See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6445309

Structure and dynamics of the hydration shells of the Al3+ ion

Article  in  The Journal of Chemical Physics · April 2007

DOI: 10.1063/1.2566868 · Source: PubMed

CITATIONS READS
42 430

4 authors:

Eric J. Bylaska Marat Valiev

Pacific Northwest National Laboratory Pacific Northwest National Laboratory
157 PUBLICATIONS   8,192 CITATIONS    104 PUBLICATIONS   5,694 CITATIONS   

SEE PROFILE SEE PROFILE

James Rustad John H Weare

University of California, Davis University of California, San Diego
148 PUBLICATIONS   5,276 CITATIONS    182 PUBLICATIONS   8,720 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

simulation in material science View project

CZT Crystal Growth Modeling View project

All content following this page was uploaded by John H Weare on 05 February 2015.

The user has requested enhancement of the downloaded file.

 THE JOURNAL OF CHEMICAL PHYSICS 126, 104505 共2007兲

Structure and dynamics of the hydration shells of the Al3+ ion

Eric J. Bylaskaa兲 and Marat Valiev
Pacific Northwest National Laboratory, P.O. Box 999, Richland, Washington 99352
James R. Rustad
Department of Geology, University of California, Davis, Davis, California 95616
John H. Weare
Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla,
California 92093
共Received 29 November 2006; accepted 17 January 2007; published online 14 March 2007兲

First principles simulations of the hydration shells surrounding Al3+ ions are reported for
temperatures near 300 ° C. The predicted six water molecules in the octahedral first hydration shell
were found to be trigonally coordinated via hydrogen bonds to 12 s shell water molecules in
agreement with the putative structure used to analyze the x-ray data, but in disagreement with the
results reported from conventional molecular dynamics using two-and three-body potentials. Bond
lengths and angles of the water molecules in the first and second hydration shells and the average
radii of these shells also agreed very well with the results of the x-ray analysis. Water transfers into
and out of the second solvation shell were observed to occur on a picosecond time scale via a
dissociative mechanism. Beyond the second shell the bonding pattern substantially returned to the
tetrahedral structure of bulk water. Most of the simulations were done with 64 solvating water
molecules 共20 ps兲. Limited simulations with 128 water molecules 共7 ps兲 were also carried out.
Results agreed as to the general structure of the solvation region and were essentially the same for
the first and second shell. However, there were differences in hydrogen bonding and Al–O radial
distribution function in the region just beyond the second shell. At the end of the second shell a
nearly zero minimum in the Al–O radial distribution was found for the 128 water system. This
minimum is less pronounced minimum found for the 64 water system, which may indicate that sizes
larger than 64 may be required to reliably predict behavior in this region. 关DOI: 10.1063/1.2566868兴

I. INTRODUCTION
An understanding of the structure and dynamics of the
water molecules in the hydration shells surrounding ions is
essential to the interpretation of many chemical processes in
aqueous solutions. X-ray and neutron scattering results have
been reported which provide direct results about shell struc-
ture for many ionic species.1,2 Information about the dynam-
ics of water molecules in this region has also been obtained
from other probes such as NMR, infrared spectroscopy, and
inelastic neutron scattering.1,2 For singly charged ions
共Na+ , Li+兲 a structured first hydration shell can be identified.
The residence time in this shell is short 共e.g., ⬍1 ps for Na+兲
and the change in the solution structure due to the presence
of the ion does not extend far beyond first nearest neighbor
water molecules.3 However, for highly charged metal ions
such as Al3+, which play a role in many important chemical
processes, experimental information providing the structure
in the region near the ion is much more difficult to obtain.
From the interpretation of x-ray, extended x-ray absorption
fine structure, NMR, and neutron scattering data there is evi-
dence of a well structured first solvation shell with six
共Al3+ , Cr3+兲 共Refs. 4–8兲 to eight 共Dy3+ , Yb3+兲 共Ref. 9兲
strongly bound water molecules with very long lifetimes.
a兲
Author to whom correspondence should be addressed. FAX: 1-509-376
3650. Electronic mail: eric.bylaska@pnl.gov
The residence times of the water molecules in the first hy-
dration shell of triply charges ions vary over many orders of
magnitude 共Ti3+, ⬇10−5, Ir3+, ⬇1010 s兲. For Al3+ the resi-
dence time is quite long 共⬇1 s兲. There is also evidence that
the first hydration shell is followed by more labile but still
structured second and possibly third hydration shells.5,7,10
The bonding pattern and the dynamics of the molecules
in the second shell are critical to the interpretation of the ion
association properties and reaction chemistry of the ion in
solution. Unfortunately, determining the structure of the sec-
ond shell and the existence of an ion related structure in the
third shell is very difficult to obtain from x-ray or neutron
data. In order to relate the scattering intensities in terms of a
shell structure, a geometric model of the molecular arrange-
ment must be invoked. The validity of the assumed model
must be demonstrated by the agreement of the scattering pre-
diction from the model with observations.5 While this type of
fitting can produce excellent agreement between the experi-
mental data and the results of model calculations, small but
important changes in structure are difficult to resolve. As a
consequence, information about the structure of the solution
beyond the second shell and the behavior of the solution as it
returns to the bulk structure11 are not available for Al3+ and
other highly charge metal ions.
Since these are strongly interacting systems, the most

0021-9606/2007/126共10兲/104505/8/$23.00 126, 104505-1

Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
104505-2 Bylaska et al. J. Chem. Phys. 126, 104505 共2007兲

TABLE I. The average parameters of the hydration shell of aqueous Al3+ from three different AIMD simula-
tions. Results from analysis of various experimental data are also included for comparison in the sixth column.

Al3+ + 6H2O Al3+ + 64H2O Al3+ + 64H2O Al3+ + 128H2O

Average parameters 共gas phase兲 共I兲 共II兲 共III兲 Expt.

Initial temperature 共K兲 400 300 300

Equilibration timea 共ps兲 3 3 2
Collection timeb 共ps兲 17.34 12.53 7.11
Final temperature 共K兲 326 263 257
rAl–OI 1.937 1.94 1.94 1.93 1.87–1.90d
␴Al–OI ¯ 0.074 0.066 0.059 0.04–0.103c
rOI–HI 0.961 0.99 0.99 0.99 0.99e
␴OI–HI ¯ 0.038 0.034 0.038 ¯
First shell ⬔HOH 107.36 107.35 107.6 107.2 108–114e 120f
␴⬔HOH 5.44 5.09 5.04 ¯
First shell ⬔Tilt 27.8g 24.98g 29.6g 34± 6h
rOI–OII 2.68 2.68 2.66 2.63h, 2.68–2.73c
␴OI–OII 0.14 0.13 0.12 0.02–0.087c
rAl–OII 4.09 4.12 4.08 4.01–4.15c,d
␴Al–OII 0.23 0.21 0.19 0.23–0.33c
⬔OI–H–OII 163.8 165 166 180f
␴⬔OI–H–OII 8.19 7.68 7.28 ¯
Second shell H2O 12 12 12 13± 1d,i
a
Constant temperature simulation.
b
Constant energy simulation.
c
References 6 and 27.
d
Reference 4.
e
From analysis of neutron diffraction data of 2.2M chromium 共III兲 percholate in water 共Ref. 28兲.
f
From analysis x-ray diffraction data of aqueous Cr3+.
g
Tilt angle was determined on the basis of rCr–OI, rOI–HI, and first shell ⬔HOH, see Eq. 共1兲.
h
From analysis 共Refs. 28 and 30兲 of aqueous Cr3+.
i
Reference 5.

reliable path to theoretical interpretation is direct simulation
at the molecular level. For singly charged ions, both first
principles 共ab initio molecular dynamics,12–14 AIMD兲 and
classical molecular dynamics 共based on empirical intermo-
lecular potentials, MD兲 methods are in reasonable agreement
with each other and with the experimental data. However,
even for these simple systems important differences in sol-
vent structure are seen between AIMD and MD
simulations.3,15 For highly charged species, such as Al3+,
conventional MD methods are even more difficult to apply,
because the strong polarization of the surrounding water
molecules from +3 ion makes it difficult to develop an accu-
rate representation of the solvent-ion interaction.16,17 AIMD
methods, on the other hand, calculate the interaction between
species directly from the electronic Schrödinger equation as
the simulation proceeds. They, therefore, do not need to in-
troduce an intermolecular potential. A serious drawback of
using AIMD methods is that they are extremely time con-
suming and they require a significant amount of computa-
tional resources. However, there has been a considerable
amount of progress in recent years in making AIMD methods
efficient,14,18 and it is now possible to simulate up to 128
water molecules with modest computational resources.
We and others have been using AIMD methods to study
the hydration and hydrolysis of aluminum,10,19,20 and with
this study we report extensive AIMD simulations of Al3+
ions in aqueous solutions using up to 128 water molecules.
Our focus is on elucidation of the transition from a highly
structured first solvation shell to the bulk structure of water,
as well as comparisons with putative structural interpreta-
tions used to analyze the x-ray data.
The article is organized as follows. In Sec. II, the com-
putational methods used in this work are described. The re-
sults of Al3+ AIMD simulations and a detailed analysis of the
hydrogen bonding structure of the first, second, and outer
hydration shells are presented in Sec. III. Concluding re-
marks are given in Sec. IV.
II. METHODS
Car-Parrinello molecular dynamics simulations,12 within
framework of density Functional Theory 共DFT兲,21 were per-
formed using the pseudopotential plane-wave program
共NWPW module兲 contained in the NWCHEM computation
chemistry package.22 The DFT equations were solved using
the gradient corrected PBE96 exchange-correlation
functional.23 Valence electron interactions were approxi-
mated using generalized norm-conserving Hamann
pseudopotentials24 modified into a separable form suggested
by Kleinman and Bylander.25 For gradient corrected calcula-
tions, the NWPW module automatically revises the pseudopo-
tentials by generating them with the specified exchange
correlation functional. The original pseudopotential param-
etrizations suggested by Hamann were too “hard,” and softer
pseudopotentials were constructed by increasing the core ra-
dii, with the following core radii, H: rcs = 0.8 a.u., rcp
= 0.8 a.u.; O: rcs = 0.7 a.u., rcp = 0.7 a.u., rcd = 0.7 a.u.; Al: rcs

Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
104505-3 Hydration shells of Al3+
FIG. 1. The radial distribution functions, gAl–O共r兲, and running coordination
number functions, nAlO共r兲, from AIMD simulations of aqueous Al3+ 共solid
lines, 400 K Al3+ + 64H2O; dashed lines, 300 K Al3+ + 64H2O; dot-dashed
lines, 300 K Al3+ + 128H2O兲.
= 1.241 a.u., rcp = 1.577 a.u., rcd = 1.577 a.u.. The electronic
wave functions were expanded using a plane-wave basis set
with periodic boundary conditions, sampled at the ⌫ point,
with a wave function cutoff energy of 72 Ry and a density
cutoff energy of 165 Ry. The triple charges at the Al centers
were neutralized by a homogeneous negative charge back-
ground. The Car-Parrinello equations of motion were inte-
grated with a time step of 0.17 fs and a fictitious mass of
750 a.u. To facilitate the integration all the hydrogen atoms
were replaced by deuterium.
In order to properly describe the second solvation shell
we have performed simulations with 64 water molecules
共256 orbitals and 512 electrons兲. The choice of this system
size was validated by limited simulations using 128 water
molecules. The simulation cell 共12.4 Å for 64 water mol-
ecules 15.6 Å for 128 water molecules兲 was chosen to pro-
vide a density of water near 1. Calculations were initiated
from a starting configuration taken from molecular dynamics
simulations using a potential developed for the Al3+ ion by
Wasserman et al.16 Car-Parrinello simulations were per-
formed in the presence of a Nosé-Hoover thermostat for the
first 2 – 3 ps, after which the data were collected in a constant
energy simulation. Simulations were initialized at tempera-
tures of 300 and 400 K. The two different temperatures were
used for comparison. During the data collection, a small
amount of electronic heating was present in the simulations,
which resulted in a slight decrease in the ionic kinetic energy
during the simulations 共⬃5 K / ps兲.
Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
J. Chem. Phys. 126, 104505 共2007兲
FIG. 2. 共Color兲 The Boys localized orbitals of the isolated Al3+ + 6H2O
species.
III. RESULTS
The hydration shell parameters obtained from several
AIMD calculations, along with measured experimental val-
ues, are given in Table I. As expected, all simulations gave
very similar results for the structural parameters and agree
quite well with available experimental data. Somewhat sur-
prisingly, raising the temperature had an almost negligible
effect, with only slight broadening of the standard deviations
for the various structural parameters in the 400 K simulation.
The Al3+ + 64H2O and Al3+ + 128H2O simulations show an
average first shell bond distances rAl–OI of 1.94, 1.94, and
1.93 Å, respectively. These distances are slightly longer than
the 1.88– 1.90 Å obtained from x-ray data,4,6,26,27 but consis-
tent with the calculated bond length of the isolated
Al共H2O兲3+ 6 共our results with the same level of DFT兲. The
peak widths, ␴Al–OI, from these simulations are 0.066, 0.074,
and 0.059 Å and in good agreement with the measured val-
ues of 0.04– 0.103 Å. The average OH bond length in the
first shell, rOI–HI, is 0.99 Å, which is slightly longer than
0.97 Å in the isolated Al共H2O兲3+ 6 共our calculation兲, but it
agrees with neutron diffraction measurements for the similar
Cr3+ ion. The second shell peaks in rAl–OII are at 4.09, 4.12,
and 4.08 Å 共␴Al–OII = 0.23, 0.21, and 0.19 Å兲, respectively.
This is also in very good agreement with results from x-ray
data 共rAl–OII = 4.01– 4.15 Å, ␴Al–OII = 0.224– 0.332 Å兲. The
predicted hydrogen bond length between the first and second
shell water molecules, rOI–OII = 2.66– 2.68 Å, is very close to
the x-ray value of 2.68– 2.73 Å.
The radial distribution functions, gAl–O共r兲, and running
coordination number functions, nAlO共r兲, from AIMD simula-
tions of 300 K Al3+ + 64H2O 共12.5 ps兲, 400 K Al3+ + 64H2O
共17.3 ps兲, and 300 K Al3+ + 128H2O 共7.1 ps兲 are given in
Fig. 1. The well defined and isolated peaks in the radial
distribution functions, gAl–O共r兲, in the range of 1.80– 2.10 Å
are evidence of a stable octahedral first coordination shell.
The near trigonal planar configuration and the presence of a
highly charged aqua ion resulted in the conjecture that the
first shell water oxygen will rehybridize from sp3 towards
104505-4 Bylaska et al.
sp2 bonding.5 Indeed, initial x-ray diffraction studies, which
reported large ⬔HOH angles of 115° and 120° for Al3+ and
Cr3+, appeared to support this hypothesis. However, these
early estimates turned out to be incorrect, because they relied
on a linear ⬔OI – HI – OII bond assumption, and subsequent
neutron diffraction measurements have since showed that
⬔HOH angles range from 108° to 114° for the Cr3+ aqua
ion.28 As expected, our simulations do not support the sp2
bonding conjecture originally proposed,5 but instead support
the more recent data.28 The average first shell ⬔HOH of ␪
= 107.2° – 107.6° is in reasonable agreement with the neutron
measurements, and plots of the electronic wave functions,
localized using the Boys localization procedure 共Fig. 2兲,
clearly identify the first shell water molecules as having sp3
bonding.
An effective angle of tilt, ␺, of the first shell water mol-
ecules can be obtained from neutron diffraction studies.29,30
The following formula, derived from the geometric model
shown in Fig. 3, can be used to determine ␺ in terms of the
first shell parameters, rAl–OI, rAl–HI, rOI–HI, and ⬔HOH, ␪:
␺ = arccos 冉 2
rAl–HI 2
− rAl–OI 2
− rOI–HI
2rAl–OIrOI–HI cos共␪/2兲
.冊
The calculated angles, ␺, from the simulations are 27.8°,
25.0°, and 29.6°. These results are in good agreement with
the value of 34° reported for Cr3+ aqua ion,28,30 and they do
FIG. 4. Left: Distribution of the first solvation shell tilt angle ␺. Solid line; ␺ determined by angle between the bisector of H2O and the Al3+ – O axis. Dashed
line: ␺ determined by angle between the plane spanned by H2O and the Al3+ – O axis. Right: Contour plot of the distribution in terms of the two different
definitions of ␺ of the first solvate shell.
Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
J. Chem. Phys. 126, 104505 共2007兲
FIG. 3. Geometric model used to derive formula for the
tilt angle ␺ in terms of rAl–OI, rAl–HI, rOI–HI, and first
shell ⬔HOH.
not appear to support the conjecture that there is little pref-
erence for tilt in these systems.29 Notwithstanding this agree-
ment, these results are rather ambiguous as they clearly de-
pend on the definition of the geometric model.
In theory, a more direct measure of this angle could be
obtained by plotting the distribution of tilt angles over all
molecular configurations from a simulation. However, in
practice these plots can also produce ambiguous results.17 In
the left panel of Fig. 4, two different distributions of tilt
angles are shown from the 400 K Al3+ + 64H2O simulation.
The dashed line shows the distribution of the angle between
the Al–OI bond and the plane spanned by the water molecule
and the solid line shows the distribution of the angle between
the Al–OI bond and the bisector of the water molecule. In the
first case a broad flat distribution of tilt angles centered about
zero with no well defined maximum is seen, whereas in the
other case a well defined maximum located between 10° and
20° is seen. As shown in the correlation diagram 共right panel
of Fig. 4兲, there is a slight anticorrelation between ␺bisector
and ␺cross product near zero, such that ␺bisector ⬎ 0 when
共1兲
␺cross product = 0. Despite these differences the average angles,
␺, from these distributions averaged to be 24.0° and 21.9°,
respectively. Similar averages for ␺ have also been seen in
the MD simulations by Spangberg and Hermansson.17
The running coordination number functions, nAlO共r兲
共dashed lines in Fig. 1兲, predict 11–12 water molecules 共12
104505-5 Hydration shells of Al3+
FIG. 5. 共Color兲 共a兲 Trigonal second shell structure and 共b兲 tetrahedral struc-
ture just beyond the second shell 共a limited number of water molecules
retained for clarity兲.
for the Al3+ + 128H2O simulation兲 in the second shell region
共2.5– 4.5 Å兲 compared to 13± 1 water molecules obtained
from the analysis of the x-ray and IR data. In Fig. 5 typical
snapshots from the 400 K Al3+ + 64H2O simulation are
shown. In the top part of Fig. 5, the molecules beyond the
second shell have been removed and the first hydration shell
partially is hidden by a constant density contour. The second
shell displays a trigonal leaf structure, with the second shell
water molecules all forming acceptor hydrogen bonds with
the first shell water molecules. 共An acceptor hydrogen bond
is formed by sharing a lone pair, and a donor hydrogen bond
is formed by sharing a proton.兲 This structure agrees with
previous AIMD simulations10,20 and is consistent with bond
valence analysis and the putative structure used to interpret
the x-ray data. The trigonal tree structure illustrated in Fig. 5
is very different from the second shell surrounding alkali
metal ions. For these ions, AIMD simulations3 suggest that
the second shell is formed by hydrogen bonding three second
shell water molecules to each of the first shell water mol-
ecules in a tetrahedral bulk water conformation. Hence, em-
bedded alkali ions cause little change in the bulk water struc-
ture beyond the first solvation shell.
Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
J. Chem. Phys. 126, 104505 共2007兲
FIG. 6. 共Color兲 Acceptor bonding with superposed gAl–O共r兲 共light brown 64,
dark brown 128兲: red squares, number of first shell acceptor bonds: blue
circles, number of second shell acceptor bonds; green triangles, number of
third shell acceptor bonds.
Our results for the second shell structure do not agree
very well with the results from conventional MD methods.
This is somewhat anticipated, since the strong polarization of
the surrounding water molecules from +3 ion makes it diffi-
cult to develop an accurate solvent-ion interaction force field.
Even though the MD model of Wasserman et al. also predicts
a highly structured second shell comprised of 12 water mol-
ecules in a trigonal leaf structure, it disagrees with our AIMD
simulations as well as results from x-ray data. This MD
model predicts a highly compressed second hydration shell
共rAl–OII = 3.83 Å, ␴Al–OII = 0.15 Å兲. More recent MD models
based on two and three body potentials including
polarization17,31 are even more inconsistent with our results.
These models predict the presence of 18–24 water molecules
in the second shell, suggesting that the second shell is in a
tetrahedral structure.
Results for the region beyond the second shell are illus-
trated at the bottom part of Fig. 5. In this figure, several
water molecules in the outer bulk region are shown and the
first and second hydration shells are partially hidden by con-
stant density contour. Water molecules in this region 共bulk兲
are tetrahedrally coordinated, forming acceptor hydrogen
bonds to both second shell and bulk water molecules and two
donor hydrogen bonds to second shell water molecules, sug-
gesting that the coordination has substantially returned to the
tetrahedral hydrogen bonding of bulk water.
To provide more quantitative insight into the structure of
the hydration shells, a shell by shell acceptor hydrogen bond
analysis is given in Fig. 6. The solid symbols are from an
analysis of a 300 K Al3+ + 128H2O simulation and the hollow
symbols are from a 400 K Al3+ + 64H2O simulation. The su-
perimposed radial distributions gAl–O共r兲 identify the shell ra-
dii. 共In these calculations a hydrogen bond is defined by the
criterion of hydrogen bond length ⬍3.2 Å and the ⬔HOH
⬎ 140°.兲 The objectives of the analysis are to provide a way
to view the structural model for the behavior of the mol-
ecules near the ion and to identify the onset of bulk water
structure, i.e., tetrahedral bonding as we move away from the
104505-6 Bylaska et al. J. Chem. Phys. 126, 104505 共2007兲

ion center. Near the ion, since the water molecules are ar-
ranged in a trigonal planar tree structure, the number of ac-
ceptor hydrogen is near 1.0. Whereas, for the bulk tetrahedral
structure there are two acceptor hydrogen bonds for each
water molecule.
The red squares, in Fig. 6, illustrate the number of ac-
ceptor bonds from a water located at a distance x from Al
atom to a first shell water molecule. As expected, in the
range of 3.7⬍ x ⬍ 4.6 Å, this number is close to 1.0. How-
ever, for the extended 400 K Al3+ + 64H2O simulation the
number of acceptor bonds to the first shell is greater than 1 in
the inner region, which has very few molecules 共x ⬵ 3.5 Å兲.
This elevated acceptor bonding results from a rare event in
which two water molecules in the first shell coordinate to a
single water molecule in the inner part of the second shell.
We do not observe these structures in the Al3+ + 128H2O cal-
culation. The blue circles, in Fig. 6, show the number of
acceptor bonds from a water located x to a second shell water
molecule. Consistent with Fig. 5, this number begins to grow
as x approaches the bulk region. The green triangles show
the growth of the number of acceptor bonds from a water
molecule at x to a bulk water molecule. In this region the
total number of acceptor bonds is about 1.80 for 300 K
Al3+ + 128H2O simulation, which is close to that for bulk
water. A similar calculation for donor bonds leads to about
1.7 in this region, slightly smaller than for acceptor bonds
suggesting a slight perturbation of the structure by the pres-
ence of the ion. For the 64 water simulation there is less H
bonding beyond the second shell and a shallower minimum
at the end of the second shell in gAl–O共r兲, suggesting that 128
water molecules may be necessary to describe this region.
The exchange of solvent water molecules from the first
to the second hydration shell has attracted a great deal of
interest. Lifetimes determined from 17O NMR extend over FIG. 7. 共Color兲 Ligand exchange mechanism between the second and third
shells seen in 300 K Al3+ + 128H2O simulation, 共a兲 intermediate structure,
many orders of magnitude 共e.g., from 105 s for Cr3+ to and 共b兲 relevant OI – OII trajectories.
⬍ 10−10 s for Ca2+ and 1 – 6 s for Al3+兲. Consistent with the
residence time of seconds for the first shell water molecules
ture. The exchange rate for the 400 K simulation appeared to
of Al3+, no transfers between the first and second shells have be at least twice as great as the 300 K simulations.
been observed in our simulations. Considerably less is Typically, ligand exchange mechanisms are classified ac-
known about the lifetimes of water molecules in the well cording to their associative 共SN1 reaction in which the ex-
structured second shell of triply charged ions. For the ex- change is initiated by a particle entering the shell with the
change between the second and third shells, Bleuzen et al. subsequent dissociation of a second ligand兲 or their dissocia-
report a lifetime of 128 ps measured from 17O NMR for the tive mechanism 共SN1 reaction in which the exchange is ini-
Cr3+ aqua ion.2 The lifetime is expected to be considerable tiated by a particle leaving the shell with the subsequent
shorter for Al3+, since the first shell lifetime for Al3+ is five association of a second ligand兲.32 All the exchanges between
orders of magnitude shorter than that for Cr3+. Our simula- the second and third shells observed in our simulations oc-
tions are consistent with this rationalization, as several trans- curred by a dissociative mechanism, in which water mol-
fers on the picosecond time scale are observed between the ecule initially leaves the second shell, followed by a water
second and third shell water molecules in the 64 water and molecule entering the second shell. This mechanism results
128 water simulations. in intermediate solvent structure of 11 water molecules in the
Six exchanges between the second and third shells were second shell rather than 12 water molecules. We note that
seen in the 17.3 ps 400 K Al3+ + 64H2O simulation, one ex- this mechanism agrees with the positive activation volumes
change was seen in the 300 K Al3+ + 128H2O 共7.1 ps兲, and observed in recent NMR experiments.2
1.5 共two entering and one leaving the second shell兲 ex- However, the exact structure of the intermediate appears
changes were seen in the 12.5 ps 300 K Al3+ + 64H2O simu- to depend on the size of the simulation. As shown in Fig. 7,
lation. Clearly there were not enough exchanges to estimate the only intermediate structures seen in the Al3+ + 128H2O
an accurate exchange rate. However, our simulations showed simulation display a rather simple structure in which the out-
that the exchange rate depends significantly on the tempera- going and incoming water molecules are contained in the

Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
104505-7 Hydration shells of Al3+
FIG. 8. 共Color兲 Ligand exchange mechanism between the second and third
shells seen in 400 K Al3+ + 64H2O simulation, 共a兲 intermediate structure,
and 共b兲 relevant OI – OII trajectories.
third shell. Contrasting this, a more complicated interchange
is observed for the Al3+ + 64H2O simulations 共see Fig. 8兲. In
these simulations, the system first evolves into a state in
which a water molecule leaves the second shell and the re-
sidual undercoordinated first shell water molecule then forms
a hydrogen bond with another second shell water molecule.
This results in an intermediate in which a single water mol-
ecule in the inner part of the second shell is hydrogen bonded
to two different first shell water molecules, as shown in Fig.
8. The system then evolves by simultaneously shifting the
shared water molecule onto just one first shell water mol-
ecule and having an entering water molecule hydrogen bond
to the other first shell water molecule.
IV. CONCLUSION
Three different AIMD simulations of Al3+ ion in aque-
ous solutions using up to 128 water molecules have been
reported. The simulations were carried out at temperatures
near 300 and 400 K in periodic supercells, comprised of ei-
ther Al3+ + 64H2O or Al3+ + 128H2O, that were chosen to pro-
vide a density near 1. These first principles calculations were
extremely demanding due to large number of water mol-
ecules and long simulation times required to fully describe
Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
J. Chem. Phys. 126, 104505 共2007兲
the chemical events in the extended hydration shells. All
three simulations gave very similar results for the structural
parameters and agreed quite well with available experimental
data.
The simulations predicted a stable octahedral structure
for first hydration shell containing six water molecules. The
second hydration shell was found to contain on average 12
water molecules trigonally coordinated via hydrogen bonds
to six first shell water molecules. This structure is consistent
with experimental estimates and it agrees quite well with the
putative structure used to analyze the x-ray data. However, it
disagrees with the results reported from conventional mo-
lecular dynamics using two- and three-body potentials. Bond
lengths and angles of the water molecules in the first and
second hydration shells and the average radii of these shells
were also found to agree very well with the results of the
x-ray analysis. Beyond the second shell, the bonding pattern
was found to substantially return to the tetrahedral structure
of bulk water in all the simulations. Water transfers into and
out of the second solvation shell were also observed. These
exchanges were found to occur on a picosecond time scale
via a dissociative mechanism, in which a water molecule
initially leaves the second shell subsequently followed by a
water molecule entering the second shell.
Limited simulations with 128 water molecules 共7.1 ps兲
were also carried out. On the whole, very good agreement
was found between the simulations using 64 water molecules
and 128 water molecules. The general structures of the first
and second solvation regions were found to be essentially the
same. However, some subtle differences were seen between
the 64 water and 128 water simulations beyond the second
shell. To reliably predict behavior in these regions, sizes
larger than 64 may be required. The most significant differ-
ence found was that the second shell showed a sharper cutoff
in 128 water simulation 关deeper minimum in gAl–O共r兲 at the
end of the second shell兴 compared to the 64 water simula-
tions. Another significant difference found was that the struc-
ture of the intermediate in the dissociative transfer between
the second and third shells was found to depend on the size
of the simulation. For the 128 water simulation, the interme-
diate has a rather simple structure in which the outgoing and
incoming water molecules are contained in the third shell. In
contrast, for the 64 water simulations, a more complicated
interchange was found which resulted in an intermediate in
which a single water molecule in the inner part of the second
shell was hydrogen bonded to two different first shell water
molecules.
ACKNOWLEDGMENTS
This research was supported by the BES Geosciences
program of the U.S. Department of Energy, Office of Science
DE-AC06-76RLO 1830, and the NSF Grant No. NSF-EAR-
0545811. The Pacific Northwest National Laboratory is op-
erated by Battelle Memorial Institute. Some of the calcula-
tions were performed on the MPP2 computing system at the
Molecular Science Computing Facility in the William R.
Wiley Environmental Molecular Sciences Laboratory
共EMSL兲 at PNNL. EMSL operations are supported by the
104505-8 Bylaska et al.
DOE’s Office of Biological and Environmental Research.
The authors also wish to thank the Scientific Computing
Staff, Office of Energy Research, and the U.S. Department of
Energy for a grant of computer time at the National Energy
Research Scientific Computing Center 共Berkeley, CA兲.
1
H. Ohtaki and T. Radnai, Chem. Rev. 共Washington, D.C.兲 93, 1157
共1993兲.
2 18
D. T. Richens, The Chemistry of Aqua Ions: Synthesis, Structure, and
Reactivity: A Tour Through the Periodic Table of the Elements 共Wiley,
Chichester, 1997兲.
3
J. A. White, E. Schwegler, G. Galli, and F. Gygi, J. Chem. Phys. 113,
4668 共2000兲; A. P. Lyubartsev, K. Laasonen, and A. Laaksonen, ibid.
114, 3120 共2001兲.
4
W. Bol and T. Welzen, Chem. Phys. Lett. 49, 189 共1977兲.
5
R. Caminiti, G. Licheri, G. Piccaluga, and G. Pinna, J. Chem. Phys. 69,
1 共1978兲.
6
R. Caminiti, G. Licheri, G. Piccaluga, G. Pinna, and T. Radnai, J. Chem.
Phys. 71, 2473 共1979兲.
7
G. W. Neilson, S. Ansell, and J. Wilson, Z. Naturforsch., A: Phys. Sci.
50, 247 共1995兲.
8
G. J. Herdman and P. S. Salmon, J. Am. Chem. Soc. 113, 2930 共1991兲; P.
S. Salmon, G. J. Herdman, J. Lindgren, M. C. Read, and M. Sandstrom,
J. Phys.: Condens. Matter 1, 3459 共1989兲.
9 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865
R. A. Mayanovic, S. Jayanetti, A. J. Anderson, W. A. Bassett, and I. M.
Chou, J. Phys. Chem. A 106, 6591 共2002兲; R. A. Mayanovic, S. Jay-
anetti, A. J. Anderson, W. A. Bassett, and I. M. Chou, J. Chem. Phys.
118, 719 共2003兲; C. Cossy, A. C. Barnes, J. E. Enderby, and A. E.
Merbach, ibid. 90, 3254 共1989兲.
10
T. Ikeda, M. Hirata, and T. Kimura, J. Chem. Phys. 119, 12386 共2003兲.
11
J. O. M. Bockris and A. K. N. Reddy, Modern Electrochemistry, 2nd ed.
共Plenum, New York, 1998兲.
12
R. Car and M. Parrinello, Phys. Rev. Lett. 55, 2471 共1985兲.
13
M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias, and J. D. Joannopo-
ulos, Rev. Mod. Phys. 64, 1045 共1992兲; M. Valiev, E. J. Bylaska, A.
Gramada, and J. H. Weare, in Reviews In Modern Quantum Chemistry: A
Celebration Of The Contributions Of R. G. Parr, edited by K. D. Sen
共World Scientific, Singapore, 2002兲.
Downloaded 01 Jun 2007 to 218.94.142.5. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
View publication stats
J. Chem. Phys. 126, 104505 共2007兲
14
D. Marx and J. Hutter, in Modern Methods and Algorithms of Quantum
Chemistry, edited by J. Grotendorst 共Forschungszentrum, Julich, Ger-
many, 2000兲, Vol. 1, p. 301.
15
F. C. Lightstone, E. Schwegler, M. Allesch, F. Gygi, and G. Galli,
ChemPhysChem 6, 1745 共2005兲; F. C. Lightstone, E. Schwegler, R. Q.
Hood, F. Gygi, and G. Galli, Chem. Phys. Lett. 343, 549 共2001兲.
16
E. Wasserman, J. R. Rustad, and S. S. Xantheas, J. Chem. Phys. 106,
9769 共1997兲.
17
D. Spangberg and K. Hermansson, J. Chem. Phys. 120, 4829 共2004兲.
E. J. Bylaska, M. Valiev, R. Kawai, and J. H. Weare, Comput. Phys.
Commun. 143, 11 共2002兲.
19
M. I. Lubin, E. J. Bylaska, and J. H. Weare, Chem. Phys. Lett. 322, 447
共2000兲; A. J. Sillanpaa, J. T. Paivarinta, M. J. Hotokka, J. B. Rosenholm,
and K. E. Laasonen, J. Phys. Chem. A 105, 10111 共2001兲; J. D. Kubicki,
ibid. 105, 8756 共2001兲.
20
T. Ikeda, M. Hirata, and T. Kimura, J. Phys. Chem. 124, 074503 共2006兲;
T. W. Swaddle, J. Rosenqvist, P. Yu, E. Bylaska, B. L. Philiips, and W. H.
Casey, Science 308, 1450 共2005兲; S. Amira, D. Spangberg, and K. Her-
mansson, J. Chem. Phys. 124, 104501 共2006兲.
21
P. Hohenberg and W. Kohn, Phys. Rev. B 136, 864 共1964兲; W. Kohn and
L. J. Sham, Phys. Rev. A 140, 1133 共1965兲.
22
E. J. Bylaska, W. A. de Jong, K. Kowalski et al., NWCHEM, a computa-
tional chemistry package for parallel computers, Version 5.0, Pacific
Northwest National Laboratory, Richland, WA 99352-0999, 2006.
23
共1996兲.
24
D. R. Hamann, M. Schluter, and C. Chiang, Phys. Rev. Lett. 43, 1494
共1979兲; D. R. Hamann, Phys. Rev. B 40, 2980 共1989兲.
25
L. Kleinman and D. M. Bylander, Phys. Rev. Lett. 48, 1425 共1982兲.
26
Y. Marcus, Chem. Rev. 共Washington, D.C.兲 88, 1475 共1988兲.
27
R. Caminiti and T. Radnai, Z. Naturforsch. A 35, 1368 共1980兲.
28
R. D. Broadbent, G. W. Neilson, and M. Sandstrom, J. Phys.: Condens.
Matter 4, 639 共1992兲.
29
G. W. Neilson and J. E. Enderby, J. Phys. C 11, L625 共1978兲.
30
M. C. Read and M. Sandstrom, Acta Chem. Scand. 46, 1177 共1992兲.
31
A. Lauenstein, K. Hermansson, J. Lindgren, M. Probst, and P. A. Bopp,
Int. J. Quantum Chem. 80, 892 共2000兲.
32
F. P. Rotzinger, Chem. Rev. 共Washington, D.C.兲 105, 2003 共2005兲.