Self-assembly of a catechol-based macrocycle at the liquid–solid interface: experiments and molecular dynamics simulations

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 11937

Cite this: Phys. Chem. Chem. Phys., 2012, 14, 11937–11943

Self-assembly of a catechol-based macrocycle at the liquid–solid

interface: experiments and molecular dynamics simulations

Javier Saiz-Poseu,aAlberto Martınez-Otero,

bThomas Roussel,

cJoseph K.-H. Hui,

d

Mavis L. Montero,eRoberto Urcuyo,

eMark J. MacLachlan,

dJordi Faraudo*

cand

Daniel Ruiz-Molina*b

Received 2nd May 2012, Accepted 10th July 2012

DOI: 10.1039/c2cp41407d

This combined experimental (STM, XPS) and molecular dynamics simulation study highlights the

complex and subtle interplay of solvent effects and surface interactions on the 2-D self-assembly

pattern of a Schiff-base macrocycle containing catechol moieties at the liquid–solid interface.

STM imaging reveals a hexagonal ordering of the macrocycles at the n-tetradecane/Au(111)

interface, compatible with a desorption of the lateral chains of the macrocycle. Interestingly,

all the triangular-shaped macrocycles are oriented in the same direction, avoiding a close-packed

structure. XPS experiments indicate the presence of a strong macrocycle–surface interaction.

Also, MD simulations reveal substantial solvent effects. In particular, we find that co-adsorption

of solvent molecules with the macrocycles induces desorption of lateral chains, and the solvent

molecules act as spacers stabilizing the open self-assembly pattern.

1. Introduction

Combined scanning tunneling microscopy (STM) and theore-

tical simulations have emerged as a powerful experimental tool

to understand molecule–metal interactions and their role in

different fields ranging from heterogeneous catalysis to molecular

electronics or photochemical processes.1 This is especially

relevant for macrocycles with more internal degrees of freedom,

for which surface interactions can modify their configuration

and, consequently, their supramolecular assembly. Several

different families of macrocycles on surfaces have been studied

by STM,2 but as far as we know, no examples of surface-

assembled catechol-based macrocycles have been described in

spite of the important role of catechols in several natural or

artificial coatings and interfaces. In fact, catechol, catechin,

phenol, and their derivatives have been explored as coatings

on metal surfaces to improve their resistance to corrosion.3

Catechol can react with surface oxygen or other oxidizing

agents to form o-quinone. Wei et al. have studied catechols on

a well-defined Cu(111) surface due to their inherent interest as

antioxidants for metal surfaces.4 Weinhold et al. have also

reported a detailed structural and spectroscopic investigation

of the interface between L-DOPA and a single-crystalline

Au(110) model surface.5 More recently, Diebold et al. have

extensively studied the growth and organization of a catechol

monolayer on TiO2 due to their relevance in dye-sensitized

solar cells (DSSCs).6 However, none of the previous examples

corresponds to a catechol-based macrocycle. Moreover, all of

these examples have been studied under strict vacuum condi-

tions and do not give insight into the role of the solvent in the

interaction of catechols with surfaces.

In this paper, we report a combined experimental and

theoretical study of the assembly of macrocycle 1 at the

n-tetradecane/Au(111) interface. The interest to study com-

pound 1 is twofold. First, it represents a nice example of a

macrocycle with three catechols rigidly linked. Second, the

macrocycle structure disrupts intermolecular hydrogen-bonding

interactions. This allows us to get more details of the role

played by molecule–surface interactions as a continuation of

previous work developed in our groups focussed on developing

a fundamental understanding of the interaction of catechols

with surfaces in the presence of solvent.7 The self-assembly

pattern is characterized by STM and the interaction of

compound 1 with the Au(111) surface is studied by X-ray

photoemission spectroscopy (XPS). The role of solvent is

investigated with full atomistic details by all-atomic molecular

dynamics (MD) simulations. We emphasize here that solvent

effects play a substantial role in self-assembly of functional

molecules at solid surfaces.8 Our combined experimental and

a Fundacion Privada ASCAMM, Unidad de Nanotecnologıa(NanoMM), ParcTecnologic del Valles, Av. Universitat Autonoma,23, E-08290 Cerdanyola del Valles, Spain

bCentro de Investigacion en Nanociencia y Nanotecnologıa(CIN2-CSIC), Campus UAB, E-08193, Bellaterra, Spain.E-mail: [email protected]; Fax: +34 935813717; Tel: +34 935814777

c Institut de Ciencia de Materials de Barcelona, ICMAB-CSIC,Campus de la UAB, E-08193 Bellaterra, Spain.E-mail: [email protected]

dDepartment of Chemistry, University of British Columbia,2036 Main Mall, Vancouver, BC V6T 1Z1, Canada

e Centro de Electroquımica y Energıa Quımica, CELEQ,Universidad de Costa Rica, San Jose, Costa Rica

PCCP Dynamic Article Links

www.rsc.org/pccp PAPER

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online / Journal Homepage / Table of Contents for this issue

11938 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 This journal is c the Owner Societies 2012

simulation study highlights the complex and subtle interplay

of solvent effects and surface interactions on the 2-D self-

assembly pattern of compound 1.

2. Methods

Compound 1 was prepared as previously described.9 Solvents

and reagents were used as received unless explicitly described.

Experimental STM

All (STM) experiments were performed at room temperature

using a PicoSPM (Agilent) in constant current mode. When the

heating of the sample was needed, a LakeShore 331 Temperature

Controller was used. Pt/Ir STM tips were prepared by mechanical

cutting of Pt/Ir wires (80 : 20, diameter 0.25 mm, Advent

Research Materials, Ltd.). The molecules were dissolved in

n-tetradecane with a concentration of approximately 2 mg mL�1.

A hot drop of the solution was applied to a freshly cleaved

graphite substrate (HOPG, grade ZYB,Momentive Performance

Materials Quartz GMBH) and the tip was immersed in it. The

STM images were then obtained at the liquid–solid interface.

It was possible to scan the underlying graphite substrate after

scanning the self-assembled monolayer. This enabled us to

correct the drift effects using the Scanning Probe Image

Processor (SPIP) software (Image Metrology ApS).

Experimental XPS

All experiments were performed in a Thermo K-Alpha with

monochromated X-rays (12 kV, 6 mA) with a spot size of

400 mm and a takeoff angle of 901 relative to the surface, with a

typical exposure time per spot (usually three spots) of 2–7 min

in total to minimize beam damage. The binding energies were

calibrated based on Au 4f7/2 at 83.96 eV, Ag 3d5/2 at 368.21 eV

and Cu LVV at 568.09 eV. Surveys were done with high pass

energy (200 eV) while high resolution spectra were acquired

with 50 eV of pass energy. Typical pressures during analysis

were below 10�8 Torr.

Methodology for molecular dynamics simulations

Molecular Dynamics (MD) simulations are based on the numerical

solution of the Newtonian equations of motion for all atoms

of a molecular system constrained to the given thermodynamic

conditions. All MD simulations (except otherwise indicated)

were performed using the NAMD software,10 version 2.7 running

in parallel at the Finisterrae Supercomputer (CESGA Super-

computing Center). The equations of motion were solved with

a 2 fs time step. In our constant temperature ensemble

simulations (NVT), the temperature was kept constant at

20 1C using the Langevin thermostat with a relaxation con-

stant of 1 ps�1. In simulations at constant pressure and

temperature (NpT), we employed the Nose–Hoover–Langevin

piston as implemented in NAMD with an oscillation period of

100 fs and a decay time of 50 fs to adjust the solution pressure

at 1 atm. Periodic boundary conditions in all directions were

employed in all our simulations.

The model for the molecules (compound 1 and solvent) was

based on the CHARMM22/CMAP force field,11 designed for

biomolecular simulations. Although compound 1 has not been

previously simulated, the selected force field has been success-

fully employed for simulations of macrocycles (see for example

ref. 12). In fact, porphyrin macrocycles were included in the

parameterization of the CHARMM22 version of the force

field. In our simulations here we employ standard CHARMM

atom types. The modular structure of this force field (con-

structed from quantum chemical calculations of the inter-

actions between model compounds) allows one to construct

the model parameters for a given organic compound from the

basic building blocks of the force field. Within this force field,

intramolecular interactions contained bonding, torsion and

dihedral potentials and intermolecular interactions were

described by electrostatic interactions (modelled with partial

charges) plus a Lennard-Jones interaction potential. A NAMD

topology file containing the values of the force field parameters

required for the simulation is available upon request. The force

field for Au atoms and their interactions with organic molecules

was taken from a new improved Lennard-Jones parameteriza-

tion, accurate for interfacial calculations and compatible with

the CHARMM force field.13

In this paper, we have performed four different simulation

runs, summarized in Table 1. The first set of simulations

(denoted simulations S1A and S1B) are designed for under-

standing of surface–molecule interactions and the role of

solvent in these interactions. In simulations S1A and S1B we

considered a small surface patch with a single pre-adsorbed

molecule of compound 1. Simulation S1A is performed under

vacuum conditions whereas simulation S1B is performed in

n-tetradecane solution (as in experiments). The second set of

simulations (simulations S4A and S4B) were designed to study

the influence of the solvent on the assembly pattern deduced

from STM experiments. In this case, we considered a larger

gold surface with four adsorbed molecules of compound 1.

Simulation S4A is performed under vacuum conditions whereas

Table 1 Details corresponding to the different MD simulations

Surface area Au atoms Solvent molecules Adsorbed macrocycles Atoms (total) Sim. time (NVT) (ns)

Simulation S1A 49.616 A � 57.529 A 299 — 1 1502 15Simulation S1B 49.616 A � 57.529 A 299 400 1 19 102 29Simulation S4A 43.17 A � 59.96 A 4320 — 4 5544 20Simulation S4B 43.17 A � 59.96 A 4320 400 4 23 144 20

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 11939

simulation S4B is performed in n-tetradecane solution (as in

experiments).

Before running the MD simulations, we have to create

suitable initial configurations of the molecules and the surface

involved in the simulations. First, we built a gold slab containing

a total of 4320 Au atoms arranged in 12 layers of Au atoms

stacked along the h111i orientation of the fcc structure. We

then relaxed the surface by performing a quenched molecular

dynamics simulation. The metallic interaction energy is modelled

by a semi-empirical many-body potential proposed and tested

in a previous work,14 derived from tight binding quantum

mechanical calculations in its second moment approximation

(TB-SMA). These calculations were not performed using

NAMD since the employed potential was not implemented

there (instead we used a non-commercial program15).

The resulting structure was a slab of 12 layers of Au(111) with

a surface with dimensions 43.17 A � 59.96 A. Once the gold

structure is generated, the Au atoms were maintained fixed at

their equilibrium positions during all the remaining simulations,

in order to speed up our costly calculations. On top of the gold

surface, we placed four molecules of compound 1 with distances

and orientations corresponding to the unit cell shown in Fig. 1b,

as determined from STM measurements. After that, the posi-

tions of the atoms of the compound 1 molecule were energy

minimized using the NAMD 2.7 software.9 The employed

minimization algorithm was a sophisticated conjugate gradient

and line search algorithm. The method of conjugate gradients is

used to select successive search directions (starting with the

initial gradient), which eliminates repeated minimization along

the same direction. Along each direction, a minimum is first

bracketed (rigorously bound) and then converged upon by

either a golden section search or, when possible, a quadratically

convergent method using gradient information.

After 104 minimization steps, the energy stabilized configuration

was saved for use in S4A simulations as the initial condition.

Using the same methodology we prepared another surface (with

a smaller number of Au atoms and Au layers, see Table 1) to be

employed as the initial condition in the S1A simulations.

In order to obtain the initial conditions (with n-tetradecane

solvent) for the S1B and S4B simulations, we prepared a slab

of pre-equilibrated liquid n-tetradecane solvent. This equili-

brated solution of pure n-tetradecane at 20 1C and 1 atm was

prepared as follows: first, we generated a system with a lattice

containing 400 n-tetradecane solvent molecules (17 600 atoms

in total). Using this configuration as the initial condition, we

have conducted an NpT simulation inside a cubic box with

periodic boundary conditions in all directions. The barostat

was adjusted to 1 atm and the thermostat to 20 1C. The liquid

solvent was considered to be equilibrated after a few nano-

seconds of simulation since the relevant magnitudes of interest

(length of each side of simulation box, pressure) were stabi-

lized and a liquid of 400 n-tetradecane molecules inside a box

of size 175.4 nm3 was obtained. The initial configuration for

simulation S4B was obtained by adding this solvent slab to the

initial condition used for simulation S4A. In this case, the full

system (surface + adsorbates + solvent) contains a total of

23 144 atoms (see Table 1). After merging the solvent with the

surface and the adsorbed macrocycles, we performed an addi-

tional energy minimization (104 steps) using the same algorithm

as described earlier in this section. The configuration obtained

after this minimization was employed as the initial configu-

ration for simulation S4B. The initial configuration for simula-

tion S1B was obtained following an analogous procedure.

All simulations reported in Table 1 were performed under

NVT conditions (with T= 20 1C in all cases). The simulations

in the presence of solvent (S1B and S4B) were run in parallel

employing 32 Itanium Montvale processors whereas the less

expensive vacuum simulations (S1A and S4A) were run on a

single processor. For the sake of comparison, we should note

that each nanosecond of simulation required 0.84 h of CPU

time in the case of S4A and 115 h of CPU time in the case

of S4B.

All of the snapshots from the simulations were obtained

employing the Visual Molecular Dynamics (VMD) software,16

version 1.9. The interaction energies, unit cell distances and

unit cell angle reported in the Results section were computed

from the MD trajectories employing the different analysis options

in VMD (options ‘‘energy’’, ‘‘labels-bond’’ and ‘‘labels-angle’’).

As usual in simulations, the values of these quantities were

observed to evolve during a few nanoseconds before reaching

stable, equilibrium values. The results reported in the paper

were computed as averages over the part of the trajectory

corresponding to these equilibrium configurations (with a

typical length of about 10–15 ns). The calculation of the

number of adsorbed carbon atoms from the different species

over Au(111) was done as follows. First, we computed with

VMD the correlation function between surface Au atoms and

C atoms. Then we identified as adsorbed C atoms all those

C atoms located at a distance from Au smaller than the

distance corresponding to the first minimum (located after

the first peak) of the correlation function. This task was done

using a homemade python script (available from the authors)

running in VMD version 1.8.6. The obtained results were

averaged over equilibrium configurations.

3. Results and discussion

STM investigations

Initially, the experimental conditions described in Fig. 1 were

used to obtain stable and reproducible images at least over three

independent surface areas for different casting experiments of

Fig. 1 (a) STM image (27 nm � 27 nm) of 1 self-assembled at the

n-tetradecane–Au(111) interface. The imaging conditions are Iset = 46 pA

and Vbias = 565 mV. (b) A tentative model of packing. Unit cell

parameters: a = 2.66 � 0.1 nm, b = 2.59 � 0.1 nm, a = 113 � 11.

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online

11940 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 This journal is c the Owner Societies 2012

freshly prepared samples and their stability assessed by taking

STM images at different time intervals. Triangular motifs with

a side length of 1.8 � 0.2 nm and a bright contrast attributed

to the aromatic rings can be observed (see Fig. 1a). These units

organize into a hexagonally centred compact packing where

each molecule is surrounded by six neighbouring molecules.

Considering this organization and the maximum space left

between adjacent molecules (B0.5 nm), it is reasonable to

assume that no alkoxy chains are physisorbed on the surface.

A schematic representation of the corresponding molecular

packing is shown in Fig. 1b, where all the molecules are

oriented in the same direction avoiding a close-packed structure.

The observed preferential orientation could be indicative of a

strong interaction with a reconstructed gold surface. Indeed,

it is known that n-alkanes in general,17 and n-tetradecane in

particular,18 induce reconstruction of the Au(111) surface. The

denser and anisotropic reconstructed surface increases the

interaction of adsorbates with the surface and stabilizes an

anisotropic self-assembly pattern.16

XPS experiments

A 1 mM solution of compound 1 in n-tetradecane was

prepared and dropped onto a 1 � 1 cm2 piece of glass-covered

Au(111). The surface was dried in nitrogen. The XPS measure-

ments were performed for the multilayer so formed. The gold

and nitrogen spectral regions were specifically observed (see

Fig. 2 and 3), to search for information regarding the chemical

state of the nitrogen and gold atoms in a multilayer of

compound 1 on Au(111).

In the Au(4f) region, a close examination of the band shape

and position reveals two major differences after the deep

coating. First, the overall peak position is shifted to a lower

binding energy (BE) by 0.3 eV. Second, there are two over-

lapping components under the band envelope. By spectral

deconvolution, we have identified two sets of Au(4f) bands,

that is, 84.1 eV and 84.8 eV for Au(4f7/2), and 87.8 eV and

88.0 eV for Au(4f5/2), see Fig. 2. This behaviour indicates the

presence of at least two oxidation states. The lower BE

component can be attributed to Au(0), whereas the higher

BE component is likely indicative of the presence of Au(I).19

Even though compound 1 has only one type of nitrogen, the

N 1s photoemission spectra (Fig. 3) also show two peaks at

402.9 and 399.7 eV. The higher BE component probably

corresponds to –CQN– nitrogen bonded to gold, whereas

the lower BE component corresponds to free imine groups,

which presents lower N 1s binding energies than that of N 1s

bonded to gold surfaces.20

This behaviour of gold and nitrogen clearly indicates that the

multilayer contains two chemically different types of compound 1

molecules (free molecules and adsorbed molecules). Moreover,

adsorbed molecules exhibit a strong interaction between their

nitrogen atoms of the iminic nitrogen –CQN– electron lone

pairs and Au(I) at the gold surface. This interaction is highly

reminiscent of the interaction observed between iminic

nitrogen in tetradodecylporphyrins and a Au(111) substrate,19

even though coordination of gold atoms by 1 is not feasible, as

shown next.

Reactivity of compound 1 with gold

Macrocycle 1 readily forms complexes with zinc, cadmium,

and first row transition metals, where the metals are bound to

the N2O2 salphen pockets.21 As it appeared possible that the

macrocycle was coordinating to Au(I) on the surface of the

gold, we attempted to synthesize a complex of Au(I) with

macrocycle 1. Macrocycle 1 was reacted with AuCl in chloro-

form in the presence of NEt3, but no evidence for complexa-

tion to the gold was evident. After 2–4 days, the 1H NMR

spectrum showed only many peaks that are typically observed

during decomposition of the macrocycle. It was not possible to

isolate a gold complex of the macrocycle. This is not surprising;

there are no well-defined salphen complexes of gold in the

literature. Au(III) complexes of salphen form a bridging struc-

ture where the ligand undergoes a large distortion.22 In salen

complexes of Au(III), the gold is accommodated in the ligand,

but only with a large distortion.23 Gold(I) complexes of

salphen are unknown and likely cannot be formed owing to

the large ionic radius of Au(I). In the case of macrocycle 1, the

rigid macrocycle will prevent bulk incorporation of Au(I) into

the salphen pockets. It is clear that the surface chemistry of

macrocycle 1 is distinct from the solution reactivity.

Fig. 2 X-ray photoemission spectrum of the gold spectral region for

a multilayer of compound 1 on a Au(111) surface. The raw data are

plotted in black and the fits to the proposed deconvolution are plotted

in colours. The blue and red curves allowed for the identification of

four peaks: 84.1 and 84.8 eV for Au(4f7/2) and 87.8 and 88.0 eV for

Au(4f5/2).

Fig. 3 X-ray photoemission spectrum of the N 1s spectral region for

a multilayer of compound 1 molecules on a Au(111) surface. The

proposed deconvolution (blue and red curves) allowed for the identifi-

cation of two peaks at 402.9 and 399.7 eV.

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 11941

Molecular dynamics simulations

In order to obtain an atomistic-level interpretation of our

experimental observations, we have performed MD simula-

tions as described in the Methods section.

A first set of molecular dynamics (MD) simulations, denoted

S1A and S1B in Table 1, was performed to establish the role of

solvent in the adsorption of a single molecule of compound 1

on the Au(111) surface. The system studied consists of a single

molecule of compound 1 adsorbed onto a patch of an Au(111)

surface at 20 1C (see Table 1 and the Methods section for

details). Simulation S1A was performed in vacuo and the

simulation S1B was performed in contact with a solution of

n-tetradecane. Typical snapshots of the simulations are shown

in Fig. 4. In the in vacuo simulation (S1A), the chains were

observed to adsorb completely onto the surface, with the

carbon atoms of the chains following the directions defined

by the atoms of the gold surface (see Fig. 4a). In the simulation

S1B (explicit solvent), the chains were observed to be mainly

desorbed and adsorption of n-tetradecane molecules is also

observed (average of 26 n-tetradecane solvent molecules

adsorbed in the space on the surface left by the adsorbed

molecule of compound 1, see Fig. 4). On average, only 3 out of

12 carbon atoms of each alkoxy chain were adsorbed at the

surface. This situation is illustrated in the snapshot shown in

Fig. 4b. A simple interpretation of this result could be that the

surface prefers to adsorb n-tetradecane solvent molecules

instead of alkoxy chains of compound 1.

In order to confirm this interpretation, we have computed

the average potential energy of interaction between the gold

surface and compound 1 in both S1A and S1B simulations and

the average potential energy of interaction between adsorbed

n-tetradecane molecules and the surface. In the case of the

Au–compound 1 interaction, we obtained�611.0� 0.1 kcal mol�1

in simulation S1A and �349.0 � 0.1 kcal mol�1 in simulation

S1B. Therefore, the desorption of alkoxy chains observed in

S4B has an energy penalty of 262 kcal mol�1. This gives a

penalty of about 4.85 kcal mol�1 per desorbed carbon atom

of the chain. In simulation S1B we obtain an average poten-

tial energy of �71.99 � 0.01 kcal mol�1 for the inter-

action between an adsorbed n-tetradecane molecule and the

Au(111) surface, which corresponds to an interaction energy

of �5.15 kcal mol�1 per carbon atom of the n-tetradecane

molecule. Therefore, our simulations predict that the adsorp-

tion of a carbon atom from the n-tetradecane solvent is more

favourable (by 0.3 kcal mol�1) than the adsorption of a carbon

atom of the alkoxy chain. The desorption (observed in simula-

tion S1B) of an average of 54 carbon atoms of alkoxy chains

(9 per chain) per molecule of compound 1 can be compensated

by the adsorption of 50.4 carbon atoms from solvent molecules,

which corresponds to 3.6 n-tetradecane molecules.

A second set of simulations (simulations S4A and S4B in

Table 1) was performed in order to study the influence of the

solvent on the self-assembly pattern of the macrocycles. Again,

we performed simulations both in the presence of n-tetradecane

solvent (S4B) and in the absence of solvent (e.g. strict vacuum

conditions, S4A), as described in the Methods section.

The snapshots from the simulations shown in Fig. 5 illus-

trate clearly the effect of the solvent. Again, in the simulation

with explicit solvent (S4B), the lateral chains of the macro-

cycles are mainly desorbed, whereas in the vacuum simulations

(S4A), chains are mostly adsorbed on the surface (compare

Fig. 4 with Fig. 5). On average, we found that in simulation

S4A, 10.4� 0.1 carbon atoms (among 12) per alkoxy chain are

adsorbed on the Au(111) surface. In the simulations with

explicit solvent (S4B), only 1.4 � 0.1 out of 12 carbon atoms

per alkoxy chain are adsorbed on the surface.

In these simulations we can also obtain information about

the effect of solvent on the spacing of adsorbed molecules.

There are clear differences between simulations S4A and S4B,

as can be seen in Fig. 5. During the simulations, we have

measured the evolution of the lattice vectors a, b and the angle

a as defined in Fig. 1 in the main text. At equilibrium, the non-

solvated structure (simulation S4A) exhibits unit cell parameters

a = 2.48 nm, b = 2.83 nm and a = 1081, significantly different

from the experimental values. On the other hand, in the simula-

tions with n-tetradecane solvent (S4B), we obtained a=2.62 nm,

Fig. 4 Snapshots from MD simulations of a single molecule of

compound 1 adsorbed onto a Au(111) surface under different condi-

tions: (a) simulation S1A (vacuum conditions) and (b) simulation S1B

(n-tetradecane solution). In (b), adsorbed solvent molecules are shown

as wireframe structure for simplicity. The color code is as follows: Au

(yellow), oxygen (red), carbon from the macrocycle core (blue), carbon

from alkyl chains (green), hydrogen (white).

Fig. 5 Snapshots from MD simulations S4A and S4B (see Table 1).

In this figure, (a) and (c) panels correspond to lateral and top views of

a snapshot from the simulations performed in the absence of solvent,

whereas the (b) and (d) panels correspond to simulations in the

presence of n-tetradecane. In (b) all solvent molecules are shown

(yellow translucent lines), whereas in (d) we shown those solvent

molecules adsorbed onto the gold surface.

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online

11942 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 This journal is c the Owner Societies 2012

b = 2.64 nm and a = 1101, which are close to those estimated

from STM images (Fig. 1).

Our simulations also predict solvent co-adsorption. As seen

in Fig. 5, n-tetradecane molecules adsorb on the surface and

play the role of spacers between molecules of compound 1 with

orientations that match the structure of the substrate. In our

simulations, we found an average of 4 n-tetradecane molecules

per adsorbed macrocycle. This result is consistent with our

experimental observations where, although no solvent molecules

can be observed at the interface working under the STM

conditions described in Fig. 1, the measured free unit cell area

(approximately 0.5 nm2) is enough to accommodate the

co-adsorption of n-tetradecane molecules as expected from

our simulations.

3. Conclusions

In summary, combined STM/XPS experiments and theoretical

calculations revealed a complex and subtle interplay of solvent

effects and surface interactions on the 2-D self-assembly

pattern of compound 1.

STM experiments show a self-assembly pattern of compound 1

at the n-tetradecane–Au(111) interface where the triangular

macrocycles orient in the same direction with hexagonal

ordering, avoiding a close-packed structure. The rather open

observed structure is compatible with desorption of the lateral

chains of the macrocycle and solvent co-adsorption, which act

as spacers between adsorbed macrocycles. Such solvent effects

have been confirmed by molecular dynamics. Our simulations

show desorption of the alkoxy chains of compound 1 in the

presence of n-tetradecane. Also, our theoretical calculations

show that adsorption of solvent molecules is energetically more

favourable than adsorption of the alkoxy chains of compound 1.

Accordingly, the unit cell parameters calculated in the presence

of solvent are in very good agreement with the experimental ones

while those obtained in the absence of solvent differ considerably.

The directional adsorption of the macrocycles might be also

influenced by a strong interaction with a possibly recon-

structed gold surface, which promotes higher anisotropy in

the self-assembly of molecules. Such strong interaction with

the Au(111) surface was confirmed by XPS experiments, which

showed a strong interaction of the imine nitrogen atoms with

Au(I) atoms at the surface. Interestingly, no interaction with

gold atoms was observed in solution, confirming the differ-

ential interaction of the macrocycle upon surface adsorption.

It must be also emphasized that even though there is a strong

interaction with the surface, there is no experimental or

theoretical (vide infra) confirmation of molecular deformation

upon adsorption.24

The exact effect of a possible reconstruction of the Au(111)

surface on the self-assembly pattern remains to be understood.

A direct and systematic study of the effect of different recon-

structions of the gold surface on the self-assembly pattern is too

computationally demanding for the MD techniques employed

here. This is due to the need to consider very large surfaces and

consequently too large liquid volumes. For this reason, we are

developing a new simulation procedure combining a simulation

procedure designed to study self-assembly of nanoobjects at

large surfaces25 and implicit solvent methods.

Our results presented here emphasize the need to combine

different techniques to unravel the subtle mechanisms respon-

sible for adhesion and self-assembly at liquid–solid interfaces,

envisaging the possibility to selectively steer their 2-D supra-

molecular organization.

Acknowledgements

D.R.-M, J.F. and T.R. thank the Ministerio de Ciencia for

financial support through projects MAT2009-13977-C03,

FIS2009-13370-C02-02 and CONSOLIDER-NANOSELECT-

CSD2007-00041. T.R. is supported by a JAEdoc contract

co-funded by CSIC and the EU. We also acknowledge the

CESGA Supercomputing Center for computational time and

technical assistance. The authors gratefully thankMs Alejandra

Sanchez for help in XPS analysis. Dr Nathalie Katsonis is also

acknowledged for helpful discussions.

Notes and references

1 (a) F. Zasada, W. Piskorz, S. Godlewski, J. S. Prauzner-Bechcicki,A. Tekiel, J. Budzioch, P. Cyganik, M. Szymonski and Z. Sojka,J. Phys. Chem. C, 2011, 115, 4134–4144; (b) R. Wen, C.-J. Yan,H.-J. Yan, G.-B. Pan and L.-J. Wan, Chem. Commun., 2011, 47,6915–6917; (c) P. Jiang, K. Deng, D. Fichou, S.-S. Xie, A. Nionand C. Wang, Langmuir, 2009, 25, 5012–5017; (d) X. Cui, Z. Wang,S. Tan, B. Wang, J. Yang and J. G. Hou, J. Phys. Chem. C, 2009,113, 13204–13208; (e) T. K. Shimizu, A. Mugarza, J. I. Cerda andM. Salmeron, J. Chem. Phys., 2008, 129, 244103; (f) E. Gomar-Nadal, J. Puigmartı-Luis and D. B. Amabilino, Chem. Soc. Rev.,2008, 37, 490–504; (g) M. In’t Veld, P. Iavicoli, S. Haq,D. B. Amabilino and R. Raval, Chem. Commun., 2008, 1536–1538.

2 (a) M. S. Dyer, A. Robin, S. Haq, R. Raval, M. Persson andJ. Klimes, ACS Nano, 2011, 5, 1831–1838; (b) T. Chen, G.-B. Pan,H. Wettach, M. Fritzsche, S. Hoger, L.-J. Wan, H.-B. Yang,B. H. Northrop and P. J. Stang, J. Am. Chem. Soc., 2010, 132,1328–1333; (c) S.-S. Li, B. H. Northrop, Q.-H. Yuan, L.-J. Wanand P. J. Stang, Acc. Chem. Res., 2009, 42, 249–259; (d) I. Kossev,W. Reckien, B. Kirchner, T. Felder, M. Nieger, C. A. Schalley,F. Vogtle and M. Sokolowski, Adv. Funct. Mater., 2007, 17,513–519; (e) G.-B. Pan, X.-H. Cheng, S. Hoger andW. Freyland, J. Am. Chem. Soc., 2006, 128, 4218–4219;(f) V. Kalsani, H. Ammon, F. Jackel, J. P. Rabe andM. Schmittel, Chem.–Eur. J., 2004, 10, 5481–5492;(g) M. Linares, P. Iavicoli, K. Psychogyiopoulou, D. Beljonne,S. De Feyter, R. Lazzaroni and D. B. Amabilino, Langmuir, 2008,24, 9566–9574.

3 T. Nakamoto, Jpn. Kokai Tokkyo Koho, Patent JP 2007237460,2007.

4 G.-X. Wei, G.-B. Pan, L.-J. Wan, J.-C. Zhao and Ch.-L. Bai, Surf.Sci., 2002, 520, L625–L632.

5 M. Weinhold, S. Soubatch, R. Temirov, M. Rohlfing, B. Jastorff,F. S. Tautz and C. Doose, J. Phys. Chem. B, 2006, 110,23756–23769.

6 (a) L.-M. Liu, Sh.-Ch. Li, H. Cheng, U. Diebold and A. Selloni,J. Am. Chem. Soc., 2011, 133, 7816–7823; (b) Sh.-Ch. Li,Y. Losovyj and U. Diebold, Langmuir, 2011, 27, 8600–8604;(c) Sh.-Ch. Li, L. NaChu, X.-Q. Gong and U. Diebold, Science,2010, 328, 882–884; (d) Sh.-Ch. Li, J.-g. Wang, P. Jacobson,X.-Q. Gong, A. Selloni and U. Diebold, J. Am. Chem. Soc.,2009, 131, 980–984.

7 (a) J. Saiz-Poseu, J. Faraudo, A. Figueras, R. Alibes, F. Busqueand D. Ruiz-Molina, Chem.–Eur. J., 2012, 18, 3056–3063;(b) J. Saiz-Poseu, I. Alcon, R. Alibes, F. Busque, J. Faraudo andD. Ruiz-Molina, CrystEngComm, 2012, 14, 264–271.

8 Y. Kikkawa, H. Kihara, M. Takahashi, M. Kanesato,T. S. Balaban and J. M. Lehn, J. Phys. Chem. B, 2010, 114,16718–16722.

9 A. J. Gallant, J. K.-H. Hui, F. E. Zahariev, Y. A. Wang andM. J. MacLachlan, J. Org. Chem., 2005, 70, 7936–7946.

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 11937–11943 11943

10 J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid,E. Villa, C. Chipot, R. D. Skeel, L. Kale and K. Schulten,J. Comput. Chem., 2005, 26, 1781–1802.

11 (a) A. D. MacKerell, et al., J. Phys. Chem. B, 1998, 102, 3586–3616;(b) A. D. MacKerell Jr., M. Feig and C. L. Brooks III, J. Comput.Chem., 2004, 25, 1400–1415.

12 C. L. Whittington, W. A. Maza, H. L. Woodcock and R. W.Larsen, Inorg. Chem., 2012, 51, 4756–4762.

13 H. Heinz, R. A. Vaia, B. L. Farmer and R. R. Naik, J. Phys. Chem.C, 2008, 112, 17281–17290.

14 V. Rosato, M. Guillope and B. Legrand, Philos. Mag. A, 1989, 59,321–336.

15 C. Mottet, G. Treglia and B. Legrand, Phys. Rev. B: Condens.Matter Mater. Phys., 1992, 46, 16018.

16 W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics, 1996,14, 33–38, 27–28.

17 Z. X. Xie, X. Xu, J. Tang and B. W. Mao, J. Phys. Chem. B, 2000,104, 11719–11722.

18 B. Hulsken, J. W. Gerritsen and S. Speller, Surf. Sci., 2005, 580,95–100.

19 M. M. Maye, Y. Lin, M. H. Engelhard, M. Hepel and C.-J. Zhong,Langmuir, 2003, 19, 125–131.

20 N. Katsonis, J. Vicario, T. Kudernac, J. Visser, M. M. Pollard andB. L. Feringa, J. Am. Chem. Soc., 2006, 128, 15537–15541.

21 (a) A. J. Gallant, J. H. Chong and M. J. MacLachlan, Inorg.Chem., 2006, 45, 5248–5250; (b) T. Nabeshima, H. Miyazaki,A. Iwasaki, S. Akine, T. Saiki, C. Ikeda and S. Sato, Chem. Lett.,2006, 1070–1071; (c) P. D. Frischmann, A. J. Gallant, J. H. Chongand M. J. MacLachlan, Inorg. Chem., 2008, 47, 101–112;(d) P. D. Frischmann and M. J. MacLachlan, Chem. Commun.,2007, 4480–4482; (e) T. Nabeshima, H. Miyazaki, A. Iwasaki,S. Akine, T. Saiki and C. Ikeda, Tetrahedron, 2007, 63,3328–3333.

22 K. S. Murray, B. E. Reichert and B. O. West, J. Organomet.Chem., 1973, 61, 451–456.

23 S. L. Barnholtz, J. D. Lydon, G. Huang, M. Venkatesh,C. L. Barnes, A. R. Ketring and S. S. Jurisson, Inorg. Chem.,2001, 40, 972–976.

24 W. Aurwarter, F. Klappenberger, A. Weber-Bargioni, A. Schiffrin,Th. Strunskus, Ch. Woll, Y. Pennec, A. Rieman and J. V. Barth,J. Am. Chem. Soc., 2007, 129, 11279–11285.

25 T. J. Roussel and L. F. Vega,Nanotechnology: Electronics, Devices,Fabrication, MEMS, Fluidics and Computational, 2011, vol. 2,ch. 9, pp. 579–582.

Dow

nloa

ded

by I

nstit

uto

de C

ienc

ia d

e M

ater

iale

s. B

iblio

teca

Man

on

24 A

ugus

t 201

2Pu

blis

hed

on 1

1 Ju

ly 2

012

on h

ttp://

pubs

.rsc

.org

| do

i:10.

1039

/C2C

P414

07D

View Online