Geochemical Rate Models An Introduction To Geochemical Kinetics Rimstidt J.D
Geochemical Rate Models An Introduction To Geochemical Kinetics Rimstidt J.D
com
https://ebookgate.com/product/geochemical-rate-models-an-
introduction-to-geochemical-kinetics-rimstidt-j-d/
OR CLICK BUTTON
DOWLOAD EBOOK
https://ebookgate.com/product/geochemical-processes-conceptual-models-
for-reactive-transport-in-soil-and-groundwater-horst-d-schulz/
ebookgate.com
https://ebookgate.com/product/global-environment-water-air-and-
geochemical-cycles-2nd-edition-elizabeth-kay-berner/
ebookgate.com
https://ebookgate.com/product/geochemical-anomaly-and-mineral-
prospectivity-mapping-in-gis-1st-edition-emmanuel-john-muico-carranza/
ebookgate.com
https://ebookgate.com/product/an-introduction-to-generalized-linear-
models-third-edition-barnett/
ebookgate.com
Mathematical models in biology An introduction 1st Edition
Elizabeth S. Allman
https://ebookgate.com/product/mathematical-models-in-biology-an-
introduction-1st-edition-elizabeth-s-allman/
ebookgate.com
https://ebookgate.com/product/introduction-to-probability-models-9th-
edition-sheldon-m-ross/
ebookgate.com
https://ebookgate.com/product/an-introduction-to-ethics-deigh/
ebookgate.com
https://ebookgate.com/product/models-of-itinerant-ordering-in-
crystals-an-introduction-1st-edition-edition-jerzy-mizia/
ebookgate.com
https://ebookgate.com/product/design-of-experiments-an-introduction-
based-on-linear-models-1st-edition-max-morris/
ebookgate.com
“Don Rimstidt brings 30 years of teaching experience at a world class geosciences
Ri M st i Dt
department to the writing of this book. With its beautifully clear exposition of the
theoretical and practical aspects of geochemical kinetics, and use of ‘real-world’
examples, this book is essential reading for students and professionals with
interests in geochemical rate models.”
David J. Vaughan, Research Professor, University of Manchester, and
President of the Mineralogical Society of America
Geochemical
Yellowstone National Park. providing a solid understanding of geochemical kinetics. It will also provide
Image © Michael J. Thompson/ researchers and professional geochemists with a valuable reference for solving
Shutterstock.com. technical and scientific problems.
Rate Models
An Introduction to Geochemical Kinetics
rimstidt
Self-study problems and solutions
Full data spreadsheets for models used in figures and examples
Geochemical Rate Models
An Introduction to Geochemical Kinetics
J. Donald Rimstidt
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9781107029972
© J. Donald Rimstidt 2014
A catalogue record for this publication is available from the British Library
Preface page ix
1 Geochemical models 1
2 Modeling tools 9
3 Rate equations 36
4 Chemical reactors 56
5 Molecular kinetics 79
6 Surface kinetics 102
7 Diffusion and advection 128
8 Quasi-kinetics 156
9 Accretion and transformation kinetics 182
10 Pattern formation 205
References 210
Index 228
vii
ix
with actual outcomes to test the validity of the underlying knowledge. Every
scientist should understand and appreciate the importance of this model-
building and -testing scenario.
This book does not delve into computer models of rate processes because
there are already many good sources of information about this approach.
Instead it focuses on fundamental and relatively simple models that are the
foundations of the models incorporated into the various computer codes.
I have enjoyed the process of writing this book and learned a great deal
from teaching this material to students. I hope that the readers will find it
equally interesting and enlightening.
Modern geochemistry studies the distribution and amounts of the chemical elem-
ents in minerals, ores, rocks, soils, waters, and the atmosphere, and the circulation
of the elements in nature, on the basis of the properties of their atoms and ions.
(Goldschmidt, 1958)
The distribution and circulation of the chemical elements in and on the Earth
is influenced by a myriad of chemical and physical factors, many of which
have changed over geological time. Understanding the role of these factors
in geological processes requires us to condense information about elemental
abundances and distributions into models. This book is about geochemical
models for situations where time plays a key role. Geoscientists have always
appreciated the importance of time in fashioning the Earth. Many geological
processes require time spans that are far too long for human observation,
but we can use models to extrapolate rates based on short-term observations
to predict geochemistry in deep time. Equally important are models that
forecast the future behavior of geochemical systems because those models
are needed for environmental management and resources recovery projects.
Some of the models described in this book were developed by geochemists
but many others come from applied sciences and engineering. Because of
this diverse provenance, the models in their original form used a confusing
mix of units, terminology, and notation. This book attempts to remedy that
problem by recasting the models using internally consistent notation, units,
and terminology familiar to geochemists. Furthermore, whenever possible
the models are developed from fundamental theory showing a sufficient
number of intermediate steps to allow the reader to follow the derivations.
Thermodynamic models have been a mainstay of geochemistry since the
early twentieth century. They are especially effective for deep earth condi-
tions where local equilibrium conditions prevail. However, at and near the
Earth’s surface extensive amounts of mass transport and low temperatures
keep many reactions from reaching equilibrium. Kinetics models are needed
to properly describe these situations. This makes the models of kinetic and
dynamic processes described in this book complementary to thermodynamic
Model construction
Models organize our knowledge about the world. Doing science is the pro-
cess of making observations, using those observations to develop a model,
and then verifying the model’s effectiveness by comparing its predictions
with additional observations. The methods and scope of model-making
vary from discipline to discipline, but the goal of creating reliable predictive
tools is always the same. Developing a useful model requires a combination
of creative thinking and disciplined use of modeling tools. There are many
good discussions of modeling strategies (Aris, 1994; Bender, 1978; Bunge,
1997; Hall and Day, 1977; Hall et al., 1977; Harte, 1988; Overton, 1977;
Rescigno and Thakur, 1988; Shoemaker, 1977) but developing an effective
model is in many ways a creative process. Good models are elegant and
powerful. Elegance means that the model is expressed in the simplest, easi-
est to understand terms. Power means that the model explains the widest
Model reliability
No model is perfect (Oreskes et al., 1994). Even the best models fail under
some conditions. In addition, making mistakes is a natural part of the mod-
eling process. Systematic methods should be used to find, analyze, and cor-
rect mistakes and to define the valid range of a model. Determining the cause
of a model’s failure and repairing the model is a key task in model building.
Models are used to understand how complex behaviors arise from the
interaction of simple processes. Ideally models are built upon reliable prin-
ciples such as the conservation of matter, energy, and charge or the prin-
ciple of detailed balancing. Often less-reliable relationships such as empir-
ical rate equations must also be used. All quantitative models require input
data and relationships that come from measurements, which always contain
some error. This error can propagate through a model in unexpected ways,
especially if the model simulates nonlinear interactions. This challenges the
modeler who must decide whether an unexpected result is a legitimate pre-
diction or simply an artifact arising from an unfortunate combination of
errors. Three classes of errors occur in models. Formal errors are incorrect
assumptions and/or formulations. They include errors in the conceptual
foundation of the model as well as errors in the input data. Structural errors
are errors in mathematical manipulations such as programming errors or
algebraic errors. They include software bugs. Computational errors are
errors in numbers caused by incorrect rounding or by addition or multi-
plication errors. Because there are so many ways that a model might fail,
models and their predictions must be verified and validated to delineate the
bounds of their reliability.
Verification tests whether the model is internally consistent, incorpo-
rates the correct relationships in the correct ways, and uses correct data.
It is a good idea to develop a set of standards and practices that can be
used in geochemistry model-making to insure the validity of the resulting
model. Table 1.1 is an example of a checklist that might be used. Models
should use a consistent system of notation and units to avoid structural
errors. Equations should be tested using dimensional analysis. The input
data should be reviewed to insure its correctness and internal consistency.
Verification should determine the expected precision of the model’s predic-
tions based on the propagation of errors through the model. Error ana-
lysis is probably the most undervalued part of model development. For
simple models, error propagation can be done using simple algebraic meth-
ods described in Chapter 2. For complex models, simple error propagation
is often not practical. In such cases sensitivity analysis is a good strategy.
Sensitivity analysis uses various schemes to systematically vary, within
the expected range of error, the values of important modeling parameters
to see how those variations affect the model’s predictions. The utility of
Verification Questions
Validation Questions
Are the predictions geologically reasonable? Are they consistent with reasonable
estimates?
Are the predictions consistent with other scientific observations and knowledge?
Are the quantitative predictions sufficiently close to the behavior of one or more
natural analogs?
into the water table, the concentration of dissolved species increases to perhaps
500 mg/L. The total mass of rock and soil dissolved away each year is 5 × 104
mg or ~50 g. If the density of the rock and soil minerals is ~3 g/cm3, then 17 cm3
(= 1.7 × 10–5 m3) of material is carried away annually. If that volume is removed
from under the 1 m2 of land surface, the surface elevation will be lowered by 1.7
× 10–5 m (~0.02 mm). So the land surface elevation is reduced by 0.02 mm/yr or
~2 mm/century.
These estimates show that chemical weathering denudation rates should be
on the order of a few millimeters per century. They might be as high as a few
centimeters per century in humid tropical settings or as low as a few hundred
microns per century in arid cold environments. Because these estimates con-
strain model predictions to fall within this range of values, they can be used as a
validation test of a chemical weathering denudation model.
Harte (1988) and Weinstein and Adam (2008) provide detailed explanations
about how to make meaningful estimations and they give a large number of
practice examples.
Interpretation of results
The results of simple deterministic models are generally easy to interpret
and apply. For example, a model of radioactive decay will predict that the
radioactivity of a substance will be 1/8 of the original value after three
half-lives have passed. As models become more complex they become more
difficult to interpret. This suggests that model building should begin with
highly idealized and simplified cases that are easily understood and tested.
The next stage of model building involves testing the effect of potentially
important variables on the model output to determine which variables are
important enough to include and which can be neglected. This process of
building models in a stepwise fashion not only eliminates a large number of
unneeded independent variables but it helps the modeler develop a concep-
tual understanding of the situation being modeled.
The most important task of model building is to communicate expert
knowledge developed by the model builder to others who can use that
knowledge to solve problems. The modeling endpoint should be a report
that contains thorough documentation of the model and an explanation
of how to use its predictions. This report should contain a clear descrip-
tion of the model’s conceptual basis in a well-written narrative section with
accompanying illustrations and explanations of the mathematical methods.
The report should illustrate relationships between the model’s predictions
and input parameters using response maps. These graphs of predictions ver-
sus parameters are much easier to understand than equations and tables.
Many of the figures in this book are response maps. The report should also
describe the implementation of the modeling algorithm, including descrip-
tions of special computational methods. The rationale for selecting the
input data, including a discussion of their uncertainty, should be described.
The range and domain of the model should be specified. Finally, the report
should explain how to interpret the model’s output.
Before any model is ready for use it must be verified. The verification step
is greatly simplified if the model is constructed using conventional com-
putational methods, notation, and units. This chapter reviews some pro-
cedures and conventions that are recommended for geochemical model
construction.
1. Write the chemical formulae for the reactant and product species on opposite
sides of an equal sign.
2. Find the element that is least represented in these species and insert integer (or
sometimes small fraction) coefficients into the equation to make equivalent
amounts of this element on each side of the equation.
3. Repeat this step with the second least represented element and so on until all the
elements are balanced. For reactions involving aqueous solutions, it is usually
necessary to add H2O, O2, H2, or H+ to balance the hydrogen and oxygen.
4. Adjust the coefficients of the charged species to produce an equivalent amount
of charges on each side of the equation.
5. Finally, divide the entire equation through by integers to reduce these coefficients
to small integer values (although sometimes small fractions are acceptable).
1. We know that during the formation of acid mine drainage, pyrite is destroyed
and goethite is formed so the first step is to put pyrite on the left side of the
equation and goethite and sulfate on the right.
FeS2 = FeOOH + SO42−
3. However, there must be two sulfates on the right-hand side to balance the two
sulfur atoms in the pyrite:
FeS2 = FeOOH + 2 SO42−
4. Next, there must be 0.5 H2O on the left-hand side to balance the H in the
goethite on the right-hand side.
FeS2 + 0.5 H2O = FeOOH + 2 SO42−
5. Because the sulfur in the pyrite is reduced, with a formal charge of −1, and the
sulfur in the sulfate is oxidized, with a formal charge of +6, we know that an
oxidant, in this case O2, must occur on the left-hand side of the reaction. There
are already 10 oxygen atoms on the right-hand side of the reaction and only
0.5 on the left so 9.5 oxygen atoms (4.75 O2) are needed on the left-hand side.
FeS2 + 4.75 O2 + 0.5 H2O = FeOOH + 2 SO42−
6. Finally, four hydrogen ions are needed on the right-hand side to charge bal-
ance the sulfate. These can come from the addition of two more water mol-
ecules to the left-hand side.
FeS2 + 4.75 O2 + 2.5 H2O = FeOOH + 4 H+ + 2 SO42−
However, this adds too much oxygen to the left-hand side, so one O2 must be
removed.
FeS2 + 3.75 O2 + 2.5 H2O = FeOOH + 4 H+ + 2 SO42−
7. This reaction is often written with decimal fractions as shown above but the
coefficients can be written as simple fractions.
FeS2 + 15/4 O2 + 5/2 H2O = FeOOH + 4 H+ + 2 SO42−
Then the relationships between the coefficients that are required by the conser-
vation of mass and charge are expressed by simple equations.
Fe: a = d
S: 2a = f
O: 2b + c = 2d +4f
H: 2c = d + e
chg: e = 2 f
These equations are solved in a stepwise fashion to find the numerical value for
each coefficient.
Let a = 1.
From Fe: d = a = 1.
From S: f = 2a = 2.
From chg: e = 2f = 4.
From H: 2c = 1 + 4 so c = 5/2
From O: 2b = 2d +4f – c = 2 + 8 – 5/2 = 15/2 so b = 15/4
In all cases, the value for one of the coefficients must be arbitrarily chosen. The
other coefficients are then scaled to that value.
Notation
The notation used in this book follows geochemistry conventions as closely
as possible but has been modified where necessary to avoid redundancy or
confusion. Geochemical modeling requires an easy to recognize and intern-
ally consistent set of notation. Inconsistent notation complicates model
verification and leads to errors. The notation used in geochemistry is often
different from that used in engineering, chemistry, and biology, and some-
times models require the introduction of new notation. In all cases, notation
should be clearly defined in the model’s documentation and it is good prac-
tice to include a table of notation, with units, in every report.
Thermodynamic functions, especially free energy, enthalpy, and entropy,
are often used in kinetics models. Because thermodynamics is so widely used
in science, many different kinds of notation have developed. The convention
chosen for this book is illustrated using Gibbs free energy as an example.
For the free energy change related to a reaction, the delta in front of ΔGr°
means that this variable reflects the change in free energy in going from the
reactants to the products. The subscript, r , designates this variable as the
free energy of reaction. The superscript, °, indicates that both the reactants
and the products are in the standard state, 298.15 K (25°C) and either 1
atm or 100 kPa (1 bar) total pressure. For the free energy of formation of
a substance from the elements, the subscript, f, associated with ΔGf° means
that this is the free energy related to the formation from the elements (this is
essentially the same as ΔGr° except that the free energy of formation of the
elements in their most stable state at 298.15 K (25°C) and either 1 atm or
1 bar total pressure is defined as zero. It is often useful to use an appended
notation for ΔGf° and ΔGr° to indicate what reaction or substance the vari-
able refers to. For example, the free energy of reaction for a reaction listed
in equation (1) would be ΔGr°(1) and the free energy of formation of quartz
would be ΔGf°(qz).
Concentration and activity notation can be very confusing. Chemists have
long used square brackets, [i], to indicate the activity of species i and paren-
theses, (i), to indicate the concentration of species i, but geochemists have
occasionally reversed this convention. To make things worse, curly brackets
are used in some engineering disciplines to indicate activity. This confusion
is best avoided by abandoning the bracket-parenthesis notation in favor of
using ai to denote activity and mi to indicate molal concentration. Although
molar concentration units are best avoided in geochemistry, when used they
can be represented as Mi. Concentrations given other units such as mol/m3,
ppm, or mg/L units are best represented as ci.
pX notation indicates the negative logarithm (base 10) of quantity X.
Most people are familiar with pH, which is the negative logarithm of the
hydrogen ion activity, i.e. pH = –log aH+. By analogy pK = –log K and
pCl = –log aCl−.
Kinetics variable notation is quite different from discipline to discipline
and often inconsistent within a single discipline. In this book, ri refers to the
change in the number of moles of species i in a system per unit time (mol/
sec), Ri refers to the change in concentration of species i per unit time (molal/
sec), and Ji refers to the flux of species across an interface (mol/m2sec).
Units
The International System of Units was devised to facilitate communication
between scientific and engineering disciplines. Using SI units is highly rec-
ommended because it minimizes errors that arise from complicated units
conversions. The International System of Units is based on seven base units
(Table 2.1). Decimal multiples of these units and derived units are named
using the prefixes listed in Table 2.2.
Concentration
Concentration units listed in Table 2.3 are frequently used by geochemists.
Thermodynamic calculations are typically based on molal concentration
and concentration should be expressed in molal units whenever possible.
Some disciplines use molar or millimolar concentration units supposedly
because of the convenience of mixing of solutions using volumetric flasks.
Unfortunately molar concentrations change whenever the solution density
changes because of changing temperature, pressure, or composition. This
SI base units
length meter m
mass gram g
time second s
electric current ampere A
temperature (thermodynamic) kelvin K
amount of substance mole mol
luminous intensity candela cd
Derived units
volume liter 1 L = 10–3 m3
force newton 1 N = 1 kg m–1
sec–2
energy joule 1J=1Nm
calorie1 1 cal = 4.184 J
pressure pascal 1 Pa = 1 N m–2
bar 1 bar = 105 Pa
atmosphere 1 atm = 101325 Pa
power watt W
electrical charge coulomb C
electrical potential volt V
1
Widely used but not recommended.
Table 2.3. The most commonly used concentration units for aqueous solutions
Unit Definition
This means that both CO2(aq) and H2CO3(aq) are present in the solution.
Because there is no simple way to determine the concentration of these
individual species, the concentration of unionized, dissolved CO2 is usu-
ally reported as the sum of the concentrations of CO2(aq) and H2CO3(aq).
This convention is quite useful for thermodynamic calculations and the
reported equilibrium constants for reactions involving H2CO3(aq) as well
as the ΔGf°(H2CO3(aq)) are based on the convention of lumping these two
species together. However, this convention may cause a problem in kinetics
models because the rate of reaction of CO2(aq) is likely to be different from
the rate of reaction of H2CO3(aq).
Dimensional analysis
A dimension is a quantity such as time, mass, temperature, or distance,
that can be expressed numerically in terms of one or more standard units.
Dimensions have homogeneous linear scales so that a 2-meter interval mea-
sured at the beginning of a 1-kilometer traverse is identical in length to a
2-meter interval measured near the end of the traverse. The magnitude of
a dimension is expressed in terms of a pure number and a unit of measure.
Units are standards to which dimensional quantities are compared to obtain
a numerical ratio, which is the pure number. For example, comparison of
The universal gas constant can be converted from volume and pressure units
to energy units.
-2
8.314x10
314 dm 3 bar 1000 cm 3 1 J/bar 8.314 J
× × = (2.5)
mol K 1 dm 3 10 cm 3 mol K
The surface area to volume ratio of a sphere is used in several models in this
book. Dimensional analysis can be used to find the relationship between the sur-
face area (A) with dimensions of [L2] and the internal volume (V) with dimen-
sions of [L3] for any convex regular solid. The surface area is related to the vol-
ume by a dimensionless constant, b.
[ L2 ] b[ 3 23
] (2.6)
A bV 2 / 3 (2.7)
The first step toward finding the value of b, substitutes the definitions of A and
V for a sphere into Eq. (2.7).
2 /3
⎛ 4 3⎞
4π 2
πr ⎟ (2.8)
⎝3 ⎠
4π 4π 1/ 3
b= 2 /3
= 2 /3
= 4.84 (2.9)
⎛ 4⎞ ⎛ 4⎞
π 2 /3
⎝ 3⎠ ⎝ 3⎠
Logarithms
Logarithmic transformations are quite common in geochemical models. The
variables used in geochemical calculations often range over many orders
of magnitude so that it is more convenient to work with the logarithms of
these numbers rather than the numbers themselves. In addition, many rela-
tionships in thermodynamics and kinetics are linearized using logarithmic
transformations.
A logarithm is the power (p) to which a base (a) must be raised to produce
a number (N).
(2.10)
Note that the largest errors associated with this method are
for the logarithms of 3 and 6, both of which are incorrect by
~0.02 log units.
Errors
There is some uncertainty in all data, and model building must take this
error into account. The first step in error management is error detection,
error reduction, and error quantification. There are three types of error:
systematic error, random error, and blunders. Improved experimental pro-
tocol can reduce all these, but designing progressively better experiments
eventually leads to diminishing returns so that at some point it is necessary
to use some kind of error analysis to manage the uncertainty in the variable
being quantified.
Systematic errors affect the accuracy but not the precision of the result.
They are usually errors in calibration or observation where the same incor-
rect protocol is applied to all measurements. They displace all measurements
from the true value by the same amount so they cannot be detected by a
statistical analysis of only one data set. However, systematic error can be
detected and reduced by comparing data sets from several different sources
using meta-analysis and systematic review (Rimstidt et al., 2012).
Random error displaces individual measurements from the true value, but
as the number of determinations increases their average becomes closer to
the true value. In a series of independent measurements that contain only
random errors, the most reliable estimate of the true value is the mean of all
the measurements. So that for a series of N measurements, x1, x2, x3, … xN,
the best estimate of the true value of x is x .
∑x i
i =1
x= (2.11)
N
∑ (x x)
2
i
σx = (2.12)
N −1
σx
σx = (2.13)
N
xs x
ts = (2.14)
σs
3. Use a probability table to determine the probability (Ps) that a normally distrib-
uted measurement will differ from the mean by ts standard deviations (Ps = 1
− Ptable).
4. Determine the number of measurements in the data set that are expected to be as
bad as xs by multiplying the total number of data (N) by the probability (Ps).
Ns NP
NPs (2.15)
5. If Ns < 0.5, then xs fails Chauvenet’s criterion and is rejected along with all data
that differ from the mean by more than xs.
6. After the data are rejected, calculate a new mean and uncertainty using the
remaining data.
7. Chauvenet’s criterion can be applied to a data set only one time.
Significant digits
The significant digits in a number give an indication of the uncertainty asso-
ciated with that value. This means that the number has been rounded to the
same order of magnitude as its uncertainty. The number of significant digits
is determined by the following rules:
1. The leftmost nonzero digit is the most significant one.
2. In numbers with no decimal point, the rightmost nonzero digit is the least
significant one.
3. In numbers with a decimal point, the rightmost digit is the least significant one,
even if it is a 0.
4. All digits between the least and most significant are counted as significant digits.
For example, 65000 has two significant digits; 0.1010 has four significant
digits; and 65000.1010 has nine significant digits. When numbers are trun-
cated to the proper number of significant digits, the least significant digit
must be rounded up or down according to the following rules:
1. If the digit following the one to be retained is 6, 7, 8, or 9, round the retained digit
up by one.
2. If the digit following the one to be retained is 0, 1, 2, 3, or 4, simply drop it with-
out changing the retained digit.
3. If the digit is 5 and the digit to be retained is even, do not change it. If it is odd,
round it up by one.
reported uncertainty, the least significant digit in the stated value should be
of the same order of magnitude (in the same decimal position) as the most
significant digit of the uncertainty. It is customary to perform the calcula-
tion using numbers with additional significant digits, to avoid introducing
rounding errors and then rounding the final answer to the correct number of
significant digits. For a number without a reported uncertainty, one should
assume that all reported digits are significant. Geochemical measurements
typically have more than 1% error so that, if no uncertainty is stated, three
significant digits are likely to be sufficient to express the reliable part of the
number.
Propagation of errors
Precision is lost as variables are manipulated through formulae. This loss of
precision is taken into account by propagating the uncertainty associated
with the model’s variables through the equations. There are several standard
references that provide detailed discussions of error propagation and man-
agement (Bevington, 1969; Deming, 1943; Mandel, 1964; Taylor, 1982).
For the propagation of uncertainties in equations with addition or sub-
traction, the uncertainty in the calculated value is the square root of the sum
of the squares of the absolute uncertainties.
xyz
q= (2.18)
abc
2 2 2 2 2 2
δq ⎛ δx⎞ ⎛ δ y⎞ ⎛ δz⎞ ⎛δ ⎞ ⎛ δb⎞ ⎛ δc ⎞
= ⎜ ⎟ +⎜ ⎟ +⎜ ⎟ +⎜ ⎟ +⎜ ⎟ +⎜ ⎟ (2.19)
q ⎝ x⎠ ⎝ y⎠ ⎝ z⎠ ⎝ a⎠ ⎝ b⎠ ⎝ c⎠
q Bx (2.20)
δq δx (2.21)
q q (x) (2.22)
dq
δq δx (2.23)
dx
dq = (10 x ) ( ) δx (2.24)
δ ( (l )
)δ log x (2.25)
log e
δq δx (2.26)
x
0.434 δx
δ( g ) δ x = 0.434 (2.27)
x x
q xn (2.28)
δq δx
= n (2.29)
q x
q f (x y z) (2.30)
2 2 2
⎛ ∂q ⎞ ⎛ ∂q ⎞ ⎛ ∂q ⎞
δq δx + ⎜ δ y δz (2.31)
⎝ ∂x ⎟⎠ ⎝ ∂y ⎠ ⎝ ∂z ⎟⎠
δq δx δ y δz
≤ + + + (2.32)
q x y z
δ q = 2 2 + 2 2 + 2 2 = 3.46 (2.33)
This means that the mass of the composited sample is 725(3.46) mg.
Multiplication and/or division: The rate of reaction in a mixed flow reaction is
calculated by multiplying the concentration (mol/g) of a chemical species in the
effluent solution by the flow rate (g/sec). If the concentration is 2.31(0.0122) ×
10−4 mol/g and the flow rate is 0.124(0.00131) g/sec, the rate is 1.53 × 10−5 mol/sec
and the uncertainty associated with this value is found using (2.19).
2 2
δq ⎛ 1.22 × 10 −6 ⎞ ⎛ 0.00131⎞
= ⎜ −4 ⎟
+⎜ ⎟ = 0.0118 (2.34)
q ⎝ 2 .31 × 1 0 ⎠ ⎝ 0.124 ⎠
The fractional error in the rate is 1.18% so the absolute error is (0.0118)(1.53 ×
10−5) = 1.81 × 10−7. This means that the rate is 1.53(0.0181) × 10−5 mol/sec.
Multiplication by a constant with no error: A common laboratory activity is to
prepare a solution by weighing a mass of solid and then dissolving it in water.
The concentration is usually reported in terms of the number of moles of solid
rather than the mass. If the mass of NaCl (WM = 58.44) is 40.00(0.01) g, the
number of moles is 40.00/58.44 = (40.00)(0.0171) = 0.6845 moles. We can assume
that there is no random error associated with the molecular weight of NaCl, so
the error for this value is found using Eq. (2.21).
This conversion shows that the amount of NaCl added to the solution is
0.6845(0.00017) mol.
Regression models
Fitting data to equations is an important modeling activity that is supported
by an abundance of computer software. The mathematical basis for this
software is described in many statistics books and will not be elaborated
here. Methodologies that are especially suited for kinetics data are reviewed
in Chapter 7 of van Boekel (2009). This section explains some strategies that
can guide fitting kinetic data to equations.
One reason to fit data to a function is to summarize the data and report it
in a way that allows easy interpolation or other mathematical manipulation.
If this is the main objective, the data can be fit to an arbitrary function so
long as that function makes a smooth and parsimonious summary of the
data and is compatible with desired mathematical manipulations. Typically
a low-order polynomial is suitable for this purpose. It is important to avoid
over-fitting the data. The goal is to filter out random error without losing
the information contained in the data. Increasing the number of regression
variables tends to fit the data better as reflected by smaller residuals and a
higher correlation coefficient. This is because the additional variables make
the fitting function more flexible, with more minima and maxima, so it can
pass closer to the data points. However, this additional flexibility can cause
the function to flex wildly and to predict values that are beyond the range
of the data. This unreasonable result can be avoided by beginning the fitting
process using the simplest feasible equation and then adding variables one
at a time until the residuals of the fit are about the same size as the errors in
the data. Sometimes data can be converted to a linear or other simple form
using a logarithmic or power transformation. The fitted equation should not
be extrapolated beyond the domain of the original data.
Table 2.4 lists Fe3+ concentration versus contact time with pyrite from a plug
flow reactor experiment (PFR10) (Rimstidt and Newcomb, 1993). The rate of
ferric iron consumption is the time derivative of these data. Finding this deriva-
tive for t = 0 is especially useful because the initial conditions of the experiment
are accurately known so their effect on the rate can be clearly established. One
strategy for finding the initial rate is to fit the first few concentration versus time
0 1.80 × 10−3
4.7 1.16 × 10−3
6.3 1.12 × 10−3
9.1 1.04 × 10−3
20.5 8.24 × 10−4
0.002
0.0015
• •
mFe3+, molal
0.001 •
0.0005
0
0 5 10 15 20 25
t, sec
Figure 2.1. Fit of the PFR10 data to a second-order polynomial (solid line, R 2 =
0.96) and a fourth-order polynomial (dashed line, R 2 = 1.00).
dmFe3+
=b ct + ... (2.39)
dt
When t = 0, the rate equals b. Figure 2.1 shows fits of a second and a fourth-
order polynomial to the data.
mFe3+ 1.73 10 3
1.23 10 4 t + 3.80 10 −6 t 2 (2.40)
mFe3+ 1.80
8 10 3
3.07 10 4 t + 5.34
3 × 10 5 t 2 4.11 10 6 t3 + 1.04 × 10 −7 t 4
(2.41)
The second-order fit predicts an initial rate of −1.23 × 10−4 m/sec and the fourth
fit predicts a much faster rate of −4.85 × 10−4 m/sec. Both functions fit the data
well and the fourth-order polynomial fits it so well that the function passes
through all the points so there are no residuals. However, the rate based on the
fourth-order polynomial appears to be too fast because that function does not
smooth out the random errors in the data. Both functions incorrectly predict
that the concentration of Fe3+ increases after ~18 seconds of reaction because
they are simply approximations for a short interval of the true concentration
versus time function.
A second reason to fit data to a function is to test whether the data are con-
sistent with a model or to use a theoretical function to extrapolate experimen-
tal results to conditions otherwise not attainable. The goodness of fit of the
data to a theoretical function can be gauged by the coefficient of determin-
ation (R2), which is the correlation coefficient squared. R2 is often interpreted
as the fraction of the variability of the response variable that is explained
by its functional relationship to the independent variable. For example, if
R2 = 0.8 then 80% of the variation in the response variable is explained by
the model and 20% of the variation is the result of factors, including ran-
dom error, that are not part of the model. Some geochemical processes occur
under conditions or over time intervals that are difficult to simulate by exper-
iments, making it necessary to use a theoretical model to predict their behav-
ior. This approach uses data from experiments performed under easily attain-
able conditions to quantify parameters that are used to calibrate a theoretical
model. These parameters are then used to predict the value of the variable at
0.8
0.6
T, °C k, sec–1
–6
–8
•
••
–10 •
–12
log k
–14
–18
100°C
–20
0.0015 0.002 0.0025 0.003 0.0035
1/T, K–1
Figure 2.3. Arrhenius plot used to extrapolate rate constants for the kaolinite to
illite transformation reaction from high temperature measurements to find the
rate constant at 100°C.
k Ae − Ea RT
(2.42)
⎛ − Ea ⎞ ⎛ 1 ⎞
log log A + ⎜ (2.43)
⎝ R ⎟⎠ ⎜⎝ T ⎟⎠
Fitting the data in Table 2.5 to this equation gives the line shown in Figure 2.3.
This fit explains 97% of the variation in log k (R2 = 0.97).
−8039( )
log k = + 5.. (1. ) (2.44)
T
δ log k = ( . )2 + ( T ) = 2.67
2
(2.45)
So log k = −16.1(2.7). Even though the data fit the equation quite well, the
uncertainty in log k is nearly 3 log units. Most of this error is due to the uncer-
tainty associated with the slope of the line. This example stands as a warning to
model builders who extrapolate data without considering error propagation.
Numerical differentiation
Kinetic processes cause a change in a quantity over time. If this quantity is
measured at evenly spaced time intervals, it is relatively easy to find the rate
by computing a numerical derivative (Pollard, 1977). The simplest method
of numerical differentiation is based on the observation that the chord of a
graph has a slope that closely approximates the tangent at an intermediate
point.
This leads to a simple scheme for calculating the numerical derivative
from a data set (Pollard, 1977). The derivative at the first point in the data
−′ ) gives more weight to the first two points.
set ( f−1
1
f −′1 = (− + − ) (2.46)
2h
The derivative at each intermediate point ( f0′ ) is the slope of the chord con-
necting the surrounding points (Figure 2.4).
1
f0′ = (− + ) (2.47)
2h
1
f+′1 = ( − + ) (2.48)
2h
f–1 f0 f+1
h
Note that the equations for the endpoints of the data set are less dependable
than the equation for the intermediate points and they tend to give unrea-
sonable results if the slope is steep. Because numerical differentiation takes
the difference between values of the dependent variable it magnifies the scat-
ter in those data. Some of this scatter can be eliminated by various smooth-
ing schemes, but they tend to bias the data and are best avoided.
Ferric iron reacts with pyrite to produce ferrous iron and sulfuric acid.
14 Fe3+ + FeS2(py) + 4 H2O = 15 Fe2+ + 2 SO42− + 8 H+ (2.49)
⎛ ⎞
−′ = ⎜
f −1
⎝(
1
)( ) ⎟⎠
(−3(9.79 × 10 ) + (9.6 −4
× 10 −4 ) − 9.45 × 10 −4 ) = −6.33 × 10 −8
(2.50)
The rate for the next point (t = 300) is found using Eq. (2.47).
⎛ 1 ⎞
f0′ = ⎜ ( −9.79 10 4 + 9.45 × 10 −4 ) = −5.67 × 10 −8 (2.51)
⎝( )( ) ⎟⎠
The subsequent points are treated the same way. The rate for the last point (t =
3000) is found using Eq. (2.48).
⎛ ⎞
+′ = ⎜
f+1
⎝(
1
)( ) ⎟⎠
(9.04 × 10 4 − 4 (8. 8 × −4
) + 3 (8.91 × 10 −4 )) = −2.50 × 10 −8
(2.52)
9.5× 10–4
concentration, molal
9.0× 10–4
8.5× 10–4
0 1000 2000 3000
time, sec
(b) 0.0
–2.0 × 10–8
rate, molal/sec
–4.0 × 10–8
–6.0 × 10–8
Figure 2.5. (a) Concentration of Fe3+ versus time from experiment BR5 (Rimstidt
and Newcomb, 1993). (b) Numerical derivatives of the concentration of Fe3+
versus time data. Early in the experiment the rates are fast but they slow to a
nearly constant rate after ~300 sec.
ebookgate.com