AN ABSTRACT OF THE DISSERTATION OF
Mélissa J. Drignon for the degree of Doctor of Philosophy in Geology presented on July 19, 2018 Title: An Experimental and Analytical Investigation of the Parameters that Influence Measured CO2 in Plagioclase-Hosted Melt Inclusions in MORBs
Abstract approved:
Roger L. Nielsen Frank J. Tepley III
To understand the processes that lead to the formation of the oceanic crust, one must know the composition and the depth at which primary melts originate. Towards this end, this dissertation focuses on plagioclase-hosted melt inclusions from plagioclase ultraphyric basalts (PUBs). Plagioclase is usually considered to be the second phase, after olivine, in the crystallization sequence of mid-ocean ridge basalts. However, the relatively low density of plagioclase makes it possible for melts to carry them from great depths. Volatiles were analyzed in plagioclase-hosted melt inclusions from PUBs from the Blanco Transform Fault using a new approach that analyzes the CO2 content trapped in both the glass and vapor bubbles of melt inclusions. This technique revealed that 60 to 90% of the CO2 is trapped within vapor bubbles.
More importantly, minimum entrapment pressures calculated from CO2+H2O analyses demonstrated upper mantle origins of the plagioclase crystals with pressures averaging 3.5 to 6.5 kbars. To obtain reliable analyses, our samples had to be homogenized to dissolve post entrapment crystallization (PEC) back into the melt of melt inclusions. The homogenization temperature used was 1230ºC for the specific sample studied, based on the anorthite content of the plagioclase crystals An83±2. These experiments, performed on a one-atmosphere vertical furnace, revealed a more primitive composition of the melt inclusions compared to typical MORB glasses (Mg# 67.53±2.48 2σ). Time-series heating experiments performed from 30 minutes to 4 days documented systematic changes in the composition of the melt inclusions. This illustrated the importance of re-equilibration processes, as well as the importance in understanding the effects of heating-experiments of melt inclusions. After 4 days, the melt inclusions experienced an average of 15% increase in CO2 in the vapor bubble. With time, the melt inclusion compositions drift toward higher SiO2 and MgO, and lower Al2O3, CaO and
Na2O. The increase in MgO in the melt inclusion is paired with diffusion of MgO from the plagioclase crystals to the melt inclusions. These diffusion profiles exist in the longer run time experiments. Crystal/melt partition coefficients for Mg (DMg) calculated in our 30-minute and
4-day runs demonstrated an anti-correlation exists between DMg and plagioclase anorthite content in the 30-min runs, and in the 4-day runs the anti-correlation is annihilated, and the
DMg decreased by 25%.
The chemistry drifts and increasing CO2 partitioning can be attributed to re- equilibration processes. Crystal relaxation corresponds to a volume increase of a mineral’s structure by remaining at high temperatures for an extensive amount of time. Crystal relaxation likely induced a pressure drop within the melt inclusions, which explains the increase in CO2 partitioning within vapor bubbles and enhanced diffusive exchange between the plagioclase crystals and their melt inclusions. Performing heating-stage experiments on a Vernadsky disk we observed that plagioclase-hosted melt inclusions behave differently from olivine-hosted melt inclusions. Instead of having the vapor bubbles disappearing past the homogenization temperature, the melt inclusions experienced episodes of bubble nucleation. At each increment of temperature starting about 40ºC above the homogenization temperature, a new episode of bubble nucleation would occur. These episodes are likely the results of a pressure gradient forming between the still pressurized melt inclusions and their relaxed host-plagioclase. Both the melt inclusions and vapor bubbles increased in volume. The melt inclusions volume increase is limited by the low coefficients of expansivity of their host-plagioclase, whereas the original amount of volatile dissolved within the melt inclusions constitutes the vapor bubbles’ expansion threshold.
©Copyright by Mélissa J. Drignon July 19, 2018 All Rights Reserved
An Experimental and Analytical Investigation of the Parameters that Influence Measured CO2 in Plagioclase-Hosted Melt Inclusions in MORBs
by Mélissa J. Drignon
A DISSERTATION
Submitted to
Oregon State University
In partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Presented July 19, 2018 Commencement June 2019
Doctor of Philosophy dissertation of Mélissa J. Drignon presented on July 19, 2018
APPROVED:
Co-major professor, representing Geology
Co-major professor, representing Geology
Dean of the College of Earth, Ocean and Atmospheric Sciences
Dean of the Graduate School
I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my thesis to any reader upon request.
Mélissa J. Drignon, Author
ACKNOWLEDGEMENTS
I would like to thank all the persons that somehow participated in the realization of this thesis. First and foremost, I am grateful to Roger Nielsen for selecting as a PhD candidate and trusting me for this study. I know it is never easy to choose someone based on an application, and even less easy to select an international student that is not completely fluent to start with. Frank, thank you for being a co-advisor and giving me the opportunity to work with you as a GRA in the EMP lab. I value my time in the EMP lab and learnt a lot. Thank to both Roger and Frank for being available thorough these few years. Thank you to the other members of my thesis committee: David Graham, Adam Kent and my graduate GCR Dominique Bachelet for being part of this project. Thank you especially to Dave for always being available to discuss my results. Robert Bodnar and Lowell Moore thanks for your help at VT and your interest in this project.
The VIPER group and VIPER seminars helped a lot with my learning and developing my communication skills, thanks to all the VIPER students and falcuties. Thanks to the administration team of CEOAS for being and available and always ready to help. I would like to especially thank some of my CEOAS friends who helped a lot with enjoying my time at OSU: Sophie Wensman, Mike Kula, and my two first housemates Mike Hutchinson and Mike Sepp.
Purple House, past and present, you guys ar great, thanks for being my Oregon family! I am thinking especially to Sarah Stahl, Andrew Morse, Rafael Bermeo Malo, Saeed Khorram, Valerio Moccia and of course Luna.
Mark Biller, thank you for being a wonderful human being.
Je tiens également à remercier Nicolas Cluzel et Laurent Arbaret pour leur aide a Clermont et Orléans. Merci a Yoann Dubernet, Pierre-Yves Tournigand, Sarah Abécassis, Eloïse Bessière, Oliver Barnault et Antoine Coperey pour votre soutient et tous nos délires.
Last but not least, comme ils disent en anglais, merci à ma famille: Papa, Romuald et Marie pour votre soutient et vos encouragements. Les lentilles et pastilles Vichy ont été particulièrement appreciées. Jocelyne, je sais que tu aurais été la première à lire cette thèse d’une traite.
CONTRIBUTION OF AUTHORS
Dr. Nielsen, Dr. Tepley and Dr. Bodnar contributed to the entire study and helped with the redaction of all chapters. Dr. Arbaret and Dr. Cluzel were involved with the analyses and the redaction of Chapter four.
TABLE OF CONTENTS Page
1. GENERAL INTRODUCTION…………………………………………………… 1
2. UPPER MANTLE ORIGIN OF PLAGIOCLASE MEGACRYSTS FROM PLAGIOCLASE ULTRAPHYRIC MORB……………………………………… 8
ABSTRACT…………………………………………………………………… 9
INTRODUCTION……………………………………………………………. 9
GEOLOGICAL SETTING AND METHODOOGY…………………………. 11
RESULTS……………………………………………………………………... 12
DISCUSSION………………………………………………………………… 13 Impact of post entrapment processes on the volatile concentrations of MI…………………………………………………………………. 13 Hydrogen diffusive loss…………………………………………. 13 Decrepitation………………………………………………………... 13 Vapor bubbles……………………………………………….…….… 14 Crystal relaxation…………………………………………………… 15 On the mantle origin of A91-R1……………………………………… 14
CONCLUSIONS……………………………………………………………… 15
ACKNOWLEDGEMENTS…………………………………………………... 15
REFERENCES CITED………………………………………………………. 16
3. RE-EQUILIBRATION PROCESSES OCCURING IN PLAGIOCLASE- HOSTED MELT INCLUSIONS FROM PLAGIOCLASE ULTRAPHYRIC BASALTS………………………………………………………………………… 23
ABSTRACT…………………………………………………………………… 24
INTRODUCTION…………………………………………………………...... 24 Statement of the problem……………………………………………… 24 Background material…………………………………………….……. 26
MATERIALS AND METHODS……………………………………………… 27 Heating experiments…………………………………………………. 27 Analytical methodology……………………………………………… 28
RESULTS……………………………………………………………………... 28
TABLE OF CONTENTS (Continued) Page
Sulfur as an indicator of MI integrity…………………………….…… 28 Effect of run time in MI composition…………………………………. 29
DISCUSSION…………………………………………………………………. 30 Plagioclase-hosted MI: Evidence for early magmatic processes……… 30 Crystal relaxation as a leading cause for chemistry changes: Comparisons between olivine and plagioclase………………………... 31 Crystal relaxation in plagioclase-hosted MI: Phase relation evidence... 33 Mg diffusion at the MI/host interface…………………………………. 34
CONCLUSIONS………………………………………………………………. 36
ACKNOWLEDGEMENTS…………………………………………………… 36
REFERENCES………………………………………………………………… 37
4. EXPERIMENTAL AND ANALYTICAL DETERMINATION OF CRYSTAL RELAXATION FROM PLAGIOCLASE ULTRAPHYRIC BASALTS AND THE EFFECTS ON PLAGIOCLASE-HOSTED METL INCLUSIONS………… 48
ABSTRACT……………………………………………………………….…… 49
INTRODUCTION………………………………………………………………49 Statement of the problem………………………………………………. 49 Background material…………………………………………………… 51
METHODOLOGY……………………………………………………………. 52 Sample provenance……………………………………………………. 52 Experimental procedure………………………………………………. 52
RESULTS……………………………………………………………………... 53 Bubble nucleation episodes during heating experiments……………... 53 Temperatures of bubble nucleation episodes…………………………. 54 2-D quantification of volumes………………………………….……... 54 Quantification correlations…………………………………………… 55
DISCUSSION…………………………………………………………………. 55 Structural differences between olivine and plagioclase………………. 55 Pressure gradient between host confining pressure and internal MI pressure………………………………………………………………... 56
TABLE OF CONTENTS (Continued) Page
Pressure differential to allow an episode of bubble nucleation………... 57
CONCLUSIONS…………………………………………………………………... 57
ACKNOWLEDGEMENTS………………………………………………….……. 58
REFERENCES ………………………………………………………………….… 58
5. GENERAL CONCLUSIONS…………………………………………………….... 69
BIBLIOGRAPHY…………………………………………………………………. 72
APPENDICES……………………………………………………………………… 79
SUPPLEMENTAL INFORMATION FOR CHAPTER TWO…………………. 79 SUPPLEMENTAL INFORMATION FOR CHAPTER THREE……………… 91 SUPPLEMENTAL INFORMATION FOR CHAPTER FOUR…………………106
LIST OF FIGURES
Figure Page
1.1 Map of the Blanco Transform Fault………………………………………….…...... 7
2.1 Frequency diagrams of the pressures in 103 bars recorded by the melt inclusions heated for 30 min and 4 days………………………………………………………...... 20
2.2 Volatiles analyzed within the glass of the melt inclusions by SIMS………...... 21
2.3 CO2 in glass (ppm) vs. CO2 density in vapor bubbles for 30 minutes and 4 days...... 22
3.1 Back scattered electron images and schematic of naturally quenched and homogenized melt inclusions…………………………………………………...... 42
3.2 S versus FeO* plot of both melt inclusions that remained intact and breached during the time-series experiments………………………………………………...... 43
3.3 Al2O3 versus MgO plots of melt inclusions heated during our time-series experiments……………………………………………………………………………… 44
3.4 Na2O/CaO versus run-times plot of the melt inclusions and host-plagioclase heated during the time-series experiments……………………………………………… 45
3.5 Alkaline (Na2O + K2O) versus. SiO2 for melt inclusions heated for 30 min and 4 days compared to PUBs glasses and MORB erupted glasses from the LEPR database…. 45
3.6 Schematic of the plagioclase phase relations based on high pressure experiments on lunar basalts and MORBs, as well as based on 1-atmosphere experiments………..… 46
3.7 Evidences for MgO diffusing from the host-plagioclase to the melt inclusions after 4 days of heating experiments…………………………………………………….... 47
3.8 Mg partition coefficients calculated between MI and host plagioclase versus the anorthite contents of the host plagioclase…………………………………………...... 47
4.1 Pictures showing the evolution of the heating experiments………………………… 61
4.2 Initial volume of melt inclusions (cm3) versus the temperature of first bubble nucleation episode (ºC)………………………………………………………………….. 63
4.3 Initial volume (cm3) versus final volume (cm3) for both melt inclusions and vapor bubbles…………………………………………………………………………………… 64
4.4 Volume increase of both the vapor bubbles and the melt inclusions (%) versus temperature at which the final episode of bubble nucleation occurred (ºC)……………... 65
LIST OF FIGURES (Continued)
Figure Page
4.5 Volume increase of both the vapor bubbles and the melt inclusions (%) versus temperature increase between the first and last episode of bubble nucleation (ºC)……… 66
4.6 Coefficients of thermal expansivity of forsterite and anorthite from previous studies. 67
4.7 Cartoon explaining the different phases through which plagioclase-hosted MI relaxe. 68
LIST OF TABLES
Table Page
2.1 Averages of the volatiles analyzed for both experimental run times within the glass of the melt inclusions by SIMS: H2O in wt.%; S, Cl, and F………………...... 19
3.1 Differences in compositions observed between melt inclusions heated for 30 Minutes, 4 hours and days………………………………………………………………. 43
4.1 Temperature evolution of each sample………………………………………………. 62
4.2 Summary of the volume estimates for melt inclusions and vapor bubbles………….. 63
LIST OF APPENDIX FIGURES
Figure Page
A.1 BSE images of MI prior to and post homogenization experiments……………….…. 85
A.2 Percentage of crystallization versus Mg# plot of both 30-minute and 4-day runs….. 90
B.1 S versus FeO* plot of both melt inclusions heated at 1200°C in the air and under FMQ+1 for 30 minutes ……………………………………………...... 94
B.2 MgO versus a) SiO2; b) Al2O3; c) FeO*, d) CaO, e) K2O, f) Na2O plots of heating experiments performed at 1200°C for 30 minutes in the air and under FMQ+1…….…. 95
B.3 Additional Mg diffusion profiles of samples heated for both 30 minutes and 4 days. 102
LIST OF APPENDIX TABLES
Table Page
A.1 Data set of all the volatile analyses performed on melt inclusions heated for 30 minutes and 4 days with their associated calculated pressures and depths……………….. 86 A.2 Results of the t-tests comparing the statiscal differences observed in between the 30-min and 4-day runs in terms of volatiles and pressures estimates……………..….. 87
A.3 Mg# as well as the percentage of crystallization undergone by melt inclusions heated for both 30 minutes and 4 days based on their MgO and FeO contents……….… 88-89
B.1 Chemical composition of the melt inclusions homogenized for 30 min at 1200°C using the 1-atmosphere vertical furnace…………………………………………………. 96
B.2 Data set of the chemistry of the melt inclusions heated for 30 min, 4 hours, and 4 days.……………………………………………………………………………………… 97
B.3 Data set of the chemistry of the melt inclusions heated for 30 min, 4 hours, and 4 days that breached during the heating experiments……………………………………… 98
B.4 Data set of the host-plagioclase heated for 30 min, 4 hours, and 4 days…………… 99
B.5. Data set used to calculate the errors based on the reproducibility of the standards… 100
B.6 Results of the t-tests on the MI chemistry………………………………….………… 101
B.7 Equations used to calculate the true MgO concentration at the MI/host interfaces…. 103
B.8 Data set detailing the DMg calculations……………………………………………... 104
B.9. Details of the mass balance calculations……………………………………………. 105
B.10 Results of the t-tests on the diffusion data set……………………………………… 106
P a g e | 1
An Experimental and Analytical Investigation of the Parameters that Influence Measured CO2 in Plagioclase-Hosted Melt Inclusions in MORBs
Chapter One
General Introduction
Mid-ocean ridges (MORs) are the most volumetric significant volcanic features on Earth. Since the discovery of the extent to which they dominate the earth’s surface, these 65,000 km long features have been the focus of extensive research. The compositions of mid-ocean ridge basalts (MORBs) erupted at MORs as well as the extensive magmatism that makes up the lower and middle crust (Coogan, 2014) have long been used as proxies for the processes active in the upper mantle. Due to their large volume and primitive character, MORBs are most valuable regarding our understanding of the heat and mass balance of the Earth’s upper mantle. Most investigations have focused on MORB glasses, which are the last products to be formed by quenching during an eruption. MORB glasses taught us a great deal on the composition of the oceanic crust and how mantle melts are modified through fractionation, mixing and assimilation (Coogan, 2014; Winter 2015). However, crustal processes have modified their character to the extent that information on magmatic processes in the mantle have been obscured. Therefore, one must look at more primitive melt samples. To find the most suitable samples, one must understand where to find them based on the structure of the oceanic crust. Current models for the formation of the oceanic crust involve melting via near-adiabatic decompression melting (Coogan, 2014). Those melts form from high extents of melting (8- 20%), and originate from the depleted upper mantle, which has been extensively and continuously melted since its formation (Klein and Langmuir, 1987). Once generated, the melts percolate through the oceanic crust via multiple melt lenses and mush zones intruded by sills and dikes forming pathways for melts to reach the seafloor. Small ephemeral mush zones have been imaged under fast spreading ridges, whereas it is thought that at slow- and intermediate- spreading ridges, melts ascend and erupt without pause on the oceanic seafloor (Coogan, 2014). This observation, together with the study of MORB compositions, brought to light the more primitive nature of MORBs erupted at slow and intermediate spreading ridges, where no P a g e | 2 reservoir provides the possibility for melts to evolve further in the upper crust. This observation agrees with the decrease of Mg#, and the Cl-enrichment from assimilated hydrothermally- altered crust at fast-spreading ridges (Nielsen et al., 2000; Rubin and Sinton, 2007; Michael and Schilling, 1989). Therefore, to capture the most primitive melts representative of the upper depleted mantle, samples coming from slow-spreading ridges are more appropriate. It is at slow and intermediate spreading ridges that plagioclase ultraphyric basalts (PUBs) erupt. PUBs are found worldwide. These rocks are defined as MORB with >15 vol.% plagioclase megacrysts (0.2-2 cm) with smaller and less abundant olivine and spinel. The crystals are entrained in a glass that has the same composition as typical aphyric MORBs. The plagioclase megacrysts, however, are highly anorthitic (An80 to An92). This primitive character is not in equilibrium with the host MORB glasses in which the plagioclase megacrysts are enclosed (Lange et al., 2013). Therefore, the plagioclase megacrysts represent a unique opportunity to understand the primitive nature of the magmatic processes responsible for the oceanic crust’s formation. For this purpose, studying melt inclusions present within the plagioclase megacrysts provides a new approach with promising outcomes. It is often assumed that melt inclusions, drops of magma trapped at depth during a mineral’s formation, remain a closed-system once formed. As a result, melt inclusions are potential examples of the composition of the melts from which minerals crystallize. This provides more direct information about melts that represent earlier stages in the evolution of the magmatic system. To assess the compositions of the most primitive melts, studies have focused mainly on olivine-hosted melt inclusions due to the primary appearance of olivine in basaltic crystallization sequence (Wanless and Shaw, 2012; Wanless et al 2014; 2015; Wallace et al., 2015; Le Voyer et al. 2017). Other minerals have been successfully studied for melt inclusions, but only a handful of studies have looked at plagioclase-hosted melt inclusions (Helo et al., 2011; Coumans et al., 2015; Neave et al., 2015, 2017). The assumption is that a mineral with two cleavage planes such as plagioclase crystals cannot accommodate for P-T condition changes without their melt inclusions breaching. One of the main goals of this study is to investigate the robustness and reliability of plagioclase as a host. Compared to olivine, plagioclase comes later in the crystallization sequence of MORBs (Coogan et al., 2014). However, plagioclase has an advantage over olivine that make this mineral more meaningful for this study: it is less dense. Plagioclase with anorthite contents ranging from 55 to 89 have densities of 2.64-2.68 g.cm-3 at 1200ºC. At this temperature a typical MORB melt would have density of 2.56-2.66 g.cm-3 (Campbell et al., 1978). Olivine, on the other hand, have densities of 3.3-3.4 g.cm-3. Thus, in a typical MORB melt, olivines sink, P a g e | 3 and plagioclase crystals float or have neutral buoyancy. It is thought that typical MORB melts travelling through trocotolitic cumulates based in the upper mantle or the lower crust have the possibility to entrain plagioclase crystals due to their low densities, but that olivine is usually left behind (depending on the magma ascent rate). Melts and plagioclase crystals can then be erupted as PUB if their velocity is not lowered by melt lenses that would filter the plagioclase crystals out (Lange et al., 2013). This last point explains why PUBs are only found at slow- and intermediate-spreading centers, and off-axis at fast-spreading centers. Thus, despite the later appearance of plagioclase in the crystallization sequence of MORBs, the minerals erupted with PUBs are supposedly more primitive than olivine-hosted melt inclusions usually studied in MORBs. Indeed, numerous investigations have focused on finding the crystallization pressures of olivines in MORBs (Wanless and Shaw, 2012; Wanless et al., 2014; Wanless et al., 2015; Wallace et al., 2015; etc). Because of the high density of olivines, melts can hardly carry them over great distances while ascending to the seafloor, whereas melts do have the potential to carry plagioclase crystals through the entire oceanic crust. Only much fewer and smaller primitive olivine crystals that crystallized in the lower crust and the mantle can be erupted (Lange et al., 2013). The purpose of this study is to gain a better understanding of the how homogenization affects the composition of melt inclusions and their hosts, with the goal of better understanding the processes leading to the formation of the oceanic crust. In this view, plagioclase-hosted melt inclusions in PUBs will provide complementary information to allow better tests of the existing models and diversity of melt compositions contributing to the crust. Sample selection for this investigation was based on three criteria: (1) the plagioclase megacrysts need to contain abundant melt inclusions; (2) the plagioclase should have a high anorthite content, so we can expect the melt inclusions to be representative of primitive MORB melts; (3) choosing homogeneous crystals would simplify the study so we can assume the crystals formed from one melt at one temperature. A91-R1 is a sample that comprises those important characteristics, notable by having a tight range of anorthite content An83±2, which shows its homogeneity (Lange et al., 2013). This sample was collected on the Blanco Transform Fault (BTF). The BTF is a 350-km long right lateral transform boundary that separates the Juan de Fuca Ridge to the NW from the Gorda Ridge to the SE. It is a relatively young feature of about 5 Ma. It is a complex feature composed of multiple depressions and faulting zones, possibly pull-apart basins, (Figure 1.1) cut in half by the Cascadia Channel. P a g e | 4
Embley and Wilson, 1991 suggested that the BTF results from a shift in plate motion initiated 5 Ma ago on the Gorda Ridge. Most of the BTF then formed over the next 5 Ma by a series of strike-slip faults and depressions that started with the formation of the Cascadia Depression and ended with The Surveyor and Blanco Depression less than 1 Ma ago. The youngest oceanic crust is on the western end of the fault and was dated at 1.5 Ma based on magnetic anomaly and sea beam bathymetry (Figure 1.1; Wilson et al., 1984; Embley and Wilson, 1991). This dissertation contains three main projects. The first two projects aim to better understand the changes in the composition of melt inclusions during experimental heating and to interpret the volatile content in terms of depth of formation. The third project documents the significance of crystal relaxation as a post-entrapment process. The first project was designed to document the minimum entrapment depths of plagioclase-hosted melt inclusions from our sample A91-R1, dredged on the East Blanco Depression (44°12.55, 129°37.93; Figure 1.1; Sprtel 1997). The calculated depth is based on
CO2 and H2O analyses, which are proxies for pressure and therefore depth of entrapment
(Newman and Lowenstern, 2002). Both the CO2 trapped within the melts and the vapor bubbles of melt inclusions were analyzed, using SIMS and Raman respectively. The necessity of analyzing CO2 trapped within vapor bubbles of melt inclusions arises from the recent recognition that vapor bubbles usually contain between 40 and 90% of the total amount of CO2 initially present within a melt inclusion (Steele-MacInnis et al., 2011; Hartley et al., 2014; Moore et al., 2015; Esposito et al., 2016). Melt inclusions heated for both 30 minutes and 4 days at 1230ºC were analyzed for volatiles. Understanding how the concentration changes with time is critical for our evaluation of the degree to which the volatiles may be lost through diffusion or redistributed due to changes in the structure of the host mineral. In addition, our ability to interpret the CO2+H2O analyses in terms of the overall petrogenesis of the sytem, the Mg# of the melt inclusions was used to document how close the melts were to primary magmas (Michael and Graham, 2015). The second project focuses on the major element chemistry (as opposed to the focus on volatiles in part 1) of plagioclase-hosted melt inclusions, and the effects of heating as a function of time. This part of the study relies on electron microprobe analyses of major and trace elements of the melt inclusions. The first order observation of the composition of the inclusions shows that the plagioclase-hosted melt inclusions are more primitive (higher Mg #) compared to typical MORB glasses. They also demonstrate the time dependent chemical drift of melt inclusions. This drift is attributed to crystal relaxation, a result of the fact that the heating takes P a g e | 5 place at 1 atm pressure, whereas the crystals formed at 3.5-6.5 kbar. Crystal/ melt partition coefficients calculated using Mg (DMg), a major element in melt inclusions, but a trace element in plagioclase, highlight the occurrence of diffusion profiles appearing at the melt inclusion/plagioclase interface after long (4 days) experimental runs. The calculated DMg shows a significant decrease after 4 days, and the elimination of the anti-correlation between DMg and the anorthite of the plagioclase near the melt inclusion/plagioclase interface. The third project constitutes a more in-depth proof of concept investigation on crystal relaxation. As a crystal relaxes under high temperature its volume increases. While this volume increases, the mineral’s structure reaches a threshold where it cannot handle its internal pressure, leading to a pressure drop. The leading hypothesis gathered from this dissertation’s first two projects is that crystal relaxation induces volatile exsolution into vapor bubbles and enhances diffusion between melt inclusions and host. The aim of this last project is therefore to document and characterize crystal relaxation. Based on Schiavi et al. (2016), who performed heating-stage experiments on olivine-hosted melt inclusions, an experimental procedure based on heating stage experiments and micro tomography volume characterization was designed. The heating-stage experiments were performed to directly observe induced crystal relaxation. The experiments revealed a difference of behavior of the plagioclase crystals compared to olivine crystals, as multiple episodes of bubble nucleation were observed passed the temperature of homogenization. Measurments made on the volume increases of melt inclusions showed their dependence to the thermal expansion coefficients of the plagioclase-hosts. The vapor bubbles volume expansion is related to the original amount of volatile dissolved within the melt inclusions.
REFERENCES
Campbell, I.H., Roeder, P.L., and Dixon, J.M.., 1978, Plagioclase buoyancy in basaltic liquids as determined with a centrifuge furnace: Contributions to mineralogy and petrology, v.67, p.369-377. Coogan, L.A., 2014, The lower oceanic crust, in Treatise on geochemistry, second edition: Elsevier, p.497-541. doi: 10.1016/B978-0-08-095975-7.00316.8. Coumans, J., Stix, J., and Minarik, W., 2015, The magmatic architecture of Taney Seamount- A, northeast Pacific Ocean: Journal of Petrology, v.56, p. 1037-1067. doi: 10.1093/petrology/egv027. Embley, R. W., Kulm, L. D., Massoth, G., Abbot, D., and Holmes, M.,1987, Morphology, structure, and resource potential of the Blanco Transform Fault Zone: Chapter Twenty- Three, p.550-561. Embley, R.W., and Wilson, D.S., 1992, Morphology of the Blanco transform fault zone-NE Pacific: Implications for its tectonic evolution. Marine geophysical researches, v.14, i.1, p.25-45. P a g e | 6
Esposito, R., Lamadrid, H.M., Redi, D., Steele-MacInnis, M., Bodnar, R.J., Manning, C., De
Vivo, B., Cannatelli, C., and Lima, A., 2016, Detection of liquid H2O in vapor bubbles in reheated melt inclusions: Implications for magmatic fluid composition and volatile budgets of magmas: American Mineralogist, v.101, p.1691-1695. doi: 10.2138/am- 2016-5689. 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 Science Letters, v.393, p.120-131. doi: 10.1016/j.epsl.2014.02.031. Helo, C., Longpre, M.A., Shimizu, N., Clague, D. A., and Stix, J., 2011, Explosive eruptions at mid-ocean ridges driven by CO2-rich magmas: Nature Geoscience, v.4, p.260-263. doi: 10.1038/NGEO1104. Klein, E.M., and Langmuir, C.H., 1987, Global correlation of ocean ridge basalt chemistry with axial depth and crustal thickness: Journal of Geophysical Research, v.92, p.089–8115. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2012, Diverse Sr isotope signatures preserved in mid-oceanic-ride basalt plagioclase. Geology, v. 41, p. 279-282. https://doi.org/ 10.1130/G33739.1. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2013, The petrogenesis of plagioclase-phyric basalts at mid-ocean ridges: Geochemistry, Geophysysical, Geosystems, v.14, p.3282-3296. doi: 10.1002/ggge.20207. Le Voyer, M., Kelley, K. A., Cottrell, E., and Hauri, E. H., 2017, Heterogeneity in mantle carbon content from CO2-undersaturated basalts: Nature communications, v. 8. doi: 10.1038/ncomms14062. Michael, P.J., and Schilling, J-G., 1989, Chlorine in mid-ocean ridge magmas: Evidence for assimilation of seawater-influenced components: Geochimica et Cosmochimica Acta, v.53, p.3131–3143. Michael, P.J., and Graham, D.W., 2015, The behavior and concentration of CO2 in the suboceanic mantle: inferences from undegassed ocean ridge and ocean island basalts. Lithos, v. 236-237, p. 338-351. doi: 10.1016/j.lithos.2015.08.020. Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. Neave, D. A., Maclennan, J., Thordarson, T., and Hartley, M. E., 2015, The evolution and storage of primitive melts in the Eastern Volcanic Zone of Iceland: the 10 ka Grímsvötn tephra series (i.e. the Saksunarvatn ash): Contribution to Mineralogy and Petrology, v. 170, p. 1-23. doi:10.1007/s00410-015-1170-3. 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 Iceland: Geochimica et Cosmochimica Acta, v.205, p.100–118. doi: 10.1016/j.gca.2017.02.009. Newman, S., and Lowenstern, J.B., 2002, VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic Excel: Computers in Geosciences, v.2, p.597-604. Rubin, K.H., and Sinton, J.M., 2007, Inferences on mid-ocean ridge thermal and magmatic structure from MORB compositions: Earth and Planetary Science Letters, v.260, p. 257– 276. Doi: 10.1016/j.espl.2007.05.035. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine-hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. P a g e | 7
Sprtel, Frank. M., 1997, Two styles of oceanic near-ridge volcanism for the Southeast Indian Ocean and the NE Pacific Ocean. [Masters. thesis]: Corvallis, Oregon State University, 53 p. Steele-MacInnis, M., Esposito, R., and Bodnar, R.J., 2011, Thermodynamic model for the effect of post-entrapment crystallization on the H2O-CO2 systematics of volatile-saturated silicate melt inclusions: Journal of Petrology, v. 52, p.2461-2482. doi: 10.1093/petrology/egr052. Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. Wanless, V.D., and Shaw, A.S, 2012, Lower crustal crystallization and melt evolution at midocean ridges: Nature Geoscience, v.5, p.651-655. doi: 10.1038/NGEO1552. Wanless, V.D., Behn, M.D., Shaw, A.M., and Plank, T., 2014, Variations in melting dynamics and mantle compositions along the Eastern Volcanic Zone of the Gakkel Ridge: insights from olivine-hosted MIs: Contribution to Mineralogy and Petrology, v.167. doi: 10.1007/s00410-014-1005-7. Wanless D. V., Shaw, A.M., Behn, M.D., Soule, S.A., Escartin, J., and Hamelin, C., 2015, Magmatic plumbing at Lucky Stike volcano based on olivine-hosted melt inclusion composition: Geochemistry, Geophysics, Geosystems, v. 16, p. 126-147. doi: 10.1002/2014GC00551. Wilson, D.S., Hey, R.N., and Nishimura, C., 1984, Propagation as a mechanism od reorientation of the Juan de Fuca Ridge, Journal of Geophysical Ressource, v.89, p384-392. Winter, J.D., 2013, Mid-ocean ridge volcanism, in Principles of igneous and metamorphic petrology, Second edition: Pearson Education, Chapter 13, p. 245.
Figure 1.1: Map of the Blanco Transform Fault Zone retrieved from GeoMapApp. The ridges appear in red, the depressions in black, and the Cascadia Channel in blue. The location of the sample used for this study A91-R1 is pinpointed with a brown star. This map was made after Embley and Wilson (1992).
P a g e | 8
Chapter Two
Upper Mantle Origin of Plagioclase Megacrysts from Plagioclase Ultraphyric MORB from the Blanco Transform Fault
Mélissa J. Drignon Roger L. Nielsen Frank J. Tepley III
Robert J. Bodnar
This manuscript is in Review in: Geology MS#G45542 3300 Penrose Place, Boulder, CO 8031 August 2018
P a g e | 9
ABSTRACT
To better interpret data from melt inclusions and their host phenocrysts it is necessary to understand the pressures at which crystallization and melt trapping occur. Pressure can be estimated based on the H2O-CO2 concentration of the melt inclusions. Most of these estimates to date are for olivine-hosted melt inclusions, largely based on the assumption that plagioclase phenocrysts do not retain volatiles and/or are the product of late-stage crystallization. We have investigated the effects of post-entrapment processes on plagioclase-hosted melt inclusions in one plagioclase ultraphyric basalt from the Blanco Transform Fault. Our results demonstrate that plagioclase-hosted melt inclusions retain their CO2 concentrations even after 4 days of homogenization experiments. Furthermore, the total reconstructed CO2 (glass+vapor) in the melt inclusions indicates that the plagioclase megacrysts crystallized at pressures of 2.3-9.1 kbars, which are at or below the local MOHO. These minimum pressures of trapping suggest that adequate re-homogenization experiments, together with the measurements of the CO2 concentrations of vapor bubbles, can counter the effects of post-entrapment processes that altered the original information contained within melt inclusions.
INTRODUCTION
Plagioclase is a common phenocryst phase in mid-ocean ridge basalts (MORBs), but it has been assumed to crystallize later in the differentiation sequence, and at a shallower depth, compared to olivine (Bryan, 1983). Therefore, most work on melt inclusions (MI) in MORBs aimed at understanding the early crystallization history has focused on olivine (Sobolev and Shimizu, 1993 ; Laubier et al., 2012; Wanless and Shaw, 2012; Le Voyer et al., 2017; and many others). In addition, it has been commonly assumed that MI in plagioclase crystals do not preserve the primary volatile concentration due to leakage along the two cleavage planes. However, little evidence exists to support the contention that plagioclase-hosted MI are more susceptible than olivine-hosted MI to CO2+H2O+S loss. This study focuses on plagioclase-hosted MI in one plagioclase ultraphyric basalt (PUB) from the Blanco Transform Fault (BTF). PUBs are characterized by a high modal % of anorthitic megacrystic plagioclase (>15 % ; Lange et al., 2013), which host abundant MI. These megacrysts are often xenocrystic (Lange et al. 2013), but their host glasses are usually similar in chemistry to the array of erupted lavas from their ridge segment. The plagioclase megacrysts and their MI are usually more primitive (more An- and Mg-rich) and compositionally diverse P a g e | 10 than late-crystallizing plagioclase crystals that are in equilibrium with their associated lavas near the surface (Sours Page et al., 2000 ; Lange et al., 2013; Neave et al., 2014).
The H2O-CO2 concentrations of MI can be related to pressures through solubility models, and therefore can give useful information on the crystallization pressures of the plagioclase megacrysts (Dixon et al., 1995 ; Newman and Lowenstern, 2002). Most existing pressure estimates are from olivine-hosted MI in MORBs and are consistent with crystallization within the oceanic crust, with a minority indicative of deeper mantle crystallization (Wanless and Shaw, 2012; Wallace et al., 2015; Le Voyer et al. 2017., Wanless and Behn, 2017). However, fewer studies have utilized plagioclase-hosted MI to determine the pressure of plagioclase crystallization, and concurrent MI entrapment has shown that crystallization depths are as deep as for olivine-hosted MI, and often even deeper (Helo et al., 2011; Coumans et al., 2015; Neave et al., 2015; 2017).
Determining pressures of MI entrapment based on H2O-CO2 concentrations is complicated by the fact that CO2 is partitioned between various phases. Specifically, CO2 is often present in both the glass and vapor bubbles within MI (Steele-MacInnis et al., 2011;
Hartley et al., 2014; Moore et al., 2015). CO2-rich vapor bubbles are formed in MI as a host crystal cools and depressurizes. CO2 solubility decreases in response to both decreasing internal MI pressure. Given that calculations of saturation pressures of MI require knowledge of the total CO2 and H2O concentrations, an accurate estimate requires that we determine the total bulk CO2 concentration of the MI. The aim of this work is to assess whether saturation pressures calculated from plagioclase-hosted MI are reliable, and how they relate to existing studies on olivine-hosted MI from MORBs. Utilizing plagioclase-hosted MI has the potential to advance our understanding of how the oceanic crust forms, and, thus, ascertaining the reliability of plagioclase-hosted MI is essential.
We present total bulk CO2 concentrations in plagioclase-hosted MI by combining the
CO2 contained in the vapor bubbles with that in the glass. Moreover, we conducted a set of homogenization experiments to evaluate the processes that influence the ability of plagioclase- hosted MI to retain their original volatile budgets. Our results document the exsolution of CO2 from the melt phase into the vapor phase as a function of time when the MI are held at high temperature. However, the total bulk CO2 concentration of the MI does not change in between our two run times. Therefore, CO2 is not lost from the MI during cooling and depressurization but, rather, is transferred from the melt into the vapor bubble. More importantly, based on CO2-
H2O equilibrium solubility models, our results indicate that the plagioclase megacrysts originate from the mantle, with minimum crystallization pressures of 2.3-9.1 kbars. P a g e | 11
GEOLOGICAL SETTING AND METHODOLOGY
PUBs are found worldwide at slow and intermediate spreading centers, as well as at off- axis seamounts and transform faults where the absence of a magma reservoir may inhibit crystals from being filtered out of their host magmas (Lange et al., 2013). We focus on the BTF, a right-lateral transform fault zone that extends for over 350 km and separates the Juan de Fuca ridge to the northwest from the Gorda Ridge to the southeast (Embley and Wilson, 1992). This study focuses on megacrysts from a single sample A91-1R located 44°12.55, 129°37.93 (Sprtel 1997), characterized by abundant (~20%) euhedral, centimeter-scale (0.2-1 cm) plagioclase megacrysts. Plagioclase crystals from this sample do not present any zoning and have a narrow range of anorthite contents (An83+2). PUB megacrysts have similar Sr isotopic signature than their host lavas (Lange et al., 2012). Experiments performed on high An-plagioclase revealed that MI are trapped during an isothermal period of growth where fast diffusion prevent the existence boundary layers (Kohut and Nielsen, 2004). We performed homogenization experiments at Oregon State University (Corvallis, OR USA) using a 1-atm vertical furnace. The naturally quenched MI observed in our samples are fully or partially crystallized by post-entrapment crystallization, and contain one or multiple vapor bubbles, making experimental homogenization a necessity (Figure A1). Individual crystals were suspended in the furnace at 1230°C for 30 minutes or 4 days and quenched (see APPENDIX A for more information). The crystals were then polished to expose the MI for Raman and Secondary Ionization Mass Spectrometry (SIMS) analyses. Raman analyses were performed at Virginia Tech (Blacksburg, VA USA) to determine the density of CO2 within the vapor bubbles. Then, the same MI were analyzed at Woods Hole Oceanographic Institution
(WHOI, Woods Hole, MA USA) by SIMS to determine the CO2, H2O, S, Cl and F concentrations within the glass of the MI. The total CO2 concentration in the MI was reconstructed by combining the CO2 from the vapor bubble and the CO2 from the glass following the technique of Esposito et al. (2011) and Moore et al. (2015). We used VolatileCalc 2.0 (Newman and Lowenstern, 2002) to calculate the minimum trapping pressures based on the
H2O-CO2 concentrations of the MI. To estimate minimum entrapment pressures using
VolatileCalc 2.0. We computed H2O-CO2 concentrations from 27 MI from 8 crystals that were homogenized 30 minutes and from 16 MI from 8 crystals that were homogenized for 4 days. Additional information of the methodology used can be found in Appendix A. P a g e | 12
RESULTS
The reconstructed MI CO2 concentrations based on both the CO2 concentration of vapor bubbles and glass for individual MI held at 1230°C for 30 minutes averages 2253±800 (1σ) ppm (ug/g) CO2, and 2714±1183 ppm CO2 for the 4-day runs (Table 2.1). The melts are anhydrous, with < 0.1 wt. % H2O for both runs. Minimum crystallization pressures calculated using VolatileCalc2.0 (Newman and Lowenstern, 2002) produced similar results for both run times, and no systematic difference was observed between MI from individual hosts. The 30- minute runs yield an average of 4089± 1184 bars (1σ), and the 4-day runs yield an average of 4732±1518 bars (Table 2.1; Figure 2.1). These uncertainties reflect sample variability, not the uncertainty in the solubility model. In terms of distribution, 71±14 % (1σ) of the CO2 is in the vapor bubbles after the 30-min run, whereas 83±10 % of the CO2 is in the vapor bubbles after the 4 day runs (Table 2.1). It is important to note here that >50% of the CO2 is usually present within the vapor bubbles for both run-times, highlighting the importance of analyzing vapor bubbles for CO2. All data are included in Table A1. T-test calculations for both the total reconstructed CO2 and the CO2 distribution within vapor bubbles revealed that : (1) the total reconstructed CO2 is not statistically significant between the 30-minute and 4-day runs ; (2) the
CO2 distribution within vapor bubbles is statistically different between the two experimental run times (Table A2). The behavior of S, Cl and F concentrations as a function of time, as determined by SIMS analyses, differed from that of H2O and CO2. S concentrations average 821±100 ppm (1σ) for the 30-minute runs, and 671±179 ppm for the 4-day runs (Table 2.1). All samples but two are saturated with S (Wallace and Carmichael, 1994). The two MI that are not S saturated were considered breached (S <90 ppm) and were not further considered for this study (Table A1). As with S, Cl concentrations are identical within error for the two run times, with 36±12 ppm and 55±63 ppm, respectively, for the 30-min and the 4-day runs. In contrast, F in the 30-minute runs averages 20±15 ppm but is below the detection limit of 1.8 ppm for the 4-day runs (Table 2.1; Figure 2.2). We do not know whether the F diffused out of the MI during the 4-day runs or if it exsolved from the melt and into the vapor bubbles (Koleszar et al., 2009). To avoid this complication, we suggest that plagioclase-hosted MI should not be held at high temperature for more than 30 minutes.
P a g e | 13
DISCUSSION
Impact of post entrapment processes on the volatile concentrations of MI
Hydrogen diffusive loss
MORBs are typically anhydrous with ≤ 0.5 wt. % H2O (Métrich and Wallace, 2008).
Our samples exhibit the same low H2O concentration for both run times, indicating that no major H2O occurred in between the two runs. However, we cannot rule out the possibility of hydrogen diffusive loss. Experiments performed on An30 plagioclase revealed an increase of OH loss with increasing run-time and temperature, such that at 1000°C, 50% of OH was lost in <1 day (Johnson and Rossman, 2013). As both the 30-minute and 4-day runs show the same low H2O concentrations and that they likely are anhydrous in nature, the resulting pressure estimates would only be slightly influenced by hydrogen diffusive loss. In addition, in H2O- poor systems, hydrogen diffusive loss does not have a strong impact on the solubilities of other volatiles such as CO2 due to the nearly flat H2O-CO2 solubility curves in P-T space (Gaetani et al., 2012; Bucholz et al., 2013). Moreover, hydrogen diffusive loss would result in an underestimation of the H2O concentration of the MI, which would underestimate the saturation pressures as described from the experiments performed by Dixon et al. (1995). Consequently, the entrapment pressures calculated for our samples would represent minimum values but are still consistent with our conclusion that MI in plagioclase studied here were trapped within the upper mantle.
Decrepitation
Decrepitation leads to the loss of part of the volatiles such as CO2, leading to a possible incorrect interpretation that the MI trapped a volatile-undersaturated melt (Maclennan, 2017). Decrepitation is more likely to happen for volatile-rich MI with high internal pressure like the MI considered for this study (Maclennan, 2017; Wanless and Behn, 2017). Maclennan (2017) suggested that decrepitation can be prevented in MI through post-entrapment cooling and depressurizing. In CO2-rich and H2O-poor systems such as ours, the formation of vapor bubbles does induce an internal pressure decrease within the MI, thus protecting them from decrepitation. As all our MI contained one or more vapor bubbles, they might have been preserved thanks to the formation of vapor bubbles.
P a g e | 14
Vapor bubbles Vapor bubbles can form due to both the cooling of MI and with post-entrapment crystallization (PEC), leading to volatile exsolution (Kent, 2008). All our naturally quenched
MI contained one or more vapor bubbles. As observed, more than 50% of the CO2 is usually present within the vapor bubbles for both experimental run-times, and this percentage increased in the 4-day runs. In addition, the MI heated for 4 days likely experienced compositional drifts that could impact the solubility of CO2 (see Chapter 3). Long experimental run times at magmatic temperatures are known to modify the major element composition of olivine-hosted MI (Danyushesvky et al., 2002). These changes in major element chemistry potentially lowered the solubilities of volatiles in the melt phase, leading to further exsolution. For example, it has been demonstrated that the solubility of CO2 decreases with decreasing CaO, MgO and alkaline contents (Brey, 1976; Moore, 2008; Vetere et al., 2014).
Crystal relaxation Schiavi et al. (2016) demonstrated that crystal relaxation occurs when a crystal is kept at magmatic temperature for long experimental run times at 1-atm. Crystal relaxation operates though the pressure differential that is established between the internally pressurized MI and their relaxed hosts, which confining pressure is 1-atm. We suggest that when the MI are instantaneously heated to 1230°C and the pressure inside the MI reaches several kilobars, the plagioclase host surrounding the MI gradually relaxed to eliminate the pressure gradient between the MI and surrounding plagioclase host, resulting in an increase in MI volume and concomitant decrease in MI internal pressure, inducing further CO2 exsolution into the vapor bubbles. This argument is supported by the increased CO2 density of the vapor bubbles after the 4-day runs compared to the 30-minute runs (Figure 2.3).
On the mantle origin of A91-R1
The best estimate for the thickness of the oceanic crust under the Blanco Transform Fault is ~6.7-10 km based on seismic imaging (McNutt, 1979; Christeson et al., 2010). The samples analyzed in this study yield a range of CO2 concentrations that indicate minimum pressures ranging from 2279 to 9104 bars, corresponding to depths of 6 to 27 km, all of which are at or below the local MOHO (Figure 2.1, Table A1). These pressures indicate that the MI were saturated with CO2 at the time of eruption as they were erupted at 320 bars (Sprtel, 1997). In contrast, most pressures calculated based on MORB glasses and olivine-hosted MI P a g e | 15 compositions indicate the olivine phenocrysts formed in the mid-crust, with a minority at deeper levels (Wanless and Shaw, 2012; Wallace et al., 2015; Wanless and Behn, 2017 and others). Previous studies of plagioclase-hosted MI in MORBs recorded both crustal (Neave et al., 2015; 2017) and mantle depths (Helo et al., 2011; Coumans et al., 2015). Current models for the formation of PUBs involve plagioclase crystallization in the upper mantle and melt interactions with troctolitic cumulates during magma ascent (Kohut and Nielsen, 2003; Lange et al., 2013). MORB melts can entrain high-An plagioclase in the upper mantle due to their similar density (~2.7 g/cm3). During ascent, the melts become less dense than the plagioclase crystals that can only be erupted if the ascent rate is fast enough (Lange et al., 2013). The density contrast between early-crystallized olivine (3.2 g/cm3) and plagioclase could result in crystal sorting and the eruption of PUBs containing little or no olivine (Kohut and Nielsen, 2003; Lange et al., 2013) assuming a crustal density of 3 g/cm3. Therefore, our mantle pressure estimates are consistent with the conceptual model of PUB formation conceptualized by Kohut and Nielsen (2003) and Lange et al. (2013). In addition, the mantle origin of our sample is supported by the near-primary compositions of the MI that have a Mg# of 67.53±2.48 (2σ; Michael and Graham, 2015). As it is beyond the scope of this paper to assess the major- and trace-element concentrations of the MI, this argument is further developed in Appendix A.
CONCLUSIONS
Based on homogenization experiments, we have documented that plagioclase-hosted MI retain their original CO2 concentrations, even after 4-day runs at 1230°C. However, during these runs conducted at 1-atm, crystals underwent time-dependent relaxation, resulting in expansion of the plagioclase-hosts and their MI. The resulting pressure drop within the MI and concomitant decrease in CO2 solubility led to the exsolution of CO2 from the melt and into the vapor bubbles of the MI. However, the total CO2 concentrations, and therefore the estimated minimum pressures of entrapment, do not change. Pressure estimates obtained from the MI indicate that plagioclase megacrysts from the BTF are the product of crystallization in the upper mantle.
ACKNOWLEDGEMENTS
This project is supported by NSF grants EAR 1634206 and EAR 1634211. We are grateful for the help received at VT by Lowell Moore and Natalie Sievers. We would like to thank David Neave and two anonymous reviewers, who’s comments greatly improved the quality of this manuscript. In addition, discussions and comments from David Graham and P a g e | 16
Adam Kent (OSU) were also beneficial to this paper. Mark Biller’s assistance with English editing was appreciated. Lastly, we thank Brian Monteleone (WHOI) for assistance on the SIMS and help with data processing.
REFERENCES
Brey, G., 1976, CO2 solubility and solubility mechanisms in silicate melts at high pressures: Contribution to mineralogy and petrology, v. 57, i. 2, p. 215-221. Bryan, W.B., 1983, Systematics of modal phenocryst assemblages in submarine basalts: Petrologic implications: Contributions to Mineralogy and Petrology, v. 83, p. 62-74. Bucholz, C. E., Gaetani, G. A., Behn, M. D., and Shimizu, N., 2013, Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions: Earth and Planetary Science Letters, v. 374, p.145-155. doi :10.1016/j.epsl.2013.05.033. Christeson, G.L., Karson, J.A., and McIntosh, K.D., 2010, Mapping of seismic layer 2A/2B boundary above the sheeted dike unit at intermediate spreading crust exposed near the Blanco Transform: Geochemistry, Geophysics, Geosystems, v. 11, no.3. doi: 10.1029/2009GC002864. Coumans, J., Stix, J., and Minarik, W., 2015, The magmatic architecture of Taney Seamount- A, northeast Pacific Ocean: Journal of Petrology, v.56, p. 1037-1067. doi: 10.1093/petrology/egv027. Danyushevsky L.V., McNeill, A.W., and Sobolev, A.W., 2002, Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications: Chemical Geology, v.183, p.5–22. Dixon, J.E., Stolper, E.M., and Holloway, J.R., 1995, An experimental study of water and carbon dioxode solubilities in Mid-Ocean Ridge basaltic liquids. Part I: calibration and solubility models: Journal of Petrology, v.36, p.1607-1631. Embley, R.W., and Wilson, D.S., 1992, Morphology of the Blanco transform fault zone-NE Pacific: Implications for its tectonic evolution. Marine geophysical researches, v.14, i.1, p.25-45. Esposito, R., Bodnar, R. J., Danyushevsky, L. V., De Vivo, B., Fedele, L., Hunter, J., and Shimizu, N., 2011, Volatile evolution of magma associated with the Solchiaro eruption in the Phlegrean Volcanic District (Italy): Journal of Petrology, v. 52, no.12, p.2431-2460. doi.org/10.1093/petrology/egr051. Esposito, R., Hunter, J., Schiffbauer, J. D., Shimizu, and N., Bodnar, R. J., 2014, An assessment of the reliability of melt inclusions as recorders of the pre-eruptive volatile content of magmas: American Mineralogist, v. 99, no.5-6, p.976-998. doi: 10.2138/am.2014.4574. 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, v. 40, no. 10, p. 915-918. doi: 10.1130/G32992.1. 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 Science Letters, v.393, p.120-131. doi: 10.1016/j.epsl.2014.02.031. Helo, C., Longpre, M.A., Shimizu, N., Clague, D. A., and Stix, J., 2011, Explosive eruptions at mid-ocean ridges driven by CO2-rich magmas: Nature Geoscience, v.4, p.260-263. doi: 10.1038/NGEO1104. Johnson, E. A., and Rossman, G. R., 2013, The diffusion behavior of hydrogen in plagioclase feldspar at 800–1000_C: implications for re-equilibration of hydroxyl in volcanic phenocrysts: American Mineralogist, v. 98, p. 1779–1787. P a g e | 17
Kent, A.J.R., 2008, Melt inclusions in basaltic and related volcanic rocks: Reviews in Mineralogy and Geochemistry, v.69, p.273-331. doi :10.2138/rmg.2008.69.8. Kohut, E.J., and Nielsen, R.L., 2003, Low-pressure phase equilibria of anhydrous anorthite bearing mafic magmas: Geochemistry, Geophysical Geosystems, v.4 i.7. doi:10.1029/2002GC000451. Kohut, E.J., and Nielsen, R. L., 2004, Melt inclusion formation mechanisms and compositional effects in high-An feldspar and high-Fo olivine in anhydrous mafic silicate liquids: Contributions to Mineralogy and Petrology, v. 147, no.6, p.684-704. Koleszar, A. M., Saal, A. E., Hauri, E. H., Nagle, A. N., Liang, Y., and Kurz, M. D., 2009, The volatile contents of the Galapagos plume; evidence for H2O and F open system behavior in MIs: Earth and Planetary Science Letters, v. 28, i.3, p.442-452. doi: 10.1016/j.epsl.2009.08.029. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2012, Diverse Sr isotope signatures preserved in mid-oceanic-ride basalt plagioclase: Geology, v.41, p.279-282. doi:10.1130/G33739.1. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2013, The petrogenesis of plagioclase-phyric basalts at mid-ocean ridges: Geochemistry, Geophysysical, Geosystems, v.14, p.3282-3296. doi: 10.1002/ggge.20207. Laubier, M., Gale, A., and Langmuir, C. H., 2012, Melting and crustal processes at the FAMOUS segment (Mid-Atlantic Ridge): New insights from olivine-hosted melt inclusions from multiple samples: Journal of Petrology, v. 53, no.4 p. 665-698. doi:10.1093/petrology/egr075. Le Voyer, M., Kelley, K. A., Cottrell, E., and Hauri, E. H., 2017, Heterogeneity in mantle carbon content from CO2-undersaturated basalts: Nature communications, v. 8. doi: 10.1038/ncomms14062. Maclennan, J., 2017, Bubble formation and decrepitation control the CO2 content of olivine‐ hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.18, no.2, p.597- 616. doi:10.1002/2016GC006633. McNutt, M., 1979, Compensation of oceanic tomography: an application of the response function technique to the Surveyor area, Journal of geophysical research, v. 84, no. 13, p.7589-7598. Métrich, N., and Wallace, P.J., 2008, Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions: Reviews in Mineralogy and Geochemistry, Mineralogical Society of America, v.169, p.363-402. doi: 10.2138/rmg.2008.69.10. Michael, P.J., and Graham, D.W., 2015, The behavior and concentration of CO2 in the suboceanic mantle: inferences from undegassed ocean ridge and ocean island basalts. Lithos, v. 236-237, p. 338-351. doi: 10.1016/j.lithos.2015.08.020. Moore, G., 2008, Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments and modeling: Review in Mineralogy and Geochemistry, v.69, p.333–362. doi:10.2138/rmg.2008.69.9. Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. Neave, D. A., Maclennan, J., Hartley, M. E., Edmonds, M., & Thordarson, T., 2014, Crystal storage and transfer in basaltic systems: the Skuggafjöll eruption, Iceland: Journal of Petrology, v. 55, i.12, p.2311-2346. doi: 10.1093/petrology/egu058. Neave, D. A., Maclennan, J., Thordarson, T., and Hartley, M. E., 2015, The evolution and storage of primitive melts in the Eastern Volcanic Zone of Iceland: the 10 ka P a g e | 18
Grímsvötn tephra series (i.e. the Saksunarvatn ash). Contribution to Mineralogy and Petrology, v. 170, p. 1-23. doi:10.1007/s00410-015-1170-3. 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 Iceland, Geochimica et Cosmochimica Acta, v.205, p.100–118. doi: 10.1016/j.gca.2017.02.009. Newman, S., and Lowenstern, J.B., 2002, VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic Excel: Computers in Geosciences, v.2, p.597-604. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine- hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. Sobolev, A. V., and Shimizu, N., 1993, Ultra-depleted primary melt included in an olivine from the Mid-Atlantic Ridge: Nature, v.363, p.151-154. doi:10.1038/363151a0. Sours-Page, R.E., Nielsen, R.L., and Batiza, R., 2002, Parental magma diversity on a fast spreading ridge: Evidence from olivine and plagioclase-hosted melt inclusions in axial and seamount lavas from the northern East Pacific Rise: Chemical Geology, v.183, p.237-262. Sprtel, Frank. M., 1997, Two styles of oceanic near-ridge volcanism for the Southeast Indian Ocean and the NE Pacific Ocean. [Masters. thesis]: Corvallis, Oregon State University, 53 p. Steele-MacInnis, M., Esposito, R., and Bodnar, R.J., 2011, Thermodynamic model for the effect of post-entrapment crystallization on the H2O-CO2 systematics of volatile- saturated silicate melt inclusions: Journal of Petrology, v. 52, p.2461-2482. doi: 10.1093/petrology/egr052. Wallace, P.J., and Carmichael, I.S.E.,1994, Sulfur speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts: American Mineralogist, v.79, p.161-167. Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. Wanless, V.D., and Shaw, A.S, 2012, Lower crustal crystallization and melt evolution at midocean ridges: Nature Geoscience, v.5, p.651-655. doi: 10.1038/NGEO1552. Wanless, V. D., and Behn, M. D., 2017, Spreading rate‐dependent variations in crystallization along the global mid‐ocean ridge system: Geochemistry, Geophysics, Geosystems, v. 18, no. 8, p.3016-3033. doi:10.1002/2017GC006924.
P a g e | 19
FIGURES AND TABLES
30 min 4 Days
H2O wt.% (glass) 0.02 ± 0.02 0.01 ± 0.006
CO2 total (ppm) 2253 ± 800 2714 ± 1183 S glass (ppm) 821 ± 100 671 ± 179 Cl glass (ppm) 36 ± 12 55 ± 63 F glass (ppm) 20 ± 15 0.73 ± 0.28 P (bars) 4089 ± 1184 4732 ± 1518 Depths (km) 12 ± 3 14 ± 5
% CO2 in bubble 71 ± 14 83 ± 10 Number of analysis 27 16
Table 2.1: Averages with 1σ errors (standard deviation) of the volatiles analyzed for both experimental run times within the glass of the MI by SIMS: H2O in wt. %; S, Cl, and F in ppm. Total CO2 (ppm) represents the calculated CO2 in the MI, combining the CO2 concentrations of the glass and vapor bubbles, together with volume estimates for both. The averaged pressures and depths calculated using VolatileCalc2.0 (Newman and Lowenstern, 2002) are also shown, as well as the percentage of CO2 present into the vapor bubbles of the MI, and the number of MI analyzed for both run times.
P a g e | 20
Figure 2.1: Kernel density plot of the pressures in kbars recorded by the MI heated for 30 minutes (black) and 4 days (grey). Kernel density plots are probability density functions where the distribution parameters are unspecified. The pressures were calculated using VolatileCalc2.0 using the measured H2O concentration from SIMS and the total reconstructed CO2 for each MI (Newman and Lowenstern, 2002; Moore et al., 2015). The errors in 2σ were computed with the kernel density plots. The range of MOHO pressures seen in grey was established from the literature (McNutt, 1979; Christeson et al., 2010). The number of MI plotted (n=x) is noted in the legend. P a g e | 21
Figure 2.2: Volatiles analyzed within the glass of the MIs by SIMS. The opened circles are individual MI homogenized for 30 minutes ; the solid black circles are the individual MI homogenized for 4 days. The errors bars are 1σ errors calculated for each SIMS analysis. All graphs are plotted against CO2 (ppm) in the glass of the MIs on the y-axis. a) H2O (wt. %) in the glass of MIs ; b) S (ppm) in the glass of MIs; c) Cl (ppm) in the glass of MIs; d) F (ppm) in the glass of MIs. The errors are consistent with those determined previously during a rigorous assessment of the reliability of volatile analyses of MIs using the same analytical methods and instrumentation (Esposito et al., 2014).
P a g e | 22
CO2 density in the vapor bubbles (g/cc) 0 0.1 0.2 0.3 0
500 30 minutes 1000 4 days
1500
in the glass (ppm) glass the in 2
CO 2000
2500
Figure 2.3: CO2 in the glass (ppm) analyzed by SIMS within the glass of the MI vs. CO2 density in the vapor bubbles (g/cc) analyzed by Raman within the vapor bubbles for both experimental run times, 30 minutes (open circles) and 4 days (black circles). The errors are displayed in 2σ.
P a g e | 23
Chapter Three
Re-equilibration Processes occurring in Plagioclase-Hosted Melt Inclusions from Plagioclase Ultraphyric Basalts
Mélissa J. Drignon Roger L. Nielsen Frank J. Tepley III
Robert J. Bodnar
This manuscript is in Review Geochemistry, Geophysics, Geosystems MS# 2018GC007795 2000 Florida Avenue N.W. Washington, DC 20009-1227 August 2018
P a g e | 24
ABSTRACT
Plagioclase-hosted melt inclusions are an underutilized indicator of magmatic processes owing to the perception that they are less robust than olivine-hosted inclusions. However, based on a set of time series experiments, we demonstrate that plagioclase can preserve a primitive magmatic signal if steps are taken to correct for post entrapment crystallization. Diffusion within plagioclase-hosted melt inclusions is sufficiently fast that the melt inclusions can be homogenized within 30 minutes through heating experiments. As heating time increases, the composition of melt inclusions changes. We attribute this longer-term phenomenon to diffusive re-equilibration driven by crystal relaxation, and associated change in pressure and partitioning behavior. The rate of re-equilibration is therefore limited by the rate of diffusion within the solid. Measured at the host/melt inclusion interface, the partition coefficient for Mg decreases by 25% from the 30-minute to the 4-day experiments. Diffusive re-equilibration is also limited by the rate of crystal relaxation. Finally, the change in melt inclusion composition from the 30- minute to 4-day experiments is distinct from what would be produced by progressive melting of the plagioclase host.
INTRODUCTION
Statement of the problem
The physical and chemical characteristics of mid-ocean ridge basalts (MORBs) have long served as a source of information on the processes responsible for mantle melt generation and crustal formation at divergent plate boundaries. Glassy lavas from mid-ocean ridge environments are the products of a complex set of mixing and differentiation processes. These processes filter and blend the array of primitive mantle melts resulting in both homogenization of the primary magmatic signal, and preferential sampling of specific components from this array. If we rely exclusively on glass analyses, these factors introduce large uncertainties into the interpretation of the roles of mantle, lower-crustal, and magma-reservoir processes. In recent years, phenocrysts and their melt inclusions (MI) have become an increasingly important source of information on the “pre-processed” magmas (Sobolev and Shimizu, 1993; Nielsen et al., 1995; Danyushevsky et al., 2002; Kent, 2008; and many others). From MI we have gained a more nuanced understanding of the range of magmas produced in the mantle, and of the processes that result in the generation of the oceanic crust (Kent, 2008). The study of MI P a g e | 25 presumes that their chemistry is characteristic of magma compositions at a time when the host- crystal formed, and prior to late stage processes such as mixing and fractionation (Roedder, 1979; Hauri et al. 2002; Wallace 2005; Métrich and Wallace, 2008; Kent 2008). Once trapped, interpretation of MI chemistry requires that they remain a closed system with respect to major and trace elements to provide reliable information on the composition of the melt at the time of entrapment. However, the composition of MI is often modified by post entrapment crystallization (PEC), diffusive re-equilibration, and the formation of vapor bubbles (Figure 3.1). These processes occur during magma transport and storage, and after eruption (Roedder, 1979; Lowenstern, 1995; Danyushevsky et al., 2000, 2002; Toplis, 2005; Steele-MacInnis et al., 2011; Schiavi et al., 2016; Maclennan, 2017). To date, most studies focus on olivine-hosted MI, based on the premise that olivine is the earliest crystallizing phase and therefore is most likely to trap primitive melts (Bryan, 1983).
However, based on the CO2 concentrations of the olivine-hosted MI, most were formed at pressures equivalent to the mid to upper crust and have arguably seen extensive modification (Wanless et al., 2014; Wallace et al., 2015; Le Voyer et al. 2017). We note, however, that most of the pressures reported from olivine-hosted MI are based only on CO2 concentrations within the glass of MI. Indeed, past studies specifically analyzed MI without vapor bubbles in MI assemblages containing both MI with and without vapor bubbles, thus constituting a sample biais. Recently, the community came to the realization that up to 90% of the CO2 can be trapped within vapor bubbles of MI, implying that entrapment pressures calculated from CO2-H2O concentrations have been largely under-estimated (Steele-MacInnis et al., 2011; Bucholz et al., 2013; Hartley et al., 2014; Moore et al., 2015; Aster et al., 2016; Esposito et al., 2016). Comparably less work has been done on plagioclase-hosted MI (Helo et al., 2011; Coumans et al., 2015; Neave et al., 2015, 2017; Chapter 2) for three main reasons. First, plagioclase is assumed to be a late crystallizing phase. Secondly, cleavage planes in plagioclase are assumed to make it more susceptible to leakage. Lastly, plagioclase is assumed to be susceptible to diffusive re-equilibration (Cottrell et al., 2002) because many elements are compatible (e.g. Sr, Eu2+) or only moderately incompatible (e.g. Ba, LREE, Mg, Ti) in plagioclase due to their higher partition coefficients compared to olivine. We focus on plagioclase-hosted MI from plagioclase ultraphyric basalts (PUBs), which are found worldwide at ultraslow- to intermediate-spreading ridges. This class of MORB is characterized by >15 vol.% plagioclase (Lange et al., 2013), with individual crystals sometimes >1 cm in diameter. Previous work on these crystals (Nielsen et al., 1995, 2017; Sours-Page et al., 2002; Nielsen, 2011; Adams et al., 2011; Lange et al., 2013) has documented their high P a g e | 26 anorthitic character as well as the diversity of their trace element and isotopic composition. In addition, new information based on the CO2 concentration of plagioclase-hosted MI suggest many of these megacrysts crystallize in the upper mantle, as deep or deeper than olivine phenocrysts found in MORBs (Chapter 2). Therefore, the chemistry of such crystals may provide new information on the parental melts responsible for the generation of the oceanic crust. However, due to questions regarding the reliability of plagioclase-hosted MI, we must first assess whether these plagioclase crystals preserve the compositional characteristics of the magmas from which they crystallized. Towards that end, we examine the characteristics of plagioclase-hosted MI using a set of time-series heating experiments. Our goal is to measure time-dependent changes in the composition of plagioclase-hosted MI and host plagioclase crystals and evaluate these changes with respect to various processes, including diffusive re- equilibration, crystal relaxation, diffusive volatile loss, MI breaching, and partial melting.
Background material
The PUB sample analyzed for this study comes from the 360 km-long Blanco Transform Fault (BTF) that links the Gorda Ridge and the Juan de Fuca Ridge in the Pacific Ocean off the coast of Oregon. The BTF is a relatively young feature that formed within the last 5 Ma due to change in plate motion (Embley et al., 1985, 1987; Embley and Wilson, 1992; Cordier et al., 2007). This study focuses on one specific sample, A91-1R, sampled at 44°N 12’55”, 129°W 37’93”. A91-R1 was sampled in the < 1 Ma East Blanco Depression. We used this single sample for three reasons: (1) the plagioclase crystals are remarkably homogeneous with respect to major elements (An83.5+2, Supplementary Table B4); (2) Lange et al. (2013) demonstrated that the plagioclase hosts in this sample are chemically more primitive than those in equilibrium with the host lava (based on trace element and isotopic data) but may be derived from the same parental melt(s). The petrogenesis of the plagioclase megacrysts was ascribed to the disaggregation of troctolites residing in the lower crust or the upper mantle (Lange et al., 2013); (3) PUBs were chosen because the crystals were volumetrically important, ≥ 15 vol.% of the rock, and each host plagioclase contains numerous MI. In addition, although they are more primitive than typical MORB glasses, they are part of the same array of liquids from which MORB originate (Lange et al., 2013). Plagioclase-hosted MI are more susceptible to PEC than olivine-hosted MI because the amount of plagioclase that precipitates per degree of cooling is about two times greater than that of olivine (Osborn and Tait, 1952). In addition, plagioclase crystals are often transported in more evolved host melts, and at as much as 100 oC lower temperatures than the temperature P a g e | 27 at which they crystallized (Lange et al., 2013). This means that plagioclase-hosted MI likely undergo significant modification before eruption. It is therefore necessary to homogenize plagioclase-hosted MI to extract information on their chemistry at the time of entrapment (Figure 3.1).
MATERIALS AND METHOD
Heating experiments
Plagioclase crystals (0.3 cm to 1 cm in diameter) were separated from sample A91-R1 by crushing, and then were hand-picked under a binocular microscope. Only fresh and clear crystals were selected. The crystals were then heated using a 1-atmosphere vertical furnace at Oregon State University (OSU). Time-series heating experiments of 30 minutes, 4 hours, and 4 days were conducted at 1230°C. A heating temperature of 1230°C was selected because it represents the temperature at which the MI were trapped, as determined from a series of experiments at incremental temperatures using the technique of Sinton et al. (1993). Selecting the correct heating temperature allows the dissolution of PEC back into the MI melts such that the resulting glass phase after quenching represents the composition of the melt that was originally trapped in the MI. Heating the MI only to the temperature at which all PEC products are returned to the melt, without overheating them, allows us to avoid modifications of glass (melt) composition by adding excess host phase to the melt. In this case, the heating temperature can also be referred to as a homogenization temperature. In our case, the 1230°C heating temperature was later confirmed by performing heating-stage experiments on a Vernadsky disk at the Laboratoire Magmas et Volcans (Clermont-Ferrand, France) where PEC homogenization consistently occurred between 1225 and 1240°C (Chapter 4).
The heating experiments were performed by placing 3 to 6 individual crystals in a platinum boat is suspended from a platinum wire, such that the sample was in the “hot spot” of the furnace. Unlike olivine-hosted MI, tests carried out on plagioclase crystals showed that their low Fe contents prevented the crystals from oxidizing in the air at magmatic temperatures and turning opaque. Indeed, heating experiments performed at 1200°C with both a controlled oxygen fugacity set at FMQ+1, and within the air revealed no composition dependence of the MI as a function of oxygen fugacity (Supplementary Table B1; Supplementary Figures B1; B2). Temperature during heating was monitored using a thermocouple suspended in the furnace at the same height as the samples. At the end of each experiment, a current was run through the P a g e | 28 platinum suspension wires, melting it and dropping the samples into water. We chose to perform experiments using a vertical 1-atmosphere furnace rather than a heating-stage for two main reasons: (1) the quenching time is faster and more efficient with a 1-atmosphere vertical furnace, which aids in preventing PEC; and (2) experimentation using a heating-stage tends to weaken the host crystals as it requires double polishing to hold the crystals flat during the experiments.
Analytical methodology
Following PEC homogenization, plagioclase-hosts were mounted in 2.54 cm diameter epoxy plugs for analyses at OSU using a CAMECA SX100 electron microprobe. Plagioclase analyses were performed at 15 kV accelerating voltage, 30 nA sample current and a 5 µm beam.
The plagioclase procedure analyzed SiO2, TiO2, Al2O3, FeO*, MgO, CaO, Na2O, and K2O and was tested for external reproducibility on our labradorite standard run as an unknown (reference USNM 115900). The MI were analyzed at 15 kV accelerating voltage, 10 nA sample current and a 10 µm beam. The MI procedure contained the following elements: SiO2, TiO2, Al2O3,
FeO*, MnO, MgO, CaO, Na2O, K2O, P2O5, S, and Cl. The alkali elements were analyzed first in the elemental sequence and were analyzed using a 0-time intercept to accommodate for the volatility of Na and K under the electron beam. This procedure was tested for external reproducibility on our basaltic glass standard run as an unknown (USNM 113498/1). For both the plagioclase-hosts and the MI analyses, results were selected with totals between 98.5 wt. % and 101.5 wt.%. Errors were calculated based on the reproducibility of our two standards (Supplementary Tables B2; B3; B4 and B5).
RESULTS
Sulfur as an indicator of MI integrity
Two characteristics of the S concentration of MORBs make its composition a useful tool in evaluating the integrity of MI. First, essentially all MORB magmas are S-saturated (Wallace and Carmichael, 1994) and fall on a Fe-S saturation line defined by analyses of our PUB glasses (Figure 3.2). Second, owing to the volatility of S, any breached MI would lose some of its S during the heating process by leakage through cracks (Nielsen et al., 1998). Based on the observation that most of the MI we analyzed lie on the sulfide saturation line, we assume that they did not leak and lose S during the heating experiments. Breached MI lie below the saturation line, indicative of S loss through leakage. We consider an MI to be compromised P a g e | 29
(has lost some of its S) if the S concentration lies >2 below the S saturation line, which is the case for ~24 % of the MI (Figure 3.2; Supplementary Tables B2; B3). Small differences in S concentrations exist between the samples heated for different run times. MI held at 1230°C for 30 minutes have 896.0 ± 155.8 ppm S in average, and 7.9 ± 0.6 wt. % FeO* (total iron). MI heated for 4 hours have S concentration averaging 890.4 ± 152.4 ppm, and FeO 8.0 ± 0.4 wt.%. The 4-day runs average of 939.8 ± 115.7 ppm for S and 8.3 ± 1.1 wt. % for FeO* (Table 3.1, Supplementary Table B2). Unpaired t-tests on the data were performed to estimate the statistical significance of the results on the different run-times. An unpaired t-test compares two datasets based on their means and standard errors. If the two datasets are unpaired, or statistically significant, the resulting P values will be close to 0; if they are paired, or statistically similar, the P value will be close to 1. T-tests performed on the data reveal that the 30-mimute and 4-day runs are statistically different in terms of both S concentration and FeO content, whereas only FeO contents are statistically different between the 30-minute and 4-hour runs (Supplementary Table B6)
Effect of run time on MI composition
In addition to the relationship observed between S concentrations and experimental run times, other elements also show time-dependent compositional changes. Two notable changes are shown by MgO and Al2O3, with MgO increasing and Al2O3 decreasing as a function of increasing run times (Figure 3.3).
These changes in MgO and Al2O3 become statistically significant after 4 hours (Supplementary Table B6). The 30-minute experiments plot on the plagioclase-side of the cotectic, whereas the data from the 4-day experiments plot on the olivine-side of the cotectic. The MI heated for 4 hours show intermediate compositions to the 30-minute and 4-day runs, and plot on either side of the cotectic (Table 3.1). MI held at 1230°C for 30 minutes averages a
MgO/Al2O3 ratio of 0.48 ± 0.04, the 4-hour run averages a MgO/Al2O3 ratio of 0.52 ± 0.06, and the 4-day runs averages a MgO/Al2O3 ratio of 0.56 ± 0.04 (Table 3.1). As MI compositions vary as a function of run time, one can question whether compositional changes can also be observed in the host plagioclase. The data show that although there is a clear decrease of the Na2O/CaO within the MI with increasing run time, no change is observed in the host plagioclase (Figure 3.4, Supplementary Table B6).
The Na2O/CaO in the host plagioclase remains constant with an average of 0.1 ± 0.01 (Table
3.1). The Na2O/CaO within the MI is 0.22 ± 0.01 after 30 minutes, 0.20 ± 0.04 after 4 hours, and 0.16 ± 0.02 after 4 days. Both Na2O and CaO contents decreased with increasing run-times. P a g e | 30
Na2O drops from 2.7 ± 0.3 wt. % after 30 minutes to 1.8 ± 0.2 wt. % after 4 days heating experiments. Conversely, CaO decreases from 12.2 ± 0.5 wt. % to 10.8 ± 0.6 wt. % between these two experimental run times. T-test reveals that although the Na2O/CaO decrease is statistically significant between the 30-minute and 4-hour runs, only the Na2O contents are significantly different in between these two run-times. The CaO contents are only statistically significant in between the 30-minute and 4-day runs (Table 3.1, Supplementary Table B6). If we compare the experimental liquid compositions from the MI in this study with MORB glasses and 1-atmosphere experimental glass compositions (Figure 3.5), we see that the 30- minute runs are closer to “typical” MORB glasses than are the longer 4-day runs. After 4 days
MI have lower Na2O + K2O 1.9 ± 0.2 wt. %, but slightly higher SiO2 52.2 ± 1.0 wt. % than MI homogenized for 30 minutes (Table 3.1). Additionally, the MI exhibit different SiO2 and alkaline contents than LEPR glasses. LEPR, or library of experimental phase relations, is a database containing published results on experimental studies applied to natural magmatic systems (Hirschmann et al., 2008). The LEPR glasses average 2.2 ± 1.3 wt. % Na2O + K2O at
50.4 ± 2.2 wt. % SiO2 (Table 3.1). MI heated for both 30 minutes and 4 days tend to have higher SiO2 contents than MORB glasses. Although the alkaline content of the MORB glasses is spread over a large range, MI homogenized for 4 days tend to have Na2O + K2O contents that are more like the experimental glasses from LEPR, compared to the MI homogenized for 30 minutes, which have an overall slightly higher alkalinity content (Figure 3.5). T-tests performed on both SiO2 and Na2O+K2O contents revealed that the results on the two run-times are statistically significant (Supplementary Table B6).
DISCUSSION
Plagioclase-hosted MI: evidence for early magmatic processes
The composition of PUB glasses is similar to aphyric MORB glasses (Lange et al., 2013), with MgO contents ranging from 4 to 10 wt. % (N-MORB to E-MORB range). The plagioclase hosts are not in equilibrium with those glasses, exhibiting more primitive, high An-contents
(An80-92), but are nonetheless arguably part of the global array of MORBs (Lange et al., 2013). Conversely, the MI are in equilibrium with their plagioclase hosts (Lange et al., 2013), and have
Mg# close to primary melts with an average of ≈ Mg#68 (Chapter 2). The MI and host PUB glass compositions should be part of the same, or at least similar liquid lines of descent. Indeed, if the lavas erupted at a ridge segment are genetically related to one another the host glasses should represent the evolved end-members of a parental melt from which the MI are related. P a g e | 31
Previous work has documented the primitive character of plagioclase-hosted MI in PUBs, which we have ascribed to be the product of crystallization in the lower crust or the upper mantle (Lange et al., 2012, 2013; Chapter 2). A central premise of our research is that the density contrast between plagioclase and basaltic melt significantly increases the possibility of their being sampled from the lower crust/upper mantle (Campbell et al., 1978). The most current model for the petrogenesis of PUBs is that they form from typical MORB melts that entrain crystals from troctolitic cumulates in the lower crust/upper mantle (Lange et al., 2013). Volatile analyses of MI in our sample confirm the upper mantle origin of most of the plagioclase crystals, with minimum entrapment pressures estimated at 2.3-9.1 kbars (Chapter 2). Troctolitic cumulates, by definition, are made up of both olivine and plagioclase crystals. The plagioclase crystals are preferentially sampled due to their lower and similar density to MORB melts at 5 kbars (~2.65 g/cm3) and consequently their greater buoyancy compared to olivine (~ 3.24 g/cm3). The ascent velocity required for the PUB melts to overcome the settling velocity of the host plagioclase within the cumulate is ~ 1 cm/s (Lange et al., 2013). Those characteristics also explain why PUBs are found worldwide at ultraslow to intermediate spreading centers, only off-axis at fast-spreading centers, and at fracture zones, because the presence of a magma lens would decrease that ascent velocity and filter the plagioclase crystals out (Lange et al., 2013). Therefore, somewhat counter-intuitively, such crystals offer a greater probability for sampling the products of lower crustal to upper mantle processes compared to olivine-hosted MI.
Crystal relaxation as a leading cause for chemistry changes: comparison between olivine and plagioclase
Diffusive re-equilibration involving Fe-loss and hydrogen loss in olivine-hosted MI are of the most commonly recognized and cited post entrapment processes related to MI (Danyushevsky et al., 2002). Fe-loss occurs when the MI and their host olivines re-equilibrate during cooling, over-heating, and/or during experimental runs exceeding 20-30 minutes. In the case of an experiment lasting longer than 20-30 minutes, magmatic temperatures induce an internal pressure drop within the MI which increases diffusive exchange with the host olivines. This results in an enrichment of the MI in Mg# due to the migration of FeO from the MI into their host olivine (Danyushevsky et al., 2002). To correct for this modification, researchers have developed several models that back-calculate the original MI chemistry by adding olivine into the MI composition until the MI and its host crystal are at equilibrium (Danyushevsky et al., 2000, 2002, 2004; Toplis 2005). P a g e | 32
Crystal relaxation is another common, yet less studied post entrapment process. We refer to crystal relaxation as the increase in volume that a crystal experiences when it is exposed to magmatic temperatures and lower pressures than its formation pressure. In the case of our experiments, the sample is held at magmatic temperatures during laboratory heating, but at an atmospheric pressure that is significantly lower than the confining pressure at the time of trapping. The same process may occur during ascent in natural systems. The importance of crystal relaxation was revealed by Schiavi et al. (2016) who conducted experiments on olivine- hosted MI to alternatively grow and dissolve a vapor bubble in multiple heating cycles using a heating stage. They observed that with repeated attempts to equilibrate the MI, the temperatures at which the vapor bubbles grew and disappeared both increased with time. Through oxidation- decoration experiments, Schiavi et al. (2016) noted the appearance of dislocation features at the MI/host olivine interfaces. Oxidation-decoration experiments work well on olivine crystals as they are based on oxidizing the Fe present in the crystals to ‘decorate’ features such as dislocations. Schiavi et al. (2016) proposed that these dislocation features appeared due to crystal relaxation. The olivine-hosts could not accommodate the pressure difference between their own crystalline structures and the slightly different elasticity behavior of their MI. The different cycles of expansion and compression induced by the high temperatures at which the experiments were performed caused the volume of the cavity representing the MI to increase, resulting in an internal pressure drop within the MI. At these lower pressures the solubility of volatiles was lower, and volatiles contained in the vapor bubble could therefore not be re- dissolved into the melt.
The decrease in Al2O3, CaO, Na2O and K2O together with the increase of SiO2 and MgO with increasing experimental run times suggest that plagioclase melting is not resposible for these compositional drifts (Figures 3.3-3.5). We propose that these compositional drifts are the result of re-equilibration processes that are enhanced by relaxation of the plagioclase structure over the 4-day runs. The expansion caused by heating induces a pressure drop within the plagioclase-hosts, thus modifying the partitioning behavior of the elements. This argument has also been suggested by Danyushevsky et al. (2002) who observed the same composition drifts in olivine-hosted MI as reported here for plagioclase-hosted MI. Crystal relaxation is the focus of the following chapter of this dissertation. Although more details will be given in Chapter 4, it is important to note a few differences between olivine and plagioclase. First, olivine’s strength, i.e. its resistance to both plastic and brittle deformation, is >10x that of plagioclase (Meade and Jeanloz, 1990; Rybacki and Dresen, 2000). As a result, plagioclase crystals are more likely to britely deformed through the formation (Zhang, 1998). P a g e | 33
However, it as also been demonstrated that plagioclase have lower thermal expansivities than olivine, restricting their volumetric expansion while relaxing (Bouhifd et al., 1996; Tribaudino et al., 2010). These differences have consequences on how crystal relaxation affects plagioclase-hosted MI (see Chapter 4).
Crystal relaxation in plagioclase-hosted MI: phase relation evidence
If changes in the phase equilibria and partitioning driven by crystal relaxation is the leading cause of the observed compositional changes between the 30-minute and 4-day runs, it implies that the internal pressure of the MI drops over the course of 4 days while the samples are held at 1230°C. The MI remains pressurized because volatiles were still dissolved within their glasses (Chapter 2). Nevertheless, the relaxation of the plagioclase-hosts likely led to a pressure decrease within the MI. Consequently, due to the pressure decrease, the phase relations applicable to MI heated for 4 days would be different than those applicable to the 30-minute runs. To test this concept, we examined the compositional changes as a function of time within the context of our understanding of the changes in phase equilibria and partitioning behavior as a function of pressure and temperature. High pressure, piston cylinder experiments performed on lunar basalts illustrate the changing phase relations of plagioclase with pressure. Nekvasil et al. (2015) found that at higher pressures, the slope of the lunar plagioclase solidus and liquidus flattens, which results in high anorthitic plagioclase in equilibrium with a more albitic melt, compared to typical 1-atmosphere plagioclase phase relations (Figure 3.6). Further, at ~1 GPa a pseudoazeotrope is present and divides the phase relations into two separate phase fields. The albitic side extends from An0 to
An84 and is roughly equivalent to the 1-atmosphere field whereby the plagioclase is more calcic than the liquid with which it is in equilibrium. However, at 1 GPa the anorthitic loop extends from An84 to An100 and the plagioclase is less calcic than the liquid. At 2 GPa, the anorthitic field expands to lower An- contents (Figure 3.6). High-pressure experiments were performed recently on MORBs (Ustunisik et al., 2017), and demonstrated similar phase relations to those reported for lunar basalts (Nekvasil et al., 2015). These experiments also showed that at 5 kbars the presence of Al-rich spinel buffers the phase relations (Ustunisik et al., 2017). Based on recent pressure estimates for plagioclase- hosted MI from the BTF (Chapter 2; Figure 3.6), we believe that our samples (30-minute runs) are best represented by the 5 kbars anorthite phase field. The plagioclase crystals are An-rich
(An83.5±2), and small spinel crystals can sometimes be found as round inclusions within the MI, or at the glass/plagioclase interface (Lange et al., 2013). If this hypothesis is true, then the 4- P a g e | 34 day runs represent lower pressure phase relations due to the pressure decrease induced by crystal relaxation. The rate at which crystal relaxation occurs is a topic of a separate paper
(Chapter 4.). However, Chapter 2 documents that the CO2 concentrations of the MI exsolved progressively from the melt into the vapor bubbles within the 4-day runs compared to the 30- min runs. If the CO2 concentration of each MI is close to equilibrium, then based on H2O-CO2 concentrations analyzed by SIMS within the glass of the MI, the pressure decrease is then equivalent to ~1 kbar over 4 days, (Chapter 2, Figure 3.6).
Mg diffusion at the MI/host interface
The effect of pressure on the composition of MI was tested by measuring the content of MgO from the core of the MI to the host plagioclase along EMP traverses (Figure 3.7a). We chose to focus on MgO because it is a major element in basaltic glass, and a trace element in plagioclase that is sufficiently concentrated so the EMP can detect small variations (Nielsen et al., 2017). In addition, we recognize that we lack information in the third dimension, which could have a significant effect on the shape and magnitude of the profile (e.g., Shea et al., 2015). Nevertheless, multiple profiles were collected. The 30-minute runs showed no decrease in MgO within the plagioclase host as one approached the MI, however, the 4-day runs showed a decrease in MgO as one approached the MI (Figure 3.7b, Supplementary Figure B8). We found that for the 4-day runs, MgO shows a change in composition from the MI/host interface to the MI interior that we interpret as diffusion profile. The increase of MgO seen within the MI heated for 4 days is the consequence of the scavange of that element from the plagioclase. Mass balance calculations reveal that the zone impacted within the plagioclase at the MI/plagioclase interface increases with the MI radius. Our calculation demonstrate that it is possible to scavange 0.5 to 1.0 wt. % MgO from a 22-27 µm plagioclase-radius at the MI/plagioclase interface for a 10-µm MI-radius (Figure 3.7c; Supplementary Table B9) The absence of profiles within the 30-minute experiments does not mean that diffusive re-equilibration does not occur, only that we cannot detect them using EMP (Supplementary Figure B3). Indeed, diffusion between MI and their hosts occurs once all the PEC is dissolved back within the MI, which are then solely left with a homogeneous melt phase and one or several vapor bubbles. We argue nevertheless that these diffusive exchanges are only significant once crystal relaxation occurs and enhance the process, which does not occur within 30 minutes of heating experiments.
We calculated Mg partition coefficients (DMg) for both 30-minute and 4-day runs for (1) MI/host interfaces and (2) between MI and their host plagioclase with MgO analyzed within P a g e | 35
the hosts 30-50 microns of any MI. We found that DMg calculated for MgO analyzed within the host plagioclase away from the interfaces showed statistically similar results for both experimental run times, with a range going from 0.020 to 0.027 over the anorthite of our samples (Supplementary Tables B7; B8; B9). These values are similar to the high-pressure data and most of the MI pairs published on MORBs for the same range of An-content by Nielsen et al. (2017). Additionally, our results show a weak anti-correlation between DMg and the An- contents of the plagioclase-hosts (Figure 3.8; Van Orman et al., 2014). Bindeman et al. (1998) and Costa et al. (2003) demonstrated the dependence of An-content on the activity of Mg. Most elements in plagioclase have partition coefficients that are anticorrelated with An, and Mg is therefore no exception (Blundy and Wood, 1991; Bédard, 2006).
However, at the MI/host interface for the 4-day runs, the DMg exhibit values from 0.014 to 0.020, which we attribute to a ~25% reduction in the partition coefficient from our 30-minute runs, and 4-day runs away from the MI/host interfaces. We interpret this reduction in partition coefficient to be the result of a drop in DMg as a function of pressure. The difference between
DMg measured at the interface and DMg measured using the host composition is statistically significant according to the t-test (Supplementary Tables B7; B8; B9; B10). In that case, the anti-correlation with An-content is nonexistent. The An-content of the plagioclase does not statistically change even when estimated at the MI/host interface within the 4-day runs
(Supplementary Table B10). This 25% decrease in DMg cannot be attributed to the addition of a stoichiometric plagioclase component into the MI after 4 days as we observed an increase of
SiO2 together with a decrease in Na2O, CaO, and Al2O3 (Figures 3.3-3.5). Instead, we attribute this decrease to a change in P-T as a function of time. Theoretically one could test this hypothesis by examining the experimentally determined Ds as a function of temperature. Unfortunately, the precision of those experiments is insufficient to make such an evaluation
(Nielsen et al., 2017). We do not know the pressure dependence of DMg well enough to evaluate the change as a function of pressure. However, we do know that CO2 partitioning between the vapor bubbles and the glasses of the MI increase with increasing run times, suggesting a pressure dependence of the MI compositions with experimental run times (Chapter 2). Diffusion may be a function of the crystallographic orientation. However, it has been shown that Mg diffusion varies negligibly between the b and c axes of plagioclase (Van Orman et al., 2014). In addition, we did not see any significant differences in the magnitude or width of the profiles in our experiments. Consequently, the orientation of the crystal at which the EMP analyses were performed is not relevant, and the profiles are interpreted to be representative of P a g e | 36 diffusion within the host plagioclase. As a result, we attribute the observed MgO diffusion within our 4-day runs to crystal relaxation.
CONCLUSIONS
We have demonstrated that plagioclase-hosted MI are reliable for estimating the range of compositions of primary magmas, as long as one understands the processes active during entrapment and homogenization experiments. Approximately 76% of plagioclase hosted MI in this study retain their S during heating experiments, with the remaining 24% having lost significant S as a result of leakage. Understanding how heating experiments affect the chemistry of MI is key to our ability to assess the chemistry of the plagioclase megacryst parental magmas. From our experiments, the MI heated for 30 minutes have compositions matching high-pressure experiments on lunar basalts and MORBs (Nekvasil et al. 2015; Ustunisik et al., 2017). However, after 4 days of heating, the composition of MI is substantially modified, which can be attributed to crystal relaxation. As the plagioclase-hosts expand during longer experimental runs at high temperature, their pressure drops to compensate for volume increase, leading to major element re-equilibration within the MI. CO2 and H2O analyses performed on MI homogenized for 30 minutes and 4 days confirmed the high-pressure origins of the MI, and the internal pressure drop hypothesized to be the cause of the compositional drift within the 4-day runs (Chapter 2). Finally, DMg unaffected by crystal relaxation show an anticorrelation with the plagioclase-host An-contents. However, after 4 days of heating, this anticorrelation disappears, and the DMg dropped by 25% as a result of MgO migration from the plagioclase-host into the MI. Overall, this study revealed the importance of understanding post entrapment processes such as crystal relaxation and diffusive re-equilibration. Equally important, our results document that the conditions of homogenization experiments may be a moving target, influenced by relaxation and diffusive re-equilibration.
ACKNOWLEDGMENT
This material is based in part on work supported by the National Science Foundation under grant numbers grants OCE-1634206 and OCE-1634211. Discussion with Adam Kent and David Graham (OSU) were highly beneficial to the preparation of this manuscript. Thank you to Mark Biller for substantially improving the quality of this manuscript. All data published with this study are available for download together with this publication. P a g e | 37
REFERENCES
Adams, D.T, Nielsen R.L, Kent., A.J., and Tepley, F.J., 2011, Origin of minor and trace element compositional diversity in MIs: Evidence from anorthitic feldspar phenocrysts from the Juan de Fuca Ridge: Geochemistry, Geophysical Geosystems, v.12, no.12. doi:10.1029/2011GC003778. Aster, E. M., Wallace, P. J., Moore, L. R., Watkins, J., Gazel, E., and Bodnar, R. J., 2016, Reconstructing CO2 concentrations in basaltic melt inclusions using Raman analysis of vapor bubbles: Journal of Volcanology and Geothermal Research, v.323, p.148-162. doi :10.1016/j.jvolgeores.2016.04.028. Bartels, K. S., Kinzler, R.J., and T. L Grove., 1991, High pressure phase relations of a near primary high alumina basalt from Medicine Lake Highland, N. California: Contribution to Mineralogy and Petrology, v.108, p.253- 270. Bédard, J.H., 2006, Trace element partitioning in plagioclase feldspar: Geochemica et Cosmocimica Acta, v.70, p.3717-3742. doi: 10.1016/j.gca.2006.05.003. Bindeman, I. N., Davis, A. M., and Drake, M. J., 1998, Ion microprobe study of plagioclase- basalt partition experiments at natural concentration levels of trace elements: Geochimica et Cosmochimica Acta, v.62, no.7, p.1175-1193. Blundy, J. D., and Wood, B. J., 1991, Crystal-chemical controls on the partitioning of Sr and Ba between plagioclase feldspar, silicate melts, and hydrothermal solutions: Geochimica et Cosmochimica Acta, v.55, no.1, p.193-209. Bouhifd, M.A., Andrault, D., Fiquet, G., and Richet, P., 1996, Thermal expansion of forsterite up to the melting point: Geophysical Research Letters, v. 23, no. 10, p. 1143-1146. Bowen, N.L.,1922, The reaction principle in petrogenesis: The Journal of Geology, v.30, no.3, p.177-198. Bryan, W.B., 1983, Systematics of modal phenocryst assemblages in submarine basalts: Petrologic implications: Contributions to Mineralogy and Petrology, v. 83, p. 62-74. Bucholz, C. E., Gaetani, G. A., Behn, M. D., and Shimizu, N., 2013, Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions: Earth and Planetary Science Letters, v. 374, p.145-155. doi :10.1016/j.epsl.2013.05.033. Campbell, I.H., Roeder, P.L., and Dixon, J.M.., 1978, Plagioclase buoyancy in basaltic liquids as determined with a centrifuge furnace: Contributions to mineralogy and petrology, v.67, p.369-377. Cordier, C., Caroff, M., Juteau, T., Fleutelot, C., Hemond, C., Drouin, M., Cotten, J., and Bollinger, C., 2007, Bulk-rock geochemistry and plagioclase zoning in lavas exposed along the northern flank of the Western Blanco Depression, Northeast Pacific: Insight into open-system magma chamber processes: Lithos, v.99, p. 289-311. doi :10.1016/j.lithos.2007.06.009. Costa, F., Chakraborty, S., and Dohmen, R., 2003, Diffusion coupling between trace and major elements and a model for calculation of magma residence times using plagioclase: Geochimica et Cosmochimica Acta, v. 67, no.12, p.2189-2200. doi:10.1016/S0016- 7037(02)01345-5 . Cottrell, E., M. Spiegelman., and Langmuir, C.H., 2002, Consequences of diffusive reequilibration for the interpretation of melt inclusions : Geochemistry, Geophysics, Geosystems v.3, no.5, p.1-26. doi: 10.1029/2001GC000205. Coumans, J., Stix, J., and Minarik, W., 2015, The magmatic architecture of Taney Seamount- A, northeast Pacific Ocean: Journal of Petrology, v.56, p. 1037-1067. doi: 10.1093/petrology/egv027. P a g e | 38
Danyushevsky, L.V., Della-Pasqua, F.N., and Sokolov, S., 2000, Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications: Contributions to Mineralogy and Petrology, v.138, p.68-83. Danyushevsky L.V., McNeill, A.W., and Sobolev, A.W., 2002, Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications: Chemical Geology, v.183, p.5–22. Danyushevsky, L. V., Leslie, R. A., Crawford, A. J., and Durance, P., 2004, Melt inclusions in primitive olivine phenocrysts: the role of localized reaction processes in the origin of anomalous compositions: Journal of Petrology, v.45, no.12, p.2531-2553.doi: 10.1093/petrology/egh080. Embley, R. W., 1985, A locally formed deep ocean canyon system along the Blanco Transform, Northeast Pacific: Geo-marine letters, v.5, no.2, p.99-104. Embley, R. W., Kulm, L. D., Massoth, G., Abbot, D., and Holmes, M., 1987, Morphology, structure, and resource potential of the Blanco Transform Fault Zone: Chapter Twenty- Three, p.550-561. Embley, R.W., and Wilson, D.S., 1992, Morphology of the Blanco transform fault zone-NE Pacific : Implications for its tectonic evolution. Marine geophysical researches, v.14, i.1, p.25-45. Esposito, R., Lamadrid, H.M., Redi, D., Steele-MacInnis, M., Bodnar, R.J., Manning, C., De
Vivo, B., Cannatelli, C., and Lima, A., 2016, Detection of liquid H2O in vapor bubbles in reheated melt inclusions : Implications for magmatic fluid composition and volatile budgets of magmas : American Mineralogist, v.101, p.1691-1695. doi: 10.2138/am- 2016-5689. Grove, T. L., and Bryan, W. B., 1983, Fractionation of pyroxene-phyric MORB at low pressure: an experimental study: Contributions to Mineralogy and Petrology, v.84, no.4, p.293- 309. 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 Science Letters, v.393, p.120-131. doi: 10.1016/j.epsl.2014.02.031. Hauri, E.H., 2002, SIMS investigations of volatiles in volcanic glasses, 2: abundances and isotopes in Hawaiian melt inclusions: Chemical Geology, v.183, p.115–141. Helo, C., Longpre, M.A., Shimizu, N., Clague, D. A., and Stix, J., 2011, Explosive eruptions at mid-ocean ridges driven by CO2-rich magmas: Nature Geoscience, v.4, p.260-263. doi: 10.1038/NGEO1104. Hirschmann, M. M., Ghiorso, M. S., Davis, F. A., Gordon, S. M., Mukherjee, S., Grove, T. L., ... and Till, C. B., 2008, Library of Experimental Phase Relations (LEPR): A database and Web portal for experimental magmatic phase equilibria data : Geochemistry, Geophysics, Geosystems, v.9, no.3. doi : 10.1029/2007GC001894. Kent, A.J.R., 2008, Melt inclusions in basaltic and related volcanic rocks: Reviews in Mineralogy and Geochemistry, v.69, p.273-331. doi :10.2138/rmg.2008.69.8. Kinzler, R. J., and Grove, T.L., 1992, Primary magmas of mid-ocean ridge basalts, 2: Applications: Journal of Geophysical Research, v.97, no. B5, p.6885-6906. Kohut, E.J., and Nielsen, R.L., 2003, Low-pressure phase equilibria of anhydrous anorthite bearing mafic magmas: Geochemistry, Geophysical Geosystems, v.4 i.7. doi:10.1029/2002GC000451. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2012, Diverse Sr isotope signatures preserved in mid-oceanic-ride basalt plagioclase: Geology, v.41, p.279-282. doi:10.1130/G33739.1. P a g e | 39
Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2013, The petrogenesis of plagioclase-phyric basalts at mid-ocean ridges: Geochemistry, Geophysysical, Geosystems, v.14, p.3282-3296. doi: 10.1002/ggge.20207. Le Voyer, M., Kelley, K. A., Cottrell, E., and Hauri, E. H., 2017, Heterogeneity in mantle carbon content from CO2-undersaturated basalts: Nature communications, v. 8. doi: 10.1038/ncomms14062. Lowenstern, J.B., 1995, Applications of silicate melt inclusions to the study of magmatic volatiles, in Thompson, J.F.H., eds., Magmas, Fluids and Ore Deposits: Mineralogical Association of Canada, Short Course 23, p. 71-99. Maclennan, J., 2017, Bubble formation and decrepitation control the CO2 content of olivine‐ hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.18, no.2, p.597-616. doi:10.1002/2016GC006633. Meade, C., Jeanloz, R., 1990, The strength of mantle silicates at high pressures and room temperature: implications for the viscosity of the mantle: Nature, v.348, p. 533-535. Métrich, N., and Wallace, P.J., 2008, Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions: Reviews in Mineralogy and Geochemistry, Mineralogical Society of America, v.169, p.363-402. doi: 10.2138/rmg.2008.69.10. Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. Neave, D. A., Maclennan, J., Thordarson, T., and Hartley, M. E., 2015, The evolution and storage of primitive melts in the Eastern Volcanic Zone of Iceland: the 10 ka Grímsvötn tephra series (i.e. the Saksunarvatn ash). Contribution to Mineralogy and Petrology, v. 170, p. 1-23. doi:10.1007/s00410-015-1170-3. 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 Iceland, Geochimica et Cosmochimica Acta, v.205, p.100–118. doi: 10.1016/j.gca.2017.02.009. Nekvasil, H., Lindsley, D.H., DiFrancesco, N., Catalano, T., Coraor, A.E., and Charlier, B., 2015, Uncommon behavior of plagioclase and the ancient lunar crust: Geophysical Reasearch Letters, v.42, p.10,573-10,579. doi :10.1002/2015GL066726. Nielsen, R.L., Crum, J., Bourgeois, R., Hascall, K., Forsythe, L.M., Fisk, M.R., and Christie, D.M., 1995, Melt inclusions in high-An plagioclase from the Gorda Ridge: an example of the local diversity of MORB parent magmas: Contributions to Mineralogy and Petrology, v.122, p.34-50. Nielsen, R. L., Michael, P. J., and Sours-Page, R.,1998, Chemical and physical indicators of compromised melt inclusions: Geochimica et Cosmochimica Acta, v.62, no.5, p.831- 839. doi :10.1016/S0016-7037(98)00024-6. Nielsen, Roger L., 2011, The effects of re‐homogenization on plagioclase hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.12, no.10. doi:10.1029/2011GC003822. Nielsen, R. L., Ustunisik, G., Weinsteiger, A. B., Tepley, F. J. III., Johnston, A. D., and Kent, A. J. R., 2017, Trace element partitioning between plagioclase and melt: An investigation of the impact of experimental and analytical procedures: Geochemistry Geophysics Geosystems, v.18, doi:10.1002/2017GC007080. Osborn, E. F., and Tait, D. B., 1952, The system diopside-forsterite-anorthite: American Journal of Science, v.250, p.413-433. Roedder, E.,1979, Origin and significance of magmatic inclusions: Bulletin de Mineralogie, v.102, p.467-510. P a g e | 40
Rybacki, E., and Dresen, G., 2000. Dislocation and diffusion creep of synthetic anorthite aggregates: Journal of Geophysical Research, v.105, p.26017–26036. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine-hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. Shea, T., Costa, F., Krimer, D., and Hammer, J. E., 2015, Accuracy of timescales retrieved from diffusion modeling in olivine: A 3D perspective. American Mineralogist, v.100, no.10, p.2026-2042. doi :10.20138/am-2015-5163. Sinton, C. W., Christie, D. M., Coombs, V. L., Nielsen, R. L., and Fisk, M. R., 1993, Near- primary melt inclusions in anorthite phenocrysts from the Galapagos Platform: Earth and Planetary Science Letters, v.119, no. 4, p.527-537. Sobolev, A. V., and Shimizu, N., 1993, Ultra-depleted primary melt included in an olivine from the Mid-Atlantic Ridge: Nature, v.363, p.151-154. doi:10.1038/363151a0. Sours-Page, R.E., Nielsen, R.L., and Batiza, R., 2002, Parental magma diversity on a fast spreading ridge: Evidence from olivine and plagioclase-hosted melt inclusions in axial and seamount lavas from the northern East Pacific Rise: Chemical Geology, v.183, p.237-262. Steele-MacInnis, M., Esposito, R., and Bodnar, R.J., 2011, Thermodynamic model for the effect of post-entrapment crystallization on the H2O-CO2 systematics of volatile-saturated silicate melt inclusions: Journal of Petrology, v. 52, p.2461-2482. doi: 10.1093/petrology/egr052. Thy, P., Lesher, C.E., and Fram, M.S., 1998, Low pressure experimental constraints on the evolution of basaltic lavas from site 917, southeast Greenland continental margin: Proc.Ocean Drill. Program Sci. Results, v.152, p.359–372. Toplis, M. J., 2005, The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems: Contributions to Mineralogy and Petrology, v.149, no.1, p.22-39. doi: 10.1007/s00410-004-0629-4. Tormey, D. R., Grove T.L., and Bryan, W. B., 1987, Experimental petrology of normal MORB near the Kane Fracture Zone: 22o-25ON, mid- Atlantic ridge: Contribution to Mineralogy and Petrology, v.96, p.121-139. Tribaudino, M., Angel, R.J., Cámara, F., Nestola, F., Pasqual, D., Margiolaki, I., 2010, Thermal expansion of plagioclase feldspars: Contribution to Mineralogy and Petrology, v. 160, p.8999-908. doi:10.1007/s00410-010-0513-3. Ustunisik, G., Nielsen, R.L., and Walker, D., 2017, Phase equilibria of plagioclase ultraphyric basalts: constraints from high-pressure experiments. Paper presented at Geological Society of America Annual meeting, Seattle, WA. Van Orman, J. A., Cherniak, D. J., and Kita, N. T., 2014, Magnesium diffusion in plagioclase: dependence on composition, and implications for thermal resetting of the 26Al–26Mg early solar system chronometer: Earth and Planetary Science Letters, v.385, p.79-88. doi :10.1016/j.epsl.2013.10.026. Wallace, P.J., and Carmichael, I.S.E.,1994, Sulfur speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts: American Mineralogist, v.79, p.161-167. 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, v.140n no.1-3, p.217-240. doi :10.1016/j.jvolgeores.2004.07.023. Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. P a g e | 41
Wanless, V.D., Behn, M.D., Shaw, A.M., and Plank, T., 2014, Variations in melting dynamics and mantle compositions along the Eastern Volcanic Zone of the Gakkel Ridge: insights from olivine-hosted MIs: Contribution to Mineralogy and Petrology, v.167. doi: 10.1007/s00410-014-1005-7. Yang, H. J., Kinzler, R. J., and Grove, T. L., 1996, Experiments and models of anhydrous, basaltic olivine-plagioclase-augite saturated melts from 0.001 to 10 kbar: Contributions to Mineralogy and Petrology, v.124, no.1, p.1-18. Zhang, Y., 1998, Mechanical and phase equilibria in inclusion-host systems: Earth and Planetary Science Letters, v.157, p.209-222.
P a g e | 42
FIGURES AND TABLE
Figure 3.1: a) Back scattered electron image of a non-homogenized MI (light grey) within its host plagioclase (darker grey). An albitic rim (in dark grey) developed at the contact between the MI and the host plagioclase. Within the MI, post entrapment crystallization (PEC) and a vapor bubble formed due to decrease in pressure and temperature; a) Simplified schematic of the BSE image shown in a). b) BSE image of two heated host plagioclase with their MI. During the heating experiment the PEC is dissolved back into the MI glasses. Note the presence of vapor bubbles in most MI, and the large vapor bubble in the upper crystal.
P a g e | 43
1200
1000
800 Sulfide Saturation Line 600 S (ppm) S Uncompromised 400 Compromised 200
0 7 8 9 10 11 FeO* (total wt.%)
Figure 3.2: S (ppm, µg/g) vs. FeO* (total, wt.%) showing the proportion of MI that retained their integrity and those that breached during PEC homogenization. The un-compromised MI are plotted with an open circle. The black diamonds represent for all MI that breached during PEC homogenization. The blue line is defined by MORBs glass analyzed by Wallace et al. (2005) and represents the sulfide saturation line. Any MI that is not within 2σ of the average S concentration is considered breached. The FeO* notation means total iron measured on the EMP, which include both FeO and Fe2O3. The errors displayed are 2σ and based on the reproducibility of our standards (Supplementary Tables B2; B3; B5).
30 minutes 4 hours 4 days S (ppm) 896.0 ± 155.8 890.4 ± 152.4 939.8 ± 115.7 FeO* (wt%) 7.9 ± 0.6 8.0 ± 0.4 8.3 ± 1.1
Na2O (wt%) 2.7 ± 0.3 2.4 ± 0.6 1.8 ± 0.2 CaO (wt%) 12.2 ± 0.5 12.1 ± 0.4 10.8 ± 0.6 MgO (wt%) 8.1 ± 0.5 8.5 ± 0.8 9.0 ± 0.6
Al2O3 (wt% 17.0 ± 0.6 16.6 ± 0.4 16.1 ± 0.6
Na2O+K2O (wt%) 2.9 ± 0.3 2.7 ± 0.6 1.9 ± 0.2
SiO2 (wt%) 51.0 ± 1.1 51.3 ± 0.8 52.2 ± 1.0
Na2O/CaO 0.22 ± 0.01 0.20 ± 0.04 0.16 ± 0.02
MgO/Al2O3 0.48 ± 0.04 0.52 ± 0.06 0.56 ± 0.04 Number of analyses 86 23 17
Table 3.1: Table showing the differences in compositions observed between MI heated for 30 minutes, 4 hours, and 4 days. The numbers represent the average for each element/ratio and their standard deviations (2σ). P a g e | 44
Figure 3.3: a) Plot of Al2O3 (wt.%) vs. MgO (wt.%) showing MI heated for 30 minutes with open circles, MI heated for 4 hours (grey diamonds) and MI heated for 4 days with black diamonds. The cotectic is shown as a solid grey line and was plotted from MORB glasses analyses from Sours-Page et al., 2002. b) Plot of MgO/Al2O3 vs. heating time in minutes showing the chemical changes occurring in the MI while run times increase. All MI displayed on this graph are un-breached. The errors displayed are 2σ and based on the reproducibility of our standards (Supplementary Tables B2; B5). P a g e | 45
Figure 3.4: Na2O/CaO vs. run times (min) showing the re-equilibration processes occurring with increasing run times. The MI are displayed with opened circles. The host plagioclase data is displayed with black squares. Only un-breached MI are displayed. The errors displayed are 2σ and based on the reproducibility of our standards (Supplementary Tables B2; B4; B5).
3.50
3.00
2.50
O (wt.%) O LEPR_1 atm 2 2.00 A91-R Glass
O + + O K 1.50 2 30 minutes
Na 4 days 1.00
0.50 47 48 49 50 51 52 53 54
SiO2 (wt.%)
Figure 3.5: Alkaline (Na2O + K2O; wt.%) vs. SiO2 (wt.%) for MI heated for 30 min (grey circles) and 4 days (black circles), compared to MORB erupted glasses from the LEPR database (open diamonds; Grove and Bryan, 1983; Tormey et al., 1987; Bartels et al., 1991; Kinzler and Grove, 1992; Yang et al., 1996; Thy et al., 1998; Kohut and Nielsen, 2003), and A91-R1 PUB Glass (black stars; Nielsen 2011). Only un-breached MI are displayed. The errors displayed are 2σ and based on the reproducibility of our standards (Supplementary Tables B2; B5).
P a g e | 46
Figure 3.6: Schematic of the plagioclase phase relations based on high pressure experiments on lunar basalts (Nekvasil et al., 2015), and MORBs (Ustunisik et al., 2017), and 1-atmosphere experiments (Bowen, 1922). The 30-minute runs relate to the high-pressure loops of ~5 kbars, whereas the 4-day runs relate to a lower pressure one. The PEC homogenization temperature is also displayed in green.
P a g e | 47
Figure 3.7: a) BSE image showing the track of a 1-µm-step analysis from core of the MI to the host plagioclase for a 4-day run. b) profil resulting from the 1-µm-step analysis. The top profil shows the full range of MgO, the bottom profi is a zoom within the MgO range of plagioclase. The yellow box corresponds to the MI, which has a 10 µm radius; the grey box corresponds to the plagioclase. c) mass balance equation showing the hypothetical radius of plagioclase scavenged (µm) for MgO according to the radius of the MI (µm). The calculations were made for a 10000- ppm MgO increase within the MI (open circles), as well as for a 5000-ppm increase (black circles). More diffusion profiles, as well as the details of the mass balance calculation can be found in supporting information Apendix B, Supplementary Figure B3, and Supplementary Tables B8; B9.
Figure 3.8: Mg partition coefficients (DMg) calculated between MI and host plagioclase versus the An-contents of the host plagioclase. Partition coefficients were calculated for 30-minutes runs at the MI/host interface (grey squares), the 30-minute experiments within the host plagioclase, away from the MI (black square), as well as for 4-day runs at the MI/host interface (open circle), and the 4-day experiments within the host plagioclase, away from the MI (green circles). Errors are reported in 2σ (Appendix B; Supplementary Table B8).
P a g e | 48
Chapter Four
Experimental and Analytical Determination of Crystal Relaxation from Plagioclase Ultraphyric Basalts and the effects on Plagioclase-hosted Melt Inclusions
Mélissa J. Drignon Laurent Arbaret Nicolas Cluzel Roger L. Nielsen Frank J. Tepley III Robert J. Bodnar
This manuscript is in prepartion for Journal of Geophysical Research
P a g e | 49
ABSTRACT
Interpretation of melt inclusion (MI) data requires that we understand the impact of the major post entrapment processes. Although crystal relaxation has been relatively rarely acknowledged as important, new evidence suggests that it has substantial impacts on MI chemistry, mineral structure and the internal pressure of MI. Crystal relaxation impacts the volatile partitioning between the melts and vapor bubbles of MI. It also likely enhances diffusive re-equilibration between MI and their mineral-hosts. Crystal relaxation occurs when a mineral spends long periods of time at magmatic temperature post-formation. It consists in the relaxation of a mineral’s structure and relates to a volume increase. To accommodate this volume increase, the internal pressure within the mineral drops. To better understand the phenomenon of crystal relaxation and quantify the volume increase, we present results of heating-stage experiments designed to induce crystal relaxation within plagioclase-hosted MI, as well as measurements of MI and vapor bubbles volumes. Experiments revealed that plagioclase-hosted MI behave differently from olivine-hosted MI to crystal relaxation. The relaxation of the host-plagioclase induces a strong pressure gradient with the MI that remain at high internal pressure. The MI become quickly supersaturated leading to volatile exsolution into vapor bubbles through episodes of bubble nucleation past a temperature threshold that is ~40ºC above their post entrapment crystallization homogenization temperature. Both MI and vapor bubbles experience a volume increase, but vapor bubbles usually see their volume increase >100%, whereas MI only experience a 30% volume increase in average. We suggest that the volume increase of MI is limited by the low coefficients of thermal expansivity of the host-plagioclase. The volume increase of the vapor bubbles is limited by the amount of volatile dissolved originally within the MI from which the vapor bubbles can grow.
INTRODUCTION
Statement of the problem
Melt inclusions (MI) are pockets of melt trapped during crystal growth. Their chemistry and volatile contents are assumed to be closer to parental melts than volcanic glasses that have been subject to fractionation processes. Because MI evolve as quasi-closed systems after entrapment, they are more likely to retain the chemistry and volatile information of the melt(s) from which their host-mineral crystallized than magmas that have subject to open system processes (Roedder, 1979; Lowenstern, 1995; Danyushevsky et al., 2000, 2002; Toplis, 2005; P a g e | 50
Steele-MacInnis et al., 2011; Schiavi et al., 2016 ; Maclennan, 2017; Chapters 2 and 3). This makes MI powerful tools for assessing the early magmatic history of igneous rocks (Kent, 2008; Métrich and Wallace, 2008). At the time of entrapment, a MI is composed of a single homogeneous liquid phase representing melt from which its mineral-host crystallized. However, MI are subjected to post- entrapment processes that alter their chemistry and modify their volatile partitioning. Post entrapment processes occur due to changing P, T conditions, the most common being post- entrapment crystallization (PEC), which consists of daughter crystals growing inward from the walls of the MI. Both PEC and decreasing P, T lead to the formation of vapor bubbles that form from volatile exsolution (Kent, 2008; Métrich and Wallace, 2008). PEC can be reversed through heating experiments, the objective of which is to recover the original chemistry and glassy texture of MI by dissolving the daughter crystals back to the melt from which they formed. Vapor bubbles, the result of exsolution of volatile species from MI, however, can resist these heating experiments (Chapter 2). Heating experiments performed on olivine-hosted MI have demonstrated that vapor bubbles can be dissolved back into the melt of MI, however either extremely high temperatures or abrupt temperature changes are required (Wallace et al., 2015; Schiavi et al., 2016). The issues with those experiments are that they irreversibly modify the chemistry of MI, requiring post experimental correction (Danyushevsky and Plechov, 2011). Interest in studying vapor bubbles has grown with the realization that they may contain substantial amounts of the total CO2, H2O, and S that were originally dissolved in MI. As the importance of vapor bubble grew, analytical techniques were improved to allow for their quantification (Steele-MacInnis et al., 2011; Hartley et al., 2014; Moore et al., 2015; Esposito et al 2016). It was found that CO2 may partition from 40 to 90% within the vapor bubbles of MI
(Moore et al., 2015; Wallace et al., 2015). Additionally, given the strong partitioning of CO2, it is difficult to determine whether the CO2 present in vapor bubbles exclusively originates from MI exsolution. It is possible that at the time of melt inclusion entrapment, a supersaturated melt containing bubbles was trapped simultaneously by the mineral-host together with the melt inclusion. To accomodate this possibility, researchers have adopted the practice of only analyzing vapor bubbles that represent ≤5% of the MI volumes (Moore et al., 2015). This approach minimizes the possibility of overestimating CO2 abundances in MI. However, vapor bubbles may still experience a volume increase after their formation if more volatiles exsolve from the melt due to changing P-T conditions. P a g e | 51
Background information
Crystal relaxation is defined as the volume increase of a mineral subjected to abrupt or prolonged changes in pressure and/or temperature. In this view, a mineral can expand while residing at magmatic temperature for long periods of time post formation such as in cumulates, magma lenses, and in transport magmas. Conversely, while ascending within its host-melt, a mineral is subjected to a pressure decrease that favors relaxation. The study of olivine-hosted MI reveals that crystal relaxation may be a major post- entrapment process that greatly affects the structure of both MI and their host-minerals (Schiavi et al., 2016). Heating-stage experiments based on repeated cycles of both increasing and decreasing temperature demonstrated the existence of crystal relaxation based on two key observations. The first observation is that the temperature at which the MI homogenize increases with time and is accompanied by a temperature increase of the vapor bubble appearance and disappearance (Massare et al., 2002; Schiavi et al., 2016). The second observation is that dislocation features form at MI/olivine interfaces due to cycles of expansion and contraction of the mineral host. These dislocation features result from the different plastic properties of the MI and their host olivines. (Schiavi et al., 2016). They are easily revealed within olivines through oxidation-decoration experiments (Kohlstedt et al., 1976). For the case of plagioclase, crystal relaxation was suggested as the cause of decreasing internal pressure due to longer experimental run-times within plagioclase-hosted MI (Chapter 2). Our time-series experiments performed from 30 minutes to 4 days on plagioclase-hosted MI in a 1-atmosphere vertical furnace revealed compositional changes in the MI with increasing run times as well as the occurrence of MgO diffusing at the MI/host interfaces (Chapter 3). The experiments also revealed a 15% increase of CO2 partitioning between melt and vapor bubbles of MI with increasing run times (Chapter 2). If crystal relaxation is indeed responsible for these changes, then the observation and quantification of such a phenomenon would help in understanding its mechanisms. It would also help us to develop a more accurate methodology for determining pressure information from CO2-H2O studies of MI in any mineral by removing the bias from only analyzing MI with small vapor bubbles. We present a study based on in-situ observation and volumetric quantification of crystal relaxation. Our goal is to both demonstrate the existence of crystal relaxation in plagioclase and understand its implications for CO2 partitioning and compositional drifts with increasing run times. Our attempt to image crystal relaxation through dislocation features in plagioclase is challenging given the paucity of Fe and its reluctance to oxidize in this system. Consequently, P a g e | 52 another methodology needs to be developed to prove the existence and impacts of crystal relaxation within plagioclase-hosted MI. In this regard, we performed heating-stage experiments to induce crystal relaxation and to quantify the volumetric increase of both MI and vapor bubbles through microscopic images. Our results reveal that crystal relaxation affects plagioclase differently from olivine, but nevertheless confirm that crystal relaxation is the leading cause for the observed volatile and chemical discrepancies with prolonged exposure at magmatic temperature post formation.
METHODOLOGY
Sample provenance
The sample chosen for this study is a plagioclase ultraphyric basalt (PUB) from the Blanco Transform Fault (BTF). The BTF is a 350 km right-lateral transform fault located off the coast of Oregon (USA), which separates the Juan de Fuca Plate to the north from the Pacific Plate to the south. The BTF split a mid-ocean ridge in two with the Juan de Fuca Ridge to the NE and the Gorda Ridge to the SW (Embley and Wilson, 1992). Sample A91-1R was chosen as the experimental focus, and it was retrieved at 44°N 12’55”, 129°W 37’93”2.2 at 3200 m below sea level (Sprtel 1997). The plagioclase megacrysts are 0.1-0.5 cm in diameter and contain multiple MI that are typically 25-30 µm in diameter and with a few that are >50 µm in diameter. The samples were double-polished using alumina powder to make sure they remain flat and immobile within the heating stage.
Experimental procedure
Following Schiavi et al. (2016), we placed a double-polished host plagioclase on a sapphire disk within a Vernadsky-type heating stage at the Laboratoire Magma et Volcans, (LMV, Université d’Auvergne, France) that allows for a stable and clear in situ observation of the MI during the experiments (Schiano, 2003). The samples were subjected to cycles of increasing and decreasing temperature in order to observe the appearance and disappearance of vapor bubbles. In our experiments, we hoped to re-create and observe MI complete homogenization (PEC + vapor bubbles) at elevated temperature, as well as cycles of vapor bubble disappearance and appearance with temperature variation to verify the process of crystal relaxation within plagioclase-hosted MI. We also aimed to observe and quantify the apparent volume increase of both MI and vapor bubbles. We performed a total of 9 experiments. The two first experiments (Experiments #1 and #2) were conducted in air, as plagioclase does not oxidize. However, we decided to perform the P a g e | 53 remaining seven experiments (Experiments #3-#9) under an oxygen fugacity of 10-9 atm by injecting a purified He flux into the heating stage to have better control over the temperature. A small amount of gold metal with a known melting temperature of 1064 °C was placed next to the crystal as a reference to monitor the temperature. In our experiments, the gold systematically melted at 1038 ºC, so 26 ºC was added to all the temperatures listed in this paper.
The temperature itself was monitored with a Pt-Pt90Rh10 thermocouple with an estimated error <5 °C. During the experiment, the temperature was increased at about 100 °C/min close to that of the MI entrapment temperature. Once at the entrapment temperature, we waited until all the PEC were dissolved back into the melt, and then, for all experiments but one (Experiment #4), the temperature was increased at about 10°C/min to observe vapor bubble behavior. Pictures were taken throughout the experimental runs to document our observations. From these pictures, we estimated long and short axes of both MI and vapor bubbles using the ‘Line Segment’ measuring tool and Adobe Illustrator©. These measurements were used to calculate volumes assuming the MI to be oblate and the vapor bubbles to be spherical. Each axis measurement had a 0.5 µm-error associated that was then propagated to the volume calculations (Moore et al., 2015).
RESULTS
Bubble nucleation episodes during heating experiments
From the first experiment onward, we observed that plagioclase-hosted MI behaved differently than MI in olivine in the sense that the vapor bubbles did not disappear, leading to the necessity of adjusting the experimental procedure. In-situ microscopic observations consistently showed the following: (1) at the initiation of the experiment, the MI are fully crystallized with PEC; (2) around 1200°C, the first daughter crystals dissolve back into the melt of the MI; (3) between 1225°C and 1240°C PEC homogenization is completed, and one or two vapor bubbles remain. The vapor bubbles tend to shrink a little with the disappearance of the last daughter crystals, but never completely redissolve into the MI; (4) ~40°C above the PEC homogenization temperature, the MI experience a sudden episode of bubble nucleation. The episodes of bubble nucleation occur from the walls of the MI inwards. The newly formed bubbles quickly migrate toward the original vapor bubble and coalesce with it, forming a single larger vapor bubble. These vapor bubbles do not nucleate from random places on the MI walls. The vapor bubbles all nucleate from only a few spots and coalesce linearly to the bigger vapor bubble, usually located at the center of the MI (Supplemental Files C1 and C2; P a g e | 54
Figure 4.1). This phenomenon occurs in a matter of seconds, most of the newly formed bubbles coalescing quickly within the original vapor bubble. The few remaining bubbles can take up to 2 hours to coalesce (Supplemental Files C1 and C2). Episodes of bubble nucleation can be repeated many times during each temperature increment without breaching the MI until a threshold temperature is reached beyond which nothing happens up to the point that the crystal melts. The MI that breach, are located near cleavage planes, and/or near the plagioclase surface. (Supplementary Files C1 and C2; Figure 4.1). Each experiment aside from Experiment #4 followed the same procedure. Experiments #1 and #2 served as observational experiments and no pictures were taken. Pictures for volume quantification were taken for Experiments #3, and #5-#9. For Experiment #4 we let the plagioclase sit for 5 hours at 1256ºC, the temperature at which a few MI experienced an episode of bubble nucleation, to see if we could observe vapor bubble disappearance with time. No additional episode of bubble nucleation nor bubble disappearance occurred.
Temperature of bubble nucleation episodes
The temperature at which the MI were homogenized for PEC varied between 1225 and 1240ºC, which agrees with previous estimates made from the anorthite content of our crystals (Sinton et al., 1993; Nielsen, 2011). All MI homogenized for PEC at 1232 ± 5 ºC (1σ). The first episode of bubble nucleation occurs ~ 40ºC above the PEC homogenization temperature, with an average of 1270 ± 16 ºC (1σ; Table 4.1). The largest MI tend to experience their first episode of bubble nucleation first, but that phenomenon is not systematic (Figure 4.2). The temperature of last nucleation recorded is in average 1335 ± 32 ºC (1σ; Table 4.1), which is ~100 ºC above the PEC homogenization temperature. The final temperature at which the experiment was quenched for MI 7_1 and 7_2 corresponds to the maximum temperature at which an episode of bubble nucleation was observed. A further increase of temperature did not induce any more episode of bubble nucleation. However, for the other samples, the final temperature at which the experiments were quenched correspond to the temperature at which the plagioclase-hosts started to melt.
2-D quantification of volumes
After experimentation, it was determined that both MI and vapor bubbles experienced volume increases. MI volumes increased on average by 32%, with a range from 6% to 88%. Vapor bubble volumes increased on average by 561% and ranged from 62 to 1451% (Table 4.2). We only considered vapor bubbles that initially represented ≤ 5% of the initial MI volumes P a g e | 55 to assume that these vapor bubbles formed only from volatile exsolution from MI and not from co-trapping of over saturated melt (Moore et al., 2015). There is no correlation between MI and vapor bubble volume increases, suggesting that there are different mechanisms explaining the MI volumes increase on one side, and vapor bubbles volumes increase on the other side.
quantification correlations
Correlations (r2) were assessed through linear regressions calculated using Excel. A strong positive correlation with r2= 0.82 exists between the initial and final volumes of MI (Figure 4.3a). This correlation is lower with r2= 0.67 for vapor bubbles (Figure 4.3b). The last temperature of bubble nucleation seems to be roughly correlated to the vapor bubble volume increase with r2= 0. 65, but temperature and MI volume increase are not correlated (Figure 4.4). The temperature difference between the first and last episode of bubble nucleation (ΔT) correlates positively (r2= 0.61) with the difference between initial and final volume of the vapor bubbles (Figure 4.5a). This correlation exists as well between ΔT and the difference between initial and final volumes of the MI but is much weaker as r2= 0.44 (Figure 4.5b).
DISCUSSION structural differences between olivine and plagioclase Crystal relaxation operates differently in plagioclase-hosted MI compared to olivine- hosted MI. Within the plagioclase crystals, crystal relaxation operated through volume increase of both MI and vapor bubbles associated with episodes of bubble nucleation. In olivine, crystal relaxation occured through the increase of temperature for both complete homogenization and bubble re-appaearance (Schiavi et al., 2016). It is now essential to understand why olivine and plagioclase behaved so differently under the same experimental conditions. Anorthite has lower coefficients of thermal expansivity than forsterite (Figure 4.6; Berman, 1988; Bouhifd et al., 1996; Abramson et al., 1997; Gottschalk, 1997; Holland and Powell, 1998; Tribaudino et al., 2010). Additionally, plagioclase coefficients of thermal expansivity decrease with increasing anorthite content (Tribaudino et al., 2010). In anorthite, most coefficients of thermal expansivity are below 2.5x10-5. K-1, whereas coefficients of thermal expansivity of forsterite are generally above this value (Figure 4.6). Consequently, plagioclase cannot deform as much as olivine, making their structure more predisposed to brittle deformation. In addition, the strength of olivine, i.e. its ability to resist both plastic and brittle deformation, is 10x higher the one of plagioclase. At room conditions, olivine has a strength of 2.7 GPa, whereas at 2 GPa, its strength is 5.9 GPa (Meade and Jeanlov, 1990). In comparison, P a g e | 56 at 1 GPa the strength of plagioclase varies between 0.17 and 0.30 GPa (Stünitz and Tullis, 2001). We argue that the key point driving those differences reside in the fact that olivine deforms more plastically and plagioclase more brittlely.
Pressure gradient between host confining pressure and internal MI pressure According to Tait (1992), episodes of bubble nucleation are expected to occur once a host-mineral starts cracking. Cracking occurs alongside pre-existing microcracks and is due to the decrease of tensile stress of a mineral at the interface with a melt inclusion. The bigger the MI, the more likely cracking will occur within the host-mineral. Tait (1992) demonstrated that minerals with lower surface energy are more subject to cracking, such as in olivines as compared to quartz. If the cracks propagate to the MI, degassing will follow, unless quenching of the MI occurs rapidly. If the cracks do not propagate to the MI, a pressure gradient develops between mineral-host and MI. Pressure gradients develop owing to different compressibility and thermal expansivity between the mineral-host and their MI (Zhang, 1998). As a result, the
MI internal pressure lowers and becomes quickly super-saturated in CO2, which leads to an episode of bubble nucleation. We did not observe cracks within the plagioclase crystals post experimentation. However, microcracks are the predominant deformation mechanism of plagioclase under low confining pressure (<1.5 kbars) and high temperature, whereas at higher confining pressures, dislocation creep becomes the prefered deformation mechanism (Tullis and Yund, 1977). Microcracks form alongside cleavage planes and near to heterogeneities such as MI (Tullis and Yund, 1977). It is therefore likely that microcracks developed during experiments but would require transmitted electron microprobe (TEM) imaging to be revealed. In addition, we suggest that the plagioclase cleavage planes could induce a pressure gradient between the confining pressure of the plagioclase and the pressurized MI just as cracks would. If a pressure gradient can be established, the MI remains at high pressure while their hosts has relaxed. We attribute the occurrence of multiple episodes of bubble nucleation to the following
(Figure 4.7). At entrapment time, the parental melt and MI are at the same pressure (Pext =Pint). In our case, we estimate this pressure to be about 5 kbars (Chapter 2). During ascent to the seafloor and quench, the MI crystallize with PEC and one or multiple vapor bubbles appear due to decreasing internal pressure (Pq
Pressure differential to allow an episode of bubble nucleation One can estimate the amount of internal pressure decrease needed within a MI to induce an episode of bubble nucleation. This internal pressure decrease represents the pressure differential between the CO2-H2O saturation pressure existing within a MI at instant t1, and the lower internal pressure existing in the MI after a rise of temperature inducing relaxation at instant t2. In the case of a subaerial eruption, Tait (1992), estimated that while the host has equilibrated to its confining pressure of 1 bar, the MI may still be pressurized at up to 90% of their entrapment pressure. It is beyond the scope of this chapter to calculate the pressure differential needed to induce an episode of bubble nucleation. Nevertheless, two studies estimated that an episode of bubble nucleation within a basaltic melt can occur with a pressure differential of 150-300 bars in the case of a heterogeneous episode of bubble nucleation, and of 500 bars in the case of a homogeneous episode of bubble nucleation (Le Gall and Pichavant, 2016b; Shea, 2017). Bubble nucleation episodes occured from the MI wall and form from a
CO2-saturated melt (Chapter 2), suggesting a heterogeneous bubble nucleation process (Le Gall and Pichavant, 2016b). In this regard, every increase in temperature induces the crystal to relax causeing a decrease of ≤ 300 bars within the MI, which is enough to induce an episode of bubble nucleation. In this regard, it should be possible to induce episodes of bubble nucleation as long as there is still some CO2 dissolved within the melt of the MI. Bubble nucleation episodes would only stop if the MI reached the confining pressure of their hosts (i.e 1 bar).
CONCLUSIONS Crystal relaxation has substantial impacts on the composition of MI. The specific effects vary according to the nature of the mineral-host. In the case of our experiments, the increase of P a g e | 58 temperature beyond the homogenization point provokes bubble nucleation episodes that can be explained by the exsolution of further volatiles from the melt of the MI into their vapor bubbles. This exsolution is the result of a pressure gradient that develops in between the plagioclase- hosts and their MI due to the crystal’s cleavage planes as well as the existence of potential microcracks. On average, MI experience a 32 % increase of their volume, whereas vapor bubbles experience on average a volume increase > 500%. We suggest that the MI are constrained by the low coefficients of thermal expansivity of anorthitic plagioclase that limit their expansion. On the other hand, vapor bubbles can experience volume increase until there are not enough dissolved volatiles within the melt of MI to be exsolved. These observations confirm that the increase of CO2 partitioning into vapor bubbles with experimental run times, as well as the existence of diffusive re-equilibration processes are indeed the result of crystal relaxation (Chapters 2 and 3) Future directions include the addition of 3-D micro-tomography data to be compared with our 2-D volume estimates, as well as TEM work to image dislocation in the plagioclase crystals caused by crystal relaxation. 3-D micro-tomography data will give stronger support to our volume calculations and correlations for asserting the extent of crystal relxation on both the MI and their vapor bubbles. With TEM we hope to image the presence of either microcracks or evidences for plastic deformation such as dislocation creeps.
ACKNOWLEDGEMENTS This material is based in part on work supported by the National Science Foundation under grant numbers grants OCE-1634206 and OCE-1634211. We are grateful to Federica Schiavi (LMV) for constructive discussions on her experimental procedure as well as on this study. Discussions and comments from Adam Kent and David Graham (OSU) were appreciated.
REFERENCES Abramson, E. H., Brown, J. M., Slutsky, L. J., and Zaug, J., 1997, The elastic constants of San Carlos olivine to 17 GPa: Journal of Geophysical Research: Solid Earth, v.102, no.B6, p.12253-12263. Berman, R.G., 1988, Internally-consistent thermodynamic data for minerals in the system Na2O-K2O-CaO-MgO-FeO-Fe2O3- Al2O3-SiO2-TiO2-H2O-CO2: Journal of Petrology, v.29, p.445–522. Bouhifd, M.A., Andrault, D., Fiquet, G., and Richet, P., 1996, Thermal expansion of forsterite up to the melting point: Geophysical Research Letters, v. 23, no. 10, p. 1143-1146. Danyushevsky, L.V., Della-Pasqua, F.N., and Sokolov, S., 2000, Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications: Contributions to Mineralogy and Petrology, v.138, p.68-83. P a g e | 59
Danyushevsky L.V., McNeill, A.W., and Sobolev, A.W., 2002, Experimental and petrological studies of melt inclusions in phe-nocrysts from mantle-derived magmas: an overview of techniques, advantages and complications: Chemical Geology, v.183, p.5–22. Danyushevsky, L. V., and P. Plechov., 2011, Petrolog3: Integrated software for modeling crystallization processes : Geochemistry Geophysics Geosystems, v.12, Q07021, doi:10.1029/2011GC003516. Embley, R.W., and Wilson, D.S., 1992, Morphology of the Blanco transform fault zone-NE Pacific: Implications for its tectonic evolution: Marine Geophysical Researches, v.14, i.1, p.25-45. Esposito, R., Lamadrid, H.M., Redi, D., Steele-MacInnis, M., Bodnar, R.J., Manning, C., De
Vivo, B., Cannatelli, C., and Lima, A., 2016, Detection of liquid H2O in vapor bubbles in reheated melt inclusions: Implications for magmatic fluid composition and volatile budgets of magmas: American Mineralogist, v.101, p.1691-1695. doi: 10.2138/am- 2016-5689. Gottschalk, M., 1997, Internally consistent thermodynamic data for rock-forming minerals in the system SiO2-TiO2-Al2O3-CaOMgO- FeO-K2O-Na2O–H2O-CO2: European Journal of Mineralogy, v. 9, p.175–223. 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 Science Letters, v.393, p.120-131. doi: 10.1016/j.epsl.2014.02.031. Holland, T.J.B., and Powell, R., 1998, An internally consistent thermodynamic data set for phases of petrological interest: Journal of Metamorphic Geology, v.16, p.309–343. Kent, A.J.R., 2008, Melt inclusions in basaltic and related volcanic rocks. Reviews in Mineralogy and Geochemistry, v.69, p.273-331. doi:10.2138/rmg.2008.69.8. Kohlstedt D. L., Goetze C., Durham W. B., and Vander Sande J., 1976, New technique for decorating dislocations in olivine: Science, v.191, p.1045–1046. Le Gall, N., and Pichavant, M., 2016, Homogeneous bubble nucleation in H2O-and H2O-CO2- bearing basaltic melts: results of high temperature decompression experiments : Journal of Volcanology and Geothermal Research, v. 327, p.604-621. doi :10.1016/j.volgeores.2016.10.004. Lowenstern, J.B.,1995, Application of silicate-melt inclusions to the study of magmatic volatiles. In J.F.H. Thompson (Eds.): Magmas, fluids and ore deposits (pp. 71-100). Maclennan, J., 2017, Bubble formation and decrepitation control the CO2 content of olivine‐ hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.18, i.2, p.597-616. doi:10.1002/2016GC006633. Massare, D., Métrich, N., and Clocchiatti, R., 2002, Hight-temperature experiments on silicate melt inclusions in olivine at 1 atm: inference on temperatures of homogenization and H2O concentrations: Chemical geology, v.183, p.87-98. Meade, C., and Jeanloz, R., 1990, The strength of mantle silicates at high pressures and room temperature: implications for the viscosity of the mantle: Nature, v.348, no.6301, p.533-535. Métrich, N., and Wallace, P.J., 2008, Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions: Reviews in Mineralogy and Geochemistry, Mineralogical Society of America, v.169, p.363-402. doi:10.2138/rmg.2008.69.10. Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. P a g e | 60
Nielsen, Roger L., 2011, The effects of re‐homogenization on plagioclase hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.12, no.10. doi:10.1029/2011GC003822. Roedder, E.,1979, Origin and significance of magmatic inclusions: Bulletin de Mineralogie, v.102, p.467-510 Schiano, P., 2003, Primitive mantle magmas recorded as silicate melt inclusions in igneous minerals: Earth-Science Reviews, v. 63, p.121–144. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine-hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. Shea, T., 2017, Bubble nucleation in magmas: A dominantly heterogeneous process?: Journal of Volcanology and Geothermal Research, v. 343, p.155-170. doi : 10.1016/j.volgeores.2017.06.025. Sinton, C. W., Christie, D. M., Coombs, V. L., Nielsen, R. L., and Fisk, M. R., 1993, Near- primary melt inclusions in anorthite phenocrysts from the Galapagos Platform: Earth and Planetary Science Letters, v. 119, no. 4, p.527-537. Sprtel, Frank. M., 1997, Two styles of oceanic near-ridge volcanism for the Southeast Indian Ocean and the NE Pacific Ocean. [Masters. thesis]: Corvallis, Oregon State University, 53 p. Steele-MacInnis, M., Esposito, R., and Bodnar, R.J., 2011, Thermodynamic model for the effect of post-entrapment crystallization on the H2O-CO2 systematics of volatile- saturated silicate melt inclusions: Journal of Petrology, v. 52, p.2461-2482. doi: 10.1093/petrology/egr052. Stünitz, H., and Tullis, J., 2001, Weakening and strain localization produced by syn- deformational reaction of plagioclase: International Journal of Earth Sciences, v.90, n.1, p.136-148. Tait, S., 1992, Selective preservation of melt inclusions in igneous phenocrysts: American Mineralogist, v. 77, p. 146-155. Toplis, M. J., 2005, The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems: Contributions to Mineralogy and Petrology, v.149, i.1, p.22-39. doi: 10.1007/s00410-004-0629-4. Tribaudino, M., Angel, R.J., Cámara, F., Nestola, F., Pasqual, D., and Margiolaki, I., 2010, Thermal expansion of plagioclase feldspars: Contribution to Mineralogy and Petrology, v. 160, p.8999-908. doi:10.1007/s00410-010-0513-3. Tullis, J., and Yund, R.A., 1977, Experimental deformation of dry Westerly granite : Journal of geophysical research, v. 82, no.36, p. 5705-5718. Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. Zhang, Y., 1998, Mechanical and phase equilibria in inclusion-host systems: Earth and Planetary Science Letters, v.157, p.209-222. P a g e | 61
FIGURES AND TABLES
P a g e | 62
Figure 4.1: Evolution of the heating experiments that intended to induce bubbles nucleation episodes (Experiments #3, #5, #6, #7, #8, and #9). The MI that were measured are identified with a name. The first digit refers to the experiment number, the second to the order to which they were measured. No pictures from experiment #4 are shown here, as this experiment did not have the purpose to induce several bubble nucleation episodes to increase the volumes of both MI and vapor bubbles.
Sample Homogenization Temperature of First Temperature of last Name Temperature (ºC) Episode (ºC) Episode (ºC) 3_1 1232 1271 1367 3_2 1232 1271 1367 3_3 1232 1271 1367 5_1 1239 1252 1295 5_2 1239 1267 1295 5_3 1239 1267 1295 6_1 1225 1287 1322 6_2 1225 1287 1322 6_3 1225 1287 1322 7_1 1231 1288 1366 7_2 1231 1288 1366 7_3 1231 1243 1366 7_4 1231 1243 1366 8_1 1230 1264 1302 9_1 1240 1264 1301 Average 1232 ± 5 1270 ± 16 1335 ± 32 (1σ)
Table 4.1: This table summarizes the temperature (ºC) evolution of each sample. The homogenization temperature is on the second column, the temperature at which the first episode of bubble nucleation occurred is displayed on the third column. Finally, the last column shows the temperature of last bubbles nucleation. The average and standard deviation (1σ) was calculated for each column.
P a g e | 63
1300
1290
1280
1270
1260
1250
1240
Temperature (ºC) nucleation episide first of Temperature 1230 0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 3.5E-08 Initial Volume of Melt Inclusions (cm3)
Figure 4.2: Initial volume of MI (cm3) versus the temperature of first bubble nucleation episode (ºC). The errors displayed are 1σ.
Table 4.2: Summary of the volume estimates for MI and vapor bubbles. The first column has the samples names. The second and third columns display respectively the initial and final volumes of the MI in cm3, calculated from 2D measurements on Illustrator and following Moore et al. (2015). The fourth column gives volume increase of the MI in %, dividing the final volume by the initial volume for each melt inclusion. The fifth and sixth columns present the initial and final volumes of vapor bubbles, and the Last column, the volume increase of vapor bubbles in %. Errors were calculated following Moore et al. (2015) and are shown as 1σ. For both MI and vapor bubbles, the initial volume was calculated at a state prior to the first nucleation episode, and the final volumes calculated after the last nucleation episode.
P a g e | 64
5.0E-08
4.0E-08 A)
3.0E-08 R² = 0.8238 2.0E-08
1.0E-08 Final Volume MI (cm³) MI Volume Final 0.0E+00 0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 3.5E-08 Initial Volume MI (cm³)
5.0E-09 B) 4.0E-09 R² = 0.6662 3.0E-09
2.0E-09 (cm³) 1.0E-09
0.0E+00 Final Volume Vapor Bubble Bubble Vapor Volume Final 0.0E+00 2.0E-10 4.0E-10 6.0E-10 8.0E-10 1.0E-09 1.2E-09 Initial Volume Vapor Bubble (cm³)
Figure 4.3: A) Initial volume of MI (cm3) versus Final volume of MI (cm3). The correlation between the two axes is symbolized by the logarithmic dotted blue line with R2 being the correlation coefficient; B) Initial volume of vapor bubbles (cm3) versus Final volume of vapor bubbles (cm3). The correlation between the two axes is symbolized by the linear dotted blue line with R2 being the correlation coefficient. The errors displayed are 1σ.
P a g e | 65
1400 A) 1380 1360 R² = 0.652 1340 1320 1300 1280
Final Temperature (ºC) Temperature Final 0 200 400 600 800 1000 1200 1400 1600 ΔVolume Vapor Bubble (%)
1380 1360 B) 1340 R² = 0.221 1320 1300 1280 0 20 40 60 80 100
Final Temperature (ºC) Temperature Final ΔVolume MI (%)
Figure 4.4: A) Volume increase of the vapor bubbles (%) versus Temperature at which the final episode of bubble nucleation occurred (ºC). The correlation between the two axes is symbolized by the logarithmic dotted blue line with R2 being the correlation coefficient; B) Volume increase of the MI (%) versus temperature at which the final episode of bubble nucleation occurred (ºC). The correlation between the two axes is symbolized by the logarithmic dotted blue line with R2 being the correlation coefficient. The errors displayed are 1σ.
P a g e | 66
2000 A) 1500
1000 R² = 0.6088
(%) 500
0
Volume Vapor Bubbles Bubbles Vapor Volume 0 20 40 60 80 100 120 140 Δ ΔT (ºC)
100 B) 80
60 R² = 0.4402 40
Volume MI (%) MI Volume 20 Δ 0 0 20 40 60 80 100 120 140
ΔT (ºC)
Figure 4.5: A) Volume increase of the vapor bubbles (%) versus Temperature increase between the first and last episode of bubble nucleation (ºC). The correlation between the two axes is symbolized by the logarithmic dotted blue line with R2 being the correlation coefficient; B) Volume increase of the MI (%) versus Temperature increase between the first and last episode of bubble nucleation (ºC). The correlation between the two axes is symbolized by the linear dotted blue line with R2 being the correlation coefficient. The errors displayed are 1σ.
P a g e | 67
5
4 )
1 3 -
.K Plagioclase
5 -
2 Olivine
(10 α 1
0 Anorthite/Forsterite
Figure 4.6: Coefficients of thermal expansivity α (10-5. K-1) for anorthite plagioclase (open circles) and forsterite olivine (black diamonds) from different studies (Berman, 1988; Bouhifd et al., 1996; Abramson et al., 1997; Gottschalk, 1997; Holland and Powell, 1998; Tribaudino et al., 2010).
P a g e | 68
Figure 4.7: For all cartoons, the plagioclase is in grey and the MI in orange. A) After entrapment and before ascent the P, T conditions are the same than at time of entrapment. The plagioclase and MI are at the same P, T conditions. In our case the external pressure (Pext) is the same than the MI internal pressure (Pint) and equals 5000 bars. Pint is also the initial MI internal pressure (Pi). The temperature is 1230 ºC. The MI is composed of a single homogeneous phase. B) During ascent and at quench time of the seafloor, the MI crystallized PEC and one or multiple vapor bubbles nucleated. PEC tend to shrink the MI volume that is now smaller than for A. The temperature on the seafloor is about 5 ºC, which correspond to the quench temperature. Pext is the pressure at which eruption occured on the seafloor, which is 320 bars for our sample. At quench time, Pint = Pq (quench pressure) < Pi. Pq is higher than Pext but cannot be estimated. C) During 1-atmosphere homogenization experiments Pext = 1 bar and the temperature applied is the same than the entrapment temperature. The Pint = Ph (homogenization pressure) < Pq but > Pext. At this temperature, the cleavage planes of the plagioclase start convaying Pext, but both the temperature and experimental run-time are insuffiscient for relaxation to be significant. PEC dissolves back within the MI melt, which almost recovers its original volume. The homogenization experiment does not allow the dissolution of the vapor bubbles. D) About 40 ºC above the homogenization temperature, the first episode of bubble nucleation occurs. Both MI and vapor bubbles increase in volume. The original volumes are in dashed lines (black for MI and white for vapor bubbles). The MI volume does not increase as much as the vapor bubble volumes. Pint keeps decreasing and is now P1 but is still higher than Pext. The cleavage planes of the plagioclase convay Pext and induce the establishment of a pressure gradient between the confining pressure Pext and P1. E) With the last episode of bubble nucleation occuring around 1330 ºC both MI and vapor bubble increased to their maximum. The dashed lines (black for the MI and white for the vapor bubble) show the evolution of volume expansion. Pf, the final MI internal pressure is still higher than Pext but lower than any previous Pint.
P a g e | 69
Chapter Five
General Conclusions
This study documents and quantifies the degree to which plagioclase-hosted melt inclusions can be affected by the process of experimental homogenization. Volatile analyses performed on both the glass and vapor bubbles of our samples showed that even after 4 days of heating, the melt inclusions still retain their CO2, H2O and S. The main difference between the
30-minute and 4-day runs is the partitioning of CO2 between vapor bubbles and glass, with an increase of 15 % after 4 days compared to the 30-minute runs. The total CO2 content for each melt inclusion calculated based on the CO2 density and the volumes of both the vapor bubbles and melt inclusions showed an average > 4000 ppm CO2 for both run-times. In addition, melt inclusions heated for both 30 minutes and 4 days exhibit S contents that plot on the iron-sulfide saturation line. Therefore, this evidence supports the idea that minimum entrapment pressures may be accurately determined based on the H2O and total CO2 contents of the melt inclusions. These pressures were calculated using VolatileCalc 2.0 (Newman and Lowenstern, 2002) and indicate that the plagioclase megacrysts formed at between 3.5 and 6.5 kbars, and therefore within or below the MOHO. The extent of degassing that might have occurred at depth prior to ascent is beyond the scope of this investigation. Nevertheless, although this would improve our knowledge of the depths at which the plagioclase crystals crystallized, the more important observation is that the plagioclase megacrysts in PUB lavas are derived from processes active in the mantle and are at least as deep as the pressures obtained from olivine-hosted melt inclusions analyzed in MORBs. Most olivine-hosted melt inclusions analyzed in MORBs have crustal origins (Wallace et al., 2015; Wanless and Shaw, 2012; Wanless et al., 2014). Our pressure estimates from the PUB lava from the Blanco Transform Fault are consistent with the model wherein, due to their low density, primitive plagioclase crystals are more likely to be sampled from the mantle compared to primitive olivines (Kohut and Nielsen, 2003; Lange et al., 2013) and that our samples likely originate from disruption of mantle troctolites. As MORB melts transitthe cumulates, they entrain plagioclase crystals and erupt as PUBs along ridge segments where there are no axial magma lenses that might filter the plagioclase crystals out prior to eruption (Lange et al., 2013). We argue that since PUB samples have mantle affinities, they are closer representative of the parental melts responsible for the formation of the oceanic crust than olivine-hosted melt P a g e | 70 inclusions usually studied in MORB. The primitive character of PUB lavas is further supported by the high Mg# of plagioclase hosted melt inclusions (average Mg#67.5±2.5). These values are close to values required to be in equilibrium with mantle assemblages as defined by Michael and Graham (2015). Due to the pervasive post entrapment crystallization characteristic of plagioclase-hosted melt inclusions, reconstruction of the primary composition requires that they be homogenized prior to any chemical analyses. To be homogenized, the entrapment temperature of the melt inclusions needs to be estimated, and the experimental run time needs to be carefully chosen. Our one-atmosphere vertical furnace and heating stage experiments estimated the entrapment temperature to be between 1225°C and 1240°C. The results of the time-series experiments documented that heating for extended periods (e.g. 4 days) will cause the composition of the plagioclase hosted inclusions to drift. This drift can be attributed to diffusive re-equilibration of the melts, driven by changes in pressure within the inclusions. This pressure change is caused by relaxation of the plagioclase structure; a result of heating at 1 atmosphere – several kb below the pressure of formation. The rate of diffusive re-equilibrations is slower in plagioclase compared to olivines, which we attribute to a more rigid structure. As a result of relaxation and diffusive re-equilibration, both the major and trace element chemistry of melt inclusions are irreversibly affected experimental runs. The MgO and
SiO2 contents of the melt inclusions are increasing whereas the Al2O3, CaO and Na2O contents are decreasing. This is reflected in the fact that the apparent DMg calculated at the melt inclusion/host interface drop by ~25% from the 30-minute and 4-day runs, An unexpected outcome of this investigation was the importance of crystal relaxation and its effects on inclusion chemistry. Our results demonstrate that crystal relaxation is responsible for CO2 partitioning into vapor bubbles with increasing experimental run times. This aspect of the research was driven by our observations of the time dependent changes in melt inclusion composition, as well as the contemporary work of Schiavi et al., 2016 on relaxation in olivine. In the final segment of this project, we adapted the experimental procedure of Schiavi to plagioclase-hosted melt inclusions in an attempt to quantify the rate and magnitude of crystal relaxation. Our results documented that stepwise heating of inclusions at or near the entrapment temperature resulted in secondary bubble nucleation. The episodes of bubble nucleation occurred at each increment of temperature past the homogenization temperature by ~ 40°C. These results contrast with those described by Schiavi et al. (2016) wherein the bubbles within olivine-hosted melt inclusions disappeared and did not re-nucleate with subsequent heating. The episodes of bubble nucleation are expected to occur if a pressure P a g e | 71 gradient develops between the relaxed host-mineral and the still pressurized melt inclusions. This is interpreted to be due to the low external pressure the melt inclusions experience during heating at one atmosphere. In the case of plagioclase hosted inclusions, they become super- saturated which explain why volatiles exsolve from the melts into the vapor bubbles. It has also been suggested that episodes of bubble nucleation will occur if the mineral-host cracks. These episodes can be repeated as long as there are dissolved volatiles available for exsolution (Tait, 1992). In this regard, the increase of vapor bubbles volumes is limited by the amount of volatile originally dissolved in the melt. Conversely, melt inclusions are more limited in their expansion because of the low coefficients of thermal expansivity of their plagioclase-hosts (Tribaudino et al., 2010).
REFERENCES Kohut, E.J., and Nielsen, R.L., 2003, Low-pressure phase equilibria of anhydrous anorthite bearing mafic magmas: Geochemistry, Geophysical Geosystems, v.4 i.7. doi:10.1029/2002GC000451. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2013, The petrogenesis of plagioclase-phyric basalts at mid-ocean ridges: Geochemistry, Geophysysical, Geosystems, v.14, p.3282-3296. doi: 10.1002/ggge.20207. Michael, P.J., and Graham, D.W., 2015, The behavior and concentration of CO2 in the suboceanic mantle: inferences from undegassed ocean ridge and ocean island basalts. Lithos, v. 236-237, p. 338-351. doi: 10.1016/j.lithos.2015.08.020. Newman, S., and Lowenstern, J.B., 2002, VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic Excel: Computers in Geosciences, v.2, p.597-604. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine- hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. Tait, S., 1992, Selective preservation of melt inclusions in igneous phenocrysts: American Mineralogist, v. 77, p. 146-155. Tribaudino, M., Angel, R.J., Cámara, F., Nestola, F., Pasqual, D., Margiolaki, I., 2010, Thermal expansion of plagioclase feldspars: Contribution to Mineralogy and Petrology, v. 160, p.8999-908. doi:10.1007/s00410-010-0513-3. Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. Wanless, V.D., and Shaw, A.S., 2012, Lower crustal crystallization and melt evolution at midocean ridges: Nature Geoscience, v.5, p.651-655. doi:10.1038/NGEO1552. Wanless, V.D., Behn, M.D., Shaw, A.M., and Plank, T., 2014, Variations in melting dynamics and mantle compositions along the Eastern Volcanic Zone of the Gakkel Ridge: insights from olivine-hosted melt inclusions: Contribution to Mineralogy and Petrology, v.167. doi:10.1007/s00410-014-1005-7.
P a g e | 72
BIBLIOGRAPHY
Abramson, E. H., Brown, J. M., Slutsky, L. J., and Zaug, J., 1997, The elastic constants of San Carlos olivine to 17 GPa: Journal of Geophysical Research: Solid Earth, v.102, no.B6, p.12253-12263. Adams D.T, Nielsen R.L, Kent., and A.J, Tepley, F.J., 2011, Origin of minor and trace element compositional diversity in MIs: Evidence from anorthitic feldspar phenocrysts from the Juan de Fuca Ridge: Geochemistry, Geophysical Geosystems, v.12, no.12. doi:10.1029/2011GC003778. Aster, E. M., Wallace, P. J., Moore, L. R., Watkins, J., Gazel, E., and Bodnar, R. J., 2016, Reconstructing CO2 concentrations in basaltic melt inclusions using Raman analysis of vapor bubbles : Journal of Volcanology and Geothermal Research, v.323, p.148-162. doi :10.1016/j.jvolgeores.2016.04.028. Bartels, K. S., Kinzler, R.J., and T. L Grove., 1991, High pressure phase relations of a near primary high alumina basalt from Medicine Lake Highland, N. California: Contribution to Mineralogy and Petrology, v.108, p.253- 270. Bédard, J.H., 2006, Trace element partitioning in plagioclase feldspar: Geochemica et Cosmocimica Acta, v.70, p.3717-3742. doi: 10.1016/j.gca.2006.05.003. Berman, R.G., 1988, Internally-consistent thermodynamic data for minerals in the system Na2O-K2O-CaO-MgO-FeO-Fe2O3- Al2O3-SiO2-TiO2-H2O-CO2: Journal of Petrology, v.29, p.445–522. Bindeman, I. N., Davis, A. M., and Drake, M. J., 1998, Ion microprobe study of plagioclase- basalt partition experiments at natural concentration levels of trace elements : Geochimica et Cosmochimica Acta, v.62, no.7, p.1175-1193. Blundy, J. D., and Wood, B. J., 1991, Crystal-chemical controls on the partitioning of Sr and Ba between plagioclase feldspar, silicate melts, and hydrothermal solutions : Geochimica et Cosmochimica Acta, v.55, no.1, p.193-209. Bouhifd, M.A., Andrault, D., Fiquet, G., and Richet, P., 1996, Thermal expansion of forsterite up to the melting point: Geophysical Research Letters, v. 23, no. 10, p. 1143-1146. Bowen, N.L.,1922, The reaction principle in petrogenesis: The Journal of Geology, v.30, no.3, p.177-198. Brey, G., 1976, CO2 solubility and solubility mechanisms in silicate melts at high pressures: Contribution to mineralogy and petrology, v. 57, i. 2, p. 215-221. Bryan, W.B., 1983, Systematics of modal phenocryst assemblages in submarine basalts: Petrologic implications: Contributions to Mineralogy and Petrology, v. 83, p. 62-74. Bucholz, C. E., Gaetani, G. A., Behn, M. D., and Shimizu, N., 2013, Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions : Earth and Planetary Science Letters, v. 374, p.145-155. doi :10.1016/j.epsl.2013.05.033. Campbell, I.H., Roeder, P.L., and Dixon, J.M.., 1978, Plagioclase buoyancy in basaltic liquids as determined with a centrifuge furnace: Contributions to mineralogy and petrology, v.67, p.369-377. Christeson, G.L., Karson, J.A., and McIntosh, K.D., 2010, Mapping of seismic layer 2A/2B boundary above the sheeted dike unit at intermediate spreading crust exposed near the Blanco Transform: Geochemistry, Geophysics, Geosystems, v. 11, no.3. Coogan, L.A., 2014, The lower oceanic crust, in Treatise on geochemistry, second edition: Elsevier, p.497-541. doi: 10.1016/B978-0-08-095975-7.00316.8. Cordier, C., Caroff, M., Juteau, T., Fleutelot, C., Hemond, C., Drouin, M., Cotten, J., and Bollinger, C., 2007, Bulk-rock geochemistry and plagioclase zoning in lavas exposed along the northern flank of the Western Blanco Depression, Northeast Pacific: Insight P a g e | 73
into open-system magma chamber processes: Lithos, v.99, p.289-311. doi :10.1016/j.lithos.2007.06.009. Costa, F., Chakraborty, S., and Dohmen, R., 2003, Diffusion coupling between trace and major elements and a model for calculation of magma residence times using plagioclase: Geochimica et Cosmochimica Acta, v. 67, no.12, p.2189-2200. doi:10.1016/S0016- 7037(02)01345-5 . Cottrell, E., M. Spiegelman., and Langmuir, C.H., 2002, Consequences of diffusive reequilibration for the interpretation of melt inclusions: Geochemistry, Geophysics, Geosystems v.3, no.5, p.1-26. doi: 10.1029/2001GC000205. Coumans, J., Stix, J., and Minarik, W., 2015, The magmatic architecture of Taney Seamount- A, northeast Pacific Ocean: Journal of Petrology, v.56, p. 1037-1067. doi: 10.1093/petrology/egv027. Danyushevsky, L.V., Della-Pasqua, F.N., and Sokolov, S., 2000, Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications: Contributions to Mineralogy and Petrology, v.138, p.68-83. Danyushevsky L.V., McNeill, A.W., and Sobolev, A.W., 2002, Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications: Chemical Geology, v.183, p.5–22. Danyushevsky, L. V., Leslie, R. A., Crawford, A. J., and Durance, P., 2004, Melt inclusions in primitive olivine phenocrysts: the role of localized reaction processes in the origin of anomalous compositions: Journal of Petrology, v.45, no.12, p.2531-2553. doi: 10.1093/petrology/egh080. Danyushevsky, L. V., and P. Plechov., 2011, Petrolog3: Integrated software for modeling crystallization processes : Geochemistry Geophysics Geosystems, v.12, Q07021, doi:10.1029/2011GC003516. Dixon, J.E., Stolper, E.M., and Holloway, J.R., 1995, An experimental study of water and carbon dioxode solubilities in Mid-Ocean Ridge basaltic liquids. Part I: calibration and solubility models: Journal of Petrology, v.36, p.1607-1631. Embley, R. W., 1985, A locally formed deep ocean canyon system along the Blanco Transform, Northeast Pacific: Geo-marine letters, v.5, no.2, p.99-104. Embley, R. W., Kulm, L. D., Massoth, G., Abbot, D., and Holmes, M., 1987, Morphology, structure, and resource potential of the Blanco Transform Fault Zone: Chapter Twenty- Three, p.550-561. Embley, R.W., and Wilson, D.S., 1992, Morphology of the Blanco transform fault zone-NE Pacific: Implications for its tectonic evolution. Marine geophysical researches, v.14, i.1, p.25-45. Esposito, R., Bodnar, R. J., Danyushevsky, L. V., De Vivo, B., Fedele, L., Hunter, J., and Shimizu, N., 2011, Volatile evolution of magma associated with the Solchiaro eruption in the Phlegrean Volcanic District (Italy): Journal of Petrology, v. 52, no.12, p.2431- 2460. doi.org/10.1093/petrology/egr051. Esposito, R., Hunter, J., Schiffbauer, J. D., Shimizu, and N., Bodnar, R. J., 2014, An assessment of the reliability of melt inclusions as recorders of the pre-eruptive volatile content of magmas: American Mineralogist, v. 99, no.5-6, p.976-998. doi: 10.2138/am.2014.4574. Esposito, R., Lamadrid, H.M., Redi, D., Steele-MacInnis, M., Bodnar, R.J., Manning, C., De
Vivo, B., Cannatelli, C., and Lima, A., 2016, Detection of liquid H2O in vapor bubbles in reheated melt inclusions: Implications for magmatic fluid composition and volatile budgets of magmas: American Mineralogist, v.101, p.1691-1695. doi: 10.2138/am- 2016-5689. P a g e | 74
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, v. 40, no. 10, p. 915-918. doi: 10.1130/G32992.1. Ghiorso, M. S., and Gualda, G. A., 2015, An H2O–CO2 mixed fluid saturation model compatible with rhyolite-MELTS: Contributions to Mineralogy and Petrology, v.169, no.6, p.53. doi: 10.1007/s00410-015-1141-8. Gottschalk, M., 1997, Internally consistent thermodynamic data for rock-forming minerals in the system SiO2-TiO2-Al2O3-CaOMgO- FeO-K2O-Na2O–H2O-CO2: European Journal of Mineralogy, v. 9, p.175–223. Grove, T. L., and Bryan, W. B., 1983, Fractionation of pyroxene-phyric MORB at low pressure : an experimental study : Contributions to Mineralogy and Petrology, v.84, no.4, p.293- 309. 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 Science Letters, v.393, p.120-131. doi: 10.1016/j.epsl.2014.02.031. Hauri, E.H., 2002, SIMS investigations of volatiles in volcanic glasses, 2: abundances and isotopes in Hawaiian melt inclusions: Chemical Geology, v.183, p.115–141. Helo, C., Longpre, M.A., Shimizu, N., Clague, D. A., and Stix, J., 2011, Explosive eruptions at mid-ocean ridges driven by CO2-rich magmas: Nature Geoscience, v.4, p.260-263. doi: 10.1038/NGEO1104. Hirschmann, M. M., Ghiorso, M. S., Davis, F. A., Gordon, S. M., Mukherjee, S., Grove, T. L., ... and Till, C. B., 2008, Library of Experimental Phase Relations (LEPR): A database and Web portal for experimental magmatic phase equilibria data : Geochemistry, Geophysics, Geosystems, v.9, no.3. doi : 10.1029/2007GC001894. Holland, T.J.B., and Powell, R., 1998, An internally consistent thermodynamic data set for phases of petrological interest: Journal of Metamorphic Geology, v.16, p.309–343. Hugues, C.J., 1982, Igneous petrology: Developments in petrology, v.7, Elsevier, p.268. Johnson, E. A., and Rossman, G. R., 2013, The diffusion behavior of hydrogen in plagioclase feldspar at 800–1000_C: implications for re-equilibration of hydroxyl in volcanic phenocrysts: American Mineralogist, v. 98, p. 1779–1787. Kent, A.J.R., 2008, Melt inclusions in basaltic and related volcanic rocks: Reviews in Mineralogy and Geochemistry, v.69, p.273-331. doi :10.2138/rmg.2008.69.8. Kinzler, R. J., and Grove, T.L., 1992, Primary magmas of mid-ocean ridge basalts, 2: Applications: Journal of Geophysical Research, v.97, no.B5, p.6885-6906. Klein, E.M., and Langmuir, C.H., 1987, Global correlation of ocean ridge basalt chemistry with axial depth and crustal thickness: Journal of Geophysical Research, v.92, p.089–8115. Kohlstedt D. L., Goetze C., Durham W. B., and Vander Sande J., 1976, New technique for decorating dislocations in olivine: Science, v.191, p.1045–1046. Kohut, E.J., and Nielsen, R.L., 2003, Low-pressure phase equilibria of anhydrous anorthite bearing mafic magmas: Geochemistry, Geophysical Geosystems, v.4 i.7. doi:10.1029/2002GC000451. Kohut, E.J., and Nielsen, R. L., 2004, Melt inclusion formation mechanisms and compositional effects in high-An feldspar and high-Fo olivine in anhydrous mafic silicate liquids: Contributions to Mineralogy and Petrology, v. 147, no.6, p.684-704. Koleszar, A. M., Saal, A. E., Hauri, E. H., Nagle, A. N., Liang, Y., and Kurz, M. D., 2009, The volatile contents of the Galapagos plume; evidence for H2O and F open system behavior in MIs: Earth and Planetary Science Letters, v. 28, i.3, p.442-452. doi: 10.1016/j.epsl.2009.08.029. P a g e | 75
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, v.450, p.210- 222. doi: 10.1016/j.chemgeo.2016.12.034. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2012, Diverse Sr isotope signatures preserved in mid-oceanic-ride basalt plagioclase: Geology, v.41, p.279-282. doi:10.1130/G33739.1. Lange, A.E., Nielsen, R.L., Tepley, F.J. III, and Kent, A.J.R., 2013, The petrogenesis of plagioclase-phyric basalts at mid-ocean ridges: Geochemistry, Geophysysical, Geosystems, v.14, p.3282-3296. doi: 10.1002/ggge.20207. Laubier, M., Gale, A., and Langmuir, C. H., 2012, Melting and crustal processes at the FAMOUS segment (Mid-Atlantic Ridge): New insights from olivine-hosted melt inclusions from multiple samples: Journal of Petrology, v. 53, no.4 p. 665-698. doi:10.1093/petrology/egr075. Le Gall, N., and Pichavant, M., 2016, Homogeneous bubble nucleation in H2O-and H2O-CO2- bearing basaltic melts: results of high temperature decompression experiments : Journal of Volcanology and Geothermal Research, v. 327, p.604-621. doi :10.1016/j.volgeores.2016.10.004. Le Voyer, M., Kelley, K. A., Cottrell, E., and Hauri, E. H., 2017, Heterogeneity in mantle carbon content from CO2-undersaturated basalts: Nature communications, v. 8. doi: 10.1038/ncomms14062. Lowenstern, J.B., 1995, Applications of silicate melt inclusions to the study of magmatic volatiles, in Thompson, J.F.H., eds., Magmas, Fluids and Ore Deposits: Mineralogical Association of Canada, Short Course 23, p. 71-99. Maclennan, J., 2017, Bubble formation and decrepitation control the CO2 content of olivine‐ hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.18, no.2, p.597-616. doi:10.1002/2016GC006633. Massare, D., Métrich, N., and Clocchiatti, R., 2002, Hight-temperature experiments on silicate melt inclusions in olivine at 1 atm: inference on temperatures of homogenization and H2O concentrations: Chemical geology, v.183, p.87-98. McNutt, M., 1979, Compensation of oceanic tomography: an application of the response function technique to the Surveyor area, Journal of geophysical research, v. 84, no. 13, p.7589-7598. Meade, C., Jeanloz, R., 1990, The strength of mantle silicates at high pressures and room temperature: implications for the viscosity of the mantle: Nature, v.348, p. 533-535. Métrich, N., and Wallace, P.J., 2008, Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions: Reviews in Mineralogy and Geochemistry, Mineralogical Society of America, v.169, p.363-402. doi: 10.2138/rmg.2008.69.10. Michael, P.J., and Chase, R.L., 1987, The influence of primary magma composition, H2O and pressure on mid-ocean ridge basalt differentiation. Contributions to Mineralogy and Petrology, v.96, i.2, p.245-263. Michael, P.J., and Schilling, J-G., 1989, Chlorine in mid-ocean ridge magmas: Evidence for assimilation of seawater-influenced components: Geochimica et Cosmochimica Acta, v.53, p.3131–3143. Michael, P.J., and Graham, D.W., 2015, The behavior and concentration of CO2 in the suboceanic mantle: inferences from undegassed ocean ridge and ocean island basalts. Lithos, v. 236-237, p. 338-351. doi: 10.1016/j.lithos.2015.08.020. Moore, G., 2008, Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments and modeling: Review in Mineralogy and Geochemistry, v.69, p.333–362. doi:10.2138/rmg.2008.69.9. P a g e | 76
Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. Neave, D. A., Maclennan, J., Hartley, M. E., Edmonds, M., & Thordarson, T., 2014, Crystal storage and transfer in basaltic systems: the Skuggafjöll eruption, Iceland: Journal of Petrology, v. 55, i.12, p.2311-2346. doi: 10.1093/petrology/egu058. Neave, D. A., Maclennan, J., Thordarson, T., and Hartley, M. E., 2015, The evolution and storage of primitive melts in the Eastern Volcanic Zone of Iceland: the 10 ka Grímsvötn tephra series (i.e. the Saksunarvatn ash). Contribution to Mineralogy and Petrology, v. 170, p. 1-23. doi:10.1007/s00410-015-1170-3. 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 Iceland, Geochimica et Cosmochimica Acta, v.205, p.100–118. doi: 10.1016/j.gca.2017.02.009. Nekvasil, H., Lindsley, D.H., DiFrancesco, N., Catalano, T., Coraor, A.E., and Charlier, B., 2015, Uncommon behavior of plagioclase and the ancient lunar crust: Geophysical Reasearch Letters, v.42, p.10,573-10,579. doi :10.1002/2015GL066726. Newman, S., and Lowenstern, J.B., 2002, VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic Excel: Computers in Geosciences, v.2, p.597-604. Nielsen, R.L., Crum, J., Bourgeois, R., Hascall, K., Forsythe, L.M., Fisk, M.R., and Christie, D.M., 1995, Melt inclusions in high-An plagioclase from the Gorda Ridge: an example of the local diversity of MORB parent magmas: Contributions to Mineralogy and Petrology, v.122, p.34-50. Nielsen, R. L., Michael, P. J., and Sours-Page, R.,1998, Chemical and physical indicators of compromised melt inclusions: Geochimica et Cosmochimica Acta, v.62, no.5, p.831- 839. doi :10.1016/S0016-7037(98)00024-6. Nielsen, Roger L., 2011, The effects of re‐homogenization on plagioclase hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.12, no.10. doi:10.1029/2011GC003822. Nielsen, R. L., Ustunisik, G., Weinsteiger, A. B., Tepley, F. J. III., Johnston, A. D., and Kent, A. J. R., 2017, Trace element partitioning between plagioclase and melt: An investigation of the impact of experimental and analytical procedures: Geochemistry Geophysics Geosystems, v.18, doi:10.1002/2017GC007080. Osborn, E. F., and Tait, D. B., 1952, The system diopside-forsterite-anorthite: American Journal of Science, v.250, p.413-433. Papale, P., Moretti, R., and Barbato, D., 2006, The compositional dependence of the saturation surface of H2O+ CO2 fluids in silicate melts: Chemical Geology, v. 229, no. 1-3, p. 78- 95. doi :10.1016/j.chemgeo.2006.01.013. Roedder, E.,1979, Origin and significance of magmatic inclusions: Bulletin de Mineralogie, v.102, p.467-510. Rubin, K.H., and Sinton, J.M., 2007, Inferences on mid-ocean ridge thermal and magmatic structure from MORB compositions: Earth and Planetary Science Letters, v.260, p. 257– 276. Doi: 10.1016/j.espl.2007.05.035. Rybacki, E., and Dresen, G., 2000. Dislocation and diffusion creep of synthetic anorthite aggregates: Journal of Geophysical Research, v.105, p.26017–26036. Saal, A. E., Hauri, E. H., Langmuir, C. H., and Perfit, M. R., 2002, Vapour undersaturation in primitive mid-ocean-ridge basalt and the volatile content of Earth's upper mantle. Nature, v. 419, p.451-455. P a g e | 77
Schiano, P., 2003, Primitive mantle magmas recorded as silicate melt inclusions in igneous minerals: Earth-Science Reviews, v. 63, p.121–144. Schiavi, F., Provost, A., Schiano, P., and Cluzel, N., 2016, P-V-T-X evolution of olivine-hosted melt inclusions during high-temperature homogenization treatment: Geochimica et Cosmochimica Acta, v. 172, p. 1-21. doi: 10.1016/j.gca.2015.09.025. Shea, T., Costa, F., Krimer, D., and Hammer, J. E., 2015, Accuracy of timescales retrieved from diffusion modeling in olivine: A 3D perspective. American Mineralogist, v.100, no.10, p.2026-2042. doi :10.20138/am-2015-5163. Shea, T., 2017, Bubble nucleation in magmas: A dominantly heterogeneous process?: Journal of Volcanology and Geothermal Research, v. 343, p.155-170. doi : 10.1016/j.volgeores.2017.06.025. Shishkina, T.A., Botcharnikov, R.E., Holtz, F., Almeev, R.R., Jazwa, A.M., Jakubiak, A.A., 2014, Compositional and pressure effects on the solubility of H2O and CO2 in mafic melts: Chemical Geology, v.388, p.112-129. doi: 10.1016j/chemgeo/2014.09.001. Sinton, C. W., Christie, D. M., Coombs, V. L., Nielsen, R. L., and Fisk, M. R., 1993, Near- primary melt inclusions in anorthite phenocrysts from the Galapagos Platform: Earth and Planetary Science Letters, v.119, no. 4, p.527-537. Sobolev, A. V., and Shimizu, N., 1993, Ultra-depleted primary melt included in an olivine from the Mid-Atlantic Ridge: Nature, v.363, p.151-154. doi:10.1038/363151a0. Sours-Page, R.E., Nielsen, R.L., and Batiza, R., 2002, Parental magma diversity on a fast spreading ridge: Evidence from olivine and plagioclase-hosted melt inclusions in axial and seamount lavas from the northern East Pacific Rise: Chemical Geology, v.183, p.237-262. Sprtel, Frank. M., 1997, Two styles of oceanic near-ridge volcanism for the Southeast Indian Ocean and the NE Pacific Ocean. [Masters. thesis]: Corvallis, Oregon State University, 53 p. Steele-MacInnis, M., Esposito, R., and Bodnar, R.J., 2011, Thermodynamic model for the effect of post-entrapment crystallization on the H2O-CO2 systematics of volatile-saturated silicate melt inclusions: Journal of Petrology, v. 52, p.2461-2482. doi: 10.1093/petrology/egr052. Sterner, S.M., and Bodnar, R.J., 1984, Synthetic fluid inclusions in natural quartz I. Compositional types synthesized and applications to experimental geochemistry: Geochimica et Cosmochimica Acta, v. 48, p.2659–2668. Stünitz, H., and Tullis, J., 2001, Weakening and strain localization produced by syn- deformational reaction of plagioclase: International Journal of Earth Sciences, v.90, n.1, p.136-148. Tait, S., 1992, Selective preservation of melt inclusions in igneous phenocrysts: American Mineralogist, v. 77, p. 146-155. Thy, P., Lesher, C.E., and Fram, M.S., 1998, Low pressure experimental constraints on the evolution of basaltic lavas from site 917, southeast Greenland continental margin: Proc.Ocean Drill. Program Sci. Results, v.152, p.359–372. Toplis, M. J., 2005, The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems: Contributions to Mineralogy and Petrology, v.149, no.1, p.22-39. doi: 10.1007/s00410-004-0629-4. Tormey, D. R., Grove T.L., and Bryan, W. B., 1987, Experimental petrology of normal MORB near the Kane Fracture Zone: 22o-25ON, mid- Atlantic ridge: Contribution to Mineralogy and Petrology, v.96, p.121-139. P a g e | 78
Tribaudino, M., Angel, R.J., Cámara, F., Nestola, F., Pasqual, D., Margiolaki, I., 2010, Thermal expansion of plagioclase feldspars: Contribution to Mineralogy and Petrology, v. 160, p.8999-908. doi:10.1007/s00410-010-0513-3. Tullis, J., and Yund, R.A., 1977, Experimental deformation of dry Westerly granite : Journal of geophysical research, v. 82, no.36, p. 5705-5718. Ustunisik, G., Nielsen, R.L., and Walker, D., 2017, Phase equilibria of plagioclase ultraphyric basalts: constraints from high-pressure experiments. Paper presented at Geological Society of America Annual meeting, Seattle, WA. Valer, M., Bachèlery, P., and Schiano, P., 2017, The petrogenesis of plagioclase-ultraphyric basalts from la Réunion Island: Journal of Petrology, v. 58, n. 4, p. 675-698. doi: 10.1093/petrology/egx030. Van Orman, J. A., Cherniak, D. J., and Kita, N. T., 2014, Magnesium diffusion in plagioclase: dependence on composition, and implications for thermal resetting of the 26Al–26Mg early solar system chronometer: Earth and Planetary Science Letters, v.385, p.79-88. doi :10.1016/j.epsl.2013.10.026. Wallace, P.J., and Carmichael, I.S.E.,1994, Sulfur speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts: American Mineralogist, v.79, p.161-167. 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, v.140n no.1-3, p.217-240. doi :10.1016/j.jvolgeores.2004.07.023 Wallace, P.J., Kamenetsky, V.S., and Cervantes, P., 2015, Melt Inclusion CO2 Contents, pressures of olivine crystallization, and the problem of shrinkage bubbles: American Mineralogist, v.100, p.787-794. doi: 10.2138/am-2015-5029. Wanless, V.D., and Shaw, A.S, 2012, Lower crustal crystallization and melt evolution at midocean ridges: Nature Geoscience, v.5, p.651-655. doi: 10.1038/NGEO1552. Wanless, V.D., Behn, M.D., Shaw, A.M., and Plank, T., 2014, Variations in melting dynamics and mantle compositions along the Eastern Volcanic Zone of the Gakkel Ridge: insights from olivine-hosted MIs: Contribution to Mineralogy and Petrology, v.167. doi: 10.1007/s00410-014-1005-7. Wanless D. V., Shaw, A.M., Behn, M.D., Soule, S.A., Escartin, J., and Hamelin, C., 2015, Magmatic plumbing at Lucky Stike volcano based on olivine-hosted melt inclusion composition: Geochemistry, Geophysics, Geosystems, v. 16, p. 126-147. doi: 10.1002/2014GC00551. Wanless, V. D., and Behn, M. D., 2017, Spreading rate‐dependent variations in crystallization along the global mid‐ocean ridge system: Geochemistry, Geophysics, Geosystems, v. 18, no. 8, p.3016-3033. doi:10.1002/2017GC006924. Wilson, D.S., Hey, R.N., and Nishimura, C., 1984, Propagation as a mechanism od reorientation of the Juan de Fuca Ridge, Journal of Geophysical Ressource, v.89, p384-392. Winter, J.D., 2013, Mid-ocean ridge volcanism, in Principles of igneous and metamorphic petrology, Second edition: Pearson Education, Chapter 13, p. 245. Yang, H. J., Kinzler, R. J., and Grove, T. L., 1996, Experiments and models of anhydrous, basaltic olivine-plagioclase-augite saturated melts from 0.001 to 10 kbar: Contributions to Mineralogy and Petrology, v.124, no.1, p.1-18. Zhang, Y., 1998, Mechanical and phase equilibria in inclusion-host systems: Earth and Planetary Science Letters, v.157, p.209-222. P a g e | 79
APPENDIX A SUPPLEMENTAL INFORMATION FOR CHAPTER TWO
Homogenization experiments and sample preparation for volatile analysis
Essentially all plagioclase-hosted MI have undergone significant post entrapment crystallization and contain one or more vapor bubbles. The reasons for the extent of post entrapment crystallization may be linked to two factors: (1) the host magmas are often significantly less primitive than the melts from which the megacrysts formed; (2) the phase equilibria of plagioclase results in twice the amount of crystallization per degree as for olivine (Kohut and Nielsen 2003). Therefore, reconstruction of the composition of the initial composition of the melt requires homogenization (Figure A1). Homogenization experiments were carried out using a 1-atm vertical furnace at Oregon State University by suspending 3 to 6 crystals in a platinum boat within the furnace at the entrapment temperature of 1230°C. That temperature was determined based on trial and error, was constrained by the olivine plagioclase cotectic and was confirmed using a heating stage (Sinton et al., 1993; Nielsen et al., 1995; Kohut and Nielsen, 2003; Nielsen, 2011). At the end of a run the samples were quenched in water. To test plagioclase crystals are reliable pressure vessels, samples were homogenized for 30 minutes, which is a typical time for homogenization experiment, and for to 4 days. The assumption was that the total reconstructed CO2 concentration of the MI should not change as a function of experimental run time. Once homogenized for post entrapment crystallization, the plagioclase crystals were polished individually by fixing them on a round glass slide with crystal bond. Coarse silica grids of 600 and 1200 µm were used to remove the crystal bond from the surface of the plagioclase crystals and expose the melt inclusions. Once the MI were exposed, alumina powders of 5, 3 and 0.1 µm were used to finish polishing and remove any silica grid left on the crystals. The crystal bond was then dissolved in acetone, followed by a cleaning in a sonic bath with an alcoholic solution for 1h. The plagioclase crystals were then pressed flat into wells filled with indium drilled into 1-inch aluminum mounts. Once mounted in indium, the MI were evaluated for the following: cracks in or around the melt inclusions; any irregular melt inclusion shapes; the presence of daughter crystals; and/or the presence of unusually large (>10 % by volume of inclusion) vapor bubbles. Any MI meeting any of those criteria (~20% of total) was discarded from further analyses (Moore et al., 2015). P a g e | 80
Raman Analysis: CO2 in the vapor bubbles
CO2 analysis within the vapor bubbles of MI were performed at Virginia Tech (VT) on a JY Horiba LabRam HR (800 mm) Raman spectrometer. The instrument was set up to work with a 514-nm argon laser, a confocal hole diameter of 400 µm, a slit width of 150 µm, and a grating of 1800 mm-1. A synthetic fluid inclusion was used as a standard to test for the reproducibility of the Fermi-diad determination (Sterner and Bodnar 1984). Each analysis was performed with three 45-second scans that were averaged. The LabSpec software was used to apply baseline correction on each analysis and make the peak fittings on the Fermi-diad by applying a Gaussian fitting. From the peak fitting results, the values of peak splitting were determined, and the CO2 densities calculated following the calibration curve ρ = -36.42(0.31)
+ 0.355 (0.01) ΔCO2 (Lamadrid et al., 2017). After determining the CO2 densities, the CO2 concentrations of the vapor bubbles, and error analyses, were performed following Moore et al.
(2015). It should be noted that no CO2 was detected in the vapor bubbles of a melt inclusion homogenized for 4 days (sample: 4_1_3). Consequently, only the volatile concentrations obtained by SIMS analyses with their associated errors are displayed.
It is important to note that S was present as SO2 within some of the vapor bubbles after the 30-minute runs based on the presence of a Raman peak at ~1151 cm-1. However, we did not attempt to quantify the amount of S contained in the vapor bubbles because the Raman spectral features of SO2 have not been calibrated as a function of pressure or density.
SIMS analysis: CO2 and other volatiles in the glass of melt inclusions
Volatile analyses (H2O, CO2, S, F and Cl) of the melt inclusions’ glasses were performed at Woods Hole Oceanographic Institution (WHOI) on a CAMECA IMS 1280, a secondary ion mass spectrometer (SIMS). These melt inclusions were the same as the ones analyzed on the
Raman at VT. A synthetic forsterite with SiO2 of 42 wt.% (SynFo) and a basaltic glass (519-4-
1) of similar CO2 concentration as expected from the MI, pressed into indium mounts, were regularly used as standards to check for instrumental drift. The mounts were cleaned, then dried in an oven at 65°C for > 2h. Once dry, the mounts were gold-coated and placed in a vacuum oven at 110°C for several hours until placed into the instrument to let them outgas. Next, each mount sat in the SIMS airlock for at least 8h at 5.0E-9 torr for further outgassing before being introduced to the sample chamber at pressures < 3.0E-9 torr. 133Cs+ was used as the primary beam (1nA, 10 µm). The primary beam focused on a 20 µm/20 µm area, and the aperture was set up to analyze a 7.5 µm/ 7.5 µm area. Each spot was pre-sputtered for 3 min, then each mass (12C, 16O 1H, 19H, 30Si, 32S, 35Cl) was analyzed over 5 cycles for a total analysis time of 9 min. P a g e | 81
Mass interferences of 17O from 16O 1H, and 29Si 1H, from 30Si were resolved by entrance and exit slit widths yielding a mass resolving power >6000. As the MI had higher S and CO2 concentrations, the spot position on the sample was checked by verifying the intensity of the 32 12 S and C. Nine standards glasses with different H2O and CO2 concentrations were used to make the calibration curves on 12C/30Si, 16O1H/30Si, 19F/30Si, 32S/30Si, and 35Cl/30Si. During analyses, the stability of the signal was carefully monitored for any sign of sample contamination. The detection limits were the following: 30 ppm for CO2; 0.02 wt.% for H2O; 1.2 ppm for S; 1 ppm for Cl; and 1.8 ppm for F. MI were determined to be “un-breached” or intact based on their S concentration. Any MI with S concentration at, or below 90 ppm was considered breached and not further considered for this study. Indeed, MORB melts are supposed to be saturated with S (Wallace and Carmichael, 1994). Based on their S concentrations, only 2 MI were considered breached, one for each run time (Table A1). The two main sources of error were investigated to perform the error analysis. A bootstrap regression was performed on the calibration to estimate the calibration errors. A time- based background analysis was used to check for any analytical drift with time on H2O. Drift, if present, was corrected and applied to the data. As a result, the volatile concentrations presented are calculated based on the corrected calibration curves. The errors shown as 2σ were established by propagation analysis of the corrected calibration curves and corrected volatile concentration for each sample.
Once the CO2 concentrations were determined from both the vapor bubbles and the melt of a single melt inclusion, total CO2 reconstruction and errors associated were calculated as per Moore et al. (2015). Following Moore et al. (2015), the first source of error is in the 2-D measurement of the long and short axes of both the melt inclusions and vapor bubbles. These errors are estimated to be 50 µm for any 2-D measurements. Errors on 2-D measurements are then propagated on the estimation of the volumes of both vapor bubbles and melt inclusions, which themselves are propagated with SIMS and Raman analytical errors to give the final errors on the CO2 concentrations as + and – for one datapoint (Table A1). Pressures were calculated using VolatileCalc2.0 (Newman and Lowenstern, 2002) using
SiO2 = 49 wt.%, T = 1230°C, as well as the H2O analyzed from the melt inclusions’ glass, and the total CO2 reconstructed from vapor bubble and glass analyses. Although more recent models are available to interpret H2O-CO2 data from MI (Papale et al., 2006; Shishkina et al., 2014; Ghiorso and Gualda, 2015), the models are all based on the same fundamental experimental data and differences between models are insignificant for the purposes of this study. P a g e | 82
All data can be found in Table A1.
T-test: statistics comparing the volatile and pressure estimates of the 30-min and 4-day runs
T-tests were calculated using the online GraphPad software to compare the volatile and pressure estimates obtained for the 30-minute and 4-day runs. The aim was to understand whether the differences observed were statistically significant. The t-tests compare one variable at a time for the two experimental run times. All the data obtained for CO2 in the glass, H2O in the glass, total CO2 reconstructed, pressures, and %CO2 present within the vapor bubbles were computed for these t-tests (Table A2). The outcome of a t-test is the P-value. The P value is comprised between 0 and 1 and represent the probability for one variable represented in two populations for the null hypothesis to be true. In other words, the closest P is to 0, the more likely the two populations sharing that same variable are statistically distinct.
Mg# of the MI and PEC calculations
To assess the primitiveness of the MI, the Mg# of MI held at 1230°C for both 30 minutes and 4 days were calculated. Mg# was calculated assuming 2+ Fe = 0.9 x FeOtotal, (Equation 1) so that Mg# = 100x(Mg/Mg+Fe2+) in atomic proportions (Equation 2; Figure A2; Table A3). This data was gathered on different set of MI than those presented in this review, but they are nevertheless from the same sample (A91-R1). Due to the small size of the MI (20-30 µm), it was impossible to perform both chemical and volatile analyses without C contamination from coating for EMP analyses. The calculated Mg# are near-primitive melt compositions to both experimental run times (Michael and Chase, 1987; Michael and Graham, 2015; Saal et al., 2002). For each dataset, the maximum Mg# was assumed to be unaffected by post-entrapment crystallization, therefore was used as the basis for the calculation of a linear regression. Based on the Mg#, the F parameter was calculated so that
F= Mg#/Mg#highest of the dataset (Equation 3)
From the F parameter, the linear regression was calculated using %crystallization= 100-(F*100) (Equation 4) The Mg# of the 30-minute runs show near-primary compositions compared to the host lavas. MI heated for 30 minutes have a Mg# of 67.53±2.48 (2σ). The Mg# are higher in the 4- day runs due to diffusive re-equilibration with their host (Figure A2; Table A3). The linear regressions indicate that the MI have seen <10% of fractional crystallization post- P a g e | 83 homogenization experiment compared to that representing the most primitive member of the array of melts. The linear regressions also show that our MI are near to being primary in nature (Michael and Graham., 2015; Table A3; Figure A2). In addition, the high Mg# of our samples suggest that Fo88.5 olivine would be in equilibrium with those melts (Hughes, 1982). We suggest that these olivine crystals are found within the same troctolite cumulates that our plagioclase crystals originated from, but that their high density prevents them from being erupted with the plagioclase megacrysts. The 4-day runs exhibit more primitive values and fit on a different linear regression than the 30-minute runs (Figure A2; Table A3). This difference is likely due to re-equilibration processes occurring with increasing run times, where Mg from the plagioclase-hosts diffuses within the MI. Therefore, the 4-day data represents anomalously primitive melts, as the original chemistry of the MI was irreversibly modified by experimentation.
REFERENCES
Ghiorso, M. S., and Gualda, G. A., 2015, An H2O–CO2 mixed fluid saturation model compatible with rhyolite-MELTS: Contributions to Mineralogy and Petrology, v.169, no.6, p.53. doi: 10.1007/s00410-015-1141-8. Hugues, C.J., 1982, Igneous petrology: Developments in petrology, v.7, Elsevier, p.268. Kohut, E.J., and Nielsen, R.L., 2003, Low-pressure phase equilibria of anhydrous anorthite bearing mafic magmas: Geochemistry, Geophysical Geosystems, v.4 i.7. doi:10.1029/2002GC000451. 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, v.450, p.210- 222. doi: 10.1016/j.chemgeo.2016.12.034. Michael, P.J., and Chase, R.L., 1987, The influence of primary magma composition, H2O and pressure on mid-ocean ridge basalt differentiation. Contributions to Mineralogy and Petrology, v.96, i.2, p.245-263. Michael, P.J., and Graham, D.W., 2015, The behavior and concentration of CO2 in the suboceanic mantle: inferences from undegassed ocean ridge and ocean island basalts. Lithos, v. 236-237, p. 338-351. doi: 10.1016/j.lithos.2015.08.020. Moore, R.L., 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 Mineralogists, v.100, p.806-823. doi: 10.2138/am-2015-5036. Newman, S., and Lowenstern, J.B., 2002, VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic Excel: Computers in Geosciences, v.2, p.597-604. Nielsen, R.L., Crum, J., Bourgeois, R., Hascall, K., Forsythe, L.M., Fisk, M.R., and Christie, D.M., 1995, Melt inclusions in high-An plagioclase from the Gorda Ridge: an example of the local diversity of MORB parent magmas: Contributions to Mineralogy and Petrology, v.122, p.34-50. Nielsen, Roger L., 2011, The effects of re‐homogenization on plagioclase hosted melt inclusions: Geochemistry, Geophysics, Geosystems, v.12, no.10. doi:10.1029/2011GC003822. P a g e | 84
Papale, P., Moretti, R., and Barbato, D., 2006, The compositional dependence of the saturation surface of H2O+ CO2 fluids in silicate melts: Chemical Geology, v. 229, no. 1-3, p. 78- 95. doi :10.1016/j.chemgeo.2006.01.013. Saal, A. E., Hauri, E. H., Langmuir, C. H., and Perfit, M. R., 2002, Vapour undersaturation in primitive mid-ocean-ridge basalt and the volatile concentration of Earth's upper mantle. Nature, v. 419, p.451-455. Shishkina, T.A., Botcharnikov, R.E., Holtz, F., Almeev, R.R., Jazwa, A.M., Jakubiak, A.A., 2014, Compositional and pressure effects on the solubility of H2O and CO2 in mafic melts: Chemical Geology, v.388, p.112-129. Doi: 10.1016j/chemgeo/2014.09.001. Sinton, C. W., Christie, D. M., Coombs, V. L., Nielsen, R. L., and Fisk, M. R., 1993, Near- primary melt inclusions in anorthite phenocrysts from the Galapagos Platform: Earth and Planetary Science Letters, v. 119, no.4, p.527-537. Sterner, S.M., and Bodnar, R.J., 1984, Synthetic fluid inclusions in natural quartz I. Compositional types synthesized and applications to experimental geochemistry: Geochimica et Cosmochimica Acta, v. 48, p.2659–2668. Wallace, P.J., and Carmichael, I.S.E.,1994, Sulfur speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts: American Mineralogist, v.79, p.161-167.
P a g e | 85
APPENDIX FIGURES
Figure A1: A) BSE images of MI prior to homogenization experiment showing post entrapment crystallization in grey and vapor bubbles in black. B) BSE image post homogenization showing MI free of post entrapment crystallization, but with vapor bubbles.
P a g e | 86
P a g e | 87
Table A1: Details of the run times, sample names, CO2 concentrations in both glass and vapor bubble of melt inclusions. The top of the table shows the data for MI that were considered for this study. The lower part of the table shows the dissolved volatile concentrations of the two melt inclusions that breached and were discarded. The H2O, SO2, Cl and F concentrations of the melt inclusion glasses are also shown. The calculated pressures from VolatileCalc 2.0 are displayed (Newman and Lowenstern, 2002), as well as the depths calculated by multiplying by 3 the pressure estimates. The errors on the volatile concentrations, the pressures and depths are shown as well.
P value Statistically different? CO2 in glass 0.7 No H2O in glass 0.08 No Total CO2 0.13 No Pressures 0.13 No %CO2 in vapor bubble 0.006 Yes
Table A2: Results of the t-tests performed to prove whether two groups of one variable are statistically different. The P values constitute the outcome of the t-test. Results from CO2 (glass), H2O (glass), Total CO2, Pressures and % CO2 in vapor bubbles are statistically compared in between the 30-minute and 4-day runs.
P a g e | 88
Mg# MgO FeO % (Fe2+) (wt.%) (wt.%) crystallized 69.55 8.62 7.48 0.00 69.50 8.32 7.23 0.07 68.63 8.43 7.64 1.32 68.52 8.73 7.94 1.48 68.49 8.43 7.68 1.52 68.48 8.47 7.96 1.54 68.44 8.32 7.60 1.60 68.33 8.40 7.88 1.75 68.30 8.49 7.81 1.79 68.29 8.19 7.53 1.80 68.25 8.44 7.75 1.86 68.08 7.88 8.21 2.11 67.94 8.43 7.88 2.31 67.86 8.11 7.61 2.43 67.49 8.59 8.19 2.96 67.30 7.99 7.55 3.23 66.94 7.75 7.72 3.75 66.81 8.51 7.76 3.94 66.76 8.40 8.28 4.01 66.52 7.95 7.65 4.36 66.03 8.20 8.35 5.06 65.96 8.29 8.48 5.16 65.65 7.81 8.09 5.60 65.52 7.99 7.81 5.79 64.69 8.14 8.80 6.99 mean 67.53 8.27 7.88 2.90 2 σ 2.48 0.52 0.70 3.57
P a g e | 89
MgO % 4 days Mg# (Fe2+) FeO (wt.%) F (wt.%) crystallized 70.17 9.43 7.94 1.00 0.00 69.85 9.45 8.08 1.00 0.46 69.78 9.27 7.95 0.99 0.56 69.62 9.39 8.11 0.99 0.78 69.46 9.32 8.12 0.99 1.01 69.19 8.98 7.92 0.99 1.39 69.04 9.03 8.02 0.98 1.62 68.83 9.03 8.10 0.98 1.92 68.83 8.93 8.01 0.98 1.92 68.76 8.98 8.08 0.98 2.01 68.65 8.93 8.08 0.98 2.17 68.52 8.93 8.13 0.98 2.35 68.03 8.85 8.24 0.97 3.06 66.01 8.67 8.84 0.94 5.93 65.62 8.67 9.00 0.94 6.49 64.65 8.73 9.45 0.92 7.87 64.17 8.59 9.50 0.91 8.55 mean 68.19 9.01 8.33 0.97 2.83 2σ 3.63 0.53 1.02 0.05 5.18
Table A3: Mg#, MgO (wt.%), FeO (wt.%), parameter F, and % crystallization calculation assessed by linear regression for MI homogenized for both 30 minutes and 4 days. Means and standard deviations (2σ) are also shown.
P a g e | 90
9
8
7 Linear Regression on 4 day data 6 = -1.425x + 100 5 30 minutes 4 4 days
3 % Crystallization % 2 Linear Regression on 30 min data 1 y = -1.438x + 100 0 63 64 65 66 67 68 69 70 71 Mg#
Figure A2: Graph showing the results of the linear regression showing Mg# versus % crystallization for MI homogenized for both 30 minutes and 4 days. .
P a g e | 91
APPENDIX B
SUPPLEMENTAL INFORMATION FOR CHAPTER THREE
Tests performed on the importance of oxygen fugacity for heating experiments on Plagioclase crystals
Preliminary tests were made before performing the homogenization experiments to understand whether a fugacity of oxygen is needed. The experiments were performed at 1200°C for 30 min in the 1-atmosphere vertical furnace. The experimental procedure is the same as described in the material and method section of the main manuscript. Some experiments were performed in air, some others using a mix H2-CO2 with a fO2=FMQ+1. The tests revealed that plagioclase crystals remain clear post experimentation, and that the chemistry does not change significantly wether or not the oxygen fugacity is controlled (Supplementary Figure B1; Supplementary Table B1). The other chemical elements do not show more differences (Supplementary Figure B2; Supplementary Table B2).
Determination of Mg-diffusion profiles
Data analyses Crystals homogenized for 30 min and 4 days within the 1-atmosphere-vertical furnace were mounted in 2.54 cm diameter epoxy plugs for EMP analyses at OSU using a CAMECA SX 100. Plagioclase analyses were performed at 15 kV, 30 nA and a 1 µm beam, to obtain precise measurements of the MgO contents. The plagioclase procedure contained: SiO2, TiO2, Al2O3,
FeO*, MgO, CaO, Na2O, and K2O and was tested on our labradorite standard (reference USNM
115900), knowing that under a 1µm beam Na2O, K2O, and SiO2 tend to volatilize. Nevertheless, the procedure tested on our labradorite standard gave good reproducible results. The melt inclusions were analyzed using the same procedure as described in the main manuscript. A basaltic glass procedure was set up using the following conditions: 15 kV, 10 nA and, and a 5
µm beam. The MI procedure contained the following elements: SiO2, TiO2, Al2O3, FeO*, MnO,
MgO, CaO, Na2O, K2O, P2O5, S, and Cl. SiO2, Na2O, CaO, K2O, were analyzed using a 0-time intercept to accommodate for the volatility of Na and K under the electron beam. To have the best detection limit possible, MgO was analyzed for 60 secs for each analysis. This calibration was tested on our basaltic glass standard as a secondary reference (USNM 113498/1). For both the plagioclase-host crystals and the MI analyses, results were selected with totals between 98.5 wt. % and 101.5 wt.%. Errors were calculated based on the reproducibility of our two standards. P a g e | 92
The MgO was analyzed using the line point picking option on the CAMECA software. A line was picked from the center of the MI to its edge using the glass procedure. From the MI/host interface to over 30 µm away from the MI, another line containing 10 points was picked using the plagioclase procedure. Overall the step sizes vary between 2.98 µm and 4.44 µm; meaning that each 2.98 to 4.44 µm a point was analyzed from the center of the MI to > 30 µm within the plagioclase.
Mg diffusion profiles Once obtained, the data were plotted on graphs showing MgO vs. Distance to MI center. The 30-min runs do not show any sign of diffusion, whereas the 4-day runs show diffusion profiles with a drop of MgO at the MI/host interface (Supplementary Figure B3).
Within the 4-day runs, a line through each dataset was fit to be able to find the exact MI/host interface with the correct MgO content at the interface. One line was fit within the data analyzed within the MI, and one line was fit using the data analyzed within the plagioclase. The intersection to both lines for one MI-plagioclase systematics gave the MgO content at the interface, knowing the interface distance to the MI center based on BSE images (Supplementary Table B7).
Mg partition coefficient calculations Two partition coefficients were calculated for each MI/host systematics. The first one is the partition coefficient calculated at the MI/host interface (Equation B1):
MgO interface Equation B1: DMg1 = MgO MI center
With DMg1 the partition coefficient at the MI/plagioclase interface. MgO interface is the MgO calculated from the intersection of both equations, fitting the MI and plagioclase datasets MgO MI center is the MgO content at the center of the MI obtained from the glass procedure.
The second partition coefficient was calculated as followed (Equation B2)
MgO plagioclase Equation B2: DMg2 = MgO MI center
With DMg2 the partition coefficient between the MI and plagioclase, > 30µm away from the MI. MgO plagioclase is the MgO analyzed within the plagioclase using the plagioclase procedure, at least 30 µm away from the MI. MgO MI center is the MgO content at the center of the MI obtained from the glass procedure. P a g e | 93
The difference ∆DMg between the two partition coefficients was determined by (Equation B3)
Equation B3: ∆DMg = DMg2 − DMg1
Finally, the error on each data point was calculated according to the standard 2σ reproducibility. The errors on DMgs were calculated by error propagation.
P a g e | 94
APPENDIX FIGURES
1200 1150 1100 1050 1000 Air 950
S (ppm) S 900 fO2 850 800 Sulfide Saturation Line 750 700 7.5 8.0 8.5 9.0 9.5 FeO* (wt.%)
Figure B1: S (ppm) versus FeO* (wt.%) of heating experiments performed at 1200°C for 30 min. The black circles represent the experiments performed in the air, the open circles the experiments performed using FMQ+1 fO2. The sulfide saturation line for MORBs is also displayed as a solid grey line. The errors are shown as 2σ were calculated based on the reproducibility of our standard.
P a g e | 95
Figure B2: MgO (wt.%) versus a) SiO2; b) Al2O3; c) FeO*, d) CaO, e) K2O, f) Na2O (wt.%) of experiments performed at 1200°C for 30 minutes. The black circles represent the experiments performed in the air, the open circles the experiments performed using FMQ+1 fO2. The errors are shown as 2σ were calculated based on the reproducibility of our standard.
P a g e | 96
Table B1. Chemical composition of the MI homogenized for 30 min at 1200°C using the 1- atmosphere vertical furnace. The table is divided in two: the first part contains the experiments performed in air, the second the experiments performed using a fugacity of oxygen FMQ+1. The errors are shown as 2σ were calculated based on the reproducibility of our standard.
P a g e | 97
Table B2. Data Set of the chemistry of the MI heated for 30 min, 4 hours, and 4 days. The errors (2σ) are based on the reproducibility of our standard.
P a g e | 98
Table B3. Data set of the chemistry of the MI heated for 30 min, 4 hours, and 4 days that breached during the experiments. The errors (2σ) are based on the reproducibility of our standard.
P a g e | 99
Table B4. Data set of the chemistry of the plagioclase-hosts heated for 30 min, 4 hours, and 4 days. The errors (2σ) are based on the reproducibility of our standard.
P a g e | 100
Table B5. Data set of the details of the error calculations based on the reproducibility of our standards.
P a g e | 101
P values 30 minutes to 4 hours P values 30 minutes to 4 days SiO2 0.0370 (yes) 0.0001 (yes) Al2O3 0.001(yes) 0.0001(yes) MgO 0.0001(yes) 0.0001(yes) CaO 0.1214(no) 0.0001(yes) Na2O 0.0001(yes) 0.0001(yes) FeO 0.0079 (yes) 0.0001 (yes) MgO/Al2O3 0.0001 (yes) 0.0001 (yes) Na2O/CaO 0.0001 (yes) 0.0001 (yes) S 0.7558(no) 0.0435(yes) Na2O+K2O 0.0001(yes) 0.0001(yes)
Table B6. Results of the t-tests performed on GraphPad. The P values indicate whether two datasets are statistically significant. The closest the P value is to 0, the more the two datasets are statistically significant. The table compares 30-minute with 4-hour data, and 30-minute with 4-day data for SiO2, Al2O3, MgO, CaO, Na2O, FeO, MgO/Al2O3, Na2O/CaO, S and Na2O+K2O contents.
P a g e | 102
Figure B3: Additionnal MgO diffusion profiles of samples held at 1230°C for 30 minutes and 4 days. The MgO contents displayed go from the MI/host interface to 40 µm away from the MI center. A drop of MgO is visible at MI/host interface for the 4-day runs, but not for the 30- minute runs. The error on MgO (2σ) is based on the reproducibility of our standard and is 0.003.
P a g e | 103
Table B7. Determination of the MgO (wt.%) at the MI/host interface with Equation 1 representing the line describing the MgO (wt.%) as a function of distance to the center of the MI within the MI. Equation 2 represents the same parameters than Equation 1, this time within the plagioclase-hosts. The intersection of the two lines gives the MgO (wt.%) at the MI/host interface as a function of the distance from the center of the MI (μm). This distance was determined optically on BSE images, and using the known step with which the EMP analyses were performed.
P a g e | 104
Table B8. Data set of the details of the calculations of the DMg with the chemistry of the MI and plagioclase-hosts analyzed using a 1 μm beam. The distance from the center of the the MI is displayed for each data-point, as well as the long and short axis of the MI analyzed.
P a g e | 105
MI MI Observed Radius Ratio Volume radius volume ∆MgO(ppm) increase cubic radius scavanged concentrations Scavanged (µm3) (µm) (µm3) (ppm) (µm) 10 4189 500 10000 20 83776 20000 27 20 33510 500 10000 20 670206 160000 54 30 113097 500 10000 20 2261947 540000 81 40 268083 500 10000 20 5361651 1280000 109 10 4189 500 5000 10 41888 10000 22 20 33510 500 5000 10 335103 80000 43 30 113097 500 5000 10 1130973 270000 65 40 268083 500 5000 10 2680826 640000 86
Table B9. Mass balance calculations estimating the plagioclase radius necessary to scavenged both 10000ppm and 5000 ppm form the plagioclase to the MI according to the MI radius. The calculations were made for MI radii from 10 to 40 µm. From the their radi, the MI volumes wre calculated assuming spherical MI for simplification. ∆MgO (ppm) represents the MgO difference measured between MgO within the plagioclase and at the MI/plagioclase interface. The observed increase (ppm) represents the MgO increase within the MI between the 30-minute and the 4-day runs. We chose a range between 5000 ppm and 10000ppm according to what we observed in our MI. Ratio concentrations is the Observed increase divided by the ∆MgO. These ratios multipled by the MI volume give the volume scavenged within the plagioclase (µm3). Re-arrangement of the calculation of a sphere gives the cubic radius, that can finally be converted into the radius scavenged at the MI/plagioclase interface to see an increase of MgO within the MI by 10000 or 5000 ppm.
P values 30 min_interface-30 P values 30 min_all_4 P values 4 days_interface_other min_plagioclase days_plagioclase data DMg 0.7815(no) 0.2151 (no) 0.001(yes) An 0.7663(no) 0.1524(no) 0.5651(no)
Table B10. Results of the t-tests performed on GraphPad. The P values indicate whether two datasets are statistically significant. The closest the P value is to 0, the more the two datasets are statistically significant. The table compares 30-minute with 4-hour data, and 30-minute with 4-day data for SiO2, Al2O3, MgO, CaO, Na2O, FeO, MgO/Al2O3, Na2O/CaO, S and Na2O+K2O contents.
P a g e | 106
APPENDIX C
SUPPLEMENTAL INFORMATION FOR CHAPTER FOUR Video File C1: Video showing two MI 6_1 and 6_2 experiencing an episode of bubble nucleation at 1287°C.This video was made from images taken every 1 second. The timeline at which the episode of bubble nucleation occurs on the video correspond to real timeline. The video shows how the episode of bubble nucleation starts from the walls of the MI to migrate inward toward the original vapor bubble. Video File C2: Video showing a bigger MI with a pear-shape from Experiment #7 experiencing an episode of bubble nucleation at 1267°C. Despite its unsual shape, this MI was chosen to illustrate how an episode of bubble nucleation occurs as it is especially clear with this MI. This video was made from images taken every 1 second. The timeline at which the episode of bubble nucleation occurs on the video correspond to real timeline. The video shows how the episode of bubble nucleation starts from the walls of the MI to migrate inward toward the original vapor bubble.