses during crys-
tallisation (1.5 to 7 times higher; Chiaradia and Caricchi 2017; Rezeau and Jagoutz 2020).
Equation 3.1 showed that for a closed system comprised of melt and an MVP in equilibrium,
the initial magmatic H2O concentration determines the mass fraction of the exsolved MVP at
any particular pressure or degree of crystallisation, moreover explicitly accounts for the mass of
metals rather than the concentration of metals in the fluid. Increasing the salinity of magmatic
fluids can increase their metal-carrying capacity (by effectively raising their DX

MVP/melt), while
increasing dissolved magma H2O concentrations will increase G. Additionally, increasing the
metal concentration in the melt at the onset of vapour saturation (MX

Melt) will result in higher
concentrations of metals partitioning into the MVP. Each factor will ultimately increase the
mass of trace metals that are fluxed to the atmosphere via volcanic gases (which scales with
MX

MVP); the question is which exercises the strongest control in natural systems? Here I de-
scribe the generation of an MVP during fractional crystallisation and decompression of magma
in the crust underlying active volcanoes to understand the wide variability in degassing fluxes
of metals. In particular, I explore which magma pathways (e.g., depth of storage, degree of
crystallisation, volatile content, chlorine content and initial melt metal concentration) optimise
the mass flux of metals to shallow levels in the crust via an MVP, and how I might elucidate
these different pathways in natural systems. In doing so, I gain an understanding of the role of
magmatic water in transporting metals from deep crustal magmatic systems to the surface (via
volcanoes) or near-surface environments (ore deposits).

4.2 Global ore-forming fluid and volcanic gas datasets

Chloride has long been recognised as an important ligand for trace metals and metalloids in hy-
drothermal ore-forming fluids (Dolejš and Zajacz, 2018; Tattitch et al., 2021; Pokrovski et al.,
2013; Williams-Jones and Heinrich, 2005). Figure 4.1a and 4.1b shows X/Cl and X/S mass
ratios for quartz-hosted fluid inclusions from a range of ore-forming granite systems (Audétat,
2019; Audétat and Pettke, 2003; Audétat and Zhang, 2019; Seo et al., 2009; Zajacz et al., 2008),
for data compilation please refer to Electronic Appendix B, Table B.4). Fluid inclusions range
from higher pressure single-phase inclusions to lower pressure brine and vapour inclusions,
formed by unmixing of the single phase supercritical fluid during decompression (Hedenquist

68



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

and Lowenstern, 1994) (Figure B.1). Also shown are data for volcanic gases (for data com-
pilation please refer to Electronic Appendix B, Table B.5). The X/Cl ratios of volcanic gases
overlap with those of low-pressure brine fluid inclusions for Sn, As, Cu, Mo and Tl (Figure
4.1a), albeit with much lower X/S ratios (Figure 4.1b), reflecting both the important role that
chlorine plays in speciation and transport of metals and the much higher salinity (10 – 70 mol%;
Shinohara 2013) of most of fluid inclusions compared to volcanic gases (Audétat and Pettke,
2003; Audétat, 2019). However, the relative abundances of chalcophile metals in volcanic gases
are overall similar to the fluid inclusions, suggesting they provide a window into magmatic flu-
ids at depth.

69



4.2. Global ore-forming fluid and volcanic gas datasets

70



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

Figure 4.1: The composition of volcanic gas/aerosol plumes compared with the composition of
deeper exsolved fluids, represented by fluid inclusions hosted by quartz. Metals on the x-axis
are ordered by increasing volatility from left to right (Audétat, 2019). Quartz-hosted fluid inclu-
sions are distinguished into single phase fluid (Audétat, 2019; Audétat and Zhang, 2019; Zajacz
et al., 2008), brine (Audétat and Zhang, 2019; Seo et al., 2009; Williams-Jones and Heinrich,
2005) and vapour (Audétat and Pettke, 2003; Seo et al., 2009; Williams-Jones and Heinrich,
2005; Zhang and Audétat, 2018) in order to represent the MVP (see Electronic Appendix B,
Table B.4 for dataset); Citations for volcanic gas and aerosol data are reported in Table B.5; a)
X/Cl ratios and b) X/S ratios, where X is a trace metal. Each point refers to the median X/Cl
and X/S of all data (from all volcanoes) and bars mark the interquartile ranges for a given metal
within a given data group; c) highlighting the entire range of X/S mass ratios for individual gas
and aerosol plumes (see legend), with median points joined with a solid line. The similar mag-
nitudes of X/Cl ratios in volcanic gases and deeper exsolved fluids (a) imply that the majority
of metals enriched in either of these phases tend to speciate with Cl. In contrast, S-normalised
metal ratios (b) are almost two orders of magnitude higher in the fluids relative to gases, re-
flecting the metal rich nature of these fluids due to their elevated salinity. Fluid inclusion S
data incur a lot of uncertainty due to interferences of 16O2 on 32S (Audétat, 2019; Zajacz et al.,
2008). Despite normalising to S to remove the effect that different degassing rates have on the
abundance of metals in volcanic gases, data still show intense variability across different arc
settings.

There are significant inter-volcano differences in the mass fluxes of metals delivered to
the atmosphere via the gas and aerosol phases outgassed from volcanoes. A global review of
volcanic gas/aerosol flux data shows that both the absolute mass flux of trace metals (Figure
4.2a), magma supply-rate normalised mass fluxes (Figure 4.2b), and the mass flux of trace
metals relative to sulfur (Figure 4.1b) varies considerably. Mass fluxes of Zn exceed 104 kg/day
from Mount Etna (Italy), compared to 103 kg/day for Ambrym and 102 kg/day for Yasur volcano
(Vanuatu) (Gauthier and Le Cloarec, 1998; Mandon et al., 2019) (Figure 4.2a). Thallium (Tl)
fluxes are 9000 kg/day for Etna, compared to 100 kg/day for Ambrym (Vanuatu) and 1 kg/day
for Yasur Volcano. All of these volcanoes are characterised by a dominant mode of ‘passive
degassing’, whereby the gas flux far exceeds that expected from the degassing of erupted magma
and requires an extensive degassing of unerupted magmas deeper in the crust (Edmonds et al.,
2022; Shinohara, 2008). These deeper, crustal magmas supply volatiles through a combination
of first and second boiling (Audétat et al., 1998; Candela, 1997), with magma convection also
likely playing a role in delivering magma to low pressures beneath the volcano (Edmonds et al.,
2022; Shinohara, 2008). The degassing magma flux varies from volcano to volcano depending
on a variety of factors, including the magma supply rate from the mantle and the magma ascent
rate beneath the volcano. However, differences in magma flux alone cannot account for the

71



4.2. Global ore-forming fluid and volcanic gas datasets

observed trace element flux variations, because there are also key differences in normalised
fluxes. For example, Zn/S and Tl/S ratios are two orders of magnitude higher in the gas/aerosol
plume from Mount Etna than for Yasur volcanoes (Figure 4.1c).

Variations between the mass flux of degassing trace metals between volcanoes suggest that
there are factors which differ between systems that optimise the outgassing of trace volatile
metals (and their enrichment in the magmatic exsolved volatile phase in general), other than
magma supply and degassing rates. It has been suggested that the two most important fac-
tors controlling variability in outgassing metal flux between volcanoes are magmatic water
(Richards, 2018)and chlorine (Shinohara and Toshimichiiiyama, 1989; Tattitch et al., 2021;
Williams-Jones and Heinrich, 2005) contents. Primitive basaltic melt inclusions from Yasur
and Etna record Cl concentrations of ∼1000 ppm (Métrich et al., 2004, 2011) and from Am-
brym up to 450 ppm (Moussallam et al., 2021). Primary basalts from Etna have a water content
of 3.4 wt% (Métrich et al., 2004), whereas primitive basalts from Yasur and Ambrym volca-
noes contain 1 – 1.5 wt% (Métrich et al., 2011; Moussallam et al., 2021). Therefore these two
volcanoes represent suitable end-members to study the effects of variable H2O and Cl contents
on variations of mass flux of degassing of trace metals.

The aim of this chapter is to explore, via modelling and the examination of natural datasets,
the effects of varying magmatic water and chlorine contents on both the composition and mass
flux of metals produced via the exsolved volatile phase, and to apply this understanding to nat-
ural volcanic systems. Ultimately this analysis can help us to understand not only volcanic
emissions and their impact on our environment, but also how chalcophile elements are pro-
cessed through the crustal magmatic system and the controls on mass yield of these elements in
exsolved fluids.

72



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

Figure 4.2: Mass fluxes (kg/day) of trace metals carried via gas and aerosol plumes at three arc
volcanoes and one hotspot. Metals ordered as in Figure 4.1. Metal outgassing fluxes in kg per
day (a) and kg per kg of magma (b) from Etna (Gauthier and Le Cloarec, 1998; Aiuppa et al.,
2003), Yasur (Mandon et al., 2019), Ambrym (Allard et al., 2016) and Kilauea (Mather et al.,
2012) show significant variation. Circles represent single mass flux measurements for a metal
at a given volcano. Published SO2 fluxes from Etna (Allard et al., 2006; Edmonds et al., 2022;
Coppola et al., 2019), Yasur (Woitischek et al., 2020), Ambrym (Allard et al., 2016) and Kilauea
(Anderson and Poland, 2016) were used to estimate the magma degassing flux. The size of each
circle is scaled to bulk magmatic water content of the associated volcano, as determined from
the maximum water contents analysed in melt inclusions. Note the greater fluxes of Cl and
Cl-speciating elements emitted from volcanoes with higher magmatic water contents relative to
more water-poor magmatic systems.

73



4.3. Methods

4.3 Methods

I model the development of the MVP during isobaric crystallisation and decompression using
five hypothetical arc magmatic systems that differ in initial magma H2O, CO2 and Cl contents
(Table 3.1 and 3.2) – see Chapter 3.1 for full details of methods used for modelling. All systems
are assumed to be sulfur-free. I use the solubility model of MagmaSat (Ghiorso and Gualda,
2015) to model the H2O and CO2 concentrations in the melt and coexisting MVP. Chlorine
and fluid-mobile chalcophile element partitioning into the MVP is modelled using closed and
open system partitioning models (Electronic Appendix B, Tables B.1 and B.2). To replicate the
different behaviours of groups of selected chalcophile elements, three hypothetical metals X,
Y and Z, distinguished by their partition coefficients are defined: X (moderately volatile and
Cl-complexing, e.g., Pb), Y (highly volatile and Cl-complexing, e.g., Bi) and Z (highly volatile
and non-Cl-complexing, e.g., Se). I assume that all metals begin with the same concentration
(10 ppm) in the primitive melts in the first set of models, so that the effect of varying the
volatile contents can be extracted from the results. I also consider the impact of changing initial
chalcophile element concentrations (Appendix B, Figure B.2 to B.5), however this does not
changed the results of our research.

4.4 Results

4.4.1 Models of degassing of trace metals during second boiling and crys-
tallisation

The behaviour of chlorine and trace metals during second boiling is shown in Figure 4.3. In
general, it can be seen that during crystallisation, while the concentration of chlorine does not
vary to a great extent in the melt phase (Figure 4.3a), concentrations of chlorine in the MVP
increase as the melt evolves from basaltic compositions to more silicic compositions (Figure
4.3b). Consequently, the mass flux of chlorine generated via the MVP (in kg Cl in the MVP
per kg of parental magma) also increases (Figure 4.3c). The most saline MVP is generated by
systems (both water-poor and water-rich) with the highest concentrations of magmatic chlorine
(A and E in Figure 4.3), and the least saline fluids are generated by the most water-rich and
chlorine-poor melts, regardless of pressure (Figure 4.3b). The highest pressures are associated
with the most saline MVPs. However, when assessing which case produces the highest mass of
chlorine in the MVP per unit of crystallising parental magma relative to the original mass of the
parental basalt (i.e. F = 1, Chapter 3.1), i.e. the flux, the role of magmatic water becomes impor-

74



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

tant. Whilst Cl concentrations (i.e. salinity) in the MVP are predominantly dependent upon the
bulk chlorine contents of a magma system, the mass fluxes of Cl via the MVP are instead con-
trolled by the magma’s bulk water contents. I find that water-rich magmas, with high-moderate
chlorine concentrations (cases A and E; Figure 4.3c), crystallising at the shallowest pressures
deliver the highest mass of chlorine in the MVP via second boiling.

In terms of trace metal systematics, I see trends that mirror those for chlorine for the met-
als that speciate as chloride (X and Y). The most Cl-rich cases are associated with the most
saline MVP and with the most X- and Y-enriched MVP (A and E, superimposed in Figure
4.3b). The highest pressure MVPs are most metal-rich (Figure 4.3). The water-rich and Cl-
poor cases (D) generate less metal-rich MVPs (Figure 4.3b, second and third columns). As
expected, the highest metal concentrations are generated in the MVPs co-existing with the most
evolved, silica-rich melts. The largest masses of X and Y are generated in the water-rich, high-
and moderate-Cl cases (C and E), regardless of pressure. For element Z, whose partitioning
behaviour does not depend on Cl, the same trends are observed: the highest concentrations of Z
are observed in MVPs generated from the most evolved melts; and the highest masses (per unit
of magma) generated in the MVP for the water-rich case (Figure 4.3c, 4th column). Concentra-
tions across the five modelled systems are identical for the non-chloride speciating element Z.
Increasing the initial metal contents of the primitive melts in the models also has the effect of
scaling up the metal fluxes in the MVP (Figure B.2 and B.3).

4.4.2 Models of decompression degassing of trace metals

For the case of first boiling, the behaviour of chlorine and trace metals is shown in Figure 4.4. To
simulate a natural scenario, these magmas were equilibrated with a fluid during crystallisation
to 50% melt remaining (approximately equivalent to a basaltic andesite), and then this fluid
was released (Chapter 3.1). This scenario represents an end-member case; if there are fluids
coexisting with the magma prior to decompression, they are described by the second boiling
cases in Figure 4.3. The most shallowly-stored magmas have the highest Cl contents (∼1000
ppm; Figure 4.4a), owing to the weaker partitioning of chlorine into the MVP at low pressure
(Shinohara and Toshimichiiiyama, 1989; Tattitch et al., 2021). The most saline MVP, with the
highest concentrations of metals X and Y, is generated by the single-step decompression of
chlorine-rich magmas (regardless of initial water contents) from the shallowest storage regions,
although the dependence of MVP composition on pressure is weak (the MVP is 10% richer in Cl
when decompressed from 80 MPa compared to one decompressed from 400 MPa with a similar
initial magma volatile content; Figure 4.4b). Total mass yields of Cl, X and Y delivered by the
MVP are highest in the water-rich Cl-rich case but are similar for decompression of water-poor

75



4.4. Results

Figure 4.3: The behaviour of Cl and trace metals during second boiling. Symbols designate the
pressure at which magmas undergo crystallisation, and colours refer to different cases defined
in Table 3.1 The model describes the composition of the melt and MVP during crystallisation
where a melt fraction of 1 represents a basalt and 0.1 represents a rhyolite. Plots show a)
concentrations of Cl, element X, Y and Z in the melt against melt fraction; b) concentrations of
Cl, element X, Y and Z in the MVP; c) cumulative mass of each element in the MVP (kg per kg
of magma) – note that the total mass of fluid (H2O + CO2) in kg per kg of magma is 10−2 for the
water-rich cases (C-E) and 10−3 for the water-poor cases (A,B). The most concentrated Cl and
metal signatures are associated with deep fractionation of Cl-rich systems – with no distinction
across different bulk volatile contents (e.g., systems A and E are superimposed). The greatest
mass yields of Cl and each element are however achieved in shallow fractionating, water-rich
systems with moderate to high concentration of Cl.

76



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

and water-rich Cl-bearing (high-moderate) magmas (Figure 4.4c), suggesting that, in contrast to
second boiling, magmatic water contents play a less prominent role in metal transport, likely due
to extensive open system degassing simulated in the model prior to decompression. Increasing
the initial metal concentrations of each magma results in a linear increase in the total mass flux
of X, Y and Z delivered by the MVP (Figure B.4c and B.5c).

4.5 Discussion

Our models of MVP development during first and second boiling highlight that chalcophile
element concentrations and fluxes via the MVP are the result of an interplay between (1) the
magma’s dissolved H2O concentration, which governs the total mass of the MVP and moderates
the salinity of the MVP reservoir, hence controlling the partitioning of Cl-speciating metals; (2)
the depth of the storage region, which controls the pressure-dependent fluid-melt partition co-
efficients of chlorine and hence the partitioning of chloride-speciating metals; (3) the extent of
volatile loss due to open system degassing prior to magma decompression; and (4) the con-
centration of metals in the melt at the onset of fluid saturation. Whereas deeply-stored Cl-rich
(H2O-poor or H2O-rich) magmas generate the most saline and metal-rich MVPs, the greatest
chalcophile element yields in the MVP (and volcanic gas fluxes at the surface) are associated
with the second boiling of shallow water-rich magmatic systems underlying active volcanoes
where magmas undergo extensive crystallisation; or with magmas that have undergone pro-
longed crystallisation at depth (which concentrates metals in the silicate melt; Figure 4.3a), and
then decompressed to the surface.

In order to generate high fluxes of chloride-speciating metals (X and Y in the model, which
represent chalcophile metals such as Pb and Bi), a moderate to high concentration of magmatic
chlorine is also required, to act as a ligand. In the case of non-chloride speciating metals and
metalloids, such as Se, As, magmatic chlorine has no impact on degassing efficiency and the flux
of metals is independent of magmatic chlorine contents, and is solely dependent on the water
contents of the magma. Mass fluxes of both Cl-speciating and non-Cl speciating elements can
be further maximised if their concentration at the point of volatile saturation is elevated, which
is essentially facilitated by increased concentrations in the initial magma.

Optimum chalcophile element mass yields via the MVP beneath volcanoes may be gen-
erated by a vertically-protracted, crystallising magmatic storage system in the crust whereby
metals and volatiles are concentrated in evolving melts at depth and these melts migrate up into
upper crustal reservoirs, generating a metal-rich saline MVP feeding persistently active vol-
canic gas plumes (Huber and Parmigiani, 2018). Eruptions may generate short-lived pulses in

77



4.5. Discussion

Figure 4.4: The behaviour of Cl and trace metals during first boiling. Each data point refers to
an MVP concentration or flux from decompression of a 50% isobarically fractionated melt (F =
0.5, see Figure 4.3). Pressures refer to the depths at which these 50% isobarically fractionated
melts were stored prior to decompression; a) decompressed MVP concentrations of Cl, X, Y
and Z with falling depth of storage prior to decompression; b) concentrations of Cl, X, Y and
Z in the decompressed melt; c) mass flux (kg per kg of magma) of Cl, X, Y, Z. The total mass
of fluid (H2O + CO2) in kg per kg of magma for the water-rich and water-poor case is ∼10−2,
but with water-rich systems reaching consistently higher fluxes. Note the lower overall fluxes
compared to second boiling (Figure 4.3c). Maximum mass yields are now satisfied by deeper
fractionating magmas followed by rapid decompression to the surface (0.1 MPa).

78



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

volatile (including chalcophile element) gaseous fluxes from otherwise persistently-degassing
volcanoes (Figure 4.4c). For water-poor mafic magmatic systems (e.g., Yasur, Vanuatu), de-
compression degassing may contribute a similar or greater (by 1–2 orders of magnitude) yield
of chalcophile elements via the MVP than the second boiling process at depth (Figure 4.3c,
4.4c), but the overall mass yield of chalcophile elements in the MVP remains less than that for
water-rich systems (Figure 4.3c, 4.4c).

4.5.1 Comparison of models with data from global volcanoes

Data from global volcanic systems show that the highest normalised metal (both chloride and
non-chloride-speciating) gas/aerosol fluxes are generated by Etna, Tolbachik (Kamchatka) and
Stromboli (Italy) volcanoes (Figure 4.1b), which are all well known for being particularly water-
rich arc volcanic systems (primary basaltic melts have 3.4, 2.9 and 3.6 wt% respectively; Allard
et al. 2000; Gauthier and Le Cloarec 1998; Portnyagin et al. 2008; Zelenski et al. 2014). The
lowest normalised metal fluxes are observed at Yasur and Ambrym volcanoes (Figure 4.1b),
which are associated with relatively water-poor basaltic magmas (1.0 and 1.5 wt%, respectively;
Allard et al. 2016; Métrich et al. 2011). In general I see that Cl-speciating metal (e.g., Pb, Zn,
Cd, Bi) fluxes are highest at volcanoes characterised by higher magmatic water contents: using
a time-average degassing magma flux measured at Etna (Allard et al., 2006), I calculate Etnean
gas and aerosol plumes emit up to 104 kg/day of Zn (∼ 10−5 kg per kg magma), >3000 kg/day
of Pb (10−6 kg per kg magma), and >500 kg/day of Bi (10−7 kg per kg magma) (Gauthier
and Le Cloarec, 1998) compared to up to 103 kg/day of Zn (10−7 kg per kg magma), ∼ 30
kg/day of Pb (10−8 kg per kg magma), and 4 kg/day of Bi (10−9 kg per kg magma) emitted
by Yasur (Mandon et al., 2019) (using the time-average degassing flux from Woitischek et al.
2020, Figure 4.5). The observed outgassing flux of chalcophile elements from Etna and Yasur
are consistent with the results from our model (Figure 4.5), which predict a flux of up to 461
Pb (10−7 kg per kg magma) and 600 Bi kg per day (10−7 kg per kg magma) for Etna and
up to 10 kg Pb (10−8 kg per kg magma) and 26 kg Bi per day (10−7 kg per kg magma) for
Yasur. Primitive basalts from Etna and Yasur record concentrations of Zn between 70-150
ppm (Dupuy et al., 1982; Ferlito and Lanzafame, 2010). Fluids generated from my models,
starting with an initial melt metal concentration of 100 ppm, deliver mass fluxes of up to 3657
kg/day for the water-rich case and 68 kg/day for the water-poor case, which adhere to Zn mass
fluxes observed in the natural datasets for Etna (112–16198 kg/day) and Yasur (2–103 kg/day)
(Gauthier and Le Cloarec, 1998; Mandon et al., 2019) (Figure B.6). In these basaltic systems,
which exsolve low salinity fluids relative to those exsolved from more evolved magmas, brine
segregation from gas prior to outgassing is likely minimal.

79



4.5. Discussion

Figure 4.5: Modelled trace metal mass fluxes compared to natural data. Relative mass fluxes
in kg metal per kg of magma (a) and observed fluxes in kg per day (b) of volatiles (H2O,
Cl) and metals (Pb, Bi, Se) listed along the x-axis, compared to fluxes predicted by isobaric
crystallisation and decompressional degassing models, with initial metal concentrations of 10
ppm. The SO2 flux from Etna is used to estimate the magma degassing flux (Allard et al.,
2006; Coppola et al., 2019; Edmonds et al., 2022) and for Yasur, a magma degassing flux of
2.1 × 108 kg magma per day (estimated by Woitischek et al. 2020) is used. Natural mass flux
data for Yasur and Etna are represented by stars (median) and range bars refer to minimum and
maximum values where data are available. Circles represent modelled data presented in Figure
4.3 and 4.4. Only fluxes from isobarically derived MVPs associated with melt fractions of
greater than 0.5 are shown in order to replicate basaltic-trachybasaltic magmas that are erupted
by Etna (Métrich et al., 2004) and Yasur (Métrich et al., 2011).

80



Chapter 4. Water-rich magmas optimise volcanic chalcophile element outgassing fluxes

4.6 Conclusions

The volcanic flux and composition of outgassing volatile trace metals is highly variable between
volcanoes. I present a set of models to investigate how the initial magma volatile content (H2O,
Cl) may influence the flux of metals outgassed from a volcano. Our models describe the compo-
sition and mass of exsolved magmatic fluids generated through isobaric crystallisation (second
boiling) and decompressional degassing under different conditions of magma volatile content.
Although chlorine is an essential ligand for many chalcophile elements, I show that it is the
water content of a magma that has a greater importance for optimising the outgassing fluxes
of trace volatile metals from volcanoes by increasing the overall mass of the volatile phase
reservoir (see also Chiaradia and Caricchi 2017, for similar conclusions on Cu). I find that
the shallowest-stored water-rich crystallising magmas exsolve fluids with relatively low con-
centrations of metals, but high mass yields of these metals in the exsolved magmatic volatile
phase (per unit of crystallising magma). Equally, extensive fractionation of the deepest stored
water-rich magmas, followed by decompression and degassing, yield the greatest mass yields
of chlorine and trace volatile metals in the exsolved magmatic volatile phase.

Therefore, whilst high magmatic chlorine contents maximise metal concentrations in ex-
solved fluids, it is water that maximises fluxes. By identifying a decoupling in the conditions
satisfying maximum trace element concentrations and mass fluxes I provide a potential ex-
planation for the lack of evidence to support that PCD-hosting mineralised magmatic systems
require particularly metal-rich fluids (Audétat, 2019; Chiaradia and Caricchi, 2017; Rezeau and
Jagoutz, 2020). Instead, I expect that fluids delivering the highest mass yield of metals (likely
derived from water-rich fractionating magmas) play a more prominent role in the formation of
magmatic-hydrothermal ores. The importance of high metal concentrations in single phase flu-
ids and hypersaline brines as a prerequisite for ore formation has recently been contested, with
evidence of high temperature quartz-hosted fluid inclusions from barren and mineralised sys-
tems (Audétat 2019 and references therein) bearing comparable concentrations of trace metals.
In some cases, barren suites have recorded even higher concentrations of metals like Cu and
Mo than those found in fluid inclusions from mineralised systems (Audétat, 2019), alluding - in
tandem with our models - concentration to be a subordinate measure of ore-forming potential.

81



5
Sulfide resorption by water-rich melt yields
copper-rich magmatic fluids

This chapter is an adapted version of the manuscript “Sulfide resorption by water-rich melt
yields copper-rich magmatic fluids” by Hogg, O.R., Edmonds, M., Wieser, P., Gleeson, M.,
Jenner, F., and Blundy, J., currently in review in the journal of Nature Communications. Sup-
porting information that will be published alongside this work can be found in Chapter 3.2 and
Appendix C.

Co-author contributions: In this manuscript, O. R. Hogg was responsible for conceptu-

alisation, methodology, investigation, visualisation, writing and editing the final publication.

M. Edmonds supervised this work and helped with the conceptualisation and methodology. P.

Wieser and M. Gleeson contributed to the model and method development. F. Jenner and J.

Blundy helped with the development of ideas for this manuscript. All co-authors above, pro-

vided initial feedback on this manuscript and contributed to editing prior to its submission.

Abstract

Crustal enrichments in copper are of interest because of increasing global demand for Cu related
to the energy transition. The abundance of Cu in silicate melts is strongly influenced by sulfide
phase saturation and volatile degassing during fractionation. The relative importance of these

82



Chapter 5. Sulfide resorption by water-rich melt yields copper-rich magmatic fluids

processes is key in controlling whether Cu and other chalcophile elements become sequestered
in crustal cumulates or transported via hydrous fluids to upper crustal reservoirs. Here I quantify
the impact of magmatic water on fractionation, degassing and sulfide saturation during crustal
evolution of arc magmas and the implications for the abundance and distribution of sulfur (S)
and Cu in magmatic fluids. My models show that ubiquitous sulfide saturation is a critical limi-
tation on the Cu and S load of exsolved magmatic fluids. However, when sulfide-undersaturated
water-rich melts percolate through and resorb accumulated sulfides in crustal reservoirs at low
pressures, Cu-rich fluids may be generated. These results are consistent with the role of long-
lived crustal mush zones being a critical component for the mass transfer of Cu, S and water to
the upper crust via magmatic fluids.

5.1 Introduction

The strive towards a carbon-neutral economy comes with increased an demand for a number of
so-called critical metals, including chalcophile elements such as Cu. Over 70% of global Cu
supply is linked to porphyry copper deposits (PCDs) that are associated with large crustal mag-
matic intrusions of intermediate composition at convergent margins (Sillitoe, 2010). Economic
enrichments of Cu in the shallow crust are thought to be formed by the precipitation of sulfides
from metal-enriched, saline fluids derived from underlying magma bodies in the crust (Silli-
toe, 2010). Mineralisation is influenced by several parameters such as tectonic regime (Loucks,
2021), fluid focusing (Park et al., 2021), long-lived thermal sustainability (Chelle-Michou et al.,
2017; Chiaradia and Caricchi, 2017) and precipitation efficiency (Chiaradia, 2022). However,
a critical preliminary step in the genesis of these deposits, is the generation of Cu-rich fluids;
therefore, quantifying how S and Cu-rich fluids form in magmas and under what conditions
their concentrations are optimised, is key to understanding why economically important PCDs
are generated above some subduction zones but not others. Understanding the magmatic path-
ways of Cu in the crust is complex, since chalcophile elements partition strongly into both
sulfide and exsolved volatile phases and their partitioning behaviour depends on factors such as
temperature, pressure, oxidation state, melt composition and magma volatile concentrations (Li
and Audétat, 2015; Zelenski et al., 2021). Additional complexity arises because of the potential
of some magmas to assimilate pre-existing sulfides during their transit through crustal mush
zones that were generated during previous magmatism (Xia et al., 2023).

Intermediate arc volcanic rocks (with 3 – 7 wt% MgO) display a wide range in bulk Cu
concentrations from <50 to >600 ppm (Richards, 2015) (Figure 5.1a). Giant PCDs require
high masses of S and Cu to form and are often found in continental settings associated with
relatively Fe-poor and Cu-poor calc-alkaline rocks, which raises questions regarding how the

83



5.1. Introduction

Figure 5.1: Global whole rock volcanic arc rock compositions (from GEOROC,
https://georoc.eu/), showing a) Cu and b) FeOt against MgO concentrations. Data are colour-
coded for individual arcs (refer to Appendix D.1 for filtering method). Models incorporating
crystallisation, sulfide saturation and degassing are shown for different pressures and water
concentrations and for ∆QFM+1, +1.2 and +1.4. Initial melt Cu and S concentrations are 75
and 1000 ppm respectively. For details of the modelling, see Chapter 3.2. Water-rich systems
(H2O > 3 wt%) follow calc-alkaline trends whereas water-poor systems (H2O < 3 wt%) follow
tholeiitic trends.

Cu was lost from the magma during fractionation and whether this process is important in the
mass transfer of these elements to sites of mineralisation (Chiaradia, 2014; Lee et al., 2012).
Primitive magmas (with MgO >7 wt%) from different tectonic settings display a limited range
in Cu (70 – 100 ppm) (Richards, 2015), comparable to the range in primitive MORB (Jenner,
2017). Hence, despite porphyries being found exclusively at convergent margins, these magmas
do not appear to be predisposed to high Cu concentrations, which hints at the importance of
crustal processing and sulfide assimilation for the development of their ore-forming potential
(e.g., Xia et al. 2023).

The range in whole rock Cu concentrations that are produced during magmatic differentia-
tion (Figure 5.1a) may be influenced by differences in the thickness of the overriding arc crust
(Chiaradia, 2014) or by differences in magmatic water concentrations (Rezeau and Jagoutz,
2020; Loucks, 2014). A thick crust may promote high pressure fractionation of garnet, that de-
pletes the melt in FeO (Lee and Tang, 2020); and high magmatic water concentrations promote
the high temperature fractionation of Fe-rich phases (e.g., amphibole), resulting in Fe-poor calc-
alkaline magmas (Barber et al., 2021; Grove et al., 2003) (Figure 5.1b). Decreasing melt FeO

84



Chapter 5. Sulfide resorption by water-rich melt yields copper-rich magmatic fluids

content during fractionation lowers the sulfur content at sulfide saturation (SCSS2−) of a melt,
consequently causing sulfide fractionation and the development of Cu-depleted calc-alkaline
melts (Barber et al., 2021; Chiaradia, 2014; Lee et al., 2012).

The dissolved water concentrations of primitive arc basalts can vary significantly, from <1
up to >6 wt% (Klein et al., 2023). Crystallisation during cooling drives magmas to even higher
water concentrations up until the point of water saturation. More water-rich magmas (i.e., those
with initially higher concentrations of dissolved water), exsolve a volatile phase at higher melt
fractions, higher crustal pressures, and exsolve far greater masses of fluids relative to water-poor
systems during differentiation (Rezeau and Jagoutz, 2020). The concentration of dissolved wa-
ter in magmas also influences the relative timing and/or depth of degassing and sulfide saturation
(Li and Zhang, 2022). Sulfur, Cu and chlorine (Cl) partition strongly into fluids exsolved from
oxidised, evolved silicate melts at mid to upper crustal pressures (Scaillet et al., 1998; Zelenski
et al., 2021), yet the effect of degassing on the partitioning of S and Cu between the melt, sulfide
and exsolved fluid phases has not been fully explored. From the perspective of understanding
the evolution of magmas that fuel PCD formation, it is of critical importance that the influence
of degassing is constrained. Previous studies have demonstrated the importance of sulfide re-
sorption for producing Cu-rich magmas and fluids (Heinrich and Connolly, 2022; Reekie et al.,
2019). Interaction between accumulated sulfides in shallow crustal reservoirs (e.g., crystal-rich
crustal mush zones) and intruding sulfide-undersaturated melts may be a mechanism for devel-
oping Cu-rich aqueous magmatic fluids and melts in settings associated with PCD formation
(e.g., Xia et al. 2023), but the process has not yet been explored quantitatively.

There have been no published attempts to model simultaneously, sulfide saturation and de-
gassing during magma fractionation to establish their effects on the mass distribution of S and
Cu among melt, fluid and sulfide reservoirs; or to extend this modelling to understand the role
of melt mixing and sulfide assimilation in generating Cu-rich magmatic fluids. Here I model a
series of hypothetical arc magma systems varying only in their initially dissolved magma water
concentrations (0 to 6 wt% H2O). I investigate the relative importance of sulfide saturation and
degassing in causing Cu depletion in evolving melts and establish the optimal conditions for
generating Cu-rich fluids. Using these models, I demonstrate that assimilation of pre-existing
sulfide cumulates in the crust by water-rich sulfide-undersaturated silicate melts can signifi-
cantly increase the Cu endowment of aqueous magmatic fluids.

5.2 Modeling chalcophile element behaviour in arc magmas

To evaluate the role of increasing primary melt H2O concentrations on the partitioning of chal-
cophile elements between melts, sulfides and fluids, I model isobaric crystallisation, degassing

85



5.3. Results

and sulfide saturation in incremental steps starting with a primitive basalt (with 12 wt% MgO).
I model the composition of the melt during crystallisation and degassing from liquidus tem-
peratures to 800◦C for a range of magma water concentrations (0.1, 1.0, 3.0 and 6.0 wt%),
storage pressures (50 MPa to 400 MPa) and oxygen fugacities ( ∆QFM+1 to +1.2) using Rhyo-
liteMELTS (Ghiorso and Gualda, 2015) run through the open-source Python tool, PetThermo-
Tools (Wieser, 2023) (Chapter 3.2). The S and Cu concentrations of the starting basaltic melt
were fixed in these models (1000 ppm and 75 ppm, respectively), although I acknowledge vari-
ation exists in nature (Chapter 3.2). The effect of varying initial melt S and Cu concentrations
are explored (Figure 3.4) but are not significant for understanding the principal controls on S
and Cu enrichment in the magmatic fluids.

Sulfur dissolved in silicate melts exists as both sulfide (S2−) and sulphate (S6+); the rela-
tively oxidised nature of arc magmas means that high proportions of total dissolved S may exist
as S6+ (Jugo et al., 2010). I use the S speciation model of Jugo et al. (2010) (Figure 3.5) imple-
mented in the open-source Python tool PySulfsat (Wieser and Gleeson, 2023) to calculate melt
S6+/Stotal which is then used to calculate melt S2− concentrations. The sulfur concentration at
sulfide saturation (SCSS2−) was calculated using the model of O’Neill (2021) and correlates
positively with melt FeO and fO2 (Chapter 3.2; Wieser and Gleeson 2023). I model S, Cl and
chalcophile element partitioning into an exsolving volatile phase using published fluid-melt par-
tition coefficients, whereby the partitioning of Cl and some chalcophile elements is linked to
fluid salinity and the partitioning of S depends on pressure (Chapter 3.2). The model tracks
the concentration of S2− in the melt and, once it exceeds SCSS2−, describes the fluid-melt par-
titioning of S2− and chalcophile elements into a sulfide phase, which is modelled as sulfide
liquid (SL) for melt FeOt concentrations of >3.6 wt%; and monosulfide solid solution (MSS)
below this threshold (Chapter 3.2 for justification). I use sulfide-melt partition coefficients from
published studies (Kiseeva and Wood, 2013; Li and Audétat, 2015) (Figure 3.3). If sulfide
saturation occurs, the model iteratively adjusts the exsolved volatile phase concentrations of
S and chalcophile elements to satisfy both fluid-silicate melt and sulfide-silicate melt partition
coefficients, creating more sulfide in each iteration as necessary (Chapter 3.2 for details).

5.3 Results

During the early stages of fractionation (MgO > 7 wt%) melt Cu concentrations increase across
all model systems at all pressures (Figure 5.1; Figure 3.8a) because the melts are not saturated in
a fluid or sulfide phase. During fractionation at sulfide-undersaturated and fluid-undersaturated
conditions, the mass of Cu contained in the melt remains constant even though its concentration
increases (Figure 3.8c). This distinction between mass and concentration is critical to under-

86



Chapter 5. Sulfide resorption by water-rich melt yields copper-rich magmatic fluids

Figure 5.2: The effect of magma water concentration and pressure on the mass distribution of
Cu in the melt-fluid-sulfide system for pressures of 100 and 400 MPa. Models shown were run
at ∆QFM+1.2. Cuphase is the mass fraction of the total Cu in the fluid, melt and sulfide. The
total mass of Cu in the system is the same in each panel. Modelling results for four different
magmatic water concentrations are shown: (a) 6 wt%; (b) 3 wt%; (c) 1 wt%; and (d) 0.1 wt%
(where no fluid exsolves from the melt).

standing Cu enrichment, mass transport and ore formation during differentiation (Hogg et al.,
2023); very large masses of low concentration phases can be more important than small masses
of very Cu-rich phases. Once a sulfide or fluid phase saturates, the mass of Cu in the melt de-
creases as it is lost to one or both phases according to its partitioning behaviour (Figure 3.8c).
Once the melt becomes sulfide-saturated, Cu partitions strongly into sulfide. Sulfide becomes
the dominant (by mass) host for Cu in the system at melt fractions of <40% for all model runs
(Figure 5.2).

The mass of Cu sequestered by sulfides is highest in water-poor systems (with initially 0-
0.1 wt% H2O) at lower pressures (Figure 5.2a-b) and by water-rich systems at higher pressures
(Figure 5.2c-d; Figure 3.8e-h). The models show that Cu begins to partition into exsolved fluids
at melt fractions of <50% for the most water-rich systems. Only fluids generated at shallow
pressures from the most water-rich magmas carry high mass fractions (up to 80%) of total Cu,
because intense degassing (Figure 3.2h) prevents a significant mass of sulfide from forming
(Figure 3.8e-h). Fluids account for only a minor mass fraction of total Cu in the system: up
to 1-2% in magmas containing 3 wt% H2O (Figure 5.2). Overall I observe that, at a given
oxidation state, fractionation of water-rich magmas generates Cu-poor melts, whereas more
water-poor magmas tend to evolve towards Cu-rich concentrations before Cu is depleted by

87



5.4. Melt interaction with sulfide cumulates generates copper-rich fluids

sulfide saturation, consistent with the trends observed in global convergent margin magmas
(Figure 5.1; Figure 3.4).

Water-rich systems (with initially 3-6 wt% H2O) experience degassing prior to sulfide satu-
ration (Figure 5.2a,b); whereas water-poor systems (0.1–1 wt% H2O) become sulfide-saturated
before they start degassing. The mass of Cu in an exsolved fluid coexisting with a sulfide-
saturated magma is low compared to that for a sulfide-undersaturated magma, reflecting the
three-way partitioning between sulfide, melt and exsolved fluid and the strong sequestration of
Cu into the sulfide phase (Figure 5.2). Consequently, from my simple models, I find that ex-
solved magmatic fluids are in general Cu-poor except in exceptional cases where either a very
high magma water concentration or high oxygen fugacity precludes the melt from becoming
sulfide saturated.

5.4 Melt interaction with sulfide cumulates generates copper-
rich fluids

High masses of Cu in fluids have been linked to the fractionation of volatile-undersaturated mag-
mas in the mid to lower continental crust, followed by transport of evolved, volatile-enriched
residual melts to upper crust (Chiaradia and Caricchi, 2017; Loucks, 2014). However, in agree-
ment with previous studies (Matjuschkin et al., 2016), these models show that sulfide saturation
is inevitable during magma fractionation, causing the concentration of Cu in most aqueous ex-
solved fluids to be low. Here I consider an alternative mechanism for generating Cu-rich fluids
that is supported by both modelling and natural observations.

Sulfide resorption has been observed directly in mineralised (Bajo de Alumbrera, Halter
et al. 2005; Park et al. 2021) and unmineralized (Kawa Igen, Berlo et al. 2014) volcanic sys-
tems and is considered to be an essential prerequisite for generating Cu-rich fluids with ore-
forming potential (e.g., Xia et al. 2023). This process has been detected cryptically through
the analysis of chalcophile element ratios (Cu/Ag, Cu/Au) in exsolved hydrous fluids (Berlo
et al., 2014; Halter et al., 2005; Nadeau et al., 2010), whole rocks (Xia et al., 2023), volcanic
glasses (Reekie et al., 2019), and through S isotope measurements and sulfide grain textures in
deep crustal sections (Holwell et al., 2022) or mafic enclaves (e.g., Georgatou et al. 2022) (for
further discussion see Appendix C, section C.2). Providing that sulfides are available to interact
with aqueous bubbles directly, S and Cu can transfer directly into these exsolving magmatic
fluids (Mungall et al., 2015).

88



Chapter 5. Sulfide resorption by water-rich melt yields copper-rich magmatic fluids

Here I model the resorption of sulfides (that may be present in a crustal mush; Jackson et al.
2018) by percolating sulfide-undersaturated melts (Figure 5.3). I show that up to 7 g of sulfide
may be assimilated per kg magma for the most primitive intruding magmas (i.e. those with
the lowest S2− concentrations) before the SCSS2− limit is reached (Figure 5.3c). Resorption
of sulfide will increase the S and Cu concentrations in both the melt and co-existing aqueous
magmatic fluid by up to 6 times, depending on the nature and composition of the sulfide phase
(Figure 5.3a). Water-rich melts that have resorbed sulfides will exhibit anomalously high Cu
concentrations (Halter et al., 2005; Park et al., 2021), and at low pressures will generate high
mass fluxes of Cu in the aqueous exsolved fluid phase (Figure 5.3).

5.5 Conclusions

I have developed models simulating crystallisation, sulfide saturation and degassing of mag-
mas under varying conditions of oxidation state, water content and pressure. The formation
of sulfides are ubiquitous in arc magmas and reduce the ability of exsolving magmatic fluids
to carry appreciable masses of Cu. This problem can be circumvented by: a) fractionation
of oxidised, hydrous magmas, where S degassing can sufficiently diminish melt S concentra-
tions and prevent sulfide saturation, thereby promoting the partitioning of Cu into the exsolving
fluids, or b) resorption of crustal sulfides (accumulated in long-lived mush systems) by sulfide-
undersaturated water-rich melts. I envisage that percolation of sulfide-undersaturated melts
through deep crustal mushes may generate Cu-rich melts in long-lived, mature and hydrous
magmatic systems (Jackson et al., 2018) and may then lead to Cu-rich magmatic fluids upon
degassing at lower pressures.

89



5.5. Conclusions

Figure 5.3: The effect of sulfide resorption on a) melt Cu concentrations (ppm); b) the concen-
tration of Cu in exsolved fluids (wt%) and c) the mass flux of Cu generated via the exsolved
fluids from the magma in g of fluid per kg magma. The composition of the intruding magma
ranges from basalt (melt fraction 1) to basaltic andesite (melt fraction 0.6) and contains 3 and
6 wt% H2O. All cases start with 75 ppm Cu and 1000 ppm S. Basaltic melts can resorb higher
masses of sulfide prior to becoming sulfide saturated compared to basaltic andesitic melts. The
most water-rich magmas generate the highest fluxes of Cu-rich fluids.

90



6
Decompression crystallisation forms copper-
rich melt signatures in arc magmas

This chapter comprises work done as part of a collaboration. O. R. Hogg was responsible for

sample preparation, data processing and modelling, conceptualisation, methodology, investi-

gation, and data visualisation. M. Edmonds supervised this work and helped with the concep-

tualisation, methodology development and editing of this chapter. F. Jenner helped with the

development of ideas for this chapter. I. Buisman, C.J de Hoog and B. Kunz assisted with data

acquisition and quality evaluation. L. Melekhova and J. Blundy undertook sample collection

for Yasur samples. Electronic Appendix A and D are relevant to this work.

Abstract

Global volcanic arc whole rock data shows that tholeiitic (TH, Fe-rich) volcanic rocks have
high Cu concentrations (up to 300 ppm) and calc-alkaline (CA, Fe-poor) rocks have lower Cu
concentrations. However, inspection of global arc melt inclusion data reveals a different pattern,
with FeO and Cu trends for TH and CA melt inclusions overlapping. Melt inclusion and whole
rock Cu concentrations reach up to 600 ppm. I present new data for melt inclusions in rocks
erupted from Yasur volcano, Vanuatu Arc. These melt inclusions preserve some of the highest
Cu concentrations (500 – 600 ppm) recorded by the global melt inclusion dataset. Diffusive
modification of the Cu contents in plagioclase-hosted melt inclusions may cause elevated melt

91



6.1. Introduction

Cu concentrations. However, this process cannot account for the scatter in the global dataset,
which is independent of phenocryst type. Using a combination of geochemical observations
and modelling, I show that degassing and subsequent crystallisation during magma ascent and
decompression can effectively suppress sulfide saturation through degassing of sulfur into ex-
solving water-rich fluids, leading to Cu enrichment in melts.

6.1 Introduction

Global volcanic whole rock data are commonly used to infer first order processes related to
magma differentiation (Chiaradia, 2014). Erupted volcanic rocks represent a mixture of phases
including crystals and melt, and represent the final product of a series of magmatic differenti-
ation and mixing processes during storage in the crust over a range of timescales (Turner and
Langmuir, 2015). It is challenging to infer liquid lines of descent (LLDs) using whole rock
data as the rocks may contain unknown quantities of mixed melt and/or accumulated crystals.
Furthermore, evidence of important magmatic processes like degassing and sulfide saturation,
which are essential for our understanding of the mass transfer of chalcophile elements like Cu
in magmatic systems, are commonly obfuscated in the whole rock record. Silicate melt inclu-
sions, formed by entrapment of melt during growth of host phenocrysts, represent snapshots
of melt evolution and hence preserve records of processes of interest that may be lost in the
compositional trends in the whole rocks (Rose-Koga et al., 2021).

A clear example of the value of melt inclusions over bulk rock compositions for deciphering
petrogenesis is demonstrated by the study of Cu behaviour in magmas (Figure 6.1). It is well
established that Cu systematics in magmas are linked intimately to the behaviour of sulfur (S)
which in turn is a function of melt FeOt and melt oxidation state (Jugo et al., 2010; Matjuschkin
et al., 2016; O’Neill, 2021). Concentrations of FeOt in melts correlate with sulfur concentrations
at sulfide saturation (SCSS2−) for a given fO2 (e.g., Wallace and Edmonds 2011). Most arc
magmas start with ∼75 – 100 ppm Cu (Figure 6.1a-b) and melt concentrations increase during
fractional crystallisation up to the onset of sulfide saturation (Cox et al., 2019) and intense
degassing (Hogg et al., in review). Copper partitions strongly into sulfide phases (sulfide-silicate
melt partition coefficients for Cu range from 102 – 104; Kiseeva and Wood 2013) and the melt
becomes depleted in Cu as crystallisation continues.

92



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Figure 6.1: Global arc melt inclusion (MI) and volcanic whole rock (WR) FeOt, and Cu con-
centrations. Data are screened for CA and TH compositions that are defined using the equation
of Miyashiro (1974). Left hand panel (a and d) show published MI data and new Yasur MI data
from this study (black). Middle panel (b and e) are published WR data. Right hand panel (c and
f) shows distribution of elements concentrations in CA (blue) and TH (pink) WRs (circles) and
MIs (diamonds). WRs have a distinct Cu-FeO-MgO demarcation that is not observed in MIs.
Density curves show MIs (dashed lines) tend to be richer in Cu than WRs (solid lines).

The global whole rock record shows that calc-alkaline (CA; Fe-poor) magmas are domi-
nated by Cu-poor signatures whilst arc tholeiites (TH; Fe-rich) are typically Cu-rich (Figure
6.1b-c). This relationship is well established in the literature (Chiaradia 2014 and references
therein) and has been explained by differences in the timing of sulfide saturation driven by Fe
systematics (Jenner et al., 2010). In CA systems, early fractionation of amphibole (± garnet)
causes depletion of Fe3+ and FeOt in the melt, thereby lowering the SCSS2− and promoting
sulfide saturation at high melt MgO concentrations (Figure 6.1b; Jenner et al. 2010; Barber
et al. 2021; Lee et al. 2012; Cox et al. 2019). The onset of sulfide saturation causes melt Cu
concentrations to plummet as Cu partitions strongly into sulfides. Tholeiitic magmas maintain
high melt FeOt concentrations due to extensive plagioclase fractionation >5 wt% MgO, causing
the SCSS2− to remain high and sulfide saturation does not occur (Jenner et al., 2015, 2010; Cox

93



6.1. Introduction

et al., 2019), allowing S and Cu to enrich in the melt during crystallisation. Sulfide saturation
may eventually be induced by the crystallisation of Fe-Ti-oxides (magnetite), which reduces the
FeOt content of the melt dramatically (Reekie et al., 2019). Just prior to sulfide saturation, melts
may contain up to 300 ppm Cu at 4 – 6 wt% MgO (Chapter 5, Hogg et al., in review). However,
some data depart from this trend and reach up to 600 ppm Cu in this MgO range (Figure 6.1b-c),
inconsistent with typical differentiation processes such as isobaric crystallisation (modelled in
previous chapters; Figure 5.1a).

The clear demarcation between CA and TH magmas in both melt FeOt and Cu concentra-
tions does not appear in the melt inclusion data (Figure 6.1a-c). Instead, melt inclusions from
TH and CA magmas show a very similar range and distribution in Cu concentrations (Figure
6.1a-c). As observed in the whole rock data, a small subset of evolved melt inclusions are
Cu-enriched, reaching up to 600 ppm (Figure 6.1a). Melt inclusions generally record higher
Cu concentrations than whole rocks - although the differences in TH whole rock and TH melt
inclusion Cu concentrations are far more subdued.

Cu-rich signatures may be generated during fractional crystallisation providing these mag-
mas remain sulfide-undersaturated throughout differentiation (Iveson et al., 2022; Deng et al.,
2022). Exhaustion of sulfides in the mantle source by high degrees of melting generates sulfide-
undersaturated primitive magmas and may be a means of generating Cu-rich arc magmas af-
ter crystallisation (Iveson et al., 2022; Deng et al., 2022). Previous work has also ascribed
melt Cu enrichment to sulfide resorption in crustal magma reservoirs (Berlo et al., 2014; Xia
et al., 2023). A mechanism to achieve sulfide resorption is by infiltrating hydrous, sulfide-
undersaturated basaltic magmas into crustal cumulate piles that contain previously accumulated
sulfides, thereby increasing the bulk Cu concentration of the magma (Hogg et al., in review;
Heinrich and Connolly 2022; Berlo et al. 2014). Resorption of in situ sulfides can also oc-
cur during magma ascent when the SCSS2− of the melt increases, combined with the effect of
extensive S degassing associated with decompression (Keppler, 2010; Nicholson et al., 2024;
Wieser et al., 2020). These processes are believed to influence the melt Cu signatures observed
at Kilauea, Hawaii, USA (Wieser et al., 2020) and Holuhraun, Iceland (Nicholson et al., 2024),
as recorded in melt inclusions. In this chapter I present a novel geochemical study of chalcophile
elements in melt inclusions from Yasur volcano, Vanuatu Arc, to add to the global dataset. The
work presented in this chapter aims to:

a) explain why the global melt inclusion TH and CA data do not show discrete trends with
respect to Cu concentrations similar to the whole rock global dataset.

b) explain the mechanisms responsible for generating Cu-rich melts (>300 ppm Cu).

94



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

By combining models of crystallisation, degassing and sulfide saturation I show that the evolu-
tion of melt Cu concentrations with MgO content during isobaric crystallisation are quite differ-
ent to those generated during decompression crystallisation. In addition, I show that extensive
sulfur degassing can largely suppress sulfide saturation and permit melts to maintain high Cu
concentrations on eruption. Post-entrapment processes may modify the compositions of melt
inclusions and lead to differences between the whole rock and melt inclusion datasets. I discuss
the mechanisms of Cu diffusion through plagioclase hosts into melt inclusions, which occurs
during melt water degassing during decompression (Zhang et al., 2023; Audétat et al., 2018),
and whether this process can account for scatter observed in the melt inclusion datasets. Ad-
ditional post-entrapment melt inclusion modifications are discussed, including post-entrapment
crystallisation (PEC) (Danyushevsky et al., 2000) and sulfide formation or resorption (Hartley
et al., 2018).

6.2 Sample preparation and methods

The major, minor, volatile and chalcophile element geochemistry of melt inclusions hosted in
olivine, plagioclase and clinopyroxene phenocrysts from Yasur were measured. Samples of
basaltic scoria were collected from Yasur by L.Melekhova and J. Blundy soon after eruption in
2016 (co-ordinates: 19◦ 31.807N,169◦ 27.005E) (Figure 2.1). Phenocrysts hosting glassy, ho-
mogenous silicate melt inclusions were prepared for analysis, as described in Chapter 2. Scoria
were crushed and sieved into different size fractions and selected phenocrysts were mounted
and polished to different grades (Figure 6.2). Major elements in glasses and hosts, and volatiles
(S and Cl) in glasses were analysed using EPMA. Spot analyses of phenocryst hosts were taken
adjacent to respective melt inclusions, and the major and minor element concentrations repre-
sent an average of three repeated measurements (Figure 6.3; Table 6.1, for complete EPMA
dataset see Electronic Appendix A3). The same procedure was taken for the rim and core anal-
yses of selected phenocrysts (Figure 6.3). H2O and CO2 concentrations in the glasses were
analysed by SIMS (Table 6.1, details of standards, methods, precision and accuracy are given
in Chapter 2 and Electronic Appendix A2, Tables A2.1 to A2.3). Finally, LA-ICP-MS was used
for analysis of chalcophile and trace elements in melt inclusions; details of the technique, stan-
dards, precision and accuracy can be found in Chapter 2 and Electronic Appendix A4, Tables
A4.1 to A4.10.

95



6.2. Sample preparation and methods

Figure 6.2: Transmitted a) and reflected b) light image of a representative homogenous silicate
melt inclusion (MI), here hosted in an olivine (ol) from sample T3o28. In c), a BSE map of
a T1 scoria thin section. Scoria were highly vesicular and generally consisted of ∼20–30%
phenocrysts of olivine, plagioclase (plag), clinopyroxene (cpx) and oxides, surrounded by a
glassy microlite poor groundmass.

96



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Ta
bl

e
6.

1:
M

aj
or

,m
in

or
,v

ol
at

ile
an

d
tr

ac
e

el
em

en
tc

on
ce

nt
ra

tio
ns

of
m

el
ti

nc
lu

si
on

s
fr

om
Y

as
ur

.C
on

tin
ue

d
on

ne
xt

pa
ge

.1

T
3o

28
m

1
T

2o
14

m
1

T
2o

18
m

2
T

2o
20

m
1

T
2o

25
m

1
T

1o
3

m
1

T
1o

3
m

2
T

3o
3

m
2

T
3o

24
m

1
T

3o
27

m
1

M
in

er
al

H
os

t
ol

ol
ol

ol
ol

ol
ol

ol
ol

ol
Sa

m
pl

e
T

3o
28

T
2o

14
T

2o
18

T
2o

20
T

2o
25

T
1o

3
T

1o
3

T
3o

3
T

3o
24

T
3o

27
Si

O
2

58
.6

6
58

.4
1

58
.4

6
58

.4
1

58
.6

8
58

.3
4

58
.3

0
58

.1
2

58
.3

6
58

.6
9

Ti
O

2
1.

01
1.

06
1.

02
1.

05
0.

99
1.

04
1.

03
1.

02
1.

02
1.

04
A

l 2
O

3
14

.6
7

14
.8

3
14

.8
0

14
.7

5
14

.5
5

14
.8

8
14

.6
0

14
.6

8
14

.5
5

14
.4

8
Fe

O
t

8.
97

9.
10

8.
92

9.
15

8.
96

8.
89

9.
25

9.
12

9.
00

9.
04

M
nO

0.
19

0.
20

0.
21

0.
19

0.
18

0.
23

0.
18

0.
20

0.
21

0.
22

M
gO

2.
71

2.
74

2.
70

2.
81

2.
76

2.
71

2.
78

2.
81

2.
76

2.
81

C
aO

5.
07

5.
03

5.
31

5.
02

5.
33

5.
28

5.
19

5.
44

5.
52

5.
31

N
a 2

O
3.

59
3.

56
3.

68
3.

57
3.

46
3.

50
3.

61
3.

61
3.

65
3.

49
K

2O
3.

86
3.

92
3.

85
3.

89
3.

84
3.

87
3.

88
3.

76
3.

74
3.

65
P 2

O
5

0.
66

0.
65

0.
66

0.
64

0.
68

0.
72

0.
66

0.
66

0.
61

0.
68

C
r 2

O
3

0.
01

0.
01

B
D

L
B

D
L

0.
01

0.
03

B
D

L
B

D
L

B
D

L
0.

02
SO

2
(p

pm
)

25
0.

97
26

5.
97

32
2.

48
32

5.
05

40
4.

74
23

7.
05

30
7.

37
14

7.
73

30
6.

55
27

9.
75

C
l(

pp
m

)
54

9.
76

45
2.

12
43

7.
15

51
1.

60
79

7.
57

51
9.

91
47

5.
92

49
5.

45
47

7.
45

54
3.

05
To

ta
l

99
.4

0
99

.5
0

99
.6

1
99

.4
7

99
.4

3
99

.4
7

99
.4

7
99

.4
2

99
.4

4
99

.4
1

M
g#

34
.9

8
35

.0
6

35
.0

5
35

.3
4

35
.4

3
35

.2
4

34
.8

5
35

.4
8

35
.3

3
35

.6
3

PE
C

m
et

ho
d

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

Pe
tr

ol
og

H
os

tF
o#

67
.0

5
67

.0
5

67
.2

1
67

.4
7

67
.2

0
67

.1
4

66
.9

3
67

.3
8

67
.2

4
67

.2
2

PE
C

(%
)

-0
.9

7
-0

.7
0

-0
.0

2
-1

.2
8

-0
.8

5
-0

.5
7

-0
.9

5
-3

.3
0

-4
.7

7
-3

.2
3

H
2O

(w
t%

,S
IM

S)
0.

45
0.

38
0.

24
0.

38
0.

43
0.

38
0.

38
0.

43
0.

42
0.

45
C

O
2

(p
pm

,S
IM

S)
58

.1
0

56
.5

3
16

.0
1

72
.5

1
46

.7
1

67
.5

2
32

.5
8

21
.7

2
57

.1
1

68
.9

0
C

u
(p

pm
)

54
8.

02
40

8.
40

45
1.

68
42

3.
42

46
0.

29
43

1.
27

42
8.

33
40

7.
39

36
7.

86
39

1.
07

2S
E

(p
pm

)
34

.0
0

17
.0

0
12

.0
0

25
.0

0
18

.0
0

49
.0

0
26

.0
0

46
.0

0
42

.0
0

50
.0

0
A

g
(p

pb
)

12
4.

83
10

2.
57

10
4.

91
93

.3
6

2S
E

(p
pb

)
13

.0
0

6.
10

5.
80

13
.0

0
C

u/
A

g
43

90
.1

7
39

81
.4

7
43

05
.4

9
45

35
.1

4

1 M
aj

or
el

em
en

ts
ar

e
re

po
rt

ed
in

w
t%

.
O

th
er

el
em

en
tc

on
ce

nt
ra

tio
n

un
its

ar
e

lis
te

d
in

th
e

ta
bl

e.
Fe

O
t

in
cl

in
op

yr
ox

ne
(c

px
)

an
d

pl
ag

io
cl

as
e

(p
la

g)
ho

st
ed

m
el

ti
nc

lu
si

on
s

is
ta

ke
n

as
Fe

O
fr

om
E

PM
A

an
al

ys
is

;F
eO

t
fo

r
PE

C
-c

or
re

ct
ed

ol
iv

in
e

(o
l)

ho
st

ed
m

el
ti

nc
lu

si
on

s
is

th
e

su
m

of
Fe

O
+

(0
.8

9
×

Fe
2O

3)
,w

he
re

Fe
2O

3
is

ou
tp

ut
by

Pe
tr

ol
og

3.
2S

E
re

fe
rs

to
th

e
in

te
rn

al
pr

ec
is

io
n,

ta
ke

n
as

th
e

st
an

da
rd

er
ro

ro
n

in
te

rn
al

st
an

da
rd

s
m

ea
su

re
d

by
L

A
-I

C
P-

M
S.

B
D

L
(b

el
ow

de
te

ct
io

n
lim

it)
.

97



6.2. Sample preparation and methods

Ta
bl

e
6.

1:
M

aj
or

,m
in

or
,v

ol
at

ile
an

d
tr

ac
e

el
em

en
tc

on
ce

nt
ra

tio
ns

of
m

el
ti

nc
lu

si
on

s
fr

om
Y

as
ur

.C
on

tin
ue

d
on

ne
xt

pa
ge

.2

T
2o

25
m

2
T

1o
4

m
1

T
1o

10
m

1
T

2o
15

m
1

T
2o

15
m

3
T

2o
18

m
1

T
3o

4
m

2
T

2o
29

m
1

T
1o

3
m

3
T

3o
27

m
2

M
in

er
al

H
os

t
ol

ol
ol

ol
ol

ol
ol

ol
ol

ol
Sa

m
pl

e
T

2o
25

T
1o

4
T

1o
10

T
2o

15
T

2o
15

T
2o

18
T

3o
4

T
2o

29
T

1o
3

T
3o

27
Si

O
2

58
.1

8
58

.5
0

58
.3

7
58

.6
1

58
.2

5
58

.3
0

58
.9

6
58

.3
5

58
.5

0
58

.5
4

Ti
O

2
0.

98
1.

02
1.

00
1.

02
1.

03
1.

00
1.

01
1.

00
1.

01
1.

00
A

l 2
O

3
14

.7
5

14
.6

2
14

.8
3

14
.6

7
14

.5
2

14
.6

7
14

.5
2

14
.7

9
14

.8
4

14
.6

7
Fe

O
t

9.
14

8.
92

8.
99

9.
00

9.
19

9.
21

8.
94

8.
89

8.
90

9.
03

M
nO

0.
22

0.
19

0.
19

0.
22

0.
17

0.
19

0.
15

0.
19

0.
21

0.
18

M
gO

2.
81

2.
76

2.
75

2.
74

2.
84

2.
86

2.
78

2.
74

2.
80

2.
73

C
aO

5.
20

5.
38

5.
47

5.
05

5.
62

5.
26

5.
32

5.
25

5.
13

5.
28

N
a 2

O
3.

59
3.

46
3.

56
3.

60
3.

52
3.

56
3.

48
3.

63
3.

55
3.

58
K

2O
3.

86
3.

80
3.

75
3.

90
3.

67
3.

79
3.

76
3.

83
3.

82
3.

78
P 2

O
5

0.
64

0.
71

0.
68

0.
66

0.
67

0.
68

0.
67

0.
65

0.
67

0.
62

C
r 2

O
3

0.
05

0.
01

B
D

L
0.

01
0.

00
0.

00
0.

01
0.

00
0.

01
B

D
L

SO
2

(p
pm

)
44

0.
83

31
2.

28
26

7.
51

22
7.

95
28

7.
64

24
5.

77
30

9.
54

36
8.

63
31

7.
46

31
9.

15
C

l(
pp

m
)

53
3.

08
43

9.
45

48
9.

07
52

3.
80

48
1.

00
45

0.
41

49
6.

32
44

6.
05

54
8.

24
47

6.
77

To
ta

l
99

.4
1

99
.3

6
99

.5
9

99
.4

7
99

.4
9

99
.5

2
99

.5
9

99
.3

1
99

.4
3

99
.4

1
M

g#
35

.3
8

35
.5

8
35

.2
4

35
.1

5
35

.5
5

35
.6

3
35

.6
8

35
.4

6
35

.8
9

34
.9

7
PE

C
m

et
ho

d
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
H

os
tF

o#
67

.3
4

67
.2

8
67

.0
7

67
.2

7
67

.1
7

67
.5

0
67

.4
2

67
.5

2
67

.8
3

66
.8

9
PE

C
(%

)
-2

.8
6

-1
.9

4
-3

.2
6

-3
.0

7
-3

.8
1

-1
.0

3
-1

.1
3

-0
.2

7
-1

.0
4

-2
.0

6
H

2O
(w

t%
,S

IM
S)

0.
44

0.
50

0.
26

0.
39

0.
37

0.
33

0.
55

0.
43

0.
45

C
O

2
(p

pm
,S

IM
S)

62
.8

0
83

.4
6

75
.4

6
86

.4
3

58
.9

8
22

.3
9

50
.7

8
16

6.
81

16
8.

58
C

u
(p

pm
)

41
4.

62
41

4.
40

43
0.

54
39

5.
76

38
6.

72
44

6.
94

41
1.

84
43

4.
57

42
7.

90
38

7.
68

2S
E

(p
pm

)
44

.0
0

59
.0

0
36

.0
0

28
.0

0
86

.0
0

98
.0

0
17

.0
0

16
.0

0
31

.0
0

18
.0

0

2 M
aj

or
el

em
en

ts
ar

e
re

po
rt

ed
in

w
t%

.
O

th
er

el
em

en
tc

on
ce

nt
ra

tio
n

un
its

ar
e

lis
te

d
in

th
e

ta
bl

e.
Fe

O
t

in
cl

in
op

yr
ox

ne
(c

px
)

an
d

pl
ag

io
cl

as
e

(p
la

g)
ho

st
ed

m
el

ti
nc

lu
si

on
s

is
ta

ke
n

as
Fe

O
fr

om
E

PM
A

an
al

ys
is

;F
eO

t
fo

r
PE

C
-c

or
re

ct
ed

ol
iv

in
e

(o
l)

ho
st

ed
m

el
ti

nc
lu

si
on

s
is

th
e

su
m

of
Fe

O
+

(0
.8

9
×

Fe
2O

3)
,w

he
re

Fe
2O

3
is

ou
tp

ut
by

Pe
tr

ol
og

.
2S

E
re

fe
rs

to
th

e
in

te
rn

al
pr

ec
is

io
n,

ta
ke

n
as

th
e

st
an

da
rd

er
ro

ro
n

in
te

rn
al

st
an

da
rd

s
m

ea
su

re
d

by
L

A
-I

C
P-

M
S.

B
D

L
(b

el
ow

de
te

ct
io

n
lim

it)
.

98



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Ta
bl

e
6.

1:
M

aj
or

,m
in

or
,v

ol
at

ile
an

d
tr

ac
e

el
em

en
tc

on
ce

nt
ra

tio
ns

of
m

el
ti

nc
lu

si
on

s
fr

om
Y

as
ur

.C
on

tin
ue

d
on

ne
xt

pa
ge

.3

T
3o

28
m

1
T

1o
7

m
1

T
1o

7b
m

1
T

2o
14

m
1

T
2o

15
m

4
T

2o
20

m
1

T
3o

24
m

2
T

1c
5

m
1

T
3c

15
m

1
T

3c
6

m
1

M
in

er
al

H
os

t
ol

ol
ol

ol
ol

ol
ol

cp
x

cp
x

cp
x

Sa
m

pl
e

T
3o

28
T

1o
7

T
1o

7b
T

2o
14

T
2o

15
T

2o
20

T
3o

24
T

1c
5

T
3c

15
T

3c
6

Si
O

2
58

.6
6

58
.0

2
58

.6
6

58
.4

1
58

.5
9

58
.4

1
59

.6
9

59
.5

5
58

.8
6

58
.4

8
Ti

O
2

1.
01

1.
08

1.
01

1.
06

1.
06

1.
05

1.
06

1.
09

1.
04

0.
78

A
l 2

O
3

14
.6

7
14

.8
5

14
.5

8
14

.8
3

14
.6

4
14

.7
5

15
.1

0
15

.3
1

14
.2

6
16

.0
8

Fe
O

t
8.

97
9.

18
9.

02
9.

10
9.

09
9.

15
8.

37
8.

93
9.

37
7.

33
M

nO
0.

19
0.

22
0.

20
0.

20
0.

20
0.

19
0.

17
0.

18
0.

21
0.

16
M

gO
2.

71
2.

83
2.

75
2.

74
2.

77
2.

81
1.

65
1.

40
2.

42
2.

71
C

aO
5.

07
5.

24
5.

32
5.

03
5.

13
5.

02
5.

72
3.

75
4.

82
4.

95
N

a 2
O

3.
59

3.
56

3.
49

3.
56

3.
51

3.
57

3.
60

3.
53

3.
40

3.
88

K
2O

3.
86

3.
82

3.
81

3.
92

3.
80

3.
89

3.
89

4.
20

3.
61

4.
29

P 2
O

5
0.

66
0.

64
0.

67
0.

65
0.

68
0.

64
0.

68
0.

77
0.

58
0.

77
C

r 2
O

3
0.

01
0.

02
0.

01
0.

01
0.

00
0.

00
B

D
L

B
D

L
B

D
L

B
D

L
SO

2
(p

pm
)

25
0.

97
34

5.
98

25
5.

65
26

5.
97

28
3.

70
32

5.
05

36
9.

33
65

.0
0

31
4.

00
40

3.
00

C
l(

pp
m

)
54

9.
76

47
8.

43
44

0.
92

45
2.

12
43

8.
89

51
1.

60
48

2.
33

50
9.

00
10

35
.0

0
50

2.
00

To
ta

l
99

.4
0

99
.4

5
99

.5
0

99
.5

0
99

.4
6

99
.4

7
10

0.
01

98
.7

4
98

.6
5

99
.5

1
M

g#
34

.9
8

35
.5

0
35

.1
9

35
.0

6
35

.1
9

35
.3

4
25

.9
9

21
.8

0
31

.4
9

39
.7

7
PE

C
m

et
ho

d
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
Pe

tr
ol

og
H

os
tF

o#
67

.0
5

67
.4

7
66

.9
7

67
.0

5
67

.0
8

67
.4

7
PE

C
(%

)
-0

.9
7

-1
.2

4
-1

.8
6

-0
.7

0
-1

.2
0

-1
.2

8
H

2O
(w

t%
,S

IM
S)

0.
45

0.
40

0.
35

0.
38

0.
39

0.
38

0.
44

0.
44

0.
44

0.
48

C
O

2
(p

pm
,S

IM
S)

58
.1

0
11

4.
90

53
.9

1
56

.5
3

58
.4

4
72

.5
1

67
.6

4
81

.6
0

18
7.

22
85

.1
3

C
u

(p
pm

)
44

1.
00

37
9.

58
43

2.
08

43
7.

21
44

6.
80

41
9.

48
46

4.
00

32
2.

00
33

6.
00

39
4.

00
2S

E
(p

pm
)

44
.0

0
20

.0
0

14
.0

0
15

.0
0

41
.0

0
16

.0
0

49
.0

0
29

.0
0

26
.0

0
32

.0
0

3 M
aj

or
el

em
en

ts
ar

e
re

po
rt

ed
in

w
t%

.
O

th
er

el
em

en
tc

on
ce

nt
ra

tio
n

un
its

ar
e

lis
te

d
in

th
e

ta
bl

e.
Fe

O
t

in
cl

in
op

yr
ox

ne
(c

px
)

an
d

pl
ag

io
cl

as
e

(p
la

g)
ho

st
ed

m
el

ti
nc

lu
si

on
s

is
ta

ke
n

as
Fe

O
fr

om
E

PM
A

an
al

ys
is

;F
eO

t
fo

r
PE

C
-c

or
re

ct
ed

ol
iv

in
e

(o
l)

ho
st

ed
m

el
ti

nc
lu

si
on

s
is

th
e

su
m

of
Fe

O
+

(0
.8

9
×

Fe
2O

3)
,w

he
re

Fe
2O

3
is

ou
tp

ut
by

Pe
tr

ol
og

3.
2S

E
re

fe
rs

to
th

e
in

te
rn

al
pr

ec
is

io
n,

ta
ke

n
as

th
e

st
an

da
rd

er
ro

ro
n

in
te

rn
al

st
an

da
rd

s
m

ea
su

re
d

by
L

A
-I

C
P-

M
S.

B
D

L
(b

el
ow

de
te

ct
io

n
lim

it)
.

99



6.2. Sample preparation and methods

Ta
bl

e
6.

1:
M

aj
or

,m
in

or
,v

ol
at

ile
an

d
tr

ac
e

el
em

en
tc

on
ce

nt
ra

tio
ns

of
m

el
ti

nc
lu

si
on

s
fr

om
Y

as
ur

.4

T
3c

4
m

1
T

2c
1

m
1

T
2c

5
m

2
T

2c
5

m
1

T
2c

8
m

1
T

3c
9

m
1

T
2c

1
m

5
T

2p
28

m
2

T
2p

28
m

4
T

2p
28

m
1

M
in

er
al

H
os

t
cp

x
cp

x
cp

x
cp

x
cp

x
cp

x
cp

x
pl

ag
pl

ag
pl

ag
Sa

m
pl

e
T

3c
4

T
2c

1
T

2c
5

T
2c

5
T

2c
8

T
3c

9
T

2c
1

T
2p

28
T

2p
28

T
2p

28
Si

O
2

57
.4

1
58

.2
6

59
.0

2
58

.6
3

58
.5

5
59

.1
5

58
.0

7
58

.6
0

59
.3

7
57

.5
3

Ti
O

2
0.

93
1.

02
1.

04
1.

03
1.

02
1.

02
1.

05
0.

98
1.

00
1.

06
A

l 2
O

3
15

.5
9

14
.7

3
14

.8
0

14
.6

0
14

.6
9

14
.8

4
14

.6
1

14
.7

1
14

.7
7

14
.7

6
Fe

O
t

9.
31

8.
80

8.
99

8.
97

8.
96

8.
51

8.
79

8.
39

8.
30

8.
90

M
nO

0.
21

0.
16

0.
20

0.
19

0.
25

0.
17

0.
19

0.
18

0.
15

0.
20

M
gO

2.
43

2.
55

2.
18

2.
31

2.
42

2.
34

2.
59

2.
44

2.
39

2.
49

C
aO

4.
93

5.
18

4.
70

4.
86

5.
01

4.
81

5.
30

4.
98

4.
87

5.
08

N
a 2

O
3.

82
3.

66
3.

49
3.

65
3.

59
3.

62
3.

57
3.

45
2.

83
3.

44
K

2O
3.

73
3.

86
3.

92
3.

76
3.

80
4.

00
3.

85
3.

95
3.

98
3.

88
P 2

O
5

0.
63

0.
63

0.
69

0.
65

0.
61

0.
69

0.
56

0.
63

0.
65

0.
69

C
r 2

O
3

0.
00

B
D

L
0.

00
0.

00
B

D
L

B
D

L
B

D
L

0.
02

0.
00

0.
00

SO
2

(p
pm

)
54

6.
00

43
7.

00
16

0.
00

22
1.

00
16

4.
00

32
7.

67
43

2.
00

27
9.

00
31

8.
50

47
9.

50
C

l(
pp

m
)

51
8.

50
42

9.
00

52
5.

00
52

3.
67

54
3.

00
44

7.
67

41
0.

00
43

3.
50

45
1.

50
49

2.
00

To
ta

l
99

.0
9

98
.9

3
99

.1
2

98
.7

3
98

.9
5

99
.2

2
98

.6
4

98
.4

0
98

.3
8

98
.1

4
M

g#
31

.8
0

34
.0

6
30

.1
8

31
.4

1
32

.5
3

32
.9

2
34

.4
2

34
.1

7
33

.8
7

33
.2

7
PE

C
m

et
ho

d
H

os
tF

o#
PE

C
(%

)
H

2O
(w

t%
,S

IM
S)

0.
44

0.
40

0.
42

0.
42

0.
37

0.
45

0.
40

0.
40

0.
41

0.
44

C
O

2
(p

pm
,S

IM
S)

12
7.

24
74

.7
5

61
.0

3
86

.8
9

66
.0

1
12

0.
26

64
.9

0
68

.0
0

77
.4

6
39

4.
11

C
u

(p
pm

)
24

7.
00

38
4.

00
34

9.
00

35
4.

00
39

7.
00

40
7.

00
38

3.
00

40
2.

00
37

3.
00

38
9.

00
2S

E
(p

pm
)

41
.0

0
20

.0
0

18
.0

0
15

.0
0

41
.0

0
17

.0
0

37
.0

0
26

.0
0

29
.0

0
29

.0
0

4 M
aj

or
el

em
en

ts
ar

e
re

po
rt

ed
in

w
t%

.
O

th
er

el
em

en
tc

on
ce

nt
ra

tio
n

un
its

ar
e

lis
te

d
in

th
e

ta
bl

e.
Fe

O
t

in
cl

in
op

yr
ox

ne
(c

px
)

an
d

pl
ag

io
cl

as
e

(p
la

g)
ho

st
ed

m
el

ti
nc

lu
si

on
s

is
ta

ke
n

as
Fe

O
fr

om
E

PM
A

an
al

ys
is

;F
eO

t
fo

r
PE

C
-c

or
re

ct
ed

ol
iv

in
e

(o
l)

ho
st

ed
m

el
ti

nc
lu

si
on

s
is

th
e

su
m

of
Fe

O
+

(0
.8

9
×

Fe
2O

3)
,w

he
re

Fe
2O

3
is

ou
tp

ut
by

Pe
tr

ol
og

3.
2S

E
re

fe
rs

to
th

e
in

te
rn

al
pr

ec
is

io
n,

ta
ke

n
as

th
e

st
an

da
rd

er
ro

ro
n

in
te

rn
al

st
an

da
rd

s
m

ea
su

re
d

by
L

A
-I

C
P-

M
S.

B
D

L
(b

el
ow

de
te

ct
io

n
lim

it)
.

100



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Ta
bl

e6
.1

:M
aj

or
,m

in
or

,v
ol

at
ile

an
d

tr
ac

e
el

em
en

tc
on

ce
nt

ra
tio

ns
of

m
el

ti
nc

lu
si

on
sf

ro
m

Y
as

ur
.T

hi
st

ab
le

is
al

so
av

ai
la

bl
e

in
E

le
ct

ro
ni

c
A

pp
en

di
x

D
.5

T
2p

23
m

1
T

2p
11

m
3

T
2p

25
m

1
T

2p
11

m
2

M
in

er
al

H
os

t
pl

ag
pl

ag
pl

ag
pl

ag
Sa

m
pl

e
T

2p
23

T
2p

11
T

2p
25

T
2p

11
Si

O
2

60
.9

6
56

.5
2

57
.2

9
57

.3
6

Ti
O

2
0.

85
1.

18
1.

08
1.

17
A

l 2
O

3
14

.9
6

14
.6

2
14

.7
0

14
.6

4
Fe

O
t

7.
58

10
.0

5
9.

52
9.

21
M

nO
0.

16
0.

21
0.

16
0.

23
M

gO
2.

11
2.

82
2.

70
2.

73
C

aO
4.

34
5.

49
5.

35
5.

19
N

a 2
O

3.
74

3.
44

3.
51

3.
50

K
2O

4.
37

3.
88

3.
87

3.
97

P 2
O

5
0.

49
0.

73
0.

70
0.

79
C

r 2
O

3
0.

01
0.

00
B

D
L

0.
01

SO
2

(p
pm

)
10

8.
00

51
4.

50
26

3.
50

42
4.

00
C

l(
pp

m
)

35
5.

00
56

5.
50

50
2.

00
48

4.
00

To
ta

l
99

.6
1

99
.0

3
98

.9
6

98
.9

0
M

g#
33

.1
1

33
.3

1
33

.5
7

34
.5

8
PE

C
m

et
ho

d
H

os
tF

o#
PE

C
(%

)
H

2O
(w

t%
,S

IM
S)

0.
39

0.
42

0.
33

0.
41

C
O

2
(p

pm
,S

IM
S)

86
.8

1
99

.5
8

19
9.

16
27

7.
01

C
u

(p
pm

)
36

5.
00

40
7.

00
47

0.
00

41
3.

00
2S

E
(p

pm
)

14
.0

0
23

.0
0

18
.0

0
15

.0
0

5 M
aj

or
el

em
en

ts
ar

e
re

po
rt

ed
in

w
t%

.
O

th
er

el
em

en
tc

on
ce

nt
ra

tio
n

un
its

ar
e

lis
te

d
in

th
e

ta
bl

e.
Fe

O
t

in
cl

in
op

yr
ox

ne
(c

px
)

an
d

pl
ag

io
cl

as
e

(p
la

g)
ho

st
ed

m
el

ti
nc

lu
si

on
s

is
ta

ke
n

as
Fe

O
fr

om
E

PM
A

an
al

ys
is

;F
eO

t
fo

r
PE

C
-c

or
re

ct
ed

ol
iv

in
e

(o
l)

ho
st

ed
m

el
ti

nc
lu

si
on

s
is

th
e

su
m

of
Fe

O
+

(0
.8

9
×

Fe
2O

3)
,w

he
re

Fe
2O

3
is

ou
tp

ut
by

Pe
tr

ol
og

3.
2S

E
re

fe
rs

to
th

e
in

te
rn

al
pr

ec
is

io
n,

ta
ke

n
as

th
e

st
an

da
rd

er
ro

ro
n

in
te

rn
al

st
an

da
rd

s
m

ea
su

re
d

by
L

A
-I

C
P-

M
S.

B
D

L
(b

el
ow

de
te

ct
io

n
lim

it)
.

101



6.3. Results

Figure 6.3: Compositions of melt inclusions (MI) and matrix glasses (MxGl) relative to host
(adjacent spot) (or for matrix glasses adjoining phenocryst) measured in natural data (Métrich
et al., 2011) and this study. Overlain are best-fitting decompression crystallization models gen-
erated using RhyoliteMELTS (Ghiorso and Gualda, 2015), using two different starting melt
compositions: Tuk3 and Tan23 (refer to Chapter 3.3 for explanation of models) (Métrich et al.,
2011). Mg numbers (Mg#) of melt inclusions and matrix glasses are compared to a) Forsterite
content (Fo) of the host olivines, b) Anorthite content (An) of the host plagioclase, and c) En-
statite content (En) of the clinopyroxene hosts. For matrix glasses, each phenocryst spot was
taken adjacent to the glass.

6.3 Results

6.3.1 Petrography and host geochemistry

Basaltic scoria contains phenocrysts of olivine, clinopyroxene and plagioclase and Fe-Ti oxides
in a glassy, microlite-free groundmass, with a modal proportion of ∼20–30% phenocrysts and
∼70–80% groundmass (Figure 6.2). Inspection of SEM backscatter images and EDS Sulfur
maps, show that crystal hosts, melt inclusions and vapour bubbles were free of a sulfide phase
(see Appendix D, Figure D.1). The forsterite content (Fo) of olivines in this study range from
66.9 to 67.8 (Figure 6.3a, Table 6.1, Electronic Appendix A3, Table A3.6), where:

Fo = 100 × Mg / (Mg + Fe2+)

Generally, core and rim analyses show no or very weak zoning, differing by less than 1 mol%
Fo (Figure 6.4a, Electronic Appendix A3, Table A3.7). Also shown are Yasur data published
by Métrich et al. (2011). Clinopyroxene crystals show a limited compositional range (En 43.1

102



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Figure 6.4: Core and rim analysis of phenocryst hosts for assessment of zoning for a) olivine,
b) plagioclase and c) clinopyroxene. Phenocrysts are generally unzoned, with core and rim
compositions differing by less than 1 mol%.

– 44.6; Electronic Appendix A3, Table A3.6) and zonation (Figure 6.3c, Table A3.7), where En
content is calculated as:

En = 100 × Mg / (Mg + Ca + Fe2+)

All but one core-rim pair are within 1 mol% En (Figure 6.4c). Only four plagioclase crys-
tals that hosted melt inclusions were analysed; cores and rims have anorthite contents (An)
ranging from 55.3 to 57.2 (Figure 6.3b, Electronic Appendix A3, Table A3.6), where:

An = 100 × Ca / (Ca + Na + K)

EPMA analysis of crystals showed that core-rim pairs varied by a maximum of ∼3 mol%
An and were absent of zoning besides one crystal that had very weak reverse zoning (Figure
6.4b, Table A3.7). Hosts adjacent to melt inclusions also showed this range in An, besides
one sample T2p11 where host spot analyses adjacent to melt inclusion #2 and #3 have An >

70, far more mafic than the rest (Figure 6.3b). Phenocrysts of magnetite were identified in
scoria samples by assessment of stoichiometry (Figure 2.6, Figure 6.2c, Electronic Appendix
A3, Table A3.9).

6.3.2 Melt inclusion geochemistry: major elements

Major element compositions of melt inclusions and matrix glasses from this study are plotted in
Figure 6.5 along with existing melt inclusion and whole rock data from Yasur and the rest of the

103



6.3. Results

Vanuatu Arc (Métrich et al., 2011; Dupuy et al., 1982; Deng et al., 2022; Sorbadere et al., 2011,
2013; Eggins, 1993, 1989; Gorton, 1974; Robin et al., 1993; Allard et al., 2016; Moussallam
et al., 2019, 2021; Sheehan and Barclay, 2016; Coulon and Maury, 1981; Carney and Mac-
farlane, 1982) (Appendix D, Figure D.3). Silicate melt inclusions are basaltic-trachyandesitic
in composition and have a limited compositional range (Figure 6.5; Table 6.1), despite being
entrapped in a variety of phenocryst hosts. For example, melt inclusions in the least evolved
olivines (Fo = 67.8) have a Mg# of ∼35.9 (Figure 6.3a), which is compositionally similar to
melt inclusions found in the most evolved plagioclases (An = 55.3) with a Mg# of 33.6 (Figure
6.3b). This implies that olivine and plagioclase (and clinopyroxene) were in equilibrium, and
that the melt inclusion compositions represent a cotectic melt. Olivine-hosted melt inclusions
were corrected for PEC using Petrolog3 (as described in Chapter 2) at ∆QFM and an initial
FeO∗ of 8.88 wt% (the average of matrix glasses acquired in this study). The extent of calcu-
lated PEC in the olivine-hosted melt inclusions is minor, reaching a maximum of 4.8%. The
PEC-corrected MgO contents of melt inclusions ranges from 2.7 to 2.9 wt% (Table 6.1). By
definition, PEC-corrected melt inclusions are in equilibrium with their olivine hosts (Figure
2.8a).

PEC corrections for plagioclase- and clinopyroxene-hosted melt inclusions are not as straight-
forward as for olivine, due to their more complex mineralogy (enabling more complex elemental
substitutions) and the processes that modify plagioclase- and clinopyroxene-hosted melt inclu-
sion compositions being less well understood (Neave et al., 2017). For a given Mg#, PEC would
increase the MgO contents of melt inclusions hosted in plagioclase relative to the general trend
of Yasur whole rocks and melt inclusions, and decrease CaO (in plagioclase and clinopyroxene-
hosted melt inclusions) and Al2O3 (in plagioclase-hosted melt inclusions) (Appendix D, Figure
D.2) (Neave et al., 2017; Nielsen, 2011). The data presented here show no obvious deviation
from existing Yasur datasets (Figure D.2). Since olivine, plagioclase, and clinopyroxene ap-
pear to be cotectic in these samples, I can tentatively extrapolate the degree of PEC observed
in olivine-hosted melt inclusions to those hosted in other phenocrysts. However, it is impor-
tant to note that the errors associated with Cu measurements, which range from 5% to 20%
(as detailed in Chapter 2 and Electronic Appendix A4), exceed the errors introduced by PEC.
Therefore, although a small degree of PEC likely occurred in plagioclase- and clinopyroxene-
hosted melt inclusions, this effect is not expected to significantly impact the results of this study.
Consequently, I have chosen not to correct these melt inclusions for PEC.

The silicate melt inclusion compositions presented here are relatively evolved compared to
other published data from Yasur (Métrich et al., 2011) and the rest of the Vanuatu Arc (Figure
6.5). Taken together, the Yasur data record a fractionation trend: as melt MgO drops, K2O in-
creases and FeOt, Al2O3 and CaO decrease, caused by concomitant fractionation of plagioclase
and clinopyroxene (Figure 6.5d-f).

104



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

105



6.3. Results

Figure 6.5: Major element systematics in melt inclusions (MIs), whole rocks (WRs) and matrix
glasses from Yasur (blue and turquoise) and the rest of the Vanuatu Arc (grey) overlain by LLDs
output by decompression crystallisation models generated using RhyoliteMELTS (Ghiorso and
Gualda 2015; Chapter 3.3). Whole rockss are shown as crosses; mlet inclusions are shown
as circles (olivine-hosted), triangles (plagioclase-hosted) and squares (clinopyroxene-hosted).
Starting compositions of models have been highlighted (Tuk3 and Tan2318; Métrich et al.
2011; Dupuy et al. 1982; Deng et al. 2022) as black circles. Tuk3 models are shown in red
and Tan2318 models in purple. Nodes on each of the modelled LLDs mark changes in melt
fraction of 0.1 (i.e. 10% crystallisation). For Tuk3 models nodes begin at a melt fraction of 0.4
(60% crystallised) and for Tan2318 models, nodes begin at 0.5 (50% crystallised). Each line
represents a single isobaric decompression model run - temperatures are highlighted in b) but
are not central to the discussion.

6.3.3 Melt inclusion geochemistry: volatile elements

The H2O and CO2 concentration of the melt inclusions ranges from 0.24 to 0.55 wt% (± 2.4%)
and 16 to 394 ppm (± 4.1 to 11.4%) respectively (Figure 6.6a; Table 6.1; Electronic Appendix
A2, Tables A2.1 and A2.3). Yasur samples T3o1, T2o7, T3o6 underwent Raman spectroscopy
analysis of the vapour bubble but not SIMS analysis of the glass due to breaching of the raster
pit with the host phenocryst. The vapour bubbles that were analysed yielded concentrations of
CO2 ranging from 72 to 983 ppm (Electronic Appendix A1, Table A1.3), suggesting that the
glass-only analysis of inclusion CO2 may be underestimates of the total inclusion CO2, as has
been found for other volcanic systems (Hartley et al., 2014; Moore et al., 2015). Concentrations
of H2O in the Yasur samples analysed by Métrich et al. (2011) range from 0.81 to 2.0 wt%
(Figure 6.6a).

Isobars are plotted on Figure 6.6a using the open source Python package VESICAL (Ia-
covino et al., 2021) using MagmaSat (Ghiorso and Gualda, 2015), which is a mixed-volatile
solubility model calibrated over a wide PT and melt compositional range (0 – 20 kbar; 550 –
1420◦C) that is appropriate for Yasur. The most primitive melt inclusion composition (with
volatile data) recorded at Yasur (Tan2326; Métrich et al. 2011) and magma temperatures of
1075◦C and 1050◦C were used to calculate the isobars, following Métrich et al. (2011). The
model in Figure 6.6a suggests that melt inclusions from this study equilibrated at pressures
of 25 – 80 MPa. Minimum estimates of entrapment pressures recorded in melt inclusions from
Métrich et al. (2011) are higher: up to 155 MPa (∼6 km) – although pre-eruptive S contents and
CO2/SO2 gas emissions from Yasur suggest volatile saturation at 290 MPa (∼11 km) (Métrich
et al., 2011).

106



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Figure 6.6: Melt inclusion (MI) and matrix glass concentrations of a) CO2 and H2O and b)
S and Cl concentrations. Isobars in a) are plotted for the melt composition of Tan2326 and
temperature of 1075◦C and 1050◦C, consistent with best-fitting decompression crystallisation
models from RhyoliteMELTS , with oxygen fugacity buffered at ∆QFM. The error on the CO2

and H2O measurements represents the propagated error (in quadrature) described in Chapter 2
and are smaller than symbols. The precision on the S and Cl were calculated from repeated
analysis on a secondary standard (Chapter 2; Table 2.4).

107



6.4. Discussion

Maximum S and Cl concentrations recorded by primitive melt inclusions at Yasur are ∼1000
ppm and 1280 ppm, respectively (Métrich et al., 2011) (Figure 6.7c-d). Melt inclusions analysed
in this study record lower S and Cl concentrations that range from 814 ppm to 65 ppm (± 5.8%)
and 1035 ppm to 355 ppm (± 10.9%), respectively (Figure 6.7c-d; Table 2.4). Concentrations
of S and Cl in matrix glasses range from 70 to 320 ppm and 272 to 498 ppm respectively (Figure
6.7c-d). Sulfur concentrations in melt inclusions drop sharply from ∼600 ppm at < 3 wt% MgO
(Figure 6.7c). Chlorine concentrations of melt inclusions increase from ∼400 – 600 ppm at 4 –
5 wt% MgO to >1200 ppm at ∼3 wt% MgO, and then decrease at lower melt MgO contents,
with most melt inclusions with less than 2 wt% MgO containing 400 – 800 ppm Cl (Figure
6.7d).

6.3.4 Melt inclusion geochemistry: Cu and Ag

Published whole rock data for Yasur (Deng et al. 2022 and references therein, Dupuy et al.
1982; Figure 6.7a) suggest Cu concentrations of ∼145 ppm for whole rocks with 7.8 wt%
MgO and Cu concentrations of just below 300 ppm for whole rocks with 2 – 3 wt% MgO
(Figure 6.7a). The Cu contents recorded by Yasur melt inclusions from this study extend to
higher concentrations, ranging from 247 to 548 ppm (Figure 6.7a). The concentrations of Cu
preserved in melt inclusions hosted by different phenocrysts are similar (Figure 6.7a). These
are the highest Cu concentrations recorded along the entire Vanuatu Arc and among the highest
globally in arc melt inclusions at such low melt MgO concentrations (Figure 6.1a; Hogg et al.,
in review). Up to now, there are no published reports of Ag concentrations for whole rocks or
melt inclusions from the Vanuatu Arc (Figure D.6). PEC-corrected melt inclusions from this
study show Ag concentrations of 93.4 to 124.9 ppb (Figure D.6). Cu/Ag ratios measured in melt
inclusions from this study range from 3981 to 4535 (Table 6.1), coinciding with the uppermost
limit of the MORB array (Figure 6.7b).

6.4 Discussion

I have shown that global arc whole rocks show clear and discrete trends in Cu-FeO-MgO sys-
tematics for CA and TH rocks, whereas melt inclusion compositions are scattered across the
entire range or to higher Cu concentrations (Figure 6.1). Some arc whole rocks reach concen-
trations > 300 ppm (Figure 6.1b), which is more than can be generated during simple isobaric
crystallisation from a parental melt with 75 – 100 ppm Cu (Chapter 5; Chiaradia 2014; Hogg et
al., in review). Melt inclusions from Yasur record some of the highest Cu concentrations in the

108



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

Figure 6.7: Cu, S and Cl systematics in Yasur magmas from melt inclusions and whole rocks.
Color coding and symbology of natural data and models are as in Figure 6.5 and Figure 6.6.
Concentrations of a) Cu; c) S and d) Cl in melt inclusions from this study and published studies.
Global MORB data shown in b) in red, and the range expressed in grey. Lighter shading shows
±10%. Decompression models show compositional evolution of the melt after three-way par-
titioning between melt-fluid-sulfide. Purple and red dashed lines in c) show the SCSStot and
solid lines show total melt S concentration ([S6+ + S2−], ppm). Blue dashed lines plotted either
side of the SCSStot curve modelled at 1075◦C represent an uncertainty of ± 10% of the SCSStot

calculation. Error bars on Cu represent the combined error on LA-ICP-MS measurements from
this study (Chapter 2; Electronic Appendix A4, Table A4.10). Error bars on S and Cl are calcu-
lated from precision on the glass secondary standard (Chapter 2; Table 2.4). Errors on S from
published Yasur melt inclusions from Métrich et al. (2011) are reported as ±10% and overlap
with the uncertainty estimate plotted for the SCSStot.

109



6.4. Discussion

global dataset (at low MgO contents), reaching up to 600 ppm (Figure 6.7a). Here, I discuss the
mechanisms acting to generate these Cu-enriched melt signatures and whether arc magmas are
predisposed to capturing them.

6.4.1 Cu diffusion into melt inclusions through phenocryst hosts

A potential explanation for the increased scatter in global melt inclusion datasets over the whole
rock datasets (Figure 6.1) is post-entrapment modification of melt inclusions. Diffusive gain of
Cu by melt inclusions is a potential mechanism for enriching their Cu concentrations and has
been studied in plagioclase-, clinopyroxene- and orthopyroxene-hosted (Audétat et al., 2018)
and more recently in olivine-hosted melt inclusions (Zhang et al., 2023). It has been proposed
that the mechanisms for post-entrapment Cu loss or gain differ depending on phenocryst type
(Audétat et al., 2018). Post-entrapment crystallisation can trigger sulfide saturation in melt in-
clusions (Danyushevsky et al., 2000) resulting in the precipitation of sulfide globules in the melt
which act as a Cu sink (Audétat et al., 2018). This precipitation initiates a concentration gradient
between the external melt and melt inclusion by means of Cu diffusion through the phenocryst
host, and results in a net increase in the Cu concentration of the melt inclusion (Audétat et al.,
2018). Audétat et al. (2018) show that orthopyroxene-hosted melt inclusions are particularly
susceptible to recording PEC-induced sulfide saturation and Cu gain. For plagioclase-hosted
melt inclusions, Cu gain by diffusion may be far more significant and may compensate for
H diffusion out of melt inclusions as they degas during magma ascent and eruption (Audétat
et al., 2018). Studies have shown that plagioclase- and orthopyroxene-hosted melt inclusions
can contain an order of magnitude more Cu than melt inclusions in co-precipitated clinopyrox-
ene and olivine (Audétat et al., 2018; Zhang et al., 2023). At temperatures between 1100 and
1000◦C, Cu in plagioclase-hosted melt inclusions can fully re-equilibrate on timescales of hours
to weeks, whereas in clinopyroxene-, orthopyroxene-, and olivine-hosted melt inclusions, diffu-
sive equilibration timescales typically range from 101 to 102 years (Audétat et al., 2018; Zhang
et al., 2023). Diffusion rates drop as temperatures decrease; by 900◦C, the time required to
re-equilibrate olivine-, clinopyroxene- and orthopyroxene-hosted melt inclusions shifts to 103

— 107 years. Hence, Cu concentrations in olivine-, clinopyroxene-, orthopyroxene-hosted melt
inclusions are expected to be least affected by diffusion (Audétat et al., 2018).

Figure 6.8 compares the distribution of Cu concentrations among melt inclusions in olivine,
clinopyroxene, orthopyroxene and plagioclase host phenocrysts from the global dataset in Fig-
ure 6.1a. Arc melt inclusions report higher mean Cu concentrations than arc whole rocks (Ap-
pendix D.1). Olivine-hosted melt inclusions show higher concentrations of Cu in TH melt in-
clusions relative to CA melt inclusions, similar to the established trends preserved in the whole

110



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

rocks record (Figure 6.1), whereas orthopyroxene-hosted melt inclusions record significantly
lower concentrations of Cu in TH melt inclusions relative to CA melt inclusions. Differences
in mean Cu concentrations of clinopyroxene- and plagioclase-hosted melt inclusions are statis-
tically insignificant (Appendix D.1) and show no dependence on the FeO content of the melt
(here taken as melt category, CA or TH).

Plagioclase-hosted melt inclusions show a wider spread distribution, relative to other phe-
nocrysts, with several potential outliers identified at Cu > 300 ppm (Figure 6.8a) which may
be consistent with post-entrapment Cu gain (Audétat et al., 2018). However, even with the re-
moval of plagioclase-hosted melt inclusions from the dataset (and despite mean differences in
Cu concentrations in CA and TH melt inclusions now becoming statistically significant, Ap-
pendix D.1) post-entrapment Cu gain cannot fully explain the scatter (and high concentrations
> 300 ppm) in Cu contents among melt inclusions (Figure 6.8c). It is likely that a number of
post-entrapment modifications occur in melt inclusions that are not well quantified, and may
contribute to the trends observed in the global melt inclusion dataset. For example, PEC will
drive Cu concentrations up in the absence of sulfide saturation. Degassing-induced melt re-
duction has been shown to trigger the formation of sulfide globules in melt inclusions (Hartley
et al., 2017), which would act as a Cu sink.

Nonetheless, while PEC modifications may account for the scatter amongst melt inclusions
globally, olivine-hosted melt inclusions from Yasur experience only a limited degree of PEC
(<5%) - which I also extrapolate to clinopyroxene- and plagioclase-hosted melt inclusions in
this study - therefore cannot explain Cu concentrations between 500 – 600 ppm in this dataset.
The following sections explore the mechanisms capable of generating Cu-enriched (> 300 ppm)
melts observed in melt inclusion datasets from Yasur (Figure 6.1).

111



6.4. Discussion

Figure 6.8: Assessment of the potential impact of melt inclusion post-entrapment Cu diffu-
sion on global datasets. a) violin and boxplots of global melt inclusion data (see Figure 6.1)
categorized according to olivine, clinopyroxene, orthopyroxene and plagioclase hosts. Out-
liers are shown in red circles. Horizontal lines mark interquartile ranges. Plagioclase-hosted
melt inclusion data show are large distribution with outliers spread to Cu concentrations > 300
ppm. b) the same data as in a), now each host is split according to melts being tholeiitic (TH)
or calc-alkaline (CA). Olivine-hosted melt inclusions have higher Cu in TH melt inclusions
than in CA melt inclusions (akin to whole rocks trends in Figure 6.1) whereas this is not the
case for orthopyroxene-, clinopyroxene- and plagioclase-hosted melt inclusions. c) Removal of
phenocryst-hosted melt inclusion data that are more susceptible to post-entrapment Cu modifi-
cations (orthopyroxene and plagioclase; Audétat et al. 2018), does not resolve the scatter among
Cu with respect to TH and CA data.

112



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

6.4.2 Isobaric crystallisation and sulfide saturation cannot reproduce Ya-
sur melt inclusion data

Isobaric crystallisation models were run starting with a primary basalt thought to represent the
composition of the parental magma feeding Yasur (Tuk3, ∆QFM, Si = 500 ppm, Cli = 500
ppm, Cu = 145 ppm, CO2 = 0.1 wt%; Métrich et al. 2011), ranging from 0.5 to 1.0 wt% H2O
(Métrich et al., 2011), at pressure of 200 MPa, to reconstruct melt Cu contents. The results of
these models are presented in Figure 6.9 and clearly highlight that Yasur melt inclusion datasets
(major and volatile elements included) do not fit the model. Modelled melts start depleting in
FeOt at ∼6.2 wt% MgO (likely marking the magnetite-in reaction) and S and Cu thereafter (at
∼5.8 wt% MgO) in response to magnetite-induced sulfide saturation (Figure 6.9b-d). Copper
concentrations in Yasur melt inclusions are 2–6 times higher than those predicted from isobaric
crystallisation models under ‘Yasur-like’ conditions, suggesting that alternative conditions are
required to explain the data. It is possible for primary arc magmas to have initially high Cu
contents (> 200 ppm) (Figure 6.1a-c). However, whole rock compositions for Yasur enable us
to constrain primitive magma Cu concentrations to ∼145 ppm (Dupuy et al., 1982) as used in
this model.

One way to achieve higher melt Cu concentrations at melt low MgO contents of ∼3 wt%
is to suppress sulfide saturation in the melt, which would allow S and Cu concentrations to
continue to rise with crystallisation. As discussed in Chapter 5, sulfide saturation in arc magmas
is ubiquitous and difficult to avoid during fractional crystallisation. This difficulty is due to the
combined effect of melt S concentrations increasing and FeOt concentrations decreasing with
crystallisation (Jenner et al., 2015; Wieser et al., 2021), causing the SCSS2− to decrease as the
temperature drops (Li and Zhang, 2022; O’Neill, 2021; Jugo et al., 2010; Smythe et al., 2017;
Hartley et al., 2018). Melt S2− concentrations therefore exceed the SCSS2−, triggering sulfide
saturation and melt Cu depletion as Cu strongly partitions into the sulfide phase.

Tholeiitic melts typically saturate in magnetite at MgO contents of <6 wt%, which causes
melt FeOt, as well as the Fe3+/Fe2+ ratio, to decrease such that melts become progressively
reduced (Jenner et al., 2010). Melt oxidation state plays a vital role on the timing of magnetite
saturation in arc magmas, with higher oxidation states promoting earlier (higher temperature or
higher MgO) saturation of magnetite (Jenner et al., 2010; Zimmer et al., 2010). Fe depletion
in the melt caused by magnetite fractionation causes a drop in the SCSS2− and reduction of
the melt caused by magnetite fractionation causes some of the S6+ present in the melt to be
converted to S2−. Both of these factors promote sulfide saturation and therefore it has been pro-
posed that magnetite fractionation is a widespread and important trigger for sulfide saturation
in tholeiitic melts (Jenner et al., 2010).

113



6.4. Discussion

Figure 6.9: Fractional crystallisation models generated using RhyoliteMELTS (Ghiorso and
Gualda, 2015) plotted alongside Yasur melt inclusion and whole rock data from this study.
Model inputs are detailed in Table 3.4 (Chapter 3). All models start with 145 ppm Cu (Dupuy
et al., 1982) and fractionate at 200 MPa (Métrich et al., 2011). Three sets of models are shown,
each run at two different melt water concentrations.

Whole rock and melt inclusion datasets demonstrate the tholeiitic nature of Yasur magmas
(Figure 6.5b), with FeOt increasing from ∼10 wt% in most primitive basalts to 13 wt% at 5-
6 wt% MgO before abruptly decreasing at lower MgO contents (Figure 6.5b). Magnetite is
present as phenocrysts and inclusions in my samples and this is reflected in the drop in FeOt

from 10.0 and 7.3 wt% MgO in the data (Figure 6.5b). Métrich et al. (2011) report magnetite in
melt inclusions with ∼5 wt% MgO. No sulfides were found in melt inclusions, phenocryst hosts
or matrix glasses for the Yasur samples nor have they been reported in published data for Yasur
besides two occurrences within the vapour bubbles of olivine-hosted melt inclusions which may
have formed from post-entrapment (Métrich et al., 2011; Hartley et al., 2014). Hence, despite
the magnetite-saturated state of the system, magmas at Yasur appear to be sulfide-undersaturated
on eruption, raising questions about how this can occur.

114



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

High melt oxidation states may suppress sulfide saturation and allow high Cu concentra-
tions to develop in evolved melts. Another set of isobaric models were run under more oxidised
conditions (∆QFM +1.5) in an attempt to reproduce the higher melt Cu concentrations (Figure
6.9 in green). Models predict that sulfide saturation, marked by the inflection in Cu concentra-
tion (Figure 6.9d), occurs relatively late (<2 wt% MgO) and replicates the concentrations of
melt inclusions well, with Cu increasing incompatibly in sulfide-undersaturated conditions, up
to ∼650 ppm. However, these models were unable to reproduce the melt inclusion major ele-
ment and S data (Figure 6.9). By fixing higher redox states, models significantly overestimated
the concentration of S dissolved in the melt (Figure 6.9c). Although the redox state of melt in-
clusions studied here are unconstrained, Métrich et al. (2011) report that melt compositions are
best matched to models performing at ∼NNO (or ∆QFM+0 – 0.5). Studies of tholeiitic magmas
from Efate (Deng et al., 2022; Raos and Crawford, 2004) and Kamchatka (Iveson et al., 2022)
argue for high oxidation states to explain enriched Cu signatures recorded by whole rocks and
melt inclusions, respectively (Figure 6.1a-b) but until now have not been modelled. Although
fractionation of highly oxidised magmas do not explain the Cu signatures in Yasur melt inclu-
sions from this study (Figure 6.9d), they may plausibly explain the subset of whole rocks data
with Cu > 300 ppm (Figure 6.1b).

Simply lowering the initial concentration of S in the melt can also delay sulfide saturation
during crystallisation. In a final subset of isobaric models (Figure 6.9 in pink), Si was reduced
from 500 to 200 ppm. These models were able to delay sulfide saturation relative to the first
model (Figure 6.9c in red), however still failed to reproduce the high Cu and S concentrations
characterising Yasur melt inclusion data (Figure 6.9d).

In Chapter 5, sulfide resorption was introduced as a means of efficiently increasing the Cu
contents of arc magmas by fluxing of mafic hydrous basalts through sulfide-saturated horizons
in the crust. This process can be recognised through elevated Cu/Ag ratios in melts and ex-
solving fluids (Berlo et al., 2014; Heinrich and Connolly, 2022). Melt inclusions from Yasur
have Cu/Ag ratios of 3981 to 4535 (Figure 6.7b), which lie within 10% of the upper limit of the
MORB array (Jenner et al., 2010). Combined with the lack of evidence of sulfides in my sam-
ples, resorption of accumulated sulfides as a mechanism for generating the Cu-rich signatures
in Yasur magmas is discounted.

6.4.3 Decompression crystallisation and degassing generates Cu-rich melts

Ascent dynamics play a fundamental and perhaps underappreciated role in shaping the chal-
cophile element signatures of magmas. Previous models show that with decreasing pressure,

115



6.4. Discussion

the SCSS2− at a given magma composition and temperature increases, making the ability to
attain/or sustain sulfide saturation difficult (Mavrogenes and O’Neill, 1999). If crystallisation
occurs during magma ascent, then the SCSS2− is influenced by the competition between chang-
ing melt composition (specifically FeOt) serving to decreases the SCSS2− and decompression
which raises the SCSS2−. Decompression is also accompanied by extensive degassing due to
the inverse relationship of volatile solubility and pressure, with the greatest mass of water- and
sulfur-rich fluids being exsolved at the shallowest pressures (Chapter 4). As discussed in Chap-
ter 5, significant degassing can promote sulfide-undersaturated conditions (Nicholson et al.,
2024).

Isothermal decompressional degassing models were run from 200 MPa, starting with either
a Yasur parental basalt (Tuk3; Deng et al. 2022; Dupuy et al. 1982) or a basaltic trachyandesite
(Tan2318; Métrich et al. 2011) with 1.2 wt% H2O and 0.1 wt% CO2 (Figure 6.5 and 6.7). Mod-
els were run over a range of temperatures (1050, 1075◦C). These initial magma compositions
are highlighted in black on Figure 6.5 and 6.7. The basaltic trachyandesite (Tan2318) is es-
timated to have crystallised 40–50% from a primary melt, consistent with calculations using
primary melt Zr and K2O contents in Yasur whole rocks and with those predicted by Rhyo-
liteMELTS (Ghiorso and Gualda, 2015). Models initiated with a Tan2318 composition produce
melts that are Fe-rich and K2O- and Al2O3-depleted compared to models starting with Tuk3
(Figure 6.5). Natural data from Yasur sit well within these predicted paths (Figure 6.5 and 6.7).

6.4.4 Sulfur degassing maintains the melt at sulfide saturation during de-
compression

The least evolved melt inclusions (MgO > 3 wt%) from Métrich et al. (2011) lie almost di-
rectly on the SCSStot predicted by the Tuk3 model run at 1075◦C (Figure 6.7c). However,
Métrich et al. (2011) reports that no sulfides are present in these melt inclusions, consistent
with observations made during this study. In order to calculate the SCSStot, an assumption on
the composition of the sulfide during decompression must be made (Chapter 3.2), which intro-
duces an additional source of error to these models. I chose to fix the Fe/(Fe+Ni+Cu) ratio of
the melt to 0.6 (Ding and Dasgupta, 2018; Wieser et al., 2020). Decompression models and
natural data show that during ascent, Yasur melts undergo up to ∼30 – 40% crystallisation (Fig-
ure 6.5). The relationship between the SCSS2−, melt FeOt, redox state and temperature output
from the decompression models were shown in Figure 3.10. The influence of melt FeOt has
a greater impact than decreasing pressure on the SCSStot (Figure 3.10c) such that in the latter
stages of decompression where modelled melts become Fe-depleted, there is a coincident drop
in the SCSS2−. Melt inclusions approximately follow the SCSStot trend (Figure 6.7c), which,

116



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

along with the absence of sulfides present, suggests that small mass sulfides may be consis-
tently formed and resorbed due to degassing, consistent with observations from other volcanic
systems (Wieser et al., 2020; Nicholson et al., 2024).

Figure 6.10: Decompression of an oxidized hydrous basalt results in a calc-alkaline composi-
tional melt trajectory (b) and high melt copper concentration (a) due to intense decompression-
induced degassing of S (c). Colour coding and symbology remains the same as described in
aforementioned figures. Previous models tailored to ‘Yasur-like’ conditions show tholeiitic
melt trends in FeOt and high Cu concentrations facilitated by decompressional degassing and
crystallisation (red and purples lines). In green, initial input magma parameters (4 wt% H2O,
∆QFM +1.5) were amended to generate calc-alkaline melt trends, and are still able to geenrate
Cu-rich melts.

The isothermal decompression degassing models predict that melts are initially sulfide-
saturated but very quickly become sulfide-undersaturated (Figure 6.7c). Melt inclusion S con-
centrations drop sharply between 2–3 wt% MgO and correspond well with the range predicted
by decompression models for both Tuk3 and Tan2318, between 1050 – 1075◦C. This is paired
with a continual increase in modelled melt Cu concentrations from the onset of decompression

117



6.5. Conclusions

crystallisation at 200 MPa through to 0.5 MPa reaching >600 ppm (Tan2318) and >700 ppm
(Tuk3) (Figure 6.7a). The steep S drop predicted by models and observed in the natural data
reflects degassing upon ascent. Sulfur solubility decreases with falling pressure (Ding et al.,
2023; Keppler, 2010), such that the proportion and mass of S exsolving into the fluid increases
during decompression. Sulfides can be resorbed, which liberates Cu and S to be partitioned
between the melt and fluid reservoirs (Chapter 5; Berlo et al. 2014; Halter et al. 2005, Hogg
et al., in review). Intense degassing of S during magma ascent lowers melt S and keeps is at
or below SCSStot, thereby preventing sulfide saturation and explaining the steep decrease in S
in melt inclusions from this study. The consequence of decompression-induced degassing and
crystallisation is not exclusive to TH magmas as demonstrated in a second set of models; de-
compressing a basalt (Tuk3) with 4 wt% H2O at ∆QFM +1.5 to generate a calc-alkaline trend
(Figure 6.10b) can still reach Cu concentrations up to 450 ppm (Figure 6.10a).

6.5 Conclusions

Analysis of global volcanic arc whole rock data show discrete trends in Cu-Fe-MgO where
tholeiitic magmas are Cu-rich and calc-alkaline magmas are Cu-poor. However, this is not
the case for global arc melt inclusions which instead show overlapping Cu concentrations be-
tween tholeiitic and calc-alkaline magmas. Some whole rock and melt inclusion data reach Cu
concentrations up to 600 ppm, which is more than can be accounted for by simple isobaric crys-
tallisation starting with average Cu concentrations of 70 – 100 ppm (Chiaradia, 2014). Melt
inclusions hosted in olivine, clinopyroxene and plagioclase, from Yasur record Cu contents
between 400 – 600 ppm. Cu diffusivity in phenocrysts can artificially increase or decrease Cu
concentrations recorded by melt inclusions. Diffusive modifications of Cu in plagioclase-hosted
melt inclusions specifically, have been shown to cause Cu gain in melt inclusions on timescales
of hours to weeks (Audétat et al., 2018). However, removal of these data cannot fully reconcile
the scatter in the global melt inclusion dataset, and does not explain Cu signatures of up to
600 ppm in Yasur melt inclusions. Instead, PEC modifications which influence the solubility
of sulfur and thereby impact Cu concentrations in melt inclusions, likely contribute towards the
scatter in the melt inclusion dataset that is otherwise absent in the whole rock record.

Several potential mechanisms help to generate Cu-rich arc magmas and typically require
magmas to be sulfide undersaturated. I show that sulfide resorption (Chapter 5) and isobaric
fractionation of highly oxidised arc magmas (Deng et al., 2022; Iveson et al., 2022) can achieve
high Cu concentrations that may be invoked to explain global Cu-rich signatures (observed in
both melt inclusion and whole rock datasets); however these mechanisms do not explain Cu
systematics at Yasur. Rather, my models demonstrate that decompression-induced crystallisa-

118



Chapter 6. Decompression crystallisation forms copper-rich melt signatures in arc magmas

tion and degassing are key to facilitating Cu-rich melt generation at Yasur, achieved through
intense degassing of S into exsolving fluids upon ascent, which consequently mitigates sulfide
saturation in the ascending melt. These results reflect arc magmas’ predisposition to record Cu-
rich signatures, attributed to their hydrous nature which enables them to degas and crystallise
upon decompression.

119



7
Conclusions and future work

This chapter synthesises the main conclusions of this thesis. Potential avenues for future work
are also discussed in relation to outstanding questions.

7.1 Watery magmas for chalcophile element-rich magmatic
and volcanic volatile phases

The results of the work I presented in Chapter 4 enabled the composition and relative mass
fluxes of volatile elements in volcanic outgassing plumes to be contextualised in terms of the
fluids and magmas that they derive from. I analysed global volcanic gas concentration and
mass flux data which gave insights into the relative roles of magmatic Cl and water contents
for shaping the geochemistry of exsolved magmatic fluids (Hogg et al., 2023). Magma wa-
ter concentrations are a primary influence on the mass of the exsolving fluid reservoir, but the
absolute affinity of several chalcophile elements for the fluid phase is determined by magma
Cl concentrations. I constructed a simple trace element partitioning model to investigate the
extent to which magma water and Cl contents controlled the concentration and mass of chal-
cophile elements partitioning into exsolving magmatic fluids during fractional crystallisation
and decompression at (and from) a range of different pressures.

A key finding of this work was the identification of a decoupling in the conditions required
to optimise the concentrations and masses of chalcophile elements transported in exsolved mag-

120



Chapter 7. Conclusions and future work

matic fluids and volcanic gases. Models show that although Cl behaves as an important ligand
for several chalcophile elements, it is the water content of a magma that has the greatest in-
fluence on maximising the mass fluxes of these elements that are outgassed by volcanoes. My
models quantifiably validate the conclusions drawn from observations in mineralised and barren
systems (Audétat and Zhang, 2019; Audetat et al., 2008) with similar enrichments of Cu and
other chalcophile elements, which suggest that differences in the mass of delivered fluids are
a critical factor in determining the ore-forming potential of these systems (Park et al., 2021).
By delineating the conditions required to generate high masses of metals in volatile phases, this
work could significantly impact the search for volcanic systems capable of producing ‘brine
mines’ (Blundy et al., 2021). The research discussed in Chapter 4 could be benefited by con-
sidering: 1. more complex magma evolution pathways: I presented two end member models
of fractional crystallisation and degassing, but in natural systems magmas will experience dif-
ferential extents of mixing, assimilation and stalling (Annen et al., 2006). Better constraints on
chalcophile element systematics in these environments would result from accounting for these
processes. 2. Interrogate the impact of sulfur on chalcophile elements in the gas plume:
for simplicity I modelled sulfur-free systems, since highly volatile species that become enriched
in volcanic gas plumes are primarily influenced by degassing (Zelenski et al., 2021; Edmonds
et al., 2018). A logical progression would be to incorporate the effect of sulfur, so that models
predicting the fate of elements like Cu, Se, and Au in the gas plume give a better representation
of the natural system.

7.2 An underappreciated role of degassing in magmatic Cu
systematics

The research presented in Chapter 5 showed that the formation of sulfides are ubiquitous dur-
ing fractionation of arc magmas and diminish the Cu-carrying capacity of exsolving magmatic
fluids. I identified two key mechanisms that can mitigate this issue: fractionation of oxidised,
hydrous magmas by degassing high masses of fluids that offset sulfide saturation, and thereby
promoting high mass proportions of Cu to partition into the evolving fluid reservoir. Alter-
natively, resorption of crustal sulfides that are accumulated over time by infiltrating oxidised
sulfide-undersaturated water-rich melts that redistribute Cu and S into fluids. These results have
important implications for the way that we understand the cycling of chalcophile elements in
magmatic and volcanic systems. Previous models have focused primarily on the effect that sul-
fide saturation has on Cu systematics (Chiaradia, 2014; Lee et al., 2012) and fail to consider the
impact that degassing plays (Wieser et al., 2020). Here, I am able to model and evaluate the
impact of these processes simultaneously.

121



7.2. An underappreciated role of degassing in magmatic Cu systematics

Fundamentally, I provided evidence to show that changes to magma water concentrations
can strongly influence chalcophile element abundances and distributions across magmatic reser-
voirs due to their control on the mass of exsolving fluids, in support of previous research
(Rezeau and Jagoutz, 2020; Chiaradia and Caricchi, 2017; Chelle-Michou et al., 2017). I ar-
gue that high concentrations of chalcophile elements, primarily Cu, in magmas and fluids are
of subordinate importance for the ore-forming potential of magmatic fluids (Du and Audétat,
2020), in contention with perspectives from previously published studies (Loucks, 2021, 2014).
Water-rich magmas could serve as a reliable indicator for locating metal-rich fluid lenses. This
factor, which has been historically underappreciated, may prompt earth scientists and industry
professionals to reassess the fundamental tenets of porphyry exploration and consider innova-
tive methods for accessing these fluids (Blundy et al., 2021). Future endeavours should focus
on the following:

1. Can deep salty fluids contend with shallow high mass fluids? Loucks (2014), argue
that exsolution of deep (>500 MPa) hypersaline, Cu-rich fluids near the base of the continental
crust are prerequisites to porphyry copper ore formation, despite early sulfide saturation. At
higher pressures, DCl

fluid/melt and consequently DCu
fluid/melt increases (Tattitch et al., 2021). How-

ever, are these fluid-melt partition coefficients sufficiently high enough to ensure that most
magmatic Cu partitions into these early-forming fluids? If so, this would imply that subsequent
degassing during magma ascent to shallower crustal levels (Chiaradia and Caricchi, 2017) con-
tributes primarily to the mass of water in the fluid reservoir, rather than the mass of Cu. To
formally address this question, the models presented in Chapter 5 must be run for higher pres-
sures (500 – 800 MPa) synonymous to the base of the crust. A holistic geochemical study
of whole rock, melt inclusion, fluid inclusion and mineral data from a trans-crustal magmatic
system could be used to glean barometric and geochemical (major, volatile, and chalcophile
element) insights that would help determine; a) if samples are fluid and sulfide saturated; b)
the depths magmas became sulfide- and volatile-saturated; and c) how concentrations change in
samples from deep and shallow reservoirs.

2. Development of proxies to trace degassing and sulfide saturation in natural datasets.
The contrasting partitioning behaviours of chalcophile elements relative to Cu, with respect to
sulfide phases (e.g., SL and MSS) and exsolving magmatic fluids, presents a unique opportunity
to develop geochemical proxies. These proxies could effectively trace systems dominated by
degassing and/or sulfide saturation, offering direct insights into the distribution of Cu within the
crust (Chapter 5; Appendix C, Figure C.2, C.3). Theoretically, fractionation of highly volatile
Se (Jenner, 2017; Zelenski et al., 2021), from moderately volatile Cu during degassing would
cause Cu/Se ratios to increase, whereas the onset of sulfide saturation would cause Cu/Se ratios
to decline (Appendix C, Figure C.2, C.3). Increasing the amount of experimental and natural
data on lesser studied but equally critical chalcophile elements – such as Se – will harbour a

122



Chapter 7. Conclusions and future work

better understanding of their concentrations and partitioning behaviours among silicate melts,
sulfides, exsolved magmatic fluids and volcanic gases.

7.3 The origins of Cu-enriched magmas

In Chapter 6, I addressed the differences in Cu concentrations recorded by global melt inclusion
and whole rock datasets. The global arc whole rock record showed that calc-alkaline (Fe-poor)
magmas are dominated by Cu-poor signatures whilst tholeiitic (Fe-rich) magmas are typically
Cu-rich. However, the distinct demarcation between CA and TH magmas in Cu-FeO-MgO
concentrations is not observed in the melt inclusion datasets (Figure 6.1a-c); instead the con-
centrations of Cu in melt inclusions trapping TH and CA magmas appear to overlap.

Recent studies have shown that Cu gain in plagioclase-hosted melt inclusions arises due to
the loss of hydrogen from the melt during degassing at the surface, with Cu being exchanged for
outwardly diffusing H+ ions. However, I demonstrate that even removing data susceptible to
Cu diffusion over short timescales cannot account for the strongly divergent Cu concentrations
observed in the melt inclusions. Crystallisation in melt inclusions after entrapment can drive
inclusion S concentrations up to the point of sulfide saturation, thereby modifying inclusion Cu
concentrations in the melt (Hartley et al., 2018). Likewise, degassing can impact the redox state
and solubility of S in melt inclusions which will also impact the concentration of Cu in the melt
(Hartley et al., 2018; Moussallam et al., 2016). Post-entrapment modifications are common in
melt inclusions and therefore may contribute to the scatter defining the global melt inclusion
data.

The results presented in Chapter 6 showcase the ability of arc magmas, attributed to their
higher magmatic water contents, to generate Cu-enriched melts. Their hydrous nature facili-
tates intense degassing and crystallisation during decompression. This mechanism successfully
explains high Cu concentrations recorded in melt inclusions from Yasur, of this study. Sulfide
resorption (discussed in Chapter 5) and isobaric crystallisation under highly oxidised condi-
tions can generate Cu-enriched melts that may plausibly explain Cu concentrations of >300
ppm observed in global whole rock and melt inclusion datasets. However, existing geochemical
studies (Métrich et al., 2011) and the work presented in Chapter 6, enable this mechanism to be
discarded for the case of Yasur. Instead, I find that intense degassing of sulfur into exsolving
magmatic fluids during decompression offsets sulfide saturation in Yasur magmas, enabling Cu
concentrations to reach up to ∼600 ppm.

It remains unclear whether differences in Cu concentrations recorded by global melt inclu-

123



7.3. The origins of Cu-enriched magmas

sion and whole rock datasets reflects the propensity of melt inclusions to preserve late-stage
decompressional degassing. To definitively constrain the origins of these compositional differ-
ences, a detailed arc-scale study would be beneficial. For example, a comprehensive study of
an arc with abundant melt inclusion and whole rock data – such as the Cascades or the Andes
– could provide valuable insights. Comparing Cu concentrations from these datasets within
the same locations would help better constrain the impacts of PEC, diffusion, and magmatic
processes, such as degassing and sulfide saturation during various magmatic pathways. This
comparison would ultimately help to evaluate the causes of the differences observed in the
global dataset.

124



Bibliography

Adams, J. V., Spera, F. J., and Jackson, M. G. (2021). Trachytic Melt In-
clusions Hosted in Clinopyroxene Offer a Glimpse Into Samoan EM2-Endmember
Melts. Geochemistry, Geophysics, Geosystems, 22(3):e2020GC009212. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2020GC009212.

Agangi, A. and Reddy, S. M. (2016). Open-system behaviour of magmatic fluid phase and
transport of copper in arc magmas at Krakatau and Batur volcanoes, Indonesia. Journal of

Volcanology and Geothermal Research, 327:669–686.

Aiuppa, A., Dongarrà, G., Valenza, M., Federico, C., and Pecoraino, G. (2003). Degassing
of trace volatile metals during the 2001 eruption of etna. GEOPHYSICAL MONOGRAPH-

AMERICAN GEOPHYSICAL UNION, 139:41–54.

Allard, P., Aiuppa, A., Bani, P., Métrich, N., Bertagnini, A., Gauthier, P.-J., Shinohara, H.,
Sawyer, G., Parello, F., Bagnato, E., Pelletier, B., and Garaebiti, E. (2016). Prodigious
emission rates and magma degassing budget of major, trace and radioactive volatile species
from Ambrym basaltic volcano, Vanuatu island Arc. Journal of Volcanology and Geothermal

Research, 322:119–143.

Allard, P., Aiuppa, A., Loyer, H., Carrot, F., Gaudry, A., Pinte, G., Michel, A., and
Dongarrà, G. (2000). Acid gas and metal emission rates during long-lived basalt de-
gassing at Stromboli Volcano. Geophysical Research Letters, 27(8):1207–1210. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/1999GL008413.

Allard, P., Behncke, B., d’Amico, S., Neri, M., and Gambino, S. (2006). Mount etna 1993–
2005: Anatomy of an evolving eruptive cycle. Earth-Science Reviews, 78(1-2):85–114.

Allen, A. G., Oppenheimer, C., Ferm, M., Baxter, P. J., Horrocks, L. A., Galle, B., McGonigle,
A. J. S., and Duffell, H. J. (2002). Primary sulfate aerosol and associated emissions from
Masaya Volcano, Nicaragua: PRIMARY SULFATE AEROSOL FROM MASAYA VOL-
CANO. Journal of Geophysical Research: Atmospheres, 107(D23):ACH 5–1–ACH 5–8.

Alletti, M., Baker, D., Scaillet, B., Aiuppa, A., Moretti, R., and Ottolini, L. (2009). Chlorine

125



Bibliography

partitioning between a basaltic melt and H2O–CO2 fluids at Mount Etna. Chemical Geology,
263(1-4):37–50.

Anderson, K. R. and Poland, M. P. (2016). Bayesian estimation of magma supply, storage, and
eruption rates using a multiphysical volcano model: Kı̄lauea Volcano, 2000–2012. Earth and

Planetary Science Letters, 447:161–171.

Annen, C., Blundy, J. D., and Sparks, R. S. J. (2006). The Genesis of Intermediate and Silicic
Magmas in Deep Crustal Hot Zones. Journal of Petrology, 47(3):505–539.

Audétat, A. and Edmonds, M. (2020). Magmatic-hydrothermal fluids. Elements: An Interna-

tional Magazine of Mineralogy, Geochemistry, and Petrology, 16(6):401–406.

Audetat, A., Pettke, T., Heinrich, C. A., and Bodnar, R. J. (2008). Special Paper: The Com-
position of Magmatic-Hydrothermal Fluids in Barren and Mineralized Intrusions. Economic

Geology, 103(5):877–908.

Audétat, A. and Simon, A. C. (2012). Magmatic controls on porphyry copper genesis. Special

Publications of the Society of Economic Geologists.

Audétat, A., Zhang, L., and Ni, H. (2018). Copper and li diffusion in plagioclase, pyroxenes,
olivine and apatite, and consequences for the composition of melt inclusions. Geochimica et

Cosmochimica Acta, 243:99–115.

Audétat, A. (2019). The Metal Content of Magmatic-Hydrothermal Fluids and Its Relationship
to Mineralization Potential. Economic Geology, 114(6):1033–1056.

Audétat, A., Günther, D., and Heinrich, C. A. (1998). Formation of a Magmatic-
Hydrothermal Ore Deposit: Insights with LA-ICP-MS Analysis of Fluid Inclusions. Science,
279(5359):2091–2094.

Audétat, A. and Pettke, T. (2003). The magmatic-hydrothermal evolution of two barren gran-
ites: a melt and fluid inclusion study of the Rito del Medio and Cañada Pinabete plutons in
northern New Mexico (USA). Geochimica et Cosmochimica Acta, 67(1):97–121.

Audétat, A. and Zhang, D. (2019). Abundances of S, Ga, Ge, Cd, In, Tl and 32 other ma-
jor to trace elements in high-temperature (350–700 °C) magmatic-hydrothermal fluids. Ore

Geology Reviews, 109:630–642.

Bani, P., Harris, A. J. L., Shinohara, H., and Donnadieu, F. (2013). Magma dy-
namics feeding Yasur’s explosive activity observed using thermal infrared re-
mote sensing. Geophysical Research Letters, 40(15):3830–3835. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/grl.50722.

126



Bibliography

Barber, N. D., Edmonds, M., Jenner, F., Audétat, A., and Williams, H. (2021). Amphibole con-
trol on copper systematics in arcs: Insights from the analysis of global datasets. Geochimica

et Cosmochimica Acta, 307:192–211.

Berlo, K., van Hinsberg, V. J., Vigouroux, N., Gagnon, J. E., and Williams-Jones, A. E. (2014).
Sulfide breakdown controls metal signature in volcanic gas at Kawah Ijen volcano, Indonesia.
Chemical Geology, 371:115–127.

Blundy, J., Afanasyev, A., Tattitch, B., Sparks, S., Melnik, O., Utkin, I., and Rust, A. (2021).
The economic potential of metalliferous sub-volcanic brines. Royal Society Open Science,
8(6):202192.

Blundy, J., Mavrogenes, J., Tattitch, B., Sparks, S., and Gilmer, A. (2015). Generation of por-
phyry copper deposits by gas–brine reaction in volcanic arcs. Nature Geoscience, 8(3):235–
240. Publisher: Nature Publishing Group.

Botcharnikov, R., Behrens, H., Holtz, F., Koepke, J., and Sato, H. (2004). Sulfur and chlorine
solubility in Mt. Unzen rhyodacitic melt at 850 °C and 200 MPa. Chemical Geology, 213(1-
3):207–225.

Brenan, J. M. (2015). Se–Te fractionation by sulfide–silicate melt partitioning: Implications for
the composition of mantle-derived magmas and their melting residues. Earth and Planetary

Science Letters, 422:45–57.

Candela, P. A. (1997). A Review of Shallow, Ore-related Granites: Textures, Volatiles, and Ore
Metals. 38(12):15.

Carney, J. and Macfarlane, A. (1982). Geological evidence bearing on the miocene to recent
structural evolution of the new hebrides arc. Tectonophysics, 87(1-4):147–175.

Chambefort, I., Dilles, J. H., and Kent, A. J. (2008). Anhydrite-bearing andesite and dacite as a
source for sulfur in magmatic-hydrothermal mineral deposits. Geology, 36(9):719–722.

Chang, J., Li, J.-W., and Audétat, A. (2018). Formation and evolution of multistage magmatic-
hydrothermal fluids at the Yulong porphyry Cu-Mo deposit, eastern Tibet: Insights from
LA-ICP-MS analysis of fluid inclusions. Geochimica et Cosmochimica Acta, 232:181–205.

Chelle-Michou, C., Rottier, B., Caricchi, L., and Simpson, G. (2017). Tempo of magma de-
gassing and the genesis of porphyry copper deposits. Scientific Reports, 7(1):40566.

Chiaradia, M. (2014). Copper enrichment in arc magmas controlled by overriding plate thick-
ness. Nature Geoscience, 7(1):43–46.

Chiaradia, M. (2020). How Much Water in Basaltic Melts Parental to Porphyry Copper De-
posits? Frontiers in Earth Science, 8:138.

127



Bibliography

Chiaradia, M. (2022). Distinct magma evolution processes control the formation of porphyry
Cu–Au deposits in thin and thick arcs. Earth and Planetary Science Letters, 599:117864.

Chiaradia, M. and Caricchi, L. (2017). Stochastic modelling of deep magmatic controls on
porphyry copper deposit endowment. Scientific Reports, 7(1):44523. Number: 1 Publisher:
Nature Publishing Group.

Coppola, D., Laiolo, M., Massimetti, F., and Cigolini, C. (2019). Monitoring endogenous
growth of open-vent volcanoes by balancing thermal and so2 emissions data derived from
space. Scientific reports, 9(1):9394.

Cottrell, E., Birner, S. K., Brounce, M., Davis, F. A., Waters, L. E., and Kelley,
K. A. (2021). Oxygen Fugacity Across Tectonic Settings. In Magma Redox Geo-

chemistry, pages 33–61. American Geophysical Union (AGU). Section: 3 eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119473206.ch3.

Coulon, C. and Maury, R. (1981). Petrology of tholeiitic lavas from tanna island (new hebrides):
importance of cumulative processes in island arc magmatism. Bulletin Volcanologique,
44:661–680.

Cox, D., Watt, S. F., Jenner, F. E., Hastie, A. R., and Hammond, S. J. (2019). Chalcophile
element processing beneath a continental arc stratovolcano. Earth and Planetary Science

Letters, 522:1–11.

Cox, D., Watt, S. F., Jenner, F. E., Hastie, A. R., Hammond, S. J., and Kunz, B. E. (2020).
Elevated magma fluxes deliver high-Cu magmas to the upper crust. Geology, 48(10):957–
960.

Danyushevsky, L. V., Della-Pasqua, F. N., and Sokolov, S. (2000). Re-equilibration of melt in-
clusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petro-
logical implications. Contributions to Mineralogy and Petrology, 138(1):68–83.

Danyushevsky, L. V. and Plechov, P. (2011). Petrolog3: Integrated software for model-
ing crystallization processes. Geochemistry, Geophysics, Geosystems, 12(7). eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2011GC003516.

Debret, B., Koga, K. T., Nicollet, C., Andreani, M., and Schwartz, S. (2014).
F, Cl and S input via serpentinite in subduction zones: implications for the
nature of the fluid released at depth. Terra Nova, 26(2):96–101. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1111/ter.12074.

Deng, C., Jenner, F. E., Wan, B., and Li, J.-L. (2022). The Influence of
Ridge Subduction on the Geochemistry of Vanuatu Arc Magmas. Jour-

128



Bibliography

nal of Geophysical Research: Solid Earth, 127(1):e2021JB022833. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2021JB022833.

DeVitre, C. L., Allison, C. M., and Gazel, E. (2021). A high-precision CO2 densimeter
for Raman spectroscopy using a Fluid Density Calibration Apparatus. Chemical Geology,
584:120522.

Ding, S. and Dasgupta, R. (2018). Sulfur Inventory of Ocean Island Basalt Source Regions Con-
strained by Modeling the Fate of Sulfide during Decompression Melting of a Heterogeneous
Mantle. Journal of Petrology, 59(7):1281–1308.

Ding, S., Plank, T., Wallace, P. J., and Rasmussen, D. J. (2023). Sulfur x: A Model
of Sulfur Degassing During Magma Ascent. Geochemistry, Geophysics, Geosystems,
24(4):e2022GC010552.

Dolejš, D. and Zajacz, Z. (2018). Halogens in Silicic Magmas and Their Hydrothermal Sys-
tems. In Harlov, D. E. and Aranovich, L., editors, The Role of Halogens in Terrestrial and Ex-

traterrestrial Geochemical Processes: Surface, Crust, and Mantle, pages 431–543. Springer
International Publishing, Cham.

Du, J. and Audétat, A. (2020). Early sulfide saturation is not detrimental to porphyry cu-au
formation. Geology, 48(5):519–524.

Dupont, C. L., Butcher, A., Valas, R. E., Bourne, P. E., and Caetano-Anollés, G. (2010). History
of biological metal utilization inferred through phylogenomic analysis of protein structures.
Proceedings of the National Academy of Sciences, 107(23):10567–10572.

Dupuy, C., Dostal, J., Marcelot, G., Bougault, H., Joron, J., and Treuil, M. (1982). Geochem-
istry of basalts from central and southern New Hebrides arc: implication for their source rock
composition. Earth and Planetary Science Letters, 60(2):207–225.

Edmonds, M., Liu, E., and Cashman, K. (2022). Open-vent volcanoes fuelled by depth-
integrated magma degassing. Bulletin of Volcanology, 84(3):28.

Edmonds, M., Mather, T. A., and Liu, E. J. (2018). A distinct metal fingerprint in arc volcanic
emissions. Nature Geoscience, 11(10):790–794.

Eggins, S. (1989). The origin of primitive ocean island and island arc basalts. PhD thesis,
University of Tasmania.

Eggins, S. (1993). Origin and differentiation of picritic arc magmas, ambae (aoba), vanuatu.
Contributions to Mineralogy and Petrology, 114(1):79–100.

Ferlito, C. and Lanzafame, G. (2010). The role of supercritical fluids in the potassium enrich-
ment of magmas at Mount Etna volcano (Italy). Lithos, 119(3-4):642–650.

129



Bibliography

Fiege, A., Holtz, F., Behrens, H., Mandeville, C. W., Shimizu, N., Crede, L. S., and Göttlicher,
J. (2015). Experimental investigation of the S and S-isotope distribution between H2O–S±Cl
fluids and basaltic melts during decompression. Chemical Geology, 393-394:36–54.

Floor, G. H. and Román-Ross, G. (2012). Selenium in volcanic environments: A review. Ap-

plied Geochemistry, 27(3):517–531.

Fortin, M.-A., Riddle, J., Desjardins-Langlais, Y., and Baker, D. R. (2015). The effect of water
on the sulfur concentration at sulfide saturation (SCSS) in natural melts. Geochimica et

Cosmochimica Acta, 160:100–116.

Frost, B. R. (1991). Chapter 1.INTRODUCTION TO OXYGEN FUGACITY AND ITS
PETROLOGIC IMPORTANCE. In Chapter 1.INTRODUCTION TO OXYGEN FUGACITY

AND ITS PETROLOGIC IMPORTANCE, pages 1–10. De Gruyter.

Gaetani, G. A., O’Leary, J. A., Shimizu, N., Bucholz, C. E., and Newville, M. (2012). Rapid
reequilibration of h2o and oxygen fugacity in olivine-hosted melt inclusions. Geology,
40(10):915–918.

Gaetani, G. A. and Watson, E. B. (2002). Modeling the major-element evolution of olivine-
hosted melt inclusions. Chemical Geology, 183(1-4):25–41.

Gauthier, P.-J. and Le Cloarec, M.-F. (1998). Variability of alkali and heavy metal fluxes re-
leased by Mt. Etna volcano, Sicily, between 1991 and 1995. Journal of Volcanology and

Geothermal Research, 81(3-4):311–326.

Georgatou, A., Chiaradia, M., and Klaver, M. (2022). Deep to shallow sulfide saturation at
nisyros active volcano. Geochemistry, Geophysics, Geosystems, 23(2):e2021GC010161.

Georgatou, A., Chiaradia, M., Rezeau, H., and Wälle, M. (2018). Magmatic sulphides in Qua-
ternary Ecuadorian arc magmas. Lithos, 296-299:580–599.

Ghiorso, M. S. and Gualda, G. A. R. (2015). An H2O–CO2 mixed fluid saturation model
compatible with rhyolite-MELTS. Contributions to Mineralogy and Petrology, 169(6):53.

Gorton, M. P. (1974). The geochemistry and geochronology of the New Hebrides. The Aus-
tralian National University (Australia).

Grondahl, C. and Zajacz, Z. (2022). Sulfur and chlorine budgets control the ore fertility of arc
magmas. Nature Communications, 13(1):4218.

Grove, T. L., Elkins-Tanton, L. T., Parman, S. W., Chatterjee, N., Müntener, O., and Gaetani,
G. A. (2003). Fractional crystallization and mantle-melting controls on calc-alkaline differ-
entiation trends. Contributions to Mineralogy and Petrology, 145(5):515–533.

130



Bibliography

Gu, Y., Yang, H., Wu, Y., Tuo, M., Xu, M., Hu, G., and Zuo, T. (2024). Regulation Mechanism
for Designing Decarbonization Pathways in the Copper Industry Toward Carbon Neutrality.
Environmental Science & Technology, 58(3):1518–1530. Publisher: American Chemical
Society.

Halter, W. E., Heinrich, C. A., and Pettke, T. (2005). Magma evolution and the formation of
porphyry Cu–Au ore fluids: evidence from silicate and sulfide melt inclusions. Mineralium

Deposita, 39(8):845–863.

Hartley, M. E., Bali, E., Maclennan, J., Neave, D. A., and Halldórsson, S. A. (2018). Melt
inclusion constraints on petrogenesis of the 2014–2015 Holuhraun eruption, Iceland. Contri-

butions to Mineralogy and Petrology, 173(2):10.

Hartley, M. E., Maclennan, J., Edmonds, M., and Thordarson, T. (2014). Reconstructing the
deep CO2 degassing behaviour of large basaltic fissure eruptions. Earth and Planetary Sci-

ence Letters, 393:120–131.

Hartley, M. E., Shorttle, O., Maclennan, J., Moussallam, Y., and Edmonds, M. (2017). Olivine-
hosted melt inclusions as an archive of redox heterogeneity in magmatic systems. Earth and

Planetary Science Letters, 479:192–205.

Hedenquist, J. W. and Lowenstern, J. B. (1994). The role of magmas in the formation of
hydrothermal ore deposits. Nature, 370(6490):519–527.

Heinrich, C. A. and Connolly, J. A. (2022). Physical transport of magmatic sulfides promotes
copper enrichment in hydrothermal ore fluids. Geology, 50(10):1101–1105.

Hogg, O. R., Edmonds, M., and Blundy, J. (2023). Water-rich magmas optimise volcanic
chalcophile element outgassing fluxes. Earth and Planetary Science Letters, 611:118153.

Holwell, D. A., Fiorentini, M. L., Knott, T. R., McDonald, I., Blanks, D. E., Campbell Mc-
Cuaig, T., and Gorczyk, W. (2022). Mobilisation of deep crustal sulfide melts as a first order
control on upper lithospheric metallogeny. Nature Communications, 13(1):573. Number: 1
Publisher: Nature Publishing Group.

Hsu, Y.-J., Zajacz, Z., Ulmer, P., and Heinrich, C. A. (2019). Chlorine partitioning between
granitic melt and H2O-CO2-NaCl fluids in the Earth’s upper crust and implications for
magmatic-hydrothermal ore genesis. Geochimica et Cosmochimica Acta, 261:171–190.

Huber, C. and Parmigiani, A. (2018). A Physical Model for Three-Phase Compaction in Silicic
Magma Reservoirs. Journal of Geophysical Research: Solid Earth, 123(4):2685–2705.

Hutchinson, M. C., Brooker, R. A., Dilles, J. H., and Blundy, J. (2020). The stability and
composition of sulfate melts in arc magmas. Contributions to Mineralogy and Petrology,
175(10):92.

131



Bibliography

Iacovino, K., Matthews, S., Wieser, P. E., Moore, G., and Bégué, F. (2021). Vesical part i: An
open-source thermodynamic model engine for mixed volatile (h2o-co2) solubility in silicate
melts. Earth and Space Science, 8(11):e2020EA001584.

Ilyinskaya, E., Mason, E., Wieser, P. E., Holland, L., Liu, E. J., Mather, T. A., Edmonds, M.,
Whitty, R. C. W., Elias, T., Nadeau, P. A., Schneider, D., McQuaid, J. B., Allen, S. E., Harvey,
J., Oppenheimer, C., Kern, C., and Damby, D. (2021). Rapid metal pollutant deposition from
the volcanic plume of Kı̄lauea, Hawai’i. Communications Earth & Environment, 2(1):1–15.
Number: 1 Publisher: Nature Publishing Group.

Iveson, A. A., Humphreys, M. C., Jenner, F. E., Kunz, B. E., Savov, I. P., De Hoog, J. C.,
Churikova, T. G., Gordeychik, B. N., Hammond, S. J., Plechov, P. Y., et al. (2022). Tracing
volatiles, halogens, and chalcophile metals during melt evolution at the tolbachik mono-
genetic field, kamchatka. Journal of Petrology, 63(9):egac087.

Iveson, A. A., Webster, J. D., Rowe, M. C., and Neill, O. K. (2019). Fluid-melt trace-element
partitioning behaviour between evolved melts and aqueous fluids: Experimental constraints
on the magmatic-hydrothermal transport of metals. Chemical Geology, 516:18–41.

Jackson, M., Blundy, J., and Sparks, R. (2018). Chemical differentiation, cold storage and
remobilization of magma in the earth’s crust. Nature, 564(7736):405–409.

Jenner, F. E. (2017). Cumulate causes for the low contents of sulfide-loving elements in the con-
tinental crust. Nature Geoscience, 10(7):524–529. Number: 7 Publisher: Nature Publishing
Group.

Jenner, F. E., Arculus, R. J., Mavrogenes, J. A., Dyriw, N. J., Nebel, O., and
Hauri, E. H. (2012). Chalcophile element systematics in volcanic glasses from the
northwestern Lau Basin. Geochemistry, Geophysics, Geosystems, 13(6). eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2012GC004088.

Jenner, F. E., Hauri, E. H., Bullock, E. S., König, S., Arculus, R. J., Mavrogenes, J. A.,
Mikkelson, N., and Goddard, C. (2015). The competing effects of sulfide saturation
versus degassing on the behavior of the chalcophile elements during the differentiation
of hydrous melts. Geochemistry, Geophysics, Geosystems, 16(5):1490–1507. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/2014GC005670.

Jenner, F. E. and O’Neill, H. S. C. (2012). Analysis of 60 elements in 616 ocean
floor basaltic glasses. Geochemistry, Geophysics, Geosystems, 13(2). eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2011GC004009.

Jenner, F. E., O’Neill, H. S. C., Arculus, R. J., and Mavrogenes, J. A. (2010). The Magnetite
Crisis in the Evolution of Arc-related Magmas and the Initial Concentration of Au, Ag and
Cu. Journal of Petrology, 51(12):2445–2464.

132



Bibliography

Jugo, P. J., Wilke, M., and Botcharnikov, R. E. (2010). Sulfur K-edge XANES analysis of
natural and synthetic basaltic glasses: Implications for S speciation and S content as function
of oxygen fugacity. Geochimica et Cosmochimica Acta, 74(20):5926–5938.

Keith, M., Haase, K. M., Klemd, R., Schwarz-Schampera, U., and Franke, H. (2017). System-
atic variations in magmatic sulphide chemistry from mid-ocean ridges, back-arc basins and
island arcs. Chemical Geology, 451:67–77.

Keppler, H. (2010). The distribution of sulfur between haplogranitic melts and aqueous fluids.
Geochimica et Cosmochimica Acta, 74(2):645–660.

Kiseeva, E. S. and Wood, B. J. (2013). A simple model for chalcophile element partitioning
between sulphide and silicate liquids with geochemical applications. Earth and Planetary

Science Letters, 383:68–81.

Kiseeva, E. S. and Wood, B. J. (2015). The effects of composition and temperature on chal-
cophile and lithophile element partitioning into magmatic sulphides. Earth and Planetary

Science Letters, 424:280–294.

Klein, B. Z., Jagoutz, O., Schmidt, M. W., and Kueter, N. (2023). A
Global Assessment of the Controls on the Fractionation of Arc Magmas.
Geochemistry, Geophysics, Geosystems, 24(5):e2023GC010888. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1029/2023GC010888.

Kouzmanov, K. and Pokrovski, G. S. (2012). Hydrothermal controls on metal distribution in
porphyry cu (-mo-au) systems. Special Publications of the Society of Economic Geologists.

Kress, V. C. and Carmichael, I. S. (1991). The compressibility of silicate liquids containing fe
2 o 3 and the effect of composition, temperature, oxygen fugacity and pressure on their redox
states. Contributions to Mineralogy and Petrology, 108:82–92.

Lamadrid, H. M., Moore, L. R., Moncada, D., Rimstidt, J. D., Burruss, R. C., and Bodnar, R. J.
(2017). Reassessment of the Raman CO2 densimeter. Chemical Geology, 450:210–222.

Lecumberri-Sanchez, P. and Bodnar, R. J. (2018). Halogen Geochemistry of Ore Deposits: Con-
tributions Towards Understanding Sources and Processes. In Harlov, D. E. and Aranovich,
L., editors, The Role of Halogens in Terrestrial and Extraterrestrial Geochemical Processes:

Surface, Crust, and Mantle, pages 261–305. Springer International Publishing, Cham.

Lee, C.-T. A., Luffi, P., Chin, E. J., Bouchet, R., Dasgupta, R., Morton, D. M., Le Roux, V.,
Yin, Q.-z., and Jin, D. (2012). Copper Systematics in Arc Magmas and Implications for
Crust-Mantle Differentiation. Science, 336(6077):64–68.

Lee, C.-T. A. and Tang, M. (2020). How to make porphyry copper deposits. Earth and Planetary

Science Letters, 529:115868.

133



Bibliography

Li, H. and Zhang, L. (2022). A thermodynamic model for sulfur content at sulfide saturation
(SCSS) in hydrous silicate melts: With implications for arc magma genesis and sulfur recy-
cling. Geochimica et Cosmochimica Acta, 325:187–204.

Li, Y. and Audétat, A. (2013). Gold solubility and partitioning between sulfide liquid, mono-
sulfide solid solution and hydrous mantle melts: Implications for the formation of Au-rich
magmas and crust–mantle differentiation. Geochimica et Cosmochimica Acta, 118:247–262.

Li, Y. and Audétat, A. (2015). Effects of temperature, silicate melt composition, and oxygen
fugacity on the partitioning of V, Mn, Co, Ni, Cu, Zn, As, Mo, Ag, Sn, Sb, W, Au, Pb, and
Bi between sulfide phases and silicate melt. Geochimica et Cosmochimica Acta, 162:25–45.

Li, Y., Audétat, A., Liu, Z., and Wang, F. (2021). Chalcophile element partitioning between Cu-
rich sulfide phases and silicate melt and implications for the formation of Earth’s continental
crust. Geochimica et Cosmochimica Acta, 302:61–82.

Lipman, P. W., Banks, N. G., and Rhodes, J. M. (1985). Degassing-induced crystallization of
basaltic magma and effects on lava rheology. Nature, 317(6038):604–607.

Loucks, R. (2014). Distinctive composition of copper-ore-forming arcmagmas. Australian

Journal of Earth Sciences, 61(1):5–16.

Loucks, R. R. (2021). Deep entrapment of buoyant magmas by orogenic tectonic stress: Its
role in producing continental crust, adakites, and porphyry copper deposits. Earth-Science

Reviews, 220:103744.

Mandon, C. L., Christenson, B. W., Schipper, C. I., Seward, T. M., and Garaebiti, E. (2019).
Metal transport in volcanic plumes: A case study at White Island and Yasur volcanoes. Jour-

nal of Volcanology and Geothermal Research, 369:155–171.

Mandon, C. L., Seward, T. M., and Christenson, B. W. (2020). Volatile transport of metals
and the Cu budget of the active White Island magmatic-hydrothermal system, New Zealand.
Journal of Volcanology and Geothermal Research, 398:106905.

Mason, E., Wieser, P. E., Liu, E. J., Edmonds, M., Ilyinskaya, E., Whitty, R. C. W., Mather,
T. A., Elias, T., Nadeau, P. A., Wilkes, T. C., McGonigle, A. J. S., Pering, T. D., Mims,
F. M., Kern, C., Schneider, D. J., and Oppenheimer, C. (2021). Volatile metal emissions from
volcanic degassing and lava–seawater interactions at Kı̄lauea Volcano, Hawai’i. Communi-

cations Earth & Environment, 2(1):1–16. Number: 1 Publisher: Nature Publishing Group.

Mason, E. M., Edmonds, M., Hammond, S., Ilyinskaya, E., Jenner, F., Kunz, B., Nicholson,
E. J., and Velasquez, G. (2024). Chalcophile element degassing at an active continental arc
volcano. Geochimica et Cosmochimica Acta, 367:72–86.

134



Bibliography

Masotta, M., Keppler, H., and Chaudhari, A. (2016). Fluid-melt partitioning of sulfur in dif-
ferentiated arc magmas and the sulfur yield of explosive volcanic eruptions. Geochimica et

Cosmochimica Acta, 176:26–43.

Mather, T., Witt, M., Pyle, D., Quayle, B., Aiuppa, A., Bagnato, E., Martin, R., Sims, K., Ed-
monds, M., Sutton, A., and Ilyinskaya, E. (2012). Halogens and trace metal emissions from
the ongoing 2008 summit eruption of Kı̄lauea volcano, Hawai‘i. Geochimica et Cosmochim-

ica Acta, 83:292–323.

Matjuschkin, V., Blundy, J. D., and Brooker, R. A. (2016). The effect of pressure on sulphur
speciation in mid- to deep-crustal arc magmas and implications for the formation of porphyry
copper deposits. Contributions to Mineralogy and Petrology, 171(7):66.

Matzen, A. K., Baker, M. B., Beckett, J. R., and Stolper, E. M. (2011). Fe–mg partitioning be-
tween olivine and high-magnesian melts and the nature of hawaiian parental liquids. Journal

of Petrology, 52(7-8):1243–1263.

Mavrogenes, J. A. and O’Neill, H. S. C. (1999). The relative effects of pressure, temperature and
oxygen fugacity on the solubility of sulfide in mafic magmas. Geochimica et Cosmochimica

Acta, 63(7-8):1173–1180.

McNulty, B. A., Jowitt, S. M., and Belousov, I. (2022). THE IMPORTANCE OF GEOLOGY IN
ASSESSING BY- AND COPRODUCT METAL SUPPLY POTENTIAL; A CASE STUDY
OF ANTIMONY, BISMUTH, SELENIUM, AND TELLURIUM WITHIN THE COPPER
PRODUCTION STREAM. Economic Geology, 117(6):1367–1385.

Miyashiro, A. (1974). Volcanic rock series in island arcs and active continental margins. Amer.

Jour. Sci., 274.

Moore, L. R. and Bodnar, R. J. (2019). A pedagogical approach to estimating the CO 2 budget
of magmas. Journal of the Geological Society, 176(2):398–407.

Moore, L. R., Gazel, E., Tuohy, R., Lloyd, A. S., Esposito, R., Steele-MacInnis, M., Hauri,
E. H., Wallace, P. J., Plank, T., and Bodnar, R. J. (2015). Bubbles matter: An assessment of
the contribution of vapor bubbles to melt inclusion volatile budgets. American Mineralogist,
100(4):806–823.

Moussallam, Y., Edmonds, M., Scaillet, B., Peters, N., Gennaro, E., Sides, I., and Oppenheimer,
C. (2016). The impact of degassing on the oxidation state of basaltic magmas: a case study
of kı̄lauea volcano. Earth and Planetary Science Letters, 450:317–325.

Moussallam, Y., Médard, E., Georgeais, G., Rose-Koga, E. F., Koga, K. T., Pelletier, B., Bani,
P., Shreve, T. L., Grandin, R., Boichu, M., Tari, D., and Peters, N. (2021). How to turn off a

135



Bibliography

lava lake? A petrological investigation of the 2018 intra-caldera and submarine eruptions of
Ambrym volcano. Bulletin of Volcanology, 83(5):36.

Moussallam, Y., Rose-Koga, E. F., Koga, K. T., Médard, E., Bani, P., Devidal, J.-L., and Tari,
D. (2019). Fast ascent rate during the 2017–2018 Plinian eruption of Ambae (Aoba) volcano:
a petrological investigation. Contributions to Mineralogy and Petrology, 174(11):90.

Mungall, J. E., Brenan, J. M., Godel, B., Barnes, S. J., and Gaillard, F. (2015). Transport
of metals and sulphur in magmas by flotation of sulphide melt on vapour bubbles. Nature

Geoscience, 8(3):216–219. Number: 3 Publisher: Nature Publishing Group.

Muth, M. J. and Wallace, P. J. (2022). Sulfur recycling in subduction zones and the oxygen
fugacity of mafic arc magmas. Earth and Planetary Science Letters, 599:117836.

Métrich, N., Allard, P., Aiuppa, A., Bani, P., Bertagnini, A., Shinohara, H., Parello, F., Di Muro,
A., Garaebiti, E., Belhadj, O., and Massare, D. (2011). Magma and Volatile Supply to Post-
collapse Volcanism and Block Resurgence in Siwi Caldera (Tanna Island, Vanuatu Arc). Jour-

nal of Petrology, 52(6):1077–1105.

Métrich, N., Allard, P., Spilliaert, N., Andronico, D., and Burton, M. (2004). 2001 flank erup-
tion of the alkali- and volatile-rich primitive basalt responsible for Mount Etna’s evolution in
the last three decades. Earth and Planetary Science Letters, 228(1-2):1–17.

Métrich, N. and Rutherford, M. J. (1998). Low Pressure Crystallization Paths of H2O-Saturated
Basaltic-Hawaiitic Melts from Mt Etna: Implications for Open-System Degassing of Basaltic
Volcanoes. Geochimica et Cosmochimica Acta, 62(7):1195–1205.

Nadeau, O., Williams-Jones, A. E., and Stix, J. (2010). Sulphide magma as a source of metals in
arc-related magmatic hydrothermal ore fluids. Nature Geoscience, 3(7):501–505. Number:
7 Publisher: Nature Publishing Group.

Namur, O., Charlier, B., Toplis, M. J., and Vander Auwera, J. (2012). Prediction of plagioclase-
melt equilibria in anhydrous silicate melts at 1-atm. Contributions to Mineralogy and Petrol-

ogy, 163(1):133–150.

Nash, W. M., Smythe, D. J., and Wood, B. J. (2019). Compositional and temperature effects
on sulfur speciation and solubility in silicate melts. Earth and Planetary Science Letters,
507:187–198.

Nayak, P. K., Yang, L., Brehm, W., and Adelhelm, P. (2018). From
Lithium-Ion to Sodium-Ion Batteries: Advantages, Challenges, and Sur-
prises. Angewandte Chemie International Edition, 57(1):102–120. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/anie.201703772.

136



Bibliography

Neave, D. A., Hartley, M. E., Maclennan, J., Edmonds, M., and Thordarson, T. (2017). Volatile
and light lithophile elements in high-anorthite plagioclase-hosted melt inclusions from Ice-
land. Geochimica et Cosmochimica Acta, 205:100–118.

Newman, S. and Lowenstern, J. B. (2002). VolatileCalc: a silicate melt–H2O–CO2 solution
model written in Visual Basic for excel. Computers & Geosciences, 28(5):597–604.

Nicholson, E. J., Wieser, P. E., Hartley, M. E., Jenner, F. E., Kunz, B. E., Ilyinskaya, E., Thor-
darson, T., and Edmonds, M. (2024). Sulfide saturation and resorption modulates sulfur and
metal availability during the 2014–15 holuhraun eruption, iceland. Communications Earth &

Environment, 5(1):164.

Nielsen, R. L. (2011). The effects of re-homogenization on plagioclase hosted melt inclusions.
Geochemistry, Geophysics, Geosystems, 12(10).

Nriagu, J. O. (1989). A global assessment of natural sources of atmospheric trace metals.
Nature, 338(6210):47–49. Number: 6210 Publisher: Nature Publishing Group.

O’Neill, H. S. (2021). The Thermodynamic Controls on Sulfide Saturation in Sili-
cate Melts with Application to Ocean Floor Basalts. In Magma Redox Geochem-

istry, pages 177–213. American Geophysical Union (AGU). Section: 10 eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781119473206.ch10.

O’Neill, H. S. C. and Mavrogenes, J. A. (2022). The sulfate capacities of silicate melts.
Geochimica et Cosmochimica Acta, 334:368–382.

Park, J.-W., Campbell, I. H., Chiaradia, M., Hao, H., and Lee, C.-T. (2021). Crustal magmatic
controls on the formation of porphyry copper deposits. Nature Reviews Earth & Environ-

ment, 2(8):542–557. Bandiera abtest: a Cg type: Nature Research Journals Number: 8
Primary atype: Reviews Publisher: Nature Publishing Group Subject term: Economic ge-
ology;Geochemistry;Petrology Subject term id: economic-geology;geochemistry;petrology.

Paton, C., Hellstrom, J., Paul, B., Woodhead, J., and Hergt, J. (2011). Iolite: Freeware for
the visualisation and processing of mass spectrometric data. Journal of Analytical Atomic

Spectrometry, 26(12):2508–2518.

Patten, C., Barnes, S.-J., Mathez, E. A., and Jenner, F. E. (2013). Partition coefficients of
chalcophile elements between sulfide and silicate melts and the early crystallization history of
sulfide liquid: LA-ICP-MS analysis of MORB sulfide droplets. Chemical Geology, 358:170–
188.

Pokrovski, G. S., Borisova, A. Y., and Bychkov, A. Y. (2013). Speciation and Transport of
Metals and Metalloids in Geological Vapors. Reviews in Mineralogy and Geochemistry,
76(1):165–218.

137



Bibliography

Portnyagin, M., Almeev, R., Matveev, S., and Holtz, F. (2008). Experimental evidence for rapid
water exchange between melt inclusions in olivine and host magma. Earth and Planetary

Science Letters, 272(3-4):541–552.

Raos, A. M. and Crawford, A. J. (2004). Basalts from the Efate Island Group, central section
of the Vanuatu arc, SW Pacific: geochemistry and petrogenesis. Journal of Volcanology and

Geothermal Research, 134(1):35–56.

Reekie, C. D. J., Jenner, F. E., Smythe, D. J., Hauri, E. H., Bullock, E. S., and Williams, H. M.
(2019). Sulfide resorption during crustal ascent and degassing of oceanic plateau basalts.
Nature Communications, 10(1):82. Number: 1 Publisher: Nature Publishing Group.

Rezeau, H. and Jagoutz, O. (2020). The importance of H2O in arc magmas for the formation of
porphyry Cu deposits. Ore Geology Reviews, 126:103744.

Richards, J. P. (2011). HIGH Sr/Y ARC MAGMAS AND PORPHYRY Cu ± Mo ± Au DE-
POSITS: JUST ADD WATER. Economic Geology, 106(7):1075–1081.

Richards, J. P. (2015). The oxidation state, and sulfur and Cu contents of arc magmas: implica-
tions for metallogeny. Lithos, 233:27–45.

Richards, J. P. (2018). A Shake-Up in the Porphyry World? Economic Geology, 113(6):1225–
1233.

Richards, J. P. (2021). Porphyry copper deposit formation in arcs: What are the odds? Geo-

sphere, 18(1):130–155.

Robin, C., Eissen, J.-P., and Monzier, M. (1993). Giant tuff cone and 12-km-wide associ-
ated caldera at ambrym volcano (vanuatu, new hebrides arc). Journal of Volcanology and

Geothermal Research, 55(3-4):225–238.

Roedder, E. (1984). Occurrence and significance of magmatic inclusions and silicate liquid
immiscibility. Acta Geologica Polonica, 34(1-2):139–178. Number: 1-2.

Roeder, P. L. and Emslie, R. F. (1970). Olivine-liquid equilibrium. Contributions to Mineralogy

and Petrology, 29(4):275–289.

Rose-Koga, E. F., Bouvier, A. S., Gaetani, G. A., Wallace, P. J., Allison, C. M., Andrys, J. A.,
Angeles de la Torre, C. A., Barth, A., Bodnar, R. J., Bracco Gartner, A. J. J., Butters, D.,
Castillejo, A., Chilson-Parks, B., Choudhary, B. R., Cluzel, N., Cole, M., Cottrell, E., Daly,
A., Danyushevsky, L. V., DeVitre, C. L., Drignon, M. J., France, L., Gaborieau, M., Gar-
cia, M. O., Gatti, E., Genske, F. S., Hartley, M. E., Hughes, E. C., Iveson, A. A., Johnson,
E. R., Jones, M., Kagoshima, T., Katzir, Y., Kawaguchi, M., Kawamoto, T., Kelley, K. A.,
Koornneef, J. M., Kurz, M. D., Laubier, M., Layne, G. D., Lerner, A., Lin, K. Y., Liu,

138



Bibliography

P. P., Lorenzo-Merino, A., Luciani, N., Magalhães, N., Marschall, H. R., Michael, P. J., Mon-
teleone, B. D., Moore, L. R., Moussallam, Y., Muth, M., Myers, M. L., Narváez, D. F., Navon,
O., Newcombe, M. E., Nichols, A. R. L., Nielsen, R. L., Pamukcu, A., Plank, T., Rasmussen,
D. J., Roberge, J., Schiavi, F., Schwartz, D., Shimizu, K., Shimizu, K., Shimizu, N., Thomas,
J. B., Thompson, G. T., Tucker, J. M., Ustunisik, G., Waelkens, C., Zhang, Y., and Zhou, T.
(2021). Silicate melt inclusions in the new millennium: A review of recommended practices
for preparation, analysis, and data presentation. Chemical Geology, 570:120145.

Scaillet, B., Clemente, B., Evans, B. W., and Pichavant, M. (1998). Redox control of sulfur de-
gassing in silicic magmas. Journal of Geophysical Research: Solid Earth, 103(B10):23937–
23949. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1029/98JB02301.

Seck, G. S., Hache, E., Bonnet, C., Simoën, M., and Carcanague, S. (2020). Copper at the
crossroads: Assessment of the interactions between low-carbon energy transition and supply
limitations. Resources, Conservation and Recycling, 163:105072.

Seo, J. H., Guillong, M., and Heinrich, C. A. (2009). The role of sulfur in the formation of
magmatic–hydrothermal copper–gold deposits. Earth and Planetary Science Letters, 282(1-
4):323–328.

Sheehan, F. and Barclay, J. (2016). Staged storage and magma convection at ambrym volcano,
vanuatu. Journal of Volcanology and Geothermal Research, 322:144–157.

Shinohara, H. (2008). Excess degassing from volcanoes and its role on eruptive and intrusive
activity. Reviews of Geophysics, 46(4):RG4005.

Shinohara, H. (2009). A missing link between volcanic degassing and experimental studies on
chloride partitioning. Chemical Geology, 263(1-4):51–59.

Shinohara, H. (2013). Volatile flux from subduction zone volcanoes: Insights from a detailed
evaluation of the fluxes from volcanoes in Japan. Journal of Volcanology and Geothermal

Research, 268:46–63.

Shinohara, H. and Toshimichiiiyama, J. (1989). Partition of chlorine compounds between sili-
cate melt and hydrothermal solutions: I. Partition of NaCl-KCl. Geochimica et Cosmochim-

ica Acta, 53(10):2617–2630.

Signorelli, S. and Carroll, M. (2002). Experimental study of cl solubility in hydrous alkaline
melts: constraints on the theoretical maximum amount of cl in trachytic and phonolitic melts.
Contributions to Mineralogy and Petrology, 143(2):209–218.

Sillitoe, R. H. (2010). Porphyry Copper Systems*. Economic Geology, 105(1):3–41.

139



Bibliography

Sisson, T. W. and Grove, T. L. (1993). Experimental investigations of the role of H2O in calc-
alkaline differentiation and subduction zone magmatism. Contributions to Mineralogy and

Petrology, 113(2):143–166.

Smith, P. M. and Asimow, P. D. (2005). Adiabat 1ph: A new public front-end to the melts,
pmelts, and phmelts models. Geochemistry, Geophysics, Geosystems, 6(2).

Smythe, D. J., Wood, B. J., and Kiseeva, E. S. (2017). The S content of silicate melts at sulfide
saturation: New experiments and a model incorporating the effects of sulfide composition.
American Mineralogist, 102(4):795–803. Publisher: De Gruyter.

Sorbadere, F., Schiano, P., Métrich, N., and Bertagnini, A. (2013). Small-scale coexistence
of island-arc-and enriched-morb-type basalts in the central vanuatu arc. Contributions to

Mineralogy and Petrology, 166:1305–1321.

Sorbadere, F., Schiano, P., Métrich, N., and Garaebiti, E. (2011). Insights into the origin of
primitive silica-undersaturated arc magmas of aoba volcano (vanuatu arc). Contributions to

Mineralogy and Petrology, 162:995–1009.

Steele-MacInnis, M., Bodnar, R. J., and Naden, J. (2011). Numerical model to determine the
composition of H2O–NaCl–CaCl2 fluid inclusions based on microthermometric and micro-
analytical data. Geochimica et Cosmochimica Acta, 75(1):21–40.

Symonds, R. B., Rose, W. I., Reed, M. H., Lichte, F. E., and Finnegan, D. L. (1987). Volatiliza-
tion, transport and sublimation of metallic and non-metallic elements in high temperature
gases at Merapi Volcano, Indonesia. Geochimica et Cosmochimica Acta, 51(8):2083–2101.

Taracsák, Z., Hartley, M. E., Burgess, R., Edmonds, M., Iddon, F., and Longpré, M.-A. (2019).
High fluxes of deep volatiles from ocean island volcanoes: Insights from El Hierro, Canary
Islands. Geochimica et Cosmochimica Acta, 258:19–36.

Tattitch, B. and Blundy, J. (2017a). Partitioning of Cu and Mo between felsic melts and saline
magmatic fluids: influence of salinity, ƒO2 and ƒS2. Applied Earth Science, 126(2):99–99.
Publisher: Taylor & Francis eprint: https://doi.org/10.1080/03717453.2017.1306300.

Tattitch, B., Chelle-Michou, C., Blundy, J., and Loucks, R. R. (2021). Chemical feedbacks
during magma degassing control chlorine partitioning and metal extraction in volcanic arcs.
Nature Communications, 12(1):1774.

Tattitch, B. C. and Blundy, J. D. (2017b). Cu-Mo partitioning between felsic melts and
saline-aqueous fluids as a function of XNaCleq, fO2, and fS2. American Mineralogist,
102(10):1987–2006. Publisher: De Gruyter.

Thomas, R. W. and Wood, B. J. (2021). The chemical behaviour of chlorine in silicate melts.
Geochimica et Cosmochimica Acta, 294:28–42.

140



Bibliography

Toplis, M. (2005). The thermodynamics of iron and magnesium partitioning between olivine
and liquid: criteria for assessing and predicting equilibrium in natural and experimental sys-
tems. Contributions to Mineralogy and Petrology, 149:22–39.

Tucker, J. M., Hauri, E. H., Pietruszka, A. J., Garcia, M. O., Marske, J. P., and Trusdell, F. A.
(2019). A high carbon content of the Hawaiian mantle from olivine-hosted melt inclusions.
Geochimica et Cosmochimica Acta, 254:156–172.

Turner, S. J. and Langmuir, C. H. (2015). The global chemical systematics of arc front stra-
tovolcanoes: Evaluating the role of crustal processes. Earth and Planetary Science Letters,
422:182–193.

Venugopal, S., Moune, S., Williams-Jones, G., Druitt, T., Vigouroux, N., Wilson, A., and Rus-
sell, J. K. (2020). Two distinct mantle sources beneath the Garibaldi Volcanic Belt: Insight
from olivine-hosted melt inclusions. Chemical Geology, 532:119346.

Wallace, P. J. (2005). Volatiles in subduction zone magmas: concentrations and fluxes based
on melt inclusion and volcanic gas data. Journal of Volcanology and Geothermal Research,
140(1-3):217–240.

Wallace, P. J. and Edmonds, M. (2011). The Sulfur Budget in Magmas: Evidence from Melt
Inclusions, Submarine Glasses, and Volcanic Gas Emissions. Reviews in Mineralogy and

Geochemistry, 73(1):215–246.

Webster, J., Kinzler, R., and Mathez, E. (1999). Chloride and water solubility in basalt and
andesite melts and implications for magmatic degassing. Geochimica et Cosmochimica Acta,
63(5):729–738.

Webster, J. D., Baker, D. R., and Aiuppa, A. (2018). Halogens in Mafic and Intermediate-Silica
Content Magmas. In Harlov, D. E. and Aranovich, L., editors, The Role of Halogens in

Terrestrial and Extraterrestrial Geochemical Processes: Surface, Crust, and Mantle, pages
307–430. Springer International Publishing, Cham.

Webster, J. D., Tappen, C. M., and Mandeville, C. W. (2009). Partitioning behavior of chlorine
and fluorine in the system apatite–melt–fluid. II: Felsic silicate systems at 200MPa. Geochim-

ica et Cosmochimica Acta, 73(3):559–581.

Webster, J. D., Vetere, F., Botcharnikov, R. E., Goldoff, B., McBirney, A., and Doherty, A. L.
(2015). Experimental and modeled chlorine solubilities in aluminosilicate melts at 1 to 7000
bars and 700 to 1250 C: Applications to magmas of Augustine Volcano, Alaska. American

Mineralogist, 100(2-3):522–535.

Wieser, G. a. (2023). ‘gleesonm1/PetThermoTools: PetThermoTools’. Zenodo. doi:
10.5281/zenodo.10258058.

141



Bibliography

Wieser, P. and Gleeson, M. (2023). PySulfSat: An Open-Source Python3 Tool for modelling
sulfide and sulfate saturation. preprint, Earth Sciences.

Wieser, P., Petrelli, M., Lubbers, J., Wieser, E., Özaydin, S., Kent, A. J., and Till, C. (2022).
Thermobar: an open-source python3 tool for thermobarometry and hygrometry.

Wieser, P. E., Jenner, F., Edmonds, M., Maclennan, J., and Kunz, B. E. (2020). Chalcophile
elements track the fate of sulfur at Kı̄lauea Volcano, Hawai’i. Geochimica et Cosmochimica

Acta, 282:245–275.

Wieser, P. E., Lamadrid, H., Maclennan, J., Edmonds, M., Matthews, S., Iacovino, K., Jenner,
F. E., Gansecki, C., Trusdell, F., Lee, R., and Ilyinskaya, E. (2021). Reconstructing Magma
Storage Depths for the 2018 kīlauean Eruption From Melt Inclusion CO 2 Contents: The
Importance of Vapor Bubbles. Geochemistry, Geophysics, Geosystems, 22.

Wilkinson, J. J. (2013). Triggers for the formation of porphyry ore deposits in magmatic arcs.
Nature Geoscience, 6(11):917–925. Number: 11 Publisher: Nature Publishing Group.

Williams-Jones, A. E. and Heinrich, C. A. (2005). 100th anniversary special paper: vapor
transport of metals and the formation of magmatic-hydrothermal ore deposits. Economic

Geology, 100(7):1287–1312.

Woitischek, J., Woods, A. W., Edmonds, M., Oppenheimer, C., Aiuppa, A., Pering, T. D.,
Ilanko, T., D’Aleo, R., and Garaebiti, E. (2020). Strombolian eruptions and dynamics of
magma degassing at Yasur Volcano (Vanuatu). Journal of Volcanology and Geothermal Re-

search, 398:106869.

Xia, W.-j., Wang, R., and Jenner, F. (2023). Sulfide resorption contributes to porphyry deposit
formation in collisional settings. Ore Geology Reviews, page 105804.

Zajacz, Z., Candela, P. A., and Piccoli, P. M. (2017). The partitioning of Cu, Au and Mo
between liquid and vapor at magmatic temperatures and its implications for the genesis of
magmatic-hydrothermal ore deposits. Geochimica et Cosmochimica Acta, 207:81–101.

Zajacz, Z., Candela, P. A., Piccoli, P. M., and Sanchez-Valle, C. (2012). The partitioning of sul-
fur and chlorine between andesite melts and magmatic volatiles and the exchange coefficients
of major cations. Geochimica et Cosmochimica Acta, 89:81–101.

Zajacz, Z., Candela, P. A., Piccoli, P. M., Sanchez-Valle, C., and Wälle, M. (2013). Solu-
bility and partitioning behavior of Au, Cu, Ag and reduced S in magmas. Geochimica et

Cosmochimica Acta, 112:288–304.

Zajacz, Z., Halter, W. E., Pettke, T., and Guillong, M. (2008). Determination of fluid/melt
partition coefficients by LA-ICPMS analysis of co-existing fluid and silicate melt inclusions:
Controls on element partitioning. Geochimica et Cosmochimica Acta, 72(8):2169–2197.

142



Bibliography

Zajacz, Z. and Tsay, A. (2019). An accurate model to predict sulfur concentration at anhydrite
saturation in silicate melts. Geochimica et Cosmochimica Acta, 261:288–304.

Zelenski, M., Malik, N., and Taran, Y. (2014). Emissions of trace elements during the
2012–2013 effusive eruption of Tolbachik volcano, Kamchatka: enrichment factors, parti-
tion coefficients and aerosol contribution. Journal of Volcanology and Geothermal Research,
285:136–149.

Zelenski, M., Simakin, A., Taran, Y., Kamenetsky, V., and Malik, N. (2021). Partitioning of
elements between high-temperature, low-density aqueous fluid and silicate melt as derived
from volcanic gas geochemistry. Geochimica et Cosmochimica Acta, 295:112–134.

Zhang, D. and Audétat, A. (2018). Magmatic-hydrothermal evolution of the barren huang-
shan pluton, anhui province, china: a melt and fluid inclusion study. Economic Geology,
113(4):803–824.

Zhang, L., Wang, F., and Ni, H. (2023). Anisotropic copper diffusion in olivine: Implications for
cu preservation in olivine and olivine-hosted melt inclusions. Geochimica et Cosmochimica

Acta, 347:116–124.

Zheng, Y.-F. (2019). Subduction zone geochemistry. Geoscience Frontiers, 10(4):1223–1254.

Zimmer, M. M., Plank, T., Hauri, E. H., Yogodzinski, G. M., Stelling, P., Larsen, J., Singer,
B., Jicha, B., Mandeville, C., and Nye, C. J. (2010). The Role of Water in Generating the
Calc-alkaline Trend: New Volatile Data for Aleutian Magmas and a New Tholeiitic Index.
Journal of Petrology, 51(12):2411–2444.

143



Appendix

Appendix A

Appendix A consists of supplementary data tables detailing the acquisition conditions used for
EPMA spot analyses (Tables A3.1.1 - A3.1.9) and the relative detection limits of these analyses
(Tables A3.2.1 - A3.2.3). Additionally, there is an electronic supplement that includes Raman
spectra (Appendix A1), SIMS (Appendix A2), EPMA (Appendix A3 which also includes elec-
tronic versions of Tables A3.1.1 to A3.2.3) and LA-ICP-MS (Appendix A4) data including
calculations of precision, accuracy and errors that are mentioned in Chapter 2 and 6.

Table A3.1.1: Acquisition conditions for EPMA spot analyses on plagioclase measured in
Session 1. All run with a 15 kV and 10 nA defocused beam with 5 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside
Al 20 TAP Corundum
Mg 90 LTAP SJ Olivine
Ca 20 PET Diopside
Ti 90 LPET Rutile

Mn 90 LIF Magnetite
Fe 60 LIF Fayalite
Cr 40 LIF Chromium
K 10 LPET Nickel
Na 10 LTAP Jadeite

144



Appendix

Table A3.1.2: Acquisition conditions for EPMA spot analyses on clinopyroxene measured in
Session 1. All run with a 15 kV and 20 nA defocused beam with 5 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 60 LTAP Corundum

Mg 20 TAP SJ Olivine

Ca 20 PET Diopside

Ti 60 LPET Rutile

Mn 40 LIF Magnetite

Fe 30 LIF Fayalite

Cr 40 LIF Chromium

K 10 LPET Nickel

Na 5 LTAP Jadeite

Table A3.1.3: Acquisition conditions for EPMA spot analyses on olivine measured in Session
1. All run with a 15 kV and 40 nA defocused beam with 1 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 150 LTAP Corundum

Mg 20 TAP SJ Olivine

Ca 150 PET Diopside

Ti 150 LPET Rutile

Mn 40 LIF Magnetite

Fe 10 LIF Fayalite

Cr 40 LIF Chromium

Ni 60 LIF Spinel

145



Table A3.1.4: Acquisition conditions for EPMA spot analyses on glass measured in Session 1.
All run with a 15 kV and 10 nA defocused beam with 10 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 30 TAP Corundum

Mg 30 LTAP SJ Olivine

Ca 30 PET Diopside

Ti 30 LPET Rutile

Mn 60 LIF Magnetite

Fe 30 LIF Fayalite

P 60 PET Apatite

K 10 LPET Nickel

Na 10 LTAP Jadeite

Table A3.1.5: Acquisition conditions for EPMA spot analyses on plagioclase measured in
Session 2. All run with a 15 kV and 10 nA defocused beam with 5 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 10 TAP Corundum

Mg 90 LTAP SJ Olivine

Ca 20 PET Diopside

Ti 90 LPET Rutile

Mn 90 LIF Magnetite

Fe 60 LIF Fayalite

Cr 40 LIF Chromium

K 10 LPET Nickel

Na 10 LTAP Jadeite

146



Appendix

Table A3.1.6: Acquisition conditions for EPMA spot analyses on clinopyroxene measured in
Session 2. All run with a 15 kV and 20 nA defocused beam with 5 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 60 LTAP Corundum

Mg 20 TAP SJ Olivine

Ca 20 PET Diopside

Ti 90 LPET Rutile

Mn 40 LIF Magnetite

Fe 30 LIF Fayalite

Cr 40 LIF Chromium

K 10 LPET Nickel

Na 20 LTAP Jadeite

Table A3.1.7: Acquisition conditions for EPMA spot analyses on olivine measured in Session
2. All run with a 15 kV and 40 nA defocused beam with 1 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 150 LTAP Corundum

Mg 20 TAP SJ Olivine

Ca 150 PET Diopside

Ti 150 LPET Rutile

Mn 40 LIF Magnetite

Fe 10 LIF Fayalite

Cr 40 LIF Chromium

Ni 60 LIF Spinel

147



Table A3.1.8: Acquisition conditions for EPMA spot analyses on glasses measured in Session
2. All run with a 15 kV and 20 nA defocused beam with 5 µm spot size.

Element Pk(s) Crystal Standard

Si 10 TAP Diopside

Al 60 LTAP Corundum

Mg 20 TAP SJ Olivine

Ca 20 PET Diopside

Ti 30 LPET Rutile

Mn 40 LIF Magnetite

Fe 30 LIF Fayalite

P 30 PET Apatite

K 10 LPET Nickel

Na 20 LTAP Jadeite

Cl 60 LPET Halite

S 60 PET Pyrite

Cr 40 LIF Corundum

Table A3.1.9: Acquisition conditions for EPMA spot analyses on oxides measured in Session
3. All run with a 15 kV and 20 nA defocused beam with 1 µm spot size.

Element Pk(s) Crystal Standard

Mg 30 TAP Periclase

Ti 30 PETL rutile

K 20 PETH KSpar

Ca 30 PETH Wollastonite

Fe 20 LIFL Fayalite

Cr 40 LIFL Cr metal

Mn 40 LIFL Mn metal

Na 10 TAPL Jadeite

Si 20 TAPL Fayalite

Al 40 TAPL Corundum

148



Appendix

Ta
bl

e
A

3.
2.

1:
D

et
ec

tio
n

lim
its

fo
rm

at
er

ia
ls

m
ea

su
re

d
in

Se
ss

io
n

1.

Se
ss

io
n

1
1

1
1

M
at

er
ia

l
pl

ag
pl

ag
ol

ol
cp

x
cp

x
gl

as
s

gl
as

s
m

ea
n

3σ
m

ea
n

3σ
m

ea
n

3σ
m

ea
n

3σ
N

a
64

6.
24

12
7.

71
24

2.
14

24
.2

9
31

5.
11

39
.1

7
K

38
7.

59
77

.2
5

25
8.

96
24

.5
6

39
9.

13
93

.9
2

Si
85

0.
12

81
.9

7
39

9.
95

25
.6

3
59

4.
82

36
.2

4
61

9.
55

39
.4

8
C

a
57

0.
94

60
.9

1
71

.9
8

1.
57

48
0.

18
30

.7
3

34
2.

95
68

.4
5

Ti
12

3.
53

5.
92

49
.0

0
0.

00
92

.2
9

2.
93

16
3.

51
8.

40
Fe

57
5.

00
66

.8
1

11
81

.9
8

84
.1

5
73

9.
43

72
.6

0
70

1.
37

10
0.

95
M

n
55

5.
65

96
.9

6
47

1.
22

41
.0

4
64

0.
57

10
6.

58
60

7.
27

90
.3

5
C

r
28

9.
46

25
.9

0
43

5.
82

63
.4

8
42

4.
83

54
.2

8
A

l
69

1.
24

69
.2

6
42

.7
8

1.
26

10
3.

75
5.

81
16

1.
02

30
.3

5
M

g
11

0.
53

5.
82

42
5.

27
18

.6
4

45
6.

00
34

.5
5

29
4.

52
66

.3
5

P
22

4.
15

44
.5

1
S

15
3.

33
18

.2
4

C
l

11
7.

89
9.

34
N

i
39

0.
10

18
.0

0

149



Ta
bl

e
A

3.
2.

2:
D

et
ec

tio
n

lim
its

fo
rm

at
er

ia
ls

m
ea

su
re

d
in

Se
ss

io
n

2.

Se
ss

io
n

2
2

2
2

M
at

er
ia

l
gl

as
s

ol
cp

x
pl

ag
m

ea
n

3σ
m

ea
n

3σ
m

ea
n

3σ
m

ea
n

3σ
N

a
64

0.
30

10
7.

29
48

6.
50

30
.8

0
67

0.
12

30
.2

7
K

56
4.

81
12

5.
88

25
6.

03
13

.2
5

39
6.

41
31

.5
6

Si
88

6.
21

87
.0

0
40

2.
37

8.
23

60
1.

76
12

.3
2

87
1.

54
28

.9
8

C
a

39
3.

47
63

.6
5

72
.2

5
1.

24
47

4.
80

12
.7

7
57

5.
04

25
.1

6
Ti

22
9.

45
17

.4
8

48
.4

8
0.

55
11

2.
36

1.
10

12
4.

62
2.

45
Fe

99
0.

99
19

6.
48

11
57

.1
6

35
.2

5
72

6.
07

22
.1

1
57

2.
40

27
.1

3
M

n
71

1.
77

13
1.

27
46

3.
44

14
.8

2
63

4.
19

30
.5

2
55

1.
98

27
.6

2
C

r
28

9.
48

8.
13

43
6.

38
17

.6
7

A
l

35
2.

67
41

.9
7

43
.5

8
2.

34
10

7.
23

4.
76

49
5.

44
15

.8
6

M
g

25
0.

00
12

7.
23

42
6.

54
15

.8
7

45
7.

58
14

.9
1

11
4.

28
5.

03
P

23
6.

26
44

.9
6

S
16

3.
27

24
.5

5
C

l
N

i
37

6.
71

5.
88

150



Appendix

Table A3.2.3: Detection limits for materials measured in Session 3.

Session 2
Material oxide

mean 3σ

Na 126.93 29.33
K 54.60 10.66
Si 183.97 42.13
Ca 60.98 12.81
Ti 51.06 8.37
Fe 96.99 18.79
Mn 59.31 10.22
Cr 64.25 9.94
Al 64.98 15.01
Mg 62.65 10.19
Ni 60.63 8.42

151



Appendix B

This section contains the supporting information for Chapter 4 that was originally published
alongside the manuscript “Water-rich magmas optimise volcanic chalcophile element outgassing
fluxes.”, Hogg. O. R., Edmonds. M., Blundy. J. (2023) in Earth and Planetary Science Letters.
DOI:https://doi.org/10.1016/j.epsl.2023.118153.

The modelling framework, global volcanic gas dataset, fluid inclusion, brine inclusion and
vapour inclusion datasets, and calculations of chlorine fluid-melt partition coefficients can be
found in the electronic version of Appendix B.

Figure B.1: P-T-X evolution of intermediate-low salinity magmatic hydrothermal fluids. The
critical curve separates regions dominated by vapour or brine. a) In deep intrusions, fluids
lie in the single phase field at 700°C (A) but during decompression and cooling undergo fluid
immiscibility and fluids lie in the two-phase field at 500°C forming a brine (30 wt% NaCl, B)
and low-density vapour (1 wt% NaCl, C). b) For the same conditions, a fluid with bulk fluid
salinity of 1 wt% (D) during decompression and cooling to 500°C remains in the single-phase
field (E) and does not intersect the two-phase field until <400°C, forming a brine (7 wt% NaCl,
F) and vapour (0.2 wt% NaCl, G).

152

DOI: https://doi.org/10.1016/j.epsl.2023.118153


Appendix

Figure B.2: The behaviour of Cl and trace metals during second boiling with initial melt con-
centrations of 20 ppm. Symbols designate the pressure at which magmas undergo crystallisa-
tion, and colours refer to different cases defined in Table 3.1. The model describes the compo-
sition of the melt and MVP during crystallisation where a melt fraction of 1 represents a basalt
and 0.1 represents a rhyolite (for details see Figure 4.3 Plots show concentrations of Cl, element
X, Y and Z in the melt (a) and MVP (b) against melt fraction. In (c) cumulative mass of each
element in the MVP (kg per kg of magma) is shown. By increasing initial melt metal concen-
trations both concentrations of X, Y and Z in the melt and fluid increase as well as generating
greater fluxes of metals relative to Figure 4.3 (10 ppm initial metal concentration). The greatest
mass yields of each element are still achieved in shallow fractionating, water-rich systems with
moderate Cl concentrations.

153



Figure B.3: The behaviour of chlorine and trace metals during second boiling with initial melt
concentrations of 100 ppm. Increasing initial melt metal concentrations to 100 ppm, increases
concentrations of X, Y and Z in the melt and fluid as well as their mass fluxes, by an order of
magnitude relative to Figure 4.3 (10 ppm initial metal concentration) and Figure B.2 (20 ppm
initial metal concentration). The sensitivity of each system to changing water and chlorine con-
tents and storage pressure remains the same with greatest mass yields of each element achieved
in shallow fractionating, water-rich systems with moderate Cl concentrations.

154



Appendix

Figure B.4: The behaviour of chlorine and trace metals during first boiling, with initial metal
concentrations of 20 ppm. Each data point refers to an MVP concentration or flux from decom-
pression of a 50% isobarically fractionated melt (F = 0.5, see Figure B.2). Pressures refer to the
depths at which these 50% isobarically fractionated melts were stored prior to decompression;
a) decompressed MVP concentrations of Cl, X, Y and Z with falling depth of storage prior to
decompression; ; b) concentrations of Cl, X, Y and Z in the decompressed melt; c) mass flux (kg
per kg of magma) of Cl, X, Y, Z. The effect that changing water and chlorine content as well as
depth of storage prior to decompression have on melt and fluid concentrations plus mass yields
in the MVP, remains the same as our original model (see Figure 4.4). Overall by increasing
the initial melt concentration from 10 ppm (Figure 4.4) to 20 ppm, fluxes of all elements are
increased.

155



Figure B.5: The behaviour of Cl and trace metals during first boiling, with initial metal con-
centrations of 100 ppm. The effect that changing water and chlorine content as well as depth
of storage prior to decompression have on melt and fluid concentrations plus mass yields in
the MVP, remains the same as our original model (see Figure 4.4) – maximum mass yields are
still satisfied by deeper fractionating magmas followed by rapid decompression to the surface.
Overall by increasing the initial melt concentration from 10 ppm (Figure 4.4) to 100 ppm, fluxes
of all elements are increased by an order of magnitude.

156



Appendix

Figure B.6: Modelled trace metal mass fluxes compared to natural data. Relative mass fluxes
in kg metal per kg of magma (a) and observed fluxes in kg per day (b) of volatiles (H2O,
Cl) and metals (Pb and Zn) listed along the x-axis, compared to fluxes predicted by isobaric
crystallisation and decompressional degassing models, where initial metal concentrations is
100 ppm.

157



Appendix C

Supporting information for Chapter 5 which has been submitted as part of the manuscript ”Sul-
fide resorption by water-rich melts yields copper-rich magmatic fluids” currently in review in
Nature Communications, with co-authors: M. Edmonds, P. Wieser, M. Gleeson, F. Jenner and
J. Blundy. The Python scripts used to construct the models presented in Chapter 5, are available
on my GitHub: https://github.com/oliviahogg/CuRichFluid

C.1 Global Arc Database Compilation

The database presented in Figure 5.1 (and Figure 6.1) compiles global arc whole rock (and
melt inclusion data in Chapter 6) data retrieved from GEOROC (https://georoc.eu/) and other
manually downloaded arc datasets. I adopted a similar screening criteria to published work
(Turner and Langmuir, 2015; Barber et al., 2021). I removed any data collected before 1960 and

if the data had no year information pertaining to the year that they were analysed. All records
that did not report an analytical method were removed; I include data acquired by XRF, TIMS,
EMPA, FTIR, SIMS, LA-ICP-MS. Methods such as WET, INAA and IGN were discarded as
these are generally older techniques (Barber et al., 2021). In some instances, individual samples
were analysed with multiple methods. To avoid duplication, these data were also removed so
that each sample name had one unique set of geochemical attributes.

C.2 Limitations for developing potential proxies for degassing and sulfide
saturation

Fingerprinting systems dominated by sulfides versus those where significant degassing also oc-
curs, requires fractionation of chalcophile elements from Cu. Selenium is more volatile than
Cu (Zajacz et al., 2008; Zelenski et al., 2021) but has a lower affinity for sulfides (Figure C.1).
Incorporating the fluid-melt and sulfide-melt partitioning data available for Se into the model
presented in Chapter 5, predicts that Cu/Se ratios decrease during sulfide saturation, and in-
crease only during degassing in the most water-rich systems (Figure C.2d, C.3d). It has been
suggested that Re has vastly different affinities for the sulfide liquid (SL) (being lower than Cu;
Figure C.1b) and mono-sulfide solid solution (MSS) (being greater than Cu; Figure C.1b) (Li
et al., 2021) hence Cu/Re trajectories may depend heavily on the form of sulfide (SL, Figure
C.2b, SL and MSS, Figure C.3b). Au exhibits a stronger affinity for SL relative to Cu (Figure
C.1b) (Li et al., 2021; Li and Audétat, 2015) causing Cu/Au to increase during sulfide liquid

158



Appendix

saturation (Figure C.2c). However, in more evolved melts (< FeO 3.6 wt%) where sulfides typ-
ically take the form of MSS (Figure C.1b), models predict Cu/Au ratios to drop (Figure C.3c).
Au analysis becomes increasingly difficult at low temperature (and MgO contents), incurring
high errors due to Ta interferences (Jenner et al., 2010). The analytical challenges (for Se, Re,
Au and other chalcophile elements) plus limited experimental work on the sulfide-silicate melt
partitioning behaviours of these elements during differentiation (Chapter 1), allow us only to
speculate the fidelity of these proxies for tracing sulfide saturation and degassing and require
more research.

Figure C.1: Relative (a) fluid-melt and (b) sulfide-melt partition coefficients for chalcophile
elements. Dfluid/melt for Cl-speciating Cu (Tattitch and Blundy, 2017b); Re (Zelenski et al.,
2021); Ag (Zajacz et al., 2008), increase with fluid salinity during magma differentiation to
lower MgO. Dfluid/melt for non Cl-speciating Se (Zelenski et al., 2021) and Au (Zajacz et al.,
2013) are high and assumed constant through magma differentiation. Bold lines in b) repre-
sent Dsulfide/melt where SL-silicate melt partition coefficients (Kiseeva and Wood, 2013; Li and
Audétat, 2015; Li et al., 2021; Patten et al., 2013) are modelled up to FeO 3.6 wt% and tran-
sition to MSS-silicate melt partition coefficients (Li and Audétat, 2015; Li et al., 2021) below
this threshold. Faint lines in b) show DSL/melt (for models assuming only SL is present).

159



Sulfide-silicate melt partitioning of Ag is better constrained (Li and Audétat, 2015; Kiseeva
and Wood, 2013) and with more natural data available (Cox et al., 2019; Jenner et al., 2015)
may be used to deconvolve degassing and sulfide saturation. Cu/Ag ratios are initially high at
high melt fractions in our models but drop towards lower MgO (Figure C.4). For reasonable
initial S concentrations (Muth and Wallace, 2022) our models reconcile the range in Cu/Ag
ratios reported for published natural data that suggest the drop in Cu/Ag tracks the conversion
of SL to MSS at low MgO (Cox et al., 2019; Jenner et al., 2015). Systems undergoing both
extensive degassing and sulfide saturation show an earlier drop in Cu/Ag (due to MSS), relative
to water-poor systems (Figure C.4), which may explain the range of Cu/Ag ratios observed
across Chilean stratovolcanoes (Cox et al., 2020). Moreover, evidence of globular sulfides in
arc basalts to dacites (Agangi and Reddy, 2016; Georgatou et al., 2018) further support that the
transition to MSS can vary, and is perhaps constrained by magma water contents.

Figure C.2: Model results for melt ratios of a) Cu/Ag b) Cu/Re, c) Cu/Au, d) Cu/Se. Conditions
modelled assume sulfide present only as SL. Cu/Ag of bulk continental crust plotted in (a). Note
that c) is plotted on logarithmic scale.

160



Appendix

Figure C.3: Results of models that were run under conditions assuming sulfide present as SL
and transitions to MSS. Cu/Ag of bulk continental crust plotted in (a). Note that c) is plotted on
logarithmic scale. Key differences from Figure C.2 manifest in a) Cu/Ag and c) Cu/Au ratios
where the relative affinity of Ag and Au compared to Cu change (Figure C.1b).

Figure C.4: Modelled melt Cu/Ag ratios for initial S concentrations of a) 1000 and b) 2000
ppm, for systems of differing water concentrations. Green line represents Cu/Ag of bulk con-
tinental crust. Arrows in a) indicate Cu/Ag as SL transitions to MSS. Arrow in b) oriented in
direction of increased degassing. Data from the Lau Basin, Tonga Arc (Jenner et al., 2012), and
various Chilean stratovolcanoes (Cox et al., 2020) are plotted for comparison.

161



Appendix D

This Appendix includes the supporting information for the work presented in Chapter 6. An
electronic version of Table 6.1 is available in Electronic Appendix D.

Figure D.1: EDS element map for the inspection of sulfide phases in Yasur scoria samples, the
example shown here is for thin section T1.

162



Appendix

Figure D.2: Inspection of PEC in plagioclase- and clinopyroxene-hosted melt inclusions from
Yasur, this study. At any given Mg# all melt inclusions align with the trends defined by existing
Yasur whole rocks and melt inclusions. PEC would increase MgO contents and reduce CaO in
clinopyroxene-hosted and Al2O3 in plagioclase-hosted melt inclusions.

163



Figure D.3: Selected major element (wt%) , Cl, S and Cu concentrations for melt inclusion
and whole rock data along the Vanuatu Arc. Global arc melt inclusion and WR data are in
grey. Blocks contain data from several volcanic islands, as described in Figure 2.1 Northern
and Southern blocks contain data from, Mota Lava, Vanua Lava, Mota, Yasur, Erromango and
Anatom. Yasur data from melt inclusions of this study are shown in turquoise stars. Central
block data are from Gaua, Ambrym and Ambae but Cl and Cu data are just for Ambrym and
Ambae.

164



Appendix

Figure D.4: Selected trace element ratios for melt inclusion and whole rock data along the
Vanuatu Arc. Data are the same as in Figure D.3.

165



Figure D.5: Selected trace element concentrations analysed in melt inclusions from Yasur in
this study.

166



Appendix

Figure D.6: Selected chalcophile element concentrations analysed in melt inclusions from Ya-
sur in this study, including the first Ag, Au, Cd, Bi data for the Vanuatu Arc.

167



D.1 Statistical analysis of melt inclusions and whole rock datasets.

To assist the global arc whole rock and melt inclusion datasets that are shown in Figure 6.1, a
series of statistical tests to compare Cu concentrations between whole rocks and melt inclusions
with tholeiitic (TH) and calc-alkaline (CA) compositions. All tests were run using an open
source Python module, statsmodels (v.0.13.2). I use a parametric statistical test to compare
differences in the mean Cu concentrations observed in two groups (CA and TH) of whole rocks
and melt inclusions (Figure D.7). I choose a Welch’s t-test due to the heterogeneity of the
variance in these datasets. Mean Cu concentrations from these data are shown in Table D.1.

Figure D.7: Comparing the distribution of CA and TH whole rocks and melt inclusions. Hor-
izontal lines mark interquartile ranges. Yellow circles mark the mean Cu concentration of the
group, as reported in Table D.1.

Mean Cu (ppm)
Tholeiitic (TH) Calc-alkaline (CA)

Whole Rock (WR) 84.0 43.2
Melt Inclusion (MI) 163.6 136.3

Table D.1: Mean Cu concentrations (ppm) for tholeiitic (TH) and calc-alkaline (CA) composi-
tions of whole rocks and melt inclusions.

I set a threshold p-value of 0.005. Welch’s t-test results for global arc whole rock data
are reported in Table D.2 and show that the mean Cu concentrations of CA whole rocks (43.2
ppm) differ significantly from TH whole rocks (84.0 ppm). The same test was performed on
global melt inclusion data (Table D.2) and show the p-value (0.41) to be above the threshold
set, indicating that differences in mean Cu concentrations in TH (163.6 ppm) and CA (136.3

168



Appendix

test t-value p-value DoF
WR (TH, CA) Welch’s t-test 20.9 2.9e-90 2660
MI (TH, CA) Welch’s t-test 0.82 0.41 356

Table D.2: Statistical test results for global arc whole rock (TH, CA) and melt inclusion (TH,
CA) datasets.

ppm) melt inclusions are statistically insignificant – in direct contrast to what is observed in the
whole rock record.

A Welch’s t-test is suitable for analysing variance between two groups, however when more
than two groups are involved, an ANOVA test must be used. To explore what might be causing
melt inclusion Cu concentrations to deviate from the Cu-FeO-MgO trend established in by arc
volcanic whole rocks, I use an ANOVA test to assess the differences in mean Cu concentrations
preserved in different phenocryst hosts. This distribution of these data are plotted in Figure
6.8a-b. Results of the ANOVA test are found in Table D.3, and show p-values that are below the
threshold value of 0.005, therefore differences in Cu concentrations among groups (of different
phenocrysts) are significant. To ensure that no one point is influencing the data, a Cooks distance
model was fitted; all data plot below 0.5 suggesting no point has a particular influence on the
ANOVA test result (Figure D.8)

df sum sq mean sq F p-value
C(Host) 3.0 1.47e+5 49073 7.48 0.000072
Residual 368 2.41e+6 6565 NaN NaN

Table D.3: Results of the ANOVA test comparing mean Cu concentrations in arc melt inclu-
sions (N=372) across different phenocryst hosts.

To further investigate if a particular group was causing the significant results of the ANOVA
test, a Tukey-HSD test was performed. Results output from python, can be found in Figure D.9
and show that plagioclase- and orthopyroxene-hosted melt inclusions differ more (significantly)
than other groups (olivine and clinopyroxene), with p-values (p-adj) < 0.005. This aligns with
our understanding of Cu diffusivity and PEC modifications affecting melt inclusions hosted in
these phenocrysts.

169



Figure D.8: Cooks Distance plot to assess the influence of individual data points on the ANOVA
test.

Figure D.9: Tukey-HSD results. P-adj column gives the p-values for each group comparison.
The null hypothesis for each pair is that no difference in the mean Cu contents between the two
groups (phenocrysts).

170


	Introduction
	Motivation
	The behaviour of chalcophile elements
	Introduction to controls on Cu systematics in magmatic systems
	Key research questions
	Thesis aims and structure

	Methods
	Geological setting
	Sample preparation
	Raman Spectroscopy
	Background
	Acquisition procedure
	Data quality

	Secondary Ion Mass Spectrometry (SIMS)
	Acquisition procedure
	Calibration and data quality

	Electron Probe Micro-Analyser (EPMA)
	Analytical procedure
	Data quality
	Oxide spot analyses

	Laser Ablation Inductively Coupled Mass Spectrometry (LA-ICP-MS)
	Acquisition parameters
	Data processing and quality

	Post-entrapment crystallisation corrections

	Modelling crystallisation, degassing and sulfide saturation in magmas
	Degassing of volatile chalcophile elements during fractional crystallisation and decompression
	Partitioning behaviours of chlorine and groups of chalcophile elements
	Modelling isobaric crystallisation and second boiling
	Modelling decompression and first boiling

	Degassing and sulfide saturation of chalcophile elements during fractional crystallisation
	Background
	Modelling fractional crystallisation and degassing of H2O and CO2
	Modelling degassing of sulfur, chlorine and chalcophile elements
	Modelling sulfide saturation
	Sulfide-silicate melt partitioning of chalcophile elements
	Modelling sulfide assimilation by hydrous magmas

	Degassing and sulfide saturation of chalcophile elements during decompression
	Background
	Modelling degassing and crystallisation during ascent
	Modelling degassing of sulfur, chlorine and chalcophile elements at Yasur
	Modelling sulfide saturation and sulfide-silicate melt partitioning at Yasur
	Modelling isobaric crystallisation at Yasur


	Water-rich magmas optimise volcanic chalcophile element outgassing fluxes
	Introduction
	Global ore-forming fluid and volcanic gas datasets
	Methods
	Results
	Models of degassing of trace metals during second boiling and crystallisation
	Models of decompression degassing of trace metals

	Discussion
	Comparison of models with data from global volcanoes

	Conclusions

	Sulfide resorption by water-rich melt yields copper-rich magmatic fluids
	Introduction
	Modeling chalcophile element behaviour in arc magmas
	Results
	Melt interaction with sulfide cumulates generates copper-rich fluids
	Conclusions

	Decompression crystallisation forms copper-rich melt signatures in arc magmas
	Introduction
	Sample preparation and methods
	Results
	Petrography and host geochemistry
	Melt inclusion geochemistry: major elements
	Melt inclusion geochemistry: volatile elements
	Melt inclusion geochemistry: Cu and Ag

	Discussion
	Cu diffusion into melt inclusions through phenocryst hosts
	Isobaric crystallisation and sulfide saturation cannot reproduce Yasur melt inclusion data
	Decompression crystallisation and degassing generates Cu-rich melts
	Sulfur degassing maintains the melt at sulfide saturation during decompression

	Conclusions

	Conclusions and future work
	Watery magmas for chalcophile element-rich magmatic and volcanic volatile phases
	An underappreciated role of degassing in magmatic Cu systematics
	The origins of Cu-enriched magmas