American Mineralogist, Volume 85, pages 217–224, 2000

A shell model for the simulation of rhombohedral carbonate minerals

and their point defects

DIANA K. FISLER,1,* JULIAN D. GALE,2 AND RANDALL T. CYGAN1,†

1
Geochemistry Department, Sandia National Laboratories, Albuquerque, New Mexico 87185-0750, U.S.A.
2
Department of Chemistry, Imperial College of Science, Technology, and Medicine, South Kensington, London, SW7 2AY, U.K.

ABSTRACT
The electronic polarization of oxygen ions has been explicitly incorporated in a shell model to
better simulate the structure of calcite and related rhombohedral carbonate minerals. Pair-potentials
for Ca2+ ions and C and O comprising the carbonate molecular ion were simultaneously fitted to
experimental lattice, elastic, dielectric, and vibrational data for calcite, and the structure and elastic
properties of aragonite. The resulting potential parameters for the CO32– group were then transferred
to models for the structures and bulk moduli of the carbonate minerals incorporating Mn, Fe, Mg,
Ni, Zn, Co, Cd, and thus a fully consistent set of interaction parameters for calculating the properties
of the carbonate minerals was obtained. Defect energies for doping the divalent cations into the
calcite structure, and for calcium and carbonate ion vacancies were calculated. In addition, various
disorder types for dolomite, including anti-site defects, stacking defects, and the energy related to
increasing the Ca/Mg ratio in the dolomite structure were simulated. The theoretical enthalpy for
dolomite ordering (34.4 kJ/mol) compares very well with experimental measurements.

INTRODUCTION mathematically equivalent hexagonal or rhombohedral struc-

The value of atomistic simulations in the earth and materials tures. The structure may also be regarded as being comprised
sciences lies in demonstrating the mechanisms of atomistic pro- of basal layers of metal cations and layers of planar CO2–
3 groups

cesses, and extending this capability to evaluate material prop- normal to the c axis. In the mixed metal cation dolomite struc-
erties to regimes where direct laboratory measurements are ture, Ca2+ and Mg2+ preferentially occupy the alternate hexago-
difficult or impossible to perform. Knowledge of the transport nal basal planes and form calcium and magnesium layers. Larger
properties of cations in carbonate minerals is necessary to evalu- cations such as Sr2+, Pb2+, and Ca2+ at higher pressures form the
ate order/disorder relations, deformation, and geothermometry, orthorhombic aragonite structure, also composed of ionic CO2– 3

but slow transport rates and the decomposition of carbonate units and the metal cations.
phases at high temperatures limit experimental evaluation. Al- Previous molecular mechanics studies (Pavese et al. 1992;
though simulations do not replace careful experimental mea- Dove et al. 1992; Catti et al. 1993; Catti and Pavese 1997) gen-
surements, a well-parameterized atomistic model can provide erally focused on parameterizing a model for only CaCO3. These
a solid framework for evaluating mechanisms and predicting studies included extensive testing of both rigid ion and shell
activation energies and rates for diffusion under conditions not models of the lattices of calcite and aragonite, and calculated
accessible to the experimentalist. This study develops a fully basic properties such as elastic and optical properties, and iso-
consistent model of the carbonate minerals based upon a shell tope fractionation. Parker et al. (1993) used the atomistic model
description of the electronic structure of the oxygen in the co- for calcite to examine surface precipitation and dissolution pro-
valent CO2– 3 group. Ultimately, the model may be used to cal-
cesses, an area of wide application for geologic problems in-
culate chemical diffusion rates, phase relations, and surface volving the alteration and weathering of carbonate rocks such
reactivity. as limestone, dolomite, and marble.
The divalent cations Ca, Mg, Mn, Fe, Cd, Zn, Co, and Ni By calculating the properties of most of the known divalent
form stable carbonate phases of the calcite structure, respec- carbonate phases, we have created a general framework for
tively, calcite, magnesite, rhodochrosite, siderite, octavite, interpreting a variety of bulk carbonate properties, and for be-
smithsonite, and cobalt and nickel carbonates. They share the ginning to address the problems of inhomogeneous zoning and
dolomitization. The fully transferable potentials for the CO2– 3
group, in particular, lend a great deal of flexibility to the
empirical models, such that substitution defect energies and
*Present address: Johns Manville Technical Center, 10100 W. disorder defect energies can be derived. Additionally, the mod-
Ute Avenue, P.O. Box 625005, Littleton, Colorado 80162-5005, els can be used to evaluate transition state geometries and pre-
U.S.A. dict Arrhenius activation energies for chemical diffusion in the
†E-mail: [email protected] carbonate phases (Fisler et al. 1998).
Brought to you by | New York University Bobst Library Technical Services
0003-004X/00/0001–0217$05.00 217
Authenticated
Download Date | 7/7/15 8:46 PM
218 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS

THEORETICAL PROCEDURE Calcium carbonate minerals

Interatomic energy and shell model The empirical fitting of model parameters often leads to a
The simulations performed in this study use a modification non-unique set of values that can easily reproduce the lattice
of the Born model description of solids, which treats the ionic parameters for a given structure. However, it is much more
material as a collection of point ions with electrostatic (Cou- desirable and accurate to incorporate as many high quality
lombic) and short-range forces acting between them. The pro- measurements of observable properties, such as elastic and di-
cedure is described at length elsewhere (Catlow and Mackrodt electric constants, as possible in the fitting of the parameters to
1982). In brief, the different energy contributions for all ion- maximize transferability. The structures of calcite (rhombohe-
ion interactions of the crystal are summed out to a maximum dral CaCO3) and aragonite (orthorhombic CaCO3), the elastic,
cut-off of 15 Å for the short-range terms (if there is no C6 term static and high frequency dielectric constants and vibrational
then these can be truncated at 10 Å with no significant loss of frequencies of the carbonate deformation modes of calcite have
accuracy) whereas electrostatic interactions are partially per- been used to fit the potential parameters using the relaxed fit-
formed in reciprocal space according to the Ewald method to ting algorithm (Gale 1996) in which the changes in the struc-
ensure proper convergence (Tosi 1964). The electrostatic po- tural parameters on optimization are used to calculate the
tential Vijel is inversely proportional to the distance between ions: residual error.
The chief improvement of our model for calcium carbonate
Vijel(rij) = e2ZiZj/rij (1) over the previous studies (Pavese et al. 1992; Dove et al. 1992;
where rij is the distance between ions having charge Zi and Zj, Catti et al. 1993; Parker et al. 1993) is recognizing that com-
and e is the charge of an electron. The short-range potential Vijsr pletely different interaction potentials are required between and
is expressed as a function of the interatomic distance based on within CO32– molecular anions. Therefore, an intermolecular O-
the Buckingham functional form: O Buckingham potential and an intramolecular O-O Buckingham
potential have been defined. A three-body O-C-O term and a
Vijsr(rij) = Aijexp(–rij/ρij) – Cijrij–6 (2) torsional O-C-O-O potential simulate additional interactions
where Aij, ρij, and Cij are the empirical parameters for the short- within the CO2– 3 molecular anion. A more extensive functional

range attractive and van der Waals energies. The potential form form for the oxygen core/shell spring constant by including a
differs from that usually used in the Born model in that elec- 4th order term has been used to accurately reproduce the vi-
tronic polarization of ions in the lattice is accounted for by the brational frequencies in calcite, while predicting the relative
addition of charged shells (Dick and Overhauser 1958). The ion stability of aragonite with respect to calcite which is critically
is therefore represented by a massless shell of charge Zs and a dependant on the oxygen polarizability. All intramolecular po-
core of charge Zc, where Zs + Zc equals the total charge of the tentials are defined to act between cores while intermolecular
ion. The total charge on each ion in these simulations is reduced interactions act on shells, where applicable. Furthermore, all
from its formal ionic charge and the partial charge on each ion is Coulomb interactions are excluded within molecules. Table 1
a fitted parameter. The core and shell are coupled by a harmonic provides the values for the optimized interatomic potential pa-
spring of analytical form Epolar = 1/2 kX2, where X represents the rameters obtained through this process.
separation distance of the shell and core of an ion. During the Simulations of the bulk structure of calcite (and metal-sub-
energy minimization calculations, the shell positions are allowed stituted derivatives) use the primitive rhombohedral cell, al-
to relax about the ion core resulting in a dipole that mimics the though lattice parameters are reported here in hexagonal
electronic polarization. Shells are only included for describing coordinates as the two representations are mathematically
the large polarization associated with the oxygen ions; all other equivalent. The aragonite and the Mg-rich huntite phases re-
atoms were represented by a rigid ion. quire an orthorhombic simulation cell. The computer program

TABLE 1. Interionic potentials obtained from fits to calcite and aragonite structures and properties
Potentials Aij rij Cij eV Cutoffs
(eV) (Å) (Å6) (Å)
Buckingham (intramolecular) O-O 4030.30 0.2455 0–2.5
Buckingham (intermolecular) O-O 79158.6 0.1983 21.844 0–15
Buckingham Ca-O 2154.06 0.2891 0–10
Buckingham Cd-O 4329.81 0.2563 0–10
Buckingham Mn-O 2000.94 0.2727 0–10
Buckingham Fe-O 2151.99 0.2651 0–10
Buckingham Zn-O 1029.39 0.2891 0–10
Buckingham Co-O 1095.60 0.2863 0–10
Buckingham Ni-O 1634.46 0.2666 0–10
Buckingham Mg-O 1039.59 0.2893 0–10
Notes: For Morse C-O, De = 5.0 eV, a = 2.5155 /Å, and r0 = 1.2025 Å.
For spring O-O, k2 = 32.194 eV/Å2 and k4 = 10000 eV/Å4.
For three body O-C-O (intramolecular), k2 = 1.7887 eV/rad2, and θ0 = 120°.
For torsion O-C-O-O, k = 0.1510 eV and n = –2.
Partial charges are ZC = +1.3435, ZO-core = +1.0185, ZO-shell = –2.1330, and ZCa = +2.0000.

Brought to you by | New York University Bobst Library Technical Services

Authenticated
Download Date | 7/7/15 8:46 PM
 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS 219

GULP (Gale 1997) was used to first derive the potential pa- implicitly by lattice summation techniques. We used the above
rameters and then to perform the energy minimizations required procedure to examine a variety of point defects in the carbon-
for the bulk structural and defect calculations. Energy minimi- ate phases, and performed several preliminary evaluations of
zations were performed at constant pressure allowing all indi- more complex defect aggregates.
vidual ionic coordinates and lattice parameters to vary, while
constraining the space group symmetry. Although the theoreti- RESULTS
cal results reported here are 0 K calculations, in practice, ther- Table 2 compares the properties of calcium carbonate used
mal effects are effectively included in the interatomic potentials to develop the shell model potentials (Table 1) by the described
by fitting to room temperature experimental data. procedure. Table 2 also reportes measured properties that were
not used in the fit for comparison with the calculated values,
Other rhombohedral carbonate minerals and calculated properties for which there are no reliable ex-
There are few high quality measurements of the structural perimental data. The vibrational frequencies were calculated
and physical properties for the other divalent carbonate phases. at the gamma point (zero wavevector) and therefore we do not
The interatomic potentials and charges for the carbonate anion report dispersion curves and TO-LO modes. For degenerate
group and the cation charge (+2) were held fixed for fitting to vibrational modes in calcite both calculated frequencies are
the structure and the bulk modulus for the other carbonate reported. In principle, all of the vibrational modes of calcite
phases; the only degree of freedom for the fitting model was in and aragonite could be used in the parameterization procedure,
the cation-oxygen Buckingham potential. Due to the non-di- but due to the cooperative motions of a large number of atoms
rectional nature of the metal cation as described by the in the low frequency region, and the uncertainty in assigning
Buckingham potential of Equation 2, there is no explicit term the eigenvalues for each experimental mode, we have chosen
that addresses the d-electrons of the transition metal cations. to rely on the high frequency localized modes of the carbonate
The transition metal cations are represented as spherical rigid ion. This ensures that the parameterization properly fits vibra-
ions (no electronic polarization) as with Ca2+ and Mg2+, and there- tional frequencies to the correct internal modes. The low fre-
fore magnetic effects are ignored. The data for the bulk moduli quencies that are primarily associated with Ca-O modes are
were obtained from Zhang and Reeder (1999). Finally, struc- incorporated in the model by fitting the elastic properties rather
tures for dolomite [CaMg(CO3)2] and huntite [Mg3Ca(CO3)4] than the vibrational frequencies.
were simulated, with no attempt to further optimize the poten- The lattice parameters differ from reported values by less
tials with respect to those observed structures. than experimental error for calcite, and those for aragonite dif-
fer by less than 1% (see Table 2). The elastic constants for cal-
Defect calculations cite are reproduced with a relative error of approximately 6.6%
A large number of defect calculations have been performed including those not used in the fitting procedure. The elastic
on oxides and halides (e.g., Catlow and Mackrodt 1982; constants of aragonite are reproduced with about the same ac-
Mackrodt 1984). Defect calculations of covalent materials are curacy as previously published rigid ion models (e.g., Pavese
significantly less straightforward, but in examining calcite we et al. 1992), with the most obvious failure in the off-diagonal
are able to treat the structure as a de facto ionic solid and create C12 and C13 constants. The relatively poor agreement in these
ionic defects involving the metal cation and the CO2– 3 molecu- off-diagonal terms is related to several factors: the emphasis of
lar anion. Mott and Littleton (1938) demonstrated that the dis- parameterization in the weighting on the diagonal terms, the
placement around a defect may be treated mathematically as a structure, and relative energies of the two phases; the relative
combination of short range and long range displacement re- inaccuracies in the experimental data for aragonite compared
gions. In this formalism, the lattice energy will almost certainly to calcite; and the relatively simple nature of the bonding model.
converge, as is not always possible if all ionic displacements in Overall, the new shell model provides a significant improve-
the full lattice are allowed. The defect formation energy can ment in the prediction of the second derivative properties, in-
then be calculated from the difference between the energies of cluding the elastic and dielectric constants and the vibrational
the defective lattice and the perfect lattice. The size of region 1 frequencies, compared to the previous models.
(the atomistic region immediately surrounding the defect) is
increased until the energy of the defect has converged satisfac- Structures of the carbonate minerals
torily. The techniques used in defect calculations resemble those The calculated lattice parameters for the end-member car-
used in bulk lattice simulations except for the high degree of bonates (used in the fit) differ from their measured lattice pa-
distortion in the environment of the defect, and the fact that the rameters by less than 1% (Table 3). The relative errors in lattice
calculation can use only point group symmetry, if any exists, parameters for the divalent carbonates are all slightly smaller
in the defective region. Region 1 typically contains up to a few than the relative error (2.1%) for calcite lattice parameters by
hundred atoms, and all coordinates of the ions in this region the rigid ion model of Pavese et al. (1992). Lattice parameters
are explicitly allowed to relax. In region 2a, which may con- for dolomite and huntite differed from measured values by less
tain several thousands of atoms, the relaxation energy is calcu- than 2%, although no attempt was made to fit to those struc-
lated assuming that the ions respond harmonically about their tures directly. The lattice energy generally increases with in-
lattice sites to the electrostatic forces due to the defect species. creasing ionic radius (Fig. 1). The minor deviations of the lattice
Beyond this in region 2b, the ions respond only to the total energies from the trend are attributed to the anomalous experi-
charge of the defect and the relaxation energy is calculated mental bulk moduli with respect to bond length (and ionic ra-
Brought to you by | New York University Bobst Library Technical Services
Authenticated
Download Date | 7/7/15 8:46 PM
220 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS

TABLE 2. Experimental and calculated properties of CaCO3

Aragonite Calcite
Experimental Calculated ∆ (%) Experimental Calculated ∆ (%)
Elastic Constants (GPa)
C11 85.0† 89.9 6.5 145.7§ 140.9 –3.3
C22 159.6† 155.3 –2.7
C33 87.0† 104.2 20 85.3§ 85.8 0.6
C44 42.7† 23.3 –45 33.4§ 33.4 0.0
C55 41.3† 36.7 –11
C66 25.6† 12.4 –52
C12 15.9†* 48.0 >100 55.9§ 63.7 14
C13 36.6†* 55.9 53 53.5§* 62.6 17
C14 –20.5§* –19.5 –4.9
C23 2.0†* 54.7

Static Dielectric
ε011 7.84 8.5|| 9.28 9.2
ε022 22.48
ε033 8.26 8.0|| 8.30 3.7

High Frequency Dielectric

ε11
∞
2.86‡* 3.05 6.6 2.75‡ 2.69 –2.2
ε∞22 2.82‡* 2.53 –10
ε∞33 2.34‡* 2.50 6.8 2.21‡ 3.02 37
–1
Vibrational frequencies (cm )
Asymmetric stretch (IR), ν3 1473# 1500 1.8 1463# 1465 3.8
1573
Symmetric C-O stretch (Raman), ν1 1086# 1124 3.5 1088# 1082 –0.1
1091
Torsional bending (IR), ν2 873# 781 –11 881# 878 6.2
994
Stretch/bend (IR), ν4 705# 627 –11 714# 612 –14
* Properties not used in the potential fitting.
† Experimental values from Hearmon (1946).
‡ Experimental values from Deer et al. (1966).
§ Experimental values from Dandekarand Ruoff (1968).
|| Experimental values from Kaye and Laby (1982).
# Experimental values averaged from White (1974).

dius) observed by Zhang and Reeder (1999) and which were cite and aragonite and the corresponding enthalpies determined.
used in the parameterization. Simple variation of the cation- Because of the differing degrees of oxygen polarization in the
oxygen potential parameters failed to satisfactorily reproduce two structures the enthalpy increases at different rates as a func-
the exact trend in bulk moduli. tion of pressure and a transition from calcite to aragonite is pre-
The lattice energy of aragonite is less exothermic per mole dicted to occur at approximately 2.4 kbar consistent with
than calcite at ambient pressure, consistent with the metasta- experimental measurements (Crawford and Hoersch 1972).
bility of aragonite at room pressure. A series of minimizations
have been performed at increasing applied pressure for both cal- Defect energies
The Mott-Littleton method requires that the defect energy
has properly converged to a value consistent with an infinitely
large region 1 volume. The energies for each type of defect
(Table 4) demonstrate convergence to within 0.02 eV for sub-
stitutional defects (0.1 eV for vacancies) with an increase in
the size of region 1 from 6 to 12 Å. Region 2a was held fixed at
20 Å; varying this radius had a negligible effect on the defect
energy. Figure 2 includes two views of a calcium vacancy in
calcite and shows the limiting extents of the different calcula-
tion regions. The expanded view (Fig. 2b) shows that the sig-
nificant distortion due to a calcium vacancy reaches less than
10 Å into the surrounding crystal. Substitutional defects (e.g.,
Mn for Ca) show much less distortion of the calcite lattice. The
amount of distortion around a calcium vacancy in calcite is
quantified by comparing oxygen-oxygen distances in the vi-
cinity of the defect with those in the perfect lattice (Fig. 3).
The O-O distances within the carbonate anion near the defect
F IGURE 1. Calculated lattice energies for the rhombohedral are seen to contract, whereas the intermolecular distances are
carbonate minerals as a function of ionic radius of the divalent cation.observed to expand and also exhibit a broader range of values.
Brought to you by | New York University Bobst Library Technical Services
Authenticated
Download Date | 7/7/15 8:46 PM
 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS 221

TABLE 3. Comparison of carbonate structural properties

Bulk modulus
Composition Lattice energy Calculated Experimental ∆ a ∆a b ∆b c ∆c
(eV) (GPa) (GPa) (%) (Å) (%) (Å) (%) (Å) (%)
CaCO3 (calcite)‡ –85.31 77 73§§ 5.5 4.99 0.01 4.99 17.06 –0.01
CaCO3 (arag)† –170.60 71 48## 47.9 5.75 0.31 4.94 0.48 7.92 0.61
CaCO3 (arag)‡‡ –85.30
CdCO3§ –87.76 98 100* –2.0 4.90 –0.51 4.90 16.39 –0.63
MnCO3‡ –89.55 108 108* 0.0 4.78 0.29 4.78 15.56 –0.51
FeCO3‡ –90.79 121 117* 3.4 4.73 0.80 4.73 15.25 –0.87
ZnCO3‡ –91.30 123 123* 0.0 4.67 0.45 4.67 14.88 –0.94
CoCO3|| –91.37 124 124* 0.0 4.67 0.25 4.67 14.89 –0.51
NiCO3|| –92.59 138 131* 5.3 4.64 0.65 4.64 14.73 –0.03
MgCO3‡ –91.16 122 107* 14.0 4.68 –0.96 4.68 14.92 0.69
MgCa(CO3)2** –88.32 94 91* 3.3 4.85 0.89 4.85 15.82 –1.16
1
⁄2[Mg3Ca(CO3)4]†† –179.68 110 9.64 1.40 9.64 7.70 –1.50
1
⁄2[Mg3Ca(CO3)4] ‡‡ –89.84
* Bulk moduli from Zhang and Reeder (1999).
† Experimental lattice parameters from dal Negro and Ungaretti (1971).
‡ Experimental lattice parameters from Effenberger et al. (1981).
§ Experimental lattice parameters from Borodin et al. (1979).
|| Experimental lattice parameters from Graf (1961).
# Experimental lattice parameters from Finger (1975).
**Experimental lattice parameters from Ross and Reeder (1992).
†† Experimental lattice parameters from Dollase and Reeder (1986).
‡‡ Aragonite and huntite are also reported per unit cell equivalent to the rhombohedral carbonates.
§§ Calculated from experimental elastic constants of Dandekar and Ruoff (1968).
## Calculated from experimental elastic constants of Hearmon (1946).

FIGURE 2. (a) Schematic of the defect region for a Ca vacancy in calcite based on the Mott-Littleton calculation method. The hexagonal-
based unit cell of calcite is also indicated. Defect calculations were performed for region 1 radius varying from 6 to 12 Å. (b) Close-up view of
an 8 Å defect region showing the distortion of CO32– ions surrounding the calcium vacancy.

TABLE 4. Point defect energies (eV) in calcite as a function of region 1 Formation energies for Ca2+ and CO2– 3 vacancies are quite
radius large and represent the energy for an isolated and charged va-
Defect 6Å 8Å 10 Å 12 Å cancy defect being removed to an infinite distance from the
Ca vacancy VCa 20.71 20.52 20.43 20.53
CO3 vacancy VCO3 28.73 28.45 28.24 28.17
crystal. Strictly, the energy for the vacancy defect should be
Schottky pair VCa + VCO3 46.776 30.478 46.378 not converged calculated for removing the ion to the surface, though creating
Cd substitution CdCa –1.19 –1.19 –1.20 –1.20 a surface was beyond the scope of the present study. Thus the
Mn substitution MnCa –1.95 –1.99 –2.00 –2.01
Fe substitution FeCa –2.48 –2.53 –2.55 –2.56 calculated defect energies represent a maximum and should be
Zn substitution ZnCa –2.64 –2.71 –2.74 –2.76 compared to experimentally derived defect energies with cau-
Co substitution CoCa –2.67 –2.75 –2.78 –2.79
Ni substitution NiCa –3.20 –3.29 –3.33 –3.35 tion. Also shown in Table 4 is the energy for a neutral Schottky
Mg substitution MgCa –2.58 –2.65 –2.68 –2.70 defect pair of a Ca2+ vacancy and a CO2– 3 vacancy. The defect

Brought to you by | New York University Bobst Library Technical Services

Authenticated
Download Date | 7/7/15 8:46 PM
222 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS

spontaneous formation of these types of point defects results

from their smaller ionic size. Defect energy of the metal sub-
stitution generally increases with the ionic radius of the metal
ion (Fig. 4).

Dolomite disorder
Several different types of disorder associated with dolomite
are easily addressed within the modeling framework and the
new potential model. The energy of a single anti-site defect
was calculated by combining a single substitution of magne-
sium for calcium paired with a substitution of calcium for mag-
nesium. Stacking defect energies (Table 5) were determined
from lattice energies of dolomite with a single magnesium and
calcium site switched in a bulk supercell that is three times the
unit-cell dimension in the c direction (total of 18 basal cation
planes). To determine dolomite ordering energies, the occu-
FIGURE 3. Radial distribution function (RDF) plot of the variation pancies of magnesium and calcium in the fully ordered dolo-
in oxygen-oxygen distance due to the presence of a calcium vacancy mite unit cell were varied from 1 (fully ordered) to 0.5 (fully
in calcite (black line) vs. that in the perfect lattice (gray line). disordered). The lattice energies (Table 6) for the relaxed con-
Intramolecular O-O distances broaden in the defect region, whereas
figurations representing this increasing disorder were calcu-
intermolecular oxygen-oxygen distances broaden and expand slightly.
The dashed lines in the insets of the CO32– groups represent the
lated using a mean field approach in which each metal site of
corresponding O-O distances for the peaks in the RDF. the dolomite experiences a potential that is the mean of all pos-
sible disordered positions. As a consequence, all possible con-
figurations are equally as likely with no preference for any
particularly stable configuration (Gale 1997). This modeling
procedure also represents the carbonate configuration that
would be expected to occur under kinetic control rather than a
process that is driven by thermodynamics.

DISCUSSION
Bulk and defect structures
The models satisfactorily reproduce the lattice parameters and
elastic properties of all the end-member rhombohedral carbon-
ates and aragonite using a uniform set of potentials for the car-
bonate anion. However, less-satisfactory lattice parameters for
dolomite and huntite result from the direct transfer of cation-
oxygen potentials that were optimized for calcite and magnes-
ite. In addition, the models have incorporated the unusually “soft”
bulk modulus reported by Zhang and Reeder (1999) for magne-
site. The experimental modulus is slightly smaller, yet similar to
FIGURE 4. Substitution defect energies of metals in calcite as a
function of ionic radius. Defect energies generally increase with
decreasing metal radius. TABLE 5. Energies for the formation of disordering defects in dolomite
Disorder types Energy (eV)
MgCa + CaMg anti-site defect 1.3
was created by the removal of neighboring Ca2+ and CO32– ions Stacking defect 1-1 0.505
aligned along the c axis of the hexagonal cell and separated by CaMg 5.03
MgCa –3.91
a distance of 4.46 Å. The results show that the formation en-
ergy for the coupled defect is less than the sum of the isolated
defects. The reduced energy is related to the lattice distortion TABLE 6. Occupancy of calcium in calcium site as a function of
lattice energy for dolomite at constant Ca/Mg = 1
that occurs about the defect sites and the electrostatic interac-
Ca per Ca site Lattice energy Ca per Ca site Lattice energy
tion of the charged defect centers in forming an associated de- (eV) (eV)
fect pair. Therefore, the long-range Coulombic contribution to 1.00* –88.32 0.75 –87.77
the Schottky defect is approximately half the vacancy defect 0.95 –88.18 0.65 –87.66
energy and the remainder of the energy is related to the accom- 0.90 –88.05 0.60 –87.63
0.85 –87.94 0.50† –87.60
panying steric distortion. 0.80 –87.85
Defect calculations involving the substitution of other metal *Fully ordered dolomite.
cations for Ca in calcite yield negative binding energies. The †Fully disordered dolomite.

Brought to you by | New York University Bobst Library Technical Services

Authenticated
Download Date | 7/7/15 8:46 PM
 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS 223

the values obtained by Ross (1997) and Fiquet and Reynard Dolomite disorder types
(1999). Further refinement of these carbonate models may help Ordering energetics in dolomite and the process of dolo-
to determine the source of the anomaly, in particular, further ex- mitization are still not fully understood. Completely disordered
ploration of the specific properties of magnesite may allow cal- phases preferentially form at low temperatures but are less stable
culation of the predicted changes in elastic properties with than fully ordered dolomite, which forms rapidly from disor-
increasing substitutions and defects in the magnesite structure. dered dolomite (Hardie 1987). Ordering and dolomitization rely
Understanding of the point defect structure of minerals is on both the driving energies for ordering and on the transport
crucial in the characterization and evaluation of many chemi- rates for cations in dolomite. Several disorder types for dolo-
cal and physical properties. Kronenberg et al. (1984) indicated mite have been calculated in this study, and may be used to
that manganese impurities are the mostly likely defect in cal- evaluate the ordering mechanisms. Burton (1987) has shown
cite, and, indeed, all of the divalent cations are soluble in cal- by cluster variation method analysis of carbonate crystals that
cite up to a few percent. The present calculations indicate that experimental phase diagrams are reproduced by taking into
the most favored impurity defect in calcite is Ni2+. However, account highly anisotropic cation ordering, such that interlayer
all substitutions occur as spontaneous defects, and the likeli- cations highly favor ordering while intralayer interactions fa-
hood of a defect will obviously be dependent on the relative vor clustering. The cluster variational model provides a sim-
energy of competing metal oxides and on the availability of plistic view of metal interactions based on a tetrahedral cluster
the cation dopant. The defect calculations also assume the ex- and two parameterized interaction terms.
istence of an isolated and uncharged defect. As the impurity The results of the Mott-Littleton defect model used in the
content increases, this dilute defect approximation is no longer present study provide a more direct measure of the energy pen-
valid and the carbonate structure begins to markedly distort alty for interlayer disorder in dolomite—this is equivalent to the
giving rise to two-phase regions. Electrical conductivity mea- energy of a single anti-sit defect, 1.3 eV. By contrast, the energy
surements by Rao and Rao (1968) indicate that lattice defects for forming a single stacking fault (a switch of full magnesium
are responsible for charge transport in calcite, and they calcu- and calcium basal planes) is approximately 0.5 eV, while the
lated formation energies between 1 and 2 eV. Although the de- energy for an isolated cation substitution is exothermic. These
fect structure of calcite is likely to be quite complex, the high calculations therefore agree that interlayer disorder is energeti-
formation energies calculated here for calcium vacancies im- cally disfavored compared to a clustering of cation defects within
ply that calcium vacancies alone are not responsible for trans- basal planes. Using varying occupancies for the magnesium and
porting charge, and that extrinsic defects are likely to be calcium site in a dolomite structure, the carbonate model allows
involved in the formation of defects in the calcite lattice. Re- a comparison in lattice energy between fully ordered dolomite
cent experimental studies of cation diffusion in calcite (Fisler and fully disordered (proto-) dolomite. The resulting lattice en-
and Cygan 1999) provide activation energies on the order of ergy (Fig. 5) can be represented by E = E0 + a[1 – exp(–bx)],
2.9 eV (at 550 to 800 °C) which are in the range appropriate where x is the fractional occupancy of Mg on the Ca sites. The
for ionic diffusion processes. However, the energies are sig- energy difference between the two fully ordered and fully disor-
nificantly smaller than the defect energies calculated here and, dered dolomite (34.4 kJ/mol) represents the driving force for
therefore, probably represent an extrinsic process controlled dolomite ordering and compares very well with the enthalpy of
by the migration energy of the cation diffusion from an occu- disordering (35.76 ± 0.44 kJ/mol) for dolomite measured ex-
pied site to a defect site. perimentally by Dooley and Navrotsky (unpublished data).

ACKNOWLEDGMENTS
The authors are grateful for the critical reviews provided by David Price,
Lars Stixrude, and an anonymous referee. This paper benefited tremendously
from their comments and suggestions. This research was supported by the U.S.
Department of Energy, Office of Basic Energy Sciences, Geosciences Research,
under contract DE-AC04-94AL85000 with Sandia National Laboratories. Julian
Gale also acknowledges the Royal Society for a University Research Fellow-
ship and funding.

REFERENCES CITED
Borodin, V.L., Lyntin, V.I., Ilyukhin, V.V., and Belov, N.V. (1979) Isomorphic
octavite-calcite series. Soviet Physic Doklady, 24, 226–227.
Burton, B.P. (1987) Theoretical analysis of cation ordering in binary rhombohedral
carbonate systems. American Mineralogist, 72, 329–336.
Catlow, C.R.A. and Mackrodt, W.C. (1982) Computer Simulation of Solids, 320 p.
Berlin, Springer-Verlag.
Catti, M. and Pavese, A. (1997) Quantum mechanical and classical simulations of
Mg-Ca carbonates. In B. Silvi and P. D’Arco, Eds., Modelling of Minerals and
Silicated Materials, 113–156. Kluwer Academic Publishing.
Catti, M., Pavese, A., and Price, G.D. (1993) Thermodynamic properties of CaCO3
calcite and aragonite: A quasi-harmonic calculation. Physics and Chemistry of
Minerals, 19, 472–479.
Crawford, W.A., and Hoersch, A.L. (1972) Calcite-aragonite equilibruim from 50°
to 150° C. American Mineralogist, 57, 995–998.
FIGURE 5. Lattice energy for a dolomite-calcite structure as a dal Negro, A. and Ungaretti, L. (1971) Refinement of the crystal structure of arago-
function of calcium composition and disordering type. nite. American Mineralogist, 56, 768.
Brought to you by | New York University Bobst Library Technical Services
Authenticated
Download Date | 7/7/15 8:46 PM
224 FISLER ET AL.: SHELL MODEL FOR CARBONATE MINERALS

Dandekar, D.P. and Ruoff, A.L. (1968) Temperature dependence of the elastic con- Hearmon, R.F.S. (1946) The elastic constants of anisotropic minerals. Reviews of
stants of calcite between 160 and 300K. Journal of Applied Physics, 39, 6004– Modern Physics, 18, 409–440.
6009. Kay, G.W.C. and Laby, T.H. (1982) Tables of Physical and Chemical Constants.
Deer, W.A., Howie, R.A., and Zussman, J. (1966) An Introduction to the Rock Form- Longmans, London.
ing Minerals. Longmans, London. Kronenberg, A.K., Yund, R.A., and Gilletti, B.J. (1984) Carbon and oxygen diffu-
Dick, B.G. and Overhauser, A.W. (1958) Theory of the dielectric constants of alkali sion in calcite: effects of Mn content and PH2O. Physics and Chemistry of Min-
halide crystals. Physical Review, 112, 90–103. erals, 11, 101–112.
Dollase, W.A. and Reeder, R.J. (1986) Crystal-structure refinement of huntite, Mott, N.F. and Littleton, M.J. (1938) Conduction in polar crystals. I. Electrolytic
CaMg3(Co3)4, with x-ray-powder data. American Mineralogist, 71, 163–166. conduction in solid salts. Transaction of the Faraday Society, 34, 485–499.
Dove, M.T., Winkler, B., Leslie, M., Harris, M.J., and Salje, E.K.H. (1992) A new Parker, S.C., Titiloye, J.O., and Watson, G.W. (1993) Molecular modeling of car-
interatomic potential model for calcite: Applications to lattice-dynamics stud- bonate minerals: Studies of growth and morphology. Philosophical Transac-
ies, phase-transition, and isotope fractionation. American Mineralogist, 77, 244– tions of the Royal Society of London Series A—Physical Sciences and Engi-
250. neering, 344, 37–48.
Effenberger, H., Mereiter, K., and Zemann, J. (1981) Crystal structure refinements Pavese, A., Catti, M., Price, G.D., and Jackson, R.A. (1992) Interatomic potentials
of magnesite, calcite, rhodochrosite, siderite, smithsonite, and dolomite, with for the CaCO3 polymorphs (calcite and aragonite) fitted to elastic and vibra-
discussion of some aspects of the stereochemistry of calcite-type carbonates. tional data. Physics and Chemistry of Minerals, 19, 80–87.
Zeitschrift für Kristallographie, 156, 233–243. Rao, K.S. and Rao, K.V. (1968) Dielectric dispersion and its temperature variation
Finger, L.W. (1975) Least squares refinement of the rigid body motion parameters in calcite single crystals. Zeitschrift für Physics, 216, 300–306.
of CO3 in calcite and magnesite and correlation with lattice vibrations. Carnegie Ross, N.L. (1997) The equation of state and high-pressure behavior of magnesite.
Institute Washington Yearbook, 74, 572–575. American Mineralogist, 82, 682–688.
Fiquet, G. and Reynard, B. (1999) High-pressure equation of state of magnesite: Ross, N.L. and Reeder, R.J. (1992) High-pressure structural study of dolomite and
New data and a reappraisal. American Mineralogist, 84, 856–860. ankerite. American Mineralogist, 77, 412–421.
Fisler, D.K. and Cygan, R.T. (1999) Diffusion of Ca and Mg in calcite. American Tosi, M.P. (1964) Cohesion of ionic solids in the Born model. In F. Seitz and D.
Mineralogist, 84, 1392–1399. Turnbull, Eds., Solid State Physics: Advances in Research and Applications, 1–
Fisler, D.K., Gale, J.D., and Cygan, R.T. (1998) Models of cation diffusion in car- 120, Academic Press, New York.
bonate minerals. EOS Transactions of the American Geophysical Union, 79, White, W.B. (1974) The carbonate minerals. In V.C. Farmer, Ed., Infrared Spectros-
S369. copy of Minerals, p. 227–284. Mineralogical Society, London.
Gale, J.D. (1996) Empirical potential derivation for ionic materials. Philosophical Zhang, J. and Reeder, R.J. (1999) Comparative compressibilities of calcite-struc-
Magazine, 73, 3–19. ture carbonates: Deviations from empirical relations. American Mineralogist,
———(1997) GULP: A computer-program for the symmetry-adapted simulation of 84, 861–870.
solids. Journal of the Chemical Society-Faraday Transactions, 93, 629–637.
Graf, D.L. (1961) Crystallographic tables for the rhombohedral carbonates. Ameri-
can Mineralogist, 46, 238–242. MANUSCRIPT RECEIVED MAY 18, 1998
Hardie, L.A. (1987) Perspectives: Dolomitization: A critical view of some current MANUSCRIPT ACCEPTED SEPTEMBER 8, 1999
views. Journal of Sedimentary Petrology, 57, 166–183. PAPER HANDLED BY LARS STIXRUDE

Brought to you by | New York University Bobst Library Technical Services

Authenticated
Download Date | 7/7/15 8:46 PM