MINERAL CARBONATION FOR SUBSURFACE CARBON STORAGE:
AN EXPERIMENTAL INVESTIGATION OF OLIVINE ((Mg,Fe)2SiO4)
DISSOLUTION AND CARBONATION
A DISSERTATION SUBMITTED
TO THE DEPARTMENT OF CHEMICAL ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPY
Natalie Caryl Johnson
August 2014
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/wx406mf6583
© 2014 by Natalie Caryl Johnson. 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.
ii
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Gordon Brown, Jr, Co-Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Thomas Jaramillo, Co-Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Katharine Maher
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Robert J. Rosenbauer
Approved for the Stanford University Committee on Graduate Studies.
Patricia J. Gumport, Vice Provost for 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 inUniversity Archives.
iii
iv
v
Abstract
Concern about effects of climate change motivates the development of carbon
capture and storage (CCS) technologies that facilitate both the capture of carbon
dioxide (CO2) from point sources, such as power plants, and injection of CO2 into the
subsurface for long-term storage. Compared with other approaches to reduce CO2
emissions, the advantages of CCS technologies include the possibility of near-term,
large-scale implementation and reduction of CO2 emissions without major grid
infrastructure changes. However, ensuring the stability of CO2 in the subsurface for
hundreds to thousands of years is a major technological challenge. One approach for
improving the stability is to convert CO2 into a mineral via chemical reactions, a
process known as "mineral carbonation." Carbon dioxide reacts with divalent cations
(e.g., Mg) to form carbonate minerals (e.g., magnesite, MgCO3) that are stable and
negatively buoyant in the subsurface. Field studies have established that mineral
carbonation occurs in nature, but the kinetics of the reaction are not fully understood.
Of particular interest is the formation of a Si-rich surface layer that may passivate the
surface of the reacting minerals.
In order to investigate the reaction kinetics of mineral carbonation reactions at
conditions representative of geologic storage, we performed a series of batch reactions
at 60 °C and 100 bar pCO2 using Twin Sisters olivine ((Mg1.84Fe0.16)SiO4) as the
reactive silicate mineral. Results demonstrate that mineral carbonation is feasible at
subsurface storage conditions, with reaction extent reaching 7 mol% carbonated over
94 days. The reaction rate is strongly affected by silicic acid concentration, [H4SiO4],
and decreases two orders of magnitude as H4SiO4 approaches amorphous silica
(SiO2(am)) saturation, but remains constant for up to 94 days after that at a rate of 8±2
x 10-14
mol cm-2
s-1
. During the time frame of these experiments, the rate does not
depend strongly on pH.
We consistently observed that a Si-rich layer forms on reacted olivine grains in
less than 2 days, persists for 94+ days, and increases in thickness with reaction extent.
A high-resolution transmission electron microscopy (HR-TEM) study of mineral
cross-sections showed two distinct types of amorphous surface layers, both rich in Si
vi
(or depleted in Mg). Analysis of olivine reacted with a 29
Si isotopic tracer using a
Sensitive High Resolution Ion Microprobe Reverse Geometry (SHRIMP-RG) resulted
in strong evidence for the presence of a precipitated layer on olivine reacted for 19+
days but not on olivine reacted for 2 days.
Observations from the HR-TEM and SHRIMP-RG studies allowed us to
generate a new model for the formation of Si-rich layers on multi-oxide silicate
minerals. Previous studies have suggested either a leached surface layer or a
precipitated surface layer. In contrast, we have evidence that both types of surface
layers occur simultaneously. The leached or "active" layer forms in the first hours to
days of reaction and persists for 19+ days; the removal of Mg and Si from this layer
controls the bulk dissolution rate. The precipitated layer forms after the bulk solution
reaches saturation with respect to am-SiO2 and grows in thickness as the reaction
continues at a constant rate, suggesting that it is does not passivate the surface. New
knowledge linking the measured dissolution rate with surface dynamics is useful for
the design of an engineered system for subsurface (or ex situ) carbon storage. The
model presented here for Si-rich surface layer formation on olivine may be applicable
to other multi-oxide silicate minerals, providing a new framework for geochemical
models that include mineral dissolution.
vii
Acknowledgements
I begin by expressing my sincere gratitude to my primary dissertation advisor,
Professor Gordon E. Brown, Jr. for his guidance, advice, and support during my time
at Stanford. Gordon's combination of an astute attention to detail and impressive
ability to see "the big picture" greatly helped me achieve focus by both motivating me
to believe in my work and helping me to understand the importance of detail in
scientific work. I'm grateful that he took me in to his research group despite my
woeful lack of geological background and shared with me a piece of his incredible
breadth and depth of geochemical knowledge. Countless times I walked into his office
confused about my data thinking "this time I'm going to stump him for sure" and each
time, he was able to suggest a reference (sometimes quite obscure) to help me find my
way. I also owe a great amount of thanks to Professor Kate Maher, who was always
willing to listen to my ideas, answer my questions, and help me work through topics
that were new to me. Kate patiently walked me through a long list of calculations and
geochemical models and seemed to always have time to help me, even when her
schedule was clearly packed.
Dr. Robert Rosenbauer and Dr. Burt Thomas at the US Geological Survey
were invaluable in teaching me how to perform experimental work. Burt learned how
to operate the lab as a post-doctoral scholar at about the time I began graduate school,
and he helped me establish lab procedures that I would follow for the remainder of my
PhD work. Bob provided useful advice and tips for running successful experiments,
gleaned from decades of experience working in the "rocker" lab. I'm grateful to
Professor Dennis Bird for providing guidance on experimental conditions, as well as
for his many helpful consultations on both geochemical modeling and manuscript
drafts. Professor Tom Jaramillo was my co-advisor in Chemical Engineering, and
while I didn't have much contact with him during my PhD work, I'm grateful to him
for his assurance early in my graduate career that he would be my advocate within the
chemical engineering department, should I need it (thankfully I didn't). Finally, my
work at Stanford would not have been possible without a generous grant from the
Global Climate and Energy Program (GCEP) at Stanford, for which I am very
viii
appreciative.
While I have already mentioned all the members of my reading committee, I
would like to acknowledge them separately here. Thank you to Gordon, Kate, Bob,
and Tom for helping me put years of work into a cohesive story and for your ideas and
comments on this dissertation. I would also like to thank Professor Sally Benson for
serving as my committee chair, her years-long support of our project, and for the
conversations she and I have had at various GCEP events.
Without support from friends, my graduate career would have been much more
difficult. I'm grateful to members of my chemical engineering class, who provided
both emotional and intellectual support during our first year of classes, research
rotations, and the pre-qualification and qualification exams. The members, past and
present, of the Brown and Maher research groups have shared many ideas, hours in
lab, and beers with me over the years, for all of which I am thankful. Thank you also
to the various roommates I've had over the years--having a welcoming community at
home greatly helped sustain a social life, even when research consumed many hours
per day. I'm grateful especially for the time spent in the beautiful California wilderness
and the people who traveled there with me, from chemical engineering friends to
housemates to fellow instructors from the Outdoor Education Program at Stanford.
Finally, college friends who I've kept in touch with have been a wonderful extended
support system as well as a welcome connection with life outside of graduate school.
I also would like to thank my family for their unconditional love, and support.
My parents especially encouraged me to apply to graduate school then accept the offer
from Stanford, despite its distance from their home in Massachusetts. I'm grateful for
the many hours they have spent listening to me talk about the frustrations of research
as well as their enthusiasm in celebrating my successes. I'm also thankful that my
brother, as a fellow chemical engineer, is often willing and excited to "talk shop" with
me.
This section would be woefully incomplete without the inclusion of my fiancé,
Ariel. Ariel has been a major force in my life for most of my time at Stanford. He has
the uncanny ability to know when I need a hug versus when I need a push, and without
ix
his unwavering belief in me, my work would be much weaker than it is today.
Together we adopted a dog, named Bear, and I can't express how wonderful it is to
return home to a slobbery, furry monster who's love is totally ignorant of the
difficulties of experimental work. Ariel is my teammate, my coach, my loudest
supporter (sometimes all at once!) and I love him very much.
x
xi
Table of Contents
Abstract ....................................................................................................................... v
Acknowledgements ...................................................................................................vii
Table of Contents ....................................................................................................... xi
List of Tables .......................................................................................................... xiii
List of Figures .......................................................................................................... xiv
Chapter 1: Introduction to mineral carbonation as a subsurface carbon storage
mechanism .................................................................................................................. 1
1.1 Climate Change .................................................................................... 2
1.2 Carbon Capture and Sequestration (CCS) ............................................. 3
1.3 Mineral Carbonation ............................................................................. 7
1.4 Objectives .......................................................................................... 12
1.5 Summary of Chapters ......................................................................... 13
1.6 References .......................................................................................... 17
Chapter 2: Olivine dissolution and carbonation under conditions relevant for in situ
carbon storage ........................................................................................................... 23
Abstract...................................................................................................... 24
2.1 Introduction .......................................................................................... 25
2.2 Materials and Methods ......................................................................... 29
2.3 Results ................................................................................................. 35
2.4 Discussion ............................................................................................ 42
2.5 Conclusions .......................................................................................... 54
2.6 Acknowledgments ................................................................................ 56
2.7 References ............................................................................................ 57
Chapter 3: The role of the Si-rich surface layer in olivine dissolution 1: High-
resolution TEM study ................................................................................................ 63
Abstract...................................................................................................... 64
3.1 Introduction .......................................................................................... 65
3.2 Methods ............................................................................................... 68
3.3 Results ................................................................................................. 73
3.4 Discussion ............................................................................................ 77
3.5 Conclusions .......................................................................................... 90
3.6 Acknowledgments ................................................................................ 93
3.7 References ............................................................................................ 94
Chapter 4: The role of the Si-rich surface layer in olivine dissolution 3: Spatially and
temporally resolved ................................................................................................... 99
xii
incorporation of an isotopic tracer (29
Si).................................................................... 99
Abstract ................................................................................................... 100
4.1. Introduction ...................................................................................... 101
4.2. Methods ............................................................................................ 103
4.3. Results .............................................................................................. 109
4.4. Discussion ........................................................................................ 114
4.5. Conclusions ...................................................................................... 129
4.6 Acknowledgments ............................................................................. 131
4.7 References ......................................................................................... 132
Appendix 1: The role of the Si-rich surface layer in olivine dissolution 2: A spatially
resolved surface kinetic model ................................................................................ 137
Abstract ................................................................................................... 138
A1.1. Introduction.................................................................................... 140
A1.2. Methods ......................................................................................... 144
A1.3. Results ........................................................................................... 148
A1.4. Discussion ...................................................................................... 150
A1.5. Conclusions.................................................................................... 171
A1.6. Acknowledgments .......................................................................... 173
A1.7. References ..................................................................................... 174
Appendix 2: Experimental Protocols ....................................................................... 179
A2.1 Gold Bag Preparation ...................................................................... 179
A2.2 Repairing Gold Bags ....................................................................... 181
A2.3 Improving the seal between Au bag and Ti collar ............................ 182
A2.4 Assembling the reactor .................................................................... 184
A2.5 Sampling ......................................................................................... 194
A2.6 Taking down an experiment ............................................................ 196
A2.7 Other Procedures ............................................................................. 198
A2.7.1 Using the Furnace ..................................................................... 198
A2.7.2 Pressurization System ............................................................... 199
A2.7.3 Coulometer Operation ............................................................... 201
Appendix 3: Redox State of Reaction ...................................................................... 203
Appendix 4: Derivation of diffusion coefficient from Lasaga (1979). ...................... 205
xiii
List of Tables
Table 2.1: Experimental data for four experiments. .................................................. 34
Table 2.2: Experimental data for four experiments. .................................................. 37
Table 2.3: Geometric surface area-normalized reaction rates for four experiments, with
BET-normalized rates in parenthesis. ....................................................................... 41
Table 3.1: Conditions of four experiments. ............................................................... 69
Table 3.2: Solution data from four experiments ........................................................ 73
Table 3.3: Olivine dissolution rates from four experiments ....................................... 79
Table 4.1: Duration, degree of mixing, and additive concentrations for five
experiments. ........................................................................................................... 105
Table 4.2: Resin cleaning and Si purification procedure adapted from Georg et al.
(2006). ................................................................................................................... 105
Table 4.3: Steady-state dissolution and precipitation rates for 19-R and 19-S ......... 118
Table 4.4: Model parameters for 19-R and 19-S ..................................................... 125
Table A1.1: Summary of experiments from Chapters 2 and 3 used in model
development. .......................................................................................................... 144
Table A1.2: Mg isotope data for experimental olivine and experiment 3 from Chapter
2 along with average values for reference materials, where n is the number of times
the individual sample was analyzed to obtain the average value and standard deviation.
............................................................................................................................... 147
Table A1.3: Model parameters and values applied to experiments 1-4, 2R, 19S, 2S,
19S. ........................................................................................................................ 160
xiv
List of Figures
Figure 1.1: Mitigation cost increase over a base scenario when various technologies
are excluded from the mitigation portfolio for two target atmospheric CO2
concentration ranges. Scenarios without CCS are the most costly, followed by
bioenergy. Limiting solar and wind to 20% of the energy portfolio and ceasing to build
new nuclear power plants have the lest effect on mitigation cost. Adapted from
Edenhofer et al., 2014. ................................................................................................ 3
Figure 1.2: Diagram of four subsurface CO2 trapping mechanisms: physical, residual,
solubility, and mineral. Figure adapted from personal correspondence with Pablo
Garcia Del Real. ......................................................................................................... 7
Figure 2.1: X-ray diffraction patterns for unreacted and reacted olivine. Magnesite
peaks are visible in the reacted spectra . .................................................................... 31
Figure 2.2: X-ray photoelectron spectroscopy data for unreacted olivine, and olivine
reacted for 2, 4, and 74 days. The peak intensity of Mg decreases with increasing
reaction time relative to the Si peak, indicating that the mineral surface becomes more
Si-rich as the reaction progresses. ............................................................................. 32
Figure 2.3: Elemental concentrations versus time for magnesium (A), silicon (B), iron
(D), and alkalinity versus time (C) for four experiments. Insets show the same y-axis
for the initial 4 days of reaction. Experiments 1 and 2 contain 0.5 M NaCl;
Experiments 3 and 4 have 0 M NaCl. ........................................................................ 35
Figure 2.4: Images of unreacted (A, B) and reacted (C-F) olivine from Experiment 2.
Unreacted olivine grains are of uniform size. After reaction, the solids are dominated
by large (>100 µm) silica sheets (C) and many grains are partially coated with silica
(D). Magnesite nodules grow via a spiral growth pattern on the silica (coatings and
sheets) (D, F), which itself has a structure composed of fused spheres (E). EDS spectra
of magnesite nodules (F) and silica (E) are inset........................................................ 38
Figure 2.3: Saturation of magnesite (A), amorphous silica (B), and siderite (C) vs. time
as well as calculated pH vs. time (D). Data presented were calculated from
Experiments 1, 2 and 4. Saturation is presented as the logarithm of the reaction
quotient Q divided by the equilibrium constant K. Dashed lines in A-C represent
equilibrium between the solid phase and the reaction solution (logQ/K=0 at Q=K). .. 40
Figure 2.6: The rate of olivine dissolution, calculated from both Mg and Si data from
Experiment 3, decreases as the silica saturation state increases during the initial 3 days
of reaction. Non-stoichiometric dissolution occurs during the first time interval (due to
incongruent dissolution) and the last (presumably due to amorphous silica
precipitation). ........................................................................................................... 46
Figure 2.7: A: Experimental rate data (normalized to geometric surface area and
xv
calculated from Mg concentrations) as a function of pH. Deviation at the lowest pH
(first time point) is due to a combination of incongruent dissolution creating an
artificially high rate and dissolution of fine particles not removed by washing.
Experimental data, particularly from 1 and 2, do not show a pH-dependence at pH >
4.5. Horizontal uncertainty is defined by the pH values at the two data points used to
calculate each rate. For comparison, one data point from Wang and Giammar (2013) is
included. This rate was measured over 8 hours at 50 °C, 100 bars pCO2, and 0 M
NaCl; we corrected it by a factor of 40 to account for the lower forsterite concentration
in their experiments (0.5 g/L compared to our 20g/L). B: Long term (10-70 days)
dissolution rates compared to three published models for forsterite dissolution
(Pokrovsky and Schott, 2000b; Hanchen et al., 2006; Rimstidt et al., 2012). Rates
from experiments with 0.5M NaCl are consistent with the models, the rate from
Experiment 4 (no added electrolyte) is lower. All rates shown here were calculated
before Mg-carbonate precipitation occurred. ............................................................ 51
Figure 3.1: Steps of IFFT analysis. First, the TEM image is taken using the
microscope (A). The image is then transformed using a Fast Fourier Transform, which
yields a second image in the frequency domain (B). Bright spots on this image
represent periodicity, so those particular frequencies are selected and the rest are
masked (C). An Inverse Fast Fourier Transform is then applied to the masked
frequency domain image to recreate the original image showing only the regions of
periodicity (D). The periodicity in (D) is distorted due to aliasing (relatively low
resolution in the image results in a high frequency periodicity from the original image
(A) appearing to have a lower frequency). This IFFT image was then overlaid on the
original TEM image (E) and false color was added to show periodic regions and the
gold surface coating (F). ........................................................................................... 72
Figure 3.2: Mg (A) and Si (B) concentrations as a function of time for four
experiments. Amorphous silica saturation is indicated by the dashed line in (B). ...... 74
Figure 3.3: Representative TEM crystallinity maps of unreacted (A) and forsterite
from experiments 2-S (B), 2-R (C), 19-S (D) and 19-R (E). Green areas on the right of
each image are crystalline, yellow-colored regions on the left are a protective gold
layer deposited on the mineral surface (colored version online). Uncolored areas are
amorphous. The scale bar in (E) applies to all reacted forsterite (B-E). ..................... 75
Figure 3.4: TEM/EDS linescans showing Mg/Si ratio as a function of depth in
forsterite reacted without mixing (A) and with mixing (B). Forsterite from all four
experiments had a Si-rich (Mg-depleted) surface layer at least 20 nm thick. Linescans
showing Fe/Si ratios are also shown for the well-mixed experiments (C) and both have
a peak of Fe associated with the Si-rich layers. Ratios were calibrated using spectra
from deep in the mineral. .......................................................................................... 76
Figure 3.5: Forsterite dissolution rate as a function of Si-rich layer thickness. The rate
does not decrease with increasing layer thickness; in fact, within the two different data
sets (rocking vs. stationary), show that the rates increase with Si-rich layer thickness.
xvi
Error bars represent 2-sigma uncertainty due to the few (2-5) data points available for
each rate calculation. No correlation between rate and layer thickness is implied
beyond the data presented here.................................................................................. 78
Figure 3.6: Cross sectional diagram of the reacted mineral surface, specifically for
experiment 19-R. The three different Si-rich layers and their defining characteristics
are visible from top down: Precipitated layer, active layer, and Mg-depleted forsterite.
................................................................................................................................. 82
Figure 3.7: Simple diffusion model in which the diffusion coefficient, D*, is a
function of Mg concentration in addition to diffusivity of protons (DH) and of Mg
(DMg), plus a structural factor (α) (A). Decreasing from α 0 to -0.3 results in flattening
the profile, and reducing it further to -0.9 creates a minimum value of D* at c≈0.1.
Changing DMg results in a shift of the concentration gradient curve (B) and changing
DH affects both the steepness and depth of the curve (C). Data from the two 19 day
experiments can be fit with this model and different values of DH (D). Vertical bars
indicate the transition from amorphous to crystalline material, and the model was
shifted to the right to align with the “active” layer. .................................................... 88
Figure 4.1: Silicon concentrations and 29
Si/30
Si ratios as a function of time for 19-R
(A) and 19-S (B), with amorphous silica saturation plotted for comparison. Ratios of
Mg/Si vs . time (C) show that the fluid composition is elevated in Mg relative to the
mineral phase. ......................................................................................................... 110
Figure 4.2: Ratios of Mg/Si as a function of depth for 2,19-R (A) and 2,19-S(B). The
depth was calibrated using the SI-rich layer thickness as measured from TEM images
(65 nm (19-R), 40 nm (19-S), 25 nm (2-S), and 20 nm (2-R)). Isotopic ratios
(29Si/30Si) are also plotted as a function of depth for the rocking (C) and stationary
(D) experiments. Each data point is an average from at least three different forsterite
grains. ..................................................................................................................... 112
Figure 4.3: Comparison of isotopic composition of the aqueous and mineral phases for
four experiments. .................................................................................................... 113
Figure 4.4: Time-resolved Mg (A) and Si (B) concentrations for three experiments:
19-R (this study), Experiment 1 (Chapter 2), and catechol (this study). ................... 114
Figure 4.5: Dissolution rates for 19-R calculated using the isotope ratio method and
dMg/dt indicates that Mg and Si release are stoichiometric at t>5 days (A), and Si
removal rates (precipitation plus exchange) are the same within uncertainty as
dissolution rates at t> 5 days (B). The same calculations for 19-S reveal slower
dissolution and a longer transient stage (C) and precipitation begins later (D). Missing
data are either less than zero or have uncertainties that span positive and negative rates
(i.e. rates are zero within uncertainty). .................................................................... 117
Figure 4.6: Kinetic model fit to concentration data from rocking experiments (A),
xvii
stationary experiments (B), and isotope data from both experiments (C). ................ 127
Figure A1.1. Summary of previous experimental data from Chapter 2 in (A)-(C) and
Chapter 3 in (D)-(E). Vertical arrows in (A) and (D) indicate time when experiments
reached MgCO3 saturation, horizontal line with arrows indicates interval when
MgCO3 precipitation is thought to have occurred in experiment 2, based on the lack of
precipitates in experiment 1 at 74 days. Horizontal dashed lines in (B) and (E) indicate
saturation with respect to amorphous silica. Error bars represent 1 s.d. ................... 149
Figure A1.2. Magnesium isotope data for the short-term experiment 3 (no NaCl) over
3 days of reaction time. Error bars represent 1.s.d. ................................................. 149
Figure A1.3. Conceptual model for the surface kinetic model. (A) High resolution
TEM image shows the general features observed in the outer 100 nanometers of
olivine after reaction for 2 days (modified from Chapter 3), with the fluid-solid
interface to the left of the image. Average Mg/Si ratios measured by EDS are provided
for reference. The black arrows indicate the migration of the active layer into the
crystalline olivine over time, as determined by the rates of the individual reactions in
(B). (B) The conceptual model for the independent rates considered in the model and
the resulting mass balance constraints. The net rate (Rnet) is the Mg exchange rate plus
the net difference between the net <SiO2 dissolution rate and the net precipitation rate
of amorphous silica. Magnesium is removed from the active layer of volume VAL,
while >SiO2 can only be dissolved or added from the surface of the active layer. (C)
Representative HR-TEM EDS linescan of the Mg/Si ratio in reacted olivine from
experiment 19R from Chapter 3 with a Lowess curve fit (green line) compared to the
model representation (bold teal line). The grey shaded region (indicated (-)) shows the
Mg depletion relative to the assumed stoichiometric composition, while the green
shaded region (indicated (+)) shows the Mg-enrichment in the active layer relative to
model Mg/Si. To maintain mass balance the total Mg in the two zones must be equal.
............................................................................................................................... 152
Figure A1.4. Model fit to experiments 1-4. Magnesium (open symbols) and Si (closed
symbols) over the full experimental duration are shown in (A), and during the initial
incongruent dissolution phase in (B). The model fit to the combination of experiments
1 and 2 in (A) is considered the reference model. The same scheme is applied in (C)
and (D) where model results for experiments 3 and 4 are presented. The changes in
rate coefficients are indicated and the thickness of the active layer (AL, nm) is
indicated. For experiments 3 and 4, a sensitivity analysis showing variation in the AL
is provided. All experiments were conducted at 100 bar CO2 and 60˚C in the presence
of 0.5 M NaCl (Chapter 2) . .................................................................................... 162
Figure A1.5. Model fit to the combined experiments 2R and 19R (well mixed) are
shown over the full experimental duration (A), and during the initial incongruent
dissolution phase in (B). Model fit to experiments for 2S and 19S (not-well mixed)
over the same durations are shown in (B) and (C). The changes in rate coefficients
relative to the reference model are indicated. All experiments were conducted at 100
xviii
bar and 60˚C in the presence of 0.5 M NaCl (Chapter 3). ........................................ 163
Figure A1.6. Rate profiles as a function of (A) time; (B) pH; and (C) H4SiO4
concentrations for the reference model. The bar at top indicates the dominant
processes occurring from initial to steady-state dissolution. .................................... 165
Figure A1.7. Sensitivity of the steady-state dissolution rate to the individual rate
coefficients. Each rate coefficient was varied independently while the others were held
constant at the values determined for the reference model (see Table 4). The best fit to
experiments 1 and 2 is shown by the grey bar. ........................................................ 167
Figure A1.8. Model magnesium isotopic and surface layer Mg/Si evolution for
experiment 3. (A) The model change in the Mg/Si of the active layer (Mg/SiAL) is
shown in comparison to the 26
Mg of the active layer (AL) and the dissolved Mg. (B)
Comparison between model and measured 26
Mg of the aqueous Mg (uncertainties are
represented as 1 s.d.). The best fit to the data is between Mg-Fo = 0.9995 and 0.9990
and variations in Mg-Fo are shown for reference. ..................................................... 169
Figure A1.9. Model evaluation of Wimpenny et al. (2010) experimental series FO2 at
pH 3 and 25 ˚C. (A) Model fit for Mg and Si concentrations (see parameters in Table
4), (B) Comparison between measured 26
Mg to model prediction using Mg-Fo =
0.999....................................................................................................................... 170
Figure A2.1: Gold bag assembly. From left to right: gold bag with Ti collar, Au/Ti
filter in Ti head with attached sampling tube, washer, flanged compression cap
containing thrust bolts. ............................................................................................ 179
Figure A2.2: Stainless steel pressure vessel with steel head, gasket, and thrust bolts.
The gold reaction cell is contained in the vessel and surrounded by a pressure fluid (DI
water). .................................................................................................................... 186
Figure A2.3: Assembling the reactor vessel. A. Attaching the sampling valve (step 8).
B. Full assembled reactor. C. Completed support structure, top view. Potential leak
points are visible in B and C. If the leak is below the valve (B), tighten the bolt
connecting the valve to the tube. If the leak is visible from a top view (C), tighten
accessible bolts and retest. Disassembly may be required. ....................................... 188
Figure A2.4: View of 4 rocking furnaces (A) showing temperature controls and
variacs. Zoomed in image of temperature controls (B) shows the temperature read-out,
power switch, and fuse for each of the four furnaces. .............................................. 192
Figure A2.5: Sampling syringe with 3-way valve, attached to sampling port. The
sampling port consists of a stainless steel, 5/8" bolt with a teflon insert, and plastic
tubing connects the insert to the 3-way valve. ......................................................... 195
Figure A2.6: Pressurization system on the pressure bench. Valves across the front are
xix
labeled (left to right): Bleed, Pump, Reservoir, Rack 3, Rack 4, Workbench. Analogue
pressure gauge hangs above the valves. The Reservoir sits below the bench and has its
own gauge. ............................................................................................................. 200
Figure A2.7: Gas delivery system for the coulometer. Sample is injected at 3-way
valve. ..................................................................................................................... 201
Figure A3.1: Element concentration (A) and H2 pressure (B) as a function of time. 203
xx
1
Chapter 1: Introduction to mineral carbonation as a subsurface carbon storage mechanism
Natalie C. Johnson
2
1.1 Climate Change
Recent reports from the Intergovernmental Panel on Climate Change
(Edenhofer et al., 2014) and the US Global Change Research Program (Melillo et al.,
2014) emphasize both the potential large-scale impacts of climate change and the
immediate need for implementation of technologies to mitigate such impacts. Climate
change is caused by high concentrations of so-called "greenhouse gases" such as CO2,
methane, and water in the atmosphere, which let sunlight pass through the atmosphere
but trap heat that would otherwise escape. Natural sources of greenhouse gases do
exist, but recent climate models suggest that the recent upward trend of global average
temperature cannot be explained by natural sources alone, the model must also include
anthropogenic sources (Huber and Knutti, 2012). Increasing levels of greenhouse
gases in the atmosphere are already affecting our climate; from rising sea levels to
melting ice sheets (Joughin et al., 2014), and rising temperatures to increasing
numbers of large storms (Melillo et al., 2014). Other consequences include increases
in U.S. summer time ozone concentrations of up to 70% by 2050 (Pfister et al., 2014)
and ocean acidification from increased dissolved CO2 (Melillo et al., 2014). Whereas
there are many unequivocal observations that our climate is already changing (e.g.
Melillo et al., 2014), reduction of emissions of CO2 and other greenhouse gases today
would reduce future changes.
Given the global scale of climate change, it's unlikely that any single
technology will have an impact large enough to sufficiently mitigate climate change
worldwide, where "mitigate" is defined as reducing the sources or improving the sinks
of CO2 and other greenhouse gases. Instead, a suite of technologies in a variety of
fields like power production, transportation, energy use, and land use will be needed.
In fact, a comparison of CO2 mitigation scenarios in which specific technologies are
excluded shows that carbon capture and storage (CCS) and bioenergy are perhaps the
most important technologies, because mitigation may cost up to 300% more than a
base scenario when one or both are not a part of the mitigation strategy (Fig. 1.1)
(Edenhofer et al., 2014). One reason that biofuels and CCS are so important is the
prediction that the carbon intensity of energy sources worldwide is expected to remain
3
constant for the next 100 years (Edenhofer et al., 2014). Thus, technologies that either
directly replace (e.g. biofuels in the transportation sector) or lower the greenhouse gas
emissions of carbon-intensive energy (e.g. CCS on carbon-based electricity generation
facilities) are essential to successful CO2 mitigation.
1.2 Carbon Capture and Sequestration (CCS)
One potential technology to mitigate climate change involves capturing carbon
dioxide from point sources (i.e. power plants, cement plants) and injecting it into the
subsurface. Storing CO2 in the subsurface rather than emitting it to the atmosphere
could substantially reduce the rate at which atmospheric CO2 concentrations are
increasing and thus reduce the probability of catastrophic climate change. Technology
Figure 1.1: Mitigation cost increase over a base scenario when various technologies are
excluded from the mitigation portfolio for two target atmospheric CO2 concentration
ranges. Scenarios without CCS are the most costly, followed by bioenergy. Limiting solar
and wind to 20% of the energy portfolio and ceasing to build new nuclear power plants
have the lest effect on mitigation cost. Adapted from Edenhofer et al., 2014.
4
for safe injection of CO2 into the subsurface already exists, and oil companies have
been using it to improve oil yields for decades as one component of enhanced oil
recovery (EOR). Currently, oil companies use CO2 extracted from the subsurface in
EOR processes, but the potential exists for waste CO2 to be used instead (Latil, 1980).
Also, CO2 could be injected into other types of rocks that do not host oil, such as
saline aquifers (sedimentary deposits) or igneous complexes (such as basalt).
The first step of CCS is the capture of CO2 from point sources such as power
plants. While much research regarding technologies for carbon capture is underway,
this document will not explicitly address the topic of capture. However, successful
carbon capture is essential to carbon sequestration, so it bears mention here.
After capture from a point source (e.g. coal or natural gas power plant), carbon
dioxide is injected as a compressed liquid (or co-injected with water to form a single,
aqueous phase (Gislason et al., 2010)) through cement-lined wells to depths of at least
800 m, but more often 1.2 to 2.5 km (Hosa et al., 2010), where the CO2 becomes a
supercritical fluid. In the case of co-injection, the carbonated water may be injected to
shallower depths, ranging from 400-700 meters (Oelkers et al., 2008). The rocks into
which CO2 is injected must be porous (or have induced permeability), and their pores
contain water, dissolved species such as NaCl, and possibly oil/natural gas if the
region is a depleted oil field. CO2 spreads from the injection point and fills the pores,
mobilizing the water (and oil, if present, which can then be removed from the
subsurface at extraction wells). Technological hurdles to large-scale implementation
of subsurface carbon storage remain, and center around long-term stability,
monitoring, induced seismicity, and water quality concerns. Other, non-technical
obstacles include the high cost of CCS, long-term liability disputes, and land rights
(Wilson et al., 2007).
The goal of CCS is to store CO2 in the subsurface for hundreds to thousands of
years, but supercritical CO2 is less dense than water and thus acts as a buoyant fluid.
Leakage to the atmosphere represents wasted effort and thus must be minimized. In
order to contain CO2 in the subsurface, one or more physical "caps" must be present
above the reservoir into which the CO2 is injected. However, leakage may still occur
5
through cracks in the caprock or old wells that were improperly retired (Wilson et al.,
2007). Site selection is very important to minimize the leakage potential, and several
levels of monitoring will be necessary to detect leaks and potential leaks early so that
they can be fixed before substantial volumes of CO2 are lost. CCS monitoring is
another area of current research, and potential technologies include gravity monitoring
(Alnes et al., 2011), surface monitoring (Madsen et al., 2009), and dissolved inorganic
carbon measurements of subsurface reservoirs (Romanak et al., 2012).
Other potential hazards caused by subsurface carbon storage include induced
seismicity with earthquakes up to magnitude 4.5 (Cappa and Rutqvist, 2011) and loss
of water quality (Little and Jackson, 2009). Both of these hazards can be prevented by
proper well construction and reservoir pressure monitoring. Induced seismicity occurs
when the pressure inside the reservoir exceeds a critical value, most likely causing
nearby faults to slip (Cappa and Rutqvist, 2011). A thorough understanding of the
horizontal and vertical stress applied to the caprock and avoidance of active faults plus
monitoring the reservoir pressure during injection will reduce the probability of large
seismic events. Water quality degradation is another potential effect of induced
seismicity, if a fault opens that connects the injection reservoir with a reservoir that
contains drinking water. Generally injection reservoirs are several thousand feet deep,
while drinking water reservoirs are only hundreds of feet deep, so the larger concern
with water quality relates to leakage through the well casings. In either case, leakage
of CO2 into drinking water may reduce the pH of the reservoir and induce
geochemical reactions which can result in increasing concentrations of heavy metals
such as arsenic (Little and Jackson, 2009). Regulations surrounding well construction
would reduce the risk of leakage through the wells themselves.
Non-technical challenges surrounding CCS center around cost, but also include
land rights and assignment of long-term liability (Wilson et al., 2007). First of all,
CCS is expensive, and as long as power plants can emit CO2 freely, they have no
economic incentive to invest in CCS. Research on new technologies may decrease the
cost of CCS, but regulation will almost certainly be necessary to make CCS
economically beneficial relative to releasing CO2 as a waste gas to the atmosphere.
6
Land rights are also a potential issue and the way that rights are assigned to the
subsurface varies by state (Klass and Wilson, 2009). However, subsurface land rights
issues have existed for many successful industries, from gold mining to oil extraction,
so they are not insurmountable. Finally, long-term liability is an unresolved issue.
Carbon Capture and Sequestration is ideally a very long-term project (hundreds to
thousands of years), and assigning liability to a company for that length of time is not
practical. It's possible that liability could rest with the injecting entity for some length
of time, then shift to the public (governmental agencies), but questions remain about
who would be responsible for the clean-up of a injection site that leaks hundreds of
years after injection (and likely after the injecting company no longer exists) (Wilson
et al., 2007). However, reducing the likelihood of leakage would make the liability
issue less prominent.
Once CO2 is in the subsurface, storage occurs via a variety of mechanisms
listed here from first (and least stable) to last (and most stable) and shown in Fig. 1.2.
Physically containing CO2 under an impermeable caprock is known as "physical
trapping" and is necessarily the first of four trapping mechanisms to occur. Thus, any
reservoir that is a candidate for subsurface storage must have several layers of
impermeable rock above the injection region in order to physically trap the buoyant
fluid. The second trapping mechanism begins to occur immediately, but becomes
increasingly important with time and is called "residual trapping." In this mechanism,
capillary forces trap CO2 in rock pores. "Solubility trapping" is the third mechanism
and describes the dissolution of CO2 into the aqueous phase, resulting in carbonic acid,
bicarbonate ions, carbonate ions, and protons. This will be discussed in detail later.
Finally, the last mechanism is "mineral trapping," which is the focus of this
dissertation. Mineral trapping occurs when dissolved CO2 reacts chemically with
dissolved cations, such as calcium, magnesium, and iron, to make carbonate minerals.
This last mechanism is the most stable over thousands to millions of years because the
CO2 is trapped as a non-buoyant solid. Under certain conditions (e.g. acidic fluids)
carbonate minerals do dissolve, but the probability of leakage is much lower if the
bulk of the CO2 injected is converted to minerals.
7
Thus, optimizing CO2 injection for mineral trapping by choosing rocks rich in
divalent cations (Ca, Mg, Fe) and with sufficient porosity and surface area may
minimize risks and costs associated with subsurface carbon sequestration. Successful
mineral trapping of CO2 may simplify liability issues, reduce public concern over the
safety of injection, and increase the effectiveness of long-term storage relative to
injection in rocks that are ill-suited for mineral trapping.
1.3 Mineral Carbonation
Mineral carbonation describes the series of chemical reactions that make up
"mineral trapping," the fourth and most stable trapping mechanism that occurs in
subsurface carbon storage. In short, host rocks dissolve in formation water, releasing
divalent cations (Ca, Mg, Fe) that react with dissolved CO2 (carbonate and bicarbonate
anions) to form solid carbonate minerals. Carbonate minerals are thermodynamically
stable; in fact, they are the most stable form of carbon. However, the kinetics of the
Figure 1.2: Diagram of four subsurface CO2 trapping mechanisms: physical,
residual, solubility, and mineral. Figure adapted from personal correspondence
with Pablo Garcia Del Real.
8
reaction are slow relative to industrial timescales. Other technical challenges involve a
large volume change of the reaction (the products are 92% larger in volume than the
reactants) and potentially passivating surface layers that form on the host minerals.
Each step of the process is described in detail in the following sections.
1.3.1 Olivine ((Mg,Fe)2SiO4) Dissolution
In order for mineral carbonation to occur, the host rocks must first release
divalent cations through a dissolution process. Silicate mineral dissolution has been
studied for many decades in order to understand surface weathering processes, but
only recently have researchers begun studying mineral dissolution at elevated
temperature and CO2 pressure, the conditions that define a subsurface storage
scenario. The dissolution rate of olivine (and multi-oxide silicate minerals in general)
increases with increasing temperature and decreasing pH (Blum and Lasaga, 1988;
Chen and Brantley, 2000; Giammar et al., 2005; Hänchen et al., 2006; Olsen and
Rimstidt, 2008; Pokrovsky and Schott, 2000a; Prigiobbe et al., 2009; Rimstidt et al.,
2012; Shirokova et al., 2012; Wang and Giammar, 2013; Welch and Banfield, 2002;
Wogelius and Walther, 1992, 1991). The addition of CO2 may enhance the reaction,
perhaps due to pH effects (Golubev et al., 2005; Wang and Giammar, 2013), though
others have found that bicarbonate ions directly enhance the rate of carbonation (Chen
et al., 2006). More data on olivine dissolution in the presence of CO2 must be
generated before the effect of CO2 can be accurately quantified.
During the dissolution of olivine (as well as other multi-oxide silicate
minerals), a Si-rich phase forms on the reacting mineral surface (Bearat et al., 2006;
Daval et al., 2011; King et al., 2010; Pokrovsky and Schott, 2000a; Sissmann et al.,
2013; Zakaznova-Herzog et al., 2008). There are two, competing theories for the
formation mechanism of this layer. The first theory is that it forms via a leaching
process, in other words, Mg (and other cations) are released preferentially and residual
silicon tetrahedra rearrange and polymerize to form the surface layer (Hellmann et al.,
1990; Pokrovsky and Schott, 2000b; Ruiz-Agudo et al., 2012; Zakaznova-Herzog et
al., 2008). Alternatively, the surface layer could be formed via a
dissolution/reprecipitation process, in which the mineral dissolves congruently but
9
local areas of high H4SiO4 result in precipitation of amorphous silica at the mineral
surface (Hellmann et al., 2013, 2012; King et al., 2011, 2010). There is also
disagreement about the extent to which the Si-rich layer passivates the olivine surface,
with some studies observing near complete termination of dissolution (Daval et al.,
2011, 2009; Hellmann et al., 2012) and others seeing continued dissolution even in the
presence of thick surface layers (King et al., 2010).
Finally, ferrous iron makes up 5-15% of the cation sites in the mineral olivine,
and may play an important role in olivine dissolution. For example, Fe2+
may oxidize
to Fe3+
during the reaction and then stabilize the Si-rich layer and slow the reaction
(Saldi et al., 2013; Sissmann et al., 2013). As the reaction progresses and conditions
become more reducing, the ferric iron may reduce back to ferrous iron, which is
associated with the rapid dissolution of the Fe-phase and resumption of olivine
dissolution (Sissmann et al., 2013).
1.3.2 Carbon Dioxide Dissolution and Dissociation
When carbon dioxide and water are in contact with each other, some CO2
dissolves into the aqueous phase and some water evaporates into the CO2 phase. At
equilibrium, the net concentrations of CO2 in water and water in CO2 remain constant,
though exchange continues to occur. The amount of dissolved CO2 depends on the
pressure and temperature of the system as well as the ionic strength of the aqueous
phase. The solubility increases with increasing pressure, but decreases with increasing
ionic strength (Duan and Sun, 2003). The solubility generally decreases with
increasing temperature, though at very high pressure (>200 bar), there exists a
solubility minimum at a temperature around 350 K and solubility increases at lower
and higher temperatures (Duan and Sun, 2003). When considering conditions relevant
for mineral carbonation, one can assume that CO2 solubility decreases with increasing
temperature.
After dissolving in water, CO2 molecules are hydrated to form carbonic acid,
H2CO3. Carbonic acid then dissociates to form protons (H+), bicarbonate ions
(HCO3-), and carbonate ions (CO3
-2). The relative concentrations of ions are
determined by equilibrium constants, which like the solubility, depend on temperature,
10
pressure, and ionic strength, though the relationships are not monotonic. The
equilibrium constants can be calculated using a model (Li and Duan, 2007) and then
used to calculate the equilibrium pH. The addition of CO2 to water causes the pH to
drop and the carbonate system acts as a pH buffer.
1.3.3 Secondary Phase Precipitation
Dissolution of silicate minerals can result in precipitation of several secondary
phases. If, as in the case of olivine, the mineral consists of only divalent cations and
silicon tetrahedra, the secondary phases are limited to metal-carbonate minerals and
silicon dioxide. If aluminum is present, clay minerals may precipitate instead of silica
and could incorporate reactive divalent cations. An advantage of performing
experiments with a relatively simple mineral such as olivine is that the complexity of
secondary phases is greatly reduced to only two phases: silicon dioxide and Mg-
carbonate. The Mg-carbonate could be present as hydrated minerals (e.g.
Hydromagnesite, Nesquehonite) or an anhydrous form (Magnesite). Silicon dioxide is
generally present as amorphous silica in low-temperature systems, where it's solubility
is low relative to other silicate minerals (Steefel and Van Cappellen, 1990) and
precipitation kinetics of less soluble SiO2 phases are prohibitively slow. Other possible
secondary phases include phyllosilicate minerals (Hellmann et al., 2013) such as
serpentine minerals (King et al., 2010; Lafay et al., 2014), especially at higher
temperatures (Sissmann et al., 2013).
Magnesite precipitation is slow at temperatures relevant for subsurface carbon
storage and has been hypothesized to limit the rate of conversion of olivine to
magnetite (Saldi et al., 2012). The rate increases with increasing temperature and
increasing saturation state and appears to be controlled by a spiral growth mechanism
(Saldi et al., 2012). However, conversion of Mg-silicates to magnesite has been
observed by several studies at a range of temperatures (Felmy et al., 2012; Giammar et
al., 2005), though at temperatures below 50 °C an intermediate, hydrated carbonate
mineral may form first and then transform to magnesite (Felmy et al., 2012;
Hopkinson et al., 2008).
11
1.3.4 Ex situ vs. in situ mineral carbonation
Mineral carbonation could be completed via an ex situ or an in situ process,
both processes have advantages and disadvantages. Ex situ mineral carbonation would
occur in an industrial facility and use an engineered process to achieve maximum
conversion of silicate minerals to carbonate minerals at an optimized cost and
conversion time. However, the silicate minerals would require extraction and transport
to the facility. The scale of the required mining operation is mammoth; an average
coal fired power plant that generates 4 MW of electricity per year would require 6.4
Mt of forsterite to mineralize all the CO2 produced (Oelkers et al., 2008). In addition,
large amounts of magnesite would be produced (2.4 Mt in the example above,
(Oelkers et al., 2008)), which would need to be disposed of and would likely contain
traces of heavy metals found in olivine, such as chromium and nickel. One result of
the mining and disposal needed for successful ex situ carbonation is an increased cost
in electricity by up to 370% (Giannoulakis et al., 2014). Life cycle assessments of ex
situ mineral carbonation repeatedly find that though it succeeds in reducing
greenhouse gas emissions, the net environmental effect is negative (Bodénan et al.,
2014; García-Gusano et al., 2013; Giannoulakis et al., 2014). The expected cost of an
optimized process for ex situ mineralization ranges from $54-$427 per ton of CO2,
depending on process considerations and the feedstock used (O’Connor et al., 2004).
Engineering an efficient ex situ mineral carbonation process is more feasible
than an in situ process because every step can be carefully controlled; however, the
negative outcomes associated with ex situ mineralization make it unappealing for large
scale CO2 storage. On the other hand, in situ mineralization has the potential to be less
expensive and have fewer negative impacts on the environment and human health
(Bodénan et al., 2014) because CO2 is transported to injection sites via pipelines and
injected directly into the subsurface. Site selection for in situ storage is critical, as the
subsurface must contain rocks with both divalent cations and sufficient permeability
for injection (Oelkers et al., 2008). Disadvantages of in situ storage surround a lack of
process control and difficulty predicting the fate of injected CO2 (e.g. time required for
complete mineralization, surface area requirements for successful reaction, etc.).
12
Site selection is key for successful in situ mineral carbonation. For example,
mineralization of 90 kg of carbonate per cubic meter of sandstone could take 100,000
years, though the rate depends heavily on the primary silicate minerals present in the
reservoir (Xu et al., 2005). On the other hand, mafic rocks such as basalt neutralize
carbonic acid and precipitate carbonate minerals more readily than sedimentary rocks
because of the relatively high concentration of divalent cations and low silica content
(Matter et al., 2007). Basalt, an igneous rock formed by rapidly cooled lava and
consisting of less than 20% quartz, makes up only 10% of the earth's surface but
accounts for 30-35% of CO2 uptake by natural weathering due to its reactivity (180 Mt
CO2/year) (Dessert et al., 2003). In summary, basalt is a highly reactive rock that is
very prevalent (it makes up much of the seafloor in addition to 10% of the continental
surface) and has the potential to mineralize large volumes of carbon dioxide.
In order to engineer an efficient, subsurface, in situ mineral carbonation
project, the kinetics of the reaction must be fully understood and the transport
considerations of the injection site must be well researched. Only then can knowledge
about transport and reaction kinetics be combined into reactive transport models,
which will allow for prediction of the fate of injected CO2 in potential storage
reservoirs.
1.4 Objectives
The major objective of this dissertation is to address questions surrounding
dissolution and carbonation rates of olivine under conditions relevant to subsurface
carbon storage; in particular, to quantify the reaction rates and improve our
understanding of the reaction mechanisms. This broad objective is divided into three
parts, each presented in a separate chapter.
1. Olivine dissolution and carbonation under conditions relevant for in
situ storage (Chapter 2)
Determine the dissolution and carbonation rates of olivine in batch
reactors over long times (3+ months), simulating a sub-surface storage
scenario. Establish experimental and analytical procedures that allow
13
for reproducible data.
2. Role of the Si-rich surface layer in olivine dissolution (Chapter 3)
Study the Si-rich surface layer that forms during olivine dissolution by
performing high-resolution transmission electron microscopy on cross-
sections of reacted mineral grains. Determine if surface layer passivates
the mineral surface and if the reaction rate is affected by mixing.
3. Formation mechanism(s) of Si-rich layer (Chapter 4)
Investigate the formation mechanism of the Si-rich layer using an
isotopic tracer. Calculate more precise dissolution and precipitation
rates using isotope dilution rates, which can be used to constrain a
novel kinetic model of olivine dissolution.
The work presented here has led to an improved understanding of the mechanisms that
control olivine dissolution and carbonation rates under subsurface storage conditions.
The rates measured as a part of this work are applicable to olivine only, which serves
as a model mineral for mafic and ultramafic rocks, but the mechanistic observations
may be generally applicable to multi-oxide silicate mineral dissolution.
1.5 Summary of Chapters
This dissertation is presented as a collection of chapters, each of which is a
manuscript that is either published, submitted, or in preparation for submission for
publication in a scientific journal. As such, each chapter has its own abstract,
introduction, methods, results, discussion, conclusions, and references section. I am
the first author on the three manuscripts and the work presented here was
predominantly completed by me, with contributions from coauthors appropriately
cited. This dissertation also includes four appendices, including a manuscript that
describes a model built on my experimental results (on which I am the second author)
and a detailed description of laboratory protocols developed during the process of
performing experimental work. The chapters and appendices are summarized below.
14
Chapter 2: Olivine dissolution and carbonation under conditions
relevant for in situ storage
Two, reproducible experiments performed at 60 °C, pCO2=100 bar, and
lasting 74-94 days yielded a long term, steady-state olivine dissolution
rate of 1.69±0.23 x 10-12
mol cm-2
s-1
(normalized by geometric surface
area) that is constant with slowly increasing pH (pH=4.5-5.5). The
dissolution rate is initially two orders of magnitude higher, but
decreases as the reaction solution approaches saturation with respect to
amorphous silica. Magnesite precipitation was observed in the 94 day
experiment but not the 74 day experiment; significant oversaturation of
that phase is required for precipitation at the reaction conditions.
Chapter 3: HR-TEM study of Si-rich layer formation during olivine
dissolution
A suite of four experiments designed to investigate the effect of mixing
on the dissolution rate and the effect of time on the Si-rich surface layer
was performed under the same conditions as in Chapter 2. Reacted
olivine cross sections were studied using high-resolution transmission
electron microscopy to quantify the chemistry and crystallinity of the
surface layer. Several different Si-rich layers were observed
simultaneously: a porous, amorphous, very Mg-depleted precipitated
layer up to 45 nm thick; an amorphous layer up to 25 nm thick with a
Mg/Si ratio that changes with depth; and a crystalline, Mg-depleted
region (only in the mixed experiments). Mixed experiments reacted 5
times faster than unmixed experiments, but the surface layers that
formed in both types of experiment were very similar.
Chapter 4: Spatially and temporally resolved incorporation of an
isotopic tracer (29
Si)
The same experiments discussed in Chapter 3 were reacted with a
solution containing an isotopic tracer (29
Si), which was tracked in both
the aqueous and solid phases to determine its fate. Precise calculations
15
of olivine dissolution and amorphous silica precipitation rates were
based on the dilution rate of the tracer by Si from the olivine. Uptake of
the tracer was greater in experiments that exceeded amorphous silica
saturation, leading to the conclusion that secondary phase precipitation
occurs only when the bulk fluid is saturated with respect to that phase.
Measured rates and other experimental parameters are used as inputs to
a novel kinetic model that describes multi-oxide silicate dissolution. A
complexing agent (Tiron, a catechol derivative) was added to one
experiment and improved the steady-state dissolution rate by a factor of
two.
Appendix 1: A spatially resolved kinetic model
New results on Mg isotope fractionation during olivine dissolution are
presented and data from chapters 2-4 were compiled in order to develop
a new model for silicate mineral dissolution. Constraints for the model
are provided by time-resolved Mg and Si concentration data from eight
experiments, TEM results, Mg isotope measurements, and Si isotope
measurements. The model describes the olivine surface as a series of
layers, and the dissolution rate is controlled by both the removal of Mg
from the "active" layer volume and the removal of Si from surface sites
at the "active" layer-precipitated layer interface. Fits of experimental
data show that the model reproduces experimental trajectories from a
range of experiments over both the short and long-term.
Appendix 2: Laboratory protocols
Experiments were performed at the United States Geological Survey in
Menlo Park, CA, in the Water-Rock Interaction Laboratory. The
primary experimental equipment used was the Dickson-style "Rocker
Bomb" and using and maintaining this equipment requires substantial
knowledge and effort. This appendix compiles all of the methods
developed for successfully performing experiments, from cleaning and
16
repairing gold reaction cells to troubleshooting leaks in the pressure
vessel.
Appendix 3: Constraints on redox state of reaction
Supercritical CO2 samples were analyzed for hydrogen concentration,
which can then be used to calculate the redox sate of the reaction and
the theoretical relative concentrations of Fe2+
and Fe3+
. This work
allows for the comparison with literature that suggests Fe3+
may
stabilize the Si-rich layer. This appendix contains preliminary data on
the reduction potential of experiments as a function of time.
Appendix 4: Derivation of Diffusion Coefficient from Lasaga (1979)
Lasaga (1979) describes diffusion coefficients for interdiffusion of
protons and monovalent cations (e.g. sodium) in glasses, and provides a
governing equation that can be applied to di- and trivalent cations. In
this section, I derive the expression for Mg2+
/ H+ interdiffusion.
17
1.6 References
Alnes, H., Eiken, O., Nooner, S., Sasagawa, G., Stenvold, T., Zumberge, M., 2011.
Results from Sleipner gravity monitoring: Updated density and temperature
distribution of the CO2 plume. Energy Procedia 4, 5504–5511.
Bearat, H., Mckelvy, M.J., Chizmeshya, A.V.G., Gormley, D., Nunez, R., Carpenter,
R.W., Squires, K., Wolf, G.H., 2006. Carbon Sequestration via Aqueous
Olivine Mineral Carbonation : Role of Passivating Layer Formation. Environ.
Sci. Technol. 40, 4802–4808.
Blum, A., Lasaga, A., 1988. Role of surface speciaiton in the low temperature
dissolution of minerals. Nature 331, 431–433.
Bodénan, F., Bourgeois, F., Petiot, C., Augé, T., Bonfils, B., Julcour-Lebigue, C.,
Guyot, F., Boukary, a., Tremosa, J., Lassin, a., Gaucher, E.C., Chiquet, P.,
2014. Ex situ mineral carbonation for CO2 mitigation: Evaluation of mining
waste resources, aqueous carbonation processability and life cycle assessment
(Carmex project). Miner. Eng. 59, 52–63.
Cappa, F., Rutqvist, J., 2011. Impact of CO 2 geological sequestration on the
nucleation of earthquakes. Geophys. Res. Lett. 38, n/a–n/a.
Chen, Y., Brantley, S.L., 2000. Dissolution of forsteritic olivine at 65°C and 2<pH<5.
Chem. Geol. 165, 267–281.
Chen, Z.-Y., O’Connor, W.K., Gerdemann, S.J., 2006. Chemistry of Aqueous Mineral
Carbonation for Carbon Sequestration and Explanation of Experimental
Results. Environ. Prog. 25, 161–166.
Daval, D., Sissmann, O., Menguy, N., Saldi, G.D., Guyot, F., Martinez, I., Corvisier,
J., Garcia, B., Machouk, I., Knauss, K.G., Hellmann, R., 2011. Influence of
amorphous silica layer formation on the dissolution rate of olivine at 90°C and
elevated pCO2. Chem. Geol. 284, 193–209.
Dessert, C., Dupré, B., Gaillardet, J., François, L.M., Allègre, C.J., 2003. Basalt
weathering laws and the impact of basalt weathering on the global carbon
cycle. Chem. Geol. 202, 257–273.
Duan, Z., Sun, R., 2003. An improved model calculating CO2 solubility in pure water
and aqueous NaCl solutions from 273 to 533 K and from 0 to 2000 bar. Chem.
Geol. 193, 257–271.
18
Edenhofer, O., Pichs-Madruga, R., Sokona, Y., Kadner, S., Minx, J., Brunner, S.,
2014. Technical Summary of Working Group III: Mitigation of Climate
Change.
Felmy, A.R., Qafoku, O., Arey, B.W., Hu, J.Z., Hu, M., Todd Schaef, H., Ilton, E.S.,
Hess, N.J., Pearce, C.I., Feng, J., Rosso, K.M., 2012. Reaction of water-
saturated supercritical CO2 with forsterite: Evidence for magnesite formation
at low temperatures. Geochim. Cosmochim. Acta 91, 271–282.
García-Gusano, D., Garraín, D., Herrera, I., Cabal, H., Lechón, Y., 2013. Life Cycle
Assessment of applying CO2 post-combustion capture to the Spanish cement
production. J. Clean. Prod. 2025.
Giammar, D.E., Bruant, R.G., Peters, C.A., 2005. Forsterite dissolution and magnesite
precipitation at conditions relevant for deep saline aquifer storage and
sequestration of carbon dioxide. Chem. Geol. 217, 257–276.
Giannoulakis, S., Volkart, K., Bauer, C., 2014. Life cycle and cost assessment of
mineral carbonation for carbon capture and storage in European power
generation. Int. J. Greenh. Gas Control 21, 140–157.
Gislason, S.R., Wolff-Boenisch, D., Stefansson, A., Oelkers, E.H., Gunnlaugsson, E.,
Sigurdardottir, H., Sigfusson, B., Broecker, W.S., Matter, J.M., Stute, M.,
2010. Mineral sequestration of carbon dioxide in basalt: A pre-injection
overview of the CarbFix project. Int. J. Greenh. Gas Control 4, 537–545.
Golubev, S. V, Pokrovsky, O.S., Schott, J., 2005. Experimental determination of the
effect of dissolved CO2 on the dissolution kinetics of Mg and Ca silicates at
25°C. Chem. Geol. 217, 227–238.
Hänchen, M., Prigiobbe, V., Storti, G., Seward, T.M., Mazzotti, M., 2006. Dissolution
kinetics of fosteritic olivine at 90–150°C including effects of the presence of
CO2. Geochim. Cosmochim. Acta 70, 4403–4416.
Hellmann, R., Daval, D., Wirth, R., 2013. Formation of Amorphous Silica Surface
Layers by Dissolution-Reprecipitaton During Chemical Weathering:
Implications for CO2 Uptake. Procedia Earth Planet. Sci. 7, 346–349.
Hellmann, R., Eggleston, C., Hochellajr, M., Crerar, D., 1990. The formation of
leached layers on albite surfaces during dissolution under hydrothermal
conditions. Geochim. Cosmochim. Acta 54, 1267–1281.
19
Hellmann, R., Wirth, R., Daval, D., Barnes, J.-P., Penisson, J.-M., Tisserand, D.,
Epicier, T., Florin, B., Hervig, R.L., 2012. Unifying natural and laboratory
chemical weathering with interfacial dissolution–reprecipitation: A study based
on the nanometer-scale chemistry of fluid–silicate interfaces. Chem. Geol.
294-295, 203–216.
Hopkinson, L., Rutt, K., Cressey, G., 2008. The Transformation of Nesquehonite to
Hydromagnesite in the System CaO-MgO-H2O-CO2 : An Experimental
Spectroscopic Study. J. Geol. 116, 387–400.
Hosa, A., Esential, M., Stewart, J., Haszeldine, S., 2010. Benchmarking worldwide
CO2 saline aquifer injections.
Huber, M., Knutti, R., 2012. Anthropogenic and natural warming inferred from
changes in Earth ’ s energy balance. Nat. Geosci. 5, 31–36.
Joughin, I., Smith, B.E., Medley, B., 2014. Marine Ice Sheet Collapse Potentially
Underway for the Thwaites Glacier Basin, West Antarctica. Science (80-. ). 1–
7.
King, H.E., Plümper, O., Geisler, T., Putnis, A., 2011. Experimental investigations
into the silicification of olivine: Implications for the reaction mechanism and
acid neutralization. Am. Mineral. 96, 1503–1511.
King, H.E., Plümper, O., Putnis, A., 2010. Effect of secondary phase formation on the
carbonation of olivine. Environ. Sci. Technol. 44, 6503–6509.
Klass, A.B., Wilson, E.J., 2009. Climate Change, Carbon Sequestration, and Property
Rights.
Lafay, R., Montes-Hernandez, G., Janots, E., Chiriac, R., Findling, N., Toche, F.,
2014. Simultaneous precipitation of magnesite and lizardite from hydrothermal
alteration of olivine under high-carbonate alkalinity. Chem. Geol. 368, 63–75.
Latil, M., 1980. Enhanced Oil Recovery. Gulf, New York.
Li, D., Duan, Z., 2007. The speciation equilibrium coupling with phase equilibrium in
the H2O–CO2–NaCl system from 0 to 250 °C, from 0 to 1000 bar, and from 0
to 5 molality of NaCl. Chem. Geol. 244, 730–751.
Little, M., Jackson, R.B., 2009. Potential Risks of Freshwater Aquifer Contaimination
from Carbon Capture and Storage, in: American Geophysical Union. p. H13A–
0914.
20
Madsen, R., Xu, L., Claassen, B., McDermitt, D., 2009. Surface Monitoring Method
for Carbon Capture and Storage Projects. Energy Procedia 1, 2161–2168.
Matter, J.M., Takahashi, T., Goldberg, D., 2007. Experimental evaluation of in situ
CO2 water-rock reactions during CO2 injection in basaltic rocks: Implications
for geological CO2 sequestration. Geochemistry Geophys. Geosystems 8,
1525–2027.
Melillo, J.M., Richmond, T., Yohe, G.W., 2014. Highlights of Climate Change
Impacts in the United States: The Third National Climate Assessment. U.S.
Global Change Research Program.
O’Connor, W.K., Dahlin, D., Rush, G., Gerdemann, S.J., Penner, L., 2004. Energy and
economic considerations for ex-situ aqueous mineral carbonation. Albany, OR.
Oelkers, E.H., Gislason, S.R., Matter, J., 2008. Mineral carbonation of CO2. Elements
4, 333–337.
Olsen, A.A., Rimstidt, J.D., 2008. Oxalate-promoted forsterite dissolution at low pH.
Geochim. Cosmochim. Acta 72, 1758–1766.
Pfister, G., Walters, S., Lamarque, J.-F., Fast, J., Barth, M.C., Wong, J., Done, J.,
Holland, G., Bruyere, C.L., 2014. Projections of future summertime ozone
over the U.S. J. Geophys. Res. Atmos. 119, 1–24.
Pokrovsky, O.S., Schott, J., 2000a. Kinetics and mechanism of forsterite dissolution at
25°C and pH from 1 to 12. Geochim. Cosmochim. Acta 64, 3313–3325.
Pokrovsky, O.S., Schott, J., 2000b. Forsterite surface composition in aqueous
solutions : A combined potentiometric , electrokinetic , and spectroscopic
approach. Geochim. Cosmochim. Acta 64, 3299 –3312.
Prigiobbe, V., Hänchen, M., Costa, G., Baciocchi, R., Mazzotti, M., 2009. Analysis of
the effect of temperature, pH, CO2 pressure and salinity on the olivine
dissolution kinetics. Energy Procedia 1, 4881–4884.
Rimstidt, J.D., Brantley, S.L., Olsen, A.A., 2012. Systematic review of forsterite
dissolution rate data. Geochim. Cosmochim. Acta 99, 159–178.
Romanak, K.D., Smyth, R.C., Yang, C., Hovorka, S.D., Rearick, M., Lu, J., 2012.
Sensitivity of groundwater systems to CO2: Application of a site-specific
analysis of carbonate monitoring parameters at the SACROC CO2-enhanced
oil field. Int. J. Greenh. Gas Control 6, 142–152.
21
Ruiz-Agudo, E., Putnis, C. V., Rodriguez-Navarro, C., Putnis, a., 2012. Mechanism of
leached layer formation during chemical weathering of silicate minerals.
Geology.
Saldi, G.D., Daval, D., Morvan, G., Knauss, K.G., 2013. The role of Fe and redox
conditions in olivine carbonation rates: an experimental study of the rate
limiting reactions at 90 and 150 °C in open and closed systems. Geochim.
Cosmochim. Acta 118, 157–183.
Saldi, G.D., Schott, J., Pokrovsky, O.S., Gautier, Q., Oelkers, E.H., 2012. An
experimental study of magnesite precipitation rates at neutral to alkaline
conditions and 100–200°C as a function of pH, aqueous solution composition
and chemical affinity. Geochim. Cosmochim. Acta 83, 93–109.
Shirokova, L.S., Bénézeth, P., Pokrovsky, O.S., Gerard, E., Ménez, B., Alfredsson, H.,
2012. Effect of the heterotrophic bacterium Pseudomonas reactans on olivine
dissolution kinetics and implications for CO2 storage in basalts. Geochim.
Cosmochim. Acta 80, 30–50.
Sissmann, O., Daval, D., Brunet, F., Guyot, F., Verlaguet, A., Pinquier, Y., Findling,
N., Martinez, I., 2013. The deleterious effect of secondary phases on olivine
carbonation yield: Insight from time-resolved aqueous-fluid sampling and FIB-
TEM characterization. Chem. Geol. 357, 186–202.
Steefel, C.I., Van Cappellen, P., 1990. A new kinetic approach to modeling water-rock
interaction: The role of nucleation, precursors, and Ostwald ripening.
Geochim. Cosmochim. Acta 54, 2657–2677.
Wang, F., Giammar, D.E., 2013. Forsterite Dissolution in Saline Water at Elevated
Temperature and High CO2 Pressure. Environ. Sci. Technol. 47, 168–173.
Welch, S.A., Banfield, J.F., 2002. Modification of olivine surface morphology and
reactivity by microbial activity during chemical weathering. Geochim.
Cosmochim. Acta 66, 213–221.
Wilson, E.J., Friedmann, S.J., Pollak, M.F., 2007. Research for deployment:
incorporating risk, regulation, and liability for carbon capture and
sequestration. Environ. Sci. Technol. 41, 5945–52.
Wogelius, R.A., Walther, J. V, 1991. Olivine dissolution at 25°C : Effects of pH , CO2
, and organic acids. Geochim. Cosmochim. Acta 55, 943–954.
Wogelius, R.A., Walther, J. V, 1992. Olivine dissolution kinetics at near-surface
conditions. Chem. Geol. 97, 101–112.
22
Xu, T., Apps, J. a., Pruess, K., 2005. Mineral sequestration of carbon dioxide in a
sandstone–shale system. Chem. Geol. 217, 295–318.
Zakaznova-Herzog, V.P., Nesbitt, H.W., Bancroft, G.M., Tse, J.S., 2008.
Characterization of leached layers on olivine and pyroxenes using high-
resolution XPS and density functional calculations. Geochim. Cosmochim.
Acta 72, 69–86.
23
Chapter 2: Olivine dissolution and carbonation under conditions relevant for in situ carbon storage
Natalie C. Johnsona,*
, Burt Thomasb, Kate Maher
c, Robert J. Rosenbauer
b,
Dennis Birdc, and Gordon E. Brown, Jr.
a,c,d
a. Department of Chemical Engineering, Stauffer III, 381 North-South Mall,
Stanford University, Stanford CA 94305, USA
b. U.S. Geological Survey, 345 Middlefield Rd, Menlo Park CA 94025, USA
c. Department of Geological and Environmental Sciences, 450 Serra Mall, Braun Hall
Building 320, Stanford University, Stanford CA 94305, USA
d. Department of Photon Science and Stanford Synchrotron Radiation Lightsource,
SLAC National Accelerator Laboratory, Menlo Park CA 94025, USA
Reprinted from Chemical Geology, 373, Johnson, N. C., Thomas, B., Maher, K.,
Rosenbauer, R. J., Bird, D., & Brown, G. E., Olivine dissolution and carbonation
under conditions relevant for in situ carbon storage, 93-105, Copyright 2014, with
permission from Elsevier.
24
Abstract
In order to evaluate the chemistry and kinetics of mineral carbonation reactions
under conditions relevant to subsurface injection and storage of CO2, olivine alteration
was studied at 60 °C and 100 bar CO2 pressure, including olivine dissolution and the
formation of carbonate minerals. Batch experiments were performed with olivine
(Fo92), water, CO2, and NaCl inside gold cells contained within rocking autoclaves.
Two reproducible experiments yielded an initial (1 hour) dissolution rate of 9.50±0.10
x 10-11
and a long-term (10-70 days) rate of 1.69±0.23 x 10-12
mol cm-2
s-1
. The long-
term rate is consistent with previously published rate laws at 4.5 < pH < 5.5. The
dissolution rates presented here are constant with increasing pH in the same range,
suggesting a pH-independent dissolution mechanism at elevated CO2(aq) and H4SiO4
at 60°C. A Si-rich surface layer forms on olivine grains within 2 days of reaction and
appears to slow dissolution by passivating the surface. The olivine dissolution rate
decreases by 2 orders of magnitude over 4 days as the system approaches amorphous
silica saturation but remains constant thereafter based on a linear increase in Mg
concentrations. Secondary phases consist of amorphous silica and magnesite, with up
to 7 mol% olivine converted to Mg-carbonate over 94 days of reaction. Magnesite
precipitation rates could not be precisely quantified due to experimental limitations.
However, our minimum estimate of 1.40 x 10-13
mol cm-2
s-1
suggests that the
precipitation rate is several orders of magnitude greater than predicted by previous
studies. Finally, the presence of 0.5 M NaCl resulted in a decrease in olivine
dissolution rate at reaction times of less than 4 days, but a significant enhancement of
the reaction rate at reaction times greater than 4 days relative to electrolyte-free
experiments. Our results suggest that geochemical models developed to predict the
behavior of subsurface CO2 storage systems in mafic and ultramafic rocks should
incorporate the effects of dissolved species, including SiO2 and NaCl.
25
2.1 Introduction
Increasing levels of atmospheric carbon dioxide (CO2) motivate the study of
carbon capture and storage (CCS) as an important component in a multinational
strategy to mitigate the risks of future climate change. Although many possible
solutions exist for carbon storage in the subsurface, the most thermodynamically
stable solution over geologic timescales is mineral carbonation, such as the chemical
reaction between CO2 and reactive silicate minerals (e.g., forsterite, Mg2SiO4) to form
carbonate minerals:
Mg2SiO4 + 2CO2 → 2MgCO3 + SiO2 (1)
Mineral carbonation occurs via a series of steps, and the overall reaction is
thermodynamically favored (ΔG = -47 kJ/mol with CO2 gas and amorphous silica as
the SiO2 phase at 60oC and pCO2 =100 bar) and occurs in nature. Mineral carbonation
steps are illustrated by reactions 2-7, although other steps may occur, such as
formation of metastable Mg-carbonate solids (e.g., hydromagnesite, nesquehonite)
(Hopkinson et al., 2008):
Mg2SiO4 + 4H+ → 2Mg
2+ + H4SiO4 (2)
H4SiO4 = SiO2(am) + 2H2O (3)
2CO2(g) = 2CO2(aq) (4)
2CO2(aq) + 2H2O = 2HCO3- + 2H
+ (5)
2HCO3- = 2CO32-
+ 2H+ (6)
2CO32-
+ 2Mg2+
→ 2MgCO3 (7)
________________________________
Mg2SiO4 + 2CO2(g)→ 2MgCO3 + SiO2(am) (8)
First, the silicate mineral dissolves and releases divalent cations, such as
magnesium, iron, and calcium, in addition to silica in the form of silicic acid (H4SiO4)
(reaction 2). The silicic acid then precipitates as (possibly hydrated) amorphous silica
(SiO2(am)) in low temperature systems (reaction 3), where quartz precipitation is
kinetically limited and amorphous silica solubility is low relative to olivine (and other
26
silicate minerals) solubility (Steefel and Van Cappellen, 1990). Carbon dioxide also
dissolves concurrently (reactions 4-5), producing carbonic acid, bicarbonate ions, and
carbonate ions whose relative quantities are pH dependent. The dissolved Mg ions
then react with bicarbonate and carbonate ions to produce carbonate minerals, which
precipitate at high saturation indices (reaction 7) (Saldi et al., 2012). Some examples
of natural mineral carbonation involving Mg-silicates, which are both abundant and
highly reactive (Matter and Kelemen, 2009), can be found in Oman (Kelemen and
Matter, 2008), Italy (Boschi et al., 2009), and California (Goff and Lackner, 1998).
Such natural analogues motivate a laboratory-based study of silicate carbonation from
the initial stages of olivine dissolution, and the attendant silica precipitation to
precipitation of a carbonate phase (reaction 7). To our knowledge, long-term
experiments covering the full sequence of reactions associated with Mg-silicate
carbonation have not been conducted previously.
Olivine ((Mg,Fe)2SiO4) is a common mineral in mafic and ultramafic rocks in
the subsurface, but due to its high solubility it weathers quickly under earth-surface
conditions (Goldich, 1938). Numerous studies of olivine weathering at surface and
near-surface conditions have shown that olivine dissolution rates increase with
increasing temperature and decreasing pH (Blum and Lasaga, 1988; Chen and
Brantley, 2000; Giammar et al., 2005; Hänchen et al., 2006; Pokrovsky and Schott,
2000b; Prigiobbe et al., 2009; Oelkers, 2001a; Olsen and Rimstidt, 2008; Shirokova et
al., 2012; Wang and Giammar, 2013; Welch and Banfield, 2002; Wogelius and
Walther, 1992, 1991). Recent studies of olivine dissolution in the presence of
dissolved CO2 have attributed decreases in the dissolution rate to changes in pH
(Golubev et al., 2005), although Chen et al. (2006) found olivine dissolution rate
enhancement in the presence of bicarbonate ions. Other work on silicate mineral
carbonation suggests that the carbonation reaction may occur in a hydrated
supercritical CO2 phase in addition to the aqueous phase (Felmy et al., 2012; Kwak et
al., 2011; Kwak et al., 2010; Loring et al., 2011). The formation of a Si-rich layer on
the olivine surface during dissolution has been observed using x-ray photoelectron
spectroscopy (XPS) (Zakaznova-Herzog et al., 2008), transmission electron
27
spectroscopy (TEM) (Bearat et al., 2006; Daval et al., 2011), and nuclear magnetic
resonance (NMR) spectroscopy (Kwak et al., 2011). The Si-rich layer may be formed
via a leaching process (Bearat et al., 2006; Daval et al., 2011; Pokrovsky and Schott,
2000a, 2000b; Oelkers, 2001b), a dissolution/reprecipitation process (Hellmann et al.,
2012; King et al., 2011), or a combination of both processes (Saldi et al., 2013). The
Si-rich layer may inhibit olivine dissolution. For example, both Daval et al. (2011) and
Pokrovsky and Schott (2000b) found that the rate of olivine dissolution decreases with
increasing SiO2(am) saturation. While many studies have examined the dissolution
behavior of olivine in aqueous solutions (e.g., Rimstidt et al. 2012 and references
therein), relatively few examined olivine dissolution in the presence of two fluid
phases (aqueous and supercritical CO2) and with high concentrations of dissolved
species, specifically H4SiO4.
Olivine dissolution is only the first step of the carbonation process. Effective
CO2 sequestration requires carbonate precipitation as well, and because the extent and
rate of carbonate precipitation will depend on the rate of olivine dissolution, these
processes are coupled. Reactions 2 through 7 show the interdependency of these steps
during olivine carbonation. In a closed system, CO2 and olivine must dissolve (and
alkalinity must increase) before Mg-carbonate can precipitate. Open systems allow for
chemical steps to occur at separate times and in different physical locations—olivine
may dissolve upstream creating a Mg-rich fluid, which later interacts with CO2 to
precipitate carbonate. Carbonation associated with an in situ storage scenario could
represent some combination of open and closed systems (Boschi et al., 2009; Kelemen
et al., 2011), thus it is important to study olivine carbonation both as a series of
coupled reactions and as distinct reaction steps. Ex-situ carbonation would likely be
engineered as an open system with each step carefully optimized for maximum
reaction and minimum cost. Several coupled olivine dissolution/magnesite
precipitation studies have been conducted, though all previous successful attempts at
precipitating magnesite from the irreversible dissolution of olivine in CO2-rich
solutions required additional chemicals, such as acids or sodium bicarbonate, and
relatively high temperatures (i.e., >80 ˚C) to enhance magnesite precipitation rates
28
(Bearat et al., 2006; Chen et al., 2006; Giammar et al., 2005; Hänchen et al., 2007;
King et al., 2010; Kwak et al., 2010). Such additives would increase the cost of
sequestration. In order to understand olivine carbonation under conditions relevant to
natural systems, longer-term studies at lower temperatures and without chemical
additives are required.
A better understanding of the dissolution pathways and kinetics is also
necessary to assess the efficacy of mineral carbonation for sequestering CO2.
Recently, Rimstidt et al. (2012) performed a systematic review of 25 independent
olivine dissolution studies and derived two equations that describe forsterite
dissolution rate as a function of pH and temperature. A 10% variance in dissolution
rates, not associated with pH or temperature, was attributed to variability in the
calculation of surface area as total surface area is less important than the number of
active sites on the mineral surface (Rimstidt et al. 2012). A key conclusion was that
insufficient data sets containing dissolved carbonate species exist to quantify the effect
of CO2 on forsterite dissolution rates (Rimstidt et al. 2012).
To address the need for additional experimental work on the influence of CO2
on forsterite dissolution (e.g., Rimstidt et al., 2012) and to provide additional
experimental insight into the consequences of the Si-rich layer, we have examined
long-term olivine carbonation kinetics under reaction conditions relevant to in-situ
storage by observing reaction progress over 94 days at 60 ˚C and 100 bars pCO2 in the
presence of water. In contrast to many dissolution studies that have used mixed-flow
reactors, we use a batch reactor to evaluate the range of reaction conditions associated
with olivine carbonation, from dissolution to saturation with respect to amorphous
silica and magnesite. Use of a batch reactor distinguishes this study from those using
mixed-flow experiments as it allows us to evaluate the reactions across a broad range
of conditions reflective of the geochemical evolution of natural systems.
By analyzing the entire carbonation reaction, from initial olivine dissolution to
secondary phase precipitation, we can evaluate the limiting step(s) in the carbonation
process. We present time-resolved elemental concentrations and detailed analysis of
reaction products, allowing us to carefully monitor the entire carbonation reaction,
29
from the dissolution of CO2 and olivine to carbonate precipitation. These data provide
new constraints on olivine dissolution rates at high pCO2 and suggest a weaker
dependence of dissolution rates on pH compared to previous studies. The data thus
contribute to the growing database of olivine dissolution rates in the presence of CO2.
We also observe that dissolved electrolytes have a complex and time-dependent effect
on the carbonation reaction. Importantly, our work indicates that dissolved SiO2 and
NaCl are important controls on the dissolution/carbonation rates of olivine, and thus
should be included in geochemical models of carbon storage in olivine-rich rocks.
We begin our analysis of olivine carbonation by confirming the production of
amorphous silica and magnesite as secondary phases and then by examining the
aqueous phase composition as a function of time in order to extract reaction rates and
thermodynamic saturation values. We finish with (1) a discussion of the effects of
amorphous silica saturation and NaCl concentration on the rate of olivine dissolution,
(2) a comparison of rates from this study with previously published rate laws, and (3)
the rate of magnesite precipitation.
2.2 Materials and Methods
2.2.1 Autoclave Reactor Experimental System
To minimize potential catalytic effects of the reaction cell surface and to allow
for the changing reactant volume throughout the experiment, we employed a flexible
and inert Au-bag reaction vessel design following the detailed description in
Rosenbauer et al. (1983) with modifications for sampling and analysis of CO2
(Rosenbauer et al., 2005). Briefly, most experiments were carried out in flexible Au-Ti
reaction cells with ~200 mL initial volume, equipped with a small bore (2.3 mm) ¼
inch OD, Ti capillary exit tube and sampling valve. Titanium parts were passivated
using a nitric acid bath followed by bake-out in air for 24 hrs at 450 ˚C. This
experimental set-up allows for serial sampling of cell contents while maintaining
constant cell pressure. The cell was contained within a steel autoclave in a cyclic
rotating furnace (Rosenbauer et al., 1983). Liquid CO2 was added to the reaction
vessel with high-pressure syringe pumps (ISCO Corp./Teledyne Inc., Model D100)
30
charged from a high purity CO2 gas cylinder (Matheson Gas Inc.) equipped with a
full-length eductor tube for liquid withdrawal. Prior to final assembly, reagent grade
chemicals were used to generate 0.5 M NaCl solutions that were combined with the
olivine. After the reactants were sealed inside the cell, the autoclave was pressurized
to ~100 bar by the annular water pressure, allowing the headspace gas to exit from the
cell and preventing oxygen dissolution during pressurization. The assembly was then
depressurized and heated to experimental temperature, after which it was pressurized
to 100 bar and a sample was taken in order to provide a baseline point. Then, CO2 was
injected into the reaction cell through the exit tube with a syringe pump at 100 bars
and 20 °C, the reaction time was started upon completion of CO2 injection.
Temperature was maintained by a proportional controller (Omega, Love™ Controls,
Model 49) and measured with a calibrated type K thermocouple. Pressure was
measured with dead-weight calibrated analog gauges.
Serial samples were obtained throughout the experiments. At the conclusion of
an experiment, CO2 was vented through the exit tube in concert with disassembly of
the stainless steel autoclave. The large heat capacity of the steel autoclave + pressure
fluid kept the reaction cell above 50 °C for the entire disassembly process. After the
autoclave and reaction cell were disassembled, the reacted solids were rinsed and
filtered under a vacuum through 2.5 m paper filters and dried overnight in air at
60 °C. The time from beginning depressurization of the reaction cell to filtration of the
products was 30 minutes or less. The pH of the filtrate ranged from 6 to 7.
2.2.2 Analytical and Sampling Procedures
Serial samples were obtained throughout each reaction via polypropylene
syringe and a three-way stopcock that allowed for an initial 0.5 mL withdrawal to
purge the exit tube. Samples were obtained for total CO2 analysis, cation analysis, and
alkalinity titration using the same 3-way stopcock. Total CO2 samples were analyzed
by coulometry with liquid injection into a phosphoric acid reaction vessel (UIC
Analytics). Total CO2 samples consisted of approximately 0.5 mL liquid sample and
10mL headspace. Sample masses were determined by high precision gravimetry of the
sampling syringe before and after sampling and injection. Alkalinity and cation
31
samples were obtained via polypropylene syringe, and filtered with a 0.45µm nylon
syringe filter into pre-weighed vials. Exact sample amounts were determined by
gravimetry. Alkalinity samples (~0.5 mL) were then diluted 10x in 15 mL serum vials
and analyzed via sulfuric acid titration to a pH endpoint of 4.2 with a calibrated pH
probe and autotitrator cartridge system (Hach Inc., USA). Cation samples (~0.5 mL)
were acidified with TM grade nitric acid to pH <2 and diluted for analysis via ICP-
AES on a Thermo Scientific ICAP6300 Duo View Spectrometer.
Reaction cell integrity was monitored by frequent refractometer salinity
analysis to ensure no mixing between the annular freshwater and the saline reaction
fluid. The reactions were run for a period of 4 to 94 days, with 7 to 10 fluid samples of
about 2 mL taken over the course of each experiment. The total volume of samples
removed was kept below 10% of the initial volume in order to maintain the water to
rock ratio, initially set at 50:1 by mass. The scCO2:water ratio was set at 1:10 by
volume at reaction temperature and pressure. A sufficient volume of CO2 was added to
Figure 2.1: X-ray diffraction patterns for unreacted and reacted olivine. Magnesite peaks are
visible in the reacted spectra .
32
the reaction vessel in order to saturate the aqueous phase (saturation = 0.93 mol/kg
with 0.5 M NaCl) and maintain a headspace equal to 1/10 the volume of the aqueous
phase at reaction conditions. The headspace allowed the reaction to maintain a
constant pressure of 100 bars.
2.2.3 Mineral and Mineral Surface Analyses
Initial and product mineral phases were analyzed using a series of techniques
before and after reaction to determine mineralogy, total carbon content, surface
chemical composition, and bulk chemical composition. Powder x-ray diffraction
(XRD) was performed on a Rigaku X-ray Diffractometer, model number CM2029
using monochromated Cu Kα radiation for 2-theta values between 10° and 70° to
determine the mineralogy of each sample. Total carbon content was determined by
coulometry using a phosphoric acid digestion and the UIC Analytics coulometer. X-
ray photoelectron spectroscopy (XPS) was performed on a PHI VersaProbe Scanning
XPS Microprobe and allowed for the study of the composition of mineral surfaces
(~10 nm deep). Bulk chemical composition was obtained by several methods:
digestion followed by ICP-AES, energy dispersive spectroscopy (on a FEI XL30
Figure 2.2: X-ray photoelectron spectroscopy data for unreacted olivine, and olivine reacted for 2,
4, and 74 days. The peak intensity of Mg decreases with increasing reaction time relative to the Si
peak, indicating that the mineral surface becomes more Si-rich as the reaction progresses.
33
Sirion SEM with FEG source with EDAX detector), and by electron microprobe
analysis (JEOLJXA-733A Superprobe). SEM images were taken with a Sirion SEM.
The composition of the olivine used in these experiments, as measured by
hydrofluoric acid digestion and ICP, is Mg1.84Fe0.16SiO4 (Fo92) and is from the Twin
Sisters Dunite, Washington. Rock powder used in our experiments contains <1%
magnetite and chromite as determined by electron microprobe analysis (neither phase
was visible in XRD patterns) (Onyeagocha, 1978). Unreacted olivine was
predominantly forsterite, with about 3% lizardite by mass. No other secondary phases
were identified by XRD (Fig. 2.1). Mineral samples were prepared by alternating
between methanol washes and sieving (3 times total) to obtain a grain size range of
38-75 μm. The grain size fraction 75-105 μm was also retained for use in other
experiments. XPS analysis of unreacted olivine indicated that washing with water
removed Mg from the olivine surface, and visual inspection of SEM images indicated
that water was also less effective at removing the fine and ultrafine particles.
Quantitative XPS analysis was used to determine that the unreacted olivine has a
stoichiometric surface composition (1.85±0.10:1 Mg:Si), and elements other than Mg,
Fe, Si, and O make up <1 mol% of the material (see Supplementary Material Fig. 2.2).
Geometric surface area was calculated from the grain diameters (38-75µm range)
using Equations 3 and 4 from the supplemental information of Rimstidt et al. (2012) to
be 342 cm2/g. The BET surface area was measured using an ASAP 2020
(Micromeritics Instrument Corporation) with a 7-point surface analysis and nitrogen
as the adsorbant. The BET area is 6410 cm2/g, which gives a roughness factor (BET
area divided by geometric area) of 18.74.
2.2.4 Experiments
Results of four experiments using the same starting material (Fo92 olivine),
temperature (60oC), CO2 pressure (100 bar), and water:rock ratio by mass (50:1) are
summarized in Table 2.1. Two experiments contained 0.5 M NaCl, whereas two had
no added electrolyte. Experimental duration ranged from 4 to 94 days. The shortest
experiments were conducted with higher sampling frequency to look at the initial
dissolution kinetics as silica saturation was approached.
34
2.2.5 Calculations and Modeling
Aqueous speciation and reaction affinity were calculated using PHREEQC
(Parkhurst and Appelo, 1999) and the included llnl.dat database, which is derived from
the EQ3/EQ6 database (Wolery et al., 1990) for each data point using measured
solution compositions. The pH was calculated from alkalinity and total inorganic
carbon measurements, using previously published CO2 solubility data from Li and
Duan (2007). Mineral saturation was calculated in the form of the log of the reaction
quotient, Q, divided by the equilibrium constant K. A mineral is at saturation if Q=K,
or log(Q/K)=0. Uncertainties in pH and saturation states were calculated using
uncertainties in experimental measurements. The largest source of uncertainty came
from the alkalinity measurements, which had a mass uncertainty of ±0.02 g and a
titration uncertainty of ±0.003 mL. We did not control or measure redox state, so Fe
speciation was calculated using total Fe only. Olivine dissolution rates were calculated
for each data point using the measured concentration and time differences between the
two points of interest. In some instances, the pH of the two data points used to
calculate instantaneous rates were different, so these rates are plotted showing the
range of pH for each calculated rate. Long-term rates (10-70 days) were calculated
using a linear regression of Mg concentration vs. time and the errors were calculated
based on the R2 value (square of the correlation coefficient), which was larger than the
5% concentration error on each data point. After the experiments were depressurized,
Mg concentrations and pH were re-measured in the solution and the total inorganic
carbon content of the solids was measured using a coulometer.
Table 2.1: Experimental data for four experiments.
Temperature
(oC)
Pressure
(bar)
Initial ionic
strength (M)
Aqueous
solution:CO2
ratio (mass)
Aqueous
solution:rock
ratio (mass)
Duration
(days)
1 60 100 0.5 34:1 50:1 74
2 60 100 0.5 34:1 50:1 94
3 60 100 0.0 34:1 50:1 4
4 60 100 0.0 34:1 50:1 98
35
2.3 Results
2.3.1 Solution Data
Elemental concentrations (Mg, Si, and Fe) and alkalinity as a function of time
are presented in Fig. 2.3 and Table 2.2 for the four experiments. Silicon (in the form
of H4SiO4) concentrations increased rapidly in all experiments over the initial 10 days.
Experiment 1 (0.5 M NaCl) displays significant variance of 30% in the measurements
from 10-74 days, Experiment 2 (0.5 M NaCl) shows a maximum H4SiO4 of 6 mM at
20 days followed by a decrease to steady-state at 45 days and 4 mM. Experiment 4 (0
Figure 2.3: Elemental concentrations versus time for magnesium (A), silicon (B), iron (D), and
alkalinity versus time (C) for four experiments. Insets show the same y-axis for the initial 4 days
of reaction. Experiments 1 and 2 contain 0.5 M NaCl; Experiments 3 and 4 have 0 M NaCl.
36
M NaCl) also shows a maximum of H4SiO4 (8 mM at 58 days). In all experiments, Mg
concentrations increased for the first 2 days, then remained constant for 2-5 days. Mg
concentrations again increased at a roughly constant rate for the remainder of
Experiments 1, 2, and 4. The Fe concentrations peaked at 20, 28, and 77 days for
Experiments 1, 2, and 4, respectively, followed by a decrease. The duration of
Experiment 3 (0 M NaCl) was short enough to expect no change in Fe concentrations
based on the trends from other experiments. Finally, the alkalinity data mirror the Mg
concentrations.
2.3.2 Mineral Products Analysis
X-ray diffraction confirmed the presence of magnesite in Experiments 1 and 2.
Phosphoric acid digestion and coulometry showed that 12±2 mol % of the olivine
converted to carbonate in Experiment 1, and 17±2 mol % in Experiment 2. These
conversions are artificially high because some carbonate precipitation occurred during
depressurization and the corresponding rise in pH. The amount of carbonate that
precipitated during depressurization was determined by measuring the Mg
concentration from samples before and after depressurization. The reduction in Mg
during depressurization was assumed to have been due to Mg-carbonate precipitation,
and this amount was subtracted from the coulometric results to yield carbonation
extent. We define carbonation extent as the mole percent of olivine converted to
magnesite during the experimental duration, before depressurization. The extent of
carbonation that occurred during the reaction was 0 % for Experiment 1 (74 days) (i.e.,
all carbonate precipitation occurred during depressurization) and 7±2 % for
Experiment 2 (94 days). No new phases other than magnesite were detected by X-ray
diffraction, and no broad SiO2(am) peak was detected (supplemental Fig. 2.1).
Fig. 2.4 shows SEM images of unreacted (Figs. 2.4A-B) and reacted (Figs.
2.4C-2F) olivine samples. Unreacted olivine grains are of uniform size and few fine
particles are visible. After reaction, amorphous silica is present both in large sheets
>100 µm across (Fig. 2.4C) and as irregular coatings on olivine grains (Fig. 2.4D).
The topography of the precipitated SiO2 phase (Fig. 2.4E) and the precipitated MgCO3
on the surface of the silica (Fig. 2.4F) are also evident. EDS scans of precipitated
37
silica (Fig. 2.4E) and magnesite (Fig. 2.4F) show that the precipitated silica contains
only silicon and oxygen, while the magnesite contains carbon, oxygen, magnesium,
and iron. A small silicon peak is also visible in the magnesite EDS spectrum, which
Table 2.2: Experimental data for four experiments.
time (days) [Mg] (mM) [Si] (mM) [Fe] (mM) alkalinity (meq) [CO2] (M)
Experiment 1 0.00 0.3 0.1 0.0 - -
60 °C, 100 bar 0.04 5.0 0.8 0.1 14.2 0.72
0.0 M NaCl 1.17 8.6 2.8 0.1 20.0 0.81
2.06 8.5 2.3 0.1 23.6 0.80
4.99 12.2 4.1 0.4 27.3 0.87
8.16 17.3 6.1 1.1 43.3 0.88
13.01 26.0 5.1 2.0 60.3 0.88
20.09 35.1 4.0 2.8 86.8 0.88
26.95 43.0 2.9 1.6 104.0 0.90
35.19 63.2 4.1 0.8 118.6 0.90
42.96 97.8 5.7 0.5 194.4 0.97
49.07 71.1 4.2 0.3 204.9 1.07
56.23 117.3 5.8 0.3 230.3 0.92
64.54 108.6 5.8 0.3 260.5 0.86
73.99 139.2 5.8 0.2 260.9 0.82
73.99 68.4 2.0 0.0 - -
Experiment 2 0.00 0.6 0.2 0.0 - -
60 °C, 100 bar 0.04 4.6 0.8 0.1 12.8 0.67
0.5 M NaCl 1.00 8.0 2.4 0.1 27.3 0.84
3.00 9.4 3.0 0.1 26.4 0.86
10.19 11.8 4.1 0.3 33.3 0.94
18.50 25.5 6.1 1.7 66.9 0.86
27.96 37.1 5.4 2.1 91.6 0.96
34.96 54.6 5.9 1.5 121.5 0.90
43.17 77.7 5.4 0.8 165.0 0.97
51.08 87.4 4.2 0.5 198.2 1.00
61.05 97.8 3.8 0.3 227.3 0.83
71.16 117.6 4.3 0.2 263.6 1.14
78.13 118.1 3.9 0.2 284.1 -
86.20 134.3 4.2 0.1 282.4 0.82
92.20 128.2 4.0 0.2 289.1 1.05
92.20 94.2 3.5 0.1 - -
Experiment 3 0.00 0.3 0.2 0.0 - -
60 °C, 100 bar 0.04 7.4 0.8 0.0 14.3 0.68
0.0 M NaCl 0.13 9.3 1.5 0.1 17.3 0.91
0.29 10.1 1.9 0.0 17.2 0.91
0.63 13.9 2.6 0.1 22.2 0.98
0.96 12.4 2.8 0.0 21.9 1.00
1.44 13.7 3.3 0.0 24.9 1.03
2.00 16.8 4.5 0.2 24.9 1.00
2.52 15.8 4.1 0.1 26.4 1.00
4.02 17.3 4.6 0.0 29.8 1.04
Experiment 4 0.00 1.7 1.1 0.5 - -
60 °C, 100 bar 0.04 7.8 1.5 0.2 12.12 0.92
0.0 M NaCl 1.00 10.9 3.1 0.2 18.64 1.06
3.09 14.5 4.2 0.3 22.37 0.93
7.25 17.8 5.1 0.3 24.38 1.06
14.08 16.0 4.5 0.0 29.83 1.2
23.04 22.0 6.1 0.1 28.47 0.99
36.15 22.3 6.3 0.3 30.57 0.99
58.27 32.1 8.0 1.1 44.74 0.95
66.40 34.5 7.3 1.3 62.14 1
77.15 38.6 6.9 1.6 48.68 1.02
91.18 44.4 6.7 1.5 68.25 1.08
93.38 48.3 7.4 1.2 76.15 1.07
98.17 47.1 6.8 0.9 84.13 1
38
we attribute to small amounts of silica precipitated on the surface of the magnesite
nodules (Fig. 2.4F). MgCO3 precipitated preferentially on silica (both coatings and
discreet particles) as opposed to olivine. No separate iron-rich phases were observed in
the reaction products of the experiments discussed here, though both EDS and XPS
showed that the Mg-carbonate phase contains iron. The BET surface area of unreacted
olivine was measured to be 6410 cm2/g. After 94 days of reaction, the BET surface
area was 13,920 cm2/g. Much of the increased surface area likely comes from
precipitated silica, which appears to hve high roughness from visual inpection of SEM
images (Fig. 2.4E).
XPS of the unreacted olivine surface showed a stoichiometric Mg:Si ratio
Figure 2.4: Images of unreacted (A, B) and reacted (C-F) olivine from Experiment
2. Unreacted olivine grains are of uniform size. After reaction, the solids are
dominated by large (>100 µm) silica sheets (C) and many grains are partially
coated with silica (D). Magnesite nodules grow via a spiral growth pattern on the
silica (coatings and sheets) (D, F), which itself has a structure composed of fused
spheres (E). EDS spectra of magnesite nodules (F) and silica (E) are inset.
39
(1.85±0.10:1) and olivine composition of Fo93±2 (supplemental Fig. 2.2). Within
uncertainty, the XPS measurement is equivalent to the ICP measurement (Fo92).
However, the ICP measurement is more precise and thus was used for all calculations.
After reaction periods as short as two days, the olivine surface region was depleted
with respect to Mg (and enriched with respect to Si) by 46±6 % (Mg:Si ratio of
1.0±0.1:1). After four days of reaction (Experiment 3), the surface was further
depleted in Mg and after 74 days of reaction (Experiment 1), Mg was not detected by
XPS (Fig. 2.2).
2.3.3 Thermodynamic Calculations
The calculated temporal evolution of magnesite, amorphous silica, and siderite
saturation states is shown in Figs. 2.5A-C for Experiments 1 and 2 (0.5 M NaCl), and
4 (0 M NaCl). All experiments were undersaturated with respect to forsterite by at
least 6 orders of magnitude (data not shown). Magnesite saturation was exceeded
between 17 and 20 days in Experiment 1, at 28 days in Experiment 2, and 91 days in
Experiment 4. Experiments reached amorphous silica saturation after only 5 days.
Siderite saturation was reached at 13 days and 28 days then dropped below saturation
in Experiments 1 and 2, respectively. Siderite saturation was not reached during
Experiment 4. Calculated pH versus time is presented Fig. 2.5D and shows a rise in
pH with time for the duration of all three experiments, though some variability is
visible for Experiment 2 after 60 days and is due to variance in the total CO2
measurements.
40
2.3.4 Rate Calculations
Olivine dissolution rates vary as a function of time, thus several rates for each
experiment are presented in Table 2.3. Initial dissolution rates were calculated using
the change in Mg concentration and the change in time between data points. Average
dissolution rates from 10-70 days were calculated with a linear regression of the data
points in that time range. Previous studies of olivine dissolution in aqueous solutions
have shown that forsterite dissolution rates are highly dependent on pH (e.g., Rimstidt
Figure 2.3: Saturation of magnesite (A), amorphous silica (B), and siderite (C) vs. time as well as
calculated pH vs. time (D). Data presented were calculated from Experiments 1, 2 and 4. Saturation
is presented as the logarithm of the reaction quotient Q divided by the equilibrium constant K.
Dashed lines in A-C represent equilibrium between the solid phase and the reaction solution
(logQ/K=0 at Q=K).
41
2012 and references therein), and thus one should avoid calculating rates when the pH
is changing, as it is in our experiments. However, although the pH is increasing
between 10 and 70 days we see no pH dependence of olivine dissolution rate as
evidenced by the linear relationship between Mg concentration and time.
The carbonation rate was calculated only for Experiment 2 because it was the
only experiment during which Mg-carbonate precipitated before depressurization, as
explained in section 3.2. Based on our experiments, we estimate the time at which
precipitation begins in an experiment with 0.5 M NaCl is between 74 and 94 days, as
an experiment lasting 74 days (Experiment 1) did not precipitate carbonate before
depressurization. Thus, assuming the observed precipitation occurred over 20 days, the
carbonation rate was calculated using the amount of carbonate formed before
depressurization and the initial geometric surface area of olivine, following Rimstidt et
al. (2012). The resulting carbonation rate is only a minimum magnesite precipitation
rate because we do not know precisely when carbonate precipitation began or the
actual surface area term. All rates presented in Table 2.3 are normalized with the
geometric surface area of unreacted olivine, with the BET-normalized rates given in
parenthesis. Experiment 4 (0 M NaCl) did not precipitate carbonate presumably
because of the lower overall olivine dissolution rate which resulted in a lower degree
of magnesite saturation compared to the experiments containing electrolyte (Fig.
2.5A).
Table 2.3: Geometric surface area-normalized reaction rates for four experiments, with
BET-normalized rates in parenthesis.
Initial dissolution rate
(first hour)
[mol olivine s-1cm-2]
Average dissolution rate
from 10-70 days
[mol olivine s-1cm-2]
Carbonation rate
[mol MgCO3 s-1cm-2]
Experiment 1 1.03 ± 0.10 x10-10
(5.50 x 10-12)
1.69 ± 0.15 x10-12
(9.02 x 10-14)
N/A (carbonate
precipitated during
depressurization)
Experiment 2 8.69 ± 0.96 x10-11
(4.63 x 10-12)
1.62 ± 0.08 x10-12
(8.64 x 10-14)
5.70 ± 0.68 x10-12
(3.04x10-13)
Experiment 3 1.61 ± 0.16 x10-10
(8.57 x 10-12) N/A N/A (no carbonate)
Experiment 4 1.35 ± 0.14 x10-10
(7.23 x 10-12)
3.26 ± 0.07 x10-13
(1.74 x 10-14) N/A (no carbonate)
42
2.4 Discussion
2.4.1 Solution Data and Secondary Phases
Solution composition data as a function of reaction time (Fig. 2.3) provide
insight into the dissolution and precipitation processes that occur during olivine
carbonation. The first day of reaction is characterized by a rapid, non-stoichiometric
release of ions into solution, particularly Mg2+
. Early in the reaction, Mg is
preferentially removed from the olivine surface, possibly by a direct exchange with
protons (Pokrovsky and Schott, 2000b). One result of preferential Mg removal is the
formation of a Si-rich (Mg-depleted) layer on the olivine surface (Bearat et al., 2006;
Daval et al. 2011; Pokrovsky and Schott, 2000b). This period of rapid release is
followed by a 3-5 day period during which olivine dissolution rate slows (as measured
by Mg, Si, or alkalinity). During this time period, dissolution is congruent with respect
to Mg and Si and the reaction fluid approaches saturation with respect to amorphous
silica. These first two periods (rapid Mg release followed by slowing dissolution) are
best illustrated by the results of Experiment 3 because more samples were obtained
during the first 4 days compared to the other three experiments. A similar change in
dissolution rate with time was also observed by Wang and Giammar (2013), although
dissolution in their experiment remained stoichiometric with respect to Mg and Si.
After 10 days, Mg concentrations again increased at a steady rate for the
remainder of Experiments 1 and 4 and until day 85 of Experiment 2 (Fig. 2.3A).
Calculations indicate that Experiment 2 was saturated with respect to amorphous silica
by day 10, although Si concentration continued to rise for several more days before
declining to saturation (Fig. 2.4B). The concentration maximum of Si is considered to
be due to kinetic limitation of silica precipitation. At 45 days and longer, the silica
precipitation rate is faster than the Si release rate and the concentration of Si in
solution remains constant. Experiment 1 showed significant variability in Si
concentration with time, which we attribute to an inconsistent sampling procedure
rather than any physical phenomenon in the reaction vessel. This effect occurs due to
the rapid temperature drop from 60 °C to room temperature (about 20 °C) which
43
occurs upon removing the sample for ICP analysis. The decrease in amorphous silica
solubility likely results in silica precipitation inside the sampling syringe in a matter of
minutes, such that a sample filtered after 5 minutes would have a lower Si
concentration than a sample filtered after 1 minute. We were able to experimentally
confirm this hypothesis after the experiments were completed, but do not have specific
data on sampling times for each data point presented here. Nevertheless, despite the
variability in Experiments 1 and 2, the trends in concentrations are consistent and the
maximum values are hence still a reliable indicator. We addressed this issue in
Experiments 3 and 4 by filtering the ICP samples within 60 seconds of withdrawing
them from the reactor.
At 10 days reaction time, the decrease in Fe concentrations followed a large
concentration maximum (both Experiments 1 and 2, Fig. 2.3D), which is indicative of
precipitation of a secondary phase; however, in this case the nature of that phase is not
clear. Saturation calculations illustrate that siderite is a possible secondary phase, but
it is slightly undersaturated for much of the experiment (>20 days). Calculations also
suggest the dissolved Fe is dominantly in the form of FeHCO3+. Based on the amount
of Fe released from olivine (and assuming stoichiometric Mg and Fe release), we
would expect to see up to 0.23 g of siderite in our reaction products. However, no
siderite was detected by either XRD or SEM (supplemental Fig. 2.1), which implies
that the Fe-rich phase is amorphous or possibly incorporated in the magnesite, similar
to other findings (Saldi et al. 2013). Analysis of the magnesite nodules with EDS
showed the presence of Fe, though the amount was not quantified due to the
complications associated with SEM/EDS quantification (Fig. 2.4F). We see no Fe in
the precipitated amorphous silica using EDS (Fig 2.4E). Another possibility is the
formation of a non-carbonate Fe phase, such as magnetite. However, no magnetite was
detected by XRD after reaction, suggesting that the majority of the Fe is incorporated
in magnesite and/or that any magnetite formed was beneath the detection limit of XRD
(1 wt. %).
Observations of secondary phases are consistent with the concept that silica
precipitation controls H4SiO4 concentration as the precipitated silica is present in large
44
(>100 m) sheets in Experiments 1, 2, and 4. (Fig. 2.4C). Amorphous silica also
formed a coating on the olivine grains visible in SEM (Fig. 2.4D), but there appears to
be space between the coating and the olivine, suggesting that the precipitated coating
did not prevent the interior olivine grain from continuing to dissolve and shrink in
size. Visual inspection of SEM images and the dramatic increase in BET surface area
(117% increase over 94 days in Experiment 2) indicate that amorphous silica formed
most of the available surface for secondary phase precipitation. The greater available
surface area of silica relative to olivine may explain why we observe the preferential
precipitation of magnesite on the silica surfaces rather than directly on the olivine.
Other possible explanations for preferential magnesite precipitation on silica as
opposed to olivine could include a lower interfacial energy between magnesite and
silica relative to magnesite and olivine (De Yoreo and Vekilov 2003), or a rougher
silica surface with more active sites for nucleation. For example, Fernandez-Martinez
et al. (2013) found that heterogeneous calcium carbonate nucleation on quartz is
favored relative to homogeneous nucleation due to the lower interfacial energy of
quartz.
In theory, one would expect Mg concentration to reach a constant value that
represents steady-state olivine dissolution and magnesite precipitation. None of the
experiments conducted in this study definitively show that behavior, though XRD and
coulometry analyses indicate that magnesite did precipitate in Experiment 2. Thus, the
rate of magnesite precipitation was likely slower than that of olivine dissolution.
Longer-duration experiments must be performed to examine the behavior of the
system over such time scales.
2.4.2 Initial Incongruent Dissolution
All experiments conducted as part of this study show initial incongruent
dissolution of olivine. Magnesium release is enhanced 3.5-4.5 times relative to Si over
the first hour of reaction, and by the seventh hour dissolution is congruent. Evidence
for initial incongruent dissolution includes solution composition as a function of
reaction time data (Fig. 2.3) and a comparison of olivine dissolution rates calculated
from Mg and Si (Fig. 2.6). Based on solution data, dissolution appears to be non-
45
stoichiometric again after several days of reaction, but this occurs only after
amorphous silica saturation is reached. Thus the deficiency of Si in solution in the late
experimental stages is more likely due to silica precipitation than a resumption of
incongruent dissolution. Dissolution rates calculated from Mg are higher than from Si
for the first 3 hours, at which point the rates become equal until silica precipitation
begins (Fig. 2.6). Initial incongruent dissolution is attributed to a rapid exchange
between Mg cations in the olivine and protons in solution (Pokrovsky and Schott,
2000a, 2000b; Oelkers et al., 2009), but it could also be due to localized saturation and
precipitation of a secondary, silicon-rich phase (Hellmann et al., 2012; King et al.,
2011). Using solution data from Experiment 3 and mass balance considerations, the
theoretical thickness of the Si-rich layer can be calculated. Residual Si is calculated by
subtracting the measured Si (at hour 7, when congruency is first observed) from the Si
expected based on released Mg. Using the geometric surface area of olivine and the
molar volume of amorphous SiO2 (29 cm3/mol), the theoretical layer thickness was
calculated to be 18 nm. This is an order of magnitude greater than estimates from
Pokrovsky and Schott (2000b), though their titrations were of a duration of only 1-8
minutes, and suggests the formation of a Si-rich, Mg-deficient surface layer occurs
rapidly and continues for at least several hours.
2.4.3 Dissolution Rate as a Function of SiO2(am) Saturation
The rate of olivine dissolution in Experiment 3 (no electrolyte) decreases
substantially as the reaction fluid approaches amorphous silica saturation (Fig. 2.6).
While the saturation index of amorphous silica increases from -1.4 to 0, the olivine
dissolution rate, as calculated from both Mg and Si release, decreases over two orders
of magnitude. This effect was also observed by Daval et al. (2011). Cations are
removed from the olivine surface, leaving behind a Si-rich layer (either leached or
precipitated) that passivates the surface. The Si-rich passivating layer grows in
thickness as more Mg is removed, and the resulting surface may be more silica-like
than olivine-like in its properties (Pokrovsky and Schott, 2000a). As the reacting fluid
reaches SiO2 saturation, the thermodynamic driving force for net dissolution of the Si-
rich layer approaches zero, and thus little to no dissolution of that layer occurs. As a
46
result, the overall dissolution rate of olivine drops significantly when calculated from
measured Mg release (some Mg likely still diffuses via interconnected pores through
the Si-rich layer, but at a slower rate) and approaches zero when calculated from the
measured Si release (because there is no net change of Si in solution). This trend is
observed in the data from all four experiments (Fig. 2.3).
2.4.4 Effect of Added Electrolyte on Olivine Dissolution
Two experiments presented here were conducted with 0.5 M NaCl background
electrolyte (Experiments 1 and 2) and two were conducted without any added
electrolyte (Experiments 3 and 4). The two groups show different behavior as a
function of time. The electrolyte-free experiments initially release Mg and Si at a
higher rate than the experiments with electrolyte, but over long times (>5 days), the
electrolyte-free experiments have a slower release of Mg and Si by a factor of 5
(Table 2.3). Slower Mg and Si release over long times (>5 days) suggests slower
kinetics in the electrolyte-free experiments, and consequently saturation with respect
to magnesite is attained much later in the experiments. The presence of 0.5 M NaCl
increases long-term rate of Mg release and thus the carbonation extent over the times
Figure 2.6: The rate of olivine dissolution, calculated from both Mg and Si data from
Experiment 3, decreases as the silica saturation state increases during the initial 3 days of
reaction. Non-stoichiometric dissolution occurs during the first time interval (due to
incongruent dissolution) and the last (presumably due to amorphous silica precipitation).
47
of our experiments (94 days). Also significant is the higher steady-state Si
concentration in Experiment 4 relative to Experiments 1 and 2. We note, however, that
while we reproduced the long-term behavior of olivine carbonation in experiments
with 0.5 M NaCl (Experiments 1 and 2), we present only one dataset showing the
long-term behavior of olivine carbonation in the absence of added electrolyte. Our
experience with reproducing both long-term behavior with NaCl and short-term
behavior with and without NaCl gives us confidence that the experiments are, in
general, reproducible.
In the absence of other factors, the experiments with 0.5 M NaCl should show
faster olivine dissolution because sodium is thought to enhance the reactivity of water
by modifying its nucleophilic properties (Icenhower and Dove, 2000). In the case of
olivine, which has a silica-like surface and a neutral to negative charge under our pH
conditions (pHIEP = 4.5, where pHIEP= isoelectric point, or the pH at which the surface
carries a net zero charge) (Pokrovsky and Schott, 2000a), cations such as Mg are
electrostatically attracted to the mineral surface. The addition of sodium may displace
Mg from the mineral-solution interface and disrupt the water-SiO2 network at the
mineral surface, thus enhancing the dissolution rate (Dove, 1999; Dove and Crerar,
1990; Dove and Nix, 1997; Icenhower and Dove, 2000). This hypothesis explains the
higher Mg release rate seen in Experiments 1 and 2 relative to Experiment 4 at longer
times, as well as previous work (i.e. Pokrovsky and Schott, 2000b). However, at short
(<4 days) times, olivine dissolution appears to be controlled by silica saturation state
(see section 4.3). Generally electrolytes such as NaCl increase the solubility of solid
phases because ion pairing reduces the activity of ions in solution, but experimental
evidence suggests the opposite is true in the case of amorphous silica and NaCl
solutions (Chen and Marshall, 1982). Solution data from the experiments presented in
this study support the conclusions of Chen and Marshall (1982), as electrolyte-free
Experiment 4 has higher steady-state Si concentrations relative to the NaCl
experiments. Depressed silica solubility in the experiments with added NaCl
(Experiments 1 and 2) correlates with lower initial Mg release, whereas the
experiments without NaCl (Experiments 3 and 4) had the fastest initial Mg release.
48
More specifically, at day 3, Experiment 2 (0.5 M NaCl) had lower concentrations of
Mg and Si (9.0 mM Mg and 3.0 mM Si) than salt-free Experiment 4 (14.5 mM Mg,
4.2 mM Si) (Table 2.4). We present the above argument as further evidence for the
direct and detrimental effect of silica saturation on the initial (0-4 days) dissolution
rate of olivine, but we also suggest that dissolved silica ceases to dominate the
dissolution kinetics at longer times (>4 days in our experiments). An important point
is that both dissolved silica and salt concentration have a large effect on the
dissolution rate of olivine, but neither is generally included in kinetic rate law
formulations.
2.4.5 Kinetics of Olivine Dissolution and Carbonation
Calculated olivine dissolution and carbonation rates based on our experiments
are presented in Table 2.3. The initial dissolution rate was calculated using d[Mg]/dt
for the first hour of reaction, which we have shown to be incongruent (Fig. 2.6).
Calculating these rates with d[Si]/dt yields a rate approximately one order of
magnitude lower. The initial rates were calculated over a short period of time (1 hour),
but the pH was changing over that time and thus we expect the rate also changed.
Thus, although each initial rate is a time-averaged rate, the time is short relative to the
duration of the experiment. Initial rates from Experiments 1 and 2 are consistent with
each other (within uncertainty); initial rates in Experiments 3 and 4 are also consistent
and approximately 50% higher than those from 1 and 2 (Table 2.3, Fig. 2.7A). We
suggest that Experiments 1 and 2 have lower initial dissolution rates because of higher
ionic strength solutions— the addition of NaCl in those experiments (0.5 M vs. 0 M in
Experiments 3 and 4) decreases the solubility of amorphous silica (Chen and Marshall
1982), and thus decreases the rate of olivine dissolution. Other researchers have found
olivine dissolution rates to increase with ionic strength (Pokrovsky and Schott,
2000b), but we consider the effect here to be related to silica solubility, which is
depressed by added NaCl (see section 4.2) (Chen and Marshall, 1982). Fig. 2.6
provides evidence that the rate of olivine dissolution depends strongly on the silica
saturation state, so decreasing the solubility of silica (by adding NaCl) would, by
extension, decrease the olivine dissolution rate under these experimental conditions. In
49
a flow-through experiment we would not expect to see the same effect, unless the fluid
residence time was substantial.
Consistent long-term (10-70 days) dissolution rates calculated from
Experiments 1 and 2 indicate high levels of experimental reproducibility (Fig. 2.7A).
These rates are about two orders of magnitude slower than the initial rates (Table
2.3.). The difference between long-term and initial rates may be due to the high levels
of dissolved SiO2 present in the reaction fluid at time >2 days, or could also be
explained by rapid fine particle dissolution and/or preferential release of Mg leading to
artificially high initial rates. Dissolution rates calculated from Experiment 4 are up to
an order of magnitude lower than those from Experiments 1 and 2 at pH > 4 (at 10-70
days). The pH range over which rates from Experiment 4 are calculated is smaller than
that of Experiments 1 and 2 because the reaction proceeded more slowly and thus less
olivine dissolved. For comparison, Fig. 2.7A also shows a data point from Wang and
Giammar (2013), who conducted experiments on olivine (Fo92) dissolution in the
presence of varying pressures of CO2, water, and varying NaCl concentrations. The
data point in Fig. 2.7A was measured at 50 °C, 100 bar pCO2, and 0 M NaCl. We
adjusted their rate by a factor of 40 to account for the higher mineral:solution ratio in
our experiments (1:50 in ours vs. 1:2000 in theirs). Their rate was calculated over a
longer time (8 hours) and thus a larger pH range (3.1-4.7) than ours. In addition, the
temperature is 10 ˚C lower, which may explain the slightly lower rate. We did not
apply a temperature correction because of uncertainty in the activation energy.
Long-term (10-70 days) olivine dissolution rates calculated for Experiments 1,
2, and 4 using the change in magnesium concentration over time and normalized by
initial geometric surface area are plotted against pH in Fig. 2.7B and compared with
three previously published pH-dependent rate laws from Hanchen et al. (2006),
Pokrovsky and Schott (2000b), and Rimstidt et al. (2012). Though the
mineral:solution ratio did change with time, the size and number of liquid samples
taken were minimized in order to maintain at least 90% of the original fluid volume,
and the surface area of the olivine was assumed to stay constant (i.e. any new surface
area that formed was considered to be due to secondary phase precipitation). Rimstidt
50
et al. (2012) carried out a systematic analysis of literature data (critical selection of
data followed by statistical analysis) and calculated the dissolution rate of olivine in
terms of pH and temperature (Equations 9 and 10, units are mol m-2
s-1
). Their analysis
combined all available data on olivine and forsterite dissolution, encompassing 0 ≤ pH
≥ 14 and 0 ≤ T ≥ 150 °C. The rate is expressed as a function of geometric surface area
(rgeo) in terms of pH and temperature (T) (Rimstidt et al. 2012).
)/T(pH....rgeo 633683200460220056log 5.6pHfor
(9)
)/T(pH...rgeo 139346502026038007.4log 5.6pHfor
(10)
Pokrovsky and Schott (2000b) developed a model based on experimental data
collected for experiments at 25 °C with variable pH (1 to 12), ionic strength (0.001 to
0.1M), CO2 (0 to 0.05M), aqueous Mg (10-6
to 0.05M), and aqueous Si (10-6
to
0.001M). They derived an olivine dissolution rate (rT) that is a function of the number
of reactive Si and Mg sites, which in turn are written as functions of adsorption and
exchange coefficients of participating aqueous species. In acidic and weakly alkaline
solutions the Si term (rSi) dominates, thus the Mg term is not included here (Equation
11). The Si component of the rate is proportional to the surface species concentration
({>Si2O-H+}) through the rate constant kSi. The surface species concentration is a
function of the activity of protons in solution (aH+), the activity of Mg in solution
(aMg2+), and two constants: Kads and Kex. Kads is the stability constant for the exchange
reaction between Mg2+
and 2H+ on the olivine surface and Kex is the stability constant
of H+ adsorption on the mineral surface. The Pokrovsky and Schott (2000b) rate was
computed using their formula, then adjusted for temperature using the activation
energy from Hanchen et al. (2006).
(11)
4
2
5.0
5.0
2
2
1
solutions acidicfor
Hex
Mg
Hads
Hads
SiSiSiT
aK
aaK
aKkHOSikrr
51
To the best of our knowledge, Rimstidt (2012) did not account for the effect of
temperature on measured pH, Hanchen et al. (2006) did calculate the pH at reaction
temperature using the ambient pH and EQ3/6 (Wolery et al. 1990). At 25 °C, a neutral
solution will have a pH of 7, but if the pH of that same neutral solution is measured at
an elevated temperature, it will be lower because log KW decreases with increasing
temperature. For example, in our experiments at 60 °C the neutral pH is 6.37. We also
do not correct the models (or our data) for the temperature dependence of pH, but
suggest that it may partially explain scatter in data obtained at different temperatures.
The models of Pokrovsky and Schott (2000b) and Rimstidt et al. (2012) both predict
that the dissolution rate will decrease by about an order of magnitude as pH increases
from 4 to 6. The slopes of the three models are similar, meaning the dependence on
the hydrogen ion activity is also similar.
Figure 2.7: A: Experimental rate data (normalized to geometric surface area and calculated from Mg
concentrations) as a function of pH. Deviation at the lowest pH (first time point) is due to a
combination of incongruent dissolution creating an artificially high rate and dissolution of fine particles
not removed by washing. Experimental data, particularly from 1 and 2, do not show a pH-dependence
at pH > 4.5. Horizontal uncertainty is defined by the pH values at the two data points used to calculate
each rate. For comparison, one data point from Wang and Giammar (2013) is included. This rate was
measured over 8 hours at 50 °C, 100 bars pCO2, and 0 M NaCl; we corrected it by a factor of 40 to
account for the lower forsterite concentration in their experiments (0.5 g/L compared to our 20g/L). B:
Long term (10-70 days) dissolution rates compared to three published models for forsterite dissolution
(Pokrovsky and Schott, 2000b; Hanchen et al., 2006; Rimstidt et al., 2012). Rates from experiments
with 0.5M NaCl are consistent with the models, the rate from Experiment 4 (no added electrolyte) is
lower. All rates shown here were calculated before Mg-carbonate precipitation occurred.
52
Data from the present study show a different pattern (Fig. 2.7A). The
dissolution rate for the first hour of reaction (3.1 < pH < 4.3) is an order of magnitude
greater than predicted by the models and may be artificially high due to incongruent
dissolution associated with the formation of the Si-rich layer and/or rapid dissolution
of fine particles. Calculating the same rate using Si instead of Mg gives an order of
magnitude lower rate (not shown). Rates measured at intermediate to long times (pH >
4.1) agree in magnitude with the model predictions, but do not show the expected pH
dependence (Fig. 2.7B). Rates derived from Experiment 4 are lower than those from 1
and 2 at pH>4.4 because these pH values are associated with longer reaction times,
when the background electrolyte increases the rate of reaction. Experiment 4 rates
appear to decrease monotonically with increasing pH in the same way as the models,
but as this experiment was slower overall, it covered a relatively short pH range.
Longer reaction time (and thus higher reaction extent) is necessary to investigate the
pH dependence of dissolution rates under salt-free reaction conditions.
The lack of pH dependence in our dissolution rate measurements (from
Experiments 1 and 2) has not to our knowledge been reported previously and likely
indicates that the experimental conditions here facilitated a different kinetically
limiting step than the one observed using flow-through reactors. There are two
primary factors that separate the experiments in this study from more traditional flow-
through kinetic studies: high CO2(aq) and high H4SiO4. Thus, we propose that one or
both of these species are responsible for the pH independence of the dissolution rates.
Other studies have suggested that carbonate species hinder olivine dissolution
(Pokrovsky and Schott, 2000b), which may be happening at pH<5, though this would
not explain the pH-independence unless carbonate species have a positive effect on
dissolution rate at pH>5. As pH increases, the balance of carbonate species shifts from
nearly all carbonic acid, H2CO3(aq), to several mol percent bicarbonate, HCO3-.
Bicarbonate ions may activate the dissolution reaction at mildly acidic pH, which
could partially explain the (relatively) enhanced rate of olivine carbonation in the
presence of sodium bicarbonate (Chen et al., 2006; Jarvis et al., 2009). In fact,
thermodynamic calculations indicate that up to 44% of the Mg in solution is
53
complexed by bicarbonate during the experiment (free Mg2+
is 50% or greater), which
lowers the activity of Mg in solution and possibly enhances olivine dissolution relative
to lower pH values where the carbonate species are predominantly carbonic acid.
While the presence of carbonate species may explain the pH independence, the more
likely cause is high dissolved SiO2. Amorphous silica dissolution rates are not pH
dependent at acidic to neutral pH, so we suggest that under our experimental
conditions (high H4SiO4 and pH 4-6), removal of silicon from the olivine structure is
rate limiting. Further study on this topic needs to be performed in order to verify or
refute our hypothesis.
The role of magnesite precipitation in our experiments is critical to the
discussion of carbonation kinetics, especially as a previous study suggests that the rate
of magnesite precipitation may limit overall carbonation kinetics (Saldi et al. 2012).
Although the work presented here was not specifically designed to quantify
precipitation rates, our results regarding precipitation provide useful information for
the design of future carbonation experiments. As we have no way of separating the
rates of olivine dissolution from that of magnesite precipitation in our current
experiments, the discussion above has assumed that all changes of Mg in solution in
the early phases of the experiments are due to olivine dissolution, an assumption that
we know to be true initially. However, Mg-carbonates in the reaction products indicate
that at 74 days and later, magnesite precipitation occurred concurrently with olivine
dissolution. The results of Experiments 1 and 2 allow us to bracket the time at which
carbonate precipitation begins in experiments containing 0.5 M NaCl (between 74 and
94 days). More specifically, Experiment 1 lasted 74 days and did not precipitate
carbonate before depressurization, but Experiment 2 lasted 94 days and precipitating
Mg-carbonate consumed 14 mol % Mg from the starting olivine. Thus, in the case of
Experiment 2, magnesite precipitation occurs over a maximum of 20 days. Further
complicating the calculation of magnesite precipitation is the lack of a known
available surface area for precipitation as the experiments were not seeded with
magnesite crystals. As an estimate of the surface area, we use the final BET surface
area, which accounts for both available olivine and silica surface areas. Taking the
54
above points into consideration, our magnesite precipitation rate is >1.40 x 10-13
mol
cm-2
s-1
, which is 5 orders of magnitude greater than predicted by Saldi et al. (2009).
For the comparison, we used the second-order rate expression from Saldi et al. (2009)
and assumed a surface area equal to the final BET area of 1.39 m2, a time of 20 days, a
saturation index of 2, and corrected the rate constant for temperature using an
Arrhenius-type relationship and an activation energy of 159 kJ/mol (from Saldi et al.
2009). In order to match our estimated precipitation rate with that predicted by
equation 5 from Saldi et al. (2009) for nucleation on pre-existing magnesite surfaces,
we would need to have a larger surface area available for magnesite precipitation
and/or lengthen the time over which the magnesite precipitated. A third possibility is
that the iron from olivine (Fo92) is incorporated into magnesite, which has been
shown to increase the rate of magnesite precipitation (Saldi et al., 2013) and is evident
in EDS data (Fig. 2.4E). An additional caveat regarding the comparison of our
estimated precipitation rate with that from Saldi et al. (2009) is that they measured
magnesite precipitation rates on magnesite crystal seeds, which have a different
surface energy than the phases in our experiments. We also acknowledge that the
factors that control crystal formation are nucleation and growth rates, neither of which
are measurable with our experiments as presented here. Nevertheless, the rates we
calculate represent a minimum bound on the rate of carbonate precipitation under
these experimental conditions, and suggest that carbonate precipitation may not limit
mineral carbonation in the long-term, once sufficient supersaturation is reached with
respect to the precipitating phase.
2.5 Conclusions
Experimental results demonstrate the temporal evolution of olivine carbonation
in a batch reactor where olivine dissolution (reaction 2), SiO2(am) precipitation
(reaction 3), CO2 dissolution/dissociation (reactions 4-6), and magnesite precipitation
(reaction 7) all occur simultaneously. Our results suggest that once a critical level of
magnesite supersaturation is achieved, magnesite precipitation occurs at a faster rate
than previously predicted (Saldi et al. 2009). In agreement with previous studies, a Si-
rich passivating layer forms on olivine surfaces initially and can persist for many days,
55
potentially limiting reaction rate (e.g., Daval et al. (2009)). In addition, we have
provided further evidence that olivine dissolution rate decreases with increasing
aqueous silica concentration as previously reported by Daval et al. (2011) and
Pokrovsky and Schott (2000b). Our rates are consistent with those predicted by
previous (CO2-free, low H4SiO4) models and studies, particularly at 4 < pH < 5 and
with 0.5 M added NaCl, but unlike the models, we do not see the rates decrease with
increasing pH. Combination of these observations suggests that either CO2 limits the
reaction at low pH but promotes dissolution at mildly acidic pH, or that the rate-
limiting step under high dissolved SiO2 conditions is the destruction of the Si-rich
layer, which is pH-independent.
Comparison of experiments with 0.5 M NaCl with those lacking NaCl in
distilled water suggests two different rate regimes, possibly governed by two different
dissolution mechanisms. The first regime extends from 0-5 days and is characterized
by rapidly slowing dissolution measured by Mg, Si, or alkalinity. In this regime, the
NaCl-free experiments show higher rates and achieve higher Mg, Si, and alkalinity
concentrations. The second regime begins at about day 10 (in experiments with
background electrolyte, later in the salt-free experiment) and is distinguished by a
constant dissolution rate (measured by Mg or alkalinity) and Si concentrations at or
above saturation with respect to SiO2(am). In this regime, the NaCl-free experiment
had a significantly slower dissolution rate than the experiments with NaCl. This effect
is not included in most commonly used rate laws for olivine dissolution.
Based on the results of this study, further work should be targeted on the
formation and destruction of the passivating Si-rich layer. Two models for formation
have been proposed: the first is a leaching model, in which Mg is preferentially
removed from the olivine surface leaving behind a polymerized network of SiO4
tetrahedra (Bearat et al., 2006; Daval et al., 2011; Oelkers, 2001b; Pokrovsky and
Schott, 2000a, 2000b); the second is a dissolution/reprecipitation model in which Mg
and Si leave the surface at the same rate, but SiO2 re-precipitates at the mineral-fluid
interface (Hellmann et al., 2012; King et al., 2011). We consider these two models to
be end members of a continuum rather than two discreet ideas, and that the layer
56
formation may be a result of a hybrid of the two. Our future work will more explicitly
address this point. Understanding the mechanism of formation of the silica-rich layer
is an important first step in developing a method to prevent its formation/encourage its
destruction and is essential for the development of accurate geochemical models
addressing carbon storage in mafic and ultramafic rocks.
2.6 Acknowledgments
We would like to acknowledge the Global Climate and Energy Project (GCEP-
48942) at Stanford University for funding this project, along with technical assistance
from Bob Jones and Guangchao Li. We also acknowledge the Stanford
Nanocharacterization Laboratory, where many of the analyses were performed.
57
2.7 References
Bearat H., Mckelvy M. J., Chizmeshya A. V. G., Gormley D., Nunez R.,
Carpenter R. W., Squires K. and Wolf G. H. (2006) Carbon Sequestration via Aqueous
Olivine Mineral Carbonation: Role of Passivating Layer Formation. Environmental
Science & Technology 40, 4802–4808.
Blum A. and Lasaga A. (1988) Role of surface speciation in the low
temperature dissolution of minerals. Nature 331, 431–433.
Boschi C., Dini A., Dallai L., Ruggieri G. and Gianelli G. (2009) Enhanced
CO2-mineral sequestration by cyclic hydraulic fracturing and Si-rich fluid infiltration
into serpentinites at Malentrata (Tuscany, Italy). Chemical Geology 265, 209–226.
Chen C-T. A. and Marshall W. L. (1982) Amorphous silica solubilities IV:
Behavior in pure water and aqueous sodium chloride, sodium sulfate, magnesium
chloride, and magnesium sulfate solutions up to 350 °C. Geochimica et Cosmochimica
Acta 46, 279–287.
Chen Y. and Brantley S. L. (2000) Dissolution of forsteritic olivine at 65 °C
and 2<pH<5. Chemical Geology 165, 267–281.
Chen Z-Y., O’Connor W. K. and Gerdemann S. J. (2006) Chemistry of
Aqueous Mineral Carbonation for Carbon Sequestration and Explanation of
Experimental Results. Environmental Progress 25, 161–166.
Daval D., Sissmann O., Menguy N., Saldi G. D., Guyot F., Martinez I.,
Corvisier J., Garcia B., Machouk I., Knauss K. G. and Hellmann R. (2011) Influence
of amorphous silica layer formation on the dissolution rate of olivine at 90 °C and
elevated pCO2. Chemical Geology 284, 193–209.
De Yoreo J. J. and Vekilov P. G. (2003) Principles of Crystal Nucleation and
Growth. Reviews in Mineralogy and Geochemistry 54, 57–93.
Dove P. M. and Crerar D. A. (1990) Kinetics of quartz dissolution in
electrolyte solutions using a hydrothermal mixed flow reactor. Geochimica et
Cosmochimica Acta 54, 955–969.
Dove P. M. and Nix C. J. (1997) The influence of the alkaline earth cations,
magnesium, calcium, and barium on the dissolution kinetics of quartz. Geochimica et
Cosmochimica Acta 61, 3329–3340.
Dove P. M. (1999) The dissolution kinetics of quartz in aqueous mixed cation
solutions. Geochimica et Cosmochimica Acta 63, 3715–3727.
58
Felmy A. R., Qafoku O., Arey B. W., Hu J. Z., Hu M., Todd Schaef H., Ilton
E. S., Hess N. J., Pearce C. I., Feng J. and Rosso K. M. (2012) Reaction of water-
saturated supercritical CO2 with forsterite: Evidence for magnesite formation at low
temperatures. Geochimica et Cosmochimica Acta 91, 271–282.
Fernandez-Martinez A., Hu Y., Lee B., Jun Y.-S. and Waychunas G. A. (2013)
In Situ Determination of Interfacial Energies between Heterogeneously Nucleated
CaCO3 and Quartz Substrates: Thermodynamics of CO2 Mineral Trapping.
Environmental Science & Technology 47, 102–109.
Giammar D. E., Bruant R. G. and Peters C. A. (2005) Forsterite dissolution
and magnesite precipitation at conditions relevant for deep saline aquifer storage and
sequestration of carbon dioxide. Chemical Geology 217, 257–276.
Goff F. and Lackner K. (1998) Carbon Dioxide Sequestering Using Ultramafic
Rocks. Environmental Geosciences 5, 89–101.
Goldich S. S. (1938) A Study in Rock-Weathering. The Journal of Geology 46,
17–58.
Golubev S. V, Pokrovsky O. S. and Schott J. (2005) Experimental
determination of the effect of dissolved CO2 on the dissolution kinetics of Mg and Ca
silicates at 25 °C. Chemical Geology 217, 227–238.
Hänchen M., Prigiobbe V., Storti G., Seward T. M. and Mazzotti M. (2006)
Dissolution kinetics of fosteritic olivine at 90–150 °C including effects of the presence
of CO2. Geochimica et Cosmochimica Acta 70, 4403–4416.
Hänchen M., Krevor S., Mazzotti M. and Lackner K. S. (2007) Validation of a
population balance model for olivine dissolution. Chemical Engineering Science 62,
6412 – 6422.
Hellmann R., Wirth R., Daval D., Barnes J.-P., Penisson J.-M., Tisserand D.,
Epicier T., Florin B. and Hervig R. L. (2012) Unifying natural and laboratory
chemical weathering with interfacial dissolution–reprecipitation: A study based on the
nanometer-scale chemistry of fluid–silicate interfaces. Chemical Geology 294-295,
203–216.
Hopkinson L., Rutt K. and Cressey G. (2008) The Transformation of
Nesquehonite to Hydromagnesite in the System CaO-MgO-H2O-CO2 : An
Experimental Spectroscopic Study. The Journal of Geology 116, 387–400.
Icenhower J. P. and Dove P. M. (2000) The dissolution kinetics of amorphous
silica into sodium chloride solutions : Effects of temperature and ionic strength.
Geochimica et Cosmochimica Acta 64, 4193–4203.
59
Jarvis K., Carpenter R. W., Windman T., Kim Y., Nunez R. and Alawneh F.
(2009) Reaction Mechanisms for Enhancing Mineral Sequestration of CO2.
Environmental Science & Technology 43, 6314–6319.
Kelemen P. B. and Matter J. (2008) In situ carbonation of peridotite for CO2
storage. Proceedings of the National Academy of Sciences 105, 17295–17300.
Kelemen P. B., Matter J., Streit E. E., Rudge J. F., Curry W. B. and Blusztajn
J. (2011) Rates and Mechanisms of Mineral Carbonation in Peridotite: Natural
Processes and Recipes for Enhanced, in situ CO2 Capture and Storage. Annual Review
of Earth and Planetary Sciences 39, 545–576.
King H. E., Plümper O. and Putnis A. (2010) Effect of secondary phase
formation on the carbonation of olivine. Environmental Science & Technology 44,
6503–6509.
King H. E., Plümper O., Geisler T. and Putnis A. (2011) Experimental
investigations into the silicification of olivine: Implications for the reaction
mechanism and acid neutralization. American Mineralogist 96, 1503–1511.
Kwak J. H., Hu J. Z., Hoyt D. W., Sears J. A., Wang C., Rosso K. M. and
Felmy A. R. (2010) Metal Carbonation of Forsterite in Supercritical CO2 and H2O
Using Solid State 29
Si, 13
C NMR Spectroscopy. Journal of Physical Chemistry C 114,
4126–4134.
Kwak J. H., Hu J. Z., Turcu R. V. F., Rosso K. M., Ilton E. S., Wang, C., Sears
J. A., Engelhard M. H., Felmy A. R. and Hoyt D. W. (2011) The role of H2O in the
carbonation of forsterite in supercritical CO2. International Journal of Greenhouse
Gas Control 5, 1081–1092.
Li D. and Duan Z. (2007) The speciation equilibrium coupling with phase
equilibrium in the H2O–CO2–NaCl system from 0 to 250 °C, from 0 to 1000 bar, and
from 0 to 5 molality of NaCl. Chemical Geology 244, 730 – 751.
Loring J. S., Thompson C. J., Wang Z., Joly A. G., Sklarew D. S., Schaef H.
T., Ilton E. S., Rosso K. M. and Felmy A. R. (2011) In situ infrared spectroscopic
study of forsterite carbonation in wet supercritical CO2. Environmental Science &
Technology 45, 6204–6210.
Matter J. M. and Kelemen P. B. (2009) Permanent storage of carbon dioxide in
geological reservoirs by mineral carbonation. Nature Geoscience 2, 837-841.
Oelkers E. H. (2001a) An experimental study of forsterite dissolution rates as a
function of temperature and aqueous Mg and Si concentrations. Chemical Geology
175, 485–494.
60
Oelkers E. H. (2001b) General kinetic description of multioxide silicate
mineral and glass dissolution. Geochimica et Cosmochimica Acta 65, 3703–3719.
Oelkers E. H., Golubev S. V, Chairat C., Pokrovsky O. S. and Schott J. (2009)
The surface chemistry of multi-oxide silicates. Geochimica et Cosmochimica Acta 73,
4617–4634.
Olsen A. A. and Rimstidt J. D. (2008) Oxalate-promoted forsterite dissolution
at low pH. Geochimica et Cosmochimica Acta 72, 1758–1766.
Onyeagocha A. C. (1978) Twin Sisters dunite : Petrology and mineral
chemistry. Geological Society of America Bulletin 89, 1459–1474.
Parkhurst D. L. and Appelo C. A. J. (1999) User’s guide to PHREEQC (
Version 2 )— A computer program for speciation, batch-reaction, one-dimentional
transport, and inverse geochemical calculations., Denver, Colorado.
Pokrovsky O. S. and Schott J. (2000a) Forsterite surface composition in
aqueous solutions : A combined potentiometric, electrokinetic, and spectroscopic
approach. Geochimica et Cosmochimica Acta 64, 3299 –3312.
Pokrovsky O. S. and Schott J. (2000b) Kinetics and mechanism of forsterite
dissolution at 25 °C and pH from 1 to 12. Geochimica et Cosmochimica Acta 64,
3313–3325.
Prigiobbe V., Hänchen M., Costa G., Baciocchi R. and Mazzotti M. (2009)
Analysis of the effect of temperature, pH, CO2 pressure and salinity on the olivine
dissolution kinetics. Energy Procedia 1, 4881–4884.
Rimstidt J. D., Brantley S. L. and Olsen A. A. (2012) Systematic review of
forsterite dissolution rate data. Geochimica et Cosmochimica Acta 99, 159–178.
Rosenbauer R. J., Bischoff J. L. and Fadtke A. S. (1983) Hydrothermal
alteration of graywacke and basalt by 4 molal NaCl. Economic Geology 78, 1701–
1710.
Rosenbauer R. J., Koksalan T. and Palandri J. L. (2005) Experimental
investigation of CO2–brine–rock interactions at elevated temperature and pressure:
Implications for CO2 sequestration in deep-saline aquifers. Fuel Processing
Technology 86, 1581–1597.
Saldi G. D., Jordan G., Schott J. and Oelkers E. H. (2009) Magnesite growth
rates as a function of temperature and saturation state. Geochimica et Cosmochimica
Acta 73, 5646–5657.
61
Saldi G. D., Schott J., Pokrovsky O. S., Gautier Q. and Oelkers E. H. (2012)
An experimental study of magnesite precipitation rates at neutral to alkaline conditions
and 100–200 °C as a function of pH, aqueous solution composition and chemical
affinity. Geochimica et Cosmochimica Acta 83, 93–109.
Saldi G. D., Daval D., Morvan G. and Knauss K. G. (2013) The role of Fe and
redox conditions in olivine carbonation rates: an experimental study of the rate
limiting reactions at 90 and 150 °C in open and closed systems. Geochimica et
Cosmochimica Acta 118, 157–183.
Shirokova L. S., Bénézeth P., Pokrovsky O. S., Gerard E., Ménez B. and
Alfredsson H. (2012) Effect of the heterotrophic bacterium Pseudomonas reactans on
olivine dissolution kinetics and implications for CO2 storage in basalts. Geochimica et
Cosmochimica Acta 80, 30–50.
Steefel C. I. and Van Cappellen P. (1990) A new kinetic approach to modeling
water-rock interaction: The role of nucleation, precursors, and Ostwald ripening.
Geochimica et Cosmochimica Acta 54, 2657–2677.
Wang F. and Giammar D. E. (2013) Forsterite Dissolution in Saline Water at
Elevated Temperature and High CO2 Pressure. Environmental Science & Technology
47, 168-173.
Welch S. A. and Banfield J. F. (2002) Modification of olivine surface
morphology and reactivity by microbial activity during chemical weathering.
Geochimica et Cosmochimica Acta 66, 213–221.
Wogelius R. A. and Walther J. V. (1991) Olivine dissolution at 25 °C : Effects
of pH, CO2, and organic acids. Geochimica et Cosmochimica Acta 55, 943–954.
Wogelius R. A. and Walther J. V. (1992) Olivine dissolution kinetics at near-
surface conditions. Chemical Geology 97, 101–112.
Wolery T. J., Jackson K. J., Bourcier W. L., Bruton, C. J., Viani B. E., Knauss
K. G. and Delany J. M. (1990) Current Status of the EQ3/6 Software Package for
Geochemical Modeling. In Chemical Modeling of Aqueous Systems II pp. 104–116.
Zakaznova-Herzog V. P., Nesbitt H. W., Bancroft G. M. and Tse J. S. (2008)
Characterization of leached layers on olivine and pyroxenes using high-resolution
XPS and density functional calculations. Geochimica et Cosmochimica Acta 72, 69–
86
62
63
.
Chapter 3: The role of the Si-rich surface layer in olivine dissolution 1: High-resolution TEM study
Natalie C. Johnsona,*
, Ariel Jacksonb, Burt Thomas
c, Robert J. Rosenbauer
c,
Dennis K. Birdd, Gordon E. Brown, Jr.
a,d,e, Kate Maher
d
a. Department of Chemical Engineering, Stauffer III, 381 North-South Mall, Stanford University, Stanford, CA 94305, USA
b. Department of Materials Science and Engineering, 496 Lomita Mall, Durand Building,
Stanford University, Stanford, CA 94305, USA
c. U.S. Geological Survey, 345 Middlefield Rd, Menlo Park, CA 94025, USA
d. Department of Geological & Environmental Sciences, 450 Serra Mall, Braun Hall Building 320,
Stanford University, Stanford, CA 94305-2115, USA
e. Department of Photon Science and Stanford Synchrotron Radiation Lightsource,
SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA
Submitted to Geochimica et Cosmochimica Acta, June 2014.
64
Abstract
Olivine (Mg2SiO4) is a reactive Mg-silicate mineral that could serve as a cation
source for subsurface carbon dioxide (CO2) mineralization as part of a global carbon
management strategy. However, olivine must dissolve in order to produce free Mg2+
,
and the dissolution process may be hindered by the formation of a Si-rich surface
layer. In order to assess the formation and evolution of the surface layer, olivine was
reacted with water and CO2 in the presence of an electrolyte at conditions relevant for
subsurface storage (60 °C, 100 bar pCO2). Parallel experiments, lasting up to 19 days,
were conducted under well-mixed and unmixed conditions. Dissolution proceeded 4
times faster in the mixed experiment relative to the unmixed after 19 days, suggesting
that the unmixed experiment was transport limited rather than surface-reaction
controlled. Transmission electron microscopy (TEM) cross sections were prepared
from reacted olivine grains after 2 and 19 days of reaction, and analyzed for
crystallinity and chemical composition as a function of depth. Several types of Si-rich
surface layers were observed: Mg-depleted olivine (crystalline, but enriched in Si), an
“active” layer (amorphous, Mg/Si changes with depth), and a newly precipitated layer
(amorphous, Mg/Si<0.3). Magnesium-depleted olivine was only present in the mixed
experiments, and the precipitated layer was only visible in experiments that exceeded
SiO2(am) saturation in the bulk fluid. The active layer was present on all reacted
olivine examined. Results suggest that the active layer forms first (in 2 days or less),
during the time that both incongruent dissolution and a decrease in dissolution rate are
observed, whereas the precipitated layer forms later and does not passivate the mineral
surface.
65
3.1 Introduction
The reaction of (Mg,Fe,Ca)-silicates with CO2-rich fluids to form stable
carbonate minerals provides a long-term storage option for anthropogenic CO2.
However, the slow reaction kinetics and the formation of a silica-rich layer on the Mg-
silicate surface may limit widespread industrial implementation of this approach
(Bearat et al., 2006; Daval et al., 2011; Oelkers et al., 2008; Chapter 2). Knowledge of
the underlying mechanisms that control olivine dissolution is required before
strategies to enhance the dissolution rate can be developed. In addition, improved
understanding of the processes associated with olivine dissolution may have
implications for other silicate minerals
Numerous studies of olivine dissolution in mixed-flow and plug-flow reactors
have demonstrated that olivine dissolution rates increase with increasing temperature,
and decrease with increasing pH (Hänchen et al., 2006; Mazzotti et al., 2010; Oelkers,
2001; Pokrovsky and Schott, 2000a; Prigiobbe et al., 2009; Wogelius and Walther,
1992, 1991). Wang and Giammar (2013) conducted batch experiments and also found
olivine dissolution rates to increase with increasing temperature, although they did not
examine rate as a function of pH. In another set of batch experiments, we observed
little effect of pH on dissolution rates at intermediate reaction times (10-70 days) and
mildly acidic pH (pH=4-6) (Chapter 2).
The differing results between experiments conducted in batch reactors and
flow through/mixed flow reactors may yield insights into the processes occurring at
the mineral surface. One important distinction is that flow reactors allow for the
careful control of the aqueous composition, whereas batch experiments allow the
aqueous phase to evolve throughout the duration of the experiment. Most flow
reactions are conducted far from equilibrium with respect to the primary phase
(olivine) and secondary phases (amorphous silica and magnesite); while batch
reactions may reach or exceed saturation with respect to multiple secondary phases.
For example, Oelkers (2001) studied the effect of dissolved Si and Mg on olivine
dissolution and found no effect on the rates in the range 0 < [Si] < 0.62 mmol/kg and 0
< [Mg] < 1.36 mmol/kg. However, these ranges do not extend to either silica
66
saturation (2 mmol/kg at 25 °C) or magnesite saturation (no carbonate species were
present), meaning the experiments were conducted far from equilibrium with respect
to likely secondary phases. In contrast, we found the olivine dissolution rate to depend
strongly on aqueous [Si] near amorphous silica saturation (Chapter 2) along with
Daval et al. (2011), and Sissmann et al. (2013),.
Understanding secondary phase formation is critical to understanding olivine
dissolution, as a Si-rich phase is known to form on olivine surfaces during dissolution
(Bearat et al., 2006; Daval et al., 2011; King et al., 2010; Pokrovsky and Schott,
2000b; Sissmann et al., 2013; Zakaznova-Herzog et al., 2008). Two hypotheses have
been proposed to explain the origin of this Si-rich surface coating. In the “leached
layer hypothesis,” the Si-rich surface layer forms via preferential removal of Mg from
the olivine surface, leaving behind a network of polymerized silica tetrahedra
(Hellmann et al., 1990; Pokrovsky and Schott, 2000a; Ruiz-Agudo et al., 2012;
Zakaznova-Herzog et al., 2008). The “precipitated layer hypothesis” describes local
areas of amorphous silica saturation near the olivine surface, resulting in a layer of
precipitated SiO2 (Hellmann et al., 2013, 2012; King et al., 2011, 2010). One line of
reasoning supporting the precipitated layer theory is that solid-state diffusion is quite
slow and leached layer formation requires the diffusion of cations through the layer,
which would limit the potential thickness of the surface layer (Hellmann et al., 2012).
To date, it is unclear whether the Si-rich layer passivates the olivine surface and
decreases the dissolution rate; some studies have shown that dissolution rates decrease
continually after the Si-rich layer develops (Daval et al., 2011, 2009; Hellmann et al.,
2012) while others have not observed a substantial effect (King et al., 2010). A
precipitated layer may be porous initially, but later secondary phases (i.e. carbonates,
phyllosilicates) may precipitate in the pores and reduce the permeability of the surface
layer (Hellmann et al. 2013). Recent work also suggests that Fe3+
may play an
important role in stabilizing the Si-rich layer and slowing olivine (Fo88) dissolution
(Saldi et al., 2013; Sissmann et al., 2013), at least until conditions become sufficiently
reducing to dissolve the Fe-phase, at which point dissolution resumes with a spike in
aqueous Fe concentration (Sissmann et al., 2013).
67
Several studies that have examined surface layer formation using high
resolution Transmission Electron Microscopy (TEM) have observed a Si-rich surface
layer, ranging from 1-1000 nm thick, on a variety of silicate minerals (Daval et al.,
2013, 2009; Hellmann et al., 2013, 2012; Sissmann et al., 2013). These surface layers
are generally observed to be amorphous, porous, and silica-rich with sharp, step-like
chemical gradients between the fresh mineral surfaces and surface layers. Transport
calculations suggest that cation diffusion through the surface layer is not the limiting
step in the overall reaction; rather, the rate is controlled by the surface reaction at the
fresh mineral-fluid interface (Daval et al., 2009; Hellmann et al., 2012). Conversely,
Schulze et al. (2004) studied surface layers that form during serpentine dissolution
using TEM and XPS, and based on their results, hypothesized that cation diffusion
through the surface layer is limiting. To our knowledge, there has been only one study
that explicitly compares dissolution rate with Si-rich layer thickness, and this study
focused on diopside (Daval et al., 2013). No such studies have yet been published for
olivine.
Although numerous studies of olivine dissolution exist in the literature, few
have been carried out in the presence of CO2. Further, many studies have observed the
formation of a Si-rich layer on the olivine surface during dissolution, but none to our
knowledge have probed the surface layer in depth in order to understand its
composition, formation mechanism, temporal evolution, and role in passivating
olivine surfaces. The objective of the present study is to use high-resolution,
nanometer-scale techniques to study the Si-rich layer that forms when olivine
dissolves in the presence of CO2 and H2O to understand both how such layers form
and how they participate in the dissolution reaction.
Our study is based on a series of four experiments designed to investigate the
effects of transport and reaction time on the Si-rich surface layer. Reaction progress in
each experiment is tracked using time-resolved solution compositions. We present
TEM data from reacted olivine cross-sections at two time intervals, which give insight
into the chemical composition, level of crystallinity, and formation mechanisms of the
observed surface layers. We finish with a simple diffusion model to show that
68
diffusion through the amorphous surface layer is sufficient to maintain the dissolution
rates we measure.
3.2 Methods
3.2.1 Autoclave reactor experimental system and sampling procedures
Experiments were carried out in a flexible, inert, Au reaction cell placed inside
a steel autoclave, as described in detail by Rosenbauer et al. (1983) with modifications
for sampling CO2 (Chapter 2; Rosenbauer et al., 2005). The Au cell was surrounded by
pressure fluid (DI water) and attached to a titanium sampling tube and valve. The
combination of a flexible reaction cell, pressure fluid, and sampling tube allows for the
removal of fluid samples from the reaction vessel without changing the pressure
inside. The reaction cell was cleaned with a progression of reagent grade acids (nitric
acid, HCl, and 1% HF in nitric acid), and then baked at 430 °C under air for at least 12
hours in order to build a thick, inert oxide layer on the titanium components. Reactants
were placed inside the Au cell (DI water, 75-105 um olivine at a 50:1 solution:rock
ratio by mass, 0.5 M NaCl), which was sealed and placed inside the steel autoclave
and the annular space filled with DI water. After sealing the vessel, the autoclave was
placed inside a rocking furnace and heated to 60 °C (Rosenbauer et al., 1983). Once
this temperature was achieved, the pressure was increased to 100 bar by injection of
liquid CO2. The volume of injected CO2 as calculated at reaction conditions (60 °C,
100 bar) was 1/10 the volume of the aqueous phase in addition to sufficient CO2 to
saturate the aqueous phase at reaction conditions (Duan et al., 2006). See Chapter 2 for
a more detailed description of the experimental system.
Four experiments comprised this study. All four were conducted at the same
temperature and pressure conditions with the same reactants. Two of the experiments
were rocked at a rate of 8 rotations/minute in the furnaces, ensuring that they were
well mixed. The other two experiments were held stationary, with run durations of 2
and 19 days (Table 3.1).
69
Aqueous samples were removed from the reactor at regular intervals
throughout the duration of each experiment and were analyzed for alkalinity,
elemental concentration, and dissolved CO2 concentration. Complete aqueous
sampling procedures are presented in Chapter 2. Briefly, fluid compositions were
measured by inductively coupled plasma optical emission spectroscopy (ICP-OES),
alkalinity measurements were made via titration with 0.16 N sulfuric acid, and
dissolved CO2 concentrations were measured using coulometry (phosphoric acid
digestion followed by electrochemical analysis of released CO2 gas). The uncertainty
of ICP measurements is ±10%. Speciation calculations were completed using the
software package PHREEQC (Parkhurst and Appelo, 1999), and the included llnl.dat
database, which is derived from the EQ3/EQ6 database (Wolery et al., 1990).
Hydrogen concentrations in the CO2 phase were also measured for one
experiment to better constrain the redox state. The scCO2 phase was sampled in the
same way as the aqueous phase (using a three-way stopcock that allowed for purging
the exit tube of water and isolation of the sample), and was drawn into a
polypropylene syringe. The three-way stopcock was removed from the reactor and
attached to a needle, which was used to puncture the rubber stopper of a 5 times
evacuated (then backfilled with He) glass vial. The glass vial containing the CO2 and
H2 was stored in a freezer until analysis. Analysis was completed on a Agilent 6890
Series Gas Chromatograph (GC) with a pulse discharge detector and He as the carrier
gas. Peak area integration was performed using Agilent ChemStation software, and
quantification was completed using a four-point calibration curve encompassing H2
concentrations from 100 to 800 ppm. Uncertainty from the calibration was
significantly lower than the standard error for each data point.
Table 3.1: Conditions of four experiments.
Experiment name Duration Well mixed?
2-S 2 days No
2-R 2 days Yes
19-S 19 days No
19-R 19 days yes
70
3.2.2 Sample Preparation
This study used olivine from the Twin Sisters Dunite in Washington, with a
composition of Mg1.84Fe0.16SiO4 (Fo92). For a detailed description of this material see
Chapter 2. The mineral powder was prepared by washing with methanol 3 times and
sieving to obtain the size fraction 75-105 µm. The BET surface area was measured
with an ASAP 2020 (Micromeritics Instrument Corporation) using a 7-point surface
analysis and nitrogen as the adsorbent and found to be 5735 ± 97 cm2/g. Dissolution
rates were calculated using the BET surface area and Mg concentrations as a function
of time.
3.2.3 Transmission electron microscopy sample preparation and analysis
Transmission electron microscopy (TEM) samples were prepared using a
Helios NanoLab 600i DualBeam FIB/SEM by FEI, equipped with a Ga+ ion beam in
addition to an electron beam. Reacted olivine samples were attached to SEM mounts
and plasma-coated with gold before placement inside the FIB. Once in the
microscope, a flat area of at least 4x10 µm was chosen from which to cut the cross
section, and the area of interest was coated with 200 nm of Pt or C using electro
deposition. An additional 1 µm thick coating of Pt or C was then deposited using the
Ga+ ion beam. The combination of electro deposition followed by ion deposition
successfully shielded the mineral surface from the destructive ion beam. Once the
surface was protected, two large rectangles of material were removed by ion milling
from the surface, leaving a 3-4 µm thick slice between them. The sample was then
rotated and ion milling was used to make a U-shaped cut on the sides and bottom of
the slice. An Omniprobe AutoProbe 200 in situ sample lift-out system was used to
remove the slice from the olivine grain and transfer it to a copper Pelco® FIB lift-out
half grid for TEM samples. The slice was welded to the grid using Pt ion deposition,
then polished with ion milling to < 100 nm thick. The ion beam was set to an
accelerating voltage of 30 kV for most of the process, but dropped to 3-5 kV for the
final polishing, which reduced the thickness of the amorphized surface layer caused by
the ion beam to about 5nm (such that a 100 nm thick slice would have 90 nm of
71
crystalline material and 5nm of amorphous material on each side).
TEM analysis was completed on an FEI Tecnai TEM at 200 kV and a FEI
Titan TEM with spherical aberration image correction at 300 kV. Energy Dispersive
Spectroscopy (EDS) line scans were collected in scanning TEM (STEM) mode using
either an EDAX SUTW or Oxford SSD EDS detector. O-K, Mg-K, Si-K, Fe-K,
and Au-L elemental peaks were detected and the EDS peak analysis was performed
using TIA (Tecnai Imaging Analysis). The bulk olivine phase at the base of each
sample was used as an internal standard to extract sensitivity factors (k-factors) for the
peaks (except for Au), allowing for quantitative analysis of the composition across the
line scan. At least two deep scans (>700 nm) were performed per sample in order to
determine sensitivity factors. The maximum achievable resolution was 2 nm between
spots and each analyzed spot had a radius of less than 1 nm. A higher spot density
resulted in artificially low Mg/Si due to beam damage, as the electron beam appeared
to evaporate the Mg without detecting it.
Cross sections were cut from two different reacted mineral grains from
experiments 2-S and 19-R to ensure that the data obtained were representative and
uniform. In all cases, multiple TEM images were taken of each cross section and
multiple EDS line scans were performed. Each TEM image shown is representative of
the olivine surface from that sample, and each EDS linescan is the average of three
linescans from the same sample. TEM images and EDS line scans were taken from the
same cross section but not from the same location on each cross section. Visual
inspection demonstrated a high level of uniformity across each 10 µm cross-section.
3.2.4 TEM image processing: IFFT analysis
Some TEM images of the olivine samples have visible crystal lattice fringes,
but at lower zoom levels those fringes are difficult to see with the naked eye. An
Inverse Fast Fourier Transform (IFFT) analysis was performed on these images to
clarify which regions were crystalline and which were amorphous using ImageJ
software (Rasband, 1997). The first step is to take a Fourier Transform of the image of
interest, which produces another image in the frequency domain. Bright spots in the
Fourier Transform image are representative of periodic regions in the original image.
72
The frequencies of the spots were selected to pass the filter, while all other frequencies
were masked, and an inverse Fourier transform was performed. The result of this
procedure is an image similar to the original, but showing only the periodic regions as
a series of regular stripes. The IFFT image was used to draw false color on the original
image to clearly show which regions are crystalline. This process is diagramed in Fig.
3.1.
Figure 3.1: Steps of IFFT analysis. First, the TEM image is taken using the microscope (A).
The image is then transformed using a Fast Fourier Transform, which yields a second image
in the frequency domain (B). Bright spots on this image represent periodicity, so those
particular frequencies are selected and the rest are masked (C). An Inverse Fast Fourier
Transform is then applied to the masked frequency domain image to recreate the original
image showing only the regions of periodicity (D). The periodicity in (D) is distorted due to
aliasing (relatively low resolution in the image results in a high frequency periodicity from
the original image (A) appearing to have a lower frequency). This IFFT image was then
overlaid on the original TEM image (E) and false color was added to show periodic regions
and the gold surface coating (F).
73
3.3 Results
3.3.1 Solution Composition
Solution compositions as a function of time are plotted for four experiments in
Fig. 3.2 and shown in Table 3.2. In all four experiments, Mg increased throughout the
duration of each experiment (Fig. 3.2A). Experiment 19-R released the most Mg, three
times more than experiment 19-S. Similarly, experiment 2-R released twice as much
Mg as experiment 2-S. Silicon also
increased in all four experiments,
though experiment 19-R reached a
maximum concentration at 6 days (Fig.
3.2B). The other three experiments
showed increasing Si for the duration of
each experiment. Experiments 19-R and
19-S exceed amorphous silica saturation
at 2 days and 14 days, respectively, but
2-R and 2-S do not reach saturation.
Silica saturation was calculated using
experimental conditions, PHREEQC
(Parkhurst and Appelo, 1999), and the
included llnl.dat database. The pH of
each experiment remained below 6, so
the saturation remains constant. All
experiments remained undersaturated
with respect to magnesite. Comparison
of solution data from experiments
conducted under identical conditions (2-
R with 19-R and 2-S with 19-S) shows
the reproducibility over two days.
Table 3.2: Solution data from four experiments
time
(days)
[Mg]
mM
[Si]
mM
2-S 0.00 0.02 0.00
0.04 2.86 0.59
1.04 5.55 1.28
2.04 5.43 1.35
2-R 0.00 0.43 0.09
0.04 6.25 1.13
0.13 7.43 1.62
0.25 9.01 2.36
0.50 10.25 2.92
1.00 9.33 2.61
1.33 10.17 2.84
2.00 11.03 3.10
19-S 0.00 0.01 0.45
0.04 2.28 0.37
1.13 5.54 1.19
4.29 8.80 2.44
6.06 9.22 2.72
14.15 13.19 4.29
18.96 14.51 4.56
18.96 13.97 4.70
19-R 0.00 3.31 1.06
0.04 7.69 1.88
1.14 10.81 3.83
2.76 13.06 5.21
6.07 22.09 7.01
14.14 32.30 4.62
18.82 51.43 6.31
18.86 49.05 5.77
74
3.3.2 IFFT Analysis: Mapping Crystallinity
IFFT analyses were performed on a TEM image of reacted olivine from each
experiment in order to distinguish crystalline domains from amorphous domains (Fig.
3.3). False color was added for clarity. All olivine cross sections were coated with
gold (after reaction) to protect the surface during imaging and cross-section
preparation; the gold is clearly visible as dark material on the left side of each TEM
image. Also visible are the amorphous regions (uncolored) and crystalline regions
(green). Unreacted olivine (Fig. 3.3A) does not show an amorphous region at the
Figure 3.2: Mg (A) and Si (B) concentrations as a function of time for four
experiments. Amorphous silica saturation is indicated by the dashed line in (B).
75
mineral surface. Olivine lattice fringes extend 10 nm into the gold coating, which may
be a result of uneven surface topography on the olivine surface and the TEM image
represents a 2D projection of the 3D structure. After 2 days of reaction (Fig. 3.3B-C),
an amorphous layer of thickness 21-26 nm formed. After 19 days of reaction (Fig.
3.3D-E), the layer expanded to 39 nm (19-S) and 65 nm (19-R) thick. In three of the
images (Fig. 3.3B,D-E), some of the gold coating appears to have penetrated the
surface layer and deposited inside pores, as evidenced by the darker spots inside the
amorphous regions. Cross-sections from two different mineral grains from both
samples 19-R and 2-S were examined to determine the uniformity of the surface layer
and were found to be in agreement with each other. TEM images of larger areas (at
lower zoom levels) also show uniformity and provide confidence that images we show
are representative.
Figure 3.3: Representative TEM crystallinity maps of unreacted (A) and forsterite from
experiments 2-S (B), 2-R (C), 19-S (D) and 19-R (E). Green areas on the right of each image are
crystalline, yellow-colored regions on the left are a protective gold layer deposited on the mineral
surface (colored version online). Uncolored areas are amorphous. The scale bar in (E) applies to all
reacted forsterite (B-E).
76
3.3.3 Chemical compositions of surface layers
The chemical compositions of reacted olivine surfaces are shown in Fig. 3.4 as
Mg/Si and Fe/Si molar ratios plotted as a function of depth. The Mg/Si ratio for the
ureacted olivine is 1.84. After two days of reaction, mineral surfaces (solid-aqueous
interface) are depleted with respect to Mg and the Mg/Si ratio changes as a function of
depth (Fig. 3.4A-B). Olivine reacted for 2 days shows an increase in the Mg/Si ratio at
the mineral-fluid interface (<10 nm depth). After 19 days of reaction, olivine is also
Figure 3.4: TEM/EDS linescans showing Mg/Si ratio as a function of depth in forsterite reacted
without mixing (A) and with mixing (B). Forsterite from all four experiments had a Si-rich (Mg-
depleted) surface layer at least 20 nm thick. Linescans showing Fe/Si ratios are also shown for the
well-mixed experiments (C) and both have a peak of Fe associated with the Si-rich layers. Ratios
were calibrated using spectra from deep in the mineral.
77
depleted in Mg at the surface (Fig. 3.4A-B) but contains two distinct regions: one in
which the Mg/Si ratio is constant with depth (0 to 50 nm depth) and one in which the
Mg/Si ratio changes with depth (19-R only). Both well-mixed experiments (2-R and
19-R) contain a region of Mg depletion deeper than the surface layer, while the
unmixed experiments (2-S and 19-S) maintain the bulk olivine Mg/Si ratio up to the
surface layer.
Iron is also an important component of the reacting olivine (Fo92), and the
Fe/Si ratios for experiments 2-R and 19-R are shown in Fig. 3.4C. A striking feature
visible in both experiments is the spike in Fe/Si at the same depth as the steep decrease
in Mg/Si, indicating an accumulation of iron at this depth. Olivine from experiment 2-
R has a constant Fe/Si at depths >40 nm and the maximum Fe/Si occurs at 16 nm
depth (the minimum Mg/Si occurs at 10 nm depth) (Fig. 3.4C). Olivine from
experiment 19-R has a constant Fe/Si at depths >60 nm with a maximum Fe/Si at 47
nm (Mg/Si reaches a minimum at 40 nm depth) (Fig. 3.4C). At depths shallower than
38 nm, the Fe/Si ratio is near zero indicating that little to no iron is present in this
region. In the case of the stationary experiments, no Fe spike is visible. Instead, the
Fe/Mg ratio is constant with depth at the bulk ratio of 0.09 in reacted olivine from both
2-S and 19-S.
3.4 Discussion
3.4.1 Rates
Olivine dissolution rates were calculated from Mg vs. time data for all four
experiments. For the 19-day experiments, the rates were calculated using a linear
regression of the [Mg] data for days 3-19 and normalized using the BET surface area.
Rates from the 2-day experiments were calculated with the last two data points
because the rate varies with time over the initial 2 days, and the rate from 2-S was
calculated using Si data because the Mg data showed no change between 24 and 49
hours. Again, all were normalized with BET surface area. Fig. 3.5 shows olivine
dissolution rate as a function of Si-rich layer thickness for all four experiments.
Interestingly, the dissolution rate does not decrease with increasing surface layer
78
thickness. In fact, separating the rocking experiments and the stationary experiments
into two sets suggests that the dissolution rate increases with increasing layer
thickness. However, we do not suggest a correlation between layer thickness and
dissolution rate, but merely highlight that layer thickness does not seem to have a
direct effect on dissolution rate in our experiments. This result suggests that the Si-rich
surface layer does not inhibit dissolution over the thicknesses of Si-rich layers we
observe. In order for the surface layer not to inhibit dissolution, diffusion through the
layer must be much be faster than the dissolution rate, which could be achieved by
connected porosity in the surface layer allowing fluid to react at the fresh mineral
surface. The rate from experiment 19-R agrees (within uncertainty) with the rates
determined previously in Chapter 2, where state dissolution occurred over periods as
long as 94 days (Table 3.3).
Mass transport at the mineral-solution interface strongly affects the rate of
olivine dissolution. Two of the experiments in this study were conducted in rotating
furnaces and rocked 180° at 8 rpm, ensuring that the solution was well mixed (2-R and
19-R). The other two experiments were conducted in the same furnaces, but they were
Figure 3.5: Forsterite dissolution rate as a function of Si-rich layer thickness. The
rate does not decrease with increasing layer thickness; in fact, within the two
different data sets (rocking vs. stationary), show that the rates increase with Si-rich
layer thickness. Error bars represent 2-sigma uncertainty due to the few (2-5) data
points available for each rate calculation. No correlation between rate and layer
thickness is implied beyond the data presented here.
79
held stationary such that no mixing occurred beyond the pressurization step (2-S and
19-S). The olivine samples in the rocking experiments dissolved faster, as measured
by both changing Mg and Si concentrations. They also reacted to a greater extent over
the same time period than the olivine samples in the stationary experiments. These
results suggest that dissolution rates from the stationary experiments were limited by
transport rather than reaction kinetics. The dissolution rates from the rocking
experiments may also be transport limited, as we did not perform experiments to
determine the rate of rocking required to remove all transport limitations. For
example, Gislason and Oelkers (2003) performed dissolution experiments at a variety
of stirring rates and found that the dissolution rate increased up to a stir speed of 550
rpm, beyond which the rate remained constant and was determined to be surface-
controlled. However, given the lower dissolution rates in our stationary experiments
relative to the rocking experiments, we can say conclusively that those experiments
are transport limited, and that the surface-controlled reaction rate is faster than the
transport rate in a non-mixed system. The transport limitation could arise from one (or
more) of the three diffusive regions in the reactor system. Carbon dioxide must diffuse
across the water-scCO2 boundary before dissolving and hydrating, and this interface is
much smaller in a non-mixed system than in a mixed one (in the rocking experiments,
we expect CO2 bubbles in the aqueous phase which increases the surface area
available for dissolution). Another possible region for diffusive limitation in the
mineral-aqueous interface, which ions must dissolve across in order to reach the bulk
fluid. In the absence of mixing, concentration and pH gradients could exist near the
mineral surface. Transport through the aqueous phase could also limit the reaction,
and an estimation of the diffusion rate is possible using the characteristic length.
Table 3.3: Olivine dissolution rates from four experiments
Experiment Rate (mol cm-2
s-1
)
1 (Chapter 2) 9.02 ± 0.80 × 10-14
2 (Chapter 2) 8.64 ± 0.43 × 10-14
19-R 1.03 ± 0.44 × 10-13
19-S 2.10 ± 0.16 × 10-14
80
The characteristic diffusion length, Lc, is defined as the length over which
diffusion occurs with diffusivity D during reaction time, t (Equation 1). The diffusion
length calculation is applicable to 19-S, in which diffusive transport through the
aqueous phase dominates, but not 19-R because mixing causes convective transport to
dominate over diffusion.
DtLc (1)
Using a diffusion coefficient of 7.1x10-6
, a typical diffusion coefficient for Mg in
water at 25 °C (Haynes et al., 2014), and reaction time of 19 days, the diffusion length
is 3.4 cm. In other words, a molecule initially at x=0 would be, on average, at x=3.4
cm after 19 days of diffusion. In our experiments, the height of the aqueous phase
separating the mineral from the CO2 is 3-6 cm, which is of the same order of
magnitude as the characteristic diffusion length. The diffusion length applies to both
Mg and Si (originating at the mineral surface) and protons (originating at the aqueous-
CO2 interface). As a result, concentration gradients of all three reacting species could
exist in the aqueous phase, even after 19 days of reaction, and diffusion is slow
enough to potentially limit the reaction rate in the absence of convective mixing.
A larger point regarding transport is that depending on injection rates and host
rock flow characteristics, some regions of an in situ mineralization project will likely
be transport limited. Based on our results, we anticipate that this limitation will not be
directly due to the formation of surface layers on reactive mineral surfaces, but rather
diffusion of ions to and/or from the dissolving mineral surface. On the pore scale, Li et
al. (2008) found that a discrepancy between well-mixed laboratory rates and natural
(simulated) rates only exists when there are both incomplete molecular diffusion and
similar reaction and transport rates, which suggests that individual pores function as
well-mixed reactors under most natural conditions. Our results are in contrast to Li et
al. (2008), but are also derived from experiments on a larger scale. Based on our data,
the kinetics of the surface-controlled dissolution reaction are less relevant than the
transport rates, and we might expect to see dissolution rates up to an order of
magnitude slower than we observe in a well-mixed laboratory reactor. In an ex situ
sequestration project, the reaction could be engineered to remove the transport
81
limitation.
Experiment 19-S was not mixed, thus may best represent the reaction rate and
mineral surface characteristics that one would expect to observe in the subsurface.
Reacted olivine from 19-S also showed the steepest chemical gradient in the surface
layer at the transition between the Si-rich surface layer and the bulk mineral. This
observation is in agreement with published studies including TEM images of natural
samples, which tend to show a very steep, step-like chemical gradient between the
surface layer and the mineral (e.g. Hellmann et al., 2013, 2012). Consequently, we
hypothesize that the absence of a shallow gradient (like that seen in 19-R) is a result of
the transport limitation. More specifically, a scarcity of protons at the mineral surface
to exchange with Mg cations could result in a slower exchange rate, and a slower
exchange rate (diffusion rate) would create a steeper chemical gradient at the mineral-
surface layer interface. More discussion on this topic is presented in Section 4.3.
3.4.2 Surface layer characterization
We observed three distinct types of surface layers from the TEM/EDS results
(Fig. 3.6). The first is 35-40 nm thick and is visible only in the 19-day experiments,
19-R and 19-S. This layer is both amorphous (Fig. 3.3D-E) and Si-rich, but the Mg/Si
ratio does not change with depth within the layer (Fig. 3.4A-B). Given that we only
observe this type of layer on olivine grains exposed to solutions supersaturated with
respect to amorphous silica, and also because of the near complete depletion of Mg
(Mg/Si = 0.1-0.3), we infer that this layer is formed by precipitation of amorphous
silica. This “precipitated layer” appears to have porosity, as the gold that was
deposited on the reacted olivine surfaces before TEM sample preparation penetrated
the surface layer and is visible as dark spots in the amorphous layer on TEM images
from 19-S and 19-R (Fig. 3.2D-E). We hypothesize that the porosity is interconnected
and does not hinder dissolution (similar to Daval et al., 2013, 2009; Hellmann et al.,
2012), because the olivine grains continue to dissolve underneath the coating at a
steady rate for up to 19 days in this study (Fig. 3.2) and up to several months in
previous work (Chapter 2). The porosity can be quantified by comparing the darkened
areas (gold) to the total area of material and the result is a porosity of up to 40%.
82
However, it's important to note that this calculation is based on a 2D projection of a
3D cross section, and there's no way to calculate the actual porosity without assuming
an average shape and size of the pores. Thus, we present only an upper bound. Further
work is currently underway to better understand the formation mechanism of this
alteration layer.
The second type of Si-rich layer sits at the mineral-fluid interface of 2-R and 2-
S (Fig. 3.6). This layer is characterized by a changing Mg/Si ratio with depth that is
about 40 nm thick. At the solid-fluid interface (<10nm), the Mg/Si ratio decreases
with depth, indicating accumulation of Mg at this interface. The Mg/Si ratio reaches a
minimum in olivine for 2-S and 2-R of 0.53 at 12 nm depth and 0.28 at 10 nm depth,
respectively. Below this layer (10-40 nm deep), the Mg/Si ratio increases until it
reaches the stoichiometric value, 1.84. Interestingly, only the top 21-26 nm of the
surface layer is amorphous (Fig. 3.3B-C), which is only about half the total thickness
Figure 3.6: Cross sectional diagram of the reacted mineral surface, specifically for
experiment 19-R. The three different Si-rich layers and their defining characteristics
are visible from top down: Precipitated layer, active layer, and Mg-depleted forsterite.
83
of the Si-rich layer. We define the 20 nm amorphous region with changing Mg/Si as
the “active layer” because it appears to be actively exchanging Mg. The characteristics
of the active layer are also visible in olivine from 19-R at a depth of 40-60 nm (Fig.
3.4B). In two experiments (2-R and 19-R), this layer also contains an increased
amount of Fe relative to the unreacted mineral and the surrounding alteration layers.
Iron in the active layer may exist in the ferric state and stabilize the layer similar to
findings from Saldi et al. (2013) and Sissmann et al. (2013). However, measurement
of hydrogen gas from the scCO2 phase of an experiment with a chemical evolution
similar to that of sample 19-S (fugacity of H2 ranged from 10-4
to 10-3
bars from 3-14
days of reaction) showed that Fe2+
concentrations were several orders of magnitude
greater than Fe3+
.
The third type of Si-rich layer occurs below the active layer (Fig. 3.6). This
layer has chemical characteristics similar to those of the active layer, but unlike the
active layer, is crystalline. We call this “Mg-depleted olivine”, which is visible in
TEM images of samples 2-S, 2-R, and 19-R (Fig. 3.4A-B). This layer extends as
deeply as 150 nm (in 19-R). For samples 2-R and 19-R we also observe maxima in the
Mg/Si ratios in the Mg-depleted olivine regions (Fig. 3.3B) at 40-50 and 70-80 nm
depth, respectively. These peaks suggest an accumulation of Mg and are not visible in
the stationary experiments.
We hypothesize that both the active layer and Mg-depleted olivine are formed
via preferential removal of Mg from the mineral surface and replacement by protons
from solution. In the active layer, protons are present to maintain charge balance, but
the remaining SiO4 tetrahedra polymerize to form a network through which Mg
cations must diffuse to reach the mineral-solution interface. This mechanism has been
proposed previously (Bearat et al., 2006; Daval et al., 2011; Oelkers, 2001; Pokrovsky
and Schott, 2000a, 2000b). In the Mg-depleted olivine regions, protons may form
hydroxyl-like bonds and maintain the olivine crystal structure. This type of Si-rich
layer has not, to our knowledge, been proposed or observed before in olivine.
However, there has been previous work on isomorphic substitution of (O4H4)4-
for
(SiO4)4-
in a variety of minerals (Aines and Rossman, 1984). Foreman (1968)
84
conducted a neutron and x-ray diffraction study on hydrogrossular and found
substitution of four protons for silicon with only slight changes to the structural
parameters; his work was later verified by computational simulations (Wright et al.,
1994) and other experiments (Aines and Rossman 1983). Hydrogen bonding occurs
within the (O4H4)4-
tetrahedra in hydrogrossular (Harmon et al., 1982). Like
hydrogrossular, olivine is a nesosilicate with isolated SiO44-
tetrahedra connected by
interstitial metal cations. One might expect that removing a silicon ion or a metal
cation would cause the structure to collapse on itself; however, at least in the case of
Si, the combination of hydroxyl and hydrogen bonding maintains the crystal structure.
We have no direct evidence that protons substitute for Mg in olivine from our
experiments, but we suggest this mechanism as a way to explain our data that show
Mg-depleted, crystalline regions far from the mineral-fluid interface (up to 150 nm
deep).
Further evidence for the plausibility of proton substitution for Mg in the M1
and M2 sites exists in the hydrous olivine literature. Olivine is generally an anhydrous
mineral (Khisina et al., 2001), but it has been observed to hold up to 0.12 wt% water
(Kohlstedt et al., 1996). Hydrous regions in olivine are known to exist in three
different forms: lamellar defects, small (10 nm) inclusions, and large (100 nm)
inclusions (Khisina et al., 2001). Khisina et al. (2001) measured the concentrations of
Mg, Fe, and Si across the inclusions and found that the Si concentration is the same as
that of the host olivine, so the regions are depleted in Mg and Fe similar to the Mg-
depleted layer in olivine from our experiments. Incorporation of protons and hydroxyl
ions into olivine should affect the structure around the inclusions/defects as well as
bond strengths (Churakov et al., 2003; Haiber et al., 1997) and may affect the
dissolution rate as well (Rimstidt et al., 2012). For example, protons are more
electronegative than Mg cations (2.2 vs. 1.3 Pauling Units, respectively), and we
would expect them to pull electron density away from other bonds in the crystal,
effectively weakening those bonds.
Based on our TEM/EDS data and crystallinity maps, in the absence of
transport limitations, we propose that all three types of Si-rich layers exist on olivine
85
surfaces that have reacted with water and CO2 for times sufficient for the bulk solution
to reach saturation with respect to amorphous silica. The sample from experiment 19-
R fits these criteria, as it exceeded silica saturation and was continuously rocked,
ensuring a well-mixed aqueous phase. Over shorter timescales (or lower reaction
extent), amorphous silica saturation is not reached and the precipitated layer does not
form (e.g., samples 2-S and 2-R). The formation of a precipitated layer does not
depend on transport beyond the effect of transport on overall reaction rate; i.e.
stationary experiments proceeded more slowly and thus took longer to reach
amorphous silica saturation. One might expect a non-mixed experiment to precipitate
amorphous silica even if the bulk silica concentration is below saturation as there
could be locally saturated regions near the olivine surface; however, we did not
observe this.
3.4.3 Diffusion Modeling
In order to investigate whether or not the measured concentration gradients
are the result of diffusion, we used a series of diffusion equations from Lasaga (1979)
and Hellmann et al. (2012). All modeling was done assuming steady state in the
mineral and alteration layer (dc/dt = 0) and thus was only performed on data from the
19-day experiments because they exhibited constant Mg release with time after 3 days
(steady-state dissolution). Modeling was based on Fick’s Second Law for diffusion
with a term added to account for the retreating reaction front as the mineral dissolves
(Equation 2), where a is the velocity at which the mineral surface retreats, c is the
normalized concentration of magnesium (concentration at depth x divided by the bulk
concentration), D* is an effective diffusion coefficient, and x is depth (Doremus, 1975;
Hellmann et al., 2012).
(2)
Equation 2 is a second-order ordinary differential equation, and since the diffusion
coefficient depends on concentration, the ODE must be solved in two steps. The first
integration results in the following expression:
0
x
ca
x
cD
xt
c *
86
constant* acdx
dcD (3)
Deep in the mineral (at large x), the change in cation concentration with depth is zero
and the normalized cation concentration is 1 (dc/dx =0, c=1). This constitutes the
boundary condition to solve for the integration constant in Equation 3, which is equal
to a, and reduces the second order ODE in Equation 1 to the first order ODE in
Equation 4.
*
)1(
D
ca
dx
dc (4)
The effective diffusion coefficient in this case is derived from Equation 15 in
Lasaga (1979), which describes interdiffusion of two ions in a silicate matrix. The
expression assumes that the silicate matrix is stable and that oxygen and silicon
diffusivity are negligible (Lasaga, 1979). In our system, the silicate matrix is not stable
and does dissolve; however, the presence of crystalline regions that are Mg-depleted
suggests that the destruction of Si-O bonds and diffusion of Si and O are slower than
the diffusion of Mg and H+. The expression for the diffusion coefficient is based on a
mean-field model, i.e. the transient effect of individual moving ions on other ions is
not included. Finally, the model assumes a fixed volume, because moving ions have a
very small volume relative to the volume of the silicate network (Lasaga, 1979). In the
case of olivine dissolution, isolated SiO4 tetrahedra must rearrange to form new Si-O-
Si bonds and the associated volume change is about 50%. However, we use the model
only to study the steady-state case, when Mg release is constant with time. Our TEM
evidence suggests that the active layer thickness does not change significantly with
time after the initial 2 days of reaction (21 nm in 2-R and 25 nm in 19-R) and thus the
volume change is negligible within the amorphous layer. The effective diffusion
coefficient D*
is presented in Equation 5, and is a function of the tracer diffusion
coefficients of Mg (DMg) and H (DH), the concentration of Mg (c) (Lasaga, 1979) and
a concentration-dependent structural factor, α, following Hellmann et al. (2012) (see
Appendix 4 for derivation).
ccDcD
cDDD
HMg
HMg
11
)1(4
)13(* (5)
87
The concentration (and thus depth) dependent structural factor, α was added by
Hellmann et al. (2012) to account for physical changes in the altered layers, i.e. the
areas with low cation concentration are also porous and accordingly are expected to
have faster diffusion and higher diffusivity. Hellmann et al. (2012) used an α value of
-0.9 to generate the largest effect of D* on concentration, such that D* is constant for
c>0.1 and then increases rapidly at c<0.1. Using Equation 5, values of α greater than
0.3 resulted in a non-monotonic increase of D* with decreasing concentration, such
that a minimum D* was achieved between c=1 and c=0, when the minimum should
exist at c=1 since this is the region of unaltered olivine (Fig. 3.7A). For the remaining
calculations, we set α=-0.3 because this was the smallest value for which D* increased
monotonically with decreasing concentration.
The dependence of D* on concentration necessitated the use of a numerical
solver to obtain the concentration gradient from Equation 4. We used Matlab and the
function “ode45” for numerically solving non-stiff differential equations. We used
Equations 1-4 to model only the "active" layer of reacted olivine grains. The model
was developed from studies of sodium diffusing in glass materials and thus is not
applicable to crystalline regions (in which diffusion occurs via a different mechanism).
Also, we assume that diffusion through the precipitated surface layer is fast relative to
solid-state diffusion, as it likely consists of cations diffusing through water in pores.
Diffusion coefficients of ions in solid phases vary greatly as a function of
temperature, diffusing species, and the matrix, and estimation is not accurate. For
example, the diffusion coefficient of hydrogen in SiO2 at 200 °C is 6.5E-10 cm2/s
(Cussler, 1997), of Na in rhyolite glass at 60 °C is 6.7E-16 cm2/s (Jambon, 1982) and
the coefficient for interdiffusion of Mg and Fe in olivine at 200 °C is 1.78E-13 cm2/s
(Buening and Buseck, 1973). Further insight regarding the values of tracer diffusion
coefficients can be obtained by observing the limits of Equation 5 as the magnesium
concentration, c, goes to zero and to one. When c=0, the equation reduces to
D*=0.7DMg, a condition that exists in the most altered (Mg-poor) region, and we
follow Hellmann et al. (2012) in assuming that diffusion in these regions is relatively
fast. Thus, we estimate that a reasonable interdiffusion coefficient in the altered region
88
is 10-12
to 10-16
cm2/sec. Setting c=1 gives us the effective diffusion coefficient deep
in the mineral, and the expression simplifies to D*=DH. We used 1E-16 cm
2/s value as
a starting point, following Hellmann et al. (2012), and then assessed the diffusivities
Figure 3.7: Simple diffusion model in which the diffusion coefficient, D*, is a function of
Mg concentration in addition to diffusivity of protons (DH) and of Mg (DMg), plus a
structural factor (α) (A). Decreasing from α 0 to -0.3 results in flattening the profile, and
reducing it further to -0.9 creates a minimum value of D* at c≈0.1. Changing DMg results in a
shift of the concentration gradient curve (B) and changing DH affects both the steepness and
depth of the curve (C). Data from the two 19 day experiments can be fit with this model and
different values of DH (D). Vertical bars indicate the transition from amorphous to crystalline
material, and the model was shifted to the right to align with the “active” layer.
89
required to fit the data. Several glass corrosion studies have verified that DMg>>DH
(e.g., Smets and Lommen, 1982). Also, while the values chosen to fit our data are
larger than diffusion coefficients from tracer diffusion studies, evidence suggests that
in altered, hydrated glass layers ionic transport is faster than in dry glass (Doremus,
1983).
Figs. 3.7B-C show concentration gradients calculated from Equation 5 with a
range of DMg and DH values and a = 1.5E-10 cm/s (calculated from the dissolution rate
of 19-R). In general, the slope is controlled by DH (a lower DH results in a steeper
slope) and the depth of the starting point of the slope is controlled both by DMg and
DH. Depending on which values of DMg and DH are selected, nearly any concentration
gradient can be generated. Thus, one can even produce a very steep, step-like gradient
using this equation, which removes the argument that a steep concentration gradient
precludes the existence of a leached layer. In fact, the slower the diffusion of protons
relative to Mg, the steeper the gradient becomes. Fast Mg diffusion relative to H
diffusion means that the removal of Mg from the mineral is limited by the availability
of protons with which to exchange.
Fig. 3.7D shows calculated concentration gradients that were fit to our EDS
data. The data were shifted and scaled using the bulk Mg/Si ratio of 1.84 such that the
precipitated layer is defined by c = 0 and the active layer-mineral interface is defined
by c = 1. The crystalline-amorphous boundaries for samples19-S (green) and 19-R
(pink) are indicated by vertical bars. For both 19-S and 19-R, D* was calculated using
DMg = 1E-14 cm2/s. The two fits differ in the values of a (velocity of the reaction
front), which were calculated using steady-state Mg release rates to be 1.5E-10 cm/s
(19-R) and 2.7E-11 cm/s (19-S), because the dissolution rate of sample 19-S was
significantly slower than that of sample 19-R. In order to fit the EDS data, different
values of DH were also used for the two data sets (9E-17 cm2/s for 19-R and 5E-18
cm2/s for 19-S). In summary, two Mg concentration gradients in the Si-rich,
amorphous surface layer of altered olivine can be successfully fit using steady-state
diffusion alone, assuming that the diffusivity of the highly altered regions is 2-4 orders
of magnitude greater than that of the minimally altered regions. Even very sharp, step-
90
like gradients can be fit using this approach.
In order to fit our data, the model requires DH to be about an order of
magnitude lower for sample 19-S than for sample 19-R. The implication is that
protons in the stationary experiment diffuse more slowly inside the altered layer than
in the rocking experiment. We are not able to measure concentration gradients in the
fluid at the mineral-fluid interface, and as a result, the diffusion coefficients we use to
fit our data are effective diffusion coefficients that incorporate the effect of any
concentration gradients in the fluid. Thus, slower proton diffusion in the stationary
experiment could be an effect of a lack of protons at the fluid-mineral interface, which
would result in a smaller concentration gradient in the altered layer. Effectively, slow
transport of protons from the bulk solution to the mineral could, in theory, result in
slower proton transport inside the altered layer and thus a smaller effective diffusion
coefficient. In the well-mixed experiments, convective transport could increase the
concentration of protons at the mineral surface, and consequently create a strong
chemical potential for diffusion into the altered layer.
3.5 Conclusions
Dissolution of olivine in a batch reactor under conditions relevant for
subsurface carbon storage results in fluid and mineral phases that were carefully
examined in order to improve our understanding of the mechanism(s) and rate-limiting
steps of the dissolution reaction. Analysis of the fluid phase revealed that amorphous
silica saturation was reached between 1 and 3 days in the well-mixed, rocking
experiment and 7 and 14 days in the unmixed, stationary experiment. By 19 days,
olivine from both experiments (mixed and unmixed) appeared to be dissolving at
steady state because the Mg release rates were constant with time between 3 and 19
days. Mineral phase analysis of thin cross sections prepared with a FIB and analyzed
by TEM yielded information about the chemical composition as a function of depth in
the mineral surface as well as the location of amorphous and crystalline regions.
Combination of the fluid and mineral phase data creates a more complete picture of
the processes that occur during olivine dissolution.
As expected, olivine dissolution proceeds faster in a well-mixed reactor than in
91
an unmixed one, and in an unmixed situation, the reaction appears to be transport
limited. Thus, in a subsurface storage project, one might expect silicate dissolution
rates to be significantly slower than the surface-controlled reaction rates measured in
laboratories. However, the thickness of the Si-rich layer does not allow prediction of
the dissolution rate. In fact, the experiment with the highest measured dissolution rate
(19-R) also had the thickest Si-rich layer (65 nm). However, we observed the
formation of a Si-rich layer after as little as 2 days of reaction, before SiO2 saturation
was reached in the bulk aqueous phase. The formation of this layer coincides with the
previously observed two order of magnitude decrease in olivine dissolution rates over
the first several days of reaction (Chapter 2; Daval et al., 2011; Pokrovsky and Schott,
2000b).
We observed several different types of Si-rich surface layers with different
characteristics. The “precipitated” layer exists at the mineral-fluid interface on olivine
from the 19-day experiments, was amorphous, and contained little to no Mg and Fe.
The precipitated layer may form via SiO2 precipitation from the bulk solution and
contains pores visible in TEM images. Further evidence for the
dissolution/reprecipitation mechanism will be presented in part 3 of this series
(Chapter 4). The precipitated layer was not present on olivine reacted for 2 days.
Beneath the precipitated layer (19 day experiments) or in the case of the 2-day
experiments, we observed a second type of Si-rich layer at the mineral-fluid interface,
also amorphous but with increasing Mg/Si ratios with depth. We call this layer the
“active” layer and suggest that it forms via a leaching mechanism in which Mg-O
bonds are broken and Mg2+
ions diffuses to the surface. Finally, the Mg-depleted
olivine is below the active layer and observed on olivine from three of the four
experiments (2-S, 2-R, and 19-R). This region is crystalline but depleted in Mg
relative to Si, suggesting that it is a hydrous olivine.
Our data indicate that the active layer forms in the first hours to days of
reaction time and the precipitated layer forms later after the bulk solution reaches
saturation with respect to amorphous silica. The correlation between active layer
formation and dissolution rate decrease in the first hours to days of reaction suggests
92
that the active layer may partially passivate the olivine surface. On the other hand,
olivine dissolution rates do not decrease with increasing precipitated layer thicknesses,
implying that the precipitated layer does not further passivate the surface.
Finally, we used a simple diffusion model (Lasaga, 1979; Hellmann et al.,
2012) to fit the Mg/Si ratios as functions of depth in the altered surface layers for two
experiments (19-S and 19-R). The model was originally developed for interdiffusion
of protons and cations in an amorphous matrix and assumes steady-state dissolution
(Doremus, 1975). We fit a curve to our data using reasonable diffusivity coefficients
and showed that many different gradients could be generated, including near step-
functions, depending on the values assumed for the diffusion coefficients. Recent
studies (e.g., Hellmann et al., 2012) have used the presence of step-like changes in
concentration to support the dissolution/reprecipitation mechanism of Si-rich layer
formation, but this justification may not be correct. While our evidence does support
silica precipitation during silicate mineral dissolution at t>2 days, our data and fit
using the simple diffusion model suggests that in the case of olivine, solid-state
diffusion through an active layer of up to 25 nm facilitates the creation and ongoing
presence of a Si-rich leached layer.
The combination of TEM/EDS and ICP data presented in this study allows us
to hypothesize about Si-rich surface layer formation mechanisms, but it does not
directly probe the mechanisms themselves. Additional work focusing on the uptake of
an isotopic Si spike from solution into the surface layer is underway (Chapter 4) and
will further illuminate the processes controlling olivine dissolution. Also important is
the need to broaden the scope of this study by applying the data and conclusions to
build a model that will allow for the prediction of olivine dissolution rates under a
range of conditions and over longer timescales than are practical to investigate in the
laboratory.
93
3.6 Acknowledgments
We acknowledge the Global Climate and Energy Project (GCEP-48942) at
Stanford University for funding the work presented here. We also would like to
acknowledge the staff of the Stanford Nanocharacterization Laboratory, where the FIB
and TEM analyses were performed, particularly Richard Chin for technical assistance.
94
3.7 References
Aines, R.D., Rossman, G.R., 1984. The hydrous component in garnets : pyralspites.
Am. Mineral. 69, 1116–1126.
Bearat, H., Mckelvy, M.J., Chizmeshya, A.V.G., Gormley, D., Nunez, R., Carpenter,
R.W., Squires, K., Wolf, G.H., 2006. Carbon Sequestration via Aqueous
Olivine Mineral Carbonation : Role of Passivating Layer Formation. Environ.
Sci. Technol. 40, 4802–4808.
Buening, D.K., Buseck, P.R., 1973. Fe-Mg lattice diffusion in olivine. J. Geophys.
Res. 78, 6852–6862.
Churakov, S. V., Khisina, N.R., Urusov, V.S., Wirth, R., 2003. First-principles study
of (MgH2SiO4)•n(Mg2SiO4) hydrous olivine structures. I. Crystal structure
modelling of hydrous olivine Hy-2a(MgH2SiO4)•3(Mg2SiO4). Phys. Chem.
Miner. 30, 1–11.
Cussler, E., 1997. Diffusion: Mass Transfer in Fluid Systems, 2nd ed. Cambridge
University Press, New York.
Daval, D., Hellmann, R., Saldi, G.D., Wirth, R., Knauss, K.G., 2013. Linking nm-
scale measurements of the anisotropy of silicate surface reactivity to
macroscopic dissolution rate laws: New insights based on diopside. Geochim.
Cosmochim. Acta 107, 121–134.
Daval, D., Martinez, I., Guigner, J.-M., Hellmann, R., Corvisier, J., Findling, N.,
Dominici, C., Goeffe, B., Guyot, F., 2009. Mechanism of wollastonite
carbonation deduced from micro- to nanometer length scale observations. Am.
Mineral. 94, 1707–1726.
Daval, D., Sissmann, O., Menguy, N., Saldi, G.D., Guyot, F., Martinez, I., Corvisier,
J., Garcia, B., Machouk, I., Knauss, K.G., Hellmann, R., 2011. Influence of
amorphous silica layer formation on the dissolution rate of olivine at 90°C and
elevated pCO2. Chem. Geol. 284, 193–209.
Doremus, R.H., 1975. Interdiffusion of Hydrogen and Alkali Ions in a Glass Surface.
J. Non. Cryst. Solids 19, 137–144.
Doremus, R.H., 1983. Diffusion-Controlled Reaction of Water with Glass. J. Non.
Cryst. Solids 55, 143–147.
Duan, Z., Sun, R., Zhu, C., Chou, I.-M., 2006. An improved model for the calculation
of CO2 solubility in aqueous solutions containing Na+, K+, Ca2+, Mg2+, Cl−,
and SO42−. Mar. Chem. 98, 131–139.
95
Foreman, D.W., 1968. Neutron and X-Ray Diffraction Study of Ca3Al2(O4D4)3, a
Garnetoid. J. Chem. Phys. 48, 3037.
Gislason, S.R., Oelkers, E.H., 2003. Mechanism, rates, and consequences of basaltic
glass dissolution: II. An experimental study of the dissolution rates of basaltic
glass as a function of pH and temperature. Geochim. Cosmochim. Acta 67,
3817–3832.
Haiber, M., Ballone, P., Parrinello, M., 1997. Structure and dynamics of protonated
Mg2SiO4: An ab-initio molecular dynamics study. Am. Mineral. 82, 913–922.
Hänchen, M., Prigiobbe, V., Storti, G., Seward, T.M., Mazzotti, M., 2006. Dissolution
kinetics of fosteritic olivine at 90–150°C including effects of the presence of
CO2. Geochim. Cosmochim. Acta 70, 4403–4416.
Harmon, K.M., Gabriele, J.M., Nuttal, A.S., 1982. Part 14. Hydrogen bonding in the
tetrahedral 04H44- cluster in hydrogrossular. J. Mol. Struct. 82, 213–219.
Haynes, W., Bruno, T.J., Lide, D.R. (Eds.), 2014. Handbook of Chemistry and
Physics, 94th ed. CRC Press, Boca Raton, FL.
Hellmann, R., Daval, D., Wirth, R., 2013. Formation of Amorphous Silica Surface
Layers by Dissolution-Reprecipitaton During Chemical Weathering:
Implications for CO2 Uptake. Procedia Earth Planet. Sci. 7, 346–349.
Hellmann, R., Eggleston, C., Hochella, Jr., M., Crerar, D., 1990. The formation of
leached layers on albite surfaces during dissolution under hydrothermal
conditions. Geochim. Cosmochim. Acta 54, 1267–1281.
Hellmann, R., Wirth, R., Daval, D., Barnes, J.-P., Penisson, J.-M., Tisserand, D.,
Epicier, T., Florin, B., Hervig, R.L., 2012. Unifying natural and laboratory
chemical weathering with interfacial dissolution–reprecipitation: A study based
on the nanometer-scale chemistry of fluid–silicate interfaces. Chem. Geol.
294-295, 203–216.
Jambon, A., 1982. Tracer Diffusion in Granitic Melts: Experimental Results for Na, K,
Rb, Cs, Ca, Sr, Ba, Ce, Eu to 1300C and a Model of Calculation. J. Geophys.
Res. 87, 10797–10810.
Khisina, N.R., Wirth, R., Andrut, M., Ukhanov, a. V., 2001. Extrinsic and intrinsic
mode of hydrogen occurrence in natural olivines: FTIR and TEM
investigation. Phys. Chem. Miner. 28, 291–301.
96
King, H.E., Plümper, O., Geisler, T., Putnis, A., 2011. Experimental investigations
into the silicification of olivine: Implications for the reaction mechanism and
acid neutralization. Am. Mineral. 96, 1503–1511.
King, H.E., Plümper, O., Putnis, A., 2010. Effect of secondary phase formation on
the carbonation of olivine. Environ. Sci. Technol. 44, 6503–6509.
Kohlstedt, D.L., Keppler, H., Rubie, D.C., 1996. Solubility of water in the α, β and γ
phases of (Mg,Fe)2SiO4. Contrib. to Mineral. Petrol. 123, 345–357.
Lasaga, A.C., 1979. Multicomponent exchange and diffusion in silicates. Geochim.
Cosmochim. Acta 43, 455–469.
Li, L., Steefel, C.I., Yang, L., 2008. Scale dependence of mineral dissolution rates
within single pores and fractures. Geochim. Cosmochim. Acta 72, 360–377.
doi:10.1016/j.gca.2007.10.027
Mazzotti, M., Zurich, E.T.H., Storti, G., Baciocchi, R., Vergata, T., 2010. Diss . ETH
no . 18932 Valentina Prigiobbe Zurich 2010.
Oelkers, E.H., 2001. An experimental study of forsterite dissoluiton rates as a
function of temperature and aqueous Mg and Si concentrations. Chem. Geol.
175, 485–494.
Oelkers, E.H., Gislason, S.R., Matter, J., 2008. Mineral carbonation of CO2. Elements
4, 333–337. doi:10.2113/gselements.4.5.333
Parkhurst, D.L., Appelo, C.A.J., 1999. User’s guide to PHREEQC ( Version 2 )— A
computer program for speciation, batch-reaction, one-dimentional transport,
and inverse geochemical calculations, US Geological Survey Water-Resources
Investigations Report 99-4259. Denver, Colorado.
Pokrovsky, O.S., Schott, J., 2000a. Kinetics and mechanism of forsterite dissolution
at 25°C and pH from 1 to 12. Geochim. Cosmochim. Acta 64, 3313–3325.
Pokrovsky, O.S., Schott, J., 2000b. Forsterite surface composition in aqueous
solutions : A combined potentiometric , electrokinetic , and spectroscopic
approach. Geochim. Cosmochim. Acta 64, 3299 –3312.
Prigiobbe, V., Hänchen, M., Werner, M., Baciocchi, R., Mazzotti, M., 2009. Mineral
carbonation process for CO2 sequestration. Energy Procedia 1, 4885–4890.
Rasband, W., n.d. ImageJ [WWW Document]. U. S. Natl. Institutes Heal. Bethesda,
Maryland, USA. URL http://imagej.nih.gov/ij/
97
Rimstidt, J.D., Brantley, S.L., Olsen, A.A., 2012. Systematic review of forsterite
dissolution rate data. Geochim. Cosmochim. Acta 99, 159–178.
Rosenbauer, R.J., Bischoff, J.L., Fadtke, A.S., 1983. Hydrothermal alteration of
graywacke and basalt by 4 molal NaCl. Econ. Geol. 78, 1701–1710.
Rosenbauer, R.J., Koksalan, T., Palandri, J.L., 2005. Experimental investigation of
CO2–brine–rock interactions at elevated temperature and pressure:
Implications for CO2 sequestration in deep-saline aquifers. Fuel Process.
Technol. 86, 1581–1597.
Ruiz-Agudo, E., Putnis, C. V, Rodriguez-Navarro, C., Putnis, A., 2012. Mechanism
of leached layer formation during chemical weathering of silicate minerals.
Geology.
Saldi, G.D., Daval, D., Morvan, G., Knauss, K.G., 2013. The role of Fe and redox
conditions in olivine carbonation rates: An experimental study of the rate
limiting reactions at 90 and 150°C in open and closed systems. Geochim.
Cosmochim. Acta 118, 157–183.
Schulze, R.K., Hill, M.A., Field, R.D., Papin, P.A., Hanrahan, R.J., Byler, D.D.,
2004. Characterization of carbonated serpentine using XPS and TEM. Energy
Convers. Manag. 45, 3169–3179.
Sissmann, O., Daval, D., Brunet, F., Guyot, F., Verlaguet, A., Pinquier, Y., Findling,
N., Martinez, I., 2013. The deleterious effect of secondary phases on olivine
carbonation yield: Insight from time-resolved aqueous-fluid sampling and FIB-
TEM characterization. Chem. Geol. 357, 186–202.
Smets, B., Lommen, T., 1982. Leaching of sodium aluminosilicate glasses studied by
secondary ion mass spectrometry. Phys. Chem. Glas. 23, 83–87.
Wogelius, R.A., Walther, J. V, 1991. Olivine dissolution at 25°C : Effects of pH ,
CO2 , and organic acids. Geochim. Cosmochim. Acta 55, 943–954.
Wogelius, R.A., Walther, J. V, 1992. Olivine dissolution kinetics at near-surface
conditions. Chem. Geol. 97, 101–112.
Wolery, T.J., Jackson, K.J., Bourcier, W.L., Bruton, Carol, J., Viani, B.E., Knauss,
K.G., Delany, J.M., 1990. Current Status of the EQ3/6 Software Package for
Geochemical Modeling, in: Chemical Modeling of Aqueous Systems II. pp.
104–116.
98
Wright, K., Freer, R., Catlow, C.R.A., 1994. The energetics and structure of the
hydrogarnet defect in grossular: A computer simulation study. Phys. Chem.
Miner. 20, 500–503.
Zakaznova-Herzog, V.P., Nesbitt, H.W., Bancroft, G.M., Tse, J.S., 2008.
Characterization of leached layers on olivine and pyroxenes using high-
resolution XPS and density functional calculations. Geochim. Cosmochim.
Acta 72, 69–86.
99
Chapter 4: The role of the Si-rich surface layer in olivine
dissolution 3: Spatially and temporally resolved incorporation of an isotopic tracer (29Si)
Natalie C. Johnsona,*, Burt Thomasb, Robert J. Rosenbauerb, Dennis K. Birdc, Gordon E. Brown, Jr.a,c,d,
Christopher E.D. Chidseye, Kate Maherc
a. Department of Chemical Engineering, Stauffer III, 381 North-South Mall, Stanford University, Stanford CA 94305, USA
b. U.S. Geological Survey, 345 Middlefield Rd, Menlo Park CA 94025, USA
c. Department of Geological and Environmental Sciences, 450 Serra Mall, Braun Hall Building 320,
Stanford University, Stanford CA 94305, USA
d. Department of Photon Science and Stanford Synchrotron Radiation Lightsource,
SLAC National Accelerator Laboratory, Menlo Park CA 94025, USA
e. Department of Chemistry, Stauffer II, North-South Mall,
Stanford University, Stanford CA 94305, USA
In preparation for submission to Geochimica et Cosmochimica Acta.
100
Abstract
The kinetics and mechanisms of forsterite dissolution are of interest because of
the potential for large-scale subsurface carbon dioxide (CO2) storage via mineral
carbonation. However, an amorphous, Si-rich surface layer forms on the forsterite
surface during reaction with water and supercritical CO2 and the characteristics of this
layer, including its potential for passivation of the mineral surface, are not fully
understood. Forsterite was reacted with solutions containing an isotopic tracer, 29
Si,
and pCO2 (100 bar) at 60 °C for 2 days and 19 days in both well-mixed and unmixed
reactors. The isotopic composition of both solid and aqueous phases were then
analyzed for to resolve incorporation of the tracer both spatially and temporally. A
precipitated surface layer (characterized by incorporation of the 29
Si tracer) exists on
forsterite reacted for 19 days (20% enrichment compared to unreacted forsterite) but is
less evident at 2 days (3% enrichment), suggesting that the precipitated surface layer
forms more rapidly after the bulk fluid reaches amorphous silica saturation. Steady-
state forsterite dissolution and amorphous silica precipitation rates were calculated
from the rate of tracer dilution to be 1.1±0.6×10-13
mol cm-2
s-1
for the mixed
experiment. The unmixed experiment reached steady-state shortly before its
conclusion because of an extended initial transient stage of reaction, the steady-state
dissolution rate was calculated to be 2.5±0.9×10-14
mol cm-2
s-1
. Data from the
experiments were used to further evaluate a surface kinetic model that includes the Si
isotopic tracer. Based on these findings, a complexing agent (Tiron) was added to a
mixed experiment to facilitate transfer of Si from the mineral surface to the bulk fluid.
The presence of a complexing agent for Si doubled the rate of steady-state dissolution
relative to the control experiment.
101
4.1. Introduction
Mineral carbonation, which describes the chemical reaction between silicate
minerals and carbon dioxide (CO2) to make carbonate minerals, is a promising
strategy for subsurface carbon storage. The main advantage of mineral carbonation (or
mineral trapping) over other storage mechanisms is its stability over thousands to
millions of years. Mineral carbonation is thermodynamically favored and occurs in
natural systems (Boschi et al., 2009; Mervine et al., 2014), motivating the study of
reaction kinetics with the ultimate goal of engineering subsurface storage within rocks
rich in reactive minerals.
Over the past two decades, there have been many studies on the dissolution
and carbonation rates of forsterite ((Mg,Fe)2SiO4), a reactive silicate mineral found in
mafic and ultramafic rocks. Flow-through experiments have shown the rate of
dissolution to increase with increasing temperature and decreasing pH (Rimstidt et al.,
2012), while studies using batch reactors report a weak dependence on pH and strong
dependence on H4SiO4 (Chapter 2; Daval et al., 2011; Wang and Giammar, 2013).
During dissolution under a range of conditions, a Si-rich, amorphous layer forms on
forsterite surfaces that may passivate the surface (Andreani et al., 2009; Daval et al.,
2011; King et al., 2010). However, the mechanism of layer formation is not fully
understood: some studies have observed preferential removal of Mg resulting in a
“leached layer”, (Bearat et al., 2006; Daval et al., 2011; E. H. Oelkers, 2001;
Pokrovsky and Schott, 2000a, 2000b), while others have observed sharp interfaces
between the surface layer and mineral (Hellmann et al., 2012), evidence of
psuedomorphic replacement, and uptake of an oxygen isotope tracer (King et al.,
2011), all of which are indicative of a dissolution/reprecipitation process. Others have
proposed that the Si-rich surface layer may form via both mechanisms (Saldi et al.,
2013).
In general, the dynamics of Si removal from dissolving olivine are poorly
understood, in part because the aqueous phase reaches saturation with respect to a
secondary phase (amorphous silica) and thus the Si(aq) reaches a steady-state
concentration and cannot be used to determine the dissolution rate. However, Si
102
removal is known to influence the overall rate (Pokrovsky and Schott, 2000a, 2000b)
and thus the study of Si surface dynamics is relevant and important. Most attempts to
accelerate forsterite (and other silicate minerals) dissolution have focused on changing
the pH or complexing cations with organic acids (Blake and Walter, 1999a; Krevor
and Lackner, 2009; Manning et al., 1992; Oelkers, 2001; Olsen and Rimstidt, 2008;
Prigiobbe and Mazzotti, 2011). In this study, we use an isotopic tracer to improve our
understanding of the role of Si dynamics in forsterite dissolution, and then attempt to
accelerate the kinetics by introducing a Si-complexing agent (catechol).
The isotopic tracer (29
Si) allows this study to constrain both primary mineral
dissolution/secondary mineral precipitation rates using the "isotope ratio method"
from Gruber et al. (2013) and to probe the formation of the Si-rich surface layer.
Analysis of the Si isotopic composition of both the aqueous Si and the surface layer of
the mineral phases allows for precise and accurate tracing of the partitioning of the
29Si spike. For example, a dissolution/reprecipitation process would incorporate
29Si
from the bulk solution, while a leaching process would not, so one would expect a
precipitated Si-rich layer to be at or near isotopic equilibrium with the aqueous phase.
The incorporation of an isotopic spike uniquely allows for the calculation of secondary
phase (SiO2(am)) precipitation rates. The isotopic and modeling approaches presented
here build on observations from a series of previously published experiments, along
with new experiments conducted to evaluate the ability of Si complexing agents to
accelerate forsterite dissolution rates under batch conditions.
Part 1 of this series used high resolution transmission electron microscopy
(HR-TEM) to image and analyze the Si-rich surface layer on reacted forsterite
(Chapter 3). Several distinct types of Si-rich layers were observed: an amorphous, Si-
rich layer with varying Mg/Si (Mg/Si=0.2-1.5) as a function of depth. This first layer
formed within 2 days of reaction, before the bulk solution reached SiO2 saturation, and
is 20-25 nm deep. The second layer to form is porous, amorphous, and very Si-rich
(Mg/Si<0.3) and formed only after the bulk solution reached saturation with respect to
amorphous silica. The third layer was crystalline, but also depleted in Mg. Chapter
three hypothesizes that the first layer formed by preferential Mg removal, or leaching,
103
and the second layer was the result of SiO2(am) precipitation. However, direct
evidence for either mechanism could not be inferred from solely the HR-TEM
analysis.
Part 2 of this study developed a "surface kinetic" model to describe the
dissolution kinetics and spatial relationships between Mg and Si observed in part 1 and
in previous work (Appendix 1). Using parameters from TEM images, we successfully
modeled the temporal evolution of solution composition by operationally defining a 25
nm "active" surface layer that forms via preferential Mg release. Newly precipitated
SiO2(am) is accounted for in the model, but is assumed to play no role in passivating
the surface. The model results suggest that under conditions of high H4SiO4 and
mildly acidic to neutral pH, the rate is controlled by both SiO2 removal from the active
layer-precipitated layer interface and the exchange of Mg for protons within the active
layer. Thus, the model provides further support for the importance of both a leached
(active) layer and a non-passivating precipitated layer, but again, offers no direct
experimental evidence. Here we provide experimental evidence of the conceptual
models developed in parts 1 and 2 of this series.
4.2. Methods
4.2.1 Isotopic tracer preparation
The 29
Si spike was prepared gravimetrically from ISOFleX 99.8% 29
Si in the
oxide form. The 29
SiO2 was dissolved in 2M NaOH, using sonication to increase the
dissolution kinetics, then neutralized with HCl. The tracer solution was then diluted to
a concentration of 1.90 mM 29
Si and 0.64 M NaCl. The tracer was further diluted
before reaction with forsterite using deionized (DI) water, and sufficient NaCl was
added to achieve a concentration of 0.5 M NaCl and 0.29-0.35 M 29
Si .
4.2.2 Experimental apparatus and sampling procedures
Experiments were performed in Dickson-style "rocker bombs" described in
detail in Rosenbauer et al. (1983) that were modified for CO2 analysis following
Rosenbauer et al. (2005). The specific experimental conditions and sampling
104
procedures are described in full in Chapters 2 and 3. Briefly, the experiments were
performed in flexible, Au reaction cells equipped with a Ti-lined sampling tube and
valve for extraction of liquid and gas samples. The reaction cell was placed inside a
stainless steel autoclave and the annular space filled with DI water to function as a
pressure fluid and sustain a pressure of 100 bar. A temperature of 60 °C was
maintained by placement of the steel vessel inside a rocking furnace (Rosenbauer et
al., 1983). The apparatus allows for serial sampling without changing the pressure or
temperature conditions.
Aqueous samples were withdrawn periodically and analyzed for elemental
concentration, alkalinity, dissolved CO2 (see Chapter 2 for details), and Si isotopic
composition (see section 2.4). The total volume of fluid removed during sampling was
10% or less of the starting volume in order to minimize the effect of changing
water:solution ratios throughout each experiment. At the conclusion of each
experiment, the stainless steel autoclave was removed from the furnace, vented, and
disassembled as quickly as possible (<30 minutes) to minimize precipitation from
quenching. The solids were filtered from the reaction solution and rinsed well, then
dried at 40 °C for several hours. Reacted mineral phases were analyzed with imaging
techniques (SEM, TEM) (Chapter 3) and for isotopic composition (see section 4.2.3).
This study includes five experiments, a suite of four experiments designed to
understand the effects of time and mixing on the Si-rich surface layer, and a fifth to
explore the idea of increasing dissolution rate by adding a complexing agent. All five
experiments contained 0.5 M NaCl and were performed at 60 °C and 100 bar. Two
experiments lasted 2 days (2-R, 2-S) and two lasted 19 days (19-R, 19-S). One
experiment of each duration was rocked inside the furnaces to ensure that the reacting
components were well mixed (2-R, 19-R), and one of each duration was not rocked to
investigate potential transport limitations (2-S, 19-S). All four of these experiments
contained a 29
Si tracer in the initial aqueous solution. These four experiments are
discussed in detail in Chapter 3 and are also presented in Table 4.1.
105
Table 4.1: Duration, degree of mixing, and additive concentrations for five experiments.
Experiment name Duration Mixing Additive
2-S 2 days No 0.35 mM 29
Si
2-R 2 days Yes 0.29 mM 29
Si
19-S 19 days No 0.35 mM 29
Si
19-R 19 days Yes 0.29 mM 29
Si
Catechol 35 days Yes 100 mM Tiron
The fifth experiment contained 100 mM of Tiron (4,5-Dihydroxy-1,3-
benzenedisulfonic acid disodium monohydrate, (OH)2C6H2(SO3Na)2·H2O). Tiron is a
sulfonated catechol with a first pKa of 7.2, and is used primarily as a colorimetric
reagent for iron, manganese, titanium, and molybdenum. Tiron has been shown to
complex with silicic acid in the neutral to elevated pH range (Bai et al., 2008). The
complex is not stable below pH of 6 (Bai et al., 2011). Thus, Tiron could potentially
form a complex with Si near the mineral surface (where the pH is high), but return to
the acid state in regions where the pH is lower (due to carbonic acid dissociation)
acting as a shuttle for Si from the mineral surface to the bulk solution.
4.2.3 Isotopic analysis of the aqueous phase
Aqueous samples were prepared for isotopic analysis using ion
chromatography designed to remove dissolved sodium, magnesium, and iron,
Table 4.2: Resin cleaning and Si purification procedure adapted from Georg et al. (2006).
Separation Stage Solution Matrix Volume (mL)
Pre cleaning 2.5 N HCl 3
4.0 N HCl 3
MQ water 1
7.0 N HCl 3
MQ water 1
4.0 N HCl 3
2.5 N HCl 3
Conditioning MQ water 6
MQ water 3
MQ water 3
Sample Load and Collect Sample 2
Collection MQ water 2
MQ water 2 mL
106
following methods established by Georg et al. (2006). Sufficient sample to contain 5
g Si was dried overnight at 100˚C, then placed in water to a total fluid volume of 2
mL. Columns were loaded with 40 g of BioRad AG 50W-X2 (200-400 mesh) resin for
a volume of 1.8-2.0 mL. The columns were prepared by a number of pre-cleaning
steps based on Georg et al. (2006) (Table 4.2), then the sample was loaded onto the
column and Si collected.
Isotopic analyses were performed on a Nu Plasma HR Multi-collector ICP-MS
where the 28
Si, 29
Si, and 30
Si were measured simultaneously in Faraday cups with 15
cycles per sample. Samples were run in duplicate or triplicate and numbers presented
are averages of all data obtained per sample (i.e. they represent 30-45 cycles of data
collection). Before each sample measurement, a baseline measurement was taken
using double distilled water, and then a standard was measured (IRMM 3.8 ppm
solution, 29
Si/30
Si=1.44). Ratios obtained from the MC-ICP-MS were corrected for
mass bias using the 28
Si/30
Si ratio from the standard run immediately before each
sample; the reproducibility of the 29/30 ratio of the standard was very high with 2σ =
0.1% over two days of measurement.
4.2.4 Isotopic analysis of the mineral phase
Mineral surfaces were analyzed for elemental concentration using the
Stanford/USGS Sensitive High Resolution Ion Microprobe-Reverse Geometry
(SHRIMP-RG) built by Australian Scientific Instruments. Fresh and reacted forsterite
grains for SHRIMP analysis were prepared by drilling several small holes in 25.4 mm
epoxy discs, which were then filled with indium metal. The disks and metal were
polished to ensure an even height of 4 mm. Reacted forsterite grains were then pushed
into the indium metal to preserve the surface layers. The mount was coated with gold
using a plasma deposition for 50 seconds and then placed in a vacuum oven for at least
12 hours to dry. Before analysis, each mount was photographed with an optical
microscope camera to aid in sample navigation.
The SHRIMP-RG directs a beam of oxygen anions onto the mineral surfaces.
Reverse geometry (RG) refers to the placement of the magnet prior to the electrostatic
analyzer (ESA), resulting in much higher mass resolution, but also the inability to
107
collect data for several masses at once. In order to minimize the effect of time-
averaging the ratios and also to maximize the number of measurements at the surface
and near surface of the mineral, we used a very low amperage primary beam (0.1-0.3
nA), consisting of O2-
focused through a 50 nm Köhler aperture with a brightness of 6
and a spot size of approximately 10 microns. Using transmission electron microscopy
images and energy dispersive spectroscopy linescans for calibration (Chapter 3), we
determined the rate of sputtering to be 3-9 nm/minute. We collected data for 26
Mg,
29Si , and
30Si by peak hopping, with the entire sampling cycle taking about 20 seconds
due to peak centering and the time allowed for the magnet to settle in between masses.
Thus, the ratios that we present are actually time-averaged ratios over about 20
seconds of collection time. Also, the beam is incident on the sample at a 45° angle,
resulting in slightly uneven milling and sampling of the pit walls; both of which have
an effect similar to depth averaging. Elemental ratio calibration was performed using
unreacted forsterite grains and standard glass, NIST 611.
4. 2.5 Calculation of Dissolution and Precipitation Rates
Dissolution rates were calculated for experiments 19-S and 19-R using two
different methods: (1) a conventional method based on the differences between Mg
concentrations at different sample times (e.g. Chapter 2), and (2) the "isotope ratio
method" from Gruber et al. (2013), which is based on the changes in Si isotopic ratios
with time. Using the conventional method, a dissolution rate, Rd, was calculated
according to Equation 1:
(1)
where the difference between Mg concentrations is divided by the associated
difference in time between the samples, and this term is multiplied by the inverse of
the stoichiometric number (XMg,d=1.84) and the mineral surface area (Ad). For all
experiments, the dissolving surface area is assumed to be constant throughout the
experiment at the measured BET surface area of 5735 cm2/g.
The isotope ratio method was developed by Gruber et al. (2013) specifically to
measure dissolution and precipitation rates of minerals under conditions near
dt
dMg
AXR
ddMg
d
,
1
108
equillibrium. The method requires an isotopically enriched initial solution and the
measurement of both the total concentration of dissolved species and the isotopic ratio
of the solution as a function of time. In our case, we are measuring the rate of Si
release from forsterite and Si uptake by amorphous silica when the system is near
equilibrium with the secondary phase, SiO2(am). The calculations are based on mass
balances for total Si, 29
Si, and 30
Si, where the rate of change of each species is equal to
the dissolution rate minus the precipitation rate. Gruber et al. (2013) show the
derivation in detail; the final result for the dissolution rate is Equation 2:
Fo
Si
Fo
SidSi
d
XXSi
SiA
dt
Sid
dt
Sid
Si
Si
R
293030
29
2930
30
29
(2)
Equation 2 also contains the mol fractions of 30
Si (X30SiFo
= 0.031) and 29
Si (X29SiFo
=
0.047) found in the forsterite. The isotopic ratios were measured for each sample, but
the rates of change of 29
Si and 30
Si are averaged over two samples by necessity, so for
the purpose of this calculation, the isotopic ratios were also averaged over two
samples and thus assumed to be linear with time between each measurement.
The isotope ratio method accounts for both dissolved Si addition and removal
rates as components of the material balance, so once the dissolution rate is calculated,
the precipitation rate can be calculated from Equation 3.
(3)
Equation 3 relates the net change in Si concentrations and dissolution rate, Rd, with the
precipitation rate, Rp (Gruber et al., 2013). This equation includes two stoichiometric
values and surface areas, one each for the dissolving phase and the precipitating phase.
In this case, the stoichiometric value is 1 for both phases (forsterite and SiO2(am)),
and we assume that the surface area available for precipitation, Ap, is the same as the
surface area available for dissolution, Ad. For comparison, we calculated a second
precipitation rate with a different set of assumptions; namely, that Si and Mg are
released congruently, and that any missing Si in solution is due to SiO2(am)
precipitation (Equation 4).
pppSidddSi RAXRAXdt
dSi,,
109
(4)
This equation represents a material balance in which the sum of the precipitation rate
and the observed net change in Si with time is equal to the expected change of Si with
time, or the forsterite dissolution rate as calculated using changes in [Mg]. We note
that the precipitation rates we calculate from Equations 3 and 4 are effectively net
removal rates that include both the new SiO2(am) precipitation rate and any net
exchange of Si(aq) into the reacted surface or Si-rich layer, where the later is
supported by the presence of 29
Si in Si-rich surface layer after 2 days of reaction. We
cannot explicitly separate the exchange rate from the secondary phase precipitation
rate, but we can use a comparison of precipitation rates calculated from different
equations to implicitly calculate each rate.
4.3. Results
4.3.1 Time-resolved solution compositions
As forsterite dissolves and releases Si into solution, the concentration of total
dissolved Si increases and the ratio of 29
Si/30
Si decreases, approaching the natural ratio
of 1.51 (Fig. 4.1A-B). In both 19-day experiments (19-R and 19-S), the solution
reaches and then exceeds saturation with respect to amorphous silica. Saturation was
reached at day 1 in experiment 19-R (Fig. 4.1A) and between 6 and 14 days in
experiment 19-S (Fig. 4.1B). Because of the lack of mixing in experiment 19-S,
forsterite dissolved more slowly than in experiment 19-R and thus a longer time was
required to reach saturation (Chapter 3). Both experiments also show dilution of the
initial 29
Si spike with time (Fig. 4.1A-B). The initial dissolved Si was 99.8% 29
Si , but
Si from the dissolving forsterite was 4.7% 29
Si , 3.1% 30
Si, and 92.2% 28
Si. Thus, the
ratio of 29
Si/30
Si decreased as more Si was released from forsterite. The rate of dilution
is greater in experiment 19-R, again because forsterite dissolution rate was greater in
this experiment.
The Mg/Si ratio was elevated in the aqueous phase for experiments 19-R and
19-S for the entirety of both experiments (i.e. the aqueous phase is enriched in Mg
pppSi
dMg
RAXdt
dSi
dt
dMg
X,
,
1
110
relative to the mineral phase) (Fig. 4.1C). In experiment 19-R, Mg/Si = 4 after 1 hour
of reaction and the ratio drops towards the stoichiometric ratio of 1.84 over the next
several days, then rapidly increases for the remainder of the experiment. In experiment
19-S, Mg/Si decreases monotonically during the experiment but remains above 1.84.
Figure 4.1: Silicon concentrations and 29Si/30Si ratios as a function of time for 19-R
(A) and 19-S (B), with amorphous silica saturation plotted for comparison. Ratios of
Mg/Si vs . time (C) show that the fluid composition is elevated in Mg relative to the
mineral phase.
111
4.3.2 Si ratios on mineral surfaces
The Mg/Si ratios (Fig. 4.2A-B) and 29
Si/30
Si ratios (Fig. 4.2C-D) were
measured on mineral surfaces as a function of time using the ion microprobe
(SHRIMP) and converted to a function of depth using transmission electron
microscope images and energy dispersive spectroscopy linescans to calibrate the rate
of ion milling. Mineral surfaces from all four experiments had lower Mg/Si ratios than
expected based on mineral stoichiometry (Mg/Si = 1.84), which is indicative of a Si-
rich surface layer. The amount of depletion varied by experiment; the level of
depletion increased from experiments 2-S and 2-R to experiments 19-S to 19-R and
correlates with reaction extent (i.e. experiments with greater reaction extents resulted
in mineral surfaces with greater Mg-depletion). The surface of forsterite grains from
all four experiments also showed enrichment in 29
Si, although the degree of
enrichment differed. The 19-day experiments (19-R, 19-S) had more surface 29
Si
compared to the 2-day experiments.
The depths of Mg depletion and 29
Si incorporation differ. For the forsterite
from experiment 19-R, the measured Mg/Si ratio reaches the ratio of bulk crystalline
forsterite (1.84) at about 60 nm depth, but the measured 29
Si /30
Si ratio reaches the
bulk ratio (1.51) at 45 nm deep. The same pattern is visible in experiment 19-S (Fig.
4.2B,D); the Mg/Si ratio reaches the bulk forsterite level at 45 nm and the 29
Si /30
Si
ratio reaches the bulk ratio at 30 nm deep.
4.3.3 Comparison of 29
Si/30
Si ratios in the mineral and aqueous phases
Comparison of the 29
Si/30
Si ratios in the solid surface and aqueous phases from
the last sample (i.e. sample taken at 2 days or 19 days) for four experiments is shown
in Fig. 4.3. The ratio at the solid surface was calculated by averaging the first
SHRIMP-RG data point over 3 to 5 mineral grains, and the error bars represent one
standard deviation. The first measurement taken represents the solid-aqueous
interface, though depth averaging does result in this value being averaged over 3 nm
of layer thickness. Reaction extent increases from left to right and corresponds with
dilution of the initial isotopic spike. Experiment 2-S had the lowest extent of reaction
112
and also has the highest 29
Si/30
Si ratio in the aqueous phase, whereas experiment 19-R
had the highest extent of reaction and the lowest 29
Si/30
Si ratio in the aqueous phase.
The explanation for this trend is the same as discussed above; as the reaction
progresses and Si is released from forsterite at 29
Si/30
Si = 1.51, the aqueous 29
Si/30
Si
decreases and approaches 1.51.
4.3.4 Reaction in the presence of a complexing agent
The addition of a complexing agent (Tiron, a catechol derivative) does increase
the rate of reaction, especially during the first 7 days of reaction (Fig. 4.4A,B). After 7
Figure 4.2: Ratios of Mg/Si as a function of depth for 2,19-R (A) and 2,19-S(B). The
depth was calibrated using the SI-rich layer thickness as measured from TEM images
(65 nm (19-R), 40 nm (19-S), 25 nm (2-S), and 20 nm (2-R)). Isotopic ratios
(29Si/30Si) are also plotted as a function of depth for the rocking (C) and stationary (D)
experiments. Each data point is an average from at least three different forsterite grains.
113
days, three times as much Mg was released by the experiment with catechol than by
the control experiments (19-R and Experiment 1, no complexing agents) (Fig. 4.4A).
Silicon release was also initially elevated in the presence of catechol. Particularly
striking is a sharp increase in Si concentrations from 0 to 2 days followed by a steep
drop to a steady-state concentration at 2 days (Fig. 4.4B). Data from the other
experiments do not exhibit this feature. Over longer times (t > 7 days), the change in
Mg concentrations with time (slope of the data in Fig. 4.4A) is approximately equal to
the rate of change in the control experiments. During that same period, the Si
concentration was constant in the catechol experiment and slightly elevated relative to
Experiment 1, though consistent with experiment 19-R (Fig. 4.4B). The lower Si
concentrations in Experiment 1 are likely due to an imperfect sampling procedure, in
which SiO2 precipitated in the sampling tube and was inadvertently filtered out before
ICP analysis (Chapter 2). Thus, the complexing agent appears to increase the initial
rates of both Mg and Si release, but does not significantly affect the long-term kinetics
or activity of Si in solution.
Figure 4.3: Comparison of isotopic composition of the aqueous and mineral phases for four
experiments.
114
4.4. Discussion
4.4.1 Conceptual Model of Surface Layers on Olivine
A conceptual model for forsterite dissolution was outlined in the first two parts
of this series (Chapter 3, Appendix 1) and is described briefly here. When forsterite is
exposed to water, Mg sites are protonated, which reduces the strength of the Mg-O
bond. As a result, Mg is removed from the forsterite surface and leaves behind a
network of silica tetrahedra and a region depleted in Mg. This process forms the
"active" surface layer and generates a reservoir of surface SiO2 sites, henceforth
referred to as >SiO2. The removal of Mg from the surface is considered irreversible
during dissolution, but H4SiO4 and >SiO2 are expected to exchange. The net rate of
Figure 4.4: Time-resolved Mg (A) and Si (B) concentrations for three experiments: 19-R (this
study), Experiment 1 (Chapter 2), and catechol (this study).
115
active layer dissolution is determined by the difference between the forward reaction
rate, or the removal of >SiO2 from the active layer/aqueous interface, the backward
reaction rate, or the attachment of H4SiO4 to the active layer interface. Aqueous SiO2
is also controlled by precipitation of a secondary phase, amorphous silica, which
precipitates on reactor and mineral grains, but does not appear to passivate the surface.
The sum of net dissolution and precipitation describes the measured changes in
H4SiO4.
Based on HR-TEM data, the active layer forms within two days of the reaction
and remains at a roughly constant thickness of 20-25 nm out to 19 days (Chapter 3).
Thus, this layer appears to control the dissolution rate. The net rate of Si removal from
the active layer/aqueous interface controls the rate at which the active layer advances
into the mineral, while the rate of Mg removal controls the rate at which the active
layer forms. The experimental data used to develop this conceptual model were
collected under one set of conditions, but understanding the kinetics would allow for
prediction of rates under a range of conditions: SiO2(am) solubility, pH, and ionic
strength. In fact, the active layer may not be Si-rich under certain conditions (e.g., high
pH). In Appendix 1, we were not able to constrain the Si dynamics beyond suggesting
the potential importance of the rate of net Si removal from the active layer as a control
on the Mg supply to the active layer and net dissolution. The addition of an isotopic
tracer allows us to uniquely quantify the H4SiO4/>SiO2 exchange rates. Constraining
these rates improves our understanding of Si dynamics on the surface of dissolving
forsterite.
The rest of the discussion is organized as follows. First, we evaluate the rates
of forsterite dissolution from experimental data using elemental and isotopic material
balances. We then analyze the temporal evolution of surface layers on the dissolving
forsterite surface in terms of experimental evidence and a surface kinetic model. The
model includes provision for the isotopic tracer in addition to constraints from solution
data and nm-scale analysis of reacted forsterite. Finally, we discuss the implications of
our findings for the observed early rate increase in the presence of catechol.
116
4.4.2 Dissolution and Precipitation Rates
Two methods were used to calculate the forsterite dissolution and
SiO2(am) precipitation rates: a conventional method based on an elemental mass
balance and the "isotope ratio method" which uses the elemental concentrations as
well as the isotopic composition of the aqueous phase (Gruber et al., 2013.). A
comparison between two methods for calculating the rates from experiments 19-R and
19-S as a function of time is shown in Fig. 4.5A-B using Equations 1-4. The first rate
(point "a") represents the transient initial stage of reaction, which is characterized by
preferential release of Mg over Si (Fig. 4.5A) resulting in an erroneously high
precipitation rate (Fig. 4.5B, point "a"). The dissolution rate calculated using dMg/dt
is four times greater than the rate calculated from the isotope ratio method. This initial
difference in release rates of Mg and Si could be due to incongruent dissolution (e.g.,
Chapter 2) and/or exchange of aqueous Si with Si in the surface layer. The
precipitation rate calculated from the isotope ratio method is zero (not shown),
indicating that only release of Si into solution is occurring during that timeframe (Fig.
4.5B). The precipitation rate calculated from Equation 4 (a material balance of
expected Si release based on the measured Mg release) does indicate precipitation
during this first time interval; however, the calculation assumes stoichiometric
dissolution, which the isotope ratio analysis has shown to be invalid for the first time
point. Thus, the combination of a calculated dissolution and precipitation rate from the
isotope ratio method for this first time point provides strong evidence for incongruent
dissolution, or preferential Mg release (i.e. leaching), during the first hour of reaction
(point "a" in Fig. 4.5 A-B).
In order to satisfy the isotopic mass balance in experiment 19-R, removal of Si
from the fluid phase must begin between 1 hour and 1 day after reaction (indicated by
non-zero precipitation rates calculated from the isotope ratio method at point "b" a,
Fig. 4.5B). However, the precipitation rate calculated from dMg/dt is zero during that
time because dissolution is stoichiometric (Fig. 4.5A, point "b"). Combining these
observations leads us to hypothesize that SiO2(am) is not precipitating between 1 hour
and 2 days of reaction, but rather the "precipitation rate" during this time actually
117
represents the exchange rate between the aqueous phase and surface layer. In other
words, the net precipitation of SiO2(am) is zero, but isotopic exchange proceeds at a
measurable rate. Amorphous silica precipitation begins to contribute to the removal of
H4SiO4 at the 3rd data point, which represents the time between 1 and 3 days of
reaction (point "c"). The bulk aqueous phase reaches SiO2(am) saturation at 1 day,
suggesting that SiO2(am) precipitation does not begin until after the system achieves
saturation.
Between 3 and 19 days, dissolution rates calculated with both methods are
constant and consistent within uncertainty (Fig. 4.5A). The Si removal rate from
Figure 4.5: Dissolution rates for 19-R calculated using the isotope ratio method and dMg/dt indicates
that Mg and Si release are stoichiometric at t>5 days (A), and Si removal rates (precipitation plus
exchange) are the same within uncertainty as dissolution rates at t> 5 days (B). The same
calculations for 19-S reveal slower dissolution and a longer transient stage (C) and precipitation
begins later (D). Missing data are either less than zero or have uncertainties that span positive and
negative rates (i.e. rates are zero within uncertainty).
118
solution (precipitation + backwards exchange) increases during that time (Fig. 4.5B),
possibly due to increasing surface area available for precipitation. The similarity of the
two rates indicates that forsterite dissolution and H4SiO4 removal (via amorphous
silica precipitation and exchange with the mineral surface) occur at the same rate
during the steady-state regime. Also, the rate of Si removal from the surface is roughly
twice the rate of Si attachment to mineral from the fluid phase. Rates of dissolution,
precipitation, and Si exchange are presented in Table 4.3 along with results from the
model (see section 4.4).
Performing the same calculations for dissolution rate for experiment 19-S
yields results similar to those from experiment 19-R, though the rates are slower (Fig.
4.5C). The initial, transient stage in which dissolution is incongruent lasts for 2 days
instead of 1 hour (the first 2 rates from 19-S show preferential Mg release), and the
steady-state rate is nearly 5 times slower than for experiment 19-R. Data from
experiment 19-S highlight the utility of the isotope ratio method for dissolution rate
calculations when the rate is very slow (Gruber et al., 2013). Magnesium
concentrations are nearly constant with time for parts of the experiment, so calculating
the rate of change for each sampling interval results in rates with large uncertainties
that span zero. However, the rate can easily be calculated using the rate of 29
Si dilution
to give consistent rates and small uncertainties (Table 4.3).
Precipitation rates for experiment 19-S were calculated using Equations 3 and
4 (Fig. 4.5D). In this case, the first two rates represent the transient, incongruent stage
of the reaction and thus the precipitation rates calculated from dMg/dt are not
meaningful (points "d" and "e"). Removal (precipitation + exchange) does not begin
Table 4.3: Steady-state dissolution and precipitation rates for 19-R and 19-S .
Forsterite dissolution rate,
Mg removal (mol cm-2 s-1)a
Forsterite dissolution
rate, Si removal (mol cm-2 s-1)b
Rate of Si disappearance from
solution (precipitation + back exchange) (mol cm-2 s-1)b
19-R Exp. 1.1 ± 0.6 × 10-13 1.2 ± 0.4 × 10-13 1.1 ± 0.2 × 10-13
19-R Model 9.4 ± 0.9 × 10-14 1.0 ± 0.4 × 10-13 1.0 ± 0.4 × 10-13
19-S Exp. 2.1 ± 0.4 × 10-14 2.5 ± 0.9 × 10-14 2.4 ± 0.6 × 10-14
19-S Model 2.1 ± 0.5×10-14 2.7 ± 0.5 × 10-14 2.6 ± 0.4 × 10-14
a. Calculated using dMg/dt
b. Calculated using Isotope Ratio Method (Gruber et al., 2013)
119
until the last time interval (14-19 days, point "f") as measured by the isotope ratio
method. The last sampling interval also represents the time when the aqueous phase is
supersaturated with respect to SiO2(am) (Fig. 4.1B). Thus, we do not see definitive
precipitation until the bulk aqueous phase exceeds saturation. Expected precipitation is
non-zero when calculated using Equation 4, though as discussed previously, the lack
of reliability of the rates calculated from dMg/dt reduces the legitimacy of any rate
based on a Mg mass balance. The precipitation rate in Table 4.3 for experiment 19-S
comes from the last time interval and Equation 3, and its similarity to the dissolution
rate suggests that experiment 19-S may also have reached steady-state forsterite
dissolution and SiO2(am) precipitation, though not until 14 days (experiment 19-R
reached steady-state dissolution/precipitation at 4 days). Values for rates with
uncertainties that spanned positive and negative values were not included in the figure.
From the rate analysis above, we can draw several conclusions. First, the
isotope ratio method is a superior method for measuring silicate mineral dissolution
rates when the rate is slow (Gruber et al., 2013). Secondly, we can extract SiO2(am)
precipitation and Si(aq) to >SiO2 exchange rates, which are useful for constraining the
surface kinetic model. Finally, we see strong evidence for a transient stage during the
first hour (19-R) to 2 days (19-S) of reaction, during which Mg is preferentially
removed from the mineral surface. In both experiments, SiO2(am) precipitation does
not begin until after the bulk aqueous fluid exceeds SiO2(am) saturation.
4.4.3 Direct evidence for both leaching and dissolution/reprecipitation mechanisms
Evidence for a "leached" surface layer can be found from both solution data
(rate calculation) and analysis of reacted forsterite surfaces. Unreacted forsterite
surfaces are stoichiometric (as measured by X-ray Photoelectron Spectroscopy in
Chapter 2); consequently preferential Mg release during the initial, transient stage of
reaction must lead to the formation of a Si-rich surface layer via a leaching
mechanism.
Further evidence for leaching is found in the SHRIMP-RG data, as the surfaces
of forsterite grains reacted for 2 days are indeed Si-rich (or Mg-depleted) (Fig. 4.2A-
B). However, these same surfaces exhibit a Si isotopic composition that is very
120
different from the aqueous phase, indicating that SiO2(am) precipitation from the bulk
fluid did not cause the Si-rich layer (Fig. 4.3). Both experiments 2-R and 2-S do show
slight enrichment of 29
Si in the surface layer relative to the bulk mineral composition
(Fig. 4.2C-D), but this is better explained by Si exchange between the aqueous phase
and the surface layer than by precipitation because the mineral surfaces are 53-75%
less enriched in 29
Si than the aqueous phase. Silicon exchange between the aqueous
phase and surface layer results in some uptake of the isotopic spike; the amount of
uptake depends on the relative rates of the forward and backward reactions. If the
backward reaction is faster than the forward reaction, the amount of tracer uptake
would be small. In the opposite situation, with a fast forward rate relative to the
backward rate, the mineral surface isotopic composition would closely resemble that
of the aqueous phase. The low degree of 29
Si enrichment in experiments 2-R and 2-S
suggests that the back reaction (Si attachment) is much slower than the forward
reaction; in other words, the net effect is removal of Si from the mineral in both the
rocking and stationary case.
After the initial, transient stage of reaction, dissolution becomes congruent and
approaches a steady-state rate (Fig. 4.5A,C). At this point, a leached, Si-rich surface
layer has formed on forsterite surfaces, and equivalent release rates of Si and Mg
require that this "leached" layer persists but does not grow in thickness. TEM/EDS
analysis supports this hypothesis (Chapter 3) The surface layers formed during
experiments 2-R and 2-S were depleted in Mg, and the Mg/Si ratios varied with depth.
The characteristic concentration gradient of the leached layer was also observed in
experiment 19-R and to a lesser extent in experiment 19-S, suggesting that the leached
layer retains its thickness and advances into the mineral as the reaction progresses.
At longer times, SiO2(am) precipitation occurs and creates a second, Si-rich
layer. Precipitation begins between 1 and 3 days (19-R) or 14 and 19 days (19-S), after
the aqueous phase reaches saturation with respect to SiO2(am) (Fig. 4.5B,D). Analysis
of forsterite reacted for 19 days indicates that the mineral surfaces are 14-15% more
enriched in Si (depleted in Mg) than 2-R and 2-S (Fig. 4.2A-B). Images and EDS
linescans from the HR-TEM study show that the precipitated layer is porous, very
121
depleted in Mg, and appears to grow in thickness with reaction extent (Chapter 3).
An interesting note is that even after precipitation begins, the mineral surface is
not isotopically equivalent to the aqueous phase. This discrepancy could have several
causes, the first being an analytical limitation of the SHRIMP-RG. Only one element
can be analyzed at any given time, and about 20 seconds are required to cycle through
the three elements analyzed in this study, meaning that each ratio shown is averaged
over 1-3 nm of depth. Further, the beam is incident on the mineral surface at a 45°
angle and secondary ions released by the beam do not all originate at the same depth.
The limitation of depth averaging means that a true surface measurement is impossible
to achieve, though we reduced the depth over which each measurement was averaged
by using a low beam current.
Secondly, the surface that we measure reflects a time-integrated fluid
composition. The isotopic composition of the fluid changes significantly throughout
the reaction, and we know that the exchange of H4SiO4 with >SiO2 is slow relative to
the forward rate of >SiO2 removal. Thus, the time required for isotopic equilibration
may be quite long, and experiments of a longer duration would be required in order to
achieve equivalence.
4.4.4 Surface kinetic model with Si isotopes
The conceptual model for forsterite dissolution presented in Chapters 3 and
Appendix 1 describes the forsterite surface as several layers, based on both TEM
imaging and a surface kinetic model constrained by Mg isotopes. The base layer is
crystalline forsterite, and an amorphous, Si-rich "active" layer sits on top. This layer is
formed via preferential removal of Mg, leaving behind a region depleted in Mg and
enriched in Si. This layer forms via process typically described as leaching, but we use
the term "active" instead because this layer is an actively dissolving reaction front that
recedes into the mineral as dissolution progresses (Chapter 3, Appendix 1). The next
layer is a precipitated SiO2 layer that forms only after the bulk fluid reaches saturation
with respect to amorphous silica. Silicon isotopes allow us to uniquely evaluate this
model and build on the concepts developed in Appendix 1.
This model differs from a traditional Arrhenius kinetic model in two major
122
ways. First, this model explicitly includes rates for both forward and backward
reactions, unlike the Arrhenius equation which accounts for only measured
concentration changes, themselves representing the net change of forward and
backward reactions. Secondly, this new model links spatial configuration (surface
layer formation and migration) with measurable changes in solution chemistry.
Briefly, protonation of Mg surface sites results in weakening of the Mg-O bond
and the irreversible release of Mg (Pokrovsky and Schott, 2000b) plus the generation
of a SiO2 surface site, >SiO2 (Reaction 1).
Reaction 1
Removal of Si from forsterite is described as a reversible reaction in which the surface
sites are hydrated to form silicic acid (Reaction 2).
Reaction 2
Finally, H4SiO4 precipitates as an amorphous silica surface layer that does not
passivate the mineral surface (Reaction 3). The precipitation step dominates H4SiO4
after the reaction reaches SiO2(am) saturation.
Reaction 3
The rate of forsterite dissolution is controlled by the rates of reactions 1 and 2, while
reaction 3 controls the concentration of H4SiO4.
The rate of reaction 1 (Equations 5) is a function of the detachment coefficient,
Mg, the volume of the active layer, VAL, and the probability that any given site
contains Mg, PMg.
ALMgMgMg VPR (5)
Appendix 1 includes expressions for both 26
Mg and 24
Mg in the rate expressions as the
model was originally developed using Mg isotopes to constrain the rates of reaction.
However, in this case, we remove the Mg isotope expressions and add Si isotope
expressions to incorporate the measurements of 29
Si and 30
Si. Thus, Si release is
described with two equations, one for each isotope measured, that are the sum of Si
detachment and attachment rates from/to the active layer surface. The expression for
29Si release from forsterite is shown in Equation 6, with an equivalent expression for
>Mg2SiO4 + 4H + ¾®¾ > SiO2 + 2H2O + 2Mg2+
> SiO2 + 2H2O¾®¾¬¾¾ H4SiO4 aq( )
SiO2 am( ) + 2H2O¾®¾¬¾¾ H4SiO4 aq( )
123
30Si and the rate of release of that isotope.
(6)
In the above expression, SiO2 is the removal rate of >29
SiO2 from the surface of the
active layer and P29SiO2 is the probability that a site contains 29
SiO2. The detachment
rates for all silicon isotopes were assumed to be the same; i.e. we ignored Si isotope
fractionation in the development of the model. Further, 28
Si was not explicitly
included in the model, but was instead assumed to follow 30
Si (i.e. both isotopes are
released from the forsterite at rates proportional to their mole fractions).
The
attachment rate is a function of a rate constant, k>29SiO2, the concentration of silicic
acid (consisting of 29
Si), and the site probability. The sum of the two rates is
multiplied by the forsterite surface area A, the site density Γ, and the molar volume,
Vmin (Appendix 1). Unlike Equation 5, this reaction does not include the term for the
active layer volume because Si is only removed from the active layer surface, whereas
Mg can be removed from the entire active layer volume.
Addition of silicon isotopes to the model requires calculation of the site
probabilities, P>29SiO2, and P>30SiO2. The probability is a function of the bulk forsterite
isotopic composition (X29Si=0.047, X30Si = 0.031), the mole fraction of Si in forsterite
(XSi/Fo = 1), and the molar volume Vmin. Again, the expression shown is for 29
Si
(Equation 7), but an analogous equation was used for 30
Si.
min
29
229V
XXP
AL
Si
AL
Si
SiO (7)
The probability does not change with time as it does in the case of Mg, as any kinetic
isotope fractionation at the per mil level during dissolution is overwhelmed by the
addition of the tracer 29
Si and thus could not be resolved.
The rate of silica precipitation is also expressed as two different equations, one
for each measured Si isotope. In both cases, the rate depends on a rate constant (kSiO2),
the solubility product of SiO2(am) (KSiO2), the concentration of silicic acid (H429
SiO4),
the mol fraction in the fluid phase (X29Siaq
), the surface area, and the molar volume
(Equation 8), following Druhan et al. (2013).
min2294
29
4229229229229 ) ( AVPSiOHkPR SiOSiOSiOSiOSiO
124
min
29229
2
29
429229229 1 Av
XK
SiOHXkR
aq
SiSiO
aq
SiSiOSiO
(8)
We assumed that negligible fractionation occurred during SiO2 precipitation, due to
the fact that the aqueous phase was greatly enriched with 29
Si from the addition of a
tracer.
The rates expressions in equations 5-8 are combined in order to link the
specific reactions to the active layer concentrations of Mg and each Si isotope
(Equations 9-11):
(9)
(10)
(11)
Equations 9-11 display the method through which the individual rates are coupled.
Because the rate expressions include the concentration terms (Px), and the
concentration terms depend on the rates, changing a single parameter in one rate
expression affects the other rates in the system. As such, the model cannot be easily
manipulated to produce a wide range of different reaction trajectories as it partly
constrained by the interrelationships of the rate expressions.
Experiments and analyses allow us to constrain the model significantly. The
active layer thickness was set to 25 nm based on HR-TEM data (Chapter 3) and the
surface area was measured by BET. Time-resolved solution compositions were used to
fit the equations to the data, by adjusting the rates of Mg and Si detachment (g,
SiO2=SiO2) and the rate constants of SiO2 precipitation (kSiO2=kSiO2).
Magnesium was used as the primary constraint because SiO2 precipitation results in a
H4SiO4 plateau, though the addition of Si isotopes in this study does allow for
calculation of a precipitation rate (actually the sum of SiO2 precipitation and SiO2
attachment to the mineral surface), which provides an additional limitation on the
model.
ALSiMgSiOSiOSiOMg
MgVXRRRR
dt
dP/)( /230229228
ALSiOSiMgSiMg
SiO VRXXRdt
dP// 229/29
229
ALSiOSiMgSiMg
SiO VRXXRdt
dP// 230/30
230
125
Two sets of model parameters were determined to fit experimental data from
experiments 19-R and 19-S (Table 4.4). The parameters resulted in steady-state rates
that are comparable to the measured rates (Table 4.3). In experiment 19-R, the rates of
Mg appearance in solution, Si appearance in solution, and Si removal from solution
agree; i.e. forsterite dissolution and SiO2 precipitation are occurring at the same rates
and dissolution is at steady state. The model indicates a slightly lower steady-state Mg
appearance rate, though the two are equivalent within uncertainty. In the case of
experiment 19-S, the rates of Si appearance in solution and Si removal are equal,
which again suggests steady-state forsterite dissolution and SiO2(am) precipitation.
The Mg removal rate (normalized by forsterite stoichiometry) is slightly lower, though
equivalent within uncertainty. Rates calculated by the model and from experimental
data agree within uncertainty.
One major conclusion from the model and relating specifically to experiment
19-S is that this experiment appears to reach steady-state only shortly before its
conclusion. The rates from the model continue to change with time for the duration of
the 20-day period over which the model was calculated. Thus, in the absence of
advective transport caused by mixing, the initial transient stage lasts up to 20 days,
which is five times longer than in the rocking (mixed) scenario. Experiment 19-S
progressed slower than experiment 19-R due to transport (diffusion) limitations
(Chapter 3), thus the extended transient stage is not surprising.
The parameters used to fit the experiment differ, with lower rate constants
needed to fit data from 19-S relative to 19-R. The parameters from 19-R were first
presented in Appendix 1, but the parameters for 19-S are different than the values
presented in the previous study. The addition of the isotopic tracer in this work
provided additional constraints on the rate of Si appearance and disappearance from
Table 4.4: Model parameters for 19-R and 19-S
νMg
(s-1
)
ν>SiO2
(s-1
)
k>SiO2
(M-1
s-1
)
kSiO2(am)
(mol m-2
s-1
)
Active layer
thickness (nm)
19-R Model* 6±3×10-4 3.8±0.3×10-5 <7.3×10-4 ~3.3×10-9 25
19-S Model 1.0±0.5×10-4 1.0±0.3×10-5 <5.0×10-4 ~3.3×10-9 21 *From Appendix 1
126
solution, so for the first time we were able to resolve forsterite dissolution rates based
on Si appearance and SiO2(am) precipitation rates based on Si disappearance. As a
result, the value for ν>SiO2 used in Appendix 1 was found to be nearly an order of
magnitude lower than required to match the measured Si appearance rate. Increasing
ν>SiO2 in the model resulted in a reasonable Si detachment rate, but it also increased
the steady-state Mg detachment rate (because the rate expressions are closely
coupled), and a reduction of the active layer thickness was needed to match the
concentration data. Results from a TEM study suggest that the active layer may be
thinner on forsterite from an unmixed reactor (Chapter 3), and thus the slight reduction
here is justified.
Fits of the model to experimental data are shown in Fig. 4.6. The fit for
experiment 19-R is quite good, with the model replicating both the short and long-
term behavior of Si and Mg concentrations as a function of time (Fig. 4.6A) as well as
the 29
Si/30
Si ratio as a function of time (Fig. 4.6C). The model also replicates the
concentration data for 19-S (Fig. 4.6B), though the model fit to the silicon isotope data
is imperfect (Fig. 4.6C). The parameters were chosen to model each experiment to fit
the data shown in Fig. 4.6 as well as match the rates of Mg appearance, Si appearance,
and Si disappearance (Table 4.3).
The 29
Si tracer was diluted more slowly in experiment 19-S than experiment
19-R due to a decreased forsterite dissolution rate; however, the poor model fit
suggests that the lower dissolution rate alone cannot account for the slower decrease in
29Si/
30Si. Thus, another factor, not accounted for by the model, must be important in
transport-limited systems. One possibility stems from the model's assumption of a
well-mixed active layer. In the experiments, attachment of H4SiO4 to surface sites on
the forsterite would initially result in a reservoir of surface >SiO2 enriched in 29
Si,
assuming negligible fractionation of the H4SiO4 (which is initially predominantly
29Si). However, the model cannot account for the additional
29Si in the top 1-2 unit
cells, and instead distributes it throughout the thickness of the active layer (12-14 unit
cells). As a result, the modeled active layer is only slightly enriched in 29
Si, whereas
the forsterite in the experiments may have significant 29
Si enrichment on the mineral-
127
fluid interface with a 29
Si/30
Si ratio resembling that of the aqueous phase. A higher
concentration of 29
Si on the mineral surface would result in slower dilution of the 29
Si
aqueous spike and may explain the discrepancy between the data and the model. In a
transport-limited system such as experiment 19-S, this effect may be stronger due to
higher rates of >SiO2 attachment relative to >SiO2 detachment (Table 4.4). Additional
spatial resolution in the model, specifically the addition of a third surface layer 1-2
unit cells thick and with an isotopic composition similar to that of the fluid, may
correct this issue.
Figure 4.6: Kinetic model fit to concentration data from rocking experiments
(A), stationary experiments (B), and isotope data from both experiments (C).
128
The possibility of a very thin (1-2 unit cells) surface layer that contains
actively exchanging >SiO2 is an important consideration in experiment that use a Si
isotopic tracer to measure silicate dissolution rates. Exchange between the surface
>SiO2 sites and H4SiO4 may result in a very thin surface layer enriched in the isotopic
tracer that persists beyond the first several hours to days of reaction. Thus, the
dissolution rate calculated from the tracer dilution at intermediate times (3-19 days in
the case of 19-S) may be lower than the actual dissolution rate. We expect this issue to
be particularly important for minerals that dissolve slowly, either due to transport
limitations as we observed in this study or due to slower reaction kinetics.
In summary, the kinetics of >SiO2 removal may not play as significant a role
as previously thought. A comparison of experiments with and without NaCl
(Appendix 1) and with and without mixing show a lack of a clear correlation between
the rates of surface detachment (ν>SiO2) and the dissolution rate. In fact, the experiment
with the second slowest dissolution rate (Exp. 3, [NaCl]=0 M) is associated with the
highest ν>SiO2 (Chapter 2, Appendix 1). In this experiment, the relatively slow
dissolution rate is due to a high >SiO2 attachment rate (k>SiO2, the backwards rate
constant). Thus, the long-term dissolution rate appears to be controlled by the
exchange between aqueous and surface SiO2, and neglecting the backwards rate (SiO2
attachment) yields an incomplete picture of forsterite dissolution, particularly under
conditions characterized by high H4SiO4.
4.4.5 Can we improve the reaction rate by complexing Si?
Numerous studies have used organic acids in order to increase the rate of
silicate mineral dissolution (Amrhein and Suarez, 1988; Bennett, 1991; Blake and
Walter, 1999b; Hänchen et al., 2006; Krevor and Lackner, 2009; Oelkers, 2001; Olsen
and Rimstidt, 2008; Prigiobbe and Mazzotti, 2011; Prigiobbe, 2009; Teir et al., 2007;
Wogelius and Walther, 1991). In particular, the dissolution rate of forsterite is
accelerated in the presence of oxalic and citric acids (Olsen and Rimstidt, 2008;
Prigiobbe and Mazzotti, 2011), which is thought to be the result of ligand-promoted
dissolution. Organic ligands may bind to Mg sites in forsterite, causing the Mg to be
overcoordinated and thus breaking the Mg-O bond (Liu et al., 2006). Thus, organic
129
acids increase the dissolution rate by enhancing the release of cations such as Mg, but
do not directly affect the release of Si from the mineral surface. This study introduced
Tiron, a catechol derivative that is known to form a complex with Si (Bai et al., 2008),
with the goal of specifically enhancing the net rate of Si removal from forsterite.
Tiron does increase the reaction rate during the initial, transient period of
reaction and has a small effect on steady-state forsterite dissolution rates. Between 8
and 35 days, forsterite dissolves at a rate of 1.8±0.2×10-13
mol cm-2
s-1
, which is 80 %
higher than the results of experiment 19-R and double the rate from Experiments 1 and
2 (Chapter 3). Rates for this comparison were all calculated using a linear regression
of [Mg] vs. time data (Chapter 3). While the rate improvement is relatively minor in
this case, we might expect a more significant effect in a transport-limited system.
Tiron was added with the anticipation that it would take advantage of pH
gradients in the aqueous phase to "shuttle" Si from the mineral surface to the bulk
fluid. More specifically, an elevated pH near the mineral surface (expected because
forsterite, an alkaline mineral, consumes protons during dissolution) would shift the
Tiron equilibrium towards the conjugate base, which would then complex with silicic
acid. As the complex moves away from the alkaline conditions near the mineral
surface, it would destabilize and re-form the acid catechol plus a separate silicic acid
molecule. However, results from both rocking and stationary experiments indicate that
the stationary experiments are transport-limited, and may have stronger concentration
gradients in the aqueous phase than rocking experiments. Stronger concentration
gradients could result in a stronger rate effect by Tiron (or another complexing agent).
Because the stationary experiments are perhaps most representative of in situ storage
in the subsurface, complexing agents should be considered as potential additives to
increase the rate of mineral carbonation, despite the relatively minor effect observed
using a well-mixed system.
4.5. Conclusions
Measurement of silicon isotopic compositions of the aqueous and solid phases
from four experiments allowed us to build upon the first two parts of this series in
developing a new kinetic model for forsterite dissolution under conditions relevant for
130
subsurface carbon storage. A suite of four reactions performed at 60 C and pCO2 =
100 bar for 2 and 19 days, with one experiment of each duration mixed (rocked) and
one of each duration not mixed (stationary) containing 29
Si as an isotopic tracer. The
Si isotopic composition of the aqueous phase was measured with MC-ICPMS and
yielded data on the rate of tracer dilution. Precise dissolution and precipitation rates
were calculated from the changing isotopic composition using the "isotope ratio
method" developed by Gruber et al. (2013). Measurements of the Si isotopic
composition of the solid phase were collected from the Stanford/USGS SHRIMP-RG.
Incorporation of 29
Si into the solid phase was 15% greater in forsterite reacted for 19
days than for 2 days, though isotopic equivalence was not observed.
The combination of aqueous and solid phase isotopic data plus HR-TEM data
from part 1 of this study (Chapter 3) resulted in the conclusion that forsterite dissolves
initially by a preferential release of Mg, resulting in the formation of a leached Si-rich
surface layer 20-25 nm thick. This layer forms during the initial, transient stage of the
reaction, which lasts 2 days (mixed experiment) to 19 days (unmixed experiment).
This "active" layer persists for 19+ days and advances into the mineral as a reaction
front. A kinetic model developed based on experimental data describes the relative
rates of Mg and Si removal from the active layer and is able to accurately predict time-
resolved solution compositions from a variety of mixed experiments (Appendix 1),
effectively linking molecular dynamics in the solid phase to bulk, measurable
dissolution rates. Further work is needed to improve the model for non-mixed
reactions, which are likely most representative of the subsurface.
Precipitation of secondary amorphous silica occurs after the bulk fluid reaches
saturation with respect to that phase, resulting in uptake of the isotopic tracer. The lack
of isotopic equivalence between the aqueous and mineral phases even after
commencement of SiO2(am) precipitation suggests that exchange between the solid
and aqueous phases is slow and longer reaction times would be required to observe
isotopic equilibrium. Precipitation occurs on the dissolving forsterite (on top of the
active layer) as well as on reactor walls, though the phase is porous and does not
appear to passivate the mineral surface (Chapter 3).
131
Finally, Tiron (a catechol derivative) was added to the reaction solution to act
as a shuttle, taking advantage of pH gradients that may exist near the mineral surface
to complex with >SiO2 and transport it away from the mineral to the bulk fluid. We
observed a modest increase in the steady-state dissolution rate with the addition of
Tiron in a mixed reaction, though we would expect a larger effect if Tiron was used in
an unmixed reactor where concentration gradients may be more pronounced.
4.6 Acknowledgments
We would like to acknowledge the Global Climate and Energy Project at
Stanford University (GCEP-48942) for funding this work. We also acknowledge the
facilities which operate and maintain the instrumentation used for this study: The
Stanford Nanocharacterization Laboratory, the Stanford-USGS SHRIMP-RG
Laboratory, and the Stanford ICP-MS/TIMS Laboratory. In particular, we
acknowledge Karrie Weaver and Caroline Harris for their technical assistance.
132
4.7 References
Abraham, K., Opfergelt, S., Fripiat, F., Cavagna, A.-J., De Jong, J.T.M., Foley, S.F.,
André, L., Cardinal, D., 2008. δ30Si and δ29Si Determinations on USGS
BHVO-1 and BHVO-2 Reference Materials with a New Configuration on a Nu
Plasma Multi-Collector ICP-MS. Geostandards and Geoanalytical Research 32,
193-202.
Amrhein, C., Suarez, D.L., 1988. The use of a surface complexation model to describe
the kinetics of ligand-promoted dissolution of anorthite. Geochim. Cosmochim.
Acta 52, 2785–2793.
Andreani, M., Luquot, L., Gouze, P., Godard, M., Hoise, E., Gibert, B., 2009.
Experimental Study of Carbon Sequestration Reactions Controlled by the
Percolation of CO -Rich Brine through Peridotites. Environ. Sci. Technol. 43,
1226–1231.
Bai, S., Tsuji, Y., Okaue, Y., Yokoyama, T., 2008. First Detection of a Silicic Acid
Complex with a Catechol Derivative under Natural Conditions. Chem. Lett. 37,
1168–1169.
Bai, S., Tsuji, Y., Okaue, Y., Yokoyama, T., 2011. Complexation of Silicic Acid with
Tiron in Aqueous Solution Under near Natural Condition. J. Solution Chem. 40,
348–356.
Bearat, H., Mckelvy, M.J., Chizmeshya, A.V.G., Gormley, D., Nunez, R., Carpenter,
R.W., Squires, K., Wolf, G.H., 2006. Carbon Sequestration via Aqueous Olivine
Mineral Carbonation : Role of Passivating Layer Formation. Environ. Sci.
Technol. 40, 4802–4808.
Bennett, P.C., 1991. Quartz dissolution in organic-rich aqueous systems. Geochim.
Cosmochim. Acta 55, 1781–1797.
Blake, R.E., Walter, L.M., 1999a. Kinetics of feldspar and quartz dissolution at 70–
80°C and near-neutral pH: effects of organic acids and NaCl. Geochim.
Cosmochim. Acta 63, 2043–2059.
Blake, R.E., Walter, L.M., 1999b. Kinetics of feldspar and quartz dissolution at 70–
80°C and near-neutral pH: effects of organic acids and NaCl. Geochim.
Cosmochim. Acta 63, 2043–2059.
Boschi, C., Dini, A., Dallai, L., Ruggieri, G., Gianelli, G., 2009. Enhanced CO2-
mineral sequestration by cyclic hydraulic fracturing and Si-rich fluid infiltration
into serpentinites at Malentrata ( Tuscany , Italy ). Chem. Geol. 265, 209–226.
133
Daval, D., Sissmann, O., Menguy, N., Saldi, G.D., Guyot, F., Martinez, I., Corvisier,
J., Garcia, B., Machouk, I., Knauss, K.G., Hellmann, R., 2011. Influence of
amorphous silica layer formation on the dissolution rate of olivine at 90°C and
elevated pCO2. Chem. Geol. 284, 193–209.
Druhan, J.L., Steefel, C.I., Williams, K.H., DePaolo, D.J., 2013. Calcium isotope
fractionation in groundwater: Molecular scale processes influencing field scale
behavior. Geochim. Cosmochim. Acta 119, 93–116.
Georg, R.B., Reynolds, B.C., Frank, M., Halliday, a. N., 2006. New sample
preparation techniques for the determination of Si isotopic compositions using
MC-ICPMS. Chem. Geol. 235, 95–104.
Gruber, C., Harpaz, L., Zhu, C., Bullen, T.D., Ganor, J., 2013. A new approach for
measuring dissolution rates of silicate minerals by using silicon isotopes.
Geochim. Cosmochim. Acta 104, 261–280. doi:10.1016/j.gca.2012.11.022
Hänchen, M., Prigiobbe, V., Storti, G., Seward, T.M., Mazzotti, M., 2006. Dissolution
kinetics of fosteritic olivine at 90–150°C including effects of the presence of
CO2. Geochim. Cosmochim. Acta 70, 4403–4416.
Hellmann, R., Wirth, R., Daval, D., Barnes, J.-P., Penisson, J.-M., Tisserand, D.,
Epicier, T., Florin, B., Hervig, R.L., 2012. Unifying natural and laboratory
chemical weathering with interfacial dissolution–reprecipitation: A study based
on the nanometer-scale chemistry of fluid–silicate interfaces. Chem. Geol. 294-
295, 203–216.
King, H.E., Plümper, O., Geisler, T., Putnis, A., 2011. Experimental investigations
into the silicification of olivine: Implications for the reaction mechanism and acid
neutralization. Am. Mineral. 96, 1503–1511.
King, H.E., Plümper, O., Putnis, A., 2010. Effect of secondary phase formation on the
carbonation of olivine. Environ. Sci. Technol. 44, 6503–6509.
Krevor, S.C., Lackner, K.S., 2009. Enhancing Process Kinetics for Mineral Carbon
Sequestration. Energy Procedia 1, 4867–4871.
Liu, Y., Olsen, A.A., Rimstidt, J.D., 2006. Mechanism for the dissolution of olivine
series minerals in acidic solutions. Am. Mineral. 91, 455–458.
Manning, D.A.C., Ttir, K.G., Ra, E.I.C., 1992. Feldspar dissolution in the presence of
organic acid anions under diagenetic conditions : An experimental study. Org.
Geochem. 19, 483–492.
134
Mervine, E.M., Humphris, S.E., Sims, K.W.W., Kelemen, P.B., Jenkins, W.J., 2014.
Carbonation rates of peridotite in the Samail Ophiolite, Sultanate of Oman,
constrained through 14C dating and stable isotopes. Geochim. Cosmochim. Acta
126, 371–397.
Oelkers, 2001. The mechanism, rates and consequences of basaltic glass dissolution: I.
An experimental study of the dissolution rates of basaltic glass as a function of
aqueous Al, Si and oxalic acid concentration at 25°C and pH = 3 and 11.
Geochim. Cosmochim. Acta 65, 3671–3681.
Oelkers, E.H., 2001. General kinetic description of multioxide silicate mineral and
glass dissolution. Geochim. Cosmochim. Acta 65, 3703–3719.
Olsen, A.A., Rimstidt, J.D., 2008. Oxalate-promoted forsterite dissolution at low pH.
Geochim. Cosmochim. Acta 72, 1758–1766.
Pokrovsky, O.S., Schott, J., 2000a. Forsterite surface composition in aqueous
solutions : A combined potentiometric, electrokinetic, and spectroscopic
approach. Geochim. Cosmochim. Acta 64, 3299 –3312.
Pokrovsky, O.S., Schott, J., 2000b. Kinetics and mechanism of forsterite dissolution at
25°C and pH from 1 to 12. Geochim. Cosmochim. Acta 64, 3313–3325.
Prigiobbe, V., 2009. Enhanced dissolution of olivine using organic substances.
Prigiobbe, V., Mazzotti, M., 2011. Dissolution of olivine in the presence of oxalate,
citrate, and CO2 at 90°C and 120°C. Chem. Eng. Sci. 66, 6544–6554.
Rimstidt, J.D., Brantley, S.L., Olsen, A.A., 2012. Systematic review of forsterite
dissolution rate data. Geochim. Cosmochim. Acta 99, 159–178.
Rosenbauer, R.J., Bischoff, J.L., Fadtke, A.S., 1983. Hydrothermal alteration of
graywacke and basalt by 4 molal NaCl. Econ. Geol. 78, 1701–1710.
Rosenbauer, R.J., Koksalan, T., Palandri, J.L., 2005. Experimental investigation of
CO2–brine–rock interactions at elevated temperature and pressure: Implications
for CO2 sequestration in deep-saline aquifers. Fuel Process. Technol. 86, 1581–
1597.
Saldi, G.D., Daval, D., Morvan, G., Knauss, K.G., 2013. The role of Fe and redox
conditions in olivine carbonation rates: An experimental study of the rate limiting
reactions at 90 and 150°C in open and closed systems. Geochim. Cosmochim.
Acta 118, 157–183.
135
Teir, S., Revitzer, H., Eloneva, S., Fogelholm, C.-J., Zevenhoven, R., 2007.
Dissolution of natural serpentinite in mineral and organic acids. Int. J. Min. 83,
36–46.
Wang, Giammar, D.E., 2013. Forsterite Dissolution in Saline Water at Elevated
Temperature and High CO2 Pressure. Environ. Sci. Technol. 47, 168–173.
Wogelius, R.A., Walther, J. V, 1991. Olivine dissolution at 25°C : Effects of pH, CO2,
and organic acids. Geochim. Cosmochim. Acta 55, 943–954.
136
137
Appendix 1: The role of the Si-rich surface layer in olivine dissolution 2: A spatially resolved surface kinetic model
Kate Maher1, Natalie C. Johnson2, Laura Nielsen Lammers1*, Abe B. Torchinsky1, Karrie L. Weaver1,
Dennis K. Bird1 and Gordon E. Brown, Jr.1,2,3
1 Department of Geological and Environmental Sciences,
Stanford University, Stanford CA 94305-2115, USA 2 Department of Chemical Engineering,
Stanford University, Stanford CA 94305-2115, USA 3 Department of Photon Science and Stanford Synchrotron Radiation Lightsource,
SLAC National Accelerator Laboratory, 2575 Sand Hill Rd., MS 69, Menlo Park, CA 94025, USA
*Now at: Department of Environmental Science, Policy, and Management,
UC Berkeley, CA 94720, USA
Submitted to Geochimica et Cosmochimica Acta, July 2014.
138
Abstract
The development of complex alteration layers on silicate mineral surfaces
undergoing dissolution is a widely observed phenomenon. Given the complexity of
these layers, most kinetic models used to predict rates of mineral-fluid interactions do
not explicitly consider their formation. As a result, the relationship between the
development of the altered layers and the final dissolution rate is poorly understood.
To improve our understanding of the relationship between the alteration layer and the
dissolution rate, we developed a spatially resolved surface kinetic model for olivine
dissolution and applied it to a series of closed-system experiments consisting of three-
phases (water (± NaCl), olivine, and supercritical CO2) at conditions relevant to in situ
mineral carbonation (i.e. 60˚C, 100 bar CO2). We also measured the corresponding
26
Mg of the dissolved Mg during early stages of dissolution. Analysis of the solid
reaction products indicates the formation of Mg-depleted layers on the olivine surface
as quickly as 2 days after the experiment was started and before the bulk solution
reached saturation with respect to amorphous silica. The 26
Mg of the dissolved Mg
decreased by approximately 0.4‰ in the first stages of the experiment and then
approached the value of the initial olivine (-0.35‰) as the steady-state dissolution rate
was approached. We attribute the preferential release of 24
Mg to a kinetic effect
associated with the formation of a Mg-depleted layer that develops as protons
exchange for Mg2+
.
We used the experimental data to calibrate a surface kinetic model for olivine
dissolution that includes crystalline olivine, a distinct “active layer” from which Mg
can be preferentially removed, and secondary amorphous silica precipitation. By
coupling the spatial arrangement of ions with the kinetics, this model is able to
reproduce the early and steady-state long-term dissolution rates, and the kinetic
isotope fractionation. In the early stages of olivine dissolution the dissolution rate is
controlled by exchange of protons for Mg, while the steady-state dissolution rate is
controlled by the removal of both Mg and Si from the active layer. Modeling results
further indicate the importance of the spatial coupling of individual reactions that
occur during olivine dissolution. The inclusion of Mg isotopes in this study
139
demonstrates the utility of using isotopic variations to constrain interfacial mass
transfer processes. Alternative kinetic frameworks, such as the one presented here,
may provide new approaches for modeling fluid-rock interactions.
140
A1.1. Introduction
The mineral olivine (i.e. (Mg,Fe,Ca)2SiO4), common in lower crustal and
upper mantle rocks, is one of the most reactive rock-forming silicate minerals under
surface and near surface conditions. As a result, the dissolution rate of olivine is linked
to the evolution of the lithosphere and upper mantle (Sleep et al., 2004), the utility of
olivine as an industrial feedstock for mineral carbonation (Oelkers and Cole, 2008),
neutralization of acid mine drainage (Jambor et al., 2007), and geoengineering using
enhanced silicate weathering (Hartmann et al., 2013; Koehler et al., 2010). Given the
high reactivity of olivine, the surface chemistry and dissolution kinetics have been
studied under a broad range of experimental conditions, including closed-system
(Daval et al., 2011a; Giammar et al., 2005; King et al., 2010) and well-mixed flow-
through dissolution experiments on powdered olivine samples (Oelkers, 2001;
Pokrovsky and Schott, 2000b), as well as single crystals (Jarvis et al., 2009; Saldi et
al., 2013). Characterization of the reaction products has also included a range of
techniques, such as x-ray photoelectron spectroscopy (Olsson et al., 2012; Pokrovsky
and Schott, 2000a; Zakaznova-Herzog et al., 2008), infrared spectroscopy (Giammar
et al., 2005; Loring et al., 2011; Pokrovsky and Schott, 2000a), transmission electron
microscopy (Bearat et al., 2006; Daval et al., 2011a) and atomic force microscopy
(Ruiz-Agudo et al., 2012). Despite the range of studies focused on olivine reactivity in
the presence of aqueous solutions, a precise description of the processes that control
olivine dissolution kinetics has been elusive (Daval et al., 2011a; Rimstidt et al.,
2012). This uncertainty has been attributed to the broad range of experimental
conditions (Rimstidt et al., 2012), combined with the implicit complexity of multi-step
heterogeneous reactions (Lasaga and Luttge, 2004). The lack of a comprehensive
model for olivine dissolution prevents the design of reliable strategies for both ex situ
and in situ mineral carbonation. In this paper we present a new spatially resolved
kinetic model for olivine dissolution constrained by time-resolved elemental
compositions, Mg isotope fractionation and High-Resolution Transmission Electron
Microscopy (HR-TEM) analysis.
Characterization studies of olivine surfaces from laboratory dissolution
141
experiments (Bearat et al., 2006; Daval et al., 2011a; Ruiz-Agudo et al., 2012) and
naturally-weathered olivine (Hellmann et al., 2012) have observed a pervasive Mg-
depleted amorphous layer on the olivine surface after exposure to water. Exploration
of the olivine-fluid interface using HR-TEM suggests this coating is typically less than
50 nm thick and forms rapidly (hours to days) after contact between the fluid and
olivine surface (Chapter 3, Bearat et al., 2006; Daval et al., 2011a; Pokrovsky and
Schott, 2000a; Ruiz-Agudo et al., 2012). However, the temporal evolution of the
coating has not been explored until recently (e.g., Chapter 3). As a consequence, the
relationship between the Mg-depleted layer and the overall mineral dissolution rate
has remained unclear.
The rapid development of Mg-depleted layers on the olivine surface during
dissolution is consistent with exchange of Mg2+
ions for protons, followed by
repolymerization of SiO4 tetrahedra. Several studies have also observed the olivine
dissolution rate to decrease as amorphous silica (SiO2(am)) saturation is approached
(Chapter 2; Daval et al., 2011b; Pokrovsky and Schott, 2000b), indicating that the
amorphous layer moderates the flux of Mg2+
and Si4+
ions from the olivine surface to
the bulk solution. In contrast, based in part on the sharp interface between the fresh
crystal and the surface layer, other studies have suggested that such coatings (which
are observed on the surfaces of many silicate minerals during dissolution) are a newly
formed precipitate that does not impact release of ions from the surface (Hellmann et
al., 2012). As noted in Chapter 3, the two processes are not mutually exclusive and
both preferential Mg removal and new precipitation of a secondary phase could occur
simultaneously. Regardless, these observations suggest that the spatial relationship
between reactions involving Mg and Si may play a role in determining the net olivine
dissolution rate. However, current models do not commonly address such spatial
considerations, and most quantitative models do not explicitly consider the multi-step
nature of the dissolution process, suggesting the need for an alternative kinetic model
for olivine, and by extension, silicate mineral dissolution.
General insights into the mechanisms associated with silicate mineral
dissolution may come from isotopic studies: kinetic isotope fractionation during
142
mineral dissolution under experimental conditions is now widely observed. For
example, fractionation has been noted for Mg isotopes during experimental olivine
dissolution (Wimpenny et al., 2010) and for Fe isotopes during experimental
hornblende (Brantley et al., 2004), micas (Kiczka et al., 2010) and goethite dissolution
studies (Wiederhold et al., 2006). Such isotope effects, which are not predicted by the
classical description of bulk mineral dissolution, provide constraints on the processes
occurring at mineral surfaces during dissolution. As a result, various models have been
developed to explain the preferential removal of ions from mineral surfaces. Brantley
et al. (2001; 2004) first observed Fe isotopic fraction associated with hornblende
dissolution both in the presence and absence of siderophores, but they did not observe
fractionation during goethite dissolution under the same conditions. The magnitude of
Fe isotopic fractionation in these studies also varied with the degree of fluid mixing
and the binding affinity of the ligand for Fe. The fractionation was attributed to a
kinetic isotope effect associated with the formation of a “leached layer” at the mineral
surface, however time resolved measurements or HR-TEM images were not available.
A number of additional studies have also demonstrated isotopic fractionation during
mineral dissolution (Pearce et al., 2012; Wimpenny et al., 2010; Ziegler et al., 2005).
Such fractionation is often associated with the initial stages of mineral dissolution,
before a steady state and/or stoichiometric dissolution rate is reached and the isotopic
composition of the dissolved fraction returns to that of the initial solid. For example,
Wimpenny et al. (2010) observed that the dissolved 26
Mg released from forsterite
under flow-through condition was approximately 0.2 to 0.4‰ lower than that of the
initial solids, an effect which was attributed to kinetic fractionation during mineral
dissolution. Such transient isotopic fractionation conveys important information about
the processes occurring at mineral surfaces and thus the extent and duration of the
isotope effects can be used to further constrain the interfacial processes controlling
mineral dissolution.
Macroscopic descriptions of mineral dissolution and precipitation are
numerous, and recent studies have extended these approaches to explain kinetic and
equilibrium isotope partitioning under surface reaction-controlled conditions
143
(DePaolo, 2011; Druhan et al., 2013). Such descriptions commonly assume that the
net reaction rate is the difference between the gross forward (precipitation) and
backward (dissolution) rates, where each rate is associated with a kinetic fractionation
(DePaolo, 2011). Although in practice it is difficult to quantify either, the backward
rate and attendant kinetic fractionation factor may be constrained by the far-from-
equilibrium dissolution rate (DePaolo, 2011). However, according to macroscopic
descriptions, persistent kinetic isotope effects during dissolution cannot be quantified
because the surface layers are removed stoichiometrically when a single net rate
constant and surface area are assumed. “Ion-by-ion” models provide a microscopic
description of growth kinetics and isotope and trace element partitioning based on
quasi-elementary ion attachment and detachment events (Nielsen et al., 2013; Nielsen
et al., 2012; Watkins et al., 2013). We use a similar surface kinetic approach here to
develop a model that links the microscopic processes at the olivine interface to the
macroscopic behavior indicated by the net dissolution rate. Because the individual
ions (i.e. Mg2+
, Si4+
and H+) are evaluated independently, but coupled via spatial and
stoichiometric considerations, the approach provides an alternative description of
mineral dissolution and the attendant kinetic isotope fractionation that is consistent
with both microscopic (HR-TEM) and macroscopic (time resolved fluid compositions)
observations.
To develop this framework, we present Mg isotope data for previously
published closed-system olivine dissolution experiments conducted at 60˚C, 100 bar
CO2, with solution compositions and rates detailed in Chapter 2. We also consider a
complementary suite of closed-system experiments conducted in the presence or
absence of fluid mixing where the reacted olivine surfaces have been characterized
using HR-TEM as described in Chapter 3. To explain the results from the four
experiments, a spatially resolved surface kinetic modeling approach is presented as an
alternative framework for describing forsterite (Mg2SiO4) dissolution kinetics. This
model reproduces the time-resolved solution compositions, the observed isotopic
fractionation, and the TEM surface structure and composition.
144
A1.2. Methods
In this study we use the experimental run products (i.e. reacted olivine grains
and resulting fluid samples) and experimental data (i.e. measured Mg and Si
concentrations, measured alkalinity, and calculated pH, saturation states and
dissolution rates) from a series of closed-system olivine dissolution experiments
conducted at 60˚C and 100 bar CO2 and detailed in Chapter 3. A brief description of
these experiments and the results is presented below. We report new data for the
26
Mg of the experimental olivine and the dissolved Mg over the duration of one
short-term experiment.
A1.2.1 Summary of olivine dissolution experiments
The experiments detailed in Chapters 2, 3, and Table A1.1 were conducted in
a flexible and inert Au-bag reaction vessel design with modifications for sampling and
analysis of CO2 (Chapter 2; Rosenbauer et al., 2005). Briefly, liquid CO2 was added to
the reaction vessel with high-pressure syringe pumps and temperature was maintained
by a proportional controller and measured with a calibrated type K thermocouple.
Serial samples were obtained throughout each experiment for total CO2 analysis,
cation analysis, and alkalinity titration (see details in Chapter 2). The olivine sample
is (Fo92) from the Twin Sisters Dunite, Washington, USA. The BET surface areas are
Table A1.1: Summary of experiments from Chapters 2 and 3 used in model development.
Experiment Name Temp. (˚C) PCO2 (bars) Duration
(days)
Electrolyte Mixing Mg
isotopes
Experiment 1a 60 100 74 0.5 M NaCl Well-mixed
Experiment 2 a 60 100 92 0.5 M NaCl Well-mixed
Experiment 3 a 60 100 4 none Well-mixed yes
Experiment 4 a 60 100 98 none Well-mixed
2-Rocking (2R)b 60 100 2 0.5 M NaCl Well-mixed
19-Rocking (19R) b 60 100 19 0.5 M NaCl Well-mixed
2-Stationary (2S) b 60 100 2 0.5 M NaCl Not mixed
19-Stationary (19S) b 60 100 19 0.5 M NaCl Not mixed
a: presented in Chapter 2
b: presented in Chapter 3
145
0.6410 m2/g for experiments 1-4 (38 – 75 m) and 0.57 m
2/g (70-105 m) for
experiments 2R, 19R, 2S, 19S, respectively. For the later series of experiments, a
larger grain size was used to facilitate characterization of the olivine surface after
reaction. The BET surface areas are used in the subsequent calculations.
The results of the eight experiments from Chapters 2 and 3 are presented in
this paper, all using the same starting material (Fo92 olivine), temperature (60oC), CO2
pressure (100 bar), and water:rock ratio by mass (50:1) as summarized in Table A1.1.
Six experiments contained 0.5 M NaCl, whereas two had no added electrolyte. Four
experiments were conducted to assess the temporal evolution of the olivine surface
under well-mixed (continuous rocking of the experimental vessel at 8
rotations/minute) and poorly mixed (stationary experimental vessel) conditions. The
reacted forsterite grains were characterized by HR-TEM, with details provided in
Chapter 3. The duration of the experiments ranged from 2 to 94 days. The shortest
experiments were conducted with higher sampling frequency to look at the initial
(non-steady state) dissolution kinetics as silica saturation is approached.
A1.2.2 Mg isotope analyses
We analyzed the isotopic composition of the starting olivine material and the
dissolved Mg from one experiment: experiment 3 (4 days, no electrolyte). For the
long-term experiments, the sample volumes were small in order to maintain
approximately constant water:rock ratios over more than 90 days. Thus, most of the
early samples for experiments 1, 4, 2R, 2S, 19R, 19S were consumed for elemental,
alkalinity, dissolved inorganic carbon (DIC) and Si isotopic analyses. For experiment
2, we found small samples combined with the presence of 0.5 M NaCl compromised
the isotopic analysis of the dissolved Mg.
All sample preparation procedures were carried out in laminar flow hoods in
the PicoTrace metal-free clean laboratory at Stanford University, and all acids were
Optima grade diluted with Milli-Q (18 M) water. Both San Carlos, AZ olivine and
the experimental olivine (WA) derived from the Twin Sisters Dunite (50 g) were
dissolved in capped screw-top Teflon beakers with 1 ml of a 3:1 mixture of
146
concentrated HF and HNO3, and left overnight at room temperature. The beakers were
then heated to approximately 90C for 24 hours before the caps were removed, and the
HF and HNO3 mixture was evaporated. The samples were then dissolved in 1 ml of
aqua regia, left overnight at room temperature, and then evaporated to dryness and
finally dissolved in 1N HNO3 prior to cation exchange chromatography. Fluid samples
were centrifuged for 10 minutes, evaporated to dryness, and dissolved in 1N HNO3
prior to cation exchange chromatography.
To purify and isolate Mg for isotopic analysis, Mg dissolved in 1N HNO3 was
loaded onto 0.5 ml of pre-cleaned Bio-Rad AG50W-X12 (200-400 mesh) resin.
Matrix elements were eluted in 1 N HNO3 and 0.1 N HF. Magnesium was eluted in 1
N HNO3. The samples were then dried and the above procedure was repeated twice.
Calcium was removed from the sample using Eichrom DGA resin (50-100 m).
Following the steps outlined above, the samples were dissolved in 2 N HNO3 and
loaded onto columns containing 0.2 ml of pre-cleaned Eichrom DGA resin. Mg was
eluted and collected with 2N HNO3 while Ca remained on the columns. The collected
samples were then evaporated to dryness to complete the Mg separation process. The
yield on the chromatographic procedure was >99.95%. Processing of the pure Mg
standards CAM1 and SU1 (internal elemental standard) with this technique did not
lead to a detectable shift in the measured Mg isotope ratio.
Mg isotope measurements were carried out on a Nu Plasma MC-ICP-MS using
medium resolution and wet plasma at Stanford University. Purified Mg samples were
dissolved in 2% HNO3, centrifuged, and diluted to 1.7 ppm prior to MC-ICP-MS
analysis. Purified Mg-bearing solution was introduced to the instrument through a
self-aspirating Glass Expansion SeaSpray nebulizer directly into the spray chamber.
Mg isotopes were measured simultaneously with 24
Mg, 25
Mg, and 26
Mg collected in
the L10, L5, and H1 Faraday cups, respectively; ion beams were positioned such that
measurements were collected on the low-mass side of the CN+ interference. Sample
measurements were bracketed by analysis of the standard DSM3, and between all
measurements a 2% HNO3 rinse solution was aspirated for approximately two minutes
until the signal on the H1 Faraday cup was less than 1x 10-4
volts (Galy et al., 2003).
147
For each analysis, (26/24
Mg)DSM3/‰ (and (25/24
Mg)DSM3/‰) values were calculated by
comparing the measured 26
Mg/24
Mg (25
Mg/24
Mg) ratio of the samples to the average
26Mg/
24Mg (
25Mg/
24Mg) ratio of the bracketing measurement standards. Hereafter, we
abbreviate the (26/24
Mg)DSM3/‰ as the 26
Mg. To constrain the accuracy and
precision of measurements made on the MC-ICP-MS, we compared the measured
isotope ratios at two standard deviations (2 s.d.) of pure Mg and natural rock reference
materials to accepted or reported values (Table A1.2). The reference material CAM1
was repeatedly measured over four months of analytical sessions with a 26
Mg value
of -2.60 0.17 (n=32), in agreement with the accepted value of -2.580.14 (Galy et al.,
2003). The isotopic ratio of an in-house synthetic measurement standard consisting of
1000 ml of High-Purity Standards pure 1000 ppm Mg elemental standard was
characterized by repeated measurements over four months. This measurement
standard, “Stanford University 1” (SU1) was analyzed 76 times and found to have a
26
Mg value of -0.55 0.14 (n=76).
The natural rock standards BHVO-1, BIR-1, and San Carlos olivine were also
within the uncertainty of accepted values. The measured 26
Mg of BHVO-1 was -0.19
Table A1.2: Mg isotope data for experimental olivine and experiment 3 from Chapter 2 along with
average values for reference materials, where n is the number of times the individual sample was
analyzed to obtain the average value and standard deviation.
Samples Time
(days)
[Mg]
(mM) 26
MgDSM3 2 s.d.
Exp. Olivine -0.35 ± 0.10 (n=8)
Exp. 3 (no NaCl) 0.04 7.36 -0.71 0.21 (n=7)
0.13 9.30 -0.64 ± 0.12 (n=4)
0.29 10.12 -0.58 ± 0.09 (n=3)
0.63 13.92 -0.52 ± 0.21 (n=5)
0.96 12.40 -0.60 ± 0.14 (n=3)
1.44 13.68 -0.58 ± 0.18 (n=3) 2.00 16.77 -0.48 ± 0.06 (n=3)
2.52 15.80 -0.47 ± 0.27 (n=6)
Standards
CAM 1 -2.60 ± 0.17 (n=32)
BHVO-1 -0.19 ± 0.09 (n=7)
BIR -0.25 ± 0.03 (n=2)
San Carlos
Olivine -0.75
±
0.09 (n=5)
SU-1 -0.55 ± 0.14 (n=76)
148
0.09 (n=7) within the previously reported values of -0.30.08 (Huang et al., 2009)
and -0.085 (Baker et al., 2005). Measured 26
Mg of BIR-1 was -0.25 0.03 (n=2),
which is consistent with the value of -0.23 0.23 (Wombacher et al., 2009) and -0.29
(Baker et al., 2005). Measured San Carlos olivine 26
Mg was -0.75 0.09 (n=5), in
agreement with other reported values of -0.73 0.06 (Teng et al., 2007). In all cases,
the measured 26
Mg values were within the uncertainty of the literature values. Each
sample reported in Table A1.2 was measured 3-7 times during different analytical
sessions, and the average values and standard deviations are reported along with the
number (n) of times that an individual sample was analyzed. Because of the small
volume of samples, duplicate samples of experiments could not be prepared and thus
the reported uncertainties reflect only the measurement uncertainty.
A1.3. Results
A1.3.1 Elemental concentrations and Mg isotope results
The evolution of Mg and Si concentrations and pH over time for the well-
mixed experiments 1 and 2 (with 0.5 M NaCl) and experiments 3 and 4 (no NaCl) is
summarized from Chapter 2 and shown in Figs. A1.1A-C. Early experimental stages
(ca. < 10 hours) are characterized by incongruent dissolution with preferential release
of Mg over Si. At longer times, the Mg release rate is constant while Si concentrations
plateau due to precipitation of SiO2(am) (Chapter 2). Magnesite saturation was
reached between 10 and 25 days in the experiments with NaCl (experiments 1 and 2)
and at 90 days in the absence of NaCl (experiment 4), although magnesite
precipitation did not occur in measurable amounts until after about 75 days in the
NaCl experiments (Chapter 2). The alkalinity increased over the duration of each
experiment in stoichiometric relation to the amount of olivine dissolution. Early and
long-term rates are markedly different between the experiments with and without 0.5
M NaCl. As suggested by the data in Fig. A1.1, olivine dissolution rates under the
closed-system conditions do not exhibit the pH dependence over the pH range 4 to 7
observed in experimental flow-through studies (e.g., Pokrovsky and Schott, 2000b).
The weak dependence on pH is an important feature of the data explored in the
149
subsequent modeling efforts.
The second series of experiments (i.e. well-mixed experiments 2R and 19R,
and unmixed experiments 2S and 19S, all with 0.5 M NaCl) are shown in Fig. A1.1D-
E. These experiments were conducted over shorter durations to examine the evolution
of the Si-rich layer under both well-mixed conditions, and in the absence of
mechanical mixing. Complementary high-resolution TEM analyses after 2 and 19 days
Figure A1.2. Magnesium isotope data for the short-term experiment 3 (no NaCl) over 3
days of reaction time. Error bars represent 1.s.d.
Figure A1.1. Summary of previous
experimental data from Chapter 2 in (A)-(C)
and Chapter 3 in (D)-(E). Vertical arrows in
(A) and (D) indicate time when experiments
reached MgCO3 saturation, horizontal line
with arrows indicates interval when MgCO3
precipitation is thought to have occurred in
experiment 2, based on the lack of
precipitates in experiment 1 at 74 days.
Horizontal dashed lines in (B) and (E)
indicate saturation with respect to amorphous
silica. Error bars represent 1 s.d.
150
of reaction constrain the evolution of the thickness and composition of the Si-layer and
olivine surface (Chapter 3). The well-mixed experiments (2R and 19R) agree well
with experiments 1 and 2, which were conducted under similar conditions. The poorly
mixed experiments show a similar transient rapid increase in Mg release over Si,
followed by near steady long-term dissolution, but at much slower rates.
Results of Mg isotopic measurement of WA olivine and the samples from
experiment 3 are presented in Table A1.2. All fluid samples from experiment 3 were
isotopically enriched in 24
Mg (26
Mg from -0.71 to -0.47) compared to WA Olivine
(26
Mg of -0.35). The 26
Mg of the first fluid sample has the lowest 26
Mg value, and
as the experiment progressed the dissolved 26
Mg values increased progressively such
that the 26
Mg of the final sample is within uncertainty of the value of unreacted WA
Olivine (Fig. A1.2). The Mg isotopic measurements of samples from experiment 3
are consistent with the preferential release of 24
Mg during olivine dissolution. The
addition of isotopically light Mg during initial phases of dissolution is consistent with
prior results from flow-through experiments (Wimpenny et al., 2010).
A1.4. Discussion
A1.4.1 Conceptual framework and spatial considerations for olivine dissolution
The experimental results considered here represent an array of conditions
likely to occur as CO2-rich fluids react with lithologies rich in olivine. The first set of
experiments detailed in Chapter 2 (experiments 1-4) considered the effect of
supporting electrolyte and found that in the presence of 0.5 M NaCl the initial olivine
dissolution rates were slower, while the long-term rates were more rapid, compared to
the experiments with no electrolyte in the initial solution. We present new 26
Mg data
for the electrolyte-free experiment. The second set of experiments (2R, 19R, 2S, 19S)
detailed in the first part of this series (Chapter 3) compared the results of experiments
conducted under well-mixed and poorly mixed conditions. In the mixed experiments,
dissolution rates were approximately five times faster compared to the poorly mixed
experiments. Cross sections of reacted olivine grains after 2 and 19 days of reaction
were also analyzed for crystallinity and chemical composition using HR-TEM. A
151
complex sequence of Mg-depleted surface layers were observed: (1) olivine which is
crystalline, but depleted in Mg relative to the bulk olivine composition, (2) an
amorphous “active layer” (AL) where Mg/Si changes rapidly with depth, and (3) a
newly precipitated amorphous layer where Mg/Si < 0.3 (Fig. A1.3). The Mg-depleted
olivine was most pronounced in the mixed experiments, where Mg depletion was
observed over zones extending up to 80 nm within crystalline material (Fig. A1.3C).
In poorly mixed experiments this zone was less than 20 nm thick (Chapter 3). The
active layer was present on all reacted forsterite examined, and was of a similar
thickness in the HR-TEM profiles from all four experiments. In contrast, the
precipitated layer was only visible in experiments that exceeded SiO2(am) saturation.
Given the 20 to 40% porosity observed in TEM images (Chapter 3), the precipitated
layer does not form an appreciable barrier to ion transport from the active layer to the
bulk fluid.
The entire sequence of alteration, from Mg-depleted crystalline olivine to the
precipitated Si layer, is consistent with the both the development of a non-
stoichiometric “leached layer” (e.g., Casey and Bunker, 1990; Casey et al., 1988;
Chou and Wollast, 1984; Hellmann, 1997; Luce et al., 1972; Stillings and Brantley,
1995), and a process of interfacial dissolution-reprecipitation (Hellmann et al., 2012;
King et al., 2010; King et al., 2014). However, given the stratigraphy evident in the
alteration layer, including a sharp transition between amorphous and crystalline
material and multiple chemical gradients indicated by the Mg/Si (and Fe/Si) profiles,
Chapter 3 further subdivided the altered layer into the zones defined above and in Fig.
A1.3. In the development of the kinetic model, we make additional simplifications.
First, we assume an average value for Mg/Si within the active layer that evolves from
an initial stoichiometric value to the value required by mass balance constraints
imposed by time resolved solution compositions and the HR-TEM profiles (Fig.
A1.3C). The initial development of the active layer should preferentially contribute
26Mg to solution, resulting in lower dissolved
26Mg relative to the bulk olivine as
observed. Second, we do not explicitly treat the process of solid-state diffusion (e.g.
Chapter 3, Hellmann et al., 2012; Yeng et al., 2009) implied by the complex Mg/Si
152
Figure A1.3. Conceptual model for the surface kinetic model. (A) High resolution TEM image
shows the general features observed in the outer 100 nanometers of olivine after reaction for 2
days (modified from Chapter 3), with the fluid-solid interface to the left of the image. Average
Mg/Si ratios measured by EDS are provided for reference. The black arrows indicate the
migration of the active layer into the crystalline olivine over time, as determined by the rates of
the individual reactions in (B). (B) The conceptual model for the independent rates considered in
the model and the resulting mass balance constraints. The net rate (Rnet) is the Mg exchange rate
plus the net difference between the net <SiO2 dissolution rate and the net precipitation rate of
amorphous silica. Magnesium is removed from the active layer of volume VAL, while >SiO2 can
only be dissolved or added from the surface of the active layer. (C) Representative HR-TEM EDS
linescan of the Mg/Si ratio in reacted olivine from experiment 19R from Chapter 3 with a Lowess
curve fit (green line) compared to the model representation (bold teal line). The grey shaded
region (indicated (-)) shows the Mg depletion relative to the assumed stoichiometric composition,
while the green shaded region (indicated (+)) shows the Mg-enrichment in the active layer
relative to model Mg/Si. To maintain mass balance the total Mg in the two zones must be equal.
153
gradients in both the amorphous active layer and the Mg-depleted olivine. Instead, we
assume a stoichiometric flux of Si and Mg into the active layer and account for the
total Mg depletion within the active layer. The consequences of these assumptions will
be discussed in a subsequent section. Conceptually, the active layer can be viewed as a
reaction front that maintains a constant thickness as it advances into the unreacted
olivine. This thickness is moderated by the balance between the removal of ions at the
fluid-active layer interface and the supply of ions from the crystalline olivine.
In summary, the key features of olivine dissolution we seek to address through
development of an alternative kinetic model include: (1) non-stoichiometric release of
Mg relative to Si during the early stages (ca. <1 day of reaction time); (2)
enhancement of the Mg release rate at early times in the absence of an electrolyte,
followed by lower steady-state dissolution rates compared to experiments with an
electrolyte present, (3) preferential release of 24
Mg to solution during the early stages,
(4) weak dependence of the dissolution rate on pH under closed-system experimental
conditions, and (5) a zone of Mg depletion between the fluid-solid interface and the
crystalline olivine. As many of these features would not be captured by existing rate
laws, we have developed a framework that considers the exchange of individual ions
within the active layer and between the surface of the active layer and the bulk
solution.
A1.4.2 Spatially resolved surface kinetic model for olivine dissolution
A1.4.2.1 Stoichiometric and mass balance constraints
For olivine, where octahedrally coordinated Mg is linked to isolated silica
tetrahedra, protonation of the magnesium surface sites is required to release
magnesium from the bulk crystal (for clarity, Fe is omitted below). The hydrolysis
reaction requires the exchange of four protons for two magnesium ions at the surface
of the mineral, followed by polymerization of the isolated surface silica tetrahedra
(Pokrovsky and Schott, 2000b), according to the following net reaction:
(1)
>Mg2SiO4 + 4H + ¾®¾ > SiO2 + 2H2O + 2Mg2+
154
where > represents a site within the active layer. At low temperatures, reaction (1) is
considered irreversible, and should depend on pH. However, the exact dependence is
difficult to constrain as protons may also be distributed between surface sites and/or
adsorbed to surface layers (Pokrovsky and Schott, 2000). Diffusion of Mg out of the
active layer is not the rate-controlling step for reaction (1) under well-mixed
conditions because the steady-state dissolution rate is independent of Mg
concentration in solution (Chapter 3).
Although reaction (1) has been included in previous models, it has not been
explicitly linked to the compositional evolution of the surface. Here, reaction (1) only
occurs within the volume (VAL, m3) of material defined as the active layer (Fig.
A1.3B), as suggested by both the HR-TEM results and the early non-stoichiometric
dissolution observed in the experiments. Penetration of protons into the olivine
structure at depths greater than one unit cell is further suggested by x-ray
photoelectron spectroscopy (XPS) of the olivine surface (Chapter 2; Pokrovsky and
Schott, 2000a). The VAL is calculated as FoAhVm, where Fo is the site density of
olivine (3.3 x 10-5
mol/m2 Wogelius and Walther, 1991), A is the physical surface
area in m2
from measured BET surface area and the mass of initial olivine, Vm is the
molar volume of olivine (4.38 x 10-5
m3/mol) and h is the number of olivine layers
from which Mg can be extracted according to equation (1). A list of parameters is
presented in A1.3. The equivalent thickness of the active layer is 20 nm (h ≈14) with
a steady-state Mg/Si of 0.3 to 1.5 determined by (1) the rates required to match the
measured Mg and Si release rates of Fig. A1.1 and (2) mass balance considerations
based on HR-TEM imaging of olivine cross-sections shown conceptually in Fig.
A1.3C.
As the active layer forms, residual silica populates the volume that was once
crystalline olivine. The volume change from crystalline olivine to amorphous silica,
on a per mole basis, is approximately 51%. Although volumetric expansion mostly
likely occurs during the formation of the active layer, we cannot constrain the extent.
Therefore, the overall mass balance is calculated on a per mol basis so the potential
volume change does not impact the model results. Our calculation of VAL based on an
155
active layer thickness of 20 nm is consistent with a volume increase of about 30%
based on the HR-TEM results. Active layer thickness is provided hereafter only as a
guide.
Once the surface silica sites are formed, the dissolution of silica can be treated
as a reversible reaction:
(2)
Unlike exchange reaction (1) above, removal of silica is largely confined to the
surface of the active layer, Fo A, rather than VAL (Fig. A1.3B). This spatial constraint
is required to reproduce the observed non-stoichiometric dissolution and to allow for a
steady state active layer thickness to emerge. Mathematically, the active layer can be
viewed as a reaction front that moves inward at a rate controlled by both the rate of
Mg exchange according to reaction (1) and the net release of Si via reaction (2).
Closed-system experiments, such as those reported here, show a plateau in Si
concentrations near SiO2(am) saturation, while Mg continues to increase, supporting
the presence of amorphous silica precipitates. Precipitation of SiO2(am) can occur
directly on the mineral surface or in other parts of the reaction vessel as shown by
reaction (4):
H4SiO4 aq( )¾®¾¬¾¾ > SiO2 + 2H2O (3)
This local equilibrium controls the aqueous Si (i.e. H4SiO4(aq)) at later stages of the
experiments once SiO2(am) saturation is exceeded (e.g., Fig. A1.1).
Reactions (1) through (3), combined with the conceptual model of Fig. A1.3
emphasize the coupled reactions that occur during the initial and steady-state periods
of olivine dissolution. The rate of reaction (1) and the net rate of reaction (2) can be
considered to control the dissolution rate of olivine, while reaction (3) represents an
indirect control on both the rates of reactions (1) and (2) at greater extents of reaction
progress as it determines the aqueous Si concentrations once amorphous silica
saturation is reached. Although previous models have considered such coupling
(Pokrovsky and Schott, 2000b), the spatial relationships between the reactions have
not been explicitly accounted for. We hypothesize that the spatial relationship between
the reactions explains the majority of the dissolution behavior outlined above.
> SiO2 + 2H2O¾®¾¬¾¾ H4SiO4 aq( )
156
A1.4.2.2 Rate equations and inclusion of isotopes as additional constraints
Similar to the assumptions commonly made to describe mineral dissolution
rates, most previous surface isotope exchange models assume the surface layer
composition is at steady state (DePaolo, 2011; Nielsen et al., 2012). Here, the
development of the active layer results in a surface layer multiple unit cells thick that
is evolving isotopically and compositionally. Such transient behavior is difficult to
solve for analytically. Thus, we use a numerical model to track the fluid and solid
compositions over time given the coupling among reactions (1), (2) and (3) enforced
by the spatial constraints and the kinetic framework presented below. The approach is
conceptually similar to that of Brantley et al. (2004) and Wiederhold et al. (2006).
However, previous studies were not able to precisely constrain the size of the active
layer, or did not require it.
In order to develop a kinetic model for the reactions (1), (2) and (3), that
includes a provision for kinetic fractionation of Mg isotopes during the early stages of
dissolution, we used a set of relationships that are based on the “ion-by-ion” growth
model developed for crystal growth and applied to calcite precipitation in Nielsen et
al. (2013; 2012). In this approach, the olivine dissolution rate is described by the net
rate Rnet, which is the difference between the ion fluxes from (Rb) and to (Rf) and the
surface, where the subscript b refers to the backward or dissolution reaction and
subscript f refers to the forward or precipitation reaction (DePaolo, 2011). As reaction
(1) is treated as irreversible, the flux of Mg is only characterized by Rb, providing an
opportunity to quantify both Rb and the associated kinetic isotope effect. In contrast,
reactions (2) and (3) are assumed to be reversible and thus only the net Si release is
observable.
The isotope-specific Mg exchange rate (RMg,, mol/s) from the active layer
associated with reaction (1) is a function of the rate coefficient for Mg exchange, Mg
(s-1
) and the concentration, Pi (mol/m3) of a given isotope or solid component, i, in the
surface layer:
ALMg
n
HMgMg VPaR 2424 (4)
The dependence of Mg exchange rate on pH is expressed via a dependence on the
157
proton activity (aH+n), where n is between 0 and 0.5 (Pokrovsky and Schott, 2000a;
Rimstidt et al., 2012). The value of n is difficult to assess because of the variable
mechanisms for proton uptake, so we used the value 0.37 suggested by Rimstidt et al.
(2012) based on synthesis of a large number of experimental studies. Similarly, for the
minor isotope the rate equation can be written in terms of the common isotope using
the kinetic fractionation factor between Mg2+
in olivine and in solution (Mg-Fo), to
describe the preferential removal of 24
Mg:
ALMg
n
HMgFoMgMg VPaR 2626 (5)
The exchange rate coefficient, Mg, along with Mg-Fo constitute the unknowns in the
rate equations. Equation (1) is used to represent the stoichiometry for mass balance
purposes. Although we track the individual rates, VAL is originally determined based
on the molar volume of olivine, and hence it is equivalent to calculating the rates in
terms of mole fraction and the number of original moles of olivine in the active layer.
Thus, any positive volume change associated with reaction (2) does not impact our
mass balance results.
Although the isotopes of Mg are not required to model olivine dissolution
using our approach, they provide a constraint on the rate of reaction (1). The model
approach we present may also be useful for considering other kinetic isotopes effects
(e.g., Pearce et al., 2012). However, to model isotopic fractionation introduces
additional calculations. First, we calculate PMg24 initially as a function of the mole
fraction of Mg that is 24
Mg, XMg24, and the stoichiometric Mg:Si ratio in olivine, XMg/Fo
(1.84 for the olivine considered here):
1
/2424 )(
mFoMgMgMg VXXinitialP (6)
An analogous equation describes PMg26. As the surface layer evolves, PMg,i is tracked
according to the individual reaction rates. In addition to the coupling between Mg and
Si associated with reaction (1), the removal of >SiO2 sites allows the reaction front to
migrate inward, supplying Mg to the active layer. Thus, PMg,i is not only a function of
the magnesium release rate but also the net rate of reaction (2). To describe the net
rate of silica release (R>SiO2) from the active layer associated with reaction (2), we
assume that the backward (detachment) rate is a function of the concentration of
158
>SiO2 at the surface only, while the forward (attachment) rate is a function both >SiO2
and the aqueous Si activity, [H4SiO4]:
mFoSiOSiOSiOSiOSiO AVPSiOHkPR 2442222 (7)
where SiO2 (s-1
) is the surface detachment rate and P>SiO2 is the concentration of Si
surface sites throughout VAL and k>SiO2 (M-1
s-1
) is the rate constant associated with
attachment of SiO2 at the surface. The key difference between equation (7) and
equations (4) and (5) is that the thickness of the surface layer is not included in this
rate expression because we assume that the attachment and detachment of Si can only
occur from the surface and not the entire volume of the active layer (i.e. h = 1). This
treatment allows for incongruent dissolution during the active layer formation, and
maintains a constant active layer thickness as suggested by HR-TEM analyses.
Equation (7) is computed for the Si associated with each Mg isotope. Initially P>SiO2 is
assumed to be zero, and is sequentially updated based on the rates from equations (4),
(5), (6) and (7).
The rate of new silica precipitation is described according to the equilibrium
constant for SiO2(am), or KSiO2, and the precipitation rate kSiO2 (mol m-2 s-1):
1
2
44
2)(2
SiO
SiOamSiOK
SiOHAkR (8)
To calculate both the non-steady state evolution of the surface layer, and the spatial
relationship between the ion-specific reactions, we couple the rate formulations above
to the active layer concentrations (Pi) according to:
ALFoMgMgSiOMg
MgVXXRR
dt
dP//24224
24
(9)
ALFoMgMgSiOMg
MgVXXRR
dt
dP//26226
26
(10)
ALSiOFoMgMgMg
SiO VRXRRdt
dP// 2/2624
2
(11)
In equations (9) and (10) the concentration of Mg in the active layer is determined by
the balance between the rate of Mg exchange and the detachment rate of >SiO2, as
required to maintain the constant VAL suggested by HR-TEM characterization.
159
Similarly, the concentration of >SiO2 is controlled by the difference between the rate
of Mg exchange, which generates >SiO2, and the net removal of >SiO2 from the
active layer. These equations emphasize the coupling between the Mg and Si removal
rates and the distribution of Mg and Si in the altered layer.
The corresponding active layer Mg/Si (Mg/SiAL) can be calculated from:
(12)
This approach uniquely allows us to track the Mg/Si and 26
Mg in the active layer and
the fluid without assuming a steady-state surface composition.
The coupled set of equations represented by equations (1) through (12),
combined with the constraints on the active layer provided by isotopic data, HR-TEM
analyses, and time-resolved solution compositions, results in the following adjustable
parameters that are fit using the experimental data (1) Mg (and Mg, which depends on
Mg), and (2) SiO2 and k>SiO2, which determine R>SiO2. The value of kSiO2 cannot be
precisely determined because we do not know the surface area associated with
SiO2(am) precipitation, although the long-term steady-state Si concentrations provide
a useful constraint on the overall rate.
A1.4.3 Application of model to experimental data
To model the closed-system experimental data of Chapters 2 and 3, we use a
forward model of the coupled set of reactions/equations (1)-(12) using the reactor
volume, the initial mass of olivine, homogenous stoichiometric olivine with Mg/Si of
1.84, and the measured BET surface area as initial conditions (see Table A1.3 for
parameter values and sources). We do not consider aqueous or solid state diffusion.
Alkalinity is calculated based on the Mg release rate and is used to solve for pH
according to “apparent” equilibrium constants for CO2 solubility and carbonate
speciation in the presence or absence of 0.5 M NaCl and at appropriate temperature
and pressures. Equilibrium constants are determined using the approach of Duan and
Sun (2003). We also assume unit activity coefficients for aqueous species.
Mg
Si AL(t) = PMg( ) t( ) / PMg t( ) / XMg/Fo( ) + P>SiO2 t( )é
ëùû.
160
Because the early release rates of both Mg and Si are controlled by the rate of
reaction (1), Mg was adjusted to match the early (< ~2 to 3 days) experimental Mg
and Si profiles. Then, because the net removal of >SiO2 according to reaction (2)
controls the supply of the Mg to the active layer via equations (9) and (10), R>SiO2 is an
additional control on Mg concentrations at steady state. As Si concentrations plateau
once the active layer is formed due to SiO2(am) precipitation, the early SiO2 and long-
term Mg profiles constrain the net >SiO2 dissolution rate. Accordingly, the rate was
adjusted to match the long-term Mg profile by varying the ratio of SiO2 to k>SiO2. The
model fit to experiments 1 and 2 is considered the “reference model” hereafter, with
parameters provided in Table A1.4. The reference model was then adjusted to fit the
additional experimental data following the above procedure.
The model is able to reproduce the major features of the data from experiments
1- 4, including the initial incongruent dissolution (Fig. A1.4). The Mg/Si in the active
layer evolves from 1.84 to 0.5 after 5 days (the isotopic and surface layer evolution
Table A1.3: Model parameters and values applied to experiments 1-4, 2R, 19S, 2S, 19S.
Parameter Description Units Value or range
Vm Molar volume m3 mol-1 4.38 x 10-5
Fo Site density mol m-2 layer-1 3.3 x 10-5 a
h Number of layers --- 14b
A Surface area m2 2.24c/2.0d
VAL Active layer volume
=FoA hVm
m3 calculated
RMg,i Mg isotope release rate mol s-1 calculated
R>SiO2 Net SiO2 release rate mol s-1 calculated
RSiO2(am) Precipitation rate mol s-1 calculated
Mg Exchange rate coefficient for
Mg
s-1 variable
SiO2 Detachment rate coefficient for
>SiO2
s-1 variable
k>SiO2 Attachment rate coefficient for >SiO2
M-1s-1 variable
kSiO2 Rate coefficient for SiO2(am) mol m-2 s-1 variable
XMg/Fo Mole fraction Mg in olivine 1.84c
XMg,i Mole fraction of i in Mg
component
calculated
Mg-Fo Kinetic fractionation factor --- variable
PMg26, PMg24 Concentration in VAL mol m-3 calculated
P>SiO2 Concentration of >SiO2 in VAL mol m-3 calculated
KSiO2 Equilibrium constant for
SiO2(am)
--- 10-2.4067e
161
will be discussed in more detail in a subsequent section). Relative to experiments 1
and 2, the early Mg release rate in the electrolyte-free experiments 3 and 4 was more
rapid, followed by slower long-term rates. To fit the data requires an increase in Mg
by a factor of 3 and a decrease in R>SiO2 by a factor of 4.5. Although we cannot
determine uniquely whether the later decrease was associated with the attachment or
detachment rate of >SiO2, to account for the lower Si concentrations appears to
require an increase in k>SiO2 by a factor of 50 because increasing k>SiO2 more strongly
impacts the steady-state Mg release rate, while decreasing >SiO2 impacts both early
and steady-state Mg and Si rates (cf. Fig. A1.1). The Mg isotopes also place
constraints on Mg as discussed in section 4.4. However, the rapid increase in Mg
concentrations appears to require an increase in the thickness of the active layer (AL)
relative to the experiments that contained a supporting electrolyte (Fig. A1.4C and D).
These changes in rate coefficients result in a decrease in the overall olivine dissolution
rate from 10-13.1
to 10-13.75
mol/cm2/sec between the reference model and the
corresponding model for experiments 3 and 4 (Table A1.4). The rate constant for
SiO2(am) precipitation of 3.3 x 10-9
mol/m2/sec, using the olivine surface area, is also
in agreement with previously published values at neutral pH (3.1 x 10-9
mol/m2/sec at
60˚C (Rimstidt and Barnes, 1980)).
The decrease in R>SiO2 in the absence of an electrolyte is consistent with the
observed decrease in rates of SiO2(am) dissolution with decreasing electrolyte
concentration (Icenhower and Dove, 2000). As noted in Chapter 2, the solubility of
silica may also be impacted by the absence of a supporting electrolyte. Explaining the
increase in the Mg rate is more difficult. The enhanced exchange of Mg in the absence
of an electrolyte is surprising as sodium is thought to modify the nucleophilic
properties of water (Icenhower and Dove, 2000), and a reduction in ionic strength
generally results in an increase in the thickness of the electrical double layer (Brown
and Parks, 2000). It is possible that the high concentrations of Na+ may interfere with
with proton exchange. The more rapid Mg exchange rate also generates more
depletion of Mg in the active layer leading to model Mg/Si of 0.05 in experiments 3
and 4, compared to 0.5 in experiments 1 and 2. Thus, the lower R>SiO2 and increase in
162
>SiO2 concentration in the active layer associated with more rapid Mg exchange
create a feedback that reduces the long-term net dissolution rate substantially, despite
the lower pH in these experiments. To determine the precise nature of the dependence
of Mg and >SiO2 on supporting electrolyte concentration would require a broader
range of experimental conditions.
Experiments 2S, 2R, 19S and 19R, conducted with larger olivine grains, show
Figure A1.4. Model fit to experiments 1-4. Magnesium (open symbols) and Si (closed symbols)
over the full experimental duration are shown in (A), and during the initial incongruent
dissolution phase in (B). The model fit to the combination of experiments 1 and 2 in (A) is
considered the reference model. The same scheme is applied in (C) and (D) where model results
for experiments 3 and 4 are presented. The changes in rate coefficients are indicated and the
thickness of the active layer (AL, nm) is indicated. For experiments 3 and 4, a sensitivity analysis
showing variation in the AL is provided. All experiments were conducted at 100 bar CO2 and
60˚C in the presence of 0.5 M NaCl (Chapter 2) .
163
similar behavior and are well described by the model after adjustment of some rate
parameters (Fig. A1.5). For experiments 2R and 19R, the data are best fit with a Mg
that is a factor of 3 greater than in the reference model but within the uncertainty,
while R>SiO2 and the associated Si rate parameters remained consistent with the
reference model. Overall rates are comparable to those determined for experiments 1
and 2. In contrast, for the poorly mixed experiments (2S and 19S), both Mg and
R>SiO2 are slower by a factor of 6 and 10, respectively, compared to the well-mixed
experiments. The experiments also yield net dissolution rates that are nearly an order
of magnitude slower than those determined for the rocking experiments. This
difference in rate may reflect chemical gradients that decrease both transport of H+ to
Figure A1.5. Model fit to the combined experiments 2R and 19R (well mixed) are shown
over the full experimental duration (A), and during the initial incongruent dissolution phase
in (B). Model fit to experiments for 2S and 19S (not-well mixed) over the same durations are
shown in (B) and (C). The changes in rate coefficients relative to the reference model are
indicated. All experiments were conducted at 100 bar and 60˚C in the presence of 0.5 M
NaCl (Chapter 3).
164
the surface and transport of SiO2 away from the surface, and are not captured using the
bulk solution properties. Under these conditions, the calculated rates are effective rates
that include transport effects. The data and modeling approach presented here do not
allow us to constrain the full dynamics of transport effects (e.g., Li and Steefel, 2000;
DePaolo, 2011). More discussion of diffusion limitation is provided in Chapter 3.
A1.4.3.1 Overall dissolution rates and comparison to previous studies
The model rate profiles for net Mg and Si release (i.e. R>SiO2 – RSiO2) are shown
as a function of time, pH, and H4SiO4 in Fig. A1.6 for experiments 1 and 2. Early in
the experiments, Mg release rates are substantially greater than Si because the rate of
Si dissolution requires formation of >SiO2 sites via the Mg exchange reaction. The
rapid early increase in >SiO2 from the start of the experiment to approximately 1 day
occurs because P>SiO2 increases more quickly than aqueous SiO2 concentraitons, where
the later controls the attachment rate. Because of the feedback between Mg release,
the creation of >SiO2 in the active layer and aqueous Si concentrations, the rates
eventually stabilize at a steady state characterized by stoichiometric dissolution, with
rates in agreement with those calculated from a linear fit to the data in Chapter 2.
Although there is a weak dependence on pH incorporated into the model, the
relationship between rates and pH in Fig. A1.6 is partly an artifacy of the correlation
between reaction progress and pH (Fig. A1.1). Experimental studies under flow-
through conditions generally observe that the overall rate of olivine dissolution
decreases with increasing pH, suggesting that Si concentrations in these experiments
were sufficiently low that Mg release rates have a stronger effect on the net dissolution
rate. As discussed in Chapter 2, a correlation between rate and pH is not observed in
the experiments presented here, likely because at late stages (after several days) the
net dissolution rate is controlled by both R>SiO2 and RMg.
Flow-through and closed-system experiments have shown an inverse
correlation between the aqueous Si concentration and both Mg and net dissolution
rates. For olivine, the breaking of Si-O bonds at low pH is typically designated as the
rate limiting reaction. Our modeling results suggest that the overall dependence of the
rate on Mg and Si dynamics is complex (Fig. A1.7). During the formation of the
165
active layer, the Mg release rate initially controls the Si release rate. Once the active
layer is established, then the net destruction of >SiO2 becomes more important, such
that R>SiO2 determines both the Mg release rate and the steady-state dissolution rate at
high degrees of Mg depletion in the surface layer, or efficient Mg exchange (e.g., Figs.
A1.4A compared to Fig. A1.4C). At lower Mg exchange rates, such as observed in
the electrolyte-free experiments, Mg exchange continues to moderate the overall
dissolution rate by regulating the re-polymerization of silica.
Figure A1.6. Rate profiles as a function of (A) time; (B) pH; and (C) H4SiO4
concentrations for the reference model. The bar at top indicates the dominant
processes occurring from initial to steady-state dissolution.
166
The assignment of single reactions as rate limiting thus appears to be of limited
utility, even for a relatively simple silicate such as olivine. The spatially resolved
surface kinetic description affords an alternative insight into how the kinetics of the
various reactions and the solution stoichiometry control the overall rate. One key
assumption of the model, that appears to be supported by the elemental and isotopic
data, as well as the TEM analysis, is the presence of an active layer. Conceptually, this
may consist of several layers of altered material where solid-state diffusion of Mg and
H+ are rapid enough to maintain the rates observed (Chapter 3), or alternatively, for
other minerals it may represent kink sites and other surface defects that allow for
depletion of one element over the other in excess of what is allowed by the BET
surface area and one monolayer of mineral. Regardless of the exact mechanism, the
coupling imposed by reactions (1), (2), and (3) exerts a strong control on the overall
rate. This is shown conceptually by individually varying the individual rate constants
and running the model to steady-state rates (Fig. A1.7). As Mg increases, the
sensitivity of the overall dissolution rate to the Mg exchange rate decreases. This is
because at high Mg, the supply of Mg to the active layer, where it can exchange with
two H+, is determined by the removal of >SiO2. Thus, there is a stronger positive
relationship between SiO2 and the overall dissolution rate, and the inverse is true for
k>SiO2. The initiation of a plateau in the overall rate at high SiO2 is presumably
because Mg exchange eventually becomes limiting. Thus, under flow-through
conditions where H4SiO4 is continually removed, the >SiO2 attachment rate may
become so small that the overall rate depends more strongly on Mg. Finally the
dependence of the overall rate on kSiO2 is weaker, but occurs because net SiO2(am)
precipitation impacts R>SiO2 by controlling the attachment rate of >SiO2 once
SiO2(am) saturation is reached.
A1.4.3.2 Consequences of spatial discretization and implications for rate coefficients
The determination of all kinetic parameters is subject to assumptions made
regarding the reactive surface area, and other factors that contribute to the observed
overall rate (e.g., the dependence of overall rate on pH, reaction affinity and ionic
167
strength). Here, the observed Mg release rate was reproduced according to the values
assigned to represent the VAL, Mg and PMg (which is partly a function of R>SiO2). As
noted previously, the HR-TEM profiles of Chapter 3 show remarkable complexity,
including a zone of Mg-depleted crystalline olivine 80-100 nm thick adjacent to the
active layer in the mixed experiments, and approximately 10-20 nm thick in poorly
mixed experiments. In contrast, the boundary conditions in the model assume that
stoichiometric olivine was supplied to the active layer as >SiO2 was removed from the
surface. This assumption was necessary to avoid solving explicitly for solid-state
diffusion. Below we qualitatively describe the uncertainty arising from this
assumption. In addition, we find that the rate coefficients differ in the presence or
absence of a supporting electrolyte. As a consequence, the rate parameters provided in
Table A1.4 should be considered “effective” rate coefficients applicable to the
specific experimental conditions.
Although the model reproduces the solution composition convincingly, we
further evaluate the solid phase mass balance because the solid phase stoichiometry is
coupled to the rate coefficients via equations (9)-(11). In order to calculate the
maximum amount of Mg removal from the HR-TEM profiles, we assume constant
volume replacement and that Si is conservative throughout the altered layers (cf. Fig.
Figure A1.7. Sensitivity of the steady-state dissolution rate to the individual rate coefficients. Each
rate coefficient was varied independently while the others were held constant at the values
determined for the reference model (see Table 4). The best fit to experiments 1 and 2 is shown by the
grey bar.
168
A1.3C). The resulting difference between the HR-TEM and the model is -24% and
12% for 19R and 19S respectively, which is reasonable given that the HR-TEM
calculations represent a maximum estimate of Mg depletion. The consequence of
preserving the mass balance through averaging the active layer concentration is that
the value of PMg is artificially low (and P>SiO2 high) and concentration gradients are
not honored. The rate of Mg exchange also differs between the Mg-depleted olivine
and the active layer. These underlying assumptions are embedded in the effective rate
coefficients. However, effective rate coefficients can still be useful if they capture the
key processes: the model successfully reproduces both the widely observed initial
incongruent dissolution and steady-state dissolution, as well as the formation of
altered surface layers. For olivine, given that most of the Mg depletion occurs within
the active layer, where Mg/Si ratios drop from ca. 1.5 to 0.2 over approximately 30
nm, our treatment of only the active layer appears reasonable. However, more studies
that combine HR-TEM constraints on composition and crystallinity of the surface as a
function of depth, combined with spatially resolved models that account for solid-state
interdiffusion (e.g., Yang et al., 2009), may yield further insights into the importance
of the Mg-depleted crystalline zone for both the evolution of the active layer and
overall dissolution kinetics.
A1.4.4. Isotopic fractionation and surface layer evolution during dissolution
The rapid increase in dissolved Mg relative to Si observed in the early stages
of the experiment is accomplished in the model through the creation of a Si-rich and
Mg-depleted active layer as indicated by HR-TEM analyses (Chapter 3). Figure
A1.8A shows the modeled change in Mg/Si of the active layer (Mg/SiAL) from
stoichiometric values of 1.84 to steady-state values of approximately 0.05. For
comparison, in experiments 1 and 2, with more rapid net dissolution rates but lower
Mg exchange rates, the Mg/SiAL is much higher and stabilizes at values closer to 0.3.
169
The 26
Mg of the active layer (26
MgAL) and dissolved Mg evolve in unison.
Values of Mg-Fo of between 0.9995 – 0.999 are consistent with the observed isotopic
variation in the dissolved Mg (Fig. A1.8B). In comparison to the dissolved Mg
profiles, the model 26
Mg of the surface layer is complex. During the initial stages of
reaction, Mg release is most rapid as indicated by the rapid decrease in Mg/Si and
corresponding increase in dissolved Mg concentration. Subtle variation in 26
MgAL
evident around 4 days when the Mg/Si depletion reaches steady state. Due to the
initially high Mg exchange rate, a maximum in R>SiO2 also occurs coincident with the
26
Mg maximum because P>SiO2 increases more quickly than aqueous SiO2
concentrations, where the later controls the attachment rate. This results in the early
enrichment of 26
Mg of the active layer. As the rate of Mg release decreases, an
increased supply of Mg to the active layer from the bulk crystal reduces the 26
MgAL
until eventually a steady-state balance develops between the net release of >SiO2 and
the supply of Mg. The slight drop in the 26
MgAL after 4 days is thus associated with
the attainment of a lower steady-state R>SiO2. Once steady-state dissolution is reached,
the dissolved Mg supplied to solution has the isotopic composition of the crystalline
olivine and the fractionation is no longer preserved in the fluid. In contrast, the surface
Figure A1.8. Model magnesium isotopic and surface layer Mg/Si evolution for experiment
3. (A) The model change in the Mg/Si of the active layer (Mg/SiAL) is shown in comparison
to the 26Mg of the active layer (AL) and the dissolved Mg. (B) Comparison between
model and measured 26Mg of the aqueous Mg (uncertainties are represented as 1 s.d.). The
best fit to the data is between Mg-Fo = 0.9995 and 0.9990 and variations in Mg-Fo are shown
for reference.
170
remains permanently enriched in 26
Mg.
Although the limited data in this study preclude full evaluation of the Mg
isotope systematics, the model suggests the isotopic composition of dissolved Mg
under closed system conditions is much more sensitive to variations in vMg compared
to variations in VAL, whereas the dissolved Mg responds identically to both. In
addition, because the early release rate of Mg determines the isotopic fractionation, the
isotopes are not appreciably sensitive to the value of SiO2. The sensitivity of the
isotopes to only one model parameter suggests that isotopic fractionation could
provide a useful tool for constraining individual ion exchange rates, even in the
absence of HR-TEM constraints. We also assume that no secondary processes such as
sorption influence the dissolved isotopic composition. As Mg would likely form weak
outer sphere complexes with amorphous silica resulting in negligible fractionation,
Figure A1.9. Model evaluation of Wimpenny et al. (2010) experimental series FO2 at
pH 3 and 25 ˚C. (A) Model fit for Mg and Si concentrations (see parameters in Table
4), (B) Comparison between measured 26Mg to model prediction using Mg-Fo = 0.999.
171
and we observe no other secondary phases in experimental products (Chapters 2, 3),
this assumption appears reasonable.
The experiments were also conducted only under closed-system conditions
where concentrations evolve with time. To explore the ability of the model to
reproduce observed Mg, Si and 26
Mg evolution under flow-through conditions, we
applied the model to data of Wimpenny et al. (2010) using their experimental
conditions and reported parameters. In the absence of HR-TEM constraints, we
assume an active layer thickness of 20-40 nm (an active layer greater than 12 nm is
required to reproduce the observed features). The model suggests a transient period of
dissolution in the early stages (<0.5 days) followed by attainment of steady-state
release (Fig. A1.9). A similar pattern is observed when the experimental pH was
subsequently lowered to 2 in the reactor, although we did not model this case because
of the uncertainty introduced by the previous experimental conditions. The model is
also consistent with the observed 26
Mg, although given the limited information and
number of measurements, the long-term behavior (i.e. attainment of bulk olivine
26
Mg values) and the value of Mg-Fo cannot be precisely constrained. Relative to the
closed-system experiments, the flow- through experiments are highly sensitive to the
Mg exchange rate and much less sensitive to the Si attachment and detachment
kinetics, as suggested in the earlier rate profile analysis (Fig. A1.9A and C).
It is likely that the kinetic isotope fractionation observed and modeled here
occurs during the dissolution of many minerals, although it may not be observable
under all conditions. However if several layers are active, then an isotopically depleted
layer may be maintained indefinitely on mineral surfaces. The slower the preferential
release rate of the element is relative to the migration of the active layer, the more
strongly the fractionation will be expressed within the mineral surface.
A1.5. Conclusions
Descriptive models of silicate mineral dissolution must be balanced against
their utility for describing complex multi-component open systems. Kinetic models for
silicate mineral dissolution have been largely based on a simplified version of
172
Transition State Theory (TST), where the overall dissolution rate is equal to the
product of the kinetic rate constant, surface area, the reaction affinity, and the activity
of ions that inhibit or accelerate dissolution. For complex silicate minerals, such as
olivine and many other major rock-forming minerals, the application of a single
kinetic rate constant and reaction affinity may not be appropriate if spatial constraints,
such as the formation of an active layer, determine the overall dissolution rate.
The model approach used here also suggests that no single reaction is rate
limiting, but rather the relationships between various components, over space and
time, determine the overall dissolution rate. For isolated SiO4 tetrahedra of olivine,
technically no Si-O bond has to be broken for the crystal structure to be disrupted.
Nevertheless, the dissolution of repolymerized SiO4 tetrahedra regulates the supply of
Mg to the active layer. Concurrently, the net exchange of Mg for protons controls the
rate of re-polymerization of tetrahedral SiO4, leading to net mineral dissolution. The
interdependency of these processes appears to explain the order of magnitude variation
in overall dissolution rates we observed and modeled. The possibility that similar
processes may occur in other silicate minerals is supported by the observation of
kinetic isotope effects during the dissolution of amphiboles and micas; however in
such cases the inter-relationship among individual components may be more complex.
The approach presented here represents a model of moderate complexity that
affords an evaluation of the coupling between the fluxes of individual ions and the
overall dissolution rate of olivine. Such an approach may be advantageous in
extracting rates from experimental data, as an interpretive tool for probing the
interconnection between key reactions. Although parameterizing the rate equations
does require more information, the approach presented here could still provide an
alternative rate law for geochemical and reactive transport models applicable at the
continuum scale (e.g., Druhan et al., 2013). Ideally, such approaches might ultimately
replace the simplified versions of net TST rate laws commonly used, leading to
improved models of mineral-fluid interactions across a range of geological
environments.
173
A1.6. Acknowledgments
We acknowledge the Global Climate and Energy Project (GCEP-48942) at
Stanford University for funding the work presented here. We are grateful to the Earth
Surface Geochemistry group at GFZ Potsdam for comments and discussions that
greatly improved the manuscript. We also acknowledge the staff of the Stanford
ICPMS, Environmental Measurements-1 and the Nanocharacterization Laboratories.
174
A1.7. References
Baker, J., Bizzarro, M., Wittig, N., Connelly, J., Haack, H., 2005. Early planetesimal
melting from an age of 4.5662 Gyr for differentiated meteorites. Nature 436,
1127-1131.
Bearat, H., McKelvy, M.J., Chizmeshya, A.V.G., Gormley, D., Nunez, R., Carpenter,
R.W., Squires, K., Wolf, G.H., 2006. Carbon sequestration via aqueous olivine
mineral carbonation: Role of passivating layer formation. Environ. Sci. Technol.
40, 4802-4808.
Brantley, S.L., Liermann, L., Bullen, T.D., 2001. Fractionation of Fe isotopes by soil
microbes and organic acids. Geology 29, 535-538.
Brantley, S.L., Liermann, L.J., Guynn, R.L., Anbar, A., Icopini, G.A., Barling, J.,
2004. Fe isotopic fractionation during mineral dissolution with and without
bacteria. Geochim. Cosmochim. Acta 68, 3189-3204.
Brown, G.E. Jr., Parks, G.A., 2001. Sorption of trace elements on mineral surfaces:
Modern perspectives from spectroscopic studies, and comments on sorption in
the marine environment. International Geology Review 43, 963-1073.
Casey, W.H., Westrich, H.R., Arnold, G.W., 1988. Surface-chemistry of labradorite
feldspar reacted with aqueous-solutions at pH = 2, 3, and 12. Geochim.
Cosmochim. Acta 52, 2795-2807.
Casey, W.H., Bunker, B., 1990. Leaching of mineral and glass surfaces during
dissolution, in: Hochella, M.F., White, A.F. (Eds.), Mineral-Water Interface
Geochemistry, Reviews in Mineralogy, 23, 397-426.
Chou, L., Wollast, R., 1984. Study of the Weathering of Albite at Room-Temperature
and Pressure with a Fluidized-Bed Reactor. Geochim. Cosmochim. Acta 48,
2205-2217.
Daval, D., Sissmann, O., Menguy, N., Saldi, G.D., Guyot, F., Martinez, I., Corvisier,
J., Garcia, B., Machouk, I., Knauss, K.G., Hellmann, R., 2011a. Influence of
amorphous silica layer formation on the dissolution rate of olivine at 90˚C and
elevated pCO2. Chem. Geol. 284, 193-209.
Daval, D., Sissmann, O., Menguy, N., Saldi, G.D., Guyot, F., Martinez, I., Corvisier,
J., Garcia, B., Machouk, I., Knauss, K.G., Hellmann, R., 2011b. Influence of
amorphous silica layer formation on the dissolution rate of olivine at 90˚C and
elevated pCO2. Chem. Geol. 284, 193-209.
175
DePaolo, D.J., 2011. Surface kinetic model for isotopic and trace element
fractionation during precipitation of calcite from aqueous solutions. Geochim.
Cosmochim. Acta 75, 1039-1056.
Druhan, J.L., Steefel, C.I., Williams, K.H., DePaolo, D.J., 2013. Calcium isotope
fractionation in groundwater: Molecular scale processes influencing field scale
behavior. Geochim. Cosmochim. Acta 119, 93-116.
Duan, Z.H., Sun, R., 2003. An improved model calculating CO2 solubility in pure
water and aqueous NaCl solutions from 273 to 533 K and from 0 to 2000 bar.
Chem. Geol. 193, 257-271.
Galy, A., Yoffe, O., Janney, P.E., Williams, R.W., Cloquet, C., Alard, O., Halicz, L.,
Wadhwa, M., Hutcheon, I.D., Ramon, E., Carignan, J., 2003. Magnesium isotope
heterogeneity of the isotopic standard SRM980 and new reference materials for
magnesium-isotope-ratio measurements. J. Anal. At. Spectrom. 18, 1352-1356.
Giammar, D.E., Bruant, R.G., Peters, C.A., 2005. Forsterite dissolution and magnesite
precipitation at conditions relevant for deep saline aquifer storage and
sequestration of carbon dioxide. Chem. Geol. 217, 257-276.
Hartmann, J., West, A.J., Renforth, P., Koehler, P., De La Rocha, C.L., Wolf-
Gladrow, D.A., Duerr, H.H., Scheffran, J., 2013. Enhanced chemical weathering
as a geoengineering strategy to reduce atmospheric carbon dioxide, supply
nutrients, and mitigate ocean acidification. Reviews of Geophysics 51, 113-149.
Hellmann, R., 1997. The albite-water system .4. Diffusion modeling of leached and
hydrogen-enriched layers. Geochim. Cosmochim. Acta 61, 1595-1611.
Hellmann, R., Wirth, R., Daval, D., Barnes, J.P., Penisson, J.M., Tisserand, D.,
Epicier, T., Florin, B., L., H.R., 2012. Unifying natural and laboratory chemical
weathering with interfacial dissolution and reprecipitation: a study based on the
nanometer-scale chemistry of fluid-silicate interfaces. Chem. Geol. 294-295, 203-
216.
Huang, F., Glessner, J., Ianno, A., Lundstrom, C., Zhang, Z., 2009. Magnesium
isotopic composition of igneous rock standards measured by MC-ICP-MS. Chem.
Geol. 268, 15-23.
Jambor, J.L., Dutrizac, J.E., Raudsepp, M., 2007. Measured and computed
neutralization potentials from static tests of diverse rock types. Environ. Geol.
52, 1019-1031.
Jarvis, K., Carpenter, R.W., Windman, T., Kim, Y., Nunez, R., Alawneh, F., 2009.
Reaction mechanisms for enhancing mineral sequestration of CO2. Environ. Sci.
176
Technol. 43, 6314-6319.
Johnson, J.W., Oelkers, E.H., Helgeson, H.C., 1992. SUPCRT92: A software package
for calculating the standard molal thermodynamic properties of minerals, gases,
aqueous species, and reactions from 1 to 5000 bar and 0 to 1000 ˚C. Comput.
Geosci. 18, 899-947.
Kiczka, M., Wiederhold, J.G., Frommer, J., Kraemer, S.M., Bourdon, B.,
Kretzschmar, R., 2010. Iron isotope fractionation during proton- and ligand-
promoted dissolution of primary phyllosilicates. Geochim. Cosmochim. Acta 74,
3112-3128.
King, H.E., Plumper, O., Putnis, A., 2010. Effect of secondary phase formation on the
carbonation of olivine. Environ. Sci. Technol. 44, 6503-6509.
King, H.E., Satoh, H., Tsukamoto, K., Putnis, A., 2014. Surface-specific
measurements of olivine dissolution by phase-shift interferometry. Am.
Mineralogist 99, 377-386.
Koehler, P., Hartmann, J., Wolf-Gladrow, D.A., 2010. Geoengineering potential of
artificially enhanced silicate weathering of olivine. Proc. National Acad. Sci. 107,
20228-20233.
Icenhower, J.P., Dove, P.M., 2000. The dissolution kinetics of amorphous silica into
sodium chloride solutions: Effects of temperature and ionic strength. Geochim.
Cosmochim. Acta 64, 4193-4203.
Lasaga, A.C., Luttge, A., 2004. Mineralogical approaches to fundamental crystal
dissolution kinetics. American Mineralogist 89, 527-540.
Loring, J.S., Thompson, C.J., Wang, Z., Joly, A.G., Sklarew, D.S., Schaef, H.T., Ilton,
E.S., Rosso, K.M., Felmy, A.R., 2011. In situ infrared spectroscopic study of
forsterite carbonation in wet supercritical CO2. Environ. Sci. Technol. 45, 6204-
6210.
Luce, R.W., Bartlett, R.W., Parks, G.A., 1972. Dissolution kinetics of magnesium
silicates. Geochim. Cosmochim. Acta 36, 35-50.
Nielsen, L.C., De Yoreo, J.J., DePaolo, D.J., 2013. General model for calcite growth
kinetics in the presence of impurity ions. Geochim. Cosmochim. Acta 115, 100-
114.
Nielsen, L.C., DePaolo, D.J., De Yoreo, J.J., 2012. Self-consistent ion-by-ion growth
model for kinetic isotopic fractionation during calcite precipitation. Geochim.
Cosmochim. Acta 86, 166-181.
177
Oelkers, E.H., 2001. An experimental study of forsterite dissolution rates as a function
of temperature and aqueous Mg and Si concentrations. Chem. Geol. 175, 485-
494.
Oelkers, E.H., Cole, D.R., 2008. Carbon dioxide sequestration: A Solution to a global
problem. Elements 4, 305-310.
Olsson, J., Bovet, N., Makovicky, E., Bechgaard, K., Balogh, Z., Stipp, S.L.S., 2012.
Olivine reactivity with CO2 and H2O on a microscale: Implications for carbon
sequestration. Geochim. Cosmochim. Acta 77, 86-97.
Pearce, C.R., Saldi, G.D., Schott, J., Oelkers, E.H., 2012. Isotopic fractionation during
congruent dissolution, precipitation and at equilibrium: Evidence from Mg
isotopes. Geochim. Cosmochim. Acta 92, 170-183.
Pokrovsky, O.S., Schott, J., 2000a. Forsterite surface composition in aqueous
solutions: A combined potentiometric, electrokinetic, and spectroscopic
approach. Geochim. Cosmochim. Acta 64, 3299-3312.
Pokrovsky, O.S., Schott, J., 2000b. Kinetics and mechanism of forsterite dissolution at
25˚C and pH from 1 to 12. Geochim. Cosmochim. Acta 64, 3313-3325.
Rimstidt, J.D., Barnes, H.L., 1980. The kinetics of silica-water reactions. Geochim.
Cosmochim. Acta 44, 1683-1699.
Rimstidt, J.D., Brantley, S.L., Olsen, A.A., 2012. Systematic review of forsterite
dissolution rate data. Geochim. Cosmochim. Acta 99, 159-178.
Rosenbauer, R.J., Koksalan, T., Palandri, J.L., 2005. Experimental investigation of
CO2-brine-rock interactions at elevated temperature and pressure: Implications
for CO2 sequestration in deep-saline aquifers. Fuel Process. Technol. 86, 1581-
1597.
Ruiz-Agudo, E., Putnis, C.V., Rodriguez-Navarro, C., Putnis, A., 2012. Mechanism of
leached layer formation during chemical weathering of silicate minerals. Geology
40, 947-950.
Saldi, G.D., Daval, D., Morvan, G., Knauss, K.G., 2013. The role of Fe and redox
conditions in olivine carbonation rates: An experimental study of the rate limiting
reactions at 90 and 150˚C in open and closed systems. Geochim. Cosmochim.
Acta 118, 157-183.
Sleep, N.H., Meibom, A., Fridriksson, T., Coleman, R.G., Bird, D.K., 2004. H2-rich
fluids from serpentinization: Geochemical and biotic implications. Proc. National
Acad. Sci. 101, 12818-12823.
178
Stillings, L.L., Brantley, S.L., 1995. Feldspar dissolution at 25˚C and pH3: Reaction
stoichiometry and the effect of cations. Geochim. Cosmochim. 59, 1483-1496.
Teng, F.-Z., Wadhwa, M., Helz, R.T., 2007. Investigation of magnesium isotope
fractionation during basalt differentiation: Implications for a chondritic
composition of the terrestrial mantle. Earth and Plan. Sci. Lett. 261, 84-92.
Watkins, J.M., Nielsen, L.C., Ryerson, F.J., DePaolo, D.J., 2013. The influence of
kinetics on the oxygen isotope composition of calcium carbonate. Earth and
Plan. Sci. Lett. 375, 349-360.
Wiederhold, J.G., Kraemer, S.M., Teutsch, N., Borer, P.M., Halliday, A.N.,
Kretzschmar, R., 2006. Iron isotope fractionation during proton-promoted,
ligand-controlled, and reductive dissolution of goethite. Environ. Sci. Technol.
40, 3787-3793.
Wimpenny, J., Gíslason, S.R., James, R.H., Gannoun, A., Pogge Von Strandmann,
P.A.E., Burton, K.W., 2010. The behaviour of Li and Mg isotopes during primary
phase dissolution and secondary mineral formation in basalt. Geochim.
Cosmochim. Acta 74, 5259-5279.
Wogelius, R.A., Walther, J.V., 1991. Olivine Dissolution at 25˚C - Effects of pH,
CO2, and Organic-Acids. Geochim. Cosmochim. Acta 55, 943-954.
Wombacher, F., Eisenhauer, A., Heuser, A., Weyer, S., 2009. Separation of Mg, Ca
and Fe from geological reference materials for stable isotope ratio analyses by
MC-ICP-MS and double-spike TIMS. J. Anal. At. Spectrom. 24, 627.
Zakaznova-Herzog, V.P., Nesbitt, H.W., Bancroft, G.M., Tse, J.S., 2008.
Characterization of leached layers on olivine and pyroxenes using high-resolution
XPS and density functional calculations. Geochim. Cosmochim. Acta 72, 69-86.
Ziegler, K., Chadwick, O.A., Brzezinski, M.A., Kelly, E.F., 2005. Natural variations
of 30
Si ratios during progressive basalt weathering, Hawaiian Islands. Geochim.
Cosmochim. Acta 69, 4597-4610.
179
Appendix 2: Experimental Protocols
A2.1 Gold Bag Preparation
1. Check bag for leaks: Assemble gold bag with titanium collar, titanium head
with the sampling valve attached, washer, and flanged compression cap containing
thrust bolts (Fig. A2.1). Tighten the bolts, a little at a time, alternating bolts to ensure
an even seal. Place the assembled bag in a large beaker full of water and pressurize
using the house air line. Look for bubbles leaving the bag. If bubbles are escaping
from a hole in the gold itself, the hole will need to be welded (see Section A2.2
Repairing Gold Bags). Mark the tear now by drawing a small circle around the tear
with a permanent marker. If bubbles are escaping from the collar/head seal, see
Section A2.3 Improving the seal between Au bag and Ti collar. If the bag is mostly
collapsed, filling it with air may cause a tear in the gold. You can reduce the chance of
this happening by pre-heating the assembled bag in the furnace for a few minutes (see
step 8 for instructions) to soften the gold.
2. Nitric acid wash: If bag does not leak, it is ready to be cleaned. Place the bag
in a glass beaker surrounded by DI water, and place the beaker on the hotplate in the
hood. Plug in the hot plate and set the temperature to 180 °C. Pour 25-50% trace metal
grade nitric acid into the gold bag, filling it to just below the top of the gold. The Ti
head and Ti-Au filter should also be washed in nitric acid; pour 2 cm of nitric acid into
a small glass beaker on the hot plate, then add the filter and head. Let all three (bag,
Figure A2.1: Gold bag
assembly. From left to
right: gold bag with Ti
collar, Au/Ti filter in Ti
head with attached
sampling tube, washer,
flanged compression cap
containing thrust bolts.
180
filter, and head) sit in hot nitric acid for about 30 minutes, label the fume hood with
"Nitric Acid" and the % concentration. The purpose of this step is to oxidize any
residual material inside the bag.
3. Rinse: After 30 the nitric acid wash, pour the acid back into the labeled
bottle. The DI water that was outside the bag can be poured down the sink (you are
confident that the gold bag is intact, because you tested it in step 1). Rinse the head
and filter, set both aside to dry. Rinse the gold bag, inside and outside VERY well and
change gloves before the next step.
4. Hydrochloric acid wash:. Place the gold bag back in the tall beaker with
fresh DI water and put the beaker on the hotplate, set the temperature to 150 °C. Pour
HCl (25-50%) into the gold bag, ensuring that the level stays 1 cm below the Au-Ti
interface. Let the bag sit for 30 minutes, but ensure that the acid does not boil (the
vapor will corrode the Ti collar). A few important notes: (1) Mixing HCl and HNO3
makes aqua regia, which dissolves gold, and (2) HCl is corrosive to titanium, so Ti
parts cannot be cleaned with HCl.
5. Rinse: repeat Step 3. If the bag was previously exposed to SiO2, go to step
6. Otherwise skip to 8.
6. HF wash: If the bag has been exposed to precipitating amorphous silica, you
need to dissolve the silica using dilute hydrofluoric acid. There may be a 1% HF in
50% nitric acid solution in the fume hood, if not you'll need to make one using
concentrated HF (in the blue cabinet) and nitric acid. Also, HF is consumed by
dissolving silica (4HF + SiO2 = SiH4 + 2H2O) and silicon tetrafluoride volatilizes
rapidly. Thus, a HF/Nitric acid solution will eventually lose its ability to dissolve
silica. HF is corrosive to both glass and Ti, so secondary containment must be teflon.
Do not put HF in a gold bag unless you are absolutely sure that there are no holes in
the gold. Once you have teflon secondary containment and a intact gold bag set up,
wash with the HF/Nitric solution the same way you completed the other washes.
Minimize contact between HF and Ti, but if there is precipitated SiO2 on a Ti part that
you can't scrape off, a brief HF wash is acceptable. Working with HF is very
hazardous. Appropriate personal protective gear must be worn, including gloves, face
181
protection, and a lab coat. Contact with skin or inhalation can result in very severe and
permanent consequences. Calcium glutamate (used to treat HF burns) is located in the
lab fridge. Obtain proper training before working with concentrated HF.
7. Rinse (see step 3).
8. Bake-out: This step is particularly important if you did a HF wash, as HF
dissolves the TiO2 coating on the titanium pieces, exposing raw Ti metal, which may
act as a catalyst for some reactions. Set the furnace in the rocker room to 420 °C (See
section A2.7.1 Using the Furnace) and place the assembled bag inside the furnace. Lay
the assembled bag diagonally so that the Ti tube fits inside the furnace. Ensure that no
metal from the bag is touching the heating coils in the furnace as this could cause a
short circuit. Close the door and let bake for at least 12 hours (24+ is best).
9. Final tightening: Before loading the reactants, it's a good idea to tighten the
center tube where it joins the head, since this bolt tends to loosen and is a major source
of leaks. Disassemble the gold bag and wrap the head with 3-4 kimwipes. Place the
wrapped head into the inside of a glove, then place in the small vice in the rocker
room. The purpose of the kimwipes and glove is to protect the head from scratches
imposed by the vice, which can permanently destroy the head. Once the head is held
by the vice, tighten the bolt at the base of the tube (clockwise) as much as possible.
Reassemble the gold bag and test one more time in the beaker with DI water and the
house air line.
A2.2 Repairing Gold Bags
1. Cleaning: If the gold bag has a tear, it will need to be fixed by welding a
small piece of gold onto the tear. In order for the patch to stick, both the bag and the
small gold piece must be very clean. The cleaning procedure mostly follows steps 2-5
of Section A2.1 Gold Bag Preparation, with the only difference being that the bag
must be cleaned inside and out. That means that for each acid wash, the bag should be
placed in a beaker of acid and then filled with acid to slightly above the level of the
tear. Again, it is very important to not mix nitric acid and HCl when cleaning the bag.
Also, the nitric acid wash must be done first, followed by a rinse, followed by the HCl
wash.
182
2. Preparing to weld: For gold, use the acetylene torch next to the workbench
with the valves. The torch has two tubes coming out of it, one goes to the oxygen tank
(green) and the other goes to the acetylene tank (red). The torch is quite bright, wear
the dark green glasses to protect your eyes (usually located on top of the red tool
cabinet). Also, tie back hair and loose clothing. Turn both tanks on, then open the
acetylene valve on the torch very slightly. Light the torch using a sparker or a lighter.
A yellow flame will appear. Slowly open the oxygen valve on the torch until the flame
turns white/blue. Too much oxygen will cause the flame to go out, but the more
oxygen you have, the hotter the flame will be. Start with a cool flame- you can always
add more heat. You can leave the lit torch in a vice attached to the lab bench for short
periods of time when you are at the bench, but if you leave the bench, turn the torch
off by first turning off the oxygen, then the acetylene.
3. Welding: Place the disassembled gold bag on the stone heat sink and make a
slight indent at the tear using your thumb. Place the small gold ball directly on top of
the tear, then direct the torch at the ball. You will see the ball heat up and start to
glow, then it will suddenly cool (after making conductive bond with the bag, allowing
heat to dissipate). Continue to heat the small ball, carefully. Eventually, it will glow,
and then suddenly melt. At that point, withdraw the heat. It may look like a very large
hole initially, but (hopefully) it's just a very smooth, reflective weld.
4. Testing: The bag will be very hot at this point, so do not attempt to move it
without using heat gloves. Depending on the size of the tear, you may be able to test
the bag by holding the opening up to the desk lamp on the lab bench and looking for
light penetrating through the bag, but the optimal test involves reassembling the bag
and inflating it in the beaker full of DI water.
5. Repeat steps 2-4 until the tear is welded closed and no longer leaks air
bubbles in the pressure test. Then proceed with bag preparation.
A2.3 Improving the seal between Au bag and Ti collar
If you notice air escaping from the head/collar assembly during the submerged
pressure test, there are a few potential causes. I'll address them in order of ease to fix.
1. Loose bolts holding the head inside the collar. Loosen the 6 bolts on top of
183
the bag assembly and then re-tighten, being sure to alternate bolts and tighten each one
a little bit at a time. Re-test.
2. Poor seal between the gold foil inside the collar and the head, part 1: Ensure
that the gold inside the collar is clean and free of debris, and that the head is also clean
and smooth. Rinsing with water and light scrubbing with a brush should be enough to
remove debris. Check for scratches on the head: horizontal ones are probably fine, but
vertical ones may be the (unfixable) culprit. Reassemble and test.
3. Loose connection between the Ti sampling tube and the head: Disassemble
the bag and take the Ti head with attached sampling tube. Wrap the head with 3-4
kimwipes and place inside a heat glove (to prevent scratches caused by the vice) and
place the head securely in the small vice attached to the lab bench in the rocker lab.
Tighten the 5/8" bolt that connects the tube to the head, it should be very tight. Do not
put the head directly in the vice without protection, the texture on the vice will dent
the head and may permanently damage it. Reassemble the bag and test again.
4. Poor seal between the gold foil inside the collar and the head, part 2:
Sometimes an extended bake-out in the furnace is sufficient to improve the seal.
Assemble the bag and place in the furnace for at least several hours, overnight is best.
Remove from furnace, let cool for about 30 minutes, then retest.
5. Poor seal between the gold foil inside the collar and the head, part 3: If the
bag has been used many times, the gold foil will become thin at the seal point and will
no longer create a seal. At this point, the only solution is to remove the collar from the
gold bag, trim off the top 2-3 cm of the bag, and put the collar back on. I've found a
screwdriver to be effective at loosening the bag from the collar, though it is an arduous
process that requires time, strength, and agility. Trimming is best done with scissors
and an even top edge will make things easier. To reinsert the bag to the collar, the bag
diameter must be reduced at the top by creating loose folds or waves in the foil. Once
the folded/waved bag is inside the collar, it can be stretched out to fit the form of the
collar. Pay attention to any sharp folds that form, those are nearly impossible to
remove later. The top of the new gold should extend at least to the slope change inside
the collar, though higher is better as you will later be pushing the gold down when you
184
reintroduce the head. Also remember that the new gold top is a thicker foil than the old
one, so inserting the head may take several iterations of pushing down (possibly
lightly pounding with a small hammer), heating to soften the gold, and manual
stretching of the gold to coat the inside of the collar. Once the bag is successfully
reassembled with the head and bolts tightened, place it in the furnace to bake for at
least overnight, but longer is better. Several days may be required to create a solid
seal, and several iterations of heating, testing, and reassembly might be necessary.
A2.4 Assembling the reactor
1. Calculate amount of reactants needed: Measure the volume of the gold bag
by taring the scale with the gold bag inside a tall beaker. Pour DI water into the bag
until the water level is at the slope change in the bag, this weight is the total mass of
water that will fit inside a sealed bag. It's safe to take that number as your volume in
cubic centimeters. To prevent the bag from bursting, set aside 20% of that volume, so
the volume you will be filling with reactants=80% of the total volume. Calculate how
much water, mineral, and gas you will be adding to the bag at this point and be sure
that the total volume <80% of the available volume (so a 100 mL bag may have 7 mL
of CO2, 70 mL of water, and 3 mL of rock). For this calculation, be sure to use your
desired rock:water and water:gas ratios. A sample calculation is below, where ρ is the
density of a substance at reaction conditions.
gaswaterrock volumevolumevolumevolumebag _*8.0
gas
gas
water
water
rock
rockmassmassmass
volumebag
_*8.0
if your water:rock ratio = 50:1 by mass, then 50
waterrock
mm
if your water:gas ratio = 10:1 by volume, then water
waterwatergas
mvv
1010
It's also a good idea to calculate the volume of CO2 that you will be adding
now, even though you won't actually add it until later. The total volume of added CO2
at reaction temperature should be less than the free volume in the bag. So if you are
185
adding 70 mL of fluid to a 100 mL bag, the total volume of CO2 at reaction
temperature should be less than 30 mL to account for the time when the CO2 is heated
to reaction temperature but not yet dissolved. See the sample calculation below.
aseseperateph
CO
dissolved
CO
tot
CO molmolmol 222
water
dissolved
CO volumeLmolsolubilitymol *)/(2
)/(_
)(_
22
molccvolumemolar
ccvolumemol
Treaction
COaseseperateph
CO
)/(_*_
2
_
2 molccvolumemolarmolvolumeTroomtot
CO
Troom
CO
)/(_*_
2
_
2 molccvolumemolarmolvolumeTreactiontot
CO
Treaction
CO
2. Add reactants: Add all solid and liquid reactants to the bag, then assemble it
(bag and collar, head with Ti/Au filter and Ti rod, washer, and flanged compression
cap) and pressure test it one more time. Be sure to record weights and/or volumes of
reactants added.
3. Attach bag to stainless steel head: When assembled, the gold bag will hang
inside the stainless steel reaction vessel from the stainless steel head. Place the
stainless steel head in the smaller vice on the lab bench and tighten securely. Brush a
small amount of nickel grease onto the Ti tube on the bag assembly about half way
along the tube and on the 5/8" bolt at the base of the tube. Push the tube through the
largest hole in the middle of the stainless steel head until the bolt contacts the bottom
of the head, tighten the bolt using a 5/8" wrench as tight as you can make it. Expect
the bench to shake a little. If this bolt is not sufficiently tight and leaks when you
pressure test the assembled apparatus, you will need to take almost everything apart to
tighten it.
186
4. Load the SS pressure vessel: Place the steel pressure vessel in the larger vice
attached to the lab bench, with the opening facing up and the rod protruding down
through the base of the vice. The vessel has two small notches at the base that allow it
to sit securely in the vice. Fill the vessel most of the way with DI water and put the
gasket in the groove on top of the vessel. Use Ni grease along the top edge of the
vessel and coat the gasket well. Next, remove the SS head (with gold bag attached)
from the small vice and place it on top of the SS vessel. Water should overflow as the
bag is lowered into the vessel. If it doesn't, add more water until it does. It is very
important to ensure that the vessel contains no air in the annular space, as this could
Figure A2.2: Stainless steel pressure vessel with steel head, gasket, and thrust bolts. The
gold reaction cell is contained in the vessel and surrounded by a pressure fluid (DI water).
187
later cause the bag to collapse. See Fig. A2.2 for a diagram of the assembled pressure
vessel.
5. Seal the pressure vessel: Generously grease the threads on the SS pressure
vessel with Ni grease and lift the stainless steel compression cap (containing 10 thrust
bolts) over the Ti tube and onto the SS pressure vessel. Ensure that the thrust bolts are
loose, then rotate the SS compression cap on the threads on the vessel until tight.
There are two metal rods that can be placed in holes on the sides of the compression
cap to reduce the forced needed to tighten. Once tight, loosen by 1/4 turn. Grease each
thrust bolt and tighten them to barely finger tight. Then, use a torque wrench to tighten
the thrust bolts in 5 lb increments and alternating bolts: tighten 1, skip 2 and 3, tighten
4, skip 5 and 6, tighten 7, skip 8 and 9, tighten 10, skip 1 and 2, tighten 3, skip 4 and 5,
tighten 6, skip 7 and 8, tighten 9, etc. After tightening all 10 thrust bolts at a given
torque, increase the force by 5 lbs. Stop at 30 lbs.
6. Attach thermocouple and pressure fluid inlet tube/valve: The two remaining
ports on the SS head are for the thermocouple and pressure fluid inlet tube and valve.
Using the 5/8" wrench, securely tighten the bolts on the thermocouple and on the tube
into the ports.
7. Build the support structure for the Ti tube and sampling valve: The Ti tube
is relatively fragile, so we build a structure that prevents us from putting any lateral
force or torque on it when moving the vessel around and sampling. Screw two vertical
rods into the two holes on top of the compression cap, which is attached to the SS
vessel. Place a tube over each rod (the tubes are slightly shorter than the rods and wide
enough to slide over them easily). Now place the horizontal strut over the top of the
rods; the Ti-tube should sit freely in the middle of the strut and the strut should rest on
the tubes encasing the two rods. Use a washer and nut (9/16") on each tube to secure
the strut in place. It should feel very tight (Fig. A2.3C). Finally, attach the metal cage
to the strut. There are two holes on the side of the strut near the Ti tube, these
correspond to two holes on the metal cage. For now, place a bolt (1/2") through the
cage and then the strut, hold in place with a nut but do not tighten. The cage should
freely rotate, once the valve is attached you'll tighten it.
188
8. Attach the sampling valve: Place a hollow bolt, threaded side facing up, on
the Ti tube and let it fall. Then, screw on a collar to the threaded top of the Ti tube.
The tube is reverse-threaded, so you need to twist it counter-clockwise to tighten. This
can require more force than you can apply with your hands, you can use a hand-held
vice to assist you (simply adjust the vice so that it clamps tightly on the collar, when
you are done squeeze the lever near the handle to release the compression). The collar
should rest slightly below the top of the Ti-tube but not nearly to the bottom of the
threads. Place the valve on top of the collar, and raise the protective cage (Fig.
A2.3A). Ensure that the sampling port on the valve is accessible with the cage up, and
take note of the required orientation for the cage to fit the valve. Drop the cage, and
bring the bolt up the Ti tube to attach to the valve. Tighten very well, using two
wrenches (5/8" on the bolt and a large adjustable one on the valve). Raise the cage,
making certain that it fits snugly around the valve (if the fit is too loose, several
washers can be used between the valve and the cage), and tighten the bolt connecting
the cage and strut. Add the second bolt next to the first. Use a hose clamp to secure the
valve in the cage. Fig. A2.3B shows the attached valve and support structure.
9. Purge the gold bag of air: Remove air from the gold bag by adding pressure
Figure A2.3: Assembling the reactor
vessel. A. Attaching the sampling
valve (step 8). B. Full assembled
reactor. C. Completed support
structure, top view. Potential leak
points are visible in B and C. If the
leak is below the valve (B), tighten
the bolt connecting the valve to the
tube. If the leak is visible from a top
view (C), tighten accessible bolts and
retest. Disassembly may be required.
189
fluid to the annular space between the bag and the SS vessel (see section A2.7 for
more information on the pressure system). First, attach the inlet/outlet tube and valve
to the workbench pressure line to the right of the large vice. Use a hollow bolt with a
plug in the third opening of the inlet/outlet valve, this valve should be closed. At the
bench with the pressurization system, close the valves for racks 3 and 4, open the
reservoir valve and the workbench valve. Use the bleed to drop the pressure to room
pressure. Next, open the valve on the bench next to the vice as well as the inlet/outlet
valve on the vessel and open the sampling valve. Use the pump valve on the
pressurization system to add pressure fluid to the vessel, air will come out the
sampling valve. When water starts to come out of the sampling valve, close it and turn
off the pump. The gold bag is now air-free.
10. Pressure test the system: With the vessel still connected to the
pressurization system and the sampling valve closed, open all valves between the
pump and the vessel (two at the vice plus "workbench" at the pressure bench) and
bring the pressure up to 100 bars. Check for leaks, which will look like slowly (or
quickly) growing pools of water on top of the vessel. If no water is visible, increase
the pressure to 200 bars and let set for 10 minutes. If there is still no water present, the
vessel is sealed and ready to go, skip to step 12. Otherwise troubleshoot the leak (step
11).
11. Troubleshooting the leak: Most leaks are easy to fix, so if in doubt, try all
of those fixes before disassembling the vessel (see Fig. A2.3 for location of leak
points).
- Water is leaking out of the sampling port of the valve: A leaky valve can't be
fixed easily, but it can be plugged using a hollow 5/8" bolt with a plug. Tighten
the plug with a wrench, but do not over torque and make sure the valve is
stabilized by the support structure so as not to damage the Ti tube.
- Water is leaking out of the junction between the valve and the Ti tube:
Tighten the bolt at this connection. You will probably need to remove the cage
to access the bolt, and use a large wrench to hold the valve steady while you
tighten the bolt with a 5/8" wrench.
190
- Water is leaking at the connection between the SS head and either the
thermocouple or the pressure fluid inlet/outlet: Tighten the connection with a
5/8" wrench.
- Water is leaking where the Ti tube comes out of the head: This is the tough
one. The leak is the result of a poor connection between the gold bag and the
SS head; the only way to fix it is to disassemble the whole reactor and
retighten the bolt (Step 3).
12. Load the vessel into the rocking furnace: Bleed off the pressure to take the
reactor back to 1 bar. Disconnect the pressure fluid inlet/outlet line and valve from the
workbench line and remove the plug from the valve. Close the valve to the left of the
vice and the "workbench" valve on the pressure bench. Tilt the rocking furnaces such
that the opening is facing out, and prop them up using a jack and a few pieces of wood
(on the floor in front of the controllers). Loosen the vice holding the SS reactor vessel
and lift it out. Place the vessel into the furnace, rod first, such that the sampling valve
is facing out and up. The strut will be vertical. The vessel is very heavy and can cause
significant injury if dropped or carried improperly. If you are not confident that you
can move it yourself, find someone to help.
13. Secure the vessel inside the furnace: Maneuver the vessel such that the rod
at the bottom sticks out through the hole in the back of the furnace. This is easiest to
do with two people: have one person stand in front and lift the rod up towards the hole
by pulling the strut downwards, and have the second person standing behind the
furnace with a long, narrow screwdriver. When the person in back sees the rod near
the opening, insert the screwdriver into the hollow rod and guide it into alignment with
the opening. The person in front can than push the vessel further into the furnace and
the rod will extend out the back. Hold the rod in place by using a 3/4" nut on the rod,
though it should not yet be fully tightened. If the nut does not easily move along the
rod but appears to be the correct size, check the thread spacing. Some rods and nuts
have wide spacing, others have narrow spacing, and they are not compatible with each
other. Be very careful when working behind the furnaces as there are bare wires
present.
14. Attach the vessel to the furnace in front: The holes at the top and bottom of
191
the strut should align with holes on the furnace, though depending on which strut you
use and which furnace, sometimes only one hole aligns. Using 1/2" bolt(s), secure the
strut to the furnace. You will need to lift the strut in order to line up the openings,
again a second person is useful. If you have trouble lifting the vessel and strut, the nut
in back may be too tight and prevent the needed movement.
15. Completing the pressure fluid inlet/outlet lines: The valve attached to the
vessel must be screwed down to the left-top of the furnace using two long screws,
such that the valve is facing outwards (it will look similar to the valve already attached
on the right side of the furnace). Once secured, the line that goes to the pressure
release valve and the line that goes to the right-side valve should be attached to the
valve connected to the vessel. The right valve should be closed, now open the left
valve and leave it open for the remainder of the experiment. This valve connects the
pressure fluid inside your experiment to the pressure release valve, and should you
have runaway heating due to a broken component, the pressure release valve will
prevent an explosion from occurring.
16. Weight the back of the furnace: In the back, tighten the 3/4" nut using a
wrench. Next, place a lead weight on the rod and then another 3/4" nut to hold the
weight in place. The weight is necessary to shift the center of mass of the furnace
towards the middle so that the motor can rock the reactor. Without it, you risk
overworking the motor.
17. Turn on the furnace: Flip the switch to the right of the temperature
controller to turn on the furnace (Fig. A2.4). Below the controllers are 4 variacs
(variable transformers), these control the power output to each furnace. Turn the
variac that corresponds with the furnace you are using up to 50-90%. The higher the
power, the faster the vessel will heat up, but you will also have greater overshoot by
the controller (so the temperature of the furnace will oscillate over a larger range).
You can always use a high power to heat the furnace and then turn the power down
later. Set the temperature on the controller by pushing down on the middle circle and
turning it. The red arrow points to the set temperature and the green arrow points
towards the current temperature. Unfortunately the numbers on the controllers do not
192
necessarily correspond with actual temperatures, so you may need to calibrate the
furnace you are using to achieve your desired temperature. Keep in mind that the
temperature used to control the furnace is measured inside the furnace itself, and will
be 5-10 degrees higher than the temperature inside the reactor.
18. Monitor pressure and temperature: While the reactor is heating, it's a good
idea to keep an eye on the pressure and temperature inside the reactor, which can only
be done if the furnace is not rocking. To monitor temperature, simply connect the
thermocouple from the reactor to the thermocouple extension cord in port 2, the read-
out temperature will be between the two sets of non-operational furnaces on the left.
Monitor pressure by opening all the valves between the reactor and the pressure
gauge, including the "reservoir" which acts as a pressure buffer. See section A2.7.2
Pressurization System for more information.
19. When desired temperature is reached, take an aqueous sample: This first
sample will be your t=0 data point. Pressurize the reactor to at least 50 bars in order to
take the sample (since you are removing liquid, the pressure will drop quickly with
only small amounts of fluid removed, and you cannot sample if the pressure inside the
reactor is equal to the outside pressure). Take a aqueous sample of about 3 mL into a
syringe, then filter 1 mL into a vial to be tested later for elemental concentration.
Figure A2.4: View of 4 rocking
furnaces (A) showing temperature
controls and variacs. Zoomed in
image of temperature controls (B)
shows the temperature read-out,
power switch, and fuse for each of
the four furnaces.
193
Acidify the sample and be sure to get the weight of the sample and of the sample +
dilution. Use the remainder of the sample to directly measure the pH. See more
detailed sampling protocol in section A2.5 Sampling), but keep in mind that this first
sample is simpler because you are only taking one aqueous sample for both ICP and
direct pH measurement. After sampling, run the pump until the reactor is slightly
below your target pressure.
20. Add CO2: Open the CO2 tank (near the bench with the vices) and the valve
to the left of both racks of furnaces. Turn on the pump, both the power switch on the
unit on the bottom shelf of the cart and the power switch on the pump on the top shelf
of the cart. Fill the pump's reservoir by pressing "refill." The reservoir will fill with
CO2 at the tank pressure, which should be around 60 bar (if it's lower, the tank is out
of liquid and should be replaced). Once the reservoir is full (volume is about 100 mL),
compress the gas to your desired pressure by selecting "constant pressure," setting
your pressure ("A", then enter desired pressure, then "enter") and pressing "run." The
volume will decrease as the pressure increases until the set pressure is reached. Open
the valve above the pump for a second (or less) to clear the line of air, then close the
valve and connect the line to the sampling port of your experiment. Tighten the bolt
well, then open the valve above the pump again. The line is now pressurized and the
volume will drop. Wait until the flow rate is below 0.1 mL/min, make a note of the
volume in the reservoir, then open the sampling valve to allow CO2 to flow into the
reactor. Use the bleed valve to slowly remove pressure fluid from the reactor in order
to make space for the injected gas. Once your desired volume of gas has been injected,
close the bleed valve, close the sampling port, then close the valve above the CO2
pump. The pump should be stored with some gas inside. Turn off both power
switches, close the valve on the wall, and close the tank. Remove the line from the
sampling port of the reactor and replace it with a plug. Close the right valve on the
reactor plus the other valves in line (the rack 3 or 4 valve + the valve on the rack next
to the reactor), then remove the pressure line from the reactor. Detach the
thermocouple and begin rocking the reactor by flipping (and holding, if necessary) the
switch on the side of the rack.
194
A2.5 Sampling
1. Stop the reactor from rocking, measure temperature and pressure: Flip the
switch next to the reactor to stop it from rocking. Use a jack and several pieces of
wood to hold the furnace into the desired position (sampling valve facing towards you
and slightly down for an aqueous sample). Attach the thermocouple to the extension
from port 2, check the temperature readout on the display between furnace racks 1 and
2. Connect the pressure fluid line to the top right valve and pressurize the line by
opening the valve for rack 4 (or 3), the reservoir, the valve on the rack, and pumping
until you reach the desired reactor pressure. Close the reservoir valve. Open the valve
on the top right of the reactor and read off the pressure.
2. Prepare for taking an aqueous sample: Open the reservoir valve and pump
the system until you reach your desired pressure (if necessary). Remove the plug from
the sampling port and replace it with the teflon sampling tube (Fig. A2.5).
3. Preweigh vials and syringes: For an aqueous sample, you will generally be
taking a fluid sample for ICP analysis and alkalinity analysis in the same syringe, plus
a sample for dissolved CO2 analysis in a separate syringe. You will first need to take a
"bleed" sample, which effectively clears the Ti sampling tube of stagnant water so that
you are sampling water that is representative of the reaction fluid. In all, you will have
three syringes (bleed, ICP/alkalinity, CO2) and two vials (ICP, alkalinity) that need to
be labeled and pre-weighed. The CO2 syringe should be pre-weighed attached to a 3-
way valve, as this will be used during sampling and left on to isolate the aqueous CO2
sample from the air. The ICP/alkalinity syringe should be weighed with the filter to be
used later for filtering the sample.
4. Take bleed: Attach the "bleed" syringe to the teflon sampling tube and
slowly open the sampling valve until water and gas begin filling the syringe. This
sample should be at least 0.5 mL of water, the gas volume will depend on your
temperature/pressure conditions.
195
5. Take ICP/alkalinity and CO2 samples: Attach both the ICP/alkalinity and the
CO2 syringes to the three-way valve, with the CO2 syringe isolated (switch pointing
towards this port). Insert the third side of the 3-way valve into the sampling port and
slowly open the sampling valve. Water and CO2 will flow into the ICP/alkalinity
syringe. Take about 0.5 mL of aqueous sample, then flip the switch on the 3-way
valve to isolate the ICP/alkalinity syringe and allow fluid and CO2 to flow into the
CO2 syringe without touching the sampling valve. Collect about 0.3 mL of fluid and
associated gas, then flip the switch back to isolate the CO2 syringe and continue taking
the ICP/alkalinity sample. In total, collect 1-2 mL of water in the ICP/alkalinity
syringe, then close the sampling valve.
6. Finish taking ICP and alkalinity samples: Moving quickly, remove the
ICP/alkalinity syringe from the 3-way valve, attach the filter, and weigh it. Pushing
the aqueous sample through the filter, inject about half into each of the pre-weighed
icp and alkalinity vials. Weigh both vials. Add approximately 10 mL of 3-5% nitric
acid to the icp vial, weigh again. Do not acidify the alkalinity sample.
7. Finish taking the aqueous CO2 sample: The CO2 syringe will still be
attached to the 3-way valve and isolated from the lab. Use pressurized house air to dry
the valve (any residual water is not part of your sample) and weigh the syringe + valve
+ sample. Record this mass. Next, inject the sample into the coulometer for analysis,
the output is in mg C which can be converted to mol CO2/kg sample (see A7.3 for
details on coulometer operation).
8. Preparing to take a gas sample: Carefully rotate the furnace such that the
sampling tube is pointing straight up (the CO2 phase is less dense than the aqueous
phase). Purge the sampling tube of water by taking a bleed, expect 1-3 mL of water to
Figure A2.5: Sampling syringe with 3-way valve, attached to sampling port. The
sampling port consists of a stainless steel, 5/8" bolt with a teflon insert, and
plastic tubing connects the insert to the 3-way valve.
196
be removed before you reach the gas phase. Use the house air line to clear water from
the teflon sampling port and dry the valve with a kim wipe. Prepare your syringe by
attaching a simple 2-way valve to the end, pre weighing is not necessary. Prepare your
storage vessel by evacuating and backfilling with an appropriate gas, such as helium.
9. Taking a gas sample: Replace the sampling port into the sampling valve,
connect the syringe to the port. Open the sampling valve slowly and fill the syringe
with gas, then close the sampling valve. Close the valve on the syringe as well to
isolate the sample from the lab air, then remove from the sampling port. Attach a
needle to the valve, which allows for the penetration of the rubber stopper on the
evacuated storage vial. Open the valve on the syringe and inject the gas sample into
the storage vial, then close the valve and slowly withdraw the needle from the rubber
stopper. Gas samples are generally stored in the freezer to minimize the reactivity of
water that may have been inadvertently introduced.
10. Finishing the sampling process: Ensure the sampling valve is closed and
replace the sampling port with a stopper (in the sampling valve). Pump the reactor if
necessary to reach the desired pressure. Close all pressure valves and remove the
pressure line from the top right (Rack 4) or left (Rack 3) valve on the furnace. Unplug
the thermocouple from the reactor. Flip the switch on the outside of the rack to turn on
the motor that rocks the furnaces.
A2.6 Taking down an experiment
To take down an experiment, repeat the set-up in reverse, starting at step 17.
Before beginning the take-down process it's worthwhile to set up your filtration
system, so that when you have opened the gold bag you can simply pour the contents
into the filter without needing to change gloves and set everything up.
1. Turn off the furnace by flipping the switch on the temperature control board.
Turn the variac down to zero (no power).
2. Remove the lead weight from the back of the reactor, using a 3/4" wrench to
remove the nut holding the weight on the rod. Remove the second nut, if present. Once
again, be cautious of bare wires and debris behind the furnaces.
3. Turn off both valves on top of the furnace. Loosen the two lines connected
197
to the "pressure release valve" (top left (Rack 4) or top right (Rack 3)). Then remove
the valve by removing the two screws securing the valve to the furnace. Unplug the
thermocouple.
4. Detach the steel strut from the furnace by loosening the bolts on the top and
bottom of the structural element. The stainless steel autoclave and attached structure
should now be free of the furnace.
5. Remove the stainless steel autoclave from the furnace, keeping in mind that
it may be very hot. The vessel is quite heavy, be sure to take appropriate precautions
(e.g. insulated gloves, a second person to assist with moving the vessel). Place the
autoclave in the large vice on the workbench and tighten securely.
6. Vent the gas phase by opening the sampling valve. You want to disassemble
the reactor as quickly as possible to minimize chemical changes during take-down
(e.g. precipitation of secondary phases). Thus, while the experiment is venting, begin
disassembling the structure that supports the sampling valve and tube.
7. Remove the hose clamp from the valve and detach the metal valve cage.
Remove the nuts that secure the horizontal strut to the vertical rods, remove the strut.
Then, unscrew the rods from the stainless steel compression cap.
8. Once the experiment has vented, remove the sampling valve using 2
wrenches, one to hold the 5/8" bolt and one to hold the valve itself (1-2"). Carefully
unscrew the bolt, putting as little torque on the sampling tube as possible. Remove the
collar from the top of the Ti-tube (it's reverse threaded) and the hollow bolt will slide
off the top.
9. Loosen the bolts connecting the thermocouple and pressure fluid inlet line to
the stainless steel head. Some water will likely leak out, if you see any gas escape then
you know that your bag is unlikely to be in good shape. It's essential that you have
vented the gas from inside the gold back before releasing pressure from the pressure
fluid, otherwise the gas inside the bag will expand and you'll create a hole in the bag.
10. Loosen the thrust bolts on top of the compression cap. Loosen one, skip
two, loosen the fourth, skip the fifth and sixth, etc. If the last bolt or two are stuck, try
tightening the bolts on each side of the stuck bolt to relieve some of the pressure and
198
reduce the force necessary to release it. Remove compression cap by unscrewing it
and then lifting it over the top of the Ti tube.
11. Carefully lift the stainless steel head with Ti-tube and gold bag attached out
of the stainless steel vessel. Place it in the smaller vice on the right side of the bench
and tighten the vice around the head, such that the bolt connecting the bag and head is
accessible. Loosen the bolt with a 5/8" wrench, then gently pull the bag away from the
head to release the Ti-tube.
12. Disassemble the bag by loosening the bolts on top of the flanged
compression collar. Often the last bolt gets stuck, tighten the two bolts on each side of
it to release the pressure.
13. Pour the contents of the gold bag into the filter. Generally you'll want to
use a vacuum filter set-up because the mineral clogs the filter quickly, plus you may
need to rinse with a lot of water. Collect some of the filtrate for analysis (pH,
alkalinity, ICP) and store it in sample bottles, acidifying and diluting the ICP sample
only. Rinse out the bag using DI water, pour into the filter. Repeat several times, until
the bag is empty of mineral/rock powder.
14. Lift the filter paper out of the filter and place on a watch glass, then leave
to dry (in either the hood or in a low temperature oven).
A2.7 Other Procedures
A2.7.1 Using the Furnace
1. Turn on power switch on the bottom right. Four numbers will appear on the
two displays on the bottom right, the two red numbers are the current temperature and
the green numbers are the programmed temperatures. The leftmost display shows the
"SP" (set point) temperature in green, this is the desired temperature and should be
420 for baking out gold bags. The rightmost display is the "Exceeded" (upper bound)
temperature and must always be greater than the set temperature. When the furnace
exceeds this higher temperature, it turns off until it cools below the set temperature.
For baking gold bags, this number should be 430.
2. Change the "SP" temperature by simply pressing the up or down arrow on
199
the leftmost display, when the desired temperature is reached, it the "set/enter" button.
Change the "Exceeded" temperature by holding down the "set/ent" button for 3
seconds, until "A1" is displayed on the top line. Adjust the temperature using the
arrows until the desired temperature is reached, which must be greater than the "SP"
temperature. Press "set/ent" again to set the upper bound temperature and then press
and hold "set/ent" for 3 seconds to return to the normal display.
3. Once the "SP" and "Exceeded" temperatures are what you'd like them to be,
start heating by holding down the "reset" button on the rightmost display until you
hear a click, at which point the readout temperature (in red) should start to increase.
A2.7.2 Pressurization System
Pressurization of reactor is achieved by compressing DI water, stored above
the pressure bench in a large plastic container (Fig. A2.6). Water is pumped by the
"pump" and can be used to pressurize several vessels. The first is the "reservoir"
which is on the second level of the pressure bench. The pump is also connected to the
furnaces on Racks 3 and 4 through valves on the pressure bench. Each of the two racks
has two pressure lines, one for the top furnace and one for the bottom furnace. These
lines are not connected to the furnaces themselves by default (just the rack structure to
the right of rack 4 and left of rack 3), and must be detached when the furnaces are
rocking to prevent shearing of the tubing. The fourth place that the pump can
pressurize is a vessel located on the main lab bench, through the "workbench" valve.
There is a second valve attached to the corner of the workbench, between the large
vice and the red tool cabinet. The 6th valve located on the pressure bench is the
"bleed" valve, this releases pressure by draining water out of the vessel to which you
are currently connected. It's generally a good idea to ensure that all valves on the
pressure bench are closed before you begin to work with the pressurization system to
prevent inadvertently affecting an experiment.
To pressurize a vessel on the workbench, connect the vessel's pressure fluid
line with the valve on the workbench (next to the vice). Ensure that all valves on the
pressure bench are closed, then use the bleed to reduce the pressure in the reservoir to
room pressure (open reservoir valve, open bleed valve, once pressure has stopped
200
decreasing, close bleed valve). Next, open the "workbench" valve and the valve next
to the vice, you now have a connection between your reaction vessel, the reservoir,
and the pump. Increase the pressure by turning the pump valve. When the desired
pressure is reached, close all valves (next to vice, "workbench", and "reservoir").
To pressurize a vessel loaded in a furnace, first ensure that the pressure fluid
inlet/outlet lines are attached to the furnace. Then, take the pressure line from the right
(Rack 4) or left (Rack 3) of the furnace and attach it to the valve on the top right (or
left) of the furnace, using a 5/8" wrench to tighten. Close all valves on the pressure
table. Then, open the reservoir valve and bleed it (if necessary) such that the pressure
is equal to what you expect the reactor pressure to be. Open all valves in line with
your experiment (Rack 3 or 4 on the table, plus one on the side of the rack, plus the
one on the top right (or left for Rack 3) of the furnace). Increase the pressure using the
pump valve on the pressure bench.
Figure A2.6: Pressurization system on the pressure bench. Valves across the front are labeled (left to
right): Bleed, Pump, Reservoir, Rack 3, Rack 4, Workbench. Analogue pressure gauge hangs above
the valves. The Reservoir sits below the bench and has its own gauge.
201
A2.7.3 Coulometer Operation
Gas delivery system: Air is taken in from the back left of the gas delivery
system (Fig. A2.7) and passes through a KOH solution to remove CO2. The air then
moves through a tube to the glass test tube, where it bubbles through the acidic
solution, picks up CO2 from the sample, and passes through a silver chloride solution
to remove sulfur. The gas then enters the coulometer. Once inside the coulometer, the
gas bubbles through a proprietary cathode solution in an electrochemical cell. CO2
reacts with the blue cathode solution to turn it clear, which increases the transmission
of the light through the electrochemical cell.. The increased transmission signals the
coulometer to run a current through the electrochemical cell and presumably reduce
the CO2 species, making the solution blue again. The coulometer measures the amount
of current needed to reduce the transmission back to 28% and converts that number
into milligrams of carbon, which is the number it outputs.
Sample injection: Wait for the coulometer reading to steady, such that the
current is at 28 and the total carbon is constant. Hit the "reset" switch to set the total
carbon measured to zero. The sample is
introduced as liquid, gas, or a combination
through a 3-way valve that connects the sample
syringe, glass test tube, and phosphoric acid
(10%) reservoir. Inject the sample by closing the
valve towards the acid reservoir, then close the
valve towards the glass test tube and backfill the
syringe with 2 pumps (4 mL) of phosphoric acid.
Close the valve towards the acid reservoir again
and inject the acid from the syringe into the glass
tube.
Calibration: The coulometer's calibration
can be checked by using test tubes with known
amounts of calcium carbonate. Inject 4 mL of
phosphoric acid into the test tube, which will
Figure A2.7: Gas delivery system for
the coulometer. Sample is injected at
3-way valve.
202
dissolve the carbonate, and the output of mg C should be comparable to the amount of
carbon you put it. To calibrate the coulometer, flip the switch on the lower left to
"calibrate". In this mode, the coulometer will NOT run a current through the cell, so
you cannot make a measurement. However, you can adjust the % of light transmitted,
such that 28% is transmitted when the electrochemical cell has no CO2 and 100% is
transmitted when it is saturated with CO2. This will affect the strength of the electrical
current applied, which affects the time required for a measurement.
Fluid replacement: When the coulometer is working well, you should see the
transmission % increase quickly after the addition of your sample, remain high for a
while, then decrease quickly to 28-30%. If you observe a tail on your measurements,
you likely need to replace some (or all) of the fluids- KOH, cathode solution, or anode
solution.
203
Appendix 3: Redox State of Reaction
We did not control for the oxidation/reduction potential in our experiments;
however, it's possible that redox state plays an important role in determining the
kinetics of olivine carbonation. Olivine contains about 10 mol% Fe, which is a redox
active element. Iron exists in the ferrous state in olivine and in its carbonate form,
siderite, but in the presence of oxygen and water, iron will oxidize to Fe3+
(Equation
A2.1).
Fe2+
+ H+ + 0.25O2 = Fe
3+ + 0.5H2O (log K=5.897 at °60 C) (A3.1)
The positive value of logK means that equilibrium favors the products, in this case,
ferric iron, over the reactants. Thus, we expect that any dissolved oxygen remaining in
the reactor will oxidize ferrous iron to ferric iron. By knowing the fugacity of oxygen
in the reactor at any given time, we can predict the partitioning of iron into the ferric
and ferrous states.
In order to determine the fugacity of oxygen in the reactor, we measure the
concentration of hydrogen gas in the supercritical CO2 phase. Measurements were
made using an Agilent 6890 Series Gas Chromatograph (GC) with a pulse discharge
detector and He as the carrier gas. ChemStation software was used to integrate peak
areas, and a four point calibration curve ranging from 100 to 800 ppm was used to
quantify the peaks. Hydrogen partial pressures are plotted as a function of time in Fig.
A3.1B and show increasing concentration as a function of time. Elemental
concentrations from the aqueous phase of this experiment are shown in Fig. A3.1A
and also show increasing concentration with time. The chemical evolution of this
experiment was very similar to experiment 19-S from Chapters 3 and 4.
Figure A3.1: Element concentration (A) and H2 pressure (B) as a function of time.
204
205
Appendix 4: Derivation of diffusion coefficient from Lasaga (1979).
We start with equation 15 from Lasaga (1979) for interdiffusion of n ions, linking the
flux of component i with the concentration gradient of component j:
o
njn
k
o
kkk
iji
o
i
ij
o
iij DD
Dcz
czzDDD
0
2
For binary interdiffusion, n=2 and i=j=1. For i=j, δ=1.
o
k
o
kkk
oo
DD
Dcz
czDDD 2
0
122
1
2
11111
oo
oooo
oo
ooo
DczDcz
czDDczDD
DczDcz
DczDczDD
22
2
211
2
1
1
2
1121
2
111
22
2
211
2
1
22
2
211
2
1111
oo
oooooooo
DczDcz
czDDDDczczDDDDczD
22
2
211
2
1
1
2
112122
2
21
2
111111
2
111
Define normalized concentrations (mol fractions) 1
21
1 xcc
c
2
21
2 xcc
c
oo
oooo
DxzDxz
xzDDDDxzD
22
2
211
2
1
1
2
112122
2
211
Set charges on Mg (z1=2) and H (z2=1).
oo
oooo
DxDx
xDDDDxD
2211
11212211
4
4
oo
oo
DxDx
xxDDD
2211
121211
4
)4(
206
Substitute for x2 to put the expression for D in terms of Mg concentration only.
121 xx
oo
oo
DxDx
xxDDD
2111
111211
14
)41(
)1(4
)13(
1211
12111
xDxD
xDDD
Rename variables (D11 becomes D*, 1 becomes Mg, 2 becomes H, x1 becomes c) and
add the structural term from Hellmann (2012) to get Equation 4.
ccDcD
cDDD
HMg
HMg
11
)1(4
)13(*
Reference:
Lasaga, A.C., 1979. Multicomponent exchange and diffusion in silicates. Geochim.
Cosmochim. Acta 43, 455–469.
Top Related