of ABSTRACT
Title Document:
SULFUR ISOTOPE FRACTIONATIONS IN BIOLOGICAL SYSTEMS: INSIGHT INTO THE PROTEROZOIC BIOSPHERE David T. Johnston, Ph.D. 2007
Directed By:
Professor J. Farquhar, Department of Geology and ESSIC
This study focuses on developing a framework for interpreting isotopic fractionations in the four stable sulfur isotopes, with a special focus on microbial effects. Calculations of low temperature equilibrium fractionations between various sulfur species were performed and suggest a tightly constrained fractionation relationship between the isotopes. This work was followed by conducting a series of batch culture experiments with two prominent sulfur metabolisms: sulfate reduction and sulfur disproportionation. The results of these experiments illustrate that measurable deviations from the equilibrium predictions exist, and further serve to highlight the ability of these measurements to differentiate between fractionations produced by these two metabolisms. This is a distinction not possible from traditional isotope measurements alone (34S/32S). In addition, models constructed to describe these observations provide insight into the intra-workings of these metabolic pathways,
with the most recent experiments suggesting that the traditionally accepted model for isotope fractionation produced by sulfate reducers be re-evaluated. The microbial fractionations were also used to calibrate global sulfur cycle box models. These models have been coupled with the measurement of natural samples to illustrate the onset of microbial disproportionation ~800 million years earlier in Earth history than previously thought and to describe the sulfur cycle of three Proterozoic marine basins.
SULFUR ISOTOPE FRACTIONATIONS IN BIOLOGICAL SYSTEMS: INSIGHT INTO THE PROTEROZOIC BIOSPHERE
By
David T. Johnston
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2007
Advisory Committee: Professor James Farquhar, Chair Professor Alan J. Kaufman Professor William McDonough Dr. Marilyn Fogel Dr. Donald E. Canfield
Copyright by David T. Johnston 2007
Preface
Much of the work presented in this thesis was previously published as a collection of peer-reviewed journal articles. I am grateful to the efforts and comments of all my co-authors on those studies, as their contributions made the papers both stronger and broader in scope. Chapters 1, 3, 6, and 7 represent original contributions, and the references for the remaining chapters are listed below (though all chapters have been modified from their original published form). Copyright approval has been granted for the republication of this material, and is available upon request. Chapter 2: David T. Johnston, James Farquhar, Boswell A. Wing, Alan J. Kaufman, Donald E. Canfield, and Kirsten S. Habicht (2005b). Multiple sulfur isotope fractionations in biological systems: a case study with sulfate reducers and sulfur disproportionators. American Journal of Science. vol. 305, p. 645-660.
I acknowledge comments on early versions of this manuscript by L. Tuit. This research was supported by grants from the NSF (EAR0348382), NASA (EXB and NAI), SOAS (International School of Aquatic Sciences) and the Danish National Science Foundation. I also thank P. Sholt and L. Salling for technical support. I appreciate the insightful reviews of J. Ague, Y. Shen and two anonymous reviewers. Chapter 4: David T. Johnston, Boswell A. Wing, James Farquhar, Alan J. Kaufman, Harald Strauss, Timothy W. Lyons, Linda C. Kah, Donald E. Canfield (2005a). Active Microbial Sulfur Disproportionation in the Mesoproterozoic. Science. vol 310, p. 1477-1479.
I acknowledge support from NSF [EAR 0348382 (J.F.), EAR 9725538 (T.W.L. and ii
L.C.K.), and EAR 0418005 (A.J.K.)], NASA (NAG512350 and NNG05GQ96G), the NASA Astrobiology Institute (D.T.J., B.A.W., and J.F.), and Danish National Research Foundation (Denmarks Grundforskningsfond, D.E.C.) for this research. I also thank J. W. Schopf for the contribution of samples labeled as PPRG (Precambrian Paleobiology Research Group). Chapter 5: David T. Johnston, Simon W. Poulton, Philip W. Fralick, Boswell A. Wing, Donald E. Canfield, James Farquhar (2006). Evolution of the Oceanic Sulfur Cycle at the end of the Paleoproterozoic. Geochimica et Cosmochimica Acta. vol. 70, p. 57235739.
I thank L. Kump, Y. Shen, J. Amend, and one anonymous reviewer for constructive and insightful comments. I acknowledge support from NSF (EAR 0348382), NASA (NAG512350), the NASA Astrobiology Institute (DTJ, BAW, JF), NAI Research fellowship (DTJ), National Science and Engineering Research Council of Canada (PWF), and Danish National Research Foundation (Danmarks Grundforskningsfond; SWP, DEC) for support of this research.
The final stages of this project were funded by the University of Maryland, Graduate School Deans Dissertation Fellowship (Fall 2006).
iii
Dedication
I dedicate this work in memory of Robert A. Miller: not only my grandfather, but also a mentor, friend, and eternal fishing buddy.
In addition, this work is dedicated to my family. The never-ending support that they have all provided has allowed me to pursue this degree with a focus that only comes with true encouragement and support. To my Mom and Dad, I cannot begin to express the debt that I owe you both for standing at my side during both the rainy and sunny days. Even though you both thought it crazy when I expressed my interest in pursuing my Ph.D., you quickly jumped on board. The respect that you had for my decisions and goals has shaped me as a person and is a trait that I only hope I can adopt. The love and friendship that you have both provided inspires me everyday, and drives me to succeed and make you proud. And to Greg, my brother and best friend. You were always my favorite outlet from the rigors of this degree. I would have visited you wherever you landed, but I cannot thank you enough for your hospitality during those long winter trips west. Those couple of weeks would keep me working the rest of the year.
iv
Acknowledgements
I must first acknowledge the patience and energy of James Farquhar, without whom this thesis would not have been possible. His endless commitment to science and the success of his students is unique, and I am grateful for having passed through his lab. In addition to acting as a mentor, I am even more pleased to have made a great friend. I must also thank Boz Wing for providing balance, taking the time to always start from first principles, and instilling in me a deeper appreciation for all things quantitative. A special thanks to Don Canfield, who opened the doors to his lab and group throughout my years at Maryland. My trips to Denmark and chats over cake made this project a richer one and taught me a great deal. To Jay Kaufman, thank you for the insightful conversations about ancient environments and stressing to me the importance of context. I must also acknowledge the work and support of the remainder of my committee (Marilyn Fogel, Bill McDonough and Phil Piccoli), as their guidance helped to shape this project. Lastly, I would like to acknowledge the primary financial support of the NASA Astrobiology Institute, The International School for Aquatic Sciences, the Danish National Research Foundation, and the University of Maryland, Graduate School (Deans Dissertation Fellowship program).
Throughout this project, I was truly fortunate to have been surrounded by supportive and genuine people. First and foremost, I am eternally indebted to Taylor Mauck. Your candor and wit always kept a smile on my face, and your support saw me through the bulk of this project. I could not have done this without you. To all those
v
who passed through the doors of 407 Beech Ave, I also give my thanks. That house quickly, and as a function of the personalities that filled it, took on a personality of its own and was a fantastic place to live (and hide from the lab). Lastly, in my opinion, science is best done in a social and happy work environment. With this in mind, I would like to acknowledge and thank all the grad students who have come and gone during my time at Maryland. A special thanks to the early days of the Craig Hebert era. I am also very thankful for the friendship and constant source of amusement that was GEOL office 4105 from 2002-05 (Yes Courtney and Leah, we are still WE). From the long-haul folks (here is looking at you Ireland-Reno et al.) to those in the new school, I thank you.
vi
Table of Contents
Preface........................................................................................................................... ii Dedication .................................................................................................................... iv Acknowledgements....................................................................................................... v Table of Contents........................................................................................................ vii List of Tables ............................................................................................................... ix List of Figures ............................................................................................................... x Chapter 1: Introductions and historical perspectives .................................................... 1 Sulfur in the Earth system......................................................................................... 1 The geologic record, insights from 34S: ................................................................. 4 The minor isotopes (33S and 36S) enter the scene:................................................... 10 This study:............................................................................................................... 11 Chapter 2: Sulfur Isotope Fractionations in Biological Systems: a case study with sulfate reducers and sulfur disproportionators .......................................................... 14 Introduction............................................................................................................. 14 Background ............................................................................................................. 14 Isotope Notation.................................................................................................. 18 Mass-dependent isotope fractionation ................................................................ 20 Modeling isotope networks..................................................................................... 24 Biological and modeling methods .......................................................................... 25 Results..................................................................................................................... 28 Sulfate reducing microorganisms ....................................................................... 29 Sulfur disproportionating microorganisms ......................................................... 30 Discussion/Conclusions .......................................................................................... 32 Interpretation of experimental data using flow networks ................................... 32 Experiments with elemental sulfur and sulfite disproportionating microorganisms................................................................................................... 35 Geological and biological applications............................................................... 37 Chapter 3: Sulfur Isotope Fractionations Produced by Sulfate Reducing Bacteria: insight into isotope systematics and metabolic models............................................... 41 Introduction............................................................................................................. 41 Background ............................................................................................................. 42 More on mass-dependence.................................................................................. 43 Metabolic models................................................................................................ 45 Biological methods ................................................................................................. 47 Discussion ............................................................................................................... 50 Evaluating metabolic models.............................................................................. 57 The relationship between 33S and 36S.................................................................. 68 Conclusions............................................................................................................. 74 Chapter 4: Active Sulfur Disproportionation in the Mesoproterozoic........................ 77 Introduction............................................................................................................. 77 Background ............................................................................................................. 77 Sulfur cycle models................................................................................................. 79 This study................................................................................................................ 82
vii
Proterozoic oxygenation ......................................................................................... 86 Conclusions............................................................................................................. 88 Chapter 5: Evolution of the Oceanic Sulfur Cycle at the End of the Paleoproterozoic ..................................................................................................................................... 90 Introduction............................................................................................................. 90 Background ............................................................................................................. 91 Setting, Systems and Methods ................................................................................ 92 Geologic setting .................................................................................................. 92 Analytical methods ............................................................................................. 96 Modeling treatment............................................................................................. 97 Results..................................................................................................................... 98 Discussion ............................................................................................................. 100 Implications of the steady-state box model approach....................................... 101 Insight into the euxinic shale environment ....................................................... 109 Dispelling non mass-dependent contributions and calibrating a new relationship ........................................................................................................................... 121 Conclusions........................................................................................................... 123 Chapter 6: Evidence for a mid-Proterozoic oxygenation......................................... 126 Introduction........................................................................................................... 126 Geologic Setting.................................................................................................... 127 Results................................................................................................................... 128 Discussion ............................................................................................................. 131 Chemostratigraphic record................................................................................ 131 Constraints on seawater sulfate......................................................................... 133 Proterozoic seawater sulfate ................................................................................. 135 The Neoproterozoic .......................................................................................... 138 The role of tectonics.......................................................................................... 140 Following the evolution of 33S versus 33S.................................................... 142 Conclusions........................................................................................................... 145 Chapter 7: Conclusions, directions, and closing thoughts ....................................... 146 General conclusions .............................................................................................. 146 Lessons learned at the cellular level ................................................................. 147 A new picture of the Proterozoic biosphere...................................................... 147 Looking forward ............................................................................................... 153 Part I.................................................................................................................. 153 Part II ................................................................................................................ 154 Part III ............................................................................................................... 155 A few closing thoughts ......................................................................................... 156 Appendices................................................................................................................ 157 References................................................................................................................. 170
viii
List of Tables
Table 1...2 Table 2.30 Table 3.31 Table 4.49 Table 5.59 Table 683-84 Table 7.....99 Table 8139-140 Table 9...137 Table 10.147 Table A1.159 Table A2.162
ix
List of Figures
Figure 1.4 Figure 2.9 Figure 311 Figure 428 Figure 533 Figure 636 Figure 738 Figure 851 Figure 954 Figure 10..56 Figure 11..60 Figure 12..64 Figure 13..67 Figure 14..70 Figure 15..72 Figure 16..74 Figure 17..80 Figure 18..81 Figure 19..85 Figure 20..87 Figure 21......93 Figure 22..97 Figure 23104 Figure 24106 Figure 25108 Figure 26110 Figure 27117 Figure 28119 Figure 29122 Figure 30128
x
Figure 31132 Figure 32134 Figure 33139 Figure 34143 Figure 35148 Figure 36149 Figure 37150 Figure A1...158 Figure A2...166 Figure A3...168
xi
Chapter 1: Introductions and historical perspectives
Sulfur in the Earth system The geologic sulfur isotope record represents one of the most robust and informative indicators of Earth surface change. This record carries a wealth of information because of the role and importance of sulfur, as an element, in surface processes. In part, sulfur serves as an effective tracer due to the wide range of reduction-oxidation (redox) states, spanning from the oxidized end-member sulfate (SO42-: +6) to the reduced hydrogen sulfide (H2S: -2). In addition, sulfur is poly-isotopic, which greatly contributes to the potential for this elemental system to carry information. The wide range of redox states promotes cycling in surface environments and catalyzes the production of a large range of isotopic fractionations, which span from +130 to 60 in 34S (Coplen et al., 2002). Further, if one considers the magnitude of an isotope effect (up to 190 in 34S) and the analytical uncertainties on such a measurement (often ~ 0.20 using a SO2 combustion technique), the interpretability and capability of this isotope system to record information is further revealed.
Before I can consider the details of the sulfur isotope system, I must understand the isotopes, their abundances, and their distribution in the Earth environment. Sulfur has four stable isotopes (32S, 33S, 34S, 36S). The stable isotopes vary greatly in abundance on Earth, with the following distribution: 32S = 94.93, 33S = 0.76, 34S = 4.29, and 36S = 0.02% (Coplen et al., 2002). The most heavily studied isotope relationship is the ratio between 34S and 32S, often described using 34S (delta notation is a measure of 1
the deviation of the ratio of 34S to 32S in a sample, compared to that of a standard, and noted in parts per thousand or : see Chapter 2 for full definition) due in large part to the high relative abundances and subsequent analytical ease in their measurement. The present study also includes analysis of 33S and 36S abundances, provides a theoretical framework for their interpretation, and focuses on the use of these isotopes in understanding natural systems.
Table 1: A list of sulfur utilizing metabolic processes and the associated isotopic fractionation in 34 32 S/ S (modified from Canfield, 2001). The fractionations represent the largest isotopic fractionation known for that process. Processes that are not included are thought to proceed through various intermediate reactions, most of which are included above. Values of zero suggest a fractionation < 1 , whereas ? represents a metabolism where isotope studies have not been conducted. References are: 1) Hoek et al. (2006), 2) Harrison and Thode (1958), 3) Kaplan and Rittenburg (1964), 4) Fry et al. (1986) 5) Fry et al. (1988), 6) Habicht et al. (1998), and 7) Wentzien and Sand (2004). This list is not exhaustive, and the reader is referred to Canfield (2001) for a more thorough review.
There are a variety of pathways and processes involved in the cycling of sulfur on the Earths surface, with some having a greater geologic relevance than others. The most prominent of these processes are biological, as microorganisms play an important role in the cycling of sulfur and impart large isotopic fractionations. These fractionations are often used to track biological activity in an environment or through time. One
2
fate of sulfur in the biosphere is assimilation for biosynthesis. The organic sulfur reservoir is generally taken to be quantitatively unimportant and is often disregarded. The more commonly considered biological sulfur cycle is that of catabolic or dissimilatory processes. These processes utilize the eight-electron sulfur redox gradient to gain energy and fuel metabolic activity. Below I provide a general list of the prominent metabolisms and associated isotopic fractionations.
To date, experimental studies with the above metabolisms indicate that sulfate reduction and sulfur disproportionation produce the largest isotope fractionations. This is in part due to the large change in the redox state of the reactants and products, as both metabolisms cover the full range (8 electron transfer). Although equilibrium isotope fractionations are also suggested to be capable of producing large fractionations, they are generally not expressed (Ohmoto, 1986). Laboratory experiments of natural sediments (cf. Canfield, 2001b) have extensively documented the large metabolic fractionations, thus biology is used to explain the observed geologic isotope record. Of these many different metabolic processes, sulfate reduction is thought to be the most prominent process and have the deepest geologic root (Shen et al., 2001). The terminal electron acceptor for this metabolism, sulfate, is the second most abundant anion in the modern ocean at ~ 29mM (Jorgensen and Kasten, 2006). From this, it is easy to understand how sulfate reduction has come to play the role that it does, providing over 50% of the energy for organic carbon remineralization in modern marine sediments (Jorgensen and Kasten, 2006). Conversely, sulfur disproportionation reactions do not have as strong a link to the
3
carbon cycle, utilize more trace sulfur redox intermediates, and are thus relegated to a less prominent role in the biogeochemistry of marine environments.
The geologic record, insights from 34S: The sulfur isotope record has long been targeted to pinpoint large-scale changes in the global climate and redox state. A recent compilation of 34S record is presented in Figure 1 (modified from Canfield, 2001). The 34S composition of the bulk Earth is inferred to be approximately 0 (Thode, et al., 1961; Ohmoto, 1986) and is the original input to the Earth surface environment. In Figure 1, the diamond symbols
Figure 1: A compilation of sedimentary sulfide 34S versus time, modified from Canfield (2001). The definition of is found later in the text. All the diamonds represent measurements of sedimentary sulfides. The two upper bounding lines represent model approximations for the isotopic composition of seawater sulfate, whereas the lower bounding line is a reference line offset by -55 . The record is divided into three stages, separated by black dashed lines. The red data describes an anoxic Archean sulfur cycle; the blue data represents an intermediate (partially oxygenated) sulfur cycle, with the green data illustrating what is thought of as a modern sulfur cycle.
4
represent measurements of sedimentary sulfides. The uppermost two lines extending through the record represent model estimates of the seawater sulfate composition and bracket a 5 window. Offset by -55 is an additional line that is used for reference on the magnitude of deviation from seawater. This record is traditionally divided into three stages, which correspond to the three dominant stages, or redox conditions of the Earths surface. I begin with a description of these subdivisions.
The first stage extends through the Archean and into the earliest Proterozoic, and is characterized by a small range of isotopic compositions. The isotopic difference between estimates of seawater sulfate and sulfide are generally less than ~ 10 and vary round a bulk Earth composition of ~ 0 (Canfield et al., 2000). Archean inputs into surface environments are thought to be dominated by volcanic and hydrothermal inputs of SO2 and H2S to the atmosphere and oceans. Many processes, including a range of inorganic mechanisms, are capable of producing the small fractionations seen in the Archean record. The small fractionations are also interpreted as providing information about relationship between the sulfur cycle and biosphere. Experimental evidence suggests that the range of fractionations seen in the Archean is indicative of the activity of sulfate reducing microorganisms operating under low oceanic sulfate concentrations (< 200 M: Habicht et al., 2002). Further, evidence suggests that sulfate reduction was active back to 3450 Mya (Millions of Years Ago), corroborating this hypothesis (Shen et al., 2001). Shen and colleagues observed fractionations of ~ 24 between bedded evaporites and evaporite hosted pyrite crystals in the 3450 Ma Dresser Fm., North Pole Australia, suggesting both the
5
activity of sulfate reduction (to produce the observed fractionation) and elevated sulfate concentrations (> 200uM), to allow for larger biological fractionations and as a source for the evaporites. The interpretation of the Dresser Formation as recording a biological signature has, however, been contested (for a review, see Shen and Buick, 2004). Debate aside, the Archean 34S record does provide valuable insight into the activity of the sulfur cycle at that time.
At ~ 2400 Mya, the range of sulfur isotope fractionation increased significantly from < 10 to > 40. This change was in response to global climatic events and the postulated Great Oxidation Event (see Holland, 2006 for a review). Widespread geologic evidence now exists from multiple continents for this profound change in the Earth environment (Karhu and Holland, 1996; Bekker et al., 2004; Kopp et al., 2005). The most direct consequence of surface oxygenation on the sulfur cycle was the activation of oxidative weathering, mobilizing terrestrial sulfides. Prior to 2400 Mya, atmospheric oxygen levels are believed to have been too low to oxidize continental sulfur (as pyrite). However, as a result of oxidative weathering, the marine reservoir would have been flooded with igneous-derived sulfur. These events increased the seawater sulfate reservoir to a level greater than 200 M, resulting in larger isotopic fractionations being expressed by sulfate reducers. For the following ~ 1400 Ma (Million years), the observed range of isotopic fractionations bordered on an upper limit of ~ 45, matching the experimental limit of sulfate reducing microorganisms. The remarkable monotony of the observed Proterozoic record, and the consistency
6
with the observed limit for sulfate reducers led to the suggestion that sulfate reduction dominated the marine sulfur cycle during this period.
The increased size of the marine sulfur reservoir from oxidative weathering and the steady production of hydrogen sulfide (the by-product of sulfate reduction) may have driven a significant change in the chemistry of the deep ocean during the monotonous 1400 Ma following the GOE. The interplay between the production and burial of sulfide versus the production and re-oxidation of sulfide to sulfate, strongly influences (in addition to the carbon cycle) the amount of atmospheric oxygen over geologic timescales (Canfield, 2004). During the Archean and early Paleoproterozoic, evidence from expansive banded iron formations points to an anoxic and ferruginous (Fe2+ rich) deep-ocean. However, as Canfield (1998) suggested, the production of sulfide in the ocean slowly titrated the available iron, and once the iron pool became depleted, sulfide began to accumulate in the deep ocean water column. This condition may have characterized a majority of the Proterozoic. Using a new iron speciation approach (Poulton and Canfield, 2005), the transition from ferruginous to sulfidic deep-ocean conditions was identified in the 1840 Mya Animike Basin, North America (Poulton et al., 2004). Further support was given to the sulfidic ocean hypothesis with the basin analysis studies of Shen et al. (2002, 2003), which illustrated the presence of sulfidic conditions in the middle Proterozoic McArthur Basin of Australia.
7
The second increase in the observed range of fractionations is seen at ~ 800 Mya, recorded in sulfides from the Chuar Gp., Grand Canyon and extends through the entirety of the Phanerozoic. This transition precedes the first of the postulated Neoproterozoic snowball glaciations, suggesting that the biospheric change that promoted the production of larger 34S fractionations was not related to the ice ages, but could however, be related to the early emergence of an environmental condition that eventually resulted in glaciation. Association to climate aside, the increase in the magnitude of fractionation was originally thought to be associated with the oxidation of the deep ocean, the activation of non-photosynthetic sulfide oxidizing microorganisms, and the onset of sulfur disproportionation reactions (Canfield and Teske, 1996; Canfield, 1998). As indicated above (Table 1), disproportionation is capable of producing much larger fractionations than sulfate reduction alone. More recent studies track the coupled co-variation in sulfate and sulfide, which indicate that the isotopic difference between the two reservoirs may not have increased beyond ~ 47 until the terminal Neoproterozoic (Hurtgen et al., 2002, 2005; Fike et al., 2006; Canfield et al., 2007). The measurement of 33S and 36S is used to directly address the onset of disproportionation in Chapters 4-6.
Mass-dependent fractionations: A vast majority of natural isotope fractionating processes result in mass-dependent fractionations. These fractionations are controlled by the relative mass difference between the isotopes, and the effects are observable in elemental systems with more than two stable isotopes. In the sulfur isotope system, which has four stable isotopes,
8
Figure 2: A plot of 33S versus 34S illustrating the principle of mass-dependent fractionation. Samples plotted are both from this dissertation and unpublished UMd data. Definitions of are presented later in the text. Samples are from environments younger than 2000 Ma. All samples are from this study, with additional unpublished data from modern marine environments. The RFL represents the reference fractionation line, which has a slope of 0.515. Though shown for 33S versus 34S, the same relationship could be illustrated in 36S versus 34S, which would plot along a RFL value of ~ 1.90.
the relative mass difference between 32S and 33S (32S is the most abundant and therefore the reference isotope) is 1 atomic mass unit (AMU), whereas the difference between 32S and 34S is 2 AMU. When these differences are compared, the relationship between the magnitude of fractionation in 33S and 34S is roughly 1AMU/2AMU or 0.515. Figure 2 illustrates the mass-dependent fractionation relationship for 33S and 34S. The same principles hold for the other sulfur isotope relationship (34S vs.36S) and results in a line of slope ~ 1.90. Since these fractionations are controlled by mass and physical chemistry, they are tightly constrained. Even though there exists a tight isotopic correlation, some small-scale variability can be measured. The variability in 9
values generally occurs at the sub-percent level and can be a function of process (ie. kinetic, equilibrium, etc.) conditions (ie. temperature, molecules involved, etc.), or a combination of multiple factors. This variability can result in the production of isotope effects in 33S and 36S that deviate slightly from a single reference array (see Chapter 2 for details). These small deviations provide information about the process(es) and/or conditions that promoted the isotopic fractionation. It is these small, but resolvable, signatures that this dissertation sets out to better understand.
The minor isotopes (33S and 36S) enter the scene: Until the late 1990s, the measurement and subsequent interpretation of the least abundant stable sulfur isotopes (33S and 36S) had been restricted to the field of cosmochemistry and the study of meteorites (Hulston and Thode, 1965; Thiemens 1999, 2006). Those pioneering works date back to the 1960s and provide a backdrop for all the work in this dissertation. Fast-forward three decades and in 1999, a postdoc working with Dr. Mark Thiemens at UCSD began measuring Archean rocks in hopes of finding the earliest signs of microbial sulfate reduction. Instead of finding a diagnostic inorganic biomarker, Dr. James Farquhar found the fingerprint of an anoxic atmosphere: mass-independent sulfur isotope fractionations, an effect that left its imprint on the geologic record until ~ 2400 Mya (Farquhar et al., 2000). The search for biological signatures continued however, with early measurements of laboratory experiments run with Archeaglobus fulgidus (a sulfate reducing microorganism) suggesting that the fractionations produced by biology were massdependent and differed from both that of equilibrium and the reference fractionation
10
array. This simple observation and the implications therein spurred much of the work included in this dissertation. With the presence of mass-independently fractionated sulfur in the Archean, the target environments for the application of mass-dependent systematics, such as shown in Figure 2, is the post-2400 Mya world (or Stage II and III of Figure 3).
Figure 3: The record of 33S versus time, modified from Farquhar et al. (2000). Capital delta notation is defined later in the text and errors are smaller than the symbols. All data is presented in Farquhar et al. (2000) and colored by Stage, which are also reviewed in Farquhar and Wing (2003, 2005). Stage I marks a period when mass-independent isotope effects dominate the sulfur isotope record. Stage II represents a transitional period preceding the establishment of a fully mass-dependent record (Stage III). These data were originally used to argue for an increase in the level of atmospheric oxygen at ~ 2450 Ma (Farquhar et al., 2000).
This study: The primary goal of this study is to determine the quantitative character of biological isotope fractionations, namely sulfate reduction and sulfur disproportionation, for the four stable sulfur isotopes and to apply this information to natural systems. In addition, this work is done with the goal of identifying a unique, inorganic biomarker 11
that can be used later in studies of deep-time and astrobiological relevance. There are two primary types of studies included. The first are experimental, where I test different metabolisms and metabolic dependencies (Chapters 2-3). The second type of study is environmental, where I use the results of the experimental studies to constrain models of different natural systems and compare model predictions with natural samples. These systems range in size and time, from large-scale changes across the Proterozoic Eon to basin-scale variability covering millions of years. What they all share is a diverse and complex biogeochemical sulfur cycle that can be studied with the multiple sulfur isotopes.
This study begins with an experimental project examining the fractionation in 33S for sulfate reducing and sulfur disproportionating microorganisms (Chapter 2). The second experimental study (Chapter 3) describes a set of focused sulfate reduction temperature-block experiments that produced a large 33S and 36S data set, which I use to test the correctness of different metabolic models. Chapter 4 describes the first environmental application of this mass-dependent multiple isotope system, where I used carbonate-associated sulfate (CAS) and evaporites from the Proterozoic to pinpoint the evolutionary onset of sulfur disproportionating microorganisms. In Chapter 5 I discuss another geological study that used sulfide minerals to reconstruct a marine basin that captured the Paleoporterozoic transition from a ferruginous to sulfidic deep-ocean. This basin-scale study of sulfide minerals is followed-up by an examination of the Mesoproterzoic McArthur Gp. (Reward Fm.) and Roper Gp. (Velkerri and McMinn Fms.), where I again use sulfide minerals to place quantitative
12
constraints on the isotopic composition of seawater sulfate, and to address whether the oxidative sulfur cycle was active at this time (Chapter 6). Lastly, Chapter 7 synthesizes the results of this study, places this work in to greater context of the field, and outlines goals and future areas of interest.
13
Chapter 2: Sulfur Isotope Fractionations in Biological Systems: a case study with sulfate reducers and sulfur disproportionators
Introduction Sulfur isotope (32S, 33S 34S, and 36S) measurements of sulfur compounds associated with dissimilatory sulfate reduction, elemental sulfur disproportionation and sulfite disproportionation indicate that different types of metabolic processes impart different sulfur isotope signatures. An established network for sulfate reduction was used previously to explain the isotope variability. Here, I revisit that treatment and expand it to branching networks representative of biological sulfur disproportionation, in an attempt to understand sulfur isotope fractionations associated with that metabolism. This context is used to interpret experimental data for both sulfate reducers and sulfur compound disproportionators. I explore the types of information about material flow through these metabolic processes that can be extracted by using isotope data. The different sulfur isotope relationships (33S and ) allow various metabolic processes to be distinguished from one another, even when 34S fractionations are similar, providing a tool that can be used to interpret and identify different types of biological sulfur fractionations in the geologic record.
Background Dissimilatory sulfate reduction likely played an important role in determining the isotopic composition of sedimentary sulfides and oceanic sulfate preserved in the 14
geologic record. It has been proposed that disproportionation of intermediate sulfur species (that is the metabolic production of both an oxidized and reduced endmember) has also played a role in determining the magnitude of the fractionation within that record (Canfield and Thamdrup, 1994). The relative impact of these two metabolisms on the global sulfur isotopic record, however, remains unknown. Habicht and Canfield (2002) attempted to constrain the relative contribution of these two metabolisms in a number of modern environments. Pinpointing a robust method to differentiate between these metabolisms will make it possible to test models of ocean chemistry and of microbial evolution. In this study, I examine the sulfur isotope (32S, 33S, 34S, and 36S) signatures of products from experiments with dissimilatory sulfate reducers and sulfur disproportionators, and then integrate these results with models of material flow through metabolic networks as a first step toward addressing these questions.
Experimental studies of sulfate-reducing bacteria under different physical and chemical conditions have established a likely biochemistry for sulfate reduction (Rees, 1973; Peck, 1959, 1961; Castro and others, 2000; Detmers et al., 2001; Canfield, 2001). Rees (1973) described sulfate reduction using the network,
1 = 4(out ) 2
SO
SO= 4(in )
4 a 5a
APS
4 b 5b
3 H S SO 2
= 3
(1)
where the brackets represent the cell wall, APS is adenosine-5-phosphosulfate, and the numbers represent reaction pathways capable of fractionating material. The upper net fractionation limit for this pathway is roughly 47 in 34S (Harrison and Thode,
15
1958; Kaplan and Rittenburg, 1986; Kemp and Thode, 1968; Chambers et al., 1976, Canfield, 2001). Since sulfate reducing bacteria impart large sulfur isotopic fractionations, and considering that dissimilatory sulfate reduction accounts for over 50% of the organic carbon remineralization in modern marine sediments, this process is considered one of the most important controls on the sulfur isotope record (Jrgensen, 1982; Thamdrup et al., 1996; Canfield, 2001; Habicht and Canfield, 2002).
Our understanding of sulfur disproportionation is less well developed. Disproportionation uses various sulfur intermediates (elemental sulfur, sulfite, and thiosulfate) and produces both oxidized (sulfate) and reduced (hydrogen sulfide) endmembers (Bak and Pfenning, 1987; Canfield and Thamdrup, 1994; Thamdrup et al., 1994; Canfield and Teske, 1996; Finster et al., 1997; Habicht and Canfield, 1998; Cypionka et al., 1998). In the case where sulfite is disproportionated, the reaction proceeds as (Frederikson and Finster, 2003), SO3=
1 2
SO3=( in ) 8 H 2S
4a 5a
= APS SO4( in ) 5b
4b
6 7
= SO4 ( out )
(2)
whereas when elemental sulfur is disproportionated,
S0
1 2 = S (0in ) 10 H 2S
4a 5a
SO3
= 5b
4b
APS
6 7
= = SO4 ( in ) SO4 ( out ) .
9
8
(3)
Fractionations in sedimentary sulfides that exceed those produced by sulfate reduction have been attributed to the disproportionation of intermediate sulfur species
16
(Jrgensen 1990a, b; Jrgensen and Bak, 1991; Canfield and Thamdrup, 1996). These intermediates are naturally present in low concentrations (Troelsen and Jrgensen, 1982; Thamdrup et al., 1994), but when they are disproportionated, they can impart large fractionations (up to 70 in 34S) between the product sulfate and sulfide (Thamdrup et al., 1993; Canfield and Teske, 1996; Habicht and Canfield, 1998; Habicht and Canfield, 2002).
Studies of modern seafloor ecosystems have used both microbial and geochemical techniques to examine the fractionation between product sulfides and residual porewater sulfate (Habicht and Canfield, 2002). This approach provides convincing isotopic evidence for the influence of disproportionation in some modern environments and further supports the argument that disproportionating bacteria played an important role in the geological record of sulfur isotope fractionations. However, it is not possible to apply this method directly to the ancient geological record due to a lack of contemporaneous sulfide and pore water sulfate, and because it is not possible to study cultures of indigenous sulfate-reducing bacteria in the same way as Habicht and Canfield (2002).
Here, a new isotope tool is used to study the relative influences of dissimilatory sulfate reduction and sulfur disproportionation on the sulfur isotope record. It is demonstrated that abundances of 33S (and 36S) can be used in conjunction with 34S to differentiate the signatures of biological sulfate reduction and biological disproportionation. In the geologic record it is suggested that these variations can be
17
distinguished from those produced by non-biological processes and hence provide constraints on environmental evolution throughout the Proterozoic Eon. This approach may allow for the evaluation of fractionation processes contributing to the sulfur isotope record in ancient and modern systems.
Isotope Notation
Isotope ratios are used to determine the fractionation factor () between two reservoirs, where S = 3x S A S 32 SB
32 3x
3x
AB
(4)
and where A and B represent the two sulfur pools of interest and 3x is either 33, 34, or 36. Specifically, 33S/32S, 34S/32S and 36S/32S are reported using standard delta notation (), where
33S = 1000 ( 33 1) , 34S = 1000 ( 34 1) ,
and
(5) (6)
36S = 1000 ( 36 1) .
(7)
The less abundant isotopes (33S and 36S) are reported using capital delta notation (), where 34S 0.515 ) 1 33S = 33S 1000 (1+ 1000 and (8)
18
34S 1.90 36 S = 36S 1000 (1 + ) 1 . 1000
(9)
The exponents in these relationships (0.515 and 1.90) define the reference fractionation line (RFL) and approximate single-step thermodynamic equilibrium isotope exchange effects at low temperature (Hulston and Thode, 1965). All isotopic normalizations are discussed in Appendix 1.
Many fractionation processes depend strongly on the relative mass difference between the isotopes, resulting in highly correlated variations for 33S and 34S (Hulston and Thode, 1965; Farquhar et al., 2003). These variations are described by
33 34
=
ln( 33 ) , ln( 34 )
(10)
which is a theoretical relationship between fractionation factors () and reflects a specific fractionation process. Current convention uses a mathematically equivalent, but functional definition that describes observed variations for 33S and 34S using ) ln(1 + ) 1000 1000 , 34S A 34S B ln(1 + ) ln(1 + ) 1000 1000 ln(1 +
33S A
33S B
33 34
=
(11)
where A and B represent different sulfur reservoirs (Miller, 2002). For simplicity, I note that when discussing a theoretically derived relationship is used, whereas when discussing an observation, is used.
In this study there is a distinction between isotopic fractionation processes, which are described using and , and observed isotopic fractionations, which are described 19
using 3xS and (Mook, 2000; Farquhar et al., 2003; Angert et al., 2003). I use and as input values in the modeling. Theta () is largely a theoretical term because it is difficult to determine experimentally, however calculations presented in Farquhar et al. (2003) and Chapter 3 serve to approximate equilibrium processes. In this chapter, is used to describe both the fractionations observed in the experiments and the fractionations predicted from the modeling. 3xS and are net quantities that can include the effects of many processes and mass-balance (Hayes, 2001; Farquhar et al., 2003).
Mass-dependent isotope fractionation
Theoretical work by Urey (1947) and Bigeleisen and Mayer (1947) provides the foundation for the understanding of equilibrium mass-dependent isotopic fractionations. The predictions made using their formulae, and those of subsequent studies that describe other physical and chemical mass-dependent isotopic fractionations, have been applied to multiple isotope systems to learn more about the processes that operate in nature (Farquhar et al., 2003). Advancements in analytical precision allow for systems such as the four stable isotopes of sulfur to be more thoroughly studied experimentally.
Isotopic fractionation processes have been traditionally grouped into several overarching categories that include both equilibrium and kinetic processes. Equilibrium fractionations described by Urey (1947) and Bigeleisen and Mayer (1947) reflect the minimization of free energy associated with isotopic exchange
20
reactions. Kinetic fractionations comprise a much larger group of processes that are characterized by unidirectional reactions and transport. Transition state theory has been applied to kinetic fractionations by Bigeleisen and Wolfsberg (1958), who describe chemical kinetic isotopic fractionations in terms of the vibrational energies of the transition state, those of the reactants, and also the reaction path and its relationship to the potential energy surface that describes the different states. Other authors describe kinetic fractionations in terms of the relationship between velocity and kinetic energy for isotopically substituted species (Craig et al., 1988; Matsuhisa et al., 1978; Mook, 2000; Young et al., 2002; Schauble, 2004). It is important to recognize that each of these treatments applies in a different physical-chemical situation, and that each of these treatments has different implications for the isotopic fractionations. Further, some studies have referred to metabolic fractionation processes as kinetic fractionations (Kaplan, 1975). I prefer to make a distinction between fractionation effects that are produced by multi-step metabolic and biogeochemical processes and those that are intrinsic to individual chemical and physical processes (cf. Urey, 1947; Bigeleisen and Mayer, 1947; Bigeleisen and Wolfsberg, 1958; Matsuhisa et al., 1978; Craig et al., 1988; Mook, 2000; Young et al., 2002; Schauble, 2004).
The possibility of different mass-dependent fractionations arising from different equilibrium and kinetic fractionation processes is well documented (Bigeleisen and Wolfsberg, 1958; Matsuhisa and others, 1978; Mook, 2000; Young and others, 2002; Schauble, 2004, Young and Galy, 2004). Values of produced by equilibrium
21
isotopic exchange of sulfur between sulfur species have been constrained using the theoretical treatment of Urey (1947) and are generally near 0.515 for 33 and 1.90 for
36
. Determinations of 3x-34 produced by kinetic processes are more variable, falling
between 0.500 and 0.515 for the example of 33, and depend on the nature of the fractionation process responsible for the isotopic fractionations.
Values of produced by multi-step metabolic processes described by Farquhar et al. (2003) depend on the structure of the network, the relative amounts of material transferred through the metabolic network, and on the isotopic fractionations ( and ) of its constituent steps. Because of the dependence on material transfer and network structure, the values of can be more variable than the values of for all constituent steps within any given network. All of the modeling efforts presented in this chapter assume a 33 = 0.5145 (36 = 1.89), which is an equilibrium value for exchange between sulfate and hydrogen sulfide at 20 C (Farquhar et al., 2003). It is recognized that the calculations may later be refined with more precise assignments of 33 (and 36) for each of the constituent network steps, but do not explore this aspect in the present study. Data for 36S is presented for experiments described in this chapter, but interpretations based on those measurements are discussed in Chapter 3.
The term SR is calculated using Equations 2 and 5, and is thus designated to describe sulfate reduction. Although describing a single process, SR does not necessarily represent a single reaction. For example, a net fractionation relationship could be calculated for the sulfate reduction metabolism (net and net or SR and SR), which 22
encompasses a series of biochemical reactions with their own unique and (such as diff and diff for diffusion, or enz and enz for an enzymatic processes). In the case where there is a theoretical understanding for a well-defined process, such as for temperature-dependent equilibrium isotope exchange, is used. Otherwise, when describing a transformation that is thought or known to encompass many steps, is used.
The variations of interest here are the net biological signatures. These signatures originally have the added effects of material balance in the system (due to the batch culture experimental design), and thus deviate from the situation whereby obs = SR (where obs is the observation). If the experiments were run in a chemostat ( steadystate bioreactor), closed system effects would not be expected in the experiments and obs should equal SR. The data is corrected to remove the isotopic effects of the closed system batch experiments using standard Rayleigh equations presented later in the chapter (also see Appendix 2). This allows for the manipulation and discussion of isotopic signatures thought to reflect the sulfate reduction process. In addition to generating non-zero 33S and 36S due to contributions from specific processes, there is also a mass-balance contribution, which is more fully discussed in Appendixes 1 and 5.
23
Modeling isotope networks In a recent study, Farquhar et al. (2003) examined how the flow of sulfur through simple metabolic networks can affect the isotopic fractionations in systems with three or more stable isotopes of the same element. These authors took a previously published metabolic network that describes sulfate reduction (Rees, 1973) and adapted it to describe the isotopic fractionations involving the four sulfur isotopes. These predictions are represented by a flow-net that is contoured by a pair of relative flux terms, f3 and f5. f3 = 3/( 3 + 2) and f5 = 5/( 5 + 3), where j is a term quatifying the amount of material moving along pathway j. f3 indicates the relative amount of material leaving the cell as hydrogen sulfide (pathway 3 in Equation 1) as opposed to leaving the cell as sulfate (pathway 2 in Equation 1), whereas f5 represents the relative internal backflow of sulfur (pathways 5a and 5b in Equation 1). These flow-nets illustrate the dependence of (and 33S) on the intercellular recycling of sulfur (f5) and the internal/external exchange of sulfate (f3).
An important result from this treatment is that values for will be less than for all values of f other than 0 and 1 if the fractionation factors () are less than 1.00. The converse is true when the fractionation factors are greater than 1.00. In more complex networks with multiple fractionation factors, some of which are greater than 1 and some that are less than 1, can be greater than or less than depending on the values of f. When values of f = 0 or 1, = . Another consequence from this treatment is that the solution to progressively more and more complex networks can be obtained by nesting the solutions for simple networks (see also Rees, 1973). In
24
other words, the fractionation factors for multiple steps are multiplicative.
Modeling disproportionation networks can be undertaken by adapting a similar treatment to the disproportionation network structure. Farquhar et al. (2003) examined a very simple branching network component without backflow and noted that the mass-dependent factor, , between the products of this network would be a weighted average of the 's for each of the branches. This treatment can be expanded to branching networks with backflow by applying the ideas above. If the fractionation factors along a branch are greater than 1.00, such as is the case along oxidative branches for sulfur disproportionation, the net values have the potential to be greater than the constituent values for all values of f other than 0 and 1. For example, to produce values greater than 0.5145 for a branching network with all = 0.5145, a back reaction along the oxidative branch of the network (similar to those seen in Equations 2 and 3) must be employed. These principles are used below to interpret the experimental results.
Biological and modeling methods Pure cultures of sulfate reducing bacteria were grown according to media recipes recommended by Deutsche Sammlung von Mikroorganismen, Braunschweig, Germany (DSM strain # 3382 and 10085 respectively below). Inoculated solutions were prepared under anaerobic conditions with a N2 atmosphere and without the initial addition of sulfate or substrate. Desulfobacterium autotrophicum was grown under a H2/CO2 atmosphere where atmospheric pressure was monitored, whereas
25
Desulfospira jorgensenii was provided an organic substrate (Na-pyruvate). Cultures were grown in batch and harvested during the late exponential growth phase. Samples were fixed and metabolic activity terminated through the addition of a 20% Zinc Acetate solution.
Sulfide concentrations ( 5% 1) are determined photometrically using a methylene blue technique (Cline, 1969), and later distilled using 6N HCl, liberating H2S, which was then reacted with AgNO3 to produce Ag2S. Sulfate and thiosulfate concentrations ( 1% SD) were determined via ion chromatography after the addition of a bicarbonate buffer. Sulfates were originally precipitated as BaSO4 and then further reduced (Thode, 1961; Forrest and Newman, 1977) to CdS, and then finally to Ag2S (see Appendix 3 for general chemical methods).
Cell growth was monitored through optical density measurements at 670nm. Optical density was calibrated to cell numbers by staining cells with DAPI and counting them under UV light. As published by Detmers et al. (2001), cell specific reduction rates (csSRR) during exponential phase growth are represented by, SO4 ( 2 ) SO4 (1) csSRR = (cn(1) + cn( 2) ) (T( 2 ) T(1) ) 2
2 2
(12)
where subscripts 1 and 2 represent different sampling times. Here, SO42-(2) is the sulfate concentration at time (2), cn is cell number and T is time. The samples used for the disproportionation experiments were generated in the studies of Canfield et al. (1998) for elemental sulfur disproportionation and Habicht et al. (1998) for
26
sulfite disproportionation. Product ratios (SO4:H2S) included in these publications serve as mass-balance parameters when interpreting isotope results.
Our modeling technique is presented in Farquhar et al. (2003), and outlined earlier in this dissertation. The model network used for dissimilatory sulfate reduction is presented in Rees (1973) and can be seen in Equation (1). In the case of dissimilatory sulfate reduction, I solved for the relationship between the initial sulfate and product hydrogen sulfide. For sulfur disproportionation, I use the networks presented in Equations (2) and (3) and a stoichiometry of 4SO32- + 2H+ disproportionation and 4So + 4H2O H2S + 3SO42- for sulfite
3H2S + SO42- + 2H+ for elemental sulfur
disproportionation to solve for both end-members (sulfate and sulfide) as well as the composition of the reactant pool (either elemental sulfur or sulfite) (Thamdrup et al., 1993; Finster et al., 1998; Canfield, 2001). All reactions are modeled with 33 values at 0.5145. The individual fractionation factors for the sulfate reduction pathway are taken as = 1.003 for diffusion of sulfate into the cell, and = 0.975 for both the internal conversion of sulfate to sulfite and the enzymatic reduction of sulfite to hydrogen sulfide (Rees, 1973). Fractionation factors for the disproportionation networks were taken partly from the sulfate reduction literature (Rees, 1973), with the remaining values being provided by theoretical calculations presented in Table 2 of Farquhar et al. (2003).
27
Results Cell specific reduction rates (measured in fmol/cell*day) for both sulfate reduction experiments match well with those published in Detmers et al. (2001) for exponential phase growth. Bacterial populations during the experiments reached cell densities of the same magnitude as seen previously with these strains. Concentrations of the sulfate and sulfide pools were monitored and co-vary (Figure 4). Isotopic compositions for the stable isotopes of sulfur as well as the cell-specific metabolic
Figure 4: Concentrations of sulfates and sulfides for sulfate reduction experiments with D. autotrophicum and D. jorgensenii are plotted against the duration of the experiment. Each symbol represents a time when the experiment was sampled. I note the stoichiometric depletion of sulfate and accumulation of sulfide. Experiments were run until ~ 50 % of the initial sulfate was consumed. Due to a low csSRR and a low tolerance for H2S, D. jorgensenii was started at a lower sulfate concentration. This concentration is still large enough not to effect the fractionation produced by the organisms (Habicht et al., 2002).
28
rates and concentration data from sulfate reduction experiments are listed in Table 2. The net fractionation between the reactant sulfate and the product sulfide for sulfate reduction experiments is reported (Table 2: sulfate sulfide). Additionally, Table 3 presents the fractionations from sulfur disproportionation experiments. In the case of disproportionation, the isotopic composition of the products (sulfate and sulfide) are first reported relative to the isotopic composition of the starting material, with the net fractionation (sulfate - sulfide) reported in gray. Calculated values of 33S, 36S, 33 and 36 are further reported for each sample. The isotopic compositions of the sulfate reduction experiments were normalized to the starting sulfate composition and, when possible, sulfur disproportionation experiments were normalized to the composition of the starting sulfur compound. In the case of the Golfo Dulce samples, where no initial sample was available, values were normalized to house UMd SF6, whose value is reported in Appendix 1.
Sulfate reducing microorganisms
For Desulfospira jorgensenii, the product sulfide was depleted in 34S and the residual sulfate was enriched with respect to the 34S values for the starting sulfate. For longer duration experiments, 34S fractionation between sulfate and sulfide (34SH2S-SO4) increased slightly. The 33SH2S-SO4 values did not change much with time, but did record a larger initial change (between time = zero and the first data recorded). The H2S-SO4 values for this experiment ranged from 0.5102 to 0.5103. Mixing lines between product sulfide and the residual sulfate intersected the initial
29
sulfate composition, therefore satisfying mass balance.
Results for experiments run with Desulfobacterium autotrophicum illustrate a sulfide reservoir with depleted 34S values and with an enriched 34S composition in the sulfate reservoir, when compared to the starting material. As the reaction proceeds, the net 34S fractionation (34SH2S-SO4) increases significantly. The 33SH2S-SO4 values record a slight variability, but overall do not record a statistically significant change. Net H2S-SO4 for this experiment evolved from 0.5102 to 0.5120. Mixing lines between the residual sulfate and product hydrogen sulfide intersect the starting composition of the initial sulfate and the data are therefore consistent with isotopic mass balance.
Table 2: Isotopic and biochemical data from sulfate reduction experiments with D. autotrophicum and D. jorgensenii. Note that isotopic compositions for time = 0 in Figure 4 are not listed, as that represents innoculum material. Isotopic data was calculated as the net fractionation between the sulfide and sulfate in an experiment, at the time listed. Isotopic uncertainties are fully discussed in Appendix 1. Cellular rates were calculated according to Detmers et al. (2001) and are for exponential phase growth. Isotopic results and calculations are discussed in the text, whereas the calculation of is presented in Appendix 4.
Sulfur disproportionating microorganisms
Two experiments with freshwater sulfite disproportionator Desulfocapsa thiozymogenes yielded fractionations of up to ~31 for 34SH2S-SO4 (Table 3). The 30
sulfide pool showed strong 34S depletions of ~23 , whereas the sulfate pool was only enriched by ~7 . The 33S values of the two reservoirs are more positive than the starting sulfite by up to 0.15 and the average SO4-H2S was 0.5163.
Two sulfite disproportionating experiments with Desulfovibrio sulfodismutans produced large fractionations for 34S, with an average 34SH2S-SO4 of ~50 between the sulfides and sulfates (Table 3). Longer duration experiments produced sulfides that were less depleted in 34S whereas the sulfates became less enriched in 34S. The 33S values in both pools recorded significant enrichments of up to 0.16 , relative to the starting sulfite. The average observed H2S-SO4 for this experiment was 0.5147.
Table 3: Isotopic results obtained for sulfur and sulfite disproportionation experiments. Data for every experiment, except those with the Golfo Dulce culture, which is normalized to UMd house SF6, are referenced to the isotopic composition of the starting material (listed under Sintermediate). The numbers listed under the organism name are the time intervals (in hours) that the experiment was run. Time intervals A and B for the Golfo Dulce samples represent different experiment durations with A preceding B. For all, the raw data is listed above and the total fractionations are boxed and presented in italics.
31
Two experiments with an elemental sulfur disproportionator from Golfo Dulce, Costa Rica produced smaller 34SH2S-SO4 fractionations, with an upper limit of ~29 (Table 3). I observe 34S sulfide values of 4 , with the complimentary sulfate pool recording a 34S enrichment of ~30 . The 33S values were depleted by up to 0.09 in the sulfates, and with SO4-H2S of up to 0.5187. Measurements of 34S highlight the large fractionations between products sulfide and sulfate that are produced by sulfur disproportionation (Thamdrup et al., 1993; Canfield and Teske, 1996; Habicht et al., 1998). The addition of minor isotope data also indicates variations in 33S, most notably between initial compositions and the products. When coupled, very large fractionations in 34S and measurable 33S allow for more precise determinations of , and hence greater resolution from RFL than for the sulfate reduction experiments.
Discussion/Conclusions
Interpretation of experimental data using flow networks
The sulfur isotopic fractionations between product hydrogen sulfide and source sulfate in a biological network depend on the structure of the network, the transfer of sulfur within the network, and the isotopic fractionations associated with the transfer pathways. Using the published network for dissimilatory sulfate reduction of Rees (1973), Farquhar et al. (2003) demonstrated that it is possible to account for variability in both and the 34SH2S-SO4 produced during the reduction of sulfate by A.
32
fulgidus. In this study, multiple sulfur isotope data from experiments with two other sulfate reducers (D. autotrophicum and D. jorgensenii) are additionally presented (Figure 5). The ranges of isotopic variations produced by these organisms yields H2S-SO4 from 0.5102 to 0.5120, which can also be represented by non-zero and negative 33S values, as is seen in Figure 5. These values are significantly less than
Figure 5: Plot of 33S versus 34SH2S-SO4 illustrating the effect of the Rees (1973) biosynthetic network (Equation 1 in text) on multiple isotopic fractionations between hydrogen sulfide and sulfate. Division of flow within the network is compactly described by f3, which represents the interaction between the enzymatic reduction of sulfite and the sulfate leaving the cell and f5, which is the proportion of back reaction from sulfite to sulfate. Relative fluxes (from 0 to 1 are noted by 0.4 and 0.8 on both the f3 and f5 contours. I include data from Farquhar et al. (2003) along with new data from this study, where the blue circles note D. autotrophicum and red circles are from D. jorgensenii. Assumptions in the construction of the model are values of 0.5145 for relating the fractionation in 33S to 34S. This value was selected on the grounds that it represents a low temperature equilibrium process. Fractionation factors used were taken from Rees (1973). See text and Farquhar et al. (2003) for further discussion of the modeling.
those predicted for equilibrium exchange between hydrogen sulfide and sulfate between 0 and 120 C (0.5146-0.5150), but similar to those predicted for
33
dissimilatory sulfate reduction (0.512) (Farquhar et al., 2003). The relationship between and the 34SH2S-SO4 reflects changes in the flow parameters of the sulfate reduction network (Equation 1). As described previously (Farquhar et al., 2003), the isotopic composition of experiments with A. fulgidus (culture grown as a continuous culture within a bioreactor) evolved along a contour that defines constant proportions of sulfur diffusion into and out of the cell (f3 0.4) and variable proportions of internal back reactions between sulfite, APS, and sulfate (f5 varies from 0.2 to 1.0) (see Figure 5). In Figure 5 I translate these arguments and data from both Farquhar et al. (2003) and this chapter to a 33S versus 34S coordinate system.
Experiments with D. jorgensenii grown in batch culture produced 34S fractionations typical of other sulfate reducing bacteria. This experiment evolved in a similar way to A. fulgidus along an extension of f3 (approximately 0.6), but produced fractionations that required f5 values slightly greater than 1.0 (Figure 5). Since it is not possible for f5 to exceed 1, these experiments suggest that at least one of the input fractionation factors associated with the internal network steps is slightly incorrect. For instance, it is possible to account for the data using the Rees (1973) network by increasing the fractionation factor for the internal back reaction (SO3= to SO4=(in)) to 35 in 34S (originally at 25 ). One possible explanation for the evolution along the f3 contour is that this organism is known to metabolize very slowly, as seen by the low specific rates calculated earlier. This may indicate a less dynamic sulfate exchange mechanism.
34
The isotopic relationships produced by D. autotrophicum evolve in a different way than A.fulgidus and D. jorgensenii, but the 34S fractionations are still consistent with sulfate reduction. This experiment follows a contour that defines a range of f5 from 0.6 to 1.0 and f3 that varies from 0.45 to 0.65 (Figure 5). I interpret variations in f3 as a response to the changes in the external sulfate concentration during the batch culture experiments, which is further investigated in Chapter 3.
This analysis illustrates the potential applications for this approach. Experimental data provides a critical check of the validity of the structure, the fractionation factors, and the flow parameters used in the proposed biological networks. At present, the data do not allow us uniquely to constrain these parameters in the networks, but future work that includes experimental determination of the isotopic compositions of intermediate sulfur pools has the potential to provide this information.
Experiments with elemental sulfur and sulfite disproportionating microorganisms
A similar approach is used to understand the sulfur isotopic fractionations produced by disproportionation networks. In the case of sulfur disproportionation, we interpret the relationship between the products (sulfide and sulfate) rather than between reactant and product. The modeling treatment of these disproportionation networks (see Equations 2 and 3) describes the general features that are observed in the experiments, but the calculations can be sensitive to the input parameters, such as and , which remain to be further constrained.
35
Sulfur isotope measurements from the disproportionation experiments produced 34S fractionations that range from 23 to 55 , consistent with the understanding of disproportionation. The experiments yield values between product hydrogen sulfide and product sulfate that range from 0.5145 to 0.5187 and 33S from 0.09 to -0.027 (Figure 6). The sample pairs with the lowest values of 33S (and ) produced the largest 34S fractionations. The models of Farquhar et al. (2003) produce similar values when there is insignificant back reaction (unidirectional flow) after the
Figure 6: Plot of 33S versus 34SH2S-SO 4 for disproportionation data from experiments with D. thiozymogenes, D. sulfodismutans, and a culture from Golfo Dulce, Costa Rica. Large 34S fractionations are typical of disproportionation reactions. Error bars represent 2 uncertainties. Values of 33S greater than zero are interpreted to reflect back flow along the oxidative branch of the disproportionation network (see text for discussion).
disproportionation branching point. For sample pairs with more positive 33S values, the 34S fractionations are smaller. I interpret these larger 33S values, in the context of the Farquhar et al. (2003) model, as indicating a back reaction in the oxidative branch (pathways 5 and 7 in Equation 2 and pathways 5, 7, and 9 in Equation 3) of 36
the disproportionation network. Although I view = 0.516 as a natural upper limit for mass-dependent processes, branching networks have been shown to produce inverse fractionation effects, or in this case > 0.516, associated with one product of a branching network (Fry, 2003). Back flow along the reductive branch of the disproportionation network conversely reduces the (yielding more negative 33S) values in a similar fashion to that of sulfate reduction (see Figure 5). The large range in values may indicate that internal back flow, primarily along the oxidative branch, is a highly variable characteristic of disproportionation biochemistry. Although the modeling approach provides insight into the first-order fractionation effects, I have been unable to reproduce the data using fractionations similar to those in the sulfate reduction network. The model can be forced to reproduce the data, but requires large (~10 in 34S/32S) fractionations associated with sulfite transport into the cell, and very large (40-50 in 34S/32S) fractionations associated with the sulfite-APS-sulfate steps. I feel that the modeling methodologies are resilient, and the focus of future work rests on determining the appropriate network for these metabolisms.
Geological and biological applications
Our experiments and their evaluation in the context of metabolic networks demonstrate that it is possible to distinguish between signatures produced by dissimilatory sulfate reducers and sulfur disproportionators, even when the magnitude of the isotopic fractionation is similar. Figure 7 illustrates the relationship between 33S values and 34S fractionations for hydrogen sulfide-sulfate pairs for experiments
37
involving dissimilatory sulfate reducers as well as sulfite and elemental sulfur disproportionators. Biological disproportionation processes produce 33S values that converge on equilibrium values for hydrogen sulfide-sulfate exchange (33S ~ 0)
Figure 7: Plot of 33S versus 34SH2S-SO4 that includes data from all experiments and the A. fulgidus experiment from Farquhar et al. (2003). Data from experiments with sulfate reducers plot at values of 33S that are between -0.12 and -0.02, while data from experiments with sulfite disproportionators and elemental-sulfur disproportionators plot at values of 33S from -0.06 to values greater than 0.05. At low 34S fractionations, such as those seen here, the X-axis is similar to that of equilibrium exchange between H2S and SO4 (Farquhar and Wing, 2003). Figures 6 and 7 illustrate two fundamental points. First, most metabolic fractionation processes cannot be approximated by equilibrium isotope exchange. Further, isotope fractionations by sulfur disproportionation are often resolvable from fractionations by sulfate reduction only by considering 33S in addition to 34S.
when 34S fractionations are large, but trends to 33S > 0 when 34S fractionations are less than approximately 35 . Fractionations between H2S and SO4 that are produced by dissimilatory sulfate reduction yield negative 33S values ( values between 0.5125 and 0.5103), significantly different from that produced by either equilibrium exchange (0.5147-0.5150) or biological disproportionation (0.5147) at
38
the same magnitude 34S fractionation, as both result in near zero 33S.
Under certain conditions, some sulfur disproportionating bacteria can operate as facultative sulfate reducers. In this scenario, I expect that the associated with reduction be similar to that of the obligate sulfate reducers. Primary disproportionators will only reduce sulfate as a means of survival under conditions where disproportionation is less favorable. In interpreting the geologic record, a sulfur disproportionating bacteria reducing sulfate should be just as diagnostic of the environmental conditions as a sulfate reducing bacteria reducing sulfate. The ability to distinguish isotopic fractionations produced by dissimilatory metabolisms from those produced by disproportionation metabolisms has potential application for testing hypotheses about the changing role of sulfur metabolisms in the terrestrial sulfur cycle over the course of Earths history. For instance, Shen et al. (2001) have argued that large 34S fractionations between pyrite and host barite in 3450 Mya samples of the Pilbara group of northwestern Australia reflect dissimilatory sulfate reduction by early bacteria. Moreover, it has been suggested that after 2450 Mya, the record of sulfur isotopes on a global scale cycle bears the signature of sulfate reducing bacteria and that the imprint of disproportionation did not become significant until after approximately 1000 Mya (Canfield, 2001; Canfield, 1998; also see Chapter 4). It has also been suggested that the large positive excursions for 34S following Neoproterozoic glaciations were caused by distillation of sulfur by sulfate reducing bacteria (Hurtgen et al., 2002). The findings suggest that determination of the multiple isotope fractionations for these and related samples may
39
provide a new test of these hypotheses as well as potential tests for evidence of metabolic activity and sulfur fractionations in samples relevant to astrobiological studies.
40
Chapter 3: Sulfur Isotope Fractionations Produced by Sulfate Reducing Bacteria: insight into isotope systematics and metabolic models
Introduction Metabolic models for fractionations produced by sulfate reducing bacteria and archea derived from experimental observations are the cornerstone of the interpretation of ancient and modern biogeochemical cycles. Although recent studies have called into question a traditionally accepted model, experimental evidence has been lacking for such a claim. I present data from all four sulfur isotopes that suggest that the internal fractionations associated with the sulfate reduction network are larger than previous estimates. Models of both a traditional sulfate reduction network, as well as a more recent incarnation of the sulfate reduction network (with multiple sulfur intermediates) are constructed to aid in the understanding of new experimental data. These data also allow for the further development of additional minor isotope relationships, one which is easily measurable in geologic settings and accurately depicts the net effect of an environment, whereas the other is more applicable to modern environments and may better illuminate the specific process(es) controlling the fractionation in those environments. This approach illustrates the uses of systems containing more than two isotopes.
41
Background Conventional sulfur isotope fractionation models for microbial sulfate reduction underpin the interpretation of the geologic sulfur isotope record and in turn, the history of Earth surface oxidation (Berner, 1991, Claypool et al., 1981; Canfield, 2001). The antiquity of sulfate reduction has been traced into the early Archean (Shen et al., 2001), and sulfate reduction is thought to have played a critical role in shaping the Earths surface sulfur cycle. Estimates on the magnitude of sulfur isotope fractionation produced by sulfate reducers rest on experimental observations (Peck, 1959, 1961; Castro and others, 2000; Detmers and others, 2001; Canfield, 2001) and decades-old theoretical frameworks (Harrison and Thode, 1958; Kaplan and Rittenburg, 1964; Kemp and Thode, 1968; Rees, 1973; Chambers and others, 1976, Canfield, 2001). The traditionally applied limit, which has been used in models of the Precambrian and Phanerozoic sulfur cycles (Berner 1991; Canfield and Teske, 1996; Canfield, 1998; Habicht et al., 2002), is ~ 47 in 34S/32S (Rees, 1973; Canfield, 2001). However, recent measurements of natural environments (Rudnicki et al., 2001; Wortmann et al., 2004) and new models (Brunner and Bernasconi, 2005) suggest that the prior estimate may underestimate the fractionation associated with sulfate reduction by as much as 30. If valid, a recalibration of most conventionally accepted sulfur cycle models might be in order. Here, I use a recently developed approach (Chapter 2; Farquhar et al., 2003; Johnston et al., 2005) to investigate isotopic fractionations associated with sulfate reduction. While this approach does not provide new limits for the total fractionation a sulfate reducer can produce, it does
42
provide a way to place limits on the magnitude of fractionation for the internal steps within the metabolic networks.
In the present study, I analyze and interpret new measurements of the sulfur isotopes (32S, 33S, 34S, 36S) from sulfate and sulfide associated with pure culture sulfate reduction experiments. This study, therefore, builds on previous research, which has highlighted the use of 33S for differentiating fractionations produced by sulfate reducing and sulfur disproportionating microorganisms (Chapter 2; Farquhar et al., 2003; Johnston et al., 2005). The experiments were designed to highlight temperature-induced physiological shifts driving the magnitude and pathways of fractionation. These results, when coupled with the associated modeling, allow us to explore the maximum isotopic fractionation capability of the sulfate reduction metabolic network. In addition, the present study will: (1) provide insight into biochemical implications of different sulfate reduction models, (2) pinpoint, and make predictions about future research experiments that can further address the questions discussed/raised in this study, and (3) elucidate the character of the massdependent relationship between 33S and 36S.
More on mass-dependence
Multiple isotope signatures, like those examined in this study, are a function of massbalance (mixing) and the specific processes producing the fractionation, each of which contribute to the observed effect. What determines whether a signature is mass-dependent (or non mass-dependent) is the specific process that is producing the
43
fractionation. In the present study, microorganisms are responsible for the observed fractionations, and similar to most Earth surface processes, they produce massdependent isotope effects. To describe the exponent that characterizes the massdependent fractionation in a system, I again use , as defined in Chapter 2, which describes the observed relationship between a pair of samples (Mook, 2000; Farquhar et al., 2003; Johnston et al., 2005). This expression can also be modified to describe a suite (more than two) of samples. These values fluctuate around the reference exponents (0.515 and 1.9 for 33S-34S and 36S-34S respectively) for a number of massdependent processes. For instance, variability in this term (on the order of a few tenths of a percent) arises for equilibrium isotope exchange at a range of geologically relevant temperatures, whereas other processes, like diffusive isotope separation and chemical kinetic isotope exchange (ie. going through a transition state), may deviate from the reference values by a few percent (Mook, 2000; Miller, 2002; Young et al., 2002; Farquhar et al., 2003; Farquhar and Wing, 2003). All of the above listed fractionations are still considered to be mass-dependent. The primary interest in this study is isolating the mass-dependent fractionations directly attributable to the catabolic reduction of sulfate. To do so, I quantify and subtract the effects associated with overall mass-balance in the batch experiments, leaving behind the fractionation directly attributable to sulfate reduction (see Appendix 2). Since the variability in is small, I carefully describe uncertainties associated with these calculations using standard error propagation techniques (see Appendix 4).
44
Metabolic models
A common approach to understanding the complex fractionation paths and processes within biological systems is through the construction of steady-state box models (Chapter 2; Hayes, 2001; Farquhar et al., 2003; Johnston et al., 2005). In contrast to traditional isotope studies that use two isotopes (13C, 18O, 34S), consideration of multiple (>2) isotopes of the same element can provide better constraints on how material flows through a system. For example, Farquhar et al. (2003) and Johnston et al. (2005) (Chapter 2) used the multiple sulfur isotopes to describe material cycling in both sulfate reducers and sulfur disproportionators. The treatment here will expand on those studies, specifically the sulfate reduction model, with the calculation of new sulfate reduction networks and the inclusion of 36S. All of the fractionation results presented below were normalized to the isotopic composition of sulfate (ie. the coexisting sulfate always plots at the origin).
Predictions from Rees (1973): The classic biological pathway employed for dissimilatory sulfate reduction, as reviewed below, was originally proposed by Harrison and Thode (1958) and expanded upon by Rees (1973) and is discussed in Chapter 2:
= 1 4(out ) 2
SO
SO= 4(in )
4 a 5 a
APS
4 b 5b
3 H S . SO 2
= 3
(13)
The four primary steps and 34S fractionations within this model are 1) transport of sulfate across the cellular membrane (341=1.003), 2) the activation of inorganic sulfate to adenosine 5 phosphosulfate (APS), 3) the reduction and subsequent
45
cleavage of the sulfate bound in APS (344b=0.975), and 4.) the reduction of sulfite to hydrogen sulfide (343=0.975). If I take these fractionations as additive, this network results in an upper estimation of the 34S fractionation capability of sulfate reduction at 34S ~ 47. This estimate is consistent with observations of fractionations produced in pure culture and natural population experiments (Harrison and Thode, 1958; Kaplan and Rittenburg, 1964; Kemp and Thode, 1968; Habicht and Canfield, 2002; Canfield, 2001; Detmers et al., 2001).
Predictions for a trithionate pathway: Recent work (Rudnicki et al., 2001; Wortmann et al., 2004) suggests that the fractionation capability of sulfate reduction may exceed the 47 upper limit proposed by Rees (1973). Explanations for higher fractionations typically involve the inclusion of the trithionate pathway (Kobayashi et al., 1974; Fitz and Cypoinka, 1990; Agaki, 1995; Cypoinka, 1995), whereby the reduction of sulfite is not unidirectional and complete, but proceeds through a series of sulfur intermediate species (S2O3, S3O6). These steps are described as reversible, and according to some models are viewed as incomplete. Further, one study (Brunner and Bernasconi, 2005) includes the presence of an internal hydrogen sulfide reservoir, which could either react back towards sulfite or be transported out of the cell1. These considerations result in a variety of potential flow networks, one of which is:
1
Although shown without transport of external sulfide back into the cell, this pathway was included in Brunner and Bernasconi (2005). If there is no fractionation associated with the back reaction, as was suggested by Brunner and Bernasconi (2005), the modeling can simplify to the network shown in equation 14.
46
SO
= 1 4(out ) 2
SO= 4(in )
4 a 5 a
APS
4 b 5b
SO
= 3 int
7
6
8 9
S
3 H S . (14) H 2 S(in ) 2 (out )
Brunner and Bernasconi (2005) suggest a different fractionation for the reduction of SO3= to H2S(in), and use equilibrium isotope exchange calculations presented in Farquhar et al. (2003) to suggest 34S fractionations of 341=1.003, 344b=0.975,
34
6 = 0.947 and 347 = 1.005. This network would allow an upper 34S fractionation
limit of ~ 75, far exceeding those observed in pure culture and natural population experiments. From a modeling perspective, the expansion of the model network requires the introduction of a third flux term, which monitors the back reaction along pathway 7 (f7= [flux 7]/[flux 6]) (see Chapter 2). Similar to the modeling presented above, the flux terms vary from zero to one, and for this network are f3, f5, and f7 and are defined as before.
Biological methods A pure culture of D. autotrophicum (DSM strain # 3382) was obtained from Deutsche Sammlung von Mikroorganismen, Braunschweig, Germany and grown according to recommended media recipes. Media solutions were prepared anaerobically with a N2 atmosphere and at sulfate and butyrate concentrations of 28 mM and 20 mM respectively. All cultures were grown in batch. Cell growth was monitored through optical density measurements at 670 nm. Calibration of optical density measurements was performed by staining cells with DAPI (4, 6-diamidino phenylindole) and counting under UV light. Cell-specific rates of sulfate reduction (csSRR: see Chapter 2) during the exponential phase were determined according to Detmers et al. (2001). 47
New experiments were inoculated from growing cultures in the late exponential growth phase, as determined by earlier establishment of reproducible growth curves and subsequent optical density measurements.
Experiments were run in a temperature gradient block (2-38C) and in quadruplicate (time series of four). Metabolic activity was terminated and sulfide captured through the addition of a 20% Zinc acetate solution. Samples were then frozen until the necessary chemical extractions could be performed. Sulfate concentrations ( 1 % 1) were determined via ion chromatography with the addition of a bicarbonate buffer. Sulfate was precipitated out of solution as BaSO4, through the addition of 1mL 1M BaCl. Sulfide concentrations ( 5 % 1) were measured photometrically using a methylene blue technique described in Cline (1969). Sulfide was later distilled using 3N HCl to liberate H2S, which is captured in an AgNO3 trap as Ag2S. All isotopic methods, uncertainties, and normalizations can be found in Appendix 1.
Results Concentrations of sulfate and sulfide were monitored in the experiments and covaried. Cell specific sulfate reduction rates vary as a function of temperature (between 2 and 38C), with rates slowing to near zero (> 15 fmol/cell*day) as the end-member temperatures were approached. The maximum cell specific reduction rate was ~ 150 fmol/cell*day at 22C. After the effects of Rayleigh distillation are removed (Appendix 2), a range in the net sulfur isotope fractionation (sulfide sulfate) is observed, from -36.36 to -16.03, 0.062 to 0.204, and -2.13 to -0.11 for 34S,
48
Table 4: Experimental sulfate reduction net fractionations. Points a and b represent a time series, with point a preceding b. Rates are calculated according to Detmers et al. (2001) and reported in units of fmol/cell*day. Isotopic fractionations are determined by relating the ratio of sulfide to sulfate (Equation 4) and then converting to isotopic compositions (Equations 2-4). Raw data is presented in Table A1. Samples labeled as NA represent temperatures where isotopic data (for either sulfate or sulfide) was unavailable. Uncertainties are listed in the text and further discussed in Appendixes 1 and 2.
49
33S and 36S respectively (Table 4; Appendix 2). The minor isotope relationships of 46 pairs of samples range from 33SR = 0.5077-0.5125 and 36SR = 1.905-1.974. I further present the data as 33S vs. 36S, where the data are defined by: 36S = -8.98 ( 0.47) *33S -0.029 ( 0.041).
Discussion Cell growth and sulfate reduction rates: The cell specific rates reported here are slightly higher than the one value reported in Detmers et al. (2001) for D. autotrophicum grown on an organic substrate, but still well within the range reported for other sulfate reducers. Previously reported optimal growth temperatures for strains of D. autotrophicum extend from 25-28C, whereas this culture grew optimally at 22C. Metabolic activity and cell growth continued above this temperature (to 34C), as is typical (cf. Hoek et al. 2006). The data follow the empirical relationship between csSRR and fractionation observed in previous studies (cf. Habicht and Canfield, 1997; Hoek et al., 2006) where csSRR and fractionation are inversely related. This is noted below as the conventional relationship. The data from Detmers et al. (2001) also falls within the values reported in this study. The conventional relationship (Figure 8) is thought to be indicative of the degree of material transfer through the cell, with high rates suggesting near quantitative transfer through the cell (small net fractionations) and low rates giving way to a more restricted flow path, that would allow for the more complete isotopic expression of the internal fractionations (higher net fractionations). This relationship has been
50
shown not to be universal, however (Canfield et al., 2006), and the data from the most extreme, and two optimal temperatures within the gradient block do not fit this model (see ellipses in Figure 8).
Figure 8: A plot illustrating the relationship between fractionation (34S) and the cell specific rate of sulfate reduction (csSRR, fmol/cell*day). Isotopic errors are smaller than the symbols. The data is consistent with a previous observation (Detmers et al., 2001) and shows the inverse relationship thought to exist for pure cultures between these parameters. However, samples from the upper two most and lower most temperature, as well as two optimal temperature points, stand in marked opposition to this trend. Studies of natural environments (Habicht et al., 2001) and a recent experimental study (Canfield et al., 2006) have challenged the universality of the inverse relationship between 34S and reduction rate. The outlying data support the lack of uniformity to this relation, and point to a physiological response in the cell as causing the deviation from the inverse trend, which is observed in the remainder of the data.
Towards the end-member temperatures (and in more extreme conditions in general), the ability for the cell membrane to stabilize is thought to be a function of the relative saturation and chain length of the fatty acids comprising the cell membrane (Scherer and Newhouse, 2002; Canfield et al., 2006). In bacteria, higher temperature generally
51
results in saturation, whereas colder temperatures begin to favor unsaturated bonds. Thus, as temperatures increase and membrane bound proteins become more loosely retained; the cellular response is to increase the proportion of saturated fatty acids in order to stabilize the membrane (Rabus et al., 2002). Organisms can also modify the chain lengths of the membrane fatty acids in order to maximize membrane fluidity.
Temperature-dependent rate and fractionation: A capability of a cell to fractionate isotopes is likely variable (cf. Harrison and Thode, 1958; Kaplan and Rittenburg, 1964; Kemp and Thode, 1968; Habicht and Canfield, 1998; Canfield, 2001; Detmers et al., 2001), as may also be the dependence of fractionation on rate (as suggested by the conventional trend noted in Figure 8). As purely a thought experiment, each of these relationships is considered in turn, where we first address the relationship between csSRR and temperature (Figure 9a). As is expected, the rate calculated from the experiments increased with temperature, reaching a maximum at ~ 22C. Above this temperature, the reduction rates begin to fall and approach zero at the upper temperature of 36.4C. Figure 9b illustrates the relationship between fractionation (in 34S) and temperature. This figure presents the calculated average fractionation measured at each temperature (average of points a and b; Table 4). No systematic relationship between temperature and fractionation is observed (the slope of the trend line is not resolvable from horizontal), as the fractionation remains constant throughout. These experiments thus differ from the expectation, which suggests temperature dependence. Of interest is also trying to understand how these two parameters (fractionation and rate) are related. In order to investigate this
52
relationship, the average fractionation (: Figure 9b) is combined with the measured reduction rates (csSRR: Figure 9a). The goal of this new expression is to examine the co-variation of these parameters (Figure 9c).
If the thought experiment is continued, the relationship plotted in Figure 9c can be used to try and better understand the outlying samples from Figure 8. First, the fractionation/rate term is interpreted as follows. When no change is observed in this ratio (a horizontal line in Figure 9c), the two variables can be seen to be changing in concert (positively correlated). This could also be interpreted as a condition where the machinery responsible for the changes in rate are associated with steps that fractionate isotopes. The converse would also hold true. As seen in Figure 8, there are two areas where the conventional relationship is not upheld: at the temperature extremes and at the optimal growth temperature (though the optimal growth temperatures could loosely be agued to be within the trend established by the remainder of the data). Comparing Figure 8 and 9c allows for an interesting distinction to be made. The optimal temperature samples fall slightly outside the Figure 8 trend, but fit well with the trend observed in Figure 9c in which they show no variation. Conversely, at the temperature extremes where the trend in Figure 8 is also not met, the relationship in Figure 9c shows a great deal of variability.
53
54
Figure 9: A series of figures illustrating the relationship between temperature and csSRR (A), fractionation (B), and a ratio of the two symbols (C). A) A plot of temperature versus cell specific sulfate reduction rate, which is calculated during the exponential phase of growth at each temperature. The data follow the expected distribution, with lower rates towards the extreme temperatures and a maximum rate towards the upper end of the growth range (~ 22C). Maximum rates are similar to those reported in Detmers et al. (2001). B) A plot of temperature versus average fractionation (in 34S). Here, I average the isotopic composition of points a and b seen in Table 1. I do not observe a temperature dependence of isotopic fractionation. C) Temperature plotted against a new relationship, the ratio of the average isotopic fractionation (34S) over csSRR. This relationship allows for investigation directly into the co-variation of rate and fractionation.
I take this simple observation to suggest two preliminary conclusions. First, the cause of the deviation from the conventional trend at the extreme temperatures is different than that at the optimal temperatures (one trend is flat in Figure 9c, whereas the other is highly variable). The positive co-variation implied by the horizontal trend in Figure 9c suggests that at the optimal temperatures, the machinery controlling sulfate reduction is tied to one (or more) of the significant fractionation steps (perhaps an enzymatic response). Conversely, at the temperature extremes where the relationship between fractionation and rate is not steady (a non-horizontal trend in Figure 9c), the deviation is likely related to a biochemical change not directly dependent on the fractionation steps associated with sulfate reduction. This may again point to a cell membrane dependency (or metabolic component completely unrelated to the dissimilatory process) at these temperatures. Though this application and these results are preliminary and perhaps speculative, I feel that this style of treatment, coupled with a detailed molecular approach (for instance, tracking gene regulation), should be considered in future studies.
55
56
Figure 10: Sulfur isotope model for the consequences of the Rees (1973) network for dissimilatory sulfate reduction (Equation 9). The flux terms seen in these models are described and defined in the text, with f3 representing mass balance on the entire cell, whereas f5 constrains the backflow of material from sulfite to sulfate. See Figure 7. a.) The model cast in 33S vs. 34S (taken from Farquhar et al., 2003; Johnston et al., 2005) with the new experimental data (red diamonds) and the recalculated previously published data (see Appendixes 1 and 5; squares = Farquhar et al., 2003; circles= Johnston et al., 2005b). The new data illustrates the inconsistency with the model prediction. b.) Same model cast in terms of 36S vs. 34S. Previous experimental data has been included where 36S data is available (see Appendix 4). As was seen in Figure 9a, the model does not fit the observation.
Evaluating metabolic models
The Rees (1973) model: The standard model (Rees, 1973) for dissimilatory sulfate reduction entails a variety of enzymatically catalyzed steps and fractionations and suggests a uni-directional, single step reduction of sulfite to sulfide. Previous work (Farquhar et al., 2003; Johnston et al., 2005) used the Rees (1973) network to describe the anticipated fractionation in 33S. Those authors presented a model and illustrated a consistency between the model prediction and the previously collected fractionation data. In Figure 10a and 10b I model the predicted isotopic compositions for the Rees (1973) network and include the data from this and previous studies. The new fractionation data, which demonstrate larger 34SH2S-SO4 fractionations than previous experimental data, fall outside the field calculated using the Rees (1973) network and the assumption that 33 and 36 for the internal sulfur cycling steps are 0.5145 and 1.90 respectively. This observation suggests that the assumption that the values for the internal fractionation steps can be approximated by equilibrium is not appropriate. The new data then suggest that either 1) the Rees (1973) network does not adequately represent the sulfate reduction metabolism, 2) one or more of the assumptions in the isotope model are invalid, or 3) a combination of both. The Rees (1973) model provides both the network and the 34S fractionations that I use here.
57
The changeable free parameters in the model, which may be adjusted to fit the data are the input major and minor isotope fractionations (, ) and the network structure. This treatments uses input 34's assigned by Rees (1973) and values that represent low temperature thermodynamic equilibrium. Variations in have been proposed for some processes, such as diffusion and chemical kinetics (Craig, 1968; Young et al., 2002), but no single set of values can account for the ranges observed in 33S and 36S. This means that, if 34 is held constant, no one single set of values defining the internal fractionation steps can satisfy all the data. In order for the data to be satisfied with the Rees (1973) 34 values, the input values would have to vary within the experiments, perhaps as a function of temperature. I suggest that at least one of the internal fractionation steps (either one of the enzymatic steps, the diffusion step, or some component of both) may be underestimated in 34 and/or has a temperature-dependence in , which could explain the observation.
A modified Rees (1973) model: The modified Rees (1973) network consists of the Rees (1973) network structure, but the fractionation factors (34) are modified to reflect low temperature equilibrium processes rather than those fractionations selected by Rees. Following in the spirit of Brunner and Bernasconi (2005), I have assigned equilibrium fractionation values calculated from vibrational spectra (for 33S, Farquhar et al., 2003 is used, however all values are listed in Table 5). The changes from the Rees (1973) fractionations are in the steps associated with both the reduction of sulfate to sulfite, and sulfite to sulfide. I adopt fractionation for these steps using equilibrium values at 0, 10, 20 and 30C. These values were chosen since they span
58
the temperature range of the experiments. I am not proposing a new fractionation regime for sulfate reduction, but rather illustrating how the models respond to different assumptions about the internal fractionations.
Table 5: Listed are 34, 33 and 36 values calculated between common sulfur compounds at low temperatures. Values of 34 and 33 are expanded from Farquhar et al. (2003), where they are noted as . All additional values are calculated in the same fashion as described in Farquhar et al. (2003) and are thus based on low temperature gas phase equilibrium.
As presented in Figures 11a and 11b, these larger fractionation factors allow a majority of the data to be explained. As the temperature increases, the net fractionation prediction becomes smaller in 34S H2S-SO4, as well as in 33S H2S-SO4 and
59
60
Figure 11: Here I present a model for a modified Rees (1973) network that adopts the structure seen in Equation 9, but using equilibrium fractionations associated with low temperature thermodynamic equilibrium calculations (see Farquhar et al., 2003 for 33S, this study, Table 5, for 36S). The definition of the flux terms is the same as seen in Figure 5. I plot the entire model for fractionation values at 30oC, but include the limits at temperatures of 0, 10, and 20C. The data are grouped to match these temperature choices. a.) The model cast in 33S vs. 34S. The increased 34S consequences of this model allow for a majority of the data to be described, with a 34S limit of ~ 70. b.) Same model cast in terms of 36S vs. 34S. Previous experimental data has been included where 36S data is available (see Appendix 4). As was seen in 33S, this model can account for most of the 36S fractionations observed in the experimental data. I note, however, that most of the data is within 2 of the model prediction but the locus of points falls outside the proper field. This modified network is presented to illustrate a point, and not to propose a modification of the Rees (1973) network.
36S H2S-SO4. In Figures 11a and 11b, the data has been binned by temperature to demonstrate the match (or mismatch) between these specific fractionation predictions and the effect of temperature in the experiments. The match between the data and the theoretical temperature prediction is fairly robust, as the grouping of data generally follows the temperature predictions. Only the lowest temperature data (0-5C) do not fit this trend. Overall, this exercise presented a better fit in 33S than 36S. In Figure 11a, 100% of the data fall within the model prediction whereas in Figure 11b, only 81% are consistent within the prediction, but 95% fall within 2. While there is no indication of an analytical artifact that is biasing the 36S measurement, this isotope is difficult to measure and all possible explanations are still considered. Longer-term studies between multiple labs will help to address this issue. The modified Rees (1973) network illustrates that, to a degree, all which is required to satisfy the data are larger internal fractionations as mass-balance can account for much of the observed variability (see Appendix 1 for more details). The requirement is that the 34S fractionation be larger (if 33 is ~ 0.515) in order to produce the observed 33S signature, even though the large 34S fractionation implicit in such a situation may go
61
unexpressed (holding 33 constant, the larger the 34S fractionation the larger the potential 33S).
An active trithionate model: Discussions of the metabolic pathway employed by dissimilatory sulfate reduction have long included the potential presence of sulfur intermediate species (cf. Kobayashi et al., 1974; Fitz and Cypoinka, 1990; Agaki, 1995; Cypoinka, 1995). In particular, the trithionate pathway suggests a step-wise reduction from SO3 to H2S (Equation 14). In Figures 12a and 12b I compare model predictions for an active trithionate pathway with the data. As can be seen, when I incorporate fractionation factors suggested by Brunner and Bernasconi (2005), this model easily satisfies the 33S measurements and a majority of the 36S data. This calculation of the pathway ascribes the trithionate associated fractionations between sulfite and the sulfur intermediate pool (fluxes 6 and 7 in Equation 14). A different formulation in Brunner and Bernasconi (2005) suggests that the tritionate-associated fractionation should be split equally between the sulfite to sulfur intermediate step (fluxes 6 and 7 in Equation 14) and the sulfur intermediate to sulfide step (fluxes 8 and 9 in Equation 14). The consequence of this change on the model is in the shape of the lower limit (boundary nearest the x-axis). This boundary line would shift such that a new cusp (convergent point of concave down bounding lines) would form with spacing directly related to the size of the two new fractionation steps. In short, the overall magnitude (in 34S) of the predicted field would not change, but the shape of the lower boundary line would. It can be seen that the predictions for the isotopic consequences of an active trithionate pathway are similar to that of the modified
62
Rees network, further illustrating that the primary parameter controlling the model bounds and accounting for the data is 34.
The consistency between the data and the trithionate pathway model is encouraging, and warrants a more detailed examination into both the model implications and the biochemical consequences. Ideally, the data would provide a diagnostic indication of the presence of this pathway, however, I am only left with the observation that a pathway with larger fractionations than Rees (1973) can satisfy the data. The adoption of the trithionate pathway, or any network with larger net fractionations, is met by a formidable body of pre-existing experimental evidence on pure cultures and natural populations of sulfate reducers illustrating fractionations that do not exceed 34SH2S-SO4 = 47 . As written here and presented in Brunner and Bernasconi (2005), the sulfate reduction network should be capable of producing fractionations up to 34SH2S-SO4 ~ 75 . It remains to be seen whether the specific flow conditions required to produce these large fractionations occurs (or are possible) in natural settings or communities. This is a target for future study.
Similar to a treatment presented in Canfield et al. (2006), the model approach that this study uses provides a means of directly testing one portion of the trithionate prediction: the reversibility and presence of an oxidative sulfur intermediate pool (such as thiosulfate). The present data set is not ideal for such a treatment, but to make a prediction and demonstrate the application of such an approach, this test is outlined. It would be expected that as sulfate reduction proceeds, the electron donor
63
64
Figure 12: Sulfur isotope model for the consequences of a sulfate reduction network with an active trithionate pathway, as presented in Brunner and Bernasconi (2005) (Equation 14). The new experimental data (red diamonds) along with previously published data (yellow squares = Farquhar et al., 2003; blue circles= Johnston et al., 2005b). a.) The model cast in 33S vs. 34S. This model can describe all the experimental data, and extends to a much larger 34S limit of ~ 75. b.) Same model cast in terms of 36S vs. 34S. Previous experimental data has been included where 36S data is available (see Appendix 4). As was seen in 33S, this model can account for a majority of the 36S fractionations observed in the experimental data (within error).
for the reaction (in this case butyrate) would become depleted. Experimental evidence suggests that as the electron donor becomes more limiting, the concentration of sulfur intermediate species increases (Chambers and Trudinger, 1978; Fitz and Cyponkia, 1990). If true, it would be expected that with the increased concentrations of S-intermediates, the back reaction from the intermediate to sulfite would become more likely.
The same principle would hold true for the presence of an internal hydrogen sulfide pool. For the model presented in Equation 14, the back reaction from the Sintermediate to sulfite has an associated 5 effect in 34S, which would suggest that as the intermediate accumulates and begins to back react, the net fractionation would increase by 5 . The increased fractionation in the model is represented by the variable f7, which constrains the material flow along the back reaction. So, as f7 varies from 0 (no back reaction) towards 1 (full back reaction), there is an isotope effect of 5 . Future experiments could be designed to target the expression of this biochemical step in the sulfate reduction process as a test of the presence of a trithionate pathway.
65
Comparing metabolic models: While the data do not uniquely support one or another network structure, if it were possible to extend the measurements of the four sulfur isotopes to the internal sulfur reservoirs, I might be able to construct a framework for evaluating the different networks. Figure 13 illustrates the predicted isotopic compositions of the internal sulfur pools (sulfite and sulfate) within the Rees (1973) network and defines a testable hypothesis for the validity of this network, provided that the isotopic composition of one internal sulfur species can be obtained. Similar predictions were made for 34S by Canfield et al. (2006). Although Rees (1973) is the only network considered, this approach can, in principle, be applied to any network and for any internal species. To illustrate this point, plotted are the predictions for f3 = 0.6 and f5 = 0.8, with the external sulfate constrained to the origin. Predictions for internal species in an active trithionate pathway are more difficult to uniquely constrain given that the model has three independent parameters.
A number of noteworthy points can be derived from the example given in Figure 13. First, the compositions of sulfide and sulfite track one another, always offset by 34SSO3-S = 25 on a slope (33SO3-S) of ~ 0.5145. One example of this is demonstrated through the example included on the model fields, which are offset by 25 on a slope of 0.5145. This reflects the input network structure, where the sulfite is assumed to be uni-directionally reduced to sulfide via the sulfite reductase enzyme. Conversely, the relationship between internal sulfate and sulfite is not constant, due to the reversibility of this reaction and the added flux and fractionation of sulfur moving
66
into and out of the cell as sulfate. As shown in Figure 13, the 34SSO4(int)-SO3 ~ 20
Figure 13: Modeling treatments also allow for the prediction of the isotopic composition of the internal sulfur pools in a network. Plotted are the solutions for the Rees (1973) network (seen in terms of 33S) where the composition of the external sulfate (plotted at origin), internal sulfate (dashed field), internal sulfite (gray field) and external sulfide (black field) are predicted. The contours within the model fields present relative fluxes, similar to that presented for the Rees (1973) network. As an example, plotted are the prediction for f3 = 0.6 and f5 = 0.8 to illustrate how the composition of the three fields should co-vary. The measurement of these reservoirs is an ideal target for future assessments of the sulfate reduction network.
with 33SO4(int)-SO3 ~ 0.5138. All the 33 values for the relationship between the internal pools vary around a value of ~ 0.5145, consistent with the original definition of 33 (= 340.5145) in the model. The relationship between the external sulfate reservoir and the internal pools varies more widely, since all these relationships incorporate a variety of steps and influences. This is largely due to internal recycling of material, and predicts a range of 33 values from 33SO4(ex)-SO3 = 0.521 to the net sulfate reduction fractionation of 33SO4(ex)-S (or 33SR) = 0.510.
67
The relationship between 33S and 36S
Defining (or developing) a tool to distinguish between two (or multiple) processes contributing to a common natural record would greatly increase the understanding of natural systems. I look to the relationships of all four isotopes to provide this tool, and pose two questions: Does the fractionation observed in 33S and 36S vary in a predictable fashion? and, If there is variability in these signatures, is it related to process? Until now, the difficulty in addressing these questions has been two-fold: analytical precision has been insufficient for the task and there has been a lack of data. Recent advancements in analytical capabilities now allow for precise measurements of 36S, and this, coupled with the feasibility of collecting large data sets (98 measurements in the current chapter), allows these questions to be examined. I approach this problem from two directions. First, I evaluate the relationship between 33S and 36S, which is dependent on the minor isotope fractionation and on the magnitude of fractionation of 34S/32S. Second, I examine the relationship between
33
and 36, which are the exponents that describe the fractionations. These terms are
calculated independent of the magnitude of the 34S fractionation and this approach may allow for the character of different fractionation mechanisms to be directly compared.
33S vs. 36S: Recent studies have suggested that relationships relating 33S and 36S
may preserve information about mass-dependent geochemical cycling (Ono et al., 2006; Johnston et al., 2006: Chapter 5). Models incorporating single-step equilibrium fractionations and mixing/distillation have been presented to argue for a linear
68
relationship with 36S = -6.85*33S (Ono et al., 2006). Modeling of Paleoproterozoic sediments with multi-step mixing and distillation processes indicate that the relationship between 33S and 36S cannot be described by a single line, but that the approximation of 36S ~ -7*33S may be reasonable (Johnston et al., 2006: Chapter 5). The interest here is to define the character of the co-variation between 33S vs. 36S for a single process rather than entire environment. In addition, the goal is to determine whether there exists a unique relationship for sulfate reduction that differs from or is similar to that previously established for global or basin scale geochemical systems, and in doing so to define a range of values that can be compared to other biological, chemical, and physical fractionation processes.
Our first attempt to define a range of proposed predictions begins with the results of the modeling treatments discussed above, meaning that I can cast the models seen in Figure 10-12 in terms of 33S vs. 36S. All of these models are built with the a priori assignment of the relationship between 33S and 36S (the input exponent values of 0.5145 and 1.9 respectively), which reflects low temperature equilibrium fractionation. As can be seen in Figure 14a (model prediction for the trithionate pathway), the model does not predict a single line, but instead reveals a narrowly defined region. The choice of 0.515 (rather than 0.5145) would result in the 33S vs. 36S model prediction collapsing onto a single line, since 0.515 and 1.90 are the values used in the definition of the values. The model results then indicate that values other than 36S = -6.85 33S, the theoretical prediction of Ono et al. (2006), can be produced when either a non 0.515 (or 1.90) exponent (or process reflected by
69
the exponent) or mass-balance effect, such as those described in Appendixes 1 and 2, are present.
Figure 14: Plotted is the model solution for the active trithionate pathway (A) and the data from this study (B), in terms of 33S and 36S. Uncertainties in 33S are smaller than the symbols. The models do not suggest much variability due to the a priori determination of the relationship (see text), and generally follow a slope suggested by equilibrium (-6.85). Although a majority of the data falls near this model prediction (or even within uncertainty), the fit is not ideal as the general trend of the data and model differ.
Figure 14b and Figure 15 present all the data in terms of 33S and 36S (Table 4). Each data point represents a sulfate-sulfide pair, such that for every data point on the figure there is an accompanying data point at the origin. The array extending through the data follows the net relationship: 36S = -8.98 ( 0.47) 33S -0.029 ( 0.041), which is > 2 from the equilibrium prediction (36S = -6.85 33S). If broken down in more detail, the slope values for each of the different temperature clusters falls within 70
1 of the net relationship defined above, and ranges from 8.74 to 9.27. It is taken from this observation that the net sulfate reduction signature (or small change in slopes) deviates slightly from equilibrium predictions and is independent of temperature. I postulate that the slope defined by the data represents the net process of sulfate reduction, rather than a single physiological step within the sulfate reduction network. This is suggested because the data cover a range of temperatures (2-38C), ruling out the possibility of a single physiological condition controlling the signature. The overall fit of the model predictions is not perfect, (cf. Fig 14) but does illustrate how natural variability, like that seen in the experiments, can arise (cf. changing internal ).
In addition, Rayleigh effects (see Appendix 4) can alter the observed relationship between 33S and 36S. Prior to the removal of the Rayleigh effects, the 33S and 36S co-variation was significantly different, and suggested a value of 36S = -10.32 33S. I document, therefore, that closed system effects can drive the 33S vs. 36S slope towards more negative values, as is demonstrated in the experiments. This suggests that both process and mass-balance can contribute to the overall observed 33S vs. 36S slope. This lesson should be carried into studies of natural basins and marine settings, where contributions are more difficult to distinguish and where the possibility of closed system behavior is more prevalent. In addition to providing valuable insight into the role of closed system behavior, this allows us to use this data to begin to place limits on the anticipated range of values measured in natural settings as extending up to ~ -11 (similar to the observed, uncorrected data).
71
Figure 15: All the data plotted as a function of the minor isotope composition (33S and 36S). A mass-dependent (MDF) prediction from Ono et al. (2006) as well as a mass-independent (MIF) prediction (from Farquhar et al., 2000) are included for reference. The data all fall to the lower right (quadrant 4), with an equal number of samples plotting at (0, 0) due to the normalization. The slope of the line determined by the data is statistically resolvable from that of theoretical equilibrium predictions (Ono et al., 2006). This is taken to suggest a process-specificity to the signature.
It is acknowledged that little is known about the 36S fractionation associated with oxidation processes, and speculate that they will produce slopes < -6.85 and expand the range proposed above. On its own, it is also intriguing that the sulfate reduction data define a well-constrained relationship that differs from a single 6.85 equilibrium line. This suggests that there may exist a predictable, linear, and process-
72
specific relationship between 33S and 36S. Future studies of both natural and experimental systems are needed to aid in this understanding.
33
vs. 36: The second relationship that I examine is between 33 and 36. In Figure
16 I plot all the sulfate reduction data in terms of 33 vs. 36. I observe that a majority of the data falls away from the reference fractionation coordinate of (0.515, 1.90). Although plotting away from the reference coordinate, the sulfate reduction data does cluster with 33 < 0.515 and 36 > 1.90. I take this to suggest that the fractionations associated with 33S and 36S may be coupled (clusters in Figure 16), but follow a relationship different than that of low temperature equilibrium (origin of the reference values). This suggests that what is expressed may be a process-specific relationship between 33S and 36S that describes sulfate reduction. Limited data for sulfur disproportionation experiments is also included in Figure 16 (see Table 3). These data suggest that the 33 vs. 36 for sulfur disproportionation experiments differ from the sulfate reduction experiments. Although preliminary, I propose this approach as a means of deciphering between the influences of sulfate reduction and sulfur disproportionation, especially when trying to understand complex environments. These data would suggest that a modern marine environment, where both sulfate reduction and sulfur disproportionation are active, would plot between the two established fields. The 33 vs. 36 relationship could provide great insight into the processes contributing to both modern marine isotope records, and also ancient records where the presence of different microbial processes is unknown or disputed.
73
Figure 16: The normalized data plotted as a function of 33 and 36, a general measure is independent of the magnitude of 34S fractionation. Errors are 2 . The giant yellow triangle represents the accepted RFL value (0.515 and 1.90), whereas the sulfate reduction data plots in the field defined by the four temperature suites listed. Symbol shapes are defined in Figure 15, with the addition of gray squares, which represent data from Chapter 2. Limited sulfur disproportionator data (from Chapter 2) is also included to illustrate the potential application of this measure as a means to distinguish between contributions from different sulfur utilizing metabolisms.
Conclusions The investigation of sulfur isotope fractionations for all four sulfur isotopes by sulfate reducing microorganisms has allowed for the traditional model of dissimilatory sulfate reduction (Rees, 1973) to be re-evaluated, indicating that the internal fractionations are larger than those assigned by Rees (1973). While the data require
74
larger internal fractionation factors, they do not uniquely support (or rule out) the possibility of an active trithionate pathway with reverse flow from a sulfur intermediate (or internal sulfide) to sulfite. The identification of this network using this approach would require lab experiments that produce larger 34S fractionations than have been observed and very specific 33S and 36S. In my opinion, studies targeting the isotopic composition of the four isotopes and concentration of internal sulfur species (namely internal sulfate and sulfite) and the fractionation associated with key enzymes (sulfite reductase) will be required to provide the information necessary to define an updated sulfate reduction isotopic fractionation pathway.
Other trace isotope relationships have also aided in the understanding of massdependent fractionations and potential environmental applications of the lessons learned from this study. I suggest that the application of the 33S vs. 36S relationship in mass-dependent systems and post 2.0 Ga environments can be a powerful means of addressing the presence of different metabolic processes or the presence of closed-system behavior in larger-scale systems, such as sedimentary basins or oceanic settings. Further, the data suggest a sulfate reduction specific slope of ~ -9, which statistically differs from the slope presented for single-step equilibrium fractionation effects (Ono et al., 2006). I also investigated a relationship between 33S and 36S that is independent of the magnitude of fractionation (34SH2S-SO4), 33 vs. 36. The difference in 33 vs. 36 for sulfate reduction and sulfur disproportionation suggests that this relationship may be very useful in studies of natural environments, where the relative contribution from these two metabolisms is poorly constrained.
75
The results of this study suggest that the sulfate reduction metabolism has larger internal fractionation steps than previously proposed. If all the fractionations associated with the sulfate reduction metabolism could be fully expressed, the data require that net fractionations would exceed 47 . However, these large net fractionations have not been observed in laboratory experiments or natural population studies, suggesting that the conditions necessary to produce such effects are difficult to achieve or unlikely in nature. The question then becomes, what is the feature limiting the expression of the larger internal fractionation steps? A similar dependence has been illustrated for the relationship between isotopic fractionation and sulfate concentrations (Thode et al., 1951; Habicht et al., 2002), and the results raise the tantalizing possibility that the lack of the full isotopic expression of sulfate reduction may actually be providing information about paleo-environmental limitations or limits for the cycling of sulfur within a sulfate reducing microorganism.
76
Chapter 4: Active Sulfur Disproportionation in the Mesoproterozoic
Introduction The environmental expression of sulfur compound disproportionation has been placed between 640 1050 Mya and linked to increases in atmospheric oxygen. These arguments are based on temporal changes in the magnitude of 34S/32S fractionations between sulfate and sulfide. Here, I present a Proterozoic seawater sulfate isotope record that includes the less abundant sulfur isotope, 33S. These measurements imply that sulfur compound disproportionation was an active part of the sulfur cycle by 1300 Mya, and that progressive Earth surface oxygenation may have characterized the Mesoproterozoic.
Background There is a strong link between the oxidation state of the Earth's surface environment and the microbial sulfur metabolisms that influence the sulfur cycle (Canfield and Teske, 1996; Canfield, 1998; Shen et al., 2001). This link is revealed through sulfur isotope studies where different microbial metabolisms contribute to the final isotopic composition of sulfur species preserved in the geologic record (Thode et al., 1961; Holser et al., 1988; Canfield, 2001). The relationship between isotopic fractionation
77
due to sulfate-reducing microorganisms (SRB2; SO42-
H2S) and seawater sulfate
concentration has been the primary tool for interpreting the sulfur isotope record of Earth surface oxidation (Perry et al., 1971; Hattori et al., 1983; Ohmoto et al., 1993; Habicht et al., 2002; Kah et al., 2004). For example, the isotopic record of sedimentary sulfides reveals that SRB may have dominated the global sulfur cycle until the Neoproterozoic. After this, greater 34S/32S fractionations cannot be explained by sulfate reduction alone (Canfield and Teske, 1996), and they likely reflect the added contribution of sulfur compound disproportionating microorganisms (SDB; S0/SO32-/S2O3 SO42-+H2S). As sulfide oxidation is responsible for the intermediate
sulfur compounds used by SDB (Troelsen and Jorgensen, 1982; Bak and Pfenning, 1987; Thamdrup et al., 1993; Canfield and Thamdrup, 1998), the widespread activity of SDB has been interpreted to indicate increased atmospheric oxygen content (Canfield and Teske, 1996). New data, however, suggests that the isotopic fractionation between seawater sulfate and sulfide in the Neoproterozoic may have been smaller than previously estimated (Hurtgen et al., 2002, 2005). This raises the prospect that the 34S record may not uniquely reveal the activities of SDB during the Neoproterozoic.
Recent experiments illustrate that SRB and SDB produce resolvable 33S/32S fractionations for similar magnitudes of 34S/32S fractionations (Chapters 2-3; Farquhar et al., 2003; Johnston et al., 2005b). In those experiments, the compositions of sulfate associated with SDB were more 33S enriched than sulfate associated with SRB. The
I chose to keep the traditional shorthand of SRB (sulfate reducing bacteria) to represent sulfate reducing microorganisms (Bacteria and Archea). Similarly, I use SDB to represent sulfur disproportionating microorganisms.
2
78
fractionations preserved in the sulfur isotope record largely reflect the combined influence of these two metabolisms (Canfield, 2001). I propose that by considering both the fractionations associated with 33S/32S and 34S/32S, as preserved in ancient marine sulfide and sulfate minerals, I can elucidate the significance of SRB and SDB on the global sulfur cycle. Here I combine a steady-state, open-system isotope massbalance model with data from sediments deposited between ~2000 and ~500 Ma to constrain how sulfur isotope signatures are transferred through a global sulfur cycle that includes SRB and SDB (Figure 17). The model tracks the sulfur isotopic composition of the seawater sulfate and reactive sulfide reservoirs as sulfur is microbially cycled between them. A fundamental assumption in the model is that any re-oxidation flux from reactive sulfide to seawater sulfate ultimately occurs through disproportionation reactions.
Sulfur cycle models Our model has a similar structure to other isotope mass-balance models of the S-cycle with two exceptions: (1) the model tracks 33S as well as 34S and 32S; and (2) it explicitly includes pathways through a sulfur reservoir of intermediate oxidation state to account for the isotopic consequences of disproportionation (Figure 17). The model was constructed with conventional approximations that result from the trace distribution of minor isotopes (33S, 34S). It was calibrated with experimentally derived fractionation factors (see Chapters 2-3) for microbial sulfate reduction and microbial sulfur disproportionation. A series of model calculations were run incorporating the whole range in 33S/32S and 34S/32S fractionations observed in pure 79
and enriched culture experiments (Chapters 2-3). The limits from over 200 individual models runs are represented by the fields presented in Figure 18. Inputs to the model are: 1) the experimentally calibrated 33S/32S and 34S/32S fractionations associated SRB and SDB, 2) the isotopic composition of the sulfate entering the model through the
Figure 17: Illustration of the box model used to constrain the isotopic composition of seawater sulfate. The arrows represent fluxes of material between reservoirs. Biologically mediated reactions are shown as red arrows, with both the reduction of sulfate to sulfide (by sulfate reducing microorganisms) and the oxidation and reduction of sulfur intermediates (by sulfur disproportionating microorganisms) involving significant fractionations. Contributions from volcanism are thought to be negligible, or contribute sulfur that is isotopically similar to continental weathering.
seawater sulfate reservoir (the origin in Figures 18-20), 3) the proportion of sulfate entering the model through the seawater sulfate reservoir that leaves the model as pyrite rather than as sulfate minerals (fpy), and 4) the proportion of sulfur entering the reactive sulfide pool that is completely re-oxidized to sulfate (fr-o). I began each calculation by choosing fractionations for SRB and SDB, and by varying fpy and fr-o a
80
unique array of relationships between the 34S and 33S of model seawater sulfate was produced.
Figure 18: 33S vs. 34S values for seawater sulfate predicted from an open-system steady-state S-cycle model. The discrete curves are calculated for a sulfur cycle that includes only SRB. Different curves are calculated for different values of experimentally constrained isotopic fractionations by SRB. The field bound by a solid red line is accessible to a strict SRB S-cycle (SRB-system). The field bound by a solid blue line is accessible to a combined SRB-SDB microbial S-cycle (SRB/SDB-system). The inset shows the direction that 33S vs. 34S trajectories evolve as fpy and fr-o increase. The different 33S vs. 34S regions accessed by the SRB- and SRB/SDB-systems are used to assess the microbial contribution to the oceanic sulfur cycle at the time of sulfate deposition.
As fpy increases in a model system including only SRB, the composition of seawater sulfate becomes 34S-enriched relative to the sulfate entering the model (Figure 18; Garrels and Lerman, 1981; Kump and Garrles, 1986; Berner, 1987; Holser et al., 1988; Kah et al., 2004). In this case, the 33S values of model seawater sulfate become more negative as 34S values increase (Farquhar et al., 2003). This is reflected in the orientation of the curves that outline the field of 33S and 34S in Figure 18 labeled SRB. When the model S-cycle is expanded with a re-oxidative 81
sub-cycle that allows for microbial sulfur disproportionation, increasing fr-o leads to seawater sulfate that is more enriched in 34S and has more positive 33S than when only SRB is included. This is reflected by the field labeled SRB-SDB (Figure 18). These model results form the basis for using the isotopic composition of proxies for seawater sulfate to distinguish the significance of microbial sulfur disproportionation within the global sulfur cycle.
This study I measured the sulfur isotopic composition of 49 Proterozoic to Cambrian sulfate samples from either carbonate associated sulfate (CAS: 35 in total) or sulfate minerals (14 in total; Table 6). Although I report 36S data in Figure 6, I save the discussion of this data until Chapter 7. CAS was extracted from powdered rock samples using a modified version of standard techniques (Burdett et al., 1989) and then reduced to Ag2S (Thode et al., 1961). Sulfate minerals were powdered and directly reduced to Ag2S. In Figure 19, the 33S and 34S values of these samples are plotted relative to fields for the modeled SRB-system and the modeled SRB-SDB-system. The model interpretation of these measurements assumes that they represent a well-mixed, homogeneous seawater sulfate reservoir whose composition is set by global processes. Although the results in Figure 19 have their basis in the assumption that the sulfate entering the model is unfractionated relative to bulk Earth estimates, the relative positions of the SRB and SRB-SDB fields are not affected by variations in the isotopic composition of the sulfate entering the model. Specific inferences about
82
83
Table 6: Sulfur isotopic measurements of seawater sulfate proxies. I focus on established proxies for seawater sulfate (i.e., evaporites; Claypool et al. 1980; Strauss, 1997; and CAS; Kampschulte et al., 2001; Hurtgen et al., 2002, 2005; Kah et al., 2004; Lyons et al., 2005). Ages listed are from Claypool et al. (1980), Schopf (1983), Strauss (1993), Kampschulte et al. (2001), Kah et al. (2004), Condon et al. (2005), and Gellatly and Lyons (2005). While I recognize that there are uncertainties associated with these radiometric constraints, they are not large enough to change the interpretation of temporal 33S variations.
the exact values of fpy and fr-o that are implied by 33S and 34S measurements of seawater sulfate proxies depend on the assumed isotopic composition of the sulfate entering the model. Because of this, for the time being I focus the interpretation of the new measurements on their general placement within the SRB and SRB-SDB fields. These conclusions are valid with reasonable isotopic variations of the incoming sulfate (34S 5 (Holser et al., 1988); 33S 0.03). The majority of the Neoproterozoic/Cambrian data in Figure 19a occupy the modeled SRB-SDB field. This 33S and 34S evidence for active microbial sulfur disproportionation is consistent with phylogenetic studies and previous interpretations of the 34S record (Canfield and Teske, 1996). This approach, however, also yields evidence for an active SRB-SDB system in the Mesoproterozoic (Figure 19b), leading to the suggestion that microbial sulfur disproportionation was not initiated in the Neoproterozoic but, instead, operated for at least part of the Mesoproterozoic (Hurtgen et al., 2005).
84
Figure 19: Measured 33S vs. 34S values for Proterozoic/Cambrian seawater sulfate proxies combined with model predictions (Figure 18). Errors are 2 . (A) Neoproterozoic-Cambrian data (1000 500 Ma) divided into older (1000 750 Ma; pink diamonds) and younger (571 500 Ma; blue diamonds) groups. Variation between the two groups likely reflects differences in fpy in a system with both SRB and SDB. (B) Paleo- to Mesoproterozoic data (2000 1000 Ma; red circles). The Paleo- Mesoproterozoic data extend across a range that is defined by the SRB system and the SRB-SDB system.
85
Proterozoic oxygenation The isotopic composition of seawater sulfate from the Mesoproterozoic Society Cliffs Formation (~1200 Ma) and the Dismal Lakes Group (~1300 Ma) show clear evidence for active microbial sulfur disproportionation (Figure 20a). The Society Cliffs data contain a strong SDB signature, and the relationship between this data and the model indicates extensive sulfur processing through disproportionation reactions. The 33S - 34S data for Dismal Lakes samples also contain a clear SDB signature and are consistent with lower proportions of sulfide re-oxidation and pyrite burial. In contrast, the isotopic compositions of CAS in the ~1450 Ma Helena Formation are consistent with a strict SRB system (Figure 20a) and do not require the influence of SDB. The data indicate that SDB became progressively more important in the global sulfur cycle over the ~250 million year time interval from 1450 to 1200 Ma. While these conclusions should be confirmed with additional data from other Mesoproterozoic basins, most pre-1300 Ma samples in the current dataset exhibit 33S - 34S values that unambiguously reflect an SRB-only system (Table 6, Figure 20b).
Thus far, samples from only one pre-1300 Ma sedimentary basin (McNamara Gp., ~1660 Ma; Figure 20b) appear to be inconsistent with the conclusions drawn above. These data, however, display some unusual isotopic characteristics. The 34S values of these samples span a wide range (~17 to 39), covering a significant portion of the entire dataset (~9 to 44). In addition, McNamara 33S values vary in a near-
86
Figure 20: Measured 33S vs. 34S values for seawater sulfate proxies from Mesoproterozoic basins and model predictions (Figure 18). (A) The ~1200 Ma Society Cliffs Formation (yellow circles) and the ~1300 Ma Dismal Lakes Group (blue circles) require active SDB at the time of their deposition. Samples from the ~1450 Ma Helena Fm. (purple circles) fit within the bounds of a strict SRB system, and do not require the presence of SDB. (B) Measured 33S vs. 34S values from pre 1300 Ma basins (circles) plot for the most part in the SRB system field. Green circles are for the ~1660 Ma McNamara Gp. span almost the complete range of values observed for Proterozoic sulfate, and exhibit a linear correlation with 33S measurements. These isotopic systematics are consistent with the exclusive operation of SRB on a limited sulfate pool.
87
linear fashion with 34S. Both of these characteristics are indicative of Rayleigh fractionation, and I can reproduce the McNamara data with such a model involving only SRB. Although I cannot rule out the possibility that the McNamara samples retain isotopic evidence of the effects of disproportionation, I hypothesize that this formation records a sulfur cycle dominated by SRB operating on a limited sulfate pool. This hypothesis is consistent with recent discussions of low sulfate levels during the deposition of the McNamara Basin sediments (Kah et al., 2004; Gellatly and Lyons, 2005), and it is testable by sulfur isotope analysis of sedimentary sulfides that formed contemporaneous with carbonates of the McNamara Group. These types of systematics are further discussed in Chapter 5.
Conclusions Taken together, the results bracket the appearance of a globally significant disproportionation pathway between 1450 and 1300 Ma. This predates prior estimates by several hundred million years (Canfield and Teske, 1996) and exposes an inherent limitation of using 34S to explore biogeochemical aspects of the sulfur cycle. Positive 34S evidence for SDB requires that the fractionations expressed in the isotope record must exceed the extreme fractionations observed for SRB (Canfield and Teske, 1996). By contrast, 33S traces the contribution of microbial disproportionation at smaller 34S/32S fractionations that would seem to be completely consistent with sulfate reduction from 34S alone.
88
While the new 33S measurements suggest a major change in the microbial regimes that controlled the isotopic composition of Proterozoic seawater sulfate, the environmental impetus for this change is less clear. The intermediate sulfur compounds required for SDB are generated by chemical oxidation of sulfide by O2 and metal oxides (Thamdrup and Finster, 1993; Canfield and Teske, 1996), by photosynthetic sulfide oxidizers (Canfield and Teske, 1996), and by O2 or nitraterespiring anaerobic non-photosynthetic sulfide oxidizers (Thamdrup and Finster, 1993; Canfield and Teske, 1996). On modern Earth, the compounds produced by these processes occur in a variety of chemical transition zones, such as at oxic/anoxic interfaces in marine sediments and stratified water columns, and within the layers of microbial mat communities (Troelsen and Jorgensen, 1982). This suggests that sulfur disproportionation dominantly occupied surface ocean and/or shelf environments where local oxidative processes were responsible for the production of sulfur intermediates.
Other indicators of an oxidative surface environment, such as 13C variations (Frank et al., 2003; Kah et al., 2004), evolutionary arguments (Knoll, 1994; Canfield, 1998), and sulfate concentration estimates (Kah et al., 2004; Gellatly and Lyons, 2004), are temporally consistent with a Mesoproterozoic onset of disproportionation. A highresolution 33S record from the critical interval between 1450 and 1300 Ma may capture this onset in action, revealing whether the rise of SDB lagged or accompanied the progressive oxygenation of Earth's surface.
89
Chapter 5: Evolution of the Oceanic Sulfur Cycle at the End of the Paleoproterozoic
Introduction Here I present new measurements of 32S, 33S, 34S, and 36S in sedimentary sulfides and couple these measurements with modeling treatments to study the sulfur cycle of a late Paleoproterozoic marine basin. I target the transition in ocean chemistry from the deposition of Paleoproterozoic iron formations (Gunflint, Biwabik, Trommald, and Mahnomen iron formations) to the inferred sulfidic ocean conditions recorded by overlying shale (Rove Formation). The data suggest that certain features of the global sulfur cycle, such as a control by sulfate reducing microorganisms, and low (mM) concentrations of oceanic sulfate, were maintained across this transition. This transition was associated with changes in the structure of the basin-scale sulfur cycle during deposition of these sediments. Sulfide data from the iron formations are interpreted to reflect sedimentary sulfides formed from microbial reduction of porewater sulfate that was supplied through steady-state exchange with an overlying oceanic sulfate reservoir. The sulfide data for the euxinic Rove Formation shales reflect the operation of a sulfur cycle that included the loss of sulfide by a Rayleighlike process. I suggest that the prevalence of large and variable heavy isotope enrichments observed in Rove Formation sulfide minerals reflect a sustained and significant net loss of sulfide from the euxinic water column, either as a result of a
90
shallow chemocline and degassing to the atmosphere or as a result of a water column pyrite sink. The inclusion of 36S measurements (in addition to 32S, 33S and 34S) illustrates the mass-dependent character of these sedimentary environments, ruling out contributions from the weathering of Archean sulfides and pointing to at least modest levels of sustained atmospheric oxygen (> 10-5 present atmospheric levels of O2).
Background Recent studies of redox sensitive elements, mineral speciation, and isotopic systems have indicated a transition from ferruginous (Fe2+-rich) to widespread euxinic (S2-rich) water-column conditions coincident with the cessation of iron formation (IF) deposition in the late Paleoproterozoic (Poulton et al., 2004). Sulfidic conditions are thought to have become widespread at the time of this transition, persisted into the middle Proterozoic, and may have had profound effects on the isotopic composition and size of the seawater sulfate reservoir (Canfield, 1998; Shen et al., 2002; 2003; Arnold et al., 2004; Canfield, 2004; Rouxel et al., 2005). The lithologic progression from iron formation (Gunflint Formation, Biwabik Formation, Trommald Formation, and the Mahnomen Formation) to shale (Rove Formation) in the Animike Basin, North America, has been interpreted as capturing the transition from ferruginous to sulfidic conditions (Poulton et al., 2004). The observation of a high proportion of highly reactive iron (FeHR) to total iron (FeTotal)(FeHR / FeTotal >0.38) points to persistent anoxic bottom water conditions throughout this succession, and the ratio of pyrite iron (FePy) to highly reactive iron indicates a change from ferruginous 91
conditions during deposition of the Gunflint Formation (FePy / FeHR = 0.0010.001) to sulfidic conditions the in upper Rove Formation (FePy / FeHR = 0.870.04). In this study I evaluate 38 new measurements of the four stable sulfur isotopes (32S, 33S, 34S and 36S) for Cr-reducible sulfur (herein referred to as sulfide) from samples of the laterally equivalent Gunflint Iron Formation (n=10), Biwabik Iron Formation (n= 3), Trommald Iron Formation (n=5), and Mahnomen Formation (n=3), with additional samples from the overlying Rove Formation (n=17) (Table 7). This approach expands on that used in prior studies of sulfur isotopes because it includes new modeling and evaluates the significance of 33S and 36S in addition to the more commonly reported 34S.
Setting, Systems and Methods
Geologic setting
At approximately 2450 Ma, a rift to passive margin developed along the southern edge of Superior Province when a land-mass to the south separated (Fralick and Miall, 1989). Later closure of the resultant ocean led to the deposition of the Animikie and North Range Groups as a backarc basin developed (Pufahl and Fralick, 1995; 2004; Hemming et al., 1995; Van Wyck and Johnson, 1997; Pufahl et al., 2000), which, with a collision, was later transformed into a foreland setting (Hoffman, 1987; Southwick and Morey, 1991; Hemming et al., 1995; Ojakangas et al., 2001; Maric and Fralick, 2005). Initial south to north flooding of the backarc led to deposition of siliciclastic tidal deposits (Ojakangas, 1983) forming the Mahnomen,
92
Pokegama and Kakabeka Formations (Figure 21). These are thickest in the south, forming only a sporadically developed basal conglomerate in the northernmost area. The chemical sediment dominated Trommald, Biwabik and Gunflint formations overlie the basal siliciclastics (Figure 21). Chemical sediments are also prominent within parts of the Mahnomen Formation. These units are primarily composed of both granular and fine-grained iron oxides, carbonates and chert. The sedimentary strata are organized into fining- and coarsening-upward successions reflecting transgressive-regressive cycles (Fralick and Barett, 1995) on an open, wave and tide dominated shelf (Ojakangas, 1983; Fralick, 1988; Pufahl and Fralick, 1995; 2004; Pufahl et al., 2000). Water depths during the deposition of these chemical sediments, even on the mid-shelf, probably did not exceed 10s of meters (Pufahl and Fralick, 2004). U-Pb age determinations of zircons from a reworked tuff in the Gunflint Formation yielded an age of 1878 1 Ma (Fralick et al., 2002).
Deposition of the iron formations probably ended in the northern area, and possibly to the south, as closure with a land-mass to the south caused upwarping and withdrawal of the sea (Addison et al., 2005) during the 1860-1835 Ma (Sims et al., 1989) Penokean Orogeny. During this interval, at approximately 1850 Ma (Krough et al., 1984), a large hypervelocity impact occurred 700 km to the east, near Sudbury, and the ejecta is present very near the top of the Gunflint and Biwabik Formations (Addison et al., 2005). This uppermost portion of the Gunflint is highly altered with intense silicification and the development of agate and pyrite veins and vugs. The alteration may be the result of subareal exposure during a hiatus inferred from a U-Pb
93
94
Figure 21: Geology and location of the Animikie Basin a.) Location map. b.) Animikie sedimentary rocks generally dip gently to the southeast with their lower unconformable contact with Neoarchean rocks exposed to the northwest. The southernmost portion of the basin is more intensely deformed, and there structural contacts are probably present. Mesoproterozoic intrusive rocks related to the Mid-Continental Rift divide the basin and Rift basalts overlie the basin. The stratigraphic sections depict the rocks present in three continuously cored drill-holes. Shales to sandstones of the Mahnomen Formation and Pokegama Quartzite underlie the Trommald, Biwabik and Gunflint iron formations. The Trommald is dominated by fine-grained iron and manganese oxides deposited in the outer shelf to slope environment (ocean to the south, shoreline to the north) (Pufahl et al., 2000). Cycles containing fine-grained chemical sedimentary rocks overlain by coarsening-upwards grainstone assemblages, composed of chert, iron oxide and iron carbonate sand, dominate these shallow shelf deposits. Two tuff layers provide chronostratigraphic markers and show that the lithofacies boundaries are diachronous and reflect south to north transgression-regression-transgression (Fralick and Barrett, 1995; Pufahl et al., 2000). A hiatus appears to occur between the iron formation and the overlying siliciclastic rocks of the Virginia and Rove Formations (Addison et al., 2005). See the text for a more detailed discussion. Figure from data in Pufahl (1996) and Maric and Fralick (2005). Sample positions are indicated to the right of the stratigrpahic column, with width of the column indicating grain size.
age on zircons of approximately 1835 Ma from tuffs immediately above the Gunflint and Biwabik formations (Addison et al., 2005). These tuffs are contained in the Virginia and Rove formations (Figure 21). The lower 100 to 150 m of these units consists of alternating shale-siltstone and black, pyritiferous shale successions, probably reflecting fluctuations in sea level (Maric and Fralick, 2005). These successions, and especially the upper black shale, likely represent a major condensed interval deposited in water ~100 to 200 m deep. Lucente and Morey (1983) ascribed sedimentation of this interval to pelagic rainout of fine-grained sediment from dilute suspension or hemipelagic processes involving diffuse turbidity currents. The presence of abundant, sub-millimeter rip-up intraclasts also denotes the operation of sporadic bottom currents (Maric and Fralick, 2005). Tidal deposits present in correlative rocks to the south of Lake Superior (Ojakangas et al., 2001) confirm open connection to the ocean. Above the upper, pure black shale interval, graded finegrained sandstones are organized into a coarsening-upward succession approximately 100 m thick that is transitional into 400 m of medium-grained, sandstone-dominated,
95
stacked parasequences (Maric and Fralick, 2005). This is overlain by lenticular to wavy bedded sandstones and shales with both wave and current ripples. The coarsening-upward to sandstone-dominated portion of the Virginia and Rove formations has been interpreted as a submarine fan (Lucente and Morey, 1983; Maric and Fralick, 2005) with the uppermost ripple laminated succession representing progradation of distal distributary mouth bars of a delta (Maric and Fralick, 2005). A sandstone sample from the submarine fan portion of the succession yielded a youngest U-Pb detrital zircon age of approximately 1780 Ma (Heaman et al., 2005). The predominantly Paleoproterozoic zircon population (Heaman et al., 2005), and paleocurrents indicating sediment derivation from the north (Morey, 1973), strongly suggest the Trans-Hudson Orogen was the source of the detritus. The northern sedimentary rocks of the Animikie Basin are essentially undeformed and unmetamorphosed, making them ideal targets for the study of low temperature biogeochemical cycling.
Analytical methods
The majority of the samples examined here were previously studied by Poulton et al. (2004) using methods described in Poulton and Canfield (2005). All additional samples were chemically prepared in the same manner as those previously studied (Poulton et al., 2004). In addition to the samples prepared for this study, splits of the Ag2S extracted as Cr-reducible sulfur (considered to be pyrite; Canfield et al., 1986) by Poulton et al. (2004) were analyzed. General wet chemical and fluorination
96
methods are presented in Appendix 3, with details of V-CDT normalization described in Appendix 1.
Modeling treatment
I use a box model of the surface sulfur cycle (c.f., Garrels and Lerman, 1981) to evaluate how sulfur moves between different terrestrial sulfur reservoirs and to make
Figure 22: A box model representation of the global steady-state sulfur cycle. Arrows represent fluxes between reservoirs (boxes). The relative placement of the boxes in the figure is independent of the actual environmental location of the pool. a.) I model a global sulfur cycle with one source (continental/riverine) and two global sinks (sulfide and sulfate minerals). The recycling of the oceanic sulfur pools (global sulfide and global sulfate minerals) is not incorporated here due to the relatively short depositional period represented by these sediments. b.) The modeled sub cycle used to describe the BIF depositional environment. Here, the global seawater sulfate reservoir feeds to a ferruginous deeper water reservoir. c.) The modeled sub cycle used to describe the shale depositional environment. In order to satisfy the data using a steady state model, multiple sulfide reservoirs were included in the sub cycle.
predictions about the 33S, 34S, and 36S of Proterozoic seawater sulfate. A previous study used a similar approach and considered the isotopic fractionations associated with sulfate reduction and re-oxidative processes such as disproportionation (see Johnston et al., 2005a for a more detailed description). The form of the model in this
97
study is presented in Figure 22, and is modified from Chapter 4 (Johnston et al. 2005a) in order to account for some new features of the data (sulfides enriched in
34
S).
Results The data are presented in Table 7. The ranges of 34S, 33S, and 36S are -0.96 to 32.92 , -0.073 to 0.078 , and -0.35 to 1.03 respectively. The 34S data parallel those presented in Poulton et al. (2004) with the IF (lower Gunflint, Biwabik Trommald, and the underlying siliciclastic Mahnomen Formations) yielding an average 34S of 8.4 4.6 (1 ), the lower ~100 m of the Rove Formation (assigned to a transitional environment between ferruginous and persistent euxinia; Poulton et al., 2004) yielding an average 34S of 13.2 11.0 (1 ), and the data for the upper sulfidic Rove Formation yielding an average 34S of 16.6 6.8 (1 ).
Comparison of new SF6 analyses with earlier SO2-based analyses (see Poulton et al., 2004; Johnston et al., 2006) of the same Ag2S precipitates suggests a correlation of 34SSF6 = 1.009 ( 0.051; 1) 34S SO2 -0.066 ( 0.711; 1). However, when three significant outliers (> 3 ; R30, R28, and R5) are removed from the comparison, the same linear regression technique suggests a correlation of 34SSF6 = 1.055 ( 0.016; 1) 34S SO2 -0.166 ( 0.210; 1). Both predicted correlations are within 2 of comparisons of these two methods (34SSF6 = 1.035 34SSO2 -0.135; Rees, 1978), in which the differences are interpreted to reflect scale compression on the SO2 measurements as a result of memory effects.
98
Table 7: All sulfur isotope data grouped by formation. Samples from the Gunflint and Rove fms. are arranged in stratigraphic order. Samples in italics were visually identified as compromised by secondary processes (silicification). These measurements are included in this table and the Results section, but not interpreted as representative of the depositional environments.
99
Our results for 33S and 36S yield 33S and 36S of -0.037 0.022 and 0.54 0.22 for the IF, -0.004 0.021 and 0.32 0.18 for the transitional shales, and 0.010 0.044 and 0.24 0.29 for the upper euxinic Rove Formation samples. The IF data cluster towards lower 34S, negative 33S, and positive 36S, whereas the shale data cover almost the entire data set. No strong stratigraphic trends are observed within the data, although a slight positive trend in 34S and 33S (negative in 36S) is observed moving up-column. The 36S and 33S values are correlated and form an array given by 36S = -6.28 ( 0.78; 1) 33S + 0.300 ( 0.029; 1) (with subtrajectories of 36S = -5.58 ( 2.12; 1) 33S + 0.336 ( 0.091; 1), 36S = -3.76 ( 3.08; 1) 33S + 0.338 ( 0.063; 1), 36S = -5.77 ( 1.116; 1) 33S + 0.301 ( 0.048; 1), and 36S = -10.31 ( 2.79; 1) 33S + 0.201 ( 0.064; 1) for the IF, transitional shale, sulfidic shale, and silicified samples respectively).
Discussion Previously published treatments of the sulfur cycle suggest that the sulfur isotope composition of the oceanic sulfate and sedimentary sulfide pools depends on material-balance constraints given by the fraction of sulfur lost to pyrite burial (fpy), the magnitude of fractionation by sulfate reducing bacteria, and when relevant, the magnitude of fractionations associated with other reaction and transfer pathways such as sulfur disproportionation (Jorgensen et al., 1977; Claypool et al., 1980; Garrels and Lerman, 1981; Berner, 1991; Strauss, 1993; Hurtgen et al., 2002; 2005; Canfield, 2004; Johnston et al., 2005a; Ono et al., 2006). Some of these studies have also
100
suggested that non steady-state effects can introduce additional isotopic effects associated with material balance (Johnston et al., 2005a; Ono et al., 2006). In the first part of the following discussion I use steady-state sulfur cycle models to make inferences about the composition of seawater sulfate and sedimentary sulfide during IF deposition. The models have a global component with a structure that is similar to that used in Johnston et al. (2005a) (Figure 22a), but also includes local recycling of sulfur (Figure 22b). In the second part of the discussion I make inferences about an expanded steady-state model (Figure 22c) and a basic non steady-state system. Here, I draw on a combination of Rayleigh (e.g., Goldhaber and Kaplan, 1975) and mixing effects to gather information about the behavior of non steady-state systems.
Implications of the steady-state box model approach
The average composition of the source of sulfur for the oceanic sulfur cycle is thought to have an isotopic composition near, or slightly more enriched (< 5 in 34S) than that of CDT (33S 0, 34S 0, and 36S 0) (Farquhar et al., 2002). When combined with constraints imposed by conservation of mass, the dominant fractionations in the surface sulfur cycle produce 34S-depleted reservoirs of reduced sulfur (ie. pyrite) and 34S-enriched reservoirs of oxidized sulfur (ie. sulfate). The observation that a majority of the sulfide data (35 of 38) from the successions examined here has positive delta values (are 34S-enriched) indicates that the global sulfur sink (Figure 22a) is not represented by this sample set. Observations of 34Senriched sulfide minerals have also been described in other Proterozoic shales (e.g., Walker and Brimblecombe, 1985; Hayes et al., 1992; 1993; Shen et al., 2002; 2003;
101
Canfield, 2004). Hayes et al. (1992) suggested that the lack of 34S-depleted sulfides at this time reflects poor preservation of deep-water settings, where isotopic fractionation associated with sulfate reduction would have been large. Those authors suggest that lower rates of sulfate reduction in deep water settings would facilitate larger isotopic fractionations and the production of 34S-depleted sulfide. These sulfides would likely be subducted and lost from the surface record (Canfield, 2004).
Conversely, continental shelf and inta-cratonic basin settings, which represent much of the sampled record during these time intervals, and where organic carbon is readily available, are locations where sulfate reduction rates are assumed high and where near quantitative reduction of available sulfate is feasible. This would result in sedimentary sulfides with isotopic compositions near that of seawater sulfate, or in the case of the Proterozoic, between 0 and 20 in 34S. I suggest that the data reflect the operation of a (basin-scale) sub-cycle similar to that presented in Figure 22B. In this model, and consistent with the Hayes (1992) model, the global cycle produces an oceanic sulfate reservoir with positive 34S (34S-enriched) that serves as the source pool for the basin-scale sub-cycle, thus allowing for the production of positive 34S sulfides on the shelves or shallower water settings.
The oceanic sulfate reservoir: The isotopic composition of the global oceanic sulfate reservoir during the late Paleoproterozoic remains poorly constrained. Traditional approaches to assess the isotopic composition of the past oceanic sulfate reservoir have been through direct analyses of proxies for oceanic sulfate (e.g., evaporite
102
sulfate, carbonate associated sulfate, marine barite; Holser et al., 1979; Claypool et al. 1980; Burdett et al., 1989; Strauss, 1993, 2004; Paytan et al., 1998; 2004; Turchyn and Schrag, 2004; Lyons et al., 2004) and models of the sulfur cycle, which often rely on sulfide data (Canfield and Teske, 1996; Canfield, 2001). In recent studies by Johnston et al. (2005a) and Ono et al. (2006), the composition of the oceanic sulfate reservoir has also been evaluated using oceanic box models. These authors argue that the composition of oceanic sulfate will occupy an isotopic field that extends from the composition of the sulfate source to the oceans (roughly zero) to more positive 34S and variable but small positive and negative 33S and 36S values. The generation of non zero 33S and 36S values is due to the continued removal of isotopically fractionated sulfide and mixing.
An expanded version of model predictions from Johnston et al. (2005a) is presented in Figure 23. Whereas earlier treatment focused on predicting the isotopic composition of seawater sulfate, the current treatment has been expanded to include predictions for the isotopic composition of sulfides. An assumption built into the model results is that the isotopic composition of the sulfur entering the system is roughly zero, and use microbial fractionation factors from Johnston et al. (2005b). Model interpretations do not change with small variations (up to ~ 5 in 34S) in the composition of the sulfur source to the oceans. The model results define an SRB field representing a sulfur cycle with only active sulfate reduction and a SRB-SDB field representing the predictions for a sulfur cycle with contributions from both sulfate reducers and sulfur disproportionators. The isotopic field defined by a sulfur
103
104
Figure 23: Triple isotope plot (33S versus 34S) illustrating the steadystate sulfur cycle modeling solutions for the isotopic composition of sulfate and sulfide. Fractionation factors are derived from experiments with SRP conducted by Johnston et al. (2005b) (Chapters 2-3). The fields outlined in red represent a sulfur cycle with only SRB active, whereas fields outlined in blue have both SRB and SDB present. This model is an extended version of those presented in Johnston et al. (2005a). The position along each respective curve represents the fraction of pyrite being buried (fpy in Johnston et al., 2005a). In the case of the sulfate curve, the intercept represents no pyrite burial with pyrite burial increasing to 100% as you move away from the origin to the right. Due to mass balance, the pyrite curve is demarcated in a similar manner with no pyrite burial at the left most point on the curve, and increasing towards 100% pyrite burial as you approach the origin. In all cases, sulfate and sulfide predictions evolve away from the origin since I assume a continental/riverine flux that is isotopically similar to bulk Earth (3iS ~ 0 ). All modeling represents steady-state box modeling of the global sulfur cycle. The same treatment applies for 36S.
cycle with only sulfate-reducing microorganisms occupies a smaller region of predicted isotopic compositions than if other microbial processes, such as sulfur disproportionation, are active as well (see Johnston et al., 2005a). Previous studies suggest that only sulfate reducers operated within the sulfur cycle during the late Paleoproterozoic (Canfield and Teske, 1996; Canfield, 1998; Johnston et al., 2005a), suggesting that the smaller SRB field defines the range of compositions for oceanic sulfate at that time.
This model is used to provide a context for interpreting isotopic data for sulfides from the Gunflint, Biwabik, Trommald and Mahnomen Formations. The steady-state box model would reproduce the IF sulfide data using fractionations associated with only sulfate reducing microorganisms and a fitted composition for oceanic sulfate of 34S ~ 17 , 33S ~ -0.085 and 36S ~ 0.60 (Figure 24), which is the minimum estimate (in 34S and 36S, maximum in 33S) to satisfy the data. This prediction also supports the lack of a prominent isotopic contribution from SDB at this time. The validity of the estimate for the isotopic composition of the seawater sulfate reservoir
105
hinges on the appropriateness of the model approach, but it also illustrates a new type of constraint for seawater sulfate compositions. This treatment provides a seawater sulfate composition that is not only consistent with earlier predictions of the 34S of
Figure 24: All the data along with a steady-state modeling result for a sulfur cycle with only active sulfate reducers. The inclusion of a sub-cycle, as presented in Figure 22b, can shift the isotopic predictions, such that the sulfide and sulfate fields would not evolve away from a continental composition (original source pool; seen in Fig. 24), but would be rooted from a composition within the original seawater sulfate field (new source pool). In this case, global processes would be controlling the isotopic composition of the overlying seawater, but local effects are controlling the isotopic composition of the sedimentary sulfides in the sediment column. Red circles represent IF sulfide data (with 2 errors), while transitional data (defined by iron speciation, depths C to V) are in yellow and sulfidic Rove data is in blue. The red lines represent the global predictions for a sulfur cycle with only SRB active and are rooted at the predicted seawater sulfate composition (gray square). This seawater sulfate prediction is nonunique and falls within the range predicted by Figure 23 (an experimentally constrained field). Since the model adequately encompasses the IF data, I suggest that these sulfides are derived from the microbially (SRB) reduction of pore-water sulfate within a sediment column under steady state conditions.
Paleoproterozoic seawater sulfate (Strauss, 2004; Gellatly and Lyons, 2005), but consistent in 33S and 36S with the predictions from steady-state global box models (see position of sulfate field Figure 23A versus the composition suggested for the 106
period of BIF deposition: Figure 24). Similar steady-state modeling cannot explain the significant scatter of the Rove Fm. sulfide data, thus a prediction of the associated seawater sulfate composition is not appropriate.
Iron formation environments: I interpret the IF data to reflect operation of a localized sulfur sub-cycle linked to the IF depositional environment (ie. pore-water sulfate and local sedimentary sulfide) (Figure 25). I envision the overlying seawater sulfate infiltrating the sediment column where microbial reduction of pore-water sulfate would generate sulfide that was captured as pyrite. In this environment, the small (~10 ) 34S fractionations between the model prediction for seawater sulfate and the measured sedimentary sulfide indicate either a high fraction of pore-water sulfate sulfur was reduced and captured as pyrite (local fpy 1, where local fpy is the ratio of
pyrite sulfur to sulfur infiltrating as pore-water sulfate), or low pore-water sulfate concentrations. The high Fe-content of the sediments, coupled with low organic matter contents, suggest that lowered pore-water sulfate concentrations was most likely the limiting constituent in this environment (Habicht et al., 2002). The requisite source of Fe for pyrite formation is taken to be similar to other Superior-type IFs, which are thought to sample water derived from a deep ferruginous ocean, and where a majority of the ferrous iron would be chemically or microbially oxidized on the shelf (e.g. Derry et al., 1992).
Transitional environments: The suite of samples that lie between the underlying IF and the overlying euxinic shales of the upper Rove Formation were grouped by
107
Poulton et al. (2004) on the basis of evidence from Fe-speciation work and designated as transitional. The transitional nature of the depositional environment from which these samples were derived is further supported by the sulfur isotope data for this horizon, which straddles both the field defined by the Gunflint IF and the field defined by the overlying euxinic Rove Formation shale. The two lowermost samples in the transitional Rove Formation shales fall within the steady-state IF field, suggesting an initial consistency with the underlying IF data (ie. deposition under ferruginous water column conditions). Data for immediately overlying samples
Figure 25: A cartoon representing the envisioned IF depositional environment. The focus of this graphic is to constrain the cycling of sulfur through the system, and in doing so, other important elements and cycles may be underappreciated. In the right-most column, a = atmosphere, o = oxic ocean, f = ferruginous ocean, and s = sediments. This environment is interpreted to have a shallow, stratified water column with a component of horizontal (possibly wind/upwelling driven) transport. There would likely be significant oxic/anoxic interaction at the chemocline, with the zone of most intense SRB activity and pyrite production in the upper portions of the sediments, where organic material would most readily accumulate. The lifetime of sulfide in this system would be very short, as the available iron would quickly scavenge the free sulfide for Fe-S formation (and later pyrite). The profiles at left represent speculative ion profiles for two different sulfur species (sulfate and sulfide). The sulfate concentration would remain relatively constant until entering the lower portion of the water column, where the compositional gradients extending from the sediment system could draw down sulfate. This is a function of the most prominent sulfate sink being in the sediments. In the sediments, sulfate would be expected to decrease through the zone of active sulfate reduction and level off as methanogensis begins to dominate. The sulfide profile is a reflection of the sulfate prediction.
108
diverge from the IF field and fall within the field for the euxinic Rove Formation shales. Further up-column, there is a brief return to coincidence with the IF field before the data terminally diverge to occupy the euxinic shale field. Noteworthy is that the transitional samples overly the Gunflint/Rove disconformity, suggesting that the change in ocean chemistry or the establishment of a different basin scale sulfur cycle (discussed below) captured by these samples, post-dated the refilling of the Animike Basin, roughly 40 Ma after the deposition of the underlying iron formations.
Insight into the euxinic shale environment
The large amount of isotopic variability for the euxinic Rove Formation sulfides appears to be independent of stratigraphic position, cannot be explained by the inclusion of active SDB, and is inconsistent with the box model approach described in Figure 22B. Similarly, the isotopic data for euxinic Rove sulfide occupy an area to the right of the sulfide field presented in Figure 24. This observation leads us to consider the possible role of other effects within the euxinic Rove Fm. sulfur cycle. Two recent studies have described how removal of sulfide by Rayleigh fractionation processes can produce sulfur isotope compositions for sulfate and sulfide that migrate to positive 34S with only small changes of 33S (Johnston et al., 2005a; Ono et al., 2006). Similar effects may be produced in systems at steady-state with continuous sulfide loss, or in systems where changes in the source or sink terms have resulted in a non steady-state condition. I will discuss both the possibility for a non steady-state
109
explanation for the euxinic Rove Fm. data, as well as the possibility that the data may be explained by a steady-state sulfur cycle with continuous loss of sulfide.
Possibility of a non steady-state sulfur cycle: Non steady-state conditions, of which Rayleigh effects are a limit, arise when variability in either the source or the sink reactions leads to a significant fluctuation in the size of one or more of the sulfur pools. I consider the temporal variability in seawater sulfate concentration as a
Figure 26: A triple isotope plot describing the consequences of Rayleigh fractionation and mixing for the simple one box model presented in the lower corner of the figure. The black and white boxes are the steady state predictions for the compositions of the sink and pool respectively. The Rayleigh fractionation trajectories (straight arrows) constrain the initial direction that the loops (which include mass balance considerations) proceed in. The right most loop represents the evolving composition of the pool, whereas the left most loop represents the evolving composition of the sink, and is offset by a fractionation factor (). Plotted is the scenario where the sink overwhelms the source.
function of the sum of fluxes in and fluxes out. The isotopic ratio of these fluxes results in the definition of a fractionation factor for the process. Thus, the isotopic consequence of a non steady-state depends on the fractionation between reservoirs
110
and the magnitude of the change in reservoir size. These can be described by a combination of Rayleigh effects and mixing. In Figure 26 I illustrate the way that a one-box system responds when the source flux decreases relative to the sink flux. In this system, the initial isotopic evolution follows a trajectory defined by conventional Rayleigh fractionation (Rf = R0 f (-1), where f is the fraction of Rayleigh distillation, and Rf and R0 are the isotopic ratios for the instantaneous composition at fraction f and for f=0)(Figure 26). As the size of the pool continues to change, the predicted isotopic evolution begins to loop back around towards the starting composition as a function of mass balance (noted as mixing (recovery) loops). This treatment predicts a drastic initial isotopic evolution as a response to non steady-state conditions, but a slow and less isotopically drastic response as the system reaches a new steady state.
In order for a non steady-state interpretation of the euxinic Rove Fm. sulfide data to be valid, it must explain the general geological and geochemical features of the euxinic Rove shale. Those features include the interpretation that deposition occurred in a marine setting, and that large (tens of ) 34S enrichments occur throughout about ~40 meters of stratigraphic section. Given the measured fractionations (>20 in 34S and highly variable 33S and 36S), the magnitude of the non-steady state effects for the Rove Fm. would require a process capable of generating significant changes in the size of the sulfate pool. This is because in order for a Rayleigh-like process to produce the observed effects, a significant fraction of the local sulfate reservoir (either bottom water sulfate or pore-water sulfate) must have been removed. This
111
estimate may represent a lower limit, since this isotopic variation is captured in the sulfide, rather than sulfate, pool. A standing pool of sulfide in a sulfidic water column has the potential to damp the magnitude of isotopic signals that are transferred to it from the sulfate pool. For example, if 10s of M sulfide are produced and assimilated into a standing sulfide pool of 100s of M sulfide, the isotopic contribution of that most recent addition will not be fully expressed, whereas if there was no standing sulfide reservoir, the composition of the recently produced sulfide would be representative of the standing pool and could then be captured as pyrite.
Prior studies have suggested that some non steady-state effects may accompany the geographic restriction of a basin, with episodic re-supply from a sulfate source (e.g., Johnston et al. 2005a). I do not favor this for the Rove Formation because it has been interpreted to have been deposited in a marine environment, and thus communicating with the open ocean. As an alternative, I explore the possibility that the euxinic Rove Formation data reflect the normal operation of a low sulfate marine environment. I focus on two parts of this system, (1) the water column and (2) the underlying sediments.
High rates of sulfate reduction are common for organic rich sedimentary environments. Thus, it may be possible, given increased organic material, for euxinic water column sulfate reduction rates to consume a significant proportion of water column sulfate and induce a non steady-state behavior within the system. This type
112
of situation can be described using a simple expression and observed rates for modern water column processes. I calculate steady-state, 1-D penetration depths to gain an order of magnitude estimate of how sulfate concentrations might vary as a function of overlying water column depths. I begin with a simple diffusion length-scale expression, where L ~ Dt (or length is related to the square root of diffusion
multiplied by time). However, if I make the assumption that the characteristic timescales of the system are set by sulfate reduction rates (SRR), which is reasonable, I can substitute time (t) with the expression: [boundary concentration/sulfate reduction rate]. This is easier to understand when considering units, where boundary concentrations are reported in moles per liter (M), and sulfate reduction rates are reported in moles per liter per unit time (time = M/[M/time]). I run this calculation using low sulfate concentrations (1-4 mM), Black Sea water column mixing rates from Neretin et al. (2001)(0.5 to 3 cm2/sec) as a proxy for effective diffusivity, and for sulfate reduction rates of 1M/day which are at the high end of those reported for euxinic water columns (e.g., 0.0035 to 1.569 M/day; Jrgensen et al., 1991; Ilchenko and Sorokin, 1991; Albert et al., 1995; Neretin et al. 2001; Srensen and Canfield, 2004). These results indicate that sulfate concentrations will approach zero at depths between ~ 50 and 350 meters below the chemocline, which spans the inferred water column depths overlying the Rove Fm. As sulfate reduction rates increase, the water column depth required to remove significant amounts of sulfate decreases. Srensen and Canfield (2004) observed a rate enhancement when water column samples were augmented with acetate and lactate. It has also been argued (Logan et al., 1996) that concentrations of suspended organic carbon (dissolved and
113
particulate) were higher in the Proterozoic, which they argue would have fostered increased rates of water column sulfate reduction during that period. Both suggestions indicate that significant water column sulfate removal is likely under supposed Proterozoic shelf conditions (euxinic with high organic carbon). These predictions assume that there is not significant lateral transport of sulfate-rich water onto the shelf, but acknowledging that lateral homogeneity is more likely in closed basins or where continental shelves are extensive.
The development of a sulfate-limited system, which is modeled as a non steady-state response in the euxinic Rove Formation sulfur cycle, may also have been initiated as a result of changes in the rates of sulfate uptake by sediments. Considerations of the contribution of active sulfate reduction in the underlying sediments, and the role that this sulfate sink can play in water column chemistry, must also be considered. In low sulfate concentration water column-sediment systems, variations in sulfate reduction may influence the size of the water column sulfate pool. The depths inferred for the Rove Formation depositional environments (~100-200 m) are on the same order as the characteristic diffusion length-scales described using the same equation described above ( Dt ) and using the above listed mixing rates. Given typical rates for uptake of sulfate measured in modern sedimentary systems (~0.2 to 50 mol/m2yr) (e.g., Jorgensen, 1977), it is predicted that there will be a significant extension of the sedimentary/ water column boundary layer that would result in sulfate draw down in the water column. For example, a 100 meter by 1m2 water column with 1mM of sulfate contains only 100 moles of total sulfate, and variability within the range of
114
observed rates of sedimentary sulfate uptake will have a significant impact on the size (2-50 % change) and isotopic composition of the sulfate pool in this system. This interpretation assumes lateral homogeneity aided by bottom water currents, and the presence of rip-up interclasts suggests that bottom water currents were, at the very least, sporadic. However, this also depends on the lateral extent (scale) of the system and whether these water currents introduced sulfate-rich water from other deep-water sources. Since active communities of sedimentary sulfate reducers (or the physical process producing water column drawdown of sulfate concentrations) are largely independent of small variances in bottom water currents, the general treatment presented above should apply. In essence, a low sulfate water column can easily become sulfate limited when sedimentary SRB communities are present, and especially where the water column is shallower and organic inputs high, such as in the Rove Fm.
For variations in the intensity of water column sulfate reduction rates, or variations in the intensity of the sedimentary sulfate drawdown mechanism to be feasible explanations, these rates must vary on timescales that are comparable to the response time of the Rove Formation sulfur cycle. This will depend, in part, on water column mixing rates, but will also be influenced by changes in controls on sulfate reduction rates, such as ambient water temperatures and nutrient availability. I infer that the Rove Formation water column had a short response time (months to years), which I estimated using the length-scale expression presented above. Another issue that must be considered in evaluating the plausibility of a sulfate-limited explanation for the
115
Rove Formation data is related to the size of the standing pool of water column sulfide. A large standing sulfide pool would have the potential to damp the isotopic signal of the sulfide reservoir. This leads us to suggest that although these scenarios have the potential to produce variations in the water column sulfate concentration profiles, they would not be sufficient to produce the sustained and significant Rayleigh isotopic signal observed for the sulfides. For these reasons, I do not favor this interpretation.
Possibility of a sulfur cycle with continuous sulfide loss: An alternative model I propose for the euxinic Rove Formation sulfide data is a steady-state sulfur cycle that includes a significant and continuous loss of sulfide. A box model with this structure is presented in Figure 22c. The field of sulfide compositions that can be produced by this box model using experimentally determined sulfur isotope fractionations (e.g., Johnston et al., 2005b; Farquhar et al., 2003) is plotted in Figure 27. In general, as more sulfate and sulfide boxes are added to the model, mass-balance requires that the composition of sulfate moves to the lower right of Figure 27, due to the removal of isotopically depleted sulfides. Sulfide removal can occur through a number of processes, such as oxidation, burial, and various transports pathways out of the system.
Figure 27a presents all the data with the steady-state predictions for the isotopic composition of seawater sulfate (right field) and global sedimentary sulfide, assuming only one sulfide sink (left field). Figure 27b show the evolution of predicted sulfide
116
compositions as additional sulfide sinks are added, with Figure 27b representing a scenario with four sulfide sinks (global pyrite plus all sulfide sinks presented in Figure 22c). Strictly speaking, the model is an incremental batch loss process rather than a Rayleigh process, but the expansion of the sulfide field reflects the loss of 32Senriched sulfur from the system, as it does in Rayleigh systems. More complex models such as those with additional steps for sulfide loss (Figure 27b) are capable of producing an even larger field for sulfide isotopic compositions and
Figure 27: A triple isotope plot of the sulfide data, with a focus on understanding the period of shale deposition. Iron formation data are represented by red circles, the transitional shale with yellow ciclres and the euxinic shale with blue circles. All errors are 2 . A scenario with four sulfide sinks, labeled 1-4 (the global sulfide, with all three additional sulfide sinks in Fig. 22c). The model is calibrated with sulfate reduction fractionations from Farquhar et al. (2003). It can be seen from the figure that as the number of sulfide sinks increases, steady state models can produce more enriched sulfide compositions.
predict isotopic compositions that more closely match that of the Rove Fm. data. This style of explanation would not require the extreme situation required for Rayleigh effects, but only multiple sulfide sinks for the system. This explanation becomes even more achievable (i.e. less sinks required) if the fractionation associated 117
with SRB increases (due to limited experimental data, the upper fractionation used in the modeling is ~ 27 in 34S). For the model represented in Figure 27 to work, it must allow a significant loss of sulfide from the system (>50%), a fraction that is significantly higher than the fraction of sulfide lost to pyrite formation in the Black Sea (e.g., Neriten et al. 2001).
Recently, Kump et al. (2005) suggested that upward excursions of the chemocline in the Proterozoic oceans may have ventilated sulfidic waters in upwelling zones, releasing hydrogen sulfide into the atmosphere. Evidence for a shallow chemocline in some Paleoproterozoic environments is given by recent biomarker evidence that placed the sulfidic/oxic chemocline within the photic zone of the McArthur basin (Brocks et al., 2005). Such a loss process may apply to the Rove Formation sulfur cycle, but would further require that sulfide delivered to the atmosphere be laterally transported away from the basin before returning to the oceanic sulfur cycle (as either sulfide, or more likely as sulfate). Transport implied by the Kump et al. (2005) model provides a way to remove this sulfur from the Rove Formation sulfur cycle; however the possibility also exists that the loss may be associated with lateral heterogeneity, where a gradual isotopic enrichment of seawater sulfate (due to the progressive removal of isotopically depleted sulfide) accompanies the transport of sulfate through the system.
118
Figure 28: A cartoon representing the envisioned shale depositional environment. The focus of this graphic is to constrain the cycling of sulfur through the system, and in doing so, other important elements and cycles may be underappreciated. In the right-most column, a = atmosphere, o = oxic ocean, e = sulfidic (or euxinic) ocean, and s = sediments. This environment is interpreted to have been deep waters than the IF environment, with a larger anoxic portion of the overall water column. In this scenario, lateral transport could move both sulfate and sulfide (sulfide in sulfidic waters). Sulfate reduction would be active in the water column of this environment, which makes this system inherently more dynamic (a sulfur sink in the water column with the capability of migrating). With this in mind, the recycling of organic carbon (OC) and the mixing/diffusion across the oxic/anoxic interface are of vital importance. Recent propositions even suggest that the chemocline could reach the surface of the ocean, venting sulfide directly to the atmosphere (Kump et al., 2005). The outgassing mechanism is a viable solution for satisfying the sulfide data (see text for more detail). The zone of most intense SRB activity and pyrite production would still remain in the upper portions of the sediments, but sulfate reduction would inhabit the water column. The lifetime of sulfide in this system would be much longer than in the IF setting, as the available sulfide would overwhelm the available iron. The profiles at left represent speculative ion profiles for two different sulfur species (sulfate and sulfide). The sulfate concentration would begin to decrease in the euxinic portion of the water column as a function of low, but present rates of microbial sulfate reduction. A similar extension of the compositional gradient would draw down sulfate more quickly as the sediment/water interface is approached. In the sediments, sulfate would be expected to decrease through the zone of active sulfate reduction and level off as methanogensis begins to dominate. The sulfide profile is similar, but not a direct reflection. Potential outgassing to the atmosphere requires a variable amount of sulfide in the surface ocean, which is represented by the wavy line atop the sulfide concentration profile. Sulfide would increase quickly below the chemocline as a function of the diffusive boundary layer with the overlying oxic waters, and due to more labile organics near the surface, followed by slow accumulation through mid-depths. The compositional gradient produced in the sediment system would still intrude into the water column, causing an increase in sulfide near this interface. Below this, sulfide concentrations stabilize and are largely represented by pyrite.
119
Further, reports of sulfide samples with strongly positive 34S from other middle Proterozoic sequences such as the Reward Formation, Wollogorang Formation, and Roper Group (e.g., Hayes et al., 1992,1993; Shen et al., 2002; 2003), may provide a link between the data and global mid-Proterozoic sulfur cycle that had a component of continuous sulfide loss. Tests for this hypothesis would include studies that include 33S and 36S in addition to 34S for these and other localities, where I would predict isotopic compositions similar to those seen in the Rove Formation.
A cartoon describing the interpretation of the Rove Fm sub-cycle is presented in Figure 28. This suggests that the local sulfate reservoir is a combination of euxinic water column sulfate and pore-water sulfate. The sulfate source to this system is seen to be the overlying oxic surface ocean seawater sulfate pool and sulfate from this overlying reservoir would vertically mix into the sulfidic portions of the water column and diffuse into underlying sediments. Sulfate reducing microorganisms populating both the euxinic portions of the water column and the underlying sediments would contribute to the variability observed in the Rove, where fluctuations in the biological controls (temperature, nutrient fluxes and organic rain from surface environments) and sulfate delivery to the sulfidic parts of the sulfur cycle would have lead to development of non steady-state concentration effects, but only minor isotopic variations if the standing pool of water column sulfide was significant. Transport of sulfide out of the system, either as a result of horizontal transfer within the ocean, or as a result of ocean - atmosphere transfer of hydrogen sulfide, would leave the residual reservoir 34S enriched and contribute to the positive
120
34S observed in the shale sulfide reservoir. As suggested above, the Rove Formation may not be unique, as similar processes may have operated in other middle Proterozoic settings.
Dispelling non mass-dependent contributions and calibrating a new relationship
The question of whether a sulfur isotopic composition is entirely the result of massdependent processes, or also possesses a component attributable to non massdependent chemistry, is not a trivial one when interpreting the Precambrian geologic record. Large ( scale) non mass-dependent isotopic signals have been observed in sulfur samples older than 2.45 billion years old. These observed effects have thus far been attributed to gas-phase photochemical reactions in a low O2 atmosphere (e.g., Farquhar et al., 2000; 2001; Farquhar and Wing, 2003; 2005; Ono et al., 2003; Hu et al., 2003; Pavlov and Kasting; 2002; Mojzsis et al., 2003; Bekker et al., 2004; Papineau et al., 2005). Some workers have further argued for the persistence of small non mass-dependent signals into the Paleoproterozoic, where a resolvable, but much smaller signature remains (for samples with ages between 2.4 and 2.0 billion years old; Stage II of Farquhar et al., 2000; Farquhar and Wing, 2003; 2005). The samples studied here are slightly younger than those defining Stage II, but can serve to test the proposed lack of non mass-dependent signatures and presence of a mass-dependent surface sulfur cycle at the time of deposition.
Previous research has suggested that the relationships between 33S and 34S (33S / 34S) and 36S and 33S (36S / 33S) can be used to distinguish between
121
fractionations produced by mass-dependent and non mass-dependent processes (Farquhar and Wing, 2003; 2005; Wing et al., 2004; Ono et al., 2006). Ono et al. (2006) recently argued that isotope effects associated with mass-dependent (MD)
Figure 29: The data plotted in terms of 36S and 33S plot (in ). A representative error bar is reported in the bottom left. The short dashed line is an observational trend noted in Archean sediments (see text for details and citations) and plots with a slope of ~ -1 and is further labeled as MIF for mass-independent fractionation. Also plotted is a regression through the data, which produces a trend of slope -6.28. The observed slope for the data is within 1 of predictions from Ono et al. (2006), who predicted a slope of -6.85. The observed consistency between mass-dependent predictions and the data, reinforced by the gross deviation from the observed Archean line, suggests that these samples were produced by only mass-dependent fractionation processes. I postulate that the observed small-scale variations are the result of a slight decoupling of 33S and 36S fractionations related to both the geochemical redistribution of material (accounted for both in this treatment and in Ono et al., 2006) and a biologically produced effect (modeled here). The later will be addressed in a subsequent study.
geochemical systems would yield a 33S / 34S ~ 0.512-0.515 and a 36S / 33S relationship that would converge on a value of ~ -7. Non mass-dependent evidence 122
suggests a much more variable 36S / 33S (ranging from 36S / 33S of -10.36 [Zmolek et al., 1999] to 1.8 [Farquhar et al., 2001]), however observation of Archean samples suggests a terrestrial record of ~ -1. These data are consistent with the massdependent criteria (e.g., Figure 29), and suggest that non mass-dependent isotopic fractionation effects, and non mass-dependent detrital contributions, can be ruled out at the time that this succession was deposited. The nonzero intercept for the 36S / 33S of ~0.3 suggests a consistent sulfur source to the system with a composition different from the normalization (V-CDT). While the observations are consistent with prevailing hypotheses about modest levels of oxygen in the early to middle Proterozoic, they provide a constraint on the extent of the hypothesized Stage II, which marks the transition period between the non mass-dependent world and one more like the modern. Important to understanding element cycles at this time is accepting the difference between what is implied by Stage II in the sulfur and iron isotope records (Farquhar et al., 2000; Rouxel et al., 2005), as the S-record is based on contributions from low O2 settings and the associated processes whereas the Ferecord is based on oceanic redox conditions.
Conclusions I have argued that the transition from ferruginous to sulfidic ocean conditions circa 1840 My ago is coincident with significant changes in the sulfur cycle. Despite these changes, the data do not require a significant change in seawater sulfate concentration at this transition. The data and models suggest that during the period of IF deposition, the sulfur cycle fell under local diagenetic control and that seawater and 123
pore-water sulfate reached an isotopic steady-state. Isotopic variability within this setting is interpreted as the result of changes in local pyrite burial associated with fluctuations in iron inputs and sulfate reducer activity. These modeling treatments also provide a new means of estimating the isotopic composition of seawater sulfate. This estimate is also consistent with a lack of prominent isotopic contributions from SDB at this time. The data suggest that low seawater sulfate concentrations, coupled with active microbial sulfate reduction and a euxinic water column, destabilized the sulfur cycle that dominated during IF deposition and ushered in a sulfur cycle during deposition of the Rove Formation that included loss of 34S-depleted sulfide. The occurrence of positive 34S sulfides throughout the entire Rove Formation with 33S that are consistent with a local loss process, and 34S-enriched sulfides in other late Paleoproterozoic sequences leads to the suggestion the sustained loss process for 34Sdepleted sulfur was widespread. This may reflect outgassing of water column hydrogen sulfide to the atmosphere (Kump et al., 2005). Further studies of all four sulfur isotopes from other Proterozoic, sulfide-bearing sulfidic successions should provide a test for this hypothesis. The data suggest that the oscillations between different modes of the sulfur cycle from the steady-state regime associated with the Gunflint Formation, Biwibik Formation, Trommald Formation, and Mahnomen Formation, to the continuous loss regime associated with the Rove Formation was not immediate and that at least one reversal preceded the permanent establishment of sulfidic conditions. I recognize that the connections between biology, ocean mixing, ocean-atmosphere exchange, and basin-scale cycling of sulfur during the Proterozoic remain to be better explored. However, the possibility of continuous loss of hydrogen
124
sulfide to the atmosphere, coupled with the potential for a pervasive shallow chemocline, would have undoubtedly influenced myriad elemental cycles and hindered the advancement /evolution of life through this time.
125
Chapter 6: Evidence for a mid-Proterozoic oxygenation
Introduction Detailed carbon and sulfur isotopic and biological studies of Proterozoic surface environments provide a growing picture of an Earth in transition between an anoxic Archean world and that of the more oxic Phanerozoic (Canfield, 2005; Holland, 2006). The character of this transition includes a proposal that the surface ocean during most of the Proterozoic was oxygen-rich due to active oxygenic photosynthesis, whereas the deep ocean remained anoxic, and likely sulfidic because of the activity of sulfate reducing microorganisms (Canfield, 1998; Shen et al., 2002; Poulton et al., 2004; Arnold et al., 2004; Holland, 2006). The stratification of the ocean may have also played a role in biological evolution as the upwelling of anoxic/sulfidic waters would serve as a means of extinguishing aerobic life on the continental shelves. Here, sulfur isotope measurements are used to track the interplay and relative oxygenation of the deep and shallow oceans as a way to better understand both the long-term evolution and oxidation state of the Earths surface and, in turn, the character and diversity of the biosphere.
Sulfur isotope studies (Canfield, 1998; Canfield and Teske, 1996) provide insight into the geochemistry of the Proterozoic oceans, as information regarding biological activity and ocean chemistry is believed to be recorded in these chemical records. Traditional studies of 34S/32S use the range of observed sedimentary sulfide isotopic compositions to infer the composition and relative concentration of seawater sulfate 126
(cf. Canfield, 1998), as sulfate is conventionally perceived as the most faithful recorder of global processes (cf. Claypool et al., 1980). Recent studies (Burdett et al., 1989; Kah et al., 2004; Gellatly and Lyons, 2005) have made strides towards developing new proxies to directly access the composition of seawater, which has reinforced the understanding of sulfate as the 'target' sulfur reservoir for understanding global processes. New developments from measurements of 33S and
36
S in pyrites have broadened the use of pyrites as paleo-environmental indicators.
This is especially important since the analysis of sedimentary sulfides (pyrite and acid volatile sulfur) remains more straightforward, both from the perspective of analytical ease and geologic distribution. In this study, I present 42 sulfur isotope (32S, 33S, 34S, and 36S) measurements of Proterozoic sedimentary sulfides and couple them with a revised model of sedimentary cycling in order 1) to place constraints on the isotopic composition of Proterozoic seawater sulfate, 2) to investigate temporal changes in the activity of the sulfur related biosphere, and 3) to develop a tool for tracking periods of progressive Earth surface oxygenation.
Geologic Setting The sulfide samples presented here are from the McArthur Basin, Northern Australia, which represents a thick and laterally extensive package of relatively unmetamorphosed and un-deformed Proterozoic sedimentary rocks (Scott et al 2000; Abbott and Sweet, 2003). The ~1637 Mya Reward Formation, McArthur Gp. (Page and Sweet, 1998), consists primarily of laminated siltstones and mudstones (Figure 30). The ~ 1431 Mya Velkerri Formation and overlying McMinn Formation
127
(age is on McMinn Fm.: Jackson et al., 1987), both part of the marine Roper Gp., are comprised of mudstones and siltstones deposited in a low energy deep marine environment (Figure 30). These formations are thought to have been deposited in an intracratonic or epicontinental sea containing sulfide-rich bottom waters (Jackson, 1987; Abbott and Sweet, 2003; Shen et al., 2002, 2003).
Figure 30: A schematic representation of the stratigraphic relationship between the McArthur Group and Roper Group, both of which are within the McArthur Basin in present day Northern Australia. Age constraints are from the McMinn Formation, Lynnott Formation and Barney Creek Formation. Shales are in white with more carbonate rich sediments in gray. Samples were taken from two drill cores through the Velkerri Formation (U3 and U4), the top and bottom of the Reward Formation and hand samples from the McMinn Formation.
Results The isotopic compositions of samples from the Reward Formation (n = 10) have a total range from 26.98 to 18.75, -0.025 to -0.056, and -0.25 to -0.59 in 34S,
128
33S and 36S respectively (Table 8). The array extending through the Reward data suggest a 33S versus 36S slope of: 36S = -6.0 ( 3.1) 33S -0.66 ( 0.13). The isotopic composition of sulfide samples from the Roper Gp. vary from 24.72 to 23.44, 0.125 to -0.053, and 0.58 to -2.17 in 34S, 33S and 36S respectively (Table 8). The sulfur isotopic compositions for shale samples (n = 18) from Urapunga 3 (U3) cover a range from 24.72 to -19.95, 0.083 to -0.053, and 0.58 to -1.99 in 34S, 33S and 36S respectively. Similarly, the range for the Urapunga 4 (U4) shales (n = 10) is 22.35 to -23.44, 0.125 to -0.037, and 0.46 to -2.17 in 34S, 33S and 36S respectively. Finally, hand samples from the overlying McMinn Formation (n = 4) range from 13.30 to -8.92, 0.103 to 0.040, and -0.91 to 2.07 in 34S, 33S and 36S respectively. The slope of the array extending through Roper Gp. data (33S versus 36S) is: 36S = -11.1 ( 1.4) 33S -0.62 ( 0.07). The carbon isotopic composition of the Roper Gp. samples was relatively constant, varying from -32.11 to -34.56.
129
Table 8: Isotopic and depth data for the Reward Formation, McArthur Group and Velkerri and McMinn Fms, Roper Gp. Reward Fm. samples were previously described in Shen et al., (2002) and Roper Gp. samples were previously studied in Shen et al. (2003).
130
Discussion
Chemostratigraphic record
I begin by interpreting the new sulfide data in a stratigraphic context. The Reward Fm. is not subjected to this analysis, since the variability recorded in that formation was small. Figure 31 presents the isotopic data (13C, 34S, 33S, and 36S) for the U3 and U4 drill cores. The chemostratigraphic records from U3 and U4 were matched, in terms of depth, based on 34S records of the two cores (with the assumption of similar rates of deposition). Previous interpretations (Logan et al., 1995; Shen et al., 2003) of 34S data from U4 and other cores through the Roper Gp. suggested that more isotopically depleted samples (34S < 0) reflect basinal settings with limited reduction of seawater sulfate (upper portion of the core), whereas heavy 34S sulfides reflect the quantitative reduction of seawater sulfate (in pore waters), likely in a shallower water environment (34S > 10) (seen lower in the core). As presented in Figure 31, I observe a monotonic rise in 33S and an inversely correlated decrease in 36S. These two isotopic relationships are correlated because they reflect the process imparting the fractionations, which in the case of the Velkerri Fm., is likely heavily controlled by microbial sulfate reduction. The organic carbon isotopic composition remains relatively constant throughout the cores. Even though a covariation between 33S and 36S is expected, the coupling of these two records and, their relationship to 13C, should receive more attention as our understanding of these systems improves.
Figure 31: Chemostratigraphic record through the Velkerri Formation, Roper Group. The entire interval reported is sulfidic mudstone. Errors are 2 . Data is presented in Table 8. See text for discussion.
131
132
Constraints on seawater sulfate
The Reward Formation: A previous 34S study of the Reward Fm. (Shen et al., 2002) suggested that the tight grouping of isotopically enriched sulfides pointed to the quantitative reduction of pore-water sulfate, which would have had the same composition as that measured in the sulfides (34S 27 ). The new 33S and 36S analyses are consistent with this hypothesis, as the Reward Fm. sulfides cluster tightly in all four isotopes and suggest a seawater sulfate composition of 34S 27 , 33S -0.05 and 36S -0.25 (gray ellipse in Figure 32). The relative isotopic homogeneity within the Reward Fm. data suggests that sulfate concentrations in this basin were high enough to not be drastically affected by any basin-scale chemical closure or active sulfate sinks, such as microbial sulfate reduction. Although this study is fully consistent with previous 34S predictions, additional information is gained through interpreting the consequences of the minor isotope prediction, as the predicted isotopic composition of seawater suggests the presence of active sulfur disproportionating microorganisms when the Reward Fm. was deposited. The environmental requirement for active disproportionation is the presence of sulfur intermediate species (S0, SO3, S2O3). These intermediates are generally understood to be the product of microbial sulfide oxidation (either phototropic or non-phototropic utilizing O2 or NO3). With biomarker evidence for active purple and green sulfur bacteria in younger portions of the McArthur Basin (Brocks et al., 2006), it is likely that the necessary sulfur intermediate species were present when the Reward Fm. was deposited.
133
Figure 32: A plot of 34S versus 33S with a model sulfide field for the Reward Fm. (gray ellipse in lower right) and Roper Gp. (symbols). Reward Fm. was previously studied in Shen et al. (2002). Roper Gp. samples are from drill cores Urapunga 3 (blue circles) and 4 (red circles), with hand samples from the McMinn Fm. (yellow squares). These sediments were previously interpreted as typical of Proterozoic shelf environments. Errors are 2 . Modeling is derived from Chapter 5. The small field (outlined in black) represents a sulfur cycle with only sulfate reduction active, whereas the larger field (dashed lines) represents a sulfur cycle with both sulfate reduction and sulfur disproportionation active.
The Roper Group: The greater 34S variability observed in the Roper Gp. samples (Velkerri and McMinn Fms.) was previously described as reflecting basinal, distal shelf, and inner shelf facies following traditional models of Proterozoic biogeochemistry (Logan et al., 1995; Shen et al., 2003). The manner in which the data cluster in 34S and 33S allows for a direct assessment of the isotopic composition of contemporaneous seawater sulfate (34S 25 , 33S -0.05 , 36S 0.60 ). As seen in Figure 32, the isotopic scatter in the Roper Gp. data collapses toward a single 34S - 33S composition as 34S increase and 33S decreases (lower right of figure). First, since one isotopic composition of seawater can explain
134
the range of observed sulfide data, the concentration of basin-scale sulfate must have been stable, despite constant drawdown from sulfate reduction. This is similar to the interpretation of the Reward Fm. and the treatment of the Gunflint IF data presented in Chapter 5. Interestingly, even though both the McArthur and Roper Gps. are understood as being deposited in an intracratonic/epicontinental environment, they do not record any indications of closed system effects.
Proterozoic seawater sulfate The 34S record of seawater sulfate through the Proterozoic suggests an overall isotopic stability or slight enrichment (Canfield, 2001; Strauss, 2004). The possible enrichment is thought to reflect the integrated burial of sulfide-rich marine sediments underlying and associated with a long-lived sulfidic ocean (Canfield, 2004). However, a new 33S (and 36S) record provides further insight into the evolutionary track of seawater sulfate, and more importantly, provides information about the processes contributing to that isotopic evolution. Namely, the inclusion of 33S and 36S allows for changes in the biosphere, as well as in the bulk isotopic composition of the surface sulfur cycle (as is called on in the 34S record) to be assessed. When the seawater sulfate predictions generated in this chapter are coupled with other recent estimates (Chapters 4-5; Johnston 2005a, 2006), a more dynamic story for the evolution of seawater sulfate evolution appears (Figure 33a), with a general progression from the ~ 1880 Mya Gunflint Iron Formation to the ~ 1200 Mya Society Cliffs Fm. (and potentially even until ~ 940 Mya Katav Fm.: see Chapter 4). This
135
observed isotopic progression through the mid-Proterozoic can be interpreted in a number of manners, two of which I address below.
Our understanding of microbial fractionations suggests that disproportionation is required to produce the seawater sulfate isotopic compositions measured and inferred for the mid-Proterozoic. This is based on model predictions presented in Chapter 4 and the distinction between a sulfur cycle with active sulfate reduction (SRB) and one with additional fractionations from microbial disproportionation (SRB-SDB). Variability within the SRB-SDB model field was originally interpreted (see Chapter 4) as reflecting changes in either relative pyrite burial (fpy) or the activity of sulfur disproportionation (fr-o). Though I feel this interpretation remains valid (and further apply it below), variability within the SRB-SDB model field could also reflect changes in the expression of microbial fractionations. Thus, the first possible explanation is that the variability within the Proterozoic record reflects changes in the isotopic fractionation produced by sulfate reducers and sulfur disproportionators and not the presence or absence of these metabolic processes. Experimental evidence presented in Chapter 2 highlights the range of disproportionation related fractionations, any of which could be representative of the Proterozoic biosphere. The variability in Proterozoic seawater sulfate may be related to the activity or expression of different types of microbial sulfur disproportionation (sulfite, thiosulfate, or elemental sulfur), as well as possibly indicating the presence of other metabolic process, such as sulfide oxidation.
136
The second possible description of the evolution in Proterozoic seawater sulfate returns to the original model interpretation, where changes in isotopic composition within the SRB-SDB model field are interpreted in terms of fpy and fr-o (also see Chapter 4). In this model, seawater sulfate becomes more isotopically enriched from 1880 1667 Mya, suggesting an increase in the activity of the oxidative sulfur cycle (increasing fr-o). The period from 1667 to 1460 Mya3 is characterized by a decrease in relative pyrite burial (fpy) and is followed by a brief interval (1460-1449 Mya) where fpy held constant but where the activity of the oxidative sulfur cycle decreased (fr-o decreased). During the remainder of the Mesoproterozoic (1429 -1200 Mya), the
Table 9: Here I present minimum estimates for flux values at the formation/time interval level. Numbers in parentheses represent the chapter in which the data was originally presented, with a * referring to Goldman et al. (in review). Uncertainties are given on the ages where n samples were averaged over a time domain. Otherwise, n refers to the number of samples measured in a given formation. Uncertainties on formation and group ages can be found in the text. Rows in regular font are direct measurements of sulfates, whereas rows in italics represent estimates of seawater sulfate from sulfide data.
oxidative sulfur cycle becomes more active, as indicated by increasing fr-o. If I take the maximum possible microbial fractionations, values for fpy and fr-o can be uniquely estimated (presented in Table 9) and the relative changes over geologic time directly
The age of the Dismal Lakes Group was recently revisited by Evans (2006) and determined to be 1460 Mya, as opposed to previously reported estimates (Kah et al., 2004; Chapter 4) of 1300 Mya.
3
137
assessed. Note that these values represent minimums. Although no strong temporal trend is recognized, estimates for environments > 1200 Mya are more so characterized by wide ranges in fr-o (as may be expected) and less variable in fpy. Conversely, changes in fpy and fr-o in younger environments (Neoproterozoic forward) are loosely correlated along a ~ slope 1 line. This observation stands contrary to the expectation that as more sulfide is oxidized (greater fr-o) less pyrite would be available to be buried (lower fpy), suggesting an inverse correlation (slope closer to -1).
The transitional period, where the depositional ages are statistically equivalent (14601429 Mya), may reflect either a true and interpretable loop in the composition of sulfate (as seen in Figure 33a) or simply show the variability within a less stable sulfur cycle that is progressively becoming more oxidizing with time. If the later is the case, then these samples represent a general intermediate state between the reduced Gunflint I.F. sulfur cycle and that of the more oxidizing environment recorded in the Society Cliffs Fm. The data does not require that the gradual midProterozoic oxygenation reflect changes in pyrite burial alone, nor does it require the change in the composition of the surface sulfur cycle (Canfield, 2004). The inclusion of 33S and 36S points to a complex and coupled interplay between pyrite burial, the activity of the oxidative sulfur cycle, and possibly tectonics.
The Neoproterozoic
Sulfur isotope data for the Neoproterozoic also suggests variability in the oxidation state of the Earths surface (Figure 33b). Rather than the transience observed for the
138
Figure 33: A predictive model for 34S versus 33S of seawater sulfate. The red lines enclose a sulfate reduction sulfur cycle whereas the blue lines include sulfur disproportionation. Samples are discussed in text, and ages/labels are taken from Table 9. Ellipses represent 1 and 1 s standard errors on a given population. Modern data (< 50 Mya) is vetted from Farquhar et al. (2002). Insets represent a more lucid picture of the evolutionary track of seawater sulfate. A) The evolution of seawater sulfate in the mid-Proterozoic. B) The evolution of seawater in the Neoproterozoic through today.
139
mid-Proterozoic, 23 measurements of Neoproterozoic seawater sulfate suggest two well defined populations, one representing the pre-Sturtian (earliest) Neoproterozoic and the second recording the post-Marinoan Neoproterozoic/Cambrian world (Kaufman and Knoll, 1995; Kaufman et al., 1997; Hoffman et al., 1998). The relative global consistency within each group of samples (defined temporally) suggests that sulfate concentrations were high enough for sulfate to be globally conservative. In terms of sulfate evolution, the transition from the Meso- into Neoproterozoic is marked by a strong decrease in relative pyrite burial (fpy), possibly associated with a step-wise oxygenation of the atmosphere (Derry et al., 1992; Kaufman et al., 1993; DesMarais et al., 1994). Limited estimates for modern seawater sulfate (Farquhar et al., 2002) suggest a similar composition to that which I observe for the early Neoproterozoic, perhaps indicating either a similarity between these two environments or pointing to an especially stable state of the surface sulfur cycle. The isotopic composition of seawater sulfate at the end of the Neoproterozoic and into the Cambrian reflects a return to a 1200 Mya-like sulfur cycle (see Figure 33b). This would require that the relative fraction of pyrite burial (fpy) increased between the early and late Neoproterozoic. Though the Phanerozoic seawater sulfate record is limited, data suggests that the isotopic composition of seawater sulfate returned to a more modern-like value in the Paleozoic, where it may have remained until today.
The role of tectonics
The observed trend of increasing oxygenation in the mid-Proterozoic is occurring over tectonic timescales. For this reason, I investigate the potential link between the 140
biosphere and the assembly and break-up of a pre-Rodinia supercontinent, Columbia (aka. Nuna). Although not well understood, Columbia was thought to have been initially assembled by 1800 Mya, experienced long-lived subduction related accretion (until ~ 1500 Mya), and have finally broke up in the mid-Proterozoic (1500 Mya until at least 1300 Mya, possibly 1200 Mya: see Lupke and Lyons, 2001 and Zhao et al., 2004 for a review). I consider, as a thought experiment, the potential link between tectonics and the biogeochemical evolution of the sulfur cycle (discussed above). To do so, I adopt a BLAG-like model (Berner et al., 1983), which would suggest that increased spreading rates would result in greater quantities of CO2 released at spreading centers and warmer oceanic crust, which would cause an overall shallowing of oceanic basins. As a consequence, the deep ocean would become shallower and the continental shelves would record marine transgression and actual become generally deeper. The coupled CO2 flux and increase in submerged continental shelf would boost primary productivity and stimulate the biosphere, since a vast majority of marine biological activity occurs in shelf environments (Jorgensen, 2006). Enhanced organic carbon production would fuel sulfate reduction and foster the production of biogenic pyrite. With larger reservoirs of both biogenic pyrite and organic carbon, the amount of available free oxygen would be expected to increase (associated with the removal of these reduced pools). Such an environmental scenario would be one way to account the inferred mid-Proterozoic oxygenation.
141
Following the evolution of 33S versus 33S
In addition to evaluating the relationship between 33S and 34S, added information is captured in 36S / 33S relationships (Johnston et al., 2006; Ono et al., 2006) (Figure 34). Recently, studies of < 2000 Mya mass-dependent environments (Chapter 4-5; Johnston et al., 2005a, 2006; Ono et al., 2006, 2007), geochemical box models (Chapter 5; Johnston et al., 2006), and independent experimental studies (Chapter 3) help to constrain and illustrate the types of isotopic variability possible within a purely mass-dependent record (Figure 34a). The apparent variability is related to both the biological pathways that are present at that time and the changes in massbalance associated with the sulfur cycle itself (model in Figure 34b). Figures 34b and 34c track the evolution the isotopic composition of seawater sulfate (in 36S versus 33S) relative to model predictions for that reservoir (Figure 34b). These data illustrate a wide range of compositions, all of which are 100% mass-dependent (see Figure 6.5a) and can be described within error by the global sulfur cycle model. The consistency of the sulfate values and the model prediction demonstrates a simple understanding of the generation of these compositions; however the scatter within the sulfide data (Figure 34a) points a record potentially rich with information.
142
143
Figure 34: A plot of 33S versus 36S for the Reward Fm and Roper Gp. A) The traditional 1-1 aspect ratio (scale) with the data and lines representing mass-dependent fractionations (MDF) and mass-independent fractionations (MIF). The Roper Gp. samples plot with a slope of: 36S = 11.1 (1.7) 33S + 0.20 (0.09), whereas the Reward Fm. samples plot in a tight cluster with a slope of: 36S = -6.0 (3.1) 33S 0.66 (0.13). B) A model field prediction for the composition of seawater sulfate, where the red line envelopes the sulfate reduction field and the blue line outlines the combined sulfate reduction-sulfur disproportionation field. Also included is a predicted evolutionary track for the composition of seawater sulfate. C) The temporal evolution of seawater, where 1 = Gunflint Fm. (1880), 2 = Reward Fm. (~ 1667 Mya; see text), 3 = Dismal Lakes Gp. (1460 Mya: Evans, 2006), 4 = Helena Fm. (1449 10 Mya: Aleinikoff et al., 1996), 5 = Roper Gp. (1429 30 Mya), 6 = Society Cliffs Fm. (1199 24 Mya: Kah et al., 2001), 7 = early Neoproterozoic, 8 = late Neoproterozoic, 9 = Paleozoic. Estimates of modern seawater from Farquhar et al. (2002) did not include 36S data.
In addition, the temporal evolution observed in Figure 34c is interesting and suggestive. First, the isotopic variability in seawater sulfate in the post-1200 Mya world is significantly less than is observed in the pre-1200 Mya world. This may reflect an increase in the concentration of seawater sulfate and as a result, the relative stabilization of the marine sulfur reservoir4. Second, the locus of the two distinct populations shifts from the upper left to the lower right of Figure 34c, suggesting one of two possible scenarios (or a combination of both). First, if the mid-Proterozoic oxygenation was significant enough to usher in a modern-like sulfur cycle, where sulfate concentrations where in fact relatively high (as suggested above), then the shift in values may reflect the unidirectional (terminal) activation of the fully active (both reduced and oxidative) sulfur cycle. An alternative hypothesis is that the isotopic difference between the two suites records a change in the overall composition of the Earths surface sulfur budget (as was proposed for 34S: Canfield, 2004). If the Gunflint Formation pyrites are characteristic of sulfides from the first half of the Proterozoic, then long-term burial of such a pool would drive the overall composition
4
Higher seawater sulfate concentrations would buffer against significant isotopic variability. However, higher relative seawater sulfate levels does not require the 29 mM indicative of a truly modern sulfur cycle, but perhaps between 1 and 10 mM.
144
of the bulk surface sulfur reservoir (and in turn seawater sulfate) in the same direction as suggested by Figure 34c.
Conclusions Phanerozoic atmospheric oxygen levels are most commonly inferred from the coupled change in the carbon and sulfur isotope records (cf. Berner, 1991). In the Proterozoic, however, reservoir effects (ie. stability and size) have limited the application of this approach and such a proxy for changes in atmospheric oxygen did not appear possible. This new multiple isotope approach, however, holds promise to open the environmental window into Precambrian atmospheric oxygen levels. The data illustrate the sensitivity of this tool and point to a progressive change oxidation state of the mid to late Proterozoic biosphere. Detailed stratigraphic studies of correlative successions from throughout the Proterozoic will answer questions about the timing and causes of the apparent change. I look optimistically to similar studies in the Phanerozoic, and in addition to tracking oxygen levels, these studies will provide a more detailed understanding of processes active in Phanerozoic diagenesis, such as bioturbation.
145
Chapter 7: Conclusions, directions, and closing thoughts
General conclusions The primary goal of this study was to characterize biological sulfur isotope effects and apply this understanding to Proterozoic biosphere reconstructions. Detailed experimental work has determined that sulfate reducing microorganisms and sulfur disproportionating microorganisms, the two metabolic processes thought to be responsible for a majority of the observed 34S fractionation recorded throughout Earth history, produce 33S/32S and 36S/32S fractionations that are resolvable from one another. These experimental fractionations were used as input parameters for a global box model in order to make predictions for the isotopic composition of different surface sulfur reservoirs, when one or both metabolisms are active. Using this approach, the onset of active microbial sulfur disproportionation was pushed back from ~ 800 Mya well into the mid-Proterozoic. Studies of two marine basins adapted this approach and developed a novel means of using sulfide minerals, rather than sulfates (which underpin the study noted above), to determine the character of the basin-scale sulfur cycle and composition of contemporaneous seawater sulfate. These latter studies will allow for a more pervasive application of this tool, as sulfide minerals are more abundant than sulfates.
146
Lessons learned at the cellular level
The results of experimental studies with both microbial sulfate reducers and microbial sulfur disproportionators are compiled in Table 10. The primary conclusion suggests that there is a metabolism-specificity to the 33S and 36S fractionations, as is best represented by the various values (see below). I take the range that these data cover as generally representative of what will be observed in future experiments with these metabolisms. This is more appropriate for sulfate reduction, where the present data set is quite large. What remains to be further understood are the specific controls and mechanisms that govern the production of these signatures. The trend established by this data loosely suggests that SRB < EP (EP: equilibrium prediction) for 33S and SRB > EP for 36S. The opposite is generally true for disproportionation. I speculate that this difference is largely dependent on the direction of the redox reaction, with RP < EP (RP: reductive processes) and OP > EP (OP: oxidative processes).
Table 10: A compilation of experimental fractionation factors (in 33S and 36S) from Chapters 2-3. Both the maximum and minimum values are presented, as well as the number of experiments (n) for each metabolic process.
A new picture of the Proterozoic biosphere
The primary environmental target for applying mass-dependent sulfur isotope fractionations has thus far been the Proterozoic. The overall contributions of this work to the Proterozoic record are seen in Figures 35-37. It can be seen that the 34S
147
record provided by this study is consistent with the range of observations previously reported (Canfield, 2001) whereas the minor isotope record is novel. Also included is
Figure 35: Here I represent a compilation originally presented in Canfield (2001) and redrafted in Figure 1. The data collected throughout this study are plotted in color over the top of the original figure. Measurements of sulfates (seen in blue) correspond well with the predictions of the upper black lines in the Canfield compilation. Plotted in red are measurements of sedimentary sulfides, which nicely correspond with the black diamonds in the original plot. Errors are smaller than the points.
an updated version of Figure 3 (Figure 36), which in addition to highlighting the contribution of this study, illustrates the work done by the community. A new look of the sulfur isotope record of the Proterozoic is presented in Figure 37. In addition to the data, I include an inferred seawater sulfate curve. Excluding one outlier from the Neoproterozoic Doushantou Fm., I observe a temporal decrease in the 33S of seawater sulfate, with greater structure in the Neoproterozoic record. Values of 36S remain relatively stable throughout the Paleo-Mesoproterozoic and fluctuate (in concert with 33S) in the Neoproterozoic. Interestingly, the 36S prediction for the Reward Fm. falls well below the overall temporal trend. The Neoproterozoic
148
structure seen in both 33S and 36S may reflect a response to global, low latitude
Figure 36: Compilation of 33S isotope data, updated from Figure 3. Red circles represent measurements included in this study, whereas gray circles represent published contributions from the community. Errors are smaller than the symbols. Data is from Farquhar et al., 2000, 2002; Hu et al., 2003; Mojzsis et al., 2003; Hu et al., 2003; Ono et al., 2003, 2006a, b, 2007; Bekker et al., 2004; Whitehouse et al., 2005; Johnston et al., 2005a, 2006; Papineau et al., 2005, 2006: Ohmoto et al., 2006; Cates and Mojzsis, 2006.
glaciation events or a change in deep ocean chemistry. In either case, the coincident excursion is provocative. Perhaps even more intriguing is periods where the records may not faithfully follow one another, such as may be the case ~ 1650 Mya. Little is known about the Cambrian/Phanerozoic and the first half of the Paleoproterozoic, as indicated in Figure 37.
The first of the Proterozoic studies focused on testing hypotheses about the timing of microbial innovation: namely the activation of sulfur disproportionation reactions in tandem with the onset of non-photosynthetic sulfide oxidizers. To address this question, numerous proxies for seawater sulfate were measured and used to construct
149
Figure 37: Here I present temporal records for both 33S (A) and 36S (B). All errors are 2 . The legend on the figure describes the details of the points. The only additional point is an estimate of modern seawater (50 Mya average) From Farquhar et al. (2002) noted in A. The solid black bounding line is a qualitative prediction for the isotopic composition for seawater sulfate. These trends are discussed further in the text.
150
a temporal record of 34S and 33S change from 2000 Mya to 500 Mya. When coupled with an experimentally constrained global box model, these data point to active sulfur disproportionation reactions extending into the middle Proterozoic. This contrasts with earlier work which had suggested a ~ 800 Mya arrival for this metabolism. The requirement of non-photosynthetic sulfide oxidation could not be specifically addressed in this study, as numerous oxidation pathways are environmentally relevant and possible.
Direct proxies for seawater sulfate, as used in this study (Chapter 4), are not as abundant in geologic record as the reduced component of the sulfur cycle, sedimentary sulfides. With this in mind, the two follow-up chapters (Chapters 5-6) developed a framework for using sulfide measurements to unveil information about more local, basin-scale sulfur cycling and the contributions from microbial processes. The first study targeted the transition from ferruginous to sulfidic deep-water conditions, which is captured in the 1900-1800 Mya Animike Basin. The sulfur isotope record in the Gunflint Iron Formation and overlying sulfidic black shale of the Rove Formation (within the Animike Basin) suggests a period of stability during the deposition of the Gunflint Fm., but a destabilization of the sulfur cycle during the period of Rove deposition. It is proposed that the destabilization may be related to the combination of low seawater sulfate concentrations and outgassing of a sulfidic water column to the atmosphere; the latter has also been suggested as a mechanism for the Permian-Triassic extinction (Kump et al., 2005). In addition to providing insight into the Paleoproterozoic sulfur cycle, these data and the modeling used to
151
explain them provides a constraint on the isotopic composition of seawater sulfate. A similar approach was applied to the younger sulfidic black shales of the midProterozoic marine Reward Fm. and Roper Gp. These data suggest that sulfur disproportionation reactions were active in these basins, and provides a prediction for the composition of seawater sulfate.
Prior to this work, the changes in the surface biosphere during the second half of the Proterozoic were not well understood. The strongest lines of evidence and associated arguments for describing the evolution of the biosphere during this period was reported in a series of studies by Canfield and friends (Canfield and Teske, 1996; Canfield 1998), which coupled biospheric change to deep-ocean conditions. This dissertation has revisited these earlier works, and expanded on previous conclusions. Taken together, this study provides a detailed picture of progressive biospheric oxidation during the second half of the Proterozoic. This style of oxygenation (progressive, as opposed to step-wise) is further supported by carbon isotope studies, which suggest a similar change (cf. Frank et al., 2003). I note that these conclusions do not negate earlier hypotheses, but begin to provide a more detailed understanding of a complex period o...