Stability, Transformations, and Fate of Residual Mercury at the Inoperative New Idria Mercury Mine, New Idria, California

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF GEOLOGICAL AND ENVIRONMENTAL SCIENCES AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPY

Adam Douglas Jew January 2013

© 2013 by Adam Douglas Jew. All Rights Reserved. Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/jj799kf8676

ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Gordon Brown, Jr, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Scott Fendorf

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Alfred Spormann

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

James J. Rytuba

Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.

iii Abstract: Mercury mining in the California Coast Range has resulted in over a thousand mercury mines of various sizes that have yet to be remediated and are potential point sources of Hg pollution. The New Idria mine, located in San Benito County, California, was the second largest mercury mine in North America and was not remediated following its closure in 1972. Only within the past two years has it been designated as an EPA Superfund site and is currently undergoing remediation. This thesis focuses on the stability of mercury remaining at the site, the types of mercury phases present at the site as well as downstream from it, and the transport of mercury from the site. The stability of cinnabar (HgS), the primary mercury-bearing phase at the New Idria mine and other mercury mines in the California Coast Range, was investigated in the presence of microorganisms found in the acid mine drainage (AMD) system at the site. Geomicrobiological and geochemical studies of the identity and effects of these microorganisms on the solubility of HgS identified Thiomonas species and showed that the AMD bacterial consortium significantly enhances the solubility of cinnabar and metacinnabar, the two common HgS crystalline phases present. -54 -52 These two phases are very insoluble (Ksp = 10 and 10 , respectively) in the absence of these bacteria and are often considered to be relatively nonbioavailable. The phases of mercury present in the waste piles and downstream sediments were analyzed using a combination of sequential chemical extractions and synchrotron-based techniques, including a new low-temperature extended x-ray absorption fine structure (EXAFS) spectroscopy method developed in this project. This new method allows, for the first time, quantification of the amount of elemental mercury present in complex mine waste samples and associated sediments that contain a variety of Hg-bearing phases. When coupled with field- and lab-based evasion studies of Hg vapor from mine wastes, this new approach showed that high mercury evasion rates into the atmosphere from the New Idria and other Hg mine sites in the California Coast Range can be positively correlated with high levels of elemental Hg in the mine wastes. Another new discovery from this study is that freshwater diatoms in the New Idria drainage system are one of the major sinks for mercury. Selective

iv chemical extraction and EXAFS studies of the species of mercury associated with diatom frustules showed that these frustules can stably sequester mercury in low bioavailability forms. These studies also showed that the abundant iron- (oxy)hydroxide (ferrihydrite) nanoparticles present in the New Idria drainage system have very little associated mercury, which contradicts long-held assumptions that sorption of Hg(II) on these nanoparticles is a major Hg sequestration and transport mechanism in the New Idria drainage system. A general conclusion from this study is that the majority of mercury in the New Idria drainage system is in relatively stable, low bioavailability forms. The understanding gained from this study of the stability, forms, and transport of mercury at the New Idria site can be extrapolated to other similar inoperative mercury mine sites in the California Coast Range and should aid in their future remediation efforts.

v

Acknowledgements

As with any major work there are more people that deserve thanks than can be adequately expressed in the space given. First and foremost I would like to thank my advisor, Gordon E. Brown, Jr., for giving me the opportunity to work on a challenging thesis that is wide ranging in scope. I would also like to thank Gordon for allowing me the flexibility to work in scientific areas outside the general expertise of the surface and aqueous geochemistry group, full financial support for this work, and help in developing my critical reasoning skills. Besides thanking Gordon I would like to thank the rest of the Brown research group, both present and past members, for their friendship and advice over the years. I would like to thank James Rytuba of the United States Geological Survey for his help with this work through numerous meetings to discuss aspects about mercury cycling at not only the field site chosen for this work, but also mercury cycling in general. I wish to thank Mark Marvin-DiPasquale and Lisamarie Windham- Myers of the United States Geological Survey for allowing me to work in their laboratory during my first summer at Stanford where I was first introduced to ultra-trace mercury analysis which allowed me to further refine my techniques in the laboratory at Stanford University for this project. Special thanks is needed for Alfred Spormann, Sebastian Behrens, and the rest of the Spormann research group for allowing me to use of their laboratory for a portion of this work as well as teaching me laboratory techniques for working microorganisms. Thanks goes to Scott Fendorf and his research group, both present and past members, for their friendship and use of some of their equipment.

I would like to thank my committee members, Gordon Brown, Jr., Scott Fendorf, James Rytuba, Alfred Spormann, and Chris Francis. Though the make- up of my committee changed prior to my defense, their direction and input on this project over the years has proven invaluable. I would also like to thank Jennifer Wilcox who served as committee chair for my thesis defense.

vi

Numerous collaborators over the years have given me better understanding of science as well as friendship: Mae Gustin (University of Nevada-Reno), Christopher Kim (Chapman University), Guangchao Li (Stanford University), Doug Turner (Stanford University), Bob Jones (Stanford University), Sam Webb (SSRL), John Bargar (SSRL), Joe Rogers (SSRL), Ruben Kretzschmar (ETH-Zurich), and Jan Wiederhold (ETH-Zurich).

The support I have received from my friends and family over the years has been immense and greatly appreciated. My parents have given me so much help and support over the years that I cannot fully express my gratitude to them. They went above and beyond to help me through a very difficult health related issue during my time at Stanford. The guidance that my brother has offered over the years has been truly appreciated. I appreciate the support and encouragement from my grandparents, some of whom are no longer around to celebrate the completion of this dissertation. To old friends and new ones that I have gained during my time at Stanford, I thank you for friendship, guidance, and support over the years.

I wish to thank the Department of Geological and Environmental Sciences at Stanford University for its support over the years. I would like to thank the various funding agencies which make research projects like mine a reality. This work has been funded by the following entities: Stanford Environmental Molecular Science Institute through NSF Grant CHE-0431425, NSF Center for Environmental Implication of Nanotechnology Grant EF-0830093, the McGee Grant, the Jahns Fellowship, and the Lokey Fellowship.

vii

Table of Contents

Abstract iv Acknowledgements vi Table of Contents viii List of Tables xiii List of Figures xv Chapter 1: Introduction to Mercury Geology, the Inoperative New Idria Mercury Mine, and Mercury Analysis Techniques______1

Introduction 2 Background 4 Mercury Geology 4 The Inoperative New Idria Mercury Mine 5 Mercury Analysis Techniques 12 Objectives 18 Summary of Thesis Research 19 Literature Cited 27 Chapter 2: Microbially Enhanced Dissolution of HgS in an Acid Mine Drainage System in the California Coast Range 33

Abstract 34 Introduction 36 Material and Methods 38 Results 48 AMD Solution Chemistry 48 Clone library, qPCR, and Bacterial Isolate 49 Bacterial Dissolution of HgS 50

viii

Effects of Hg on Iron Oxidation 54 Microbial Isolate Impact on HgS Dissolution 55 Discussion 55 Clone Library, qPCR, Isolate, and Carbon Cycling 55 HgS Dissolution 56 Effects of Hg on Fe-oxidation 62 Conclusions 62 Acknowledgements 64 Literature Cited 65 Tables 69 Figures 72 Chapter 3: A Sequential Chemical Extraction and Spectroscopic Assessment of the Potential Bioavailability of Mercury Released From the Inoperative New Idria Mercury Mine, San Benito Co., CA 78

Abstract 79 Introduction 80 Mineralogy and Petrology of the New Idria Mine Site 82 Experimental Methods 83 Results 91 Total Hg Results 91 SCE Results for Field Samples 92 Ferrihydrite Adsorption and SCE Results 93 Diatom Results 94 Spectroscopic Results 96 Discussion 98 Acknowledgements 107 References Cited 108

ix

Tables 111 Figures 113 Supporting Information 118 2-line Ferrihydrite Synthesis, Characterization 119

XRF and -XAFS 119 References Cited 123 Figures 124 Chapter 4: A New Technique for Quantification of Elemental Hg in Mine Wastes and Its Implications for Mercury Evasion Into the Atmosphere 127

Abstract 128 Introduction 129 Experimental Section 132 Results and Discussion 139 Mercury Speciation in Mine Waste and Sediment Samples 139 Correlation Between Hg Speciation and Evasion 140

Implications of Low-temperature Hg LIII-edge EXAFS Spectroscopy for Hg Evasion Data 142 Acknowledgements 144 Literature Cited 145 Table 150 Figures 151 Supplemental Material 154

Structure of -Hg(0) 155 Data Analysis 156

Contamination of -HgS and -HgS with liquid Hg(0) 157 Literature Cited 159

x

Figures 160 Table 166 Appendix I: Ultra-trace Mercury Analysis Protocols 167 Sample Containers and Cleaning Protocol 167 Field Sampling Protocols 170 Tekran 2600 CVAFS Instruction Manual 173 Troubleshooting 187 Literature Cited 192 Appendix II: Theory, Design, and Operation Protocol for Pyrolysis Mercury Analyzer 193 Machine Theory 193 System Schematic and Components 195 Component Specifications 195 Component Settings 202 ChronTrol Programming Scheme 203 System Start-up and Operation 205 Troubleshooting 212 Appendix III: XAS Protocols 217 General Beam Run Protocol 217 Background Subtraction Protocol for EXAFS using SixPACK 228 Least Squares Fitting Protocol 236 Protocol for Making FEFF Theoretical Pathways for Shell-by-shell Fitting of EXAFS Spectra 240 FEFF EXAFS Fitting 245 Appendix IV: X-ray Diffraction Software: JADE Diffraction Analysis Software Tutorial 251 Main Page, Background Subtraction, and Peak Identification 251

xi

Mineral Identification (1 and 2 component samples) 257 Extra Features 258

xii

List of Tables

Chapter 1: Introduction to Mercury Geology, the Inoperative New Idria Mercury Mine, and Mercury Analysis Techniques

Chapter 2: Microbially Enhanced Dissolution of HgS in an Acid Mine Drainage System in the California Coast Range

Table 1: S-oxidizing microorganism isolation medium. 69

Table 2: Real-time PCR primers used in this study. 70

Table 3: Selected metals and anions in the New Idria

AMD water used in the HgS dissolution experiment. 71

Chapter 3: A Sequential Chemical Extraction and Spectroscopic Assessment of the Potential Bioavailability of Mercury Released From the Inoperative New Idria Mercury Mine, San Benito Co., CA

Table 1: Sample descriptions, ages, and bulk mineralogy. 111

Table 2: Hg concentrations and SCE recoveries

for samples used in SCE experiments. 112

Chapter 4: A New Technique for Quantification of Elemental Hg in Mine Wastes and its Implications for Mercury Evasion Into the Atmosphere

Table 1: Sample information for selected Hg-mine

sediments including results from Hg evasion studies

and EXAFS analysis (at both 298K and 77K). 150

Supplemental Material

Table 1: Results of Hamilton’s R-factor ratio

test for all 8 samples obtained by the additions

of -Hg(0) to the fit. 166

Appendix I: Ultra-trace Mercury Analysis Protocols Appendix II: Theory, Design, and Operation Protocol for Pyrolysis Mercury Analyzer

xiii

Appendix III: XAS Protocols Appendix IV: X-ray Diffraction Software: JADE Diffraction Analysis Software Tutorial

xiv

List of Figures

Chapter 1: Introduction to Mercury Geology, the Inoperative New Idria Mercury Mine, and Mercury Analysis Techniques

Chapter 2: Microbially Enhanced Dissolution of HgS in an Acid Mine Drainage System in the California Coast Range

Figure 1: Summary of the bacterial 16S rRNA gene clone library

results of the AMD pond water. 72

Figure 2: 16S rRNA gene copy numbers per mg sediment in a

Sediment core from the New Idria AMD pond as determined by

real-time PCR. 73

Figure 3: Mercury release from HgS bearing materials for oxic

microcosms using the New Idria microbial community. 74

Figure 4: Mercury release from HgS containing materials in

anoxic microcosms using the New Idria microbial community. 75

Figure 5: Oxidation of Fe(II) during aerobic incubation of

living cells. 76

Figure 6: Effect of added Hg to Fe(II) oxidation of the

New Idria microbial community. 77

Chapter 3: A Sequential Chemical Extraction and Spectroscopic Assessment of the Potential Bioavailability of Mercury Released From the Inoperative New Idria Mercury Mine, San Benito Co., CA

Figure 1: Photograph of the New Idria field site where

sample NI-4 was collected along with SEM images of

diatoms in sediments from sample NI-4 and Silver Creek. 113

xv

Figure 2: Plots of sequential chemical extraction results for

New Idria AMD sediment samples. 114

Figure 3: Plots of sequential chemical extraction results for

ferrihydrite adsorbed with Hg(II), diatomaceous earth

samples, and diatom-rich Silver Creek samples. 115

Figure 4: Bulk Hg LIII-edge EXAFS spectra of an orange

AMD pond sediment and Sample NI-2. 116

Figure 5: Shell-by-shell fittings of Hg LIII-edge EXAFS

spectrum for sample NI-2 and a diatom-rich sample from

Harley Gulch. 117

Supporting Information

Figure S1: Map of sample locations at the inoperative

New Idria mine and along San Carlos Creek. 124

Figure S2: X-ray fluorescence maps of AMD settling

pond sediments across an orange/gray sediment boundary.

figure includes Fe and Hg maps along with a Hg-Fe

correlation plot. 125

Figure S3: Hg LIII-edge -XANES data and fit of a

colloidal particle from Figure S2D. 126

Chapter 4: A New Technique for Quantification of Elemental Hg in Mine Wastes and its Implications for Mercury Evasion Into the Atmosphere

Figure 1: Hg LIII-edge EXAFS spectra for

elemental Hg(0) at both room-temperature (liquid Hg(0))

and 77K (-Hg(0)) showing enhancement of amplitudes in

the Hg LIII EXAFS spectrum obtained by slow cooling. 151

xvi

Figure 2: Comparison of k3-weighted EXAFS spectra

of -Hg(0), -HgS, and -HgS. 152

Figure 3: Plots of light:dark evasion data versus both

total Hg concentrations and % liquid Hg(0) for selected

silica-carbonate and hot-springs deposits. 153

Supplemental Material

Figure S1: Shell-by-shell fitting of the -Hg(0) Hg

LIII-edge EXAFS spectrum. 160

Figure S2: Plots of EXAFS spectra for HgS ( and ) at

77K, HgS ( and ) at 298K, and elemental mercury

at 77K with arrows demarking contributions from

elemental Hg to the 77K spectra. 161

Figure S3: X-ray diffractogram of “ultra-pure”

(99.999% pure, trace metal basis) of -HgS sample

from Alfa Aesar. 162

Figure S4: Linear combination fits of both -HgS and -HgS

at 77K using HgS ( and ) spectra at 298K and

elemental Hg at 77K. 163

Figure S5: Fits of Hg LIII-edge EXAFS spectra for eight

samples with various light:dark Hg fluxes. 164

Figure S6: Residual plots for all 8 samples analyzed

for -Hg(0). 165

Appendix I: Ultra-trace Mercury Analysis Protocols

xvii

Appendix II: Theory, Design, and Operation Protocol for Pyrolysis Mercury Analyzer Figure 1: Schematic of pyrolysis Hg analyzer 195 Appendix III: XAS Protocols Appendix IV: X-ray Diffraction Software: JADE Diffraction Analysis Software Tutorial Figure 1: Main window in JADE XRD analysis software. 253

Figure 2: Background subtraction dialog box in JADE. 254

Figure 3: X-ray diffraction peaks identified in JADE. 255

Figure 4: Zoom in on the diffractogram to see what

peaks are missing from initial peak identification. 256

Figure 5: Search/Match dialog box in JADE. 260

Figure 6: Results page for a general search/match

of a New Idria sample. 261

Figure 7: A one component fit for quartz. 262

Figure 8: Zoom in on the diffractogram focusing in

on the major un-matched peaks for the sample. 263

Figure 9: Fit of the New Idria sample with a

secondary component. 264

Figure 10: Final XRD fit of the New Idria acid

mine drainage evaporite deposit. 265

Figure 11: Dialog box for unit cell refinement. 266

Figure 12: Analysis of particle size for ZnS nanoparticles. 267

xviii

Chapter 1

Introduction to Mercury Geology, the Inoperative

New Idria Mercury Mine, and Mercury Analysis

Techniques

Adam D. Jew

Surface and Aqueous Geochemistry Group, Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA

1

Introduction

Mercury is a potent neurotoxin that is widespread throughout the world [1,

2]. Human occupational exposure to Hg or ingestion can lead to significant kidney and neurological impairment that can ultimately result in death. Two famous examples of Hg toxicity to humans include ‘Mad Hatter’ syndrome (made famous by Lewis Carroll in Alice in Wonderland) caused by Hg used to make fur hats and Minamata Disease, which occurred in Minamata, Japan as a result of Hg released into the bay from a nearby chlor-alkali plant [3]. Different forms of mercury show dramatic differences in toxicity to mammals, with organic Hg- compounds (primarily methylmercury) being significantly more toxic than inorganic Hg-compounds [1, 2]. Not only is methylmercury (MeHg) one of the most toxic forms of Hg, but it also has the ability to bioaccumulate up the food chain. Methylmercury in high trophic level animals (tuna, shark, swordfish, etc.), accounts for the majority of current mercury exposure to humans [2]. Because

MeHg is of such prime concern, research into MeHg formation, stability, transformations, and trophic transfer make up a significant portion of the current research on Hg [3-21].

Though MeHg is important, determining the sources of Hg that can potentially be methylated is equally important. Mercury is a unique element because it is the only known metal to be in a liquid state at most temperatures found on the Earth’s surface and easily volatilizes as a gaseous form. The gaseous form of elemental Hg is the reason that areas of the world that have no natural or anthropogenic sources of Hg (e.g., the Arctic and the Florida

2

Everglades) have environmentally harmful concentrations [22-29]. Currently, the majority of Hg being released into the atmosphere is from anthropogenic sources, primarily coal-fired power plants [1]. Other anthropogenic activities contribute large amounts of Hg into the environment including cement manufacturing, steel manufacturing, petroleum refining, municipal waste disposal, chlor-alkali production, and artisanal gold mining [1]. Even though chlor-alkali manufacturing was responsible for the poisoning in Minimata, Japan, in the

1950’s and 1960’s, contamination from chlor-alkali manufacturing is less of an issue today as better catalyst technology and increased environmental regulations have been developed.

Mining operations are responsible for the release of significant quantities of Hg into both the atmosphere and in watersheds [30-44]. The use of mercury in both gold and silver mining has had a lasting effect on the environments where the Hg was mined and where it was used for amalgamation of fine-grained gold and silver. Significant amounts of Hg were mined from European mines (namely

Almaden in Spain, Idrija in Slovenia, and Amiata in Italy) and shipped to the New

World for mining gold and silver, predominately silver, in South America [35].

The mining carried out at Almaden and Idrija has had a lasting environmental impact and is the subject of numerous research projects to determine the phases of

Hg within the leftover material [45-52].

3

Background

Mercury Geology

In North America, Hg mining was dominated by two different Hg mining districts within the California Coast Range. The Hg deposit types found within

California are silica-carbonate- and hot springs-type deposits, both of which are associated with the end of subduction along the California coast and subsequent fracturing of the subducting slab [35, 53]. The breakup of the subduction slab allowed hot fluids from the upper mantle to intrude into the crust, forming silica- carbonate- and hot springs-type Hg deposits. The silica-carbonate alteration and subsequent mercury mineralization resulted in the two largest North American Hg deposits (New Almaden and New Idria) [35]. New Almaden and New Idria differ in both deposition environment and associated minerals from the Almaden-type deposit of the Almaden mine in Spain and the hydrothermal dominated Idrija mine in Slovenia [35, 46, 50]. Because of the differences in geology between the

European Hg deposits and the California deposits, research done at Almaden and

Idrija do not serve as good proxies for determining the stability, transformations, and migration of Hg-bearing waste material from California Hg mines. Due to heavy mining of Hg in the California Coast Range in order to supply Hg for the gold rush in the Sierra Nevada and Klamath-Trinity Mountains in Northern

California in the 1850’s, it is estimated that > 1,000 abandoned Hg mines are scattered through the California Coast Range [31, 33-36]. In order to determine the long-term impact of the un-remediated Hg mines, it is important to select a field site that would serve as a good proxy for other mine sites in California.

4

Though New Almaden and New Idria started production in the 1850s and closed in the early 1970s, the New Almaden mine has been completely remediated while New Idria has had little to no remediation [54, 55]. The remediation of New Almaden has resulted in the site being of little use in understanding Hg behavior in un-remediated sites. Because New Idria has not been remediated it provides a living laboratory that is useful for comparison with other California Hg mines.

The Inoperative New Idria Mercury Mine

Since its closure in 1972, the New Idria mine has had almost no remediation, resulting in the development of an acid mine drainage (AMD) system. An AMD system is common for sulfide-containing ore deposits as the sulfide is oxidized, either biotically or abiotically, to sulfuric acid resulting in a lowering of pH and the release of heavy metals into solution [56]. Due to high concentrations of Fe typically found in AMD waters, abundant Fe-

(oxy)hydroxides precipitate from solution. The Fe-(oxy)hydroxides, usually ferrihydrite or goethite, are known to trap heavy metals (either from entrapment or adsorption) and remove them from solution and it has been presumed that sorption of Hg by ferrihydrite or goethite controls Hg concentrations in AMD impacted streams. The California Coast Range Hg deposits contain several sulfide-bearing phases that lead to the formation of AMD systems. In almost all

Hg mines the main Hg-bearing phases are the HgS polymorphs (cinnabar and metacinnabar) [35, 57, 58]. Within a few Hg mines, particularly the Socrates mine in Sonoma County, CA, the main ore-bearing phase mined was elemental

5

Hg [59]. Other than cinnabar and metacinnabar, common sulfide minerals associated with Hg mines are pyrite and marcasite [35, 58]. These FeS2 polymorphs, along with the HgS polymorphs, are responsible for the AMD systems found at New Idria as well as at other inoperative Hg mines spread throughout the California Coast Range.

The inoperative New Idria mine situated in San Benito Co., CA was the second largest Hg mine in North America [58]. This mine was in operation from

1854 to 1972. The New Idria Hg district was composed of several Hg mines, but the New Idria mine comprised > 98% of the total Hg production for the district

(production data, 1858-1966) [58]. The New Idria mine is situated on the northern edge of a serpentinite massif within a silica-carbonate altered zone along the New Idria thrust fault [58, 60]. The country rock of the New Idria mine is composed of a serpentinite massif (up gradient of the mine), the main Hg ore body (silica-carbonate alternated rock), the Panoche formation (gray shale and brown massive, concretionary sandstone) which surrounds the silica-carbonate rock, and the Franciscan group (Massive, arkosic sandstone, minor shale, chert, and greenstone) found downstream of the mine [58]. Mining at the site was done underground with nearly 100 miles of tunnels in the mineralized zone of the thrust fault. When the mine was closed, almost all the entrances to the mine were sealed with the exception of a single mining adit (Portal #10) that is open to the atmosphere. After mining ceased, an AMD system developed at the site starting underground and exiting through the Portal 10 adit. The AMD system flows over

6 a large mine tailings pile before entering a settling pond and ultimately meeting

San Carlos Creek.

Portal 10 adit (above, August, 2006) is the only exit for water draining the underground mine workings. Hydrogen sulfide gas is constantly released from this adit at all times of the year. The pH and temperature of the water exiting the mines is fairly constant at pH 5.5 and 35oC. The white material seen in the foreground is a magnesium sulfate efflorescence that is mostly present during the dry season.

7

The above photograph is the main settling pond of the New Idria AMD system

(March, 2009, courtesy of Dr. Reuben Kretzschmar). The gray/brown waste pile on the right is the main tailings pile for the mine. The pH of this pond varies throughout the year from pH 2.5-4.5. The orange colored sediment is a mixture of ferrihydrite and schwertmannite that actively precipitates from solution.

8

The above photograph is of former Brown group member sampling San Carlos

Creek ~100 m above the confluence of San Carlos Creek and the New Idria AMD

(July, 2006). San Carlos Creek drains areas that include several remediated Hg mines along with the serpentinite massif. The presence of magnesium silicates and magnesium hydroxide bearing rocks upstream causes San Carlos Creek to have a pH ranging from 9-10.

9

Downstream of the confluence of San Carlos Creek and the New Idria AMD

(above, July 2006) is a wetland environment where cinnabar and elemental Hg droplets are commonly found in bottom of pools by panning. Though the introduction of the New Idria AMD waters should cause the pH of San Carlos

Creek to decrease, the mixing of the AMD waters with the basic waters of San

Carlos Creek and the interaction of the creek with calcite cemented sandstones keep the pH of San Carlos Creek between 7.5 and 8.5.

10

San Carlos Creek ~10 km downstream (above, February, 2010) is usually devoid of any ferrihydrite derived from the New Idria AMD system. Ferrihydrite is found in this portion of the system and downstream in Silver Creek during abnormally rainy years. The pH of this site is generally pH 8 because San Carlos

Creek continues to be in contact with calcite cemented sandstone.

The New Idria field site has allowed a wide range of research projects to be done as part of this thesis. Because the AMD system contains large quantities of bacterial biofilms, the impact of microorganisms on the stability of the main

Hg ore material (cinnabar and metacinnabar) was investigated. The amount of research on microbial interactions with HgS has been minimal. The role of microbes actively breaking down one of the most stable forms of Hg (HgS) has long been hypothesized, but no experimental evidence was provided for this

11 assumption [11, 61, 62]. Attempts to look at microbial interactions with HgS were done using European Hg mines, primarily Idria, Slovenia [51, 52, 63]. The sites selected for these projects have a significantly different geology than the

North American sites making them poor analogs for California mines.

Mercury Analysis Techniques

The form of Hg-bearing phases is important to both the risk of Hg mines to ecosystems downstream of mines and the global environment. Mercury sulfides are considered very stable at the earth’s surface, but other forms

(specifically chloride salts) are quite soluble in solution. Elemental Hg is fairly insoluble in solution but is easily volatilized into the atmosphere. Because of aqueous and gaseous transport is dependent on the form of Hg an understanding of the form present at the New Idria mine site is necessary to determine the stability of different Hg-bearing phases at New Idria and other mine sites.

Currently, there are many direct and indirect methods for determining Hg speciation within different areas of the New Idria field site. X-ray spectroscopy offers a direct method of determining Hg speciation within Hg-contaminated sediments and tailings, but is constrained by concentration limitations. Indirect methods do not have the concentration limitation but are hindered by difficulties in data interpretation. To have a better grasp of the Hg-bearing phases present at

New Idria, a combination of direct and indirect methods is necessary. By coupling numerous techniques a more complete understanding of Hg cycling at the New Idria is seen.

12

Numerous analytical techniques have been used to determine the speciation of mercury in sediments, but the inherent limitations of each technique make such a study of low concentrations of Hg-bearing phases in mine wastes and

Hg-polluted sediments difficult. Ambient-temperature extended x-ray absorption fine structure (EXAFS) spectroscopy has been successfully used to quantitatively determine the speciation of crystalline Hg-bearing phases within Hg-polluted sediments and sediment leachates when Hg concentrations are ≥ 50 ppm [31, 33,

34]. We have recently developed a new low-temperature EXAFS spectroscopy technique that allows determination of elemental Hg, Hg(0), within sediments

[32], but the detection limit for this phase in Hg-polluted sediments is similar to that for crystalline Hg-bearing phases using ambient-temperature EXAFS spectroscopy (≥ 50ppm, depending on sample matrix). Application of this method to New Idria mine wastes and sediments showed that Hg(0) commonly comprises

10% of the total Hg-bearing phases [32].

Synchrotron-based x-ray fluorescence (XRF) and μ-x-ray absorption near edge structure (μ-XANES) spectroscopy are useful techniques for locating areas of high mercury concentration in representative thin-sections of mine waste samples and identifying different Hg-bearing species; however, like EXAFS spectroscopy, Hg concentration is an important limitation to analysis by -

XANES spectroscopy. When trying to analyze a single particle by -XANES spectroscopy, particle size (> 2 m) becomes an additional limitation, with particles smaller than 2 m in diameter being below the spatial resolution of this technique. Microscopic techniques, including both scanning electron microscopy

13

(SEM) and transmission electron microscopy (TEM), can be used to locate and determine the identity of Hg-bearing phases (using selected area diffraction patterns and energy dispersive compositional analysis), but both SEM and TEM are hampered by the low concentration of Hg-bearing phases, and may be statistically limited because of the relatively small amount of sample examined

[34]. The use of SEM and TEM to identify Hg-bearing phases is further limited due to the necessity of high vacuum conditions during analysis, which could negatively impact the detection of Hg(0) within samples due to potential Hg volatilization.

Two major indirect methods for determining Hg speciation in sediments are pyrolysis and sequential chemical extractions (SCEs). Pyrolysis involves heating a sample and measuring the temperature at which Hg is released in order to identify the Hg-bearing phases present in the sample [42, 47, 48, 64-67].

However, because of the overlap of Hg release peaks for different Hg-bearing phases during pyrolysis, quantitative identification of these phases is difficult [64,

65]. Sequential chemical extraction involves treating samples with chemicals of increasing harshness, after which the supernatant of each chemical extractant step is analyzed for total Hg. The major advantage of the SCE method is that the detection limit is much lower than that of most other techniques, with a range of

0.1-5 ng/g for all extracted fractions. This sensitivity allows for analysis of Hg in nearly all natural sediments and tailings, making it the best technique currently available for samples with low Hg concentrations [34, 36, 67-69]. The Hg-SCE method of Bloom et al. [69] was developed to determine the biogeochemically

14 relevant fractions of inorganic Hg within sediments rather than the identity of specific Hg-bearing phases. Although the SCE technique has a very low detection limit, there are major drawbacks for all SCE protocols. It has been shown that even for a simple set of chemicals, such as those used in the Bloom et al. method, new phases of Hg not present in the original sample are formed [68, 69].

Another major drawback is that multiple Hg-containing phases are released with most of the chemicals extractants, making identification of specific Hg phases difficult. Nonetheless, the Hg-SCE method developed by Bloom et al. [69] does provide some insight into the biogeochemically relevant Hg fractions within Hg- bearing systems such as the New Idria mine drainage system. Work by Kim et al.

[68] showed that the water soluble Hg(II) and HgS/HgSe fractions are not affected by the reagents, allowing one to distinguish among water soluble Hg(II), outer-sphere adsorbed Hg(II), and HgS/HgSe fractions with a high degree of confidence.

Though there are several techniques available for determining Hg speciation in sediments, each has limitations thus requiring the use of more than one technique to get the best possible understanding of Hg speciation. XAS techniques are among the best available for Hg analysis since they allow direct identification of Hg species, but XAS has a detection limit of ≥ 50 ppm for Hg and requiring that ≥ 5% of a Hg species is present in order to detect it and identify the Hg-bearing species. In order to get around this limitation, it is necessary to combine XAS with other, indirect, speciation techniques. Pyrolysis is a potential compliment to XAS, but given the special equipment required for analysis and the

15 difficulty in data interpretation, this technique is not ideal. A better approach is to couple with XAS is SCE. SCE provides the low detection limit (both total Hg concentration and percentage of phase present) that XAS does not. Because of the difficulty in interpreting the operationally defined SCE fractions, XAS use is required whenever possible.

One situation where coupling XAS and SCE is necessary is for a hypothetical mine waste sample containing Hg that is released during the aqua regia step, which dissolves both HgS and HgSe, if present. SCE analysis will help determine if HgS (cinnabar, metacinnabar, or both), HgSe, or both HgS and

HgSe are present in the sample, but it is not capable of distinguishing between

HgS and HgSe. XAS analysis on the same sample, assuming the total Hg concentration is high enough, can determine whether cinnabar, metacinnabar, or

HgSe are present in the sample and what percentage of each phase is present. The other side of this hypothetical scenario is for the Hg fractions that are below 5% of the total Hg present in the sample. Suppose that 2% of the total Hg in the sample is released by the relatively mobile “simulated stomach acid” SCE step.

This fraction of the Hg would be of significant environmental concern, since it is considered to be highly bioavailable, but it would not be detected by XAS because the phase is too low in amount to be detected. SCE analysis provides speciation information for Hg in samples, and thus information on the overall stability of Hg in a sample that XAS (combined with thermodynamic literature values) can only hint at or may not be able to assess when Hg concentrations or

Hg-containing phases are too low for XAS analysis.

16

Besides SCE coupled with XAS analysis, imaging techniques (SEM and

TEM) are useful for Hg speciation studies. These techniques allow the imaging and identification of Hg-bearing particles on a nanometer scale. By using EDAX, it is also possible to identify elements that are associated with the Hg-bearing particles. The drawbacks of these techniques are numerous however. Both techniques are carried out in an UHV environment that does not allow for the analysis of wet samples or samples containing elemental Hg due to its volatility under UHV conditions. Given the small spatial area investigated with these imaging techniques, finding Hg-bearing particles in highly dispersed heterogeneous materials is very time consuming and difficult. Despite these limitations both imaging techniques should be used in conjunction with XAS and

SCE to provide a more complete picture of Hg speciation.

As mentioned earlier, previous work done on New Idria samples, including tailings, calcine, and sediments, using EXAFS spectroscopy has shown that the main Hg-bearing phases are cinnabar, metacinnabar, corderoite, eglestonite, montroydite, and elemental Hg [32-34]. However, the identities of dominant Hg-bearing phases in the New Idria AMD system are not known. Such information is essential for assessing the site’s potential as a continuing point source for Hg for downstream sites.

Unlike Hg release through watersheds that affect downstream environments, atmospheric release of Hg from mine sites is a global problem.

Though mine sites have lower Hg release rates to the atmosphere than a coal-fired power plant, the larger aerial extent of mine waste material makes mines a larger

17 concern than most realize. It has been shown that different mines as well as different types of waste material have radically different volatilization fluxes [30,

41, 42, 70-72]. The difference in volatilization fluxes becomes even more pronounced when samples are exposed to light compared with fluxes in the dark

[42]. Trends between the Hg fluxes and the Hg-phases present in waste materials have been difficult to determine. As part of this thesis, a new low-temperature x- ray technique was developed for this work that allows the direct quantification of elemental Hg within sediments [32]. This new x-ray analysis method gives a more complete picture of Hg release from abandoned Hg mines within the

California Coast Range.

Objectives:

This thesis consists of the following three major objectives separated into three different chapters:

1) Determine the stability of residual Hg at the New Idria inoperable Hg

mine in an aerobic AMD system. The prime focus of this work is to

determine the stability of HgS in the presence of microorganisms

contained in the AMD system.

2) Determine the speciation of Hg in sediments downstream of the mine site

in the AMD system. This work is to determine the type of Hg being

released from the site and its potential bioavailability to downstream

ecosystems.

18

3) Understand the role of Hg volatilization to the atmosphere at New Idria.

This work includes the development of a new low-temperature EXAFS

technique to identify elemental Hg in waste materials and compare Hg

speciation to volatilization studies to better understand what major Hg

phase is released into the atmosphere from these mine sites.

The work carried out in these three areas has led to an improved understanding of

Hg stability, transformation, and transport in the New Idria AMD system. By using New Idria as a case study, it can serve as a proxy for the 1,000+ inoperative

Hg mines spread throughout the California Coast range.

Summary of Thesis Research

The following thesis is composed of three major chapters that are written and structured as manuscripts that are published, submitted, or ready for submission in scientific journals. For two of the chapters supporting information that was originally included in the manuscript is included at the end of the chapter.

Following the chapters are four appendices detailing protocols developed throughout the course of this dissertation for ultra-trace Hg analysis, x-ray absorption spectroscopy (XAS), and x-ray diffraction (XRD) analysis. All work presented was conducted primarily by the author, with co-authors being cited accordingly. Each chapter is written as a complete manuscript with its own abstract, methods sections, results, discussion/conclusions, and references. The subject matter of each chapter or appendix is as follows:

19

Chapter II: Microbially Enhanced Dissolution of HgS in an Acid Mine Drainage

System in the California Coast Range

Chapter 2 investigates the effect that microbial biofilms present in the

New Idria AMD system have on the stability of HgS crystals remaining in waste material. The ability of microorganisms to dissolve stable HgS in aerobic systems has been hypothesized, but until now it has not been experimentally proven.

Mine waste from New Idria was incubated with the addition of microorganisms from the biofilm, and total Hg was measured over time to determine if microorganisms would increase the low solubility of HgS. To try and determine which bacterium or bacteria could be responsible for the dissolution of HgS, a

16S rRNA clone library was constructed of the AMD biofilm. Once a likely candidate was found in the clone library, qPCR was completed using the genetic sequence of the candidate in order to know the abundance of the candidate bacterium in the AMD biofilm community. Because Fe cycling is important to the New Idria AMD system, the efficiency of Fe-oxidation was tested with increasing Hg concentrations to see if HgS dissolution could impact Fe-oxidation.

The New Idria microbial biofilm has the capacity to dissolve HgS and release Hg into solution. The biofilm increased in HgS solubility, when compared to the idealized solubility products of cinnabar and metacinnabar (10-54 and 10-52, respectively) [73, 74], by 28 to 31 orders of magnitude throughout the experiments. The results of the 16S rRNA clone library showed the biofilm containing primarily Fe- and S-oxidizing microorganisms. A single S-oxidizing bacterium was detected in the clone library, Thiomonas sp., and was isolated.

20

HgS dissolution using only the Thiomonas sp. isolate were unsuccessful suggesting that a specific chemical or chemicals was missing from the growth medium and/or another bacterium or group of bacteria are required for HgS dissolution. Iron oxidation in the system was retarded with the addition of Hg(II) to the system with Hg concentrations of 100 g/L or greater inhibiting Fe- oxidation.

Chapter III: A Sequential Chemical Extraction and Spectroscopic Assessment of the Potential Bioavailability of Mercury Released From the Inoperative New Idria

Mercury Mine, San Benito Co., CA

Chapter 3 is a detailed wet chemical and spectroscopic study of AMD sediments at the mine site as well as samples taken tens of kilometers downstream of the site. Sequential chemical extractions (SCEs) of AMD sediments taken during the wet and dry season were done to investigate the potential bioavailability of Hg in sediments as well as to try and identify Hg speciation.

The presence of ferrihydrite throughout the system made it necessary to carry out

Hg adsorption and subsequent SCE experiments to synthetic and natural (New

Idria derived) ferrihydrite. Due to the presence of fresh-water diatoms in the system, SCE experiments were conducted on a suite of diatomaceous earth and diatom-rich samples to determine which SCE extraction step removed Hg from diatoms. For AMD samples containing > 50 ppm Hg, several synchrotron based techniques (EXAFS, -XANES, and -XRF) were used to further refine Hg speciation.

21

Sequential chemical extractions of the New Idria sediments resulted in the majority of Hg (> 97%) in all samples was removed with 1M KOH or harsher chemical extractant. Mercury uptake onto synthetic and natural ferrihydrite and subsequent SCE analysis showed that Hg is not binding very strongly to either material. The maximum adsorption to synthetic ferrihydrite was 60% at pH 7.45 and 65% (also at pH 7.45) for natural ferrihydrite. Greater than 90% of the Hg adsorbed to the ferrihydrite was removed by a weak chemical extractant (1M

MgCl2), further emphasizing the lack of strong adsorption. SCE analysis of diatomaceous earth and diatom-rich field samples showed that the vast majority of the Hg in the samples was released by 1M KOH. Because of the Hg in diatoms being dominated by the 1M KOH fraction and lack of significant Hg in the 1M

KOH fraction for the Hg adsorbed to natural ferrihydrite, diatoms are most likely removing Hg(II) from solution either through adsorption or incorporation and storing the Hg as a relatively stable Hg-organic phase.

Chapter IV: A New Technique for Quantification of Elemental Hg in Mine

Wastes and Its Implications for Mercury Evasion Into the Atmosphere

Chapter 4 details the development of a new EXAFS technique to identify for the first time elemental Hg contained in mine waste samples. A slow cooling of samples through the freezing point of elemental Hg was required to have liquid

Hg(0) crystallize into -Hg(0). By forming -Hg(0), EXAFS spectra could be collected for field samples which allowed us to quantify the amount of elemental

Hg in each sample by linear combination fitting (LCF) of EXAFS spectra from mine waste samples. This new low-temperature EXAFS analysis method was

22 applied to samples previously analyzed by Kim et al. [31, 33], which had been previously analyzed at ambient-temperature, to determine if the Hg speciation identified was correct. Hg volatilization studies from the same mine waste samples were compared to the Hg speciation using low-temperature EXAFS analysis to see if there is any identifiable relationship between speciation and volatilization rates.

By slow cooling elemental Hg through its crystallization temperature, a well structured EXAFS was collected. Shell-by-shell fitting of the -Hg(0) spectra confirmed that the spectra collected was indeed pure -Hg(0) without any other Hg phases present. Using this new -Hg(0) spectra in linear combination fitting of slow cooled field samples, elemental Hg was able to be quantified for the first time using EXAFS analysis. When low-temperature fits were compared to ambient-temperature fits of the same samples, it was apparent elemental Hg was missing from the Hg speciation of numerous samples. The Hg speciation was compared with Hg volatilization rates from samples to determine if there was a correlation between the rate of Hg volatilization and a specific Hg species identified in samples. Because of the large difference in Hg volatilization fluxes between samples and the drastic difference in flux values when samples were exposed to light versus dark conditions, it is useful to compare Hg speciation to the ratio of Hg fluxes during light and dark analysis, Light:Dark. When the atom

% of elemental Hg(0), in the form of -Hg(0), is compared to the Light:Dark ratio, two distinct linear relationships based on deposit geology are seen. Hot- springs-type deposits have a steeper linear relationship than silica carbonate-type

23 deposits. The steeper linear relationship for the hot-springs-type deposit is likely due to smaller elemental Hg bead sizes and less encapsulation of the elemental Hg in rock particles.

Appendix I: Ultra-trace Mercury Analysis Protocols

This appendix is a detailed explanation of what is necessary for ultra-trace mercury analysis and the operation of the Tekran 2600 CVAFS Hg analysis system. The protocols cover the type of containers to use for sample storage, cleaning protocols, types of filtration equipment to use, and sample preservation.

The section on the Tekran 2600 includes detailed instructions for ultra-trace analysis, instrument instructions, and troubleshooting.

Appendix II: Theory, Design, and Operation Protocol for Pyrolysis Mercury

Analyzer

This appendix describes the theory, design, and operation of a furnace- based Hg analyzer built in the laboratory. The Tekran 2600 is designed for liquid samples, whereas the pyrolysis analyzer uses heat to remove Hg from samples for analysis in solid dry samples. Because this system was built in the laboratory, a detailed list of all parts, schematics, and system settings is included in this appendix.

Appendix III: XAS Protocols

Due to the large amount of XAS analyses carried out in this thesis research, a detailed discussion of the protocols for XAS analysis is included. This

24 appendix is separated into several sections, including: 1) Beam line operation, 2)

Data averaging and energy calibration, 3) Background subtraction, 4) Least squares fitting (Linear combination fitting), 5) FEFF theoretical EXAFS pathway creation, and 6) Shell-by-shell fitting using the FEFF generated pathways. The beam line operation portion of the appendix is specific for data collected on

Stanford Synchrotron Radiation Lightsource (SSRL) beam line 11-2. It is a step- by-step protocol to be used to collect high quality data, along with explanations of the purpose of de-tuning the beam, windowing the detector, etc. Detailed discussion of data averaging using the SixPACK XAS software [75, 76] is included to point out the importance of dead time correction to the data, determining energy glitches, and proper energy calibration of data using an energy calibration foil. Because poor background subtraction can create or remove features from the EXAFS data, a detailed explanation of the background subtraction procedure for both LCF and shell-by-shell fitting is presented. In most natural systems, the dominant fitting protocol is LCF. This section goes step by step through the process of fitting data along with outlining additional features of the program. Shell-by-shell fitting is used primarily to analyze adsorbed metals and disorder in crystalline materials. Shell-by-shell fitting of EXAFS data requires the use of FEFF generated theoretical pathways [77]. FEFF is a computer code that was developed to use crystallographic information combined with core- hole effects to create theoretical models for EXAFS, full multiple-scattering calculations for various x-ray absorption spectra (XAS), projected local densities of states (LDOS), and x-ray near edge structure (XANES) [78]. These theoretical

25 pathways are then used in shell-by-shell fitting to determine the backscattering atom, the distance between the absorbing and backscattering atoms, the number of backscattering atoms at a given distance, and the amount of thermal/static disorder of the system. This appendix includes a detailed discussion of how to go from the crystallographic data for a mineral or molecule to creating FEFF theoretical pathways for shell-by-shell fitting. Due to the complexity of shell-by- shell fitting a detailed protocol is included for fitting EXAFS data. This section includes definitions of the parameters used in fitting, how changing values of the parameters changes the fit, and what constitutes a good fit.

Appendix IV: X-ray Diffraction Software: JADE Diffraction Analysis Software

Tutorial

The JADE diffraction analysis software is one of the most user unfriendly programs available. This tutorial was developed to help with background subtraction, peak identification, and phase identification of diffractograms collected using either a bench x-ray diffractometer or x-ray diffraction data collected from a synchrotron source. This tutorial includes how to fit one and multi-component samples, refinement of unit cell parameters, and determining crystal sizes for pure specimens with crystal sizes > 100 nm.

26

Literature cited

1. United States Environmental Protection Agency, Mercury Study Report to Congress Washington D.C., 1997; p 1811. 2. United States Department of Health and Human Services, Toxicological Profile of Mercury, Agency for Toxic Substances and Disease Registry, Atlanta, 1999; Vol. 1, p 676. 3. Takizawa, Y.; Osame, M., Understanding of Minamata Disease: Methylmercury Poisoning in Minamata and Niigata, Japan. Japan Health Association: Tokyo, 2001; p 154. 4. Balogh, S. J.; Nollet, Y. H.; Swain, E. B., Redox Chemistry in Minnesota Streams during Episodes of Increased Methylmercury Discharge. Environmental Science & Technology 2004, 38, (19), 4921-4927. 5. Benoit, J. M.; Gilmour, C. C.; Mason, R. P.; Heyes, A., Sulfide Controls on Mercury Speciation and Bioavailablity to Methylating Bacteria in Sediment Pore Waters. Environmental Science & Technology 1999, 33, (6), 951-957. 6. Benoit, J. M.; Mason, R. P.; Gilmour, C. C., Estimation of Mercury-Sulfide Speciation in Sediment Pore Waters Using Octanol-Water Partitioning and Implications for Availability to Methylating Bacteria. Environmental Toxicology and Chemistry 1999, 18, (10), 2138-2141. 7. Compeau, G. C.; Bartha, R., Sulfate-Reducing Bacteria: Principal Methylators of Mercury in Anoxic Estuarine Sediment. Applied and Environmental Microbiology 1985, 50, (2), 498-502. 8. Drott, A.; Lambertsson, L.; Bjorn, E.; Skyllberg, U., Importance of Dissolved Neutral Mercury Sulfides for Methyl Mercury Production in Contaminated Sediments. Environmental Science & Technology 2007, 41, (7), 2270-2276. 9. Drott, A.; Lambertsson, L.; Bjorn, E.; Skyllberg, U., Effects of Oxic and Anoxic Filtration on Determined Methyl Mercury Concentrations in Sediment Pore Waters. Marine Chemistry 2007, 103, 76-83. 10. Drott, A.; Lambertsson, L.; Bjorn, E.; Skyllberg, U., Do Potential Methylation Rates Reflect Accumulated Methyl Mercury in Contaminated Sediments? Environmental Science & Technology 2008, 42, (1), 153-158. 11. Fagerstrom, T.; Jernelov, A., Formation of Methyl Mercury from pure Mercury Sulphide in Aerobic Organic Sediment. Water Research 1971, 5, 121-122. 12. Fleming, E. J.; Mack, E. E.; Green, P. G.; Nelson, D. C., Mercury Methylation from Unexpected Sources: Molybdate-Inhibited Freshwater Sediments and an Iron- Reducing Bacterium. Applied and Environmental Microbiology 2006, 72, (1), 457-464. 13. Hayashi, K.; Kawai, S.; Ohno, T.; Maki, Y., Photomethylation of Inorganic Mercury by Aliphatic -Amino-acids. Journal of the Chemical Society-Chemical Communications 1977, 158-159. 14. Hintelmann, H.; Keppel-Jones, K.; Evans, R. D., Constants of Mercury Methylation and Demethylation Rates in Sediments and Comparison of Tracer and Ambient Mercury Availability. Environmental Toxicology and Chemistry 2000, 19, (9), 2204-2211.

27

15. King, J. K.; Kostka, J. E.; Frischer, M. E.; Saunders, F. M., Sulfate-Reducing Bacteria Methylate Mercury at Variable Rates in Pure Culture and in Marine Sediments. Applied and Environmental Microbiology 2000, 66, (6), 2430-2437. 16. Mason, R.; Bloom, N.; Cappellino, S.; Gill, G.; Benoit, J.; Dobbs, C., Investigation of Porewater Sampling Methods for Mercury and Methylmercury. Environmental Science & Technology 1998, 32, (24), 4031-4040. 17. Qian, J.; Skyllberg, U.; Tu, Q.; Bleam, W. F.; Frech, W., Efficiency of Solvent Extraction Methods for the Determination of Methyl Mercury in Forest Soils. Fresenius' Journal of Analytical Chemistry 2000, 367, 467-473. 18. Simmons-Willis, T. A.; KOH, A. S.; Clarkson, T. W.; Ballatori, N., Transport of a Neurotoxicant by Molecular Mimicry: The Methylmercury-L-Cysteine Complex is a Substrate for Human L-type Large Neutral Amino Acid Transporter (LAT) 1 and LAT2. Biochemistry Journal 2002, 367, 239-246. 19. Skyllberg, U.; Qian, J.; Frech, W.; Xia, K.; Bleam, W. F., Distribution of Mercury, Methyl Mercury and Organic Sulphur Species in Soil, Soil Solution and Stream of a Boreal Forest Catchment. Biogeochemistry 2003, 64, 53-76. 20. Ullrich, S. M.; Tanton, T. W.; Abdrashitova, S. A., Mercury in the Aquatic Environment: A Review of Factors Affecting Methylation. Critical Reviews in Environmental Science and Technology 2001, 31, (3), 241-293. 21. Windham-Myers, L.; Marvin-DiPasquale, M.; Krabbenhoft, D. P.; Agee, J. L.; Cox, M. H.; Heredia-Middleton, P.; Coates, C.; Kakouros, E., Experimental removal of wetland emergent vegetation leads to decreased methylmercury production in surface sediments. Journal of Geophysical Research 2009, 114, 1-14. 22. Aiken, G.; Haitzer, M.; Ryan, J. N.; Nagy, K., Interactions between dissolved organic matter and mercury in the Florida Everglades. Journal of Physics IV France 2003, 107, 29-32. 23. Benoit, J. M.; Mason, R. P.; Gilmour, C. C.; Aiken, G. R., Constants for Mercury Binding by Dissolved Organic Matter Isolates from the Florida Everglades. Geochimica et Cosmochimica Acta 2001, 65, (24), 4445-4451. 24. Drexel, R. T.; Haitzer, M.; Ryan, J. N.; Aiken, G. R.; Nagy, K. L., Mercury(II) sorption to Two Florida Everglades Peats: Evidence for Strong and Weak Binding and Competition by Dissolved Organic Matter Released from the Peat. Environmental Science & Technology 2002, 36, (19), 4058-4064. 25. Ravichandran, M.; Aiken, G. R.; Ryan, J. N.; Reddy, M. M., Inhibition of Precipitation and Aggregation of Metacinnabar (Mercuric Sulfide) by Dissolved Organic Matter Isolated from the Florida Everglades. Environmental Science & Technology 1999, 33, (9), 1418-1423. 26. Reddy, M. M.; Aiken, G. R., Fulvic Acid-Sulfide Competition for Mercury Ion Binding in the Florida Everglades. Water, Air, and Soil Pollution 2001, 132, 89- 104. 27. Douglas, T. A.; Sturm, M.; Simpson, W. R.; Blum, J. D.; Alvarez-Aviles, L.; Keeler, G. J.; Perovich, D. K.; Biswas, A.; Johnson, K. P., Influence of Snow and Ice Crystal Formation and Accumulation on Mercury Deposition to the Arctic. Environmental Science & Technology 2008, 42, (5), 1542-1551. 28. Horton, T. W.; Blum, J. D.; Xie, Z.; Hren, M.; Chamberlain, C. P., Stable Isotope Food-web Analysis and Mercury Biomagnification in Polar Bears (Ursus maritimus). Polar Research 2009, 28, 443-454.

28

29. Sherman, L. S.; Blum, J. D.; Johnson, K. P.; Keeler, G. J.; Barres, J. A.; Douglas, T. A., Mass-independent Fractionation of Mercury Isotopes in Arctic Snow Driven by Sunlight. Nature Geoscience 2010, 3, 173-177. 30. Gustin, M. S.; Coolbaugh, M. F.; Engle, M. A.; Fitzgerald, B. C.; Keislar, R. E.; Lindberg, S. E.; Nacht, D. M.; Quashnick, J.; Rytuba, J. J.; Sladek, C.; Zhang, H.; Zehner, R. E., Atmospheric Mercury Emissions from Mine Wastes and Surrounding Geologically Enriched Terrains. Environmental Geology 2003, 43, 339-351. 31. Kim, C. S.; Brown Jr., G. E.; Rytuba, J. J., Characterization and Speciation of Mercury-Bearing Mine Wastes Using X-ray Absorption Spectroscopy. The Science of the Total Environment 2000, 261, 157-168. 32. Jew, A. D.; Kim, C. S.; Rytuba, J. J.; Gustin, M. S.; Brown Jr., G. E., A New Technique for Quantification of Elemental Hg in Mine Wastes and Its Implications for Mercury Evasion Into the Atmosphere. Environmental Science & Technology 2011, 45, (2), 412-417. 33. Kim, C. S.; Rytuba, J. J.; Brown Jr., G. E., Geological and Anthropogenic Factors Influencing Mercury Speciation in Mine Wastes: An EXAFS Spectroscopy Study. Applied Geochemistry 2004, 19, 379-393. 34. Lowry, G. V.; Shaw, S.; Kim, C. S.; Rytuba, J. J.; Brown Jr., G. E., Macroscopic and Microscopic Observations of Particle-Facilitated Mercury Transport from New Idria and Sulphur Bank Mercury Mine Tailings. Environmental Science & Technology 2004, 38, (19), 5101-5111. 35. Rytuba, J. J., Mercury from Mineral Deposits and Potential Environmental Impact. Environmental Geology 2003, 43, 326-338. 36. Slowey, A. J.; Johnson, S. B.; Rytuba, J. J.; Brown Jr., G. E., Role of Organic Acids in Promoting Colloidal Transport of Mercury from Mine Tailings. Environmental Science & Technology 2005, 39, (20), 7869-7874. 37. Slowey, A. J.; Rytuba, J. J.; Brown Jr., G. E., Speciation of Mercury and Mode of Transport from Placer Gold Mine Tailings. Environmental Science & Technology 2005, 39, (6), 1547-1554. 38. Blum, M.; Gustin, M. S.; Swanson, S.; Donaldson, S. G., Mercury in Water and Sediment of Steamboat Creek, Nevada: Implications for Stream Restoration. Journal of the American Water Resources Association 2001, 37, (4), 795-804. 39. Engle, M. A.; Gustin, M. S.; Zhang, H., Quantifying Natural Source Mercury Emissions from the Ivanhoe Mining District, north-central Nevada, USA. Atmospheric Environment 2001, 35, 3987-3997. 40. Fisher, P.; Gustin, M. S., Influence of Natural Sources on Mercury in Water, Sediment and Aquatic Biota in Seven Tributary Streams of the East Fork of the Upper Carson, River, California. Water, Air, and Soil Pollution 2002, 133, 283- 295. 41. Gustin, M. S., Are Mercury Emissions From Geologic Sources Significant? A Status Report. The Science of the Total Environment 2003, 304, 153-167. 42. Gustin, M. S.; Biester, H.; Kim, C. S., Investigation of the Light-Enhanced Emission of Mercury from Naturally Enriched Substrates. Atmospheric Environment 2002, 36, 3241-3254. 43. Gustin, M. S.; Taylor, G. E., Jr.; Leonard, T. L., Atmospheric Mercury Concentrations Above Mercury Contaminated Mill Tailings in the Carson River Drainage Basin, NV. Water, Air, and Soil Pollution 1995, 80, 217-220.

29

44. Nacht, D. M.; Gustin, M. S.; Engle, M. A.; Zehner, R. E.; Giglini, A. D., Atmospheric Mercury Emissions and Speciation at the Sulphur Bank Mercury Mine Superfund Site, Northern California. Environmental Science & Technology 2004, 38, (7), 1977-1983. 45. Bernaus, A.; Gaona, X.; Esbri, J. M.; Higueras, P.; Falkenberg, G.; Valiente, M., Microprobe Techniques for Speciation Analysis and Geochemical Characterization of Mine Environments: The Mercury District of Almaden in Spain. Environmental Science & Technology 2006, 40, (13), 4090-4095. 46. Higueras, P.; Oyarzun, R.; Lillo, J.; Sanchez-Hernandez, J. C.; Molina, J. A.; Esbri, J. M.; Lorenzo, S., The Almaden District (Spain): Anatomy of One of the World's Largest Hg-Contaminated Sites. Science of the Total Environment 2006, 356, 112-124. 47. Biester, H.; Gosar, M.; Covelli, S., Mercury Speciation in Sediments Affected by Dumped Mining Residues in the Drainage Area of the Idrija Mercury Mine, Slovenia. Environmental Science & Technology 2000, 34, (16), 3330-3336. 48. Biester, H.; Gosar, M.; Muller, G., Mercury Speciation in Tailings of the Idrija Mercury Mine. Journal of Geochemical Exploration 1999, 65, 195-204. 49. Foucher, D.; Ogrinc; Hintelmann, H., Tracing Mercury Contamination from the Idrija Mining Region (Slovenia) to the Gulf of Trieste Using Hg Isotope Ration Measurements. Environmental Science & Technology 2009, 43, (1), 33-39. 50. Palinkas, L.; Strmic, S.; Spangenberg, J.; Prochaska, W.; Herlec, U., Ore-forming Fluids in the Grubler Orebody, Idrija Mercury Deposit, Slovenia. Schweizerische Mineralogische und Petrographische Mitteilungen 2004, 84, 173-188. 51. Baldi, F.; Filippelli, M.; Olson, G. J., Biotransformation of Mercury by Bacteria Isolated from a River Collecting Cinnabar Mine Waters. Microbial Ecology 1989, 17, 263-274. 52. Baldi, F.; Olson, G. J., Effects of Cinnabar on Pyrite Oxidation by Thiobacillus ferrooxidans and Cinnabar Mobilization by a Mercury-Resistant Strain. Applied and Environmental Microbiology 1987, 53, (4), 772-776. 53. Rytuba, J. J., Cenozoic Metallogeny of California. In Geology and Ore Deposits of the American Cordillera: Symposium Proceedings, Coyner, A. R.; Fahey, P. L., Eds. Geological Society of Nevada: Reno, 1996; Vol. 2, pp 803-822. 54. Coleman, R. G., New Idria Serpentinite; a Land Management Dilemma. Environmental and Engineering Geoscience 1996, 2, (1), 9-22. 55. Clair, D. J. S., New Almaden and California Quicksilver in the Pacific Rim Economy. California History 1995, 73, (4), 278-295. 56. Madigan, M. T.; Martinko, J. M.; Parker, J., Brock Biology of Microorganisms. Prentice Hall: Upper Saddle River, 2003. 57. Boctor, N. Z.; Shieh, Y. N.; Kullerud, G., Mercury Ores From the New Idria Mining District, California: Geochemical and Stable Isotope Studies. Geochimica et Cosmochimica Acta 1987, 51, 1705-1715. 58. Linn, R. K., New Idria Mining District. In Ore Deposits of the United States, 1933- 1967, Ridge, J. D., Ed. The American Institute of Mining, Metallurgical, and Peroleum Engineers, Inc.: New York, 1968; Vol. 2, pp 1623-1649. 59. Bailey, E. H., Quicksilver Deposits of the Western Mayacamas District, Sonoma County, California. California Journal of Mines and Geology, Report XLII of the State Mineralogist 1946, 199-230.

30

60. Myers, W. B.; Eckel, E. B.; Forbes, J. M., "5 East" Area, Levels 3, 4, 5, 6, and 8 of the New Idria Mine, San Benito County, California. In State Mineralogist's Report XLII, Division of Mines, United States Geological Survey: 1946. 61. Wood, J. M., Biological Cycles for Toxic Elements in the Environment. Science 1974, 183, (4129), 1049-1052. 62. Jensen, S.; Jernelov, A. Behaviour of Mercury in the Environment; Vienna, 1972; pp 43-47. 63. Baldi, F.; Semplici, F.; Filippelli, M., Environmental Applications of Mercury Resistant Bacteria. Water, Air, and Soil Pollution 1991, 56, 465-475. 64. Biester, H.; Nehrke, G., Quantification of mercury in soils and sediments-acid digestion versus pyrolysis. Fresenius' Journal of Analytical Chemistry 1997, 358, 446-452. 65. Biester, H.; Scholz, C., Determination of Mercury Binding Forms in Contaminated Soils: Mercury Pyrolysis versus Sequential Extractions. Environmental Science & Technology 1997, 31, (1), 233-239. 66. Biester, H.; Zimmer, H., Solubility and Changes of Mercury Binding Forms in Contaminated Soils after Immobilization Treatment. Environmental Science & Technology 1998, 32, (18), 2755-2762. 67. Sladek, C.; Gustin, M. S.; Kim, C. S.; Biester, H., Application of Three Methods for Determining Mercury Speciation in Mine Waste. Geochemistry: Exploration, Environment, Analysis 2002, 2, 396-376. 68. Kim, C. S.; Bloom, N. S.; Rytuba, J. J.; Brown Jr., G. E., Mercury Speciation by X- ray Absorption Fine Structure Spectroscopy and Sequential Chemical Extractions: A Comparison of Speciation Methods. Environmental Science & Technology 2003, 37, (22), 5102-5108. 69. Bloom, N. S.; Preus, E.; Katon, J.; Hiltner, M., Selective Extractions to Assess the Biogeochemically Relevant Fractionation of Inorganic Mercury in Sediments and Soils. Analytica Chimica Acta 2003, 479, (2), 233-248. 70. Engle, M. A.; Gustin, M. S., Scaling of Atmospheric Mercury Emissions from Three Naturally Enriched Areas: Flowery Peak, Nevada; Peavine Peak, Nevada; and Long Valley Caldera, California. The Science of the Total Environment 2002, 290, 91-104. 71. Gustin, M. S.; Engle, M.; Ericksen, J.; Lyman, S.; Stamenkovic, J.; Xin, M., Mercury exchange between the atmosphere and low mercury containing substrates. Applied Geochemistry 2006, (1913-1923). 72. Johnson, D. W.; Benesch, J. A.; Gustin, M. S.; Schorran, D. S.; Lindberg, S. E.; Coleman, J. S., Experimental Evidence Against Diffusion Control of Hg Evasion from Soils. The Science of the Total Environment 2003, 304, 175-184. 73. Krauskopf, K. B.; Bird, D. K., Introduction To Geochemistry. 3rd ed.; WCB/McGraw-Hill: 1995. 74. Faure, G., Principles and Applications of Geochemistry. Prentice Hall: Upper Saddle River, 1991; Vol. 2, p 600. 75. Webb, S. SixPACK, 0.63; Stanford Synchrotron Radiation Laboratory: Menlo Park, 2006. 76. Webb, S. M., SIXPack: a graphical user interface for XAS analysis using IFEFFIT. Physica Scripta 2005, T115, 1011-1014.

31

77. Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D., Real-space Multiple- scattering Calculation and Interpretation of X-ray Absorption Near-edge Structure. Physica Review B 1998, 58, (12), 7565-7576. 78. Rehr, J. J.; Ankudinov, A.; Ravel, B., User's Guide, FEFF v 8.40. Department of Physics, University of Washington: 2006.

32

Chapter 2

Microbially Enhanced Dissolution of HgS in an Acid Mine Drainage System in the California

Coast Range

Adam D. Jew*1, Sebastian F. Behrens2, James J. Rytuba3, Andreas Kappler2, Alfred M. Spormann4, and Gordon E. Brown, Jr1,4,5

1) Surface & Aqueous Geochemistry Group, Department of Geological & Environmental

Sciences, Stanford University, Stanford, CA 94305-2115, USA

2) Center for Applied Geoscience, Eberhard KarlsUniversity of Tuebingen, D-72076,

Germany

3) Mineral Resources Program, U.S. Geological Survey, 345 Middlefield Road, MS 901,

Menlo Park, CA 94025, USA

4) Department of Chemical Engineering, Stanford University, Stanford, CA 94305, USA

5) Department of Photon Science and Stanford Synchrotron Radiation Lightsource, SLAC

National Accelerator Laboratory, 2545 Sand Hill Road, MS 69, Menlo Park, CA 94205,

USA

Submitted for publication in Geobiology

33

Abstract: Mercury sulfides (cinnabar and metacinnabar) are the main ores of Hg

-54 -52 and are relatively stable under oxic conditions (Ksp = 10 and 10 , respectively). However, until now their stability in the presence of microorganisms inhabiting acid mine drainage (AMD) systems was unknown.

The AMD microbial community of the inoperative Hg mine at New Idria, CA was used to test their ability to dissolve crystalline HgS. A 16S rRNA gene clone library revealed that the AMD microbial community was dominated by Fe- oxidizing and S-oxidizing bacteria. Although supersaturated with respect to cinnabar and metacinnabar, microcosms inoculated with the AMD microbial community were capable of releasing significantly more Hg into solution compared to inactivated or abiotic controls. Four different Hg-containing materials were tested for bacterially enhanced HgS dissolution: pure cinnabar, pure metacinnabar, mine tailings, and calcine material (processed ore). In the microcosm with metacinnabar, the presence of the AMD microbial community resulted in an increase of dissolved Hg concentrations up to 500 μg/L during the first 30 days of incubation. In abiotic control microcosms, dissolved Hg concentrations did not increase above 100 ng/L. When Hg concentrations were below 50 μg/L, the Fe-oxidizing bacteria in the AMD microbial community were still capable of oxidizing Fe(II) to Fe(III) in the AMD solution, whereas concentrations above 50 g/L resulted in inhibition. Although sulfur-oxidizing bacteria of the group Thiomonas sp. comprise a significant fraction (up to 50%) of the total bacterial community, as determined by quantitative PCR, batch experiments using a Thiomonas sp. isolate from the field site revealed that these

34 microorganisms do not contribute to HgS dissolution under the tested growth conditions. However, our experiments show that the AMD microbial community contributes to the dissolution of mercury sulfide minerals. These findings have major implications for risk assessment and management of inoperative Hg mines worldwide.

35

Introduction: Mercury from inoperative mine sites often poses a major environmental risk to ecosystems downstream of the mines. In California, legacy waste from historical mercury mining in the California Coast Range is found at thousands of abandoned mines (Rytuba, 2003). Although elemental Hg is the main ore material at some California mercury mines (e.g., the Socrates Mine,

West Mayacamus District, Sonoma, Co.), the majority of mines have both cinnabar and metacinnabar as primary ore minerals (Linn, 1968; Kim et al., 2000;

Rytuba, 2003; Kim et al., 2004; Lowry et al., 2004). Ore processing was done by roasting HgS-containing ore material at ~700°C (a process known as calcining), which causes HgS to breakdown to elemental Hg that was then condensed (Linn,

1968). This process was reasonably effective at recovering Hg from the ore, but a significant amount of HgS was left in the calcined material. The gangue material

(tailings) and roasted ore waste (calcine) were dumped in large piles of waste on the surface.

At Earth’s surface, cinnabar and metacinnabar are quite stable, with solubility products of 10-54 and 10-52, respectively (Faure, 1991; Krauskopf &

Bird, 1995). A number of studies have shown that HgS stability is influenced heavily by reduced sulfur species, polysulfides, and thiol-rich organic matter, but these past studies were conducted under anoxic conditions (Paquette & Helz,

1995; Paquette & Helz, 1997; Ravichandran et al., 1998; Ravichandran et al.,

1999; Jay et al., 2000; Benoit et al., 2001; Reddy & Aiken, 2001; Drexel et al.,

2002; Aiken et al., 2003; Haitzer et al., 2003; Waples et al., 2005). Because HgS

36 is considered stable in oxic systems whereas reduced sulfur species are not, the stability of HgS in highly oxic AMD systems has not been examined.

Mercury, presumably in the form of zero-valent ion pairs, is converted to methylmercury in anoxic sediments as a co-metabolic product of S-reducing and

Fe-reducing bacteria (Benoit et al., 1999a; Benoit et al., 1999b; Compeau &

Bartha, 1985; Fleming et al., 2006; King et al., 2000; Ullrich et al., 2001), both with and without added amino acids; however, methylation by microorganisms in oxic systems is unknown (Fagerstrom & Jernelov, 1971; Hayashi et al., 1977).

Both types of microorganisms are generally restricted to anoxic ecosystems or niches, where polysulfide, reduced sulfur species, and thiol-rich organic matter are generally considered to play a more dominant role in HgS dissolution than microorganisms due the strong binding constants and high abundance of the abiotic functional groups in anoxic settings.

A number of studies of Hg-resistant bacteria in the presence of different

Hg species have been carried out, but only a few have focused on HgS (Baldi &

Olson, 1987; Baldi et al., 1989; Baldi et al., 1991; Kalyaeva et al., 2001; Mindlin et al., 2001; Barkay et al., 2003). It has long been hypothesized that any S- oxidizing bacterium could slowly dissolve HgS, but there is no conclusive experimental support for this hypothesis (Wood, 1974; Madigan et al., 2003).

Baldi and Olson looked at the effect of adding cinnabar to pyrite in the presence of a Hg-resistant strain of Thiobacillus ferrooxidans; this study is one of the few which investigated HgS interaction with bacteria (Baldi & Olson, 1987).

However, this study did not show any bacterially induced dissolution of HgS

37 when cinnabar was the sole mineral substrate. Wiatrowski et al. showed that mixed-valence Fe-oxides can impact Hg oxidation state, with magnetite being able to convert Hg(II) to Hg(0) (Wiatrowski et al., 2009). Because of the potential impact of Fe-cycling on Hg speciation shown by Wiatrowski et al. and the fact that Baldi and Olson found no evidence for Hg release into solution from

HgS alone, it is possible that Hg release in systems containing FeS2 and HgS is due more to abiotic Fe redox-cycling than direct microbial activity. However, no mechanisms have been proposed for this release (Wiatrowski et al., 2009). The focus of the present study is to show for the first time the direct biological impact microorganisms have on the dissolution of relatively stable HgS minerals in oxic mine settings where HgS is considered to be stable.

Material and Methods:

Field Sampling

Samples were taken from the New Idria AMD settling pond on Aug. 13, 2008.

Water and sediment samples were collected in sterile 50 mL polypropylene tubes and immediately stored on ice. Inflow water to the settling pond (pH 3.25) was used in the microcosm experiments to best approximate the AMD settling pond environment. The water was filter-sterilized in the field using 0.1 μm Anatop

Plus® filters (Whatman Inc.) into acid-cleaned Pyrex bottles. Numerous water samples from the same location were taken for analysis of total carbon, sulfide,

Fe(II), total metals, and total Hg content. All water samples, except those collected for total carbon analysis, were filtered using 0.02 μm Anatop Plus®

38 filters into ultra-trace clean, clear borosilicate vials with Teflon® lined lids.

Cleaning of glassware for ultra-clean use involved a three step process: 1) detergent bath, 2) 1N HCl acid bath, and 3) 10% BrCl (vials were inverted to have

BrCl covering the Teflon® lined lids). Vials were left in each cleaning agent overnight. Between each step, vials and lids were triple rinsed with DDI water and dried in an oven that has never contained Hg-bearing materials at 120oC.

Samples taken for total carbon analysis were filtered using 0.2 μm polyethersulfone filters (Whatman Inc.) into ultra-trace metal clean amber borosilicate vials. Total Hg concentration samples were preserved using 0.5% bromine monochloride following EPA method 1631 (U.S. EPA, 2002). Samples taken for total metal and major anion concentrations were preserved with 2 mL of

® Ultrex brand 1N HNO3 and sent to the United States Geological Survey in

Boulder, CO for analysis.

Sulfur-oxidizing Microorganism Isolation

We attempted to isolate sulfur-oxidizing bacteria from the New Idria AMD pond water in order to investigate their potential role in HgS dissolution in defined microcosm experiments. The sampling of the pond water was done as described above. Agar plates spiked with 1 ppm Hg(NO3)2 or 1 ppm HgCl2 were used for isolation. Agar plates were prepared using the medium defined in Table 1 (Starr,

1981). To aid in bacterial isolation, bromothymol blue, a pH indicator, was added to the plates. The initial pH of the plates was about 7. Bromothymol blue changes from blue to yellow at pH 4, allowing detection of acid-producing sulfur- oxidizing microbial colonies that grew on the plates. These acid-producing

39 colonies were then transferred to fresh agar plates for further purification. After four consecutive transfers of single yellow colonies, one isolated colony was transferred into liquid media in which the Bacto-agar, bromothymol blue, HgCl2 were replaced by 0.5 g/L yeast extract. Growth of the isolate in liquid medium was also possible using peptone, but no other organic carbon sources resulted in growth of the sulfur-oxidizing isolate. The following carbon sources were used: acetate, lactate, ethanol, pyruvate, citrate, fructose, oxalate, glucose, peptone, and yeast extract. 16S rRNA gene sequencing identified the obtained isolate as belonging to the bacterial genus Thiomonas.

Clone library construction and sequencing.

The 16S rRNA gene library was constructed from the top 2 cm of the AMD settling pond sediment. DNA was extracted as described previously following a modified protocol from Zhou et al. (Zhou et al., 1996; Behrens et al., 2008).

Domain-specific primers were used to amplify almost-full length16S rRNA genes from the extracted chromosomal DNA by PCR; for Bacteria, primers GM3F

(Escherichia coli 16S rRNA position 0008) (Muyzer et al., 1993) and Uni1392R

(Lane et al., 1985) were used, and for Archaea, primers 20f (DeLong, 1992) and

Uni1392R or 20f and Arch958R (Massana et al., 1997) were used. PCRs were performed as follows: Amplifications were carried out in 50-µL volumes containing a final concentration of 0.5 µM of each primer, 200 µM of each deoxynucleoside triphosphate, 0.5 U of Taq polymerase (Qiagen GmbH,

Germany), 200 µg bovine serum albumin (Sigma-Aldrich, St. Louis, MO), and 1×

Qiagen PCR buffer containing 1.5 mM MgCl2 (pH 8.0). One-microliter amounts

40 of the undiluted and 1:10-, 1:100-, and 1:1,000-diluted environmental DNA were used as templates. The PCR amplification parameters included an initial denaturation at 94°C for 5 min, followed by 25 cycles of 94°C for 1 min, 48°C for

1 min using the bacteria domain-specific primers, and 58°C for 1 min using the

Archaea domain-specific primers, followed by an elongation step at 72°C for 1 min. The last cycle was followed by a final extension step at 72°C for 9 min. PCR amplifications were performed in a PTC-200 gradient cycler (MJ Research, Inc.,

Watertown, MA). No PCR amplicons were obtained with both Archaea primer combinations. The bacterial PCR products were purified using a QIAquick PCR purification kit (Qiagen GmbH, Germany) and ligated into the pCR4 TOPO vector (Invitrogen, Carlsbad, CA). E. coli XL10-Gold ultracompetent cells

(Stratagene, La Jolla, CA) were transformed with the plasmids following the manufacturer’s recommendations. Sequencing was performed by MCLab (South

San Francisco, CA) by using Taq cycle sequencing with a model ABI 3730XL sequencer (Applied Biosystems). Sequence assembly was done with the program

DNA Baser. The presence of chimeric sequences in the clone libraries was determined with the programs Bellerophon and Mallard version 1.02 (Huber et al., 2004; Ashelford et al., 2006). Potential chimeras were eliminated before phylogenetic analysis. Sequence data were analyzed with BLAST and the ARB software package using the SILVA 103 database (release date, 6 Jun 2010)

(Ludwig et al., 2004; Pruesse et al., 2007).

41

Nucleotide sequence accession numbers.

The 16S rRNA gene sequences from the clone library have been submitted to

EMBL and assigned the following accession numbers: HE587052 to HE587299.

The 16S rRNA gene sequence of the isolated sulfur-oxidizing bacterium can be found under the following accession number: HE587300.

Real-time quantitative PCR.

The quantification of 16S rRNA genes in the sediment from the AMD settling pond was performed using iQ Sybr Green Supermix (Bio-Rad Laboratories,

Hercules, CA), Thiomonas-specific primers and domain-specific primers for

Bacteria (see Table 2). Each sample mixture had a 30-µl reaction volume containing 1× iQ Sybr Green Supermix, forward and reverse primers at a concentration of 500 nM, and 2 µl of the prepared DNA. PCR amplification and detection were conducted in an iQ5 Cycler (Bio-Rad Laboratories, Hercules, CA).

Real-time PCR conditions were as follows: 3 min at 95°C followed by 40 cycles of 10 s at 95°C and 45 s at 61.5°C. As real-time PCR standard dilutions of a plasmid (pCR4 TOPO vector; Invitrogen, Carlsbad, CA) containing a Thiomonas sp. 16S rRNA gene from the clone library was used. Real-time PCR data was analyzed using the iQ5 Optical System Software (Version 2.1; Bio-Rad

Laboratories, Hercules, CA). Because the closest sequenced relative of the isolated Thiomonas strain from the AMD microbial community is Thiomonas intermedia K12, whose genome contains only one 16S rRNA gene operon, the quantified 16S rRNA gene copy numbers obtained with the Thiomonas primers

42 were not divided by a certain factor to account for multiple rRNA gene operons.

Total Bacteria 16S rRNA gene copy numbers were divided by a factor of 4.07 as the average number of rRNA gene operons of the domain Bacteria documented by the ribosomal RNA database (Lee et al., 2009).

HgS Dissolution Experiment

Sample preparation

Mercury sulfide dissolution experiments were carried out in 250 mL Erlenmeyer flasks that were acid washed in 1N trace-metal grade HCl and then heated at

o 500 C to remove any residual Hg on the glass and to sterilize the flasks. Growth media for all batch reactors consisted of water from the New Idria AMD system

(pH 3.25) that was filter-sterilized in the field with 0.1μm Anatop Plus® filters

(Whatman Inc.) and then re-sterilized in the laboratory with 0.1μm Anatop Plus® filters under proper sterile lab conditions. Four different types of HgS-containing materials were selected for study: cinnabar, metacinnabar, tailings, and calcine.

Tailings and calcine material were collected from the New Idria site adjacent to the AMD system, and cinnabar and metacinnabar samples were purchased from

Alfa Aesar with a purity of > 99.99%. Tailings and calcine material were dried in a dessicator and then sieved in stainless-steel sieves to a particle size of 89-124

μm. Due to the softness of the HgS crystals, particle size could not be controlled because any manipulation of the HgS powder, such as crushing in a mortar and pestle resulted in an uncontrollable change in particle size. The solids (cinnabar, metacinnabar, tailings, and calcine) as well as half of the biological samples from

43 the AMD system were sterilized with a 137Cs gamma irradiation source at the

Stanford School of Medicine at an exposure rate of 2602 R/min for 18 hours.

Sterilization by gamma irradiation was chosen so that cell lysing could be kept to a minimum. Because of the high metal content of the AMD system, autoclaving samples was not done because of the possibility of significant thermodynamic changes to the system resulting in conditions in the batch reactors being different from the conditions of the AMD environment. Batch reactors consisted of 150 mL of AMD water (oxic microcosms) or 100 mL of AMD water (anoxic microcosms and abiotic controls), 2 g of mineral substrate, and non-abiotic controls were inoculated with 100 μL of either living or killed biofilm material

(total protein added to reactors was 7.88 μg and 8.12 μg, respectively). All batch reactors were incubated at ambient temperature (21oC) without agitation to mimic the AMD settling pond conditions. Anoxic batch reactors were constructed outside of an anoxic glove box and then immediately placed in the glove box.

Anoxic batch reactors were removed from the glove box to be sampled under sterile conditions and then were immediately returned to the glove box. This procedure was used for two important reasons: 1) rubber stoppers commonly used in anoxic work will adsorb large amounts of Hg from reaction vessels, and 2) sterile conditions using a flame are not possible in an anoxic glove box. Because oxygen diffusion into solution is fairly slow, the increase of oxygen in abiotic reaction vessels should be minimal. However, the AMD water was not purged of oxygen before the experiment, resulting in some residual oxygen being left in the system for the first 3 days of the experiment run.

44

Microcosm Sampling and Sample Preservation

Sampling of the batch reactors occurred every 3 days with samples taken for non- purgeable organic carbon (NPOC, 5 mL), total carbon (TC, 4 mL), total Hg (4 mL), sulfide (3 mL), and iron (Fe2+ and total Fe, 1 mL). Samples of the oxic microcosms were taken over a total of 30 days, while anoxic microcosms were followed for a total of 18 days. NPOC and TC samples were filtered through 0.2

o μm polyethersulfone (PES) filters into amber vials and stored at 5 C until analysis. Mercury, sulfide, and iron samples were filtered with 0.02 μm Anatop

Plus® filters (Whatman Inc.) to minimize colloidal HgS uptake. Numerous experiments done in the lab using the Anatop Plus® filters (Whatman Inc.) showed no retention or carryover of Hg, sulfide, and Fe. Mercury samples were filtered into clear borosilicate vials with Teflon® lined lids and preserved with

1.5% BrCl. Sulfide samples were filtered into amber borosilicate vials containing

100 μL of 10N NaOH to raise the pH of the solution from 4.1 to 10.0, resulting in

- the trapping of sulfide as HS as well as causing metals to precipitate from solution. Iron samples were filtered into amber borosilicate vials and analyzed for

2+ Fe and FeTot.

Sample Analysis

Background concentrations for total metals were analyzed by inductively coupled plasma atomic emission spectrometry (ICP-AES) using a Perkin Elmer Optima

3300 dual view analyzer and inductively coupled plasma mass spectrometry (ICP-

MS) using a Perkin Elmer Sciex Elan 6000 analyzer. Background nitrate,

45 chloride, fluoride, and sulfate were measured by ion chromatography (IC) using a

Dionex DX-100 Ion Chromatograph. The concentration of protein within the biofilm material added to batch reactors was determined using a BioRad® fluorimetric protein assay kit. Iron (II) and FeTot concentrations were determined by ferrozine analysis using a Hewlett-Packard model 8452A UV/Vis spectrophotometer with a lower detection limit of 0.5 mg/L (Stookey, 1970).

Total carbon and non-purgeable organic carbon concentrations were determined using a Shimadzu model TOC-5000A TC/NPOC analyzer with a lower detection limit of 0.1 mg/L. During analysis, we found that the high metal concentrations contained in the AMD water resulted in passivation of the catalyst in the

TC/NPOC analyzer after 4-5 samples. To address this issue, 10 μL of 10N NaOH were added to the samples to precipitate the metals in solution. The samples were then centrifuged at 10,000 rpm for 5 minutes to settle out the metal precipitate.

The supernatant was then taken and 10 μL of 10N HNO3 was added to drop the sample pH to < 4 for NPOC analysis. Several samples were run with and without metal stripping to determine if a significant amount of carbon was being removed from solution as the metal precipitated. The results of the simultaneous sample runs showed less than 1% variation in carbon between samples where the metals were removed vs. those where metal was left in solution. Total carbon within the tailings and calcine material was measured using a Carlo-Erba model NA 1500 C,

N, and S analyzer. Sulfide samples were analyzed within 1 hour of sampling following the Cline’s method for sulfide analysis with a lower detection limit of

0.03 M (Cline, 1969). Mercury concentrations were measured on a Tekran®

46

2600 cold vapor atomic fluorescence spectrometer (CVAFS) following EPA method 1631 with a lower detection limit of 0.05 ng/L (U.S. EPA, 2002).

Hg Impact on Fe-oxidation

Water and biofilm sampling was done on February 10, 2010, with sampling methodology the same as described above for the HgS dissolution experiment, solution pH = 4.0. Because this experiment was designed to determine the inhibition effects of Hg on Fe-oxidizing microorganisms, killed controls were not done. Batch reactors contained filter sterilized AMD water with Hg added to determine Hg toxicity to the Fe-oxidizing portion of the community. No Hg containing solids were added to the reaction vessels. By not adding Hg- containing solids, only Fe(II) oxidation in solution is assessed and the Hg concentrations of the vessels are controlled only by the amount of Hg added and not impacted by the solids. The background Hg concentration for the experiment was 15.7 ± 0.4 ng/L, which is much lower than the background levels of 100-400 ng/L measured during the summer months. Mercury was added to the batch reactors in the form of Hg(NO3)2 at the following concentrations: abiotic control, no Hg added (ran in duplicate), 500 ng/L, 1,000 ng/L, 5,000 ng/L, 10,000 ng/L,

50,000 ng/L, and 100,000 ng/L. The experiment was sampled every 2 days for a total of 18 days. Samples were filtered through 0.02 μm Anatop plus filters and analyzed following the ferrozine protocol on a Hewlett-Packard 8452A UV/Vis spectrophotometer. Samples were analyzed within 1 hour of sampling and triplicates of samples showed error to be < 0.1%.

47

Isolate impact on HgS containing substrates

The S-oxidizing microorganism isolated from the New Idria AMD system was used in an experiment similar to the AMD microbial community experiment described above. The medium was changed from AMD water to the specific growth medium of the isolate (Table 1). The liquid medium used with the HgS- containing materials was identical to the liquid growth medium except that the thiosulfate used in the growth medium was removed and the pH was titrated from

7 to 4 using H2SO4. By using a well-defined medium in which the isolate has been growing, the chances of the isolate dissolving the HgS minerals should increase. Glassware cleaning, sample procedure, and analysis were identical to those described above in the microcosm study. Samples were taken every two days for this experiment to better refine the Hg release at the beginning of the experiment.

Results:

AMD Solution Chemistry

Total metal concentrations in the New Idria AMD system are typical of other

AMD systems (Table 3). Due to Fe(II) being one of the dominant metals in solution it is assumed that Fe(II) oxidation is the dominant metabolic pathway for the system. ICP analysis carried out over several years (data not shown) show that the concentrations of metals in solution are dependent on the amount of rain received during the rainy season, which can cause concentrations to vary by 3- fold. Dry winters result in an increase in metal concentrations during the

48 following summers due to evaporation, whereas wet winters tend to dilute the metals when the water discharge from underground mine workings increases.

The areal extent and thickness of the AMD biofilm vary inversely with metal concentration. Mercury concentrations within the AMD waters are quite low,

100.8 ± 5.2 ng/L, and vary inversely with the concentrations of other metals in solution between seasons. The dominant anion in solution is sulfate, which is produced by the oxidative dissolution of sulfide-bearing minerals found at the site.

Clone library, qPCR, and bacterial isolate

The 16S rRNA gene clone library of the top 2 cm of the AMD settling pond sediment revealed that the community was dominated by Fe-oxidizing bacteria of the order Ferritrophicales. Clones affiliated with this order comprised more than

60% of the 248 bacterial clones that were obtained (Fig. 1). The second largest group detected in the 16S rRNA gene clone library comprised relatives of the sulfur-oxidizing bacterial genus Thiomonas (about 14% of all clones in the library). A member of this bacterial taxa has also been successfully isolated and used in defined HgS dissolution experiments as described below. Contrary to the

16S rRNA clone library, quantitative PCR analysis with Thiomonas-specific primers revealed a relative abundance of this group of sulfur-oxidizing bacteria of about 55% of all bacteria in the top 1 cm of the AMD pond sediment (Fig. 2).

The total bacterial 16S rRNA gene numbers were 1 x 108 ± 2 x 107 copies per g

(wet sediment) in this layer. Other bacterial groups also found with relatively high sequence abundance in the 16S rRNA gene clone library were affiliated with the

49 bacterial orders Gallionellales (7% of all clones in the library), Xanthomonadales

(6%), and Rhodospirillales (4%). Altogether, the total AMD microbial community was dominated by bacterial taxa known to contain mostly iron-oxidizing, sulfur- oxidizing, and acidophilic chemoorganoheterotropic bacteria.

Bacterial Dissolution of HgS

Hg Concentrations

The background Hg concentration of the New Idria AMD water used in the experiments was 100.8 ± 5.2 ng/L. Several preliminary experiments with a similar experimental setup in which fewer parameters were measured, though

Hgtot and total sulfide were always measured, all had Hgtot concentrations that were four times higher than the current study, but showed similar trends. Oxic microcosms of living biofilm cells resulted in a significant release of Hg from all four HgS-containing materials (cinnabar, metacinnabar, tailings, and calcine), with Hg concentrations rising from 100.8 ± 5.2 ng/L to as high as 516.5 ± 23.8

μg/L over the course of 30 days (Fig. 3). More Hg was released from the calcine material than metacinnabar during the course of the experiments. Extended x-ray absorption fine structure (EXAFS) spectroscopic analysis of the New Idria calcine material showed the presence of eglestonite and montroydite, in addition to HgS polymorphs, within the calcine material (Kim et al., 2000; Kim et al., 2004;

Lowry et al., 2004). Because Hg-chlorides and oxides are more soluble than Hg- sulfides, the higher Hg concentration detected in the calcine batch reactor experiment versus the experiment with metacinnabar is most likely due to more

50 soluble Hg phases dissolving into solution as well as HgS dissolution. EXAFS analysis of the tailings material showed a mixture of cinnabar and metacinnabar

(29% and 61% of the Hg speciation, respectively) with 10% HgO (montroydite).

The high proportion of metacinnabar and the presence of montroydite are consistent with the tailings material releasing more Hg than the cinnabar alone and is less than that released by pure metacinnabar. At day 18, Hg concentrations began to level off in the cinnabar and tailings oxic living microcosms, but they continued to increase within the calcine and metacinnabar reactors. Gamma- sterilized AMD water samples incubated aerobically also released Hg from the mineral matrix, but total Hg concentrations were significantly lower than in the biologically active microcosms (Fig. 3). Hg concentrations in abiotic controls started at background levels and dropped to below detection limit within the first

9 days of experimentation indicating that Hg was adsorbing to the mineral matrix

(Data not shown). Hg release in the inactivated controls (killed) mirrored the living aerobic batch reactors, but the total concentrations were much lower (Fig.

4). Anoxic microcosms of living cells had an initial release of Hg at the start of the experiment, which was assumed to be caused by residual oxygen left in the batch reactors. In anoxic batch reactors Hg concentrations are several orders of magnitude below those in the oxic microcosms. After the initial release of Hg,

Hg concentrations either leveled off or dropped to levels below detection (Fig. 4).

All anoxic batch reactors had Hg concentrations below 1.1 ppb, which is significantly lower than that of most of the batch reactors incubated under oxic

51 conditions. These results show that oxygen is necessary as the electron acceptor in the dissolution of HgS in the presence of the biofilm material.

Sulfide Concentrations

Sulfide was detectable at the time of sampling but was not detectable in water collected from the site after 2 days. The background sulfide concentration at the settling pond was 0.8 μM, which was significantly lower than the preliminary experiments discussed above (up to 45 M). Sulfide concentrations in previous experiments were measurable for up to a week in reaction vessels due to high levels of Zn in solution, which stabilized the sulfide as ZnS(aq), as determined by

The Geochemist’s Workbench® and Visual Minteq® (Bethke, 2002; Gustafsson,

2009). When approaching the main adit of the New Idria mine, H2S gas is commonly smelled and emanates from the water exiting the underground mine.

The H2S(g) is usually noticeable up to 100 m from the entrance, but at the time of sampling, the odor was barely noticeable at the adit entrance, suggesting a drop in

H2S(g) production from the underground mine workings.

Total and Non-purgeable Organic Carbon

The New Idria AMD system has a pH ranging between 2 and 5.5 annually.

Because of this low pH, the total carbon and non-purgeable organic carbon

(NPOC) concentrations are the same. The initial concentration of the NPOC was

2.5 mg/L. The concentration of all batch reactors stayed between 1.2 mg/L and

3.5 mg/L, with no clear trends (data not shown). Ion chromatography (IC) analysis conducted on the AMD water showed no detectable organic species

52 present with an elution time > 30 minutes. This finding suggests that carbon within the AMD system is in the form of high molecular weight biomolecules, stemming from lyzed microbial cells that have difficulty moving through the IC column. AMD water was also analyzed with UV/VIS spectroscopy to look for humic and fulvic acids in the water. The lack of enhanced absorbance in the

UV/VIS spectrum specific for humic and fulvic acids, as outlined by Weishaar et al., suggests that humic and fulvic material are not large contributors to the carbon in the AMD system (Weishaar et al., 2003). The use of the BioRad® protein quantification assay revealed that protein accounted for approximately

60% of the total carbon in the AMD water used as the microcosm medium, suggesting that cellular biomass is the main source of carbon in the ecosystem.

Iron Oxidation

The background Fe concentration (369.4 ± 0.4 mg/L) of the water entering the

New Idria AMD settling pond is typical of AMD systems. The total Fe concentration was measured with ICP-OES and Ferrozine, both resulting in values within 1% of each other. The oxidation state of the Fe entering the pond is

100% Fe(II) as determined by Ferrozine. In the oxic living batch reactors the

Fe(II) was all oxidized to Fe(III) within 18 days for reactors containing cinnabar, tailings, and calcine material (Fig. 5). All other batch reactors in the experiment behaved like the metacinnabar-only batch reactor, with an initial drop in Fe(II), which also corresponds to Fetot concentrations, and then a leveling off of Fe(II) for the remainder of the experiment (Fig. 5). The initial drop in Fe(II) is considered to be caused by adsorption of Fe to either the solids or the glass wall

53 of the flask since Fe(II) is still 100% of the total Fe detected in the reaction vessels. In the tailings and calcine batch reactors containing living cells, the mineral substrate turned yellow during the experiment, indicating that the pyrite and/or marcasite present was oxidized by the biofilm material and precipitated as an Fe(III)-bearing phase. Because of the significant difference in Fe redox behavior between the reactors containing metacinnabar vs. the other three mineral substrates, an additional set of experiments were carried out that examined the effects of Hg solution concentration on Fe-oxidation.

Effects of Hg on Iron Oxidation

The background FeTot concentration for this experiment was 181.3 ± 0.1 mg/L.

Similar to the previous experiments, 100% of the Fe was in the form of Fe(II).

Fe(II) oxidized to Fe(III) in 14 days of incubation in all reaction vessels but the abiotic control and the 100,000 ng/L reaction vessel (Fig. 6). Although an exact

Hg concentration required to completely inhibit Fe-oxidation in the reaction vessels was not established, the Hg concentration required is between 50,000 ng/L and 100,000 ng/L for New Idria AMD biofilm. As expected, as Hg concentration increases, the rate of Fe(II) oxidation is retarded (Fig. 6). The inhibition concentration for the Fe-oxidizing bacteria in the New Idria AMD biofilm is much lower than the tolerance of the S-oxidizing bacterium (20 mg/L Hg) isolated from the site.

54

Microbial isolate impact on HgS dissolution

Isolation of a Thiomonas sp. strain was performed on agar plates containing about

1 ppm HgCl2. The bacterium has been grown in solution with pH of 0 to 7.

Thiomonas sp. is a heterotrophic bacterium that has only been grown in the lab using agar, peptone, or yeast extract; the isolate was not able to grow on simpler organics, which is consistent with results reported in the literature for related bacteria (Kelly & A.P., 2005). The mercury tolerance of the Thiomonas sp. isolate in the laboratory is at least 20 mg/L. The experiment using the Thiomonas sp. isolate showed no increase in Hg above control samples for any of the living oxic microcosms. This finding shows that the growth conditions in the experiment were not conducive to the isolate dissolving HgS on its own.

Discussion:

Clone library, qPCR, Isolate, and Carbon cycling

The microbial biofilm in the New Idria AMD system is dominated by two different metabolic types of bacteria: Fe-oxidizing and S-oxidizing. We observed a large difference between the relative abundance of Fe-oxidizing bacteria and S- oxidizing bacteria based on 16S rRNA clone library and qPCR. Although the 16S rRNA clone library shows that the Fe-oxidizing bacteria of the order

Ferritrophicales make up > 60% of all clones in the clone library, S-oxidizing bacteria of the genus Thiomonas (order Burkholderiales) account for only 14% of all clones in the top 1 cm of the AMD sediment. Since relative sequence abundance in clone libraries often does not reflect natural abundances we also

55 performed quantitative PCR analysis with Thiomonas-specific primers. qPCR revealed a relative abundance of this group of sulfur-oxidizing bacteria of about

55% of all bacteria in the top 1 cm of the AMD pond sediment. The relative underrepresentation of the Thiomonas sp. sequences in the clone library compared to the qPCR results can be explained by the use of primer sets with different specificity for both analyzes. The bacterial isolate, Thiomonas sp., is a mixotroph that in our study requires a complex organic substrate, namely peptone and yeast extract, for growth in liquid medium (Kelly & A.P., 2005). Attempts to grow the

Thiomonas sp. isolate on simple organic substrates and autotrophically with only

CO2/O2 were unsuccessful. Based on the necessity of complex organics for growth, we hypothesize that the organics used by the heterotrophic portion of the biofilm comes from cellular biomass of the primary producing bacteria in the biofilm. Due to the presence of a S-oxidizing bacterium in the system with a high

Hg tolerance, it is possible that the Thiomonas sp. bacterium is the main driving force for HgS dissolution.

HgS Dissolution

The dissolution of HgS within the New Idria AMD system occurs solely in the oxic regime. All anoxic batch reactors showed a minor spike in Hg due to residual oxygen in the AMD water that was not purged prior to the start of the experiment (Fig. 4). Once the oxygen in the anoxic batch reactors was used up, the Hg concentrations either leveled off or dropped below detection. An increase in Hg released from abiotic and killed reaction vessels incubated in an oxic

56 environment is most likely due to the organic matter contained in solution and mine waste materials (discussed later).

One possibility for the increased Hg in batch reactors is the presence of polysulfides, either present in solution or produced by the microorganisms. Due to the low pH of the system and the lack of an increase in Hg in anoxic microcosms, polysulfides are not considered to be the driving factor for HgS dissolution in our experiments. Based on literature findings, polysulfides (and other reduced sulfide species) should not be stable in the New Idria AMD system because polysulfide formation and stability require neutral to basic pH as well as anoxic conditions (Paquette & Helz, 1995; Paquette & Helz, 1997; Jay et al.,

2000). Although it is well known that biofilms can create microenvironments different than the surrounding environment, the fact that anoxic microcosms did not show any increase in Hg indicates that the AMD microbial community is not capable of producing an anoxic environment required for stable polysulfides or reduced sulfur species.

An explanation for why HgS(s) is degrading in the batch reactors based solely on thermodynamics is not adequate. Both the cinnabar and metacinnabar polymorphs of HgS are quite insoluble, with solubility products of 10-54 and 10-52, respectively (Faure, 1991; Krauskopf & Bird, 1995). Based on thermodynamic calculations of conditions within the batch reactors, sulfide concentrations would need to drop to levels below 10-33 M in order to have HgS undersaturated at the

New Idria site. A sulfide concentration of 10-33 M is unrealistic for any natural system and would result in supersaturation of the AMD solution with respect to

57

HgS, making the abiotic release of Hg thermodynamically unfavorable. A study by Bura-Nakić et al. shows that even in oxic waters that are at or near O2 saturation, concentrations of total reduced sulfur species are ~10-8 M (Bura-Nakic et al., 2009). Several research groups have published values for HgS stability in the presence of H2S and its deprotonated forms (Paquette & Helz, 1995; Paquette

& Helz, 1997). Due to the low pH of the system, the dominant sulfide species will be H2S. Calculated Hg concentrations, using background sulfide concentrations at the New Idria site, necessary for HgS(s) equilibrium would be

~10-12 M, which is significantly lower than the 10-8.5 M detected at the start of the experiment. Nearly all of these chemical reactions require H2S as a reactant, which would result in HgS becoming more stable as H2S concentrations decrease.

Because the experiment begins with conditions in which cinnabar and metacinnabar are supersaturated, a purely abiotic thermodynamic explanation for the increase in Hg concentration in the microcosm is incorrect.

Another possible explanation for enhanced HgS dissolution at New Idria is that the S-oxidizing microbe is oxidizing H2S in solution to sulfate, resulting in a drop in sulfide concentrations and leading to dissolution of HgS based on solubility. This hypothesis suggests that every S-oxidizing bacterium in the environment would be capable of dissolving HgS. It has been assumed that S- oxidizing bacteria can slowly dissolve HgS, but there is little to no experimental evidence supporting this assumption (Wood, 1974; Madigan et al., 2003).

Research by Baldi and Olson using Hg-sensitive and Hg-tolerant Thiobacillus ferrooxidans strains with a mixture of pyrite and cinnabar showed that this Hg-

58 tolerant bacterium was not capable of using cinnabar as the sole energy source

(Baldi & Olson, 1987). Although experiments with the Hg-tolerant Thiobacillus ferrooxidans strain in a mixture of pyrite and cinnabar showed that Hg is released into solution as both Hg(II) and Hg0, no Hg was released in experiments with cinnabar only. The release of Hg into solution was potentially attributed to back reactions of pyrite oxidation with HgS, since pyrite oxidation and Hg release were found to be linked; however, no mechanism for this reaction was proposed (Baldi

& Olson, 1987). Although not directly related to the work of Baldi and Olson,

Wiatrowski et al. showed that magnetite can reduce Hg(II) to Hg0 in anaerobic systems (Baldi & Olson, 1987; Wiatrowski et al., 2009).

The most likely hypothesis for HgS degradation in the New Idria AMD system, and potentially other inoperative Hg mine sites in California, is direct oxidation of the sulfide in the HgS crystals. Binding constants between Hg and S vary greatly depending on the oxidation state of S. As sulfur becomes more oxidized, the binding constants reduce greatly as seen in the following sequence:

- 37.71 2- 29.93 2- 22.85 2- 1.34 HS (10 ), S2O3 (10 ), SO3 (10 ), and SO4 (10 ) (Smith & Martell,

2004). During oxidation of sulfide in HgS to sulfate, Hg should be released into solution from the mineral surface. The presence of sulfate at levels > 2,600 mg/L in the New Idria AMD system (Table 3) clearly demonstrates the oxidation of sulfur. Because the sulfide-oxidizing isolate Thiomonas sp. had a high Hg tolerance (>20 mg/L Hg), we evaluated the potential for our isolated Thiomonas strain being responsible for the observed biological dissolution of HgS in the

AMD system. The HgS dissolution experiment using the bacterial isolate instead

59 of the AMD microbial community showed that given the growth conditions provided to the microorganism, Thiomonas sp. alone was not capable of dissolving HgS. There are several possible explanations as to why this experiment did not work as hypothesized. Due to the high complexity of the

AMD water at the New Idria site, there is a distinct possibility that a specific chemical in the AMD solution critical to the isolate’s ability to dissolve HgS was missing. Another possibility is that more than one microorganism is necessary for

HgS dissolution. The other microorganism(s) could change the redox conditions of the system in a way not reproduced in the laboratory or could provide an important growth factor or trace nutrient that was not present in our growth medium. Another possibility is that HgS dissolution is coupled with Fe-cycling.

The growth medium of the isolate contained little to no reduced Fe. The experiment using the AMD microbial community and the settling pond water showed that Fe-cycling was important in those microcosms. The potential of Fe- cycling helping the microorganisms with the dissolution of HgS is intriguing and requires additional work. If Fe-cycling coupled with Hg-cycling at the site is key to HgS dissolution in the New Idria AMD system, then additional bacteria would be needed to control the Fe-oxidation portion of the system.

Regardless of the actual mechanism for HgS dissolution, it is evident that the New Idria biofilm has a profound influence on the solubility of HgS, both cinnabar and metacinnabar. Although the bioreactors are far from equilibrium, it is useful to compare the activity quotient for HgS dissolution in the AMD system with that of the idealized solubility constant for both HgS polymorphs. Activities

60 for both sulfide and Hg2+ were calculated using both Geochemist’s Workbench and Visual Minteq by using the sulfide concentration (metacinnabar alive, Day 3 sample, 0.1 M), Hgtot (metacinnabar alive, Day 3 sample, 18.7 g/L), and the solution chemistry for the system (Table 3) (Bethke, 2002; Gustafsson, 2009).

By using these calculated activities, a LogQ for HgS dissolution of -23.5 was calculated. Comparing activity quotient to the idealized HgS solubility for cinnabar and metacinnabar (LogK = -54 and -52, respectively), the New Idria biofilm is capable of increasing the solubility of HgS by 28.5 to 30.5 orders of magnitude for the bioreactor experiments.

Release of Hg from abiotic and killed oxic controls is most likely due to organic matter from solution and the mine waste material used in the experiment.

It is possible that outer membrane proteins not denatured during the gamma irradiation were able to dissolve the HgS, but because abiotic and killed controls mirrored each other closely, the most likely cause of HgS dissolution in these batch reactors is due to trace amounts of thiol groups associated with the organic matter (Fig. 4). Studies by numerous research groups have shown that organic matter rich in thiol groups such as humic and fulvic acids can actively dissolve

HgS (Ravichandran et al., 1998; Ravichandran et al., 1999; Waples et al., 2005).

Since cell lysate, which is assumed to be the source of the organics in this system, have thiol containing functional groups, residual organic material in the system likely plays a minor role in HgS dissolution.

61

Effects of Hg on Fe-oxidation

Iron in the batch reactors containing living biofilm material (oxic microcosms) was oxidized from Fe(II) to Fe(III) quickly for the cinnabar, tailings, and calcine reactors (Fig. 5). In the metacinnabar batch reactor, Fe(II) concentrations were similar to controls, suggesting inhibition of the Fe-oxidizing bacteria in the batch reactor. The metacinnabar batch reactor with living cells had an initial spike of

Hg released into solution more than 24-fold greater than that of any other batch reactor for day 8 of the HgS dissolution experiments (Fig. 3). The experiment investigating the effect of Hg concentrations on Fe(II) oxidation showed that concentrations close to that of the metacinnabar reactor caused a severe retardation in Fe-oxidation. As the HgS experiment progressed, the metacinnabar

Hgtot concentrations increased to levels above those required to inhibit Fe- oxidizing bacteria. The initial pulse of Hg into solution appears to have inactivated the Fe-oxidizing bacteria before they had a chance to grow and develop a higher Hg tolerance within the batch reactor. The cinnabar, tailings, and calcine batch reactors showed no such pulse in Hg, which probably gave the

Fe-oxidizing bacteria enough time to adapt and better withstand elevated Hg concentrations.

Conclusions: Although the primary microbial metabolism, based on the high

2- concentrations of Fe (both II and III) and SO4 at the inoperative New Idria Hg mine, is the breakdown of FeS2 minerals (primarily marcasite), the breakdown of

HgS is of greater importance to the ecosystems surrounding this and potentially other HgS-containing AMD systems in the California Coast Range. It has long

62 been assumed that all S-oxidizing bacteria have the capability of dissolving HgS

(Wood, 1974); however, this assumption has not been verified for systems with

HgS as the sole mineral substrate. The present study has shown that the microbial community in the New Idria AMD system is capable of dissolving HgS and releasing Hg into solution as a soluble form. This biofilm has demonstrated its ability to greatly increase the solubility of HgS by 28.5 to 30.5 orders of magnitude compared to the idealized solubility constants for both cinnabar and metacinnabar. Although the S-oxidizing Thiomonas sp. is the most likely candidate for HgS dissolution, it is unknown whether additional bacterial species are required for HgS dissolution. Unlike experimental studies on the interaction of organic and reduced sulfur species with HgS in anoxic systems, the breakdown of

HgS in aerobic systems is novel with wide-ranging implications for remediation of active and inactive Hg mine sites. Even though the dimensions of both the

New Idria AMD system and microbial community are small and its Hg-release is thought to have a minor impact on the surrounding ecosystem, the results of this study serves as an indicator of potential Hg release at other inoperative Hg mine sites in the California Coast Range that have developed an AMD system.

63

Acknowledgements

We would like to thank Ian Marshall of the Spormann research group and Phi

Luong of the Brown research group for help with field sampling at the New Idria mine. We also wish to thank Petra Kühner for technical assistance with the qPCR experiments. Funding for this research came from the Stanford Environmental

Molecular Science Institute through NSF Grant CHE-0431425.

64

Literature Cited

Aiken G., Haitzer M., Ryan J. N., and Nagy K. (2003). Interactions between dissolved organic matter and mercury in the Florida Everglades. Journal of Physics IV France. 107: 29-32. Ashelford K. E., Chuzhanova N. A., Fry J. C., Jones A. J., and Weightman A. J. (2006). New Screening Software Shows that Most Recent Large 16S rRNA Gene Clone Libraries Contain Chimeras. Applied and Environmental Microbiology 72 (9): 5734-5741. Baldi F., Filippelli M., and Olson G. J. (1989). Biotransformation of Mercury by Bacteria Isolated from a River Collecting Cinnabar Mine Waters. Microbial Ecology 17: 263-274. Baldi F. and Olson G. J. (1987). Effects of Cinnabar on Pyrite Oxidation by Thiobacillus ferrooxidans and Cinnabar Mobilization by a Mercury-Resistant Strain. Applied and Environmental Microbiology 53 (4): 772-776. Baldi F., Semplici F., and Filippelli M. (1991). Environmental Applications of Mercury Resistant Bacteria. Water, Air, and Soil Pollution 56: 465-475. Barkay T., Miller S. M., and Summers A. O. (2003). Bacterial Mercury Resistance from Atoms to Ecosystems. FEMS Microbiology Reviews 27: 355-384. Behrens S., Azizian M. F., McMurdie P. J., Sabalowsky A., Dolan M. E., Semprini L., and Spormann A. M. (2008). Monitoring Abundance and Expression of "Dehalococcoides" Species Chloroethene-Reductive Dehalogenases in a Tetrachloroethene-Dechlorinating Flow Column. Applied and Environmental Microbiology 74 (18): 5695-5703. Benoit J. M., Gilmour C. C., Mason R. P., and Heyes, A. (1999a). Sulfide Controls on Mercury Speciation and Bioavailablity to Methylating Bacteria in Sediment Pore Waters. Environmental Science & Technology 33 (6): 951-957. Benoit J. M., Mason R. P., and Gilmour C. C. (1999b) Estimation of Mercury-Sulfide Speciation in Sediment Pore Waters Using Octanol-Water Partitioning and Implications for Availability to Methylating Bacteria. Environmental Toxicology and Chemistry 18 (10): 2138-2141. Benoit J. M., Mason R. P., Gilmour C. C., and Aiken G. R. (2001). Constants for Mercury Binding by Dissolved Organic Matter Isolates from the Florida Everglades. Geochimica et Cosmochimica Acta 65 (24): 4445-4451. Bethke C. M. (2002). The Geochemist's Workbench. Bura-Nakic E., Helz G. R., Ciglenecki I., and Cosovic B. (2009). Reduced sulfur species in a stratified seawater lake (Rogoznica Lake, Croatia); seasonal variations and argument for organic carriers of reactive sulfur. Geochimica et Cosmochimica Acta 73: 3738-3751. Cline J. D. (1969). Spectrophotometric Determination of Hydrogen Sulfide in Natural Waters. Limnology and Oceanography 14 (3): 454-458. Compeau G. C. and Bartha R. (1985) Sulfate-Reducing Bacteria: Principal Methylators of Mercury in Anoxic Estuarine Sediment. Applied and Environmental Microbiology 50 (2): 498-502. DeLong E. F. (1992). Archaea in coastal marine environments. Proceedings of the National Academy of Sciences of the United States of America 89 (12): 5685- 5689.

65

Drexel R. T., Haitzer M., Ryan J. N., Aiken G. R., and Nagy K. L. (2002). Mercury(II) sorption to Two Florida Everglades Peats: Evidence for Strong and Weak Binding and Competition by Dissolved Organic Matter Released from the Peat. Environmental Science & Technology 36 (19): 4058-4064. Fagerstrom T. and Jernelov A. (1971). Formation of Methyl Mercury from pure Mercury Sulphide in Aerobic Organic Sediment. Water Research 5: 121-122. Faure G. (1991). Principles and Applications of Geochemistry. Upper Saddle River, Prentice Hall. Fleming E. J., Mack E. E., Green P. G., and Nelson D. C. (2006). Mercury Methylation from Unexpected Sources: Molybdate-Inhibited Freshwater Sediments and an Iron-Reducing Bacterium. Applied and Environmental Microbiology 72, (1): 457- 464. Gustafsson J. P. (2009). Visual MINTEQ. Haitzer M., Aiken G. R., and Ryan J. N. (2003). Binding of Mercury(II) to Aquatic Humic Substances: Influence of pH and Source of Humic Substances. Environmental Science & Technology 37 (11): 2436-2441. Hayashi K., Kawai S., Ohno T., and Maki Y. (1977). Photomethylation of Inorganic Mercury by Aliphatic -Amino-acids. Journal of the Chemical Society-Chemical Communications: 158-159. Huber T., Faulkner G., and P. Hugenholtz (2004). Bellerophon: a program to detect chimeric sequences in multiple sequence alignments. Bioinformatics 20 (14): 2317-2319. Jay J. A., Morel F. M. M., and Hemond H. F. (2000). Mercury Speciation in the Presence of Polysulfides. Environmental Science & Technology 34 (11): 2196-2200. Kalyaeva E. S., Kholodii G. Y., Bass I. A., Gorlenko Z. M., Yurieva O. V., and Nikiforov V. G. (2001). Tn5037, a Tn21-like Mercury Resistance Transposon from Thiobacillus ferrooxidans. Russian Journal of Genetics 37 (8): 972-975. Kelly D. P. and A.P. W. (2005). Genus incertae sedis XVIII. Thiomonas. Bergey's Manual of Systematic Bacteriology. New York, Springer. 2, part C: 757-759. Kim C. S., Brown Jr. G. E., and Rytuba J. J. (2000). Characterization and Speciation of Mercury-Bearing Mine Wastes Using X-ray Absorption Spectroscopy. The Science of the Total Environment 261: 157-168. Kim C. S., Rytuba J. J. , and Brown Jr. G. E. (2004). Geological and Anthropogenic Factors Influencing Mercury Speciation in Mine Wastes: An EXAFS Spectroscopy Study. Applied Geochemistry 19: 379-393. King J. K., Kostka J. E., Frischer M. E., and Saunders F. M. (2000). Sulfate-Reducing Bacteria Methylate Mercury at Variable Rates in Pure Culture and in Marine Sediments. Applied and Environmental Microbiology 66 (6): 2430-2437. Krauskopf K. B. and Bird D. K. (1995). Introduction To Geochemistry, WCB/McGraw-Hill. Lane D., Pace B., Olsen G., Stahl D., Sogin M., and Pace N. (1985). Rapid determination of 16S ribosomal RNA sequences for phylogenetic analyses. Proceedings of the National Academy of Sciences of the United States of America 82 (20): 6955- 6959. Lee Z. M.-P., Bussema III C., and Schmidt T. M. (2009). rrnDB: documenting the number of rRNA and tRNA genes in bacteria and archaea. Nucleic Acids Research 37 (suppl_1): D489-493.

66

Linn R. K. (1968). New Idria Mining District. Ore Deposits of the United States, 1933- 1967. J. D. Ridge. New York, The American Institute of Mining, Metallurgical, and Peroleum Engineers, Inc. 2: 1623-1649. Lowry G. V., Shaw S., Kim C. S., Rytuba J. J., and Brown Jr. G. E. (2004). Macroscopic and Microscopic Observations of Particle-Facilitated Mercury Transport from New Idria and Sulphur Bank Mercury Mine Tailings. Environmental Science & Technology 38 (19): 5101-5111. Ludwig W., Strunk O., Westram R., Richter L., Meier H., Yadhukumar, Buchner A., Lai T., Steppi S., Jobb G., Forster W., Brettske I., Gerber S., Ginhart A. W., Gross O., Grumann S., Hermann S., Jost R., Konig A., Liss T., Lussmann, R. May M., Nonhoff B., Reichel B., Strehlow R., Stamatakis A., Stuckmann N., Vilbig A., Lenke M., Ludwig, T. Bode A., and Schleifer K.-H. (2004). ARB: a software environment for sequence data. Nucleic Acids Research 32 (4): 1363-1371. Madigan M. T., Martinko J. M., and J. Parker (2003). Brock Biology of Microorganisms. Upper Saddle River, Prentice Hall. Massana R., Murray A. E., Preston C. M., and Delong E. F. (1997). Vertical distribution and phylogenetic characterization of marine planktonic Archaea in the Santa Barbara channel. Applied and Environmental Microbiology 63 (1): 50-56. Mindlin S., Kholodii G., Gorlenko Z., Minakhina S., Minakhin L., Kalyaeva E., Kopteva A., Petrova M., Yurieva O., and Nikiforov V. (2001). Mercury Resistance Transposons of Gram-Negative Environmental Bacteria and Their Classification. Research in Microbiology 152: 811-822. Muyzer G., de Waal E. C., and Uitterlinden A. G. (1993). Profiling of Complex Microbial Populations by Denaturing Gradient Gel Electrophoresis Analysis of Polymerase Chain Reaction-Amplified Genes Coding for 16S rRNA. Applied and Environmental Microbiology 59 (3): 695-700. Paquette K. and Helz G. (1995). Solubility of Cinnabar (Red HgS) and Implications for Mercury Speciation in Sulfidic Waters. Water, Air, and Soil Pollution 80: 1053- 1056. Paquette K. E. and Helz G. R. (1997). Inorganic Speciation of Mercury in Sulfidic Waters: The Importance of Zero-Valent Sulfur. Environmental Science & Technology 31 (7): 2148-2453. Pruesse E., Quast C., Knittel K., Fuchs B. M., Ludwig W., Peplies J., and Glockner F. O. (2007). SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Research 35 (21): 7188-7196. Ravichandran M., Aiken G. R., Reddy M. M., and Ryan J. N. (1998). Enhanced Dissolution of Cinnabar (Mercuric Sulfide) by Dissolved Organic Matter Isolated from the Florida Everglades. Environmental Science & Technology 32 (21): 3305-3311. Ravichandran M., Aiken G. R., Ryan J. N., and Reddy M. M. (1999). Inhibition of Precipitation and Aggregation of Metacinnabar (Mercuric Sulfide) by Dissolved Organic Matter Isolated from the Florida Everglades. Environmental Science & Technology 33 (9): 1418-1423. Redd, M. M. and Aiken G. R. (2001). Fulvic Acid-Sulfide Competition for Mercury Ion Binding in the Florida Everglades. Water, Air, and Soil Pollution 132: 89-104. Rytuba J. J. (2003). Mercury from Mineral Deposits and Potential Environmental Impact. Environmental Geology 43: 326-338.

67

Smith R. M. and Martell A. E. (2004). NIST Critically Selected Stability Constants of Metal Complexes Database. NIST Standard Reference Database. Washington D.C., National Institute of Standards and Technology. Version 8.0. Starr M. P. (1981). The Prokaryotes: A handbook on habitats, isolation, and identification of bacteria. Berlin, Springer-Verlag. Stookey L. L. (1970). Ferrozine-A New Spectrophotmetric Reagent for Iron. Analytical Chemistry 42 (7): 779-781. Ullrich S. M., Tanton T. W., and Abdrashitova S. A. (2001) Mercury in the Aquatic Environment: A Review of Factors Affecting Methylation. Critical Reviews in Environmental Science & Technology 31 (3): 241-293. U.S. EPA (2002). Method 1631, Revision E: Mercury in Water by Oxidation, Purge and Trap, and Cold Vapor Atomic Fluorescence Spectrometry. U. S. Environmental Protection Agency. Washington D.C. Waples J. S., Nagy K. L., Aiken G. R., and Ryan J. N. (2005). Dissolution of Cinnabar (HgS) in the Presence of Natural Organic Matter. Geochimica et Cosmochimica Acta 69 (6): 1575-1588. Weishaar J. L., Aiken G. R., Bergamashi B. A., Fram M. S., Fujii R., and Mopper K. (2003). Evaluation of Specific Ultraviolet Absorbance as an Indicator of the Chemical Composition and Reactivity of Dissolved Organic Carbon. Environmental Science & Technology 37 (20): 4702-4708. Wiatrowski H. A., Das S., Kukkadapu R., Ilton E. S., Barkay T., and Yee N. (2009). Reduction of Hg(II) to Hg(0) by Magnetite. Environmental Science & Technology 43 (14): 5307-5313. Wood J. M. (1974). Biological Cycles for Toxic Elements in the Environment. Science 183 (4129): 1049-1052. Zhou J., Bruns M. A., and Tiedje J. M. (1996). DNA Recovery From Soils of Diverse Composition. Applied and Environmental Microbiology 62 (2): 316-322.

68

Table 1: S-Oxidizing microorganism isolation medium (Starr, 1981)

Chemical g/L

(NH4)2SO4 1.0

KCl 0.1

Ca(NO3) 0.01

. 0.2 MgSO4 7 H2O 0.5 K2HPO4 3.2 Na2S2O3

Trace Element Solution 1 mL/L

Trace Element Solution g/L

EDTA 50

ZnSO4 2.2

CaCl2 5.5

. 5.1 MnCl2 4H2O

. 5.0 FeSO4 7H2O 1.1 . (NH4)6Mo7O24 4H2O 1.6 CuSO . 5H O 4 2 1.6 . CoCl2 6H2O

Extra Chemicals for plates g/L

Bacto-Agar 15

Bromothymol Blue 0.02

HgCl2 1.0

69

Table 2: Real-time PCR primers used in this study.

Name Sequence (5’->3’) Target Anneali Fragme Referen group ng Temp nt ce length [bp]

Thio636 GGATGACTATCCGAC Thiomon 55°C 218 This F TGG as sp. study

Thio836 TACTGAACAGTTGCC Thiomon This R CGT as sp. study

Bact341 CCTACGGGAGGCAGC Bacteria 60°C 194 Muyzer F AG et al., 1993

Bact534 ATTACCGCGGCTGCT Bacteria Muyzer R GGC et al., 1993

70

Table 3: Selected metals and anions in the New Idria AMD water used in the HgS dissolution experiment.

Concentration Analyte (mg/L) Al 88.6 Ca 222 Fe 368 K 45.4 Mg 233 Na 364 Zn 1.8 Mn 3.8 Silica 19.3 Cl 211 F 4 NO3 <.08 SO4 4587

PO4 <0.01

71

Ferritrophicales (157) Burkholderiales; Genera incertae sedis Thiomonas sp. (34) Gallionellales (19) Xanthomonadales (16) Rhodospirillales (9) Others (13)

Fig. 1: Summary of the bacterial 16S rRNA gene clone library results of the AMD pond water. Classification was based on the Greengenes classifier using the NCBI taxonomy. Taxonomic orders shown in the pie chart had more than 3% sequence abundance in the clone library. All other obtained sequences are grouped as ‘others’. The total number of nearly full length (>1300 bp) 16S rRNA gene clones obtained was 248.

72

Thiomonas sp. and total Bacteria abundance in AMD pond sediment core

0 Thiomonas sp. 2 Total Bacteria

4

6

8

10

sediment depth [cm] depth sediment 12

14

16 102 103 104 105 106 16S rRNA gene copies per mg sediment

Fig. 2: 16S rRNA gene copy numbers per mg sediment in a sediment core from the New Idria AMD pond as determined by real-time PCR. Filled circles are Thiomonas sp. 16S rRNA gene copy numbers; Open circles are total bacterial 16S rRNA gene copy numbers. Shown are average values and standard deviations from two independent DNA extractions and triplicate real-time PCR reactions (a total of six values per data point).

73

600000 Cinnabar Alive

Tailings Alive

500000 Calcine Alive Metacinnabar Alive Cinnabar Killed 400000 Tailings Killed Calcine Killed Metacinnabar Killed 300000

200000 Hg Concentration Hg (ng/L) 100000

0 0 5 10 15 20 25 30 Incubation Time (Days)

Fig. 3: Mercury release from HgS bearing materials for oxic microcosms using the New Idria microbial community. Solid lines denote biologically active AMD microbial community material added (7.88 μg of protein added), while dashed lines denote gamma irradiated cells added (8.12 μg of protein added). Error bars are smaller than symbols, error is ± 8% (n=3). Experiment conducted during

August, 2008.

74

1200 Cinnabar Alive

1000

Tailings Alive

800 Calcine Alive

Metacinnabar 600 Alive Cinnabar Killed 400 Tailings Killed

200 Hg Concentration (ng/L) Concentration Hg Calcine Killed 0 0 5 10 15 Metacinnabar Killed Incubation Time (Days)

Fig. 4: Mercury release from HgS containing materials in anoxic microcosms using the New Idria microbial community material. Solid lines denote biologically active microbial community material added (7.88 μg of protein added), while dashed lines denote gamma irradiated cells added (8.12 μg of protein added). The initial pulse of Hg in the experiment is due to the growth media not being purged of O2 prior to beginning of experiment. Error bars are smaller than the symbols, and estimated standard errors are ± 8% (n=3). The experiment was conducted during August, 2008.

75

400 Cinnabar Alive

350 Metacinnabar Alive

300 Tailings Alive Calcine Alive 250 Cinnabar Killed 200 Metacinnabar Killed Tailings Killed 150 Calcine Killed 100 Water 50 Cinnabar Abiotic Metacinnabar Abiotic

Fe (II) Concentration (mg/L) Concentration(II) Fe 0 0 5 10 15 20 25 30 Tailings Abiotic Incubation Time (Days) Calcine Abiotic

Fig. 5: Oxidation of Fe(II) during aerobic incubation of living cells. Error bars are smaller than the symbols, and the error is ± 0.1% (n=3). The trend seen in the metacinnabar batch reactor is identical to anaerobic incubation, killed controls, and abiotic controls. The experiment was conducted during August, 2008.

76

200 Abiotic 180 NoHg1

160 NoHg2 140 500ng 120

1,000ng 100

80 5,000ng

60 10,000ng

40 50,000ng Fe(II) Concentration (mg/L) Concentration Fe(II) 20 100,000ng 0 0 2 4 6 8 10 12 14 Incubation Time (Days)

Fig. 6: Effect of added Hg to Fe(II) oxidation capability of the New Idria microbial community. Mercury was added to reaction vessels as Hg(NO3)2 and pH was held at pH = 4. Error bars are smaller than symbols, error is ± 0.1% (n=3).

Initial drop in Fe(II) concentration is assumed to be adsorption of Fe to the sides of the reaction vessel. Fe(II) accounts for 100% of the Fe at the beginning of the experiment. Experiment conducted during February, 2010.

77

Chapter 3

A Sequential Chemical Extraction and

Spectroscopic Assessment of the Potential

Bioavailability of Mercury Released From the

Inoperative New Idria Mercury Mine, San Benito

Co., CA

Adam D. Jewa*, Phi N.M. Luonga, James J. Rytubab, and Gordon E. Brown Jr.a,c,d

a Surface and Aqueous Geochemistry Group, Department of Geological & Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA b Mineral Resources Program, U.S. Geological Survey, 345 Middlefield Road, MS 901, Menlo Park, CA 94025, USA c Department of Chemical Engineering, Stanford University, Stanford, CA 94305, USA d Department of Photon Science and Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, 2545 Sand Hill Road, MS 69, Menlo Park, CA 94025, USA

Submitted for publication to Geochimica et Cosmochimica Acta

78

Abstract:

The inoperative New Idria mercury mine in San Benito County, California, is a potential point source of Hg to the Central Valley of California and serves as a model system for studying Hg transport from similar silica-carbonate mercury deposits in the California Coast Range. To determine the phases of Hg present in stream bed deposits downstream of the mine, sequential chemical extractions

(SCEs) targeting Hg-bearing phases and synchrotron x-ray-based spectroscopic and imaging techniques were used on sediment samples taken from the acid mine drainage (AMD) system and downstream, where pH is circum-neutral. Also characterized using the same methods were Hg adsorbed to ferrihydrite (both synthetic 2-line and natural) in the laboratory, and Hg associated with diatom-rich field samples; both associations could affect mercury transport in the New Idria drainage and other similar systems. In all field samples examined, mercury sulfide was found to be the dominant inorganic Hg-bearing phase present, with no detectable Hg associated with the natural ferrihydrites. SCE analyses of Hg(II) sorbed to synthetic 2-line and natural ferrihydrite in the laboratory showed that

Hg(II) did not bind strongly to either material under conditions similar to those in the New Idria drainage system. Mercury not associated with HgS is associated mainly with freshwater diatoms (30-60% or 2.3-4.2 ppm Hg) found in the downstream sediments. For all samples collected from the New Idria system, removal of >97% of the Hg required 1M KOH or harsher chemical treatments, indicating that most of the Hg, other than elemental Hg, in these sediment samples is environmentally stable and not readily bioavailable.

79

1. Introduction

Mercury from abandoned Hg mines in the California Coast Range is an environmental and potential human health concern, particularly in ecosystems downstream of these mines (RYTUBA, 2003). Different Hg-bearing phases have significantly different solubilities (KRAUSKOPF and BIRD, 1995), different potentials for conversion into monomethylmercury by sulfate- and iron-reducing bacteria (MOREL et al., 1998), and different toxicities in natural aquatic and soil systems (CLARKSON and MAGOS, 2006), making it important to identify Hg- bearing phases at abandoned Hg mines. The New Idria Hg mine is located in San

Benito County, California, and was the second largest Hg producer in North

America, operating from 1854 to 1972 and resulting in mine wastes containing significant amounts of mercury (KIM et al., 2000; LINN, 1968; LOWRY et al.,

2004; RYTUBA, 2003). Although recently designated a Superfund Site by the US

EPA, the mine site has not been remediated since its closure, making it a good proxy for other abandoned Hg mines in California (LINN, 1968; RYTUBA, 2003).

After the mine closed, an acid mine drainage (AMD) system developed at the site.

This system begins in the underground mine workings and discharges at a single mine adit on top of the tailings waste pile. The pH of the AMD system varies from ~5.8 at the mine adit to as low as 2-4 (depending on the time of year) in a settling pond ~250 m from the adit. Less than 500 m downstream from the settling pond, waters from the AMD system mix with the waters of San Carlos

Creek (pH 9-10), raising the pH of the AMD water to ~7 (Supporting

Information Fig. S1). Stream pH ranges from 7 to 8.5 throughout the rest of the

80 drainage system due to buffering by calcite-cemented sandstone of the Panoche

Formation (LINN, 1968).

Past research has identified Hg-bearing phases as a function of particle size at the New Idria site. For example, Lowry et al. (2004) combined sequential chemical extractions (SCEs), colloid separation, transmission electron microscopy

(TEM), and Hg LIII-edge extended x-ray absorption fine structure (EXAFS) spectroscopy to show that the main Hg-bearing phase being released from waste piles at the site is HgS in the form of nanoparticles 20-30 nm in diameter. Kim et al. (2000; 2004a) used ambient-temperature EXAFS methods to show that the main Hg-containing phases at this site include cinnabar (hex-HgS), metacinnabar

(cub-HgS), montroydite (HgO), corderoite (Hg3S2Cl2), and eglestonite

(Hg6Cl3O(OH) or Hg4Cl2O). Both cinnabar and metacinnabar are relatively insoluble, whereas the other three phases are relatively soluble and thus have higher bioavailabilities than either form of HgS. More recently, we developed a new low-temperature sample preparation protocol that results in the crystallization of Hg(0) in complex mine wastes to -Hg, which can be detected and quantified using low-temperature EXAFS spectroscopy (JEW et al., 2011).

Use of this new method on New Idria samples, including tailings, calcines, and sediments revealed that some materials contain up to 10% of the total Hg in the form of Hg(0). Ganguli et al. (2000) investigated Hg speciation downstream from the New Idria mine by analyzing total Hg, monomethylmercury, and dissolved gaseous Hg concentrations. However, none of these studies determined the speciation of Hg associated with solids in downstream sediments. The main focus

81 of the present study is to identify these Hg-bearing species and to determine if significant amounts of Hg(II) are sorbed on the abundant ferrihydrite nanoparticles in the New Idria drainage system. Such information is essential for assessing the site’s potential as a continuing point source for Hg for downstream sites and for developing remediation strategies for the New Idria site and similar

Hg-containing AMD systems in California that drain into surrounding watersheds.

Selective chemical extractions were used to achieve this goal because the concentration of Hg-bearing phases in the New Idria drainage system sediments is well below the lower limit of pyrolysis methods or molecular-level methods such as EXAFS spectroscopy.

2. Mineralogy and Petrology of the New Idria Mine Site

The country rock of the New Idria mine consists of (1) a serpentinite massif (up gradient of the mine), (2) silica-carbonate rock altered from the serpentinite that hosts the main Hg ore body, (3) the Panoche formation (gray shale and brown massive, concretionary sandstone), which surrounds the silica- carbonate rock, and (4) Franciscan group rocks (massive, arkosic sandstone, minor shale, chert, and greenstone) found downstream of the mine (LINN, 1968).

Bulk mineralogy of the New Idria samples is grouped into two categories depending on whether the samples were taken from the low pH or the circum- neutral pH portion of the New Idria drainage system. The low pH regime samples

(NI-1 and NI-2) have mineralogy similar to most AMD systems, which are dominated by quartz and iron-(oxy)hydroxides (Table 1) (CORNELL and

82

SCHWERTMANN, 2003). For downstream sediment samples collected in circum- neutral pH waters, there are no detectable iron-(oxy)hydroxides during the wet season due to their removal by stream erosion, but these phases are present in dry season sediments. One significant difference in mineralogy between the wet season and dry season samples is that the former contain calcite whereas the latter contain aragonite. Mixing of the low pH AMD water with the high pH water of

San Carlos Creek during the dry season results in the precipitation of aragonite along with ferrihydrite and schwertmannite. Aragonite has not been observed during the wet season. Calcite in the wet season samples is thought to be derived from the weathering of the calcite-cemented sandstone country rock. Diatom-rich sediments as well as diatomaceous earth is also found in this drainage system, and the dominant solid phases in the latter consist of opal-A and opal-CT.

3. Experimental Methods

3.1 Field Sampling

Sediment samples were collected on February 10, 2010 and September 6, 2011

(downstream locations only) within the New Idria AMD system (pH 2-5.5), San

Carlos Creek before mixing with the AMD water (pH 9-10), and San Carlos

Creek below the mixing point (pH 7-8.5) (Supporting Information Fig. S1). All sediment samples were collected at a depth of 0-2.5 cm. Sediments were drained before being stored in acid-cleaned borosilicate glass containers and immediately placed on ice to limit microbial changes to Hg within the samples. Prior to drying, sediments were decanted of excess water in sample containers. To limit

83 the loss of elemental Hg through thermal drying or loss by sublimation during freeze drying, samples were placed in a covered desiccator at room temperature.

The dessicator was attached to the building’s vacuum system and the vacuum valve was opened only slightly; samples were dried for 3 days under these conditions. Dried samples were sieved to <100μm size fraction for SCE analysis.

Prior to drying samples, a portion of the sediment was used to determine the

o percent water in the samples by weight loss during drying at 105 C, with percent water varying from a maximum of 90 ± 0.3% for sediments in the AMD system to a minimum of 14 ± 3% for downstream sediments.

Bulk mineralogy was determined by x-ray diffraction (XRD). Samples were analyzed on a Rigaku Model CM2029 powder x-ray diffractometer using a

o Cu Kα x-ray source over the 2 range 5° to 70 . Analysis of the resulting diffractograms was done using the Jade X-ray Diffraction Software package

(MATERIALS DATA, 2002). Identification of minerals present was done by peak matching the four peaks with the highest intensity for each phase. Results of

XRD analysis are presented in Tables 1.

3.2 2-line Ferrihydrite Synthesis, Characterization, and Hg Uptake

Experiments

2-line ferrihydrite was synthesized for Hg uptake experiments (synthesis and characterization methods detailed in Supporting Information). Because naturally occurring ferrihydrite from New Idria is much more complex than synthetic 2-line ferrihydrite (Cismasu et al., 2011), ferrihydrite samples collected

84 in the field were also used in Hg uptake experiments. All stock solutions, products of adsorption uptake experiments, and samples were stored in ultra-trace clean borosilicate vials with Teflon® lined lids to reduce the adsorption of Hg to the sides and lids of the sample vials. Due to the low binding affinity of Hg to

- NO3 , Hg(NO3)2 was chosen for the adsorption experiments. A 0.282 M

Hg(NO3)2 stock solution (Ricca Chemicals) was diluted to 0.565 mM Hg(NO3)2 at which point the stock was separated into two batches that were adjusted to either pH 3.5 (using H2SO4) or pH 7.45 (using NaOH). These stock solutions were used to represent the two major pH regimes in the New Idria drainage system. Final

Hg(II) concentrations of the two solutions were 0.564 mM for pH = 3.5 and 0.560 mM for pH 7.45. These solutions were added directly to 0.4000 ± 0.0070g of synthetic or natural New Idria ferrihydrite, resulting in one sample of each material exposed to Hg(II) at pH 3.5 and pH 7.45. Samples were mixed on an end-over-end tumbler for 24 hours. After exposure, the vials were centrifuged at

2,500 rpm and filtered through 0.45μm glass microfiber syringe filters and preserved with 0.5% bromine monochloride. The leftover ferrihydrite pellet was dried without washing as described above. The percentage of Hg(II) uptake was determined based on the concentration of Hg in solution before and after exposure to the ferrihydrite. The mass balance of Hg uptake included the recovery of Hg from the SCE analysis, a total digest of a portion of the ferrihydrite sample exposed to Hg, and a total digest of a portion of the ferrihydrite sample prior to exposure to aqueous Hg(II).

85

3.3 Diatomaceous Earth and Diatom-rich Samples

Abundant diatoms were detected in sediments from the New Idria AMD system including downstream sediments by SEM (Fig. 1). SCE analyses were carried out on Hg-bearing samples of diatom-rich stream sediments and diatomaceous earths from New Idria and other California localities as well as on commercially available, Hg-containing, diatomaceous earth samples (Table 1). A major objective here was to determine the speciation of Hg associated with the diatom- rich samples. Field samples from Hg-impacted sites included two samples from

Silver Creek (denoted as SC, New Idria drainage, San Benito County, California) and four samples from Harley Gulch (denoted as HG, drainage of the remediated

Turkey Run Mercury Mine, Lake County, California). The other natural diatom- rich samples containing Hg came from a variety sources. Samples D-1 to D3 are field samples of diatomaceous earth from the Petaluma Formation, Sonoma

County, California. Samples D-4 to D-8 are commercially available diatomite from various California localities. Sample D-4 is Wolf Creek Ranch food grade diatomaceous earth. Sample D-5 is diatomaceous earth from Safer Brand Ant and

Crawling Insect Killer. Sample D-6 is diatomite in Blue Ribbon DE Premium Cat

Litter. Sample D-7 is diatomite in the commercial product Concern DE Crawling

Insect Killer, and Sample D-8 is Pooltime DE Filter Powder. Samples D-4 to D-8 have ages that range from Pliocene to Miocene. Sample D-8 (Pooltime DE Filter

Powder) was heated to temperatures in excess of 980oC for one hour to remove any impurities and convert amorphous silica to cristobalite. Due to the age of the diatomaceous earth samples (samples denoted with D) and the heating of Sample

86

D-8, it is hypothesized that all organic material has been removed from these samples. To confirm this hypothesis, total carbon and total nitrogen for these samples were determined using a Carlo-Erba brand Dumas combustion analyzer.

3.4 Total Hg Determination

Total Hg in sediment samples was extracted using an aqua regia digest (8 mL

HCL: 2 mL HNO3, preserved with 0.5% BrCl) and Hg concentrations were determined using a Tekran 2600 series cold vapor atomic fluorescence spectrometer (CVAFS) following EPA method 1631 (U.S. EPA, 2002). The total

Hg concentrations for the sediments were compared with the summed concentrations determined for different SCE fractions (detailed below) and showed between 89.3 and 106.2% recovery.

3.5 Sequential Chemical Extractions

Sequential chemical extractions (SCE’s) followed the method of Bloom et al.

(2003) with minor changes. The chemical extraction steps are operationally defined to target specific Hg-bearing phases. The Bloom et al. (2003) method consists of exposing samples to a sequential series of 5 separate chemical extraction steps in the following order (with the expected Hg-phase released indicated in parentheses): (F1) distilled water (water-soluble Hg), (F2) 0.1M acetic acid + 0.01M hydrochloric acid (‘human stomach acid’-soluble Hg), (F3)

1M potassium hydroxide (organo-chelated Hg), (F4) 12M nitric acid (elemental

Hg), and (F5) aqua regia digest (mercuric sulfide/selenide). The different

87 extraction steps are designed to target specific Hg-bearing phases and provide information on the potential bioavailability of Hg in a sample containing one or more Hg-bearing phases. However, the Bloom et al. (2003) SCE procedure applied to complex Hg mine waste samples is not without some ambiguity in certain cases. For example, a 1M KOH extraction step was chosen by Bloom et al. to target organo-chelated Hg, but organo-chelated Hg compounds such as monomethyl-Hg are also removed in part by their F2 extraction, as defined below.

Moreover, Bloom et al. (2003) also found that their F4 step (12M HNO3), which was designed to extract elemental Hg, also extracts Hg bound to humic acids, further adding to the complexity of interpreting the SCE data when Hg-bearing samples are taken from organic-rich settings such as wetlands.

In the present study we added an additional step to the Bloom et al. extraction protocol between steps F1 and F2 consisting of 1M magnesium chloride which should remove outer-sphere Hg(II) sorption complexes. To minimize confusion between the Bloom et al. extraction steps and extraction steps in the present study, our steps are denoted as SCF1 (distilled water - water- soluble Hg), SCF2 (1M MgCl2 - outer-sphere adsorbed), SCF3 (0.1M acetic acid

+ 0.01M hydrochloric acid - ‘human stomach acid’-soluble Hg), SCF4 (1M potassium hydroxide - organo-chelated Hg), SCF5 (12M nitric acid - elemental

Hg), and SCF6 (aqua regia digest - mercuric sulfide/selenide). Although Kim et al. (2003) showed that extraction steps F2 and F3 of the Bloom et al. protocol

(2003) produce Hg-bearing phases that were not present in the original sample, the 1M MgCl2 wash should not result in formation of any new Hg-bearing phases

88 in our samples. For example, the potential for creating solid Hg2Cl2 during the 1M

MgCl2 extraction step is of minor concern because MgCl2 should not cause reduction of Hg(II) to Hg(I). In addition, because Hg2Cl2 has been shown to be extracted by three different extraction steps (F2, F3, and F4), a reliable determination of Hg2Cl2 concentration is difficult by the BLOOM et al. (2003)

SCE procedure. However, we are confident in the standard interpretation of the

SCF1 (distilled water), SCF2 (1M MgCl2), and SCF6 (HgS/HgSe) extraction steps. Although the SCF3, SCF4, and SCF5 extraction steps are more difficult to interpret, they do provide important information about the stability of Hg-bearing phases in the various samples analyzed.

Sequential chemical extractions were performed by placing 0.4 ± 0.04g of dried sediment into borosilicate vials with Teflon®-lined lids. Extractants were added to the sample-containing vials and mixed on an end-over-end tumbler at

120 rpm for 24 ± 2hrs. The reacted samples were then centrifuged at 2,600 rpm for 20 minutes after which the supernatant was filtered using 0.45μm glass microfiber (GMF) syringe filters, placed into ultra-trace clean borosilicate bottles with Teflon®-lined lids, and preserved with 1% v/v BrCl (8% for the KOH fraction). GMFs were used because these filters do not adsorb Hg or dissolve during the final two steps of the SCE protocol (pH < 0), allowing for filtering of all SCE fractions. A rinse step was performed by adding additional extractant and re-suspending the solid followed by centrifugation and subsequent filtration.

Samples were analyzed on a Tekran 2600 series CVAFS following EPA method

1631 (U.S. EPA, 2002). Though great care was taken to homogenize samples,

89 recoveries greater than 100% are due to heterogeneities in the samples and are similar to the SCE recoveries and errors for duplicate samples reported by Bloom et al. (2003).

3.6 Spectroscopic Data Collection and Analysis

To determine the Hg-bearing phase(s) present in the SCF6 fraction (HgS/HgSe fraction), samples were analyzed at the Stanford Synchrotron Radiation

Lightsource (SSRL) using micro x-ray fluorescence (-XRF), micro x-ray absorption fine structure (μ-XAFS), and bulk EXAFS spectroscopy. A description of the -XRF and -XAFS methods used and the resulting data are included in the Supporting Information.

Bulk Hg LIII-edge EXAFS spectra were collected at ambient temperature for the New Idria AMD settling pond sediments and at 77K for diatom-rich samples collected downstream of New Idria (Silver Creek, San Benito County), a sample from Harley Gulch (downstream of the remediated Turkey Run mine,

Lake County, California), and sample NI-2 on SSRL wiggler beamline 11-2 using

o a LN2-cooled Si (220) monochromator in the phi = 90 orientation and a 30- element Ge array detector. EXAFS spectra were collected only for these samples because the total Hg present in the other samples from the drainage system was too low in concentration (< 50 ppm Hg) for Hg LIII-edge EXAFS spectroscopy.

90

3.7 Analysis of Spectroscopic Data

EXAFS data averaging, background subtraction, and data fitting were carried out using the SixPACK XAS analysis software package (WEBB, 2005; WEBB, 2006).

EXAFS spectra were analyzed by linear combination fitting (LCF) and shell-by-

-1 shell fitting over a k-range of 3-9.5 Å . Previous LCF of Hg LIII-edge EXAFS spectra for physically mixed Hg model compounds at ambient-temperatures has shown an accuracy of ~5% and ~10% in quantifying the phase composition of two- and three-Hg-component systems, respectively (KIM et al., 2000). The quality of fits was determined by calculating a residual of the following form:

n 2 S (vdata - v ) t = 1 fit Residual = n where “v” is the k3-weighted (k) value in the EXAFS spectra (for both the data and fit) and “n” is the number of data points in the fit.

4. Results

4.1 Total Hg Results

Total Hg concentrations for all sediment samples are listed in Table 2. Samples

NI-4 (wet season) and NI-5 (dry season) were run in duplicate, with recoveries of

98.6 and 100.0% (Sample NI-4) and 95.6 and 97% (Sample NI-5). Because silica becomes more insoluble as pH decreases (Krauskopf and Bird, 1995), acid digestions of the diatom-rich samples resulted in the incomplete release of Hg from these samples. Due to this complication, the total Hg concentrations for the

91 diatomaceous earth samples (Samples D-1 to D-8: Table 2) were calculated as the sum of all six extraction steps.

4.2 SCE Results for Field Samples

Duplicate analyses of field samples showed variations of <15% for SCE fractions comprising >1% (by weight) of the total Hg. Due to multiple Hg-bearing phases being extracted by the different extractants, it is difficult to determine quantitatively by SCE which Hg-bearing phases are present in samples except for the water-soluble Hg (SCF1), outer-sphere adsorbed Hg (SCF2), and HgS/HgSe

(SCF6) fractions. Nevertheless, it is still possible to assess qualitatively the environmentally important Hg fractions (BLOOM et al., 2003). Based on the findings of Kim et al. (2003), the aqua regia-extracted Hg species (HgS and

HgSe) (step SCF6) are considered reliable, and there is good agreement in our study between the results of SCE and EXAFS analyses for sample NI-2, which had a Hg concentration sufficiently high for EXAFS analysis (a sample from the

New Idria AMD pond, ~2 m from sample NI-2, was also analyzed by EXAFS spectroscopy but was not analyzed by SCE).

In all field samples analyzed, SCE fraction SCF6 (HgS/HgSe) comprised

>40% of the total Hg in the samples (Fig. 2). In addition, >97% of all Hg in all samples analyzed was removed by the three harshest chemical extractants (1M

KOH, 12M HNO3, and aqua regia). Samples NI-1 and NI-2 (both from the low pH portion of the New Idria AMD system) were the only ones found to contain

Hg-bearing phases extractable by simulated ‘human stomach acid’ (3% and 0.1%

92 of the total, respectively). The absence of these phases in the higher pH portions of the New Idria drainage system suggests that the fraction of Hg that should be considered bioavailable is not stable as pH increases. Bloom et al. (2003) showed that a significant amount of HgO (montroydite) in some Hg-bearing samples they examined is removed by the simulated ‘human stomach acid’ extraction step, which is similar to the findings of Lowry et al. (2004), who reported that HgS and

HgO are the main colloidal forms of Hg released during leaching experiments on

New Idria calcine material. The SCF3 ‘simulated stomach acid’ fraction in the low pH AMD system at New Idria was not found in other sampling localities at

New Idria. A significant amount of Hg from downstream sites during the dry season is removed by the SCF4 extraction step, suggesting that it is organo- chelated Hg (Fig. 2).

4.3 Ferrihydrite Adsorption and SCE Results

The phase identity and purity of our synthetic 2-line ferrihydrite (5Fe2O3∙9H2O)

(Cornell and Schwertmann, 2003) were confirmed by XRD. Characterization of the nanoparticles by TEM showed that 2-line ferrihydrite particles ranged in size

2 from 1-3 nm, with an average surface area of 328 m /g, as determined by N2 BET measurements. Total Hg adsorption to synthetic 2-line ferrihydrite is 5.25 mg/g and 6.75 mg/g for pHs 3.5 and 7.45, respectively (Table 2). Total Hg adsorption to ferrihydrite/schwertmannite (Fe16O16(OH)y(SO4)z∙nH2O) (Cornell and

Schwertmann, 2003) collected from New Idria showed no significant variation with pH, with Hg concentrations of 7.25 ± mg/g for both pH 3.5 and 7.45. This

93 pH independence of Hg adsorption implies that the Hg is not adsorbing to the inorganic portion of the New Idria ferrihydrite but is adsorbing instead to the organic matter associated with the ferrihydrite. Though Hg adsorption to some types of organic matter has a pH-dependency, the binding constants of Hg to dissolved organic matter are so high, generally 1017 to 1032, that pH-dependence is a minor effect (BENOIT et al., 2001; HAITZER et al., 2003). The dominant controls on Hg binding to organics are the type of organic material present (AIKEN et al.,

2003; DREXEL et al., 2002; RAVICHANDRAN, 2004) and the ratio of Hg to organic matter (BENOIT et al., 2001; HAITZER et al., 2002). SCE analyses show that the majority of the Hg (>90%) from both synthetic 2-line ferrihydrite and natural

New Idria ferrihydrite was removed by distilled water and 1M MgCl2 solutions, suggesting that it was weakly sorbed (Fig. 3A). Because the samples were not washed after exposure to Hg(II), the resulting Hg extracted from the SCF1 extraction step is considered to be Hg left in residual water that was not removed prior to drying. When the SCF1 step is ignored, the SCF2 (MgCl2) extraction step results in the removal of 89 to 91% of the total Hg for synthetic 2-line ferrihydrite and 80 to 82% for natural ferrihydrite (Fig. 3A).

4.4 Diatom Results

Dumas combustion analysis, showed a carbon content of <1% (by weight) for samples D-1 to D-8 and <10% (by weight) for Silver Creek and Harley Gulch samples. Due to the high temperature of the Dumas combustion technique,

~1,080oC, carbonate-containing species will be detected as total carbon. All

94 diatomaceous earth and diatom-rich samples had no detectable nitrogen. Because there are relatively few naturally occurring minerals with nitrogen as a main constituent, we consider the lack of detectable nitrogen in the samples as evidence for the organics having been removed from the samples by either diagenesis or in the case of Sample D-8, by high temperature processing.

The lack or low level of organics in the freshwater diatoms suggests that the associated mercury is not bound to organic matter. Alternative modes of Hg association with these freshwater diatoms will be discussed in section 5. New

Idria drainage samples were investigated by SEM to determine the phase(s) with which Hg is associated. Abundant freshwater diatoms were found to be present in all samples taken from New Idria (both low pH and circum-neutral pH) and are associated with abundant iron-(oxyhydr)oxides (Fig. 1). To determine if the

SCF4 fraction of the SCE analysis of New Idria samples was due to Hg adsorption to organics within the diatoms, we carried out numerous SCE analyses of diatom-rich sediments from both Hg-mine impacted and commercial sources.

Reproducibility of SCE results for these diatom samples was very good, with less than 5% difference in Hg concentration in all fractions as well as for the sum of the six Hg fractions. Total Hg concentrations for the different diatom samples were highly variable, ranging from 52.3 ppb to13.8 ppm (sum of SCE fractions)

(Table 2).

Most of the Hg in the diatom samples was removed during the SCF4 and

SCF6 extraction steps (Fig. 3B). The majority of the samples (13 of 15 samples) had more than 50% of the Hg removed by the SCF4 extraction step (1M KOH,

95 organo-chelated). All diatom samples (including field samples) had no detectable

Hg in the SCF1-SCF3 fractions. These results suggest that Hg associated with diatoms, even for samples heated in excess of 900oC, is quite stable and not highly bioavailable.

4.5 Spectroscopic Results

Although SCE analysis of sample NI-2 (directly below the mine tailings pile) resulted in the HgS/HgSe Hg fraction comprising >98% of the total Hg-bearing phases, -XRF analysis and -XAFS spectroscopy of the AMD settling pond core sample were used to confirm the SCE findings and to provide information on the sizes of HgS/HgSe particles (results in Supporting Information). EXAFS spectroscopic analyses were carried out on sediments taken directly below the mine waste pile (sample NI-2), an AMD settling pond core sample taken a little more than a year earlier, and diatom-rich samples from both Harley Gulch (Lake

County, California) and Silver Creek (San Benito County, California).

Linear combination fits of the bulk EXAFS spectra of the AMD pond sediments (Fig. 4A) and Sample NI-2 (Fig. 4B) resulted in percentages of cinnabar and metacinnabar ranging from 75-82% and 18-25%, respectively. LCF analysis of an EXAFS spectrum of the tailings material collected by slow cooling the sample to 77K using the method of Jew et al. (2011) indicates 32% cinnabar and 68% metacinnabar (spectrum not shown), whereas Lowry et al. (2004) reported that leached colloids consist of 36% cinnabar, 46% metacinnabar, and

22% montroydite based on ambient temperature EXAFS analysis. Though elemental Hg has been detected using a hand lens within tailings material at the

96 site, when using the slow cooling EXAFS method for the tailings sample, we found no spectroscopic evidence for elemental Hg in the specific tailings sample analyzed. Due to the high heterogeneity of material at the site, the lack of detectable elemental Hg(0) by EXAFS spectroscopy on several samples does not indicate that it is not present in other samples from the New Idria waste piles.

The lack of evidence for metacinnabar in the New Idria AMD settling pond sample, based on analysis of the μ-XAFS spectra (Supporting Information), suggests that metacinnabar is too low in concentration to be detected by -XRF and μ-XAFS in our samples. There were no additional identifiable Hg-bearing species based on μ-XAFS or EXAFS analysis in the mine waste samples; however, because the samples were analyzed at ambient temperature, α-Hg(0) would not have been detected (JEW et al., 2011). Based on SCE analysis of sample NI-2 (235.3 ± 4.7 ppm Hg) and assuming that all of the Hg detected in fraction SCF5 (1.1% of the total Hg in the sample, Fig. 2) was solely in the form of α-Hg(0) (i.e. formed by slow cooling of liquid mercury; see Jew et al., 2011),

Hg(0) would not have been detectable by low-temperature EXAFS. This is the case here because LCF of low-temperature EXAFS data typically requires that an identifiable Hg-bearing phase comprises at least 5% of the total Hg-bearing phases in the sample.

LCF analysis of the EXAFS spectra of the diatom-rich Harley Gulch

(Lake County, California) and Silver Creek (San Benito County, California) samples showed that they were practically identical and contained 75% cinnabar and 25% metacinnabar. Shell-by-shell fitting of the Harley Gulch sample resulted

97 in a Hg-S pair correlation with Hg coordinated by 3.50 ± 0.43 sulfurs at a distance of 2.43 ± 0.01 Å (Figs. 5A, 5B, 5C). This Hg-S distance is slightly longer than the Hg-S bond distance in cinnabar and shorter than that in metacinnabar (2.37 Å and 2.53 Å, respectively) (RAMSDELL, 1925; WYCKOFF, 1963) and thus is a weighted average of the Hg-S distances in cinnabar and metacinnabar. Although there is the possibility of Hg-thiol bonds in the sample, the difference in bond distance between cinnabar and Hg-thiol is ~0.03 Å, which is too small to differentiate with any confidence (HOLLOWAY and MELNIK, 1995; RAMSDELL,

1925).

5. Discussion

A combined analytical, SCE, and spectroscopic approach was used to determine the main Hg-bearing phases present at and transported downstream from the New Idria mercury mine site. The main release of Hg from the site is from the waste piles (both tailings and calcine) instead of the underground mine workings. The main evidence for this assertion is the differences in Hg concentrations and speciation between samples NI-1 (above the main tailings pile) and NI-2 (below the main tailings pile) (2.6 ± 0.2 ppm and 235.3 ± 4.7 ppm, respectively). The New Idria drainage system is separated into two regions: (1) areas with abundant ferrihydrite (Samples NI-1, NI-2, NI-4, and NI-5 – dry season) and (2) areas with no ferrihydrite (Samples NI-3, NI-6, and diatom-rich samples). For all New Idria samples, SCE analyses show that Hg in the sediments is relatively stable in terms of potential for release, with the majority of Hg

98

(>97%) being released by the 1M KOH extraction step (SCF4) or stronger chemical leach (SCF5 and SCF6 steps). This finding suggests that most Hg is in the form of organic-bound Hg(II), elemental Hg, or crystalline HgS.

Sample NI-1 was taken at the entrance to the mine adit, is above the main tailings pile, and has the lowest total Hg concentration, making it an outlier when compared to the other ferrihydrite-rich samples. However, for all ferrihydrite-rich samples, including NI-1 (Samples NI-1, NI-2, NI-4, and NI-5), the HgS polymorphs are the dominant inorganic Hg phases (>40% of the total Hg, Fig. 2), with Hg concentrations ranging from a minimum of 2.6 ppm for sample NI-1 to a maximum of 235.3 ppm for sample NI-2 (wet season). In this part of the drainage system, crystalline HgS particles (both cinnabar and metacinnabar) <10 m in diameter are intermixed with ferrihydrite nanoparticles, as shown by TEM imaging and EXAFS spectroscopy in our previous study (LOWRY et al., 2004) and by -XRF mapping (Supporting Information, Fig. S2) and bulk EXAFS analysis (Fig. 4A and 4B) in the present study.

The majority of the remaining Hg identified in samples taken from the ferrihydrite-rich portion of the system is released in the SCF4 (organo-chelated) and SCF5 (elemental Hg) extraction steps. When these two fractions are combined, they comprise up to 57% of the total Hg in some of the field samples

(e.g., Sample NI-1) (Fig. 2). The concentrations of Hg in the combined

SCF4/SCF5 fractions for samples NI-2, NI-4, and NI-5 are between 2.7 ppm and

3.3 ppm for wet season samples and between 7.2 ppm and 7.5 ppm for dry season samples (Samples NI-4 and NI-5), indicating that Hg concentrations in the

99

SCF4/SCF5 combined fractions increase during the dry season. Kim et al.

(2003) found that quantitative determination of the Hg phases present in the F3 and F4 fractions in the Bloom et al. protocol can be difficult, but SCE analyses of both the Hg-sorbed ferrihydrite (natural and synthetic) and diatom-rich samples, can provide useful insights about the SCF4 and SCF5 fractions of the field samples as described below.

The amount of mercury weakly adsorbed to ferrihydrite or to organics associated with ferrihydrite is below detection using the SCE approach in the natural New Idria drainage system, as indicated by the lack of significant Hg release from ferrihydrite-rich field samples during the SCF2 (MgCl2) extraction step (Fig. 2). In contrast, based on our laboratory uptake experiments, relatively high concentrations of Hg(II) sorb to synthetic 2-line ferrihydrite (5,250 ppm at pH 3.5 and 6,750 ppm at pH 7.45) and natural New Idria ferrihydrite (7,250 ppm at pH 3.5 and pH 7.45) (Table 2). However, mercury did not sorb strongly in these laboratory experiments to either synthetic or natural ferrihydrite; the SCF2

(MgCl2) extraction step resulted in the release of 79 to 82% of the total Hg for

Hg-sorbed natural ferrihydrite and 89 to 92% for Hg-sorbed synthetic 2-line ferrihydrite (Fig. 3A), suggesting that the majority of sorbed Hg(II) is present as weakly bound outer-sphere complexes. The magnesium chloride extraction step

(SCF2) has been used to quantify outer-sphere metal ion sorption complexes for

Cd, Co, Cu, Ni, Pb, Mn, Fe, U, etc. (OSTERGREN et al., 1999; SINGER et al., 2009;

TESSIER et al., 1979), and it is assumed that this extractant also causes desorption of outer-sphere sorbed Hg. This assumption is consistent with our EXAFS results

100 for sample NI-2 (see Figs. 5D, 5E, and 5F), which showed only Hg-S first- neighbor and Hg-Hg second-neighbor pair correlations, characteristic of crystalline HgS. Importantly, we did not detect Hg-O first-neighbor or Hg-Fe second-neighbor pair correlations, which would be expected if Hg(II) were present as inner-sphere sorption complexes on ferrihydrite. Sample NI-2 was our only ferrihydrite-rich sample with sufficient Hg (235 ppm) for EXAFS analysis.

Currently, there are no good chemical extractants for assaying only inner-sphere

Hg(II) sorption complexes in compositionally complex samples like the New Idria ferrihydrites, so we assume that our EXAFS results for Sample NI-2, indicating dominantly HgS, are representative of the other ferrihydrite-rich New Idria drainage samples with much lower Hg concentrations (2.6 to 19.1 ppm for wet- season samples and 13.5 to 27.2 ppm for dry-season samples, Table 2). This assumption is consistent with our SCE results for the ferrihydrite-rich New Idria samples (NI-1, NI-2, NI-4, and NI-5), which show release of Hg only in the relatively harsh SCF4, SCF5, and SCF6 extraction steps.

Earlier uptake studies of Hg(II) on ferrihydrite and other iron-

(oxyhydr)oxides show some pH dependence (DZOMBAK and MOREL, 1990; KIM et al., 2004b; KIM et al., 2004c). However, our mercury uptake experiments on natural New Idria ferrihydrite showed no pH dependence, with similar sorbed Hg concentrations for both pH 3.5 and 7.45, suggesting that Hg is sorbed mainly to a phase other than ferrihydrite, such as associated organic matter or amorphous Al- hydroxide (CISMASU et al., 2011), or is present in a separate phase in this sample

(e.g., nanoparticulate HgS).

101

SCE analyses of the diatom-rich field samples (Fig. 3B) showed that the combined SCF4 (organo-chelated) and SCF5 (elemental Hg) fractions comprise over 60% of the total Hg for all but three of the samples analyzed. However, the

SCF4 and SCF5 fractions vary in their dominance in the diatom-rich field samples (Fig. 3B). Although elemental Hg can be found as mm-sized droplets in mine waste samples and in sediments in stream pools in the upper reaches of the

New Idria drainage system, the percentage of the SCF5 fraction detected is difficult to quantify. Mercury concentrations from the combined SCF4/SCF5 extractions of the New Idria drainage samples increase during the dry season, which suggests that the majority of Hg extracted in the SCF5 fraction is from Hg associated with diatoms. If elemental Hg were the dominant form of Hg in the

SCF5 fraction, concentrations would be expected to be higher in the wet season, as high water flow would transport more of the high density elemental Hg downstream. The higher SCF4/SCF5 Hg concentrations in the dry season suggests that as water flow goes down, more diatoms can grow and adsorb or incorporate Hg(II). These results suggest that significantly more Hg is associated with diatoms than with ferrihydrites in the downstream sediments of the New

Idria drainage system (Fig. 3B).

The SCF6 (HgS/HgSe) fraction for both the New Idria samples containing no ferrihydrite (samples NI-3 and NI-6) and ferrihydrite-rich sediments (samples

NI-1, NI-2, NI-4, and NI-5), shows that both contain crystalline HgS (>40% of the total Hg). The SCF6 fraction of sample NI-6 has Hg concentrations of 2.1 ppm and 16.0 ppm, respectively, for the wet and dry seasons. Moreover, the

102 combined SCF4/SCF5 fractions of sample NI-6 also show a small concentration difference between the wet and dry seasons of 2.7 ppm and 3.9 ppm, respectively.

Sample NI-3 is from San Carlos Creek above its confluence with the New Idria

AMD system. This portion of San Carlos Creek drains areas containing remediated (e.g., the Aurora Mine) and unremediated (Molina Mine) Hg mines up gradient and should not be impacted by the New Idria mine. The F6 fraction for sample NI-3 contained 89% of the total Hg (Fig. 2) in the form of 16.3 ppm

HgS/HgSe, whereas the combined SCF4/SCF5 fractions contained 2.0 ppm organo-chelated and elemental Hg.

The lack of strong Hg association with ferrihydrites in the New Idria drainage system is surprising and contrasts with the strong Hg(II) uptake by ferrihydrite cited by Dzombak and Morel (1990) and found in our laboratory uptake experiments discussed above. This observation suggests that transport of

Hg(II) adsorbed to ferrihydrite nanoparticles is not as important in the New Idria drainage system as suggested by these laboratory uptake studies. Although Hg(II) has been found to bind strongly to goethite as dominantly inner-sphere complexes in laboratory uptake experiments (KIM et al., 2004b; KIM et al., 2004c), EXAFS analysis of sample NI-2 (Figs. 5D, 5E, and 5F) revealed no significant Hg(II) associated with iron-(oxy)hydroxides) in the form of inner-sphere complexes.

Assuming sample NI-2 is representative of the other New Idria drainage samples, then inner-sphere Hg(II) sorption complexes on iron-(oxyhydr)oxides are not significant species in this system. One possible reason for this conclusion is the repulsive interaction between Hg(II) ions and the positively charged iron-

103

(oxyhydr)oxide particle surfaces at the low to near-neutral pH values of this drainage system. The pH point of zero charge (pHpzc) for iron-(oxy)hydroxides ranges from 7.8 to 7.9 for ferrihydrite and from 7.5 to 9.5 for goethite (CORNELL and SCHWERTMANN, 2003).

Another possible reason for the lack of significant association of Hg(II) with ferrihydrites in the New Idria drainage system is the presence of significant concentrations of chloride (4 x 10-4 to 6 x 10-3 M). Kinniburgh and Jackson

(1978) showed that Cl- in solution reduces the adsorption of Hg(II) on hydrous ferric oxide gels from 91.5% to 68% at pH 4.5 because of the formation of

(2-x)+ HgClx aqueous complexes. Their study was carried out with a chloride concentration of 1 x 10-4 M, which is close to that in the New Idria drainage system. Although Kim et al. (2004b; 2004c) showed that Hg(II) adsorbs readily to goethite at low pH and that increasing sulfate concentration over the range 10-5 to

0.9 M enhanced Hg(II) adsorption (New Idria drainage sulfate concentration ~ 27 mM), these studies used Hg concentrations (0.5 mM) greatly exceeding environmental concentrations (0.2 to 2 nM). In addition, the adsorption studies carried out in this earlier work did not investigate the effects of competing

2+ divalent cations such as Mg . The fact that 1M MgCl2 removed the majority of

Hg from the laboratory uptake samples in the present study suggests that Mg2+,

Cl-, or a combination of the two have a major effect on the stability of sorbed

Hg(II).

The present study has confirmed results from our earlier but more spatially limited study of Hg speciation at the New Idria mine site (LOWRY et al., 2004),

104 which showed that the dominant (>75%) form of Hg released from the calcine waste material is colloidal HgS (mainly cinnabar with lesser amounts of metacinnabar) with some colloidal HgO (montroydite) as well. The HgS/HgSe fraction dominates 6 of the 9 samples analyzed (>60% of total) in the present study. The present study revealed for the first time that freshwater diatoms play a significant role in Hg-cycling in this system by removing Hg from solution and adsorbing/incorporating Hg in diatoms. The diatoms in Silver Creek (samples

SC-1, and SC-2) as well as in Harley Gulch (samples HG-2 and HG-4) have SCE release patterns similar to the commercial diatomite samples with little, if any associated organic matter (see section 3.3). Therefore, release of major levels of

Hg (30-60% of the total Hg, Fig. 3B) from the diatom-rich field samples during the 1M KOH extraction step (Fraction SCF4) is not due to release from organic matter. Instead, we suggest that this release is due to dissolution of the diatom silica frustules at this high pH value (14), accompanied by release of incorporated

Hg, possibly present, in part, in the form of Hg-silicate minerals such as edgarbaileyite (Hg6Si2O7), which forms under ambient conditions and was first found at New Idria (ANGEL et al., 1990). This reasoning suggests that not all of the Hg is present as monomethyl Hg (MMeHg) in the cytoplasm of the diatoms.

MMeHg is bioaccumulated in diatoms, with MMeHg bioaccumulation factors of up to 100,000 times the concentration of MMeHg in water (PICKHARDT and

FISHER, 2007). Upon death and accumulation of diatoms in sediment, MMeHg in the diatoms degrades to Hg, which may then react with sulfide generated by sulfate-reducing bacteria to form HgS (see Rytuba et al., 2011). The Hg SCE

105 release patterns also indicate that HgS is consistently present in diatomite and in recent accumulations of diatoms in Silver Creek downstream from the New Idria mine.

Our findings of Hg associated with freshwater diatoms and the lack of significant Hg association with ferrihydrites and other Fe(III)-(oxy)hydroxides in the New Idria AMD system are important for evaluating the natural passive remediation plans proposed by environmental firms at other similar Hg mines throughout the California Coast Range. Though Hg association with the inorganic fraction of diatoms is novel, numerous other metals, including Cd and

Zn, have been shown to be associated with the silica frustules of diatoms

(GELABERT et al., 2006; GELABERT et al., 2007). The present study also shows the importance of the additional MgCl2 extraction step for the SCE analytical protocol used in determining Hg speciation in complex Hg mine waste samples.

If we had not added the MgCl2 extraction step to the SCE sequence, the results would have been interpreted as indicating that Hg(II) sorbs to both synthetic 2- line ferrihydrite and New Idria ferrihydrite. Finally, this study detected few if any

Hg-bearing species, other than elemental mercury, that would be considered potentially bioavailable in the New Idria drainage system.

106

Acknowledgements

We wish to thank Dr. John Bargar, Mr. Joe Rogers, and Dr. Samuel Webb of the

Stanford Synchrotron Radiation Lightsource for technical help on beamlines 11-2 and 2-3 during data collection. We also wish to thank Prof. Chris Kim and members of the Environmental Geochemistry Laboratory at Chapman University for collecting the EXAFS spectra for the diatom-rich samples. The LN2 cryostat used in the low-temperature work was designed and built by Dr. Steve Conradson of Los Alamos National Laboratory, and we thank him for allowing us to use it.

Funding for this research was provided by the Stanford Environmental Molecular

Science Institute through NSF Grant CHE-0431425 and by the NSF Center for

Environmental Implications of Nanotechnology (CEINT) through NSF Grant EF-

0830093. Portions of this research were carried out at the Stanford Synchrotron

Radiation Lightsource, a National user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences, with additional support from the National Institute of Health.

107

References Cited

Aiken G., Haitzer M., Ryan J. N., and Nagy, K., 2003. Interactions Between Dissolved Organic Matter and Mercury in the Florida Everglades. Journal of Physics IV France 107, 29-32. Angel R. J., Cressey G., and Criddle, A., 1990. Edgarbaileyite, Hg6Si2O7: The Crystal Structure of the First Mercury Silicate. American Mineralogist 75, 1192-1196. Benoit J. M., Mason R. P., Gilmour C. C., and Aiken G. R., 2001. Constants for Mercury Binding by Dissolved Organic Matter Isolates from the Florida Everglades. Geochimica et Cosmochimica Acta 65, 4445-4451. Bloom N. S., Preus E., Katon J., and Hiltner M., 2003. Selective Extractions to Assess the Biogeochemically Relevant Fractionation of Inorganic Mercury in Sediments and Soils. Analytica Chimica Acta 479, 233-248. Cismasu A. C., Michel F. M., Tcaciuc A. P., Tyliszczak T., and Brown Jr. G. E., 2011. Composition and Structural Aspects of Naturally Occurring Ferrihydrite. Comptes Rendus Geosceience 343, 210-218. Clarkson T. W. and Magos L., 2006. The Toxicology of Mercury and Its Chemical Compounds. Critical Reviews in Toxicology 36, 609-662. Cornell R. M. and Schwertmann U., 2003. The Iron Oxides. Wiley-VCH GmbH & Co. KGaA. Drexel R. T., Haitzer M., Ryan J. N., Aiken G. R., and Nagy K. L., 2002. Mercury(II) sorption to Two Florida Everglades Peats: Evidence for Strong and Weak Binding and Competition by Dissolved Organic Matter Released from the Peat. Environmental Science & Technology 36, 4058-4064. Dzombak D. A. and Morel F. M. M., 1990. Surface Complexation Modeling: Hydrous Ferric Oxide. Wiley, New York. Ganguli P. M., Mason R. P., Abu-Saba K. E., Anderson R. S., and Flegal A. R., 2000. Mercury Speciation in Drainage from the New Idria Mercury Mine, California. Environmental Science & Technology 34, 4773-4779. Gelabert A., Pokrovsky O. S., Viers J., Schott J., Boudou A., and Feurtet-Mazel A., 2006. Interaction Between Zinc and Freshwater and Marine Diatom Species: Surface Complexation and Zn Isotope Fractionation. Geochimica et Cosmochimica Acta 70, 839-857. Gelabert A., Pokrovsky O. S., Schott,J., Boudou A., and Feurtet-Mazel A., 2007. Cadmium and Lead Interaction with Diatom Surfaces: A Combined Thermodynamic and Kinetic Approach. Geochimica et Cosmochimica Acta 71, 3698-3716. Haitzer M., Aiken G., and Ryan J. N., 2002. Binding of Mercury(II) to Dissolved Organic Matter: The Role of Mercury-to-DOM Concentration Ratio. Environmental Science & Technology 36, 3564-3570. Haitzer M., Aiken G. R., and Ryan J. N., 2003. Binding of Mercury(II) to Aquatic Humic Substances: Influence of pH and Source of Humic Substances. Environmental Science & Technology 37, 2436-2441. Holloway C. E. and Melnik M., 1995. Mercury Organometallic Compounds-Classification and Analysis of Crytallographic and Structural Data. Journal of Organometallic Chemistry 495, 1-31. Jew A. D., Kim C. S., Rytuba J. J., Gustin M. S., and Brown Jr. G. E., 2011. A New Technique for Quantification of Elemental Hg in Mine Wastes and Its Implications for Mercury Evasion Into the Atmosphere. Environmental Science & Technology 45, 412-417.

108

Kim C. S., Brown Jr. G. E., and Rytuba J. J., 2000. Characterization and Speciation of Mercury-Bearing Mine Wastes Using X-ray Absorption Spectroscopy. The Science of the Total Environment 261, 157-168. Kim C. S., Bloom N. S., Rytuba J. J., and Brown Jr. G. E., 2003. Mercury Speciation by X- ray Absorption Fine Structure Spectroscopy and Sequential Chemical Extractions: A Comparison of Speciation Methods. Environmental Science & Technology 37, 5102-5108. Kim C. S., Rytuba J. J., and Brown Jr. G. E., 2004a. Geological and Anthropogenic Factors Influencing Mercury Speciation in Mine Wastes: An EXAFS Spectroscopy Study. Applied Geochemistry 19, 379-393. Kim C. S., Ruytuba J. J., and Brown Jr. G. E., 2004b. EXAFS study of mercury(II) sorption to Fe- and Al-(hydr)oxides I: Effects of pH. Journal of Colloid and Interface Science 271, 1-15. Kim C. S., Rytuba J. J., and Brown Jr. G. E., 2004c. EXAFS study of mercury(II) sorption to Fe- and Al-(hydr)oxides II: Effects of chloride and sulfate. Journal of Colloid and Interface Science 270, 9-20. Kinniburgh D. G. and Jackson M. L., 1978. Adsorption of Mercury(II) by Iron Hydrous Oxide Gel. Soil Science Society of America Journal 42, 45-47. Krauskopf K. B. and Bird D. K., 1995. Introduction To Geochemistry. WCB/McGraw-Hill. Linn R. K., 1968. New Idria Mining District. In: Ridge, J. D. (Ed.), Ore Deposits of the United States, 1933-1967. The American Institute of Mining, Metallurgical, and Peroleum Engineers, Inc., New York. Lowry G. V., Shaw S., Kim C. S., Rytuba J. J., and Brown Jr. G. E., 2004. Macroscopic and Microscopic Observations of Particle-Facilitated Mercury Transport from New Idria and Sulphur Bank Mercury Mine Tailings. Environmental Science & Technology 38, 5101-5111. Materials Data Inc., 2002. Jade XRD Pattern Processing Ver. 6.5. Morel F. M. M., Kraepiel A. M. L., and Amyot M., 1998. The Chemical Cycle and Bioaccumulation of Mercury. Annual Review of Ecological Systems 29, 543-566. Ostergren J. D., Brown Jr. G. E., Parks G. A., and Tingle T. N., 1999. Quantitative Speciation of Lead in Selected Mine Tailings from Leadville, CO. Environmental Science & Technology 33, 1627-1636. Pickhardt P. C. and Fisher N. S., 2007. Accumulation of Inorganic and Methylmercury by Freshwater Phytoplankton in Two Contrasting Water Bodies. Environmental Science & Technology 41, 125-131. Ramsdell L. S., 1925. The Crystal Structures of Some Metallic Sulfides. American Mineralogist 10. Ravichandran M., 2004. Interactions Between Mercury and Dissolved Organic Matter-A Review. Chemosphere 55, 319-331. Rytuba J. J., 2003. Mercury from Mineral Deposits and Potential Environmental Impact. Environmental Geology 43, 326-338. Rytuba J. J., Hothem R. L., Brussee B. E., and Goldstein D. N., 2011. Impact of Mine and Natural Sources of Mercury on Water, Sediment, and Biota in Harley Gulch Adjacent to the Abbott-Turkey Run Mine, Lake County, California: U.S. Geological Survey Open File Report 2011-1265. United States Geological Survey. Singer D. M., Maher K., and Brown Jr. G. E., 2009. Uranyl-chlorite Sorption/desorption: Evaluation of Different U(VI) Sequestration Processes. Geochimica et Cosmochimica Acta 73, 5989-6007.

109

Tessier A., Campbell P. G. C., and Bisson M., 1979. Sequential Extraction Procedure for the Speciation of Particulate Trace Metals. Analytical Chemistry 51, 844-851. United States Environmental Protection Agency, 2002. Method 1631, Revision E: Mercury in Water by Oxidation, Purge and Trap, and Cold Vapor Atomic Fluorescence Spectrometry. Washington D.C. Webb S. M., 2005. SIXPack: a graphical user interface for XAS analysis using IFEFFIT. Physica Scripta T115, 1011-1014. Webb S., 2006. SixPACK. Stanford Synchrotron Radiation Laboratory, Menlo Park. Wyckoff R. W. G., 1963. Crystal Structures 1, 2nd Edition. Interscience New York.

110

Table 1: Sample descriptions, ages, and bulk mineralogy

Sample Bulk Mineralogy (Wet Bulk Mineralogy (Dry Sample Sample Type Age Season) Season)

New Idria AMD, taken 1 m from New Quartz, Goethite, 2-line NI-1 Recent N/A Idria Mine adit, Portal #10 Ferrihydrite

New Idria AMD, taken 10 m from the Quartz, Goethite, NI-2 base of the tailings pile in AMD Recent N/A Jarosite, Kaolinite drainage San Carlos Creek, taken 100 m up Quartz, Plagioclase, NI-3 stream of San Carlos Creek/ New Idria Recent N/A Orthoclase, Chrysotile AMD confluence San Carlos Creek, taken ~1 km Quartz, Plagioclase, Quartz, 2-line NI-4 downstream of San Carlos Creek/New Recent Orthoclase, Chrysotile, Ferrihydrite, Idria AMD confluence Kaolinite, Calcite Aragonite San Carlos Creek, taken ~1 km Quartz, Plagioclase, NI-4 downstream of San Carlos Creek/New Recent Orthoclase, Chrysotile, N/A Duplicate Idria AMD confluence Kaolinite, Calcite San Carlos Creek, taken ~4 km Quartz, 2-line Quartz, Plagioclase, NI-5 downstream of San Carlos Creek/New Recent Ferrihydrite, Chrysotile, Calcite Idria confluence Aragonite San Carlos Creek, taken ~4 km Quartz, 2-line NI-5 downstream of San Carlos Creek/New Recent N/A Ferrihydrite, Duplicate Idria confluence Aragonite Quartz, Plagioclase, San Carlos/Silver Creek, taken ~6.5 km Quartz, Plagioclase, Orthoclase, NI-6 downstream of San Carlos Creek/New Recent Orthoclase, Chrysotile, Chrysotile, Calcite, Idria confluence Calcite Aragonite Late Opal-A, Opal-CT, D-1 Diatomite, Petaluma Formation N/A Pliocene Quartz Late D-2 Diatomite, Petaluma Formation Opal-A, Quartz, Albite N/A Pliocene Late Opal-A, Opal-CT, D-3 Diatomite, Petaluma Formation N/A Pliocene Quartz Wolf Creek Ranch Food Grade D-4 Pliocene Opal-A, Opal-CT N/A Diatomaceous Earth Safer Brand Ant and Crawling Insect D-5 Pliocene Opal-A, Opal-CT N/A Killer Opal-A, Opal-CT, D-6 Blue Ribbon DE Premium Cat Litter Miocene N/A Quartz D-7 Concern DE Crawling Insect Killer Pliocene Opal-A, Opal-CT N/A D-8 Pooltime DE Filter Powder Pliocene Cristobalite N/A Quartz, Orthoclase, SC-1 Silver Creek, diatom-rich sediments Recent Plagioclase, Kaolinite, N/A Aragonite Quartz, Orthoclase, SC-2 Silver Creek, diatom-rich sediments Recent Plagioclase, Kaolinite, N/A Aragonite Quartz, Plagioclase, HG-1 Harley Gulch, diatom-rich sediments Recent N/A Aragonite Quartz, Plagioclase, HG-2 Harley Gulch, diatom-rich sediments Recent N/A Aragonite Quartz, Plagioclase, HG-3 Harley Gulch, diatom-rich sediments Recent N/A Aragonite Quartz, Orthoclase, HG-4 Harley Gulch, diatom-rich sediments Recent N/A Plagioclase, Aragonite

111

Table 2: Hg concentrations and SCE recoveries for samples used in SCE experiments. SCE % SCE % Water [Hgt] (ppm) Recovery Wet [Hgt] (ppm) Recovery Sample pH Wet Season Season Dry Season Dry Season NI-1 4.3 2.6 ± 0.2 102.4 N/A N/A NI-2 3.7 235.3 ± 4.7 106.2 N/A N/A NI-3 8.3 18.4 ± 1.1 89.3 N/A N/A NI-4 7.7 10.3 ± 0.7 98.6 27.2 ± 0.1 97.1 NI-4 Duplicate 7.7 10.3 ± 0.7 100 27.2 ± 0.1 N/A NI-5 7.8 19.1 ± 0.9 98.7 13.5 ± 0.1 95.5 NI-5 Duplicate 7.8 19.1 ± 0.09 N/A 13.5 ± 0.1 97 NI-6 8 4.8 ± 0.4 103.6 19.8 ± 0.4 96.5 D-1 N/A 0.35 ± 0.01 N/A N/A N/A D-2 N/A 0.50 ± 0.02 N/A N/A N/A D-3 N/A 1.32 ± 0.03 N/A N/A N/A D-4 N/A 0.05 ± 0.01 N/A N/A N/A D-5 N/A 0.69 ± 0.01 N/A N/A N/A D-6 N/A 0.07 ± 0.01 N/A N/A N/A D-7 N/A 0.06 ± 0.01 N/A N/A N/A D-8 N/A 0.06 ± 0.01 N/A N/A N/A SC-1 N/A 7.77 ± 0.4 N/A N/A N/A SC-2 N/A 8.34 ± 0.31 N/A N/A N/A HG-1 N/A 13.00 ± 0.57 N/A N/A N/A HG-2 N/A 13.92 ± 0.55 N/A N/A N/A HG-2 Duplicate N/A 13.58 ± 0.46 N/A N/A N/A HG-3 N/A 9.80 ± 0.27 N/A N/A N/A HG-4 N/A 10.08 ± 0.01 N/A N/A N/A Fsyn 7.45 7.45 6.75 ± 0.03 mg/g 97 N/A N/A NIF 7.45 7.45 7.25 ± 0.02 mg/g 99 N/A N/A Fsyn 3.5 3.5 5.25 ± 0.01 mg/g 97.5 N/A N/A NIF 3.5 3.5 7.25 ± 0.04 mg/g 100.8 N/A N/A

112

A B

2 m

C D

50 m 10 m Figure 1: A) Photograph of site locality NI-4 of the New Idria drainage system. B) SEM images of Sample NI-4 showing a diatom intermixed with ferrihydrite particles. C and D) SEM images of diatom-rich sediments taken from Silver Creek approximately 20 km downstream of the New Idria Mine.

113

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% NI-1 NI-2 NI-3 NI-4 NI-4 NI-5 NI-5 NI-6 NI-6 (2.6 (235.3 (18.4 (10.3 (28.0 (19.1 (13.5 (4.8 (19.8 ppm) ppm) ppm) ppm) ppm) Dry ppm) ppm) Dry ppm) ppm) Dry Wet Wet Wet Wet Wet Wet % SCF6 (Aqua Regia, HgS/HgSe) % SCF5 (12M Nitric Acid, elemental Hg) % SCF4 (1M Potassium Hydroxide, organo-chelated) %SC F3 (0.1M Acetic Acid + 0.01M Hydrochloric Acid, stomach acid soluble) % SCF2 (1M Magnesium Chloride, outer-sphere complex) % SCF1 (Distilled Water, water soluble)

Figure 2: Sequential chemical extraction results for New Idria AMD sediment samples. Samples taken during the dry season are noted as ‘Dry’ and samples taken during the wet season are denoted as ‘Wet’. Total mercury concentrations are in brackets. Errors for all Fractions that are > 1% of the total are ≤ 8% for all samples. Fractions below 1% of the total are < 15% for all samples.

114

100% A 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Fsyn NIF Fsyn NIF Fsyn NIF Fsyn NIF pH 7.45 pH 7.45 pH 3.5 pH 3.5 pH 7.45 pH 7.45 pH 3.5 pH 3.5 (6.8 mg/g) (7.3 mg/g) (5.3 mg/g) (7.3 mg/g) no F1 no F1 no F1 no F1 (6.8 mg/g) (7.3 mg/g) (5.3 mg/g) (7.3 mg/g)

100% B 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

Figure 3: Sequential chemical extraction results for Ferrihydrite (synthetic and natural) and diatom-rich samples. A) SCE analysis of ferrihydrite both with and without F1 fraction plotted, Fsyn denotes synthesized 2-line ferrihydrite and NIF denotes New Idria samples. Samples plotted with F1 fraction added (left 4 samples) and without (right 4 samples). B) SCE results of diatom-rich samples. Mercury concentrations are in brackets. Errors for all Fractions are ≤ 7% for all samples.

115

4 A Data Fit 3

2

1

chi(k)

3 0 k

-1

-2

-3 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 -1 k (Å )

4 Data B Fit 3

2

1

0

chi(k)

3

k -1

-2

-3

-4 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 -1 k (Å )

3 Figure 4: k -weighted bulk Hg LIII-edge EXAFS spectra of orange AMD pond sediment and Sample NI-2. -XRF and -XAFS results for the same samples are in Supporting Information. Figure A is the AMD pond sediment sample consisting of 81% cinnabar and 19% metacinnabar (residual 0.335). Figure B is Sample NI-2 consisting of 76% cinnabar and 24% metacinnabar (residual 0.395). The difference in EXAFS oscillation position between LCF and shell-by-shell fitting (Fig. 5) in k-space is due to a different

E0 being used in background subtraction for spline removal.

116

5 A Data B 4.5 4 Data 3 Fit

3.5 Fit

3 ) 1 k 2.5 2 -1

chi( 1.5

3 Amplitude 1 k -3 0.5 0 -5 0.5 1.5 2.5 3.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 R + Δ Å k (Å-1)

C

Pathway CN R (Å) σ2

Hg-S 3.50 ± 0.43 2.43 ± 0.01 0.0117 ± 0.0015

2 Δ Eo 5.35 ± 1.22 Red-chi 5.33

2 So 0.90 R-factor 0.135

4.5 D 4 E Data 4 Data 3 Fit 3.5 Fit

2

)

k 3 1

2.5 chi(

0

3 2 k

-1 1.5 Amplitude -2 1 -3 0.5 -4 0 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 0.5 1.5 2.5 3.5 -1 k (Å ) R + Δ Å F

Pathway CN R (Å)  Hg-S 2.0 ± 0.05 2.38 ± 0.01 0.006 ± 0.001 Hg-S 1.5 ± 0.35 3.17 ± 0.02 0.022 ± 0.004 Hg-Hg 2.0 ± 0.80 3.8 ± 0.05 0.020 ± 0.005 2  Eo 5.08 ± 0.28 Red-chi 0.63 2 So 0.90 R-factor 0.007

Figure 5: Shell-by-shell fittings of Hg LIII-edge EXAFS spectrum for Sample NI-2 and a diatom- 3 rich sample from Harley Gulch using theoretical pathways generated by FEFF 8.2: (A) fit of k - weighted EXAFS of a diatom-rich Harley Gulch sample, (B) fit of Fourier Transform of diatom- rich Harley Gulch sample, and (C) table showing fitting parameters for diatom-rich Harley Gulch sample, (D) fit of k3-weighted EXAFS of Sample NI-2, (E) fit of Fourier Transform of Sample NI-

2, (F) table showing fitting parameters for Sample NI-2. The difference in EXAFS oscillation position between LCF (Fig. 4) and shell-by-shell fitting in k-space is due to a different E0 being used in background subtraction for spline removal.

117

Supporting Information

A Sequential Chemical Extraction and Spectroscopic Assessment

of the Potential Bioavailability of Mercury Released From the

Inoperative New Idria Mercury Mine, San Benito Co., CA

ADAM D. JEW, PHI N.M. LUONG, JAMES J. RYTUBA, AND GORDON E. BROWN, JR.

118

2-line Ferrihydrite Synthesis, Characterization

2-line ferrihydrite was synthesized following the protocol outlined by

Schwertmann and Cornell (SCHWERTMANN and CORNELL, 1991), in which 570

. mL of 1 M NaOH is rapidly added to 1 L of 0.2M Fe(NO3)3 9H2O and the pH is adjusted to 7.5 ± 0.2. The precipitate was aged overnight. The samples were triply washed by centrifugation (RCFmax = 1400g for 15 minutes) and re- suspension using DDI water. Samples were dried in a covered dessicator at room temperature under a light vacuum for three days. After drying, the synthesized 2- line was ground with an agate mortar and pestle. Portions of synthesized 2-line ferrihydrite were analyzed for phase purity using XRD, surface area (BET analysis), and particle size (transmission electron microscopy). Adsorption of

Hg(II) onto ferrihydrite was done using synthetic 2-line ferrihydrite and a natural sample containing ferrihydrite/schwertmannite taken where the low pH AMD water mixes with the higher pH waters of San Carlos Creek (downstream of samples NI-2 and NI-3) (Fig. S1).

XRF and -XAFS

X-ray fluorescence (XRF) microprobe analysis was carried out on a sediment core extracted from the New Idria AMD settling pond in November 2008. This analysis was designed to determine (1) if the cyclical orange/gray layering present in the AMD settling pond sediments is a compositional change or a redox boundary, and (2) if there was any preferential separation of Hg within the layers.

A 1 m push core was taken, sealed immediately, frozen in the field with liquid

119 nitrogen, and kept frozen with dry ice. Upon returning to Stanford University (6-

o 8 hours later), the core was sectioned in a -20 C cold room 2 cm above and below the orange/gray sediment boundaries. After sectioning, samples for -XRF and micro-x-ray absorption fine structure (-XAFS) were prepared by fixing sections with epoxy and cutting to a thickness of ~0.5 mm perpendicular to the sediment layering. A subset of these layers, prior to epoxy impregnation, was taken for bulk extended x-ray absorption fine structure (EXAFS) spectroscopy).

X-ray fluorescence microprobe and μ-XAFS analyses of core samples were performed at SSRL on bending magnet beamline 2-3 using a LN2-cooled Si

o (220) monochromator in the phi = 90 orientation. Beamline 2-3 has a spot size of 2μm x 2μm and is equipped with a 3-element Ge detector for both XRF mapping and μ-XAFS spectroscopy, plus a charge-coupled device detector for μ-

XRD. Individual Hg-bearing particles ≤ 2 μm were not detectable by XRF or μ-

XAFS, making individual Hg-bearing colloids/nanoparticles undetectable using this approach. Besides particle size limitations, the XRF technique has a detection limit of ~50 ppm for the 2 m x 2 m spot size. Sample thicknesses were ~0.5 mm, making core samples too thick for collection of μ-XRD by Laue diffraction. XRF maps were collected with a beam size of 2.5 μm x 2.5 μm at an energy of 14 keV using the 3-element Ge detector. Analysis of XRF data was accomplished by using the SMAK x-ray microprobe analysis software package

(WEBB, 2006b). Areas of the XRF maps with high Hg concentrations were selected for μ- XAFS analysis. Mercury LIII-edge - XAFS spectra were collected over an energy range of 200 eV below to 200 eV above the absorption

120 edge. - XAFS data averaging, background subtraction, and data fitting were carried out using the SixPACK XAS analysis software package (WEBB, 2006a;

WEBB, 2005). Mercury phase determination was completed using linear combination fitting. Shell-by-shell fitting of the diatom-rich sediment was completed using SixPACK with theoretically derived scattering pathways using

FEFF 6L (ANKUDINOV et al., 1998; WEBB, 2006a; WEBB, 2005).

A XRF map for the Fe K1 encompassing the orange/gray transition is shown in Fig. S2A. The XRF map shows a sudden drop in the Fe concentration across the boundary indicating that the boundary is compositional in nature and not a redox boundary. The XRF map for the Hg L1 emission for the same region

(Fig. S2B) shows no correlation between Fe and Hg, indicating that Hg is not preferentially distributed between the layers. XRF maps of Hg hot spots (Figs.

S2C and S2D) show that the detectable Hg occurs as small colloids from 2-10

m in diameter. More evidence for the lack of a correlation between Fe and Hg in the samples is seen in a correlation plot of Fe vs. Hg (Fig. S2E) for the area represented in Fig. S2A.

The result of - XAFS LCF is that the Hg colloids are 100% cinnabar

(Fig. S3). These results are different than the bulk EXAFS analysis (manuscript

Fig. 4A and 4B), which show that 76-81% of the HgS is cinnabar and that 19-

24% is metacinnabar. This discrepancy is probably due to two factors: (1) the size of metacinnabar crystals is too small for - XAFS analysis and/or (2) metacinnabar is present as a thin coating on cinnabar crystals that is not detectable

121 with - XAFS. The sediments in the New Idria settling pond have a cyclical layering that is the result of deposition of different materials during the dry summer season and the wet winter season. The orange layer is dominated by ferrihydrite and schwertmannite that actively precipitate during the summer. The gray-colored sediments are Fe-deficient silt from the tailings pile deposited during the rainy winter months. Mercury in the settling pond is in the form of HgS colloids > 10m diameter and are not associated with either sediment type.

122

References Cited

Ankudinov A. L., Ravel B., Rehr J. J., and Conradson S. D., 1998. Real-space Multiple- scattering Calculation and Interpretation of X-ray Absorption Near-edge Structure. Physica Review B 58, 7565-7576. Schwertmann U. and Cornell R., 1991. Iron Oxides in the Laboratory: Preparation and Characterization. VCH Publishers. Webb S. M., 2005. SIXPack: a graphical user interface for XAS analysis using IFEFFIT. Physica Scripta T115, 1011-1014. Webb S., 2006a. SMAK: Sam's Microprobe Analysis Kit Ver. 0.50. Webb S., 2006b. SixPACK. Stanford Synchrotron Radiation Laboratory, Menlo Park.

123

N N

Figure S1: Sample locations at the inoperative New Idria mine and along San

Carlos Creek.

124

A Orange Layer C Hg Map D Hg Map

  Gray Layer E Fe Map Hole 40 m Orange Layer B

Gray Layer

Hg Map Hole 40 m

Figure S2: X-ray fluorescence maps of AMD settling pond sediments across an orange/gray sediment boundary. (A) Fe K 1 XRF map with line denoting boundaries between orange layer, gray layer, and hole in the sample. (B) Hg L1 XRF map of the same region in Fig. S3A showing no correlation between Fe and Hg concentrations. (C) Blow up of box in Fig. S3A showing Hg XRF map. (D) Close up of Hg colloid in the box of Fig. S3C showing a typical sized HgS colloidal particle ~ 4μm x 6 μm in size. (E) Correlation plot of Fe and Hg showing no correlation in concentration (n = 32,000 pixels)



125

1.2 Data

1 Fit

0.8

0.6

0.4

Normalized Absorption Normalized 0.2

0 12200 12250 12300 12350 12400 12450 Energy (eV)

Figure S3: Hg LIII-edge μ-XANES data and fit of colloidal particle from Figure S2D. The particle is 100% cinnabar and is representative of all colloidal particles analyzed using μ-XANES.

126

Chapter 4

A New Technique for Quantification of Elemental

Hg in Mine Wastes and Its Implications for

Mercury Evasion Into the Atmosphere

ADAM D. JEW1*, CHRISTOPHER S. KIM2, JAMES J. RYTUBA3, MAE S. GUSTIN4, AND GORDON E. BROWN, JR.1,5

1 Surface and Aqueous Geochemistry Group, Department of Geological & Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA

2 Environmental Geochemistry Lab, Department of Chemistry, Chapman University, One University Drive, Orange, CA 92866, USA

3 USGS Mineral Resources Program, U.S. Geological Survey, 345 Middlefield Road, MS 901, Menlo Park, CA 94025, USA

4 Department of Natural Resources and Environmental Science, University of Nevada, Reno, NV 89557, USA

5 Department of Photon Science and Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, 2545 Sand Hill Road, MS 69, Menlo Park, CA 94025, USA

Published in Environmental Science & Technology, 2011 Vol. 45, No. 2 (412-417)

127

Abstract: Mercury in the environment is of prime concern to both ecosystem and human health. Determination of the molecular-level speciation of Hg in soils and mine wastes is important for understanding its sequestration, mobility, and availability for methylation. Extended x-ray absorption fine structure (EXAFS) spectroscopy carried out under ambient P-T conditions has been used in a number of past studies to determine Hg speciation in complex mine wastes and associated soils. However, this approach cannot detect elemental (liquid) mercury in Hg- polluted soils and sediments due to the significant structural disorder of liquid Hg at ambient-temperature. A new sample preparation protocol involving slow cooling through the crystallization temperature of Hg(0) (234K) results in its transformation to crystalline -Hg(0). The presence and proportion of Hg(0), relative to other crystalline Hg-bearing phases, in samples prepared in this way can be quantified by low-temperature (77K) EXAFS spectroscopy. Using this approach, we have determined the relative concentrations of liquid Hg(0) in Hg mine wastes from several sites in the California Coast Range and have found that they correlate well with measured fluxes of gaseous Hg released during light and dark exposure of the same samples, with higher evasion ratios from samples containing higher concentrations of liquid Hg(0). Two different linear relationships are observed in plots of the ratio of Hg emission under light and dark conditions vs. % Hg(0), corresponding to silica-carbonate- and hot springs-type

Hg deposits, with the hot springs-type samples exhibiting higher evasion fluxes than silica-carbonate type samples at similar Hg(0) concentrations. Our findings

128 help explain significant differences in Hg evasion data for different mine sites in the California Coast Range.

Introduction

Mercury is a widespread global pollutant that is highly toxic in certain forms to both humans and aquatic ecosystems (1). Significant sources of Hg impacting ecosystems in California include areas containing inoperative mercury mines, such as New Almaden and New Idria, as well as several thousand smaller Hg mines occurring throughout the California Coast Range (2) and placer Au deposits in the Sierra Nevada foothills and the Klamath-Trinity Mountains of

California (3). The potential bioavailability and toxicity of mercury vary dramatically depending on its speciation (4), and thus it is important to have detailed knowledge of Hg speciation in Hg-polluted sediments and mine wastes in order to determine a site’s potential environmental hazard.

Common methods for determining Hg speciation in contaminated materials include sequential chemical extractions (SCE) (5), pyrolysis measurements (6), extended x-ray absorption fine structure (EXAFS) spectroscopy (7-9), and transmission electron microscopy (TEM) coupled with energy dispersive analysis (10, 11). SCEs and pyrolysis are both indirect methods that are destructive to the samples and subject to varying interpretation. Although

SCEs are widely used to estimate Hg speciation in contaminated soil and sediment samples (e.g., (5)), SCEs can cause new Hg species to form as a result of the chemical treatments (10). Pyrolysis causes thermal desorption of Hg with

129 increasing temperature, and the amount released at different temperatures can be measured (6, 12-14). Two problems in using pyrolysis for Hg speciation are that different Hg species can have overlapping release temperatures making quantification difficult, and it is destructive to the sample. In contrast, EXAFS spectroscopy is a direct, non-destructive method for determining Hg speciation, when Hg concentrations are above 50 ppm, but up to this point, it has not been capable of detecting liquid Hg(0) in complex Hg-polluted samples. TEM analysis requires an ultra-high vacuum sample environment, which can result in dehydration of the sample and loss of volatile Hg species, such as Hg(0). In addition, sampling is statistically limited during TEM analysis because of the relatively small amount of sample examined.

Ambient-temperature EXAFS studies of Hg mine wastes and associated sediments have shown that Hg is primarily in the form of cinnabar (-HgS) or metacinnabar (-HgS) with minor proportions of highly soluble Hg salts (7, 15).

Other EXAFS studies of Hg-impacted gold placer mining deposits have shown that even though liquid Hg(0) was used during the recovery of fine-grained Au, mainly colloidal HgS was detected and presumably formed during Hg transport through watersheds (16). However, because Hg is a liquid and thus disordered at ambient- temperature and pressure, these past EXAFS studies were unable to detect liquid

Hg(0) in Hg-impacted mine wastes and sediments. Here we report a new protocol for converting liquid Hg(0) in such samples into a crystalline form (-Hg) by slow

o cooling through its crystallization temperature of -38.83 C (234K). We show that it is possible to determine quantitatively the percentage of liquid Hg(0) in complex

130

Hg-polluted samples prepared by this protocol using low-temperature (77K)

EXAFS spectroscopy. Based on past experience, we estimate that a Hg concentration ≥ 50 ppm is required for detecting liquid Hg(0) using this approach in Hg mine wastes and associated Hg-polluted sediments and soils.

Studies of Hg evasion from Hg-impacted sediments and Hg mine wastes have shown significant differences in the amount of gaseous Hg released from a given sample when it is exposed to light versus dark conditions (17). Gustin et al. hypothesized that photoreduction of Hg(II) in Hg-containing sulfide, chloride, and oxide phases to elemental Hg causes these differences, with samples containing corderoite (Hg3S2Cl2), metacinnabar (-HgS), and Hg bound to organic and inorganic phases exhibiting higher light-enhanced emissions relative to samples containing predominantly cinnabar (17). What was not known in this earlier study, however, is that significant quantities of Hg(0) were present in some of the samples studied (e.g., from the Sulphur Bank Mercury Mine, Clear Lake, CA) which cannot be accounted for by photoreduction alone. Here we show that the concentrations of liquid Hg(0) in Hg-polluted soils, sediments, and mine wastes from the California Coast Range determined using this new sample preparation protocol and low-temperature EXAFS spectroscopy can help explain light- enhanced emission of Hg as measured on similar samples under controlled light and dark conditions.

131

Experimental Section

Samples and Mercury Concentration Measurements. Samples of mine waste and calcine material from the Knoxville, New Almaden, New Idria, and Sulphur

Bank mines (all located in the California Coast Range) were collected and separated, using stainless steel sieves, into a number of different size fractions in the laboratory, from which 500-2000µm, 75-125µm, and <45µm diameter size fractions were selected for additional study. The selected size fractions were split into three sets for (1) analysis of total Hg concentrations, (2) Hg volatilization experiments, and (3) ambient- and low-temperature EXAFS studies. One split was sent to ChemEx Laboratories (Sparks, NV) to determine total Hg concentration by aqua regia digestion and cold vapor atomic fluorescence spectrometry (CVAFS) following US EPA method 1631, which has estimated error limits of 5-10% (18). Another split was used in Hg volatilization experiments under controlled light-dark conditions at the University of Nevada-

Reno. Hg flux was measured for waste materials using a single pass gas

o exchange system at 35 C, a Tekran® 2537A CVAFS, and a Tekran® automated dual sampling unit (17). Samples were allowed to equilibrate in the chamber 24 h before data collection. Hg flux was calculated using the following equation:

F = Q (Co – Ci)/A

2 Where F is the total flux in ng Hg/m h; Co and Ci are the Hg concentration of outlet and inlet air in ng Hg/m3, respectively; A is the surface area of substrate in m2; and Q is the flow of air through chamber in m3/h (17). The third split was

132 used for low temperature EXAFS spectroscopy analysis employing the new sample preparation protocol described below. Following the initial ambient- temperature EXAFS study on the field samples by Kim et al. (7), they were stored in the dark at constant temperature (25°C) in the original sample holders until the new slow cooling sample preparation protocol and low-temperature EXAFS study were carried out four years later. Such sample storage conditions should have minimized any changes in the Hg phases present in the samples over time.

EXAFS Methods. Low-temperature EXAFS data were collected at the

Stanford Synchrotron Radiation Lightsource (SSRL) on wiggler beamline 11-2

o using a LN2-cooled Si (220) monochromator in the phi=90 orientation. Spectra for all field samples and -Hg(0) (due to sample thickness) were collected in fluorescence mode using a 30-element Ge detector while spectra for all other Hg reference materials were collected in transmission mode using three ion chambers. A HgCl2 reference foil was placed between the second and third ion chamber for continuous energy calibration. When α-Hg(0) EXAFS spectra were collected, the SSRL storage ring was at an energy of 3.0 GeV and a current of ~90 mA, with an x-ray beam size of 1mm x 1mm. An aluminum filter was placed between the sample and the Ge detector to minimize the fluorescence signal from lighter elements into the detector. The incident x-ray beam was detuned by 30% to reduce higher harmonics. An aluminum cold finger LN2 cryostat was used to slowly freeze samples for the low-temperature (77K) EXAFS studies. An -

Hg(0) reference sample was prepared by very slow cooling of a liquid Hg(0)

o sample through its crystallization temperature of -38.83 C in a 1.5mm thick

133

® ® Teflon holder sealed with Kapton tape. Plunging the sample into a LN2 bath, as is typically done prior to inserting a sample into a liquid-He cryostat for

EXAFS work at 10K, results in production of amorphous solid Hg(0), which has an EXAFS spectrum similar to that of Hg(0) in the liquid form. To ensure that α-

Hg(0) was produced a combination of pre-cooling the cryostat (Described Below) and a Teflon® sample holder was used. A Teflon® sample holder was selected because liquid Hg(0) will not react with the Teflon® and it has a low thermal conductivity of 0.221W/mK (19).

Before EXAFS data collection for the -Hg(0) reference sample, the cryostat was first pre-cooled so that the reference sample could slowly cool and begin crystallizing while the sample chamber was evacuated. Cooling was accomplished by filling the empty cryostat with LN2 to cool the cold finger to

77K. Once this was achieved, the LN2 was removed, and the cold finger block was heated with a heat gun until the ice build up on the Al block of the cold finger had just started melting. After wiping away the accumulated water, the liquid

Hg(0) sample was attached to the Al block with Kapton® tape, then the sample chamber was immediately evacuated for 20 minutes. Because the empty LN2 reservoir of the cryostat was still near 77K, the sample cooled much more slowly than if the reservoir had been filled with LN2, resulting in crystallization instead of quenching to an amorphous solid. Once the sample chamber was fully evacuated, LN2 was slowly added by pouring ~100mL of LN2 into the reservoir at

5 minute intervals for a half hour then finally filling the reservoir completely with

LN2.

134

Previous attempts at freezing the sample faster resulted in the production of Hg(0) glass that had an EXAFS spectrum very similar to that of liquid Hg(0) at ambient temperature, which is characterized by low amplitude oscillations. The spectra of slowly cooled (described above) and faster cooled of both reference materials (excluding -Hg(0)) and field samples showed no difference in the spectra making slow cooling only necessary for -Hg(0). High purity -HgS, -

HgS, HgO, HgCl2, HgSO4, and HgSe reference materials were purchased from

Alfa Aesar’s Puratronic chemical line, with a reported purity of ≥99.99% (trace metal basis). All solid Hg reference materials were diluted with boron nitride

(BN) and placed in Al sample holders sealed with Kapton® tape. EXAFS spectra of mercury reference materials were collected at both 77K and 298K to determine if the reference materials were contaminated with liquid Hg(0). Self-absorption corrections were carried out on the -Hg(0) spectra using algorithms from both

Athena and GR-xafsX software packages (20, 21), and self-absorption effects were found to be minimal. Three individual EXAFS scans were averaged for all model compounds except for -Hg(0), in which case 12 scans were averaged.

The number of individual EXAFS scans averaged for the mine waste samples ranged between 6 and 18, depending on Hg concentration.

EXAFS Data Analysis. Data averaging, background subtraction, and data fitting were carried out using the SixPACK EXAFS analysis software package

(22, 23). A slightly different E0 for the start of the spline was used in linear combination fitting compared to shell-by-shell fitting accounting for a shift in the data in k-space as seen when comparing spectra from Fig. S5 and all other

135

EXAFS figures. Decomposition of Hg LIII-edge EXAFS data from field samples was accomplished by linear combination least squares fitting over a k-range of 3-

9.5 Å-1 using a library of EXAFS spectra taken at 298K for the reference materials mentioned above. The use of 298K spectra was because of the presence of -Hg(0) in the HgS reference materials (Discussed Below and in Supplemental

Material). This approach has previously been shown to be effective for determining the identity and relative percentages of crystalline Hg-containing phases in Hg-polluted sediments (15). At higher k values, the α-Hg(0) EXAFS spectrum has one maximum that is aligned with a maximum in the 298K cinnabar

EXAFS spectrum at k = 8.7 and with a maximum in the 298K metacinnabar

EXAFS spectrum at k = 9.5 (Fig. 2). These alignments result in the α-Hg(0) fraction being zero or negative when using the automated fitting protocol in

SixPACK for samples that contain cinnabar, metacinnabar, and α-Hg(0) because of the heavier weighting given to the high-k portion of the EXAFS data in the automated fitting. To compensate for this problem, spectra were fit using the automated fitting protocol for SixPACK for all Hg species, excluding α-Hg(0), and then by manually adding the α-Hg(0) spectrum to the fit at increasing fractions of α-Hg(0). A residual was calculated (see below) for each increase in the α-Hg(0) fraction and plotted versus the percent of α-Hg(0) added (see Fig. S6 in Supplemental Material). The quality of fit was determined by calculating the residual of the fit versus the data using the following formula:

n 2 S (vdata - v ) t = 1 fit Residual = n

136 where “v” is the k3-weighted chi(k) value in the EXAFS spectra (for both the data and fit) and “n” is the number of data points in the fit. For samples containing α-

Hg(0), the residuals plot as a parabola, whereas samples not containing α-Hg(0) showed a minimum at 0% -Hg(0), as shown in Figures S6 (Supplemental

Materials). For detailed sample fitting protocol refer to Supplemental Material.

Previous use of linear combination fitting of Hg LIII-edge EXAFS spectra for mixtures of Hg model compounds has shown an accuracy of ~5% to ~10% in quantifying the percentages of Hg-bearing phases in two- and three-component systems, respectively (15). This method works well only when reference spectra used in the fitting procedure are significantly different in phases and amplitudes

(see, e.g., (24)). As shown by Kim et al. (15), this condition is met for EXAFS spectra of a number of Hg-containing reference materials, and as will be shown below, this is also true for the low-temperature EXAFS spectrum of -Hg(0).

Because liquid Hg(0) is difficult to homogenize in a mixture of solid phases, a two- or three-component test of the linear combination fitting approach using liquid Hg(0) has not been possible up to this point in assessing the accuracy of this approach for liquid Hg(0)-containing samples. However, accuracy of data fitting was possible due to the presence of considerable Hg(0) within the “ultra- pure” α-HgS and β-HgS reference materials, 6.2 atom % and 8.2 atom% respectively (detailed discussion in Supplemental Materials). Other Hg- containing reference materials were found to have undetectable amounts of liquid

Hg(0). Due to the presence of liquid Hg(0) in HgS reference samples, data fitting was carried out using the 298K spectra for all Hg reference samples (excluding α-

137

Hg(0)). Linear combination fitting of the low-temperature EXAFS spectra of the

Hg(0)-containing -HgS and -HgS (Fig. S4 in Supplemental Material, plotted using lower E0 for consistency) reference materials showed that an accuracy of

~5% can also be obtained when this liquid phase is crystallized. Therefore, we estimate that our linear combination fitting results for Hg LIII-edge EXAFS spectra of field samples have an accuracy of ~5% of the phases detected.

To confirm the formation of -Hg(0) in our slow cooling protocol, we compared the experimental low temperature EXAFS spectrum of our -Hg(0) sample with theoretical EXAFS spectra generated using FEFF8.2 (25) and the structure of -Hg(0) reported by Barrett (26). The -Hg(0) model used for FEFF pathway generation was created using the CrystalMaker software package (27) and crystallographic data for -Hg(0) from Barrett (26), who specified only the crystal system (rhombohedral) but not the space group symmetry. Therefore, all seven possible rhombohedral space groups (R3, R3, R32, R3m, R3c, R3m, and

R3c) were tested for α-Hg(0) using CrystalMaker (27), and the d-spacings of the resulting structures generated using CrystalDiffract (28) were compared with the experimental d-spacings for -Hg(0) reported by Barrett (26) (results in

Supplemental Material). We also compared Hg-Hg distances and multiple scattering pathways derived from fits of the experimental low-temperature

EXAFS spectrum of -Hg(0) with those derived by fitting the theoretical EXAFS spectrum of -Hg(0) derived using FEFF 8.2 and the structural model consistent with Barrett’s (26) d-spacings. In this fitting, FEFF pathways were generated for

138

-Hg(0) using FEFF 8.2, with a self-consistency function included to better fit the pure metal spectrum (25).

Results and Discussion

Mercury Speciation in Mine Waste and Sediment Samples. Previous speciation studies of mercury in Hg mine waste and sediments based on ambient- temperature EXAFS studies (7,15) resulted in the finding of relatively insoluble

HgS species being dominant in all samples examined, with minor amounts of water-soluble phases such as montroydite (HgO), schutteite (Hg3(SO4)O2), eglestonite (Hg6Cl3O(OH)), corderoite (Hg3S2Cl2), and HgCl2. To properly fit the

Hg LIII-edge EXAFS spectra of natural samples collected at 77K with light:dark

Hg flux ratios greater than ~3.5, it was necessary to include the -Hg(0) reference spectrum in the least squares fitting. The addition of -Hg(0) resulted in a minimization of fit residuals in a statistically significant fashion. Plots of EXAFS spectra for all 8 samples are seen in Fig. S5 while the residuals are plotted in Fig.

S6 (Supplemental Materials) show the quality of the linear combination fits of the

Hg EXAFS spectra that included -Hg(0). Fits of all the spectra taken at 77K are in reasonable agreement with the results of ambient-temperature EXAFS studies, except that the ambient-temperature speciation data did not include -Hg(0)

(Table 1). Results of linear combination fitting of EXAFS spectra of slowly cooled samples taken at 77K resulted in the detection of significant concentrations of elemental Hg in most of the samples (Table 1). Based on these results, we conclude that low-temperature EXAFS data are superior to ambient-temperature

139 data for linear combination fitting of EXAFS spectra of Hg mine wastes because of the superior signal-to-noise ratios of the former, the damped thermal motion of atoms in the samples at 77K, and the ability to detect liquid Hg(0) using the low temperature sample preparation protocol and 77 K EXAFS.

Correlation Between Hg Speciation and Evasion. Results from mercury evasion studies of the eight Hg mine field samples used in the present study show differences between silica-carbonate- and hot springs-type deposits, but no clear trend between Hg fluxes under both light and dark exposures and total Hg concentration is discernable (Table 1 and Fig. 3A). Earlier EXAFS studies at ambient-temperature had difficulty explaining differences in light:dark Hg fluxes of these samples when HgS minerals were the only dominant Hg phases detected.

Gustin et al. suggested that photoreduction of Hg sulfide- and Hg chloride- containing phases along with Hg bound to organics and Fe oxides could produce

Hg(0) (29, 30). These earlier studies also noted that photoreduction and/or physical desorption of Hg(0) is important.

A number of variables can affect the rate of mercury evasion from Hg- contaminated samples. For example, several groups have shown that soil moisture, temperature, atmospheric Hg concentrations, and light have strong impacts on Hg volatilization rates (31-33). The samples used in our experiments were dried at room-temperature before volatilization experiments were conducted, making the soil moisture and atmospheric Hg concentrations of little consequence in this study. Since the Pyrex sample chamber used in the flux chambers did not transmit light of wavelength < 500nm, exposure of the samples to UV light is not

140 the cause of the enhancement of Hg volatilization (29). Because the penetration depth of light of wavelength >500nm into sediment grains is 50-100µm in most soils (34), it is likely that Hg fluxes should decrease over time as exposed sediments become depleted in liquid Hg(0). However, movement of gaseous elemental Hg by diffusion towards the surface (35, 36), in addition to sediment turnover from slope failure, winter rain run-off, and/or aeolian processes, could provide a means by which new material containing Hg(0) would be exposed to sunlight, resulting in more or less continuous gaseous Hg volatilization over time.

We found that samples with light:dark Hg flux ratios >3.5, -Hg(0) comprised between 9% and 22% of the total Hg species (Table 1), whereas in those with light:dark Hg flux ratios < 3.5 there was no detectable Hg(0) based on

EXAFS linear combination fitting. When light:dark Hg flux ratio is plotted versus

% liquid Hg(0), linear relationships emerge for both silica-carbonate- and hot springs-type Hg deposits (Fig. 3B). The trends in Fig. 3B offer another reason for differences in Hg evasion rates from different mine waste samples, which is the presence of relatively abundant liquid Hg(0) (up to 22 atom%) in some mercury mine wastes, part of which is native and not the result of photoreduction of higher valent Hg species.

In the California Coast Range, silica-carbonate deposits were formed by alteration of serpentinite rocks, which tended to result in Hg-bearing minerals of larger crystal size when compared with the younger hot springs-type deposits (2).

A possible explanation for the lower slope of the silica-carbonate-type trend relative to the hot springs-type trend (Fig. 3B) is that the sizes of the Hg(0)

141 globules are generally larger in silica-carbonate-type Hg deposits (or are entrained/encapsulated within larger particles) and thus limit Hg(0) evasion to the atmosphere relative to the smaller Hg(0) globules in hot springs-type deposits.

Consistent with this explanation is Sladek and Gustin’s observation that less

Hg(0) volatilized during pyrolysis of larger Hg(0) beads relative to smaller beads

(37). The higher Hg(0) concentrations observed in hot springs-type relative to silica-carbonate-type deposits (Fig. 3B) may be related to the generally smaller grain size of Hg-containing minerals from hot springs-type deposits (Rytuba, pers. comm.), which should make them more susceptible to weathering and conversion into liquid Hg(0). Another explanation for this observation is that the higher chloride content in both HgS crystals and hydrothermal solutions in hot springs- type Hg deposits leads to enhanced photosensitivity of HgS crystals, which should result in higher levels of Hg(0) produced by photoreduction (17, 38, 39).

Although liquid Hg(0) globules were not detected in Sulphur Bank samples by

Varekamp and Buseck (39), liquid Hg(0) is clearly present in large concentrations

(up to 22 atom %), and, as a consequence, Sulphur Bank samples have the largest light:dark Hg flux ratio relative to the other samples examined in the present study.

Implications of Low-temperature Hg LIII-edge EXAFS Spectroscopy for Hg

Evasion Data. Using low-temperature EXAFS spectroscopy and a new sample cooling protocol, we have shown that a significant amount of liquid Hg(0) (up to

22 atom %), previously undetectable by conventional EXAFS spectroscopy methods, is present in mercury mining wastes associated with abandoned Hg

142 mines in the California Coast Range. The speciation data obtained using this new approach help explain the higher light:dark Hg flux ratios from Hg mine wastes and sediment samples containing higher concentrations of liquid Hg(0).

However, complete understanding of Hg evasion rates has not yet been achieved because only a limited number of samples have been studied to date using these new experimental protocols. Additional Hg speciation vs. light:dark Hg flux ratio data are needed to verify the two linear correlations seen in Fig. 3B, and such studies are underway. Based on this study, however, it is clear that previous ambient-temperature EXAFS analyses of Hg-impacted sediments and mine wastes do not result in a complete picture of Hg cycling in Hg mining environments because the concentration of liquid Hg(0) is missing from the speciation information.

143

Acknowledgements

We wish to thank Joe Rogers and John Bargar of the Stanford Synchrotron

Radiation Lightsource for help with data collection on beamline 11-2. The LN2 cryostat used in the low-temperature work was designed and built by Steve

Conradson of Los Alamos National Laboratory. GEB thanks George Parks

(Stanford University) for helpful discussions over a number of years about the environmental chemistry of mercury. Funding for this research comes from the

Stanford Environmental Molecular Science Institute through NSF Grant CHE-

0431425. Portions of this research were carried out at the Stanford Synchrotron

Radiation Lightsource, a National user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Science, with additional support from the National Institute of Health.

144

Literature Cited

(1) Ullrich, S. M.; Tanton, T. W.; Abdrashitova, S. A. Mercury in the aquatic

environment: A review of factors affecting methylation. Critical Reviews

in Environmental Science and Technology 2001, 31 (3), 241-293.

(2) Rytuba, J. J. Mercury from mineral deposits and potential environmental

impact. Environmental Geology 2003, 43, 326-338.

(3) Hylander, L. D. Global mercury pollution and its expected decrease after a

mercury trade ban. Water, Air, and Soil Pollution 2001, 125, 331-344.

(4) Morel, F. M. M.; Kraepiel, A. M. L.; Amyot, M. The chemical cycle and

bioaccumulation of mercury. Annual Review of Ecological Systems 1998,

29, 543-566.

(5) Bloom, N. S.; Preus, E.; Katon, J.; Hiltner, M. Selective extractions to

assess the biogeochemically relevant fractionation of inorganic mercury in

sediments and soils. Analytica Chimica Acta 2003, 479 (2), 233-248.

(6) Biester, H.; Scholz, C. Determination of mercury binding forms in

contaminated soils: Mercury pyrolysis versus sequential extractions.

Environmental Science & Technology 1997, 31 (1), 233-239.

(7) Kim, C. S.; Rytuba, J. J.; Brown Jr., G. E. Geological and anthropogenic

factors influencing mercury speciation in mine wastes: An EXAFS

spectroscopy study. Applied Geochemistry 2004, 19, 379-393.

(8) Skyllberg, U.; Bloom, P. R.; Qian, J.; Lin, C.-M.; Bleam, W. F.

Complexation of mercury(II) in soil organic matter: EXAFS evidence for

145

linear two-coordination with reduced sulfur groups. Environmental

Science & Technology 2006, 40 (13), 4174-4180.

(9) Skyllberg, U.; Qian, J.; Frech, W. Combined XANES and EXAFS study

on the bonding of methyl mercury to thiol groups in soil and aquatic

organic matter. Physica Scripta 2005, T115, 894-896.

(10) Kim, C. S.; Bloom, N. S.; Rytuba, J. J.; Brown Jr., G. E. Mercury

speciation by x-ray absorption fine structure spectroscopy and sequential

chemical extractions: A comparison of speciation methods. Environmental

Science & Technology 2003, 37 (22), 5102-5108.

(11) Lowry, G. V.; Shaw, S.; Kim, C. S.; Rytuba, J. J.; Brown Jr., G. E.

Macroscopic and microscopic observations of particle-facilitated mercury

transport from New Idria and Sulphur Bank Mercury Mine tailings.

Environmental Science & Technology 2004, 38 (19), 5101-5111.

(12) Biester, H.; Gosar, M.; Covelli, S. Mercury speciation in sediments

affected by dumped mining residues in the drainage area of the Idrija

Mercury Mine, Slovenia. Environmental Science & Technology 2000, 34

(16), 3330-3336.

(13) Biester, H.; Gosar, M.; Muller, G. Mercury Speciation in tailings of the

Idrija Mercury Mine. Journal of Geochemical Exploration 1999, 65, 195-

204.

(14) Biester, H.; Nehrke, G. Quantification of mercury in soils and sediments-

acid digestion versus pyrolysis. Fresenius' Journal of Analytical

Chemistry 1997, 358, 446-452.

146

(15) Kim, C. S.; Brown Jr., G. E.; Rytuba, J. J. Characterization and speciation

of mercury-bearing mine wastes using x-ray absorption spectroscopy. The

Science of the Total Environment 2000, 261, 157-168.

(16) Slowey, A. J.; Rytuba, J. J.; Brown Jr. G. E., Speciation of mercury and

mode of transport from placer gold mine tailings. Environmental Science

& Technology 2005, 39 (6), 1547-1554.

(17) Gustin, M. S.; Biester, H.; Kim, C. S. Investigation of the light-enhanced

emission of mercury from naturally enriched substrates. Atmospheric

Environment 2002, 36, 3241-3254.

(18) U.S. Environmental Protection Agency, Method 1631, Revision E:

Mercury in water by oxidation, purge and trap, and cold vapor atomic

fluorescence spectrometry. EPA: Washington, DC, 2002.

(19) Thermal and mechanical properties of Teflon (polytetra fluorethylene).

www.yutopian.com/Yuan/prop/Teflon.html (accessed Feb 2, 2009).

(20) Ravel, B. Athena, version 0.8.056, 2002.

(21) Farges, F. GR-xafsX, version 2.4, 2008.

(22) Webb, S. SixPACK, version 0.67; Stanford Synchrotron Radiation

Lightsource: Menlo Park, 2006.

(23) Webb, S. M. SixPACK: a graphical user interface for XAS analysis using

IFEFFIT. Physica Scripta 2005, T115, 1011-1014.

(24) Catalano, J. G.; Brown Jr., G. E. Analysis of uranyl-bearing phases by

EXAFS spectroscopy: Interferences, multiple scattering, accuracy of

147

structural parameters, and spectral differences. Am. Mineral. 2004, 89,

1004-1021.

(25) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Real-space

multiple-scattering calculation and interpretation of x-ray-absorption near-

edge structure. Physica Review B 1998, 58 (12), 7565-7576.

(26) Barrett, C. S. The structure of mercury at low temperatures. Acta

Crystallographica, 1957, 10, 58-60.

(27) Palmer, D. and Conley, M., CrystalMaker for Windows version 2.0.7.

CrystalMaker Software Ltd. 2008.

(28) Palmer, D. and Conley, M., CrystalDiffract for Windows version 1.1.2.

CrystalMaker Software Ltd. 2007.

(29) Gustin, M. S.; Nacht D.; Engle, M. A. Speciation of mercury above

naturally and anthropogenically mercury enriched substrate. Abstract with

Program, American Chemical Society National Meeting, 2002.

(30) Gustin, M. S. Are mercury emissions from geologic sources significant? A

status report. The Science of the Total Environment 2003, 304, 153-167.

(31) Moore, C.; Carpi, A. Mechanisms of the emission of mercury from soil:

Role of UV radiation. Journal of Geophysical Research 2005, 110

(D24302), 1-9.

(32) Xin, M.; Gustin, M. S.; Johnson, D. Laboratory investigation of the

potential for re-emission of atmospherically derived Hg from soils.

Environmental Science & Technology 2007, 41 (14), 4946-4951.

148

(33) Xin, M.; Gustin, M. S. Gaseous elemental mercury exchange with low

mercury containing soils: Investigation of controlling factors. Applied

Geochemistry 2007, 22, 1451-1466.

(34) Ciani, A.; Goss, K.-U.; Schwarzenbach, R. P. Light penetration in soil and

particulate minerals. European Journal of Soil Science 2005, 56, 561-574.

(35) Zhang, H. and Lindberg, S. E. Processes influencing the emission of

mercury from soils: A conceptual model. Journal of Geophysical

Research 1999, 104 (D17), 21889-21896.

(36) Gustin, M. S. and Stamenkovic Effect of watering and soil moisture on

mercury emissions from soils. Biogeochemistry 2005, 76, 215-232.

(37) Sladek, C. and Gustin, M. S. Evaluation of sequential and selective

extraction methods for determination of mercury speciation and mobility

in mine waste. Applied Geochemistry 2003, 18, 567-576.

(38) McCormack, J. K. The darkening of cinnabar in sunlight. Mineralium

Deposita 2000, 35, 796-798.

(39) Varekamp, J. C. and Buseck, P. R. The speciation of mercury in

hydrothermal systems, with applications to ore deposition. Geochimica et

Cosmochimica Acta 1984, 48 (1), 177-185.

149

Table 1: Sample information for selected Hg-mine sediments including results from Hg evasion studies and low-temperature EXAFS analysis, including linear combination EXAFS fitting results, compared with ambient-temperature linear combination EXAFS fitting results. The estimated accuracies of the linear combination fit results in the column labeled “Hg Speciation (77K)” are ±5%. - HgS = cinnabar; -HgS = metacinnabar.

Dark Light Hg Hg light: Hg Hg Location Hg Flux Flux Sample Sample Concen- dark Speciation Speciation Residual (deposit @35oC @35oC Name Type tration ratio (298K) (77K) (77K) type) (ng/m2h) (ng/m2 (ppm) h)

Knoxville 32% -HgS 2257 ± 36.0 ± 22% -HgS KN1B (hot Calcine 277 78 ± 19 52% -HgS 0.381 50 0.2 78% -HgS springs) 16% Hg(0)

Calcine 34% -HgS 36%  -HgS New Idria (500- 20857 12307 1.7 ± 43% -HgS 47%  - HgS NISB (silica- 350 0.09 2000 ±160 ± 307 0.1 carbonate) 23% 17% μm) Eglestonite Eglestonite

35% -HgS 29% -HgS New Idria Calcine 48% -HgS 14390 ± 3874 ± 3.7 ± 56% -HgS NIS3 (silica- (75-125 690 7% 0.099 450 148 0.1 15% carbonate) μm) Montroydite Montroydite 10% Hg(0)

New Idria Calcine 56% -HgS 45% -HgS 15778 ± 3655 ± 4.3 ± NIS6 (silica- (<45 770 44% -HgS 46% -HgS 0.061 131 381 0.1 carbonate) μm) 9% Hg(0)

Waste Sulphur Rock 48% -HgS SBW9 Bank 4301±33 43.9 65% -HgS (500- 1580 98 ± 14 30% -HgS 0.126 S3 (hot 6 ± 0.2 35% -HgS 2000 22% Hg(0) springs) μm) Sulphur Waste SBW9 Bank Rock 7334±17 3851 ± 1.9 ± 82% -HgS 81% -HgS 22310 0.168 S6 (hot (75-125 3 114 0.1 18% -HgS 19%  - HgS springs) μm) Sulphur Waste 61% -HgS SBW9 Bank Rock 9032±17 334 ± 27.1 78% -HgS 8120 30% -HgS 0.12 S9 (hot (<45 6 212 ± 0.1 22% -HgS springs) μm) 9% Hg(0) New Conden Almaden 46586± 15058 3.1 ± NAC3 ser 19500 100% -HgS 100% -HgS 0.051 (silica- 946 ± 2614 0.2 Soot carbonate)

150

2 Alpha Hg(0) 77K 1.5 Liquid Hg(0) 298K

1

)

k 0.5

chi(

3 0 k -0.5

-1

-1.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 -1 k (Å )

Figure 1: Hg LIII EXAFS spectra for elemental Hg(0) at both room temperature (liquid Hg(0)) and 77K (-Hg(0)) showing the enhancement of amplitudes in the

Hg LIII EXAFS spectrum obtained at 77K after slow cooling of the sample. These spectra were averaged from 12 scans each.

151

3

2

1

) k

0

chi(

3 k -1

-2 Cinnabar Metacinnabar Alpha Hg(0) -3 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 k (Å-1)

Figure 2: Comparison of k3-weighted EXAFS spectra of -Hg(0), -HgS, and - HgS. -Hg(0) spectra was taken at 77K, while -HgS and -HgS spectrum were taken at 298K. The -Hg(0) sample was prepared using the new sample cooling protocol discussed in the text.

152

A 70

60 Silica-carbonate Deposit

50

40 Hot Springs Deposit

Lt:Dk 30

20

10

0 0 5000 10000 15000 20000 25000 Concentration Hg (mg/kg)

B

50 45 Silica-carbonate Deposit y = 1.8x + 4.6 40 r² = 0.93 35 Hot Springs Deposit

30 25

Lt:Dk 20 15 10 y = 0.2x + 1.70 r² = 0.91 5 0 0 5 10 15 20 25 % Elemental Hg

Figure 3: (A) Plot of light:dark evasion data versus total Hg concentration separated into Silica-carbonate and Hot Spring Deposits. (B) Plot of Light:Dark (Lt:Dk) evasion data versus % liquid Hg(0) in each sample, with data points grouped into Silica-carbonate and Hot Spring Deposits. Plot B shows the linear relationship between Lt:Dk and % liquid Hg(0) for the two separate deposits. y-intercepts not equaling 1 are considered to be the result of encapsulation of liquid Hg(0) in sediment grains. Lt:Dk error bars are smaller than the symbols used.

153

Supplemental Material

A New Technique for Quantification of Elemental Hg in Mine

Wastes and Its Implications for Mercury Evasion Into the

Atmosphere

ADAM D. JEW, CHRISTOPHER S. KIM, JAMES J. RYTUBA, MAE S. GUSTIN, AND GORDON E. BROWN, JR.

154

Structure of -Hg(0).

Barrett reported that α-Hg(0) is rhombohedral, but a specific space group for α-

Hg(0) was not identified in that study (1). Crystal structures generated using

CrystalMaker for space groups R3, R32, R3c, and R3c resulted in theoretical diffraction patterns derived using CrystalDiffract that are inconsistent with those reported by Barrett (1-3). In contrast, Barrett’s α-Hg(0) structure (1) is consistent with theoretical diffraction patterns, generated using the R3, R3m, and R3m space groups, all of which are identical. Among these three, space group R3m was arbitrarily selected for use in FEFF generation of single and multiple scattering paths for -Hg(0) (4). Results of shell-by-shell fitting for Hg-Hg interatomic distances were close to literature and model values, with first-shell Hg-Hg distances of 2.95 ± 0.01 Å and 2.96 ± 0.01 Å and a second-shell Hg-Hg distance of 3.50 ± 0.02 Å. These values are similar to those generated by FEFF 8.2 (2.99

Å and 3.46 Å, respectively) (Fig. S1) (1, 4). A coordination number of six Hg is expected for both the first and second atomic shells around a central Hg. The

EXAFS-derived first-shell coordination number for Hg is 6.01 ± 0.57 Hg, whereas the second shell was under-coordinated (4.7 ± 0.9 Hg) when compared to the ideal structure. Individual FEFF-generated chi(k) spectra for α-Hg(0) showed phase cancellation between the larger amplitude second-shell distance of 3.46 Å and the next single-scattering pathway at 4.88 Å. This phase cancellation resulted in the absence of signal for backscattering atoms beyond the second shell (since the amplitude of chi(k) lessens with distance) and a dampening of the magnitude of the second shell in the Fourier transform of the α-Hg(0) EXAFS spectrum.

155

During the course of shell-by-shell fitting analysis of α-Hg(0), it was discovered that numerous values for α-Hg(0) cell parameters taken from online databases and reference literature are incorrect, resulting in either an incorrect α-Hg(0) crystal structure or poor fitting of the experimental -Hg(0) EXAFS spectrum. The unit cell parameters for α-Hg(0) used in creating the theoretical model were a =

o 2.992Å and α = 70.74 . α-Hg(0) was observed in the cinnabar (-HgS) and metacinnabar (-HgS) reference EXAFS spectra at 77K (additional discussion below). In addition to clearly resolved features in the α-HgS spectrum at 77K caused by -Hg(0), the main oscillation produced by the Hg-S pair correlation is in the same position for both the 77K and 298K EXAFS data (Fig. S2).

Comparison of the spectra for α-Hg(0) (77K), α-HgS (298K), and β-HgS (298K) shows that they are out of phase relative to one another, thus allowing them to be used in linear-combination fitting of the field samples (Fig. 2).

Data Analysis.

To properly fit field samples measured at 77K with -Hg(0), a manual fitting minimization was done. Initially, α-Hg(0) was added from 0-100% with step sizes of 10% until a minimum was found. The fits were re-calculated over a range of ± 5% of the minimum in the previous fits with a step size of 1% in order to refine the fraction of α-Hg(0) contained in the sample. During fitting, minimization of the residuals at 10% and 1% step size was found to result in identical α-Hg(0) fractions in the residual versus α-Hg(0) plot. Figure S6 shows the residuals of all 8 samples from 0-40% -Hg(0) with a step size of 0.01% -

156

Hg(0). To determine if the addition of -Hg(0) to the fits was significant

Hamilton’s R-factor ratio test was carried out on all eight samples (6). The five samples with detectable -Hg(0) showed that the improvement in linear combination fitting resulting from addition of the EXAFS spectrum of -Hg(0) at

77K was significant at the 99.5% confidence level (Table S1).

Contamination of -HgS and -HgS with liquid Hg(0).

During the course of this work, we found by low-temperature EXAFS spectroscopy that there were subtle differences between the 298K and 77K -HgS and -HgS EXAFS spectra (Fig. S2). These subtle differences are caused by the presence of -Hg(0) in the 77K EXAFS spectra of the HgS reference materials.

A linear combination fitting of the 77K -HgS and -HgS reference spectra

(99.999% pure, trace metal basis) using the 298K -HgS and -HgS EXAFS spectra and the -Hg(0) spectrum resulted in the finding that -HgS and -HgS contain 6 atom % and 8 atom %, respectively, of liquid Hg(0). This apparent contradiction between sample purity and liquid Hg(0) contamination can be understood when taking into account that the purity is based on metal analysis of the dissolved material and not on phase purity. Before the -HgS and -HgS samples were taken to SSRL for EXAFS analysis, XRD analyses of the samples were performed on the materials to confirm phase purity. Samples were analyzed

o on a Rigaku Model CM2029 using a Cu Kα x-ray source over a range of 5-70 (2- theta). Analysis of the resulting diffractograms was done using the Jade X-ray

Diffraction Software package (5). As seen in the diffractogram for -HgS

157

(99.999% pure, trace metal basis) (Fig. S3), about 15 wt.% -HgS was present in the -HgS sample purchased from Alfa Aesar, but no -HgS was detected in the

-HgS sample purchased from Alfa Aesar. The -HgS Alfa Aesar sample was sent back to the manufacturer for replacement prior to low-temperature EXAFS data collection, and a new -HgS sample was received. XRD analysis of the new

-HgS sample showed that it contained no detectable -HgS. Due to the poorly crystalline nature of liquid Hg(0), determining the presence of liquid Hg(0) in these HgS samples by XRD at 298K was not possible. EXAFS analysis of the

EXAFS spectra of the other reference materials showed no evidence for -Hg(0).

However, Hg(0) was found in the -HgS and -HgS samples, as discussed above.

The manufacturer was contacted about the presence of liquid Hg(0) within the

HgS samples. The response from the manufacturer was that the HgS (both - and

-) routinely have 5-10 atom % liquid Hg(0) as a result of manufacturing procedures (pers. comm.). Although Alfa Aesar confirmed the presence of significant quantities of liquid Hg(0) in their HgS products, they would not comment on the testing method used to determine the concentration of liquid

Hg(0). Alfa Aesar also noted that since purity is based on metal analysis of the dissolved sample, -HgS can contain -HgS, -HgS, elemental S, and elemental

Hg(0) and would still qualify for the 99.999% purity designation.

158

Literature Cited

(1) Barrett, C. S. The structure of mercury at low temperatures. Acta

Crystallographica 1957, 10, 58-60.

(2) Palmer, D. and Conley, M. CrystalMaker for Windows version 2.0.7.

CrystalMaker Software Ltd. 2008.

(3) Palmer, D. and Conley, M. CrystalDiffract for Windows version 1.1.2.

CrystalMaker Software Ltd. 2007.

(4) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Real-space

multiple-scattering calculation and interpretation of x-ray-absorption near-

edge structure. Physica Review B 1998, 58 (12), 7565-7576.

(5) Materials Data Inc.. Jade XRD Pattern Processing Ver. 6.5, 2002.

(6) Hamilton, W. C., Significance test on the crystallographic R factor. Acta

Crystallographica 1965, 18 (3), 502-510.

159

2 A Data 1.5 Fit

1

) 0.5 k

0

chi(

3

k -0.5 -1 -1.5 -2 3 4 5 6 7 8 9 10 11 12 -1 k (Å )

B 1.6

1.4 Data

1.2 Fit 1

0.8

0.6 Amplitude 0.4 0.2

0 0 1 2 3 4 5 6 R + Δ (Å) 2 C Pathway CN R (Å) σ Hg-Hg 1st Shell 4.01 ± 0.50 2.95 ± 0.01 0.0119 ± 0.0010

Hg-Hg 1st Shell 2.00 ± 0.27 2.96 ± 0.01 0.0117 ± 0.0018

Hg-Hg 2nd Shell 4.74 ± 0.86 3.50 ± 0.02 0.026 ± 0.0048

2 Δ Eo -2.26 ± 0.60 Red-chi 45.62

2 So 0.78 R-factor 0.0684

Figure S1: Shell-by-shell fitting of the -Hg(0) Hg LIII-edge EXAFS spectrum using theoretical pathways generated by FEFF 8.2: (A) fit of k3-weighted EXAFS, (B) fit of

Fourier Transform, and (C) table showing fitting parameters for α-Hg(0).

160

A

)

k

chi(

3 3 k

Cinnabar 77K

Cinnabar 298 K

Alpha Hg(0) 3 4 5 6 7 8 9 10 11 12 -1 k (Å )

B

)

k

chi(

3 k

Metacinnabar 77K Metacinnabar 298K Alpha Hg(0)

3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 -1 k (Å ) Figure S2: A) EXAFS spectra of cinnabar at 77K, cinnabar at 98K, and elemental Hg at 77K, with arrows demarking the extra features in cinnabar (77K) due to elemental Hg. B) EXAFS spectra of metacinnabar at 77K, metacinnabar at 298K, and elemental Hg at 77K, with arrows demarking the extra features in cinnabar (77K) due to elemental Hg. Linear combination fitting results of cinnabar 77K using cinnabar 298K and -Hg(0) resulted in the detecting the presence of -Hg(0) at 6 atom %; the manufacturer confirmed that such reference materials normally contain 5-10% elemental Hg.

161

Fig. S3: X-ray diffractogram of “ultra-pure” (99.999% pure, trace metal basis) of -HgS sample from Alfa Aesar. The diffractogram shows the presence of -HgS in the sample, illustrating that the purity based on metal analysis does not guarantee phase purity.

162

A

B

Fig. S4: A) Linear combination fit of 77K “ultra-pure” -HgS using 298K -HgS spectra and -Hg(0) spectra. Fit parameters are 94% -HgS and 6% -Hg(0) with a residual of 0.184. B) Least squares fit of 77K “ultra-pure” -HgS using 298K -HgS and -Hg(0) spectra. Fit parameters are 92% -HgS and 8% - Hg(0) with a residual of 0.076. Manufacturer states that both - HgS and -HgS contain 5-10% liquid Hg(0) (pers. comm.)

163

KN1B Data KN1B Fit

NISB Data NISB Fit

NIS3

Data

) NIS3 Fit k NIS6

chi( Data

3 NIS6 Fit k SBW9S3 Data SBW9S3 Fit SBW9S6 Data SBW9S6 Fit

3.5 4.5 5.5 6.5 7.5 8.5 9.5 -1 k (Å ) Fig. S5: Fits of eight samples with various light:dark Hg fluxes.

All spectra collected at 77K. A different E0 for the spline was used for these data compared to the other figures, accounting for a slight shift in k-space. Parabolic residual plots for all 8 samples are found in Fig. S6.

164

Fig. S6: Residual plots for all 8 samples analyzed for -Hg(0). All spectra were collected at 77K.

165

Table S1: Results Hamilton’s R-factor ratio test (6) for all 8 samples obtained by the addition of -Hg(0) to the fit. When b = 1 and n = 120, the R-factor ratio for 99.5% level of confidence is 1.034.

Sample R-factor Ratio

KN1B 2.07

NISB 1.0

NIS3 1.32

NIS6 1.07

SBW9S3 1.08

SBW9S6 1.0

SBW9S9 2.02

NAC3 1.0

166

Appendix I

Ultra-trace Mercury Analysis Protocols

Sample Containers and Cleaning Protocol:

Ultra-trace level Hg analysis is difficult because of two competing problems with samples and reagents. The first problem is Hg adsorption to sample containers and the second is contamination of samples, either from the environment or the sample containers themselves. Mercury adsorption to sample containers is caused by two different things: 1) the material of the container and

2) the preservative used.

There are 4 main types of containers that can be used polypropylene, high density polyethylene (HDPE), Teflon®, and borosilicate glass. A detailed study by Parker and Bloom showed that there were severe problems with HDPE allowing atmospheric Hg to travel through the sides of container resulting in an increase in total Hg (1). Polypropylene containers have similar problems to

HDPE in that Hg concentrations increase significantly when bromine monochloride (BrCl) is used as a preservative. In the laboratory it has been seen that storing samples in polypropylene without BrCl resulted in an irreversible loss

167 of > 80% Hg by adsorption to the container walls within 20 minutes. For ultra- trace Hg work all types of plastics should be avoided unless the sample will be in contact with the plastic for less than 3 minutes, such as in a syringe. Teflon® containers are good for Hg storage as long as they are preserved with BrCl.

Though Teflon bottles are good for long term storage there are some issues with using Teflon®. Teflon® bottles are quite expensive, they are not easy to clean for ultra-trace work, and in some cases Hg can still adsorb to the Teflon®, especially when used for acid digestions. The best container to use for Hg samples is clear borosilicate vials with Teflon® lined lids (I-Chem® brand is recommended).

Though not well understood, amber borosilicate vials have been seen in the lab to produce poor accuracy and precision for total Hg concentrations. Ideally borosilicate vials with borosilicate tops would be used but it would not result in an air tight seal which would allow atmospheric Hg to penetrate into the bottle.

Borosilicate glass with Teflon® lined lids is cheap, air tight, and easy to clean making it ideal for Hg analysis.

Proper cleaning procedures for Hg are essential for attaining high accuracy and precision of samples. To clean sample containers made solely of borosilicate glass sample containers can be cleaned by placing them in a Hg-free furnace at >

500oC for 4 hours to burn off any adsorbed Hg (this is for Hg and C only, other metals need to be cleaned by acid bath). For the cleaning of Teflon or borosilicate glass for Hg analysis a 3 step cleaning process should be followed. The three step process is intended mainly for borosilicate glass with Teflon® lined lids with the final cleaning step is mandatory even for brand new vials/bottles. When

168 cleaning vials and bottles it is important to triple rinse everything with DI water to ensure little to no carryover between steps. Since this cleaning protocol works for all Hg this cleaning protocol works well for methylmercury cleaning making the use of EPA methods for methylmercury unnecessary.

1) Detergent bath: aimed at removing all the large contaminants, residual

organics, etc. in order to keep the trace metal acid bath as clean as

possible.

2) 1N hydrochloric acid bath (trace metal grade): an overnight soaking is

to remove most metals from the containers. The acid bath is enough to

clean Hg off of borosilicate glass but not enough to clean off the Hg

from the Teflon® lined lids.

3) 5mL of 10% (BrCl): vials are filled, capped, and inverted overnight so

that there is constant contact between the Teflon® and the BrCl. In

general no matter what size the vial/bottle being cleaned 5mL should

be enough, the important thing is that the Teflon® be submerged with

about 1 cm of BrCl on top. Ensure that the first rinse after BrCl

wash is done in fume hood.

4) Oven drying: loosely cap the vials and place in Hg clean oven

overnight at a temperature between 70oC and 100oC, any hotter will

begin to melt plastic in the caps.

Once the caps are matched up with a vial/bottle for the BrCl cleaning they should always be paired together. If caps and vials start getting mixed it is very difficult

169 to determine where potential contamination is coming from. After initial cleaning of vials if sample concentrations are < 200ng/L (determined by Tekran analysis) a

10% BrCl cleaning is all that is required afterwards. Sample vials that are used for acid digestions need to have the Teflon® liner replaced because the level of Hg adsorbed to the liner is nearly impossible to clean off in order to reach ultra-trace level analysis again. The amount of Hg adsorbed is not a detectable loss for the amount of Hg in a digestion tube, but results in high levels of carryover if used as a sample tube for cold vapor atomic fluorescence spectrometry (CVAFS) analysis.

Field Sampling Protocols:

Proper field sampling methods are crucial for accurate measurement of total Hg concentrations in aquatic systems. To ensure good analytical results 3 major considerations need to be addressed: 1) the sample container, 2) the type of syringe/pipette, and 3) the filter type. The best sample containers are the same described above, clear borosilicate glass with Teflon® lined lids. Sample preservation is done with the addition of 0.5% BrCl resulting in a slight yellow color (depending on the amount of organics) that can be stored at least a year without any special storage considerations.

The type of syringe or pipette (plastic or glass) is of minor concern for Hg sampling as long as the plunger tip does not contain either rubber or latex. For

170 low Hg concentrations, < 1 μg/L, all Hg is lost to the plunger tip with a single fill and discharge. The adsorption of Hg by rubber plunger tips for higher concentrations, 1 μg/L to 50 mg/L, will occur at a rate of ~80% loss with a single fill and discharge. Plastic syringes and pipettes that are taken directly from the package are suitable for Hg sampling as long as the syringe/pipette is pre- contaminated with the sample and that the sample is in the syringe/pipette for less than 5 minutes. Pre-contaminating the syringe/pipette is done by filling and emptying the syringe/pipette several times with the water that is going to be filtered or placed directly into the sample vial. This pre-contamination step results in flushing anything adsorbed to the plastic that the water chemistry will de-sorb and any Hg that will adsorb to the plastic should completely fill the adsorptive capacity of the plastic. The time that the sample stays in the syringe/pipette is more crucial than the pre-contamination step. Since loss of Hg to the plastic is detectable for ng/L concentrations within 20 minutes, the less contact time with the plastic the better. Numerous field sampling trips have shown that samples filtered within 5 minutes of being placed in a plastic syringe have no detectable loss of Hg to the syringe.

The type of filter used is the most important part of proper field sampling.

There are 5 major types of filters used for field work: 1) glass microfiber (GMF),

2) polyethersulfone (PES), 3) cellulose, 4) Teflon®, and 5) Anatop® (alumina matrix).

1) Glass microfiber filters are ideal for Hg samples since adsorption to

glass is non-detectable and the GMF filters can be used for

171

concentrated acids. There are 3 main drawbacks to these filters: 1)

adsorption of organics to these filters is large, 2) the smallest pore size

is 0.45 μm, and 3) after a large amount of liquid volume filtered the

microfibers re-align opening up the pore size to larger than 0.45 μm.

These filters are considered to be the worst for organic carbon with

measured losses in the lab of > 90% upon filtering.

2) Polyethersulfone filters are ideal for nearly all metals and carbon

samples with the exception of Hg. PES filters readily adsorb Hg, have

a minimum pore size of 0.2 μm, and cannot handle concentrated acids.

3) Cellulose filters work well for Hg with little to no loss but cannot

tolerate concentrated acids and have a minimum pore size of 0.2 μm.

4) Teflon® filters can be used for Hg work but are often difficult to use.

These filters are made with both hydrophilic and hydrophobic Teflon®

which can make filtering natural water samples difficult. Trace

amounts of Hg will adsorb to the filters but the largest drawback is that

minimum pore size is 0.2 μm.

5) Anatop® filters are an alumina matrix filter that is ideal for work with

Hg because the Hg does not adsorb to alumina. The Anatop® plus

version of these filters contains a GMF pre-filter to clean out larger

particles prior to the alumina membrane. The minimum pore size is

0.02 μm which gives these filters the smallest pore size for a filter

syringe. The main drawback with these filters is that the alumina

172

begins to degrade when strong acids are used a pH < 1, which results

in an opening of membrane pores.

After samples are filtered and properly preserved for lab work the samples

can then be run for total Hg on a Tekran 2600 CVAFS or similar system.

Do not do total Hg analysis with ICP-MS, ICP-AES, or ICP-OES. The

results are inconsistent and the detection limit is very poor. It is estimated

that for every 100,000 atoms of Hg inserted into an ICP-MS, only 2-8

atoms actually strike the detector.

Tekran 2600 CVAFS Instruction Manual

This instruction manual is to aid in the measurement of ultra-trace levels of Hg in environmental samples using the Tekran 2600 CVAFS total Hg analyzer.

The two accepted methods for total Hg analysis by the US EPA and the general

Hg community are CVAFS (ultra-trace concentrations) and CVAAS (absorbance instead of fluorescence, for trace concentrations). Since the CVAFS uses a fluorescent signal instead of an absorption signal for detecting Hg, the detection limit is significantly lower and has greater reproducibility. Refer to EPA method

1631 for a more detailed outline of the technique (2).

Total Hg analysis is done by digesting solid samples in mixture of strong acids to release Hg into solution. Depending on the sample matrix, different acid mixtures are used. The two main acid mixtures are aqua regia (HCl/HNO3, for

173 most soils) and a H2SO4/HNO3 mixture (organic materials). Protocols for specific acid digestions provided by Tekran are saved as .pdf files on the computer attached to the Tekran 2600 CVAFS. Even though digestion chemicals such as aqua regia are strong oxidizers the sample still needs to be preserved with a stronger oxidizer (BrCl, bromine monochloride) in order to keep Hg in solution as

Hg2+, since it will not volatilize like Hgo and keep Hg from adsorbing to the glassware. Because water samples have even less oxidizing conditions than acid digestates, samples also need preservation with BrCl. The purpose of this is to

o + oxidize all the Hg (Hg , HgCl2 complexes Hg , Hg bound to organics, Hg bound to minerals, etc.) in the sample to Hg2+, thus destroying speciation information but giving consistent total Hg values. Preserved samples have a slight yellow color to them which comes from the free BrCl halogens in solution, before analysis these free halogens are neutralized by adding hydroxalamine hydrochloride (HAH) to bind the halogens which turns the solution color clear. Once samples are introduced into the CVAFS the sample is reacted with a strong reductant,

o Stannous Chloride (SnCl2), to produce Hg , which is then stripped from the solution as a gaseous phase with Aro gas. Since Hg amalgams with Au, the machine uses Au coated quartz sand to trap the Hg and then concentrates the Hg by desorbing the Hg from the first Au trap and then trapping it on a second Au trap. The point of these traps is to concentrate the gaseous Hg so that when the

Hg reaches the analyzer it is a spatially small concentrated pulse of Hg. The analyzer uses a UV source that causes fluorescence of the Hg in the analyzer

174 chamber, which is then detected using a photo multiplier tube (Consult Tekran

2600 CVAFS manual for diagram).

The Tekran 2600 CVAFS is a high precision/accuracy instrument that is capable of running numerous samples quickly and accurately. The general working range for the machine is 0.5-100 ng/L, but the machine is capable of measuring 0.05-400 ng/L (samples above 100 ng/L should be diluted to avoid carryover). The machine uses about 35 mL of solution (use either 60 mL polypropylene centrifuge tubes (not recommended) or 40 mL borosilicate vials with Teflon lined lids (I-chem® brand highly recommended) and is an automated process with an analysis time of ~4 minutes per sample. Triplicate analysis using borosilicate glass has errors > 3%, while polypropylene tubes have errors of up to

25%. Though the machine is automated it should be checked periodically, once every 45 minutes or so. When something goes wrong with the machine it goes wrong fast and very badly.

Generalized instructions:

Reagents and Sample Preparation

1) ALWAYS use nitrile gloves, NEVER, NEVER use latex, vinyl, or

rubber gloves when working with Hg. Latex and rubber provide little to

no protection against Hg, you might as well not be wearing gloves if you

use latex or rubber.

175

2) Samples to be run must first be preserved with BrCl at least 1 hr before

analysis, preferably 1 day or more of oxidation time (Refer to EPA

method 1631 for BrCl recipe. It is recommended that powder reagents are

placed in a Hg-clean oven (ie. No Hg containing materials have ever been

placed in the oven before) @180oC overnight, a Hg-clean oven is very

important, even a trace amount of Hg will contaminate the BrCl.

Concentrated BrCl will breakdown Teflon so the reagent needs to be made

in a glass bottle with a Teflon® lined lid, the lid will need to be

periodically replaced since BrCl vapor will begin to breakdown Teflon®

(to cut down on the amount of lids that need replacing I suggest placing a

glass bead on top of the bottle while initially making the BrCl. It is

recommended that you use glass stir bars when making BrCl since the

Teflon stir bars will start to degrade if left in concentrated BrCl for too

long, > 1 day, if you use Teflon® try to remove the stir bar within 2-3

hours after making BrCl. Once BrCl is made in a bottle it should never

leave the fume hood it is made in. California regulations actually state

that concentrated BrCl needs to be diluted by at least 50% before it can be

transported anywhere.

3) BrCl is usually added at a concentration of 0.5% (v/v), but this depends on

the amount of organic carbon (usually DOC) in the sample. If there is a

lot of DOC the BrCl will be neutralized and fail to oxidize all of the Hg in

the sample. 0.5% BrCl is usually enough for most samples, samples high

in DOC should be fully oxidized with 1% (there will be a slight yellow

176

color to the sample when enough BrCl is added). Try not to exceed 2%

BrCl, BrCl tends to get contaminated with Hg after awhile but at low

concentrations (2% or lower) the added Hg is negligible and ~2% BrCl is

easily subtracted off as background.

4) Note: ALWAYS MAKE and use BrCl IN A FUME HOOD (this is for

any concentration of BrCl)!!! The quantities of chemicals are as

follows: 250 mL Trace-metal grade HCl, 2.7 g KBr, and 3.8 g of KBrO3.

Because BrCl can become contaminated with Hg quickly, it is

recommended that 250 mL batches be made instead of larger quantities.

Dissolve the KBr into the HCl and cap loosely (dissolution will take

upwards of an hour). Once KBr is completely dissolved add KBrO3 very

slowly, about a pin-sized amount at a time, more than this can lead to boil

over (reaction is very exothermic) and it also releases large amounts of

Br2, Cl2, and BrCl gas. Be very careful with BrCl, it is corrosive, toxic,

strong oxidizer, and potential carcinogen. BrCl can easily become

contaminated with atmospheric Hg, very obvious when it happens because

reagent blanks become very high. It is suggested that once the

concentrated BrCl is made that bottles of 10% and 0.5% BrCl be made for

cleaning, preserving, and reagent blank purposes. It is important that the

bottle of concentrated BrCl is opened as little as possible and for as small

of a time as possible. Warning: Never move your arm, especially bare

skin, over the top of the 10% or full strength BrCl bottles, you will get

an acid burn.

177

5) All samples should be stored and analyzed using borosilicate glass with

Teflon® lined lids. Tekran states that polypropylene centrifuge tubes are

sufficient for use, but years of lab work has shown that error increases

significantly compared to borosilicate and Hg loss to the plastic happens

quickly, more than 80% loss in 20 minutes unless >2 % BrCl is used for

storage.

6) Any sample dilution should be done into 0.5% BrCl to make sure that Hg

stays in solution and does not start adsorbing to the sides of the container.

Diluting samples in water is not recommended since it has been seen in

the lab that error increases from < 5% for n=3 to ~15%.

7) Any sample containing BrCl must be reacted with hydroxalamine

hydrochloride (HAH) before running in the Tekran 2600, failure to do so

can severely damage the Au traps and the machine, consult EPA method

1631 for the recipe. Remember to scale the amount of HAH with the

amount of BrCl. 5:2 ratio is what should be remembered, 0.5% BrCl

requires 0.2% HAH, 1.5% BrCl requires 0.6% HAH, etc. Just assume

0.2% v/v HAH of 40 mL, you can re-adjust the volume difference in the

spreadsheet.

Machine setup

1) The soda lime trap on the machine needs to be exchanged every 7-10 days.

If the machine has been sitting around for a week or more without being

178

used the soda lime trap needs to be replaced. To replace lift the top of the

machine and pull off the friction fittings for the trap. There is a piece of

5/16” Teflon tube on top of the sample box so that a bridge can be made

where the trap was. Empty out the trap into the trash and refill with new

soda lime. If the quartz tube is abnormally dirty it can either be scrubbed

out with water or submerged in 10% nitric acid for 15 minutes to clean.

Soda lime is very hydroscopic so make sure the quartz tube is completely

dry before adding soda lime and that the main soda lime bottle is open as

little as possible. Once the trap is refilled connect one end of the trap to

the latex tubing next to the machine and purge overnight with Ar. The

flow should be enough to where you see a bubble every second when the

opposite end is put in 10 mL of water.

2) Turn on Ar gas tank, Ar gas needs to be ultra high purity (5.0) and the tank

can never have contained He, VERY IMPORTANT, if He was ever used

in the take the residual He significantly decreases the life span of the

instrument lamp. Below the regulator on the Ar tank is a 3-way valve that

splits the gas stream to either the pyrolysis “P” Hg analyzer or the Tekran

“T”, make sure the valve is turned toward the “T”. Downstream of the

regulator the gas line is split with 3 T-unions so that the Ar stream can be

used to purge the stannous chloride and sodalime trap the night before

without turning on the instrument. When Ar tank gets to 200 psi a new

tank needs to be ordered, you will have enough gas for a minimum of 1

day, usually 2. It is absolutely necessary that Ar gas be flowing to the

179

Tekran BEFORE the machine is turned on, failure to do so will harm

the detector cuvette.

3) Power up the machine, all the machines are run through the same power

strip, all that needs to be done is turn on the power strip and all of the

machines should turn on at the same time. When the machine first starts

up make sure that the flow on the MFC (the left LCD display) is at

least at 15 mL/min (it will take about 1 min. for flow meter to

stabilize). Note: If the lamp light on the front of the machine does not

go off after about 5 minutes or turns on during the run stop the run

the lamp voltage is low and you will not get good or reliable results.

To correct this problem open the lamp wizard in the TK-MDS 2.0

software and follow the instructions.

4) The main reagent for the machine is 3% Stannous Chloride solution

(SnCl2). For this reagent there is a specific 2L Pyrex jar for the SnCl2

solution, the reagent is the only thing that goes in this bottle. For 2L of

reagent (good for about 10 hrs of analysis) measure out 60 g of SnCl2 and

add to the bottle (if crystals have a yellowish color or tint to them discard

the crystals and get a new batch). Using a graduated cylinder that is only

for use with the Tekran (the glassware is labeled), fill to 230mL with DI

water and then add 20 mL of trace metal grade HCl. Add this solution to

the 2 L bottle and swirl until the crystals are dissolved. Fill bottle to the

glass ring on the outside of the bottle (2 L mark) with DI water and shake

180

bottle to mix SnCl2. Purge the bottle with Ar gas for a minimum of 1 hr

(overnight preferable, there is a separate gas line for purging).

5) While SnCl2 is purging make the standards for the day. Standards used

are generally as follows: 0.5ng/L, 1ng/L, 5ng/L, 10ng/L, 25ng/L, 50ng/L,

and 100ng/L, 0.1ng/L can be made but usually not necessary, >100ng/L

(up to 400ng/L) can be done but not recommended due to carryover.

These standards should be made in 250mL bottles (I-Chem® brand). Once

standards are made they are stable for at least one year. These standards

are to be poured directly into specific vials in the Tekran sample rack (fill

to the base of the neck) and then add the requisite HAH. There is a

specific distilled water bottle either on the machine or in the fume hood

for the Tekran only, please use this bottle only for applications dealing

directly with the Tekran. To try and eliminate any unknown

contamination try to use this bottle so there is consistency. Run template

making standards and samples is at the end of this section.

6) Before running samples remove phase separator rod from housing (be

Very, Very careful the rod and housing are delicate and very expensive)

and clean the rod by rinsing with methanol and wiping with a Kimwipe.

Replace rod and tighten until snug. If sample wand is so dirty that

methanol will not clean it go to the troubleshooting section.

7) Put tension on all the lines of the pump. DO NOT overtighten, the tabs

should all be at only about 45o.

181

8) Check under the counter that the water reservoir is at least 2/3 full,

NEVER let the water reservoir get below 1/3 full a vapor lock situation

can occur resulting in the rinse station overflowing.

9) Check to make sure that there is plenty of room in the waste container; the

machine produces about 1 L of waste per hour of running.

10) Once steps 6 and 7 are done flip the switch on the pump unit from off to

local and make sure that the water flows evenly down the phase separator

wand and that the wash station fills up.

11) Start up the computer and open the Tekran software, TK-MDS 2.0. Create

either a new worksheet or use one that you have previously used. If you

are making a new worksheet, make sure you select Method 1631.

12) When adding samples to the worksheet there are several options that can

be done.

1) Clean- This only burns the Au traps to remove Hg

2) Wash- This runs water from the rinse station instead of a sample

3) Calblank- This tells the computer that a calibration blank has been

added

4) STDXX.X- This tells the computer a standard of XX.X concentration

was put in

5) “Sample”- This is the sample name you determine for your samples

Right clicking on the sample box will allow you to select certain types of

samples without typing them in by hand. Sample location can be either

typed in using A1-A12, B1-B12, C1-C12, and D1-D12 or by clicking on

182

the Location box and then clicking on the new box that shows up and

selecting the sample location.

13) Add into sample ID and sample position section this set of samples at the

beginning of the day.

Initial sample setup:

Sample 1-2: Clean

Sample 3-5: Wash

Sample 4: 5% Aqua Regia

Sample 5: Wash

Sample 6: 10% BrCl

Sample 7: Wash

Samples 8-11: Repeat steps 4-7

Sample 12-13: Wash

Sample 14-16: Calibration blanks

Sample 17-23: Standards (in increasing concentration)

Sample 24-25: Wash

Sample 26-X : Your samples

(it is suggested that after every 10-15 samples run a wash, standard (your

choice of which one), and wash be done to determine if there is any

appreciable floating of the machine signal). I personally do not do this

because the machine runs so consistent that standards curves barely

change over the course of 5 runs.

183

14) Before any sample with BrCl is run HAH must be added to neutralize

any free halogens, which can severely damage Au traps. For a 40 mL

sample preserved with 0.5% BrCl (v/v), 0.2% (v/v) of HAH needs to be

added (generally 10-15 minutes before sample is run). This ratio is

proportional, if 1% BrCl is used 0.4% HAH needs to be added or if 20 mL

of 0.5% BrCl sample is added to 20 mL DI water (40 mL total) then 0.1%

HAH is needed. If dilution is done it should always be done into 0.5%

BrCl and 0.2% HAH should be added. Remember: too much HAH is

alright, but too little can damage the machine.

15) Remove front panel of autosampler case and load samples, sample

positions start at the far back left and move forward. Remember that

Clean and Wash commands do not require a sample position so samples

will start with BrCl cleans. Replace front panel and switch the pump

switch from local to remote.

16) Since the MFC on the machine does not work you need to manually turn

up the flow on the MFC to 80mL/min. before running samples. When

running, the flow to the Phase separator should be set to between 400-500

mL/min. and the purging flow should be set to a level that allows for a

decent amount of bubbling (if purging flow is being used).

17) Click start run and determine that range of samples being run is what you

want. Note: The photomultiplier tube has been replaced and the new one

is very sensitive to the lights in the room. To have the best results and

184

most stable baseline it is necessary to turn off the lights once a sample run

is started.

18) While samples are running any samples added to the table will

automatically be added to the run so make sure that there are tubes in the

sample positions before the machine gets to them.

19) If a very concentrated sample goes through the machine, Peak area >

200,000 (the machine will tag the sample with OL and the peak will

plateau), immediately stop the run and start a new run using only sample

lines 3-4 to flush out the lines. Stopping the sample run occurs by

clicking on the arrow next to “Stop Run” in the tool bar and selecting

either Stop Run or Stop run after current row. It is preferable that

“Stop run after current row” is selected so that any residual sample is

flushed out of the lines. There are 4 tags that you will see during a

sample run: OL, OK, FB, and ED. OL stands for over limit meaning the

detector has been saturated and stopped collecting data. OK means the

sample ran fine. FB is forced bound which means the computer had

difficulty finding the end of the peak for integration and forced the peak

end. The FB tends to happen when there is a lot of noise in the lamp

which is a function of the stability inherent to the lamp, different lamps

give different noise. The last tag is ED which stands for edited, this

occurs when you go into the software window to look at the sample and

you adjust the peak bounds that the computer is using to integrate the

peak.

185

20) When analyzing a sample multiple times it is important to use the same

sample vial for all replicates, this cuts down on analytical error.

21) The program has the ability to make a standards curve for you and tell you

what your sample concentrations are, this rarely works. I have put on the

computer a template worksheet for your samples with functions worked

out to figure out the dilution factors and recalculating concentration of

your samples in solution and sediments.

Machine shutdown

1) When the day’s run is over always finish with 2-3 washes.

2) After cleaning swap the SnCl2 reagent with the 2 L bottle of DI water.

Flip the switch on the pump from remote to local and let run for 15

minutes. If this is not done then SnCl2 will start precipitating as SnO2 in

the lines and all of the lines will have to be replaced.

3) Once 15 minutes have passed flip the pump switch to off. THIS NEXT

STEP IS A MUST!!! Make sure you release the tension on all of the

tubes or irreparable damage can be done to the tubing and it needs to be

replaced (not cheap). Power off the machine and turn off the gas. After 7-

10 days of running the soda lime trap needs to be replaced following the

above protocol.

186

Troubleshooting

1) High integrated peak areas for samples, standards, or blanks:

The largest difficulty with ultra-trace Hg analysis is contamination of

reagents and samples. When the peak areas for standards or blanks

become abnormally large it is necessary to check that the reagent

bottles are clean. The BrCl or any chemical derived from the BrCl is

often the culprit for Hg contamination. BrCl will constantly draw in

Hg from the atmosphere due to its high oxidative potential.

Contamination of BrCl is not linear, generally the BrCl will be clean

for a long period of time and then it will suddenly become highly

contaminated.

2) Lamp light will not turn off or it turns on during a run: The lamp

for the CVAFS is set on a feedback loop where the machine will

increase the drive voltage to the lamp to keep intensity constant as the

lamp ages. After a time the voltage is driven out of range of the

machine specifications and needs to be manually adjusted back into

range. In the TK-MDS 2.0 software there is a lamp wizard module in

the pull down menu that will show you step by step how to adjust the

voltage. On top of the machine is a tool for readjusting the voltage, if

it is gone a razor blade works equally well, just be careful.

3) The standards curve is linear but peak values are significantly

lower than previous curves: Two things might be causing this: 1) the

187

sensitivity was adjusted or 2) the UV lamp is wearing out. The

sensitivity dial of the machine should be set to around 6 and should not

need adjusting unless a new lamp is put in. If the sensitivity dial is not

changed then the lamp is wearing out. The main ways of determining

if the lamp is wearing out are 1) background noise is increasing and 2)

the signal has dropped significantly without the lamp light coming on.

If the lamp needs to be replaced follow the steps outlined in the Tekran

manual.

4) There is excessive noise in the baseline but peak area is still

correct: This issue is generally only a problem when trying to detect

< 0.5 ng/L samples. There are two main causes for this problem: 1)

the lamp is noisy and 2) the detector cuvette needs cleaning (this is

very rare). In general the problem is a noisy lamp and cannot be fixed

without replacing the lamp or increasing the sensitivity. Replacing the

lamp is fairly time consuming and expensive so this is generally not

recommended unless you truly need the really low detection limit. Re-

adjusting the sensitivity requires recalibration after the adjustment is

made. To change the sensitivity adjust the sensitivity dial and then

adjust the offset dial (this is the baseline value) until it reads 0.1 V on

the right LCD panel. If this does not work take out the detector

cuvette and clean it following the steps in the Tekran manual.

5) There is excessive noise in the baseline and the peak areas are

inconsistent with concentrations: When peak area becomes very

188

noisy, voltage changes of 0.05 (right LCD panel) every half second,

and peak area is inconsistent (ie. Higher concentration samples have

lower peaks than lower concentration samples) there is probably a

break in the detector cuvette. This problem is extremely rare but does

happen. A break in the quartz cuvette that the sample is detected in

generally occurs if the gas pressure exceeds 60 psi, the cuvette was

damaged during opening and shutting of cuvette chamber, or rough

handling of the unit. A break in the cuvette affects analysis by two

major ways: 1) gas flow is inconsistent through the cuvette (generally

vents to atmosphere) and 2) Hg in the gas flow is lost to the

atmosphere instead of being detected. If a broken cuvette is a

possibility for bad data, remove the detector cuvette and inspect

following the steps in the Tekran manual.

6) All samples give the same peak area: This problem often occurs

when the heating coils are not reaching the correct temperature for

releasing all of the Hg amalgamed to the Au traps. The result of this is

that the peak will be retarded in the time it takes to reach the detector

cuvette and the peak area will be nearly the same for every sample

regardless of the concentration. Retardation of the samples can be so

severe that the sample is not completely through the detector cuvette

before the analyzer stops recording. Click on the peak button in the

tool bar for the various samples that are giving the same peak area,

usually around 5,000. When the peak shows up in the right window

189

make sure the time for the peak is correct compared to samples ran

before the problem manifested. If the peak maximum of questionable

samples is retarded more than 5 seconds (usually closer to 20) then the

heating coils need to be replaced. When replacing the heating coils

make sure that the coils are separated out an not touching each other or

arcing of the electric current will occur resulting in inefficient heating.

Heating coils need to be replaced every couple of years. If this does

not completely fix the problem then the issue is probably the bridge

rectifier. This is an extremely rare problem for Tekran analyzers but

has been an issue for this machine. The bridge rectifier controls how

much voltage is being directed to each heating coil and if it is not

working properly the heating coils will not reach proper temperature.

Once the problem is fixed it is necessary to run several “Cleans” on

the machine, usually 4-5, to remove all the built up Hg that was not

removed from previous sample runs.

7) The values for washes are high: This occurs when too much

material has precipitated in the tubing, usually after the sample and

SnCl2 have mixed, or new tubing is put on. New tubing needs to

periodically be put on to avoid permanent kinking of the tubing or

splits. When new tubing is put on there is a lot of residual Hg left

from manufacturing that needs to be cleaned off. This is generally

done by using a series of 5% aqua regia and 10% BrCl until the wash

levels drop and level out. The 5% aqua regia is for removing the large

190

non-Hg material on the tubing while the 10% BrCl is used to clean

only the Hg off the lines. Both are needed to properly clean tubing.

An alternative is to run a large number of 0.5% reagent blanks to

slowly clean out the system and bring it down to a stable background

level. If the problem is dirty reagent lines the best way to clean the

lines is to remove them and run them through the bath sonicator with

or without detergent. This technique works particularly well for the

Teflon® tubing. Once the tubing is clean put it back on the machine

and clean it like new tubing.

8) Reagents are not flowing evenly down phase separator or there is

material stuck to the wand: These problems tend to happen when

samples are not properly digested and organics precipitate along the

wand. When the phase separator becomes very dirty placing the rod

(sometimes the housing as well) in 20% nitric acid/20% sulfuric acid

at 120oC overnight will be necessary for cleaning. Never sonicate

phase separator or housing to remove material, the glass will

break.

191

Literature Cited

1) Parker, J. L.; Bloom, N. S., Preservation and Storage Techniques for Low- Level Aqueous Mercury Speciation. Science of the Total Environment 2005, 337, 253-263. 2) United States Environmental Protection Agency, Method 1631, Revision E: Mercury in Water by Oxidation, Purge and Trap, and Cold Vapor Atomic Fluorescence Spectrometry. In U.S. EPA Technical Manuals 2002.

192

Appendix II

Theory, Design, and Operation Protocol for

Pyrolysis Mercury Analyzer

Machine Theory:

The pyrolysis Hg analyzer works on the concept that if Hg is heated to

o 0 >700 C, all the Hg within a sample will go through a pyrolytic reduction to Hg and volatilize in a gaseous state. This reduction step happens regardless of both the speciation of the Hg and the sample matrix. The system is simple in design

o consisting of a furnace set at a temperature of 750-800 C, a catalyst within the furnace to oxidize organic material, soda lime trap to trap water vapor, and two

Au traps that amalgam the Hg before being detected by a cold vapor atomic fluorescence spectrometry (CVAFS) detector. This machine is designed to analyze solid samples without any sample preparation and to minor extent water samples (if water analysis is being done the Tekran 2600 CVAFS is a better choice). There are two Au traps (Au coated sand and 24k Au beads) to amalgam the elemental Hg created in the furnace after which the Hg is desorbed by heating the Au traps in an Ar gas stream to carry the Hg to the CVAFS detector. O2, CO2,

N2, and organics absorb within the same UV range that the CVAFS uses for Hg

193 fluorescence requiring the removal of organics from the gas stream. Removal of organics is carried out by decomposition of samples in an O2 rich environment after which the gas stream is run through a catalyst that degrades the organics to

CO2, NOx, SO2, and H2O. To keep these gases from reaching the CVAFS two 3- way valves are installed, the first valve switches the gas stream from O2 rich to Ar and the second valve sends gas to the second Au/detector or vents combustion products to the atmosphere. After the Hg is amalgamed to the Au traps it is heated by heating coils and is carried to the CVAFS detector with an Ar gas stream. The CVAFS detector sends the signal out as a voltage differential that is converted from an analog signal to a digital signal by way of a 22-bit converter.

This signal goes to the computer where the data can be analyzed one of two ways:

1) integration of the peak area (PA) or 2) total peak height (PH). The integration of peak area is the most accurate way of dealing with data (especially for broad peaks) but it is also the more difficult and time consuming way of dealing with data. The better way is to use peak height. Since the peaks are very narrow, ~4 seconds wide, the peak height works as well as peak area. The CVAFS detector is extremely sensitive, for example when the sensitivity set almost to 0.5 out of 10 a sample containing 3 ng of Hg is easily detectable. This system is not fully automated, but straight forward to run. Because of the automation and better detection limit of the Tekran 2600 it is suggested that the Tekran 2600 is used for aqueous samples.

194

System Schematic and Components:

Fig. 1: Schematic drawing of pyrolysis Hg analyzer and components. All valves are stainless steel.

Note: Drawing not to scale

Component Specifications: This section is divided into numerous sections. The last section is an itemized list of all the components used in the machine with appropriate sizes and where they were purchased. The sections prior to the components list discuss why the specific materials were used in construction of the pyrolysis analyzer.

195

Tubing: All tubing for the pyrolysis Hg analyzer is FEP Teflon tubing. The most common polymer of Teflon is the PTFE polymer. This polymer has a lower friction coefficient that FEP, but has a higher gas permeability (for a summary of polymer properties

(http://www.zeusinc.com/UserFiles/zeusinc/Documents/Zeus_SOP_english.pdf).

Because the pyrolysis analyzer has only gas flowing through the lines a lower friction coefficient is less important than lower gas permeability. There are only two different sizes of tubing used in the analyzer, 1/8” and 5/16” (both inner diameter). The 1/8” tubing is identical to that used in the Tekran 2600 CVAFS detector and is easy to purchase. The 5/16” FEP tubing is considered an “Odd” size but it is necessary in order to have air-tight seals to both the soda lime traps and the Au traps. The 5/16” sized tubing is also used in connecting the FEP tubing to the furnace so that diameter of the tubing has a more gradual transition to 1/8” tubing for high temperature gases from the furnace to travel through.

Tubing Connections: The connectors used in the portion of the system not in the furnace are from two companies, Swagelok and Tekran. The Tekran fittings came with the Au traps and the soda lime trap and should not need changing or replacing. The Tekran fittings are composed of friction Teflon® fittings that are pre-assembled on the components. The fittings on the Au trap are heat shrink Teflon® and need to be trimmed flush with the end of the quartz tube they are on in order to provide a solid air-tight seal when the 5/16” FEP tubing is slid over the heat shrink tubing. The ends of the soda lime trap also slide into the

196

5/16” FEP tubing to provide a solid seal that is easily removable and does not need Teflon® tape or any gluing material.

The other type of tube connections are from Swagelok Inc. Two different types of materials are used for the Swagelock fittings (Brass [B] and Stainless

Steel [SS]). The brass fittings are all used upstream of the sample chamber or the incoming gas lines. The reason to use brass is because brass amalgams elemental

Hg and is used to clean Hg out of the incoming gas stream. Because brass will amalgam elemental Hg it is imperative that brass connections are not used along the flow path for the gas stream downstream of the furnace. Contrary to brass, stainless steel does not amalgam Hg. All connections downstream of the sample requiring Swagelok fittings are stainless steel. The size of Swagelok fittings are all based on a multiple of a 1/16” sized fitting. A size 200 Swagelok fitting is the same as 2/16” (1/8”) while a size 500 is a 5/16” sized tube. All fittings used are either size 200 or 500 in either [B] or [SS]. When FEP tubing needs to be replaced new ferrules for the Swagelok fittings are needed. Sometimes ferrule sets can be reused but the metal usually deforms too much to be reused during initial installation. These ferrules are made of a front and back portion and can be purchased in sets of 10 (denoted as an arbor) from Swagelok. Because size 500 is an uncommon size delivery takes 1-2 weeks from when the order is placed so make sure there are plenty of spares. Make sure that the correct material for the ferrule is used when replacing tubing (ie. No brass ferrules along the sample flow path).

197

There are several types of Swagelok fittings used for the pyrolysis analyzer. The three main types of fittings are male unions, 3-way valves, and reducing unions. The male unions have a male threaded end with a normal tube connector. The male unions are mostly size 200 [B] for all the gas fittings coming from the gas tanks to the analytical system, but there is a single size 500 [SS] that connects the outlet of the furnace chamber to the FEP tubing. The 3-way valves

(size 200 [SS]) used in the system are used for several purposes: 1) change gas flow from O2 rich to Ar gas, 2) vent combustion products to the atmosphere so that the gases do not pass through the detector, and 3) protect the pure Au trap from combustion gases and water vapor. Because Swagelok does not make size

500 3-way valves the size 200 were used with reducing unions to change the size of the tubing coming from the furnace from size 500 to size 200.

Flowmeters: There are two different flowmeters for the pyrolysis system.

The first flowmeter is a ball flowmeter that controls the flow of the O2 rich gas to the furnace to aid in combustion of organic samples. The second flowmeter is built into the Tekran 2500 CVAFS detector and it is used for the Ar gas stream for analysis. The consistency of the gas flow rate for the furnace portion of the system is not critical since the loading time for Hg onto the first Au trap is sufficient to fully load the trap. The consistency of the Ar gas flow is critical for consistent analysis. The Ar gas flow is controlled by a mass flow controller

(MFC) built into the Tekran detector and it constantly monitors the flow rate and adjusts the flow rate to keep the flow within a tenth of a mL/min flow.

198

Gold traps, heating coils, and variacs: The Au traps used in the system are the exact same ones used in the Tekran 2600 CVAFS analyzer. The Au traps are highly susceptible to water condensation. When water condenses on the traps

Hg does not amalgam to the Au which allows some of the Hg to go directly through the trap either venting to the atmosphere or through the second Au trap directly to the detector. The first trap is Au coated sand while the second trap is

24K Au. The cost of the second trap is about 5x that of the Au coated sand so it is very important that the 3-way valve safeguard put into the gas stream is used.

The Au traps should not need replacing for several years unless a sample with un- neutralized acids is introduced into the system (this should not happen since all solution analysis should be done with the Tekran 2600). The soda lime trap is placed in the system mainly to absorb water vapor and some acidic fumes, but too much acid in the sample will damage the Au trap which is seen by a discoloration of the Au coated sand. Around the Au traps are heating coils that are used in the

Tekran 2600 CVAFS as well. These heating coils are connected to variacs by an extension cord spliced onto alligator clips. The variacs adjust the incoming voltage from 110V to ~10V used by the heating coils. The variacs also change

o the current from AC to DC to allow the heating coils to warm to > 500 C.

Because the first Au trap is made of Au coated sand the trap cannot handle a consistently high temperature which is why a computer fan is installed to cool down the trap quickly after heating.

Chrontrol and A/D converter: The Chrontrol device is used to control the timing of the variacs and computer fan. The Chrontrol was purchased from

199

ChronTrol Corporation (www.chrontrol.com). This device consists of a control box with 4 outlets in the back. The box plugs into the wall and the two variacs and fan plug into the Chrontrol box. The box is programmed to turn on each circuit at specified times for predetermined lengths of time. The programs saved in the box are lost whenever the box is unplugged without the backup battery installed or a power outage lasting more than the length of the battery (~4 hours).

Due to the complexity of re-programming the Chrontrol box a separate section on programming is below.

The CVAFS detector outputs voltage as a voltage differential which is an analog signal. This signal cannot be directly understood by the computer. To convert the data from an analog to a digital signal the analog output is connected to a 22-bit analog-digital converter. This converter, OMB-DAQ-54, was purchased from Omega Engineering, Inc. (www.omega.com). The converter can handle up to 10 different channels of data but for the pyrolysis analyzer only one channel of data is converted and sent to the computer via a USB connector. Data acquisition software came with the converter and is on a CD in the laboratory.

Furnace Components and Catalyst: All components and fittings dealing with the furnace, inside and outside, were made by the Stanford Machine Shop in the basement of the Varian Physics building. The inner diameter of the stainless steel tube inside the furnace is 1” with variable outer diameters. The main furnace unit consists of a stainless steel block with a 1” hole cut for the sample and catalyst.

This block has two stainless steel pipes that are welded to the block and exit the furnace through two holes cut in the furnace door. In order to not melt the FEP

200 tubing and allow manipulation of the brass sample door (the stainless steel pipe

o exits the furnace at ~750 C) a ceramic spacer is in the sample line to insulate the

FEP connection and brass. This ceramic goes by the trade name Macor and it is a machinable ceramic allowing it to be threaded with ordinary machine shop machines unlike most other ceramics. The ceramic is threaded to go inside of the threaded furnace pipe. If the Macor was on the outside, the steel would expand during heating more than the ceramic causing it to crack. With the Macor inside it will expand and create and air-tight seal as it presses upon the steel. The sample chamber door and O-ring are basic vacuum system parts that can be purchased at Stanford Physics Store. Sample boats for analysis are pure Ni manufactured by the Stanford Machine Shop. For replacements take an extra sample boat to the machine shop to use as a template.

The catalyst tube used in the furnace is a 4” long stainless steel tube with

~1” OD (slightly smaller to fit in furnace tube) and ~3/4” ID. The tube has two end plugs with holes (~1/4” thick) with 3 thin stainless steel meshes to keep the catalyst materials from mixing. The catalyst material is taken from the design of the catalyst for the Carlo Erba C/N/S analyzer in the Stanford Environmental

Measurements lab. The catalyst material degrades complex organics to CO2, SO2, and NOx which will not affect the Au coated sand trap. The catalyst is made of two different chemicals: 1) Chromium (III) Oxide (0.5-1.7mm) and 2) Cobalt (II,

III) Oxide (20-50 mesh). These catalysts are purchased from Costech Analytical

(www.costechanalytical.com) and has been seen in the laboratory to be a far better than a catalyst used in the Leco AMA254 direct combustion mercury

201 analyzer. Catalysts used in commercial pyrolysis systems act as a sink for organics from samples, absorbing rather than completely oxidizing the organics.

By absorbing the organics rather than degrading them the catalysts need to be exchanged regularly. The catalyst made for this system should not need changing for a very long time. In the manual for the Carlo Erba C/N/S analyzer the catalyst needs to be exchanged after ~300 samples. The manual for the Erba machine

o states that the catalyst is held at 1050 C so that the sample boats melt causing the destruction of the sample. The melting of the sample boat passivates the catalyst but since the catalyst in the pyrolysis analyzer only encounters Hg and organics this passivation will not occur. The catalyst is made with packing 1.7” of the

Cobalt (II, III) Oxide on the downstream side of the catalyst tube, placing one of the meshes between the two catalysts and then filling with 2.3” of Chromium (III) on the upstream side of the catalyst tube.

Component Settings:

1) The regulators on the gas tanks used for the system should be set at 30

psi, do not let the Ar regulator get above 60 psi or the sample cuvette

in the detector can break.

2) The ball flowmeter should be set to 0.15 SCFH (Standard Cubic Feet

per Hour), which is equivalent to 70 mL/min.

3) The MFC in the Tekran 2500 CVAFS should be set to 100 mL/min.

o 4) The temperature of the furnace should be kept between 750 C and

o 800 C, this should be a setting of just above 3 on the furnace dial. Do

202

o not exceed 800 C because the Macor ceramic has a maximum working

o temperature of ~850 C and should not be exceeded.

5) The variacs that control the heating coils should be set at 10V (this can

be adjusted) and should never exceed 12V.

6) Sensitivity and offset on the detector are adjustable. Sensitivity

adjusts how much amplification of the fluorescence signal there is, this

should be set extremely low for samples in the ppm concentration

range. The offset only controls the baseline the detector is outputting.

Once the sensitivity is set use the offset to lower the output voltage to

~0.01 volts. Any change to the sensitivity requires recalibration.

7) The Chrontrol sample cycle should be 1 minute for heating coil 1, 15

seconds all circuits off, then 1 minute for heating coil 2 and the

cooling fan (instructions for programming in next section).

ChronTrol programming scheme:

The ChronTrol unit works on the premise of one program controlling other programs. A single program can be made to run a specific circuit (the outlets on the back) for a specific time. But in order to have a circuit wait for a specific amount of time and then turning on for an interval a program has to be made for the waiting time to control the program that turns on the circuit for a certain interval. The programming protocol below is designed for the pyrolysis unit but can be easily changed to have circuits run for different amount of times.

For more detailed instructions on the function of each button pressed in the programming scheme below, check the manual (programming the time and date,

203 etc.). With regards to the programming, anything within “ ” is a key stroke on the front of the ChronTrol unit. If you need to delete a program or made a mistake that is not related to the interval time programmed it is easier to pull the plug for the ChronTrol for a few minutes to blank out the system than trying to figure out how to delete a circuit, program, etc.

1) To program ‘Program 1’ which is responsible for controlling all the

other programs downstream press “Enter”, “1”, “Program”, “2”,

“Program”, “3”, “Interval”, “2”, “Second”, “30”, then “Enter”. The

box should show a 0 0 1 when Enter is pushed. This program means

that the box will run Programs 2 and 3 for an interval of 2 minutes and

30 seconds. When finished press “Time” to exit from any menu at

any time.

2) To program ‘Program 2’ press “Enter”, “2”, “Circuit”, “1”, “Interval”,

“1”, then “Enter”. This program means that circuit 1, the first Au trap

heating coil, will turn on for 1 minute.

3) To program ‘Program 3’ press “Enter”, “3”, “Program”, “4”, “Off”,

“On”, “Interval”, “1”, “Second”, “15”, then “Enter”. This program

means that there is a delay of 1 minute and 15 seconds before Program

4 is turned on.

4) To program ‘Program 4’ press “Enter”, “4”, “Circuit”, “2”, “Circuit”,

“3”, “Interval”, “1”, then “Enter”. This program will turn on circuit 2

(the second Au trap heating coil) and circuit 3 (the cooling fan for trap

1) for 1 minute.

204

To run a program type “Program”, “X”, then “On”.

System components:

There are several parts of the system that should not be disturbed. There are several settings that if disturbed will not only require re-calibration but will also hurt reproducibility. If you change any of these things contact the tech in charge of the machine so they can reset things and re-calibrate the machine. Do not change the following things: 1) the gas setting on the flowmeter, 2) the temperature setting on the furnace, 3) any settings on the Chrontrol controller, 4) either of the variacs in back, and 5) any knobs on the CVAFS detector. The parts of the system that you will be using are the front cover of the furnace pipe, the three gas valves before and after the furnace, and the Chrontrol controller. The

Chrontrol controller is a timer box set specifically to turn on and off the heating elements for the Au traps at pre-determined intervals. If there is a power outage the Chrontrol will need to be re-programmed, contact the tech so that they can re- program the Chrontrol. The soda lime trap is present to trap any excess water from the sample so that it does not condense on the Au traps and it only needs to be exchanged every 7-10 days of running at most. Software for the machine will include the use of 3 programs: 1) pDaqview, PeakFit, and Microsoft Excel.

System start-up and operation:

The stainless steel sample chamber in the furnace should not be thermally shocked during heating up for an analysis run (this can lead to cracks in the chamber). To avoid heating the furnace too quickly do not adjust the front

205 temperature setting (just above the number 3 on the dial), this setting is for a

o holding 750-800 C. When turned on it takes about 3 hours to get up to temperature and the CVAFS takes 1 hour for warm up so either start the furnace when you first arrive or turn it on the night before. Note: The ceramics do a

o good job at keeping down the heat, but the brass cover plate is still at ~120 C, so wear nylon gloves under nitrile. There is a leather glove by the furnace that you can use to keep from burning yourself.

1) Take the front cap of the brass tube off of the furnace of the inlet before

turning on the furnace.

o 2) Flip on the furnace and wait until the temperature reaches at least 700 C

o (analysis can start when 700 C is reached) before running any samples

(this should take about 3 hours).

3) There are two different gas tanks needed for running the analyzer (20% O2

tank balanced with either air or nitrogen and a UHP Ar tank). Note:

Higher O2 concentrations can be used, actually preferred, check with

EH&S prior to using higher O2 concentrations. The Ar gas tank is also

used for the Tekran 2600 CVAFS detector and has a 3-way valve at the

gas regulator. For the pyrolysis detector set the 3-way valve to “P”, if you

are running the Tekran set the valve to “T”. Make sure the Ar gas is

flowing before you turn on the CVAFS detector, failure to do so can

damage the UV lamp and the sample cuvette. Then turn on the CVAFS

detector, the gas must be on while the detector is on. Let the detector

warm up for about an hour.

206

4) Turn on the computer. Start up the three programs that will be used: 1)

pDaqview, 2) Peakfit, and 3) Excel. pDaqview is used to acquire the data,

PeakFit gives the peak height, and Excel is for data storage and

manipulation. In pDaqview you will only use three buttons in the tool bar.

The first is a set of three folders on the button, this is where you choose

where to save the data and the file name (Note: file number does not

automatically update so a new data label must be added for each sample).

There are several viewers to see real-time data (the scrolling viewer is the

most useful) and the triangle with the gun is clicked to start recording data

(aka triggering the data collection).

5) When the furnace is up to temperature it is important to do an initial burn

of the sample boats being used for the day since they collect Hg from the

atmosphere at detectable levels after about 24 hours of sitting on the

counter top.

6) Since the goal is to completely clean the Ni sample boats all the boats

being used for the day can be burned before burning the Au traps. To

clean the boats insert them into the sample chamber, close the front, and

turn on the gas valve leading to the front cover. Note: make sure the first

3-way valve is pointing towards the furnace to allow loading onto the first

Au trap and that the second 3-way valve is pointing away from the second

Au trap so the gas is venting to the room.

7) After 3 minutes of the gas being on remove the Ni sample boat and put in

the next one; repeat until all boats are burned. Note: Ni boats are at

207

o 750 C when they are removed from furnace, very dangerous. Make

sure they are placed directly on the aluminum cooling rack to cool. If

you drop a sample boat resist the inclination to catch it in the air and

pick it up off the ground quickly using the tongs only.

8) To run the collected Hg (this is mainly to clean the Au traps) turn off the

first gas valve, turn the two 3-way valves in the opposite direction (let the

gas flow level off to ~100 mL/min., the left LCD display on the CVAFS

detector), start recording in pDaqView, and then press following buttons

in order “Program” “1” “On” to start burning the Au traps. If you want to

see real time data turn on the scroll monitor in pDaqView. Do not be

surprised if multiple peaks are seen or there is excessive noise this is

generally caused by water that condensed on the Au traps while the

machine was turned off. These multiple peaks should only be present

during the first burn of the day.

Standards Curve:

The CVAFS and pyrolysis system is very consistent from day to day as long as no settings are changed. It is not necessary to make a new standards curve with each analysis run. An initial standards curve was made by using a Fish certified reference material (DORM-3, 382 ng/g) from the Canadian Research Council; this standard is chosen due to its high homogeneity compared to soil samples.

This standards curve is saved on the desktop of the computer and can be used as long as no settings are changed. Every time a setting is changed (gas flow, sensitivity, offset, heating coil voltage, etc.) a new standards curve is needed; this

208 is true even if a setting is changed and then changed back. Currently, the fish standards curve goes from 0.01 to 0.1 V for the peak height but the linearity of the detector goes up to ~1.32 V after which the detector saturates. The analyzer linearity is consistent no matter what the sensitivity setting is set at. The fish

CRM is only 382 ppb in concentration. To have a working range of the analyzer that can handle samples in the tens to hundreds of ppm the CRM part of the standards curve needs to cover the lower part of the full range of the detector.

The reason the fish CRM is not extended farther is that above 0.05g of fish sample the pressure from the combustion gases causes a loss of gas through the sample chamber door. At the beginning of a sample run a Fish CRM of 0.02g should be run to make sure the standards curve still holds. The sample recovery should be 85-115% in order for the standards curve to still be considered good.

Note: It is important to burn the Ni boats prior to first analysis and to let the boats reach room temperature before putting in a sample to keep volatile Hg species from being lost.

Running a Sample:

1) Weigh out a sample into the Ni sample boats (make sure sample boats

are at or below room temperature). Sample amount should be between

0.001 and 0.1g. For fish samples 0.04g should be the maximum

sample size due to pressurization of the system caused by combustion

gases. Note: Make sure the sample boats are at room temperature or

lower before adding a sample to prevent volatile Hg species from

being lost.

209

2) Make sure that the first gas valve is turned off, the first 3-way valve is

turned to the furnace, and the second 3-way valve vents to the

atmosphere.

3) Remove brass cover, insert sample, and close (quickly). Put the clamp

for the cover back on and start the timer.

4) Wait for 2 minutes then turn on the gas that goes into cover plate. If

the gas is started immediately the decomposition products will pass too

quickly through the catalyst and will not properly breakdown. The

pressure caused by the decomposition/combustion products will

slowly push most of the gasses released through the catalyst for proper

breakdown.

5) Wait 3 minutes more then turn off the valves in the following order: 1)

the valve to the brass cover plate, 2) the first 3-way valve so the Ar gas

is flowing, and 3) the second 3-way valve so the gas runs through the

second Au trap instead of venting to the atmosphere.

6) Once the Ar gas flow stabilizes at ~100 mL/min (the left LCD display

on the detector), start recording with pDaqView, and press “Program”

“1” “On” on the Chrontrol to run the sample. The complete sample

cycle will take 2 minutes 15 seconds to complete. The cycle is 1

minute heating of Au trap one, 15 seconds off, and then 1 minute

heating of the second trap while running a cooling fan on Au trap one.

Even though the sample peak will go through the detector at ~40

210

seconds for the second Au trap cycle do not turn off the Ar gas flow

until the Chrontrol turns off.

7) When sample cycle is done turn the 3-way valves the opposite way

and remove the sample boat from the furnace. Place the sample boat

on the Al cooling rack, rinse when the boat cools down, and wipe

clean with a paper towel or Kimwipe. Once the boat is clean, dry the

boat by putting it on top of the furnace, and then place on the

aluminum rack to allow it to come back to temperature.

Data analysis:

1) After a sample is run open the file in PeakFit.

2) When the file opens make sure that there is a single peak present in the

spectra. Numerous peaks mean that there were variable release times

for the Hg making the data useless. Numerous peaks are generally

created by condensation of water on the Au trap or organics coating

the Au traps. These multiple peaks usually happen during the first

burn of the Au traps of the day as collected water from the atmosphere

is driven off.

3) The first page that opens up gives information about the spectra (range

along x, range of y, minimum signal, maximum signal, etc.) We are

only interested in the maximum signal or the range.

4) Type in the y-max value for the peak height into the Excel spreadsheet

to calculate concentration. The baseline barely shifts throughout the

day, less than 0.1% shift on average. If offset is changed or you feel

211

that the baseline shift is an issue you can use y-range instead of y-max

to negate the baseline or changes in the offset.

Shutdown:

1) If you plan on running the system the next day the furnace can be left

on overnight.

2) If the system is being completely turned off for the day the shutdown

procedure is easy. Take the brass cover off, turn off the furnace, make

sure the 3-way valves are pointed to the furnace and atmosphere, turn

off the detector, turn off the computer (monitor as well), and finally

close the gas tanks.

Troubleshooting:

The biggest issue for the machine is humidity. Running the machine when it is raining will result in inconsistent results because of constant water condensation on the Au traps. To rectify this problem the Au traps need to be

o kept at ~110 C. Regretfully this problem has yet to be fixed for several reasons.

It is important to not change the variac voltage settings so that consistent results

o will be achieved. In order to maintain ~110 C a second set of variacs with separate heating coils would need to be installed. To install the heating coils properly it would be necessary to intertwine the two sets of heating coils without letting them touch to keep arcing from occurring.

212

1) Multiple peaks are seen when the first burn of Au traps is

done: This is the result of water that has condensed on the Au

traps. When the first Au trap is burned the water and Hg are

initially released at the same time, but all of the water is released

before all of the Hg. When the water and Au get to the second Au

trap the water inhibits Hg amalgamation so some of the Hg gets

through and goes to the detector. Since Hg is still arriving at the

second Au trap after the water has already condensed the Hg can

amalgam to the Au and be released when the second trap is burned.

This water inhibition results in at least 2 peaks sometimes 3 peaks

since water can also inhibit the amalgamed Hg from releasing at a

consistent time. The first burn of the day should correct this

problem unless a large number of samples with a lot of water are

run.

2) Multiple peaks are seen during running a highly organic rich

sample or several organic rich samples: This problem is similar

to the water issue. If the gas flow to the furnace is turned on too

high or the gas was turned on too early the sample gases will not

have enough time to interact with the catalyst and will pass

through to the first Au trap. When the first Au trap burns some of

the organics will move to the second Au trap and inhibit

amalgamation. Generally if this is happening fairly regularly the

glass wool in the first Au trap will darken as the organics trapped

213

in the wool burn as the heating coil turns on. When you see this

happening, immediately replace the glass wool so that the burned

organics don’t absorb Hg (more on this later). To clean out the

organics on the Au traps a series of 4 burns are needed. First take

out any sample boat remaining in furnace before doing the

cleaning burns. There are two types of burns required to clean the

Au traps: 1) with O2 and 2) with Ar. Turn the 3-way valves so that

the furnace goes to the first Au trap and the other goes to the

detector (it is important that gas is running through the Au traps as

they heat up). Do a burn of the traps (monitor the peaks in

pDaqView to make sure when Hg is hitting the detector, just run

without recording) with O2 to combust organics from the Au traps.

Turn the gas to Ar and burn again to drive off Hg that might be

coated by the organics. Repeat the two burns alternating between

O2 and Ar until a low stable baseline is reached and a single peak

is seen at the appropriate time.

3) The CVAFS detector signal is not returning to baseline: This is

almost always an organic build up problem. Refer to step 2 for

how to fix.

4) Samples were running extremely well then all of the sudden

replicates were coming in low: This is usually due to organics

depositing in the lines, or possibly water condensation, that is

absorbing Hg in the gas stream. To rectify replace all the glass

214

wool (this is where the organics tend to collect) and then burn the

traps according to step 2.

5) Background has suddenly become noisy and sample peak

height is extremely low: This usually is an issue with the cuvette

in the CVAFS detector being either dirty or cracked. Remove the

cuvette following the Tekran manual and inspect. If cleaning is

necessary follow the Tekran manual for cleaning.

6) A replicate suddenly has a massive peak compared to previous

runs: This happens for several reasons: 1) heterogeneity in the

sample, 2) a buildup of Hg within the tubing of the system that

suddenly releases a large ‘Plug’ of Hg, or 3) the sample boat was

not burned clean earlier in the day. Sample heterogeneity should

be the first suspect when this occurs making it important to run

another sample to make sure this is not the case. If it is a buildup

of Hg the course of action is to remove the sample boat and run the

system as a blank (this might take a few runs).

7) A sample with a large amount of Hg (> 300 ng) results in a

peak height lower than a sample an order of magnitude lower

in Hg: This result is due to self-absorption of the fluorescence

signal. When a large concentrated pulse of Hg enters the cuvette

the UV light from the fluorescing Hg getting re-absorbed by the

rest of the Hg resulting in a lower signal seen by the detector.

Generally the amount of Hg in the cuvette is so low that self-

215

absorption is not detectable except for when a very concentrated

‘Plug’ of Hg enters the cuvette. The main ways of fixing this are

three-fold: 1) put smaller amount of sample into the furnace, 2)

dilute the sample with silica powder, or 3) lower the gas flow on

the MFC. A lower gas flow will spread out the ‘Plug’ of Hg

resulting in a widening of the peak and less or no self-absorption.

Note: If the gas flow rate is changed a new standards curve needs

to be made.

8) Inconsistent results between different types of material: Every

material releases Hg differently. For each type of material a

different standards curve is necessary. If the sample is soil, use a

standards curve constructed using the soil CRM, fish needs a fish

CRM, etc. This is necessary for each main type of material: Soil,

fish, plant, coal (activated carbon) all need their own standards

curves. This issue also occurs on commercially available

machines.

9) After a long time running the sample reproducibility becomes

erratic: This is most likely due to ash build up within the system.

Let the furnace come down to room temperature before doing

this. Take the furnace part of the system apart, remember to

remove the catalyst tube, and clean out the ash from the stainless

steel tube.

216

Appendix III

XAS Protocols

I: General Beam Run Protocol

Beam runs are different for every person; this protocol is for general running of the beamline. Please tailor to your specific needs. This protocol is written from the viewpoint of doing Hg LIII EXAFS using the Ge-30 solid state detector

(Fluorescence mode) on SSRL beamline 11-2. This protocol will also work for all other XAS beamlines at SSRL, though there is a slight difference for windowing the solid state detector if it has analog or digital settings. After the main protocol there will be a short bit on Transmission mode.

Beam Run Start-up

1) The first things that need to be done when arriving (besides logging onto

the computer and setting up where data will be saved) is to window the Ge

detector, calibrate the monochromator, and collect a deadtime curve (make

sure a deadtime curve is collected at the beginning of the run before

running any samples, don’t let the beamline scientist leave until this is

collected). The importance of a deadtime curve will be discussed in the

data averaging and background subtraction portions of this appendix.

2) The correct monochromator crystals should already be in place prior to the

start of a beam run. Generally beam line 11-2 runs with a Si (220) crystal

217

o in the Phi = 90 for the Hg LIII edge. This crystal set has the fewest

glitches within the Hg EXAFS region. Though the Si (220) phi = 90o is

generally used, crystals will wear down over time and it might be

necessary to switch to the phi = 0o if the glitches are bad. Make sure that

the crystal set installed will work for the entire run, if not notify the beam

line scientist immediately to schedule a crystal change.

3) Even though this should already be done double check alignment of the

table by using Io, the ion chamber that measures the incoming beam prior

to the beam interacting with the sample, and the vertical table motor. Scan

the table ± 5mm or (use a larger range if necessary) and set the position

for the center of the plateau you should get in the scan window. When the

beam hasn’t been topped up for a long time (missed 1 or more top ups) re-

align the table, when the beam is re-injected at a much higher energy than

it was before, electron path tends to change, which also means that the x-

ray trajectory is different than before. This also goes if there was a beam

dump and the beam has to be injected from scratch. Aligning the table is

less of an issue since SSRL has been running in continuous top up mode,

but a beam dump can still happen and every now and then SSRL does not

run in continuous top up mode.

4) Check gains on the Ion chambers. Anything over 5V is saturating the

electronics meaning that the detector is not getting the full signal, detuning

will not be done properly, and amplitude of the EXAFS oscillations will

be severely damped. In general I1 and I2 have the same voltage when there

218

is no sample in the sample chamber and Io is a little higher than both of

them. If you are too low or high with regards to voltage adjust the gains in

the ion control window. Keep in mind that the highest voltage you will

get is when you have no sample in the path of the detectors, slits at their

max, and a new beam. When you are changing slit sizes and/or a beam

was inserted (new beam or top up) take a quick look to make sure the

gains for all the ion chambers is fine. Check I1 and I2 when you change

samples because different samples will allow more beam through than

others.

5) Detune the beam. The x-ray beam coming in off the monochromator

crystals contains higher harmonics than the ideal x-ray beam for good

quality data. Detuning is a way of getting rid of the higher harmonics.

Detuning is a slight misalignment between the two monochromator

crystals and should be done by detuning the beam down by 25-30% of

peak beam intensity. This is done by moving the monochromator motor

(Relative movement ONLY, if you move it using Absolute the beam

will more than likely be lost completely and the beam line scientist will

be needed to re-align the monochromators) in relative movements by

100 and looking for a maximum in the Io chamber (display on the wall).

Once you have found a maximum voltage calculate what 70-75% of that

value is and then move the monochromator down (moving down on the

motor is better than up because of motors at the end station) in steps of -

100 until you reach that value (you can switch to -50s when you get close

219

but smaller than that will take a long time to get to the value desired).

Periodically check the detuning, after every beam fill, 8 hrs., etc.

Calibration Foils

Even though this should be done before you even arrive by the beamline scientist, you need to calibrate the monochromator with a reference foil. This calibration foil should be placed on the front of I2 ion chamber between I1 and

I2 chambers (Transmission set-up). If you need to make a foil from scratch

(Ex. HgCl2) mix ~15 mg of reference material with ~70 mg boron nitride

(BN), this ratio is good for every pure reference material you want to run in

Transmission mode. In general the reference foil will be left in place for the rest of the beam run (more on why that is important in the data background subtraction protocol). Be careful (check this periodically) that the Ion chambers (I1 and I2) are back far enough so that fluorescence from the reference foil does not strike the Ge detector.

Windowing Ge Detector

To window each channel of the Ge detector put in a reference material

(HgCl2) in position for Fluorescence collection. Windowing of the detector should be done with every sample in order to optimize the data collected.

Align the sample using the sample motors (both vertical and horizontal) making sure you are well above the absorbance edge (the Hg LIII edge is

12,284 eV, so set the monochromator to ~12,400 eV). Make sure your slits are fairly small since this is a very concentrated sample (1mm x 1mm).

220

When you turn on the beam click “Start” on the computer on that controls the

Ge detector, it will be running the Xmap GUI. This is done to make sure that the detector is not receiving too much signal. Each individual channel is color coded, no coloring means the signal is fine, orange means that the channel is seeing a decent amount of signal but is still within specifications, and red means that the channel is seeing a lot of signal. Seeing some red channels is not too bad, just make sure the Incoming Count Rate (ICR) for each channel is lower than 150,000 (ideally your highest channel should be ~100,000). If you see >200,000 counts on the ICR immediately turn off the beam then either close the slits or move the detector back. If you wait to move the detector or the slits the detector is constantly being hit with x-rays, it is much faster to just turn off the beam. Once the slits and/or the detector are in a position where all the channels are within the tolerable range for the ICR, it is time to window each channel. Make sure that you set the monochromator to below the absorption edge (12,000 eV for Hg) and then click “Start” on the Xmap Gui.

You should see elastic scattering peak (this will be at a fluorescence energy above that of Hg) but not much else for the reference material. Other samples have the possibility of showing fluorescence from other elements in the sample. Set the monochromator energy above the Hg edge (12,400 eV) and click “Start” again on the Xmap Gui. You should now see a strong peak before the elastic scattering peak that was seen when the energy was below the absorption edge. Set the window by making a box that encompasses the entire width of the peak, it is usually easier having the windowing set-up for

221 center and width than lower limit and upper limit. For best data collection set

3 SCA windows, generally make the width of each 10-20 eV smaller with the same center. By setting 3 windows you can pick and choose which SCA gives you the better signal:noise ratio for data analysis, plus sometimes you have fluorescence peaks from other elements (As) that bleed into the Hg peak so having smaller windows set a little narrower than the main peak allows for only having Hg signal and no As. Set the window for all channels, click apply

SCA to all, and then go through each channel individually to double check that everything is set right. Save the SCA files by looking at the instructions on the Dry erase board above the computer. It is good practice to check the windows before running a new sample, spending 5 extra minutes to get great data is better than saving 5 minutes and getting not so good data.

6) Double check the energy calibration by going to regions file and selecting

Hgedgequick file and making it current. Run the scan (takes about 2

minutes) and make sure the inflection point of the edge is at 12284 eV, do

this by looking at the first peak of the first derivative in the maths window

of the plotter. In plotter you need to plot using SCA1/RTC (that is the Ge

detector vs. the real time clock).

7) Now that you are windowed and calibrated the detector, have the beamline

scientist collect a deadtime curve for you. Deadtime curves are based on

the fact that the solid state detector does saturate when it is hit by too

much signal. When the detector saturates the electric potential that builds

on the Ge detector does not fully discharge before the next signal is

222

collected and recorded. This results in having too low of signal when the

ICR gets above ~150,000 counts. By collecting a deadtime curve this

saturation can be corrected. The ICR for the best data is an ICR of around

100,000, which is just outside of the linear portion of the fluorescence vs.

concentration curve. With a deadtime curve collected, a polynomial fit to

the signal vs. ICR is calculated during data averaging and the polynomial

is applied to the data to correct for the detector saturation and force data

onto a new theoretical signal vs. ICR line that is now linear.

Depending on the concentration of the sample to be run data will be collected in either Transmission mode (uses Ion chambers, generally > 1,000 ppm) or

Fluorescence mode (generally the Ge detector is used but for samples with concentration of 50-1,000 ppm the Lytle can be used), the Lytle detector (for use with concentrations between using Ge detector and Ion chambers), is run almost exactly like using the Ion chambers. The Lytle detector has a display next to the displays for the ion chambers and the gain settings for the Lytle detector are the same as the ion chamber (keep below 5V at all times).

8) Now the beam line is prepared to start running samples. If a LN2 cryostat

is being used, place the sample on the holder, turn on the turbo pump, and

let the pump run for about 10 minutes before adding liquid nitrogen. This

keeps frost from forming on the inside of the sample chamber which will

really result in very poor spectra. While the sample chamber is pumping

223

down the sample can be aligned with the beam on, take care to only do

this for samples that are immune to beam damage (As is a prime example

of a sample that will change in a beam at room temperature), Hg does not

have problems with beam damage. Once the chamber is evacuated add

LN2, keep in mind that the vertical position of the table will change a little

bit because of the extra weight and the table position should be adjusted

accordingly.

9) If you are using a multiple sample holder it is important to scan the

vertical sample chamber motor over a wide range above the Hg edge and

make a note either on the white board or in a notebook of where the

absolute positions of the sample holes. By noting the absolute positions

for samples it will be easier to move from sample to sample for the rest of

the beam run. When scanning over the vertical range for the sample

positions make sure that the x-ray beam is turned on and the

monochromator is set to an energy above the edge (12,400 eV). Align the

sample looking for the max fluorescence and hopefully a flat plateau. A

flat plateau in the fluorescence signal means the sample is homogenous in

concentration and sample height in the holder. Make sure you align with

both vertical and horizontal “Sample” motors.

10) Before running the sample re-window the detector. The detector is seeing

fluorescence from several different elements in the sample. This

fluorescence adds to the ICR for each channel. For most environmental

samples the prime element that is a problem is Fe. By removing the Fe

224

signal reaching the detector the incoming beam intensity can be increased

resulting in more Hg signal reaching the detector before the channels

saturate. It is very important to knock down fluorescence from other

materials in the sample. In general Al and Ga filters work the best for Hg

samples. To see how good of a job the filters are doing check the %

Fluorescence after changing the filter set-up. % Fluorescence =

SCA/ICR*100 use the Xmap Gui to figure this out. For low concentration

samples < 250 ppm expect the % Fluorescence to be < 5%, it just means a

lot of scans need to be done. Once % Fluorescence is maximized and

fluorescence of different elements is knocked down as much as possible

you are ready to start running samples. If you are having problems with

getting % Fluorescence up take out the Sola-slits and move in the detector,

making sure that total ICR is not too high. The Sola-slits are there to

focus the fluorescence into the detector, but sometimes it is more

important to have the detector closer to the sample chamber than have the

Sola-slits inserted. To maximize % Fluorescence you move the detector,

remove Sola-slits, change filters, and change slit size for the incoming

beam. Usually opening slits gives more fluorescence of everything but

using filters can help with that. Just be sure to watch ICR on the detector

because you can’t filter out elastic scattering because it is above Hg

fluorescence energy.

11) Now you can set-up your regions file. Generally for Hg a maximum range

of k = 12 is the highest you can get. Open up the region file for k = 12

225

long and look at it. The file is split into 3 sections: 1) Pre-edge region, 2)

XANES/NEXAFS, and 3) EXAFS. You generally don’t need to change

anything in sections 1 and 2 but you might in section 3. In section 3 make

sure the power = 3. The EXAFS part has a max count time specified for

the EXAFS region (generally 20 sec.) and a minimum (generally 1 sec.),

which means that the first sample point in the EXAFS region has a 1 sec.

collection time and the last point has 20 sec. collection time. What the

‘power’ does is it makes the seconds for each point a cubic relationship

where the first point is 1 sec. and the last is 20 sec., which means that in

the earlier part of the EXAFS the scan time is less (which is fine since

there is a stronger signal in that region) and the end part of the EXAFS is

really long, with the last point being 20 sec. This has the advantage over

power = 2 (squared relationship) of spending less time in the region of the

EXAFS that gives plenty of signal and spending more time in the area that

requires more collection time. Power = 3 is generally best for optimizing

time spent on a sample and data quality.

12) Once your region file is set-up start your run, sample should take ~40

minutes per scan. Always set run for a large number of scans (40 or

more), this way if the experimentalist does not get back to the beam line

when they want too (overslept, took longer to eat than expected, etc.), data

will still be collected. Collecting more than enough data for a sample is

better than the sample sitting without any data collection occurring. Plus,

226

a data collection can be terminated at anytime, if data collection is set for

40 scans, it is not necessary to actually collect 40 scans.

13) During the run (especially low concentration samples) do averaging and

background subtraction (See background removal protocol below) after

every scan to determine how many more scans are needed before changing

samples. Though it is labor intensive it helps to maximize time on the

synchrotron.

Transmission Mode

1) Instead of putting the sample at a 45o to the X-ray beam it is placed

normal to the x-ray beam.

2) Since the Ge detector is not being used make sure it is capped.

3) In this situation all three ion chambers are being used so check the

gains and adjust accordingly, also double check that the calibration foil

is on the front of I2 and not the back or front of I1.

4) Because the sample needs to be highly enriched in the element of

interest for Transmission mode fewer scans are necessary, generally

less than 5. Five scans is not a hard rule, it is necessary to do data

averaging and background subtraction of the data to determine the

quality of the data prior to changing samples.

227

II: Background Subtraction Protocol for EXAFS using

SixPACK

This protocol is for averaging files and doing background removal of XAS spectra collected at a synchrotron. The software that this protocol pertains to is SixPACK which is readily available freeware that is found either by searching for SixPACK on the internet or by downloading the software suite mentioned in the FEFF model making protocol (see below). There are several different software programs that will do the same thing as SixPACK, some do better background and fitting depending on the situation, but most are not as user friendly. There are

5 main parts of SixPack that will be used: 1) SamView (this is where averaging scans of a sample to make an average file for background removal), 2)

Background removal (this is where data de-glitching, pre- and post-edge normalization, spline control, and various other commands to get an optimized mu, chi, and rdf file of the data), 3) Least Squares Fitting (this is the

“Fingerprinting” method of determining what phases and fractions are present based on reference materials), 4) Feff EXAFS Fitting (this is shell-by-shell fitting using theoretical pathways created using the FEFF software), and 5) Make SS

FEFF Pathway (this is used when a single scattering pathway using the absorbing atom, scattering atom, and path distance is used to make a FEFF pathway complete crystallographic information is missing).

SamView Window (Scan averaging)

228

1) The SamView window is opened in order to average numerous scans of

data, throw out bad detector channels, do deadtime corrections, and set

energy calibration. To add scans click on either the ‘Add File’ or ‘Add

Many Files’.

2) When all the files are selected to open, click “Open” and a small window

will pop up asking if Lytle or Ge was used, click whichever detector was

used for collecting fluorescence data (if data was only collected in

Transmission mode, either selection will be fine) was used and then click

either OK or apply to all. If you click OK it will ask the same question for

every single scan selected. If Ge was selected and analysis of another

sample collected with the Lytle detector is to be averaged it is necessary to

exit all the way out of the program and open SixPACK again or it will

open all of your files as Ge instead of Lytle. After the detector is selected

the program will ask if SCA1, SCA2, or SCA3 is to be used, this prompt

only shows up if Ge was selected. If all three windows were set up for the

scans then choose the SCA channel that works best. To get the best

quality data and fit it is important to do data averaging and background

subtraction for all three SCA channels.

3) Click on each of the files to make sure they look similar since all scans in

the left column will be averaged together. If a file looks really bad

highlight it and click ‘Remove File’. Note by left clicking and dragging

on the plot a zoom box can be created, to zoom out just click the right

mouse button (this works for all plots in SixPACK).

229

4) Next do a deadtime correction. Go to the bottom left and click on ‘New

DT File’. Click load file and find the DT file that the beamline scientist

collected (If there is a problem with loading the file it is usually that (–)

was used instead of (_) in the file name). Once the file is loaded click ‘Fit

Deadtimes’. The computer will now fit a polynomial through each of the

sets of data in the graph to use for the correction. Once the fits are done

just go through each channel to make sure the polynomial looks like a

good fit and save the DT fit. Close the window and find the saved fit in

the deadtime file box. Then click the box ‘Do the deadtime correction’.

Watch the y-axis of the graph when ‘Do the deadtime correction’ is

clicked the y-axis scale increases and glitches will be significantly

diminished.

5) Now you will do the E0 calibration. Under the plot there is a Plot Type

box with 4 options: 1) muF shows the Fluorescence data, 2) muT1 shows

Transmission data for a sample between I0 and I1, 3) muT2 shows

Transmission data for a sample between I1 and I2 (generally the reference

foil), and 4) Column, which allows the Column Selection box to become

available. For E0 calibration use muT2. First click on the file that is to be

calibrated. Then click on muT2, this will show a dramatically different

plot because it is HgCl2 instead of the sample. Now plot the first

derivative in order to identify the first inflection point of the graph. Do

this by clicking first derivative, then re-plot, and finally find data E0. The

graph will show all several green lines and one red line. The red line is

230

the apparent energy selected. If there is no line going through the max of

the first main peak raise the number of points to smooth and re-plot,

continue until a line goes through the max of the first peak. Make the line

through the first peak red, then type in the actual energy based on the x-

ray absorbance of 12284 eV (most important thing is to be consistent

between samples, if 12282 eV is used that’s fine just make sure same

energy is used for all Hg spectra), hit enter, and then click ‘Apply shift to

data’. This will shift the sample data by the same amount that the

reference material was shifted. You will need to do this for all of the scan

files one by one, BE CONSISTENT.

6) Set the plot back to 0 points to smooth and none for the derivative. Now it

is important to look at each channel individually and zeroing bad channels.

This is done by highlighting ‘Column’ in Plot type. In the Column

Selection box there are several choices: 1) i0 shows the signal for I0

(helpful to look at where the glitches are in this plot, look for sharp drops

in the signal and type the eV value for the glitches into a notepad

document for de-glitching), 2) i1 shows I1 signal, 3) i2 shows I2 signal, 4)

iFsum shows the sum of all the Ge fluorescence channels, 4) iFind shows

each channel individually, and 5) iFall plots all channels on the same

graph. Click ‘iFind’ and go through each channel individually by clicking

on the channel in the Fluorescence Channel box. If a single channel looks

really bad click ‘Zero Sel. Chan.’ to zero the channel taking it out of the

sum for all the channels.

231

7) Once all the channels for each scan file have been evaluated, plot the data

as muF with no smoothing or derivatives. Now click Average Scans and

save the average file. The whole process is tedious but gives the best

quality data for fitting.

8) The gains can be changed on I0 but it is not recommended, it is not

recommend that any of the last 4 options of the tool bar should be touched.

Background Subtraction

1) Add files the same way as in SamView. To look at several different

sample files at once this is the best place to do it. Examples would be to

compare SCA1, SCA2, and SCA3 against each other or different reference

compounds. To plot many samples on the same graph double click the file

until an “x” appears next to it. Once an “x” is next to all the files of

interest, right click on the plot type of interest. Note this only works for

mu, chi, and R plots.

2) For background subtraction of a file a few things on the main window

need to be changed. For linear combination fitting type in 12300 for E0

(this sets the beginning of the spline at 12,300 eV which is above the edge

for every single Hg spectra that will be analyzed, it doesn’t need to be

12300 but it works well, BE CONSISTENT). For shell-by-shell fitting

use the E0 supplied by the software (if the E0 is set to 12,300 eV for shell-

by-shell fitting then the E0 of the fit will be shifted well out of the range,

232

discussed below). Another parameter that needs to be changed is dk (dk

is a smoothing function, it is always best to fit raw unchanged data), this

should be 0. Make sure k-weight is 3, Rbkg is automatically set to 1 (this

is a signal to noise adjustment that tends to change the peak position in

your Fourier Transform, so don’t mess with it too much). Rbkg subtracts

out low frequency oscillations in the EXAFS that are not true scattering

pathways. An Rbkg of 1 removes all oscillations that are the result of a

pathway between an absorbing atom and a backscattering atom at a

distance of 1 Å, in general the Rbkg should not be set to more than ½ of

the nearest neighbor distance. Other parameters that can be changed are

the clamps which are initially set to none (these are mainly for dealing

with older EXAFS fitting software), and FT window which is set to

Kaiser-Bessel (this generally works the best). After changing these

parameters for one file, the rest of the files can be set exactly the same by

clicking ‘Parameters’ in the tool bar and select ‘Set all to current’.

3) Once this is done it is time to de-glitch the data. Glitches are caused by

the cut and the physical properties of the monochromator crystals. The

monochromator works by changing the angle of the crystals that when

Bragg’s law is fulfilled gives different x-ray energies. This works well but

it is an imperfect process so there are sudden drops in intensity of I0 that

show up in sharp peaks in the data (that’s because the data is absorbance

and will behave inversely to I0 intensity). When de-glitch is selected from

the tool bar a new window will open up. Both Mu and Chi can be edited

233

(it is highly discouraged to edit Chi). When Mu or Chi is de-glitched it

will affect the other, but editing Chi is editing data that has had math done

to it, editing Mu is editing the raw data which is preferable. Look for any

sharp features in both the Mu and Chi and remove them. Glitches tend to

range from 1-3 data points. Try not to de-glitch more than 2 consecutive

points because that becomes questionable. Double check glitch positions

by looking at the notepad file of I0 glitches constructed earlier during data

averaging. BE VERY, VERY CAREFUL when de-glitching, poor de-

glitching can easily make up data, introduce new features, or remove real

features.

4) When finished deglitching click ‘Accept’ if this is not clicked then all the

work will not be saved.

5) Pre- and Post-Edge Normalization normalizes the edge jump to have a

magnitude of one between pre-edge average and the post-edge average

that forms the spline of the EXAFS data. Generally a Linear fit for the

Pre-edge and Quadratic fit for the Post-edge work the best. Look at the

PreEdg plot to see how good the fit is, then look at the mu(E) plot. If the

Normalization is correct the pre-edge line should be perfectly flat and the

post-edge should be perfectly flat along a y-value of 1. If this is not seen

then change the energies used for the normalization (this will happen more

often for the Pre-edge than the Post-edge). The energy values are eV from

the edge position (E0), meaning -200 in the pre-edge normalization is the

same as 12100 eV (assuming 12,300 eV was used earlier). The

234

normalization is done by using two points of the data and running either a

linear or quadratic function through the data points of the corresponding

energies.

6) Once a good mu(E) is attained it is time to look at the spline. The spline is

a polynomial that is run through the EXAFS where each point along the

spline is designated as zero in the EXAFS amplitude for the chi(k) plot. It

is of prime importance that the spline does not add or subtract to the

oscillations in the chi or have any sharp features (sharp features generally

happen at the end of the spline). The spline region goes above the energy

that you set for the scan because it gives a cleaner spline. Generally the

spline is properly set by the computer but if the data needs to be cleaned

up change the region of the spline, by reducing the upper limit of the

spline data starts getting cut of so that the data fitting range is smaller, but

it can really clean up the EXAFS spectra at higher k. Double check the

spline by using the spline, chi(k), and R plots. The spline needs to be a

nice smooth line that does not add or detract much from the raw data.

Changing the number of spline knots (a spline knot is a point where the

spline polynomial is recalculated for the region between knots, 5 spline

knots would result in 4 polynomials calculated for the entire spline) can be

done but it is not recommended. In the R window (Fourier Tranform or

Radial Distribution Function, all the same thing) you are looking for nice

clean peaks with little noise. Do keep in mind that depending on the

sample there can be shoulders on main peak caused by different bonding

235

distances, different backscattering elements, and/or multiple-scattering

pathways.

7) Once all the background subtraction is done save a mu(E), chi(k), and R

file for the sample.

III: Least Squares Fitting Protocol

This protocol is for using the Least Squares Fitting (or linear combination fitting) module for ‘Fingerprinting’ element phases in samples. The quality of fit solely depends on the quality of reference material spectra collected at the beamline. It is of prime importance that averaging and background subtraction for the reference spectra is done carefully, properly, and consistently. If reference spectra are of poor quality data fitting will be difficult and of very poor quality.

This module in the SixPACK software has a lot of glitches and unnecessary options. The end of the protocol will mention what parts not to use in this module. One important note of LCF is that data collected on a beamline that was de-tuned is not comparable to data collected on a beamline that uses harmonics mirrors to remove upper harmonics. This is because de-tuning throws out part of the primary harmonic that is used in analysis while the harmonics mirrors do not.

This results in a difference in amplitude for the EXAFS oscillations and will result in either over or under calculation of phase proportions when reference and sample spectra from the two different types of beamlines are mixed. For shell-by-

236 shell fitting, the least constrained variable is the coordination number. Since the coordination number affects the amplitude of the EXAFS oscillations, a damping down of the oscillations by de-tuning is not a major problem.

1) Open the file to be fit by clicking on the box next to the green ‘Load’ box.

Open the file and click ‘Load’. Note: There is a glitch with opening files,

it will initially give an error box saying “Missing x-limits”, click OK. If

on the right side under fit parameters has zeros for both xmin and xmax,

close the module, re-open the file, and re-load until a value other than zero

appears in the xmax box. The xmin will still be zero so type in the k value

that is being used for fitting (generally a lower limit of 3 is used). If a zero

in xmax appears and another value is entered for xmax the program will

not work properly.

2) Once the file is loaded adjust the xmin value and change x-weight to 3.

This is the same as k-weighting of the EXAFS spectra that was used in

background subtraction and is the preferred weighting for EXAFS data

fitting. k-weighting is k3*chi(k) this is the x axis value (k) cubed times the

y axis value, this makes the EXAFS amplitude of the lower k values lower

and the higher k values larger making the EXAFS oscillations into a

summation of several sine waves instead of a sum of several dampened

sine wave.

3) Every time a parameter in the ‘Fit parameters’ box is changed the ‘Load’

button must be clicked to save the changes.

237

4) In general do not check ‘Non-negative Fit’, ‘Sum comps to one’, and

‘Allow Data E to float’. They generally don’t help the fit.

5) In the Components section click ‘Add Comp’ and add the reference

material to be included in the fit.

6) Click ‘Fit’ to fit with the components in the Component box. Files can be

removed from the fit by double clicking spectra name file until the file

becomes faded.

7) Plot options allow for several options: 1) the raw data, 2) the fit of the

data, 3) Residuals (residual = valuedata – valuefit), 4) Waterfall (this is used

to plot all the components used at the same time, with the relative

amplitude used based on the Fraction). SA corrected (Self Absorption)

and E shifted are not helpful in fitting.

8) The Chi Sq and Red Chi Sq are supposed to be useful for ‘Goodness of

Fit’, but it really doesn’t help with Least Squares Fitting (it’s really for

FEFF fitting).

9) The real way to find the ‘Goodness of Fit’ is the residual for the entire fit,

regretfully the program doesn’t really calculate a residual for the entire fit.

To calculate the residual for the full fit use the following formula:

2 . . It is best to calculate this in

Excel. To get the residual values for the calculation you go into the Save

tool bar and save the residuals.

10) The software does a decent job with the ‘Fit’ command but it doesn’t

always do the best job with the fitting. Instead the ‘Construct’ command

238

can be used to make fits with user specified fractions. By doing this the fit

of the data can be refined even more, especially if a certain component is

definitely in your spectra but not fitting correctly (negative number or

really small) by constructing and checking the residuals.

11) In order plot the data in Excel (looks better) on the plot window go to

copy data to clipboard and paste the data into Excel. Note: This will plot

the y axis as k-weighted so that you don’t have to k-weight it yourself.

12) To determine % Component just divide the Fraction by the component

sums.

13) Make sure you save your workspace so that you can re-load it late without

having to re-do the fitting.

Additional features:

Two new fitting features have been added to SixPACK. They are matrix and cycle fit. Matrix fit will do fits of every single combination of reference spectra loaded into the fitting window. If five reference spectra are loaded the program will do a fit with one component, then two components using every possible combination, then three components, etc. This can be useful, but be careful some fits that have a lower reduced chi have components that are either negative or near zero, so these components are not truly in the spectra.

Matrix fitting is useful, but caution should be used while using it. Cycle fit is similar to matrix fit. Cycle fits by using the non-faded spectra as fixed components of the fit and cycling in only one of the other faded spectra into the fit. If two spectra are non-faded and three are faded the cycle fit will have

239

three fits total. Each of the three fits will include the non-faded spectra plus

one of the faded spectra.

Notes: Do not use the Self Absorption correction in this module, use a different program for this. The Interpolate tool bar is not useful either for least square fitting. The Function tab is not generally useful in fitting. The results tab only works if you use the Fit command not Construct.

IV: Protocol for Making FEFF Theoretical Pathways for

shell-by-shell Fitting of EXAFS Spectra

Because shell-by-shell fitting of EXAFS data is dependent on the construction of theoretical scattering pathways made using the FEFF software, the protocol for making theoretical FEFF pathways will be covered prior to the protocol for shell- by-shell fitting of EXAFS spectra.

First, make sure that the entire Ifeffit software suite is downloaded from the following website:

http://cars9.uchicago.edu/~ravel/software/downloads.html#winxp

This includes software suite includes FEFF6L, atoms, and SixPack which will be used in model building and data analysis. Other programs such as Athena and

Artemis do the same data analysis as SixPACK, but are not as user friendly.

240

Athena and Artemis have some advantages over SixPACK, especially by being able to Fourier-filter and back transform, so it is useful to know how to use part of or all of these two programs. Hephaestus is useful for element information, very helpful at the beamline for determining what filters to use for removing fluorescence of other elements (Note: the filter information is in version 0.18 of

Hephaestus). At the end of this protocol there is a small section on how to use the

“Make FEFF SS Paths” module in SixPACK if a crystallographic model that includes a specific bond cannot be found but a bond distance is available.

1) Find crystallographic data for model compounds from the literature. It is

necessary to have space group information and atom positions of the

model compounds. Two important websites for crystallographic model

compounds for making FEFF theoretical pathways are:

http://www.minsocam.org/MSA/Crystal_Database.html

(this website is only for minerals)

And http://icsd.ill.fr/icsd/index.html or http://icsd.ill.fr/icsd/index.php

(this website is for synthetic materials including organic molecules bound

to metals, make sure that the models selected are useful, sometimes alloys

will have mixing ratios with unusual metals that will not be useful for

fitting)

2) Make an input file in notepad for the model you want and save it as a .inp

file (an example for Cinnabar is below). Saving the input file as .inp or

.txt does not matter the atoms program will be able to use either of them,

241 for simplicity it is best to make input files in a program like Notepad rather than Word. It is helpful to use a template and just alter the file between models. Make sure you put in the correct space group, unit cell parameter (lengths and angles), and atom positions. First column in atom position is the type of atom the next three columns are x,y,z coordinates

(respectively), the fifth column is the atom designation (Hg1, Hg2, S1, S2, etc.). Very important, FEFF will generate a large amount of files when finished running (in most cases more than 100). Make sure an individual folder for each mineral/molecule model is created. Example: Make one folder for Pyrite and a separate one for Marcasite. Make sure that your input file is in the folder for that specific model compound because FEFF will store all output files in the same folder as the input file. title name: HgS Cinnabar title formula: HgS title sites: Hg1,S1 title refer1: Ramsdell title refer2: title schoen: title notes1: space = P3221 a = 4.145 b = 4.145 c = 9.496 alpha= 90 beta= 90 gamma= 120 rmax = 6.00 core = Hg1 atom Hg 0.71980 0.00000 0.66666 Hg1 S 0.48890 0.00000 0.16666 S1 ------

242

3) Once the input file is made, open Ifeffit folder in programs (assuming you

are using Windows) and click on More Programs. This will bring up a

box with several programs in it, Click on Atoms. This prompt will ask to

open the input file. Atoms will do a lot of calculations and finish with a

message saying that the program ran successfully. Open the output file

and it should show all the different pathways the program made based on

the atom positions and symmetry. It is helpful to double check the Atoms

output file by constructing the model in CrystalMaker or Vespa and

comparing the interatomic distances. If there is an error, look on the black

screen to figure out where the problem occurred, usually it is the space

group designation (remember to remove any underscores in space group

designations, use – instead of _).

4) Once Atoms is run a FEFF input file will be made. Open the file in

notepad since it will need to be edited. In the input file make sure that the

Control line is all ones, the Print line should be 1 0 0 3 (this is to ensure

that all FEFF files are written), remove the * in front of criteria and add 10

10 before “Curved Plane” (to make cutoffs for the curved plane), and

make sure that nleg is set to 4 (it can be made shorter to limit multi-

scattering, but cannot be less than 2). Save this file after modifications are

completed.

5) Now open FEFF6L in the Ifeffit folder and open the feff input file that you

corrected. The program will generate FEFF pathways that it will put in

the same folder as the input file, depending on the number of pathways

243

and the Rmax distance selected (generally 6 Å) over a hundred pathways

can be generated. When FEFF6L is finished it will state how many

pathways it made. During the fitting process it is useful to open the

pathway files in order to see what the file contains as far as the pathway

backscatters and the total distance of the pathway.

Single Scatter Path construction using SixPACK

1) Open up the Single Scattering FEFF path maker in SixPACK.

2) Left click on absorbing atom on the periodic table and right click on the

scattering atom.

3) Type in the distance to the scattering atom in the box.

4) Select the edge that is desired and the geometery of the binding, if

absorbing atom is a metal it is usually octahedral, but double check the

geometry. In general don’t change what is in the AFOLP and Exchange

boxes.

5) Very important!!! Select FEFF6-7 in the FEFF version and type in the

FEFF Command box ‘feff6L’. If this is not done the program will not

work properly because the Ifeffit software suite downloaded earlier only

contains Feff6L. If FEFF8 or FEFF7 is on computer then FEFF8 or

FEFF7 can be selected, make sure that the FEFF8 or FEFF7 programs are

in the Ifeffit folder so that the software can find them. FEFF8 must be

purchased and FEFF7 is not used that often because there are few

advantages versus FEFF6 (generally only helps with actinide elements).

244

6) Click Run FEFF and the program will make a file in the fefftemp folder in

the SixPACK folder. If this file is not moved to a different folder or

renamed, the next time a SS pathway is created it will overwrite this file

the next time another SS pathway is created.

7) Single scattering pathways have the same output file form as FEFF6L and

are used in the same fashion as FEFF6L, FEFF7, and FEFF8.2 generated

pathways.

V: FEFF EXAFS Fitting

Feff EXAFS fitting is done by fitting the chi files created during background subtraction with theoretical pathways constructed by using FEFF. Fitting is a very iterative process in which the first coordination shell (ligands bound directly to the metals) fit, followed by second coordination shell, third shell (or multiple- scattering pathways), etc. It is necessary that the fit of the first coordination shell is done well prior to adding additional pathways. Do not add additional pathways until the first (shortest pathway distance) is properly fit.

1) Open the FEFF fitting module, find the data file to be fit and load it. Set

the fit parameters for the k range preferred, make sure the k-weight is 3,

and the dk (smoothing) is set to 0.

2) Choose the first path to fit and make sure both “Use Path 1 in fit” and “Set

DEGEN=1” are both selected. If DEGEN is not set to one then the

software will assume that the degeneracy of the pathway in the FEFF

245

output file is correct and will base all fits off of that degeneracy. By not

specifying DEGEN = 1 the coordination number of pathway is not

calculated for the fitting.

3) The path label, which is not a necessary input, is the type of metal-ligand

(Hg-Cl, Hg-S, Hg-O, etc.), pathway created by FEFF. This information is

seen when the individual pathway files are opened in notepad.

4) Next go to “Templates” in the toolbar and select ‘CNs-Unique’, then go

to “Path Vars” and select both ‘Add REFF’ and ‘Add ABS to Sigma2’.

These changes make every pathway have its own individual coordination

number, the software use the effective radius of the ion, and adds

absorbance to sigma2 (the thermal and static disorder portion of the fit).

5) Click on the Variables tab and then click “Add all variables”. This will

add five boxes: 1) N1, this is the coordination number of that pathway

(start off with a reasonable value 4-6), 2) s02, this is an amplitude

reduction value (set initially to 0.9), 3) e0_1 is the ΔE, basically the shift

from the theoretical value (initially set to 0), 4) R1 is the distance between

absorber and backscatterer (set with the pathway value), and 5) sig2

(debye-waller) is the measure of thermal and static disorder in the system

(initially set for 0.005).

6) When fitting variables can allow things to float by selecting “guess”, be

set using “set”, and link/set values using “define”. Be very careful when

allowing variables to float together, since disorder and coordination

number influence each other they should not be let to float together. Same

246

with e0 and R. With an initial fit make sure that all variables are set to

“set”. The ff2chi button at the top can be used to calculate the initial

values (this is not fitting) but it is generally not that helpful.

7) Do an initial fit with everything set to determine the fit quality. The

“Goodness of fit” is taken from the Chi Sq. (chi square) and Red Chi Sq.

(reduced chi square). In general the Red chi sq. needs to be below 100 but

if value is below 1 the fit is very questionable, nearly perfect fits are

around 10 for the Red chi Sq. R factor is also helpful in determining fit

quality but the best gauge is the red chi sq.

8) In general, start off by fitting a reference sample so that the CN and bond

distance are already known.

9) Plot the data and the fit and look at the results. When looking at chi, if the

shape FEFF generated EXAFS oscillations are of a different phase or have

a different envelope shape, there is a good chance that the pathway file

selected was incorrect.

10) Once the initial fit is completed and the correct backscatterer is

determined start letting some variables float while fixing others. In

general let the variables float in this order: R, e0, N, and finally sig1. Let

s02 float once in a while to make sure value stays close to 0.9. Go through

this sequence several times until the chi sq and red chi sp come to a

minimum. Then to make sure all the parameters are stable allow all

variables to float at the same time, if the fit is solid all the variables should

247

change very little. If the values for chi sq and red chi sq ever become 0.0

then something went wrong and fit needs to start from scratch.

11) If values become really strange (list of acceptable ranges for variables at

the end) then wrong pathway file was likely selected. It could be that the

wrong distance for the backscatterer for the correct backscatterer or the

wrong backscatterer entirely was selected.

12) Plots: Since the plot of k has no math manipulation done to the data this is

the main plot used in evaluating fit quality. R is also heavily used since it

is the RDF formed by the FT of the chi data, this plot is useful because it

give a visual gauge on the path distance and CN. q is fourier filtered,

back-transform of the EXAFS, it gives you a general idea of what is going

on but don’t use it for actual fitting, it is basically fitting data that has had

2 sets of mathematical calculations done. The Paths plot shows the

contribution of each path (just like waterfall in least square fitting)

involved in fitting. With the RDF plot there are real and imaginary parts

to the EXAFS function that can be plotted underneath the main curve. To

determine how good the fit really is make look at the quality of the fits for

the real and imaginary parts.

13) When the path is fit the best it can be it is time to add another pathway. In

the paths tab click “Add Path” then do the same thing with the initial

pathway for setup except for one thing. In the E0 of the paths tab for path

2 change e0_2 to e0_1, this links the ΔE of the pathways to pathway 1.

For a single sample the pathways must have the same e0.

248

14) Keep going with the fits adding paths when necessary, be very careful

though because if fitting parameters are not constrained properly anything

can be made to fit the data (the fit might not be realistic but it can be made

to look like a great fit).

Fit parameter values

N: This value needs to be realistic for the absorbing atom, if it forms

octahedral bonding environments than the sum of the first shell needs to

sum up to 6 ± 1. The CN is generally ±0.5. Coordination number is the

least trustworthy of all the parameters. Even though the results tab will

give a lower error than 0.5, in real terms the error is usually ±0.5. In R

space, this value manifests as the area underneath the curves. Since the

debye-waller factor controls the FWHM of the peak, the two variables

affect each other and should not be allowed to float together.

S02: This value needs to be as close to 0.9 as possible, 0.8 is basically as

low as this parameter can go. If it is over 1 then the pathway is definitely

wrong because that means that the kinetic energy of the ejected photon is

over 100% when compared to the kinetic energy of the incoming photon.

E0: ±10 is the range for this. Since this value shifts the EXAFS

oscillations of the pathway right or left if it is pushed too far then the

pathway selected is incorrect. The E0 slides the pathway right and left

without any changes to the phase. R on the other hand affects the phase,

similar to fixing one end of a spring and pulling on the other. The

249 oscillations slide right and left similar to E0, but the translational change increases in k space as the phase of the oscillations are being changed.

R: This needs to be approximately what the pathway file says. Suppose there are two pathways of Hg-O one at 2.2Å and 2.8Å and the real value is

2.3Å. If the 2.8Å is used the computer will force the fit down to 2.3Å, but the shift is not realistic. It is better to start with the lower distance and have the software move the fit to a larger distance, if the distance becomes too large of a change, then select a different pathway with a longer distance.

Sig2: This value is the error of R, measures thermal and static disorder. It is a Gaussian curve that is being fit for the ± R. Because this is a squared term a negative number is always incorrect, not really possible to have negative disorder. A value below 0.0005 is too low, meaning that the curve being fit in R space is basically a vertical line, but a value of 0.1 is definitely too high. Common sense on the error value is required for this one.

250

Appendix IV

X-ray Diffraction Software: JADE Diffraction Analysis Software Tutorial

JADE is a software package used in the analysis of diffractograms collected using either a lab-based x-ray or synchrotron radiation source. This tutorial is meant for basic data analysis and fitting of one component and two component natural samples. JADE is capable of doing calculations on lattice stress of the diffractogram versus the diffractograms in the database to determine structural disorder along with several other crystal and surface structure calculations, but this is beyond the scope of this tutorial. Most examples in this tutorial are from a natural evaporite sample taken at the New Idria mercury mine, San Benito Co.,

CA. The tutorial will take you through background subtraction, peak identification, and matching peaks with peaks contained in the software database.

Diffractogram fitting is often times not straight forward, so a section on determining phases not immediately identified in the initial peak matching is included.

I: Main page, background subtraction, and peak identification

Note: Figure 1, main JADE page on Page 3

1) Open a pattern by clicking on the “File” menu and clicking on “Patterns”. Once the folder with your data is open, you can open any file in the folder by double clicking or dragging the file from the left top column. It is possible to open and overlay more than one

251

diffractogram by dragging the diffractograms you want to overlay into the main area. 2) Navigation on the main page can be done in several ways. When the arrow with a Z box is selected, as seen in Fig. 1, a zoom box can be made by clicking and dragging in the main window. Another way to navigate is to use the buttons on the bottom right of the diffractogram. The buttons allow for changing the x-axis scale, moving the diffractogram right and left, changing the y-axis, etc. Note: Changing the axis scale is done by either right or left clicking on the button. To un-zoom completely just click in the black area then press the space bar. The navigation box can be docked in different portions of the main window. It may not be in the bottom right position. To move the box go to the “View” tab and go to the selection “Dock Toolbox” and select your personal preference. 3) To do a background subtraction click on the “Analysis” menu and click on “Background Subtraction”, this can also be done with the shortcut keys denoted by the “Background Subtraction” button. 4) When “Background Subtraction” is selected a new dialog box will pop up (Fig. 2). Almost all the default settings are fine to use. The only thing that you change is along the top of the box. The default is a cubic spline and the circles next to the cubic spline box denote the number of knots selected for background subtraction. Other options for splines can be selected but these are rarely useful because of the higher background at the beginning of a diffractogram caused by x- rays reflecting off the sample at 2-θ <14o. 5) When choosing the number of spline knots, it is important to not use too many or too few knots. It is important that the background removal line is smooth and does not add or subtract from the peaks. Generally, a selection of one of the three spline knot selections closest to the cubic spline box works best. 6) Once a good background is selected click “Apply”, then “Remove”, and finally “Close” 7) Next find peaks by selecting “Find Peaks” in the “Analyze” menu. Generally the defaults work well for finding most of the peaks. If you want you can change the selection for # of peaks to a higher number so the computer will find more peaks. Other than this option it is not necessary to change the defaults. Once the peaks are found they will be seen as dashed blue vertical lines as seen in Fig. 3. 8) Often times the software will miss peaks or assign peaks that are not there. It is important to widen the x-axis scale (zoom in) and look at

252

the peaks individually. By selecting the peak diffractograms box, the image showing peaks with the crossing blue lines (third button from the left in toolbox on the main window) you can add peaks by left clicking below the diffractogram or remove peaks by right clicking.

Fig. 1: Main JADE window. Top left column is data in folder currently open, bottom left is recently used files, and black area is file currently being worked on. Bottom axis units can be changed by clicking on the 2-Theta (o) box at the bottom. The toolbox on the right is for navigation while the upper toolbox is for adjusting the background, data, peak overlay, and most importantly adding and subtracting identified peaks.

253

Fig. 2: Background subtraction dialog box. The defaults almost never need changing except for the number of spline knots which is seen by the row of circles at the top. In general a good background removal is fairly flat and does not add or remove to peak height.

254

Fig. 3: Software identified peaks, denoted by dashed blue lines. As seen from figure the software did not identify all the peaks in the diffractogram. It is necessary to zoom in and manually add in peaks using the peak addition box in the toolbox at the top of the black window.

255

Fig. 4: Zoom in on the diffractogram to see what peaks are missing from initial peak identification. Peaks are added by selecting the peak edit button (already selected in the top toolbox) and then right clicking or subtracted by left clicking on a dashed blue line.

256

II: Mineral Identification (1 and 2 component samples)

1) Once peaks are identified, a search-match can be run. It is of prime importance that the entire diffractogram is shown in the black window when doing the search match. The software will only match the peaks seen in the black window so if you only want to match half of the diffractogram then zooming in on half the diffractogram will accomplish this. 2) After un-zooming click on the “Identify” menu then click on “Search Match”. This will bring up another dialog box (Fig. 5) that allows you to select what databases the software will look through and whether filters are being used (description on filters will be discussed later). In general, selecting the inorganics and minerals databases will work for natural samples. The currently available JADE databases allow searching for inorganics, minerals, ceramics, organics, detergents, explosives, forensics, etc. Other databases that are often used are PDF and ICSD. These are additional databases that are updated often and are useful especially for material science applications. 3) On the dialog menu you can select whether you want to match major, minor, and trace phases for your matching. Depending on which selection is made the search results will change. 4) The results generated from the search need to be assessed with high scrutiny. The software is just matching up identified peaks with those in a database, so common sense is necessary for realistic mineral/inorganic identification. The new page that is brought up (Fig. 6) with the fits shows 40 hits (this can be changed) and the third column is the figure of merit (FOM) which is a measure of how good the fit is. In general the FOM is not very helpful since many unrealistic mineral or inorganic patterns will have a low FOM, but the elemental composition does not fit with the system studied. 5) To add the PDF (basically the lines used in the fit) to the main screen, check the box next to the phase, and then close the window. To just see the fit without checking the PDF just click on the name and it will show where the peak lines are in the data. 6) If the sample is a single phase such as quartz, the fit is easily accomplished with one component as seen in Fig. 7. Even though hexahydrite fits the diffractogram quite well, it is obvious from Fig. 6 that several peaks are not fit on the pattern, so a minimum of one more phase is present in the sample.

257

7) To properly fit the remaining phase in the diffractogram, it is important to zoom in on the section where the major peaks are not properly being fit then running the search/match again (Fig. 8). 8) Once again common sense is required for a proper fit. Even though sodium chromium sulfate hydrate does not have the lowest FOM, the pattern fits the rest of the diffractogram that hexahydrite does not. The sample was white with a brownish tint which explains the presence of a transition metal, and the solution chemistry for the site had relatively high chromium concentrations. Wattevilleite has a better FOM than sodium chromium sulfate and the composition makes sense, but as seen from Fig. 9, its inclusion produces a worse fit and lines up with peaks already associated with hexahydrite. 9) Complete fit of evaporite sample is a combination of hexahydrite . (MgSO4 6H2O) and sodium chromium sulfate hydrate . (NaCr(SO4)2 12H2O). This fit is seen in Fig. 10. To determine if a mineral/inorganic phase is a good fit, a good rule of thumb is that the first 4-5 peaks (based on maximum diffraction intensity) are fit with the database pattern. As seen in Fig. 10 more than 5 peaks for both phases are fit with relative intensities in the proper order. 10) In order to see the crystallographic faces responsible for the diffraction peak click on the “h” in the menu at the bottom right of the black window.

III: Extra features

1) The mineral/inorganic phase highlighted is shown in the long white box directly above the black window. This box shows the color of the plot (left of the mineral name) and the PDF number for the pattern. The buttons to the right of the box allow you to cycle through the highlighted mineral species, delete the mineral species highlighted, change the color, and allow shifting of the peaks. Shifting of the peaks is done when sample preparation is not ideal, when the sample is packed too high or low in the sample holder, the peaks shift to the right or left. This peak shift button shifts all the peaks for the highlighted phase at the same time; peaks should not be shifted more than 0.5o in 2-θ. 2) In order to overlay diffractograms of different samples, drag and drop the diffractogram that you want to overlay onto your workspace. You can cycle through different diffractograms by using the toolbar on the opposite side of the navigation tool bar that was used to add or subtract peaks. 3) In order to look at mineral species that might be in the sample but are not found in search/match, you can type in the mineral name in the box in the tool bar (in Fig. 5 it has the number 25-0922), and the software will look for the

258

mineral. There are numerous patterns for the same mineral (different localities, twinned phases, and distorted lattices) and patterns can be added by dragging and dropping from the dialog box. The box with a bluish crystal next to the box with 25-0922 denotes what database the software is looking through. If the mineral or pattern you are looking for does not appear when you search for it, click the box with the crystal in it to change the database and run the search again. 4) Filters can be used by selecting the “PDF” menu and selecting the desired filter. The filter typically used is the chemistry filter, which allows you to search for patterns only including selected elements or all patterns excluding selected elements. 5) For analyzing data not collected with a Cu x-ray tube select “Edit” from the menu, then “Preferences”, and then select “Instrument” in the dialog box. This dialog box allows you to select the type of tube used or if synchrotron radiation is used, you can manually enter the wavelength with which the data were collected. 6) To export data for figures go to edit. If you select “Copy Bitmap” it will copy the image for opening in “Paint” or another similar graphics programs. By using “Copy Pattern Data”, the 2-θ and intensity data can be pasted directly into Excel or other database software. 7) In order to refine the cell parameters for a specific phase, ex. sodium chromium sulfate hydrate, make sure the phase is highlighted and then click “Options” then “Cell Refinement”. The software will show the cell parameters for the phase selected (Fig. 11). When you click on the “Reflections” tab JADE will show the crystallographic planes and their d- spacing. By clicking “Refine” in the dialog box the software will refine the unit cell parameters to what is calculated from the peaks identified for the search/match and calculate the difference between the database pattern and your diffractogram data for cell parameters and d-spacing. 8) For nano-particle work, JADE can approximate the particle size of a pure specimen. This is accomplished by first adding the PDF for the phase in question (See part three of this section). Once the phase is loaded into the software, go to “Analyze”. In the “Analyze” tab select “Simulate Pattern”. This will bring up a new dialog box with several options. The easiest way to approximate particle size is to type in the particle size that is to be investigated into the particle size box (values are in Å, 20 in the box represents 2 nm). Hit enter and then click “Overlay”. This will use the PDF selected and simulate the pattern of diffractogram for that PDF at the specified size (Fig. 12). At this point it is useful just eyeballing the results and typing in different sizes until the sample diffractogram and the simulated pattern align. The software

259

can calculate the actual particle size, but unless all the defaults are adjusted properly for the sample investigated, the calculated particle size calculated is generally 10-fold larger. By comparing the eyeball method to fitting of the data using Scherrer’s equation there is generally a difference in particle size of ± 0.2 Å. Lattice strain % can be adjusted for a better fit, but since this fitting is for approximation purposes it is not necessary to adjust the lattice strain %.

Fig. 5: Search/Match dialog box. Left side is for the databases selected to search and the right side is for using different filters to refine the search of the databases.

260

Fig. 6: Results page for a general search/match of sample. The third column is the figure of merit (FOM) which is a measure of how good the fit is to the diffractogram. Even though Hexahydrite does not have the lowest FOM it is an extremely good fit for the diffractogram and makes sense for the sample which was taken from an acid mine drainage system with high levels of magnesium and sulfate in the system.

261

Fig. 7: One component fit for quartz.

262

Fig. 8: Zooming in on the diffractogram focuses in on the major un-matched peaks for the sample. Inorganic phase that fits best is sodium chromium sulfate hydrate. Quality of fit is assessed by looking at the entire diffractogram instead of zoomed in area.

263

Fig. 9: Fit of a sample with a better FOM than sodium chromium sulfate, but the mineral is definitely not in the sample since most of the peaks it is fitting with are fit perfectly with hexahydrite.

264

Fig. 10: Final fit of the New Idria acid mine drainage evaporite deposit. . Sample is properly fit with using hexahydrite (MgSO4 6H2O) and sodium chromium sulfate hydrate (NaCr(SO ) .12H O). 4 2 2

265

Fig. 11: Cell refinement dialog box. This box allows for the refinement of the unit cell parameters for a selected species along with calculating differences between the database values and the experimental values. This box also allows for the refinement of d-spacing for the various crystallographic planes of the species selected.

266

Fig. 12: Analysis of ZnS nano-particle size. The easiest method for approximating particle size is adjusting the Crystalline Size (currently 27 Å), clicking overlay, and eyeballing the results. Lattice strain % can also be adjusted for a better fit, but this particle size fitting is just for approximating the size of nano-particles.

267