Boro Wsu 0251E 12891.Pdf (10.93Mb)

Home , Magma chamber

RECHARGE AND MOBILIZATION OF CRYSTAL MUSH TO PRODUCE AND ERUPT

A ZONED MAGMA CHAMBER – THE TSHIREGE MEMBER

OF THE BANDELIER TUFF, VALLES CALDERA,

NEW MEXICO, USA

JOSEPH ROBERT BORO

A dissertation submitted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

WASHINGTON STATE UNIVERSITY School of the Environment

DECEMBER 2019

To the Faculty of Washington State University:

The members of the Committee appointed to examine the dissertation of JOSEPH

ROBERT BORO find it satisfactory and recommend that it be accepted.

John Wolff, Ph.D., Chair

Owen Neill, Ph.D.

Peter Larson, Ph.D.

Ren Thompson, Ph.D.

ACKNOWLEDGMENTS

Where do I start? The community that surrounds you during the pursuit of a Ph.D. is as instrumental in success as having proper instrumentation and a robust data set. In this case, one of the most influential people was Owen Neill and the countless (frustrating?) hours he spent teaching me how to run the electron microprobe, without which, this thesis would be malnourished. In a similar manner, Arron Steiner and Scott Boroughs greatly assisted in data acquisition, idea development, and in especially in the case of Scott, played devil’s advocate in testing the validity of my ideas. Alex Iveson and Tom Shea assisted greatly in directing me to resources for areas which I lacked information and helped with idea development.

Obviously, an advisor is supposed to provide support during the pursuit of an advanced degree, but John Wolff’s advisement was unmatched. He didn’t just edit my work and guide my research, he taught me how to think. He taught me the importance of examining all data with first principles and the beauty of keeping interpretations simple. He also put up with my unending ability to find just the right way to write something badly. Mostly though he provided support to me during some of the roughest moments of my life. His guidance went beyond academic, reminding me that somethings will happen in life and there’s nothing you can do about it, so keep focus on the things you can control. I’ll forever be influenced by John as a person and a scientist. Words can’t describe my thanks, maybe a six-pack of Wrecking Ball?

Thank you, John, thank you.

Mallory Bedwell (Mal) has provided unmatched and unending emotional support, comfort and encouragement during the past year and half. Her kindness and exuberance have stoked my desire to do well and finish this. I appreciate you so much, Mal.

iii

I’d like to thank my room mates, Bailey Page and Jacob Hillard for quickly developing

into great friends and making me cookies and food and just being the nicest damn people I’ve

ever met. My life is better for having gotten to know you, and this dissertation has been made

infinitely easier with your support.

Lastly, I need to thank my parents, John and Martha Boro. I was never the easiest kid to

raise. I imagine finding patience for me as a teenager and young adult while I discovered my way

in life must have been more challenging than the work presented in this dissertation. Your support (emotionally and financially) has gotten me to where I am, and I wouldn’t be here without you. You’ll probably say you’re proud of me for doing this, but I could have been making money instead of accruing debt… so don’t be too proud.

RECHARGE AND MOBILIZATION OF CRYSTAL MUSH TO PRODUCE AND ERUPT

A ZONED MAGMA CHAMBER – THE TSHIREGE MEMBER

OF THE BANDELIER TUFF, VALLES CALDERA,

NEW MEXICO, USA

Abstract

Joseph Robert Boro, Ph.D. Washington State University December 2019

Chair: John Wolff

The 1.26 Ma Tshirege Member of the Bandelier Tuff is the second of two major (~400 km3, dense rock equivalent) compositionally zoned rhyolitic eruptions from the Valles caldera.

This eruption was triggered by a dacitic recharge magma which mixed and mingled with the rhyolite prior to eruption contributing heat and H2O from second boiling. The recharge magma interacted with a heterogenous crystal mush pile which was produced by the progressive fractional crystallization of a rhyolitic magma body. The chemistry of this rhyolite was recorded in the form of melt inclusions found within quartz in the mush pile. Faceting of melt inclusions and modeling diffusive relaxation of [Ti] across growth boundaries in quartz shows that the

Tshirege magma was fractionated over a period of ~1,000-10,000 years and then the mush was incrementally mobilized over the period of 100-1,000 years. This mobilization was mostly affected by the addition of H2O from second boiling of the invading recharge dacitic magma.

Continued recharge of the system eventually over pressurized it to a point of assured eruption.

TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... iii

ABSTRACT ...... v

LIST OF TABLES ...... ix

LIST OF FIGURES ...... x

CHAPTER

CHAPTER ONE: INTRODUCTION ...... 1

Current Paradigm ...... 3

References ...... 7

CHAPTER TWO: ANATOMY OF A RECHARGE MAGMA ...... 9

Abstract ...... 9

Introduction ...... 10

Geologic Setting...... 11

Methods...... 16

Whole-pumice chemistry ...... 17

Mineral Chemistry ...... 21

Thermobarometry ...... 36

Glass Chemistry ...... 39

Discussion ...... 42

Conclusions ...... 52

References ...... 53

CHAPTER THREE: TIMING FOR MUSH DEVELOPMENT AND REACTIVATION ...... 63

Abstract ...... 63

Introduction ...... 64

Geologic Background ...... 67

Data Acquisition ...... 69

Results ...... 75

Discussion ...... 85

Implications...... 89

References ...... 94

CHAPTER FOUR: PETROGENESIS OF THE TSHIREGE MAGMA BODY ...... 104

Abstract ...... 104

Introduction ...... 104

Geologic Setting...... 107

Methods...... 111

Chemistry ...... 112

Petrology and Crystal Chemistry ...... 115

Thermobarometry ...... 128

Discussion ...... 134

Interpretations ...... 145

Conclusions ...... 159

References ...... 160

CHAPTER FIVE: CONCLUSIONS ...... 170

APPENDIX I: Whole-pumice XRF+ICP-MS Geochemistry ...... 173

APPENDIX II: EPMA + ICP-MS Glass Chemistry ...... 186

vii

EPMA Data ...... 187

LA-ICP-MS Data ...... 195

APPENDIX III: EPMA Mineral Data ...... 206

Rhyolite-hosted Feldspars ...... 207

Dacite-hosted Feldspars ...... 221

Amphiboles ...... 230

Biotite ...... 232

APPENDIX IV: Supplemental Figures and Modeling Results ...... 231

Fig. A.4.1 Rhyolite Amphibole Compositions ...... 235

Table A.4.1 Biotite PFU calculation results ...... 233

Fig. A.4.2 Alkali glass mobility plot...... 236

Fig. A.4.3 Whole-rock addition to Figure 3.19b ...... 237

Table A.4.2 FC Modeling Parameters ...... 238

Table A.4.3 FC Modeling Results ...... 239

viii

LIST OF TABLES

Page

Table 1.1: Representative dacite whole-pumice XRF/ICP-MS data ...... 18

Table 1.2: Petrographic characteristics ...... 23

Table 1.3: Representative biotite analyses from type B and C pumice ...... 35

Table 1.4: Amphibole-Fluorine equilibrium ...... 43

Table 2.1: Whole-pumice Analyses ...... 71

Table 2.2: LA-ICPMS vs WDS-EMPA ...... 74

Table 2.3: Diffusion modeling results ...... 79

Table 3.1: Petrographic characteristics ...... 117

Table 3.2: Various geothermometers for the Tshirege Ignimbrite Sequence ...... 129

LIST OF FIGURES

Page

Figure 1.1: Geologic Map and location of sample collection ...... 14

Figure 1.2: Spider diagram for dacite REE chemistry ...... 19

Figure 1.3: Major element chemistry ...... 20

Figure 1.4: Trace element discrimination diagrams ...... 22

Figure 1.5: Feldspar and Biotite BSE images and chemical profile ...... 24

Figure 1.6: Pumice types...... 25

Figure 1.7: Plagioclase rim explanation...... 27

Figure 1.8: Feldspar ternary diagram ...... 28

Figure 1.9: X-ray maps of ternary feldspars ...... 31

Figure 1.10: X-ray slide map ...... 32

Figure 1.11: Amphibole barometry results ...... 38

Figure 1.12: CIPW normative glass compositions ...... 40

Figure 1.13: Halogen concentrations for dacite and rhyolites ...... 41

Figure 1.14: Schematic for rapid growth of ternary feldspar ...... 46

Figure 1.15: Biotite resorption graph ...... 48

Figure 1.16: Schematic cross section ...... 51

Figure 2.1: CL images of quartz ...... 66

Figure 2.2: Whole-pumice trace element graph ...... 70

Figure 2.3: Bivariate plots of melt inclusion and external glass chemistry ...... 77

Figure 2.4: Representative diffusion models ...... 78

Figure 2.5: Figure explaining faceting strength ...... 86

Figure 2.6: Melt inclusion chemistry, location, and facet strength ...... 87

Figure 2.7: Diagram describing faceting and recharge history for the Tshirege magma ...... 91

Figure 3.1: Geologic map and sample locations ...... 110

Figure 3.2: Various Harker variation diagrams ...... 113

Figure 3.3: Enrichment diagram ...... 114

Figure 3.4: BSE images of pumice types ...... 116

Figure 3.5: Large feldspar from LSR ...... 118

Figure 3.6: Resorbed quartz with melt inclusion ...... 120

Figure 3.7: Decrepitated quartz with exploding melt inclusion ...... 122

Figure 3.8: Feldspar chemistry ...... 123

Figure 3.9: CL images and feldspar chemistry ...... 124

Figure 3.10: Pyroxene summery ...... 126

Figure 3.11: Biotite aggregates ...... 127

Figure 3.12: Glass chemistry diagram ...... 132

Figure 3.13: F vs Fe plot ...... 133

Figure 3.14: Rhyolite-MELTS modeling results ...... 136

Figure 3.15: Recharge heating calculations ...... 138

Figure 3.16: Quartz volume of resorption...... 143

Figure 3.17: Biotite contributions to melting ...... 146

Figure 3.18: Bivariate plots for melt inclusion chemistry and FC modeling results ...... 148-151

Figure 3.19: Petrogenesis interpretation graphs ...... 153

Figure 3.20: Schematic diagram for petrogenesis of the Tshirege magma body ...... 155-158

This dissertation is dedicated to Kevin Dunn,

A man of unending support.

You told me to never stop pursuing my dreams; I didn’t.

You’re severely missed, Kevin; you’ll never be forgotten.

Kevin, holding me a few days after my birth.

xii

CHAPTER ONE: INTRODUCTION

The definition of a magma chamber is constantly in flux (Bachmann and Huber, 2019), but currently finds its roots in early work by authors such as Reginald Daly and can be simply summarized as a body of molten liquid with or without solid crystals, which when present, act to control the chemistry of the host liquid. During crystallization, the liquid can be thought of as evolving. These crystals can settle under the force of gravity and affect distillation of the leftover fractionated fluids. In fact, many volcanic root systems are currently thought of as a large column of crystal mush which acts to produce progressively more evolved, less dense fluids which make their ways to the surface to feed volcanoes.

“Any differentiation which depends on the sinking of crystals under gravity

belongs necessarily to a somewhat early stage of crystallization, when the bulk of the

magma was still in a liquid condition. At a later stage, when the crystals formed are so

numerous or so large as to touch and support one another, the condition may be likened

to that of a sponge full of water; it is easy to picture a partial separation being effected

by the straining off or squeezing out of the residual fluid magma from the portion already

crystalized.”

The Natural History of Igneous Rocks, Alfred Harker (1909)

Magma chambers can be thought of as open systems which are in communication with any of the surrounding country rock, underlying recharge sources, and overlying eruption pathways. At any given point in time, a magma chamber will be gaining or losing heat (+heat from recharge or -heat from conduction or eruption) and changing chemical signature (inheriting

signature from recharge magma or evolving signature from crystallization). This leads to a dynamic environment for distilling evolved magmatic liquids. In the case of multiple injection events of less evolved liquids (recharge) overpressurization may occur and can ‘break the seal’ causing a volcanic eruption. This connects the concept of the magma chamber to the volcanoes they feed.

Calderas likely represent the final stage of large ‘magma-distillation columns’ (Ganne et al., 2018), which can be thought of as a series of magma chambers that increase in felsic affinity upwards in the crust from the mantle. Each chamber finds a density equilibrium where assimilation and fractional crystallization processes take place. Through this, these columns affect production of progressively more felsic magmas. Tracking the evolution of these systems is most easily done by looking at the eruptive products at caldera volcanic centers, due to the large volume of material which is expelled during the eruption, which is thought to represent the structure of the magma chamber, only deposited inversely on the surface under the laws of superposition, where the top portion of the magma chamber is the first erupted and first deposited, so on and so fourth until the latest erupted material is emplaced at the top of the deposit.

The reason caldera volcanoes are of such importance is the inherent hazard associated with large volcanic eruptions. The hazard any volcano poses can be judged on its volcanic explosivity index (VEI), where VEI0 is nonexplosive and ejects 0-0.001 km3 of tephra, VEI5 is equivalent to the May 18th, 1980 Mount St. Helens eruption (1-10 km3), VEI7 (100-1,000 km3) are large caldera eruptions such as Tambora 1815 C.E., and VEI8 (>1,000 km3) are so-called super eruptions such as or Long Valley 760 ka, Yellowstone 640 ka (Newhall et al, 2018), or the

Valles Caldera 1.61 and 1.26 Ma.

VEI7+ eruptions have the potential to cause a massive loss of life and greatly affect the

climate by injecting SO2 gases directly into the atmosphere, causing large disruptions in the

radiative heat balance of the planet, with consequent short or long-term cooling affects (Newhall

et al, 2018). Historically this has caused food shortages due to low crop production (Tambora,

1815, “the year without a summer”, Stommel and Stommel, 1983) and in today’s socioeconomic

climate, could cause even more chaos. Volcanoes are the currently one of the largest natural

threats to humanity and being able to accurately forecast a potential VEI7+ eruption would allow

for societal preparedness and is a main driving factor behind the research presented here.

Current Paradigm

Crystal settling from magmas was the paradigm, but questioned through realizing a convecting body of magma may prevent this settling; this led to ideas such as boundary layer crystallization and sidewall transport (McBirney, 1980; Turner, 1980; Brandeis and Jaupart,

1987; Tait & Jaupart, 1992); a shift back to the crystal settling paradigm occurred in the mid-

2000s with the development of the magma-mush model which accounted for convection

(Bachmann and Bergantz, 2004; Hildreth, 2004). This model invokes convective heat loss, where crystals float freely and are coupled to liquid motions in a convecting, bulk-homogenous magma chamber. This process continues as crystals become bigger and more numerous until the crystals interfere with each other. At this point the system mechanically locks and the convection shuts off. Compaction and settling under the force of gravity separates the crystals from the fluid and creates a large body homogenous, crystal poor magma that resides over the cumulate crystal mush it produced. This melt has chemical signatures that point to a large

crystallization event, which is similar to the ‘squeezed sponge’ idea from Alfred Harker, quoted

earlier.

Once these mush piles develop they are semi-stable and exist near the granite minimum

(in rhyolitic systems) for as long as the heat flux to that system is sufficient to balance conductive heat loss and maintain an interstitial liquid, which is aided by latent heat of crystallization (Gelman et al, 2013), thus a small amount of heat flux or disturbance in equilibrium (i.e. adding H2O or raising the temperature) can cause resorption of crystals (Rubin

et al., 2017; Chapter 2). Melting these cumulate piles with recharge magmas acting as a ‘heating

element’ has been proposed as a mechanism to produce the chemical zonation present in many

compositionally zoned tuffs (Wolff et al, 2015). In this model, suggested remelting of cumulates

produces a more dense zone of liquid which is chemically related to the overlyi2 ng liquid lens

from which the crystals were derived.

Ubiquitous worldwide, these compositionally zoned tuffs often contain a minor more

mafic component, mingled with the more evolved components of the deposit (Sparks et al.,

1977). These mafic magmas have been interpreted to represent recharge of the system prior to

eruptions, and in some cases, represent the eruption-triggering magma. These recharge magmas

may provide a window into deeper portions of the root systems, where they are likely stored for

at least some time prior to interaction with the magma chambers that produce zoned tuffs.

The Tshirege Member of the Bandelier Tuff is a well-preserved example of a caldera

volcano that produced a compositionally zoned tuff with a minor mafic component, which is

thought to have triggered the eruption.

“The systematic eruptions of the Bandelier ash-flow sheets, and others like them,

are remarkable, and careful studies should lead to more refined knowledge of the

processes actually at work in magma chambers.”

Smith and Bailey (1966)

This dissertation will examine the compositionally zoned Tshirege Member of the

Bandelier Tuff located at the Valles caldera. The caldera was proceeded by ~20 m.y. of volcanic activity due to extensional magma production coincident with the opening of the Rio Grande rift.

This led to a thermally mature crust, which hosted a magma-distillation column capable of caldera-volume production.

The Valles Caldera formed in two VEI8 eruptions, separated by ~350 ka, that produced the Bandelier Tuff. The first eruption produced the Otowi Member ignimbrite sheet, which is composed of highly evolved high-silica rhyolite. The second major erupted unit, the Tshirege

Member, is zoned from high-silica rhyolite to low-silica rhyolite and contains a dacitic recharge component which is thought to have triggered the eruption. Because of this, it serves as a prime candidate to test the ‘heating element’ model and if the model holds true, should allow for a better understanding the chemical and petrological signatures to look for in these systems, especially chemical changes which might occur directly prior to an eruption, which may be able to assist in eruption forecasting.

This study starts by examining a dacitic recharge magma which is thought to have triggered the Tshirege eruption. The dacite holds clues on how recharge mechanisms of the system operates and sets a base for examining the chemical and petrologic variations of the rest of the deposit. Next, trace-element chemistry in quartz crystals and melt inclusions hosted

within them are used to work out how long it takes to activate a crystal-mush bearing system and prime it for eruption. Lastly, using petrology, mineral chemistry, whole-pumice chemistry, and melt inclusion chemistry, interpretations are made for the petrogenetic evolution from genesis of the earliest magmas, formation of crystal mush, and then mobilization to prime the eruption.

REFERENCES:

Bachmann, O., Bergantz, G.W. (2004) On the Origin of Crystal-poor Rhyolites: Extracted from

Batholithic Crystal Mushes. Journal Petrology 45-8, 1565-1582.

Bachmann, O., Huber, C. (2019) The Inner Workings of Crustal Distillation Columns; the

Physical Mechanisms and Rates Controlling Phase Separation in Silicic Magma

Reservoirs. Journal Petrology 60-1, 3-18.

Brandeis, G., Jaupart, J. (1987) The kinetics of nucleation and crystal growth and scaling laws

for magmatic crystallization. Contributions to Mineralogy and Petrology 96-1, 24-34.

Ganne, J., Bachmann, O., Feng, X. (2018) Deep into magma plumbing systems: Interrogating the

crystal cargo of volcanic deposits. Geology 46(5).

Gelman, S.E., Gutierrez, F.J., Bachmann, O. On the longevity of large upper crustal silicic

magma reservoirs. Geology 41-7, 759-762.

Harker, A., 1909. The Natural History of Igneous Rocks. MacMillan, New York, 384 pp.

Hildreth, W. (2004) Volcanological perspectives on Long Valley, Mammoth Mountain, Mono

Craters: several contiguous but discrete systems. Journal of Volcanology Geotherm Res,

136, 169-198.

McBirney, A.R. (1980) Mixing and unmixing of magmas. Journal of Volcanology and

geothermal Research 7, 357-371.

Newhall, C., Self, S., Robock, A. (2018). Anticipating future Volcanic Explosivity Index (VEI) 7

eruptions and their chilling impacts. Geosphere, 14(2), 572-603.

Rubin, A.E., Cooper, K.M., Till, C.B., Kent, A.J.R., Costa, F., Bose, M., Gravley, D., Deering,

C., Cole, J. (2017) Rapid cooling cold storage in a silicic magma reservoir recorded in

individual crystals. Science, 356, 1154-1156.

Sparks, R.S.J., Sigurdsson, H., Wilson, L. (1977) Magma mixing: a mechanism for triggering acid

explosive eruptions. Nature 267, 315-318.

Stommel, H., Stommel, E., (1983) Volcano Weather: The story of 1816, The Year without a

Summer: Newport, Seven Seas Press, 177 p.

Tait, S., Jaupart, C. (1992) Compositional convection in a reactive crystalline mush and melt

differentiation. Journal of Geophysical Research 97 B5.

The Natural History of Igneous Rocks, Alfred Harker (1909)

Turner, J.S. (1980) A fluid-dynamical model of differentiation and layering in magma chambers.

Nature 285, 213-215.

Wolff, JA, Ellis, BS, Ramos, FC, Starkel, WA, Boroughs, S, Olin, PH, Bachmann, O (2015)

Remelting of cumulates as a process for producing chemical zoning in silicic tuffs: a

comparison of cool, wet hot, dry rhyolitic magma systems. Lithos 236-237, 275¬286.

CHAPTER TWO: ANATOMY OF A RECHARGE MAGMA

ABSTRACT

The 1.25 Ma Tshirege Member of the Bandelier Tuff is the second of two major (~400 km3,

dense rock equivalent) compositionally zoned rhyolitic eruptions from the Valles caldera. It

contains a minor component of silicic dacite pumice (~69% SiO2) with a phenocryst assemblage dominated by feldspar and hornblende, subdivided into two geochemically distinct groups and three petrographic types which result from various degrees of interaction between dacite and highly porphyritic (mushy) rhyolite. Most amphibole phenocrysts yield pressures of ~0.2 GPa but a sub-population of crystal cores grew at ~1 GPa. Plagioclase phenocrysts in the least- modified dacite show correlated oscillatory fluctuations in Fe and Mg contents, evidence for recharge by still more mafic magma prior to mixing with rhyolite. Together with the high-P amphiboles, the plagioclase crystals record a 'prehistory' of dacite storage, most probably in the lower crust, prior to secondary storage below the rhyolite. Consequences of mixing with mushy rhyolite include (i) cooling and partial crystallization of dacite; (ii) growth of large, dendritic feldspars with ternary compositions; (iii) ingestion and melting of feldspar and quartz from the mush; (iv) enrichment in fluorine due to melting of biotite in the mush; (v) enrichment in light

REE contents due to melting of allanite and/or chevkinite in the mush. The dacite was likely injected into the rhyolite over a protracted period and eventually triggered the Tshirege eruption.

INTRODUCTION

Large silicic ignimbrites that erupt from continental calderas are the products of long- lived crustal magmatic zones (Karakas et al., 2017). An emerging paradigm for these systems places them at the tops of deep-rooted columns of crystal mush that may extend through the crust to the MOHO and beyond (Sparks et al., 2019). The large volumes of eruptible rhyolitic magma

(~50% crystals or less), often close to granite-minimum composition, that form extensive tuff sheets associated with calderas may be transient phenomena, generated and erupted from the mush column on timescales less than 104 years. The much longer-lived mush column (106 – 107 years) exists in a static, near-solidus state in which repeated small amounts of mass and heat flux to the system can rejuvenate and mobilize it to an eruptible condition (Annen et al., 2015;

Caricchi et al., 2014; Gelman et al., 2013; Rubin et al., 2017; Wolff et al., 2015). The ultimate triggering event for an eruption may also involve renewed magmatic pressure in the form of a magmatic recharge, but whether a single recharge event can both mobilize magma and produce a major eruption is an open question.

Petrologic data on rhyolitic ignimbrites supply the bulk of our information on the conditions of the ejected portions of these systems prior to eruption. Trace element signatures indicating extensive crystal-liquid separation are ubiquitous and include compositional variations within single zoned eruptive units (Deering and Bachmann, 2010; Hildreth and Wilson, 2007;

Wolff et al., 2015). The common presence of minor, more mafic recharge magma, texturally exhibiting mingling and mixing features, in silicic tuffs has long been recognized (Sparks et al.,

1977) as evidence of recharge directly prior to eruption, quenched by the less primitive, cooler magma body or the eruption itself. The Tshirege Member of the Bandelier Tuff contains at least two such mafic magmatic components: a long-known hornblende dacite pumice (~1% by

volume) distributed through most of the unit (Bailey et al., 1969), and trace amounts of a more

recently-reported andesite component which is present as inclusions in parts of an areally

restricted late-erupted tuff sub-unit (Goff et al., 2014).

Prior to this study, the only work dedicated to the dacite pumice was that of Stimac

(1996), who described the mineral assemblages, reported on whole-rock and mineral chemistry, noted similarity to enclaves in silicic lavas and plutonic rocks, and offered some ideas on how the dacite and rhyolites interacted directly prior to the eruption, along with suggesting the possibility of the dacite triggering the eruption.

This paper provides a more detailed treatment of the dacite pumice component, providing whole-rock, glass and mineral chemistry, a detailed petrographic description, geothermometry and barometry, and comparing its relationship to the dominant volume of chemically zoned rhyolite. We highlight evidence of complex dacite-rhyolite mixing and mingling features and discuss the role the dacite may have played in triggering an eventual eruption.

GEOLOGIC SETTING

The Valles Caldera, Jemez Mountains Volcanic Field (JMVF), New Mexico, USA, is the

type location for resurgent calderas (Smith and Bailey, 1968). A large literature exists on the

geology, stratigraphy, geochronology, petrology and volcanology of the volcanic field and the

two members of the Bandelier Tuff (e.g. Campbell et al., 2009; Gardner et al., 1986, 2010;

Heiken et al., 1986; Self et al., 1986, 1996; Goff et al., 1989, 1990, 2011, 2014; Jie et al., 2019;

Kelley et al., 2013; Phillips et al., 2007; Rowe et al., 2007; Smith and Bailey, 1966, 1968; Smith

et al., 1970; Stix et al., 1988; Wilcock et al., 2013; Wolff et al., 2000, 2002, 2005; Wolff and

Ramos, 2003, 2014).

Jemez Mountains Volcanic Field:

The JMVF sits on the western shoulder of the Española basin, one of several N-S

trending Cenozoic en-echelon sedimentary basins that make up the Rio Grande Rift. Intersecting

the rift at this location is the Jemez lineament, a northeast-southwest trending alignment of

Cenozoic volcanic fields which represent a crustal weakness resulting from a south-dipping

Proterozoic lithospheric suture zone between the Southern Yavapai and Mazatzal provinces

(Shaw and Karlstrom, 1999; Karlstrom et al., 2001; Magnani et al., 2004). The JMVF rests on a substrate of Phanerozoic sedimentary rocks overlying Proterozoic (1.62-1.44 Ga) basement which consists of granitoids and metavolcanic amphibolites (Eichelberger and Koch, 1979;

Brookins & Laughlin, 1983; Laughlin et al., 1983).

The Valles caldera represents the climax to a long history of mafic, intermediate and silicic volcanism extending sporadically back to the early Miocene, with JMVF edifice construction underway by about 10 Ma (Kelley et al., 2013; Woldegabriel et al., 2013). The principal pre-caldera formations making up the JMVF edifice are the Lobato Formation (basalt,

NE JMVF, ~13.5-9.5 Ma); Paliza Canyon Formation (dominantly andesite and trachyandesite, central to S JMVF, 10-7 Ma); the La Grulla Formation (dominantly andesite, NW JMVF, ~8-7

Ma); the Bearhead Rhyolite (N-S JMVF, 7-6 Ma); the Tschicoma Formation (dominantly dacite,

NE JMVF, 5.5-2 Ma); and the flanking Cerros del Rio and El Alto Basalts (SE and NE JMVF,

3 ~3-1.5 Ma). Overall magma production rates are low (~0.2-0.6 km per ka) compared to other

regions that have produced large, silicic volcanic systems (Wolff and Thompson, in press).

The Bandelier Tuff

The present-day Valles Caldera was formed by two major caldera collapse events which

led to the deposition of the 1.60 ± 0.02 Ma Otowi Member (Wolff and Ramos, 2014) and the

1.256 ± 0.010 Ma Tshirege Member (Phillips et al., 2007) of the Bandelier Tuff. The two

calderas are nested and slightly offset: the Toledo Caldera (Otowi eruption) was mostly

overprinted by the younger Valles Caldera (Tshirege eruption), but some portions of the earlier structure remain in the north and north-west walls (Self et al., 1986; Goff et al., 2011). The ~ 400 km3 Otowi Member (Cook et al., 2016) consists exclusively of crystal-poor (~5 – 25%), highly

evolved high-silica rhyolite with strong zonation in trace elements, and represents melt that

separated from a large volume of alkali feldspar + quartz crystalline residue (Wolff and Ramos,

2014; Wolff et al., 2015). The Valle Toledo Member rhyolites, a series of relatively small

rhyolitic domes and pumice fallout deposits, were emplaced during the 350 ka between the two

major eruptions (Gardner et al., 2010; Goff et al., 2011; Heiken et al., 1986; Jacobs et al., 2016).

Tshirege Member:

The ~400 km3 Tshirege Member is also compositionally zoned (Goff et al., 2014; Self et

al., 1996; Smith and Bailey, 1966; Wilcock et al., 2013). It consists of a plinian fallout, the

Tsankawi Pumice Bed (Qbts), followed by a series of ignimbrite units (Qbt1-5, Figure 1.1;

nomenclature of Warren et al., 2007) which are overall more densely welded and thermally

altered upwards, albeit with reversals in degree of welding. It exhibits a broader compositional

range than the Otowi Member, from high-silica rhyolite (HSR) to low-silica rhyolite (LSR) with

Fig. 1.1 Location map of Valles Caldera with Bandelier Tuff outcrops and sample locations labeled with red dots. Distribution of Bandelier Tuff from Gardner et al. (2010). Upper inset: Regional map modified from Goff and Gardner (2004): Jemez lineament, yellow line; other Cenozoic volcanic fields, light red areas; JMVF, tan; Valles Caldera outline, black dashed line; Rio Grande rift, light yellow. Lower inset: generalized cross section of the Tshirege Member, modified from Goff et al., 2011; unit names from Warren et al. (2007). Average Nb content from single whole-pumice analyses of rhyolite (this study).

a minor (~1% by volume) dacite component. The LSR pumice is absent from the Tsankawi and

lower flow units; it has overall lower concentrations of incompatible elements (e.g. Rb, Nb, Th,

U) and higher compatible trace elements (e.g. V, Ba, Sr) compared to the HSR. Individual glassy

fiammé within the welded zones contain crystals which are chemically and morphologically akin

to those found in LSR and HSR pumices. Crystal content of pumice increases upwards from

~10% in the Tsankawi to <45% in late-erupted LSR fiammé; LSR pumices contain characteristic

large crystal clots and lenses of sanidine up to 2 cm, plus ortho- and clinopyroxene, with total

crystallinities of 25-45%. Some large sanidine clots appear to be aggregate glomerocrysts, while others are partly digested crystals consisting of optically continuous fragments. Such increases in crystallinity with eruption progress in silicic tuffs have been inferred to imply, by extrapolation, the existence a non-erupted volume of complementary crystal mush represented by the aggregate grains, both specifically for the Bandelier Tuff (Wolff et al., 2015), the similar Bishop Tuff

(Hildreth, 2004; Hildreth and Wilson, 2007) and more generally (Bachmann and Bergantz, 2004,

2008).

Petrography of dacite pumice

The hornblende dacite pumice is found in the upper half of the Qbts and throughout the

Tshirege ignimbrite sequence, typically as discrete pumice clasts regarded as 'erupted enclaves'

by Stimac (1996). They range in size from <1 to 15 cm, with some fragments bearing a remnant

carapace of rhyolitic pumice, while clasts of dacite mingled with rhyolite also occur.

Dacite pumices are various shades of grey to grey-green and typically very finely (<1 mm) vesicular (avg. ~70 ± 12% vesicularity) with an average melt fraction of ~74 ± 6 %. The crystal assemblage consists of plagioclase, hornblende, Fe-Ti oxides, sparse biotite and trace amounts of clinopyroxene. Stimac (1996) reports large (several mm) clino- and orthopyroxene;

we have not found these in our set of 30 dacite samples, but large pyroxenes exist in LSR

pumices. These LSR samples look similar to some of the mingled dacite samples, but they

contain primary phenocrystic sanidine, which is absent from dacite samples. Phenocrysts in the

dacite are set in a matrix of glass with abundant microlites and microphenocrysts (10-50 μm) of

plagioclase and hornblende. Mingled samples contain alkali feldspar and quartz crystals,

presumably derived from the rhyolite fraction. Especially prominent in some dacites are large

(≤1 cm), heavily fritted feldspars of dominantly sodic plagioclase to ternary compositions.

METHODS

Whole pumice clasts were analyzed for major, minor and trace elements in the Peter Hooper

GeoAnalytical Lab at Washington State University by X-ray fluorescence and inductively

coupled plasma mass spectrometry, using procedures described at

https://environment.wsu.edu/facilities/geoanalytical-lab/technical-notes/. Mineral phases and glasses were analyzed, and back-scattered electron images and X-ray maps obtained, using a

JEOL 8500-F field emission microprobe. Plagioclase, amphibole, pyroxene, biotite and sanidine

were analyzed with an accelerating voltage of 15 kV, 10-30 nA current, and spot sizes varying

from 1-10 μm depending on textural constraints. Ilmenite and titanomagnetite were analyzed

using 20 kV, 30 nA and varying spot sizes. X-ray maps were made using higher currents (150-

200 nA) and short dwell times (10-60 ms), depending on the scale of the map. F and Cl in glass,

biotite and amphibole were also analyzed using the JEOL 8500-F and checked against blank

check standards and a variety of other standards containing a range of F concentrations. All

mineral and chemical data are in Appendix I.

WHOLE-PUMICE CHEMISTRY

Representative single whole-pumice analyses are given in Table 1. The full data set can

be found in the Appendix I. Whole-pumice compositions fall into two groups: Group I pumice

exhibits lower light–middle rare earth element (L–MREE) concentrations, and Group II pumice

is comparatively enriched in L–MREE (Fig. 1.2). Heavy REE concentrations overlap among all

samples. Group I dacite pumices are found throughout the deposit including the upper half of the

Qbts, whereas Group II pumices are found in units Qbt2–3 but are absent from the Qbts.

Major-element chemistry

Selected elements are plotted vs. SiO2 in Figure 1.3 along with LSR and HSR from the

Tshirege Member. On average, Group II dacites are enriched in SiO2 and depleted in mafic

components such as TiO2, MgO and CaO compared to Group I, with considerable overlap; there

is less overlap between Group II and LSR pumices for many elements, despite close similarity of

REE contents (Fig. 1.2).

Trace-element chemistry

Thorium is an incompatible element in the Bandelier magma system (Boro, 2019; Wolff et al., 2015) and hence serves as an index of magma evolution. The LREE enrichment among

Fig. 1.2 Rare earth elements, normalized to bulk silicate Earth (McDonough and Sun, 1995), for chemical Group I and Group II dacite pumices with shaded regions showing total range for each type, and dashed red line showing low-silica rhyolite range.

Fig. 1.3 Selected Harker variation diagrams from some major and trace elements in Tshirege dacitic and rhyolitic whole pumice.

Group II dacites is clearly independent of the degree of evolution of the dacite (Fig. 1.4a).

The ration Zr/Hf, a monitor of zircon fractionation, shows a smooth baseline decrease vs. Th

with a minor offset to higher values among LSR, a subset of HSR, and some Type II dacites

(Fig. 1.4b). Heavy REE are also incompatible in the Bandelier system (Wolff et al., 2015); HSR

show a bimodal distribution in incompatible element abundances (Fig. 1.4) with overall strong depletion in elements compatible into alkali feldspar (Sr, Ba, Eu; Figs. 3, 4c; see also Balsley,

1988). Group II dacites overlap with LSR in their incompatible element concentrations, but have overall higher compatible element abundances (Figs. 2–4).

MINERAL CHEMISTRY

Group I and Group II dacites share phase assemblages dominated by plagioclase feldspar,

ternary composition fritted feldspars of varying sizes, amphibole, and accessory amounts of

titano-magnetite, ilmenite, apatite, and ortho- and clinopyroxene (Table 2). Group II dacite additionally contains quartz, sanidine, and trace amounts of biotite and pyroxene which exhibit reaction textures (Fig. 1.5b and d). Amphibole-plagioclase intergrowths with apatite inclusions are common in Group I pumice. (Fig. 1.5a). Modal abundances given in the following sections are calculated for dense rock (vesicle free) equivalence.

The dacite pumice can be divided into three petrographic types: A, B, and C, which form a continuum of mixing with rhyolite where type A is closest to the pure dacite end member (see

Fig. 1.6). Type A are in whole rock geochemical Group I, whereas type B pumices fall in both

Group I and Group II categories, and type C pumices have exclusively Group II compositions

Fig. 1.4 Trace element variation diagrams for Tshirege dacite and rhyolitic whole-pumice analyses. a. Ce vs Th. b. Th vs Zr/Hf, with zircon fractionation trend labeled. c. Yb vs Eu/Eu*.

Fig. 1.5. Back-scatter electron (BSE) images of dacite pumice. a. Plagioclase (P)-amphibole (Am) intergrowth with apatite (Ap) inclusion. b. Ternary feldspar (F) in reaction texture with magnetite (Mt), biotite (Bt) and amphibole (Am). c. Plagioclase phenocryst from Type-I dacite showing complex oscillatory zoning; note some zones have rounded or ragged boundaries. d. Similar to b., but pyroxene present. e. Plagioclase phenocryst with accompanying WDS FeO and MgO chemical data from transect core to rim transect (green line).

Fig 1.6. Three petrographic dacite pumice types. There exists a spectrum from type A to type C; the images show exemplars of each. a. Type A pumice: plagioclase and amphibole crystals are in good condition and samples lack fritted feldspars. b. Type B pumice with some small fritted feldspars (FF), plagioclase rims show some reaction features and groundmass crystals are smaller. Some biotite is present; inset in this panel shows a plagioclase with reaction rim and an adjacent small biotite (Bt). c. Type C pumice with large fritted feldspars.

(Table 2). Type A pumice (Fig. 1.6a) contains euhedral phenocrysts with sharp crystal

boundaries and corners (Fig. 1.5a, c, e), which exhibit no reaction features. Type B is a reacted

dacite, where plagioclase crystals have rounded or resorbed crystal rims which have been

mantled by K-rich albite or sanidine (Fig. 1.6b inset; Fig. 1.7). Some small (<100 um in long

dimension) anhedral crystals of quartz and sanidine are present with small populations of fritted

feldspars (Fig 6-b). Type C pumice samples have visible banding with rhyolitic pumice in some hand samples, heavily reacted rims on plagioclase with mantling features similar to type B, contain large phenocrysts (>500 µm) of quartz and sanidine with reaction rims, and large fritted

feldspars of ternary composition are abundant. Type A pumices are rare and restricted to the

Qbts; only two samples fall into this category, whereas types B and C are mostly absent from the

Qbts and are more abundantly found throughout the rest of the ignimbrite sequence.

The crystals present in the various pumice types can be divided into two general size

distributions of phenocrysts and groundmass crystals. We use the word “phenocryst” solely as an

indicator of size without any implications of magmatic provenance or relation to enclosing glass.

The phenocrysts in Types-b and c pumices a show a variety of reaction features. For a full list of

petrographic characteristics, see Table 2.

Feldspars

Feldspars fall into three main populations. (1) Plagioclase, as phenocrysts with complex

oscillatory zoning and small groundmass crystals; (2) Sanidine existing both as phenocrysts

inherited from the rhyolite and also as euhedral groundmass crystals (<20 µm, long axis); (3)

Generally large (<1-30 mm, long axis), fritted ternary feldspars which occur alone or growing on

Fig. 1.7 d. BSE image of plagioclase rim from types B and C pumice. White square shows location of x-ray maps for a. K, b. Ca, c. and Ba.

Fig. 1.8 Feldspar compositions from various pumice types in the Tshirege Member; solvus lines at 0.1 GPa from Fuhrman and Lindsley (1988).

sanidine or magnetite. Compositions of different types of feldspars found in dacitic and rhyolitic

pumice are shown in Figure 1.8.

Plagioclase

Plagioclase crystals occur in two populations: larger phenocrysts (0.5–2 mm) which show complex oscillatory zoning, which ranges in composition from An61 to An24 across all samples,

with any individual crystal showing ranges ~20% An content, and groundmass feldspars (20–100

µm), which are sometimes zoned and range in composition from An55 to An19.

A zoned plagioclase phenocryst is shown in Figure 1.5e, with FeO and MgO

concentrations shown through a core-rim transect. FeO and MgO vary by a factor of 2 and are

clearly correlated, indicating fluctuating mafic character in the melt from which the feldspar was

growing, or repeated crystal transfer between different melts. This type of zoning suggests that

the feldspars in this category grew in a magma which was subjected to periodic recharge events

or were transferred between co-existing melts of different compositions. Some zone boundaries

in these feldspars have rounded corners or are ragged (Fig. 1.5), interpreted as representing a

resorption-regrowth event, rather than a boundary layer chemical gradient driven by

crystallization.

‘Clean’, near-euhedral plagioclase phenocrysts, as shown in Figs. 5a, c, e, are found in

type A pumice and are less common in type B. Type C pumice contains no ‘clean’ plagioclase;

the most common plagioclase phenocrysts in types B and C pumices exhibit a texture where An-

rich zones (~An60) bridge a porous, melt-free gap between ~An35 feldspar with an irregular outer

boundary and a feldspar rim of more ternary compositions with up to 1.5 wt. % BaO present in

discrete, splotchy zones (Fig. 1.7). We interpret this texture as a resorption and rapid-regrowth

feature following a resorption event, as dacite mixed and mingled with water-rich rhyolite and

attempted to reach a new equilibrium before being quenched during the eruption, and may

represent an intermediate step for the formation of the large fritted feldspars seen predominately

in type C pumices. The high Ba may be derived from resorption of potassium feldspar or biotite

from the rhyolite crystal mush (see below).

Small, groundmass plagioclase grains range from An50 to An15 and trend towards ternary

compositions similar to the large fritted feldspars (see below) in these samples and are found coexisting with sanidine groundmass crystals. These small plagioclase grains are interpreted to represent continuous crystallization during complex mixing and interaction with the rhyolite.

Sanidine

Sanidine in the dacite is restricted Types B and C pumice and occurs as groundmass crystals (20–100 µm) with compositions of Or25 to Or45, similar to the range seen amongst

sanidine found in the LSR and HSR when plotted on feldspar ternary diagrams (Fig. 1.8).

Fritted Feldspars

Large (5–30 mm) fritted feldspars of dominantly ternary composition (Fig. 1.9) are found

in both type B and, in greater abundance, in type C pumice, but are absent from type A. Fine-

scale fritting (or sieving) is shown in Figure 1.9a. The fritted zones show micron-scale crystal-

glass intergrowth (Fig. 1.9b). Glass compositions within these fine-scale segregations of the

fritted feldspars match the glass compositions outside the feldspars, and do not show any

enrichment in feldspar components. Some grains show dendritic morphology in which adjacent

‘branches’ are not in optical continuity (Fig. 1.9f). These features suggest that the fritted feldspar

Fig. 1.9 Images of fritted feldspars in disequilibrium. a. representative fritted feldspar commonly found in type B pumice. b. BSE image of ternary feldspar interstices and rim. Dark grey areas are glass, lighter grey areas are feldspar. Note the fine-scale intergrowth and dendritic character of growth. c. K x-ray map of ternary feldspar growing on sanidine core. d. Ca x-ray map showing ternary feldspar in early stages of growth mantling sanidine. e. Al X-ray map of large ternary feldspar showing a core of oscillatory zoned plagioclase. f. Cross-polarized light petrographic image showing dendritic habit; note optical discontinuity.

orange in -

ray map and BSE image (inset) of LSR crystal clot sample. Large sanidines in BSE show up yellow crystalshow imageLarge in LSR clot sanidines BSE BSE (inset)rayand sample. map of - K X

ray map. Map shows enrichment of feldspar components in enrichment ofglass surroundingray components Mapfeldspar crystals. map. shows Fig. 1.10 K X -

texture resulted from rapid growth, rather than resorption as proposed by Stimac (1996). In

contrast, Fig. 1.10 shows a K X-ray map of sanidine, with a different, ‘blocky’ internal texture,

from the LSR which exhibits enrichment of feldspar components within the glass interstices of

the feldspar, interpreted to represent resorption of the sanidine. Fritted feldspar compositions

span from An29-Ab67-Or4 to An3-Ab69-Or28, compositions consistent with temperatures up to 850

°C (Fig. 1.8), although we do not suggest that these feldspars attained equilibrium with the melt

prior to quenching. Instead, they are interpreted to have grown rapidly as the dacite and rhyolite

mixed and mingled while the mix attempted to reach equilibrium.

Quartz

Quartz is a rare phase in type B and C dacites. Occasional large (>500 um) grains are

restricted to samples that exhibit mingled and mixing textures, and are derived from the rhyolite,

as indicated by the presence of HSR melt inclusions in some grains. Most samples have <<1%

modal abundance of anhedral groundmass quartz (10µm-50µm) which may also be remnants from admixed rhyolite or may have formed during ascent of the magma.

Oxides

Titanomagnetite and ilmenite make up <1% modal abundance and typically exist as touching pairs, or large exsolved aggregates located in abundance within and near fritted feldspars. Most grains show exsolution features.

Amphibole

Amphiboles account for ~5-10% modal abundance in both Group I and Group II pumice groundmass and give the pumices their common greenish appearance. They occur as larger (>1 mm in long dimension) phenocrysts of magnesio-hornblende, which are sometimes rimmed in

tschermakite, and smaller (< 50 µm) crystals of tschermakite. Mg X-ray maps in mingled samples show that amphibole is more common in Mg-rich bands of the pumice, suggesting it is

the mafic main phase growing from the dacite, but continued to nucleate during mixing.

Biotite

Biotite is scarce (< 1% modal abundance) and is only found in types B and C pumices,

although not all contain it. Biotite generally occurs in two distinct size populations exhibiting

different textures. The first consists of ‘clean’, larger crystals (>200 µm in long dimension) that

have normal totals of 93-95 wt. % oxides. The second group of biotites are generally smaller

(<30 µm) and have deformed, bent or broken, morphologies (Fig. 1.5b, d) and an overall mottled

appearance with totals ranging from 83-93 wt. % oxides. The latter population is generally seen

in reaction texture with amphibole ± pyroxene and magnetite ± feldspar present. There is some

overlap between the two populations and biotites have F contents ranging from 0.5-1.5 wt. %

(Table 3); the lower end of this range occurs only in the latter population. The range of biotite

textures appears to capture a process of progressive breakdown to ultimately form

pyroxene/amphibole, magnetite, and feldspar.

Pyroxene

Pyroxenes are rare in the samples collected for this study, with only four crystals of

hypersthene seen across all samples. Stimac (1996) found both ortho- and clinopyroxene and

describes them as early formed phenocrysts that often occur as crystal aggregates with

plagioclase up to 1 cm long. In this study, one orthopyroxene has been seen as a discrete

euhedral crystal which appears to be in equilibrium with the melt. The rest are intergrown with

amphibole and magnetite ± feldspar in biotite breakdown aggregates. One crystal with mottled

appearance was found in the groundmass, and appeared to be out of equilibrium. No

clinopyroxene has been seen in these samples.

THERMOBAROMETRY

Because dacite pumice type C shows abundant evidence for mixing and partial re-

equilibration with rhyolite, we have restricted the application of mineral thermobarometers to

mostly type A and some type B pumices in which there appeared to be limited mixing in whole- rock geochemistry (e.g. samples falling into chemical Group I), with the caveat that some of the type B data should be treated with caution.

Two-oxide thermometry

Over 40 touching Fe-Ti oxide pairs were analyzed from type A pumice, of which most exhibit exsolution; only one pair passed the Bacon and Hirshman (1988, hereafter BH) test for equilibrium. A temperature of 781-802 ± 10 °C and fO2 = -14.68 ± 0.25 were acquired using

ILMAT (Lepage, 2003 and ref. therein). We term this value ‘accepted’. All calculated two-oxide temperatures can be split into four categories. i. lower than ‘accepted’, not passing BH; ii. Same as ‘accepted’, not passing BH; iii. Same as ‘accepted’, passing BH; iv. Higher than accepted, not passing BH. Categorizing the data in this way allows for comparisons of other two-oxide thermometers.

These data were additionally passed through the thermometers of Ghiorso and Evans

(2008) and Sauerzapf et al. (2008). Ghiorso and Evans (2008) gives lower temperatures than

ILMAT for i data and agrees with ii, iii, and iv. Sauerzapf et al. (2008) temperatures agree with i, ii, and iii, but calculated higher temperatures for iv. With this it appears that the temperatures calculated for both ii and iii of ~770-790 °C are reasonable for type A pumice, but may not represent the magmatic temperature of the dacite, rather a partial quenching temperature resulting from immersion in the rhyolitic body.

Our dataset was limited by the amount of material physically available for collection and therefore we were not able to gather a statistically significant number of samples for proper error calculations on our temperature (Jolles and Lange, 2019). For that reason, the error calculated in temperature is based on the inherent error in the microprobe data.

Amphibole Barometry

The amphibole-liquid equilibrium barometer of Putirka (2016) was applied to 50 separate hornblende-glass pairs. The barometer expression involves P concentrations in the melt phase, but this is close to detection limits in the Tshirege dacite groundmass glass (Appendix I), thus leading to some large error bars.

Barometer results are summarized in Figure 1.11. Groundmass and phenocrystic amphiboles were analyzed for ground mass, rim, and core compositions. A few of the larger crystals have cores which yield pressures of 0.8-1 GPa. The rims of these amphiboles plot with most other amphiboles probed for this study; the average pressure of the lower-pressure population of data is 0.27 ± 0.09 GPa.

Fig. 1.11 Amphibole pressures vs Fe# of amphibole. Pressures calculated using eqn. 7a in Putirka (2016). Inset: image of representative amphibole phenocryst from dacites used for barometry efforts.

We conclude that there are at least two pressures recorded in these amphiboles for storage

regimes of the dacite. One higher pressure, representing lower crustal storage at 25-35 km and

one lower pressure representing amphibole growth at a shallower level of around 9 km, either

directly below the rhyolite body or in the case of type B pumice, during mixing and mingling

with rhyolitic liquid and crystals.

GLASS CHEMISTRY

CIPW normative dacite glass compositions corrected for natural silicate melts (Blundy

and Cashman, 2001) for Group I and Group II dacites are plotted in Figure 1.12 at P(H2O) =

PTotal. Glass data are in Appendix I. Glasses in one Group I pumice record higher pressures near

the 1.0 GPa cotectic curve; the same sample contains amphibole crystals with cores which gave

pressures of 0.8 – 1 GPa. Most of the glass data is in agreement with the amphibole results at

~250 MPa. The glass data on Figure 1.12 indicate a drop in P(H2O) and/or polybaric

crystallization, likely due to magmatic uprise and fully consistent with the amphibole barometry.

The second grouping of data at much lower pressure (<50 MPa) is discussed later.

Fluorine in glasses

Fluorine in magmas acts a strong network modifier due to its high electronegativity and

generally does not act as a volatile due to high solubility in silicate liquids. Fluorine contents of

metaluminous dacites generally do not exceed 1,500 ppm (e.g. Mt. St. Helens, Webster et al.,

2018.). Fluorine is abundant in the Tshirege Member glasses. Type A dacite pumice glasses have

average [F] of ~0.2–0.6 wt. % and contain no biotite. Types B and C pumice have 0.2–1.2 wt. %

Fig. 1.12 BSE image of microlites (a. quartz; b. plagioclase) in pumice glass, represented in all dacite samples, except type A. c. CIPW normative dacite glass compositions corrected for silicic melts after Blundy and Cashman (2001), plotted on Qz-Ab-Or diagram with H2O saturated minima and eutectics after Tuttle and Bowen (1958), Luth et al. (1964), Ebadi and Johannes (1991) and Blundy and Cashman (2001).

Fig. 1.13 F vs Cl wt. % for dacite and rhyolite glasses from the Tshirege, and late-erupted Otowi glass. The rhyolites are split into two groups (see text); rhyolite which is relatively enriched in incompatible trace elements in gray, and relatively depleted rhyolite in blue.

F (Fig. 1.13). Fluorine in LSR and HSR glasses ranges 0–1.7 wt. % F. Amphibole-glass equilibrium calculations using experimental KD values from Iveson et al. (2018) based on similar

compositions suggest that the amphibole grew in a melt with [F] ranging from ~60-600 ppm

(Table 4), indicating strong enrichment of F in all dacite glasses after amphibole growth.

Chlorine appears to be in equilibrium with amphiboles in all dacite pumice types.

DISCUSSION

The Tshirege hornblende dacite pumice provides insight into the interaction of recharge

and host magmas in a large silicic magma system. The range in pumice textures and chemical

compositions preserves several stages in the process of mingling between rhyolite and dacite,

together with some evidence for the history of the dacite prior to its introduction into the rhyolite

body.

The dacite pumice is distributed throughout the Tshirege Member, with the exception of

the lower half of the Tsankawi fallout, and yet the whole Tshirege Member displays strong

compositional zoning in the dominant rhyolite. Type A dacite pumice samples, which exhibit the

least evidence for mixing and disequilibrium, are taken to most closely represent the original

composition and mineralogy of dacite before interacting with the rhyolite. Type A is most

abundant in the early-erupted material and is the only dacite type found in the Tsankawi fallout

deposit. These observations indicate that dacite intrusion did not destroy pre-existing chemical

and mineralogical gradients in the rhyolitic Tshirege magma column, of which the uppermost

portion was crystal-poor and highly evolved. Assuming initial intrusion of dacite into the base of

the system, its rise to near the top may have been facilitated by partial quenching and

crystallization, and consequent second boiling-induced buoyancy, first suggested by Stimac

(1996). Although intrusion of hotter, more primitive magma at the base of a system may induce plume convection and overturn in the host fluid (Bachmann and Bergantz, 2008), there is no evidence in the Tshirege Member for wholesale disruption of zoning by intrusion of the dacite magma.

From glass chemistry in type A pumice and amphibole barometry, we can infer the dacite magma was initially stored separately from the rhyolite, at lower crustal depths, consistent with geophysical and other petrological evidence for a lower crustal mafic to intermediate magma storage zone beneath the JMVF (Aprea et al., 2002; Steck et al., 1998; Wolff et al., 2000, 2005).

This storage system was itself subjected to various recharge events recorded by the larger plagioclase feldspars (Fig. 1.5c, e). The final signature recorded in type A pumice phenocrysts was more mafic in character, as all plagioclase phenocrysts in these samples here have An-rich rims. This suggests a ‘mafic kick’ which may have started the dacite upwards to eventually interact with the rhyolite and, as argued below, its crystal mush.

Dacite-rhyolite interaction

Mineral-scale evidence

Pumice types B and C provide the richest record of dacite – rhyolite interaction. The most striking feature of these pumices is the large, fritted ternary-composition feldspars with very fine-scale glass-feldspar growth regions (Fig. 1.9).

The Tshirege Member is the result of evacuation of a silicic magma system consisting of an evolved crystal-poor cap overlying less evolved melt produced by melting of quartz-sanidine cumulate mush, in turn underlain by non-erupted crystal mush (Balsley, 1988; Goff et al., 2014;

Wolff et al., 2015; Boro, 2019). Even at water saturation of the melt, a high-crystallinity (>50%) mush, consisting of predominately anhydrous crystals, perforce has low bulk water content.

Anorthite content in plagioclase is positively correlated with temperature and water concentration in the melt (Lange et al., 2009; Waters and Lange, 2015). During the recharge event, the dacite is subjected to conditions of lower temperature and lower ambient P(H2O) as it mingles with and partly melts water-poor crystal mush, promoting the growth of albitic (ternary) feldspar. However, some portions of crystal mush contained biotite, which upon decomposition could contribute H2O to a rejuvenated mush which may contribute to the wide range in ternary feldspar compositions seen in this study (Fig. 1.8).

The envisaged sequence of fritted feldspar growth and development is depicted in Figure 1.14:

Stage 1: During interaction of H2O-saturated dacite with H2O-undersaturated rhyolite

crystal mush, Ab-rich feldspar nucleates, at least some on pre-existing surfaces (feldspar,

quartz or magnetite) and grows rapidly.

Stage 2: Dendritic growth continues producing skeletal structure (cf. Shea and Hammer,

2013). During the rapid growth and formation of these feldspars, P(H2O) within

interstices of the feldspar will become oversaturated and the melt will start to experience

second boiling.

Stage 3: Second boiling pushes the skeletal tendrils apart and pushing glass from the

interiors as water vapor escapes, leaving fritted feldspars with large gaps, partly bridged

by the fine-scale glass-feldspar growth regions.

These ternary feldspars are present in a majority of the dacite samples – including those with no chemical signature derived from admixed rhyolite, consistent with our interpretation that they

Explanation of how ternary feldspars in types B and C pumice samples formed, with BSE example of each ternaryexample pumicehow withofstage. ExplanationfeldsparsC BSE oftypesformed, andsamples in B Fig. 1.14

grew from dacite melt and are not resorbed feldspars derived from the rhyolite, which barely

overlap with the most Or-rich fritted feldspars (Fig. 1.8). The glass in the interstices of these

feldspars also is chemically identical to the glass outside these regions, also inconsistent with

resorption.

Plagioclase phenocrysts in types B and C pumice also record response to changing

conditions. The edges of most plagioclase phenocrysts have rims of high-K albite, extensively

overlapping with fritted feldspar compositions, connected to ~An35 interiors by bridges of ~An60

plagioclase (Fig. 1.7). These high-K rims contain ‘blotches’ with elevated levels of Ba. The An60

bridges may then represent rapid growth during undercooling when the dacite first interacted

with the rhyolite before the high-K rims formed, which is interpreted to have happened during

interaction of H2O-saturated dacite with H2O-undersaturated rhyolite crystal mush causing a

rapid resorption and regrowth event; the high Ba may be derived from melting alkali feldspar- dominated rhyolitic mush, with some contribution from decomposing biotite.

Biotite is restricted to pumice types B and C and is likely therefore derived from the admixed rhyolite. Biotite textures record progressive breakdown to a mixture of feldspar, magnetite and amphibole ± pyroxene in aggregate grains (Fig. 1.5), texturally distinct from larger amphibole phenocryst seen elsewhere in the dacite. Type A glasses have 0.2 – 0.6% F by weight, while Types B and C are more variable and significantly enriched in F at 0.2 – 1.2%, similar to the range seen in rhyolite glasses (Fig. 1.13). The variable enrichment is attributed to liberation of F from biotite breakdown, and incomplete homogenization of F throughout the melt. The amount of biotite that has been consumed ranges up to several tens of modal % in the

Fig. 1.15 Graph showing the amount of melted biotite needed to cause enrichment above 0.3 wt. %. Example: at 50% melt fraction, it requires 10% modal biotite with 1 wt. % F to enrich a system from 0.3 wt. % biotite to 0.5 wt. % biotite (0.3 + 0.2 black line). The range of biotite analyzed in dacite samples is labeled. To achieve enrichments to 1.2 wt. % (maximum from all samples) from 0.3% (baseline from quenched sample) it would require ~ 40% modal biotite resorption.

case of the most F-rich glasses (Fig. 1.15); note that the other components in biotite (K, Fe, Mg)

have been captured in the breakdown assemblage of feldspar + magnetite + a mafic silicate.

Other phases that might dissolve or decompose to liberate F are apatite, and amphibole.

There is no textural evidence for amphibole breakdown. Apatite is probably not a significant

source of F because there is no correlation of P with F, and P is present only in very low

concentrations in the glass.

Evidence from whole-pumice chemistry

The concentrations of several major and trace elements in chemical Group II pumices,

consisting of petrographic types B and C, overlap with Group I and trend towards the rhyolites

(Fig. 1.3). The light REE contents of Group II pumices, however, are relatively elevated and

duplicate those of the low-silica rhyolite (Fig. 1.2). On the basis of whole-pumice chemistry,

Wolff et al. (2015) regard the LSR, and perhaps the less incompatible element-rich HSR, as partly consisting of melted feldspar-rich cumulate mush. The following observations support that suggestion, and provide some insight into dacite-rhyolite-cumulate interactions:

1. Whole-pumice trace-element chemical trends suggest that Group II dacites are the

product of mixing between Group I dacite and the LSR (Figs. 3, 4).

2. Low-silica rhyolite compositions have been affected by mixing with dacite, but LSR

is not itself the product of mixing between dacite and HSR (Fig. 1.3). The crystal

mush may be represented by LSR sample SKJM-173, which has ~50% sanidine

present in large crystal clots (Fig. 1.10).

3. The LREE enrichment in Group II and LSR pumices is not correlated with major

element composition (Fig. 1.4). One possible origin of the enrichment is melting of a

chevkinite-bearing cumulate, consistent with LREE-depletion in the early erupted,

more evolved HSR (Balsley, 1988; Wolff et al., 2015; Table 1.2).

A model for recharge of the Tshirege magma body

Figure 1.16 summarizes the findings of this study explained by the text below.

It is envisioned that leading up to the generation of the Tshirege magma:

a. Dacite is stored for at least some time at a lower crustal level ~25-35 km depth and is

then stored in a second chamber in the upper crust at ~9 km of depth. Both storage

depths correlate with present-day low seismic velocity zones (Aprea et al., 2002; Steck et

al., 1993). This second storage is long enough to crystallize much of the amphibole;

plagioclase phenocrysts are probably crystallized at this time or at the lower depth, which

in either case, show a history of interaction with multiple magmas through recharge

events or being transferred between coexisting melts.

b. Although two-oxide equilibrium temperatures yield results of 780-800 °C, two-oxide

equilibrium develops late, after most of the other phases have crystallized, thus this

temperature may represent a dacite-rhyolite interaction temperature. This suggests that

the higher storage level of dacite was directly beneath or within the rhyolite system.

c. Prior to recharge, the shallow silicic system consisted of a mushy residue with a

supernatant liquid lens. During initial recharge, parts of the dacite quench against the

colder rhyolite mush and buoyantly rise to the top of the chamber (Type A dacite

pumice).

Fig. 1.16 Simplified schematic cross section of the Valles magma system prior to the 1.26 Ma Tshirege caldera-forming eruption; see text for explanation.

d. Heating and loosening of the mush are allows the dacite to chemically and physically mix

and mingle with it (Types B and C dacite pumice). Mush remobilization and melting

produces a range of rhyolite liquids (interstitial liquid, water rich; cumulate melt, water-

e. poor; local biotite-rich cumulate melt, F-rich) which mix and mingle with dacite. During

this mixing event, rapid-growth disequilibrium textures form in the dacite.

f. Continued recharge eventually leads to an over pressurization of the system and eruption.

CONCLUSIONS

The hornblende dacite component present in the Tshirege Member of the Bandelier Tuff is a

remarkable example of an eruption-triggering recharge magma that preserves textural and

chemical evidence of polybaric crystallization and magma-mush interaction. The dacite

originated in a lower-crustal intermediate magma storage zone, similar to pre-caldera JMVF intermediate magmas (Wolff et al., 2000, 2005), which itself experienced mafic recharge events.

The dacite rose to encounter the stratified Bandelier magma system in the upper crust. It

invaded and melted the cumulate pile at the base of a stratified rhyolitic melt-mush system and

experienced chemical and textural modification. These events triggered the Tshirege eruption.

REFERENCES:

Annen, C, Blundy, J.D., Leuthold, J., Sparks, S.J. (2015) Construction evolution of igneous

bodies: Towards an integrated perspective of crustal magmatism. Lithos 230, 206-221.

Aprea, C.M., Hildebrand, S., Fehler, M., Steck, L., Baldridge, W.S., Roberts, P., Thurber, C.H.,

Lutter, W.J., 2002, Three-dimensional Kirchhoff migration: Imaging of the Jemez

Mountains volcanic field using teleseismic data. Journal Geophysical Research 107, B10,

15pp.

Bachmann, O., Bergantz, G.W. (2004) On the Origin of Crystal-poor Rhyolites: Extracted from

Batholithic Crystal Mushes. Journal Petrology 45-8, 1565-1582.

Bachmann, O., Bergantz, G.W. (2008) The Magma Reservoirs That Feed Supereruptions.

Elements 4, 17-21.

Bacon, C. R., Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between

coexisting Fe-Ti oxides. American Mineralogist 73, 57-61.

Bailey, R.A., Smith, R.L. Ross, C.S. (1969). Stratigraphic nomenclature of volcanic rocks in the

Jemez Mountains, New Mexico. U.S. Geological Survey Bulletin 1274-P, 1-19.

Balsley, S.D. (1988) The petrology and geochemistry of the Tshirege Member of the Bandelier

Tuff, Jemez Mountains volcanic field, New Mexico, U.S.A: M.S. thesis, University of

Texas at Arlington.

Blundy, J., Cashman, C. (2001) Ascent-driven crystallisation of dacite magmas at Mount St

Helens, 1980–1986. Contributions to Mineralogy and Petrology 140-6, 631-650.

Boro, J.R., Wolff, J., Neill, O., Ramos, F. Shea, T. (2019) Recharge Remobilization of a Messy

Crystal Mush Pile: The Bandelier Tuff. Goldschmidt Abstract.

Brookins, D.G., Laughlin, A.W. (1983) Rb/Sr geochronologic investigation of Precambrian

samples from deep geothermal drill holes, Fenton Hill, New Mexico. Journal Volcanology

Geotherm Research 15, 43-58.

Deering, C.D., Bachmann, O. (2010) Trace element indicators of crystal accumulation in silicic

igneous rocks. Earth Planet Science Letters 297, 324-331.

Campbell, M.E., Hanson, J.B., Minarik, W.G., Stix, J. (2009) Thermal history of the Bandelier

magmatic system: evidence for magmatic injection recharge at 1.61 Ma as revealed by

cathodoluminescence titanium geothermometry. Journal Geology 117, 469–485.

Caricchi, L., Simpson, G., Schaltegger, U. (2014). Zircons reveal magma fluxes in the Earth's

crust. Nature 511, 457–461.

Ebadi, A., Johannes, W. (1991) Beginning of melting composition of first melts in the system Qz-

Ab-Or-H2O-CO2. Contributions to Mineralogy and Petrology 106(3), 286-295.

Eichelberger, J.C. Koch, F.G. (1979). Lithic fragments in the Bandelier Tuff, Jemez Mountains,

New Mexico. Journal Volcanology Geotherm Research 5, 115-134.

Fuhrman, M.L. Lindsley, D.H. (1988) Ternary-feldspar modeling thermometry. American

Mineralogist 73, 201-215.

Ganne, J., Bachmann, O., Feng, X. (2018) Deep into magma plumbing systems: Interrogating the

crystal cargo of volcanic deposits. Geology 46(5).

Gardner, J.N., Goff, F., Kelley, S.A. Jacobs, E. (2010). Rhyolites associated deposits of the

Valles-Toledo caldera complex. New Mexico Geology 32, 3-18.

Gardner, J.N., Goff, F., Garcia, S. Hagan, R.C. (1986). Stratigraphic relations lithologic variations

in the Jemez volcanic field, New Mexico. Journal Geophysical Research 91, 1763-1778.

Gelman, S.E., Gutierrez, F.J., Bachmann, O. (2013) On the longevity of large upper crustal silicic

magma reservoirs. Geology 41(7).

Goff, F., Gardner, J.N. (2004) Late Cenozoic geochronology of volcanism mineralization in the

Jemez mountains Valles caldera, north central New Mexico. New Mexico Geological

Society, The Geology of Mexico, A geologic History, 295-312.

Goff, F., Gardner, J.N. Valentine, G. (1990). Geology of St. Peter’s Dome area, Jemez

Mountains, New Mexico. New Mexico Burrow of Mines Mineral Resources, Geologic

Map 69, scale 1:24,000.

Goff, F, Gardner, JN, Baldridge, WS, Hulen, JB, Nielson, DL, Vaniman, D, Heiken, G, Dungan,

MA, Broxton, D (1989) Excursion 17B: Volcanic hydrothermal evolution of Valles

Caldera Jemez volcanic field. New Mexico Burrow of Mines Mineralogist Resource Mem

46:381-434.

Goff, F., Gardner, J.N., Reneau, S.L., Kelley, S.A., Kepter, K.A. Lawrence, J.R. (2011). Geologic

map of the Valles caldera, Jemez Mountains, New Mexico. New Mexico Burrow of Mines

Mineralogist Resource Geologic Map 79, scale 1:50,000.

Goff, F., Warren, R.G., Goff, C.J., and Dunbar, N. (2014) Eruption of reverse-zoned upper

Tshirege member, Bandelier Tuff from centralized vents within Valles caldera, New

Mexico. J. of Volcanology and Geotherm Res, 276, 82–104.

Heiken, G (1986) Introduction to the special section on the Valles caldera the Jemez Mountains

volcanic field. Journal Geophysical Research 91:1741.

Hildreth, W., Wilson, C.J.N (2007) Compositional Zoning of the Bishop Tuff. Journal Petrology

48, 951-999.

Hildreth, W. (2004) Volcanological perspectives on Long Valley, Mammoth Mountain, Mono

Craters: several contiguous but discrete systems. Journal of Volcanology Geotherm Res,

136, 169-198.

Iveson, A.A., Webster, J.D., Rowe, M.C., Neill, O.K. (2017) Major element halogen (F, Cl)

mineral-melt-fluid partitioning in hydrous rhyodacitic melts at shallow crustal conditions.

Journal of Petrol, 12, 2465-2492.

Jie, W., Rowe, M., Conin, S., Wolff, J. (2019) Mineralogical evidence of pre-caldera magma

petrogenesis in the Jemez Mountains volcanic field, New Mexico, USA. Journal Petrol, in

press.

Jolles, J.S. Lange, R.A. (2019) High‑resolution Fe–Ti oxide thermometry applied to single‑clast

pumices from the Bishop Tuff: a re‑examination of compositional variations in phenocryst

phases with temperature. Contributions to Mineralogy and Petrology, 174:70, 43 pp.

Karakas, O., Degruyter, W., Bachmann, O., Dufek, J. (2017) Lifetime size of shallow magma

bodies controlled by crustal-scale magmatism. Nature geoscience, 10, 446-450.

Karlstrom, K.E., Åhäll, K., Harlan, S.S., Williams, M.L, McLelland, J., Geissman, J.W. (2001)

Long-lived (1.8–1.0 Ga) convergent orogen in southern Laurentia, its extensions to

Australia and Baltica, and implications for refining Rodinia. Precambrian Research 111, 5-

30.

Kelley, S.A., McIntosh, W.C., Goff, F., Kempter, K.A., Wolff, J.A., Esser, R., Brashayko, S.,

Love, D. Gardner, J.N. (2013). Spatial temporal trends in pre-caldera Jemez Mountains

volcanic fault activity. Geosphere 9, 614-646.

Koning, D.J., Karlstrom, K.E., Salem, A., Lombardi, C. (2007) Preliminary Geologic Map of the

La Madera Quadrangle, Rio Arriba County, New Mexico. New Mexico Burrow of Mines

Mineral Resources. Geologic map 141, scale 1:12,000.

Lange, R.A., Frey, H.M., Hector, J. (2009) A thermodynamic model for the plagioclase-liquid

hygrometer/thermometer. American Mineralogist 94(4), 494-509.

Laughlin, A.W., Eddy, A.C., Laney, R. Aldrich, M.J. (1983). Geology of the Fenton Hill, New

Mexico, hot dry rock site. Journal of Volcanology Geotherm Research 15, 21-41.

Lepage, L.D. (2003) ILMAT: an Excel worksheet for ilmenite magnetite geothermometry

geobarometry. Computers Geosciences, 29(5), 673-678.

Luth, W.C., Jahns, R.H., Tuttle, F.O. (1964) The granite system at pressures of 4 to 10 kilobars.

Journal Geophysical Research 69, 1896-1977.

Magnani, M.B., Miller, K.C., Levander, A. Karlstrom, K. (2004). The Yavapai-Mazatzal

boundary: a long-lived tectonic element in the lithosphere of southwestern North America.

Geological Society of America Bulletin 116, 1137-1142.

McDonough, W.F. Sun, S. -S. (1995) Composition of the Earth. Chemical Geology, 120, 223-

253.

Phillips, E. H., Goff, F., Kyle, P. R., McIntosh, W. C., Dunbar, N. W., Gardner, J. N., 2007,

The40Ar/39Ar age constraints on the duration of resurgence at the Valles caldera, New

Mexico: Journal of Geophysical Res, 112, B08201.

Putirka, K. (2016) Amphibole thermometers barometers for igneous systems some implications

for eruption mechanisms of felsic magmas at arc volcanoes. American Min, 101, 841-859.

Ross, C.S. Smith, R.L. (1961). Ash-flow tuffs: their origin, geologic relations, identification.

U.S. Geological Survey Professional Paper 366, 1-81.

Rowe, M.C., Wolff, J.A., Gardner, J.N., Ramos, F.C., Teasdale, R. Heikoop, C.E. (2007).

Development of a continental volcanic field: petrogenesis of pre-caldera intermediate

silicic rocks origin of the Bandelier magmas, Jemez Mountains (New Mexico, USA).

Journal of Petrology 48, 2063-2091.

Rubin, A.E., Cooper, K.M., Till, C.B., Kent, A.J.R., Costa, F., Bose, M., Gravley, D., Deering,

C., Cole, J. (2017) Rapid cooling cold storage in a silicic magma reservoir recorded in

individual crystals. Science, 356, 1154-1156.

Self, S., Goff, F., Gardner, J., Wright, J. Kite, W. (1986). Explosive rhyolitic volcanism in the

Jemez Mountains: Vent locations, caldera development relation to regional structure.

Journal of Geophysical Research 91, 1779-1798.

Self, S, Heiken, G, Sykes, ML, Wohletz, K, Fisher, RV, Dethier, DP (1996) Field excursions to

the Jemez Mountains, New Mexico. Bulletin of the New Mexico Burrow of Geology &

Mineralogist Resource 134, 72 pp.

Shaw, C.A. Karlstrom, K.E. (1999). The Yavapai-Mazatzal crustal boundary in the southern

Rocky Mountains. Rocky Mountain Geology 34, 37-52.

Shea, T., Hammer, J.E. (2013) Kinetics of cooling- and decompression-induced crystallization in

hydrous mafic-intermediate magmas. Journal Volcanology Geotherm Research 260, 127-

145.

Smith, R.L. Bailey, R.A. (1968). Resurgent cauldrons. In Coats, R.R. et al. (eds) Studies in

volcanology. Geological Society of America Memoir 116, 613-662.

Sparks, R.S.J., Sigurdsson, H., Wilson, L. (1977) Magma mixing: a mechanism for triggering acid

explosive eruptions. Nature 267, 315-318.

Sparks, R.S.J., Annen, C., Blundy, J.D., Cashman, K.V., Rust, A.C., Jackson, M.D. (2019)

Formation and dynamics of magma reservoirs. Phil Trans R Soc A, 377, 20180019.

Smith, R.L., Bailey, R.A. Ross, C.S. (1970). Geologic map of the Jemez Mountains, New

Mexico. U.S. Geological Survey, Miscellaneous Geological Investigations Map I-571,

scale 1:125,000.

Steck, L.K., Thurber, C.H., Fehler, M.C., Lutter, W.J., Roberts, P.M., Baldridge, W.S., Stafford,

D.G., Sessions, R.K., 1998. Crust upper mantle P wave velocity structure beneath Valles

caldera, New Mexico: Results from the Jemez teleseismic tomography experiment: Journal

of Geophysical Research 103, 24,301-24,320.

Stimac, J.A., 1996, Hornblende-dacite pumice in the Tshirege member of the Bandelier Tuff:

Implications for magma chamber eruptive processes. New Mexico Geological Society,

47th Field Conference Guidebook, 269–274.

Stix, J., Goff, F., Gorton, M.P., Heiken, G. Garcia, S.R. (1988). Restoration of compositional

zonation in the Bandelier silicic magma chamber between two caldera-forming eruptions:

geochemistry origin of the Cerro Toledo Rhyolite, Jemez Mountains, New Mexico. Journal

of Geophysical Research 93, 6129-6147.

Tuttle, O.F., Bowen, N.L., 1958, Origin of granite in the light of experimental studies in the

system NaAlSi3O8-KalSi3O8-SiO2-H2O: Geological Society of America Memoir 74, 153

pp.

Warren, R.G., Goff, F., Kluk, E.C., Budahn, J.R., 2007, Petrography, chemistry, mineral

compositions for subunits of the Tshirege Member, Bandelier Tuff, within the Valles

caldera Pajarito Plateau: New Mexico Geological Society, 58th Field Conference

Guidebook, 316–332.

Waters, L. E Lange, R.A. (2015) An updated calibration of the plagioclase-liquid hygrometer-

thermometer applicable to basalts through rhyolites. American Mineralogist, 100(10),

2172–2184.

Webster, J.D., Baker, D.R., Aiuppa, A. (2018) Halogens in Mafic Intermediate-Silica Content

Magmas. In Harlov, D.E., Aranovich, A. (eds) The Role of Halogens in Terrestrial

Extraterrestrial Geochemical Processes. Springer Geochemistry, special volume.

Wilcock, J., Goff, F., Minarik, W.G., Stix, J., 2013, Magmatic recharge during formation

resurgence of the Valles caldera, New Mexico, USA: evidence from quartz compositional

zoning geothermometry: Journal of Petrol, 54, 635–664.

Woldegabriel, G., Koning, D.J., Broxton, D., Warren, R.G., 2013, Chronology of volcanism,

tectonics, and sedimentation near the western boundary fault of the Española basin, Rio

Grande rift, New Mexico, in Hudson, M.R., Grauch, V.J.S., eds., New perspectives on Rio

Grande Rift Basins: From Tectonics to Groundwater: Geological Society of America

Special Paper 494, 221–238.

Wolff, J.A., Thompson, R.A. (2018) The Jemez Mountains and Valles caldera. In: Thompson,

R.A., Turner, K., Lipman, P.W., Wolff, J.A., Dungan, M.A. and Bachmann, O., Field

guides to the Rio Grande rift, Jemez Mountains, and San Juan volcanic field, New Mexico

and Colorado. U.S. Geological Survey Scientific Investigations Report, 5022-R, in press.

Wolff, J.A. Ramos, F.C. (2003). Pb isotope variations among Bandelier Tuff feldspars: no

evidence for a long-lived silicic magma chamber. Geology 31, 533–536.

Wolff, JA, Ramos FC (2014). Processes in caldera-forming high-silica rhyolite magma: Rb–Sr

and Pb isotope systematics of the Otowi Member of the Bandelier Tuff, Valles Caldera,

New Mexico, USA. Journal Petrology 55, 345–375.

Wolff, J.A., Ramos, F.C., Davidson, J.P. (1999) Sr isotope disequilibrium during differentiation

of the Bandelier Tuff: Constraints on the crystallization of a large rhyolitic magma

chamber. Geology, 27, 495-498.

Wolff, J.A., Balsley, S.D., Gregory, R.T., 2002, Oxygen isotope disequilibrium between quartz

sanidine from the Bandelier Tuff, New Mexico, consistent with a short residence time of

phenocrysts in rhyolitic magma: Journal of Volcanology Geotherm Res, 116, 119–135.

Wolff, J.A., Brunstad, K.A., Gardner, J.N., 2011, Reconstruction of the most recent volcanic

eruptions from the Valles caldera, New Mexico. Journal of Volcanology Geotherm Res,

199, 53–68.

Wolff, JA, Ellis, BS, Ramos, FC, Starkel, WA, Boroughs, S, Olin, PH, Bachmann, O (2015)

Remelting of cumulates as a process for producing chemical zoning in silicic tuffs: a

comparison of cool, wet hot, dry rhyolitic magma systems. Lithos 236-237, 275¬286.

Wolff, J.A., Heikoop, C.E., Ellisor, R. (2000). Hybrid origin of Rio Grande rift hawaiites.

Geology 28, 203–206.

Wolff, J.A., Rowe, M.C., Teasdale, R., Gardner, J.N., Ramos, F.C., Heikoop, C.E. (2005).

Petrogenesis of pre-caldera mafic lavas, Jemez Mountains Volcanic Field (New Mexico,

U.S.A). Journal of Petrology 46, 407–439.

CHAPTER THREE: TIMING FOR MUSH DEVELOPMENT AND REACTIVATION

ABSTRACT

Many rhyolites contain quartz crystals with relatively Ti-rich rims and Ti-poor cores, with a sharp interface between the zones, attributed to partial dissolution followed by overgrowth following a heating event due to mafic recharge of the system. Quartz crystals in the compositionally zoned high-silica rhyolite 1.26 Ma Tshirege Member of the Bandelier Tuff,

Valles caldera, New Mexico, show a range in zoning styles, with Ti-rich rims becoming more abundant upwards in the ignimbrite sheet among progressively less evolved compositions. Here we compare timescales obtained by applying Ti diffusion coefficients to [Ti] profiles in Tshirege

Member quartz crystals with those from cathodoluminescence (CL) brightness profiles and show that panchromatic CL provides only a rough proxy for [Ti] in this unit. Timescales derived from

[Ti] profiles range from 55-2200 years, indicating heating and mobilization events at different times prior to the eruption. These events did not trigger the Tshirege eruption but may have played a role in destabilizing the system over the ~103 years prior to the eruption.

Melt inclusions in the Ti-poor cores of late erupted quartz are chemically akin to the early

erupted melt compositions, while adhering and groundmass glasses more closely reflect the

composition of the host pumice. The heating and mobilization events identified from quartz Ti

zoning are thus linked to overall compositional zoning of the tuff, which may have been

produced by repeated episodes of melting of a crystal cumulate cognate to the early-erupted,

evolved rhyolite. Quartz-hosted melt inclusion faceting suggests the development of a crystal

mush over a minimum time frame of 1,000 – 10,000 years prior to recharge events that produced

the eruptible Tshirege magma.

INTRODUCTION

Diffusion modeling of element concentration profiles in crystals is a fruitful way of extracting timescales of processes at elevated temperatures in geologic systems. The approach relies on some assumptions, such as starting conditions for the diffusion episode and that the concentration profiles being modeled are indeed the result of ionic diffusion. In studies of volcanic systems, diffusion profiles can help determine residence times of crystals in magma based on compositional zoning of the minerals of interest. Diffusion modeling can also be used to determine the time elapsed between an identifiable event in the history of a crystal (e.g. overgrowth on a resorption surface) and eruption (Costa and Dungan, 2005; Morgan and Blake,

2006; Wark et al., 2007; Costa et al., 2008; Till et al., 2015). Timescales deduced from the diffusive blurring of an overgrowth boundary can provide insights on how magma systems respond to disturbance, such as rejuvenation and heating by magmatic recharge, shortly before an eruption. This approach has supplied evidence for short activation timescales (<1 year – 1000 years) preceding past catastrophic silicic caldera-forming ‘super-eruptions’ (Wark et al., 2007;

Matthews et al., 2012; Till et al., 2015; Gualda and Sutton, 2016; Cooper et al., 2017).

In rhyolitic volcanic systems, quartz is an attractive and widely-used target for diffusion modeling because quartz crystals are usually large and of effectively constant major element composition. Trace element substitutions, particularly that of Ti4+ for tetrahedral Si4+ in the quartz structure (Götze, 2012; Leeman et al., 2012), result in enhanced cathodoluminescence

(CL) brightness, in principle allowing easy identification of trace cation concentration zoning, and because the behavior of, in particular, Ti in quartz is well-studied. Peppard et al. (2001) provided the first detailed description of CL zoning in quartz from a high-silica rhyolite (the

Bishop Tuff). Wark and Watson (2006) calibrated [Ti] in quartz as a geothermometer, while

Cherniak et al. (2007) reported results for the diffusion coefficient of Ti (DTi) in quartz; these two studies spawned a significant literature linking quartz [Ti] variations, CL zoning, and magmatic temperature to estimates of crystal residence timescales and the rejuvenation of large stagnant silicic magma bodies to an eruptible state, as well as new calibrations of the Ti-in-quartz

(TitaniQ) thermometer (Hayden and Watson, 2007; Wark et al., 2007; Shane et al., 2008;

Campbell et al., 2009; Thomas et al., 2010; Gualda et al., 2012; Huang and Audétat, 2012;

Matthews et al., 2012; Thomas and Watson, 2012; Wilson et al., 2012; Wilcock et al., 2013;

Gualda and Sutton, 2016; Pamukcu et al., 2015; Seitz et al., 2016; Cooper et al., 2017). The temperature sensitivity of DTi requires accurate knowledge of temperature during the diffusion episode in order to extract timescales from modeled diffusion profiles, placing demands on geothermometry, while the application of the Ti-in-quartz geothermometer requires precise knowledge of the activity of Ti in the melt, a parameter which is not always well constrained

(Huang and Audétat, 2012; Matthews et al., 2012; Thomas and Watson, 2012; Wilson et al.,

2012).

In addition to the zoning recorded in quartz during crystal growth, melt may become trapped as a melt inclusion (MI; Fig. 2.1, black areas inside of crystals); MIs are generally thought to record the liquid composition at the time of crystal growth (Lowenstern & Mahood

1991; Dunbar & Hervig 1992; Bacon et al. 1992), and are often abundant in quartz phenocrysts from rhyolites. These can then be used to make interpretations about the growth history of the quartz, and petrogenetic relations within the magma body.

Fig. 2.1. Cathodoluminescence grayscale images of representative quartz crystals from four different samples of the Tshirege Member, Bandelier Tuff, arranged in stratigraphic order. Note increasing complexity of CL-zoning upwards through the Tshirege Member. Crystallized melt inclusions appear bright; glassy inclusions are near-black.

This study examines measured [Ti] profiles in quartz phenocrysts from the Bandelier

Tuff, Valles caldera, NM, USA, and compares Ti zoning with MI chemistry. Using ﬁnite-

difference diffusion modeling procedures (Costa et al., 2008), the results are applied to

estimating crystal residence times between initiation of overgrowth following a crystal resorption

event and eruptive quenching, assuming diffusional relaxation of [Ti] in quartz crystals from an

initial step function distribution. Melt inclusions hosted within those crystals are used to

determine the petrogenetic history of the quartz. Wilcock et al. (2013) provide a detailed dataset

on the CL and Ti zoning in quartz from the Tshirege Member. This study expands on those data

with higher spatial resolution analyses and MI chemistry.

GEOLOGICAL BACKGROUND

The ~400 km3 Tshirege (upper) Member of the Bandelier Tuff erupted to form the Valles

caldera, Jemez Mountains, NM, USA at 1.26 Ma (Bailey et al., 1969; Phillips et al., 2007;

Gardner et al., 2010; Goff et al., 2014), and has been the subject of several petrological and

volcanological studies (Smith and Bailey, 1966; Self et al., 1986, 1996; Balsley, 1988; Warshaw

and Smith, 1988; Caress, 1996; Stimac, 1996; Warren et al., 2007; Wilcock et al., 2013; Goff et

al., 2014; Wolff et al., 2015). It is strongly compositionally zoned from early-erupted high-silica rhyolite to late-erupted low-silica rhyolite, with complex compositional variations in the last- erupted tuff (Smith and Bailey, 1966; Balsley, 1988; Goff et al., 2014; Boro, 2019). The

Tshirege Member event was preceded by eruption of the similarly-sized, chemically-zoned high- silica rhyolite Otowi (lower) Member of the Bandelier Tuff at ~1.60 Ma, which formed a caldera now largely overprinted by the later eruption (Goff et al., 2011, 2014; Wolff and Ramos, 2014;

Wolff and Thompson, in press). During the ~340,000 y interval between the two caldera-forming

events, several minor eruptions produced the Valle Toledo Member rhyolite domes and

pyroclastic flow and fallout deposits of the Cerro Toledo Formation (Gardner et al., 2010).

The Tshirege Member consists of a widespread plinian fallout unit, the Tsankawi Pumice

Bed, overlain by non- to densely welded ignimbrite, emplaced by numerous pyroclastic density currents (Bailey et al., 1969; Self et al., 1986, 1996). The ignimbrites are divided into mappable units, recognized over a wide area and designated Qbt1 through Qbt5; some numbered units are additionally subdivided (Warren et al., 2007; Goff et al., 2014). All units contain a range of compositions shown by bulk chemical analysis of single pumices and bulk tuff, representing magmas at different degrees of evolution (Balsley, 1988; Self et al., 1996; Warren et al., 2007;

Goff et al., 2014). The Tshirege Member consists of three magmatic components: volumetrically dominant high-silica rhyolite, low-silica rhyolite, and hornblende dacite (Bailey et al., 1969; Self et al., 1996; Stimac, 1996; Boro, 2019). The dacite occurs as scattered pumice clasts through most of the Tshirege Member, and may represent the recharge event that triggered the Tshirege eruption (Stimac, 1996; Goff et al., 2014; Boro, 2019). Goff et al. (2014) identify an additional component of andesite in late-erupted units. Despite the overall zoned character of the Tshirege

Member, there is a wide variation in pumice composition at any one stratigraphic level (Self et al., 1996). Several processes, such as overturn driven by gas exsolution or thermal disturbance before eruption (Bachmann and Bergantz, 2006; Burgisser and Bergantz, 2011), extraction effects during eruption (Blake and Ivey, 1986a,b; Trial et al., 1992), or non-sequential deposition of the different magmatic components (Torres et al., 1996; Branney and Kokelaar, 2002) may limit, scramble or obscure the record of zoning in the final deposits. Hence, we find the abundances of incompatible elements in single pumice clasts to be the most useful indicators of degree of magmatic evolution and hence likely vertical position in the pre-eruptive magma body,

assuming them as proxies for melt water content and hence density; this assumption is supported

by water contents of melt inclusions in quartz and feldspar (Dunbar and Hervig, 1992).

Whole-pumice chemistry (Fig. 2.2) shows three distinct rhyolites were present in the

Tshirege magma body at the time of eruption (Boro, 2019): two high-silica rhyolites, one

enriched in incompatible trace elements (~130-140 ppm Nb, 27-41ppm Th, and 217-331 ppm

Rb: HSR-e) and one depleted in those elements (65-80 ppm Nb, 12-23 ppm Th, and 97-198

ppm Rb: HSR-d), and a low-silica rhyolite (~25-44 ppm Nb, 9-13 ppm Th, and 49-98 ppm Rb).

The quartz grains analyzed in this study were extracted from these three compositional groups of

rhyolite pumice. The HSR-e is representative of the Tsankawi Pumice Bed and the lowermost

part of Qbt1, and the HSR-d corresponds to Unit Qbt2 and Qbt3 of Warren et al (2007). One

additional sample, 88-1, is a hornblende dacite pumice containing quartz scavenged from

rhyolite, which is texturally identical to the quartz in the rest of the rhyolite and so has been

included in this study. Whole-pumice data for selected major and trace elements are reported in

Table 2.1.

DATA ACQUISITION

We analyzed oriented quartz grain separates, cut normal to the c axis, from the Tshirege

Member and Valle Toledo unit A. Grain mounts were made using >500 µm size-fraction crystal separates. Quartz grains are frequently zoned when viewed in CL (Fig. 2.1). The style of zoning varies, but is dominantly ‘reverse’ in character, where crystal cores appear dark (lower Ti, consistent with more evolved melt or lower temperature) and rims appear bright in CL (higher

Ti). The full range of zoning styles and statistical analyses of their distribution in the Tshirege

Member is described by Wilcock et al. (2013); our study finds a similar distribution (Fig. 2.1).

Fig. 2.2. Whole-pumice Th vs Nb plotted for the Tshirege. Notice the bimodal distribution of the high-silica rhyolite magmas, allowing distinction of three magma types (enriched high-silica rhyolite, HSR-e; depleted high-silica rhyolite HSR-d, and low-silica rhyolite, LSR; see also Table 2.1).

Titanium concentrations in quartz were measured by wavelength-dispersive electron

probe microanalysis along core-rim transects orthogonal to CL zone boundaries. Measurements

were made on the JEOL JXA-8500F field emission electron microprobe located in the Peter

Hooper GeoAnalytical Lab at Washington State University, using an accelerating potential of

20kV, a beam current of 400nA, and a spot size of 2-10 µm. Measurements were made following

the procedures described by Donovan et al. (2016) for analysis of trace elements in simple

matrices such as SiO2; to summarize, measured peak X-ray intensities were corrected for continuum intensity using the mean atomic number (MAN) method of Donovan and Tingle

(1996), and then blank corrected using the procedure of Donovan et al. (2011) to eliminate systematic errors in accuracy due to background artifacts. Quartz crystals from Herkimer, NY, with Ti concentrations of <30 ppb (Kohn and Northrup, 2009; Kidder et al., 2013) were used as a blank standard. Several mounts of Herkimer quartz were prepared, so that blank standards and unknowns could be carbon coated together prior to analysis, to minimize differences in coating thickness between blank standards and unknowns.

We acquired data for a cumulative time of 18 minutes using three spectrometers.

Titanium concentration proﬁles were measured with 4 or 8 μm spacing between analysis spots.

Measurement of [Ti] in this manner allows for detection limits of <2 ppm and analytical precision (based on counting statistics) of <5% (2-sigma) for each individual measurement. This combination of spot size and uncertainty is comparable to those obtained using synchrotron XRF on similar samples (Matthews et al., 2012; Gualda and Sutton, 2016), but with improved (i.e. shallower) depth averaging.

Measurements were made either on grain mounts, using quartz crystals separated from the rock matrix, or in thin sections on areas far enough from adhered Ti-bearing glass or Ti-rich

phases, to avoid artifacts from continuum fluorescence (cf. Fournelle 2007). To minimize effects

arising from obliquely or randomly oriented zone boundaries and off-center sections (Shea et al.,

2015), crystals were mounted such that profiles were measured normal to the c axis across the broadest part of quartz bipyramids (Fig. 2.1). Modeling of the effects of continuum fluorescence across the boundaries between low- and high-Ti zones within the quartz crystal using

PENEPMA (Llovet and Salvat, 2016) suggests that any effects of secondary fluorescence of the higher-Ti zone across the boundary are negligible (~1ppm or less).

Grayscale CL images were obtained using a Gatan MiniCL cathodoluminescence detector mounted on the WSU JEOL JXA-8500F using a 50 nA beam current, 10 kV, and 60 second capture time producing images with an effective image resolution of 1 µm. ImageJ software was used to produce grey-scale intensity graphs along EPMA transect lines.

Titanium concentrations were also measured in a subset of quartz crystals by laser ablation inductively-coupled plasma mass spectrometry (LA-ICP-MS), to confirm the accuracy of the EPMA measurements. The LA-ICP-MS measurements were performed on zones in quartz crystals that were uniformly CL-bright or CL-dark, where EPMA measurements yielded uniform

Ti concentrations. Data were collected using a Teledyne Analyte Excite Excimer 193 nm laser

ablation system attached to a Finnigan Element2 ICP-MS. A 20 μm spot size was used with 7

J/cm2 beam energy, and counts collected for 30 seconds along transects that ran parallel to

crystal faces and zone boundaries, staying within the zones defined by CL imaging. Calibration

was achieved using NIST-612 and BCR-G glasses, plus a well-characterized internal standard

quartz with 9 ppm Ti. Results of LA-ICP-MS and EPMA measurements of [Ti] are

indistinguishable within analytical uncertainty (Table 2.2).

Melt inclusions in quartz, adhering glass, and groundmass glasses in the HSR-e, HSR-d, and LSR samples were analyzed for Ti, Zn, Rb, Sr, Y, Zr, Nb, Cs, Ba, REEs, Hf, Ta, Pb, Th, and

U using LA-ICPMS. Some additional thin sections with exposed quartz melt inclusions were analyzed. A New Wave UP-213 laser was used for ablation with a 20 Hz rep rate and laser power from 3-3.5 J/cm2. Ablations were analyzed with a with a sampling rate of every ~0.9 sec

with an Agilent 7700 Series quadrupole mass spectrometer. Laser track widths varied from 12-30

µm depending on available glass and melt inclusion size. Inclusions <50 µm in diameter were

avoided. Data were normalized to NIST-610 using 28Si as an internal reference element.

Whole-pumice clasts were analyzed for major, minor and trace elements in the Peter

Hooper GeoAnalytical Lab at Washington State University by X-ray fluorescence and

inductively coupled plasma mass spectrometry (Methods and procedures described at

https://environment.wsu.edu/facilities/geoanalytical-lab/technical-notes/).

RESULTS

Cathodoluminescence

CL grayscale images were obtained for >50 quartz crystals. Titanium concentration

proﬁles plus CL grayscale profiles were obtained for 20 quartz crystals from samples taken at

several stratigraphic heights within the Tshirege Member, and two Valle Toledo Member unit A

samples. Generally, CL zoning can be split into three types (Fig. 2.1): 1. Non-zoned; 2.

Reversely zoned; 3. Complex. In cases where there were multiple zones (i.e. Fig. 2.1, panel 1, 3,

and 5), the outermost zone was modeled for diffusion.

Glass trace element chemistry

Trace-element abundances in glasses are somewhat bimodal, similar to the distribution of whole-pumice compositions, but do not always reflect the bulk chemistry of the clast from which they were extracted (Fig. 2.3). Melt inclusions in quartz from HSR-e plot with glasses external to quartz crystals and groundmass glass remote from crystals in thin section (collectively referred to as “external glasses”) from those samples. Melt inclusions from Ti-poor quartz cores in the HSR-

d and LSR plot with HSR-e glasses, whereas MIs which plot with HSR-d and LSR glasses are

found in CL-bright rims of crystals or in crystals which show no zoning in the HSR-d or LSR.

Some MIs fill a gap between the two groups of data and may have trapped a liquid composition

that is not represented by whole-pumice data or groundmass glass data.

[Ti] vs. CL grayscale intensity

Four representative [Ti] proﬁles measured across CL brightness boundaries are plotted

with CL grayscale intensity (Fig. 2.4), shown as 3 µm pixel averages, together with estimated

times for development of each type of profile assuming an initial step function (next section).

Resolution of [Ti] data at the effective spot size of 2 µm does not allow for accurate modeling of

diffusional boundaries thinner than ~ 16 μm, (corresponding to ~70 years diffusive relaxation at

790 °C) because 4-5 points are needed to define a diffusion sigmoid; however in each case the

[Ti] slope is less steep than this limit (Fig. 2.4). In most cases (Fig. 2.4, Table 2.3) the CL

grayscale profile is ‘sharper’, i.e. closer to a step function, than the [Ti] profile; in a few cases

they are indistinguishable within uncertainty, and in two cases the [Ti] profile is steeper than CL

grayscale, albeit still overlapping within uncertainty.

Fig 2.3. Trace element variations among Tshirege glasses; EG and MI in symbol key stand for external glass and melt inclusion, respectively. a. Th vs. Nb; b. La vs. Hf; c. Hf/La vs Rb. d. Rb/Sr vs Ba/La, with faceted inclusions identified – see text.

3, -

ansect can can ansect

lines), calculated from and (4) (5). calculated lines), eqns. e included for each sample. CL raw and averaged averaged and raw CL sample. for each e included

3. - 1) with modeled diffusion curves ([Ti]: solid black lines; CL: dashed grey grey dashed CL: lines; black solid ([Ti]: curves diffusion modeled with 1) 3, 88 - Titanium concentration and CL grayscale intensity profiles for five representative quartz crystals from the Tshirege (158 Tshirege the from crystals quartz representative five for profiles intensity grayscale CL and concentration Titanium 1, 119 - Fig. 2.4. 141 - tr full zone; relaxation diffusional the to cropped is data Transect size. symbol than smaller are points [Ti] for bars Error 17 for sample data compared be found in supplementary data. CL grayscale images and transect locations ar locations grayscale and besupplementary transect found data.images in CL

Estimation of temperature

In Figure 2.4, CL grayscale and [Ti] profiles are compared at a fixed T = 790 °C. We

attempt to additionally estimate diffusion times from [Ti] profiles using temperatures derived

from thermometry. Due to the generally high activation energies associated with cation diffusion

in silicate minerals, timescales calculated from diffusion profiles are very sensitive to

temperature (e.g. Cherniak et al., 2007). For example, a ΔT of ~50 °C, which is similar to the

error range (i.e. ±25 °C) arising from calibration and microprobe data uncertainty for many

mineral geothermometers, may result in an order of magnitude change in calculated relaxation

time (Table 2.3). It is therefore necessary to constrain temperature during diffusive relaxation as

closely as possible. The Tshirege Member contains magnetite and sporadic ilmenite; we did not

find any pairs that pass the equilibrium test of Bacon and Hirschmann (1988). The first-erupted parts of both the Otowi and Tshirege Members of the Bandelier Tuff (respectively, the Guaje and

Tsankawi Pumice Beds) have compositions that are close to haplogranite, are homogeneous, exhibit the least textural evidence for internal disequilibrium within their respective members, and plot very near the minimum in Q–Ab–Or at 2 kb water saturation pressure (Wilcock et al.,

2013; Wolff and Ramos, 2014), consistent with a magma temperature close to 700 °C (Tuttle and

Bowen, 1958) and the measured water contents of melt inclusions (Dunbar and Hervig, 1992).

Warshaw and Smith (1988) give a temperature of 697 °C for the Tsankawi magma, based on an early version of the QUILF thermometer (Andersen and Lindsley, 1988), but do not report uncertainties. Wilcock et al. (2013) review several lines of evidence for storage depth of the

Tshirege magma and conclude that a pressure of 200 ± 50 MPa is most appropriate. We concur and assume this pressure and the temperature of 700 °C for the Tsankawi Pumice to be robust.

The TitaniQ thermometer (Wark and Watson, 2006, equation 1 below; Thomas et al.,

2010, equation 2 below; Huang and Audétat, 2012, equation 3 below), which relies on the

temperature dependence of [Ti] in quartz, can be applied to quartz grains chosen for diffusion

modeling, thus avoiding the potential problem in applying temperatures estimated from other

mineral phases in high-silica rhyolites which may not be in equilibrium with quartz (Evans and

Bachmann, 2013; Evans et al., 2016):

(° ) 273 (1) −3765 . 𝑞𝑞𝑞𝑞𝑞𝑞 𝑋𝑋 𝑇𝑇 𝐶𝐶 𝑇𝑇𝑇𝑇 − log�𝑎𝑎 �−5 69 𝑇𝑇𝑇𝑇𝑂𝑂2

ln = (60952 ± 3122) + (1.520 ± 0.04) ( ) (1742 ± 63) ( ) + ln quartz TiO2 𝑇𝑇𝑇𝑇𝑂𝑂2 𝑅𝑅𝑅𝑅 𝑋𝑋 − ∗ 𝑇𝑇 𝐾𝐾 − ∗ 𝑃𝑃 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 𝑅𝑅𝑅𝑅 𝑎𝑎(2)

. log Ti(ppm) = 0.27943 660.53 + 5.6459 4 0 35 (3) 10 𝑃𝑃 𝑇𝑇 𝑇𝑇 In each case, the thermometer is− experimentally∗ − calibrated∗ for melts saturated with rutile,

hence a(TiO2)melt = 1; the chief difficulty in its application therefore lies in the estimation of

a(TiO2)melt because the vast majority of volcanic silicate liquids are not rutile-saturated. Hayden

and Watson (2007) provide a method for estimating a(TiO2)melt based on rutile saturation as a

function of melt composition and temperature; however this is only calibrated at 1 GPa, at which

pressure rutile solubility is likely less than at the mid- to upper-crustal storage depths of most

rhyolites (Thomas and Watson, 2012). Rhyolite-MELTS (Gualda et al., 2012) calculates a(TiO2)melt but the values for the exceptionally Ti-poor glasses of the Tshirege Member are

extremely low (~10-5), yielding unreasonably high temperatures.

Instead, we have calculated a(TiO2)melt for the Tshirege Member by ‘internally

calibrating’ to an assumed temperature of 700 °C for the first-erupted Tsankawi pumice, and

setting a(TiO2)melt to a value of 0.34, which yields 700 °C for the average compositions of

Tsankawi quartz (~22.5 ppm Ti) using Eqn. (3). The same result is obtained from the Huang and

Audétat (2012) calibration (Eqn. 5) using this a(TiO2)melt if P is set to 263 MPa. The Thomas et

al. (2010) calibration (Eqn. 4) requires a different a(TiO2)melt of 0.0775. The Tsankawi-derived a(TiO2)melt value is then treated in two ways. In the first approach, it is assumed constant

throughout the Tshirege Member high-silica rhyolite magma. This assumption can be justified on

the basis that titanomagnetite is a ubiquitous, if trace, phenocryst phase (Warren et al., 2007) and

hence may buffer a(TiO2)melt. However, it is unclear whether or not quartz and titanomagnetite

were co-precipitating phases, or whether titanomagnetite compositions or abundances may have

changed during quartz growth, an issue that has led to conflicting interpretations of temperature

estimates for the Bishop Tuff , a similar voluminous high-silica rhyolite eruption (Evans and

Bachmann, 2013; Gualda and Ghiorso, 2013; Evans et al., 2016; Jolles and Lange, 2019).

Therefore, when estimating a(TiO2)melt for the Tshirege Member magmas, buffering cannot be

assumed. Instead, the a(TiO2)melt value of 0.34 is used to calculate an effective TiO2 activity

coefficient for measured TiO2 in Tsankawi glasses, which is then applied to compositions of glasses in contact with quartz rims throughout the Tshirege Member. The assumption of constant

γ(TiO2)melt, where γ denotes the activity coefficient, is justified by the essentially constant major-

element composition of Tshirege high-silica rhyolite.

Modeled diffusion relaxation times (next section) are calculated using three different temperature assumptions: (i) constant temperature of 790 °C (Fig. 2.4); (ii) temperature estimated using the Wark and Watson (2006) and Huang and Audétat (2012) calibrations at 263

MPa, assuming constant a(TiO2)melt; (iii) temperature estimated using the Wark and Watson

(2006) and Huang and Audétat (2012) calibrations at 263 MPa, and applying calculated

γ(TiO2)melt to Ti contents of glasses adhering to quartz crystals to estimate a(TiO2)melt. In cases

(ii) and (iii), the temperature calculated from the rim composition of the quartz is assumed to

apply to diffusion at the inner boundary of the rim.

Modeled diffusion relaxation times

We use the 1-D finite difference diffusion method (Costa et al., 2008, equation 4), to

model Ti diffusion in quartz as a function of time and distance of travel in the lattice:

, , , , = , + (4) 𝐷𝐷∆𝑡𝑡�𝐶𝐶𝑖𝑖+1 𝑗𝑗 −2𝐶𝐶𝑖𝑖 𝑗𝑗 +𝐶𝐶𝑖𝑖−1 𝑗𝑗� 2 𝐶𝐶𝑖𝑖 𝑗𝑗+1 𝐶𝐶𝑖𝑖 𝑗𝑗 � ∆𝑥𝑥 � Where: , = the concentration at the current lattice location; , = the concentration at the

𝐶𝐶𝑖𝑖 𝑗𝑗 𝐶𝐶𝑖𝑖 𝑗𝑗+1 current lattice location one time step forward; , = the concentration one lattice step towards

𝐶𝐶𝑖𝑖+1 𝑗𝑗 the diffusion zone; , = the concentration one lattice step away from the diffusion zone; =

𝑖𝑖−1 𝑗𝑗 diffusion coefficient𝐶𝐶 for the diffusing species in lattice of interest; = change in time for on𝐷𝐷e step in the model; = step size. ∆𝑡𝑡

The diffusion∆𝑥𝑥 coefficient of Ti in quartz parallel to the c axis is given by (Cherniak et al.,

2007, equation 5):

-8 -1 2 -1 DTi = 7 x 10 exp(-273± 12kJ mol /RT) m sec (5)

Where: DTi = diffusion coefficient of Ti in quartz, R is the universal gas constant and T is

absolute temperature, with little anisotropy apparent from seven additional experiments

measuring Ti diffusion in quartz normal to c. Regression of the latter over the temperature range

750 – 1101 °C gives an Arrhenius relation that yields results indistinguishable from Eqn. (5),

which we use here. Errors for the CL and [Ti] profiles at constant temperatures are calculated

from the inherent error in the DTi which, when applied to these data, lead to large (40-60%)

errors of modeled times. In Table 2.3, uncertainties of diffusion times for the [Ti] profiles

calculated using temperatures from TitaniQ that arise from fitting sigmoids to the data are reported separately.

In most cases, Ti diffusive relaxation times calculated from CL grayscale profiles are shorter than those calculated from [Ti] data, by up to an order of magnitude (Table 2.3, Fig. 2.4).

In some cases, the two profiles yield identical results within error. Previous authors have presented discrepant CL grayscale and [Ti] data, consistent with our findings (e.g. Fig. 2.9 of

Matthews et al., 2012), but have discounted the significance for calculated timescales. We also note that, overall, grayscale profiles are smoother than those for [Ti] from the same crystal (Fig.

2.4), resulting in many cases in decreased uncertainties associated with calculated timescales

(Table 2.3). Furthermore, while spectroscopic measurements of quartz CL emissions at specific wavelengths, such as ~454 nm, are very well correlated with Ti content (MacRae et al., 2013,

2018), greyscale CL is not a reliable proxy for Ti concentration. For many panchromatic CL detectors, the collection optics accept light from a wide variety of angles, and therefore light from adjacent areas, which are still emitting from the initial electron bombardment, or light transmitted through the sample and scattered off of microcracks in the sample, which may be included in a CL measurement that is supposed to represent only the CL emitted from an individual pixel (MacRae et al., 2013). Also, as described by Leeman et al. (2012) and MacRae et al. (2018), the greyscale intensities include not only CL from Ti dopants, but also from the intrinsic CL produced by quartz, by crystallographic defects, by aluminum dopants, and by non- bridging oxygen holes in the crystal structure. In fact, certain CL emissions from quartz, such as those from non-bridging oxygen holes, are actually inversely correlated with Ti contents (see

Fig. 2.4 of Leeman et al., 2012). Therefore, the diffusion profiles used for the modeling

presented herein are based on the Ti concentrations measured by EPMA; profiles based on CL

intensity are not considered.

DISCUSSION

Melt inclusions occur as faceted and non-faceted forms. Faceting in melt inclusions occurs as an initially round inclusion attempts to acquire a negative crystal form, accomplished through lateral diffusion of silica along the edges of the inclusion (See Fig. 2.3 in Gualda et al.,

2012). The time which an inclusion takes to go from non-faceted to faceted is positively

correlated with the size of the inclusion and negatively correlated with temperature. Melt

inclusions in the Tshirege quartz vary from almost perfectly spherical to strongly faceted (Fig.

2.5). Inclusions have been assigned a faceting strength (FS) value from 0 = unfaceted, spherical

inclusion to 2 = strongly faceted, negative bipyramidal crystal shape (Fig. 2.5d). In crystals

which contain multiple facets with different FS values, higher FS values are found towards the

core of the crystal (Fig. 2.6.). Barium and Sr, compatible elements in rhyolite, are correlated with

the degree of faceting, where no melt inclusions above FS = 1 have [Ba] above ~15 ppm; FS =

<1 have [Ba] of 5-66 ppm, with unfaceted HSR-e inclusions showing the highest Ba and Sr

concentrations.

Highly incompatible elements (i.e. Nb and Th) show positive linear correlations among

all glasses (Fig. 2.3a). On most bivariate plots, glasses and MIs from HSR-e pumices plot

together while glasses and MIs from the HSR-d and LSR pumices plot together (Fig. 2.3b &c);

Fig. 2.5 a. Facet strength (FS) vs Ba concentration; explanation of FS given in text. b. Crystal containing melt inclusions with b. FS = 0 and 0.5. c. FS = 1 and 1.5. d. FS = 2. Note that not all crystals exhibit all FS values.

Fig. 2.6. Back-scatter electron image of a quartz crystal from the HSR-d pumice with four melt inclusions (light gray). White numbers indicate FS rating; each inclusion is labeled with Ba and Sr concentrations. Gray dashed line shows interpreted original crystal boundary, before regrowth event. Note elevated Ba and Sr in the low FS inclusions in the interpreted rim. This crystal has no CL zoning.

some HSR-d and LSR MIs plot with HSR-e external glasses and MIs. Figure 2.3-d shows Rb/Sr vs Ba/Nb and a field of strongly faceted melt inclusions. This field contains all melt inclusions with a FS ≥ 1 and none with FS < 1; generally, when faceted and unfaceted inclusions occur together in the same crystal, strongly faceted inclusions occur in the cores of crystals, where weakly or unfaceted inclusions occur in the rim. MIs plotting with external glasses from LSR and HSR-d pumice are only found in CL-bright rims and are unfaceted. The strongly faceted MIs have clearly formed from a move evolved liquid.

Faceted inclusions with a FS = 2 are 50-125 µm in radius, suggesting FS = 2 inclusions were trapped and stored in quartz on the order of at least >1,000-10,000 years (Gualda et al.,

2012) at an average storage temperature of ~750 °C. Once an inclusion is strongly faceted, no further changes occur, so these are minimum times. Some of the largest (r = 150 µm) faceted melt inclusions in the Tsankawi pumice (700 °C) may have taken on the order of 104 years to

develop.

Temperatures calculated using rim compositions of zoned quartz are invariably higher

than 700 °C; this result will hold regardless of what temperature is assumed for the Tsankawi

magma. Excluding dacite 88-1, the average temperature for quartz rims in pumice clasts with 64

– 73 ppm Nb is 774 ± 14 °C (a(TiO2) method) or 740 ± 17 °C (γ(TiO2) method). This reinforces

the qualitative value of the TitaniQ thermometer, but the difficulties associated with a(TiO2)

estimation somewhat negate its quantitative value as a thermometer applicable to rhyolites (see

also Wilson et al., 2012); the majority of rhyolites with high magmatic water contents erupt at

temperatures between 700 and 800 °C (Ellis et al., 2013), so the temperature estimates in Table

2.3 are reasonable but hardly surprising.

Sample 88-1 is a dacite pumice containing quartz texturally indistinguishable from that in

the high-silica rhyolite (Fig. 2.1); the quartz is thought to have been incorporated as dacite mixed

with rhyolite immediately prior to the eruption (Stimac, 1996; Boro, 2019). The quartz rim

compositions yield temperatures 799-825 °C. The dacite contains Fe-Ti oxides with equilibrium

compositions that yield similar temperatures of 780-802 ± 10 °C (Boro, 2019).

Despite the higher temperature, dacite 88-1 produces the longest modeled Ti diffusion times (see 790 °C Ti model-time column, Table 2.3). The Ti diffusion results strongly suggest that an event or events involving reheating of the system occurred over a range between 100 and

1,000 years (order of magnitude) prior to eruption, but this preceded recharge by the dacite, which itself may have acted as the trigger for the eruption (Stimac, 1996; Goff et al., 2014; Boro,

2019) on a much shorter timescale. Some quartz crystals (Fig. 2.1) have rounded edges and appear to have gone through an additional resorption event, truncating existing zoning patterns, without subsequent growth of new quartz prior to the eruption, so the relevance of the calculated times in Table 2.3 to the sequence of events cascading towards the caldera-forming eruption is unclear.

IMPLICATIONS

Compositional zoning of the Tshirege Member magma

Most quartz cores from HSR-d pumice have near-constant and low Ti contents,

indistinguishable from non-zoned quartz in HSR-e Tsankawi pumice. The MIs hosted by these

cores are typically faceted and have trace element abundances identical to HSR-e pumice,

whereas the occasional MIs in the Ti-rich rims of quartz from HSR-d pumice have HSR-d chemistry and are non- or weakly faceted (Fig. 2.6). These observations link overall bulk zoning

in the Tshirege Member to melt inclusion chemistry (Figs. 2, 3), quartz growth, and quartz

residence time. The most evolved, first-erupted melt compositions (HSR-e chemistry) and low-

Ti quartz existed before the less evolved, late-erupted melt (HSR-d and LSR chemistries) and

high-Ti quartz developed in the erupted portion of the system. This is consistent with a simple model of crystallization, cumulate formation, and subsequent melting and remobilization of the cumulate to create a compositionally zoned magma body (Wolff et al., 2015). In this model, which is an extension of the ‘mush model’ (Bachmann and Bergantz, 2004; Hildreth, 2004) the quartz-feldspar cumulate pile beneath a cognate crystal-poor melt lens is heated and remobilized by recharge, with little mass contribution from the recharge magma (Wark et al., 2007; Wolff and Ramos, 2014). The result is a relatively water-poor rhyolite, of higher density and with lower incompatible element concentrations than the initial rhyolite melt lens. The new melt, with a cargo of crystals inherited from the cumulate pile, pools beneath the initial melt lens to form a thicker, compositionally zoned body of eruptible magma. The lack of quartz zoning in HSR-e samples suggests an enriched liquid lens that was shut off from the thermal and chemical effects of invading recharge magmas, likely shielded by a crystal mush which was melted and mobilized to form the HSR-d magma.

Figure 2.7 depicts four stages in the development of zoning in the Tshirege system:

a. Initial state of the system is a crystal mush overlain by a lens of magma, which is

highly enriched in incompatible trace elements, and depleted in compatible trace

elements – especially those which strongly partition into sanidine (i.e. Sr and Ba).

ents. b.

more primitive magma starts to to starts magma primitive more lt inclusion chemistry for the Tshirege magmas. magmas. Tshirege the for chemistry lt inclusion on event become faceted. c. become on event

her concentrations of compatible trace elements (i.e. Ba and Sr). This step step This Sr). and (i.e. Ba elements trace compatible of her concentrations 10,000 years, meltinitial fromaccumulati inclusions the - A simplified four stage model for the development of and simplifiedme thezoning A the stagemodel for four CL development after ~1,000 Fig. 2.7. are in quartz compatible depleted trace in elem inclusions ofStage occurs withwhich melt trapping one,accumulation crystal pointeruption. ofthe overto continues the system pressurizes until during continued recharge, quartz grows at higher temperatures and recharge system,whichresorption point the atquartzhigher temperatures and occurs. d.continued during recharge,at of quartz grows melt with inclusions trappinghig diffusion profiles develop,

This system is producing quartz crystals which trap melt from the evolved body and

settle to contribute to the mush.

b. The mush is around 70% crystallinity with a bulk low water content (~1.5 wt. %

H2O) and consists of crystals with melt inclusions which were derived from the

overriding liquid. These inclusions were stored for 1,000-10,000 years allowing for

strong negative crystal facets to develop.

c. Underplating and incorporation of recharge magma into the lower portions the

crystal mush column acts as a heating element and may add water to the system from

second boiling which eventually starts to loosen the rigid mechanically locked mush

(Boro, 2019), causing quartz crystals to resorb. During this time, partial melting of

feldspar adds Ba and Sr to the melt. Crystals in the HSR-e continue to grow and

capture the recharge event with unfaceted melt inclusions which are enriched in Ba

and Sr.

d. Quartz crystals regrow as they find a new equilibrium at higher temperatures forming

CL-bright rims with higher [Ti]. Continued recharge of the system mobilizes the

mush and overpressure triggers the eruption; from diffusion modeling this step took

on the order of 100-1,000 years.

Eruption forecasting

An important application of petrologic study of past eruptions is to draw links between events recorded in crystals and melts, particularly those interpreted as destabilizing to the system, and monitoring signals that might be detected at the surface in the run-up to a future eruption (Blundy and Cashman, 2008). Recent diffusion modeling studies (e.g. Gualda and

Sutton, 2016; Cooper et al., 2017) have emphasized short timescales of crystal regrowth and

residence following recharge events prior to a supereruption. In the case of the Tshirege

Member, the events recorded in quartz imply disturbances over ~103 years prior to eruption, but

likely preceded the intrusion of dacite magma into the system that was the probable immediate

trigger for the eruption (Goff et al., 2014; Boro, 2019). If recharge events can be detected at the

surface (e.g. inflation, increased gas flux), it is unclear how the critical eruption trigger could be

identified, but a series of events over a long term might indicate destabilization of a static magma

body. In this context, both long- and short-term preparation is needed to mitigate the effects of a

VEI 7 or greater eruption (Newhall et al., 2018). The accumulation of accurate estimates of crystal growth and residence times from many different systems may help inform these efforts in areas at risk of a large eruption.

REFERENCES:

Anderson, D.J., Lindsley, D.H. (1988) Internally consistent solution models for Fe-Mg-Mn-Ti

oxides; Fe-Ti oxides. American Mineralogist 73, 714-726.

Bachmann, O., Bergantz, G.W. (2004) On the Origin of Crystal-poor Rhyolites: Extracted from

Batholithic Crystal Mushes. Journal of Petrology 45-8, 1565-1582.

Bachmann, O., Bergantz, G.W. (2006) Gas percolation in upper-crustal silicic crystal mushes as

a mechanism for upward heat advection and rejuvenation of near-solidus magma bodies.

Journal of Volcanology and Geothermal Research, 149, 85-102.

Bacon, C. R., Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between

coexisting Fe-Ti oxides. American Mineralogist 73, 57-61.

Bacon, C.R. and Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between

coexisting Fe-Ti oxides. American Mineralogist, 73, 57–61.

Bacon, C.R., Newman, S., Stopler, E.M. (1992) Water, CO2, Cl, and F in melt inclusions in

phenocrysts from three Holocene explosive eruptions, Crater Lake, Oregon. American

Mineralogist, 77(9-10), 1021-1030.

Bailey, R.A., Smith, R.L. and Ross, C.S. (1969) Stratigraphic nomenclature of volcanic rocks in

the Jemez Mountains, New Mexico. U.S. Geological Survey Bulletin, 1274-P, 1–19.

Balsley, S.D. (1988) The petrology and geochemistry of the Tshirege Member of the Bandelier

Tuff, Jemez Mountains volcanic field, New Mexico, U.S.A. M.S. thesis, University of

Texas at Arlington, 188 p.

Blake, S., Ivey, G.N. (1986a) Magma-mixing and the dynamics of withdrawal from stratified

reservoirs. Journal of Volcanology and Geothermal Research, 27, 153-178.

Blake, S., Ivey, G.N. (1986b) Density and viscosity gradients in zoned magma chambers, and

their influence withdrawal dynamics. Journal of Volcanology and Geothermal Research,

30, 201-230.

Blundy, J., Cashman, K. (2008) Petrologic Reconstruction of Magmatic System Variables and

Processes. Reviews in Mineralogy and Geochemistry 69, 179-239.

Boro, J.R. (2019) Recharge and mobilization of crystal mush to produce and erupt a zoned

magma chamber – the Tshirege Member of the Bandelier Tuff, Valles caldera, New

Mexico, USA: Ph.D. dissertation, Washington State University.

Branney, M.J. and Kokelaar, P. (2002) Pyroclastic Density Currents and the Sedimentation of

Ignimbrites. Geological Society. Geological Society Memoir, 27, 143 pp.

Burgisser, A., Bergantz, G.W. (2011). A rapid mechanism to remobilize and homogenize highly

crystalline magma bodies. Nature, 471, 212-215.

Campbell, M.E., Hanson, J.B., Minarik, W.G., and Stix, J. (2009) Thermal history of the

Bandelier magmatic system: evidence for magmatic injection and recharge at 1.61 Ma as

revealed by cathodoluminescence and titanium geothermometry. Journal of Geology,

117, 469–485.

Caress, M.E. (1996) Zonation of alkali feldspar compositions in the Tshirege Member of the

Bandelier Tuff in Pueblo Canyon, near Los Alamos, New Mexico. New Mexico

Geological Society, 47th Field Conference Guidebook, 275–283.

Cherniak, D.J., Watson, E.B, and Wark., D.A. (2007) Ti diffusion in quartz. Chemical Geology,

236, 65–74.

Cooper, G.F., Morgan, D.J., and Wilson, C.J.N. (2017) Rapid assembly and rejuvenation of a

large silicic magmatic system: insights from mineral diffusive profiles in the Kidnappers

and Rocky Hill deposits, New Zealand. Earth and Planetary Science Letters, 473, 1–13.

Costa, F. and Dungan M.A. (2005) Short timescales of magmatic assimilation from diffusion

modelling of multiple elements in olivine. Geology, 33, 837–840.

Costa, F., Dohmen, R., and Chakraborty, S. (2008) Time scales of magmatic processes from

modeling the zoning patterns of crystals. Reviews in Mineralogy and Geochemistry, 69,

545–594.

Donovan, J.J., Lowers, H.A., and Rusk, B.G. (2011) Improved electron probe microanalysis of

trace elements in quartz. American Mineralogist, 96, 274–282.

Donovan, J.J., Singer, J.W. and Armstrong, J.T. (2016). A new EPMA method for fast trace

element analysis in simple matrices. American Mineralogist, 101, 1839–1853.

Donovan, J.J., Tingle, T.N., 1996. An improved mean atomic number background correction for

quantitative microanalysis. Journal of Microscopy Society of America. 2, 1-7.

Dunbar, N.W and Hervig, R.L. (1992) Volatile and trace element composition of melt inclusions

from the lower Bandelier Tuff; implications for magma chamber processes and eruptive

style. Journal of Geophysical Research, 97, 15,151-15,170.

Ellis, B.S., Wolff, J.A., Boroughs, S., Mark, D.F., Starkel, W.A., Bonnichesen, B. (2013)

Rhyolitic volcanism of the central Snake River Plain: a review. Bulletin of Volcanology

75:745.

Evans, B.W. and Bachmann, O. (2013) Implications of equilibrium and disequilibrium among

crystal phases in the Bishop Tuff. American Mineralogist, 98, 271–274.

Evans, B.W., Hildreth, W., Bachmann, O., and Scaillet, B. (2016) In defense of magnetite-

ilmenite thermometry in the Bishop Tuff and its implication for gradients in silicic

magma reservoirs. American Mineralogist, 101, 469–482.

Fournelle, J.F. (2007) The Problem of Secondary Fluorescence in EPMA in the Application of

the Ti-in-Zircon Geothermometer and the Utility of PENEPMA Monte Carlo Program.

Microscopy and Microanalysis, 13(S02), 1390-1391.

Gardner, J.N., Goff, F., Kelley, S.A. and Jacobs, E. (2010) Rhyolites and associated deposits of

the Valles-Toledo caldera complex. New Mexico Geology, 32, 3–18.

Goff, F., Gardner, J.N., Reneau, S.L., Kelley, S.A., Kempter, K.A., and Lawrence, J.R. (2011)

Geologic map of the Valles caldera, Jemez Mountains, New Mexico. New Mexico

Bureau of Geology and Mineral Resources Geologic Map, 79, scale 1:50,000.

Goff, F., Warren, R.G., Goff, C.J., and Dunbar, N. (2014) Eruption of reverse-zoned upper

Tshirege member, Bandelier Tuff from centralized vents within Valles caldera, New

Mexico. Journal of Volcanology and Geothermal Research, 276, 82–104.

Götze, J. (2012) Application of cathodoluminescence microscopy and spectroscopy in

geosciences. Microscopy and Microanalysis, 18, 1270–1284.

Gualda, A.R.G, Sutton, S.R. (2016) The Year Leading to a Supereruption. PLoS ONE 11, 18pp.

Gualda, G.A.R, Pamukcu, A.S., Ghiorso, M.S., Anderson, A.T Jr., Sutton, S.R., Rivers, M.L.

(2012) Timescales of Quartz Crystallization and the Longevity of the Bishop Giant

Magma Body. PLoS ONE 7(5).

Gualda, G.A.R. and Ghiorso, M.S. (2013) The Bishop Tuff giant magma body: an alternative to

the standard model. Contributions to Mineralogy and Petrology, 166, 755–775.

Hayden, L.A. and Watson, E.B. (2007) Rutile saturation in hydrous siliceous melts and its

bearing on Ti-thermometry of quartz and zircon. Earth and Planetary Science Letters,

258, 561–568.

Hildreth, W. (2004) Volcanological perspectives on Long Valley, Mammoth Mountain, and

Mono Craters: several contiguous but discrete systems. Journal of Volcanology and

Geothermal Research 136, 169-198.

Huang, R. and Audétat, A. (2012) The titanium-in-quartz (TitaniQ) thermombarometer: a critical

examination and re-calibration. Geochimica et Cosmochimica Acta, 84, 75–89.

Kidder, S., J.-P. Avouac, and Y.-C. Chan (2013), Application of titanium-in-quartz

thermobarometry to greenschist facies veins and recrystallized quartzites in the

Hsuehshan range, Taiwan, Solid Earth, 4(1), 1–21, doi:10.5194/se-4-1-2013

Kohn MJ, Northrup CJ. 2009. Taking mylonites’ temperature. Geology 37(1), 47–50

Leeman, W.P., MacRae, C.M., Wilson, N.C., Torpy, A., Lee, C.-T.A., Student, J.J., Thomas,

J.B., and Vicenzi, E/P. (2012) A study of cathodoluminescence and trace element

compositional zoning in natural quartz from volcanic rocks: mapping titanium content in

quartz. Microscopy and Microanalysis, 18, 1322–1341.

Llovet, X., and Salvat, F. (2016) PENEPMA: a Monte Carlo programme for the simulation of X-

ray emission in EPMA. IOP Conf. Ser.: Mater. Sci. Eng.109012009

Lowenstern, J.B., Mahood, G.A. (1991) New data on magmatic H2O contents of pantellerites,

with implications for petrogenesis and eruptive dynamics at Pantelleria. Bulletin of

Volcanology 57-1, 78-83.

MacRae, C.M., Wilson, N.C., Torpy, A. (2013) Hyperspectral cathodoluminescence. Mineralogy

& Petrology 107, 429-440. Doi: 10.1007/s00710-013-0272-8

MacRae, C.M., Wilson, N.C., Torpy, A., Delle Piane, C. (2018) Soft X-ray and

cathodoluminescence measurement, optimization and analysis at liquid nitrogen

temperatures. IOP Conference Series: Materials Science and Engineering, Volume 304,

Issue 1, pp. 012010 (2018).

Matthews, N.E., Huber, C., Pyle, D.M., Smith, V.V. (2012) Timescales of magma recharge of

large silicic systems from Ti diffusion in quartz. Journal of Petrology, 53, 1385–1416.

Morgan, D.J. and Blake, S. (2006) Magmatic residence times of zoned phenocrysts: introduction

and application of the binary element diffusion modeling (BEDM) technique.

Contributions to Mineralogy and Petrology, 151, 58–70.

Newhall, C., Self, S., Robock, A. (2018). Anticipating future Volcanic Explosivity Index (VEI) 7

eruptions and their chilling impacts. Geosphere, 14(2), 572-603.

Pamukcu, A.S., Ghiorso, M.S., Gualda, G.A. (2016) High-Ti, bright-CL rims in volcanic quartz:

a result of very rapid growth. Contributions to Mineralogy and Petrology 172, 9pp.

Peppard, B.T, Steele, I.M., Davis, A.M., Wallace, P.J., Anderson, A.T. (2001) Zoned quartz

phenocrysts from the rhyolitic Bishop Tuff. American Mineralogist 86(9), 1034-1052.

Phillips, E.H., Goff, F., Kyle, P.R., McIntosh, W.C., Dunbar, N.W., and Gardner, J.N. (2007)

The 40Ar/39Ar age constraints on the duration of resurgence at the Valles caldera, New

Mexico. Journal of Geophysical Research, 112, B08201, 15 p.

Seitz, S., Putlitz, B., Baumgartner, L.P., Escrig, S., Meibom, A., and Bouvier, A.-S. (2016) Short

magmatic residence times of quartz phenocrysts in Patagonian rhyolites associated with

Gondwana breakup. Geology, 44, 67–70.

Self, S., Goff, F., Gardner, J.N., Wright, J.V. and Kite, W. (1986) Explosive rhyolitic volcanism

in the Jemez Mountains: Vent locations, caldera development and relation to regional

structure. Journal of Geophysical Research, 91, 1779–1798.

Self, S., Heiken, G., Sykes, M.L., Wohletz, K., Fisher, R.V., and Dethier, D.P. (1996) Field

excursions to the Jemez Mountains, New Mexico. Bulletin of the New Mexico Bureau of

Geology and Mineral Resources, 134, 72 pp.

Shane, P., Smith, V.C., Narin, I. (2008) Millennial timescale resolution of rhyolite magma

recharge at Tarawera volcano: insights from quartz chemistry and melt inclusions.

Contributions to Mineralogy and Petrology 156, 397-411.

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, 100, 2026–2042.

100

Smith, R.L. and Bailey, R.A. (1966) The Bandelier Tuff - a study of ash-flow eruption cycles

from zoned magma chambers. Bulletin Volcanologique, 29, 83–104.

Stimac, J.A. (1996) Hornblende-dacite pumice in the Tshirege member of the Bandelier Tuff:

Implications for magma chamber and eruptive processes. New Mexico Geological

Society, 47th Field Conference Guidebook, 269–274.

Thomas, J.B. and Watson, E.B. (2012) Application of the Ti-in-quartz thermobarometer to rutile-

free systems. Reply to: a comment on: ‘TitaniQ under pressure: the effect of pressure and

temperature on the solubility of Ti in quartz’ by Thomas et al. Contributions to

Mineralogy and Petrology, 164, 369–374.

Thomas, J.B., Watson, E.B., Spear, F.S., Shemella, P.T., Nayak, S.K., and Lanzirotti, A. (2010)

TitaniQ under pressure: the effect of pressure and temperature on the solubility of Ti in

quartz. Contributions to Mineralogy and Petrology, 160, 743–759.

Till, C.B., Vasqez, J.A. and Boyce, J.W. (2015) Months between rejuvenation and volcanic

eruption at Yellowstone caldera, Wyoming. Geology, 43, 695–698.

Torres, R.C., Self, S. and Martinez, M.M.L. (1996) Secondary pyroclastic flows from the June

15, 1991, ignimbrite of Mount Pinatubo. In C.G. Newhall, R.S. Punongbayan, eds., Fire

and Mud: eruptions and lahars of Mount Pinatubo, Philippines. Philippine Institute of

Volcanology and Seismology and University of Washington Press, pp. 665–678.

Trial, A.F., Spera, F.J., Greer, J. and Yuen, D.A. (1992) Simulations of magma withdrawal from

compositionally zoned bodies. Journal of Geophysical Research, 97, 6713–6733.

101

Tuttle, O.F. and Bowen, N.L. (1958) Origin of granite in the light of experimental studies in the

system NaAlSi3O8-KAlSi3O8-SiO2-H2O. Geological Society of America Memoir, 74,

153 pp.

Wark, D.A. and Watson, E.B. (2006) TitaniQ: a titanium-in-quartz geothermometer.

Contributions to Mineralogy and Petrology, 152, 743–754.

Wark, D.A. and Wolff, J.A. (2006) Mafic recharge eruption trigger for the Bandelier and Bishop

Tuffs. EOS, Transactions American Geophysical Union, 87, abs. V33C-0682.

Wark, D.A., Hildreth, W., Spear, F.S., Cherniak, D.J., and Watson, E.B. (2007) Pre-eruption

recharge of the Bishop magma system. Geology, 35, 235–238.

Warren, R.G., Goff, F., Kluk, E.C., and Budahn, J.R. (2007) Petrography, chemistry, and mineral

compositions for subunits of the Tshirege Member, Bandelier Tuff, within the Valles

caldera and Pajarito Plateau: New Mexico Geological Society, 58th Field Conference

Guidebook, 316–332.

Warshaw, C.M. and Smith, R.L. (1988) Pyroxenes and fayalites in the Bandelier Tuff, New

Mexico: temperatures and comparison with other rhyolites. American Mineralogist, 73,

1025–1037.

Wilcock, J., Goff, F., Minarik, W.G., and Stix, J. (2013) Magmatic recharge during the formation

and resurgence of the Valles caldera, New Mexico, USA: evidence from quartz

compositional zoning and geothermometry. Journal of Petrology, 54, 635–664.

Wilson, C.J.N., Seward, T.M., Allan, A.S.R., Charlier, B.L.A. and Bello, L. (2012) A comment

on: ‘TitaniQ under pressure: the effect of pressure and temperature on the solubility of Ti

102

in quartz’, by Jay B. Thomas, E. Bruce Watson, Frank S. Spear, Philip Shemella, Saroj K.

Nayak and Antonio Lanzirotti. Contributions to Mineralogy and Petrology, 164, 359–

368.

Wolff, J.A. and Ramos, F.C. (2014) Processes in caldera-forming high-silica rhyolite magma:

Rb-Sr and Pb isotope systematics of the Otowi Member of the Bandelier Tuff, Valles

Caldera, New Mexico, USA. Journal of Petrology, 55, 345–375.

Wolff, J.A. and Thompson, R.A. (in press) The Jemez Mountains and Valles caldera. In:

Thompson, R.A., Turner, K., Lipman, P.W., Wolff, J.A., Dungan, M.A. and Bachmann,

O., Field guides to the Rio Grande rift, Jemez Mountains, and San Juan volcanic field,

New Mexico and Colorado. U.S. Geological Survey Scientific Investigations Report,

5022-R, in press.

Wolff, J.A., Ellis, B.S., Ramos, F.C., Starkel, W.A., Boroughs, S., Olin, P.H. and Bachmann, O.

(2015) Remelting of cumulates as a process for producing chemical zoning in silicic

tuffs: a comparison of cool, wet and hot, dry rhyolitic magma systems. Lithos, 236–237,

275–286.

103

CHAPTER FOUR: PETROGENESIS OF THE TSHIREGE MAGMA BODY

ABSTRACT:

VEI7+ eruptions pose a significant threat to humanity and understanding the processes

which operate to prime one of these systems for eruption is imperative as a society. The 1.26 Ma

eruption of the Tshirege Member of the Bandelier Tuff serves as a perfect window into a system

which produced once such eruption. Chemical and mineral variations of this system indicate that

a rhyolitic magma body was fractionally crystalized to produce a crystal mush pile. The

fractionating system changed over time and evolved to more water-rich compositions capable of supporting biotite and amphibole. The evolution of the system is tracked in faceted quartz-hosted

melt inclusions. Addition of water from recharging magmas caused resorption and regrowth and

mobilized the crystal mush incrementally prior to the eruption. This mobilization event is

tracked in unfaceted melt inclusions. This created a temperature gradient of ~100 °C at a

pressure of ~0.18-0.20 GPa.

INTRODUCTION:

Caldera volcanoes have long served as natural laboratories for studying the processes

which produce some of the most violent and environmentally devastating geologic events in

history (Newhall et al., 2018). Volcanic deposits produced by calderas often have chemical and

mineral heterogeneities and can be described as ‘Bishop-type’ zoned tuffs, which pays homage

to the landmark paper by Hildreth (1979) (Wolff et al., 2015). The zonation in these tuffs is

assumed to be inverse representations of the magma chambers which produced them, and the

104

chemistry of those magma chambers has been modified and heavily influenced by crystal

fractionation (Cameron, 1984).

The idea of crystals and the fractionated left-over liquids controlling the chemical zoning

in volcanic deposits was slowly modified into ideas such as boundary layer crystallization and

side-wall transport with convection processes (e.g. Brandeis and Jaupart, 1987; Tait & Jaupart,

1992); a paradigm shift in thinking occurred in the early 2000’s with the development of the

magma-mush model (Bachmann and Bergantz, 2004; Hildreth, 2004). This model invokes

convective heat loss and crystallization until the magma chamber mechanically locks due crystal-

edge interferences, at which point compaction and settling of crystals under the influence of

gravity and liquid expulsion produces a large, crystal-poor, semi-homogenous melt body with an underlying crystal-mush pile. Further fractional crystallization and crystal settling has been the assumed mechanism for chemical zonation in these systems.

Active systems which contain mush piles probably exist near the granite minimum for as long as the temperature is high enough to maintain an interstitial liquid, thus a small amount of heat flux or disturbance in equilibrium (i.e. adding H2O) can cause rapid resorption of crystals

(Rubin et al., 2017). Melting these cumulate piles with recharge magmas acting as a ‘heating

element’ has been proposed to be a possible mechanism to produce the chemical zonation

present in many ‘Bishop-type’ tuffs (Wolff et al, 2015). In this model, suggested remelting of

cumulates produces a denser zone of liquid which is chemically related to the overlying liquid

lens which produced the proceeding crystals. The Tshirege Member of the Bandelier Tuff,

Valles caldera, New Mexico, has long been recognized as a zoned ignimbrite (Smith and Bailey,

1966) and hence serves as a test case for the model.

105

Long-lived or high-heat-flux volcanically active terrains capable of producing highly-

evolved silicic volcanic centers that culminate in caldera-forming eruptions (Karakas et al.,

2017) have recently been discussed as the result of a large, deep-seated ‘magma-distillation piles’ (Ganne et al., 2018). In this model it is assumed there are multiple distinct storage regimes, likely controlled by a density stratification in the crust; each at different stages in evolution and each with their own crystal mush piles. The latest erupted member of the Valles caldera contains a more mafic dacitic component which is thought to represent an invading recharge magma, the source of which is thought to be deeper seated in a lower-crustal reservoir

(Chapter 1). Similar quenched mafic magmas seem to be ubiquitous in the root systems of felsic magmatic terrains (Vernon, 1984; Hildreth and Wilson, 2007)

This study examines chemistry and petrography of the Tshirege Member of the Bandelier

Tuff and tests the model of remobilization of crystal mushes as a mechanism for producing a chemically zoned magma chamber, as well as utilizing glass and melt-inclusion chemistry to constrain evolution of the system. Using this information, inferences are made on the petrogenetic processes which occur prior to mush generation. Additionally, we will use these data to make interpretations about the crystal mush pile structure and make inferences about the plumbing systems which produce caldera volcanoes.

GEOLOGIC SETTING:

The Jemez Mountains Volcanic Field (JMVF) in North-central New Mexico, USA houses

the Valles Caldera (Type location resurgent calderas, Smith and Bailey, 1968), which has served

as a fruitful locality for studying caldera volcanism over the past half a century, and thus has an

106

accompanying large set of literature on the geology, stratigraphy, geochronology, and petrology

of the Bandelier Tuff (e.g. Campbell et al., 2009; Gardner et al., 1986, 2010; Heiken et al., 1986;

Self et al., 1986, 1996; Goff et al., 1989, 1990, 2011, 2014; Jie et al., 2019; Kelley et al., 2013;

Phillips et al., 2007; Rowe et al., 2007; Smith and Bailey, 1966, 1968; Smith et al., 1970; Stix et

al., 1988; Wilcock et al., 2013; Wolff et al., 1999, 2002, 2005; Wolff and Ramos, 2003, 2014).

Jemez Mountains Volcanic Field:

The Española basin is one of many N-S trending Cenozoic en-echelon sedimentary basins that define the Rio Grande Rift. The JMVF resides here and accents the location where the rift intersects the Jemez lineament, which itself is a northeast-southwest trending line of Cenozoic volcanics that represents crustal weakness leftover from a Proterozoic suture between the

Southern Yavapai and Mazatzal provinces (Shaw and Karlstrom, 1999; Karlstrom et al., 2002;

Magnani et al., 2004). Phanerozoic sedimentary rocks and Proterozoic granitoids and metavolcanic amphibolites (1.62-1.44 Ga) make up the basement substrate, which the JMVF rests on (Eichelberger and Koch, 1979; Brookins & Laughlin, 1983; Laughlin et al., 1983).

The Valles caldera represents the climax to a long history of mafic, intermediate and silicic volcanism extending sporadically back to the early Miocene, with JMVF edifice construction underway by about 10 Ma (Kelley et al., 2013; Woldegabriel et al., 2013). The principal pre- caldera formations making up the JMVF edifice are the Lobato Formation (basalt, NE JMVF,

~13.5-9.5 Ma) (dominantly andesite and trachyandesite, central to S JMVF, 10-7 Ma); the La

Grulla Formation (dominantly andesite, NW JMVF, ~8-7 Ma); the Bearhead Rhyolite (N-S

JMVF, 7-6 Ma); the Tschicoma Formation (dominantly dacite, NE JMVF, 5.5-2 Ma); and the flanking Cerros del Rio and El Alto Basalts (SE and NE JMVF, ~3-1.5 Ma)..Overall magma

107

3 production rates are low (~0.2-0.6 km per ka) compared to other regions that have produced

large, silicic volcanic systems (Wolff and Thompson, in press).

Bandelier tuff and caldera development:

The Valles caldera resulted from two ignimbrite-forming eruptions that produced the Otowi

(1.60 ± 0.02 Ma, Wolff and Ramos, 2014) and Tshirege (1.256 ± 0.010 Ma, Phillips et al., 2007)

Members of the Bandelier Tuff; the resulting calderas from each eruption are nested. The younger Valles Caldera (Tshirege) overprints most of previous Toledo caldera (Otowi), which is preserved in the north-west and north walls of the current structure (Self et al., 1986; Goff et al.,

2011). The ~ 400 km3 Otowi Member (Cook et al., 2016) consists exclusively of crystal-poor (~5

– 25%), highly evolved high-silica rhyolite with strong zonation in trace elements, and represents melt that separated from a large volume of alkali feldspar + quartz crystalline residue (Wolff and

Ramos, 2014; Wolff et al., 2015). The Valle Toledo Member rhyolites, a series of relatively small rhyolitic domes and pumice fallout deposits, were emplaced during the 350 ka between the two major eruptions (Gardner et al., 2010; Goff et al., 2011; Heiken et al., 1986; Jacobs et al.,

2016).

Tshirege Member:

The ~400 km3 Tshirege Member is also compositionally zoned (Goff et al., 2014; Self et al., 1996; Smith and Bailey, 1966; Wilcock et al., 2013). It consists of a plinian fallout, the

Tsankawi Pumice Bed (Qbts), followed by a series of ignimbrite units (Qbt1-5, Figure 3.1; nomenclature of Warren et al., 2007), and exhibits a broader compositional range than the Otowi

Member, from high-silica rhyolite (HSR) to low-silica rhyolite (LSR) with a minor (~1% by volume) dacite component. The LSR pumice is absent from the Tsankawi and lower flow units;

108

it has overall lower concentrations of incompatible elements (e.g. Rb, Nb, Th, U) and higher compatible trace elements (e.g. V, Ba, Sr) compared to the HSR. Upper flow units become progressively more densely welded and thermally altered. Individual fiammé within the welded zones contain crystals which are chemically and morphologically akin to those found in the non- altered LSR and HSR pumices. Crystal content of pumice increases upwards from ~10% in the

Tsankawi to <45% in Qbt4 LSR fiammé; LSR pumice contain characteristic large crystal clots and lenses of sanidine (0.1-2 cm - long axis) plus ortho- and clinopyroxene, with total crystallinities of 25-45% (this chapter). Some large sanidine clots appear to be aggregate glomerocrsts, while others are partly digested crystals which are optically continuous. Such increases in crystallinity with eruption progress in silicic tuffs have been inferred to imply, by extrapolation, the existence a non-erupted volume of complementary crystal mush, both specifically for the Bandelier Tuff (Wolff et al., 2015), the similar Bishop Tuff (Hildreth, 2004;

Hildreth and Wilson, 2007) and more generally (Bachmann and Bergantz, 2004, 2008).

Tshirege Dacite Pumice:

A complementary minor hornblende dacite pumice is found in all but the lower half of the Qbts and represent a recharge magma which likely triggered the Tshirege ignimbrite-forming eruption (Stimac, 1996; Chapter 1). Clasts are generally small (<1 to 15 cm), periodically bearing a rhyolitic pumice carapace and more rarely exhibiting mingling and mixing textures in hand sample.

109

Fig. 3.1. Map showing the location of the Valles Caldera with mapped Bandelier Tuff in orange (Gardner et al., 2010); sample locations for this study labeled with red dots. The inset regional map is modified from Goff and Gardner (2004) with the Rio Grande Rift (Tan), Jemez Lineament (red line), proximal Cenozoic volcanic fields (light red areas), the JMVF in orange, and the Valles caldera (dashed line) labeled. Stratigraphic column in lower left is modified from Goff et al. (2011) using unit names from Warren et al. (2007). The average whole-pumice Nb contents are also labeled for each unit (this study).

110

The dacite pumices without mixing or mingling textures are grey to grey-green in hand sample, given color by abundant hornblende phenocrysts and microlites. They are very finely

(<<1mm) vesicular with vesicularity averaging 70% with melt fractions of ~75%. Pumices with mixing and mingling features are most easily recognized by having large (<1 cm) fritted feldspars, which formed from rapid growth during undercooling of the dacite while it interacted with the rhyolite prior to the eruption.

METHODS:

Whole pumice clasts were analyzed for major, minor and trace elements in the Peter

Hooper GeoAnalytical Lab at Washington State University by X-ray fluorescence and

inductively coupled plasma mass spectrometry (Methods and procedures described at

https://environment.wsu.edu/facilities/geoanalytical-lab/technical-notes/). Mineral phases and glasses were analyzed for major and minor elements using a JEOL 8500-F field emission microprobe at Washington State University, using a range of voltages and currents with varying beam sizes (1-10 µm) depending on size constraints of phases. Full conditions and chemistries can be found in the electronic appendix.

LA-ICPMS data for melt inclusions and adhered glass was collected on c-axis oriented grain mounts of quartz as well as some thin sections. A New Wave UP-213 laser was used for ablation with a 20 Hz rep rate and laser power from 3-3.5 J/cm2. Ablations were analyzed with a

with a sampling rate of every ~0.9 sec with an Agilent 7700 Series Quadrupole Mass

Spectrometer. We use varying track widths from 12-30 µm, depending on available space.

111

Inclusions <50 µm in diameter were avoided. Data were normalized to NIST-610 using 28Si as

an internal reference element.

CHEMISTRY:

Whole pumices, extracted from non-welded tuff close to contacts, were analyzed for major

and trace elements. Representative samples can be found in Appendix I. Tshirege rhyolite

pumices fall into three distinct chemical groups: 1. Low-silica rhyolite (LSR), which has 70-73

wt. % SiO2; 2. High-silica rhyolite HSR-e (76-77 wt. % SiO2), relatively enriched in

incompatible trace elements; 3. High-silica rhyolite HSR-d (76-80 wt. % SiO2), relatively

depleted in incompatible elements (Fig. 3.2). The Tsankawi fallout consists

exclusively of HSR-e; the proportion of HSR-d increases upwards through the Tshirege ignimbrite; LSR is only found in Qbt3-Qbt5 material.

Figure 3.3 shows an ‘enrichment plot’ (Hildreth, 1979) of rhyolites normalized to the most primitive LSR. Vanadium, light rare earth elements (LREEs), Eu, Ba, and Sr all show relative depletion and the heavy rare earth elements (HREEs), Nb, Ta, Th, U, Rb and Cs are comparatively enriched. Areas on the plot represent the range in data of respective categories.

112

Fig. 3.2. Selected Harker variation diagrams for major and some trace elements with the three magmas labeled.

113

Fig. 3.3. Enrichment diagram after Hildreth (1979). Trace elements normalized to the most primitive LSR sample. Note heavy depletions in Eu, Ba, V and Sr for the HSR-e and -d pumice and the enrichment in HREE in Th, Nb, Ta, and Hf in the HSR-e.

114

PETROLOGY AND CRYSTAL CHEMISTRY:

Petrographic characteristics of pumice samples falling into the LSR, HSR-d, and HSR-e geochemical groups are distinctive (Figs 3.4 a-c; Table 3.1). Modal percentages reported in this section are normalized to ‘vesicle free’ dense rock equivalent:

HSR-e – Pumice that fall into this category are generally white to light grey in hand sample and are relatively crystal poor with total phenocryst modal abundances ~10-20%.

Phenocrysts are predominantly euhedral to subhedral quartz (<100 um to 2 mm) and sanidine

(<50 um - to 1 mm) in roughly equal abundance, with minor amphibole, reacted pyroxenes

(100-500 um), and trace amounts of fayalite (~100-300 um), magnetite, biotite, and zircon.

HSR-d – These pumice are white to grey and have total modal abundances of 20-30% of predominately sub- to anhedral quartz (<500 um to 4 mm) and sanidine (500 µm to 5 mm) with trace amounts of magnetite and zircon. Sanidine occurs in crystal clots or as glomerocrysts with quartz and magnetite. Cores of some sanidine crystals in the HSR-d are An65 plagioclase.

LSR - samples are distinctly different from the HSRs with an overall salmon red/pink to

grey color and contain characteristic large crystal clots and lenses of sanidine (0.1-2 cm - long

axis) plus ortho- and clinopyroxene, with total crystallinities of 25-45%. Some large sanidine

clots appear to be aggregate glomerocrsts, while others are single crystals with small amounts of

melting which are optically continuous (Fig. 3.5).

115

LSR LSR d pumice - HSR

b. e pumice showing low crystallinity and feldspar + quartz. crystallinity e and+ pumicefeldspar quartz. showing low - Qbt4 ignimbrite with fiammé and very large crystal clots of feldspar + pyroxene pyroxene + feldspar of clots crystal large very and fiammé with ignimbrite Qbt4 HSR a. th feldsparpyroxeneslabeled. d. th and Representative BSE slide maps for slide maps BSE Representative

pumice wi Fig. 3.4. phases.the in andtheand labeled.also c. feldspar crystaloccurringgroups ofwith Notedecrepitation larger quartz sizes labeled.

116

117

Fig. 3.5. Large (>1 cm long axis) sanidine from the LSR pumice exhibiting slight melting texture.

118

Figures 3.4-d shows a BSE map of ignimbrite from the Qbt4 portion of the tuff. This

sample consists ~ 20% fiammé with or without large sanidine crystals and crystal clots, set in a

matrix of ash and crystal shards. The fiammé contain anhedral quartz with strong resorption features. Biotite is present within large crystal aggregates within the ignimbrite matrix, and absent from the other three petrographic groups. The Qbt4 ignimbrite can be regarded as mixture

between the three pumice types, with fiammé most closely resembling the LSR group; the

ignimbrite contains additional components (e.g. biotite) which are rare or absent from the other

three pumice types.

Quartz:

Quartz is abundant in nearly all samples, except some LSR, which lack quartz

completely. Idiomorphism of quartz decreases moving upwards in the tuff and where found in

the ignimbrite fiammé, quartz is well rounded (Fig. 3.6). For a detailed study of quartz

distribution, cathodoluminescence (CL)-zoning types, and [Ti] diffusion modeling see Chapter 2,

and Wilcock et al. (2013). In summary, CL-zoning becomes more abundant moving up through

the deposit; [Ti] in crystals from the HSR-e and crystal cores from the HSR-d are the same, and

CL- bright rims have higher [Ti]; the boundary between these layers has been modeled for

diffusion resulting in a range of times from 50-2200 years (Chapter 2).

In HSR-e samples, quartz is ≤ 2.5 mm, with most crystals being ≤ 500 µm and is present in modal abundances of about ~5%. In HSR-d samples, quartz is much larger and very few crystals are < 1 mm in size and are a maximum 5 mm and are present in modal abundances of

~15-20%; in LSR samples quartz is either present in modal abundances of up to 10% or not present at all.

119

Fig. 3.6. Two anhedral quartz from the late erupted ignimbrite showing melting texture. Note faceted melt inclusion in one of the crystals (labeled).

120

Quartz in HSR-e samples are often euhedral and melt inclusions are intact, while in HSR-

d and LSR samples they usually exhibit strong resorption textures, are decrepitated, and melt inclusions within these crystals often show nucleation of bubbles and, in some cases, look to be

exploding the crystals apart (Fig. 3.7).

Feldspars:

Feldspars are found in all samples in the Tshirege and compositions vary systematically between pumice types (Fig 3.8). HSR-d and HSR-e feldspars are Or36-Or45 with a minor

anorthite component, where for a given Or content, HSR-d analyses have higher An. LSR

crystals are Or28-Or37 and up to An5. In HSR-e samples, feldspars are <1.5 mm in modal abundances of ~10%; in HSR-d samples, feldspar modal abundances is ~10-15%, and can be divided into two groups:

1. Individual phenocrysts ranging from >1-5 mm.

2. Large crystal clots or glomerocrysts (>5 mm up to 1.5 cm) of multiple large feldspars.

Crystals in both groups are almost always decrepitated and often show strong resorption

textures (Fig. 3.9); rarely some of these crystals have An65 plagioclase cores. LSR feldspars

show similar textures to the second type from HSR-d, but crystal clots are even larger, up to 3 cm in the long dimension, although crystals in the LSR lack plagioclase cores. Additionally, some feldspars in the LSR are microcline with checker-board twinning.

Sanidines in HSR-d and LSR samples have CL-bright rims with from 0.1 to 0.25 wt. %

BaO. Sanidines in the HSR-e pumice have BaO concentrations below detection limits and show no CL-bright rims.

121

Fig. 3.7. Quartz from the HSR-d pumice which has been exploded from an interior melt inclusion. Yellow line outlines exploded melt inclusions, also note preserved bubbles in other melt inclusions.

122

Fig. 3.8. Anorthite plotted vs orthoclase content for alkali feldspars from the Tshirege ignimbrite sequence.

123

Fig. 3.9. Two CL images of sanidine from the HSR-d pumice a. Non-rimmed sanidine from the HSR-d with plagioclase core. b. Sanidine from HSR-d with CL-bright rim which corelates to the rims labeled on the graph to the left. c. Graph of BaO wt. % plotted vs Anorthite content for LSR, ignimbrite feldspars inside of fiammé, and HSR-d pumices. HSR-e feldspars below detection limit in BaO.

124

Amphibole:

Amphibole is a rare phase (<<1% modal) in the Tshirege rhyolites, and only has been

found in the Qbt1 HSR-e, where it is present as small <50 µm crystals of pargasite, edenite,

tschermakite, and hornblende compositions when plotted on Si vs (Na+K)A-site on a PFU basis

(Appendix table A.4.1).

Pyroxene:

Pyroxene chemical data is shown on Figure 3.10. Pyroxenes are present in modal

abundances of <5% in the LSR pumice and Qbt4 ignimbrite fiammé and one crystal was found

in the HSR-e pumice. Pyroxene is absent in the HSR-d. Pyroxenes in the LSR pumice occur as large (300-500 µm) zoned phenocrysts (Fig. 3.10b) and smaller (50-150 µm) touching pairs of

ortho and clinopyroxene. Additionally, Qbt4 ignimbrite contains discrete pyroxenes within the

fiammé glass and inside glomerocrysts of sanidine, biotite, magnetite and amphibole where they

are interpreted to be the product of the breakdown of biotite. Figure 3.10d shows the single

exsolved pyroxene from the HSR-e rhyolite where it contains a melt inclusion that houses a

single biotite crystal.

Biotite:

Aside from the biotite inside of the melt inclusion in the HSR-e magma from the previous

section, biotite is absent from the HSR-e, -d, and LSR and is only found in the Qbt4 ignimbrite.

When present it is found within large (>1 mm) crystal aggregates, exhibiting breakdown reaction textures to magnetite and pyroxene ± amphibole (Fig. 3.11). Biotites here contain ~1.2 ± 0.07 wt. % F and have been calculated to have ~2.3 ± 0.06 wt. % H2O by PFU difference (appendix

4).

125

Fig. 3.10. a. Chemistry of pyroxenes from the Tshirege plotted on pyroxene quadrilateral. b. Rim to core transect of pyroxene from LSR showing decrease in En towards rim. c. Temperatures from two-pyroxene thermometer of Putirka et al. (2008) plotted vs Mg-Fe exchange to test for equilibrium, with equilibrium regions for subsolidus and magmatic situations labeled and represented by shaded regions. d. Single pyroxene found in HSR-e showing exsolution and containing melt inclusion which hosts a biotite crystal.

126

erupted ignimbrite. Crystal Crystal ignimbrite. erupted - BSE slide map with series of zoomed images showing a crystal aggregate from the late the aggregate from a withofcrystal slideimages series zoomed showing BSE map

11.

Fig. 3. magnetite. and pyroxene form to reacting is it where texture breakdown exhibiting biotite of consists aggregate

127

Oxides and Minor phases:

Oxides are present as magnetite and titanomagnetite which sometimes exhibit exsolution to ilmenite + magnetite. Oxides are found in trace amounts in all three magmas between 10 <

100 µm in size. Fayalitic olivine is present in trace amounts (one crystal found in sample SKJM-

59) in the HSR-e magma and has been previously used to calculate magnetite-fayalite

temperatures by Warshaw and Smith (1988).

THERMOBAROMETREY

Results from this study along with previous authors are summarized in Table 3.2 (Warren

et al., 2007, Warshaw and Smith, 1988; Wilcock et al., 2013).

Two-pyroxene:

Two pyroxene thermometry is summarized in Figure 3.10c. All touching pairs found in this

study are in Qbt3-4 LSR and welded ignimbrite samples within fiammé and in the groundmass

of the ignimbrite. Shaded regions on Figure 3.10c represent equilibrium for both magmatic Fe-

Mg partitioning between ortho- and clino-pyroxene and subsolidus exchange. Pairs are generally

slightly out of equilibrium and plot within the subsolidus equilibrium field, but are interpreted to be of magmatic affinity and not from a subsolidus source, as pyroxenes in the Qbt3-4 LSR and

Qbt4 ignimbrite have euhedral crystal shapes and show no reaction textures. Compare to

pyroxenes within the Qbt1 HSR-e, which show strong reaction textures and generally would not be described as ‘happy’ (Fig. 3.10d).

128

129

Ti-in-quartz:

Multiple calibrations of the Ti-in-quartz thermometer on quartz-glass pairs from the

HSR-d magma (Chapter 2) give temperatures between 748-786 °C for a constant a(TiO2) for the

system, assuming buffering of Ti by oxides and 727-762 °C if a(TiO2) is calculated by using the

[Ti] in the adhered glass.

Feldspar-liquid thermometer:

The feldspar liquid thermometer of Putirka (2008) yields averages of 736 °C for the

HSR-e, 783 °C for the HSR-d, and 785 °C for the LSR. We attempted to estimate water

concentrations using the K-feldspar-liquid hygrometer of Mollo et al. (2015), but it yields

unreasonably high water concentrations (10-13 wt. % H2O).

Amphibole Barometer:

The amphibole-liquid barometer of Putirka (2016) yields pressures of .16-.20 GPa for amphiboles in the HSR-e magma and a temperature of 725 °C. This data set is limited to the 4 amphiboles that were found and 8 analyses.

Glass chemistry and water concentration estimates:

Glass analyses have been plotted using CIPW normative calculations corrected for silicate

melts after Blundy and Cashman (2001) (Fig. 3.12a). Figure 3.12b shows F plotted vs Cl for the

same glasses. Glasses which plot to lower Cl and higher F labeled ‘biotite rich source?’ in

Figure 12b are the same as the glasses which trend towards the Ab corner on Figure 3.12a.

Figure 3.12c shows average F for individual samples of HSR-d rhyolites plotted vs Σ(HREEs:

130

Gd-Lu) for the whole-pumice chemical data. This correlation suggests melting of a mafic phase

present and contributing to the melt during F enrichment.

Fluorine in materials with iron can be difficult to acquire accurate quantitative analyses using wave-dispersal spectroscopy because the K-α x-ray of F and the L-α x-ray of Fe occupy almost the same energy spectrum. We do not believe our F data is a function of Fe L-α interference as there is no correlation between Fe and F in these glasses (Fig 3.13) and glasses with no F and similar Fe contents show no F enrichment.

Water estimates:

A water saturated rhyolite at pressures of 0.2-0.3GPa should yield 6.0-7.6 wt. % H2O (Eqn.

2 in Holtz et al., 2001). To reach water concentrations suggested by Mollo et al. (2015) k- feldspar-liquid hygrometry, it would require a system at >0.5 GPa. This pressure is unrealistic for the Tshirege magma body based on amphibole barometry, but also this pressure is unrealistic for evolved silicic systems in general, as water saturation at these pressures favors less silicic compositions (Gualda et al., 2012).

We use micro-Raman spectroscopy to calculate the H2O concentrations of melt inclusions

within the HSR-e and HSR-d magmas using the calibration of method of Le Losq et al. (2012)

and data acquisition methods of Shea et al. (2014). Melt inclusions are a natural choice for

estimating the original melt water concentration as the low partition coefficient of H in quartz

acts as a bottleneck to limit the degassing loss of water (Myers et al., 2018). We find moderate

concentrations of 1.7-3.7 wt. % H2O with an average of 2.74 ± 0.58 wt. % across HSR-e and

HSR-d pumice-hosted quartz. There is no correlation between [H2O] and pumice type.

131

Fig. 3.12. a. CIPW normative glass analyses corrected for silicate melts after Blundy and Cashman (2001). Pressures are plotted for water saturation at 2 kbar. b. F wt. % plotted vs Cl wt. % for individual glass analyses. c. Average glass analyses for induvial pumices plotted vs whole- pumice Σ(HREE: Gd-Lu) from the HSR-d.

132

Fig. 3.13. F wt. % plotted vs FeO wt. % for individual glass analyses as a test for Fe L-α interference.

133

DISCUSSION:

The Tshirege ignimbrite sequences contains pumice produced from three geochemically

distinct rhyolitic magmas (Fig. 3.3). The chemical and petrologic data presented in this paper and

previous studies allow for detailed reconstruction of the Tshirege magma chamber and its pre-

eruptive history.

Petrology:

Quartz and sanidine increase in size and abundance and become less idiomorphic with

increasing textural evidence of melting (Figs. 3.4–7) upwards through the tuff. The phase

assemblages in each distinct pumice type (HSR-e: Quartz + Sanidine + Amphibole + Magneite;

HSR-d: Quartz + Sanidine; LSR: Sanidine, ± Quartz, two-pyroxene) suggest the crystal mush was produced from a more primitive magma which progressively evolved during mush production to become more felsic and more water-enriched. This configuration also suggests pre-recharge stratification of melt and cognate crystal pile which represents the order of fractional crystallization.

Although biotite is no longer present in the HSR-d, high [F] in these samples suggest its present prior to recharge. Using minor mafic phase (pyroxene, biotite, amphibole), it can be determined that the system started off relatively water poor to produce pyroxene and increased in

H2O over time and biotite became stable. The lack of biotite and presence of amphibole in the

HSR-e is curious and may be explained by lower water contents in the HSR-e because of

communication with off-gassing events or water loss to the country rock.

Decrepitated crystals from exploded melt inclusions suggest semi-rapid depressurization

or heating of the system, faster than which would allow for diffusive loss of H+ through crystal

134

lati (Bindeman, 2005; Myers et al., 2018; Zalinge et al., 2018). The amount of decrepitated crystals increases as you move down into the chamber, which may indicate the higher portions of the system underwent slower degassing, which could be accomplished through or pre-eruption degassing, indicating that the top of the system was leaking gasses, which is a common phenomenon that proceeds many volcanic eruptions (Lowenstern et al., 2012; Preece et al., 2014;

Stock et al., 2018).

Chemistry:

The enrichment in Si in the HSR-d past the accepted granitic minimum of 77.8 wt. % are impossible to explain by a simple model of fractional crystallization combined with mixing of recharge magmas. Chapter 2 presents melt inclusion chemical data which suggest that most of the Tshirege magma is the result of partially melting a cumulate pile of quartz and sanidine with minor phases present.

Melt generation:

Here we attempt to quantify the melting mechanism of the cumulate pile and present the results of 50 Rhyolite-MELTS modeling experiments of LSR and test if the cumulate melt model of Wolff et al. (2015) is a reasonable explanation for the generation of parts of the Tshirege magma. LSR was chosen because it shows the highest crystallinity and likely is the closest to the original cumulate pile before recharge events (Fig. 3.4c). Figure 3.14 summarizes the Rhyolite-

MELTS experiments. Each line represents different water saturation conditions with the behavior of the entire crystallinity of the system (blue lines) and quartz (grey lines) tracked. We assume a starting condition of ~1-1.5 wt %. H2O for a water-saturated crystal mush hosting sanidine and quartz, which melts modeling suggest would exist at ~70% crystallinity (reasonable

135

mush crystallinity) at ~780 °C, consistent with the feldspar-liquid and two-pyroxene thermometers used to estimate LSR temperatures.

A few important observations can be made from this graph:

1. Quartz is a predicted phase in quantities of ~15-20% at 800 °C and 1.0-2.0 wt. % H2O,

but is absent from most LSR samples, and where it is present shows strong melting

textures (i.e. Fig. 3.6).

Fig. 3.14. Rhyolite-MELTS modeling for LSR magma. Explanation in the text.

136

2. Temperature addition alone does not cause large changes in crystallinity until around

820-850 °C and quartz stays present, even up to 900 °C at 1.0 wt. % H2O. A simple

heat budget calculation using heat capacities calculated by MELTS was used to

determine what reaching these temperatures starting from 750 °C would require in

terms of volume of recharge magma (Equation 1).

= 𝑄𝑄( 𝑚𝑚𝑚𝑚Δ𝑇𝑇 ) (1)

𝑓𝑓 𝑖𝑖 𝑓𝑓 𝑖𝑖 𝑄𝑄 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑄𝑄 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 − 𝑄𝑄 𝑟𝑟ℎ𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦 − 𝑄𝑄 𝑟𝑟ℎ𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦

Q = heat content (J), m = mass (g), derived using density from MELTS; c = heat

capacity (J g-1 °C-1) calculated in MELTS, = change in temperature.

Figure 3.15 shows three different changesΔ𝑇𝑇 in temperature and the volumes of

recharge magma required to produce those changes. The y-axis describes the heat lost

from the recharge magma, thus at low volumes high heat losses from the recharging

magma are required to heat up the Tshirege magma. The dacite magma has been

calculated to have a temperature of 780-800 °C using two-oxide thermometry, but this

is likely an equilibrium temperature the dacite achieved while mixing and mingling

with the rhyolite (Chapter 1) and may serve as a lower limit, while the storage

temperature of the dacite could be higher. For the sake of this argument, we can

assume a temperature of 950 °C which is likely higher than the actual temperature of

the dacite, but can serve as an upper limit.

Magma flux rates vary from system to system and at the JMVF historically range

from ~0.2-0.6 km3 1ka-1 (Wolff and Thompson, in press). Although it may be logical to

137

Fig. 3.15. Temperature curves for the volume of recharge magma (x-axis) required to raise the temperature of the mush by 50, 100, and 150 °C. The y-axis shows the amount of heat lost from the recharge magma.

138

think caldera volcanoes capable of producing super eruptions require extraordinary flux

rates, moderate flux rates over long periods of time can produce thermally mature crust

with distillation columns capable of producing caldera volcanoes (Krakas et al., 2017).

3 In this case, to produce 400 km of magma in the 350ka year hiatus between the

eruption of the Otwoi and the eruption of the Tshirege it would require flux rates ~2-4

times higher than historically recorded at the JMFV.

In Chapter 2 it is shown that the Tshirege magma was reactivated on the order of

100-1,000 years. Thus, even if the dacite was initially 950 °C it would require over 200

km3 to achieve temperature increases of 100 °C (e.g. that which is required to fully

resorb quartz), it would require flux rates >200 km3/1,000 years; there is no reason to

believe that was even remotely possible for this system. Heating alone would require

>2 times the total erupted volume of the Tshirege.

3. MELTS modeling suggests an alternative mechanism in the form of water addition,

which could be sourced from second boiling of invading recharge magmas. Adding 0.5

wt. % H2O to a dry (1.5 wt. % H2O) mush environment can achieve a mass resorption

of ~20%. Additional water enrichments would only enhance this effect (Fig. 3.14).

The rhyolite crystal mush is thought to make up to half the volume of the system

(<200 km3), although this is a minimum as there is generally unseen portions of the

system in the erupted material which were retained in the chamber during the eruption.

The dacite is thought to have ~5.0 wt. % H2O from plagioclase hygrometry (Waters and

Lange, 2015), so to increase the water concentration in the rhyolite mush by ~1 wt. %

3 H2O would require 40 km dacite losing around half of its water to second boiling, for

139

which there is ample evidence in the form of polybaric crystallization and rapid growth

feldspars of ternary composition (Chapter 1).

4. This helps explain the over-enrichment in SiO2 in some HSR-d samples. If quartz is

preferentially melting during mobilization, some parts of the system may become

enriched in liquids which were derived from melting mostly quartz, and thus produce a

melt composition which would be impossible to acquire through fractional

crystallization alone.

Fluorine enrichment:

Fluorine has a strong affinity for silicate melts and generally does not act as a volatile due to its high solubility (Dolejs and Baker, 2007). The [F] of rhyolites ranges widely among eruptive centers around the world and is known to partition strongly into amphibole, biotite, and apatite in these systems (Iveson et al., 2018; Stock et al., 2018).

The [F] content of the Tshirege rhyolites ranges from below detection limits in some of the LSR and ignimbrite fiammé to almost 1.6 wt. % in some HSR-d pumices (Fig. 3.12b). The averages are LSR = 0.23 ± 0.17 wt. % F, HSR-d = 0.56 ± 0.26 wt. % F, and HSR-e = 0.28 ± 0.02 wt. % F. Generally, no correlation exists between F and other elements, except in HSR-d pumice where there is a fairly strong correlation (R2 = 0.73) between F and HREEs, suggesting F in

these samples was added in conjunction with consuming a mafic phase, or the phase which

housed F also contained HREEs. Biotite is found in the Qbt4 ignimbrite, where it is seen in

breakdown reaction textures and has [F] of ~1.2 wt. %. We suggest the crystal mush contained locally high concentrations of biotite which was melted or reacted out to add F to the melt prior to eruption.

140

Glass chemistry:

CIPW normative glass data show trends which have been interpreted as polybaric

crystallization (Blundy and Cashman, 2001; Chapter 1). Figure 3.12a shows CIPW normative rhyolite glass compositions corrected for silicic melts after Blundy and Cashman (2001); the following statements can be made:

1. The slightly translucent black oval in Fig. 3.12a represents a quartz + feldspar

cumulate mush melt at 50 MPa P(H2O), such as might be generated from a ~ 70%

crystal mush at water saturation.

2. During incremental melting of anhydrous phases in a mush, the liquid becomes

progressively more water-poor, pushing the glass towards lower P(H2O) and thus

towards the Q corner. Resorption of quartz creates SiO2 enriched melt.

3. This trend can also be achieved by polybaric crystallization as the magma makes its

way to the surface, but we see no microlites in these samples (Blundy and Cashman,

2001); although decompression crystallization may be achieved by new growth on

pre-existing crystals, most of the crystals in the LSR and HSR-d magmas show

anhedral habits or strong melt textures (Fig 3.4-7).

4. One set of HSR-d glass analyses plots towards the Albite corner. This is not due to

Na or K mobility (Appendix Fig. A.4.2), as the samples show no correlation between

K and Na vs total oxide wt. %, rather we interpret this to represent a portion of mush

with a different assemblage of crystals and higher P(H2O), and because it corresponds

to the highest [F] analyses, was likely capable of supporting biotite prior to recharge.

141

Volume problem:

CL brightness in quartz is attributed to higher [Ti] in quartz (Thomas et al., 2010; Leeman et al., 2012). Quartz in the Tshirege occurs as both zoned and non-zoned in CL with the proportion of crystals with zoned rims increasing upwards in the stratigraphic succession (see Wilcock et al., 2013; Chapter 2). This increase in brightness records higher-temperature or higher a(TiO2) conditions, which are usually linked to a recharge event (Wark et al., 2007).

Figure 3.16 shows a simplified model for calculating the volume of resorbed quartz from cross-sectional cuts of individual crystals. The model assumes that quartz are euhedral bipyramids prior to a resorption event (Fig. 3.16 white, dashed lines) and that the quartz melts equally in all directions to form spherical shapes. Using this model, initial resorption of just the edges of CL dark quartz grains would have resulted in <23% volume reduction of the quartz, as complete edge resorption is not recorded in CL dark crystal cores. This volume reduction corresponds to a 5% modal reduction quartz. If the melting of quartz is driven by the addition of water, then the process is buffered by melting, as melted quartz dilutes the water concentrations.

These resorbed cores serve as starting points for new growth of CL-bright rims (Fig. 3.16, green outlines).

142

Fig. 3.16. Figure showing model for melting bipyramidal quartz. Complete resorption of edges with equidistant melting causes ~23% volume loss of quartz. CL boundaries highlighted in green show interpreted resorbed boundary and white dashed lines show interpreted original boundary with volume losses always less than 23%. The morphologies seen here show that complete edge resorption was not always achieved, and the volume reduction of 23% is assumed a maximum in that case, but in cases where complete edge resorption is observed, then 23% would be a minimum.

143

Rhyolite-MELTS modeling predicts volume reductions ~10-20% at starting crystallinities of

70% and thus these results are in good agreement with the chemical model for melting

mechanism; (e.g. crystal edges would no longer mechanically lock the mush).

CL-bright rims are anywhere from 1-2.5 times the volume of cores. If this is a crystal mush and only the outside edge of cores have been resorbed, the system should run out of space to compensate new growth of CL-bright rims. A few possibilities could account for this discrepancy.

1. It could be that over the course of many small resorption and regrowth events, the

system favors the production of fewer, larger crystals and thus some crystals are

sacrificed to make room for their larger counter parts.

2. During injection events it is possible that bubble nucleation pushes crystals apart for

long enough times to compensate new crystal growth.

3. The assumption that crystal cores are resorbing uniformly may be too simple. If the

bulk system is taken from 1.5 wt. % H2O to 3.0 wt % H2O it will experience a volume

reduction of ~75% for quartz, which is three times the volume reduction calculated

earlier. This would require either more off-gassing from the recharge magma, or a

higher volume of recharge magma with the same proportions of second boiling, but

would easily create enough room for the larger overgrowth of CL-bright rims.

4. For inequant phases such as feldspars, Mushes may exist as rigid bodies with as little as

25% crystallinity (Holness, 2018), and since we’re only seeing a 2-D cross section of

any given crystal or set of crystals, it could be as simple as there is no space problem;

it’s an observational artefact. However, with a 2:5 ratio of quartz to feldspar, a mush of

25% is probably an unreasonable starting point for this system.

144

Second boiling and heating:

An obvious contradiction exists in the HSR-d magma and some of the LSR magmas where we suggest water addition and biotite breakdown in the same sample, where water addition should make biotite more stable, not less. The breakdown can still be accomplished if the recharging process is a two-part process whereby the system feels a relative reduction of [H2O] due to dilution of water from crystal resorption first, causing biotite to breakdown. This process also causes biotite to release water in proportions of ~2x the amount of [F] (PFU calculation, appendix 4), which in the most enriched samples requires local concentrations of 45% biotite

(Fig. 3.17), which creates complexities in the breakdown process which are unclear. Although sometimes the breakdown regions contain amphibole, so the water possibly gets partially taken up by that phase during the reaction. Next, the heat transfer from the recharge magma to the mush causes crystallization in the recharge magma which induces second boiling and releases water into the system where the biotite has already broken down and thus water addition causes higher-volume melting of feldspar and especially quartz.

INTERPRETATIONS

Chemical evolution:

Melt inclusions in Tshirege quartz come in two main forms: faceted, located in crystal cores and unfaceted, located in crystal rims (Chapter 2). Depletion of compatible trace elements

145

Graph b. O additionof O to melt resorption during 2 redicted enrichment diagrams due to the breakdown/melting of biotite with 1.2 wt. % and ofF 2.2 wt.%. biotite 1.2 wt. the with redicted to breakdown/melting enrichmentdue diagrams P

biotite. Fig. 3.17. a. calculationratio of PFU ~1:2 F:H showing of relation to water enrichment F using of

146

in crystal core inclusions (e.g. Ba), and more abundant compatible trace elements in crystal rim

inclusions leads to the interpretation that faceted inclusions in crystal cores formed early and the

unfaceted melt inclusions in crystal rims formed late during melting of cumulates.

Figure 3.18a-d shows bivariate plots of select trace elements and trace element ratios for

adhered glass (circles), and both faceted (squares) and unfaceted (triangles) melt inclusions.

Fractional crystallization models are fit to faceted melt inclusions testing if they are tracking the

evolution of a fractionating body.

A crystalizing assemblage of Quartz (11%), Biotite (47%) and K-feldspar (40%) with

trace amounts of apatite, magnetite can produce trends recorded in faceted melt inclusions and

require ~50% fractional crystallization to produce the most evolved melt inclusions. The FC

trend described is different than any trends recorded in the whole-pumice chemistry (appendix

Fig. A.4.3) Full modal abundances and partition coefficients used for modeling are included in

Appendix Table A.4.2-4.3. We note that the crystalizing assemblage changed during the evolution, but the phases we use represent many of the petrographic observations made about portions of the interpreted mush.

In this case, biotite was probably not stable until higher water contents during later stages of evolution and is a predicted phase in the HSR-e using Rhyolite-MELTS at 700°C during the last 10% of crystallization. This suggest that early fractional crystallization may have been lacking biotite and helps explain why many of the LSR and some of the HSR-d samples contain low [F] or it is below detection limits and why some HSR-d samples contain very high F contents, as when biotite became stable it started depleting the system in F, which would have been enriching up to this point.

147

Fig. 3.18. a-d. Bivariate plots of various trace elements and ratios for melt inclusions and adhered glasses for Tshirege quartz crystals. Squares are faceted melt inclusions, triangles are unfaceted inclusions, and circles adhered or embayed glass. Fractional crystallization models are labeled with black lines and 10% FC increments fit to faceted melt inclusions. Explanation of models are in text. a. Ti/Th vs Rb.

148

Fig. 3.18. a-d (cont.) b. Th vs Ti.

149

Fig. 3.18. a-d (cont.) c. Th vs Zr/Hf.

150

Fig. 3.18. a-d (cont.) d. Nb vs Yb.

151

In all panels of Figure 3.18, there is a bimodal grouping where adhered melt and some unfaceted melt inclusions from LSR and HSR-d quartz plot together. These groupings are thought to represent melted cumulate material, which itself was derived from fractionating HSR- e or an earlier magma body.

Figure 3.19 shows Ti/Th and La/Th plotted vs Rb/Ba for adhered glass, faceted, and unfaceted melt inclusions. Faceted LSR inclusions were used as a starting point for modeling the evolving liquid during fractionation because they are the latest erupted and thus probably the earliest fractioned. Similar to the FC models in Figure 3.18, this predicts ~50% crystallization to produce the most evolved faceted MI. Although in this graph, there seems to be three distinct groupings, whereas previous graphs (Figure 3.18) showed only two groups. This can be explained by the following situation:

1. The pre-fractional crystallization original melt is recorded in faceted LSR inclusions.

2. Faceted inclusions located in the center of crystals in the HSR-e and HSR-d record

this melt as it evolves following the FC model.

3. The evolved melt mixes to varying degrees with melted ~5% k-feldspar and 5%

quartz which enriches that fluid in feldspar components (Ba and Sr) and drastically

decreases Rb/Ba, and unfaceted melt inclusions in HSR-e quartz capture that liquid.

Ti/Th and La/Th ratios are not affected due to low percent mixes of only melted

feldspar and quartz.

4. Melting of more of the mush including chevkinite, force the fluid along a negative

slope away from the evolved fluid for both Ti/Th and La/Th. This does not mix with

HSR-e melt.

152

Rb/Ba vs La/Th. Note similar trends in both graphs. Ti/Th v Rb/Ba Rb/Ba Ti/Th in v Note similarboth vstrendsgraphs. Rb/Ba La/Th. b. d and fits well to faceted melt inclusions. La/Th vs Rb/Ba graph shows La/Th facetedvsgraph Rb/Ba fitsto meltshows inclusions. d and well - Rb/Ba vs Rb/Ba Ti/Th and a. e liquid during fractionation. Faceted LSR inclusions are used as the starting condition for FC condition FC the for are inclusions asstarting e used liquidfractionation.Faceted during LSR - d melt and unfaceted inclusions record melting of feldspar + chevkinite in the crystal mush. crystal the in chevkinite + feldspar of melting record inclusions unfaceted and melt d - olution of HSR bivariate plots with with plots bivariate Two

Fig. 3.19. modelFig. for 18a graph same using has FC parameters ev the for trends and HSR LSR modeling.

153

Evolution Model:

Data presented here and in Chapters 2 and 3 allow the development of a model of cumulate formation followed by recharge and mobilization of the mush to generate an eruptible system. We present a four-stage model which describes the events that lead to the formation of the Tshirege magma (Fig. 3.20):

1. The precursor to HSR-e magma started crystallizing an assemblage of quartz,

feldspar, ± biotite, trace amounts of magnetite, ilmenite, and chevkinite at a

temperature of 697-725 °C and pressure of ~0.18 GPa (this study; Warshaw and

Smith, 1988). Crystallization was accompanied by hindered settling and crystal

repacking during fractional crystallization, melt would become trapped in quartz

cores as melt inclusions.

2. During long-term storage (1,000-10,000 years, Chapter 3) these inclusions

became faceted; these faceted inclusions track the evolution of the HSR-e melt.

At some point during this evolution biotite became a stable phase, likely at a later

stage after F became slightly elevated in the HSR-e, and then acted to sequester F

into portions of the mush. This stage likely fractionally crystalized ~50% of the

HSR-e magma and produced a scenario where the mush is equal ~ half the

volume of the total system.

3 3. Incremental recharge of minimum ~40 km dacite with ~5 wt. % H2O reactivated

the system. Some of these recharge magmas made their way into the system and

equilibrated with parts of the mush, evidenced by An65 plagioclase cores in some

sanidine crystals in the HSR-d, while others stalled, added heat, crystallized, and

154

155

156

157

cted to ), recording a 2 d magma, - TiO ( a

:and ± Quartzfeldspar pyroxene with early fractionated material at the bottom bottom the at material fractionated early with

Stage 1 s quartz grows rimshighers at quartz grows or T ture is seen in all of the mush. At this stage the mush is is mush the stage this At mush. the of all in seen is ture magma, it crystallizes and second boils to releases water water releases to boils second and crystallizes it magma,

Recharge magma invades the root adds invadesthemush zonesthe and magma Recharge of O] by at least 0.5 wt. %, causing the mobilization and O] by at%,mobilization formation ofHSR causingthe leastthe 0.5 wt. 2 Continued recharge causes mobilization and LSR generation. Crystals in the LSR show the show the Crystals generation. in LSR Continuedand causes mobilization LSR recharge e. New melt inclusions form in quartz a quartz in form inclusions melt New e. - Stage 4: Stage 3: O) andconstant roughlyStage temperature. 2 Four stage diagram showing evolution of the Tshirege magma chamber. ofmagma thechamber. Four evolution stage Tshirege showing diagram After prolonged fractional crystallization, heterogeneous crystal mush exists mush crystal heterogeneous crystallization, fractional prolonged After

1.5 wt. % H 1.5 cumulate signature. signature. cumulate - Figure 20. phases theof withinclusions the evolution crystalizeminorand trap meltsystem. during chevkinite, later fractionally biotite and Stage 2: predi is polygons) green (flat Biotite storage. during faceted become inclusions melt above; material fractionated later and signa mush; chevkinite the areas of local in is present and phase a late be dry (~1 - recharge the from lost is heat As to resorb. biotite causing some heat upwards theraising into [H system the overlap. may 4 Stage and 3 Stages recharge. by affected most the were likely and textures melting strongest some the mixes with HSR of which melted

158

second boiled, which contributed water to facilitate in melting. It is unclear if the

temperature stratification of (~100 °C) shown in geothermometers existed prior to

recharge or was created during recharge.

4. During continued recharge, the LSR and eventually HSR-d was mobilized; the

three distinct (HSR-e, HSR-d, and LSR) magmas stayed chemically separate,

probably due to density stratification. Post-recharge crystallization evidenced by

CL-bright rims on quartz and sanidine suggest crystallization volume of 1-2.5x

the original crystal volume. Unfaceted melt inclusions which record a ‘melted

mush’ signature in these rims and suggest somewhat complex resorb/regrowth

cycles. Continued recharge eventually loosened any new mechanical stability

obtained by new crystal rim growth, the last of which was preserved during

mixing and mingling with portions of the mush and was the straw that broke the

camel’s back and triggered the eruption.

CONCLUSIONS

The Tshirege ignimbrite records chemical and mineralogical changes which can only be the result of a cycle of fractional crystallization combined with melting of those cumulates. The mechanism of melting in this case is interpreted to be the addition of both heat and water through second boiling of recharge magmas; this may be a ubiquitous mechanism for mush mobilization in caldera systems, as heating alone requires unreasonably high flux rates ~(>0.5 * mush volume)/ 1,000 years, depending on starting temperatures and temperature of recharge magmas.

159

REFERENCES:

Bachmann, O., Bergantz, G.W. (2004) On the Origin of Crystal-poor Rhyolites: Extracted from

Batholithic Crystal Mushes. Journal of Petrology 45-8, 1565-1582.

Bachmann, O., Bergantz, G.W. (2008) The Magma Reservoirs That Feed Supereruptions.

Elements 4, 17-21.

Bindeman I.N. (2005) Fragmentation phenomena in populations of magmatic crystals, American

Mineralogist v.90, 1801-1815

Blundy, J., Cashman, C. (2001) Ascent-driven crystallisation of dacite magmas at Mount St

Helens, 1980–1986. Contributions to Mineralogy and Petrology 140-6, 631-650.

Brandies, G., Jaupart, C. (1987) The kinetics of nucleation and crystal growth and scaling laws

for magmatic crystallization. Contributions to Mineralogy and Petrology 96, 24-34.

Brookins, D.G., Laughlin, A.W. (1983) Rb/Sr geochronologic investigation of Precambrian

samples from deep geothermal drill holes, Fenton Hill, New Mexico. Journal Volcanology

Geotherm Research 15, 43-58.

Cameron, K. (1984) Bishop Tuff Revisited: New Rare Earth Element Data Consistent with

Crystal Fractionation. Science 224, 1338-1340.

Campbell, M.E., Hanson, J.B., Minarik, W.G., and Stix, J. (2009) Thermal history of the

Bandelier magmatic system: evidence for magmatic injection and recharge at 1.61 Ma as

revealed by cathodoluminescence and titanium geothermometry. Journal of Geology, 117,

469–485.

160

Dolejs, D. Baker, D.R. (2007) Liquidus Equilibria in the System K2O–Na2O–Al2O3–SiO2–

F2O−1–H2O to 100 MPa: II. Differentiation Paths of Fluorosilicic Magmas in Hydrous

Systems. Journal of Petrology 48, 807-828.

Eichelberger, J.C. Koch, F.G. (1979). Lithic fragments in the Bandelier Tuff, Jemez Mountains,

New Mexico. Journal Volcanology Geotherm Research 5, 115-134.

Ganne, J., Bachmann, O., Feng, X. (2018) Deep into magma plumbing systems: Interrogating the

crystal cargo of volcanic deposits. Geology 46(5).

Gardner, J.N., Goff, F., Garcia, S. Hagan, R.C. (1986). Stratigraphic relations lithologic

variations in the Jemez volcanic field, New Mexico. Journal Geophysical Research 91,

1763-1778.