Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Molecular dynamics modelling of hydrated
mineral interlayers and surfaces: structure and dynamics
R. J. KIRKPATRICK*, A. G. KALINICHEV AND J. WANG
Department of Geology, 1301 W. Green St., University of Illinois, Urbana, Il 61801, USA
ABSTRACT
This paper reviews the results of recent molecular dynamics (MD) modelling studies of the interaction
of water and solute species with mineral surfaces and their behaviour in mineral interlayers. Emphasis
is on results for single and double hydroxide phases. Computational results are presented for water and
anions in the interlayers of the Ca2Al, Mg2Al, and LiAl2 layered double hydroxides and on the surfaces
of the Ca2Al phase. Detailed results for water on the (001) surface of brucite (Mg(OH)2) are presented
and compared to published results for other phases. In all these cases, hydrogen bonding and the
development of a hydrogen-bond network involving the H2O molecules and the solid substrate play
very significant roles. The MD methods are especially effective for investigating the structure and
dynamics of mineral-fluid interfaces and mineral interlayers, because they can be applied to systems
containing hundreds to thousands of atoms and for extended durations of the order of nanoseconds.
KEYWORDS: molecular modelling, mineral interlayers, hydroxide phases, brucite, hydrogen bonding, mineral-
¯uid interfaces.
Introduction
THE interaction of water with mineral surfaces and
its intercalation in structural cavities and inter-
layer spaces are among the most important
geochemical and mineralogical processes.
Despite considerable experimental and computa-
tional effort, many aspects of the structure,
dynamics, and energetics of water-mineral inter-
a c t i on a r e i n comp l e t e l y unde r s t ood .
Computational chemistry using both ab initio
quantum mechanical and semi-empirical poten-
tial-based methods has made signi®cant contribu-
tions to addressing these issues. This paper
reviews the results of our recent computational
MD modelling studies of water and ionic species
in mineral interlayers and on mineral surfaces,
and comparison is made with experimental,
mostly spectroscopic, results. Our main focus is
on studies of hydrophillic single and double
hydroxides, but comparison is made with results
for other hydrophillic and hydrophobic surfaces.
A key point of these results is that evaluation of
molecular-scale dynamical effects are essential to
understanding these interactions, and the structure
of mineral-associated water cannot be understood
independently of dynamical behaviour over a
wide range of frequencies. The structural,
dynamical and energetic effects of hydrogen
bonds (H bonds) and H-bond networks involving
water molecules, surface species (principally OH
groups for the hydroxide phases described here),
and solute species play key roles in controlling
these interactions. The MD methods provide an
effective method to investigate H-bond networks,
because relative to quantum chemical methods the
useable system sizes are larger and the simulation
durations longer, and because well tested
potentials are readily available (e.g. Kalinichev,
2001; Guillot, 2002). Spectroscopic methods,
including infrared (IR), nuclear magnetic reso-
nance (NMR), X-ray absorption (XAS), neutron
and sum-frequency generation (a non-linear
optical method that probes interfacial regions;
Miranda and Shen, 1999) and neutron and X-ray
scattering methods have provided important
insights into local structure and dynamics of
* E-mail: [email protected]
DOI: 10.1180/0026461056930251
Mineralogical Magazine, June 2005, Vol. 69(3), pp. 287±306
# 2005 The Mineralogical Society
mineral-solution interfaces and mineral inter-
layers. These chemical environments are,
however, statically and dynamically disordered,
and many details are dif®cult to probe experi-
mentally. Quantum chemical approaches,
including quantum MD, are limited to relatively
small systems and relatively short simulation
times, despite the rapid growth of computational
power (e.g. Odelius et al., 1997; Marx, 2004).
Several papers in the recent `Reviews in
Mineralogy and Geochemistry, volume 42
(Cygan and Kubicki, editors, 2001)' discuss
many important aspects of molecular modelling
theory and methods applied to aqueous solutions
and mineral-solution interfaces, and Kirkpatrick
et al. (2005) discussed the details of calculating
vibrational dynamics of surface and interlayer
species.
Our approach has been to use classical MD
methods and the CLAYFF force ®eld (Cygan et
al., 2004), which is speci®cally optimized for
low-temperature hydrous minerals and that does
not require a priori de®nition of most chemical
bonds. This non-bonded (pseudo-ionic) approach
allows the study of large, complex and disordered
systems containing thousands of atoms in solid
and ¯uid phases and at solid-¯uid interfaces. It is
intrinsically less accurate than ab initio and
quantum MD methods, but it is able to capture
the complex and cooperative interactions that are
critical in these situations. The absence of de®ned
chemical bonds for most interatomic interactions
allows effective and relatively simple treatment of
solids, ¯uids and interfaces and proper accounting
of energy and momentum transfer between the
¯uid phase and the solid. It also keeps the number
of interaction parameters small enough to allow
modelling of large and highly disordered systems.
The only de®ned bonds in CLAYFF are O-H in
H2O, OH-groups in the solid, and the bonds in
aqueous oxyanions (e.g. SO4
2ÿ). The ¯exible SPC
(simple point charge) water model (Berendsen et
al., 1981; Teleman et al., 1987) is used to describe
the H2O and OH behaviour. This model has been
well tested in many situations (e.g. Jorgensen et
al., 1983; Kalinichev, 2001; Guillot, 2002; Head-
Gordon and Hura, 2002). Cygan et al. (2004),
Kalinichev and Kirkpatrick (2002), and
Kirkpatrick et al. (2005) provide examples that
demonstrate the effectiveness of this overall
approach in modelling complex mineralogical
systems.
There are many other effective approaches to
molecular modelling of minerals and mineral-
¯uid systems, and each has its own particular
advantages. For example, treating the atoms as
®xed on a rigid lattice while still allowing for the
degrees of freedom associated with swelling and
lateral displacement of the lattice as a whole can
save substantial amounts of computer time,
because the degrees of freedom associated with
the motion of the atoms in the solid are excluded
from the calculations (Delville, 1995; Boek et al.,
1995; Chang et al., 1995, 1998; Bridgeman et al.,
1996; Karaborni et al., 1996; Desiqueira et al.,
1997; Greathouse and Sposito, 1998; Smith, 1998;
Greathouse et al., 2000; Sutton and Sposito,
2001). Such models obey all fundamental
conservation laws, but due to the immobility of
the lattice atoms, the exchange of energy and
momentum among the interacting atoms of the
substrate and the molecules of the pore ¯uid
(inelastic interactions) is not possible. For
hydrous phases, the characteristic time scales for
vibrational (~3600 cmÿ1
) and librational
(~450 cmÿ1) motions of surface OH groups are
comparable to those of similar motions of H2O
molecules and hydrated ions in the aqueous phase.
Thus, accurate representation of the dynamics of
such processes as hydrogen bonding, adsorption,
surface hydration and complexation may be
limited if the atoms of the substrate layer are
considered completely immobile. Surface diffu-
sion rates of ions and water molecules may also
be overestimated, and the computed structure of
the aqueous layers at the interface may be
distorted.
In a different approach using force ®elds with
bonded interaction terms, Teppen et al. (1997)
and Bougeard et al. (2000) modelled ionic
complexes in clays with all atoms in the system
movable. The force-®eld parameters are based on
charge assignment from quantum chemical
calculations. In this approach, bonds must be
identi®ed and evaluated for all possible metal-
oxygen coordinations. Thus, this approach is quite
accurate but limited to relatively small-scale
simulations of relatively well known mineral
structures, because of the large number of force-
®eld parameters needed to describe the bonded
states. The numerous parameters of a bonded
force ®eld are not easily transferred from
relatively simple and well known materials to
systems with complex and ill-de®ned structures,
and application of such a force ®eld can lead to
signi®cant over-parameterization due to lack of
experimental data to constrain all the necessary
terms.
288
R. J. KIRKPATRICK ET AL.
In the CLAYFF force ®eld, the partial atomic
charges are derived from periodic DFT (Density
Functional Theory) quantum chemical calcula-
tions for simple oxide, hydroxide, and oxyhydr-
oxide model compounds with well known
structures, and the empirical parameters
describing the Lennard-Jones attractive and
repulsive terms are optimized based on known
mineral structures. Thus, in contrast to the work
of Teppen et al. (1997) and Bougeard et al.
(2000), this approach incorporates a set of
experimental crystal-structure re®nements of
model phases to parameterize the empirical
force ®eld, rather than rely on quantum-mechan-
ical calculations alone. Oxygen and hydroxyl
charges vary depending on their occurrence in
water molecules or hydroxyl groups and on the
nearest-neighbour cations. For instance, oxygens
in Si-O-Si, Si-O-Al[4], Si-O-2Al[6] linkages have
different partial charges. All metal-oxygen inter-
actions are based on a simple Lennard-Jones
(12-6) potential combined with electrostatics.
Only harmonic terms are included to describe
the bond-stretch and bond-angle bending terms
associated with water molecules, hydroxyls and
polyatomic anions.
Because the OÿH bonds of water molecules
and OH groups are de®ned in the CLAYFF
approach, models using this force ®eld cannot
account for reactions involving ligand exchange
at mineral surfaces, and we thus limit our
applications to problems where such reactions
are not signi®cant. Development of generally
applicable reactive force ®elds capable of
addressing these situations is a signi®cant need
in geochemistry (e.g. Rustad, 2001; Rustad et al.,
2003).
Methods
MD calculations are performed by building the
desired structure in the computer, assigning the
individual atoms or molecules initial positions
and velocities, and then allowing the system to
evolve according to the laws of classical
Newtonian mechanics and the imposed force
®eld. The techniques and algorithms are well
developed, and the details of the methods are well
established (e.g. Allen and Tildsley, 1987;
Heinzinger, 1990). The modelled structure can
consist of up to many thousands of atoms,
depending on the needs of the problem and
available computer resources. These structures
can be built atom-by-atom, but for crystalline
phases they are more typically based on the
positional parameters of known structures. These
structures can be modi®ed as needed to account
for positional disorder over a crystallographic site,
for instance. Three-dimensional periodic
boundary conditions are applied and the Ewald
summation is used to account for long-range
Coulombic interactions. Mineral-¯uid interfaces
are generated by cleaving the mineral model
structures, often in the middle of an interlayer,
and ®lling all or part of the remainder of the
simulation box with water with or without
dissolved solute atoms (Fig. 1). In most cases,
the number of H2O molecules in this layer is
chosen to give a ¯uid density of ~1 g/cm3. In our
simulations, the thickness of the water layer is
typically >30 AÊ to minimize interaction of one
surface with another in the periodic layered model
structure. Typically, the time step is 0.001 ps, and
the `dynamic trajectory' of the simulated system
in its `phase space' (an ideal multidimentional
space in which the 6N coordinate dimensions
represent the positions and velocities of all N
atoms constituting the system) is recorded for
analysis every 0.004 ps. As in all computational
approaches to molecular-scale problems, care
must be taken to adequately sample the phase
space of the system to adequately ensure that it is
not trapped in a local energy minimum. For
crystalline phases this is not usually a dif®cult
problem, because the starting con®guration is
normally a known structure. For aqueous ¯uids
and solid-¯uid interfaces without solute, the
reorientational and diffusional correlation times
of water molecules are relatively short, and the
system properties typically converge to their
thermodynamic equilibrium values over a few
10s of ps. For systems containing a solid-¯uid
interface and dissolved solute, we position the
ions in the aqueous phase at distances not less
than 8ÿ10 AÊ (~3 molecular diameters of H2O)
from the solid surface and carefully monitor their
dynamic evolution and any adsorption onto the
surface. A typical simulation normally consists of
a pre-equilibration stage in which the atoms move
under only an energy minimization algorithm, a
further pre-equilibration period of MD simulation
lasting 50ÿ500 ps during which the system
reaches its equilibrium thermodynamic state, and
a ®nal equilibrium MD period of typically 100 ps
to 1 ns during which the trajectories of all atoms
are recorded for further statistical analysis. For
solid-¯uid systems containing solute species,
those atoms that become associated with the
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
289
interface typically move to it during the pre-
equilibration stage but often undergo exchange
with the solution during the MD runs. This allows
evaluation of surface-site lifetimes. Quantitative
results for structural parameters such as radial
distribution functions (RDFs), interatomic
distances and angles, and H-bond con®gurations;
dynamic parameters such as diffusion coef®-
cients, adsorption-site lifetimes, and power
spectra of atomic motion; and energetic para-
meters such as bulk system energy and energies of
adsorption are obtained only from analysis of the
equilibrium stage of the MD trajectories. The
power spectra (total dynamical density of states)
of the entire system, individual species and even
the motion of individual species in particular
directions are calculated by Fourier transforma-
tion of the velocity autocorrelation function
(Wang et al., 2003; Kirkpatrick et al., 2005).
The criteria for the existence of an H bond used
here are those often used for bulk liquid water:
intermolecular O_H distances <2.45 AÊ and
angles b, between O_O and O-H <30ë (Luzar,
2000). Surface OH groups are treated in the same
way as O-H of water molecules for the purpose of
HB calculations. The threshold of RO_H
4 2.45 AÊ is used because it corresponds to the
®rst minimum in the O-H radial distribution
function for SPC water at ambient conditions,
and b 4 30ë includes 90% of the angular
distribution of H bonds in water under the same
conditions (Teixeira et al., 1990; Luzar, 2000).
Results and discussion
Interlayer and surface structure and dynamics of
hydroxide phases
Many mineral surfaces are hydroxylated under
low-temperature geochemical conditions, and
much of our work is, thus, focused on hydroxide
phases. The single hydroxides brucite (Mg(OH)2),
gibbsite (Al(OH)3) and portlandite (Ca(OH)2)
have structures containing charge-neutral hydro-
xide sheets with the metals in octahedral
coordination and provide examples of trioctahe-
dral (brucite and portlandite) and dioctahedral
(gibbsite) structures. Layered double hydroxides
(LDHs) are a diverse group of phases with
positive structural charges that interact quite
differently with H2O and solute species than do
single hydroxides. Their layered structures can be
thought of as based on those of brucite,
portlandite, or gibbsite and consist of metal
hydroxide octahedral sheets and interlayer
galleries containing anions and associated water
molecules (Fig. 2). The hydroxide sheets develop
permanent positive structural charge due to
heterovalent substitution of, e.g. Al3+
for Mg2+
or Ca2+
, or Li+for vacancies, that is charge-
FIG. 1. A typical MD modelling cell used in computa-
tions of surface interactions. The top and bottom crystal
structures are hydrocalumite and the central region is
water containing Clÿ
, SO4
2ÿ, and Na
+. The water layer is
~30 AÊ thick in this model.
290
R. J. KIRKPATRICK ET AL.
balanced by the interlayer anions. Thus, these
phases are sometimes known as anionic clays.
The LDHs have a wide range of applications in
catalysis, environmental remediation and medi-
cine (e.g. Miyata, 1983; Cavani et al., 1991;
Kagunya et al., 1996, 1998; Newman and Jones,
1998; Choy et al., 1999; Basile et al., 2001) and
are being increasingly recognized as important
phases in many low-temperature natural and
anthropogenic geochemical environments (e.g.
Bish, 1980; Trolard et al., 1997; Ford and
Sparks, 1998; Ford et al., 1999; Thompson et
al., 1999; Gade et al., 1999, 2000; Genin et al.,
2001). Many tens of LDH phases are known (Hou
et al., 2003; Braterman et al., 2004), and they
offer a wide range of opportunities to investigate
the effects of hydroxide layer composition and
anion charge, size and conformation on the
development of interlayer H-bond networks and
dynamics.
The LDH phase with the best known structure
is the mineral hydrocalumite, [Ca2Al(OH)6]
Cl´2H2O, also known as Friedel's salt, and MD
modelling of its interlayer and surface structure
and the diffusional, translational and librational
dynamics of its interlayer and surface species
illustrates well the capabilities of MD methods.
The MD results demonstrate that our techniques
described above reproduce the experimentally
determined hydroxide layer and interlayer struc-
tures well, that librational dynamics of the
interlayer water molecules play a key role in its
experimentally observed structural phase transi-
tion, and that the low-frequency translational
dynamics of interlayer and surface species are
similar in many ways to that in bulk aqueous
solutions (Kalinichev et al., 2000). This phase is
unique among well known LDHs, because it has
both an ordered Ca,Al distribution in the
hydroxide layer and a well ordered Clÿ
and
water structure in the interlayer space (Fig. 1;
Terzis et al., 1987). The interlayer order is due to
coordination of the water molecules to Ca in the
hydroxide layer, which results in an unusual
7-coordinate Ca environment (Fig. 3d). The
interlayer Clÿ
ions form almost regular triangles
of a 2-D hexagonal net, and the water molecules
are located at the centre of each triangle. Each
Clÿ
has six nearest neighbour (NN) H2O, and
each water molecule has three NN Clÿ
, but is
frustrated because it can only donate two H bonds
at any instant. Each Clÿ
is coordinated through H
bonds to six OH-groups: three one each side of
the interlayer (Fig. 3c).35Cl NMR spectroscopy
(Kirkpatrick et al., 1999; Andersen et al., 2002)
shows the presence of a phase transition at which
the symmetry at Clÿ
changes from triaxial to
uniaxial or nearly so due to reorientation of the
electrical ®eld gradient at35Cl at frequencies
>~105Hz.
MD modelling shows that in the low-temp-
erature phase the water molecules are statically
disordered among three possible positions
donating H bonds to Clÿ
, whereas in the high-
temperature phase they librate (hop) among these
positions at frequencies that result in a strong IR
band near 500 cmÿ1. The observed phase
transition is due to the onset of this libration,
which causes the time-averaged symmetry of the
Clÿ
to change from triaxial with four static H
bonds from water molecules to uniaxial with six
2/3-occupied H bonds from water molecules.
FIG. 2. View of the Clÿ
hydrotalcite, (Mg2,Al)
(OH)6Cl´2H2O, parallel to the layering showing the
single Mg2,Al octahedral sheets and interlayer Clÿ
and
H2O. As in hydrocalumite, the interlayer water in this
phase occupies two sublayers due to H-bond donation
from the OH-groups to interlayer Clÿ
and H2O and from
interlayer H2O to Clÿ
.
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
291
Each Clÿ
anion is also coordinated by six OH
groups (three from each adjacent hydroxide layer)
under all conditions. In our simulations, these H
bonds immobilize the Clÿ
on the ns timescale,
even at temperatures as high as 300ëC.
Experimental35Cl NMR results suggest that
similar dynamics occurs in the less ordered
interlayer of the Mg,Al LDH hydrotalcite with
interlayer Clÿ
(Fig. 2; Kirkpatrick et al., 1999).
The structural environments of water and Clÿ
on the hydrocalumite basal surface are similar in
some ways to those in the interlayer but are more
disordered both statically and dynamically
(Kalinichev et al., 2000; Kalinichev and
Kirkpatrick, 2002). MD simulations show that
Clÿ
is associated with the surface principally as
inner sphere complexes due to their large
Coulombic interaction with the hydroxide
sheets. In contrast to the highly ordered interlayer,
however, the Clÿ
and H2O are disordered over
sites comparable to the `Clÿ
'and `H2O' sites in
the interlayer (Fig. 3). The `H2O' site is directly
coordinated to Ca, whereas the `Clÿ
' site is
coordinated to OH groups by H bonds.
Dynamically, the mean residence time of a Clÿ
on a surface site is ~50 ps, resulting in computed
diffusion coef®cients of ~1.6610ÿ6
cm2sÿ1
for
inner sphere Clÿ
and 7.5610ÿ6
cm2sÿ1
for outer
sphere Clÿ
. These values are intermediate
between interlayer diffusion coef®cients that are
too small to compute (<<10ÿ7
cm2sÿ1) and a bulk
s o l u t i o n d i f f u s i o n c o e f ® c i e n t o f
2.3610ÿ5
cm2sÿ1. The interlayer Cl
ÿ
and H2O
do not hop among sites during 100 ps MD runs.
FIG. 3. Computed positions of water molecules and Clÿ
on mineral surfaces. (a) Water located above the vacant
tetrahedral sites of the octahedral sheets of Mg(OH)2 or Ca(OH)2 and receiving three H bonds from the surface
hydroxyls. (b) Water located near the OH sites of the octahedral sheets of Mg(OH)2 or Ca(OH)2 and receiving one H
bond from it. These waters also typically have four NN H2O. (c) Clÿ
occupying the `Cl' position on the surface of
hydrocalumite. This position receives three H bonds from the surface and is located above a Ca that is displaced
downwards from the centre of the octahedral sheet in this view. (d) Clÿ
occupying the `H2O' position on the surface
of hydrocalumite. This position coordinates a Ca that is displaced upwards from the centre of the octahedral sheet in
this view.
292
R. J. KIRKPATRICK ET AL.
The vibrational dynamics of surface Clÿ
and H2O
and the librational dynamics of the surface H2O
are also quite different from those species in the
interlayer or bulk water. In the interlayer,
cooperative motion of the Clÿ
and H2O leads to
low-frequency translational bands centred near
50 cmÿ1
for motion dominantly parallel to the
hydroxide sheets (comparable to H-bond bending
in bulk water) and near 150 cmÿ1
for motion
dominantly perpendicular to the hydroxide sheets
(comparable to H-bond stretching in bulk water).
On the surface, the relative intensity of the band
due to motion perpendicular to the hydroxide
sheets is greatly reduced due to the absence of the
second wall of the interlayer and is more similar
to the power spectra in bulk solution (Fig. 4). The
computed band for surface H2O libration is
signi®cantly broader than that of interlayer H2O
due to the structural disorder and is quite similar
to that of bulk water. The MD calculations can
also be used to calculate the full vibrational power
spectrum of a phase, and recent work has shown
that the computed frequencies for hydrotalcite and
other LDHs in the far infrared region correlate
well with the observed band positions and that
MD simulations can effectively assist with band
assignment in this frequency range (Wang et al.,
2003; Kirkpatrick et al., 2005).
The computed results for the structure and
dynamics of H2O and Clÿ
associated with the
charge neutral (001) surface of portlandite
(Ca(OH)2) are quite different from the positively
charged hydrocalumi te (001) sur face .
Structurally, the water molecules are associated
with the portlandite surface both via donation of
H bonds to the O of surface OH-groups and
acceptance of H bonds from the surface OH
groups (Fig. 3a,b; Kalinichev and Kirkpatrick,
2002). Surface OH groups also bend towards the
Clÿ
to form inner sphere sorption sites. Such sites
are consistent with35Cl NMR T1 relaxation rate
data, which show signi®cant Clÿ
association with
the portlandite surface (Yu and Kirkpatrick,
2001). In the simulations, however, only a
fraction of the dissolved Clÿ
is associated with
the portlandite surface. The computed mean Clÿ
FIG. 4. Computed low-frequency power spectra for the motion of Clÿ
on the surface of Ca(OH)2 (inner-sphere and
outer-sphere sites), on the surface and in the interlayer of hydrocalumite (Ca/Al LDH), and in bulk aqueous solution.
The bands near 50 cmÿ1
involve mostly H-bond bending and are due to motion parallel to the layers in
hydrocalumite interlayers. The bands near 150 cmÿ1
involve mostly H-bond stretching and are due to motion
perpendicular to the layers in hydrocalumite interlayers.
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
293
surface site residence time is only ~20 ps, and the
computed diffusion coef®cients are signi®cantly
greater than for hydrocalumite, 3.7610ÿ6
cm2sÿ1
for inner sphere Clÿ
and 1.5610ÿ5
cm2sÿ1
for
outer sphere Clÿ
. The computed low-frequency
translational power spectrum of surface-asso-
ciated Clÿ
on portlandite is dominated by the H-
bond bending band near 50 cmÿ1
and is quite
similar to that in bulk solution (Fig. 4).
The LiAl2(OH)6Aÿ
´nH2O LDH phases can be
thought of as being derived from gibbsite by Li+
for vacancy substitution and in many ways have
different interlayer structures than the Mg2Al or
Ca2Al LDHs because of the small Li+
ionic
radius. As for hydrocalumite, MD methods have
played an important role in understanding their
interlayer structure and dynamics (Hou et al.,
2002). The LiAl2 LDHs have a highly ordered
octahedral Li,Al distribution, and in the anhy-
drous Clÿ
phase the interlayer Clÿ
lies directly
above and below the octahedral Li and is also
highly ordered as shown by XRD and35Cl NMR
data (Besserguenev et al., 1997; Hou et al., 2002).
The MD simulations reproduce this structure very
well. For the hydrated phase, the maximum water
content, n = 1, and the room temperature35Cl
NMR spectra of substantially hydrated samples
show a broad component similar to that of the
anhydrous sample and a much narrower compo-
nent representing disordered and dynamically
averaged Clÿ
(Hou et al., 2002). At temperatures
above 70oC, all of the interlayer Cl
ÿ
undergoes
dynamical averaging. Diffraction data show that
the interlayer structure of the hydrated phase is
disordered but the details are poorly known. The
MD results provide a quite detailed picture of this
statically and dynamically disordered interlayer
structure and demonstrate the high degree of
similarity between the most common NN Clÿ
environments in the interlayer and in bulk
aqueous solution (Fig. 5). The Clÿ
and H2O are
located at the centre of the interlayer along the c
direction, and the H-O-H plane of the water is
parallel to the hydroxide layers. The H2O are
located near the OH groups and receive two H
bonds from them, one from each side of the
interlayer. The Clÿ
receive H bonds from OH-
groups or H2O, and there is a well developed H-
bond network in the interlayer. A few of the Clÿ
are located near the cross-layer Li-Li vectors at
sites comparable to those in the anhydrous phase
and also in trigonal prisms of OH groups above
and below the vacant tetrahedral sites on the
octahedral sheet. In both these con®gurations,
they receive H bonds principally from OH groups.
Most Clÿ
, however, are located on distorted
octahedral sites above and below the vector
connecting nearest neighbour OH groups. These
FIG. 5. Computed interlayer structure of the LDH LiAl2(OH)6Cl´H2O. The interlayer H2O/Clÿ
ratio is limited to 1/1,
because the horizontally oriented water molecules are in stable, tetrahedral, ice-like NN coordination with two
accepted and two donated H bonds, and the Clÿ
are in 6-fold coordination receiving 6 H bonds.
294
R. J. KIRKPATRICK ET AL.
Clÿ
receive two H bonds from OH-groups of one
side of the interlayer, two from the other side, and
two from H2O (Fig. 5b). This arrangement allows
each H2O to be in a highly stable tetrahedral
H-bond environment very similar to that in bulk
solution and in ice Ih. The mean ClÿH distance is
2.16 AÊ for H of H2O and 2.12 AÊ for H of OH,
compared to ~2.22 AÊ in bulk solution. The
stability of this structure explains the maximum
1:1 Clÿ
to H2O ratio for the LiAl2 LDH, because
there are no additional sites where H2O can be in
stable, H-bonded tetrahedral coordination. This
contrasts with the Ca2Al and Mg2Al LDHs, in
which the water molecules can form two
sublayers, and the Cl/H2O ratio can approach 2
(Fig. 2). Dynamically, the interlayer H2O under-
goes restricted librational motion (hindered
rotational hopping) among H-bonded con®gura-
tions at approximately 1011
Hz (near 500 cmÿ1),
as observed in other LDH phases. The site
hopping frequencies are ~107Hz for H2O and
36108Hz for Cl
ÿ
, consistent with the observed
narrowing of the35Cl NMR resonances. The MD
computed power spectrum of this phase is in
particularly good agreement with the observed
far-IR spectrum, which shows bands for motion of
Li, Al and OHÿ
as well as the interlayer species
(Kirkpatrick et al., 2004).
Water structure at mineral surfaces
It has long been known that the structure and
physical properties of water near mineral surfaces
can be substantially different from those of bulk
water, that surfaces can perturb the ¯uid structure
and properties up to several molecular diameters
from the surface, and that these differences can be
key to understanding mineral surface chemistry
(Packer, 1977; Israelachvili and Pashley, 1983;
Hochella and White, 1990; Israelachvili and
Wennerstron, 1996; Brown et al., 1999;
Criscenti and Sverjensky, 1999; Nandi et al.,
2000; Raviv et al., 2001; Zhu and Granick, 2001;
Brown, 2001 Michot et al., 2002). Despite
decades of study, the structure, dynamics and
physical properties of this near-surface water
remains incompletely understood (e.g. McCarthy
et al., 1996; Bridgeman and Skipper, 1997; Spohr
and Hartnig, 1999; StoÈckelmann and Hentschke,
1999; Kalinichev et al., 2000; Greathouse et al.,
2000; Dore, 2000; Fenter et al., 2000a,b; Cheng et
al., 2001; Teschke et al., 2000; Bellissent-Funel,
2001, 2002; Fouzri et al., 2002; Park and Sposito,
2002; Sakuma et al., 2003). Computational
methods are playing an important role in
advancing understanding of near-surface water
(Lee and Rossky, 1994; McCarthy et al., 1996;
Bridgeman and Skipper, 1997; Hartnig et al.,
1998; Spohr and Hartnig, 1999; StoÈckelmann and
Hentschke, 1999; Kalinichev et al., 2000;
Greathouse et al., 2000; Gordillo and MartõÂ,
2000; Cygan, 2001; Gallo et al., 2002; Park and
Sposito, 2002; Michot et al., 2002; Kalinichev
and Kirkpatrick, 2002; Rustad et al., 2003;
Sakuma et al., 2003; Wang et al., 2004), and
MD methods can be especially useful, because
they can be applied to systems large enough to
capture the complex correlations of molecular
motions at the surface over much longer time- and
length-scales than is currently possible with ab
initio calculations, as described above. Here we
illustrate the capabilities of MD methods in this
regard with recent results concerning water
structure at the brucite (Mg(OH)2) (001) surface
(Wang et al., 2004) and compare these results
with published computational results for water at
other surfaces.
Solid surfaces can perturb the water structure
and dynamics by affecting its molecular packing,
orientation, rotation and translation. These effects
arise due to the `hard wall' effect, the presence of
electrostatic ®elds, and local surface-speci®c H-
bonding donor and acceptor sites (e.g. Odelius et
al., 1997; Cheng et al. 2001; Wang et al., 2004).
The `hard wall' or `excluded volume' effect
creates near-surface layering due to the spatial
geometric constraint that no part of an atom or
molecule at the surface can penetrate it. This
effect occurs for all con®ned ¯uids (e.g. Abraham,
1978). Thus, the structure of water in an
interfacial region re¯ects a delicate balance of
the ordering due to `hard wall' or `excluded
volume' effects of packing H2O molecules at a
solid surface, surface-speci®c H bonding and
orientational ordering of the water molecules,
and disordering due to thermal motion. This
structure is typically different and more disor-
dered than the tetrahedral, H-bonded structure of
ice Ih, the evidence of which is still prominently
present in the more disordered short-range
tetrahedral H-bonding molecular arrangements
in bulk liquid water (e.g. Eisenberg and
Kauzmann, 1969; Soper, 2000; Errington and
Debenedetti, 2001; Head-Gordon and Hura,
2002).
For hydrophyllic phases such as hydroxides, H
bonding between the surface and water molecules
and among water molecules both play central
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
295
roles, whereas for hydrophobic phases such as
carbon nano-tubes and talc, H bonding among the
water molecules is important, but H bonding
between the surface and water is much less
signi®cant. Many studies of surface-water struc-
ture rely on computed pro®les of atomic density
variations with distance from the surface,
comparable to radial distribution functions, some-
times supplemented by pro®les of molecular
water orientation. We have found that computed
statistics for NN coordination, H bonding, and
order parameters related to NN structure (Chau
and Hardwick, 1988; Errington and Debenedetti,
2001) provide important additional information
that leads to a greatly improved, atomistically
detailed understanding.
The MD results for water at the brucite and
portlandite surfaces show that these hydrophyllic
substrates signi®cantly in¯uence the near-surface
water structure, with both H-bond donation to the
surface oxygen atoms and H-bond acceptance
from the surface hydrogen atoms in the ®rst
surface layer of H2O molecules playing key roles
(Kalinichev and Kirkpatrick, 2002; Wang et al.,
2004). The oxygen and hydrogen atomic densities
deviate from those of bulk water to distances as
large as 10 AÊ (Fig. 6). The distances between
maxima in the O-density pro®les are not equally
spaced, as would be expected from hard wall
effects alone, clearly demonstrating that surface
structure, charge distribution and H bonding must
play important roles in controlling the near-
surface water structure. The H2O dipole orienta-
tions show structuring to as far as 15 AÊ (~5
molecular water layers) from the surface (Fig. 7).
The average number of H bonds per H2O
molecule changes from 3.8 in the near-surface
layer to 3.5 (approximately the value for bulk
FIG. 6. Computed atomic density pro®les for water con®ned between brucite layers. Curves are displaced vertically
by 0.03 AÊÿ3
(oxygen atomic density) and 0.07 AÊÿ3
(hydrogen atomic density) to avoid overlap. The position of the
surface (0.0 in these plots) is computed as the average position of the brucite surface oxygen atoms. The system size
labels are the increases in the brucite c-axis dimension used to generate the systems, and the actual water layer
thicknesses vary somewhat from these values.
296
R. J. KIRKPATRICK ET AL.
SPC water) at ~10 AÊ from the surface, and there
are signi®cant oscillations in this value (Fig. 8).
The MD simulations show that nm-scale
con®nement in slit-like pores at least 15 AÊ thick
and less leads to signi®cant overlap of the
structural effects of the two surfaces, as described
in detail by Wang et al. (2004). For thin pores, the
structure of the entire water volume is substan-
tially perturbed compared with bulk water, and
the effects of the surface depend signi®cantly on
pore thickness.
For the uncon®ned brucite surface, the varia-
tion in atomic density re¯ects the presence of
three important, well de®ned layers. These are a
high atomic density, highly structured, near-
surface layer centred near 2.5 AÊ that contains
molecules that are directly coordinated to the
surface; a low atomic density, transitional layer
centred near 4.0 AÊ from the surface; and a region
extending from ~5 to 15 AÊ from the surface in
which the structure becomes progressively more
similar to that of bulk water. Figure 9 illustrates
schematically the most common orientations of
water molecules in these layers. The layer nearest
the surface contains two principal types of water
molecules, both of which are directly coordinated
FIG. 7. Computed angular distributions of the orientations of water molecules con®ned in 30 AÊ thick pores in brucite.
This thickness is large enough that the two surfaces do not perturb each other. jD (left) and jHH (right), are the
angles between the water dipole (H-end = positive) and brucite [001], and between the water H-H vector (no sign
convention) and brucite [001] respectively. Each curve was normalized, rescaled, and displaced vertically by 1.0 to
®t the ®gure. The values listed in scales a to j are the distances from the surface.
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
297
to surface OH groups. These two types have
different orientations, NN coordinations and
H-bond con®gurations. They are, on average,
located at slightly different distances from the
surface but are intimately mixed with each other
in the plane parallel to the surface. Large domains
of one structural type cannot form, because
H-bond formation between neighbouring water
molecules prevents each type of environment
from extending more than three molecules in the
plane parallel to the surface. Type 1 molecules are
on average slightly closer to the surface (mean
distance ~2.3 AÊ ) and are predominantly oriented
with the positive (hydrogen) end of their
molecular dipole towards the surface, making an
average angle of ~130ë with the surface normal
(Fig. 7). These molecules typically have NN
coordinations of six, three surface OH-groups
and three H2O. On average, they accept ~0.5 H
bonds from surface OH groups, donate 1.0 H bond
FIG. 8. (a) Variation of the average total number of H bonds per water molecule and the contributions of various H
bond types to this number with distance from the brucite surface for the system with 30 AÊ of water. (b) Variation of
the average fraction of water molecules with different numbers of H bonds with distance from the brucite surface for
the system with 30 AÊ of water.
298
R. J. KIRKPATRICK ET AL.
to surface OH-groups, accept ~1.3 H bonds from
other water molecules, and donate ~0.8 H bonds
to other water molecules. Type 2 molecules are on
average somewhat further from the surface
(~2. 6 AÊ ) and are predominantly oriented with
the positive ends of their dipoles oriented away
from the surface, making angles of ~30ÿ80ë with
the surface normal. These molecules typically
have NN coordinations of ®ve, one surface
OH-group and four water molecules. On
average, they accept ~0.5 H bonds from surface
OH-groups, donate none to surface OH groups,
accept ~1.4 H bonds to other water molecules and
donate ~1.7 H bonds to other water molecules.
The (type 1)/(type 2) abundance ratio is ~5/4.
These two types of H2O occur on very different
surface sites. Type 1 molecules are preferentially
located above the vacant tetrahedral sites of the
trioctahedral sheet (triangles formed by surface
OH-groups; Fig. 3a,b). They form a reasonably
well ordered and dynamically averaged 2-dimen-
sional hexagonal net that re¯ects the underlying
brucite structure (Fig. 10). There appear to be
four local potential energy minima on which they
occur. One of these is at the centre of the OH
triangle, and three are near the middle of the line
connecting two nearest-neighbour OH sites. The
water molecules spend, on average, about half of
the time at the OH-triangle centre, where they
accept 1 HB, and 1/6 of the time at each of the
other sites, where they donate 1 and accept 1 HB.
This site hopping, libration, and formation and
breaking of HBs result in the computed average of
1.5 HBs with surface OH groups. The type 2
water molecules are preferentially located above
the surface OH-groups, and their distribution is
much less ordered than for the type 1 molecules
(Fig. 10). Only half accept one H bond from
surface OH groups at any instant.
Further from the surface, the low-density
region between 3 and 5 AÊ from the surface
provides the essential transition between the
near-surface layer, with a structure largely
controlled by the substrate surface, to a more or
less bulk-like water structure (Fig. 6). This
transition occurs by gradual adjustment of the
second neighbour con®guration in a distorted, but
locally tetrahedral structure. The NN coordination
in this region is ~4.4, similar to that of bulk liquid
water at the same temperature and density.
Molecules in this region have two different
preferred orientations, and as in the ®rst layer,
these types are mixed on a molecular scale across
the surface. The orientational order is, however,
much less than in the ®rst layer. Type 3 molecules
have their positive ends generally oriented
towards the surface, making angles between 90
and 180ë with the surface normal (Fig. 7). Type 4
molecules have their positive ends generally
oriented away from the surface, making angles
between 40 and 140ë with the surface normal. The
(type 3)/(type 4) abundance ratio varies from 2/1
at 3 AÊ from the surface, to 4/1 at 4 AÊ , and back to
2/1 at ~5 AÊ .
The greatest degree of the tetrahedral (ice-like)
ordering occurs at ~4 AÊ from the surface, where
the O-density has its minimum value. Indeed, at
this distance the water molecules are locally more
ordered (more similar to ice Ih) than in bulk
water. The fraction of molecules with exactly 4 H
bonds reaches a maximum at this position and the
average orientational order parameter, q, as
de®ned by Errington and Debenedetti (2001) has
a value greater than bulk water. Such structuring
is in agreement with the notion that the number of
water molecules with four H bonds increases with
decreasing density at liquid-like densities (Geiger
FIG. 9. Schematic diagram illustrating the orientations
and H bonding of water molecules in the different near-
surface layers for brucite. Black balls and sticks are
water molecules or surface OH groups. Dotted balls and
sticks are water molecules with different orientations,
which schematically show the orientational ranges in the
different layers. The dotted lines connecting water
molecules are H bonds.
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
299
and Stanley, 1982; Kalinichev, 2001; Paulo et al.,
2002), and similar structuring is also observed for
water con®ned in pores in hydroxylated silica
glass (Gallo et al., 2002).
Beyond ~6 AÊ from the surface, the water
structure is generally similar to that of bulk water,
with an average of ~3.5 H bonds/molecule
(Jorgensen et al., 1983) and an average NN
FIG. 10. Atomic density maps for oxygen of water in the ®rst atomic density maximum near the brucite surface. The
upper map is for type-1 water molecules, and the lower map for type-2 water molecules. The ®lled squares and
circles are surface Mg and O atoms respectively, and the thin lines are contours of water oxygen probability density.
The most probable positions of the type-1 molecules re¯ect the underlying brucite structure.
300
R. J. KIRKPATRICK ET AL.
coordination of ~4.5. The atomic density pro®les
(Fig. 6) show statistically meaningful variation to
~10 AÊ from the surface, and small but statistically
meaningful variations in the molecular orienta-
tions occur out to ~15 AÊ from the surface (Fig. 7).
These changes of angular distribution are due to
adjustment of the orientations of individual water
molecules to ®t their local environments, which
are perturbed indirectly by the surface through its
effects on H2O molecules next to it.
The water structure in the layer closest to the
brucite surface does not resemble those of bulk
water at ambient conditions, ice Ih, or water at
low temperatures, as has been proposed for water
con®ned in Vycor glass and silica gel based on
neutron diffraction studies (Dore, 2000;
Bellissent-Funel, 2001). This is clearly shown
by the 5- and 6-fold NN coordinations of the near-
surface molecules and by this coordination being
as much as 2.2 larger than the H bond number. In
ice Ih, the NN coordination and H bond number
are both ~4.0 (Eisenberg and Kauzmann, 1969),
and in bulk SPC liquid water at ambient
temperature and pressure the NN coordination is
~4.4 and the H bond number is ~3.5. Cooling of
bulk liquid water causes the average number of H
bonds to increase with the NN coordination
remaining more or less constant (Rapaport,
1983; Paulo et al., 2002). The structure of the
®rst layer of water does share some similarities
with those of high-pressure ice phases and liquid
water at elevated pressure. With increasing
pressure, the average NN coordination for liquid
water increases more rapidly than the average
number of H bonds (Kalinichev et al., 1999;
Paulo et al., 2002), and the structural changes can
best be interpreted in terms of an increasing
number of interstitial (non-H bonded) water
molecules in the NN coordination sphere
(Bagchi et al., 1997; Kalinichev et al., 1999;
Saitta and Datchi, 2003). In the crystalline ice
phases, there are always four H-bonded nearest
neighbours at intermolecular distances of
2.7ÿ2.9 AÊ , but the number and intermolecular
distances of the non-H-bonded molecules are
different for different phases. There are zero non-
H-bonded molecules in ice Ih (stable up to
0.3 GPa), 3.75 at 3.1ÿ3.3 AÊ in ice IV (a
metastable phase at 0.4ÿ0.55 GPa, Engelhardt
and Kamb, 1981), and 4 at 2.74 AÊ in ice VIII
(stable above 2.1 GPa, Kuhs et al., 1984). The
latter distance is shorter than the H-bond distance
(2.88 AÊ ) in that phase, paralleling a similar trend
for liquid water under pressure (Schwegler et al.,
2000). At the brucite surface, the coordination
number of type 1 molecules is 6 and number of H
bonds is ~3.8, qualitatively following the trends
for liquid water and the crystalline phases with
increasing pressure.
The MD modelling for water at the brucite
surface adds to a growing body of computational
and experimental studies that demonstrate that
different types of surfaces can have substantially
different effects on surface water structure and
that this structure should not be thought of as
simply `ice-like'. For instance, previously
published MD simulations for water at a variety
of oxide and hydroxide surfaces show the
presence of molecules with two different and
well de®ned orientations in the ®rst layer and that
the local structural environments or orientations
are different for different phases. The coexistence
of water molecules with different orientations
mixed and interconnected in the plane parallel to
the surface appears to allow the development of
an interconnected H-bond network involving the
water molecules and surface atoms. The inter-
facial water on the portlandite, Ca(OH)2, (001)
surface is similar to that for brucite due to the
similarity in their structures and involves H-bond
donation and acceptance to/from the solid surface
(Kalinichev and Kirkpatrick, 2002). MD simula-
tions for water at the magnetite (001) surface
using a potential model that allows the surface
protonation state to change during the MD
simulation show that H-bond donation and
acceptance between the surface and H2O are
important in this situation also (Rustad et al.,
2003). On this surface, interfacial water mole-
cules accept H bonds from several surface
functional groups, of which about a half are[6]FeOH2 sites (doubly protonated O-atoms
coordinated to one octahedral Fe). About three
quarters of the H bonds donated by H2O
molecules to surface sites go to[4]FeOH (singly
protonated O-atoms coordinated to tetrahedral
Fe). These results suggest that different surface
functional groups can play different roles in
developing interfacial H-bonding networks. In
contrast, our model brucite (001) surface contains
only one type of surface functional group
([6]Mg3OH) that serves as both an H-bond donor
and acceptor.
Lee and Rossky (1994) have proposed two
idealized H-bond structures for water in the ®rst
hydration layer of a hydroxylated silica surface,
and these are quite different from those for
brucite. One type has the positive end of its
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
301
dipole oriented away from the surface and accepts
one H bond from and donates one H bond to
surface OH groups. The other type has the
positive end of its dipole oriented towards the
surface and accepts one H bond from and donates
two H bonds to surface OH groups. The
differences between water orientations at the
brucite and hydroxylated silica surfaces may be
caused by the differences in the substrate surface
structure. On the silica surface modelled by Lee
and Rossky, the Si-OH groups are 5.0 AÊ apart,
whereas for brucite the shortest MgOHÿMgOH
distance is only 3.1 AÊ , approximately the
diameter of a water molecule. MD simulations
for water con®ned in pores in hydrated Vycor
glass show that the ice Ih-like NN and H-bond
geometry of bulk water is destroyed near these
surfaces (Gallo et al., 2002). MD simulations of
water at the NaCl (100) surface show a lattice-like
2-D distribution parallel to the surface, and as for
brucite this 2-D structure re¯ects the underlying
NaCl crystal structure (StoÈ ckelmann and
Hentschke, 1999).
The presence of both donating and accepting
H-bond con®gurations at hydroxylated surfaces is
not universal. H2O on the surface of hydro-
calumite described above has the positive ends of
its dipole pointing only away for the surface due
to the positive structural layer charge of this
phase, and these waters accept H bonds from the
surface OH-groups but do not donate any to them
(Kalinichev and Kirkpatrick, 2002).
The extensive H bonding between the surface
and near-surface H2O computed for brucite,
portlandite and other hydrophyllic surfaces
contrasts signi®cantly with the water structure
near hydrophobic surfaces. For instance, at the
surfaces of carbon nano-tubes H bonding to the
surface is not signi®cant and the ice Ih-like NN
and H-bond geometry of bulk water is destroyed.
The water molecules nearest to the surface
participate on average in only ~2.5 H bonds,
although this number increases to ~3.5 at
distances >4 AÊ from such surfaces (Gordillo and
MartõÂ, 2000). At the talc (001) surface, water
dipoles have two predominant orientations that
appear to correspond to molecules that either
accept H bonds from OH groups in the talc
octahedral sheet or donate H bonds to the surface-
bridging oxygens (Bridgeman and Skipper, 1997).
Our recent calculations for the talc (001) surface
show only weak H-bond donation to the bridging
oxygens at ambient conditions and essentially
none at elevated temperatures (Wang, 2004).
Acknowledgements
This research was supported by DOE Basic
Energy Sciences Grant DEFGO2-00ER-15028.
Computation was partially supported by the
National Computational Science Alliance (Grant
EAR 990003N) and utilized NCSA SGI/CRAY
Origin 2000 computers and the Cerius2-4.6
software package from Accelrys. J. Wang also
acknowledges a fellowship from the University of
Illinois at Urbana-Champaign. We gratefully
acknowledge detailed discussion of the experi-
mental data with Xiaoqiang Hou and Ping Yu and
fruitful discussion of molecular modelling with
R.T. Cygan and J.D. Kubicki.
References
Abraham, F.F. (1978) The interfacial density pro®le of a
Lennard-Jones ¯uid in contact with a (100) Lennard-
Jones wall and its relationship to idealized ¯uid/wall
systems: A Monte Carlo simulation. Journal of
Chemical Physics, 68, 3713ÿ3716.
Allen, M.P. and Tildesley, D.J. (1987) Computer
Simulation of Liquids. Clarendon Press, Oxford, UK.
Andersen, M.D., Jakobsen, H.J. and Skibsted, J. (2002)
Characterization of the alpha-beta phase transition in
Friedels salt (Ca2Al(OH)6Ca.2H2O by variable
temperature27Al NMR spectroscopy. Journal of
Physical Chemistry A, 106, 6676ÿ6682.
Bagchi, K., Balasubramanian, S. and Klein, M.L. (1997)
The effects of pressure on structure and dynamical
properties of associated liquids: Molecular dynamics
calculations for the extended simple point charge
model of water. Journal of Chemical Physics, 107,
8561ÿ8567.
Basile, F., Campanati, M., Serwicka, E. and Vaccari, A.
(2001) Hydrotalcites. Introduction to the special
issue. Applied Clay Science, 18, 1ÿ2.
Bellissent-Funel, M.-C. (2001) Structure of con®ned
water. Journal of Physics: Condensed Matter, 13,
9165ÿ9177.
Bellissent-Funel, M.-C. (2002) Water near hydrophilic
surfaces. Journal of Molecular Liquids, 96ÿ97,
287ÿ304.
Berendsen, H.J.C., Postma, J.P.M., van Gunsteren, W.F.
and Hermans, J. (1981) Interaction models for water
in relation to protein hydration. Pp. 331ÿ342 in:
Intermolecular Forces (B. Pullman, editor). Riedel,
Dordrecht, The Netherlands.
Besserguenev, A.V. et al. (1997) Synthesis and structure
of the gibbsi te intercalat ion compounds
[L iA l 2 (OH) 6 ]X {X=C l , B r , NO 3 } and
[LiAl2(OH)6]Cl´H2O using synchrotron X-ray and
neutron powder diffraction. Chemistry of Materials,
302
R. J. KIRKPATRICK ET AL.
9, 241ÿ247.
Bish, D.L. (1980) Anion-exchange in takovite: applica-
tions to other hydroxide minerals. Bulletin de
Mineralogie, 103, 170ÿ175.
Boek, E.S., Coveney, P.V. and Skipper, N.T. (1995)
Monte Carlo molecular modeling studies of hydrated
Li-, Na- and K-smectites ÿ Understanding the role
of potassium as a clay swelling inhibitor. Journal of
the American Chemical Society, 117, 12608ÿ12617.
Bougeard, D., Smirnov, K. and Geidel, E. (2000)
Vibrational spectra and structure of kaolinite: A
computer simulation study. Journal of Physical
Chemistry B, 104, 9210ÿ9217.
Braterman, P.S., Xu, Z.P. and Yarberry, F. (2004)
Chemistry of layered double hydroxides. Pp.
373ÿ474 in: Handbook of Layered Materials (S.A.
Auerbach, K.A. Carrado and P.K. Dutta, editors).
Marcel Dekker, New York.
Bridgeman C.H. and Skipper, N.T. (1997) A Monte
Carlo study of water at an uncharged clay surface.
Journal of Physics: Condensed Matter , 9,
4081ÿ4087.
Bridgeman, C.H., Buckingham, A.D., Skipper, N.T. and
Payne, M.C. (1996) Ab-initio total energy study of
uncharged 2:1 clays and their interaction with water.
Molecular Physics, 89, 879ÿ888.
Brown, G.E. (2001) Surface science ÿ How minerals
react with water. Science, 294, 67ÿ69.
Brown, G.E., Henrich, V.E., Casey, W.H., Clark, D.L.,
Eggleston, C., Felmy, A., Goodman, D.W., GraÈtzel,
M., Maciel, G., McCarthy, M.I., Nealson, K.H.,
Sverjensky, D.A., Toney, M.F. and Zachara, J.M.
(1999) Metal oxide surfaces and their interactions
with aqueous solutions and microbial organisms.
Chemical Review, 99, 77ÿ174.
Cavani, F., Tri®ro, F. and Vaccari, A. (1991)
Hydratalcite-type anion clays: preparation, proper-
ties and applications. Catalysis Today, 11, 173ÿ301.
Chang, F.R.C., Skipper, N.T. and Sposito, G. (1995)
Computer simulations of interlayer molecular
structure in sodium montmorillonite hydrates.
Langmuir, 11, 2074ÿ2082.
Chang, F.R.C., Skipper, N.T. and Sposito, G. (1998)
Monte Carlo and molecular dynamics simulations of
electrical double-layer structure in potassium-mon-
tmorillonite hydrates. Langmuir, 14, 1201ÿ1207.
Chau, P.-L. and Hardwick, A.J. (1998) A new order
parameter for tetrahedral con®gurations. Molecular
Physics, 93, 511ÿ518.
Cheng, L., Fenter, P., Nagy, K.L., Schlegel, M.L. and
Sturchio, N.C. (2001) Molecular-scale density
oscillations in water adjacent to a mica surface.
Physical Review Letters, 8715, 156103ÿ1-4.
Choy, J., Kwak, S., Park, J., Jeong, Y. and Portier, J.
(1999) Intercalative nanohybrids of nucleoside
monophosphates and DNA in layered metal hydro-
xide. Journal of the American Chemical Society,
121, 1399ÿ1400.
Criscenti, L.J. and Sverjensky, D.A. (1999) The role of
electrolyte anions (ClO4
ÿ
, NO3
ÿ
, and Clÿ
) in
divalent metal (M2+) adsorption on oxide and
hydroxide surfaces in salt solutions. American
Journal of Science, 299, 828ÿ899.
Cygan, R.T. (2001) Molecular modeling in mineralogy
and geochemistry. Pp. 1ÿ36 in: Molecular Modeling
Theory and Applications in the Geosciences (R.T.
Cygan and J.D. Kubicki, editors). Reviews in
Mineralogy and Geochemistry 42, Mineralogical
Society of America and the Geochemical Society,
Washington, D.C.
Cygan, R.T. and Kubicki, J.D., editors (2001) Molecular
Modeling Theory and Applications in the
Geosciences. Reviews in Mineralogy and
Geochemistry, 42, Mineralogical Society of
Amer ica and the Geochemica l Soc ie ty ,
Washington, D.C.
Cygan, R.T., Liang, J.J. and Kalinichev, A.G. (2004).
Molecular models of hydroxide, oxyhydroxide, and
clay phases and the development of a general force
®eld. Journal of Physical Chemistry B, 108,
1255ÿ1266.
Delville, A. (1995) Monte Carlo simulations of surface
hydration ÿ An application to clay wetting. Journal
of Physical Chemistry, 99, 2033ÿ2037.
Desiqueira, A.V.C., Skipper, N.T., Coveney, P.V. and
Boek, E.S. (1997) Computer simulation evidence for
enthalpy driven dehydration of smectite clays at
elevated pressures and temperatures. Molecular
Physics, 92, 1ÿ6.
Dore, J. (2000) Structural studies of water in con®ned
geometry by neutron diffraction. Chemical Physics,
258, 327ÿ347.
Eisenberg, D. and Kauzmann, W. (1969) The Structure
and Properties of Water. Oxford University Press,
Oxford, 296 pp.
Engelhardt, H. and Kamb, B. (1981) Structure of ice IV,
a metastable high-pressure phase. Journal of
Chemical Physics, 75, 5887ÿ5899.
Errington, J.R. and Debenedetti, P.G. (2001)
Relationship between structural order and the
anomalies of liquid water. Nature, 409, 318ÿ321.
Fenter, P., Teng, H., Geissbuhler, P., Hanchar, J.M.,
Nagy, K.L. and Sturchio, N.C. (2000a) Atomic-scale
structure of the orthoclase (001)-water interface
measured with high-resolution X-ray re¯ectivity.
Geochimica et Cosmochimica Acta, 64, 3663ÿ3673.
Fenter, P., Geissbuhler, P., DiMasi, E., Srajer, G.,
Sorensen, L.B. and Sturchio, N.C. (2000b) Surface
speciation of calcite observed in situ by high-
resolution X-ray re¯ectivity. Geochimica et
Cosmochimica Acta, 64, 1221ÿ1228.
Ford, R.G. and Sparks, D.L. (1998) Potential formation
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
303
of secondary hydrotalcite-like precipitates during Zn
and Cu sorption to pyrophyllite. Mineralogical
Magazine, 62A, 462ÿ463.
Ford, R.G., Scheinost, A.C., Scheckel, K.G. and Sparks,
D.L. (1999) The link between clay mineral weath-
ering and the stabilization of nickel surface
precip i ta tes . Environmental Science and
Technology, 33, 3140ÿ3144.
Fouzri, A., Dorbez-Sridi, R. and Oumezzine, M. (2002)
Water con®ned in silica gel and in vycor glass at low
and room temperature, X-ray diffraction study.
Journal of Chemical Physics, 116, 791ÿ797.
Gade, B., Riedmiller, J., Westerrmann, H., Heindl, A.
and Pollmann, H. (1999) Mineralogical investiga-
tions and chemical equilibrium calculations on the
hazardous waste land®ll of Raindorf/Germany.
Neues Jahrbuch fuÈr Mineralogie Abhandlungen,
174, 249ÿ275.
Gade, B., Heindl, A., Westermann, H. and Pollmann, H.
(2000) Secondary mineral inventory of hazardous
waste land®lls. Proceedings of the 6thInternational
Congress on Applied Mineralogy, 2, 539ÿ542.
Gallom P., Rapinesi, M. and Rovere, M. (2002)
Con®ned water in the low hydration regime.
Journal of Chemical Physics, 117, 369ÿ375.
Geiger, A. and Stanley, H.U. (1982) Low-density
`patches' in the hydrogen-bond network of liquid
water: Evidence from molecular-dynamics computer
simulations. Physics Review Letters , 49 ,
1749ÿ1752.
Genin, J.-M.R., Refait, P., Bourrie, G., Abdelmoula, M.
and Trolard, F. (2001) Structure and stability of the
Fe(II)-Fe(III) green rust "fougerite" mineral and its
potential for reducing pollutants in soil solutions.
Applied Geochemistry, 16, 559ÿ570.
Greathouse, J. and Sposito, G. (1998) MC and MD
studies of interlayer structure in Li(H2O)3-smectites.
Journal of Physical Chemistry B, 102, 2406ÿ2414.
Greathouse, J., Refson, K. and Sposito, G. (2000)
Molecular dynamics simulations of water mobility
in magnesium-smectite hydrates. Journal of the
American Chemical Society, 122, 11459ÿ11464.
Gordillo, M.C. and MartõÂ, J. (2000) H-bond structure of
liquid water con®ned in nanotubes. Chemical
Physics Letters, 329, 341ÿ345.
Guillot, B. (2002) A reappraisal of what we have learnt
during three decades of computer simulations on
water. Journal of Molecular Liquid, 101, 219ÿ260.
Hartnig, C., Witschel, W. and Spohr, E. (1998)
Molecular dynamics study of the structure and
dynamics of water in cylindrical pores. Journal of
Physical Chemistry, 102, 1241ÿ1249.
Head-Gordon, T. and Hura, G. (2002) Water structure
from scattering experiments and simulation.
Chemical Reviews, 102, 2651ÿ2670.
Heinzinger, K. (1990) Molecular dynamics simulation
of aqueous systems. Pp. 357ÿ394 in: Computer
Modelling of Fluids, Polymers and Solids
(C.R.A.Catlow et al., editors). Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Hochella, M.F. Jr. and White, A.F. (1990) Mineral-water
interaction geochemistry: An overview. Pp. 1ÿ16 in:
Interface Geochemistry (M.F. Hochella and A.F.
White, editors). Reviews in Mineralogy, 23,
Mineralogical Society of America, Washington, D.C.
Hou, X., Kalinichev, A.G. and Kirkpatrick, R.J. (2002)
Interlayer structure and dynamics of Clÿ
-LiAl2
layered double hydroxide:35Cl NMR observations
and molecular dynamics modeling. Chemistry of
Materials, 14, 2078ÿ2085.
Hou, X., Bish, D.L., Wang, S.-L., Johnston, C.T. and
Kirkpatrick, R.J. (2003) Hydration, Expansion,
Structure and Dynamics of Layered Double
Hydroxides. American Mineralogist, 88, 167ÿ179.
Israelachvili, J.N. and Pashley, R.M. (1983) Molecular
laying of water at surfaces and origin of repulsive
hydration forces. Nature, 306, 249ÿ250.
Israelachvili, J.N. and WennerstroÈn, H. (1996) Role of
hydration and water structure in biological and
colloidal interactions. Nature, 379, 219ÿ225.
Jorgensen, W.L., Chandrasekhar, J., Madura, J.F.,
Impey, R.W. and Klein, M.L. (1983) Comparison
of simple potential functions for simulating liquid
water. Journal of Chemical Physics, 79, 926ÿ935.
Kagunya, W., Hassan, Z. and Jones, W. (1996) Catalytic
properties of layered double hydroxides and their
calcined derivatives. Inorganic Chemistry, 35,
5970ÿ5974.
Kagunya, W., Baddour-Hadjean, R., Kooli, F. and Jones,
W. (1998) Vibrational modes in layered double
hydroxides and their calcined derivatives. Chemical
Physics, 236, 225ÿ234.
Kalinichev, A.G. (2001) Molecular simulations of liquid
and supercritical water: thermodynamics, structure
and hydrogen bonding. Pp. 83ÿ129 in: Molecular
Modeling Theory and Application in the Geosciences
(R.T. Cygan and J.D. Kubicki, editors). Reviews in
Mineralogy and Geochemistry, 42, Mineralogical
Society of America and the Geochemical Society,
Washington, D.C.
Kalinichev, A.G. and Kirkpatrick, R.J. (2002) Molecular
dynamics modeling of chloride binding to the
surfaces of Ca hydroxide, hydrated Ca-aluminate
and Ca-silicate phases. Chemistry of Materials, 14,
3539ÿ3549.
Kalinichev, A.G., Gorbaty, Y.E. and Okhulkov, A.V.
(1999) Structure and H-bonding of liquid water at
high hydrostatic pressure: Monte Carlo NPT-
ensemble simulations up to 10 kbar. Journal of
Molecular Liquid, 82, 57ÿ72.
Kalinichev, A.G., Kirkpatrick, R.J. and Cygan, R.T.
(2000) Molecular modeling of the structure and
304
R. J. KIRKPATRICK ET AL.
dynamics of the interlayer and surface species of
mixed-metal layered hydroxides: chloride and water
in hydrocalumite (Friedel's salt). American
Mineralogist, 85, 1046ÿ1052.
Karaborni, S., Smit, B., Heidug, W., Urai, J. and
Vanoort, E. (1996) The swelling of clays ÿ
Molecular simulations of the hydration of montmor-
illonite. Science, 271, 1102ÿ1104.
Kirkpatrick, R.J., Yu, P., Hou, X. and Kim, Y. (1999)
Interlayer structure, anion dynamics, and phase
transitions in mixed-metal layered hydroxides:
variable temperature35Cl NMR spectroscopy of
hydrotalcite and Ca-aluminate hydrate. American
Mineralogist, 84, 1186ÿ1190.
Kirkpatrick, R.J., Kalinichev, A.G., Wang J., Hou, X.
and Amonette, J.E. (2005) Molecular modeling of
the vibrational spectra of interlayer and surface
species of layered double hydroxides. Pp. 239ÿ285
in: The Application of Vibrational Spectroscopy to
Clay Minerals and Layered Double Hydroxides (J.T.
Kloprogge, editor). CMS Workshop Lectures, 13,
The Clay Minerals Society, Aurora, Colorado, USA.
Kuhs, W.F., Finney, J.L., Vettier, C. and Bliss, D.V.
(1984) Structure and hydrogen ordering in ice VI,
VII, and VIII by neutron powder diffraction. Journal
of Chemical Physics, 81, 3612ÿ3623.
Lee, S.H. and Rossky, P. J. (1994) A comparison of the
structure and dynamics of liquid water at hydro-
phobic and hydrophilic surface ÿ a molecular
dynamics simulation study. Journal of Chemical
Physics, 100, 3334ÿ3345.
Luzar, A. (2000) Resolving the H-bond dynamics
conundrum. Journal of Chemical Physics, 113,
10663ÿ10675.
Marx, D. (2004) Throwing tetrahedral dice. Science,
303, 634ÿ636.
McCarthy, M.I., Schenter, G.K., Scamehorn, C.A. and
Nicholas, J.B. (1996) Structure and dynamics of the
water/MgO interface. Journal of Physical Chemistry,
100, 16989ÿ16995.
Michot, L.J., VillieÂras, F., Francois, M., Bihannic, I.,
Pelletier, M. and Cases, J.-M. (2002) Water
organization at the solid-aqueous solution. Comptes
Rendus Geoscience, 334, 611ÿ631.
Miranda, P.B. and Shen, Y.R. (1999) Liquid interfaces:
a study by sum-frequency vibrational spectroscopy.
Journal of Physical Chemistry B, 103, 3292ÿ3298.
Miyata, S. (1983) Anion-exchange properties of
hydrotalcite-like compounds. Clays and Clay
Minerals, 31, 305ÿ311.
Nandi, N., Bhattacharyya, K. and Bagchi, B. (2000)
Dielectric relaxation and solvation dynamics of
water in complex chemical and biological systems.
Chemical Review, 100, 2013ÿ2045.
Newman, S. and Jones, W. (1998) Synthesis, character-
ization and applications of layered double hydro-
xides containing organic guests. New Journal of
Chemistry, 22, 105ÿ115.
Odelius, M., Bernasconi, M. and Parrinello, M. (1997)
Two dimensional ice adsorbed on mica surface.
Physical Review Letters, 78, 2855.
Packer, K.J. (1977) The dynamics of water in
heterogeneous systems. Philosophical Transactions
of the Royal Society of London B, 278, 59ÿ87.
Park, S.-H. and Sposito, G. (2002) Structure of water
adsorbed on a mica surface. Physical Review Letters,
89, 085501.
Paulo, A.N., Starr, F.W., Barbosa, M.C. and Stanley,
H.E. (2002) Relation between structure and dynami-
cal anomalies in supercooled water. Physics A, 314,
470ÿ476.
Raviv, U., Laurat, P. and Klein, J. (2001) Fluidity of
water con®ned to subnanometre ®lms. Nature, 413,
51ÿ54.
Rapaport, D.C. (1983) H-bonds in water: Network
organization and lifetimes. Molecular Physics, 50,
1151ÿ1162.
Rustad, J.R. (2001) Molecular models of surface
relaxation, hydroxylation and surface charging at
oxide-water interfaces. Pp. 169ÿ197 in: Molecular
Modeling Theory and Applications in the
Geosciences (R.T. Cygan and J.D. Kubicki, editors).
Reviews in Mineralogy and Geochemistry 42,
Mineralogical Society of America and the
Geochemical Society, Washington, D.C.
Rustad, J.R., Felmy, A.R. and Bylaska, E.J. (2003)
Molecular simulation of the magnetite-water inter-
face. Geochimica et Cosmochimica Acta, 67,
1001ÿ1016.
Saitta, A.M. and Datchi, F. (2003) Structure and phase
diagram of high density water: The role of interstitial
molecules. Physics Review E, 67, 020201.
Sakuma, H., Tsuchiya, T., Kawamura, K. and Otsuki, K.
(2003) Large self-diffusion of water on brucite
surface by ab initio potential energy surface and
molecular dynamics simulations. Surface Science,
536, L396ÿ402.
Schwegler, E., Galli, G. and Gygi, F. (2000) Water
under pressure. Physical Review Letters, 84,
2429ÿ2432.
Smith, D.E. (1998) Molecular computer simulations of
the swelling properties and interlayer structure of
cesium montmorillonite. Langmuir, 14, 5959ÿ5967.
Soper, A.K. (2000) The radial distribution functions of
water and ice from 220 to 673 K and at pressure up
to 400 MPa. Chemical Physics, 258, 121ÿ137.
Spohr, E. and Hartnig, C. (1999) Water in porous
glasses. A computer simulations study. Journal of
Molecular Liquids, 80, 165ÿ178.
StoÈckelmann, E. and Hentschke, R. (1999) A molecular-
dynamics simulation study of water on NaCl (001)
using a polarizable model. Journal of Chemical
MOLECULAR DYNAMICS OF WATER-MINERAL INTERACTION
305
Physics, 110, 12097ÿ12107.
Sutton, R. and Sposito, G. (2001) Molecular simulation
of interlayer structure and dynamics in 12.4 AÊ Cs-
smectite hydrates. Journal of Colloid and Interface
Science, 237, 174ÿ184.
Teixeira J., Bellissent-Funelt, M-C. and Chen, S.-H.
(1990) Dynamics of water studied by neutron
scattering. Journal of Physics: Condensed Matter,
2, SA105ÿSA108.
Teleman, O., JoÈnsson, B. and EngstroÈm, S. (1987) A
molecular dynamics simulation of a water model
with intramolecular degrees of freedom. Molecular
Physics, 60, 193ÿ203.
Teschke, O., Ceotto, G. and de Souza, E.F. (2000)
Interfacial aqueous solutions dielectric constant
measurements using atomic force microscopy.
Chemical Physics Letters, 326, 328ÿ334.
Teppen, B.J., Rasmussen, K., Bertsch, P.M., Miller,
D.M., and Schafer, L. (1997) Molecular dynamics
modeling of clay minerals.1. Gibbsite, kaolinite,
pyrophyllite, and beidellite. Journal of Physical
Chemistry B, 101, 1579ÿ1587.
Terzis, A., Filippakis, S., Kuzel, H.-J. and Burzlaff, H.
(1987) The crystal structure of Ca2Al(OH)6Cl´2H2O,
Zeitschrift fuÈr Kristallographie, 181, 29ÿ34.
Thompson, H.A., Parks, G.A. and Brown, G.E. (1999)
Ambient-temperature synthesis, evolution, and char-
acterization of cobalt-aluminum hydrotalcite-like
solids. Clays and Clay Minerals, 47, 425ÿ438.
Trolard, F., Genin, J.-M.R., Abdelmoula, M., Bourrie,
G., Humbert, B. and Herbillon, A. (1997)
Identi®cation of a green rust mineral in a
reductomorphic soil by MoÈssbauer and Raman
spectroscopies. Geochimica et Cosmochimica Acta,
61, 1107ÿ1111.
Wang, J. (2004) Molecular scale structure, diffusional
dynamics, and hydration energetics of nano-con®ned
water and water on mineral surfaces. PhD thesis,
University of Illinois at Urbana-Champaign, USA.
Wang, J., Kalinichev, A.G., Amonette, J E. and
Kirkpatrick, R.J. (2003) Interlayer structure and
dynamics of Cl-Hydrotalcite: Far infrared spectro-
scopy and molecular dynamics modeling. American
Mineralogist, 88, 398ÿ409.
Wang, J., Kalinichev, A.G. and Kirkpatrick, R.J. (2004)
Molecular structure of water con®ned in brucite.
Geochimica et Cosmochimica Acta, 68, 3351ÿ3365.
Yu, P. and Kirkpatrick, R.J. (2001)35Cl NMR relaxation
study of cement hydrate suspensions. Cement and
Concrete Research, 31, 1479ÿ1485.
Zhu, Y. and Granick, S. (2001) Viscosity of interfacial
water. Physical Review Letters, 87, 09614.
[Manuscript received 22 August 2004:
revised 7 January 2005]
306
R. J. KIRKPATRICK ET AL.