The Application of Silica Δ'17o and Δ18o Towards Paleo- Environmental Reconstructions

The application of silica Δ'17O and δ18O towards paleo- environmental reconstructions

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Liljestrand, Frasier. 2019. The application of silica Δ'17O and δ18O towards paleo-environmental reconstructions. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:42013151

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA ′ The application of silica Δ 17O and δ18O towards paleo-environmental reconstructions

a dissertation presented by Frasier L. Liljestrand to The Department of Earth and Planetary Sciences

in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Earth and Planetary Sciences

′ The application of silica Δ 17O and δ18O towards paleo-environmental reconstructions

Abstract

The oxygen isotope composition of silica provides a valuable record of the temperature at which it precipitated and the isotopic composition of the fluid

18 with which it equilibrated. This temperature-δ OH2O pair, however, is non-unique. The chert δ18O record shows a robust quasi-linear increase through time, but this signal has been interpreted as reflecting either cooling oceans, ′ shifting marine δ18O composition, or diagenesis. We suggest that Δ 17O measurements may resolve this uncertainty. In chapter 2 we collect Precambrian chert from many well preserved Precambrian deposits and measure their δ18O ′ and Δ 17O. Given equilibrium thermodynamic predictions for chert O-isotope evolution as a function of precipitation temperature, ocean O-isotope evolution, and diagenesis, we applied a Monte Carlo re-sampling model to test these alternate hypotheses. The results indicate the chert O-isotope record is best described as a product of alteration with higher-temperature, meteoric-derived groundwater. Neither changes in seawater temperature nor changes in seawater O-isotope composition are required. While this result is applicable to the Precambrian, it may not be useful in interpreting Phanerozoic data. We must first determine whether biological ′ fractionation imposes a significant vital effect in Δ 17O. In chapter 3 we measure ′ the δ18O and Δ 17O of diatom frustules from semi-continuous cultures. To

iii Thesis advisor: David T. Johnston Frasier L. Liljestrand ascertain the degree of natural variability we measured the O-isotope composition from two different diatom species across a range of growth rates and temperatures. Variations in the cultured growth rate and species produced no ′ significant signal in the frustule δ18O or Δ 17O. Temperature variations in the culture, however, do produce a measurable O-isotope fractionation. The absolute ′ value of the measured Δ 17O compositions were systematically negative with respect to the thermodynamic prediction. This offset we interpret to represent a vital effect which may be produced by the silica concentrating mechanisms diatoms use to precipitate their frustules. These results confirm the utility of ′ diatom δ18O in reconstructing temperature, but suggest that Δ 17O may not be a ′ suitable temperature proxy for diatomaceous sediment. Instead, Δ 17O may be used to distinguish biologically from abiologically precipitated silica. In chapter 4 we analyze carbon isotope fractionations from the stratigraphy immediately following the Neoproterozoic Marinoan snowball Earth event. We observe a temporary excursion where εp is anomalously small due to an interval

13 of heavy δ Corg deposition. This signal may reflect the unique biogeochemical conditions that persisted in the aftermath of snowball Earth. To explain this record, we developed a model that tracks the fluxes and isotopic values of carbon between the surface ocean, deep ocean, and atmosphere. From this, we conclude the post-Marinoan conditions will not intrinsically generate the observed

13 isotopic signal. Reproducing the heavy δ Corg requires the progressively diminishing contribution of an additional anomalous source of organic matter.

iv Contents

Abstract...... iii Contents ...... v Author List ...... vii ListingofFigures ...... viii Dedication ...... x Acknowledgements ...... xi

1 Introduction 1 1.1 Ocean and geologic record ...... 2 1.2 Evolution of the Earth’s surface ...... 3 1.3 Isotope methods ...... 5 1.4 Carbon cycle ...... 9 1.5 Oxygen cycle ...... 10 1.6 Preservation of sedimentary isotope records ...... 11 1.7 Focus and summary of the current thesis ...... 12

2 The triple oxygen isotope composition of Precambrian chert 14 2.1 Abstract ...... 15 2.2 Introduction ...... 16 2.3 Methods ...... 18 2.4 Results and Discussion ...... 22 2.5 Geological Scenarios ...... 27 2.6 Conclusion ...... 44

v 2.7 Additional Information ...... 47

3 Calibrating the triple oxygen isotope vital effect in cultured diatoms 69 3.1 Abstract ...... 70 3.2 Introduction ...... 71 3.3 Methods ...... 75 3.4 Results and discussion ...... 85 3.5 Environmental significance ...... 96 3.6 Conclusion ...... 101 3.7 Additional Information ...... 102

4 Isotopically anomalous organic carbon in the aftermath of the Marinoan Snowball Earth 107 4.1 Abstract ...... 108 4.2 Introduction ...... 109 4.3 Geological setting ...... 111 4.4 Analytical methods ...... 114 4.5 Geochemical results ...... 115 4.6 Discussion ...... 117 4.7 Conclusion ...... 132 4.8 Additional information ...... 134

5 Conclusion and future directions 155

References 188

vi Author List

The following authors contributed to Chapter 2: Frasier L. Liljestrand, Andrew H. Knoll, Nicholas J. Tosca, Phoebe A. Cohen, Francis A. Macdonald, Yongbo Peng, David T. Johnston. The following authors contributed to Chapter 3: Frasier L. Liljestrand, Anxhela Hania, Mario Giordano, David T. Johnston. The following authors contributed to Chapter 4: Frasier L. Liljestrand, Francis A. Macdonald, Daniel P. Schrag, Thomas A. Laakso, David T. Johnston.

vii Listing of figures

2.4.1 Chert δ18O temporal trend ...... 23 2.4.2 Chert silica fraction ...... 25 2.5.1 Temperature change and ocean O-isotope change hypothesis . . 28 2.5.2 Evolution of seawater Δ’17O...... 31 2.5.3 Model of diagenetic alteration ...... 33 ′ 2.5.4 Alteration of chert in δ18O-Δ 17O space ...... 36 2.5.5 Distribution of stochastic model ...... 39 2.5.6 Results of stochastic model ...... 41 2.5.7 Comparison of δ18O from stochastic prediction and measured value 43 2.7.1 Harvard-LSU oxygen isotope comparison ...... 57 2.7.2 Fifteenmile O-isotope stratigraphy ...... 58 2.7.3 Chert Δ17O equilibrium temperature prediction ...... 59 2.7.4 Monte Carlo best fit precipitation temperature ...... 59 2.7.5 Comparison of chert δ30Si to O-isotope compositions ...... 60

3.3.1 O2 pressure evolution during prefluorination ...... 82

3.3.2 O2 isotope evolution during prefluorination ...... 83 3.4.1 Carbon isotope composition of diatom biomass ...... 87 3.4.2 18 vs temperature relationship of diatom frustules ...... 91 ′ 3.4.3 Δ 17O sensitivity of diatom species and growth rate ...... 92 ′ 3.4.4 Δ 17O sensitivity of diatom temperature ...... 94 3.4.5 Relationship between measured θ and temperature ...... 95

viii 3.7.1 18O temperature sensitivity of each individual diatom species . . 105 3.7.2 Residual of 18O with respect to temperature ...... 106 3.7.3 Residual of 18O with respect to temperature ...... 106

4.3.1 Stratigraphic and isotope data from the Sheepbed and Ol formations ...... 112 4.6.1 Structure of the carbon isotope model ...... 119 4.6.2 carbon isotope model results ...... 122 4.8.1 Literature C-isotope data ...... 135 4.8.2 Temporal evolution of carbon cycle model ...... 137 4.8.3 Temporal evolution of carbon cycle model ...... 142 4.8.4 Carbon model sensitivity testing part 1 ...... 146 4.8.5 Carbon model sensitivity testing part 2 ...... 149 4.8.6 Carbon model sensitivity testing part 3 ...... 151 4.8.7 Carbon model sensitivity testing part 4 ...... 153

ix To my mother Blinda. Who is never afraid to lead but at the same time taught me to explore my own path. To my father Howard. Who in quiet moments and when I need it the most supports me more strongly than anyone. To my sister Emily. My best friend who is always there for me, my closest companion with whom I always overcome obstacles and share triumphs.

x Acknowledgments

First and foremost I would like to thank my advisor Dave Johnston for guiding me through this process. He was always patient and considerate with me. He taught me how to think as a scientist and I am grateful for his efforts shaping the person I am today. I wish I had developed the habit of watching the Red Sox; our conversations would probably have flowed more easily if we’d had that to share. I want to deeply thank all the Lab members of the Johnston lab, present and former. Thank you Ben Cowie for mentoring me when I decided to study triple oxygen isotope systematics before I really understood how isotopes worked. You were sometimes so supportive of me that I thought you were maybe being sarcastic. Thank you Andy Masterson for showing me its possible to be bothan excellent grad student and a mature functioning adult. Thank you Erin Beirne for sharing your office with me, you made it a joy to come to work every day. Thank you Emma Bertran for every time you shared your food, growing up with a twin I am used to having someone with me experiencing life at the same pace, you being in the same year, along with me for the ride, helped me greatly over the course of my PhD. Thank you Anna Waldek for always being able to laugh and making me do the same, in particular thank you for giving your Monday evenings a few summers ago to play frisbee with me. Thank you Sabinna Cappo for assisting me without question whenever I needed you, even if you made fun of me while doing it. Thank you Jordon Hemingway for so effortlessly helping me both inthe lab and out of it. Thank you Lewis Ward, I missed you whenever you were away,

xi but your travels inspired me to think about possibilities in my life beyond the obvious. Thank you Katherine Keller for showing me all the ways stubbornness can be an asset. Thank you Haley Olson, watching your enthusiasm as you developed your scientific skills helped me refocus on the aspects of this work that make me happy. Thank you Maya Gomes you were the perfect office mate forthe time we shared and I have missed you continuously since you went away. Thank you Erik Sperling, even though we only overlapped a short while you always welcome me wholeheartedly when we meet outside Harvard. It is a pleasure to be at a university where professors value the success and happiness of all the students. My committee in particular has helped me along my academic path. When I had a science question there was nobody better to go to than Ann Pearson, somehow you always knew the answer and were excited to share it. Whenever I visited Andy Knoll to give him an update on my research progress I always left his office feeling more confident in both my research andmy own self worth. Selfishly, I hope you never truly retire so that every future student can be, at least a little bit, guided by you. Other professors also helped me greatly. Most of my samples come from Francis Macdonald, it must have been hard to lug all the rocks off the mountain just to give them away, and yet you did, always with a smile. Thank you John Shaw for keeping this department running so smoothly and for encouraging a friendly atmosphere, leading by example. Stein Jacobsen, thank you for believing in my potential when I did not yet see a path for myself. Dozens of fellow students have given me help with science or even just company when I needed it. This is only a small list of the people I am grateful to, Tom Laakso, Jocelyn Fuentes, Ana Gonzalez Valdes, Blake Hodgin, Ned Kleiner, Kimee Moore, Will Steinhardt, Yanpeng Sun, Franklin Wolfe, Daniel Cusworth, Athena Eyster, Eugenia Hyung, Harriet Lau, Marena Lin, Jenny Middleton, Rita Parai, Tamara Pico, Judy Pu, Ben Thompson, Scott Wieman, and Elise Wilkes. I do in particular want to thank Evelyn Powell, you have cared for my well-being often more than I cared for it myself. Your friendship has become deeply important to me.

xii All the EPS staff have worked tirelessly to keep the department running and to smooth the path for me. Thank you to Sarah Colgan, Maryorie Grande, Esther James, Paul Kelley, Chenoweth Moffatt, Annika Quick, and Summer Smith for everything you’ve done (not just giving me cookies every day). I would like to thank Lynn Kirby and Alicia Ruch-Flynn, the teachers who, while they may not know it, set me on the path to becoming a geologist as I am now. With them I never felt like just one of many pupils, they always had time to teach me what they could and encouraged me to learn myself whenever my curiosity outstripped their knowledge. My path may have been decided when no one else signed up for the meteorology competition and Lynn handed me a textbook and told me to give it my best shot. It would be irresponsible of me to thank the professors who taught me here at Harvard and not also acknowledge my professors at Rice University, in particular Rajdeep Dasgupta, Helge Gonnermann, Cin-Ty Lee, and most of all Carrie Masiello. It was her enthusiasm for the subject that made me want to study biogeochemistry, and it was her generosity that invited me into her lab and helped me along my path. They all seemed genuinely happy to be there, teaching, researching, and supporting their students. Finally I want to thank my family, mom, dad, and Emily, for getting me this far. I still remember how after my qualifying exams a giant basket of fruit and chocolate was waiting for me, a gift from you in my office. I don’t even knowhow you managed the logistics of that but it was clear that you were happy for my successes, even from hundreds of miles away.

xiii 1 Introduction

1 1.1 Ocean and geologic record

It is easy, when viewing a sweeping canyon or imposing mountain range, to feel the age of the Earth and consequently to view geology as a static unchanging thing. This is however a misconception stemming from the long timescales commonly associated with geologic processes. In fact, the Earth is constantly evolving, something that becomes more apparent after studying the river that dug the canyon or the volcanic eruptions that formed the mountain range. Chemical elements constantly flow between the atmosphere, geosphere, and biosphere, sometimes slowly like in the movement of the continents, sometimes quickly like in the aerobic metabolism that sustains complex life. Appropriately, this flow is centered around the hydrosphere; water is a critical component of every other of Earth’s ”spheres”. From the broadest perspective, my research addresses a facet of these interactions, in particular I seek to better understand how the ocean builds the sedimentary record, and then use that record to understand past environments. The ocean can be treated as a chemical reservoir that balances the chemical input to the ocean, largely from rivers, with the chemical export from the ocean, via sedimentation and reactions with oceanic crust [132]. The ocean is net-depositional rather than erosional, sediments will fill available accommodation space. From a historical geology perspective, the sediments precipitating from the ocean are the most important record of past environments. A fraction of sediments, originally deposited in the ocean and carrying a signal of

2 that deposition, become lithified and preserved as a sedimentary veneer on continental or oceanic crust. Through the laws of superposition, original horizontality, and nuclear chemistry, sedimentary rocks can be dated, providing a temporal record of the depositional conditions. It is then the onus of a geologist to rigorously measure some aspect of the sedimentary record and use those results to develop a logical interpretation of past environmental conditions. This record will largely reflect the history of the ocean, evidence of biological or atmospheric processes will only be preserved insofar as those processes interact with the marine realm.

1.2 Evolution of the Earth’s surface

The sedimentary rock record has long been used to tell stories about the evolution of the Earth. For instance, with respect to the silicon cycle, changes in the locus of silica precipitation and fossil evidence of silica secreting organisms indicate that dissolved silica concentration in the ocean has decreased through time [109, 135]. The long term steady state of silica in the ocean indicates that, throughout this time, silica sinks from the ocean have balanced the inputs. Without necessarily changing the net flux, the evolution of silica precipitating organisms caused silica to be more efficiently removed from the ocean, decreasing the marine SiO2 concentration. Other environmental records are currently less constrained. Temperature for example is a crucial component in maintaining the habitability of Earth, but no definitive temporal record has been developed. Models of stellar evolution

3 indicate that solar luminosity was 75% of the modern value when Earth formed and only slowly brightened to its present intensity, the so called ”faint young sun problem” [62]. This observation should correspond with geologic evidence of low temperatures, but evidence of widespread glaciation is irregular, occurring only during specific periods of the Precambrian84 [ ]. Additionally the oxygen isotope record can be interpreted as reflecting warm oceans early in geologic history, an interpretation that is inconsistent with the observed glaciation [96]. This question of evolving surface temperatures is intimately linked to

reconstructions of the carbon cycle. Atmospheric CO2 is one of the primary greenhouse gasses regulating Earth’s temperature. Interpretations of past temperatures then make predictions about the co-occurring carbon cycle and vice versa. Reconstructions of the carbon cycle, however, show no secular change through time. This is in stark contrast with other records that do exhibit possible variations through time, like temperature and biodiversity [77]. The inconsistencies among these complementary chemical records suggest additional complexity and potential for fruitful investigation, in particular when pursuing a marine temperature record.

1.2.1 Transition points in geologic history

Many times throughout geologic history changes in environmental conditions did not occur through the accumulation of incremental steps. Instead, geologic records will often reveal large perturbations. Sometimes after the perturbation the relevant record returns to the previous value exhibiting a strong negative

4 feedback, but other times the perturbation results in a permanent shift. A famous example is the Cambrian explosion, where after billions of years of minimal physiological development new animal body plans rapidly emerged [138]. In periods of biogeochemical stability it is hard to determine how different element systems interact with each other. Only when one element system changes can we observe the resulting perturbation in the geologic record of a second phase. Studying the late Neoproterozoic is therefore ideal. In addition to variations in physiology, perturbations also occur in Earth’s surface redox, climate, and carbon records [77, 85, 186].

1.3 Isotope methods

One of the most powerful tools to learn about past environments is through the analysis of the stable isotope composition of geologic samples. Classically five isotope systems have been most closely studied: hydrogen, carbon, nitrogen, oxygen, and sulfur. In this work I will focus on interpreting carbon and oxygen isotope records. Carbon has two stable isotopes 12C and 13C, approximately 98.93% and 1.07% abundant respectively, which differ only in the number of neutrons in the atom. All natural carbon bearing species are composed of subtly different combinations of the two isotopes. The isotopic composition of a substance is defined using delta notation which compares a ratio of the heavy to light isotope in a sample to

5 the isotope ratio of a defined standard: (( ) ) 13C 12 13 = ( C )sample − ∗ . δ C 13C 1 1000 (1.1) 12C VPDB Isotope effects can be generated in two ways, through equilibrium fractionations and nonequilibrium (kinetic) fractionations. Equilibrium fractionations occur when isotopes exchange freely between two phases with different molecular vibration states. Substitution of the heavy isotope into each molecule will tend to stabilize the bonds, and the degree to which a phase is stabilized is defined by the change in zero-point energy, an attribute specific to the bonding environment [16, 27, 161, 204]. Even with no net reaction, the heavy isotope will become concentrated in one phase in order to minimize the free energy of the system, thus generating an isotope fractionation between the two phases. Importantly, this tendency is temperature dependent, an observation that allows equilibrium isotope effects to serve as paleothermometers56 [ ]. One example of an equilibrium fractionation process is the exchange of 13C between dissolved CO2 and bicarbonate [145, 221]. The heavy isotope will tend to become concentrated in the molecule with stronger chemical bonds, so in the

12 13 − ⇐⇒13 12 − isotope exchange reaction CO2 + H CO3 CO2 + H CO3 bicarbonate will become enriched in 13C [220]. Nonequilibrium isotope fractionations occur in incomplete reactions or unidirectional processes in which the reaction rate of one isotope differs from the other. In kinetic isotope reactions the light isotope tends to react more readily leaving a product depleted in the heavy isotope while the reactant becomes

6 correspondingly enriched [17, 42]. One such example of a kinetic isotope effect is the discrimination of 13C by the enzyme RuBisCO during carbon fixation [61, 76]. It is often a fair assumption that geologic reactions occur at equilibrium due to their long timescales, but the opposite is not necessarily true. Equilibrium between isotope species can be established quickly, for example the previously described bicarbonate reaction reaches isotopic equilibrium on the order of 10 seconds [220]. Properly distinguishing between kinetic and equilibrium isotope reactions has become important in modern higher order isotope studies [54] and will be relevant in the subsequent chapters of this thesis. Oxygen has three stable isotopes 16O, 17O, and 18O, approximately 99.76%, 0.04%, and 0.20% abundant respectively. Between the two minor isotopes, most studies only report the isotopic composition of 18O (using the delta notation described previously, δ18O)[40]. δ17O can be measured, but is usually neglected for several reasons. 1) 18O is more abundant than 17O, resulting in correspondingly stronger signals during mass spectrometry. 2) When measuring

O-isotopes in CO2 (particularly common in carbonate chemistry) it is difficult to distinguish between 17O and 13C substitution. 3) Most importantly, the magnitude of mass dependent isotope fractionations scale to the relative mass difference between the heavy and light isotopes. The approximately 2AMU difference between 16O and 18O produces δ18O fractionations that are approximately twice as large as a corresponding δ17O fractionation [216].

7 Specifically the fractionation factors are related by the equation:

17α = (18α)θ (1.2)

Where α is a fractionation factor between two oxygen bearing species (e.g.

18 16 18 16 α = ( R/ R)sample/( R/ R)standard), and θ is the corresponding mass law (typically has a value of about 0.52). The larger δ18O fractionations are correspondingly easier to resolve than the δ17O fractionations, and the second measurement will usually provide no additional information. It was later discovered, however, that not all fractionation processes follow a mass law of θ ≈ 0.52. Carbonaceous chondrites, a class of meteorites, were measured with a mass-independent θ of approximately 1, a result of high energy reactions during solar system formation [35, 198, 218]. Mass independent signals were later found in atmospheric CO2 and O2 due to reactions from high energy radiation in the stratosphere [199]. Once in the troposphere, the mass independent oxygen compositions are unaffected by subsequent mass dependent reactions and therefore serve as a conservative atmospheric signal [9, 125, 217]. It is possible however, even without the contribution of any mass-independently fractionated component, for the δ17O of terrestrial oxygen bearing samples to vary relative to the δ18O fractionations. The θ of an oxygen isotope fractionation varies as a function of the temperature, the mineralogy of the oxygen bearing species, and the type of reaction (e.g. kinetic vs equilibrium) [10, 26, 78, 143]. This variability is subtle compared to a mass-independent ′ fractionation signal and is therefore reported using Δ 17O notation, quantifying

8 17 18 deviations from a proscribed δ O-δ O relationship (θRFL):

( ( ) ( )) 17 18 ′ δ O δ O Δ 17O = 1000 ∗ ln VSMOW + 1 − θ ∗ ln VSMOW + 1 (1.3) 1000 RFL 1000

Subsequently, terrestrial oxygen bearing minerals, particularly minerals like chert which precipitated at low temperature, were found with mass dependent ′ ′ variability in Δ 17O [119, 152, 176]. Δ 17O of mineral samples carries a signal of ′ the environmental conditions during formation. Measuring Δ 17O in addition to δ18O therefore provides a second independent axis of isotope variability with which to reconstruct paleo-environmental conditions. This tool is potentially powerful, but is still under-explored.

1.4 Carbon cycle

In Earth’s surface conditions carbon exists in a wide variety of valence states,

most commonly -4 in methane (CH4), 0 in organic carbon (CH2O), and +4 in carbon dioxide (CO2). Microbial metabolisms regularly catalyze the transformation between these oxidation states, a process that is accompanied by a significant δ13C isotope fractionation [61]. The primary sources of carbon tothe

Earth’s surface are through volcanic emissions of CO2 and rock weathering. Once emitted, carbon exchanges rapidly between organic and inorganic pools before being deposited for long timescales as sedimentary organic matter and mineral carbonate [29]. Isotopically, this system is classically defined where the volcanic

9 h CO2 input flux has a composition of -5 which is balanced by the export flux of h h -25 organic matter and 0 carbonate. By isotope mass balance fo, defined as the ratio of carbon buried as organic matter, is then approximately 0.2. These ratios only apply over long timescales when the corresponding fluxes are balanced. On shorter timescales, variations in some component of this cycle can be preserved in the geologic record as potentially interpretable perturbations in

13 13 δ Corg or δ Ccarb

1.5 Oxygen cycle

Oxygen in the Earth’s surface is most commonly found in the -2 oxidation state, and does not experience the same redox transformations as carbon. Correspondingly, oxygen does not experience the same fractionations carbon does during its redox reactions. The fractionation factor between any two species (for example an oxygen bearing mineral and water) is then most strongly dependent on temperature which makes it particularly suitable as a paleothermometer [204]. Mineral carbonate is most commonly used as a temperature proxy [55, 219], but phosphate [18] and chert [107, 176] are equally viable. At the Earth’s surface, oxygen cycles readily between different organic and inorganic phases, but is isotopically buffered through exchange with ocean water. The ocean’s δ18O can temporarily be altered during glacial periods by trapping isotopically light water in polar ice caps, but the ocean composition can only be shifted on long timescales through interaction with the95 crust[ ]. Constraining the isotopic composition of the equilibrium fluid in a mineral-water

10 pair has continued to be a main source of uncertainty in O-isotope based paleothermometers [207].

1.6 Preservation of sedimentary isotope records

Carbon and oxygen isotope records are only as valuable in preserving a record of environmental conditions as they are they are capable of resisting diagenesis subsequent to deposition [4]. Carbon isotopes in carbonate and organic matter resist diagentic alteration because the carbon concentration of the alteration fluid (primarily water) is much lower than the rocks carbon concentration. Carbonate oxygen, however, even during early diagenesis, is susceptible to isotopic resetting

18 by pore water [171]. For this reason δ Ocarb records are very rarely extended into the Proterozoic; it is assumed that no pristine oxygen isotope compositions remain. Oxygen in chert, however, is more strongly bound, with two covalent bonds to adjacent silicon atoms, and is therefore less susceptible to alteration on long time scales. Even chert will become altered when exposed to water at high temperature for long periods of time and for this reason it is sometimes assumed that the heaviest chert measured from an outcrop, rather than the average composition, shows the initial unaltered composition [163]. In subsequent chapters the possibility of diagenesis must be addressed before any rock’s isotope composition is accepted as primary.

11 1.7 Focus and summary of the current thesis

This thesis had 3 body chapters, all of which explore how environmental conditions become preserved in the geologic record. In the first chapter we describe how the third stable isotope of oxygen, 17O, in chert can be used, in conjunction with the existing δ18O record, as an environmental proxy for the Archean and Proterozoic. Specifically we disentangle the effect of temperature, diagenesis, and evolving marine isotope composition on the chert O-isotope record. Focusing solely on ancient samples, however, introduces a host of complicating variables to our environmental reconstruction. In the second ′ chapter we therefore investigate how Δ 17O signals are produced in the modern ′ diatom dominated ocean. We measure frustule Δ 17O from cultured diatoms to calibrate the relationship between the tightly controlled environmental conditions of a culture and the resulting O-isotope composition. Trying to connect these two isotope records, the Precambrian chert and the modern diatom records, inevitably draws attention to a period in the late Neoproterozoic where biomineralization began to evolve and the silica cycle started transitioning from inorganic to biological control. Maybe not coincidentally, this is a period of extreme biogeochemical upheaval, characterized by perturbations in climate, redox, and carbon isotopes, as well as being the foundation of the Cambrian explosion. To understand this turbulent period better, in the third chapter we interpret in detail the anomalous carbon isotope

12 records of two stratigraphic sections deposited immediately after the Marinoan snowball Earth event. We build a quantitative model to link the sediment record to the unique environmental conditions in the aftermath of the Marinoan. While the following three chapters do not all share an isotope system, they do share a methodological approach, building quantitative models from measured isotope data, and share a central motivation; using novel geologic records to reconstruct environmental conditions.

13 2 The triple oxygen isotope composition of Precambrian chert

In review for Earth and Planetary Science Letters

14 2.1 Abstract

The temperature and chemistry of early seawater have both been inferred from the isotopic composition of Precambrian chert (SiO2), a precipitated mineral formed on or within marine sediments. The δ18O of chert shows a robust quasi-linear increase through time - a signal that has been interpreted in a number of conflicting ways. For example, changing δ18O has been hypothesized to reflect the product of cooling surface ocean temperatures, a signature of evolving seawater δ18O composition, or the product of later stage diagenesis (where measured δ18O reflects the composition of diagenetic fluids). We suggest this uncertainty can be resolved through the additional measurement and ′ interpretation of the minor oxygen isotope 17O (noted as Δ 17O) in conjunction with δ18O. In this study we present a suite of triple oxygen isotope data on stratigraphically constrained Precambrian chert (both peritidal chert nodules in carbonates and iron formation silica). These mineralogically well-defined data

17 allow for the first stratigraphic tests of the fidelity of O in SiO2. We apply a Monte Carlo resampling model approach to test the features of the competing hypotheses noted above but here include critical constraints from 17O. The most parsimonious interpretation of these data suggests that secondary alteration with higher-temperature, meteoric-derived groundwater has skewed an original geochemical signature. Further, our 17O measurements suggest limited (if any necessary) change in the isotopic composition of seawater itself and an equitable, modern-like Archean surface ocean temperature.

15 2.2 Introduction

The temperature dependence of the oxygen isotope equilibrium between chemical sediments and seawater was one of the first and still most profound ways isotope geochemistry has been applied toward the understanding of past environments [204]. Famously, the δ18O record of foraminiferal carbonates was successfully used to determine the temperature and glacial history of the Cenozoic Era [175, 219]. Deeper in Earth’s history, however, it becomes harder to uniquely identify a single process or equilibrium capable of explaining observed secular changes in δ18O records. For example, it is widely accepted that the Precambrian carbonate δ18O record is prone to diagenetic alteration, which destroys the primary signal of the depositional environment [170].

Sedimentary chert (SiO2) also precipitates in temperature-dependent isotope equilibrium with the surrounding seawater or pore waters, and the 3500 million year geologic record of chert δ18O captures a structured, quasi-linear increase in preserved compositions [106]. Chert is hypothesized to be more resistant to diagenetic alteration than carbonate and so may provide a more reliable record of Proterozoic and Archean environments [107]. Most directly, the equilibrium temperature-dependence suggests the evolving δ18O composition reflects a change in ocean temperatures, from a hot Archean world towards present day temperature [96, 107]. Alternatively, however, the same δ18O record can be explained as a proportional change in the δ18O of seawater without invoking any change in surface ocean (i.e. precipitation) temperature [95, 102, 207]. Finally,

16 the chert δ18O record could reflect progressive diagenetic alteration, overprinting original depositional signals [47]. Of course, each of the three hypothesis is consistent with the chert record, as they are derived directly from it. Thus, additional δ18O observations are unlikely to resolve mechanistic questions about the genesis of this record. Previous research has attempted to address this challenge by coupling the oxygen isotope composition of cherts with other, independent constraints on either temperature or the oxygen isotope composition of seawater. The δD [92, 107] and δ30Si records of chert [163], and the δ18O record of phosphates [18] have all been invoked as Precambrian temperature proxies, but these techniques, while informative, have not resolved the debate. The clumped isotope record of carbonates has been used to constrain the temperature and δ18O of Phanerozoic seawater, but has yet to provide as sharp of conclusions in the Precambrian [49, 54, 68, 79]. The underlying source of the chert δ18O signal is therefore still uncertain. Measuring the 17O of Precambrian chert in addition to δ18O may provide important new insight in this debate. The triple oxygen isotope (16O, 17O, 18O) composition of oxygen bearing mineral phases - like sulfate - has recently been championed in paleoclimate studies [9, 43]: mass-independently fractionated, atmospherically derived oxygen is incorporated via oxidative weathering of sulfides into the product sulfate. One major benefit of this approach is thelarge signal to noise ratio that accompanies the mass-independent source/signal. In parallel, detailed, high precision work on geological oxides reveals that chemical

17 sediments do indeed carry a resolvable mass-dependent signal that is most sensibly related to the minerals’ life histories [119]. The same temperature-dependent equilibrium isotope effects that have been argued to explain the δ18O of chert have strict, and definitive, predictions for the companion 17O signal. With that in mind, we report the triple oxygen isotope compositions of a suite of Precambrian chert. The mass-dependent variations in 17O provide an additional axis of variation with which to interpret these now classic δ18O records and allow us to distinguish among the three primary hypotheses advanced to explain δ18O data.

2.3 Methods

2.3.1 Isotope notation

The 17O of oxygen-bearing minerals is not routinely measured, as it is often assumed that the δ17O carries a constant, mass-dependent offset from the canonical δ18O. As a result, the extra effort required to acquire 17O data was long thought to carry no additional interpretable signal [40]. However, high-precision fluorination methods have changed this mindset, whereby mass laws and additional information can be gleaned from 17O measurements on mass-dependent materials. A suite of simple terms then allows for these small mass-dependent effects be quantified and represented visually. At its most simple, the relationship between 17O and 18O during a fractionating process is

18 described by the equation:

17 18 θ αA−B = ( αA−B) (2.1) where

x x RA − = . αA B x (2.2) RB

Here, R is the ratio of the heavy isotope relative to 16O. The θ is the relative isotope effect of 18O compared to 17O for a specific reaction. Values for θ are approximately 0.5, but vary at the percent level as a function of process and temperature [26, 216]. These values range from θ = 0.5305 for high-temperature equilibrium [26] down to θ ≈ 0.5 for the kinetic fractionation of massive oxygen bearing compounds and forces like gravity [216]. These different mass laws then manifest as resolvable and measurable anomalies at the ppm level and presented in standard delta notation: ( ( ) ) xO 16 x = ( O) sample − ∗ δ O xO 1 1000 (2.3) 16O standard ′ where x is 17 or 18. In order to visualize small variants, Δ 17O is employed. Here,

( ( ) ( )) 17 18 ′ δ O δ O Δ 17O = 1000 ∗ ln VSMOW + 1 − θ ∗ ln VSMOW + 1 (2.4) 1000 RFL 1000

17 where θRFL is the reference line to which the δ OVSMOW values are normalized.

19 17 We chose to calculate Δ’ O using a θRFL of 0.5305 because it is a value derived from a thermodynamic prediction, not the average slope of a set of natural samples, and will therefore not change as a consequence of increased or improved measurements. All data provided is placed on a VSMOW scale by replicate measurement of three common calibrated silicates (NBS-28, UWG-2, and San Carlos Olivine) [3, 119, 152, 176, 195, 217, 218].

2.3.2 Sample suite

The target of this work is to assay the long-term change in the oxygen isotopic composition of Precambrian chert in order to evaluate its information richness. To do so, we collected and analyzed 90 chert-bearing samples that range in age from early Paleoarchean to Cambrian, with highest sample density in the Neoproterozoic. Within the Neoproterozoic stratigraphy, we also target and perform a stratigraphic test. Finally, from among these samples, we sub-sampled some rocks with distinct chert phases or nodules, resulting in 105 total sample measurements. Importantly, all these target localities predate the advent of biologically driven opal precipitation, and as such reflect chemical (abiotic) precipitation [183]. A full sample list and locality information is included within the supplement, however two depositional environments dominate the stratigraphy represented. First, most Proterozoic samples reflect peritidal environments, where they precipitated as nodules during early diagenesis [108]. In such environments, supersaturated silica forms an amorphous gel which matures into microcrystaline

20 quartz [135]. Due to the early crystallization, the host fluid in these sediments is often assumed to be comparable in temperature and isotopic composition tothe overlying ocean. Juxtaposed to these settings, we also include chert from early Proterozoic and Archean iron formations. Here, the exact depositional environment can vary but is generally basinal, with silica adsorbed onto iron-bearing precipitates from seawater[12, 63, 160, 200]. Other early Archean cherts reflect diagenetic alteration of primary sediments, at least in part under the influence of hydrothermal fluids137 [ ].

2.3.3 Chemical methods

The goal in the chemical pre-processing of each sample was to isolate the quartz fraction from any contaminating oxygen bearing phases. For each sample, approximately 1 gram of chert was subsampled and powdered using a tungsten carbide shatterbox. Where texturally appropriate, multiple subsamples were powdered to access separate nodules or visually distinct textures. The powdered samples were acidified in 2 N HCl to remove carbonate phases, then filtered,

o dried, and baked at 500 C to remove organic matter. This leaves 2a SiO rich insoluble residue. Detailed mineralogy, on both the pre- and post- acidified samples were conducted at Oxford University (See SOM).

18 17 The δ O and δ O of the SiO2 were measured by laser fluorination at Louisiana State University [8] and Harvard University [38]. At LSU, the samples are

prefluorinated overnight in a reaction chamber under a BrF5 atmosphere, and

then heated by a 30 W CO2-laser to generate O2. Oxygen from the samples is

21 −→ converted to O2 by the reaction 2SiO2 + 3BrF5 2O2 + 2SiF3 + 3BrF3 [33].

The O2 yield, calculated by comparing O2 pressure to the initial mass of chert, is consistently above 90%. The purified O2 is introduced into a separate Finnigan MAT 253 isotope ratio mass spectrometer and analyzed in dual inlet mode. Measurements of silicate standards, NBS-28, UWG-2, and San Carlos Olivine, ′ suggest a reproducibility of 0.3h and 0.018h for δ18O and Δ 17O, respectively. At Harvard, the samples are again prefluorinated overnight, but combusted bya

50 W CO2-laser under a pure F2 atmosphere. Sample purification is similar to above, however with the inclusion of an in-line gas chromatograph before being directly introduced to a Thermo Scientific MAT 253 gas source isotope ratio mass spectrometer. Repeated measurements of the same silicate standards ′ generated a reproducibility of 0.4h and 0.015h for δ18O and Δ 17O, respectively. For chert samples analyzed via both methods, the agreement is well under the quoted, conservative, estimates of error from silicate standards (Figure 2.7.1).

2.4 Results and Discussion

In the present study, we add to the existing Precambrian record of δ18O in chert and expand by including the minor oxygen isotope, 17O. As seen in Figure 2.4.1, the new measurements of δ18O nicely align with previous work. Although the exact nature of the change in δ18O through time is debated, the progressive increase in δ18O is undeniable. As our new measurements also capture this trend, the accompanying 17O data can confidently be used to assess the source of the full

22 Figure 2.4.1: Geologic record of measured chert δ18O from this study (circles, squares are from [119]) with a backdrop of 500 Ma bins from literature data [106]. The red line shows median value, the box represents the 25th and 75th percentile, the whiskers are 1.5 times the scale of the box. The color coding of the data corresponds to the age of the sample (x axis) and is consistent in subsequent figures.

23 oxygen isotope signal. It is critical to appreciate that the fluorination methods used here will access all oxygen-bearing phases within a sample, including any contaminants. As such, we begin with an evaluation of the mineralogical content of the targeted material (Figure 2.4.2). Here, about 80% of the chemically treated samples consisted of more than 90% quartz. Our chemical preparation method effectively removed carbonate, but the remaining 20% of samples had significant fractions of primarily clay and iron oxide. These other oxygen bearing phases risk

contaminating the primary SiO2 signal. In fact, samples with significant non-silica phases are on average 7.2h lighter in δ18O (and only 0.02h heavier in ′ Δ 17O). We thus focus only on the samples with greater than 90% quartz fraction1. The complete mineralogy data is presented in Table 2.7.3. From the now curated dataset, a clear secular increase in the δ18O of chert is apparent, from Archean compositions averaging below 20h toward late ′ Precambrian values above 30h (Figure 2.4.1). The corresponding Δ 17O composition is consistently negative, ranging from -0.04 to -0.26 h. The negative

values are a product of our using θRFL = 0.5305, the maximum possible slope, as ′ our reference when calculating Δ 17O and are consistent with the existing literature measurements [119]. The isotope measurements are presented in Table 2.7.1 and Figure 2.4.2.

1It is noteworthy that other environmentally relevant information could be locked within these coexisting mineral phases.

24 Figure 2.4.2: All of the measured δ18O and Δ’17O data color coded by the quartz fraction. The most quartz poor samples also correspond to the anomalously light δ18O samples in the Neoproterozoic. The majority of samples are more than 90% chert and the model results in the main text correspond with these values.

2.4.1 A stratigraphic test: the Fifteenmile Formation

Before we can investigate the differing models proposed to explain the information locked within the δ18O of Precambrian chert, we must first test the scale at which chert is vulnerable to isotopic change. Put simply, if there exists a wide compositional range within a given hand sample or across a well mixed paleo-basin, interpreting the same magnitude of signal across geological time would not be prudent. The first test, variability on the hand sample scale, was performed by comparing the isotopic composition of different nodules cut from ′ the same sample. The mean difference of measured Δ 17O values is 0.027h. This is slightly larger than the reproducibility of repeated measurements from the same standard (but < 2 σ), but still far smaller than the overall range of compositions in the dataset. The later of these tests, capturing how robust the 17O

25 in chert is across a single basin, is conducted on the Tonian aged Fifteenmile Formation. We present data from two parallel stratigraphic sections. The lion’s share of the samples come from a 20 meter section of Mt. Slipper (section P1401, see [36]). The remainder of the data was generated on samples from the contemporaneous beds within the Fifteenmile, roughly one kilometer away (section P1405, [36]). These data test for variability within a given stratigraphic ′ section and across a paleo-basin. The mean Δ 17O from the P1401 section is -0.200 ± 0.022h, whereas section P1405 carried a mean of -0.206 ± 0.024h. The two sections are thus statistically indistinguishable. For further context, the reproducibility of standards is near the statistics on these natural populations. Hence, the variability we observe within a single section is indistinguishable from ′ the variability we expect based on our capacity of make precise Δ 17O measurements. This does not imply that the samples are unaltered, but simply that a cross section of samples at that geographic scale can share a loosely similar life history. Indeed, there is resolvable δ18O variability within the basin, however governing mass laws (or θ values) will not manifest an appreciable change in ′ Δ 17O over such a small-scale δ18O range. This would suggest that any alteration process must carry a θ between 0.5305 and 0.527 in order not to have generated ′ an accompanying resolvable Δ 17O. It is noteworthy that meteoric water falls a long a line of 0.528, a point considered more deeply below.

26 2.5 Geological Scenarios

The new data generated through this study are in keeping with previous δ18O measurements, these samples can be interpreted as a ’representative’ sample suite ′ with which to test pre-existing hypotheses in light of the new Δ 17O data. We do so in turn. We first individually consider changes in ocean temperature and the δ18O composition of seawater itself, before ending with a discussion of diagenetic over-printing.

2.5.1 A change in global temperature

The new δ18O measurements in this study could be explained by a progressive decrease in mean ocean temperature over the last 4 billion years. The decreasing temperature manifests as a change in the equilibrium isotope fractionation factor 18 h (αchert−water), offset relative to an ocean of constant composition (δ O = 0 ). In this scenario, the median δ18O in the Archean corresponds to a temperature of over 120oC, which subsequently evolves to a 38oC Phanerozoic ocean. The physical interpretation of this temperature is itself controversial ([92, 196, 207]); ′ however, an agnostic approach is to simply address the consistency of the Δ 17O prediction. The existing δ18O record corresponds to a unique temperature trend through geologic time, and these supposed temperatures in turn predict a corresponding ′ Δ 17O composition assuming equilibrium fractionation [176]. The equilibrium θ, ′ which controls the Δ 17O, becomes smaller at lower temperatures, which

27 ′ Figure 2.5.1: The Δ 17O in Precambrian chert can be used to test various ′ hypotheses related to the change in Δ 17O over time. Here, predictions for changing ocean temperature (in red) and changes in the composition of actual seawater - a function of high versus low temperature alteration (in green) - are captured as residual plots. That is, the difference between modeled predictions and the suite of measured samples. Importantly, in the case of the evolution of seawater, geochemical model is involved in setting the prediction.

28 ′ ultimately produces a concave downward trend in Δ 17O versus δ18O (see Fig. ′ 2.7.3). The degree of variability in Δ 17O does not allow for a test on this ′ predicted concavity, however the measured Δ 17O are consistently more negative than the predicted trend (on average, ≈ 0.05 h: Figures 2.5.1 and 2.7.3). Based on this mismatch between predicted and measured compositions, the hypothesis that variability in the chert record is produced strictly by changes in the ocean ′ temperature is not uniquely supported by the Δ 17O data. It is noteworthy that most studies that advocate for elevated temperatures make the additional assumption that the heaviest δ18O composition for each time-period, not the median value, reflects the primary composition163 [ ] - this allows for some degree of secondary alteration. Based on this interpretation, the Archean ocean would still be at 70-80oC, but no longer nonphysically hot. Unfortunately, this interpretive approach does not further minimize the ′ mismatch between predicted and measured Δ 17O values. Changing the ′ temperature prediction would shift the Δ 17O prediction along an equilibrium fractionation array, but it would not shift the array towards the field of measured ′ compositions (see Fig. 2.7.3). The calculated equilibrium Δ 17O being offset from the measured data means, regardless of the temperature, a single equilibrium formation process will not capture the bulk of the geologic record. Although not a perfect fit, what is unclear is the efficacy of the temperature model as a solution when placed relative to the other existing hypotheses.

29 2.5.2 changing the composition of seawater

18 Another possible explanation for the change in the δ O of SiO2 through the Precambrian (and by extension, throughout Earth history) is that the actual δ18O of seawater is evolving. Here, the equilibrium offset may be unchanged (i.e. no change in surface temperatures), but the apparent effect in chert drifts asa function of changing seawater composition. The δ18O of seawater is set by interactions at both high and low temperature with basaltic oceanic crust [150]. High-temperature exchange occurs at ridges and deep in the ocean crust whereas low-temperature exchange happens during weathering of continental crust and upper oceanic crust. This manifests isotopically because high-temperature reactions are characterized by 18α=1.000 and low temperature reactions are characterized by 18α=1.020 - 1.045 as a function of temperature. Continental and oceanic crustal isotopic values are buffered by recycling with a mantle, assumed to be infinitely large with constant δ18O composition. Modelling treatments ([95]) then allow the evolution of seawater by changing the dominant mode of the ocean-rock interaction from low-temperature exchange in the Archean to high-temperature exchange in the present. Here we rebuild a popular model ([95]) - designed to describe the δ18O of carbonate - with the eventual extension to chert 17O to test this hypothesis. This required the usage of a low-temperature 18α=1.028, which then enables a test of the variable contribution from different weathering regimes. First, reproducing the entire chert δ18O range requires more than 90% of the Archean ocean crust

30 Figure 2.5.2: As the seawater composition, shown in the black dashed line, evolves in δ18O there is a corresponding change in Δ’17O. The predicted chert composition parallels this change in seawater, but is offset from the measured data. This demonstrates a prediction for how the composition of seawater itself would change with time (array in upper right), but does leave a significant residual between the data and model.

exchange to occur at low-temperature and more than 90% of the Phanerozoic ocean crust exchange to occur at high temperature (Figure 2.5.1). The test comes with the assignment of 17O fractionation factors to each of the geologic pools and to each of the exchange fluxes based on the predicted temperature of that ′ exchange [152, 176]. This then predicts a Δ 17O composition of seawater in parallel to the δ18O value. The evolving ocean oxygen isotope composition, as well as the chert composition of precipitated chert in equilibrium with that ocean, are shown in Figure 2.5.2. Like with the temperature prediction discussed previously, the predicted chert trajectory parallels the triple isotope trend of the data, systematically offset towards more positive values (or leaving a negative residual, see Figure 2.5.1). The isotopic mismatch of the changing seawater

31 hypothesis is smaller than the changing temperature hypothesis, but still insufficient by itself as a mechanism for explaining the Precambrian chert record. It is also noteworthy that an unchanged seawater δ18O composition is in keeping with Phanerozoic studies of carbonate clumped isotopes [79].2

2.5.3 Diagenetic alteration

The previous hypotheses forwarded to explain the isotopic composition of

Precambrian chert rely on the capacity of SiO2 to retain a primary depositional signal despite potential later interactions with geological fluids. In fact, in the previously described case where only the heaviest δ18O are interpreted for an extracted temperature record, some degree of diagenetic vulnerability is inferred. Great efforts have been taken (cf. [4, 50, 136]) to understand the degree to which the geochemical and isotopic composition of rocks and minerals have been altered through time. The oxygen isotope composition of geologic samples are particularly susceptible to diagenetic resetting given the water-rich, high oxygen content of alteration fluids. This case becomes even more exacerbated when thinking about a largely peritidal facies, which would be subjected in even the shortest term to changes in alteration fluid as a function of sea level change. This

2It is worth noting that the slope of the oxygen isotope trajectory of seawater in Δ’17O- δ18O space is controlled primarily by the θ of the low-temperature exchange processes. If low- temperature ocean-rock exchange is driven using a larger θ, closer to the high-temperature end- ′ member of 0.5305, then there would be less contrast in Δ 17O between the high- and low- ′ temperature exchange. The oceans δ18O could then evolve without changing Δ 17O, thereby causing the chert trajectory to intercept the bulk of our measurements. Due to the temperature dependence of both α and θ it is impossible however to independently change one without affecting the other, and using a larger value for θ would require a correspondingly smaller value of α, which would make it impossible to reproduce the δ18O record.

32 Figure 2.5.3: The simplest approach to model alteration of geological units is to assume a random, time-dependent exposure to fluids with time. For this exercise, the end-members are a modern chert composition and a high- temperature alteration fluid that would carry a less enriched δ18O. In gray is the δ18O data from this study. Colored lines reflect examples of individual model simulations, whereas the heavy red line is the mean of 1000 simulations.

proposed diagenetic alteration could come anytime between lithification and the field season the sample suite was collected. A simple stochastic exercise demonstrates such a phenomenon. In a purely synthetic sense, if alteration is random and allowed to occur at variable (random) magnitudes and rate through time, a time-dependent record (Figure 2.5.3) is predicted. Importantly, this exercise is devoid of geological observables, but demonstrates alteration-potential and as meant as simply an example. Diagenetic alteration generally carries a non-unique, rather than specific prediction for the isotopic composition of chert. The most basic approach is first

33 adopting the null hypothesis that all SiO2 precipitated with an isotopic composition comparable to today, and any deviation from this composition is the result of alteration. From here, three primary variables control the ultimate composition of chert: the temperature and isotopic composition of the secondary fluid, and the degree of alteration (water to rock ratio). Not surprisingly, each variable can be tuned to reproduce the geologic record of δ18O ′ and Δ 17O. ′ Like δ18O, diagenesis carries a suite of predictions for the associated Δ 17O. Here, mixing between two isotopic compositions produces a concave upward array, distinct from equilibrium fractionation lines. The curvature becomes more pronounced as the two end-members become more isotopically different. Due to the concavity, mixing between a modern-like initial chert composition and SiO2 reset by interactions with higher-temperature fluids can easily satisfy the δ18O ′ and Δ 17O of Precambrian chert. As discussed more below, this is in part because of the flexibility in parameters noted above, which allow the data to essentially be fit, minimizing the residual between model predictions and the data. The goodness of the model fit serves to inform the mechanisms acting onthe original isotopic composition of SiO2. It is apparent that alteration with a fluid carrying a modern seawater-like seawater composition does not best fit the Precambrian data. Early diagenesis at some stage in the life-history of the sedimentary unit may indeed be seawater-buffered, but later diagenesis can be driven by fluids sourced from meteoric water with high variable compositions [120]. Alteration with meteoric water has two important consequences in the

34 context of this hypothesis. First, this fluid is lighter in δ18O than seawater, thereby forming a larger mixing curve with the primary chert that subsequently results in ′ more negative Δ 17O compositions. Second, meteoric water has a range of compositions rather than a unique value, therefore producing a field of possible solutions rather than a single unique solution. The single mixing line which minimizes the residual between model and data, shown in Figure 2.5.4, defines the initial chert to be precipitated at 30oC and have been altered by a fluid (likely groundwater) groundwater with a δ18O composition of -16.5h. Both of these values are reasonable on the modern Earth, and this solution supports a uniformitarian perspective of Earth history whereby the temperature and isotopic composition of the ocean has remained stable. The possibility of alteration is not a bi-modal distinction, but the suggestion that more than a primary depositional condition is recorded in the composition of the targeted chert. Admittedly, this hypothesis suggests that some primary geochemical information is irrevocably lost. However, even the most altered Archean samples only experienced a W-R ratio of 1 (discussed in detail below), and thus the majority of the signal is still primary. While it will be difficult to identify a unique initial composition for any individual chert, especially without a tightly constrained composition of alteration fluid, a substantial fraction of the signal remains.

35 Figure 2.5.4: Alteration with high-temperature fluids generates a new δ18O composition that then can effectively mix with an original depositional composition. Here, we model such behavior. We allow for alteration fluid to fall along a meteoric water line (mwl). The offset from this line, set in magenta, is the high-temperature equilibrium fractionation factor for the new, now modified SiO2 composition. From here, the mixing line ties this value to the original, set as a black square (lower right of the data population). The solid gray line captures a best fit for the entire data set, requiring a water value of- 16.5h, alteration at 200oC, and a primary precipitation temperature of 30oC. The gray dashed lines capture different meteoric water values.

36 2.5.4 a multivariate approach

Part of the difficulty in distinguishing between the three primary hypotheses surrounding the long-term change in δ18O is that they are not mutually exclusive. ′ All three mechanisms could, in conjunction, influence the δ18O and Δ 17O composition of the sampled chert. Above, we analyzed the hypotheses in ′ isolation, showing how the mechanisms would be expressed in Δ 17O - this suggested that diagenetic alteration provides the most statistically robust single answer. This comes as little surprise, given the flexibility (in parameter space) associated with that solution. Moving forward, however, we can evaluate the relative contribution of these different environmental solutions/parameters in parallel. Here we provide a synthetic resampling routine that allows all of the features of interest to vary within reasonable bounds. The variables, shown in Table 2.7.4, include the degree of ocean evolution, the degree of glacial or evaporative seawater distillation, the temperature of precipitation, the temperature and composition of the alteration fluid, and the degree of alteration (water to rock ratio). The final chert composition is calculated using the equation:

(( ) ) ∗ − ∗ 1x 1x W/R αf−c,highT 1 1000 + δ Of,lowT + δ Oc,lowT 1x δ O − = (2.5) c altered W/R + 1

where W/R is the water to rock ratio, the superscript 1x represents either 18O or 17O, the subscript f represents the alteration fluid, and the subscript c represents

37 the chert. This style of resampling analysis provides information about both what variables are most likely to control the output isotopic composition and also over what range of values each variable is most likely to cover. In this approach, the analysis uniformly generates random values for the six ′ primary variables and derives a predicted chert δ18O and Δ 17O composition for that combination of variable inputs. If the calculated synthetic chert composition corresponds to an actual measured chert composition in the dataset to within ′ 0.1h for δ18O and 0.01h for Δ 17O (conservative bounds, given our uncertainty), the values of the six variables are saved in association with that chert sample. This generates a curated suite of variable sets, or mathematically viable solutions for each individual data point. In fact, each datum is described by many sets of the six variables. As with a Monte Carlo resampling regime, the curated results will carry a distribution that reflects the specificity or sensitivity of that variable to the overall chert composition. The superposition of these individual datasets captures the average environmental conditions required to reproduce the Precambrian triple oxygen isotope composition of chert. This analysis was run 5 ·107 times, with 3.5% of the random solutions satisfying the composition of at least one measured value. The full, uncurated field of solutions is presented in Figure 2.5.5, with our measured data plotted on top of it. The synthetic resampling routine generated a field of interesting, and interpretable information. Figure 2.5.6 provides a probability density function for each of the parameters tested. In cases where the histogram carries structure, there is a specificity within that parameter space necessary to account for the

38 Figure 2.5.5: Plot of δ18O vs Δ’17O showing the complete probabilistic results of the stochastic model - in red - compared to the measured chert isotope data. The shade of red reflects the density of results for that value. The tested variables and values thereof are presented in Table 2.7.4. The grey and black envelopes show the position of 95% and 66% of the results respectively. The centroid of the synthetic results plots above the measured data because, as established in the main text, temperature and ocean change driven fractionations produce Δ’17O values more positive than the measured data.

39 observed chert compositions. In the cases where little structure is present (i.e. a flat line), that variable - according to this analysis - does not carry much sensitivity or likelihood of contributing to the overall preserved chert values. Of the histograms shown, some directly correspond to the three primary hypotheses discussed above. First, Figure 2.5.6a, in blue, shows the predicted equilibrium precipitation temperature, which in our model corresponds to contemporaneous ocean temperatures. Here the suggestion is for an equitable, modern-like climate where the median seawater temperature varies around about 23 oC. Few solutions recorded elevated primary surface ocean temperatures. Next, and as captured in Figure 2.5.6b, in red, we evaluate the chances that changes in the δ18O of seawater itself are actually driving the observations captured in the chert record. Interpreting this variable brings with it more ambiguity, as the data can all be satisfied without a change in seawater δ18O, however some moderate change to low:high temperature alteration does describe the median solution. As alteration hypotheses regarding the balance of high and low temperature weathering regimes through time are rooted in geophysical arguments related to heat loss, the likelihood of an intermediate state or weathering balance/solution should be ′ evaluated on those grounds3. Small changes in seawater δ18O and Δ 17O can also be accommodated with variable ice volume, captured in Figure 2.5.6f, but the

3With respect to seawater δ18O evolution through time (Figure 2.5.6b), the analysis has seem- ingly little dependence on this constraint. However, to further probe the model of[95], the model was run twice, with the glacial fraction defined as 0 and the ocean evolution defined as either 1or0. The ocean with evolving isotopic composition had only77% as many model solutions as the isotopically ”constant” ocean. Additionally, the evolving ocean model did not produce solutions for 10 of the measured samples, whereas the ”constant” ocean carried solutions for all but 4 measured samples. These results further suggest that an isotopically evolving ocean can adequately reproduce the measured data, but an ocean of constant composition is better statistically supported.

40 data is better fit with a modest glacial fraction. Finally, Figure 2.5.6c-e in green captures the principle components of diagenesis. For instance, we present predictions for water to rock ratio, alteration fluid composition and temperature. The W-R prediction suggests that, while some chert can be explained as relatively unaltered - i.e. W-R ratio below 0.1 - the large majority require diagenetic resetting. Further, the predictions here is that alteration occurred at higher temperatures (where equilibrium fractionation factors vary less per unit T) with fluids of a meteoric-like composition - none of which is geologically unreasonable. There is an important linkage here between the temperature predictions and geologic observables. We define high-temperature alteration as elevated over seawater temperatures, but still could be as low as a couple hundred degrees (or less). Thus, scaling to something like metamorphic grade is challenging. Simply presenting a histogram of compiled W-R ratios as in Figure 2.5.6c is not sufficient to establish that variable alteration drives the classic chert δ18O record. We must show that the progressively lighter δ18O compositions are matched in the model with progressively more alteration. To do this we found

Figure 2.5.6 (following page): Compiled results of the Monte Carlo model for each tested variable. In each plot the x axis shows the allowed range of variation for the specific variable, the y axis shows the relative abundance of that value among inputs that reproduced measured data. The water to rock ratio shown in c is displayed with a logarithmic x axis, but like with the other variables, the inputs are generated such that each value in this plotted space is equally probable.

41 42 Figure 2.5.7: Measured δ18O of the chert against the best fit of the modeled W-R ratio.

43 the median predicted W-R ratio for each datapoint in our resampling model, the value that is most representative of the degree of alteration, and plotted it against the measured δ18O (Figure 2.5.7). No corresponding relationship exists between δ18O and the modeled precipitation temperature (Figure 2.7.4). This result shows ′ that the δ18O and Δ 17O records in chert have an important diagenetic component, a conclusion that could not be reached with δ18O record alone.

2.6 Conclusion

The oxygen isotopic composition of Precambrian chert has long stood as oneof the most robust empirical records of Earth surface change, but interpretations of this record are mired in controversy. Complicating these interpretations is the observation of similar temporal patterns in the oxygen isotope composition of Precambrian phosphate and carbonate minerals. Further, arguments based on Si [28] and D/H [92] isotopes within chert must be accommodated by any prescribed life history of this material required for the O isotopes. Here we provide another isotopic axis along which to test these various hypotheses - the inclusion of 17O. The triple oxygen isotope data of Precambrian chert places more specific requirements for various models employed to describe the composition of chert. For instance, through these measurements and associated models we show that changes in the isotopic composition of seawater, and/or elevated Archean surface temperatures are less likely the singular solution for the δ18O record - differential diagenetic alteration has almost certainly played

44 an important role. In fact, the most parsimonious solution calls upon equitable surface temperatures, and secondary fluids that look very meteoric in composition (≈ - 15 h) at common subsurface temperatures of a couple hundred degrees. This implies that some, but not necessarily all original information is lost from these phases. These predictions are consistent with existing Si and D/H isotope data. Inthe case of Si, there is no correlation (at the 2σ level) in the 23 cases where oxygen and Si data are available on the same sample (Figure 2.7.5). This is to be expected, as the silicic acid content of alteration fluids would favor the resetting of oxygen long before similar changes were witnessed in silicon. The hydrogen isotope record in chert has been used to argue for low Archean ocean temperatures [92] - a prediction from this work as well - but arrived at such a conclusion through invoking a temporal change in ocean δ18O. It is possible that the δD of chert, not being a structural component of the quartz mineral, preserves a different environmental window than that of oxygen. If, as we argue, the oxygen isotopic composition of Precambrian chert carries a significant component of alteration, then the adjoined prediction is that phosphate and carbonate experienced similar histories. These parallel records show a shift in δ18O of comparable magnitude over the same time period, but the different oxygen-bearing phases carry differing susceptibilities to alteration. Further work comparing the thermodynamic properties of adjacent mineral phases, perhaps using 17O in these actual minerals, or more generally using the model results presented above, will serve to help quantify these co-occurring

45 changes. ′ In summary, the Δ 17O record of Precambrian chert is not solely a primary depositional signal. The question remains of how to see through this alteration to a primary composition. Of course, this model (and statistics therein) allow for some change in the δ18O of seawater, but are significantly less in keeping with major changes in surface ocean temperatures through time. This contributes to the long-standing debate in the geologic community about the origin of this signal and demonstrates the utility of mass-dependent, multiple isotope systematics to Earth history questions.

46 2.7 Additional Information

2.7.1 Geological descriptions

1. The chert-rich, lime-mudstone and grey shale ofthe Fifteenmile Formation, exposed at Mt. Slipper in west-central Yukon, Canada is our most heavily sampled unit [36]. This section is 810.7+-5.8 Ma according Re-Os geochronology data on the organic rich shale horizon 4.15 meters below the base of our section. Additionally, we report data from 7 samples from a parallel Fifteenmile section one kilometer along strike. This section represents a mixed redox, semi-restricted basinal environment. Samples were deposited below storm wave base in a distal slope setting. Chert is present among all lithologies in the section, allochthonous and autochthonous, but is variable in abundance, distribution, and form. Four samples are specifically from the Hard Luck Creek outcrop and are sourced from a yellow weathering dolomite with common intraclast breccias, black chert nodules, shale interbeds, and molar tooth structures [130]. Like most other Neoproterozoic cherts these were peritidal early replacement nodules. This group was intruded by mafic dikes, butthe samples are not taken from directly adjacent. Additional samples are from the peritidal reefal assemblage, a mixed carbonate shale unit slightly lower in the section [72]. Two samples are from Mt Slipper, which are laterally correlative with the Hard Luck Creek samples. These 2 samples are at the top of a dark gray limestone rhythmite unit and are interpreted as subtidal

47 in a shallowing upwards sequence. These Yukon samples experienced Anchizone grade metamorphism corresponding to temperatures of 200-300 degrees as part of the orogeny that formed the mountain range [127].

2. Data is provided from the Rocky Harbour Formation in Newfoundland. These are chert clasts from a pistachio colored siliceous siltstone located immediately above 579 Ma age Gaskiers glaciation diamictite [158]. The siltstone contains wavy laminations and dewatering structures on a storm wave ripple surface indicating a deep water depositional environment.

3. Data is provided from the Neoproterozoic San Juan Formation in southern Peru. The section has undergone greenschist facies metamorphism during an Early Paleozoic orogenesis [32]. Stratigraphically, the measured sample (n=1) is immediately above the inferred Gaskiers cap carbonate and represents the base of the Shuram anomaly. Lithologically, the sample is from a limestone unit above a black shale. Due to a lack of higher energy sedimentary structures this sample is most likely deep marine, but bedding was subsequently disturbed indicating possible alteration (Hodgin personal communication).

4. Data is presented from the Khesen Formation in Khubsugul, Mongolia. These units are peritidal, early replacement chert nodules that precipitated in a dark grey limestone rhythmite. Stratigraphically this unit sits above a glacial diamictite with its associated cap carbonate and disconformably

48 below a phosphorite deposit [128].

5. Data is presented from the Zuune Arts Formation in central Mongolia. Samples were taken from a 100 meter thick unit of blue limestone rhythmite and ribbonite. The same unit that includes the black chert nodules also includes meandering ichnogenera. The oldest chert sample from this unit (≈ 550 Ma) is a massive deepwater strataform, suggesting the possible contribution of hydrothermal silica, while the younger two samples (≈ 544-545 Ma) are peritidal early replacement nodules [129].

6. Data are presented from the Shuurgat Formation, a carbonate dominated sequence located between the Zuune Arts Formation above and the post-glacial Ol formation below. Deposition of chert nodules occurred at the maximum flooding surface of a carbonate ramp. Based on the presence of graded beds of flat-laminated micritic limestone with minor grain flows of redeposited ooids these samples formed in a subtidal environment below fair-weather wave-base [6]. The compaction of laminae surrounding the chert nodules indicate that the nodules formed during early diagenesis.

7. Data are presented from the Ol Formation, a carbonate dominated unit deposited immediately after the Marinoan Khongor diamictite. This unit was deposited as part of a transgressive systems tract on an isolated carbonate platform, analogous to the modern Bahama bank [19]. This chert nodule is also a peritidal early replacement nodule, but considering the conditions after the Marinoan may have formed under

49 higher-temperatures and equilibrated with low δ18O meltwater (Liljestrand, in prep).

8. Data are presented from the Cryogenian Taishir Formation, a largely carbonate unit deposited between the Sturtian and Marinoan glaciations. One sample is from a slightly older outcrop (≈ 660 Ma) of limestone calcisiltite and thin-bedded micrite. The sedimentary structures suggest this sample is an early replacement nodule from a sublitoral microbialite, slightly deeper than the common peritidal samples [19]. The 2 younger samples (≈ 650 Ma) are deposited in a calsiltite unit above a flooding surface. These samples are peritidal early replacement nodules and were formed in the midst of a negative δ13C excursion. Note, these samples as well as the Ol, and Shuurgat, are part of the same Taagaan Olom Group. It and the Zuune Arts formation located unconformably above were weakly metamorphosed as part of the Central Asian Orogenic Belt (CAOB) [19].

9. Data are presented from the Rasthof Formation in Namibia. The Rasthof is a carbonate sequence that immediately postdates the Sturtian diamictite and includes the cap dolostone as well as 200-400 meters of dark gray rhythmite and microbialite [157]. There is no evidence for wave action, which indicates this sample precipitated in the sublittoral zone below wave base. The sample itself was taken from a microbialite so it maybe influenced by the climate immediately after snowball and more localized biological processes.

50 10. Data are presented from the Beck Springs Formation in Death Valley CA. The Beck Springs is a dolomite unit forming in a broadly transgressive system [185]. The chert samples themselves are peritidal early replacement nodules. Stratigraphically the beck springs immediately predates the Kingston Peak formation which is a diamictite that has been correlated with the Marinoan glaciation.

11. Data are presented from the Callison Lake Formation, a mixed siliciclastic and carbonate deposit located in the Yukon territory of Canada dated to about 745 Ma. The chert samples themselves are sourced from a dolostone unit characterized by abundant microbial lamination, domal stromatolitic bioherms, and cross-bedded oolitic grainstone, as well as the early diagenetic chert [188]. Callison Lake strata were formed in a peritidal to shallow subtidal depositional environment in a periodically restricted marginal marine basin.

12. Data are presented from the Akademikerbreen Group, a ≈ 750 Ma Neoproterozoic unit in Svalbard, Norway. The Akademikerbreen group is a 2000 meter succession that hosts a variety of carbonate facies. Petrographic and microfossil evidence suggests that the silica nodules replaced carbonate early in diagenesis, probably on a timescale no more than a few thousand years [28]. Like other early diagenetic chert, these samples probably formed in isolated evaporative coastal environments.

13. Data are presented from the ≈ 1200 Ma Aorferneq member of the

51 Narssarssuk Formation in northwest Greenland. The chert from this unit are early diagenetic in origin, replacing preexisting carbonate microbial mats. These samples likely formed in a broad and geologically relatively stable tidal flat environment varying from subtidal to supratidal189 [ ].

14. Data are presented from the Sukhaya Tunguska Formation in northwestern Siberia. This represents a 1000-1100 year old mixed limestone and dolostone with abundant nodular chert. Cherts show evidence of early cementation and environments are largely semi-restricted peritidal with some subtidal and lower intertidal zones. [174]

15. Data are presented from the Billyakh Group in northern Siberia, which samples the Kotiukan and the overlying Yusmastakh formation. The Kotiukan samples are from a chert-rich layer in a broader unit characterized by stromatolitic peritidal dolostones. The Yusmastakh Formation is laminated to medium bedded dolomicrite with abundant though not fossil-rich chert. The Billyakh cherts were likely formed in peritidal environments, possibly semi-restricted and evaporative [173].

16. Data are presented from the ≈ 1860 Ma Ironwood Iron Formationin Michigan. The Ironwood Formation is a Superior type iron formation deposited on a relatively stable cratonic margin. The specific samples are jaspery chert in a granular iron formation, indicating a relatively shallow depositional environment, likely influenced by some current or wave

52 motion [159].

17. Data are presented from the ≈ 1860 Ma Sokoman Iron Formation in Labrador, Canada. The Sokoman is composed of a variety of textures and sedimentary structures [110]. These were likely deposited in relatively shallow water in contact with the open ocean and have been only lightly metamorphosed. Sokoman cherts are granular and show evidence of early silicification, making them more likely to preserve primary isotopic signals [135]. Data are also presented from the Wishart Formation, which underlies the Sokoman Formation. The mineralogy and petrography of the sample is similar.

18. Data are presented from the ≈ 1860 Ma Biwabak Iron Formation in Minnesota. The outcrop has a stromatolitic texture which places a limit on the possible depth of deposition, primary chert has possibly been subsequently remineralized [154].

19. Data are presented from the ≈ 1860 Ma Gunflint Iron Formation in Ontario, Canada. The Gunflint is typical of early Proterozoic iron formations, one sample is clearly stromatolitic indicating a low energy shallow depositional environment, the other is more granular indicating a higher energy. The Gunflint Formation has experienced a low degree of metamorphic alteration and has abundant microfossils preserved [135].

20. Data are presented from the 1865 Ma Duck Creek Formation. The Duck Creek is primarily dolomite with minor associated iron formation. The

53 sample comes from chert pods that are themselves associated with the iron formation [214].

21. Data are presented from the 2060 Ma Franceville Group in Gabon. Not primarily an iron formation, instead this sample comes from a stromatolitic black chert.

22. Data are presented from the 2709 Ma Rietgat Formation in the Ventersdorp Supergroup in South Africa. The Ventersdorp Supergroup is primarily volcanic, the Rietgat is a silicified carbonate located stratigraphically between two flood basalt units. The sample is stromatolitic in origin and probably represents a peritidal environment [23].

23. Data are presented from the 3250 Ma Fig Tree Group in South Africa. The Fig Tree Gp. is a mixed shallow water and volcanic sequence. The samples are from relatively minor banded iron formations [89].

24. Data are presented from the 3000 Ma Cleaverville Formation from the George Creek Group in Australia. This sample is from a massive chert bed, possibly deposited in a relatively deep subtidal environment and silicified quickly. The unit is bounded by volcanic rocks, but has experienced only low grade metamorphism [105].

25. Data are presented from the 2684 Ma Jeerinah Formation from the Fortescue Group in Australia. The sample is from a carbonate unit in the

54 Jeerinah that immediately overlies a sandstone bed and precedes a black shale. The black chert formed as a replacement of the carbonate andthe formation has experienced minor low temperature metamorphism.

26. Data are presented from the 3390 Ma Strelley Pool Formation in Australia. The Strelley Pool formation covers a range of shallow water to terrestrial depositional environments and associated lithologies. This chert probably formed as an early diagenetic carbonate replacement in a shallow water evaporitic environment [190].

27. Data are presented from the 3462 Ma Kromberg Formation in the Onverwacht Group South Africa. The Kromberg formation has both bedded (sampled here) and vein associated chert[89].

28. Data are presented from the 2550 Ma Wittenoom Formation from the Hamersley Supergroup in Australia. The specific sample measured here is a chert nodule in basinal carbonates. The sample is relatively well preserved for an Archean age unit [184].

29. Data are presented from the Tower and Apex formations, both of which are in the 3000 Ma Warrawoona Supergroup. The apex chert is a minor phase bedded in a larger basalt unit, the Tower Formation is immediately below the Apex Formation stratigraphicaly [206].

55 2.7.2 XRD information

Samples were analyzed at the University of Oxford on a Panalytical Empyrean Series 2 powder X-ray diffractometer using a Co Ka source at 40 kV and 40mA. Samples were prepared as randomly oriented aggregates by mixing with ethanol and spread on zero background silicon single crystal substrates. Background subtraction, peak statistics, phase identification, and quantification were performed using HighScore Plus® software. Quantification was performed with the reference intensity ratio method using scale factors recently compiled in the International Center for Diffraction Data (ICDD) Powder Diffraction File 4+.

2.7.3 Supporting figures

56 Figure 2.7.1: Comparison of δ18O and Δ17O for Archean chert samples that were measured at both LSU and Harvard. A) Shows the raw data with associated error bars of ±0.019h from Harvard and ±0.023h. B) Shows the same data as a box and whisker plot to better compare the relative means. The points represent the means, the boxes represent 25-75 percentile, the whiskers represent the 5-95 percentile.

57 Figure 2.7.2: Oxygen isotope data of the stratigraphic test at the Fifteenmile formation. The red and gray data are from different parallel outcrops of the same formation. As can be seen, the oxygen isotope signal, especially in the minor isotope, is indistinguishable.

58 Figure 2.7.3: The model prediction with no diagenesis, no change in seawater chemistry, but changing surface temperatures. Together these temperature predictions combine in Δ17O- δ18O space to form concave down arrays.

Figure 2.7.4: Measured δ18O of the chert samples against the Monte Carlo predicted precipitation temperature. The statistics reflect the linear best fit for each dataset.

59 Figure 2.7.5: Comparison of the previously measured δ30Si data with the new oxygen isotope compositions [28]. Red points are cherts associated with Paleoproterozoic BIF, Blue points are Proterozoic peritidal chert. The only statistically significant relationship< (P 0.05) exists between δ18O and δ30Si specifically when the BIF samples are included.

60 2.7.4 Data tables

Table 2.7.1: This table shows the complete list of measured samples, their age, their location of origin, the measured values of δ17O and δ18O, and the calculated Δ’17O

Sample ID Age (Ma) Origin δ17O δ18O Δ’17O M602-607 a. 525 Khesen Formation, Mongolia 7.48 14.52 -0.196 M602-607 b. 525 Khesen Formation, Mongolia 11.18 21.56 -0.198 M620-1.1 a. 540 Khesen Formation, Mongolia 8.52 16.52 -0.206 M620-1.1 b. 540 Khesen Formation, Mongolia 8.07 15.63 -0.186 M622-169 544 Zunne Arts Formation, Mongolia 13.07 25.11 -0.168 M622-146.5 545 Zunne Arts Formation, Mongolia 12.23 23.36 -0.091 M622-85 a. 550 Zunne Arts Formation, Mongolia 13.43 25.84 -0.199 M622-85 b. 550 Zunne Arts Formation, Mongolia 13.43 25.90 -0.225 F726-165B a. 600 Shuurgat Formation, Mongolia 12.41 23.88 -0.186 F726-165B b. 600 Shuurgat Formation, Mongolia 12.28 23.65 -0.198 F723-58 a. 632 Ol Formation, Mongolia 8.40 16.13 -0.123 F723-58 b. 632 Ol Formation, Mongolia 14.41 27.79 -0.232 F704-330 a. 650 Taishir Formation, Mongolia 10.16 19.56 -0.173 F704-330 b. 650 Taishir Formation, Mongolia 10.68 20.60 -0.200 F704-331 a. 650 Taishir Formation, Mongolia 11.59 22.32 -0.188 F704-331 b. 650 Taishir Formation, Mongolia 10.88 20.97 -0.189 F702-100 660 Taishir Formation, Mongolia 12.45 24.02 -0.221 B1419-921.0 a. 555 San Juan Formation, Peru 3.35 6.60 -0.143 B1419-921.0 b. 555 San Juan Formation, Peru 6.02 11.80 -0.220 TP1 579 Rocky Harbour Formation, Newfoundland 5.29 10.26 -0.139 TP2 579 Rocky Harbour Formation, Newfoundland 4.72 9.08 -0.090 TP3 579 Rocky Harbour Formation, Newfoundland 4.81 9.30 -0.117 F8013A a. 660 Rasthof Formation, Namibia 12.01 23.07 -0.164 F8013A b. 660 Rasthof Formation, Namibia 10.69 20.64 -0.199 F901-33.8 739 Beck Spring Formation, California 10.89 20.89 -0.137 F901-20 740 Beck Spring Formation, California 11.39 21.91 -0.167 F930-5.5 740 Callison Lake Formation, Yukon 14.52 27.96 -0.213 F927-64.5 745 Callison Lake Formation, Yukon 17.16 33.03 -0.227 F929-37.5 a. 750 Callison Lake Formation, Yukon 12.53 24.17 -0.211 F929-37.5 b. 750 Callison Lake Formation, Yukon 13.14 25.21 -0.153 81B-550 750 Akademikerbreen Group, Norway 14.71 28.29 -0.192 86P-108-3 750 Akademikerbreen Group, Norway 15.57 30.00 -0.234 86P-96-2A 750 Akademikerbreen Group, Norway 15.64 30.14 -0.236 81P-4705-2B 750 Akademikerbreen Group, Norway 13.85 26.74 -0.244 81B-625-2A 750 Akademikerbreen Group, Norway 14.52 27.99 -0.226 T-714-2 770 Fifteenmile Formation, Alaska 12.91 24.86 -0.206 T-714-1 770 Fifteenmile Formation, Alaska 14.11 27.16 -0.208 T705-18 a. 805 Fifteenmile Formation, Canada 15.62 30.03 -0.199

61 Table 2.7.1 – continued Sample ID Age (Ma) Origin δ17O δ18O Δ’17O T705-18 b. 805 Fifteenmile Formation, Canada 16.48 31.74 -0.224 T705-18 c. 805 Fifteenmile Formation, Canada 15.19 29.15 -0.172 T705-154.2 812 Fifteenmile Formation, Canada 14.47 27.80 -0.181 T705-123.6 815 Fifteenmile Formation, Canada 15.01 28.80 -0.161 P1405-11.8 810.7 Fifteenmile Formation, Canada 15.29 29.36 -0.175 P1405-13.6 810.7 Fifteenmile Formation, Canada 16.21 31.18 -0.210 P1405-19.9 810.7 Fifteenmile Formation, Canada 15.99 30.72 -0.188 P1405-23.3 810.7 Fifteenmile Formation, Canada 15.86 30.50 -0.203 P1405-24.2 810.7 Fifteenmile Formation, Canada 16.30 31.43 -0.250 P1405-26.8 810.7 Fifteenmile Formation, Canada 15.31 29.48 -0.219 P1405-27.3 810.7 Fifteenmile Formation, Canada 15.00 28.84 -0.195 P1401-15.1 810.7 Fifteenmile Formation, Canada 14.80 28.46 -0.193 P1401-18.5 810.7 Fifteenmile Formation, Canada 14.30 27.53 -0.205 P1401-19.4 810.7 Fifteenmile Formation, Canada 13.96 26.82 -0.176 P1401-22.3 810.7 Fifteenmile Formation, Canada 13.87 26.68 -0.197 P1401-24.4 810.7 Fifteenmile Formation, Canada 14.95 28.76 -0.204 P1401-24.6 810.7 Fifteenmile Formation, Canada 11.39 21.97 -0.200 P1401-24.8 810.7 Fifteenmile Formation, Canada 15.79 30.44 -0.242 P1401-24.9 A 810.7 Fifteenmile Formation, Canada 14.70 28.24 -0.179 P1401-24.9 B 810.7 Fifteenmile Formation, Canada 14.91 28.64 -0.182 P1401-25.1 810.7 Fifteenmile Formation, Canada 15.07 28.97 -0.192 P1401-25.9 810.7 Fifteenmile Formation, Canada 13.86 26.64 -0.185 P1401-27.4 810.7 Fifteenmile Formation, Canada 15.21 29.32 -0.233 P1401-30.1 810.7 Fifteenmile Formation, Canada 14.47 27.83 -0.194 P1401-36.1 810.7 Fifteenmile Formation, Canada 15.77 30.36 -0.223 P1401-37.4 810.7 Fifteenmile Formation, Canada 14.54 27.96 -0.191 F833-167 a. 827 Fifteenmile Formation, Canada 13.84 26.56 -0.159 F833-167 b. 827 Fifteenmile Formation, Canada 12.29 23.69 -0.210 F833-149.3 828 Fifteenmile Formation, Canada 14.454 27.79 -0.199 F833-146 829 Fifteenmile Formation, Canada 14.30 27.46 -0.173 F833-133.8 830 Fifteenmile Formation, Canada 13.18 25.40 -0.210 K95-97 1050 Sukhaya Tunguska Formation, Siberia 14.44 27.75 -0.190 K95-94 1050 Sukhaya Tunguska Formation, Siberia 13.16 25.38 -0.223 K95-99 1050 Sukhaya Tunguska Formation, Siberia 13.47 26.02 -0.250 KS78-2 1200 Narssarssuk Formation, Greenland 15.36 29.65 -0.261 KS78-14 1200 Narssarssuk Formation, Greenland 16.44 31.73 -0.264 KS78-15 1200 Narssarssuk Formation, Greenland 16.40 31.56 -0.218 KS78-18 1200 Narssarssuk Formation, Greenland 16.11 31.03 -0.232 KG92-30 1475 Kotuikan Formation, Siberia 12.77 24.57 -0.191 KG92-34 1475 Kotuikan Formation, Siberia 14.28 27.55 -0.240 KG92-57 1475 Yusmastakh Formation, Siberia 13.31 25.73 -0.252 KG92-59 1475 Yusmastakh Formation, Siberia 14.14 27.27 -0.235 KG92-67 1475 Yusmastakh Formation, Siberia 15.52 29.84 -0.201 DYI-54 1850 Wishart Formation, Labrador 8.65 16.66 -0.157 DYI-148 1860 Sokoman Formation, Labrador 10.89 20.96 -0.168

62 Table 2.7.1 – continued Sample ID Age (Ma) Origin δ17O δ18O Δ’17O G99-4 1860 Sokoman Formation, Labrador 9.68 18.64 -0.158 J99-1 1860 Ironwood Formation, Michigan 6.92 13.32 -0.130 J102-3 1860 Ironwood Formation, Michigan 2.03 4.04 -0.115 K71-4 1860 Biwabik Formation, Minnesota 12.73 24.62 -0.257 GF-70-38 1860 Gunflint Formation, Ontario 10.32 19.91 -0.188 GF-59-11 1860 Gunflint Formation, Ontario 11.93 22.98 -0.188 K-9 1865 Duck Creek Formation, Australia 25.35 13.16 -0.212 FV-26 2060 Franceville Group, Gabon 11.08 21.32 -0.172 92821 2550 Wittenoom Formation, Australia 11.22 21.52 -0.144 BK-75-4 2684 Jeerinah Formation, Australia 11.10 21.12 -0.179 IV-16 2709 Rietgat Formation, South Africa 7.43 14.56 -0.142 BK-75-1 3000 Cleaverville Formation, Australia 7.20 13.42 -0.155 BSA-78-7 a. 3000 Apex Chert, Australia 7.92 15.21 -0.138 BSA-78-7T a. 3000 Apex Chert, Australia 7.72 15.03 -0.137 BSA-78-7T b. 3000 Apex Chert, Australia 8.15 15.54 -0.169 BK-75-S5 a. 3250 Fig Tree Group, South Africa 4.78 9.33 -0.130 BK-75-S5 b. 3250 Fig Tree Group, South Africa 8.40 16.07 -0.141 EB-71-1 a. 3250 Fig Tree Group, South Africa 7.72 14.79 -0.144 EB-71-1 b. 3250 Fig Tree Group, South Africa 1.16 2.40 -0.114 K-07-SP1 3390 Strelley Pool Formation, Australia 8.46 16.07 -0.169 BRS-78-27a 3462 Kromberg Formation, South Africa 9.91 18.76 -0.152 BSA-78-1A 3460 Towers Formation, Australia 9.28 17.17 -0.135

63 Table 2.7.2: Triple oxygen isotope data for Archean cherts as measured at LSU and Harvard. All data is reported in h units relative to VSMOW and Δ17O values are normalized relative to λ=0.5305. Both datasets were normalized relative to the same literature values of San Carlos Olivine, NBS-28, and UWG-2. The mean difference between Δ17O as measured at Harvard and LSU is 0.0012h

LSU Harvard Sample ID Age (Ma) δ17O δ18O Δ’17O δ17O δ18O Δ’17O 92821 2550 11.22 21.52 -0.144 BK-75-4 2684 10.84 20.87 -0.175 11.10 21.37 -0.183 IV-16 2709 7.81 15.06 -0.156 7.43 14.30 -0.135 BK-75-1 3000 6.68 12.89 -0.135 7.20 13.95 -0.175 BSA-78-7 a. 3000 7.89 15.17 -0.135 7.92 15.25 -0.141 BSA-78-7T a. 3000 7.89 15.22 -0.149 7.72 14.84 -0.124 BSA-78-7T b. 3000 7.94 15.31 -0.150 8.15 15.77 -0.188 EB-71-1 a. 3250 7.74 14.92 -0.144 EB-71-1 b. 3250 1.16 2.40 -0.114 BK-75-S5 a. 3250 4.84 9.42 -0.144 4.78 9.25 -0.116 BK-75-S5 b. 3250 8.25 15.90 -0.151 8.40 16.15 -0.136 K-07-SP1 3390 8.19 15.83 -0.174 8.46 16.31 -0.164 BSA-78-1A 3460 8.60 16.50 -0.121 9.28 17.84 -0.146 BRS-78-27a 3462 9.61 18.51 -0.170 9.91 19.01 -0.134

64 Table 2.7.3: Complete mineralogy of chert samples

Raw mineralogy After sample processing Quartz Quartz Sulfates Phosphates Sulfates Other Silicates Phosphates Oxides Non Oxygen Bearing Oxides Non Oxygen Bearing Other Silicates Carbonate Sample ID Carbonate M602-607 a. 98 2 98 2 M602-607 b. 97 1 100 M620-1.1 a. 32 68 78 21 M620-1.1 b. 36 65 44 55 2 M622-169 96 4 98 2 M622-146.5 84 16 100 M622-85 a. 100 100 M622-85 b. 100 100 F726-165B a. 85 15 1 96 4 F726-165B b. 82 3 14 1 100 F723-58 a. 3 97 24 76 F723-58 b. 98 2 100 F704-330 a. 100 100 F704-330 b. 100 100 F704-331 a. 39 62 100 F704-331 b. 8 92 75 25 F702-100 87 13 99 1 B1419-921.0 a. 57 44 52 48 B1419-921.0 b. 21 78 24 44 20 12 TP1 28 72 30 70 TP2 22 68 10 29 1 64 6 TP3 27 1 72 1 26 1 74 F8013A a. 80 4 16 74 26 F8013A b. 1 99 14 86 F901-33.8 97 3 100 F901-20 99 1 100 F930_5.5 99 1 100 F927-64.5 100 100 F929_37.5 a. 100 100 F929_37.5 b. 100 100 81B-550 99 1 100 86P-108-3 81 19 93 7 86P-96-2A 96 4 100 81P-4705-2B 95 3 2 100 81B-625-2A 93 7 100 T-714-1 65 36 36 64 T-714-2 26 38 35 51 49 T705-18 a. 98 2 98 2

65 Table 2.7.3 – continued Raw mineralogy After sample processing Sulfates Carbonate Phosphates Sulfates Non Oxygen Bearing Quartz Oxides Other Silicates Quartz Other Silicates Phosphates Oxides Non Oxygen Bearing Sample ID Carbonate T705-18 b. 100 100 T705-18 c. 62 37 100 T705-154.2 96 5 100 T705-123.6 98 2 100 P1405-11.8 95 4 100 P1405-13.6 93 7 100 P1405-19.9 97 3 97 3 P1405-23.3 95 5 100 P1405-24.2 96 4 100 P1405-26.8 83 8 1 9 92 8 P1405-27.3 85 12 3 100 P1401-15.1 72 28 100 P1401-18.5 79 21 78 18 4 P1401-19.4 23 77 100 P1401-22.3 88 10 2 83 15 2 P1401-24.4 90 10 100 P1401-24.6 30 70 73 27 P1401-24.8 93 4 1 2 100 P1401-24.9 A 94 1 4 100 P1401-24.9 B 100 100 P1401-25.1 97 2 1 100 P1401-25.9 82 18 100 P1401-27.4 77 13 10 100 P1401-30.1 92 5 4 100 P1401-36.1 98 2 100 P1401-37.4 77 23 100 F833-167 a. 79 21 100 F833-167 b. 14 86 78 3 19 F833-149.3 98 2 100 F833-146 96 4 100 F833-133.8 59 41 96 4 K95-97 86 12 2 100 K95-94 86 12 2 100 K95-99 86 14 100 1 KS78-2 100 100 KS78-14 56 44 99 KS78-15 92 8 100 KS78-18 79 11 8 1 100 KG92-30 47 9 44 1 96 3

66 Table 2.7.3 – continued Raw mineralogy After sample processing Quartz Other Silicates Oxides Non Oxygen Bearing Quartz Other Silicates Oxides Sulfates Phosphates Sulfates Phosphates Non Oxygen Bearing Carbonate Sample ID Carbonate KG92-34 97 3 86 13 2 KG92-57 99 1 100 KG92-59 83 17 100 KG92-67 100 99 1 DYI-54 64 30 6 91 9 DYI-148 81 5 10 3 87 10 2 G99-4 88 11 93 6 J99-1 76 1 3 21 91 9 J102-3 13 87 88 1 11 K71-4 91 2 6 92 8 GF-70-28 96 2 2 100 GF-59-11 94 4 2 100 K-9 95 4 1 98 2 FV-26 97 1 1 100 92821 97 3 100 BK-75-4 99 1 100 IV-16 95 2 3 100 BK-75-1 100 100 BSA-78-7 a. 97 2 98 2 BSA-78-7T a. 100 100 BSA-78-7T b. 100 100 BK-75-S5 a. 23 17 12 48 65 35 BK-75-S5 b. 93 4 3 96 4 EB-71-1 a. 98 1 1 99 1 K-07-SP1 100 100 BRS-78-27a 100 100 BSA-78-1A 100 100

67 Table 2.7.4: Variables included in the Monte Carlo model of chert formation

Variable Allowed Range Description ocean change -0.5 to 1.5 The degree to which the ocean composition is fraction allowed to evolve, held constant when at 0, follows described ocean change prediction when at 1, and scales proportionally for other values. glacial fraction 0 to 0.2 Fraction of the ocean volume stored as ice, also approximates the degree of evaporation in an enclosed basin. precipitation 0 to 100 Equilibrium temperature at which the chert temperature originally precipitates. Constrained to temperatures at which the ocean is liquid. water to rock 0.01 to 10 Defines the degree to which samples are altered, ratio value can theoretically range from zero to infinity, but the three orders of magnitude here cover reasonable variability. groundwater -25 to 0 δ18O Isotopic composition of groundwater with δ18O which alteration occurs. Groundwater Δ17O is constrained by this value according to the relationship described in Luz et al. 2010 [124]. diagenesis 100 to 1000 Equilibrium temperature at which the chert is temperature diagenetically altered. Lower bound is set at 100 to ensure alteration always occurs at a temperature above original precipitation.

68 3 Calibrating the triple oxygen isotope vital effect in cultured diatoms

In preparation for Frontiers in Earth Science

69 3.1 Abstract

The 18O/16O of precipitated silica in natural environments is posited to record the temperature of that ambient environment. Measurements of 17O/16O in silica

18 16 (SiO2), when paired with O/ O, provide an opportunity to further constrain the environmental conditions at the time of mineral precipitation. On Cenozoic timescales, siliceous frustule precipitation by diatoms has served as the primary sink of silica from the ocean. Frustule precipitation is a biologically catalyzed

process from an ocean undersaturated in SiO2. This precipitation is therefore out of chemical and isotopic equilibrium, and therefore not represented by equilibrium fractionation models. Here we measure the 17O/16O and 18O/16O in diatom frustules from semi-continuous cultures with two different diatom species and across a range of growth rates and temperature. To further enable these measurements, we developed a laser fluorination method to reproducibly measure the triple oxygen isotope composition of silica frustules. While the δ13C of the organic matter scaled predictably with growth rate, neither the δ18O or ′ Δ 17O of the frustule silica varied significantly with respect to variations in growth rate or species. Temperature, however, does carry a control on the ensuing isotopic composition. That is, we observe an inverse relationship between δ18O ′ and temperature (and the opposite for Δ 17O), as predicted by equilibrium ′ isotope theory. However, the measured Δ 17O compositions were systematically negative with respect to the corresponding equilibrium isotope prediction, a signal we interpret as a diatom vital effect. This vital effect may be caused bythe

70 silica concentrating mechanisms diatoms use to precipitate their frustules. These results confirm the utility of diatom δ18O in reconstructing temperature, but ′ suggest that Δ 17O may not be a suitable temperature proxy for diatomaceous ′ sediment. Instead, Δ 17O may be used to distinguish biologically from abiologically precipitated silica.

3.2 Introduction

The oxygen isotopic composition of diatom frustules has been used as a proxy for sea surface temperature and ice volume on glacial-interglacial timescales

[177, 192]. The utility of diatom frustules (made of 2SiO ) is elevated, for example, in instances where geological records lack viable carbonate samples [1, 133]. A reliable polar ocean temperature record - the exact regions where the abundance of diatoms thrive today - may be particularly useful given the understood climate sensitivity in the high latitudes [90, 142]. In places where the diatom silica and foram carbonate record exist in parallel, the independent temperature relationships can be used to further disentangle the effect of temperature and ice volume on the isotopic record [177]. These environmental and geological records, however, are only as useful as their calibration is precise. To date, the lions share of effort has targeted the carbonate system104 [ ]. In contrast, the oxygen isotope composition of frustules is less well calibrated and can only be interpreted given a firm quantitative understanding of the link between temperature and the diatom frustules.

In forming frustules, diatoms take up oxygen isotopes into SiO2 in a

71 temperature-dependent fashion offset from the oxygen isotope composition of the surrounding water. The equilibrium δ18O (defined below) relationship between quartz and water has been calculated from thermodynamic first principles [78]. Diatoms, however, precipitate opaline hydrated amorphous silica · (or biogenic opal, SiO2 nH2O) to form their frustules [118]. Without a predictable crystal structure, it is difficult to calculate an equilibrium fractionation ab initio. In place of this, temperature relationships are calibrated by measuring both environmental SiO2 and laboratory diatom cultures [21, 41, 51, 147]. Although there is general agreement, with a slope of about 0.2h/oC, there are also open questions about the relative role of spceies specificity, the role of hydration, all against a backdrop of analytical challenges in reproducibly making these measurements [21, 51, 107, 176]. Biological processes, such as shell formation, can induce non-equilibrium isotope effects known generally as vital effects. Vital effects famously play a role in carbonate δ18O temperature reconstructions, requiring the use of species-specific calibrations [140, 187, 209, 222]. Though less studied, most previous natural environment studies show no significant δ18O offset between different diatom species growing in the same conditions [7, 21, 147, 148, 178]. A single laboratory experiment supported this general conclusion [21]. In contrast, Swann et al. (2007) measured a significant isotope difference in two diatom size fractions from sediment in the north Pacific and attributed this difference to a vital effect between the dominant diatom species in each fraction [193]. Due to the difficulty in isolating frustules from a single species in natural samples, most

72 approaches target measuring different size fractions, an acceptable though imprecise way to distinguish between species [7, 147, 178, 193]. Through the compilation of this work it has become clear that additional investigation may be warranted. In other elemental and isotopic systems, physiological features intrinsic to the organism play an important role in controlling observed isotope fractionation in the biological product (a shell, organic matter, etc.). For example, the δ13C in organic matter is classically interpreted as being controlled during carbon fixation by the rate-dependent contribution of a reversible diffusion and irreversible fixation step60 [ ]. At high growth rates a larger fraction of the DIC available in a cell becomes fixed, thereby decreasing the expression of the enzymatic RuBisCO fractionation and consequently decreasing the overall fractionation between DIC and organic carbon (εp). This growth rate relationship has been quantified in2 species of diatoms [114] along with a variety of other autotrophic organisms [15, 113, 156, 212]. Unlike carbon, however, oxygen isotopes involved in silica precipitation do not experience the same metabolically driven redox transformation. Nonetheless, the incorporation of silica into diatom frustules can still be broadly described as a two step diffusion-fixation process, and therefore δ18O may have some growth rate dependence. In the comparable carbonate δ18O system, kinetic disequilibrium scales with growth rate, though the resulting dependence is smaller than the temperature effect140 [ , 222]. The possible influence of growth rate on δ18O of diatom silica remains an open question. Most recently and in parallel to work on diatoms, increasing focus has been

73 placed on measurements of the third stable isotope of oxygen, 17O. Here, with analytical advances it can be argued that there is additional information in the 17O/16O when paired with the 18O/16O (cf. [119]). For example, the 17O/16O in barite (BaSO4) minerals is used as a valuable proxy for atmospheric O2 and CO2 [9, 43]. Among silicate minerals, sedimentary chert is consistently offset from the bulk terrestrial 17O composition and therefore has been interrogated as a potential paleoenvironmental temperature proxy [119, 152]. In fact, the 18O/16O of chert stands as a classic archive for arguments pertaining the Earth’s earliest climate, and evolution since [107]. Bedded, or early diagenetic chert nodules from Precambrian rocks are, however, inorganic precipitates and can avoid complications of shell or frustule biological effects. In complement, the 17O temperature effect for the inorganic system can be, and was, calculated78 [ , 176]. No study yet has assessed in detail the effect of biological precipitation on the ′ Δ 17O of diatom frustules. We performed a set of diatom culture experiments in which we grew two diatom species, at varied growth rates, in different temperature solutions. We further developed a method to reproducibly measure the 16,17,18O composition of diatom silica. The use of two species also allows for a direct test ofa species-specific effect, as well as the control of growth rate on the ensuing2 SiO . From this, we compare our physiologically rooted temperature relationship to previous 18O temperature calibrations and then extend that treatment into 17O. This calibration stands as the first step toward unlocking the 17O of diatom frustules, and turning this measure into a viable proxy for addressing the last

74 100+ million years of earth’s climate history.

3.3 Methods

3.3.1 Culture

We cultured two different marine diatoms, one centricThalassiosira ( pseudonana CCMP1335) and one pennate (Cylindrotheca fusiformis NEPCC417). The diatom cultures were grown in the Giordano lab at Marche Polytechnic University (Hania in prep.). The growth medium, described in [59], was Artificial Multipurpose COmplement for the Nutrition of Algae (AMCONA) buffered at a pH of 8.2. All cultures were performed in 250 mL flasks with200 − − mL of medium and constant illumination of 100 µmol photons*m 2*s 1. The culture temperature was controlled at 15o, 20o, and 30oC, reflecting a range of natural conditions. For each diatom species and at each temperature, cultures

were performed at maximum growth rate (μmax) and half of the maximum

(0.5μmax). Growth rate was controlled by delivery rate of the growth medium. The cultures were allowed to acclimatize to the culture conditions for at least 4 generations before sample collection. Together, the 2 diatom species, 3 different temperatures, and 2 growth rates results in 12 combinations, with each set of conditions cultured 4 times in replicate. With the conservative assumption that 10% of the total biomass (organic matter plus frustule) would be silica, we grew cultures large enough to generate at least 20 mg of material. Depending on the treatment, acquiring this amount of biomass took one to four months of regular

75 extraction from the semi-continuous culture.

3.3.2 Chemical treatment

The isotope methods used herein and described below access all the oxygen ina given sample. This means that much care and attention must be given to preparation methods. For instance, it is imperative we remove any organic matter from the silica frustules prior to measurement. We adapted published cleaning/prep methods [51, 176], making allowances for the high organic matter concentration that comes with culture experiments (and in contrast to natural

o sediment [51, 176]). For this work, samples were immersed in 30% H2O2 at 90

for at least 24 hours, or until any visible reaction (inferred CO2 evolution) ceased. The samples were then rinsed with milli-Q, reacted% with5 HCl for at least 12

hours at room temperature, rinsed again, then reacted with 15M HNO3 for at least 12 hours at 90o. Following a final rinse, samples were allowed to dry and stored at room temperature. Our samples are pure cultures, so it was not necessary to physically separate the diatom frustules from any background sediment. Each sample was weighed before and after cleaning to determine the approximate silica mass fraction.

76 3.3.3 Isotope notation

We first report all isotope compositions using the standard delta notation. This eases comparison with existing oxygen isotope literature, and results in: ( ( ) ) 18O 16 18 = ( O) sample − ∗ . δ O 18O 1 1000 (3.1) 16O VSMOW Other isotope ratios (δ17O and δ13C) can be calculated in the same manner. The δ17O scales closely with the relative mass difference and carries a characteristic ′ relationship with δ18O. We thus report the δ17O variance in Δ 17O notation, in which the 17O composition is normalized relative to a prescribed δ17O- δ18O relationship of 0.5303 (the θRFL in equation 3.2). This is the established high-temperature equilibrium [216]. This results in a logarithmic definition [143]:

( ( ) ( )) 17 18 ′ δ O δ O Δ 17O = 1000 ∗ ln + 1 − θ ∗ ln + 1 . (3.2) 1000 RFL 1000

Isotope data can be reported relative to an established standard, like VSMOW, or relative to another composition. In this work, offsets between measured frustules and experimental waters anchor the system. When both are reported on a VSMOW scale, the 18O/16O offset can be reported as:

( ) 18 + 18 δ OSiO2 1000 − × h ε = 18 1 1000 , (3.3) δ OH2O + 1000

77 and

17 ′ ′17 − ′17 E = Δ OSiO2 Δ OH2O. (3.4)

This normalization eases comparison to temperature predictions and equilibrium isotope effects.

3.3.4 Isotope methods

We measured δ13C of each diatom sample on a ThermoFisher Scientific flash IRMS connected to a delta V plus Isotope Ratio Mass Spectrometer in continuous flow mode (at Harvard University). The resulting isotopic compositions were offset corrected and peak size corrected using the L-glutamic acid standards USGS40 and USGS41a. Repeated measurements of these ′ standards resulted in a standard deviation of 0.09h for δ13C. The Δ 17O and δ18O measurements of the DI and Milli-Q water used to create the growth medium were made in the University of Washington stable isotope analysis Isolab using a Picarro L2140i water isotope analyzer. This method produced a δ18O precision of ′ 0.046h and a Δ 17O precision of 0.0095h. For the oxygen isotope measurements, 2 mg of sample were loaded into one of twenty one wells on a stainless steel plate (in a stainless vessel), enclosed in a

chamber with a BaF2 window, and evacuated to < 5 mTorr. When sufficient frustule mass was available, numerous measurements were made. Recall that each experimental condition yielded four replicate samples. For replicates with less than 2 mg of silica mass, the maximum available sample mass was used. The

78 sample tray plus the samples were then prefluorinated with2 F for > 5 hours at 40

Torr. The prefluorination timescale was calibrated based on the rateofO2 gas evolution and is discussed further below. After prefluorination, each sample was individually lased with a 50 W CO2-laser under a pure F2 atmosphere. The product oxygen was separated from excess F2 and other fluorinated products first by a series of cryogenic traps, and then by purification on an in-line gas chromatograph. The oxygen was directly introduced to a Thermo Scientific MAT 253 gas source isotope ratio mass spectrometer where it was measured against a calibrated lab O2 standard. For full method details and calibration, see [38]. A critical component of this study was the development of a refined method for measuring the oxygen isotope composition of fresh diatom frustules. Note that diatom frustules have a heterogeneous composition, with oxygen contributions from structural oxygen (oxygen atoms bound to silicon), hydroxyl − oxygen (OH bound to a single silicon atom), and adsorbed water [118]. Only the structural oxygen fraction reliably reflects the conditions under which the silica precipitated, and so this is our target reservoir. The hydroxyl oxygen and adsorbed water fractions are susceptible to exchange with water in the laboratory (or diagenetically in the field). For this work, this is particularly true given the series of pre-treatments and exposure to elevated temperatures. Fortunately,

exposure to F2 at room temperature (called prefluorination) effectively removes non-structural oxygen. Thus, we developed and applied a prefluorination method for the analysis of diatom frustules.

Prefluorination yields2 O , in the same fashion as analyzing a sample. That

79 product O2 was captured and measured in order to create a time-series of the evolved oxygen isotope composition. Prefluorination is complete when the product oxygen maintains a stable isotope composition between adjacent prefluorinaiton steps. The sample trays consisted of 12-14 frustule samples along with 2-3 NBS-28 quartz standards. The NBS-28 quartz standard was unreactive during prefluorination, so the measured2 O reflected an average composition of

the silica in the tray at each time-step. Figure 3.3.1 shows the volume of O2 evolved, measured from the peak integrated area (P.I.A.) of the in-line GC, as a

function of the samples exposure to F2. The C. fusiformis samples were

prefluorinated for less total time to increase2 O yields, as described in the

supplementary information. The rate at which our samples react to F2 decreases rapidly after the first minutes of exposure, which confirms the efficacy of

prefluronation as a means of removing2 H O and loosely bound hydroxyl-oxygen from the sample. The average laser fluorinated yield of T. pseudonana and C. ± ± fusiformis was 57 7% and 46 8% based on O2 yields and a perfect SiO2 stoichiometry. This is slightly lower than the 74% yield reported in the literature

[118], but our use of more reactive F2 and increased prefluorination time likely accounts for the difference.

Importantly the rate of O2 evolution never reaches zero even at the defined end of prefluorination. This indicates that exposure of samples toF2 will slowly react even the structural oxygen. Because the entire tray is exposed to F2 while a single sample is being lased, this suggests that, especially for the first samples measured, there will be a background contribution of O2 from the remaining

80 samples in the tray. On average the laser heating was completed in 8 minutes,

which minimizes the contribution of interfering O2 to any given sample. Based

on the observed rate of O2 evolution, less than 5% of any measurement consists

of O2 from other samples. We tested this by lasing an identical quartz standard at the beginning and end of every sample tray and we observed no systematic isotopic compositional difference.

The observed O2 evolves in successive prefluroination steps, progressively

18 increasing in δ O (Figure 3.3.2). The initial O2 aliquot in each prefluorination had a composition around 15h, whereas the average sample composition is above 25h. This trend has been observed in previous studies that employed stepwise fluorination30 [ , 51], though not to the same degree as herein. The water used in the diatom culture experiments has a measured δ18O composition of -8.8h. The lab water used during the chemical pre-treatment hasa δ18O composition of -6.8h. The light oxygen in early prefluorination steps therefore shows the contribution of adsorbed water and exchangeable hydroxyl oxygen, which may have re-equilibrated at high temperature with lab water during chemical pre-treatment [120, 203]. This reactive oxygen is consumed during prefluorination; both diatom species ultimately reach a steady state in δ18O before the samples were laser fluorinated. We do note that the C. fusiformis steady state is slightly offset from the sample composition. ′ This is the first study to systematically report the Δ 17O of diatom oxygen during stepwise fluorination. That noted, we would expect a similar trend to that

81 Figure 3.3.1: Much attention was paid to calibrating the prefluorination of frustules. That is, the ambient generation of O2 prior to laser analysis. Cap- tured in the top frame, we can track this O2 generation via the combination of a highly reactive fraction, and a background, less reactive fraction. P.I.A. is the peak integrated area. The cumulative F2 Pressure * Time is a measure of the total F2 exposure to the tray of samples and is calculated as the product of the F2 pressure in the chamber (Torr) and the time of the preflurination step (hours). Each species was measured over two trays shown as the red and blue data.

82 18 ′17 Figure 3.3.2: The δ O (a) and Δ O (b) of the product O2 evolved from the frustules during prefluorination as a function2 ofpF and fluorination time. The circles shows the composition of the prefluorination2 O fraction, the squares show the average composition of the samples in the tray (i.e. the value we expect the sequential prefluorination to approach). The pressure in the chamber varied slightly over time, but not vary by more than 10 % and would not influence the resulting data.

83 ′ seen for δ18O. The distilled water used as the culture medium hasa Δ 17O ′ composition of 0.050h and the Harvard lab water has a Δ 17O composition of ′ 0.04h. A decreasing contribution of this reactive Δ 17O heavy water should result in gradually decreasing isotopic compositions in successive prefluorination steps before ultimately reaching a steady state. As shown in Figure 3.3.2, this is broadly the case but there is additional unexplained structure. Figure 3.3.2 shows

′17 the Δ O of the prefluorination O2 as a function of F2 exposure. In three out of ′ four prefluornation trajectories the Δ 17O increases dramatically around 80 Torr*hour. This spike is not paralleled in the δ18O data, after 2 hours at 40 Torr the δ18O has already approached the final equilibrium composition. Large ′ mass-dependent changes in Δ 17O usually accompany large fractionations in δ18O. ′ This Δ 17O peak, however, suggests that diatom frustules include some mass

18 fraction with moderate reactivity to F2 that matches the δ O of the structurally bound oxygen but has a different characteristic θ. Laboratory distilled water does ′ have a sufficiently heavy Δ 17O composition (= 0.05 h) to generate this peak, but such a significant contribution would be accompanied by correspondingly light δ18O value (lab water δ18O = -6 h). Our goal of defining a frustule temperature effect is clearly predicated on the degree to which we successfully isolated and measured only the structural oxygen fraction. Further we must confirm that the structural oxygen isotope composition is not being reset during chemical treatment prior to fluorination, as was observed in previous experimental work [203]. In that study, Tyler et al. (2017) heated two diatom samples in water of distinct isotopic composition and

84 estimated that, even after stepwise fluorination, samples heated at 80o are 40% re-equilibrated after 1 week203 [ ]. The high organic content of our samples necessitated comparable heating, so oxygen isotope resetting is indeed possible. The chemical pre-treatment on which we based our method was previously tested using δ18O spiked water and no significant re-equilibration was observed between the water and structural oxygen [51]. That study however measured environmental samples; our cultured samples have not been subjected to any degree of diagenesis and therefore could be more susceptible to alteration. At 90oC we expect a δ18O fractionation of 23h relative to our measured lab water with a composition of -6.8h. This resulting contribution would tend to drive measured frustule δ18O towards lighter compositions - we see no evidence of this. We can therefore assume, due to lack of evidence to the contrary, that measured oxygen isotope compositions reflect diatom growth conditions. Finally, replicate fluorination of diatomaceous material anchors our quoted ′ precision on both δ18O and Δ 17O. We feel this is a conservative approach given the matrix effects and sensitivity of fluorinating this material. Here then, we ′ report 0.05h and 0.015h precision for δ18O and Δ 17O, respectively. This is based on replication of 27% of the sample set, and is in keeping with error on other more crystalline silicate materials.

3.4 Results and discussion

The semi-continuous experiments presented here were monitored. Growth rates, as well as the carbon isotopic composition of the biomass are shown in Table

85 − 3.4.1. Growth rates are known to within a 1σ of 0.05 day 1, which reflects the measured reproducibility between replicates. A similarly conservative approach yields a 1σ on δ13C of 0.9h, though replication of standards is much more precise. Using these data, we calculate the photosynthetic carbon isotope fractionation (εp) within our diatom cultures. However, the local, experimental DIC δ13C was not directly measured. Given the known culture conditions, we calculated the fractionation factor between atmospheric CO2 and dissolved CO2

13 [221] using the solubility of dissolved CO2 [210], and an atmospheric CO2 δ C h composition of -8 for central Italy. With these supplemental constraints, the εp

was then calculated [65]. This method of calculating εp may be slightly imprecise - recall that these are semi-continuous culture experiments open to atmosphere.

If in the growing diatom culture, CO2 consumption outpaces resupply, disequilibrium could be induced. Here, diatoms would consume isotopically

13 light CO2, possibly lowering [CO2] and increasing δ CDIC relative to equilibrium. The resulting values are therefore best suited for establishing trends and comparing relative values. The calculated εp are shown in Figure 3.4.1 compared to literature values.

For marine phytoplankton, εp scales to the surrounding CO2, where high

[CO2] corresponds with larger εp [48]. It has been established more specifically

in chemostat experiments that εp correlates positively with growth rate and

inversely with [CO2], a relationship that holds for a variety of eukaryotic species when corrected for different cell sizes113 [ , 156, 212]. The linear relationship

86 isotope data.png isotope data.bb

Figure 3.4.1: Bulk organic carbon δ13C isotope data of cultured diatoms from this study compared to literature data [156, 162]. The Popp et al. (1998) data (**) are from a chemostat experiment, whereas the Riebesell et al. (2008) data (*) are from a batch culture. Data from this study are from semi- continuous cultures.

87 Table 3.4.1: Isotopic composition of the organic matter fraction from the diatom cultures. Each value is the averaged composition of the four replicate cultures grown under each unique condition

Culture Growth δ13C − Description Rate (d 1) (h) T.p. 15oC max 0.16 -19.8 T.p. 15oC 50 0.08 -22.3 T.p. 20oC max 0.53 -21.5 T.p. 20oC 50 0.27 -20.5 T.p. 30oC max 0.56 -16.2 T.p. 30oC 50 0.28 -19.9 C.f. 15oC max 0.18 -22.0 C.f. 15oC 50 0.09 -23.0 C.f. 20oC max 0.30 -19.6 C.f. 20oC 50 0.15 -19.4 C.f. 30oC max 0.22 -20.3 C.f. 30oC 50 0.11 -21.4

between μ/[CO2] and εp (Figure 3.4.1) was attributed to the relative expression of the RuBisCO carbon fixation end-member (y axis intercept) and2 aCO transport controlled end-member. While this relationship is robust among chemostat culture studies, the results from batch cultures tend to produce smaller

εp fractionations with a weaker dependence on μ/[CO2] [24, 25, 162]. There is no single agreed upon explanation for this difference between batch and chemostat culture data, though it may be due to additional rate limiting steps during carbon fixation (Wilkes et al. in review). Specifically, in light replete and nutrient limited chemostat cultures, a carbon concentrating mechanism upstream of carbon fixation imposes a kinetic barrier to carbon acquisition. This then causesa significant isotope fractionation. In light-limited batch cultures, the RuBisCO fractionation is more strongly expressed resulting in smaller εp.

88 The bulk organic δ13C data from our semi-continuous diatom cultures is broadly compatible with existing batch culture data data (Figure 3.4.1). Type ID RuBisCO has a characteristic fractionation of 18.5h, slightly larger than the y

intercept of our εp trend [20]. This difference may be due to the contribution ofa − carbon concentrating mechanism delivering HCO3 to the RuBisCO active site, or due to uncertainties in our estimate of DIC δ13C. While our data do not span a

broad range of μ/[CO2], the εp data for both diatom species does exhibit a negative correlation with growth rate. This indicates that the diatoms are responding predictably to the environmental conditions in culture and suggests that the oxygen isotope results can also be treated as typical for diatom cultures.

3.4.1 The oxygen isotope composition of diatom frustules

The triple oxygen isotope composition of diatom frustules from our experiments is listed in Table 3.4.2. Each culture description, reflecting a single combination of species, temperature, and growth rate, averages the values of the four replicate cultures grown under those conditions. The standard deviation shows the variability among the replicate cultures and is noted above. The temperature-dependence of the frustule δ18O fractionation between water and silica is well established. Indeed, experimental data generated in this study parallels this relationship, as captured in Figure 3.4.2, and shows the inverse relationship between temperature and isotope fractionation (here noted as 18ε

89 and 17E’). The best fit for our data follows the equation:

· ± · 6 −2 ± 1000 lnα(diatom−water) = 1.84( 0.6) 10 T + 15.08( 6.92). (3.5)

This is a shallower slope than that reported in most other studies, parallel to,but offset from Crespin et al. 2010 which generated a temperature relationship using natural lacustrine diatoms [41]. In the two other cases where culture data was included, only a weak temperature dependence was observed [21, 169]. Further, our data captures larger than typical fractionations over this temperature range, though still within the bounds of existing studies and comparable to the scant Δ17O literature [176]. Some of the reported differences are likely a result of methodological choices in measuring δ18O as well as differences in sample origin/generation. In moving forward, direct comparison to work where the 17O is also included and generated via similar fluorination methods176 [ ] is most appropriate. We do not observe any significant difference in δ18O between equivalent cultures grown at different growth rates (Figure 3.4.3). The 0.5μmax growth rate 18 h δ O values are on average 0.4 heavier than the corresponding μmax cultures, but the standard deviation of this difference is 0.7h, far larger than any signal. As described in the supplementary materials but clear from Figure 3.4.3, there is no significant difference in the δ18O-temperature trend between the two diatom

90 Figure 3.4.2: The predicted temperature dependence of the δ18O effect between diatom silica and water (noted as 18ε). Measured data from this study is represented by the box and whisker plots where the red line shows the median, the box shows the 25th and 75th percentiles, and the whiskers extend to the maximum and minimums. Data from both diatom species and growth rates is combined. The black like captures the best fit regression, and is compared to temperature relationships from (1-7) [21, 41, 51, 107, 116, 147, 176].

91 Figure 3.4.3: The predicted temperature dependence of the δ18O effect between diatom silica and water (noted as 18ε), here denoting the lack of a dis- tinguishable rate relationship. The vital effects are discussed more thoroughly in Figure 3.4.4.

92 Table 3.4.2: Measured oxygen isotopic composition of the silica frustules from the diatom cultures. Each value is the averaged composition of the four replicate cultures grown under each unique condition

17 18 ′17 Culture δ Odiatom δ Odiatom Δ Odiatom Silica Biomass Description (h) (h) (h) Fraction (%) T.p. 15oC max 15.2 38.4 -0.292 52.2 T.p. 15oC 50 15.0 38.0 -0.273 46.8 T.p. 20oC max 13.7 35.5 -0.262 27.3 T.p. 20oC 50 14.0 36.0 -0.261 35.0 T.p. 30oC max 13.8 35.6 -0.263 31.4 T.p. 30oC 50 14.0 36.0 -0.249 21.6 C.f. 15oC max 15.1 38.3 -0.284 7.4 C.f. 15oC 50 15.4 38.8 -0.317 8.0 C.f. 20oC max 14.7 37.5 -0.293 5.3 C.f. 20oC 50 14.5 37.2 -0.290 11.1 C.f. 30oC max 13.6 35.4 -0.260 5.1 C.f. 30oC 50 14.4 36.9 -0.304 4.2

species. ′ The Δ 17O compositions of all cultured diatoms are strongly negative, a result that is compatible with the limited diatom measurements [119, 176] and more

extensive study of Precambrian chert (also SiO2) (Liljestrand et al. in review). The diatoms have an average 17E’ of -0.279 and range from -0.249 to -0.317 (Figure 3.4.4). These values are, however, more negative than the previously reported equilibrium predictions [78, 176]. Consistent with the equilibrium prediction though, there is a temperature-dependence within the measured diatom cultures such that as temperature increases, the 17E’ decreases. Predicted mass-dependent fractionation of 17O follows a power law relationship with respect to 18O. Here, 17/16α = (18/16α)θ [34, 143]. Like with the classic major isotope δ18O system, θ is temperature-dependent and decreases

93 Figure 3.4.4: Cultured diatom data binned by temperature. a) The 18ε and 17E’ colored according to growth temperature (circles) along with the equilibrium fractionation prediction at those temperatures (squares) [176]. Errors are 1σ based on internal reproducibility of diatom replicates. In (b), the residual offset between the temperature prediction and the measured frustules. Colored regressions show the measured fractionation factor (cf. vital effect) in 17θ. The error on these regressions is calculated from Johnston et al. (2007) [98]. The 15o and 20o are well within 2σ of equilibrium theory, whereas the 30o is at 3σ (all errors calculated on mean of temperature suite). However, in the 30o case, the error is heavily weighted by a possible outlier.

94 Figure 3.4.5: The relationship between temperature and 17θ between water and SiO2 for our cultured diatoms frustules. The error represents 1 σ calculated from Johnston et al., (2007) [98]. The blue line is the best fit between the thermodynamic high-temperature limit and culture data, whereas the red line is the equilibrium best fit from Sharp et al. 2016 [176].

from a maximum of 0.5305 at high temperatures towards lower values at Earth surface conditions [26, 78, 176]. Given the measured fractionation factors for

each sample relative to the composition of growth H2O, we determine the characteristic θ for each experiment (Figure 3.4.5). The best fit line for θ with respect to temperature follows the relationship:

− ± θSiO2−H2O = 2.316( 0.088)/T + 0.5305). (3.6)

The significance of this trend is dependent on anchoring the high temperature limit of 0.5305. Letting the fit move freely results in high-temperature prediction

95 of 0.5235, or no appreciable temperature dependence.

3.5 Environmental significance

The primary focus of this work is to evaluate the fidelity of frustule δ18O (and

′17 Δ O) to serve as a proxy for the temperature at which the SiO2 precipitated. At first pass, this is a simple comparison between theoretical predictions foran equilibrium isotope effect and measurements from diatom culture experiments. In performing this test, it is clear that temperature does in fact play a significant role in setting the triple oxygen isotope composition. It is also clear that kinetics, likely involving diatom physiology, also contributes in a meaningful way to the net isotope effect. So much so that it is unclear how ’sharp’ a tool this isotope effect will be in reconstructing paleotemperatures. To dig deeper, we first untangle the isotope further, considering the possible role for biological kinetics. From there, we consider the physiology of the organism as it relates both to forming Si0O bonds and further how this signal ripens in the environment. The measured oxygen isotope effects in diatom frustules from the presented semi-continuous cultures are not systematically offset from equilibrium predictions. That is, the 15oC and 20oC δ18O compositions are isotopically more negative than the equilibrium prediction, while the 30oC δ18O compositions are isotopically heavier. There are however numerous different temperature calibrations, so if we simply use these data as the first robust experimental calibration, clearly the data fall around a much tighter calibration (see Figures 3.7.1, 3.7.2 and equation 3.6). The same sort of relationship between the

96 ′ equilibrium prediction [176] and 18O is observed for 17O. In fact, the Δ 17O composition is statistically distinct, with an average offset of ≈ -0.045h. This offset is fairly constant across growth temperatures and lends itself to defining a slope (noted as 17θ, Figure 3.4.5) that carries a distinct temperature dependence from equilibrium predictions. At the magnitude of 18ε recorded by the diatom frustules, the error in these slope estimates will be ≈ 0.006 (calculated from [98], noted at the bottom left of Figure 3.4.5). In directly comparing the measured triple oxygen isotope composition to the predicted equilibrium composition we can calculate the difference, a fractionation we might attribute to biological control on precipitation or vital effects (Figure 3.4.4b). An alternative approach to interpreting this relationship is to consider it as a two step reaction: step 1 is the pure, inorganic equilibrium isotope effect, with step 2 being a purely kinetic, biological effect whose magnitude is the offset. One could then assign a 18α and 17θ to this inferred vital effect fractionation. The difference in 17E’ (which again is equivalent in scale to ′ Δ 17O) is large compared to the difference in the companion δ18O. This would require the putative vital effect fractionation to have extreme 17θ value [216]. Typical mass-dependent fractionation processes carry characteristic 17θ values between 0.52 and 0.53, varying at the < 1% level [119, 152]. As noted in Figure 3.4.4, the 17θ calculated here range from 0.558 ± 0.18 at 15oC to 0.42 ± 0.35 at 30oC - far outside the reasonable range of mass-dependent effects. However, as the δ18O of the measured frustule approaches the predicted equilibrium composition (18ε gets small), uncertainties in calculated 17θ will grow non-linearly

97 [98]. This results in the 15oC and 20oC regressions as within 2σ of typical mass-dependent effects, and theo 30 C at the 3σ level. We thus stop short of defining these offsets as anything other than typical mass-dependent isotope effects (see also Figure 3.7.3). The apparent non mass dependent θ ascribed to the vital effect suggests that this separation of the overall fractionation into equilibrium and biological components is not suitable for the diatom silica system. Instead we can ask what factors might change the θ of the initial precipitation process such that it does not match the equilibrium prediction. While equilibrium isotope effects have a strict temperature prediction in both δ18O and 17θ, kinetic isotope effects can express a wider range of values46 [ , 216]. For example, diffusion carries a θ as low as 0.50. That noted, due to the absence of any discernible growth rate effect in δ18O, it is unlikely that diffusion has a significant influence on the net triple oxygen isotope effect in frustules. More likely isthat the organic substrate of frustule precipitation, the long chain polyamines and silaffins, express some kinetic selectivity when adsorbing silica. Such an effect would have to be minor in order to not skew the δ18O composition relative to the equilibrium prediction (or is coincidentally similar in scale), but may be ′ significant enough to generate some of the measured Δ 17O offset. At the cellular scale, silica δ18O would become depleted relative to the equilibrium fractionation if the precipitation was limited by the rate of oxygen delivery to the active site of precipitation [74]. Diatoms have an obligate silicon requirement for growth, and both of the cultured diatom species have silicon

98 transporter genes to actively uptake dissolved silica during frustule formation

[22, 82, 83]. Oxygen in silicic acid (Si(OH)4), exchanges rapidly with the surrounding water immediately before silica precipitation, giving rise to an ideal temperature recorder. This then differentiates the information in δ18O from that of δ30Si. Further, neither the cell wall or the membrane of the silica deposition

vesicle (SDV), where precipitation occurs, are sufficiently impermeable2 toH O to influence the overall fractionation, even at the highest growth rates53 [ , 191]. In the modern ocean, silicic acid is undersaturated with respect to silica precipitation [201]. Due to this, silica precipitation in diatoms is catalyzed and regulated by an organic substrate composed of long chain polyamines and silaffins. Both polyamines and silaffins will drive silica precipitation in vitro from silica-bearing solutions and different biomolecule compositions will produce correspondingly different silica morphologies [111]. Again, silica precipitation in diatoms is therefore not a strictly equilibrium reaction. Despite this, diatomaceous silica is routinely used to reconstruct paleotemperature based on equilibrium predictions [78, 107, 176] and used as a means to infer paleo

[Si(OH)4][182].

3.5.1 Sedimentary Preservation

Abiotic silica precipitation occurs at equilibrium with the surrounding water, so ′ our observed δ18O- Δ 17O offset may potentially serve to differentiate biological precipitation. This will only be true in the environment, however, if the non-equilibrium vital effect persists during early diagenesis. Diatom silica

99 undergoes a ripening process in sediment before it is ultimately preserved in the geologic record as microcrystaline quartz [103]. Solid state NMR is commonly used to determine the degree of silica maturation by measuring the average bond state of Si atoms. Specifically, this distinguishes between Si atoms that are linked through bonds with oxygen to 4 adjacent Si atoms, and Si atoms that are linked to 3 or fewer Si atoms (noted as Q4 vs Q3), with the remaining bonding sites taken by hydroxyl groups. Intracellular silica has a Q4/Q3 ratio that evolves from 1.5 to about 2 as the new valves develop, which then reaches 2.5-2.8 as the silica is extruded out of the SDV [14, 69]. The silica becomes increasingly coordinated in the sediment and rock records ranging in value from 2.7 to 6.5 [31, 67, 169]. The progressive increase in Q4/Q3 ratio indicates a relative loss of oxygen compared to silicon, most likely at the expense of the exchangeable hydroxyl surface layer. In the δ18O, this diagenetic ripening is often characterized by an oxygen isotope enrichment in mature samples. Several studies have observed an enrichment in sedimentary chert δ18O relative to diatoms in the overlying water column on the order of 7h [52, 148, 169]. It is not clear whether this shift reflects an isotopic re-equilibration at sedimentary conditions, or whether it reflects the partial dissolution of silica. This distinction could determine the ′ possible preservation of a biological Δ 17O signal. Dissolution may preserve the primary oxygen isotope composition, much like prefluorination successively shifts δ18O while maintaining the structural oxygen fraction/composition. Re-equilibration, however, would tend to drive O-isotope compositions towards the previously established equilibrium fractionations [78, 176]. Artificial diatom

100 frustule aging experiments have shown a δ18O shift in postmortem diatom silica without any corresponding dissolution, which suggests that the oxygen isotope composition may indeed reset during diagenesis [52, 169]. This effect may also explain the difference between diatom temperature calibrations derived from drill core samples [176] and the culture experiments presented here. And finally, dissolution however doubtlessly affects diatoms in modern undersaturated porewater and will influence the final isotopic composition11 [ ]. More work needs to be done to determine the variability of θ in natural diatom samples. The vital effect influenced small θ has only been observed in this study ′ which exclusively measured cultured samples. If the light Δ 17O compositions are observed in sediment, and if the signal survives early diagenesis, it may inform us ′ about the evolution of biogenic opal formation. Specifically Δ 17O may provide a proxy which can distinguish between biological and abiotic silica precipitation in samples otherwise indistinguishable by δ18O. The marine silica cycle transitioned to biological control in the early Phanerozoic with the evolution of silica ′ precipitating organisms [134]. With Δ 17O measurements we might track the evolution of silica precipitating sponges, radiolaria, and diatoms, and the corresponding evolution of the silica cycle.

3.6 Conclusion

In order to assess and calibrate the temperature-dependence of the oxygen isotope composition of diatom frustules, and in doing so identify any possible vital effects, we cultured two different species across growth rates and

101 temperature. With the cultured biomass we developed a prefluorination scheme ′ to reproducibly measure the δ18O and Δ 17O of the structural fraction of the silica frustules. Neither the variable species or growth rates produced a discernible ′ δ18O or Δ 17O effect - only temperature variations produced a systematic trend. This result makes diatom silica more valuable as a paleo-climate proxy as possible confounding variables play no role in the final O-isotope composition of the frustules. The measured relationship between δ18O and temperature matches ′ previous estimates, but Δ 17O compositions were consistently more negative than ′ the equilibrium predictions. This Δ 17O vital effect is likely a product of the silica concentrating mechanism in diatoms selecting against 17O incorporation in the silica. This selectivity is expressed as an average θ between water and silica of 0.5227 ± 0.006, lower than the equilibrium prediction. The vital effect generated ′ during silica precipitation may allow Δ 17O to be used as a measure of biological contribution to geologic SiO2. The Phanerozoic oxygen isotope record may then have unexplored structure that will reveal not only temperature trends but also the biological control over the marine silica cycle.

3.7 Additional Information

3.7.1 Prefluorination test

The fluorination method we used to remove hydroxyl-oxygen, water, etc (i.e. all non-structural oxygen) from the bulk isolated silica is new to this study. We first

calibrated the amount of F2 required to sufficiently prefluorinate the frustule

102 samples. We loaded up to 21 samples, exposed all to F2, periodically extracting

the liberated O2. Multiple samples were prefluorinated simultaneously, so the O2 reflects the average composition of the samples in the tray. The molar quantity of extracted O2 was measured using the peak integrated area of the in-line GC, which allowed us to calculate the rate of O2 accumulation. The oxygen isotope composition of the evolved O2 was measured to track any changes in its provenance. We predicted the rate of O2 generation would decrease while the ′ δ18O and Δ 17O values would approach a stable composition.

18 The δ O of the evolved O2, shown in Figure 3.3.2 were initially light, less than 15h, then asymptotically approached a heavier composition more representative of the sample composition. This trend matches previous observations of stepwise fluorination of diatoms and likely reflects the dehydration of the sample [139]. Every measurement was heavier than the previous, suggesting there is no strict point where all the reactive oxygen has been removed, but there is a clear inflection point after about 400 Torr*hour. The ′ corresponding Δ 17O data (Figure 3.3.2) does not show a distinct trend over time. Instead the prefluorination aliquots seem to oscillate around the average measured composition of samples in the tray. But the majority of prefluorination

′17 O2 are within the 1σ uncertainty of 0.02 for our Δ O measurements.

The evolved O2 shows an exponentially decreasing trend as a function of time during this test prefluorination (Figure 3.3.1). This decrease is best fit as a sum of two exponential functions, a quickly decreasing reactive fraction − ∗ (y = 129.7 ∗ e( 0.01585 x)) along with a slowly decreasing less reactive fraction

103 − ∗ (y = 40.04 ∗ e( 0.0006077 x)). We postulate that the reactive fraction is composed of easily exchangeable oxygen like hydroxyl groups and water, while the less reactive fraction is composed of the structural oxygen, which is our target. Even after more than 24 hours of preflurination at 402 Torr,O was still evolving from the sample. However, all sample would be lost to fluorination if more time was allocated. Instead, we prefluorinated samples sufficiently to remove the reactive

O2 fraction based on this model fit. For the T. pseudonana, which had a higher silica biomass fraction and therefore more available SiO2, we chose to prefluorinate for 11 hours at 40 Torr. For the C. fusiformis with lower silica biomass fractions and less sample mass, we chose to prefluorinate for 5 hours at 40 Torr. The increased measurement accuracy that results from the larger less fluorinated samples makes up for the increased possibility of a contribution from exchangeable reactive O2.

3.7.2 Species specific δ18O relationships

Figure 3.4.2 shows the effect of temperature on diatom δ18O, but it does not show the species specific relationships. Figures 3.7.1 captures the box and whisker data for each of the diatom species individually. The best fit equation describing the T. pseudonana is:

· ± · 6 −2 ± 1000 ln[α(diatom−water)] = 1.73( 0.92) 10 T + 15.97( 10.58), (3.7)

104 Figure 3.7.1: The measured, species-specific δ18O of frustules. At left in blue is data is from C. fusiformis, whereas at right in red is data from from T. pseudonana.

while the best fit equation describing C. fusiformis is:

· ± · 6 −2 ± 1000 ln[α(diatom−water)] = 1.96( 0.73) 10 T + 14.11( 8.47). (3.8)

These equations are equivalent to each other within the% 95 confidence interval, demonstrating no species-dependence to the diatom temperature relationship. Figure 3.7.2 shows the individual δ18O values normalized to the best fit line with each point colored by species. The 20oC bin shows a significant offset between the two species which may lead to distinct temperature relationships between the two.

105 Figure 3.7.2: Measured δ18O of diatom silica normalized relative to the best fit line derived in the main text. Red data isfrom T. pseudonana and blue data is from C. fusiformis. The gray bars reflect 1 and 2 σ errors on any given measurement.

Figure 3.7.3: The relationship between temperature and 17θ between water and SiO2 for our cultured diatoms frustules. The error represents 1σ (heavy black), 2σ (light black), and 3σ (dashed gray) - all calculated from [98].

106 4 Isotopically anomalous organic carbon in the aftermath of the Marinoan Snowball Earth

In review for Geobiology

107 4.1 Abstract

Throughout most of the sedimentary record, the marine carbon cycle is interpreted as being in isotopic steady state. This is most commonly inferred via isotopic reconstructions, where two export fluxes (organic carbon and carbonate) are offset by a constant isotopic fractionation of ∼ 25h (termed

εorg−carb). Sedimentary deposits immediately overlying the Marinoan snowball Earth diamictites, however, stray from this prediction. In stratigraphic sections from the Ol Formation (Mongolia) and Sheepbed Formation (Canada) we observe a temporary excursion where the organic matter has anomalously heavy δ13C and is grossly decoupled from the carbonate δ13C. This signal may reflect the unique biogeochemical conditions that persisted in the aftermath of snowball Earth. For example, physical oceanographic modeling suggests that a strong density gradient caused the ocean to remain stratified for about 50,000 years after termination of the Marinoan snowball event, during which time the surface ocean and continental weathering consumed the large atmospheric CO2 reservoir. In an attempt to explain the observed carbon isotope record, we developed a model that tracks the fluxes and isotopic values of carbon between the surface ocean, deep ocean, and atmosphere. By comparing the model output to the sedimentary data, stratification alone cannot generate the anomalous observed isotopic signal. Reproducing the heavy δ13C in organic matter requires the progressively diminishing contribution of an additional anomalous source of organic matter. The exact source of this organic matter is unclear.

108 4.2 Introduction

The Neoproterozoic Era (1000-541 Ma) is a period in geologic history characterized by innovation in biology [36, 57, 109], the breakup of the supercontinent Rodinia [121], and a series of snowball Earth events that covered the globe in ice [85]. Perturbations have been noted in the carbon, sulfur, and oxygen cycles and in certain cases are stratigraphically linked to global glaciation [9, 85, 94, 165]. Here, much focus has been given to the carbon cycle (cf. [71, 85]), where arguments suggest a record of strong climatic and geochemical changes based on the δ13C of carbonate. Although less well represented, similar studies have closed the loop providing organic carbon δ13C in complement to carbonate carbon in an effort to better understand the overall biogeochemical response to these extreme climatic events [44, 97, 100, 141, 167, 194, 208]. Here we focus on the younger of the two glaciations (the Marinoan) and use the behavior of the surface carbon cycle as a vehicle to assay the nature of the biosphere and environmental change at this time. The Marinoan snowball Earth event> ( 640-635 Ma) is characterized in the geologic record by evidence of glaciation at equatorial latitudes [85]. One hypothesis is that weathering of low latitude continents decreased atmospheric

CO2 causing a runaway glaciation (cf. [39]). Glaciation only terminated when

volcanic emissions increased atmospheric CO2 sufficiently for greenhouse

warming to overcome the ice albedo [84]. The subsequently warm, high2 CO conditions accelerated the hydrologic cycle, rapidly weathered the post-glacial

109 continents, delivered alkalinity to the ocean, and drove global carbonate precipitation [81]. In the geologic record, this is observed globally in the form of cap carbonates overlying the glacial diamictite [85]. Synglacial carbonates are rare, so these cap carbonate deposits and the sediments above are the best window to understand both the environmental conditions adopted from a synglacial ocean and the lasting biogeochemical consequences that extended into the post-glacial world. Recent work suggests that immediately after the Marinoan deglaciation, the ocean was intensely stratified215 [ ]. Here, a large volume of relatively warm fresh meltwater would cover the dense, previously glacial, cold, and salty ocean. This plumeworld hypothesis was originally proposed to explain the globally preserved cap carbonates [5, 181]. Sulfur isotope measurements [179], as well as magnesium and strontium isotope measurements [122] of the post-Marinoan cap dolostone are all consistent with the idea that the carbonate precipitated in a globally stratified ocean. These ideas have been further bolstered by2 Ca + and Mg2+ isotope work [2]. Estimates of the stratification timescale vary, and have recently been extended to be as long as 50,000 years (with lower estimates at 8,000 years) [122, 215]. Related to these estimates, the inferred deposition rates of cap carbonates are also uncertain, ranging from 102 years [85] up to 105 years [64, 81, 202]. Previous studies have addressed general carbon cycling in Neoproterozoic stratigraphic sections immediately postdating the Marinoan glaciation [84, 85, 167]. These units characteristically capture evolving carbonate carbon

110 isotope compositions, but (as noted above) are less commonly complemented by the carbon isotope composition of corresponding organic matter. When this has been done, for example, isotopic variability in the offset between carbonate and

organic matter has been used to argue for low post-glacial2 CO [167] - this is in stark contrast to climate models [87, 155] and minor oxygen isotopes in Mainoan cap-barite crystal fans [9]. Following on this work, in this study we present δ13C data of carbonate and organic matter from two post-Marinoan

13 sections that show particularly anomalous δ Corg signals - an isotopic character yet unseen in the Neoproterozoic.

4.3 Geological setting

We focus on two well-studied post-Marinoan stratigraphic sections, the Ol Formation in Mongolia [129] and the Sheepbed Formation in the Yukon [100] [131]. These two stratigraphic units cover similar periods of time (and are anchored on glacial deposits), but also carry important lithological differences that allow further interpretive insight. Overall, however, these sections provide a window into carbon cycling on platform margins during the immediate aftermath of the Marinoan snowball Earth. The geological and regional setting, along with basin interpretations are provided elsewhere [129, 131], however we do provide a facies description and stratigraphic columns in Figure 4.3.1. The Ol Formation sharply overlies glacial deposits of the Khongor Formation and is composed of ∼7 m of dolostone succeeded by ∼20 m of limestone. This

111 lte oehra ih c,wt ifrn xsscales. Yaxis different with (c), right at together plotted oiino ukTCfo hl ( shale from TOC bulk of position info h uo speoiatyslccatcwt nebde carbonates interbedded are with Forma- circles siliciclastic Sheepbed Red predominantly the (b). is whereas Yukon (a) the carbonate from is tion Formation Ol Mongolian The raisfo h abnts( carbonates the from organics iue4.3.1: Figure K Ol Formation Shur. Glaciogenic diamictiteGlaciogenic Micrite Calcsilite Grainstone Intraclast breccia a b (c) (b) (a) -30 2 1 0 -10 -20 δ 13 rsne r h etoscnann anomalous containing sections the are Presented C δ 13 Barite and aragonite fans aragonite and Barite Limestone Dolomite carbonateminor with Shale sandstoneCoarse-grained 0 m 20 40 60 80 C carb

aus pnbu ice r h opsto of composition the are circles blue Open values. IBR Sheepbed Formation δ 13 C δ 3 2 1 0 -10 -20 -30 org 13 C − 112 c org δ hl lsdbu ice r h com- the are circles blue closed while ) 13 − C Carbonate Carbonate TOC (wt%) s .The ). > 0.05 0.20 < x > 0.05 > 0.20 0 m 40 80 120 160 200 ε org

− Yukon (interbedded carbonate) carb 0 m 160 200 120 40 80 0 o ohscin are sections both for 02 040 30 20 10 δ 13 13 C ε org signals.

0 m 20 40 60 80 Mongolia cap dolostone consists of buff to pink-colored, largely recrystallized micropeloidal dolomite, while the overlying limestones start with buff limestone ribbonite (nodular bedded calcisiltite), succeeded by gray rhythmite (flat, graded beds of micrite and calcisiltite), iron-rich siltstone interbedded with marly carbonate, and gray grainstone [129]. Above the cap dolomite, the carbonate rocks are mostly limestone, with the occasional presence of partially-dolomitized carbonates [19]. Sea floor cements made mainly of aragonite pseudomorphs are observed at the limestone-dolostone transition, present either as individual blades growing into overlying sediment or as crystal fan shrubs [126]. The cap dolostone, ribbonite, and rhythmite are interpreted to represent a transgressive systems tract culminating with a maximum flooding surface in the iron-rich siltstone, and succeeded by a highstand systems tract in the overlying grainstone. Importantly, Neoproterozoic carbonate strata were deposited on an isolated carbonate platform margin and ramp [19], analogous to the modern Bahama bank, and there is little evidence of terrestrial input throughout the succession. The Sheepbed Formation overlies the ∼1 m Hayhook limestone, the ∼30 m thick Ravensthroat cap dolostone and laterally discontinuous wedges of the Marinoan Stelfox diamictite. The thickness of the Sheepbed Formation varies between 200 and 600 m and is composed of shale with thin limestone interbeds. The shale from the lower 100 m of the Sheepbed is the most fissile and is likelyto represent the maximum flooding surface of the post-Marinoan snowball recovery [100, 131]. In the highstand systems tract of the upper Sheepbed Formation, siltstone and dolostone interbeds become more common before transitioning

113 into dolomite of the Gametrail Formation.

4.4 Analytical methods

Isotope data are presented using standard delta notation with units of permil (h) reported relative to the Vienna PeeDee Belemnite (VPDB) standard.

13 13 Measurements of δ Ccarb and δ Corg were made separately, but where possible, are from the same hand sample. Carbonate samples were first cut, then micro-drilled on the exposed surface to obtain 5-20 mg of powder. Samples that showed visible secondary alteration, veining, fractures, and large siliciclastic components were avoided. Analyte CO2 was collected cryogenically from the

o powdered samples by reaction in a common, purified3 H PO4 bath at 90 C. The

13 subsequent δ Ccarb analyses were made with a VG Optima dual-inlet mass spectrometer attached to a VG Isocarb preparation device with a precision of ± 0.2h. Organic matter was extracted after decalcification in concentrated HClfor 48 hours. This residue was then analyzed with a Carlo Erba Elemental Analyzer attached to a ThermoFinnigan Delta V configured in continuous flow mode. External error of this measurement was ± 0.3h for δ13C and 0.05 wt% for TOC.

13 For the Sheepbed Formation, we measured both the δ Corg from the organic

13 matter in carbonate lenses and the δ Corg from the organic matter in the shale.

13 The Ol Formation had no significant siliciclastic beds, sothe δ Corg measurements all come from carbonates.

114 4.5 Geochemical results

The δ13C data of both carbonate and organic matter from our measured sections is presented in Figure 4.3.1. The Ol and Sheepbed formation carbonates are both isotopically light - a characteristic of post snowball Earth deposits. The Sheepbed carbonate averages -5.9h with a total range of 3.6h. The Ol carbonate averages h 13 -3.2 , and carries a slight trend of increasing δ Ccarb upsection, with a range of 4.5h.

13 13 The Ol and Sheepbed formation δ Corg, unlike the coeval δ Ccarb, have strongly variable compositions over these same intervals. Immediately after cap

13 carbonate termination, the Ol Formation δ Corg carries a heavy composition of h 13 -13.3 . Over the subsequent 10 meters (moving upsection), the δ Corg ≈ h 13 h decreases by 12 , reaching a more characteristic δ Corg value of -25 . In the 13 h subsequent 80 meters, the δ Corg remains stable, with an average value of -24 . In the Sheepbed Formation, where we extracted and measured organic matter from both shale and interbedded carbonate lenses, only the carbonate derived

13 organic matter showed a similar anomaly to the Ol Formation. The δ Corg−c (the δ13C of organic matter sourced from the carbonate lenses) reaches its maximum value of -11.8h immediately above the cap carbonate, then over 20 meters, the isotopic composition decreases before stabilizing at an average composition of h 13 13 -29.5 . The δ Corg−s (δ C of organic matter sourced from the shale) in contrast remains constant with a composition of -31h throughout the section.

13 13 The strongly variable δ Corg and near constant δ Ccarb in both of these

115 sections results in a correspondingly variable εorg−carb (the isotopic offset between phases). While the stratigraphic length-scale over which this change occurs is different between the sections, both show a trend of initially sharply decreasing

εorg−carb that subsequently stabilizes at more typical values [77]. The Sheepbed h section approaches a εorg−carb of 24.5 while the Ol approaches a εorg−carb of h ≈ h 21.5 . Most notably, the εorg−carb of 8 , as measured at the base of both sections, are extremely unusual in the geologic record. Isotope data is accompanied by measurements of organic carbon content, noted as TOC. The TOC in both the Ol and Sheepbed are low, where only two samples from the Ol exceeded 0.4%, and all of the Sheepbed samples are <0.3%.

13 The Ol sample with the greatest TOC is also an isotopic outlier, with a δ Corg 4.7h lighter than the next lightest sample. While all the samples have low TOC, the anomalous samples in the Ol formation are relatively enriched in TOC compared to the overlying 40 meters. This enrichment however is only barely resolvable given the 0.05% precision of the TOC measurement. The Sheepbed does not show the same trend, the TOC of the anomalous samples, 0.06%, is not statistically different from the TOC of the section above.

The trend of increasing εorg−carb upsection in the Sheepbed and Ol formations stands in contrast with typical post-cap carbonate sections where εorg−carb remains

13 constant (see Figure 4.8.1). The light δ Ccarb is characteristic of this time [85],

13 however the trend in δ Corg requires further explanation. Observing a similar

13 δ Corg signal in two sections deposited in different paleo-basins and lithologies suggests that the process creating these signals may share an origin and may

116 reflect some fundamental aspect of carbon cycling during the post-Marinoan world. However, the counterpart, wherein all other post-Marinoan sections do not carry this behavior must also factor into any model describing these data.

4.6 Discussion

It is common for sedimentary organic matter to carry an isotopic composition offset from carbonate by roughly 20-30h [77, 99]. For the majority of the stratigraphy examined for this study, the εorg−carb is indeed within this window with a mean of 22.6h and in keeping with observations from sedimentary units younger than 800 million years old [77]. These more typical observations are

contrasted with exceedingly small εorg−carb values lower in the Ol and Sheepbed formations - these minimal isotopic offsets require additional, or perhaps even atypical behavior within the marine carbon cycle. These observations can thus be considered in a variety of fashions, from interrogating the variability within fractionation factors themselves, to physical changes in the ocean-atmosphere system during deglaciation that might enable the production and preservation of the observed signal. To enable this discussion and quantitatively consider the consequences of a post-glacial ocean, we developed a model that simulates the ocean-atmosphere system immediately after the termination of snowball Earth (Figure 4.6.1). This model updates and extends previous work on the hydrology and carbon geochemistry of the post Marinoan recovery [81], and stands in complement to recent Ca2+ and Mg2+ isotope modeling [2]. The explicit goal is to quantitatively

117 13 explain both the more common δ C values as well as the anomalous εorg−carb. To do so, we track the flux and isotopic composition of carbon - as well as the fluxof other biogeochemically significant elements like phosphorus and calcium - between three boxes: the atmosphere, surface ocean, and deep ocean. Organic matter and carbonate produced in the surface ocean box contribute to a modeled sedimentary record, which can be compared to the measured sections. Given modern inputs, the model is conservatively tuned to approach a steady state that reflects roughly modern conditions for atmospheric2 CO concentration, forg, dissolved inorganic carbon (DIC) partitioning, carbon export flux, and carbon isotope fractionation. Further, when modeling the post-glacial recovery, the model incorporates new thinking on marine stratification timescales215 [ ]. The initial parameterization of the model best reflects the conditions immediately after snowball Earth9 [ , 101, 167]. A full model description and extensive sensitivity test are provided in the supplemental materials. The treatment presented herein is set to target the carbon isotope outputs of the ocean and atmosphere system. Importantly, recent work highlights the dynamics between surface and deep ocean exchange in the snowball aftermath [215]. For instance, it has been suggested that the stratified post-snowball ocean could last for upwards 50,000 years before modern-like circulation resumes - meaning that the ocean could be strongly heterogeneous with respect to the isotopic composition of DIC. We track this feature, as mixing is initially set such that the deep ocean overturns with an e-folding time of 50,000 years, and then relaxes to a residence time of 2,000 years (Figure 4.8.2). It is also expected that in

118 1 Atmosphere

CO2, H2O 2

Con¡ nent 2+ H2O DIC, Alk, P, Ca , H2O (1) Volcanic CO2 CO2

(2) Evapora on, scales to 4 5 atmospheric CO2 (3) Weathering (carbonate DIC + P OM CO2

and siliciclas c) (4) Riverine input Surface Ocean

(5) CO2 equilibra on 8 (Weiss 1974) 2+ 6 CaCO3 <- Ca + DIC (6) Downwelling of H2O and

dissolved cons tuents (7) Upwelling of H2O and 7 upwelling downwelling

dissolved cons tuents

(8) Organic ma¢ er produc on 9

remineraliza on, and DIC + P OM Deep Ocean

sedimenta on

(9) Carbonate produc on and

sedimenta on

OM CaCO3 Sediment

Figure 4.6.1: The boxes and fluxes of the carbon isotope model. The volcanic CO2 flux is the only net source of carbon to the system (1) whileor- ganic matter and carbonate deposition is the only net sink (8-9). Water and all other species of interest - primarily carbon but also phosphate, salinity, and oxygen - are exchanged between the boxes through riverine weathering (3-4), gas exchange (5), hydrothermal circulation (6-7) and particle sinking (8-9)

119 the immediate post-snowball Earth, high atmospheric CO2 drives the surface ocean to low pH. On a longer time scale, the DIC, alkalinity, and pH of the surface ocean recover as a function of continental weathering inputs. The strong density stratification between the surface and deep ocean inhibits immediate thermohaline circulation and isolates the surface ocean. This means that very little of the DIC increase originates from the deep ocean upwelling. Due tothis gradual buildup of alkalinity, carbonate precipitation does not occur immediately after deglaciation, instead riverine fluxes must increase surface ocean2 [Ca +] and − [CO3 ] sufficiently to reach supersaturation. This time gap is a feature not expressed in a well-mixed single box ocean model and may help explain the cap carbonates characteristically abrupt basal contact [84]. For most marine chemical signatures (salinity, DIC, pH, etc), the model illustrates heterogeneity between the surface and deep ocean, which quickly converge after the resumption of vigorous thermohaline circulation. In interpreting carbon isotopes, it is important to recognize that both the oxidized (carbonate) and reduced (organic matter) components of the carbon cycle are represented in the geological record. As carbonate is interpreted as reflective of DIC, it is appropriate to begin with an analysis of sedimentary

13 13 δ Ccarb. As captured in Figure 4.3.1, carbonate carbon initially carries a δ C of -2h, before decreasing to a minimum of -5h. With time, the composition of carbonate gradually increases. These same stratigraphic observations have been previously explained by varying features like the temperature at which the carbonate precipitates [81]. However, temperature is a key component of our

120 model and linked to atmospheric CO2, and cannot be independently varied to satisfy carbonate δ13C. Instead, the composition of carbonate is accounted for via Rayleigh fractionation during the rapid draw down and subsequent precipitation

of the large atmospheric CO2 pool, [85], and by decreasing forg, a mechanism

13 classically invoked to shift δ Ccarb [77]. In addition to that listed above, we consider contributions from carbonate vs. silicate weathering, the composition of weathered carbonates, changes in atmospheric CO2 and volcanic out-gassing, and variable phosphate availability. A full sensitivity test of these parameters is available in supplemental materials. After considering model implications for external perturbations to the marine carbon cycle, we are left looking at the inter-workings of the system itself, namely how organic carbon is generated and how it adopts its associated isotopic composition. The fractionation factor describing this process (εorg−carb) can be roughly divided into two main components, the direct biological contribution of

13 carbon fixation to εorg−carb and the inorganic carbon speciation that sets δ CDIC

(acknowledging the role of diagenesis in slightly altering εorg−carb)[151]. Quantitatively, fractionation factors respond most strongly to different

environmental factors. The inorganic component of εorg−carb is driven by the temperature-dependent equilibrium fractionation between carbon species. For example, higher temperatures produce a smaller fractionation. Due to the elevated greenhouse gas concentration (as demonstrated in Figure 4.8.2) the surface ocean temperature where the carbonate precipitation occurs was greatest

shortly after deglaciation and quickly decreases as pCO2 is drawn down. This

121 100 (a) CO2 depen. ε (b) variable T (c) exogeneous C 80

m abovem capcarbonate 0 100, 32, 12 (x103 ppm) 530 60 oC -40 -30 -20 -10 0 -40 -30 -20 -10 0 -40 -30 -20 -10 0 δ13 C δ13 C δ13 C

Figure 4.6.2: Comparisons between various model predictions and the measured data. Note that all model outputs carry significant detail, as outlined in the SOM, and lead to some non-linear behavior as it relates to the scaling of the variables in question. (a) When the εp is parameterized as a function of CO2 concentration, it carries a large consequence for εorg−carb. (b) Here, εp is fixed and surface ocean temperature is prescribed. Higher temperatures gen- 13 erate isotopically lighter δ Corg (c) The prediction based on anomalous carbon inputs (see text) for the middle scenarios in a and b. The model behavior near the base of the stratigraphic sections reflects the post glacial disequilibrium and how the system approaches carbonate saturation, surface ocean – atmospheric equilibrium, all of which are proportional to reservoir size.

decrease in temperature, and corresponding increase in equilibrium fractionation, corresponds in directionality to what is preserved in the geologic record, but falls short of a silver bullet (Figure 4.6.2b). It has also been argued that the δ13C of DIC and the resulting carbonates are diagenetically altered along geochemical gradients associated with the vigor of pore water versus seawater exchange [2, 80]. One exciting and directly applicable result of that work are revised estimates of the δ13C of DIC in the post- Marinoan surface ocean. These estimates, which extend down to -11 h, only make understanding the δ13C of organic matter in our measured sections more difficult to explain.

The biological component of εorg−carb, primarily εp, is also environmentally

122 dependent. Much experimental work illustrates that this fractionation is

dependent on CO2 concentration, microbial growth rate, and surface to volume ratio of the organism or cell [156]. We directly incorporate this CO2 dependence (Figure 4.6.2a), where growth rate is set as a function of continental weathering,

and surface to volume ratio is assumed constant. At elevated extracellular CO2

concentrations, consistent with the post-Marinoan world, the εp becomes larger.

Therefore at the beginning of the model simulation, when2 CO concentrations are greatest, the εp value is correspondingly large. Large εp would tend to produce large εorg−carb, an observation that is strongly incompatible with the presence of

13 the δ Corg anomaly. One might also consider a local disequilibrium driven by rapid CO2 drawdown by microbial mats. This could aid in changing the observed

13 εorg−carb value [86, 123], but would have counter-acting effects on the δ C of local DIC [2, 58].

In more detail, environmental controls, and specifically ties to pCO2, have counter-acting effects on the isotopic offset between carbonate and organic matter. High temperatures tend to decrease equilibrium isotope effects, while

high dissolved CO2 concentrations tend to increase εp. One might envision the physiological response as acting as a stronger lever, however this would require an

initial CO2 concentration of 400 ppm followed by a gradual increase to 1,500 ppm to satisfy the observations. This contradicts both the geologic evidence of cap carbonates as well as the geochemical/climatological models of elevated

CO2. If we forgo the environmental dependence of εp and instead impose a single fractionation factor for primary production, temperature becomes the dominant

123 variable controlling εorg−carb. However, realistic temperature changes (for instance cooling the ocean from 30oC to 5oC) only accommodates about 3h of the required 15h anomaly captured in the organic matter. This also suggests that extracting paleo-barometric data from these records is beyond the current calibration of the carbon isotope system. [167]. Through the analysis of model results, what is clear is that an atypical carbon cycle is required to explain the anomalous isotopic composition of organic matter. A full model sensitivity analysis, demonstrating the isotopic leverage of changes in statification timescales, hydrologic cycling, weathering intensity and protolith, riverine and P fluxes, volcanism, carbonate precipitation dynamics, and many more (see Figures 4.8.4-4.8.7), is captured in the supplemental materials. In short, these affiliated geochemical changes cannot account for the δ13C observations. It has been previously suggested an additional organic matter flux (such as from weathering) could contribute to preserved marine records [99]. Given isotope mixing, this additional organic matter must have a reservoir size and prescribed isotopic composition. For example, in Figure 4.6.2, the model above produces an organic carbon flux with a composition of h-35 but the overall 13 h sedimentary δ Corg is about -12 . This overall value can be produced either by mixing a small amount of very heavy organic carbon or a large amount of relatively lighter (though still very heavy) organics. For instance, mixing in an anomalous composition of 20h would require a 42% contribution to the overall 13 h sedimentary δ Corg whereas mixing in an anomalous composition of -10

124 would require a 92% contribution (recall that marine primary production is set by the geochemical component of the model). The only strict constraint is that

13 the anomalous carbon must be heavier than the heaviest sedimentary δ Corg value of -11.8h. To fit the stratigraphic trend in the data (Figure 4.6.2c), we model this anomalous contribution following an exponential decay function as some still theoretical pool of carbon is gradually depleted in mass. The combined

13 δ Corg (isotopically normal organic matter plus an exogeneous source) is described by the following function:

13 − · δ C − J · δ + J · (e k t) · δ org combined = org org anomalous anomalous , · −k·t (4.1) dt Jorg + Janomalous (e ) where Jorg is the flux of normal organic matter to the sediment, δorg is the isotopic

value of that normal organic matter, Janomalous is the initial flux of anomalous carbon, k is a decay constant tied to the anomalous flux, and δanomalous is the isotopic value of the anomalous carbon pool. This is simply a fitting exercise where calculating an anomalous contribution assumes that the magnitude of any anomalous flux is dependent only on the time elapsed and the initial flux.A linear mixing model could also be sufficient to explain the data, but given the nature of the upsection trend, we prefer an exponential relationship. Further, we imply no specific age model for these sections or the fit, but would imply thatthe fit suggests relatively consistent sediment deposition in each locality. This

13 assumption can be revisited if the source of the δ Corg anomaly is uniquely identified. To that point, the solution to this mixing problem is non-unique. Practically, however, it is still difficult to produce the heavy organic matter being

125 contributed to the marine realm, so we conservatively adopted an anomalous composition of -10h. Given this composition, the sedimentary record can be fit by varying the sedimentation rate of the whole sequence and the decay constant (k) of the anomalous flux (Figure 4.6.2). The source of this organic matter and what controls it delivery, remains open.

4.6.1 Environmental significance

13 The simple fact that the anomalous δ Corg is preserved in both Mongolia and the Yukon, but not elsewhere in the world, differentiates this anomaly from other, carbonate hosted excursions. More typical carbonate - organic matter relationships are preserved in Arctic Alaska (see SOM), China [44, 97, 141, 179, 208], Brasil [167], and are all compiled in supplemental

13 materials. Put differently, the δ Corg in Northwest Canada and Mongolia is an outlier to that seen elsewhere, however it does speak to a marine carbon cycle vulnerable (at least on the local scale) to exogenous contributions [99] or differential diagenesis2 [ , 80]. The answer to the mechanism behind this signal

13 may in fact be encoded in where we find (and do not find) the heavy δ Corg values. Importantly, we note:

1. these sections are not located in the same paleo-basin, and

2. the signal is even manifest in different host lithologies.

The cause of the anomaly must be common enough to have independently arisen in two locations without being so prolific that it becomes globally distributed.

126 This still begs the question of why Mongolia and the Yukon are similar to one another, yet different from other sampled post-Marinoan margins. The Ol and Sheepbed formations, in addition to being lithologically different, may also capture different depositional timescales. The Sheepbed Fm. carbon anomaly is formed over 20 meters of siliciclastic deposition, while the Ol carbon anomaly is formed over only 10 meters of largely carbonate dominated strata. This either speaks to the relative depositional rates (and the same overall timefor the compared stratigraphy), or heterogeneity between sites. As carbonate production should outpace normal siliciclastic deposition, the stratigraphic reach of the anomaly in both sections is further exacerbated. This temporal variation, in addition to the spatial heterogeneity, further discourages the idea that a single global process could produce the observed anomalies. The available data also allow for interrogation of how much organic matter is present within a sample (or section) as it relates to the accompanying isotopic

13 composition. In the Ol Formation, samples with anomalous δ Corg have significantly greater TOC than samples from the immediately overlying strata. This observation could be explained by a low background organic matter flux that mixed with a larger anomalous flux, summing to more TOC and a heavy isotopic composition. The relative standard deviation is large on the TOC measurements, but observing the degree to which TOC increases in the anomalous section, the

13 anomalous δ Corg possibly accounts for more than 80% of the organic matter in this section. This large contribution of anomalous organic matter to the observed

13 sedimentary δ Corg would suggest that the original composition of the

127 anomalous organic matter did not deviate significantly from the observed value of -13h. This framework helps explain the Ol Formation, however the Sheepbed Formation data does not lend itself to the same mixing analysis. Recall the

13 differences in the δ Corg between the host siliciclastics and the carbonate lenses, where only carbonate lenses record a signal similar to the Old Formation.

4.6.2 Decoupling the carbon cycle

A common mathematical way to vary the εcarb−org in geochemical models is to decouple the organic matter production from DIC and carbonate precipitation. If, for example, there was a large reservoir of dissolved organic carbon in the deep ocean, it may be possible that the carbonate sampled one source while the organic matter sampled another165 [ ]. Although theoretically plausible, our model cannot accommodate the observation, even with the imposition of a large post-Marinoan DOC reservoir (see SOM). In fact, even if deep ocean DOC was residually heavy from some previous event, there is no physical justification for suddenly precipitating/depositing this pool at the time of the excursion, and only on certain margins. Finally, the excursion occurs during a period of marine stratification in which the deep ocean is less, not more, likely to influence surface ocean and sedimentary processes. If it is not possible to decouple the organic matter production from carbonate precipitation would it instead be possible to change the environmental

conditions in order to vary εcarb−org? Some microbial metabolisms can produce unusually heavy organic matter, but not nearly as heavy as is required to satisfy

128 these observations [205]. As discussed earlier, producing an εcarb−org less than 10h solely by changing environmental conditions would require either low and

increasing CO2 concentrations or unrealistically high temperatures. Smaller isotopic signals however may be explained by differences in environmental conditions [156] or differential diagenesis [2]. It is possible that, for instance, the

Ol Formation was deposited in warmer waters, satisfying the smaller εcarb−org in the Ol than the Sheepbed (21.5h vs. 24.5h), but not the extremely small

εcarb−org at the base of both sections. Ultimately, the most parsimonious model

results match the sedimentary record more closely when εp is held constant, suggesting that either CO2 was actually invariant over this time period, not likely, or that εp was not strongly dependent on CO2. Still in search of an explanation for generating such isotopically heavy organic matter, we can consider the idea that this organic material is either 1) heavily altered, 2) not marine in origin, 3) not contemporaneous with the sediments, and/or 4) some combination of all. Alteration at high-temperature can be ruled out given the relatively low grade of these units (see supplement for discussion).

13 Isotopically heavy organic matter with δ Corg values comparable to the observed post Marinoan anomaly can be found in modern hypersaline microbial mats [168]. The evaporative environments increase nutrient concentrations and microbial growth rates until CO2 becomes limiting thereby reducing εp. Our model includes a parameter of increased evaporation due to the increased temperature and rapid hydrologic cycling, but the surface ocean salinity is also greatly reduced by stratification and glacial runoff during post-glacial time.

129 Additionally, there is no geologic evidence of evaporite formation or microbialites in the measured sections. Even without the specific hypersaline conditions a sufficiently thick microbial mat may inhibit2 CO diffusion and produce small εp values, but if this process was common we would expect to observe such small fractionation over the 3 billion year history of microbial mat formation. Anoxic conditions caused by prolonged stratification would have promoted the remineralization of sedimentary organic matter by methanogenesis, with a possible isotopic consequence for the residual organic carbon. Methanogenesis does produce methane depleted in 13C, but porewater DIC, not the residual organic matter, adopts a complementary enriched 13C composition [37, 211]. Producing the anomalous organic matter via methanogenesis would require the subsequent fixing of this DIC in sediment [172]. Even if such a process occurred, it would not occur in sufficient magnitude necessary to produce the observed anomalies. The most parsimonious explanation for these anomalies is still the addition of some unknown, exogenous, anomalous organic carbon flux to the background organic carbon deposition. This is taken from the clear, exponential decay of the signal upsection - a characteristic feature of diminished mixing of adjacent, isotopically distinct pools. However, our model does not assign a specific origin to this flux, instead it provides a generic pool of heavy organic carbon that mixes with the background production. As noted earlier, the magnitude of this hypothetical flux is dependent on its isotopic composition. Simply reproducing

130 the organic matter anomaly does not necessarily require the ultimate source of this anomalous OC be identified. Instead, as in [99], we may provide a general solution that is applicable regardless of the source of the anomaly, noting that the real challenge remains outstanding - how to generate such enriched organic matter. One possible source of the anomalous organic carbon is the newly exposed continents. Glacial scouring during the snowball Earth event may have newly exposed an organic matter source that was subsequently imported to these marine basins by physical weathering [91]. The anomalous carbon weathering

13 from the continent also explains the two different δ Corg values in the Sheepbed

13 13 section. The δ Corg−s values, consistently offset from δ Ccarb, probably sample

13 organic matter produced contemporaneously in the photic zone. The δ Corg−c, which carries the carbon isotope anomaly in the carbonate lenses, may have formed during pulsed weathering events. It is also critical to note the differing alteration histories of these two lithologies. In particular is the acknowledgement that the carbonate was surely subject to open system diagenesis, as evidenced by

18 δ Ocarb value [131] and arguments derived from heavier isotope systems [2]. Whatever the case, this decoupling disappears quite rapidly. Returning to the source of the organics, excess riverine input would deliver alkalinity to the DIC-rich surface ocean forming the carbonate lenses while simultaneously delivering the heavy organic matter weathered from the continent. For a moment, this may seem to satisfy the stratigraphic distribution, but leaves open the mechanism generating the organic matter.

131 Challenging this interpretation, the Ol formation has no geologic evidence of significant terrestrial input. Thus, it is unlikely that a significant fraction ofits organic matter comes from terrestrial sources. Alternatively, another possible transient source of organic matter to the sediment is from organic matter in cryoconite pans [88]. This organic matter, closely associated with the glacier, would be released into the ocean as the ice receded. This organic matter would

probably form in close equilibrium with atmospheric CO2 and would therefore 13 h have typical light δ Corg values, not the -10 characteristic of the anomaly. This mechanism too is insufficient.

4.7 Conclusion

We present a coupled carbonate -organic matter record through post-Marinoan sedimentary successions in the Northwest Territories of Canada and Mongolia. The δ13C of carbonate is characteristic of global trends, whereas the isotopic composition of the organic matter diverge from this global prediction. After a thorough tour of potential mechanisms that could drive such an effect, we are left without a unique environmental solution. The long term stable stratification of the ocean in the immediate aftermath of the Neoproterozoic snowball Earth events is possibly unique in geologic history, but this environmental condition

13 would not necessarily generate the anomalous δ Corg observed in the Ol and

13 Sheepbed formations. Prior explanations of anomalous δ Corg signals relied on the presence of a large deep ocean pool of DOC, but our model shows that even if such a pool existed the stable stratification would make such a pool inaccessible

132 to surface ocean processes [165]. Further these isotope signals do not reflect the

13 effect of changing pCO2 on algal physiology. Instead the heavy δ Corg data can only sensibly be described by the addition of an exogenous, anomalous organic carbon flux to the sediment99 [ ]. We lack a unique solution for the generation of this organic matter. Perhaps the anomaly is generated by weathering and subsequent deposition of heavy organic matter from the newly exposed adjacent continent. This also has challenges, however a non-marine solution would avoid the constraint imposed by buffering with a marine DIC budget. In the end, we are left with a robust, but challenging observation at an anomalous, but alluring period in Earth history.

133 4.8 Additional information

4.8.1 Literature data

13 The post Marinoan δ Corg data is compiled from the literature and compared to our new measurements in figure 4.8.1. Each of these sections show the δ13C composition of organic matter and carbonate in post cap deposits comparable to the Ol and Sheepbed sections. The two arctic Alaska sections were measured by our lab in addition to the Ol and Sheepbed sections addressed in the text. The two arctic Alaska sections, while they are not as well sampled as the Sheepbed or

13 Ol formations, clearly do not show the same anomalous trend in the δ Corg. In

13 both of the measured sections the δ Ccarb becomes gradually lighter over the h h 13 approximately 30 meter section, from 3 to -2 . The δ Corg meanwhile varies widely ranging from -14h to -32h, but do not show any clear trends. The h h εorg−carb correspondingly varies between 17 and 32 . Most sections only report approximately 10 meters of δ13C data, but it is clear even compared to this limited data that the systematic lightening upward trend of the Ol and Sheepbed formations are unique.

4.8.2 Metamorphism

When a sedimentary section undergoes a high degree of metamorphic alteration, the corresponding organic matter may fractionate and reach values as heavy as those observed here. This would occur due to the metamorphic removal ofnewly

134 Figure 4.8.1: This figure compares our newly measured sections to the post Marinoan carbon isotope data from the literature.

formed, isotopically light methane from the residual, increasingly carbon-rich,

13 organic matter75 [ ]. The δ Ccarb would not experience the same degree of alteration due to the isotopic inertia of the larger volume of carbon in carbonate. Metamorphic alteration may sometimes be difficult to identify in carbonates, which would make this a plausible explanation for the Ol formation anomaly. But the Sheepbed shale has not reached the necessary thermal maturity. Two other observations also suggest that metamorphism is not the cause of the Ol and Sheepbed anomalies. First, the anomalous zone is limited to a few meters above the cap carbonate whereas metamorphism should affect the sequences more

13 broadly. Second, and perhaps more revealing, the δ Corg−s from the Sheepbed maintains a stable, light, and presumably unaltered composition (recall that the anomaly is preserved in thin carbonate lens within the siliciclastic unit). Metamorphism would affect all pools of organic matter within a section or region.

135 4.8.3 Model details and discussion

For each timestep, the model tracks the hydrologic fluxes between each ocean box, the weathering flux from the continents, speciation of dissolved inorganic carbon (DIC), carbonate saturation state / precipitation, and organic carbon burial. The only net sources of carbon to the model is a volcanic fluxtothe atmosphere on the order of 1012 mol/year and the bicarbonate produced from carbonate rock during terrestrial weathering. The only net sink comes from the deposition of carbonate and organic carbon in sediment [213]. These fluxes are

allowed to evolve with time and accommodate the high initial atmospheric pCO2.

The initial concentration of atmospheric2 CO must be elevated for a sufficiently large greenhouse effect to overcome the high ice9 albedo[ , 101, 167]. We allow this initial value to vary when fitting the isotopic data, with sensitivity test available in the supplemental materials. For the baseline model, however, we

use an initial pCO2 of 32,000 ppm. Throughout the time domain of the model, the atmospheric and surface ocean temperatures are determined by the

atmospheric pCO2 [112, 155]. Using a single average global temperature for the two boxes may be nonphysical due to latitudinal variations, but is still the most accurate approximation given the constraints of a 3 box model. The deep ocean is initially set at 0oC, reflecting the temperature of the previously subglacial ocean. The ocean is initially partitioned as a 1500m surface ocean overlying a2500m deep ocean. The initial thick surface ocean is caused by the large volume of freshwater that would be produced as the snowball melted [215]. A normal surface ocean depth of 100m is reestablished concurrently with normal

136 H O Residence Time 2 4 Atmospheric CO Concentration 105 #10 2 A 3 B 2

ppm 1

0 0 25,000 50,000 75,000 100,000 -1 103 yr Organic Matter Burial Fraction 1 C

org 0.5 f

Surface Ocean Deep Ocean 0 101 0 25,000 50,000 75,000 100,000 0 25,000 50,000 75,000 100,000 Time (years) Time (years)

Figure 4.8.2: Shows the residence time of water in the surface and deep ocean pools as well as the trend of atmospheric CO2 and forg during a 100,000 year model run. a) Global ocean stratification ends after 50,000 years increasing thermohaline circulation and sharply decreasing the residence time. The deep ocean residence time remains larger due to the deep oceans greater size. b) Atmospheric CO2 is drawn down quickly first by dissolution into the freshwater surface ocean, then by carbonate precipitation. c) Organic mat- 13 ter burial is initially low to better reproduce the measured δ Ccarb trend, and subsequently returns to more typical values.

thermohaline circulation, once the enhanced postglacial stratification is overcome. Both the shallow and deep ocean are initially saturated and at equilibrium with the atmospheric pCO2. During the glaciation, only a small amount of ice-free ocean is required for the ocean to equilibrate with the atmosphere, justifying our choice of initial conditions [115]. The oxygen content of the Neoproterozoic atmosphere is less well known, thus we set pO2 at 0.1% of the modern ocean [186]. Gas fluxes between the surface ocean and atmosphere are always setto equilibrium, and consider the salinity and temperature of the surface ocean

137 [66, 210]. Weathering in the aftermath of the Marinoan glaciation was elevated, but the exact value is poorly constrained [93]. Our geochemical model addresses weathering by calculating the fluxes of salinity, calcium ions, DIC, alkalinity, and phosphorus from the continents to the surface ocean at each timestep. These fluxes are dependent on weathering intensity (the degree to which the continents are more easily weathered than in the modern) and hydrologic intensity (the degree to which the hydrologic cycle is more vigorous relative to the modern).

Each of these factors themselves scale to pCO2 such that a modern concentration of 400 ppm produce modern fluxes (1x). The elevated pCO2 at the inception of the model correspond to some multiplier of modern values (for instance 5x; Figure 4.8.4). Scaling the weathering fluxes in this way implicitly assumes that the compositional makeup of the continents in the Neoproterozoic is similar to today [197] and assumes weathering is ultimately regulated by pCO2. Our generic scalar is also assumed to capture other features important to chemical weathering (temperature, exposed mineral surface area, exposed continental area). Like with pCO2, these factors would also decrease through time in the model, consistent to previous parameterizations [81]. The weathering flux of DIC additionally scales to the ratio of carbonate/silicate weathering, as

carbonate weathering delivers two units of DIC per CO2 consumed, whereas silicate weathering delivers only one. The model tracks the conservative properties of DIC and alkalinity in boththe

− 2− surface and deep oceans. From these, [CO2], [HCO3 ], [CO3 ], and pH are

138 determined. The equilibrium constants K1 and K2 controlling this speciation are pressure and temperature dependent and are calculated for each ocean reservoir at each timestep [144, 166]. The flux of DIC and alkalinity during normal circulation between the surface and deep ocean is directly related to the concentration and the flux of water between those reservoirs:

· FDIC = FH2O [DIC]. (4.2)

2+ Each of the other species (PO4,O2, Ca , salinity) are treated the same way and follow hydrologic exchange between marine reservoirs. Carbonate precipitation is determined by its supersaturation based on [Ca2+]

2− and [CO3 ] concentrations at each timestep. Taking from the modern, the model has an additional multiplicative factor that determines the degree of supersaturation:

· 2+ · 2− SS [Ca ] [CO ] Ω = 3 , (4.3) Ksp

where SS is the allowable degree of supersaturation. The Ksp is a function of temperature, pressure, and salinity and is calculated at each timestep [144, 149]. Additionally calcite and aragonite have different equilibrium constants, the

model default is calcite precipitation at a SS of 5, and these factors can be adjusted (Figure 4.8.6). Organic matter production (i.e. primary production) is controlled by phosphate availability on geologic timescales. In the photic zone of the modern

139 ocean, phosphate is perpetually drawn down to negligible concentrations by biological activity, and our model treats the post-glacial OM productivity in the same fashion. DIC is fixed as organic matter in proportion to the total availability of phosphate in the photic zone (top 100 meters), after consideration of the Redfield ratio (106:1). Oxygen is produced at the same time at a ratio of 138:1, and the organic matter subsequently sinks into the deep ocean. In our model, organic carbon in the deep ocean is only remineralized through oxic respiration.

If there is abundant O2 in the deep ocean, as is the case in the modern, some fraction of this OM can be remineralized with the balance being deposited in sediment. Conversely, if there is insufficient O2 in the deep ocean,

remineralization proceeds until dissolved O2 is fully consumed, leaving the remaining organic matter to be preserved in sediment. This does overlook the presence of other metabolic pathways, however, considering their lower relative rates and more modest overall contribution, this assumption is warranted [73]. In parallel to tracking geochemical constituents and fluxes, the model also tracks the δ13C of all carbon fluxes and reservoirs. The isotopic value ofthe h volcanic CO2 input, as well as the initial atmospheric CO2, is set to -5 [77]. Most carbon transport fluxes, including weathering and hydrothermal circulation, are assumed to not express an isotope effect. Air-sea gas exchange is an exception, with fractionations taken from the literature [146, 221]. Given a

13 δ C composition of the DIC pool, the ultimate sedimentary εorg−carb value is controlled by the fractionation factors associated with four processes: DIC speciation, calcium carbonate precipitation, organic carbon fixation (εp), and

140 early sedimentary diagenesis [77]. Isotopic fractionation associated with early diagenesis is small and relatively invariant, and as such, is neglected in our model [117]. The DIC speciation fractionations are a function of temperature and salinity, and are calculated for both ocean boxes at each timestep as those environmental conditions change [221]. Like DIC speciation, calcium carbonate precipitation is also a function of temperature, but additionally depends on the mineralogy of the precipitate [164]. The model can calculate the ultimate εorg−carb for both calcite and aragonite precipitation. Of the four fractionation processes listed here, the δ13C fractionation associated with primary production has the largest potential range of values, and therefore has the largest potential to influence the ultimate sedimentary εorg−carb. Studies have shown this fractionation to vary between the RuBisCO-controlled end-member of -27h and the diffusion controlled end-member of -0.7h [74]. Specifically, culture experiments have shown this fractionation to be dependent on physiological/environmental variables like CO2 concentration, algal growth rate, and the organismic surface to volume ratio [156]. Surface to volume ratio is assumed to remain constant, but algal growth rate scales to the weathering intensity. As these features are largely impossible to know for the post-Marinoan world, the model can be run using either an invariant εp fractionation or an environmentally determined εp fractionation. There is also more to learn about the behavior of the carbon cycle through a

13 13 sensitivity analysis of the variables that directly influence the δ Ccarb. The δ Ccarb values of -5h are atypical in the geologic record and must in some fashion be

141 A B

13 Figure 4.8.3: Shows how the minimum value of sedimentary δ Ccarb is af- fected by A) the isotopic composition of volcanic carbon and the initial atmospheric CO2 concentration B) the fraction of carbonate weathering and the 13 δ Ccarb of the weathered carbonate. In each plot the variables of the other 13 part are held constant in order to isolate the isotopic effect. δ Ccarb is mini- 13 13 mized with low volcanic δ C, low carbonate weathering fraction, low δ Ccarb of the weathered carbonate, and high (though not maximized) initial CO2 con- 13 h centration. The modeled best fit was produced with δ Cvol = -5 , phosphate 9 flux = 10 mol/year, initial atmospheric CO2 = 32,000 ppm, % carbonate 13 h weathering = 50%, and weathered δ Ccarb = 0 . further related to the following variables (Figure 4.8.3):

13 h 1. The δ Cvol: The modern composition of volcanic2 CO is -5 , and is presumed to be similar throughout Earth history. However, changing this value will proportionally shift the composition of all downstream pools,

13 including δ Ccarb [77].

2. Phosphate availability: According to isotope mass balance, carbonate burial must dominate the export flux in order to approach theh -5 composition of the volcanic input flux. In keeping with previous work

[81], reducing forg for the duration of cap carbonate precipitation is

142 accommodated by decreasing the flux of phosphate - the primary nutrient.

3. Carbonate vs. silicate weathering: Weathering of continental carbonate produces a flux of carbon to the ocean that carries an isotopic value heavier h than the likely -5 composition of the volcanic CO2 flux. A relative increase in carbonate weathering then provides an additional flux of heavy

13 carbon to the ocean, ultimately resulting in greater δ Ccarb [77]. It is

13 therefore possible to achieve the light δ Ccarb observed after snowball Earth by decreasing carbonate weathering. 1

13 13 4. δ Ccarb of weathered carbonate: On long timescales the δ Ccarb of weathered carbonate is in steady state equilibrium with the carbonate exported from the ocean, but this value can vary on this models timescale.

13 Measured δ Ccarb from the Trezona anomaly preceding the Marinoan snowball event are abnormally light [70]. The degree to which this light carbonate was weathered in the aftermath of the Marinoan snowball event influences the isotopic composition of the subsequently precipitated 13 h carbonate. In a standard model the weathered δ Ccarb is set at 0 .

5. Initial atmospheric pCO2: Rayleigh fractionation will shift the isotopic value of the precipitated carbonate based on the degree to which the

surface ocean/atmosphere CO2 pool is drawn down. The model

weathering rate is a function of pCO2, so higher atmospheric pCO2 will

13 produce relatively more carbonate thereby depleting δ Ccarb. 1This model is initialized after all glacial ice has been returned to the ocean, rather thanwhen the continental margins are still exposed [81]

143 13 In both the Mongolian and Yukon sections, the δ Ccarb increases gradually above this minimum higher in the section, eventually returning to a typical Neoproterozoic δ13C value of 5h [19, 100]. This is primarily accommodated via the gradual re-balancing of carbonate weathering and forg to normal values. For other control variables, the model smoothly transitions from one value to another with an equation of the form:

− V2 V1 Vt = − + V1, (4.4) 1 ·(t−t ) 1 + e Smt s

Where: Vt is the variable at time t, V1 is the variable during the stratified period,

V2 is the variable after stratification, Smt is a smoothness term influencing the speed of the transition, and ts is the timescale of the stratification. This equation is applied to upwelling and downwelling water fluxes, the carbonate to silicate ratio of the weathered continent, and the baseline riverine phosphorus flux. Hydrologic intensity and Weathering intensity are thought to scale to atmospheric CO2 in a non-linear way, but the exact functional form of the relationship is not clear. The atmospheric CO2 and the weathering/hydrologic intensity at time 0 are input variables therefore they provide one constraint. Two additional constraints are given based on the observation that a modern CO2 concentration ( 400 ppm) produces modern weathering/hydrologic intensities

(1). Also, as an endmember, 0 CO2 should produce no weathering or hydrologic activity. I fit the constraints to a function of the following form:

1 · 2 · (Hi/Wi)t = a [CO2]t + b [CO2]t + c, (4.5)

144 Where a, b, and c are constants, [CO2]t is the atmospheric CO2 concentration at

time t, and (Hi/Wi)t is the hydrologic or weathering intensity at time t. Solving for the constants a, b, and c given our constraints gives the following equations:

− · [CO2]0 400 (Hi/Wi)0 a = 1 , (4.6) · − · 2 20 ([CO2]0 20 [CO2]0 )

1 · − 2 20 (Hi/Wi)0 [CO2]0 b = 1 , (4.7) · − · 2 20 ([CO2]0 20 [CO2]0 )

c = 0, (4.8)

These values can then be plugged into the above equation to determine the weathering and hydrologic intensity at any atmospheric CO2 concentration given any combination of initial conditions (Hi/Wi)0 and [CO2]0. The pools of each element in the surface ocean, deep ocean, and atmosphere are iterated each timestep (year) based on the flux of each element into and out of that pool. For example the deep ocean carbon pool will be described at each timestep by:

− (DICDO)n = (DICDO)n−1 + FDICSO−DO FDICDO−SO, (4.9)

The element fluxes are then supplemented by the isotopic composition ofeach flux and pool, for example:

145 13 · 13 · 13 (δ CDICDO)n = ((DICDO)n−1 (δ CDICDO)n−1+FDICSO−DO (δ CDICSO)n−1

− · 13 · −1 FDICDO−SO (δ CDICDO)n−1) ((DICDO)n) (4.10)

4.8.4 Model Variable Sensitivity

Figure 4.8.4 (following page): 1) Shows the modeled δ13C output of carbonate and organic matter as a function of the timescale of ocean stratification. 2) Shows the modeled δ13C output of carbonate and organic matter as a function of the abruptness of ocean overturning. Smoothness corresponds approximately with the years required to end stratification. 3) Shows the modeled δ13C output of carbonate and organic matter as a function of the hydrologic intensity, the riverine flux relative to the modern. 4) Shows the modeled δ13C output of carbonate and organic matter as a function of the weathering intensity, the degree to which the minerals tend to weather relative to the modern. 5) Shows the modeled δ13C output of carbonate and organic matter as a function of the baseline riverine DIC flux. Increased weathering also cor- 13 responds with increased CO2 drawdown. 6) Shows the modeled δ C output of carbonate and organic matter as a function of the early riverine P flux. This represents the flux of phosphorus during the time that the ocean was stratified. Increased weathering also corresponds with increased CO2 drawdown.

146 Table 4.8.1: Variables tested in model sensitivity analysis

Figure # Variable Units Standard Description Model Run Value 1 Ocean years 50000 time that the model remains stratified before stratification reestablishing normal circulation time 2 Smoothness 2000 abruptness with which normal circulation resumes 3 Hydro 5 the magnitude that defines how weathering intensity and river flux is increased relative to the modern 4 Weathering 5 the magnitude to which rocks are more intensity susceptible to weather and river ion concentrations increase 5 C flux mol/year 1013 baseline riverine carbon flux as modified by the Hydro intensity and Weathering intensity 6 P flux1 mol/year 109 baseline riverine phosphorus flux during the stratified period of the model 7 P flux2 mol/year 4 ∗ 1010 baseline riverine phosphorus flux after stratification 8 Remin rate 0.95 maximum deep ocean remineralization rate 9 C S r1 0.5 carbonate silicate weathering ratio during stratified period (1 is entirely carbonate) 10 C S r2 0.5 carbonate silicate weathering ratio after stratified period (1 is entirely carbonate) 11 Atm CO2 ppm 32000 initial atmospheric CO2 concentration at time 0 12 Atm O2 ppm 235 initial atmospheric O2 concentration at time 0 13 CarbMin 1 binary variable that describes whether calcite or aragonite precipitates from the ocean (1=calcite, 0=aragonite) 14 SuperSat 5 degree to which ocean must become supersaturated before carbonate precipitates 15 d13C Vol h -5 isotopic value of volcanic CO2 input to atmosphere 16 d13C Carbrock h 0 isotopic value of weathered carbonate 17 W o 1.2 relative contribution of organic matter weathering compared to silicate and carbonate 18 d13C OrgCont h -20 isotopic value of weathered continental organic matter 19 VS um 1 volume to surface area ratio of carbon fixing microbes − 20 u day 1 2 growth rate of photosynthesizing microbes 21 d13C Org2 h -10 The isotopic value of the anomalous organic matter added to the background flux 22 Vorg 2 mol 5 ∗ 1016 The total volume of the anomalous organic matter added to the background flux 23 k 0.0002 The rate constant with which the anomalous organic carbon mixes with the background organic carbon flux.

147 stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 0 0 0 1: SensitivityofOceanStratificationTimescale 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 4 3 2 1 0 -10 -20 -30 -40 5: SensitivityofRiverineCarbonFlux 3: SensitivityofHydrologicIntensity 2*10 10 5*10 δ δ δ 70,000 years 50,000 years 30,000 years 13 13 13 13 13 12 mol/year C C C 8 5 2 mol/year mol/year 148

stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 0 0 0 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 0 -10 -20 -30 -40 2: SensitivityofTransitionSmoothness 6: SensitivityofEarlyRiverinePFlux 4: SensitivityofWeatheringIntensity 10 10 10 δ δ δ 10,000 years 2,000 years 100 years 13 13 13 10 9 8 mol/year mol/year C C C mol/year 8 5 2 Figure 4.8.5 (following page): 7) Shows the modeled δ13C output of carbonate and organic matter as a function of the later riverine P flux. This represents the flux of phosphorus after the time that the ocean was stratified. Increased weathering also corresponds with increased CO2 drawdown. 8) Shows the modeled δ13C output of carbonate and organic matter as a function of the organic matter remineralization rate in the deep ocean. The model results are indistinguishable because this variable only controls remineralization if there is excess oxygen in the deep ocean. In each model run the oxygen concentration reaches 0 and as such there is no modeled change. 9) Shows the modeled δ13C output of carbonate and organic matter as a function of the relative amount of carbonate weathering compared to total weathering (including silicate fraction) during the stratified period. The sustained change 13 in δ Corg is due to the increased atmospheric CO2 drawdown from increased weathering of silicates. 10) Shows the modeled δ13C output of carbonate and organic matter as a function of the relative amount of carbonate weathering compared to total weathering (including silicate fraction) after the stratified 13 period. The sustained change in δ Corg is due to the increased atmospheric CO2 drawdown from increased weathering of silicates. 11) Shows the modeled δ13C output of carbonate and organic matter as a function of the initial at- 13 mospheric CO2. 12) Shows the modeled δ C output of carbonate and organic matter as a function of the initial atmospheric O2. There is negligible isotopic effect because O2 only influences organic matter remineralization in the deep ocean, and in all cases the organic matter delivery to the deep ocean is large enough to completely consume O2.

149 stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 9: SensitivityofEarlyCarbonateWeatheringFraction 0 0 0 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 0 -10 -20 -30 -40 11: SensitivityofInitialAtmosphericCO 7: SensitivityofLaterRiverinePFlux δ δ δ 100,000 ppm 32,000 ppm 12,000 ppm 13 13 13 10 4*10 10 C C C 0.8 0.5 0.2 11 10 10 2 150

stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10: SensitivityofLaterCarbonateWeatheringFraction 0 0 0 :Sniiiyo rai eieaiainRate 8: SensitivityofOrganicRemineralization 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 0 -10 -20 -30 -40 12: SensitivityofInitialAtmosphericCO δ δ δ 235,000 ppm 235 ppm 0.235 ppm 13 13 13 0.999 0.95 0.5 C C C 0.8 0.5 0.2 2 Figure 4.8.6 (following page): 13) Shows the modeled δ13C output of carbonate and organic matter as a function of the mineral phase of carbonate that is precipitated, calcite vs aragonite. 14) Shows the modeled δ13C output of carbonate and organic matter as a function of the supersaturation of carbonate required before precipitation. 15) Shows the modeled δ13C output of carbonate and organic matter as a function of the isotopic value of the CO2 emitted from volcanoes. 16) Shows the modeled δ13C output of carbonate and organic matter as a function of the isotopic value of the carbonate weathered from the continents. 17) Shows the modeled δ13C output of carbonate and organic matter as a function of the relative amount of organic matter weathered from the continents. The values provided here are multipliers of the mineral (carbonate and silicate) weathering flux. 18) Shows the modeled δ13C output of carbonate and organic matter as a function of the δ13C of weathered organic matter.

151 stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 0 0 0 17: SensitivityofWeatheredContinentalOM 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 0 -10 -20 -30 -40 13: SensitivityofCalcitevsAragonite 15: SensitivityofVolcanic δ δ δ 13 13 13 Aragonite Calcite C C C 1.3 1.2 1.1 -3 -5 -7 δ 13 C 152

153 stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 100 -10 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 0 0 0 19: SensitivityofV/SRatioPrimaryProducers 23: SensitivityofOCMixingRateConstant 4 3 2 1 0 -10 -20 -30 -40 0 -10 -20 -30 -40 0 -10 -20 -30 -40 21: Sensitivityof δ δ δ δ 13 13 13 13 0.0004 0.0002 0.0001 -20 -10 0 C ofAnomalousOC C C C 2 1 0.5 o / oo o o / / oo oo 154

stratigraphic position above cap carbonate (m) stratigraphic position above cap carbonate (m) 100 100 -10 -10 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 0 0 20: SensitivityofGrowthRatePrimaryProducers 22: SensitivityofTotalVolumeAnomalousOC 4 3 2 1 0 -10 -20 -30 -40 4 3 2 1 0 -10 -20 -30 -40 δ δ 5*10 5*10 5*10 13 13 4 day 2 day 1 day C C 17 16 15 mol mol mol -1 -1 -1 5 Conclusion and future directions

155 In Chapter 2 we address a longstanding question in isotope geochemistry. What is the primary cause of the secular evolution of chert δ18O through time? The three hypotheses, high-T precipitation, evolving ocean δ18O composition, and diagenesis, cannot be distinguished with solely a more detailed chert δ18O ′ record. To resolve this problem, our innovation was to measure Δ 17O of chert in ′ addition to δ18O. Whereas the chert δ18O fractionations are identical, chert Δ 17O fractionates slightly differently in response to each of the three primary mechanisms. ′ The chert Δ 17O data set, presented here, is far more complete than any previous study that has reported comparable values. Most only measured the ′ Δ 17O of four to five samples [119, 152, 176]. The data set was sufficiently large that we could develop a Monte Carlo resampling model with which we simultaneously tested all the factors that likely influenced the final chert oxygen isotope composition. From this model we determined that diagenesis is ′ necessary to produce the observed Δ 17O compositions, and that the degree of diagenetic alteration correlated directly to the measured δ18O composition. The data actively supports the hypothesis of diagenetic alteration, but conversely the data explicitly predicts no change in the chert precipitation temperature. Finally the data slightly favors the interpretation of stable marine δ18O compositions. If, as we argue, the oxygen isotopic composition of Precambrian chert carries a significant component of alteration, then the corresponding prediction is that phosphate and carbonate δ18O trends were generated via comparable diagentic alteration. The alternate two mechanisms are both global processes which, if

156 present, we would observe in all mineral records. There is no way, for example, that only phosphate minerals precipitated in high ocean temperatures while co-occurring carbonates and cherts experienced a temperate ocean. While these parallel isotope records show a shift in δ18O of comparable magnitude over the same time period, the oxygen-bearing phases are differently susceptible to alteration. To distinguish whether diagenesis indeed produces the observed isotope ′ composition, one possible approach is to separate and measure the Δ 17O of two adjacent minerals with different compositions. The two minerals would have experienced the same diagenetic history and, if oxygen exchange occurred, would be at equilibrium. The two minerals would still have different susceptibilities to alteration and should therefore become fractionated at different rates. Observing this resulting isotope difference would confirm the importance of diagenesis, otherwise it suggests that the mineral compositions are primary. This approach, however, faces some technical hurdles. In order for the two mineral phases to be initially at equilibrium they would likely have to be fine grained and interspersed with each other. It would then be difficult to separate the two mineral phases without resorting to a chemical method likely to impose a oxygen isotope fractionation on at least one mineral. Further, the two minerals that would be most promising for this study, carbonate and phosphate, are difficult to measure by laser fluorination153 [ ]. A better approach may be to measure chert O-isotope compositions along with iron oxide oxygen isotope compositions from banded iron formations. Iron oxide, while still more

157 challenging to measure than silica, is relatively straightforward to measure. Chapter 2 addresses a debate that has been unresolved for decades, namely has the δ18O composition of the ocean changed through geologic time [95, 107, 207]. ′ By measuring Δ 17O compositions of chert and interpreting them through a Monte Carlo model we contributed a seawater δ18O prediction to this debate. Our results, however, only slightly favored an ocean with an invariant O-isotope composition. An ocean with an evolving O-isotope composition is also broadly ′ compatible with our chert Δ 17O and δ18O data. Fortunately this is a question which could be resolved by making additional measurements. Old, Archean age, chert samples are best able to address this problem. Assuming the ocean δ18O evolution hypothesis is true, the Archean cherts would have initially precipitated in seawater offset by more than 15 h relative to modern values. In contrast, the Neoproterozoic cherts would have initially precipitated in seawater only offset 3-4 h relative to the modern. Taken to the logical extreme, chert precipitated in the modern ocean would have the same isotope composition whether or not the ocean δ18O evolution hypothesis is true, and therefore would have no bearing in resolving the central question. To maximize the specificity of our ocean δ18O prediction, we could focus future investigations on chert samples from the Archean cratons in South Africa and western Australia. Our current data set instead focuses on chert from relatively young Neoproterozoic age formations. To address the problem of seawater δ18O evolution we could also include other ′ emerging methodologies in conjunction with the Δ 17O data set. For example,

158 many groups have been developing the capacity to make precise measurements

of clumped isotopes in carbonate [13, 45, 79]. Δ47 measurements of carbonate, when made in conjunction with δ18O, provide a paleothermometer which does not require a priori knowledge of the water’s δ18O composition. Given this temperature constraint, it is then trivial to solve for the marine δ18O composition.

The weakness of this method, however, is that Δ47 is strongly susceptible to being reset at high temperature [180]. Finding carbonates that have remained sufficiently cool becomes increasingly difficult in older formations. Clumped carbonate measurements can produce a high resolution δ18O record for the Phanerozoic, but the data is unlikely to extend further. Importantly though, the Phanerozoic clumped carbonate reconstructions show no significant change in 18 h δ OH2O whereas the ocean evolution hypothesis predicts a change of 2-3 over the same interval. The clumped carbonate data then supports our conclusion of no marine δ18O evolution. Alternatively, if we could measure the δ18O composition of a mineral phase that does not respond strongly to changes in temperature, that isotope record could serve as a direct proxy for the ocean’s δ18O composition. Nir Galili at the Weizmann Institute of Science has recently started developing such a method using the δ18O of iron oxides, which he argues are not temperature sensitive. Using petrographic methods he identifies only oxides that show no evidence of diagenetic alteration and measures their O-isotope composition. The δ18O record does show a progressive decrease as a function of age which supports the ocean δ18O change hypothesis.

159 The ocean adopts its δ18O composition through interactions with the oceanic crust. While models (including my own) tend to treat the oceanic crust O-isotope reservoir as infinitely large due to its rapid recycling in the mantle, the oceanic crust does itself become fractionated in its interaction with seawater. Ophiolites are pieces of oceanic crust that were uplifted and preserved on the continents. They can therefore, if properly analyzed, serve as a proxy for the co-occurring ocean δ18O composition. Ben Johnston at UC Boulder has performed this analysis and his results suggest that the ocean may have become more depleted in δ18O through time, opposite of the trend suggested by the ocean change hypothesis. Including our results we now have four geochemical methods giving different solutions for the same problem. Clearly, despite this increased focus, the longstanding question regarding the possible evolution of the ocean’s δ18O composition is still far from being resolved. I cannot predict where the next breakthrough will come from, but at least for now there is room to continue ′ exploring the potential of chert Δ 17O. Increased measurements of Archean samples would narrow the marine δ18O evolution prediction. In chapter 3 we shifted our focus to the silica cycle in the modern ocean. Silica precipitation is presently dominated by diatom frustule formation. This process occurs near the surface with biological intervention from an ocean undersaturated in silica, whereas Precambrian cherts precipitated inorganically at or near the sediment water interface from supersaturated seawater. To control the environmental conditions, we grew the diatoms in culture, two species with two

160 growth rates at three different temperatures. Neither the variable species nor ′ changing growth rates produced a discernible δ18O or Δ 17O effect - only temperature variations produced a systematic trend. The diatom δ18O temperature relationship was comparable to previous studies ′ of temperature dependent O-isotope fractionation, but Δ 17O compositions were consistently more negative than the equilibrium predictions. This offset we attributed to a vital effect, possibly the result of kinetic influence from thelong chain polyamines and silaffins that provide a template for silica precipitation. If ′ this Δ 17O fractionation is in fact ubiquitous among diatomaceous silica samples, it may be used to distinguish organic from inorganically precipitated chert in the geologic record. One possible future direction to take this study is to investigate the precipitation mechanism in more detail. Both silaffins and long chain polyamines will catalyze silica precipitation in vitro. It would be informative to perform such ′ an experiment and to measure the δ18O and Δ 17O composition of the precipitated silica. We could then identify whether the organic matter contribution does in fact produce the observed vital effect. ′ Chapter 2 and chapter 3 both describe studies that measure Δ 17O in silica, but despite this, the results are not easily comparable. The Precambrian cherts precipitated at equilibrium while the diatom cultures precipitated with a significant kinetic isotope effect. Even if the diatom frustule samples were exposed to the same environmental and diagenetic conditions as the cherts, there is no guarantee that the same isotope composition would ultimately be produced.

161 ′ In order to develop the complete temporal story of Δ 17O in the marine silica cycle a series of intermediate studies may be warranted. First I would test how the diatom frustule samples age under controlled laboratory conditions. Would cultured diatom samples exposed to different temperature waters change their oxygen isotope composition? At the same rate? Is there a species dependence? Is it associated with silica dissolution? I would then validate the results of the ageing experiments performed in the laboratory ′ with data collected from natural sedimentary environments. The Δ 17O composition of the silica may vary with respect to depth in the sediment. Would the transformation from opal-A to opal-CT then microcrystaline quartz be ′ associated with a Δ 17O fractionation? Diatoms have only been the dominant silica precipitating organisms since the late Cretaceous, before that radiolarians prevailed, and before that siliceous sponges. Do these organisms precipitate silica differently than diatoms? With all this information one could tell the complete ′ story of Δ 17O in the silica cycle as it transitioned from inorganic to biological control. Of course, it is possible that biologically precipitated silica adopts an equilibrium O-isotope composition shortly after reaching the sediment, which would render many of these questions moot. In chapter 4 we address a specific highly unusual time period immediately after the Marinoan glaciation. δ13C from organic carbon and carbonate were measured in two sections and an anomalous organic carbon trend was observed. We built an isotope mass balance model for the conditions after the Marinoan, but could not produce the observed anomaly with any reasonable combination of

162 13 simulated environmental factors. The heavy δ Corg data can only sensibly be described by the addition of an exogenous, anomalous organic carbon flux to the sediment, but we do not uniquely identify the source of organic carbon [99]. In this study, because most contemporaneous sections do not share the organic carbon isotope anomaly, measuring additional sections is unlikely to reinforce or expand upon the existing narrative. The most fruitful approach may be to characterize the composition of the anomalous organic carbon more specifically. Ideally, some characteristic biomarker or the nitrogen isotope composition would shed light on the origin of the anomaly. ′ Measurements of Δ 17O in biogeochemistry have the potential to unlock environmental information, far beyond the applications described here. In this ′ thesis we constrained ourselves to Δ 17O measurements of chert, but there are dozens of other geologically relevant oxygen bearing species that could be measured. Even when just looking at chert in this work, one study provides a solution to a long standing problem in isotope geochemistry, the other study identifies a novel fractionation factor that could potentially be used as anew geochemical tracer. ′ The main factors limiting the widespread use of Δ 17O are 1) the relatively small signal and 2) the difficulty in making the measurement. The first problem can be addressed by measuring samples formed with both mass-dependent and mass-independent isotope fractionations. While in these studies we limited ourselves to exploring mass-dependent isotope fractionations, mass-independent fractionations tend to be larger and easier to measure. The second problem can

163 be solved with patience and perseverance. As long as the science is interesting, there will be people stubborn enough to pursue it.

164 References

[1] Andrea Abelmann, Rainer Gersonde, Gregor Knorr, Xu Zhang, Bernhard Chapligin, Edith Maier, Oliver Esper, Hans Friedrichsen, Gerrit Lohmann, Hanno Meyer, et al. The seasonal sea-ice zone in the glacial southern ocean as a carbon sink. Nature communications, 6:8136, 2015.

[2] Anne-Sofie C Ahm, Christian J Bjerrum, Clara L Blättler, Peter K Swart, and John A Higgins. Quantifying early marine diagenesis in shallow-water carbonate sediments. Geochimica et Cosmochimica Acta, 236:140–159, 2018.

[3] Insu Ahn, Jong Ik Lee, Minoru Kusakabe, and Byeon-Gak Choi. Oxygen isotope measurements of terrestrial silicates using a co 2-laser brf 5 fluorination technique and the slope of terrestrial fractionation line. Geosciences Journal, 16(1):7–16, 2012.

[4] JR Allan and RK Matthews. Isotope signatures associated with early meteoric diagenesis. Sedimentology, 29(6):797–817, 1982.

[5] Philip A Allen and Paul F Hoffman. Extreme winds and waves in the aftermath of a neoproterozoic glaciation. Nature, 433(7022):123–127, 2005.

[6] Ross P Anderson, Sean McMahon, Uyanga Bold, Francis A Macdonald, and Derek EG Briggs. Palaeobiology of the early ediacaran shuurgat formation, zavkhan terrane, south-western mongolia. Journal of Systematic Palaeontology, 15(11):947–968, 2017.

[7] Hannah L Bailey, Andrew CG Henderson, Hilary J Sloane, Andrea Snelling, Melanie J Leng, and Darrell S Kaufman. The effect of species on lacustrine δ18odiatom and its implications for palaeoenvironmental reconstructions. Journal of Quaternary Science, 29(4):393–400, 2014.

165 [8] Huiming Bao and Mark H Thiemens. Generation of o2 from baso4 using a co2- laser fluorination system for simultaneous analysis of δ18o and δ17o. Analytical chemistry, 72(17):4029–4032, 2000.

[9] Huiming Bao, JR Lyons, and Chuanming Zhou. Triple oxygen isotope evidence for elevated co 2 levels after a neoproterozoic glaciation. Nature, 453(7194):504, 2008.

[10] Huiming Bao, Xiaobin Cao, and Justin A Hayles. The confines of triple oxygen isotope exponents in elemental and complex mass-dependent processes. Geochimica et Cosmochimica Acta, 170:39–50, 2015.

[11] Richard W Battarbee, AW Mackay, DH Jewson, DB Ryves, and M Sturm. Differential dissolution of lake baikal diatoms: correction factors and implications for palaeoclimatic reconstruction. Global and Planetary Change, 46(1-4):75–86, 2005.

[12] Andrey Bekker, John F Slack, Noah Planavsky, Bryan Krapez, Axel Hofmann, Kurt O Konhauser, and Olivier J Rouxel. Iron formation: the sedimentary product of a complex interplay among mantle, tectonic, oceanic, and biospheric processes. Economic Geology, 105(3):467–508, 2010.

[13] Kristin D Bergmann, Seth Finnegan, John Eiler, and Woodward Fischer. Estimating cambro-ordovician seawater temperature and isotopic composition using carbonate clumped isotope paleothermometry on calcitic and phosphatic fossils.

[14] R Bertermann, N Kröger, and R Tacke. Solid-state 29 si mas nmr studies of diatoms: Structural characterization of biosilica deposits. Analytical and bioanalytical chemistry, 375(5):630–634, 2003.

[15] Robert R Bidigare, Arnim Fluegge, Katherine H Freeman, Kristi L Hanson, John M Hayes, David Hollander, John P Jasper, Linda L King, Edward A Laws, Jeffrey Milder, et al. Consistent fractionation of 13c in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochemical Cycles, 11(2):279–292, 1997.

[16] Jacob Bigeleisen and Maria Goeppert Mayer. Calculation of equilibrium constants for isotopic exchange reactions. The Journal of Chemical Physics, 15(5):261–267, 1947.

166 [17] Jacob Bigeleisen and Max Wolfsberg. Theoretical and experimental aspects of isotope effects in chemical kinetics. Advances in Chemical Physics, pages 15–76, 1957.

[18] Ruth E Blake, Sae Jung Chang, and Aivo Lepland. Phosphate oxygen isotopic evidence for a temperate and biologically active archaean ocean. Nature, 464(7291):1029, 2010.

[19] Uyanga Bold, Emily F Smith, Alan D Rooney, Samuel A Bowring, Robert Buchwaldt, Francis Ő Dudás, Jahandar Ramezani, James L Crowley, Daniel P Schrag, and Francis A Macdonald. Neoproterozoic stratigraphy of the zavkhan terrane of mongolia: The backbone for cryogenian and early ediacaran chemostratigraphic records. American Journal of Science, 316(1):1–63, 2016.

[20] AJ Boller, PJ Thomas, CM Cavanaugh, and KM Scott. Isotopic discrimination and kinetic parameters of r ubis co from the marine bloom-forming diatom, s keletonema costatum. Geobiology, 13(1):33–43, 2015.

[21] Mark E Brandriss, James R O’Neil, Mark B Edlund, and Eugene F Stoermer. Oxygen isotope fractionation between diatomaceous silica and water. Geochimica et Cosmochimica Acta, 62(7):1119–1125, 1998.

[22] Mark A Brzezinski, Robert J Olson, and Sallie W Chisholm. Silicon availability and cell-cycle progression in marine diatoms. Marine ecology progress series, pages 83–96, 1990.

[23] Adam J Bumby, Patrick G Eriksson, Octavian Catuneanu, David R Nelson, and Martin J Rigby. Meso-archaean and palaeo-proterozoic sedimentary sequence stratigraphy of the kaapvaal craton. Marine and Petroleum Geology, 33(1):92–116, 2012.

[24] Steffen Burkhardt, Ulf Riebesell, and Ingrid Zondervan. Effects of growth rate, co2 concentration, and cell size on the stable carbon isotope fractionation in marine phytoplankton. Geochimica et Cosmochimica Acta, 63(22):3729–3741, 1999.

[25] Steffen Burkhardt, Ulf Riebesell, and Ingrid Zondervan. Stable carbon isotope fractionation by marine phytoplankton in response to daylength,

167 growth rate, and co2 availability. Marine Ecology Progress Series, 184: 31–41, 1999.

[26] Xiaobin Cao and Yun Liu. Equilibrium mass-dependent fractionation relationships for triple oxygen isotopes. Geochimica et Cosmochimica Acta, 75(23):7435–7445, 2011.

[27] Thomas Chacko, David R Cole, and Juske Horita. Equilibrium oxygen, hydrogen and carbon isotope fractionation factors applicable to geologic systems. Reviews in mineralogy and geochemistry, 43(1):1–81, 2001.

[28] Ramananda Chakrabarti, Andrew H Knoll, Stein B Jacobsen, and Woodward W Fischer. Si isotope variability in proterozoic cherts. Geochimica et Cosmochimica Acta, 91:187–201, 2012.

[29] IPCC Climate Change et al. The physical science basis. Contribution of Working Group I to the fourth assessment report of the Intergovernmental Panel on Climate Change, 996, 2007.

[30] Bernhard Chapligin, Hanno Meyer, H Friedrichsen, A Marent, E Sohns, and H-W Hubberten. A high-performance, safer and semi-automated approach for the δ18o analysis of diatom silica and new methods for removing exchangeable oxygen. Rapid Communications in Mass Spectrometry, 24(17):2655–2664, 2010.

[31] Bernhard Chapligin, Melanie J Leng, Elizabeth Webb, Anne Alexandre, Justin P Dodd, Akira Ijiri, Andreas Lücke, Aldo Shemesh, Andrea Abelmann, Ulrike Herzschuh, et al. Inter-laboratory comparison of oxygen isotope compositions from biogenic silica. Geochimica et Cosmochimica Acta, 75(22):7242–7256, 2011.

[32] David Chew and Christopher Kirkland. The Chiquerío Formation, southern Peru. 2011.

[33] Robert N Clayton and Toshiko K Mayeda. The use of bromine pentafluoride in the extraction of oxygen from oxides and silicates for isotopic analysis. Geochimica et cosmochimica acta, 27(1):43–52, 1963.

[34] Robert N Clayton and Toshiko K Mayeda. Oxygen isotope studies of achondrites. Geochimica et Cosmochimica Acta, 60(11):1999–2017, 1996.

168 [35] Robert N Clayton, Lawrence Grossman, and Toshiko K Mayeda. A component of primitive nuclear composition in carbonaceous meteorites. Science, 182(4111):485–488, 1973.

[36] Phoebe A Cohen, Justin V Strauss, Alan D Rooney, Mukul Sharma, and Nicholas Tosca. Controlled hydroxyapatite biomineralization in an~ 810 million-year-old unicellular eukaryote. Science advances, 3(6):e1700095, 2017.

[37] Ralf Conrad, Peter Claus, and Peter Casper. Characterization of stable isotope fractionation during methane production in the sediment of a eutrophic lake, lake dagow, germany. Limnology and Oceanography, 54(2): 457–471, 2009.

[38] Benjamin R Cowie and David T Johnston. High-precision measurement and standard calibration of triple oxygen isotopic compositions (δ18o, δ′ 17o) of sulfate by f2 laser fluorination. Chemical Geology, 440:50–59, 2016.

[39] Grant M Cox, Galen P Halverson, Ross K Stevenson, Michelle Vokaty, André Poirier, Marcus Kunzmann, Zheng-Xiang Li, Steven W Denyszyn, Justin V Strauss, and Francis A Macdonald. Continental flood basalt weathering as a trigger for neoproterozoic snowball earth. Earth and Planetary Science Letters, 446:89–99, 2016.

[40] Harmon Craig. Isotopic standards for carbon and oxygen and correction factors for mass-spectrometric analysis of carbon dioxide. Geochimica et cosmochimica acta, 12(1-2):133–149, 1957.

[41] Julien Crespin, Florence Sylvestre, Anne Alexandre, Corinne Sonzogni, Christine Paillès, and Marie-Elodie Perga. Re-examination of the temperature-dependent relationship between δ 18 o diatoms and δ 18 o lake water and implications for paleoclimate inferences. Journal of Paleolimnology, 44(2):547–557, 2010.

[42] Robert E Criss et al. Principles of stable isotope distribution. Oxford University Press on Demand, 1999.

[43] Peter W Crockford, Justin A Hayles, Huiming Bao, Noah J Planavsky, Andrey Bekker, Philip W Fralick, Galen P Halverson, Thi Hao Bui, Yongbo

169 Peng, and Boswell A Wing. Triple oxygen isotope evidence for limited mid-proterozoic primary productivity. Nature, 559(7715):613, 2018.

[44] Huan Cui, Alan J Kaufman, Shuhai Xiao, Chuanming Zhou, and Xiao-Ming Liu. Was the ediacaran shuram excursion a globally synchronized early diagenetic event? insights from methane-derived authigenic carbonates in the uppermost doushantuo formation, south china. Chemical Geology, 450:59–80, 2017.

[45] Renata C Cummins, Seth Finnegan, David A Fike, John M Eiler, and Woodward W Fischer. Carbonate clumped isotope constraints on silurian ocean temperature and seawater δ18o. Geochimica et Cosmochimica Acta, 140:241–258, 2014.

[46] Nicolas Dauphas and Edwin A Schauble. Mass fractionation laws, mass-independent effects, and isotopic anomalies. Annual Review of Earth and Planetary Sciences, 44:709–783, 2016.

[47] Egon T Degens and Samuel Epstein. Relationship between o18/o16 ratios in coexisting carbonates, cherts, and diatomites. AAPG Bulletin, 46 (4):534–542, 1962.

[48] ET Degens, RRL Guillard, WM Sackett, and JA Hellebust. Metabolic fractionation of carbon isotopes in marine plankton—i. temperature and respiration experiments. In Deep Sea Research and Oceanographic Abstracts, volume 15, pages 1–9. Elsevier, 1968.

[49] Kate J Dennis and Daniel P Schrag. Clumped isotope thermometry of carbonatites as an indicator of diagenetic alteration. Geochimica et Cosmochimica Acta, 74(14):4110–4122, 2010.

[50] Louis A Derry. A burial diagenesis origin for the ediacaran shuram–wonoka carbon isotope anomaly. Earth and Planetary Science Letters, 294(1-2):152–162, 2010.

[51] Justin P Dodd and Zachary D Sharp. A laser fluorination method for oxygen isotope analysis of biogenic silica and a new oxygen isotope calibration of modern diatoms in freshwater environments. Geochimica et Cosmochimica Acta, 74(4):1381–1390, 2010.

170 [52] Justin P Dodd, Zachary D Sharp, Peter J Fawcett, Adrian J Brearley, and Francis M McCubbin. Rapid post-mortem maturation of diatom silica oxygen isotope values. Geochemistry, Geophysics, Geosystems, 13(9), 2012.

[53] Ryan W Drum and H Stuart Pankratz. Post mitotic fine structure of gomphonema parvulum. Journal of ultrastructure research, 10(3-4): 217–223, 1964.

[54] John M Eiler. “clumped-isotope” geochemistry—the study of naturally-occurring, multiply-substituted isotopologues. Earth and Planetary Science Letters, 262(3-4):309–327, 2007.

[55] Cesare Emiliani. Pleistocene temperatures. The Journal of Geology, 63(6): 538–578, 1955.

[56] Samuel Epstein, Ralph Buchsbaum, Heinz A Lowenstam, and Harold C Urey. Revised carbonate-water isotopic temperature scale. Geological Society of America Bulletin, 64(11):1315–1326, 1953.

[57] Douglas H Erwin, Marc Laflamme, Sarah M Tweedt, Erik A Sperling, Davide Pisani, and Kevin J Peterson. The cambrian conundrum: early divergence and later ecological success in the early history of animals. science, 334(6059):1091–1097, 2011.

[58] ES Falk, W Guo, AN Paukert, JM Matter, EM Mervine, and PB Kelemen. Controls on the stable isotope compositions of travertine from hyperalkaline springs in oman: Insights from clumped isotope measurements. Geochimica et Cosmochimica Acta, 192:1–28, 2016.

[59] Andrea Fanesi, John A Raven, and Mario Giordano. Growth rate affects the responses of the green alga t etraselmis suecica to external perturbations. Plant, cell & environment, 37(2):512–519, 2014.

[60] Graham D Farquhar, Marion H O’Leary, and Joe A Berry. On the relationship between carbon isotope discrimination and the intercellular carbon dioxide concentration in leaves. Functional Plant Biology, 9(2): 121–137, 1982.

[61] Graham D Farquhar, James R Ehleringer, and Kerry T Hubick. Carbon isotope discrimination and photosynthesis. Annual review of plant biology, 40(1):503–537, 1989.

171 [62] Georg Feulner. The faint young sun problem. Reviews of Geophysics, 50 (2), 2012.

[63] Woodward W Fischer and Andrew H Knoll. An iron shuttle for deepwater silica in late archean and early paleoproterozoic iron formation. Geological Society of America Bulletin, 121(1-2):222–235, 2009.

[64] E Font, A Nédélec, RIF Trindade, and C Moreau. Fast or slow melting of the marinoan snowball earth? the cap dolostone record. Palaeogeography, Palaeoclimatology, Palaeoecology, 295(1):215–225, 2010.

[65] Katherine H Freeman and JM Hayes. Fractionation of carbon isotopes by phytoplankton and estimates of ancient co2 levels. Global Biogeochemical Cycles, 6(2):185–198, 1992.

[66] Herncin E Garcia and Louis I Gordon. Oxygen solubility in seawater: Better fitting equations. Limnology and oceanography, 37(6):1307–1312, 1992.

[67] Aıcha Gendron-Badou, Thibaud Coradin, Jocelyne Maquet, François Fröhlich, and Jacques Livage. Spectroscopic characterization of biogenic silica. Journal of Non-Crystalline Solids, 316(2-3):331–337, 2003.

[68] Prosenjit Ghosh, Jess Adkins, Hagit Affek, Brian Balta, Weifu Guo, Edwin A Schauble, Dan Schrag, and John M Eiler. 13c–18o bonds in carbonate minerals: a new kind of paleothermometer. Geochimica et Cosmochimica Acta, 70(6):1439–1456, 2006.

[69] C Gröger, M Sumper, and E Brunner. Silicon uptake and metabolism of the marine diatom thalassiosira pseudonana: solid-state 29si nmr and fluorescence microscopic studies. Journal of structural biology, 161(1): 55–63, 2008.

[70] Galen P Halverson, Paul F Hoffman, Daniel P Schrag, and Alan J Kaufman. A major perturbation of the carbon cycle before the ghaub glaciation (neoproterozoic) in namibia: Prelude to snowball earth? Geochemistry, Geophysics, Geosystems, 3(6):1–24, 2002.

[71] Galen P Halverson, Paul F Hoffman, Daniel P Schrag, Adam C Maloof, and A Hugh N Rice. Toward a neoproterozoic composite carbon-isotope record. Geological Society of America Bulletin, 117(9-10):1181–1207, 2005.

172 [72] Galen P Halverson, Francis A Macdonald, Justin V Strauss, Emily F Smith, Grant M Cox, and Lucie Hubert-Théou. Updated definition and correlation of the lower fifteenmile group in the central and eastern ogilvie mountains. Yukon Exploration and Geology, pages 75–90, 2011.

[73] Hilairy E Hartnett, Richard G Keil, John I Hedges, and Allan H Devol. Influence of oxygen exposure time on organic carbon preservation in continental margin sediments. Nature, 391(6667):572–575, 1998.

[74] JM Hayes. Factors controlling 13c contents of sedimentary organic compounds: Principles and evidence. Marine Geology, 113(1-2): 111–125, 1993.

[75] JM Hayes, KW Wedeking, and IR Kaplan. Precambrian organic geochemistry-preservation of the record. 1983.

[76] John M Hayes. Fractionation of carbon and hydrogen isotopes in biosynthetic processes. Reviews in mineralogy and geochemistry, 43(1): 225–277, 2001.

[77] John M Hayes, Harald Strauss, and Alan J Kaufman. The abundance of 13c in marine organic matter and isotopic fractionation in the global biogeochemical cycle of carbon during the past 800 ma. Chemical Geology, 161(1):103–125, 1999.

[78] Justin Hayles, Caihong Gao, Xiaobin Cao, Yun Liu, and Huiming Bao. Theoretical calibration of the triple oxygen isotope thermometer. Geochimica et Cosmochimica Acta, 235:237–245, 2018.

[79] Gregory A Henkes, Benjamin H Passey, Ethan L Grossman, Brock J Shenton, Thomas E Yancey, and Alberto Pérez-Huerta. Temperature evolution and the oxygen isotope composition of phanerozoic oceans from carbonate clumped isotope thermometry. Earth and Planetary Science Letters, 490:40–50, 2018.

[80] JA Higgins, CL Blättler, EA Lundstrom, DP Santiago-Ramos, AA Akhtar, AS Crüger Ahm, O Bialik, C Holmden, H Bradbury, ST Murray, et al. Mineralogy, early marine diagenesis, and the chemistry of shallow-water carbonate sediments. Geochimica et Cosmochimica Acta, 220:512–534, 2018.

173 [81] John A Higgins and Daniel P Schrag. Aftermath of a snowball earth. Geochemistry, Geophysics, Geosystems, 4(3), 2003.

[82] M Hildebrand, K Dahlin, and BE Volcani. Characterization of a silicon transporter gene family in cylindrotheca fusiformis: sequences, expression analysis, and identification of homologs in other diatoms. Molecular and General Genetics MGG, 260(5):480–486, 1998.

[83] Mark Hildebrand, Benjamin E Volcani, Walter Gassmann, and Julian I Schroeder. A gene family of silicon transporters. Nature, 385(6618):688, 1997.

[84] Paul F Hoffman and Daniel P Schrag. The snowball earth hypothesis: testing the limits of global change. Terra nova, 14(3):129–155, 2002.

[85] Paul F Hoffman, Alan J Kaufman, Galen P Halverson, and Daniel P Schrag. A neoproterozoic snowball earth. Science, 281(5381):1342–1346, 1998.

[86] Paul F Hoffman, Francis A Macdonald, and Galen P Halverson. Chemical sediments associated with neoproterozoic glaciation: iron formation, cap carbonate, barite and phosphorite. Geological Society, London, Memoirs, 36 (1):67–80, 2011.

[87] Paul F Hoffman, Dorian S Abbot, Yosef Ashkenazy, Douglas I Benn, Jochen J Brocks, Phoebe A Cohen, Grant M Cox, Jessica R Creveling, Yannick Donnadieu, Douglas H Erwin, et al. Snowball earth climate dynamics and cryogenian geology-geobiology. Science Advances, 3(11): e1600983, 2017.

[88] PF Hoffman. Cryoconite pans on snowball earth: supraglacial oases for cryogenian eukaryotes? Geobiology, 14(6):531–542, 2016.

[89] A Hofmann, R Bolhar, B Orberger, and F Foucher. Cherts of the barberton greenstone belt, south africa: Petrology and trace-element geochemistry of 3.5 to 3.3 ga old silicified volcaniclastic sediments. South African Journal of Geology, 116(2):297–322, 2013.

[90] Marika M Holland and Cecilia M Bitz. Polar amplification of climate change in coupled models. Climate Dynamics, 21(3-4):221–232, 2003.

174 [91] Eran Hood, Jason Fellman, Robert GM Spencer, Peter J Hernes, Rick Edwards, David D’Amore, and Durelle Scott. Glaciers as a source of ancient and labile organic matter to the marine environment. Nature, 462 (7276):1044, 2009. [92] MT Hren, MM Tice, and CP Chamberlain. Oxygen and hydrogen isotope evidence for a temperate climate 3.42 billion years ago. Nature, 462(7270):205, 2009. [93] Kang-Jun Huang, Fang-Zhen Teng, Bing Shen, Shuhai Xiao, Xianguo Lang, Hao-Ran Ma, Yong Fu, and Yongbo Peng. Episode of intense chemical weathering during the termination of the 635 ma marinoan glaciation. Proceedings of the National Academy of Sciences, page 201607712, 2016. [94] Matthew T Hurtgen, Michael A Arthur, Neil S Suits, and Alan J Kaufman. The sulfur isotopic composition of neoproterozoic seawater sulfate: implications for a snowball earth? Earth and Planetary Science Letters, 203 (1):413–429, 2002. [95] Jasmine BD Jaffrés, Graham A Shields, and Klaus Wallmann. The oxygen isotope evolution of seawater: A critical review of a long-standing controversy and an improved geological water cycle model for the past 3.4 billion years. Earth-Science Reviews, 83(1-2):83–122, 2007. [96] Ph Jean-Baptiste, JL Charlou, and M Stievenard. Oxygen isotope study of mid-ocean ridge hydrothermal fluids: Implication for the oxygen-18 budget of the oceans. Geochimica et Cosmochimica Acta, 61(13): 2669–2677, 1997. [97] Ganqing Jiang, Xinqiang Wang, Xiaoying Shi, Shihong Zhang, Shuhai Xiao, and Jin Dong. Organic carbon isotope constraints on the dissolved organic carbon (doc) reservoir at the cryogenian–ediacaran transition. Earth and Planetary Science Letters, 299(1-2):159–168, 2010. [98] David T Johnston, James Farquhar, and Donald E Canfield. Sulfur isotope insights into microbial sulfate reduction: when microbes meet models. Geochimica et Cosmochimica Acta, 71(16):3929–3947, 2007. [99] David T Johnston, Francis Alexander Macdonald, BC Gill, PF Hoffman, and Daniel P Schrag. Uncovering the neoproterozoic carbon cycle. Nature, 483(7389):320–323, 2012.

175 [100] DT Johnston, SW Poulton, NJ Tosca, T O’brien, GP Halverson, DP Schrag, and FA Macdonald. Searching for an oxygenation event in the fossiliferous ediacaran of northwestern canada. Chemical Geology, 362: 273–286, 2013.

[101] Simone A Kasemann, Chris J Hawkesworth, Anthony R Prave, Anthony E Fallick, and Paul N Pearson. Boron and calcium isotope composition in neoproterozoic carbonate rocks from namibia: evidence for extreme environmental change. Earth and Planetary Science Letters, 231(1):73–86, 2005.

[102] James F Kasting, M Tazewell Howard, Klaus Wallmann, Ján Veizer, Graham Shields, and Jasmine Jaffrés. Paleoclimates, ocean depth, and the oxygen isotopic composition of seawater. Earth and Planetary Science Letters, 252(1-2):82–93, 2006.

[103] Miriam Kastner, JB Keene, and JM Gieskes. Diagenesis of siliceous oozes—i. chemical controls on the rate of opal-a to opal-ct transformation—an experimental study. Geochimica et Cosmochimica Acta, 41(8):1041–1059, 1977.

[104] Sang-Tae Kim and James R O’Neil. Equilibrium and nonequilibrium oxygen isotope effects in synthetic carbonates. Geochimica et cosmochimica acta, 61(16):3461–3475, 1997.

[105] Shoichi Kiyokawa, Takashi Ito, Minoru Ikehara, and Fumio Kitajima. Middle archean volcano-hydrothermal sequence: bacterial microfossil-bearing 3.2 ga dixon island formation, coastal pilbara terrane, australia. Geological Society of America Bulletin, 118(1-2):3–22, 2006.

[106] L Paul Knauth. Temperature and salinity history of the precambrian ocean: implications for the course of microbial evolution. pages 53–69, 2005.

[107] L Paul Knauth and Samuel Epstein. Hydrogen and oxygen isotope ratios in nodular and bedded cherts. Geochimica et Cosmochimica Acta, 40(9): 1095–1108, 1976.

[108] AH Knoll. Exceptional preservation of photosynthetic organisms in silicified carbonates and silicified peats. Philosophical Transactions of the Royal Society of London. B, Biological Sciences, 311(1148):111–122, 1985.

176 [109] Andrew H Knoll. Biomineralization and evolutionary history. Reviews in mineralogy and geochemistry, 54(1):329–356, 2003.

[110] Andrew H Knoll and Bruce Simonson. Early proterozoic microfossils and penecontemporaneous quartz cementation in the sokoman iron formation, canada. Science, 211(4481):478–480, 1981.

[111] Nils Kröger, Rainer Deutzmann, Christian Bergsdorf, and Manfred Sumper. Species-specific polyamines from diatoms control silica morphology. Proceedings of the National Academy of Sciences, 97(26): 14133–14138, 2000.

[112] Andrew A Lacis, Gavin A Schmidt, David Rind, and Reto A Ruedy. Atmospheric co2: Principal control knob governing earth’s temperature. Science, 330(6002):356–359, 2010.

[113] Edward A Laws, Brian N Popp, Robert R Bidigare, Mahlon C Kennicutt, and Stephen A Macko. Dependence of phytoplankton carbon isotopic composition on growth rate and [co2) aq: Theoretical considerations and experimental results. Geochimica et cosmochimica acta, 59(6):1131–1138, 1995.

[114] Edward A Laws, Robert R Bidigare, and Brian N Popp. Effect of growth rate and co2 concentration on carbon isotopic fractionation by the marine diatom phaeodactylum tricornutum. Limnology and Oceanography, 42(7): 1552–1560, 1997.

[115] Guillaume Le Hir, Gilles Ramstein, Yannick Donnadieu, and Yves Goddéris. Scenario for the evolution of atmospheric pco2 during a snowball earth. Geology, 36(1):47–50, 2008.

[116] Anne Juillet Leclerc and Laurent Labeyrie. Temperature dependence of the oxygen isotopic fractionation between diatom silica and water. Earth and Planetary Science Letters, 84(1):69–74, 1987.

[117] Moritz F Lehmann, Stefano M Bernasconi, Alberto Barbieri, and Judith A McKenzie. Preservation of organic matter and alteration of its carbon and nitrogen isotope composition during simulated and in situ early sedimentary diagenesis. Geochimica et Cosmochimica Acta, 66(20): 3573–3584, 2002.

177 [118] Melanie J Leng and Philip A Barker. A review of the oxygen isotope composition of lacustrine diatom silica for palaeoclimate reconstruction. Earth-Science Reviews, 75(1-4):5–27, 2006.

[119] Naomi E Levin, Timothy D Raub, Nicolas Dauphas, and John M Eiler. Triple oxygen isotope variations in sedimentary rocks. Geochimica et Cosmochimica Acta, 139:173–189, 2014.

[120] Shuning Li, Naomi E Levin, and Lesley A Chesson. Continental scale variation in 17o-excess of meteoric waters in the united states. Geochimica et Cosmochimica Acta, 164:110–126, 2015.

[121] Zheng-Xiang Li, SV Bogdanova, AS Collins, Anthony Davidson, Bert De Waele, RE Ernst, ICW Fitzsimons, RA Fuck, DP Gladkochub, J Jacobs, et al. Assembly, configuration, and break-up history of rodinia: a synthesis. Precambrian research, 160(1):179–210, 2008.

[122] Chao Liu, Zhengrong Wang, Timothy D Raub, Francis A Macdonald, and David AD Evans. Neoproterozoic cap-dolostone deposition in stratified glacial meltwater plume. Earth and Planetary Science Letters, 404:22–32, 2014.

[123] D Lorian, J Erez, and B Lazar. Stable carbon isotopes in the reef ecosystem of the gulf of eliat—red sea. In Proceedings of 7th international coral reef Symposium. University of Guam, Mangilao, Guam, volume 364, pages 243–247, 1992.

[124] Boaz Luz and Eugeni Barkan. Variations of 17o/16o and 18o/16o in meteoric waters. Geochimica et Cosmochimica Acta, 74(22):6276–6286, 2010.

[125] Boaz Luz, Eugeni Barkan, Michael L Bender, Mark H Thiemens, and Kristie A Boering. Triple-isotope composition of atmospheric oxygen as a tracer of biosphere productivity. Nature, 400(6744):547, 1999.

[126] Francis A Macdonald. The tsagaan oloom formation, southwestern mongolia. Geological Society, London, Memoirs, 36(1):331–337, 2011.

[127] Francis A Macdonald and Phoebe A Cohen. The tatonduk inlier, alaska–yukon border. Geological Society, London, Memoirs, 36(1): 389–396, 2011.

178 [128] Francis A. Macdonald and David S. Jones. Chapter 30 the khubsugul group, northern mongolia. 36(1):339–345, 2011. doi: 10.1144/M36.30.

[129] Francis A Macdonald, David S Jones, and Daniel P Schrag. Stratigraphic and tectonic implications of a newly discovered glacial diamictite–cap carbonate couplet in southwestern mongolia. Geology, 37(2):123–126, 2009.

[130] Francis A Macdonald, Phoebe A Cohen, Francis Ő Dudás, and Daniel P Schrag. Early neoproterozoic scale microfossils in the lower tindir group of alaska and the yukon territory. Geology, 38(2):143–146, 2010.

[131] Francis A Macdonald, Justin V Strauss, Erik A Sperling, Galen P Halverson, Guy M Narbonne, David T Johnston, Marcus Kunzmann, Daniel P Schrag, and John A Higgins. The stratigraphic relationship between the shuram carbon isotope excursion, the oxygenation of neoproterozoic oceans, and the first appearance of the ediacara biota and bilaterian trace fossils in northwestern canada. Chemical Geology, 362: 250–272, 2013.

[132] Fred T Mackenzie and Robert M Garrels. Chemical mass balance between rivers and oceans. American Journal of Science, 264(7):507–525, 1966.

[133] Edith Maier, Bernhard Chapligin, Andrea Abelmann, Rainer Gersonde, Oliver Esper, Jian Ren, Hans Friedrichsen, Hanno Meyer, and Ralf Tiedemann. Combined oxygen and silicon isotope analysis of diatom silica from a deglacial subarctic pacific record. Journal of Quaternary Science, 28(6):571–581, 2013.

[134] Robert G Maliva, Andrew H Knoll, and Raymond Siever. Secular change in chert distribution: a reflection of evolving biological participation in the silica cycle. Palaios, pages 519–532, 1989.

[135] Robert G Maliva, Andrew H Knoll, and Bruce M Simonson. Secular change in the precambrian silica cycle: insights from chert petrology. Geological Society of America Bulletin, 117(7-8):835–845, 2005.

[136] Johanna Marin, Marc Chaussidon, and François Robert. Microscale oxygen isotope variations in 1.9 ga gunflint cherts: assessments of

179 diagenesis effects and implications for oceanic paleotemperature reconstructions. Geochimica et Cosmochimica Acta, 74(1):116–130, 2010. [137] J Marin-Carbonne, F Robert, and M Chaussidon. The silicon and oxygen isotope compositions of precambrian cherts: A record of oceanic paleo-temperatures? Precambrian Research, 247:223–234, 2014. [138] Charles R Marshall. Explaining the cambrian “explosion” of animals. Annu. Rev. Earth Planet. Sci., 34:355–384, 2006. [139] Ronald K Matheney and L Paul Knauth. Oxygen-isotope fractionation between marine biogenic silica and seawater. Geochimica et Cosmochimica Acta, 53(12):3207–3214, 1989. [140] Ted McConnaughey. 13c and 18o isotopic disequilibrium in biological carbonates: I. patterns. Geochimica et Cosmochimica Acta, 53(1):151–162, 1989. [141] Kathleen A McFadden, Jing Huang, Xuelei Chu, Ganqing Jiang, Alan J Kaufman, Chuanming Zhou, Xunlai Yuan, and Shuhai Xiao. Pulsed oxidation and biological evolution in the ediacaran doushantuo formation. Proceedings of the National Academy of Sciences, 105(9): 3197–3202, 2008. [142] CAPE-Last Interglacial Project Members. Last interglacial arctic warmth confirms polar amplification of climate change. Quaternary Science Reviews, 25(13-14):1383–1400, 2006. [143] Martin F Miller. Isotopic fractionation and the quantification of 17o anomalies in the oxygen three-isotope system: an appraisal and geochemical significance. Geochimica et Cosmochimica Acta, 66(11): 1881–1889, 2002. [144] Frank J Millero. Thermodynamics of the carbon dioxide system in the oceans. Geochimica et Cosmochimica Acta, 59(4):661–677, 1995. [145] WG Mook. 13c in atmospheric co2. Netherlands Journal of Sea Research, 20(2-3):211–223, 1986. [146] WG Mook, JC Bommerson, and WH Staverman. Carbon isotope fractionation between dissolved bicarbonate and gaseous carbon dioxide. Earth and Planetary Science Letters, 22(2):169–176, 1974.

180 [147] Robert Moschen, Andreas Lücke, and Gerhard H Schleser. Sensitivity of biogenic silica oxygen isotopes to changes in surface water temperature and palaeoclimatology. Geophysical Research Letters, 32(7), 2005.

[148] Robert Moschen, Andreas Lücke, Jörn Parplies, Ulrich Radtke, and Gerhard H Schleser. Transfer and early diagenesis of biogenic silica oxygen isotope signals during settling and sedimentation of diatoms in a temperate freshwater lake (lake holzmaar, germany). Geochimica et cosmochimica acta, 70(17):4367–4379, 2006.

[149] Alfonso Mucci. The solubility of calcite and aragonite in seawater at various salinities, temperatures, and one atmosphere total pressure. American Journal of Science, 283(7):780–799, 1983.

[150] Karlis Muehlenbachs and RN Clayton. Oxygen isotope composition of the oceanic crust and its bearing on seawater. Journal of Geophysical Research, 81(23):4365–4369, 1976.

[151] Amanda M Oehlert and Peter K Swart. Interpreting carbonate and organic carbon isotope covariance in the sedimentary record. Nature Communications, 5:4672, 2014.

[152] Andreas Pack and Daniel Herwartz. The triple oxygen isotope composition of the earth mantle and understanding δo17 variations in terrestrial rocks and minerals. Earth and Planetary Science Letters, 390: 138–145, 2014.

[153] Benjamin H Passey, Huanting Hu, Haoyuan Ji, Shaena Montanari, Shuning Li, Gregory A Henkes, and Naomi E Levin. Triple oxygen isotopes in biogenic and sedimentary carbonates. Geochimica et Cosmochimica Acta, 141:1–25, 2014.

[154] EC Perry, FC Tan, and GB Morey. Geology and stable isotope geochemistry of the biwabik iron formation, northern minnesota. Economic Geology, 68(7):1110–1125, 1973.

[155] Raymond T Pierrehumbert. Principles of planetary climate. Cambridge University Press, 2010.

[156] Brian N Popp, Edward A Laws, Robert R Bidigare, John E Dore, Kristi L Hanson, and Stuart G Wakeham. Effect of phytoplankton cell geometry

181 on carbon isotopic fractionation. Geochimica et cosmochimica acta, 62(1): 69–77, 1998.

[157] Sara B Pruss, Tanja Bosak, Francis A Macdonald, Marie McLane, and Paul F Hoffman. Microbial facies in a sturtian cap carbonate, the rasthof formation, otavi group, northern namibia. Precambrian Research, 181 (1-4):187–198, 2010.

[158] Judy P Pu, Samuel A Bowring, Jahandar Ramezani, Paul Myrow, Timothy D Raub, Ed Landing, Andrea Mills, Eben Hodgin, and Francis A Macdonald. Dodging snowballs: Geochronology of the gaskiers glaciation and the first appearance of the ediacaran biota. Geology, 44 (11):955–958, 2016.

[159] Peir K Pufahl and Philip W Fralick. Depositional controls on palaeoproterozoic iron formation accumulation, gogebic range, lake superior region, usa. Sedimentology, 51(4):791–808, 2004.

[160] Birger Rasmussen, Janet R Muhling, and Woodward W Fischer. Evidence from laminated chert in banded iron formations for deposition by gravitational settling of iron-silicate muds. Geology, 47(2):167–170, 2019.

[161] Pascal Richet, Yan Bottinga, and Marc Javoy. A review of hydrogen, carbon, nitrogen, oxygen, sulphur, and chlorine stable isotope fractionation among gaseous molecules. Annual Review of Earth and Planetary Sciences, 5(1):65–110, 1977.

[162] Ulf Riebesell, Steffen Burkhardt, Anke Dauelsberg, and Bernd Kroon. Carbon isotope fractionation by a marine diatom: dependence on the growth-rate-limiting resource. Marine Ecology Progress Series, 193: 295–303, 2000.

[163] François Robert and Marc Chaussidon. A palaeotemperature curve for the precambrian oceans based on silicon isotopes in cherts. Nature, 443 (7114):969, 2006.

[164] Christopher S Romanek, Ethan L Grossman, and John W Morse. Carbon isotopic fractionation in synthetic aragonite and calcite: effects of temperature and precipitation rate. Geochimica et Cosmochimica Acta, 56 (1):419–430, 1992.

182 [165] Daniel H Rothman, John M Hayes, and Roger E Summons. Dynamics of the neoproterozoic carbon cycle. Proceedings of the National Academy of Sciences, 100(14):8124–8129, 2003.

[166] Rabindra N Roy, Lakshimi N Roy, Kathleen M Vogel, C Porter-Moore, Tara Pearson, Catherine E Good, Frank J Millero, and Douglas M Campbell. The dissociation constants of carbonic acid in seawater at salinities 5 to 45 and temperatures 0 to 45 c. Marine Chemistry, 44(2-4): 249–267, 1993.

[167] P Sansjofre, M Ader, RIF Trindade, M Elie, J Lyons, P Cartigny, and ACR Nogueira. A carbon isotope challenge to the snowball earth. Nature, 478 (7367):93–96, 2011.

[168] Manfred Schidlowski, Udo Matzigkeit, and Wolfgang E Krumbein. Superheavy organic carbon from hypersaline microbial mats. Naturwissenschaften, 71(6):303–308, 1984.

[169] Mark Schmidt, Reiner Botz, D Rickert, Gerhard Bohrmann, SR Hall, and S Mann. Oxygen isotopes of marine diatoms and relations to opal-a maturation. Geochimica et Cosmochimica Acta, 65(2):201–211, 2001.

[170] Martin Schobben, Clemens Vinzenz Ullmann, Lucyna Leda, Dieter Korn, Ulrich Struck, Wolf Uwe Reimold, Abbas Ghaderi, Thomas J Algeo, and Christoph Korte. Discerning primary versus diagenetic signals in carbonate carbon and oxygen isotope records: An example from the permian–triassic boundary of iran. Chemical Geology, 422:94–107, 2016.

[171] Daniel P Schrag, Donald J DePaolo, and Frank M Richter. Reconstructing past sea surface temperatures: Correcting for diagenesis of bulk marine carbonate. Geochimica et Cosmochimica Acta, 59(11):2265–2278, 1995.

[172] Daniel P Schrag, John A Higgins, Francis A Macdonald, and David T Johnston. Authigenic carbonate and the history of the global carbon cycle. science, 339(6119):540–543, 2013.

[173] Vladimir N Sergeev, Andrew H Knoll, and John Peter Grotzinger. Paleobiology of the mesoproterozoic billyakh group, anabar uplift, northern siberia. Journal of paleontology, 69(S39):1–37, 1995.

183 [174] VN Sergeev, AH Knoll, and P Yu Petrov. Paleobiology of the mesoproterozoic-neoproterozoic transition: the sukhaya tunguska formation, turukhansk uplift, siberia. Precambrian Research, 85(3-4): 201–239, 1997. [175] Nicholas J Shackleton. Paleotemperature history of the cenozoic and the initiation of antarctic glaciation: oxygen and carbon isotope analyses in dsdp sites 277,279, and 281. Initial Reports of Deep Sea Drilling Project, 29: 743–756, 1975. [176] ZD Sharp, JA Gibbons, O Maltsev, V Atudorei, A Pack, S Sengupta, EL Shock, and LP Knauth. A calibration of the triple oxygen isotope fractionation in the sio 2–h 2 o system and applications to natural samples. Geochimica et Cosmochimica Acta, 186:105–119, 2016. [177] A Shemesh, CD Charles, and RG Fairbanks. Oxygen isotopes in biogenic silica: global changes in ocean temperature and isotopic composition. Science, 256(5062):1434–1436, 1992. [178] Aldo Shemesh, Lloyd H Burckle, and James D Hays. Late pleistocene oxygen isotope records of biogenic silica from the atlantic sector of the southern ocean. Paleoceanography, 10(2):179–196, 1995. [179] Bing Shen, Shuhai Xiao, Alan J Kaufman, Huiming Bao, Chuanming Zhou, and Haifeng Wang. Stratification and mixing of a post-glacial neoproterozoic ocean: evidence from carbon and sulfur isotopes in a cap dolostone from northwest china. Earth and Planetary Science Letters, 265 (1):209–228, 2008. [180] Brock J Shenton, Ethan L Grossman, Benjamin H Passey, Gregory A Henkes, Thomas P Becker, Juan Carlos Laya, Alberto Perez-Huerta, Stephen P Becker, and Michael Lawson. Clumped isotope thermometry in deeply buried sedimentary carbonates: The effects of bond reordering and recrystallization. GSA Bulletin, 127(7-8):1036–1051, 2015. [181] Graham A Shields. Neoproterozoic cap carbonates: a critical appraisal of existing models and the plumeworld hypothesis. Terra Nova, 17(4): 299–310, 2005. [182] Raymond Siever. The silica budget in the sedimentary cycle. American Mineralogist: Journal of Earth and Planetary Materials, 42(11-12): 821–841, 1957.

184 [183] Raymond Siever. The silica cycle in the precambrian. Geochimica et Cosmochimica Acta, 56(8):3265–3272, 1992. [184] Bruce M Simonson, Kathryn A Schubel, and Scott W Hassler. Carbonate sedimentology of the early precambrian hamersley group of western australia. Precambrian Research, 60(1-4):287–335, 1993. [185] Emily F Smith, Francis A MacDonald, James L Crowley, Eben B Hodgin, and Daniel P Schrag. Tectonostratigraphic evolution of the c. 780–730 ma beck spring dolomite: Basin formation in the core of rodinia. Geological Society, London, Special Publications, 424(1):213–239, 2016. [186] Erik A Sperling, Charles J Wolock, Alex S Morgan, Benjamin C Gill, Marcus Kunzmann, Galen P Halverson, Francis A Macdonald, Andrew H Knoll, and David T Johnston. Statistical analysis of iron geochemical data suggests limited late proterozoic oxygenation. Nature, 523(7561): 451–454, 2015. [187] Howard J Spero, Jelle Bijma, David W Lea, and Bryan E Bemis. Effect of seawater carbonate concentration on foraminiferal carbon and oxygen isotopes. Nature, 390(6659):497, 1997. [188] Justin V Strauss, Alan D Rooney, Francis A Macdonald, Alan D Brandon, and Andrew H Knoll. 740 ma vase-shaped microfossils from yukon, canada: Implications for neoproterozoic chronology and biostratigraphy. Geology, 42(8):659–662, 2014. [189] Paul K Strother, Andrew H Knoll, and Elso S Barghoorn. Micro-organisms from the late precambrian narssâ rssuk formation, north-western greenland. Palaeontology, 26(Part 1):1–32, 1983. [190] Kenichiro Sugitani. Strelley pool formation. Encyclopedia of Astrobiology, pages 1–7, 2014. [191] Manfred Sumper and Eike Brunner. Silica biomineralisation in diatoms: the model organism thalassiosira pseudonana. ChemBioChem, 9(8): 1187–1194, 2008. [192] George EA Swann, Mark A Maslin, Melanie J Leng, Hilary J Sloane, and Gerald H Haug. Diatom δ18o evidence for the development of the modern halocline system in the subarctic northwest pacific at the onset of major northern hemisphere glaciation. Paleoceanography, 21(1), 2006.

185 [193] George EA Swann, Melanie J Leng, Hilary J Sloane, Mark A Maslin, and Jonaotaro Onodera. Diatom oxygen isotopes: evidence of a species effect in the sediment record. Geochemistry, Geophysics, Geosystems, 8(6), 2007.

[194] Nicholas L Swanson-Hysell, Catherine V Rose, Claire C Calmet, Galen P Halverson, Matthew T Hurtgen, and Adam C Maloof. Cryogenian glaciation and the onset of carbon-isotope decoupling. Science, 328 (5978):608–611, 2010.

[195] Ryoji Tanaka and Eizo Nakamura. Determination of 17o-excess of terrestrial silicate/oxide minerals with respect to vienna standard mean ocean water (vsmow). Rapid Communications in Mass Spectrometry, 27 (2):285–297, 2013.

[196] R Tartèse, M Chaussidon, A Gurenko, F Delarue, and F Robert. Warm archean oceans reconstructed from oxygen isotope composition of early-life remnants. Geochemical Perspectives Letters, 3:55–65, 2017.

[197] Stuart Ross Taylor and Scott M McLennan. The geochemical evolution of the continental crust. Reviews of Geophysics, 33(2):241–265, 1995.

[198] Mark H Thiemens and John E Heidenreich. The mass-independent fractionation of oxygen: A novel isotope effect and its possible cosmochemical implications. Science, 219(4588):1073–1075, 1983.

[199] Mark H Thiemens, Teresa Jackson, Edward C Zipf, Peter W Erdman, and Cornel van Egmond. Carbon dioxide and oxygen isotope anomalies in the mesosphere and stratosphere. Science, 270(5238):969–972, 1995.

[200] Nicholas J Tosca, Stephen Guggenheim, and Peir K Pufahl. An authigenic origin for precambrian greenalite: Implications for iron formation and the chemistry of ancient seawater. Bulletin, 128(3-4):511–530, 2016.

[201] Paul Treguer, David M Nelson, Aleido J Van Bennekom, David J DeMaster, Aude Leynaert, and Bernard Queguiner. The silica balance in the world ocean: a reestimate. Science, 268(5209):375–379, 1995.

[202] RIF Trindade, E Font, MS D’Agrella-Filho, ACR Nogueira, and C Riccomini. Low-latitude and multiple geomagnetic reversals in the neoproterozoic puga cap carbonate, amazon craton. Terra Nova, 15(6): 441–446, 2003.

186 [203] Jonathan J Tyler, Hilary J Sloane, Rosalind EM Rickaby, Eileen J Cox, and Melanie J Leng. Post-mortem oxygen isotope exchange within cultured diatom silica. Rapid Communications in Mass Spectrometry, 31(20): 1749–1760, 2017.

[204] Harold C Urey. The thermodynamic properties of isotopic substances. Journal of the Chemical Society (Resumed), pages 562–581, 1947.

[205] Marcel TJ Van Der Meer, Stefan Schouten, Jan W De Leeuw, and David M Ward. Autotrophy of green non-sulphur bacteria in hot spring microbial mats: biological explanations for isotopically heavy organic carbon in the geological record. Environmental Microbiology, 2(4):428–435, 2000.

[206] Martin J Van Kranendonk. Volcanic degassing, hydrothermal circulation and the flourishing of early life on earth: A review of the evidence from c. 3490-3240 ma rocks of the pilbara supergroup, pilbara craton, western australia. Earth-Science Reviews, 74(3-4):197–240, 2006.

[207] Jan Veizer and Andreas Prokoph. Temperatures and oxygen isotopic composition of phanerozoic oceans. Earth-Science Reviews, 146:92–104, 2015.

[208] Xinqiang Wang, Ganqing Jiang, Xiaoying Shi, and Shuhai Xiao. Paired carbonate and organic carbon isotope variations of the ediacaran doushantuo formation from an upper slope section at siduping, south china. Precambrian Research, 273:53–66, 2016.

[209] Steve Weiner and Patricia M Dove. An overview of biomineralization processes and the problem of the vital effect. Reviews in mineralogy and geochemistry, 54(1):1–29, 2003.

[210] R_F Weiss. Carbon dioxide in water and seawater: the solubility of a non-ideal gas. Marine chemistry, 2(3):203–215, 1974.

[211] Michael J Whiticar. Carbon and hydrogen isotope systematics of bacterial formation and oxidation of methane. Chemical Geology, 161(1-3): 291–314, 1999.

[212] Elise B Wilkes, Susan J Carter, and Ann Pearson. Co2-dependent carbon isotope fractionation in the dinoflagellate alexandrium tamarense. Geochimica et Cosmochimica Acta, 212:48–61, 2017.

187 [213] Stanley N Williams, Stephen J Schaefer, V Marta Lucia Calvache, and Dina Lopez. Global carbon dioxide emission to the atmosphere by volcanoes. Geochimica et Cosmochimica Acta, 56(4):1765–1770, 1992. [214] Jonathan P Wilson, Woodward W Fischer, David T Johnston, Andrew H Knoll, John P Grotzinger, Malcolm R Walter, Neal J McNaughton, Mel Simon, John Abelson, Daniel P Schrag, et al. Geobiology of the late paleoproterozoic duck creek formation, western australia. Precambrian Research, 179(1-4):135–149, 2010. [215] Jun Yang, Malte F Jansen, Francis A Macdonald, and Dorian S Abbot. Persistence of a freshwater surface ocean after a snowball earth. Geology, pages G38920–1, 2017. [216] Edward D Young, Albert Galy, and Hiroko Nagahara. Kinetic and equilibrium mass-dependent isotope fractionation laws in nature and their geochemical and cosmochemical significance. Geochimica et Cosmochimica Acta, 66(6):1095–1104, 2002. [217] Edward D Young, Laurence Y Yeung, and Issaku E Kohl. On the δ17o budget of atmospheric o2. Geochimica et Cosmochimica Acta, 135: 102–125, 2014. [218] Edward D Young, Issaku E Kohl, Paul H Warren, David C Rubie, Seth A Jacobson, and Alessandro Morbidelli. Oxygen isotopic evidence for vigorous mixing during the moon-forming giant impact. Science, 351 (6272):493–496, 2016. [219] James Zachos, Mark Pagani, Lisa Sloan, Ellen Thomas, and Katharina Billups. Trends, rhythms, and aberrations in global climate 65 ma to present. Science, 292(5517):686–693, 2001. [220] Richard E Zeebe and Dieter Wolf-Gladrow. CO2 in seawater: equilibrium, kinetics, isotopes. Number 65. Gulf Professional Publishing, 2001. [221] J Zhang, PD Quay, and DO Wilbur. Carbon isotope fractionation during gas-water exchange and dissolution of co2. Geochimica et Cosmochimica Acta, 59(1):107–114, 1995. [222] P Ziveri, H Stoll, I Probert, Christine Klaas, Markus Geisen, G Ganssen, and J Young. Stable isotope ‘vital effects’ in coccolith calcite. Earth and planetary science letters, 210(1-2):137–149, 2003.

188 Colophon

his thesis was typeset using LATEX, originally developed by Leslie TLamport and based on Donald Knuth’s TEX. The body text is set in 11 point Arno Pro, designed by Robert Slimbach in the style of book types from the Aldine Press in Venice, and issued by Adobe in 2007. A template, which can be used to format a PhD thesis with this look and feel, has been released under the permissive mit (x11) license, and can be found online at github.com/suchow/ or from the author at [email protected].

189