Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Article

Cite This: J. Chem. Theory Comput. 2019, 15, 1278−1292 pubs.acs.org/JCTC

Molecular Dynamics Simulation of Polarizable Gold Nanoparticles


Interacting with Sodium Citrate
Olga A. Perfilieva,† Dmitrii V. Pyshnyi,*,†,‡ and Alexander A. Lomzov*,†,‡

Institute of Chemical Biology and Fundamental Medicine, SB RAS, 8 Lavrentiev Avenue, Novosibirsk 630090, Russia

Novosibirsk State University, 2 Pirogova Street, Novosibirsk 630090, Russia
*
S Supporting Information

ABSTRACT: To study the structure of a citrate-capped gold


Downloaded via CENTO BRASIlEIRO PESSQUISAS FISICAS on March 2, 2020 at 13:54:38 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

nanoparticle and forces involved in citrate−gold interactions,


we performed a molecular dynamics simulation of a truncated-
octahedron nanoparticle containing Au(111) and Au(100)
surfaces with sodium citrate. In this paper, we employed an
approach to the modeling of interactions of a gold
nanoparticle with citrate molecules taking into account the
image charge effect in the metal. First, we built models of 6
and 14 nm nanoparticles, which can reproduce the polar-
ization effects, based on a rigid-rod gold model and the GolP-
CHARMM force field. To verify the simulation results, we analyzed density plots, radial distributions, distributions
perpendicular to Au(111) and Au(100) surfaces, the electric potential of the system, and the dynamics of citrate crown
formation. We observed formation of a stable citrate crown around the nanoparticle and detected nonuniform surface
distribution of citrate ions with the preference for Au(111) facets over Au(100) ones. Testing of the model of the citrate-capped
gold nanoparticle in a simulation at high concentrations of Na+ and Cl− ions (0.8 M) showed incorporation of chloride anions
into the citrate crown. We compared the results of citrate crown formation between polarizable and nonpolarizable gold models
and noticed a difference in the citrate distribution on the surface of the gold nanoparticle. We found that polarization effects in
the metal are involved in the mechanism of interaction of the gold nanoparticle and citrate ions. The obtained results are in
good agreement with experimental data and computer simulations from a number of other studies, which prove the validity of
the proposed model.

■ INTRODUCTION
The structure and properties of the gold nanoparticle coated
enable a detailed study of a system at the atomic level, but the
computational cost for such methods is high.22,23
The classical MD approach involving different force fields
with a citrate shell (also called a crown) or other ligands have
makes it possible to investigate large molecular assemblies
been actively studied by different groups of researchers.1−9
(millions of atoms) for long periods (up to microseconds).23
Among the various procedures used in these studies, in silico This type of system consideration and representation is
research methods play an important role.10−16 In particular, suitable for studies on large nanoparticles (larger than 10
such methods deal with the system at the nanoscale level and nm), which can serve as a substrate platform for polymer
represent a useful tool for identifying the mechanisms of molecules such as DNA or peptides.
interaction of a citrate molecule with the surface of a gold To analyze the interaction of a gold surface with other
nanoparticle and for determining the structure of the citrate molecules using the MD approach, a number of special force
crown around the nanoparticle. This type of simulation fields have been developed. CHARMM-METAL24 is an
requires a special approach to parametrization of the example of a force field that does not take into account the
interactions in an ensemble of molecules. polarization effect and is suitable for the simulation of peptide
Methods of quantum chemical calculations are employed to adsorption on gold. For uncharged molecules with a small
study the interaction of a surface with surrounding molecules dipole moment, the contribution of the polarization effect to
in small systems (up to several hundred atoms). For example, the adsorption energy is often negligible, whereas for charged
ab initio Car−Parrinello molecular dynamics (MD) simulation molecules, such as citrate, this effect should be taken into
of an aqueous Au(111) surface has revealed orientation of the account.25−27
water molecule on the atop site of gold atoms.17 Density Polarization in a metal can be examined in different ways, for
functional theory (DFT) calculations of adsorbate binding example, classically by introducing an imaginary plane behind
energy and of interaction geometry in vacuum are often carried
out for parametrization of force fields, especially in the absence Received: April 16, 2018
of experimental data.18−21 Quantum mechanical calculations Published: December 21, 2018

© 2018 American Chemical Society 1278 DOI: 10.1021/acs.jctc.8b00362


J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

which a charge of the opposite sign is located.27 The Citrate has also been modeled with a 3 nm gold nanoparticle
capacitance−polarizability interaction model and force field having (111), (100), and (110) facets in the ReaxFF39 force
can serve as an example of another more complex approach field, which takes into account events of chemical bond
that was successfully applied to describe interactions of an formation and breakage.40 It was shown that citrate is highly
icosahedral gold nanoparticle (consisting of 309 atoms) and reactive with the gold surface and can form strong complexes
water.28 with gold adatoms by capturing them from the surface of the
Recently, polarization effects have been introduced into the nanoparticle. It was concluded that a fully deprotonated form
gold model using elements of a Drude oscillator.29 In this of citrate is preferable and important for the interaction with
model, a dummy electron is added to a positively charged core the gold surface.
of a gold atom forming a dipole with two opposite charges The starting point of the present study includes building of a
connected by a harmonic spring. Changing of bond lengths in gold nanoparticle with polarization effects because electrostatic
the oscillator reproduces the image charge effect in the metal. interaction of the charged citrate molecule with the metal may
Each atom of a gold dipole also has parameters for a van der affect the adsorption process and energy. Another key
Waals interaction. The model can be implemented with parameter is the nanoparticle size suitable for a comparison
standard force fields for simulation of biological molecules with a typical nanoparticle used in experiments (usually ∼5−
without additional parametrization. 20 nm in diameter [d]). Therefore, we consider a simple way
In the rigid-rod model, gold atoms are represented by to simulate polarization in the gold model; this approach does
dipoles with a fixed bond length between charges.30,31 Dipoles not dramatically change the standard MD formalization of
can freely rotate around the mass center of a real atom, thus force fields.
simulating polarization effects in the metal. The early force Thus, the main purpose of our research was to study the
field used with the rigid-rod model was GolP,30 which is interactions of a polarizable model of a gold nanoparticle with
compatible with the OPLS-AA32 force field, developed to sodium citrate by the classical method of MD simulation. We
model interactions of a protein with a gold surface. chose the approach proposed for formalization of the image
charge effect in gold via the rigid-rod model.30,31
Nonetheless, analysis of water molecules in this field predicts
The main steps of our study are listed below:
an incorrect adsorption geometry, with the hydrogen atom
directed toward the surface; this arrangement did not match • gold nanoparticle morphology that can be realized in a
the data of quantum mechanical calculations.17 Then, GolP- simulation force field (GolP-CHARMM) was determi-
CHARMM20,21 compatible with the CHARMM33 force field ned;
for Au(111) and Au(100) surfaces was developed and took • a rigid-rod model of a gold nanoparticle with virtual sites
into account the adsorption geometry of water and of small for Au(111) and Au(100) surfaces was built;
organic molecules. Application of the GolP-CHARMM field is • the obtained gold nanoparticle in an aqueous medium
limited to these two types of surfaces because the gold model with sodium citrate was simulated, and the results were
includes virtual sites with a specific Lennard−Jones interaction analyzed.


for each type of the surface. The rigid-rod gold model has been
successfully utilized to construct Au(111) and Au(100) slabs METHODS
and to study the interactions of interface gold atoms with small
MD Simulation. The MD simulations were carried out
organic molecules and proteins,34−36 but nanoparticle models
using Gromacs41 package version 5 in the GolP-CHARMM
of this type with Au(111) and Au(100) surfaces have not yet
force field. The nanoparticle was placed in the center of the
been devised.
cuboid simulation cell with the size of 13.0 × 13.0 × 13.0 nm3
During studies on interactions of some molecules with gold
for a 6 nm nanoparticle and 22.84 × 22.84 × 22.84 nm3 for a
nanoparticles, it is necessary to take into consideration that 14 nm nanoparticle. Citrate anions (fully deprotonated
under experimentally observed conditions, gold nanoparticles C6H5O73−) were added into the system in the amount of
are not naked but covered by various ligands to prevent 451 and 2756 molecules (0.34 and 0.39 M) for nanoparticles
aggregation. In the Turkevich method37 widely used for the with d of 6 and 14 nm, respectively. These values were chosen
synthesis of gold nanoparticles, sodium citrate acts as a to promote formation of a local concentrated layer around the
stabilizing agent for gold colloids. nanoparticles that is sufficient for assembly into a crown
Investigation of citrate interaction with the gold surface is an structure (with 84% of the surface coverage for d = 6 nm and
important field of research on nanoparticle properties. 100% for d = 14 nm). The percentage of coverage and the
Recently, a citrate model with bonded force field parameters number of citrate molecules were calculated according to data
adapted for a simulation in water was developed.38 The same reported by Park et al. (the density of citrate ligands was 2.8 ×
group of authors also developed the GolP-CHARMM force 10−10 mol/cm2, which was 45% of the surface coverage).1 For
field, which contains the parameters for a nonbonded the neutralization charge of the simulation cell, sodium ions
interaction of citrate with Au(111) and Au(100) surfaces.20,21 were added into the system in an amount of three Na+ cations
The interaction of sodium citrate with the Au(111) surface in per citrate molecule. The simulations were performed in
an explicit water shell has been simulated by means of the explicit water in the NVT ensemble with three-dimensional
GolP-CHARMM force field.36 As a result, an amorphous, (3D)-periodic boundary conditions by means of a Nosé
multilayer organization of citrate overlayers with common Hoover thermostat.42,43 The cutoff Verlet scheme44 was used
“stripe” and “islands” motifs was observed, as were more with particle mesh Ewald (PME)45 electrostatic summation
complex structures such as loops and straight and curved and was truncated at 11 Å for Lennard−Jones and Coulomb
chains. It was found that Na+ ions play an important role in the interactions.
structure of overlayers and participate in the formation of During the simulations, all real atoms and virtual sites were
citrate bilayers on the surface of gold. fixed in space, and only dipole particles were free to rotate.
1279 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Citrate was modeled by bonded38 and nonbonded20,21 GolP- Because modeling by the method of MD was carried out
CHARMM force field parameters. A modified TIP3P46,47 under periodic conditions, it was necessary to take into
water model and nonbonded parameters for the Na+ ion were account the influence of atoms from neighboring cells during
also taken from the GolP-CHARMM force field. calculation of the potential. This is especially important for the
The system was first equilibrated by slow heating from 0 to atoms located at corners of the cell. Under 3D-periodic
300 K with a reduced time step of 0.5 fs for 1 ns. A productive conditions, each simulation cell is surrounded by its 26 copies.
simulation was conducted via at least 10 cycles of annealing For a large system, the inclusion of all atoms from the cell in
with a time step of 1 fs in a manner similar to that mentioned question and from neighboring cells during calculation of the
by other authors36 testing the same model of gold. Such a total field potential at each grid node requires large
simulation method speeds up the process of cell equilibration. computational resources. Because the potential from a given
The temperature of gold was maintained at 300 K, while the charge decreases with distance as 1/r, at a given node of the
rest of the system including water, citrate ions, and Na+ ions grid, the nearest charges from a neighboring cell make a greater
followed a temperature cycle (300, 400, 500, 600, 500, 400, contribution to the intensity than do distant charges from the
and 300 K) of 700 ps with time checkpoints (0, 200, 225, 425, same cell. To optimize the calculation process, only those
625, 650, and 700 ps). charges that fall into a cube with a side equaling to half of the
An additional simulation was performed with sodium citrate cell size and centered at the given node are taken into account
and a nonpolarizable gold nanoparticle model (d = 6 nm), during estimation of the total electric field potential of a node;
which consisted only of real atoms with van der Waals this strategy is consistent with the Coulomb cutoff of 11 Å (see
interactions. The Lennard−Jones parameters for gold were the section MD Simulation).
borrowed from the CHARMM-METAL24 force field. The rest The obtained values of the electric field potential at the grid
of the conditions in the simulation, e.g., the temperature nodes were visualized by means of the points randomly
regime, the citrate model, and the amount of citrate in the distributed near a given node and not overlapping with
system, were the same as those for the polarizable gold model neighboring nodes. Intensity of the potential was determined
as described above. via the concentration of points by means of a filter system.
For simulation of the interaction of a citrate-capped Filters were the values from the absolute values of the intensity,
nanoparticle with Na+ and Cl− ions, the results of crown which determine the number of points to be placed near the
modeling for the polarizable nanoparticle with d = 6 nm were node. The system of filters forms a profile of the electric
employed. A snapshot of the MD trajectory obtained after a 7 potential intensity.
ns simulation without a chloride ion served as an initial The results of calculating the 3D potential distribution were
structure for modeling. Na+ and Cl− were added to a transformed into .gro format by representing the points of the
concentration of 0.8 M each (1323 Na+ ions and the same grid as oxygen and nitrogen atoms. The sign of the potential is
indicated by different colors of the nitrogen and oxygen atoms.
number of Cl− ions) into the enlarged 14.0 × 14.0 × 14.0 nm
When viewed in the VMD software,49 the red color
simulation cell. The simulation protocol was the same as that
corresponds to the positive potential, and the blue color
of the modeling system without the salt. Nonbonded
corresponds to the negative potential. This way of
interaction parameters of chloride ions with gold were
representation enables readers to visualize a given ensemble
calculated according to Lorentz−Berhelot mixing rules48
σi + σj of molecules together with the 3D potential in the VMD
(σij = 2 , εij = (εi*εj) ) based on GolP-CHARMM force software.
field parameters of Au−Au interaction. The parameters Computation of the Distribution of Citrate Ions
obtained for Cl− were as follows: σ = 0.362 nm and ε = Perpendicular to the Surfaces of the Nanoparticle
0.9033 kJ/mol for the Au(111) surface and σ = 0.372 nm and and Geometrical Radial Distribution. To determine the
ε = 1.3256 kJ/mol for Au(100). distribution of molecules around a nanoparticle, we considered
Calculation and Visualization of the Electrostatic the position of their mass centers in the simulation cell.
Potential in the System. The electric field potential was For distribution perpendicular to the surfaces, we inves-
visualized as a 3D grid with intensity of the potential at nodes tigated the space above Au(111) and Au(100) surfaces (see
placed inside the analyzed ensemble. The magnitude of the Figure S1). The number of molecules N (centers of mass)
electrostatic potential was revealed by the concentration of located in the current volume was determined by going from
points corresponding to certain intensity. the surface of the nanoparticle along the normal to the plane
The obtained result includes a method for taking into with specified distance step dr. The concentration of molecules
consideration the periodicity of a given ensemble, a method for at a specific distance c(r) = N(r)/V was obtained by
optimizing the calculation and visualization of the potential normalization of the number of molecules N(r) to volume V
intensity, and a procedure for presenting the results in.gro = dr × S, where S is an area of the surface under consideration
format. [Au(111) or Au(100)]. To calculate the graphs of dynamics
The electric potential at each grid point i (Vi) is calculated as (Figures S15−S18), each time point was averaged over the
temperature interval 300−340 K. The graphs of 0 and 28000
the sum of the potentials from atomic charges of all molecules
ps were calculated using seven and three frames of the
qj trajectory, respectively; for all other time points, 10 frames
Vi = ∑ were averaged.
j,j≠i
rij (1) A geometrical radial distribution was computed according to
the same principles as those for the distribution perpendicular
where qj is the charge of the jth atom, and rij denotes the to the surface. The concentration of molecules was determined
distance from the charge of the jth atom to the ith point of the via the formula c(r) = N(r)/V(r), where N(r) is the number of
grid in which the potential is being considered. centers of mass dropped onto the spherical layer with thickness
1280 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Figure 1. Unit cell of the face-centered cubic lattice (FCC) with additional virtual sites for {100} and {111} planes (four sites shown). Real atoms
of the lattice are represented by orange balls. Virtual sites are shown as red and blue stars for {100} and {111} planes, respectively. Examples of
{100} and {111} planes are yellow.

of distance step dr, in a thin spherical layer of volume V(r). shape is the most thermodynamically stable52,53,64 and retains
The starting point (zero value) for distance r was the center of its structure the best with increasing temperature.64
the nanoparticle. When the shape of a nanoparticle is being determined, the
Density map calculations were carried out by means of the ratio of {111} and {100} surfaces is an important parameter. It
AmberTools17 package.50 is known that a nanoparticle tends to increase the presence of a

■ RESULTS AND DISCUSSION


Choosing a Shape of the Gold Nanoparticle. To
lower-energy facet {111} for its stabilization.63 According to
thermodynamic reconstruction52,60 in vacuum for the 14 nm
nanoparticle, facets {111} should constitute 90.5% of the total
construct a gold nanoparticle for simulation, it is necessary to surface area. In contrast, with the increasing temperature and
determine common morphological features of the nanoparticle. taking into account some kinetic factors in the model,53,54 the
Depending on the experimental conditions for obtaining theory predicts an increasing percentage of the {100} surface
nanoparticles and their sizes, the dominant equilibrium shape area, followed by the transformation of a truncated octahedron
of the nanoparticle can change.51 By means of theoretical into a cuboctahedron. Via this approach and a phase diagram54
models, researchers attempt to predict and explain the for a gold nanoparticle with d of 14 nm at 300 K, the
structure of a nanoparticle under given conditions on the proportions of the {111} and {100} planes will be
basis of experimental data (e.g., direct images of a nanoparticle approximately 66% and 34%, respectively.
obtained by microscopic techniques).52 In our simulation, we also consider a smaller nanoparticle (6
The morphological diversity of nanoparticles can be nm) sharing the morphology of a truncated octahedron as a
subdivided into two types: nanoparticles having a single face- useful system for calculations involving moderate numbers of
centered cubic (FCC) crystal lattice and twinned nano- atoms, although nanoparticles of such diameters may be
particles.51,53 The simple FCC shapes formed by {111} and stabilized in an icosahedral shape.54
{100} crystal planes include a truncated octahedron, Building of the Gold Nanoparticle Model. To create a
cuboctahedron, cube, and truncated cube. Twinned nano- gold nanoparticle, we chose a model of gold that takes into
particles are polycrystalline and consist of single crystal units consideration the polarization effect. The model was developed
connected by a mirror plane. An icosahedron and decahedron for Au(111) and Au(100) surfaces but has not been previously
are typical examples of twinned nanoparticle shapes. A applied to construction of a gold nanoparticle. In the Gromacs
multitwinned icosahedral nanoparticle exposes 20 surfaces software, this gold model was used with GolP30,31 and GolP-
with an arrangement of atoms similar to the {111} crystal CHARMM force fields20,21 (based on the CHARMM force
plane. A decahedron possesses 5-fold symmetry with 10 field33) for proteins and with GolDNA-AMBER19 (based on
surfaces of the {111} type. Ino and Marks decahedral the AMBER force field65) for DNA and the Au(111) surface.
nanoparticles also include {100} facets. The special features of this model deal with representation
Various theoretical approaches are used to determine the
of gold atoms in the form of dipoles that can rotate around a
equilibrial dominant form, e.g., DFT calculations, thermody-
real center of an atom simulating the induced charges. This
namic, and kinetic approaches.52−62 From the thermodynamic
point of view, an equilibrium nanoparticle of a given diameter group of atoms is responsible for electrostatic interactions. In
has such a ratio of {111}, {110}, and {100} surfaces (with addition, there are virtual atoms on the surface that determine
different surface energies) that minimizes the total energy. For the van der Waals interaction of gold atoms with other
FCC, this shape corresponds to a truncated octahedron.55,60,63 molecules. Virtual sites are introduced into the model in such a
Nevertheless, if we consider the kinetic approach,53 the most way that they lead to a correct position of an adsorbed
rapidly formed and kinetically stable is the icosahedral shape. molecule relative to a gold atom (atop, bridge, and hollow
Due to the combined effects of thermodynamic and kinetic sites). All these features greatly complicate the model because
influences, the shape of the decahedron is stabilized at greater instead of one type of gold atom, there are several types: real
diameters of nanoparticles. As for the size 14 nm, which we will gold atoms, virtual dipoles associated with them, and virtual
consider in our simulation, it can be concluded54 that such sites of (111) and (100) surfaces with their own van der Waals
nanoparticles are in a transition region from the decahedral parameters.
shape to FCC shapes. The choice of truncated-octahedron The process of setting up a model is represented by the
morphology can be made on the basis of the findings that this following steps:
1281 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Table 1. Positions of (111) Virtual Sites in the Unit Cella


(111) (111̅) (1̅1̅1) (1̅1̅1̅) (11̅1) (11̅1̅) (1̅11) (1̅11̅)
ij 1 1 4 yz ij 1 1 2 yz ij 5 5 4 yz ij 5 5 2 yz ij 1 5 4 yz ij 1 5 2 yz ij 5 1 4 yz ij 5 1 2 yz
jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz
k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{
jij 2 ; 2 ; 2 zyz jij 2 ; 2 ; 4 zyz jij 4 ; 4 ; 2 zyz jij 4 ; 4 ; 4 zyz jij 2 ; 4 ; 2 zyz jij 2 ; 4 ; 4 zyz jij 4 ; 2 ; 2 zyz jij 4 ; 2 ; 4 zyz
j z j z j z j z j z j z j z j z
k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{
ij 1 4 1 yz ij 1 4 5 yz ij 5 2 1 yz ij 5 2 5 yz ij 1 2 1 yz ij 1 2 5 yz ij 5 4 1 yz ij 5 4 5 yz
jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz
k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{
ij 4 1 1 yz ij 4 1 5 yz ij 2 5 1 yz ij 2 5 5 yz ij 4 5 1 yz ij 4 5 5 yz ij 2 1 1 yz ij 2 1 5 yz
jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz jj ; ; zz
k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{ k6 6 6{
a
The header shows different types of {111} crystallographic planes. Coordinates of virtual sites in a column belong to the corresponding plane
(unit cell size is 1 × 1 × 1).

• construction and multiplication of the unit cell (with real array can be considered a 3D grid of an unknown geometric
atoms and virtual sites) to form a cubic crystal of the scale size with real or virtual atoms that can be located in its
required size; node. On the basis of visual representation of a unit cell, each
• obtaining a nanoparticle from a crystal by cutting off atom is exposed in the array. This method of data organization
excess atoms with eight planes (111) and then six planes is effective and visually representative for solving the problem
(100) and determination of surfaces Au(111) and of nanoparticle construction. At a later stage, the array can be
Au(100) and types of gold atoms for surfaces; transformed into a geometric structure according to the lattice
• construction of the final model of the nanoparticle by parameters of a gold crystal unit cell.
specifying it in 3D geometrical space, addition of virtual Obtaining a Nanoparticle from a Cubic Crystal and
atomic dipoles to real gold atoms, and preparation of Determining the Surfaces. The next step was to cut off
files for MD simulations. redundant atoms in the cube with planes to produce the
Construction of the Unit Cell and Its Multiplication. The truncated-octahedron geometry. For this purpose, we analyzed
cubic crystal from which the nanoparticle will be derived can an atom’s position relative to the planes of the nanoparticle
be considered a set of points in 3D space with a definite surface. The equations for all 14 planes (eight planes of the
position. This set of points can be described by a 3D array of {111} type and six planes of the {100} type) for the truncated
(nx + 1) × (ny + 1) × (nz + 1) size, where nx, ny, and nz are octahedron should be defined (see Figure 2 and Supporting
numbers of unit cell blocks along the x-, y-, and z-axes. Information, section “Symmetric equation for {111} and {100}
An elementary cell can also be initiated as a 3D array, which planes”). The diameter d and a ratio of total (111) and (100)
in turn can be considered geometrically analogous to an surface areas (SΣ{111}:SΣ{100}) can be used to identify the
elementary FCC unit cell. Virtual sites for all the {100} and positions of the planes when building a nanoparticle model.
{111} planes were added to the real atoms of the unit cell For cutting off excess atoms, the cube was assumed to be
(Figure 1). Virtual sites for {100} surfaces were set in the placed in a Cartesian coordinate system with the origin in the
middle of the cube edges. For the {111} surface, 4 × 8 virtual
site points were defined as centers of triangles on the
corresponding plane. If we have a unit cell with the side
equal to 1, then the coordinates of virtual atoms for {100} are
as follows: (0.5; 0; 0), (0; 0.5; 0), (1; 0.5; 0), (0.5; 1; 0), (1; 0;
0.5), (1; 1; 0.5), (0; 0; 0.5), (0; 1; 0.5), (0.5; 0; 1), (0; 0.5; 1),
(1; 0.5; 1), and (0.5; 1; 1). Taking into account the
denominator of every virtual site (111) and (100) inside the
unit cell, it is reasonable to represent each elementary cell in
the form of a 6 × 6 × 6 3D-array (du = 6) in accordance with
the position of virtual sites (see Table 1).
According to the proposed structural organization, the unit
cell can be easily extended to a cuboid crystal of any size. By
multiplying a unit cell, a researcher can create a large cubic
crystal (nx = ny = nz = n) of (du*n + 1)3 size in the form of a
3D array. Instead of a continuous list of atoms with
coordinates, the array can be used to describe the crystal
structure. The array contains only numbers that denote the
types of atoms. To label a real atom in the array, the number Figure 2. Geometry of the nanoparticle obtained from the cube
“2” was employed, and virtual atoms for {100} and {111} crystal (thin black lines). Cutting planes {111} (thick black lines)
form an octahedron in the center of the cube. The truncated
planes were represented by numbers “5” and “7”, respectively. octahedron is obtained by removing excess atoms by means of {100}
The array initialization consisted of filling it out with zeros planes (pink lines). Changing the cutting position of {100} planes
(“0”), which means that there was no atom in any unit cell relative to the center of the cube results in a different ratio of (111)
node. If we denote the array as [a], then each element of the and (100) surface areas and can lead to different shapes of the
array is designated as a(i,j,k). The maximum number of nanoparticle (an octahedron, truncated octahedron, cuboctahedron,
elements in each dimension of the array is (6 × n) + 1. The or cube).

1282 DOI: 10.1021/acs.jctc.8b00362


J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

center of the cube. Each element of the array a(i,j,k) was Nonetheless, array [a] did not contain dipoles. Dipoles were
centered by shifting all coordinates by half of the cube side (3 added to the model during determination of the Cartesian
× n) to use a symmetric equation for {111} and {100} planes. coordinates of atoms. Atoms AUC with the charge 0.3e were
An octahedron was obtained from the cube by changing the added to each real atom at the distance of 0.07 nm along the x-
array of elements according to their positions relative to eight axis to build a gold dipole in accordance with the rigid-rod
planes {111}. For the selected plane and each element of array model of gold.20,30,31 The rigid-rod dipole with the AUC atom
[a], the status of the point in the cube was redetermined: if was fixed in the topology of the system to take into account the
point a(i,j,k) was beyond plane {111} forming an octahedron, polarization effect during simulation. Topology of the system
then a(i,j,k) was redefined by zeros (“0”). Then the parameter also contained information about charges and masses of
of {100} equations (see Supporting Information, section different gold atom types.
“Symmetric equation for {111} and {100} planes”) determines Finally, output files were converted to such a format that
at what position the {100} planes are settled and enables enables using them in the simulation of the nanoparticle in the
precise cutting of excess atoms. Thus, varying the ratio of Gromacs package: a .gro file with coordinates; a file of the
(100) and (111) planes, one can obtain nanoparticles of molecule topology (.itp) with correct atom types, charges,
different shapes (an octahedron, truncated octahedron, masses, and dipole constraints; and an index file (.ndx) in
cuboctahedron, truncated cube, or cube). which some atoms that had to be fixed during the simulation
The next step was to detect the surface in the nanoparticle were listed.
model. A surface of the nanoparticle was considered for each As a result, we built the nanoparticle with surfaces (111) and
plane and nonzero element of array [a]. Elements in an array (100) for which the surface atoms were defined: virtual sites
that belong to cutting planes also lie on the surface of the responsible for short van der Waals interactions [white for
nanoparticle. We define these elements as surface atoms and (111) and red for (100)] and the real atoms with dipoles [gold
change the status of an element in array [a] according to the for (111) and blue for (100)] responsible for electrostatic
plane type: {111} or {100}. For example, instead of values “2”, interactions. Inside the nanoparticle there were only real atoms
“5”, and “7” in the array, we inserted others: “3” and “37” for with dipoles. Figure 3 depicts the typical structure of the
the real and virtual atom of surface (111), respectively, and “4” truncated-octahedron gold nanoparticle.
and “47” for the real and virtual surface atoms of (100). The
usage of different numbers for different surfaces enabled us to
give atoms of the surface their own interaction parameter if
necessary.
After that, array [a] containing zeros where there are no
nanoparticle atoms and containing numbers where the atoms
are present was subjected to the following modification:
internal virtual atoms were deleted from the array (in our case
they are denoted by the values “5” and “7”) by setting them to
“0”.
Making of a Geometrical Space Model with a Unique
Topology Taking into Account Polarization Effects. The
information from 3D array [a] (containing representation of
nanoparticles in the form of zeros and numbers in space
coordinates (i,j,k) of array [a]) had to be transformed into 3D
geometric coordinates (x,y,z) and into certain types of gold Figure 3. Illustration of the typical structure of the truncated
atoms used in the rigid-rod gold model.20,30,31 octahedron nanoparticle model (d = 3 nm). Facets, edges, and
The real atoms and virtual sites of a (111) surface indicated vertices of the nanoparticle are shown. Colors represent different
types of gold atoms: white and red are (111) and (100) virtual sites,
by values “3” and “37” of the array become AUS and AUI atom whereas blue and gold are real atoms. Dipole atoms are not shown.
types of gold, respectively, in the model. For a (100) surface,
the real atoms “4” and virtual sites “47” should be redefined as
AUSS and AUII, respectively. All internal real atoms “2” were Two gold nanoparticle models with d = 6 and 14 nm and the
renamed to the AUB type. Each type of gold atom had its own surface ratio SΣ{111}:SΣ{100} of 66%:34% were constructed for
mass and charge according to the rigid-rod model, which is simulation. Because the truncated-octahedron nanoparticle is
useful for making system topology files.20,30,31 not spherical, we defined d of the nanoparticle as a distance
A fixed distance between elements of the array [a] 3D grid between two opposite (100) planes. For simulation stability,
was denoted as the scale factor (sf) variable the vertex atoms were removed (see Supporting Information,
section “Preparing the nanoparticle system for simulation”).
us Given that our attention was focused on binding to the
sf =
du (2) metal surface, and Au(111) and Au(100) facets occupy most
of the nanoparticle interface, we did not consider in detail van
where us = 0.40782 nm is a unit cell size parameter of the Au der Waals interactions at the edges and vertices. In GolP-
lattice,66 and du = 6 is the number of unit cell partitions (see CHARMM,20,21 such parameters for corners are absent.
Construction of the unit cell and its multiplication) Perhaps for a more careful analysis of the effects of edges
Using sf as the scale factor of the grid, we were able to and vertices, these interactions should be parametrized
calculate Cartesian coordinates of all real gold atoms and separately. In principle, the proposed algorithm of nanoparticle
virtual sites. If a(i,j,k) is an element of array [a], then x = i × sf, construction makes it possible to redefine the types of gold
y = j × sf, and z = k × sf. atoms for edges and give them special values if necessary.
1283 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Figure 4. Dynamics of radial distributions of citrate (A), Na+ ions (B), and water (C) for the nanoparticle of d = 6 nm. The 21 ns trajectory (with a
time step of 200 ps, and a radial distance step of 0.1 nm) was smoothed using a moving average of five time points. The position of the origin of the
coordinates is at the center of the nanoparticle.

Figure 5. Dynamics of radial distributions of citrate (A), Na+ ions (B), and water (C) for the nanoparticle of d = 14 nm. The 28 ns trajectory (with
a time step of 200 ps and a radial distance step of 0.1 nm) was smoothed using a moving average of five time points. The position of the origin of
the coordinates is at the center of the nanoparticle.

Figure 6. Final snapshot of citrate crown formation after 28 and 21 ns simulations of gold nanoparticles of d = 14 nm (A) and d = 6 nm (B),
respectively. Atoms are colored according to elements: red corresponds to oxygen, green to carbon, white to hydrogen, ochre to real gold atoms,
yellow to virtual gold sites, and blue to sodium ions. Water molecules and gold dipoles are not shown.

Simulation of Gold Nanoparticles with Sodium both 6 and 14 nm nanoparticles. This finding can be easily
Citrate. The simulation of gold nanoparticles with d = 6 illustrated by the analysis of radial geometrical distribution for
and 14 nm in an explicit water shell with sodium citrate was different molecules (Figures 4 and 5). The analysis of radial
performed for 21 and 28 ns, respectively. Formation of a stable distribution dynamics for citrate and sodium ions reveals the
sodium citrate coverage was observed during the simulation for formation of narrowing peaks around the nanoparticle. The
1284 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

maxima of peaks reach a plateau at the same time for citrate


and sodium ions; this situation corresponds to the formation of
a stable sodium citrate crown. It formed faster for the 6 nm
nanoparticle (after ∼10 ns) and took twice the time for the 14
nm nanoparticle (∼20 ns); these data are consistent with the
size of the analyzed systems. The maxima and standard
deviation of the citrate and sodium ions radial distributions
averaged over the last 3 ns of the simulation were the same: 4.3
± 0.45 nm and 8.6 ± 0.85 nm from the center of the small and
big nanoparticle, respectively (see also Supporting Informa-
tion, section “The estimation of citrate crown thickness”).
Dynamics of water molecules along the MD trajectory
indicated a replacement of water around the nanoparticle by
citrate and sodium ions (Figures 4C and 5C). The water radial
distribution curve for the small nanoparticle consists of a peak,
then a valley, followed by an increase to the plateau. The depth
of the valley correlated with the growth of ion distribution
peaks with time. A similar time trend was observed for the big
nanoparticle, but the first peak of water was absent due to the
extended surface of the nanoparticle on radial distribution
graphs (see Supporting Information section “The estimation of
citrate crown thickness”). The presence of water near the
surface of both nanoparticles was confirmed by density maps
(see Figure 8).
Snapshots of final frames for both nanoparticles are
presented in Figure 6. Almost all citrate ions were involved
in the formation of the shell around the nanoparticle at the end
of the simulation, thus indicating that the process of crown
formation ran successfully (100% of the initial citrate amount
was in the crown for d = 6 nm and 93% for d = 14 nm).
In general, we can conclude that not all the nanoparticle
surface is covered by citrate, thus resulting in heterogeneity of
the citrate crown.1,3 Citrate is organized as a multilayered shell Figure 7. A) An average normalized distribution of citrate mass
that is consistent with the experimentally estimated thickness centers perpendicular to the surface: for the nanoparticle with d = 6
of citrate coverage1 and a study on high-resolution microscopic nm (solid curves) and for the 14 nm nanoparticle (dashed curves).
images of nanoparticles.3 Blue and red curves denote Au(111) and Au(100) surfaces,
The overall distribution of citrate ions above the nano- respectively. The graph distance step is 0.1 nm, and the position of
zero is at the surface level. Typical structures of citrate overlayers
particle surface has a nontrivial structure (see Figures 7B−E
above the surface of Au (100) for d = 14 nm (B) and for d = 6 nm
and S3−S10 for more details). Nevertheless, in the (D) and Au(111) for d = 14 nm (C) and for d = 6 nm (E). Citrate is
composition of the crown, readers can see motifs of citrate represented by blue lines, and red balls indicate centers of mass of
corresponding to simulation data obtained by Wright et al. for citrate ions.
a polarizable Au(111) surface: linear and branched chains and
bilayers connected through Na+ ions.36 In the same paper,
those authors identified three adsorption states of citrate in the noticed a large so-called “third layer”, which accounted for an
amorphous overlayers: “directly adsorbed” citrate had close asymmetrical form of distribution. According to our data, the
van der Waals contacts with gold (up to 5 Å from the surface), third layer is diffuse and merges with the previous layer, in
“indirectly adsorbed” formed a bilayer through a sodium ion agreement with complex 3D structural organization of citrate
with adsorbed species (approximately 5−8 Å), and the “third ions (see Figures S3−S10). Averaged distributions above the
layer”. Lee et al. obtained a clear-cut image of citrate layers Au(100) surface reveal a significant difference from Au(111)
around a 10 nm nanoparticle by transmission electron because of a decreased number of adsorbed citrate ions and a
microscopy techniques and also concluded that the citrate more extended diffuse layer. The analysis of the visual model of
shell around the 10 nm nanoparticle consists of 2−3 layers citrate overlayers on the Au(100) surface revealed that in a
with the estimated distance between them equal to 3−3.5 Å.3 typical case, citrate ions hung over the middle part of the
The final result on citrate distribution curves perpendicular surface without binding to it. On average, citrates were closer
to the surface of the nanoparticle indicates a layered structure to the Au(100) surface of the 14 nm nanoparticle and had a
of the citrate crown too (see Figure 7A). The graph is averaged greater number of direct contacts with the surface area.
over all eight Au(111) and all six Au(100) facets. The Moreover, in the case of the 14 nm nanoparticle, some surfaces
distributions of citrate ions above Au(111) look similar Au(100) (namely 01̅0) were populated by citrate ions (see
between 6 and 14 nm nanoparticles. The first peak at a Figures S9 and S10).
distance of 0.5 nm from the surface can be attributed to the Higher magnitude of the distribution perpendicular to the
directly adsorbed citrate layer. The next high-magnitude point Au(111) surface as compared to the Au(100) surface (Figure
appears at 0.7 nm which probably corresponds to indirectly 7A) means a preference of the citrate shell for Au(111), in line
adsorbed citrates according to reference data.36 We also with the data reported by Park and coauthors.1 In that paper, a
1285 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

model of a protonated citrate form (H2Citrate−) was analyzed citrate crown and compared it with the values from radial
on Au(111), Au(100), and Au(110) surfaces to fit distribution graphs (see Supporting Information, section “The
experimental results including microscopy images of citrate estimation of citrate crown thickness” for details). The overall
layers. It was concluded that the preference for Au(111) is due result of 1.0−1.2 nm is close to experimentally measured values
to the possible binding of the terminal carboxyl group of citrate of 0.6−1.2 nm.1,3
to the bridge site of the Au(111) surface. On the other hand, To detail the localization of different molecules and atoms in
according to the arrangement of gold atoms on different the simulation cells, an analysis of the density distribution was
surfaces, formation of a citrate network is possible on both performed next. For both nanoparticles (6 and 14 nm), the
Au(100) and Au(111) surfaces in contrast to the Au(110) picture looks very similar. According to our results, the citrate
surface.1 ions concentrate mainly near nanoparticle edges and vertices,
Differential affinity to facets of a gold nanoparticle has also as proposed by Park et al.1 on the basis of measurements of
been demonstrated for other capping ligands.10−13 It is well- citrate shell thickness (see Figure 8A). Sodium ions are tightly
known that surfactant cetyltrimethylammonium bromide bound to the negatively charged citrate ions. The arrangement
(CTAB) has uneven density at side and end surfaces of a of Na+ ions in the structure (Figure 8A) indicates their major
gold nanorod. MD simulations showed that structural participation in the assembling and stabilization of the crown,
organization of CTAB on the Au(111) surface resulted in including the possibility of formation of citrate motifs. The
more open water−ion channels which provided additional analysis of citrate ion orientations relative to the gold surface
transport of growth components to nanorod ends.10−12 revealed a preferential interaction of CH2 groups with gold
Accordingly, elongation of a gold nanorod is probably atoms. Our results match the adsorption geometry of citrate on
associated with heterogeneity of surface coverage, namely a Au(111) surface reported by Wright et al.,36 because we
with high density of CTAB at the side facets ({100}, {110}) study the same model of citrate (Cit3− with optimized
and lower binding efficiency to the end facets of the {111} parameters for an aqueous medium38), the rigid-rod model
type. of polarization in gold,30,31 and the same GolP-CHARMM
We then used the distribution perpendicular to the surface force field.20,21
to estimate the number of citrate molecules directly adsorbed Water molecule density maps calculated during the last 7 ns
on the surface of a nanoparticle. The criterion for evaluation of of 28 ns MD trajectories uncovered their strong interaction
citrate molecules in close contact with the surface was the and specific orientation near the gold surfaces (Figure 8B).
distance on distribution graphs less than or equal to 0.5 nm. The water density map outlines the contours of a nanoparticle
Similar geometric principles were applied in calculations for surface, most sharply for the Au(100) surfaces (Figure 8B).
citrate molecules on the Au(111) surface.36 Table 2 shows that Detailed consideration of the water oxygen density map with
the gold surfaces showed a preferential orientation of water
Table 2. Analysis of Direct Contacts of Citrate Ions with the molecules relative to the gold atoms. For the Au(111) surface,
Surface of a Nanoparticle from the Distribution the plane of water molecules was preferentially parallel to the
Perpendicular to the Surface of the Nanoparticle for the gold surface (see Figure S19B and S19D). In the case of the
Last Three Frames of the MD Trajectory (see Methods)a Au(100) surface, parallel and, less likely, perpendicular
6 nm 14 nm
orientations were observed (see Figures 8C, S19A, and
S19C). The orientation of water on different surfaces and
surface type {111} {100} total {111} {100} total the preference for Au(100) over Au(111) are in agreement
average number 21 1 22 142 14 156 with quantum chemical calculations.20,67 The presence of more
of directly ad-
sorbed mole- tightly interacting water may influence the binding of citrate to
cules the Au(100) surface in comparison with Au(111). It can make
average number 159 94 253 1404 279 1683 an additional contribution to the predominant adsorption of
of molecules citrate on the Au(111) surface. In addition, water is most
above surface
number of mole- 451 451 451 2756 2756 2756
tightly interacting with nanoparticle edges via the oxygen atom
cules in cell without preferential orientation of the hydrogens. We propose
% of direct con- 4.7% 0.2% 4.9% 5.2% 0.5% 5.7% that the orientation of water on edges of a nanoparticle
tacts with sur- originates from tight interactions of the positively charged
face
% of analyzed 35.3% 20.8% 56.1% 51.0% 10.1% 61.1%
dipole atoms AUC with a negatively charged O atom (see
molecules below).
a
The molecule is regarded as directly adsorbed if the distance In summary, as a result of simulation, a structure of the
between a citrate mass center and gold atoms is less than or equal to citrate crown around a gold nanoparticle with d = 6 nm or d =
0.5 nm. 14 nm was successfully formed. We characterized the final
citrate distribution, dynamics of its formation, and final citrate
distribution for both nanoparticles. The main features of citrate
only ∼5% of citrate molecules in the simulation cell were coverage (such as heterogeneity, the presence of free gold
located directly on the surface of the 6 and 14 nm surfaces, and the thickness) were consistent with experimental
nanoparticles. Most citrate molecules adsorb indirectly on results.
the surface through Na+ ions.36 The percentage of citrate Analysis of the Electrostatic-Potential Distribution.
molecules being analyzed above the surface was ∼60% for both The resulting distribution of the citrate crown around a gold
nanoparticles (see Figure S1). The rest of the citrate molecules nanoparticle is difficult to explain without consideration of the
were associated with the edge volume of each nanoparticle. forces governing the interaction between a metal nanoparticle
To complete the analysis of the distribution perpendicular to and surrounding molecules. It seems that electrostatic forces
the surface of a nanoparticle, we estimated the thickness of the arising from polarization effects in the nanoparticle model
1286 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Figure 8. Density maps of different atoms in systems along the last 7 ns of the MD trajectory: A) citrate O carboxyl atom (red) and Na+ ions (blue)
for the nanoparticle of d = 14 nm; B) a water oxygen atom (O) for the nanoparticle of d = 14 nm; C) a water oxygen atom (green) and water
hydrogen atoms (red) above Au(100) for the nanoparticle of d = 6 nm. Real gold atoms are shown as yellow spheres.

Figure 9. Electric potential of a gold nanoparticle of d = 14 nm in vacuum (A) and in aqueous media: with only water molecules (B) or in the
presence of citrate (C). The negative potential is blue, and the positive potential is red.

could have a considerable influence on a charged citrate negatively charged oxygen atom of water. After that, gold
molecule during the simulation. dipoles are oriented according to the geometry of a
To reveal the possible mechanism behind the crown nanoparticle and affect water adsorption on different surfaces.
formation process, intensity of the electric potential was Nevertheless, it is worth mentioning that citrate forms the
analyzed for the gold nanoparticles in different environments. crown structure through sodium ions; therefore, more complex
We compared the electrostatic field potentials of the 14 nm electrostatic interactions with the surface are possible. In
gold nanoparticle between vacuum, water, and the citrate- addition, the arrangement of gold atoms on the surface can
capped model system simulated in the periodic cuboid cell. affect the binding to different surfaces of the nanoparticle via
The electrostatic potential of the nanoparticle was visualized in van der Waals forces.
all cases as described in Methods. On the basis of all these results, we can propose a
In vacuum, the nanoparticle had no distinguishable charge mechanism of formation of a nonisotropic citrate crown near
distribution profile at the surface or corners (Figure 9A). The a nanoparticle. First, the closest water molecules cause
analysis of the potential of a gold nanoparticle in a water box redistribution of gold nanoparticle dipoles. Then, the
revealed the development of heterogeneity. The Au(100) negatively charged citrate ions get attracted to the positively
surface carried a stronger negative potential above the surface charged edges through the long-range electrostatic interactions
than Au(111) did (Figure 9B). To compensate the potential and form a network around the nanoparticle. After citrate
near the uncharged nanoparticle, the field potential near the reaches the surface of the nanoparticle, the van der Waals
edges became positive. When a nanoparticle from the system interactions orient the molecule on the surface in addition to
with sodium citrate was evaluated (Figure 9C), a similar electrostatic forces. It can be assumed in the first
potential distribution was observed: the nanoparticle edges approximation that citrate prefers to interact with the
acquired a positive charge, while the surfaces became Au(111) surface over Au(100) owing to a weaker negative
negatively charged with higher magnitude for {100} as field above Au(111).
compared to {111}. This finding can probably explain the Similar observations related to the presence of a positive
origin of citrate concentration near the edges of a nanoparticle. electrostatic potential on nanoparticle corners with a low
Because water density maps (Figure 9B) uncovered the coordination number have been obtained for gold and
presence of water molecules near surfaces and edges of a platinum clusters.68 This observation explains the high catalytic
citrate-capped nanoparticle and a possibility of parallel water activity of these sites toward Lewis bases. It seems that the
orientation near Au(100) (Figures 8C and S19), we assumed nanoparticle morphology itself determines the charge dis-
that these data are in agreement with the electrostatic potential tribution and offers an additional opportunity for electrostatic
distribution. We believe that small water molecules could get interactions with the surrounding molecules.
close to more open atoms at the edges and vertices and induce Comparison of Simulation Results between the
a positive charge at these sites under the influence of the Polarizable and Nonpolarizable Model of a Gold
1287 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Figure 10. A) The simulation results on the 6 nm citrate-capped gold nanoparticle interacting with NaCl. Na+ and Cl− ions are not depicted for
clarity. The arrow marks the place where a big group of citrate molecules lost contact with the crown. B) The last frame of radial distribution graphs
for citrate (CIT) in a system with and without NaCl (red and blue curves, respectively). The radial distance step is 0.1 nm, and the position of zero
is at the center of the nanoparticle.

Nanoparticle. To clarify the involvement of the polarization especially the small nanoparticle (d = 6 nm), the Au(100)
effect in the process of citrate crown formation, we performed surface area was virtually unpopulated by citrate. We assume
simulation of the nonpolarizable model of a gold nanoparticle that the negative electrostatic field above Au(100) in the
(d = 6 nm) with sodium citrate. When only van der Waals polarizable gold model strongly hinders the efficiency of
interactions with gold were considered, the citrate crown also landing on the Au(100) surface of the nanoparticle.
completely formed after 21 ns of the simulation (Figure S21A). The orientations of citrate with respect to the surface of the
Moreover, the coverage had a heterogeneous multilayer polarizable and nonpolarizable gold model were similar, with
structure with the presence of naked gold surfaces as in the CH2 groups being preferentially exposed to the nanoparticle
case of the polarizable nanoparticle model. In contrast, the surface, and the COOH group pointed away. Apparently, this
patterns of arrangement of citrate molecules above the surface binding mode of citrate facilitated the assembly of “linear
of the nanoparticle were significantly different. chains”, “bilayers”, and “island” structural motifs, which have
The main difference, which is clearly seen in the distribution been described for a polarizable Au(111) surface.36
graphs perpendicular to the surface of a nanoparticle, is the In summary, a citrate crown was formed in both cases (in
number of close van der Waals contacts with the gold surface the simulation with and without the image charge effect in the
(32% of all citrate molecules in the case of the nonpolarizable metal). Nevertheless, the necessity of including the polar-
model versus 5% in the case of the polarizable model; see ization effects in the nanoparticle model is based on theoretical
Table S1, Figure S21C, and Table 2). We can propose a principles because citrate is a charged molecule (see
possible explanation of these results on the basis of
Introduction).
electrostatic-potential analysis of the polarizable gold nano-
Testing of the Citrate-Capped Nanoparticle Model
particle (see above). Inclusion of gold dipoles in the
with Sodium Chloride. One of methods for testing and
nanoparticle model caused changes in the charge distribution,
studying gold nanoparticles is salt-induced aggregation. It is
with the negative field potential being on the surfaces and the
known that at a high salt concentration, gold nanoparticles
positive field potential near edges and vertices of the
nanoparticle. During the process of crown formation, aggregate with a visible color change of the solution from red
negatively charged citrate is affected by repulsive forces from to blue.69 The aggregation mechanism is often associated with
a surface having a negative electrostatic field in the model with the removal of the negatively charged citrate crown from a
polarization. On the other hand, in the simulation of the nanoparticle under the influence of electrostatic forces and
nonpolarizable gold model, a citrate molecule could easily bind uncovering of the hydrophobic gold surface for van der Waals
to the center of nanoparticle facets, thereby resulting in an interactions with the surface of another nanoparticle.70
increasing number of close van der Waals contacts with the To study the suitability of the newly developed gold model
gold surface. for investigation of nanoparticle properties, we carried out a
Another feature of the simulation of the nonpolarizable simulation of the citrate-capped nanoparticle of d = 6 with ions
model of gold is comparable affinity of citrate for both types of Na+ and Cl− at a concentration of 0.8 M for 35 ns. We chose a
nanoparticle surfaces with no significant preference for 7 ns snapshot of the MD trajectory as a starting point of the
Au(111) (see Figure S21B−D). It should be noted that the simulation. This snapshot corresponds to 95% of citrate ions
simulation was conducted under conditions of periodically located near the nanoparticle when the formation of the crown
rising temperature, which increased the number of collision is clearly detectable but is not fully completed. This choice was
events between citrate molecules and the gold surface. governed by the necessity of decreasing the simulation time to
Nonetheless, in the case of a nanoparticle with polarization, observe possible citrate shell reorganization.
1288 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

Figure 11. A) The density map of Cl− (yellow) and citrate O carboxyl atom (cyan) with the nanoparticle of d = 6 nm along the last 7 ns of the MD
trajectory. B) Radial distribution of the last trajectory frame for citrate (CIT: blue curve), Cl− (CL: red curve), and Na+ ions (NA: yellow curve).
The radial distance step is 0.1 nm, and the position of zero is at the center of the nanoparticle.

During the simulation, we observed numerous changes in particle. A study on displacement of oligonucleotides (non-
the structure of the citrate crown. Typical processes along the covalently attached to a gold nanoparticle) by thiols points to
MD trajectory were as follows: repeating rearrangement of the the existence of various regimes involving only partial
structure, detachment of a small number of citrate ions from displacement of DNA from the surface of a gold nanoparticle
the surface, and clustering of citrates into new groups by high-molecular-weight thiols and complete removal of DNA
(compare Figure S22A and S22B). After 32 ns, a large group by low-molecular-weight thiols.71 A layer-by-layer approach
of citrate ions (28) separated from the shell (Figure 10A). In applied to obtain the gold nanoparticles with multilayer
the absence of NaCl, we never observed the detachment of coverage (consisting of positively and negatively charged
citrate clusters of such size from the crown during the ligands) indicates the possibility of mixing the layers and
simulation. formation of a shell with heterogeneous mosaic structure.72
Although rearrangement of the crown clearly takes place in The existence of selective surface-capping agents63,73 as well as
the structure along the trajectory, we cannot make a firm production of nanostar structures during nanoparticle over-
conclusion about the radical changes in the thickness of the growth74,75 also confirms the nontrivial nature of binding to
crown during the simulation from the radial distribution the surfaces and edges of a nanoparticle. The MD method can
dynamics (Figures 10B and S23). Analysis of density maps serve as a useful tool for investigating the interactions with
suggests that Cl− ions interacted with the citrate shell and got metal nanoparticles. Analysis of our simulation results
incorporated into the crown (Figure 11A). Moreover, uncovered formation of a citrate crown around the gold
localization of chloride ions at the edges of the nanoparticle nanoparticle. This finding is in good agreement with
is detectable. We assume that this phenomenon originated experimental data on its thickness and structure.
from a tight interaction of the positively charged dipole atoms
(AUC) with the negatively charged chloride ion, which has
smaller size and higher mobility than the citrate ion does. It
■ CONCLUSIONS
We developed a gold nanoparticle model with the polarization
can be concluded from radial distribution graphs (Figures 11B effect using the original method of unit cell initiation, cutting a
and S23) that sodium and chloride ions reach the same nanoparticle out of a crystal cube, and detecting the surfaces
plateau, which corresponds to the same bulk concentrations of and different types of gold atoms. As a result of modeling of
ions in solution. the gold nanoparticle with sodium citrate, a citrate-capped gold
As a result, it can be concluded that during the 35 ns nanoparticle model was obtained for the first time. The model
simulation, we did not detect the complete removal of the was verified according to the literature data. The model
citrate crown. Even though Cl− ions are incorporated into the reproduces such important properties as the structure and
crown, Na+ ions are still located there and, apparently, thickness of the citrate crown and rearrangement of citrate
maintain its structure (Figure S22D). Nevertheless, the results molecules under the influence of NaCl. It was demonstrated
indicate that there are changes in the structure of the crown that taking into account the effects of polarization in the model
under the influence of NaCl. of a gold nanoparticle influenced the simulation results
In summary, gold nanoparticles hold promise for various concerning the distribution of citrate ions on the metal
chemical and physical interactions and cause adsorption of a surface. On the basis of electric potential visualization, the
molecule from the surrounding medium. Besides, the presence mechanisms of interaction of polarizable gold nanoparticles
of surfaces with different arrangements of atoms as well as and citrate molecules were proposed.
specific sites with low coordination such as edges and vertices We believe that it is possible to use this model as a starting
further complicates the modes of binding to a nanoparticle. It simulation point to investigate the interactions of citrate-
seems that depending on their functional groups and charges, capped gold nanoparticles with other molecules, including
molecules can interact with different regions of the nano- proteins and nucleic acids.
1289 DOI: 10.1021/acs.jctc.8b00362
J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article


*
ASSOCIATED CONTENT
S Supporting Information
(8) Nelson, E. M.; Rothberg, L. J. Kinetics and Mechanism of Single-
Stranded DNA Adsorption onto Citrate-Stabilized Gold Nano-
particles in Colloidal Solution. Langmuir 2011, 27, 1770−1777.
The Supporting Information is available free of charge on the (9) Mahmood, M.; Casciano, D.; Xu, Y.; Biris, A. S. Engineered
ACS Publications website at DOI: 10.1021/acs.jctc.8b00362. nanostructural materials for application in cancer biology and
Preparing nanoparticle system for simulation; symmetric medicine. J. Appl. Toxicol. 2012, 32, 10−19.
equation for {111} and {100} planes; volumes above (10) Meena, S. K.; Sulpizi, M. Understanding the Microscopic
Au(111) and Au(100) (Figure S1); final result of radial Origin of Gold Nanoparticle Anisotropic Growth from Molecular
distribution graphs (Figure S2); estimation of citrate Dynamics Simulations. Langmuir 2013, 29, 14954−14961.
(11) Meena, S. K.; Celiksoy, S.; Schäfer, P.; Henkel, A.; Sönnichsen,
crown thickness; structure of citrate overlayers (Figures
C.; Sulpizi, M. The role of halide ions in the anisotropic growth of
S3−S10); normalized distribution of citrate mass centers gold nanoparticles: a microscopic, atomistic perspective. Phys. Chem.
perpendicular to surfaces of the nanoparticle (Figures Chem. Phys. 2016, 18, 13246−13254.
S11−S18); citrate dynamics perpendicular to surface of (12) Meena, S. K.; Sulpizi, M. From gold nanoseeds to nanorods:
nanoparticle; density maps of water oxygen and The microscopic origin of the anisotropic growth. Angew. Chem., Int.
hydrogen atoms (Figure S19); visualization of electric Ed. 2016, 55, 11960−11964.
field potential (Figure S20); simulation results of (13) Almora-Barrios, N.; Novell-Leruth, G.; Whiting, P.; Liz-
nonpolarizable gold nanoparticle model (Figure S21); Marzán, L. M.; López, N. Theoretical description of the role of
analysis of direct contacts of citrate molecules with halides, silver, and surfactants on the structure of gold nanorods. Nano
nonpolarizable nanoparticle model (Table S1); density Lett. 2014, 14, 871−875.
maps of different atoms in simulation of nanoparticle (14) Tavanti, F.; Pedonea, A.; Menziani, M. C. A closer look into the
ubiquitin corona on gold nanoparticles by computational studies. New
(Figure S22); dynamics of system radial distribution
J. Chem. 2015, 39, 2474−2482.
with NaCl (Figure S23) (PDF)


(15) Feng, J.; Slocik, J. M.; Sarikaya, M.; Naik, R. R.; Farmer, B. L.;
Heinz, H. Influence of the shape of nanostructured metal surfaces on
AUTHOR INFORMATION adsorption of single peptide molecules in aqueous solution. Small
Corresponding Authors 2012, 8, 1049−1059.
(16) Lin, W.; Insley, T.; Tuttle, M. D.; Zhu, L.; Berthold, D. A.; Král,
*Phone: +7 383-363-5134. Fax: +7 383-363-5134. E-mail:
P.; Rienstra, C. M.; Murphy, C. J. Control of protein orientation on
pyshnyi@niboch.nsc.ru (D.V.P.). gold nanoparticles. J. Phys. Chem. C 2015, 119, 21035−21043.
*Phone: +7 383-363-5134. Fax: +7 383-363-5134. E-mail: (17) Cicero, G.; Calzolari, A.; Corni, S.; Catellani, A. Anomalous
lomzov@niboch.nsc.ru (A.A.L.). Wetting Layer at the Au(111) Surface. J. Phys. Chem. Lett. 2011, 2,
ORCID 2582−2586.
Dmitrii V. Pyshnyi: 0000-0002-2587-3719 (18) Rosa, M.; Corni, S.; Di Felice, R. Interaction of Nucleic Acid
Bases with the Au(111) Surface. J. Chem. Theory Comput. 2013, 9,
Alexander A. Lomzov: 0000-0003-3889-9464
4552−4561.
Funding (19) Rosa, M.; Corni, S.; Di Felice, R. Enthalpy-Entropy Tuning in
This work was done in accordance with the State assignment the Adsorption of Nucleobases at the Au(111) Surface. J. Chem.
by the Russian State funded budget project (VI.62.1.4, 0309- Theory Comput. 2014, 10, 1707−1716.
2016-0004), and the structural analysis of the interaction (20) Wright, L. B.; Rodger, P. M.; Corni, S.; Walsh, T. GolP-
between citrate ions and a gold nanoparticle (A.A.L.) was in CHARMM: First-Principles Based Force Fields for the Interaction of
part supported by the Russian Science Foundation (grant No. Proteins with Au(111) and Au(100). J. Chem. Theory Comput. 2013,
16-15-10156). 9, 1616−1630.
(21) Wright, L. B.; Rodger, P. M.; Walsh, T. R.; Corni, S. First-
Notes principles-based force field for the interaction of proteins with
The authors declare no competing financial interest. Au(100) (5 × 1): an extension of GolP-CHARMM. J. Phys. Chem. C

■ REFERENCES
(1) Park, J.-W.; Shumaker-Parry, J. S. Structural study of citrate
2013, 117, 24292−24306.
(22) Mahmoudi, M.; Lynch, I.; Ejtehadi, M. R.; Monopoli, M. P.;
Bombelli, F. B.; Laurent, S. Protein−nanoparticle interactions:
layers on gold nanoparticles: role of intermolecular interactions in opportunities and challenges. Chem. Rev. 2011, 111, 5610−5637.
stabilizing nanoparticles. J. Am. Chem. Soc. 2014, 136, 1907−1921. (23) Ozboyaci, M.; Kokh, D. B.; Corni, S.; Wade, R. C. Modeling
(2) Park, J.-W.; Shumaker-Parry, J. S. Strong resistance of citrate and simulation of protein-surface interactions: achievements and
anions on metal nanoparticles to desorption under thiol functionaliza- challenges. Q. Rev. Biophys. 2016, 49, No. e4.
tion. ACS Nano 2015, 9, 1665−1682. (24) Heinz, H.; Vaia, R. A.; Farmer, B. L.; Naik, R. R. Accurate
(3) Lee, Z.; Jeon, K.-J.; Dato, A.; Erni, R.; Richardson, T. J.; Simulation of Surfaces and Interfaces of Face-Centered Cubic Metals
Frenklach, M.; Radmilovic, V. Direct imaging of soft-hard interfaces Using 12−6 and 9−6 Lennard-Jones Potentials. J. Phys. Chem. C
enabled by graphene. Nano Lett. 2009, 9, 3365−3369. 2008, 112, 17281−17290.
(4) Piella, J.; Bastús, N. G.; Puntes, V. Size-controlled synthesis of (25) Braun, R.; Sarikaya, M.; Schulten, K. Genetically engineered
sub-10-nanometer citrate-stabilized gold nanoparticles and related gold-binding polypeptides: structure prediction and molecular
optical properties. Chem. Mater. 2016, 28, 1066−1075. dynamics. J. Biomater. Sci., Polym. Ed. 2002, 13, 747−757.
(5) Dinkel, R.; Braunschweig, B.; Peukert, W. Fast and Slow Ligand (26) Heinz, H.; Farmer, B. L.; Pandey, R. B.; Slocik, J. M.; Patnaik, S.
Exchange at the Surface of Colloidal Gold Nanoparticles. J. Phys. S.; Pachter, R.; Naik, R. R. Nature of molecular interactions of
Chem. C 2016, 120, 1673−1682. peptides with gold, palladium, and Pd−Au bimetal surfaces in
(6) Carnerero, J. M.; Jimenez-Ruiz, A.; Castillo, P. M.; Prado-Gotor, aqueous solution. J. Am. Chem. Soc. 2009, 131, 9704−9714.
R. Covalent and Non-Covalent DNA-Gold-Nanoparticle Interactions: (27) Heinz, H.; Jha, K. C.; Luettmer-Strathmann, J.; Farmer, B. L.;
New Avenues of Research. ChemPhysChem 2017, 18, 17−33. Naik, R. R. Polarization at metal-biomolecular interfaces in solution. J.
(7) Zhang, X.; Servos, M. R.; Liu, J. Surface Science of DNA R. Soc., Interface 2011, 8, 220−232.
Adsorption onto Citrate-Capped Gold Nanoparticles. Langmuir 2012, (28) Li, X.; Ågren, H. Dynamics Simulations Using a Capacitance−
28, 3896−3902. Polarizability Force Field. J. Phys. Chem. C 2015, 119, 19430−19437.

1290 DOI: 10.1021/acs.jctc.8b00362


J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

(29) Geada, I. L.; Ramezani-Dakhel, H.; Jamil, T.; Sulpizi, M.; (50) Roe, D. R.; Cheatham, T. E., 3rd. PTRAJ and CPPTRAJ:
Heinz, H. Insight into induced charges at metal surfaces and Software for Processing and Analysis of Molecular Dynamics
biointerfaces using a polarizable Lennard-Jones potential. Nat. Trajectory Data. J. Chem. Theory Comput. 2013, 9, 3084−3095.
Commun. 2018, 9, 716. (51) Sun, Y.; Changhua, A. Shaped gold and silver nanoparticles.
(30) Iori, F.; Di Felice, R.; Molinari, E.; Corni, S. GolP: an atomistic Front. Mater. Sci. 2011, 5, 1−24.
force-field to describe the interaction of proteins with Au(111) (52) Barnard, A. S.; Lin, X. M.; Curtiss, L. A. Equilibrium
surfaces in water. J. Comput. Chem. 2009, 30, 1465−1476. morphology of face-centered cubic gold nanoparticles > 3 nm and
(31) Iori, F.; Corni, S. Including image charge effects in the the shape changes induced by temperature. J. Phys. Chem. B 2005,
molecular dynamics simulations of molecules on metal surfaces. J. 109, 24465−24472.
Comput. Chem. 2008, 29, 1656−1666. (53) Barnard, A. S. Direct comparison of kinetic and thermodynamic
(32) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. Development influences on gold nanomorphology. Acc. Chem. Res. 2012, 45, 1688−
and Testing of the OPLS All-Atom Force Field on Conformational 1697.
Energetics and Properties of Organic Liquids. J. Am. Chem. Soc. 1996, (54) Barnard, A. S.; Young, N.; Kirkland, A. I.; van Huis, M. A.; Xu,
118, 11225−11236. H. Nanogold: a quantitative phase map. ACS Nano 2009, 3, 1431−
(33) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.; 1436.
(55) Barmparis, G. D.; Lodziana, Z.; Lopez, N.; Remediakis, I. N.
Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.;
Nanoparticle shapes by using Wulff constructions and first-principles
Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T.; Mattos,
calculations. Beilstein J. Nanotechnol. 2015, 6, 361−368.
C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; (56) Barnard, A. S.; Curtiss, L. A. Predicting the Shape and Structure
Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; of Face-Centered Cubic Gold Nanocrystals Smaller than 3 nm.
Watanabe, M.; Wiórkiewicz-Kuczera, J.; Yin, D.; Karplus, M. All-atom ChemPhysChem 2006, 7, 1544−1553.
empirical potential for molecular modeling and dynamics studies of (57) Li, J.; Li, X.; Zhai, H.- J.; Wang, L.-S. Au20: a tetrahedral
proteins. J. Phys. Chem. B 1998, 102, 3586−3616. cluster. Science 2003, 299, 864−867.
(34) Hoefling, M.; Monti, S.; Corni, S.; Gottschalk, K. E. Interaction (58) He, X.; Chen, Z.-X. A study on the morphology and catalytic
of β-sheet folds with a gold surface. PLoS One 2011, 6, No. e20925. activity of gold nanoparticles by the kinetic Monte Carlo simulation.
(35) Brancolini, G.; Kokh, D. B.; Calzolai, L.; Wade, R. C.; Corni, S. Appl. Surf. Sci. 2016, 370, 433−436.
Docking of ubiquitin to gold nanoparticles. ACS Nano 2012, 6, (59) Baletto, F.; Ferrando, R.; Fortunelli, A.; Montalenti, F.; Mottet,
9863−9878. C. Crossover among structural motifs in transition and noble-metal
(36) Wright, L. B.; Rodger, P. M.; Walsh, T. R. Structure and clusters. J. Chem. Phys. 2002, 116, 3856−3863.
properties of citrate overlayers adsorbed at the aqueous Au(111) (60) Barmparis, G. D.; Honkala, K.; Remediakis, I. N. Thiolate
interface. Langmuir 2014, 30, 15171−15180. Adsorption on Au(hkl) and Equilibrium Shape of Large Thiolate-
(37) Turkevich, J.; Stevenson, P. C.; Hillier, J. A Study of the covered Gold Nanoparticles. J. Chem. Phys. 2013, 138, No. 064702.
Nucleation and Growth Processes in the Synthesis of Colloidal Gold. (61) Viswanathan, V.; Wang, F.; Pitsch, H. Monte Carlo-based
Discuss. Faraday Soc. 1951, 11, 55−75. approach for simulating nanostructured catalytic and electrocatalytic
(38) Wright, L. B.; Rodger, P. M.; Walsh, T. R. Aqueous citrate: a systems. Comput. Sci. Eng. 2012, 14, 60−69.
first-principles and force-field molecular dynamics study. RSC Adv. (62) Gilroy, K. D.; Puibasset, J.; Vara, M.; Xia, Y. On the
2013, 3, 16399−16409. Thermodynamics and Experimental Control of Twinning in Metal
(39) Keith, J. A.; Fantauzzi, D.; Jacob, T.; van Duin, A. C. T. Nanocrystals. Angew. Chem., Int. Ed. 2017, 56, 8647−8651.
Reactive forcefield for simulating gold surfaces and nanoparticles. (63) Xia, Y.; Xiong, Y.; Lim, B.; Skrabalak, S. E. Shape-Controlled
Phys. Rev. B: Condens. Matter Mater. Phys. 2010, 81, 235404. Synthesis of Metal Nanocrystals: Simple Chemistry Meets Complex
(40) Monti, S.; Barcaro, G.; Sementa, L.; Carravetta, V.; Ågren, H. Physics? Angew. Chem., Int. Ed. 2009, 48, 60−103.
Characterization of the adsorption dynamics of trisodium citrate on (64) Huang, R.; Wen, Y.-H.; Shao, G.-F; Zhu, Z.-Z.; Sun, S.-G.
gold in water solution. RSC Adv. 2017, 7, 49655−49663. Single-crystalline and multiple-twinned gold nanoparticles: an atom-
(41) Abraham, M. J.; Murtola, T.; Schulz, R.; Páll, S.; Smith, J. C.; istic perspective on structural and thermal stabilities. RSC Adv. 2014,
Hess, B.; Lindahl, E. GROMACS: High performance molecular 4, 7528−7537.
simulations through multi-level parallelism from laptops to super- (65) Salomon-Ferrer, R.; Case, D. A.; Walker, R. C. An overview of
the Amber biomolecular simulation package. WIREs Comput. Mol. Sci.
computers. SoftwareX 2015, 1, 19−25.
2013, 3, 198−210.
(42) Nosé, S. A molecular-dynamics method for simulation in the
(66) Maeland, A.; Flanagan, T. B. Lattice spacings of gold−
canonical ensemble. Mol. Phys. 1984, 52, 255−268.
palladium alloys. Can. J. Phys. 1964, 42, 2364−2366.
(43) Hoover, W. G. Canonical dynamics: Equilibrium phase-space
(67) Hughes, Z. E.; Walsh, T. R. Non-covalent adsorption of amino
distributions. Phys. Rev. A: At., Mol., Opt. Phys. 1985, 31, 1695−1697.
acid analogues on noble-metal nanoparticles: influence of edges and
(44) Páll, S.; Hess, B. A flexible algorithm for calculating pair
vertices. Phys. Chem. Chem. Phys. 2016, 18, 17525−17533.
interactions on SIMD architectures. Comput. Phys. Commun. 2013, (68) Stenlid, J. H.; Brinck, T. Extending the σ-Hole Concept to
184, 2641−2650. Metals: An Electrostatic Interpretation of the Effects of Nanostructure
(45) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N· in Gold and Platinum Catalysis. J. Am. Chem. Soc. 2017, 139, 11012−
log(N) method for Ewald sums in large systems. J. Chem. Phys. 1993, 11015.
98, 10089−10092. (69) Zhang, Z.; Li, H.; Zhang, F.; Wu, Y.; Guo, Z.; Zhou, L.; Li, J.
(46) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. Investigation of Halide-Induced Aggregation of Au Nanoparticles into
W.; Klein, M. L. Comparison of Simple Potential Functions for Spongelike Gold. Langmuir 2014, 30, 2648−2659.
Simulating Liquid Water. J. Chem. Phys. 1983, 79, 926−935. (70) Pfeiffer, C.; Rehbock, C.; Hühn, D.; Carrillo-Carrion, C.; de
(47) Neria, E.; Fischer, S.; Karplus, M. Simulation of Activation Free Aberasturi, D. J.; Merk, V.; Barcikowski, S.; Parak, W. J. Interaction of
Energies in Molecular Systems. J. Chem. Phys. 1996, 105, 1902−1921. colloidal nanoparticles with their local environment: the (ionic)
(48) Halgren, T. A. The representation of van der Waals (vdW) nanoenvironment around nanoparticles is different from bulk and
interactions in molecular mechanics force fields: potential form, determines the physico-chemical properties of the nanoparticles. J. R.
combination rules, and vdW parameters. J. Am. Chem. Soc. 1992, 114, Soc., Interface 2014, 11, 20130931.
7827−7843. (71) Epanchintseva, A.; Vorobjev, P.; Pyshnyi, D.; Pyshnaya, I. Fast
(49) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular and Strong Adsorption of Native Oligonucleotides on Citrate-Coated
Dynamics. J. Mol. Graphics 1996, 14, 33−38. Gold Nanoparticles. Langmuir 2018, 34, 164−172.

1291 DOI: 10.1021/acs.jctc.8b00362


J. Chem. Theory Comput. 2019, 15, 1278−1292
Journal of Chemical Theory and Computation Article

(72) Shashkova, V. V.; Epanchintseva, A. V.; Vorobjev, P. E.; Razum,


K. V.; Ryabchikova, E. I.; Pyshnyi, D. V.; Pyshnaya, I. A. Multilayer
Associates Based on Oligonucleotides and Gold Nanoparticles. Russ. J.
Bioorg. Chem. 2017, 43, 64−70.
(73) Grzelczak, M.; Pérez-Juste, J.; Mulvaney, P.; Liz-Marzán, L. M.
Shape control in gold nanoparticle synthesis. Chem. Soc. Rev. 2008,
37, 1783−1791.
(74) Wang, Z.; Zhang, J.; Ekman, J. M.; Kenis, P. J.; Lu, Y. DNA-
Mediated Control of Metal Nanoparticle Shape: One-Pot Synthesis
and Cellular Uptake of Highly Stable and Functional Gold
Nanoflowers. Nano Lett. 2010, 10, 1886−1891.
(75) Niu, W.; Chua, Y. A.; Zhang, W.; Huang, H.; Lu, X. Highly
Symmetric Gold Nanostars: Crystallographic Control and Surface-
Enhanced Raman Scattering Property. J. Am. Chem. Soc. 2015, 137,
10460−10463.

1292 DOI: 10.1021/acs.jctc.8b00362


J. Chem. Theory Comput. 2019, 15, 1278−1292

You might also like