This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

Cite this: DOI: 10.1039/c0cp01516d

Pre-nucleation dynamics of organic molecule self-assembly investigated

by PEEM

Alexander J. Fleming,* Stephen Berkebile, Thomas Ules and Michael G. Ramsey

Received 17th August 2010, Accepted 13th December 2010

DOI: 10.1039/c0cp01516d

Pre-nucleation dynamics, nucleation and templated self-assembly of a conjugated planar aromatic

molecule are investigated by photoemission electron microscopy (PEEM). The high resolution of

individual molecular layers in PEEM, in combination with a numerical simulation, reveals the

dynamic behaviour of molecules during the pre-nucleation deposition period and their

temperature dependence. The in situ deposition of p-sexiphenyl (6P) molecules on Cu(110) and

Cu(110) 2 � 1–O surfaces in ultrahigh vacuum, when monitored by PEEM in real-time allows (a)

layer densities, (b) meta-stable layer filling by 6P molecules, (c) dynamic surface redistributions

during layer filling and (d) critical density spontaneous dewetting to be accurately measured.

The comparison of 6P deposited on Cu(110) to Cu(110) 2 � 1–O enables temperature dependent

6P nucleation processes on Cu(110) to be elucidated from PEEM. The interplay between

energetically stable molecular arrangements and kinetically stabilised arrangements is shown to

dominate the pre- and post-nucleation processes. In combination with additional data obtained

during post-nucleation deposition times, such as surface diffusion anisotropies and nucleation

energies, it is concluded that the pre-requisite for 6P nucleation, in a lying down orientation, is

the formation of a double tilted layer with at least one layer being meta-stable.

1. Introduction

Numerous studies conducted on 6P to date, such as electronic

band formation1,2 and doping,3,4 have demonstrated that this

molecule is exceptionally well suited to be a text book example

of organic molecule phenomenology in the condensed phase.

Our focus recently has been to comprehensively explain

growth phenomena of 6P.5 Having accrued many years of

experience in preparing high quality single crystals of 6P on

surfaces for various electron spectroscopy6,7 and surface probe

techniques,8,9 we have with our detailed in situ growth study

by PEEM, systematically examined the three growth

processes: pre-nucleation, nucleation and 3D growth. Our

novel contribution, to the study of technologically relevant

organic molecule crystal growth for the electronics industry, is

the inclusion of pre-nucleation dynamics in our description of

growth.5

Growing organic electro-active molecular crystals on metal

or metal oxide surfaces is a key process in the construction of

organic electronic devices.10,11 This decides the morphology

and orientation of molecular crystals grown, and can either

enhance or detrimentally affect device characteristics. In

order to study the connection between morphology and

performance, a first step is to grow uniformly oriented and

ordered crystals. To achieve this, molecular crystals are grown

on single crystal surfaces in an ultra-high vacuum (UHV)

chamber. The process of molecular crystal growth can be

observed by various techniques after nucleation has begun,

however in this paper we will highlight the crucial events that

occur prior to nucleation and their importance in determining

final crystal morphologies. PEEM has been used previously to

measure the post-nucleation growth of pentacene12 on SiO2,

chloroaluminium phthalocyanine on MoS2 film,13 anthracence

on Si(111),14 PTCDA on Ag(111),15 to study the dynamic CO

oxidation on platinum16 and other dynamic processes.17 In

our case PEEM is used to examine the pre-nucleation

dynamics of 6P molecules. It will be shown that the

pre-nucleation dynamics does determine the growth and

quality of 6P nano-crystals. In fact, our results indicate that

in order to grow well ordered, unique orientation molecular

crystals,18 as required for organic electronics, a complete

control over the pre-nucleation deposition period will be

necessary. In this respect our current aim is to control and

manipulate this pre-nucleation deposition period.

In this paper, technologically relevant 6P (also known as

hexaphenyl and p6P) molecules are deposited on Cu(110)

2� 1–O and Cu(110) surfaces. The different 6P nanostructures

that form when 6P is deposited on Cu(110) and Cu(110)

2 � 1–O are shown in Fig. 1(a) and (b), respectively. The

deposition characteristics of 6P on these two substrates will be

Surface and Interface Physics, Karl-Franzens Universitat Graz,Universitatsplatz 5, 8010 Graz, Austria.E-mail: sandy_fleming@yahoo.com

PCCP Dynamic Article Links

www.rsc.org/pccp PAPER

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

compared and contrasted with the aim of elucidating the

dominant controlling mechanisms of nucleation and growth

of 6P thin films. Like many weakly bound organic crystals of

short chain-like molecules, 6P readily nucleates into an

energetically stable herring-bone crystalline structure. When

6P nucleates on a surface, a combination of kinetics and

energetics determines which plane of the molecular herring-

bone crystal contacts the substrate surface. In Fig. 2, several

relevant crystal contact planes of a 6P crystal are given. It is

important to note that the molecular orientation with respect

to the crystal plane can be described as tilted, flat, or combina-

tions of tilted and flat. When 6P is deposited on weakly

interacting anisotropic substrates (Cu(110) 2 � 1–O, TiO2

and organic crystals) a tilted wetting layer covers the

surface.2,5,6,9 On top of this tilted wetting layer all subsequent

molecules arrange into crystals with the (20�3) plane contacting

the wetting layer. On the other hand, when 6P is deposited on

stronger interacting substrates8,19 (Cu(110) and Al(111)) a

flat-lying wetting layer covers the surface. On top of this

wetting layer crystals grow with alternate layers arranged in

tilted/flat/tilted intra-layer orientations. This corresponds to

the (21�3) or (�629) crystal plane contacting the flat-lying wetting

layer. Henceforth, 6P nanostructures that form are described

as (21�3) crystals assembling on Cu(110) and (20�3) crystals

assembling on Cu(110) 2 � 1–O.

To study the conditions required for nucleation, 6P mole-

cules are deposited in situ in a PEEM instrument. The

real-time acquisition of PEEM images is used to monitor

precisely (a) the deposition amount, (b) layer filling by 6P

molecules, (c) dynamic surface density redistributions during

layer filling and (d) critical density induced meta-stable layer

dewetting. It is important to note that most of the PEEM data

presented here are acquired during the deposition time prior to

nucleation. It will be shown that, by studying this crucial

pre-nucleation deposition period, the requirements for critical

nucleation can be understood. A numerical simulation of

PEEM photoemission intensity will be shown to help

determine layer filling mechanisms and maximise the amount

of useful data that can be extracted from PEEM. Further-

more, data obtained during post-nucleation deposition allow

nucleation energies to be obtained from temperature

dependent trends of nucleation rates.

It will be shown that although the deposition of 6P on both

substrates results in entirely different molecular crystals

growing ((20�3) versus (21�3)) during the post-nucleation deposi-

tion period, during the time defined by the pre-nucleation

deposition period there are some remarkable similarities.

These include the gradual filling of meta-stable layers to a

similar density of tilted molecules followed by the formation,

in both cases, of (20�3) critical nuclei. Although this may seem

counter-intuitive, it is only after nucleation that differences

start to become apparent. The nucleation induced sponta-

neous dewetting produces two distinct patterns in PEEM

images that reveal differences in surface topology in each case.

The topology can be easily distinguished by comparing the

anisotropy of molecular diffusion on these surfaces. The

temperature dependence of this topological divergence

determines the post-nucleation molecular assembly process

(of the crystals) and is readily observed as a progressive change

to the nanostructure shape. As will be shown, all these pheno-

mena have their origin in the pre-nucleation deposition period.

2. Experimental

The experiments were performed in a custom-designed,

combined Omicron VT-AFM/scanning tunnelling microscopy

(STM)-PEEM UHV instrument with a base pressure o 2 �10�10 mbar. The Cu(110) crystal was cleaned in five steps,

namely (a) flashing of the crystal to above the desorption

temperature of 6P (4220 1C) to reduce the amount of 6P on

the crystal before sputtering, (b) Ar+ sputtering (10 mA,

1.5 kV, 45 min) followed by (c) annealing for 5 min at

500 1C, (d) cooling in oxygen partial pressure at 5 � 10�7 mbar

for 5 min and (e) flashing to 500 1C. Several cycles of the

cleaning steps (b) through to (e) were performed with the final

cleaning cycle omitting (d) and (e). These cleaning steps result

in a Cu(110) surface. To form the 2 � 1 � O reconstruction

50 L of O2 are dosed in the chamber at 2 � 10�7 mbar.

Following cleaning, the sample was transferred, via a

magnetically coupled transfer rod, to the PEEM.

The molecular evaporator, mounted in situ and facing the

sample holder of the PEEM, consists of a Knudsen type

Fig. 1 PEEM image of two types of 6P self-assembled nanostructures

that form when 6P is deposited on (a) Cu(110) and (b) Cu(110)

2 � 1–O surfaces. Typical heights of the nanostructures are in the

range 50–500 nm. Insets: top-view of crystalline structure of the

nanostructures grown and top view of the Cu(110) and Cu(110)

2 � 1–O substrates with arrows indicating surface corrugation

direction. Both surfaces are anisotropic and the surface corrugations

are orthogonally oriented with respect to the other.

Fig. 2 Crystal planes (�629), (20�3) and (21�3) of the 6P herring-bone

crystal. The surface densities required for full-coverage for each plane

relative to the (20�3) crystal plane are given in italic.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

effusion cell resistively heated by current supplied to a filament

by a constant-current source. The 6P in powder form, supplied

by TCI chemicals (Japan) and introduced previously into the

Knudsen cell, is degassed thoroughly by extended heating just

below the sublimation temperature of 6P. The first few 6P

films deposited are not used for measurements. There is no

quartz thickness monitor available due to space restrictions so

no instantaneous deposition rate monitoring is possible.

However, the constant-current supply allows the heating of

the evaporator to be stabilised to a given temperature, which

when combined with the accurate photo-emission intensity

versus time curves (see results) allows for an accurate

a posteriori calculation of the stabilised deposition rate.

Typical deposition rates of molecules are between 0.25 to

1 ML min�1. However, it will be shown that the deposition

rate does not influence the layer filling in the time-scales

explored during deposition. Unless otherwise stated, the

molecular evaporator shutter is open throughout the entire

deposition. After nucleation, and once dewetting is complete,

the shutter is opened and closed periodically to probe mole-

cular surface diffusion.

PEEM measurements were carried out using a FOCUS/

Omicron PEEM instrument in a high magnification mode

using a Mercury HBO 103W/2 discharge lamp (maximum of

hn = 4.9 eV) for illumination at an angle of incidence of 251.

For the molecule/substrate combination used here the image

contrast in PEEM makes use of laterally distributed local

workfunction differences that result in varying yields of photo-

electrons. In the threshold mode, the ionisation potential of 6P

(6.12 eV) is too large for the Hg lamp excitation source

(4.9 eV) thus no photoemission is observed from 6P. However,

the workfunction of the metal substrate is less than 4.9 eV,

therefore photoemission from the metal surface is observed.

Depositing 6P (a dielectric) onto the metal substrate reduces

the surface dipole component of the metal surface workfunc-

tion and hence the photoemission intensity increases. How-

ever, once the first monolayer is complete no more changes to

the surface dipole of the metal occur so any further molecules

deposited do not change the workfunction but instead attenu-

ate the photoemission originating from the metal. The

attenuation is due to inelastic scattering of photoelectrons by

6P molecules (the electronic bandgap isB3.9 eV). An aperture

size of 50 mm together with a fully open iris gave the optimal

resolution for images with a pre-calibrated (in-house) field-of-

view of 20–70 mm. Extractor voltages of 10.5–13.5 kV were

employed, with a sample–extractor distance of 1.8 mm. No

adverse effects, such as decomposition or desorption, arising

from the UV light or high electric field were observed for 6P on

Cu(110). Photoelectron intensities plotted in the Results

section were acquired directly from intensity averaged areas

of the PEEM images. Area-averaging of the intensity works

well up to the point when dark structures form on the surface

(nucleation). Once dark structures start growing, area-

averaging the intensity to determine the surface molecular

density will become distorted by the coverage of three dimen-

sional dark structures. At 140 1C the ratio of dark structures to

bright area is small; therefore the error introduced is also

small. A video camera exposure time of 500 ms at an acquisi-

tion rate of 2 Hz was found to be adequate to monitor

dynamic processes on the surface. In situ sample heating was

achieved by indirect heating with a filament following a

pre-determined and calibrated temperature–time curve.

To observe dynamic changes in surface molecular density

during the pre-nucleation deposition period, some post-

processing of PEEM images is required. First all PEEM

images of a video sequence are normalised to the average

intensity of the first image (clean surface before deposition)

and batch processed to produce a difference image sequence

with respect to the first image. Furthermore, any contrast

enhancement observed can be quantised by batch processing

the PEEM image sequence to determine the intensity rms

deviation for each image and plot it versus deposition time.

3. Results and discussion

Highly ordered uni-axially aligned molecular crystallites form

when 6P molecules are deposited on Cu(110) and Cu(110)

2 � 1–O. The templating of the first molecular layer by the

metal surface corrugation into an ordered and aligned layer

leads to anisotropic crystallite growth and anisotropic surface

diffusion of molecules. Shown in Fig. 1 are typical examples of

the different nanostructures that form when 6P is deposited on

Cu(110) and Cu(110) at 140 1C. Straight (20�3) needles form

when 6P is deposited on Cu(110) 2 � 1–O, whereas crossed

(21�3) needles form when 6P is deposited on Cu(110). It is the

binding/sticking anisotropy of plano-linear 6P molecules

which results in a preferred co-facial stacking. Hence upon

stacking, crystallites grow perpendicular to the molecular

orientation within the molecular crystal. As will be shown

later, this growth direction is also perpendicular to the fastest

molecular diffusion direction.

In order to explain the behaviour of 6P pre-nucleation,

nucleation and growth, general observations of the combined

6P/substrate system are presented first. These include compar-

isons of the photoemission intensity evolution with deposition

time and general observations, such as condensation at steps

and dewetting, that distinguish the behaviour of 6P deposited

on Cu(110) 2 � 1–O from 6P deposited on Cu(110).

Subsequently, a description and explanation of the tempera-

ture dependence of various phenomena related to 6P deposited

on Cu(110), that are not observed for 6P deposited on Cu(110)

2 � 1–O, are presented. To do this, a simulation of PEEM

photoemission intensity is constructed to interpret the

temperature dependence; STM images and measurements of

the molecular surface diffusion are also included. Finally, the

nucleation energies of both systems are compared and

contrasted.

3.1 Photoemission intensity evolution of 6P deposited on

Cu(110) and Cu(110) 2 � 1–O

From dynamic PEEM data and other static techniques5,6,8 6P

deposited on Cu(110) 2 � 1–O is known to form a tilted first

layer (a stable wetting layer) followed by a tilted meta-stable

second layer that spontaneously forms (20�3) nuclei when a

critical surface density (at B1.95 ML) is reached. Upon

nucleation, dewetting of the metastable layer proceeds

whereby 6P molecules diffuse to the nuclei, and the nuclei

grow into long needles. On the other hand, for 6P deposited on

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

Cu(110) the information available thus far is limited. Several

static techniques, applied after nucleation, have determined

that 6P forms a flat-lying first layer (a permanent wetting

layer) and a static second layer with three-dimensional

crossed-needles growing on top. X-Ray diffraction studies

indicate that the nanostructures have a (�629) structure when

grown at RT. However, no studies have examined the

pre-nucleation deposition period, hence the motivation for

this work.

Typical area averaged photoemission intensity versus time

plots which correspond to 6P deposited on (a) Cu(110)

2 � 1–O and (b) Cu(110) are given in Fig. 3. Both plots

undergo a similar general trend which involves (1) a sharp rise

in intensity, (2) a maximum with a subsequent decrease in

intensity (this decrease is more prolonged for 6P on Cu(110)

since more molecules must be deposited), and (3) upon

reaching a critical deposited amount, a sharp cusp followed

by a constant gentle rise in intensity thereafter. The transition

from increasing to decreasing photoemission intensity upon

completion of the first layer is a general characteristic of

photoemission obtained in the threshold regime.

To facilitate comparison, the photoemission intensity

curves, shown in Fig. 3, are double normalised to the

maximum intensity and to one complete monolayer as

explained in more detail by Fleming et al.5 To facilitate

interpretation of the double normalised plots, the inset in

Fig. 3 is a plot of the same data where the normalised

photoemission intensity is plotted versus the total surface

density of molecules deposited. This better illustrates how

more material must be deposited to form a tilted first layer

on Cu(110) 2 � 1–O compared to a flat-lying first layer on

Cu(110). A noticeable difference between the evolution of the

photoemission intensity for 6P on Cu(110) and 6P on Cu(110)

2 � 1–O up to 1 ML deposited material is the presence of a

re-arrangement kink at B0.2 ML in the double normalised

plot for 6P deposited on Cu(110) 2 � 1–O. A kink describes a

sudden change in the rate of intensity increase/decrease with

deposition time. Thus during the initial deposition of 6P on

Cu(110) 2 � 1–O, PEEM correctly indicates that the dilute

layer undergoes a spontaneous re-arrangement from flat lying

to tilted 6P orientation. This agrees with the observation, from

STM8 and LEED, that the 6P wetting layer molecules are

tilted at an angle of 371 with respect to the surface. It can

therefore be deduced from the monotonic intensity increase

during the initial deposition of 6P on Cu(110) (i.e. without a

re-arrangement kink between 0 ML and 1 ML) that molecules

in the wetting layer remain flat-lying. This is supported by

STM8 and valence-band photoemission1,6 of the full-coverage

wetting layer which shows that the wetting layer consists of

flat-lying 6P molecules strongly bound to the surface. Thus,

unlike with Cu(110) 2 � 1–O, no change in orientation is

observed nor expected for the first layer deposition.

During second layer filling, on the other hand, photo-

emission intensity trends of both systems are not monotonic.

The re-arrangement kinks, observed at B1.2 ML for 6P

deposited on Cu(110) 2 � 1–O and at B1.1 ML for 6P

deposited on Cu(110), indicate that re-arrangements of dilute

second layer molecules occur. The implication is that for both

systems the second layers are initially filled with flat-lying

molecules which, at a certain surface density, tilt to eventually

form a densely packed layer of tilted molecules. This filling

process is temperature independent for 6P deposited on

Cu(110) 2 � 1–O and it proceeds until a maximum surface

density is reached (when the tilt angle of the second layer

reaches B331). At this maximum surface density, critical

nucleation occurs, followed by dewetting.

For 6P on Cu(110) on the other hand, neither nucleation

nor dewetting occurs during second layer filling. Instead layer

filling continues uninterrupted up to B3.8 ML. However,

there is a significant kink at 2.4 ML. This is not a re-arrangement

kink as described above, rather it indicates when a layer is

completed and next layer filling commences (the temperature

dependence of this kink is described in detail later). The

surface density of a full-coverage 6P tilted layer (at 331) is a

factor B1.4 greater than the density of a full-coverage

flat-lying layer (at 01). Hence the 1.4 factor difference in the

density between a flat-lying and a tilted layer should be

taken into account when interpreting the time evolution of the

averaged area photoemission intensity for coverages 41 ML

in the double normalised plot. This means that the normalised

time required to fill identical tilted layers will be 1.4 ML for 6P

deposited on Cu(110) and 1.0 ML for 6P deposited on

Cu(110) 2 � 1–O. As can be seen from the plot, there is a

kink in the intensity evolution at B2.4 ML. This kink at

2.4 ML indicates that a second layer, composed of tilted

molecules, is completed. More significantly, the important result

is that both second layers, on Cu(110) 2 � 1–O and on Cu(110),

have the same density (as can also be seen in the inset in Fig. 3).

Once the tilted second layer is complete a subsequent third

layer begins to fill for 6P deposited on Cu(110). This layer, in

common with the second layer, also has a re-arrangement kink

located at B2.5 ML which indicates a change from flat-lying

to tilted orientation. As can be seen in Fig. 3, third layer filling

continues until a critical density is reached and spontaneous

nucleation occurs. For 6P deposited on Cu(110), the critical

surface density is reached at 3.8 ML (normalised time) which

Fig. 3 Normalised area-averaged photoemission intensity versus

normalised time for 6P deposited on Cu(110) 2 � 1 � O at 140 1C

(black line) and on Cu(110) at 125 1C (grey line). Inset: same data

re-plotted with normalised intensity as a function of total surface

molecular density.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

corresponds to two completed tilted 6P layers (1.4ML+1.4ML)

deposited on top of a flat-lying first layer (1 ML). Thus critical

nucleation occurs at the point in time when the third layer

is just completed. Similarly, for 6P deposited on Cu(110)

2 � 1–O, critical nucleation begins when the second layer is

just completed at B2 ML. Note that the second and third

layers on Cu(110) and the second layer on Cu(110) 2 � 1–O,

all have the same density. This will be shown later to be the

deciding factor that determines what crystal structure critical

nuclei have.

To summarise, comparison of area-averaged photoemission

intensity versus deposition time curves for both substrates

indicates that more 6P must be deposited on Cu(110) to

initiate critical nuclei formation. 6P deposited on Cu(110)

2 � 1–O fills two layers with tilted-lying-down molecules:

the first layer is a permanent tilted wetting layer and the

second is a tilted layer. This second layer is meta-stable since

it dewets upon critical nucleation. On the other hand, 6P

deposited on Cu(110) fills three layers: the first is a flat-lying

permanent wetting layer and the second/third layers together

form a tilted-lying-down double layer. This double layer

will be also shown to be meta-stable since it partially dewets

upon critical nucleation to form a permanent second layer

reconstruction. Interestingly, the photoemission intensity

curves also show that at 125 1C all tilted layers on any of

the systems have approximately the same density.

3.2 Dynamic surface density re-distributions during deposition

of 6P on Cu(110) and Cu(110) 2 � 1–O

In this section a brief introduction to other phenomena,

observed during the pre-nucleation and post-nucleation

deposition period, which highlight the differences between

the two systems are presented and discussed. For instance,

when 6P is deposited on Cu(110) an enhanced contrast of

Cu(110) crystal steps and terraces is observed that does not

occur when 6P is deposited on Cu(110) 2 � 1–O. Further

differences, observed during the spontaneous nucleation

induced dewetting, that draw attention to the different surface

topologies during dewetting are also given. These phenomena

also have a temperature dependence that is discussed in more

detail later.

3.2.1 6P condensation at steps during pre-nucleation deposition

period for 6P on Cu(110). In situ real-time monitoring by

PEEM enables surface density re-distributions to be moni-

tored during pre-nucleation and post-nucleation deposition

periods. The difference image sequence shown in Fig. 4 (A–F)

for 6P deposited on Cu(110) reveals large changes in PEEM

image contrast which indicate that 6P re-distributes on the

surface during the pre-nucleation deposition period. The

contrast enhancement observed, quantised by batch

processing, is plotted versus normalised deposition time and

displayed in tandem with the corresponding photoemission

intensity curve in Fig. 4(b). Large variations in image contrast

for the different layers are explained by variations in molecule

to surface interactions during this pre-nucleation deposition

period. The surface on which molecules diffuse and condense

changes depending on which layers are completed. This

explains the varying behaviour as a function of deposition

time in Fig. 4(b). For deposition times o 1 ML 6P molecules

diffuse on Cu(110) whereas 41 ML 6P molecules diffuse on

either a flat-lying (first layer) or a tilted layer (second and third

Fig. 4 (a) Difference image sequence (A to F) for 6P deposited on Cu(110) that illustrates the contrast enhancement provided by 6P condensation

at steps during second layer and third layer deposition. (b) Combined plot of area-averaged photoemission intensity for 6P on Cu(110) deposited at

125 1C (grey) and corresponding area-calculated rms intensity deviation plotted versus normalised deposition time in monolayers (black). Note

that (A to F) correspond to the image sequence. (c) Schematic diagram of condensation at steps starting from the bottom of the step (condensation

starting from top of step is also possible but not drawn) for A, C and E.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

layers) of 6P. No 6P condensation at steps is observed when

deposited on Cu(110) 2 � 1–O.

Examining Fig. 4(b) to compare the spatial rms intensity

versus the area-averaged photoemission intensity plots reveals

some trends. Namely, spatial rms intensity minima observed at

1 ML, 2.4 ML and 3.8 ML coincide in time with the

completion of each successive layer (these are indicated by

B, D and F, respectively, in Fig. 4). Reciprocally, spatial rms

intensity maxima coincide, with the exception of the first layer,

with the point in deposition time when a layer is exactly half

filled (indicated by C and E in Fig. 4). Thus when the

condensation at steps reaches 50% coverage of terraces the

contrast observed is maximised, whereas 50% coverage with

no condensation at steps, whereby molecules are distributed

homogeneously over the surface results in no contrast enhance-

ment. Likewise 100% coverage of terraces (completed layers)

results in no contrast enhancement, but serves to confirm that

the normalised deposition times for layer completion (1 ML,

2.4 ML and 3.8 ML), extrapolated from the area-averaged

photoemission intensity curves in Fig. 3, are correct. The

reasons why condensation at steps varies during deposition

are explained next.

For depositions times r 1 ML 6P molecules interact

strongly with the metal surface when 6P is deposited on

Cu(110) and thus adopt a flat-lying orientation. This strong

binding interaction with the surface prevents a strong binding

interaction between molecules (there is a minimal inter-

molecular orbital overlap in the flat-lying configuration)

so no preferred condensation at steps is observed. This is

illustrated in Fig. 4 A.

On the other hand, 6P molecules filling the second layer on

Cu(110) do not interact with the metal surface and instead

experience weaker van der Waals interactions with flat-lying

molecules from the first layer. It is this relative weakness which

allows molecules to re-orient into a tilted orientation. The

process of tilting is kinetically driven in the presence of a high

surface density of molecules and is explained later in more

detail. Since condensed and tilted 6P molecules on top of the

flat-lying layer present a hydrogen terminated surface towards

the aromatic plane of the first layer, as shown in Fig. 4 C, the

intra-layer inter-molecular overlap of the second layer is

comparatively greater and hence the van der Waals interaction

within the second layer will dominate. It is important to note

that the flat-lying layer also presents a relatively small surface

topology corrugation compared to the bare substrate. This

combination facilitates the 6P condensation at steps process

since molecules will diffuse on terraces and collect at stronger

binding sites such as steps. It should be noted that condensa-

tion at steps does not produce a bulk type crystal arrangement

(in fact the layers are meta-stable), but it is nevertheless

energetically favourable during the pre-nucleation deposition

period to condense.

From Fig. 4 it is clear that third layer condensation at steps

produces a weaker contrast enhancement than the second

layer. Although condensed molecules in the third layer also

present a hydrogen terminated surface to the second layer,

the matching surface topology (they can adopt a 1 : 1

commensurate spacing) to the second layer hydrogen

termination results in molecules condensing anywhere on the

terraces including at step edges. Thus this more sympathetic

interaction leads to only a partial condensation at steps for the

third layer. It should be noted that due to the differential layer

photoemission attenuation, the maximum contrast enhance-

ment from the third layer condensation at steps will be half

that of the contrast enhancement of the second layer. Apart

from confirming some of the conclusions garnered from

photoemission intensity curves, the agreement between the

positions of rms minima and layer completion kinks indicates

that at 125 1C the layer filling mode, in the pre-nucleation

deposition time, is layer-by-layer.

3.2.2 Spontaneous dewetting during post-nucleation

deposition period. Thus far, it has been shown that compar-

isons between the intensity evolutions of 6P on Cu(110)

2 � 1–O and 6P on Cu(110) allow pre-nucleation layer filling

mechanisms to be established. The post-nucleation deposition

period can also be studied. When 6P is deposited on Cu(110)

2 � 1–O, second layer molecules form a meta-stable layer that

spontaneously dewets at a critical density to form 6P (20�3) 3D

needles.5 No condensation at steps is observed during deposi-

tion of the second layer, since in this case the layer is in a

homogenous liquid state up until a critical surface density is

reached at B1.95 ML. At this critical density, (20�3) nuclei

form and spontaneous dewetting of the layer to form 3D

nuclei proceeds. As described above, the deposition timeline

for 6P deposited on Cu(110) is different—layer filling

continues up to 3.8 ML (one flat and 2 tilted layers); upon

when spontaneous nucleation induced dewetting occurs. It will

be shown later that unlike with 6P deposited on Cu(110)

2 � 1–O, the (20�3) critical nuclei formed when 6P is deposited

on Cu(110) do not determine the final structure of the nano-

structures that grow. This is because the process of dewetting

leads to a molecular layer reconstruction, of areas surrounding

critical nuclei, which changes the favourable crystal structure

to self-assemble from (20�3) to (21�3) for temperatures 4 RT.

Shown in Fig. 5(a) and (b) are post-nucleation PEEM

images of 6P deposited on Cu(110) and Cu(110) 2 � 1–O,

respectively. In both cases, 6P crystals that form are darker in

PEEM than the surrounding area. This is due to their 3D

Fig. 5 Anisotropic dewetting of 6P from (a) meta-stable second and

third layers of 6P on Cu(110) to form crossed-needles and (b) meta-

stable second layer of 6P on Cu(110) 2 � 1–O to form needles. Areas

with lower surface density of 6P molecules appear bright since there is

less attenuation of the photoemission from the Cu(110) 2 � 1–O or

Cu(110) surface beneath. The fastest diffusion direction is along the

Cu(110) 2 � 1–O or Cu(110) surface corrugations. Note that in both

cases, the permanent first layer never dewets.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

structure of multiple 6P layers which, in total, attenuate

secondary photo-electrons very strongly. Notice that areas

immediately surrounding nuclei are brighter than the rest of

the image. These bright areas indicate regions of low surface

density where molecules have dewetted the surface by diffusing

away (in the presence of a concentration gradient) to nuclei.

The shape of these low surface density regions is determined

by how anisotropically 6P diffuses on the surface. As can be

seen in Fig. 5(a), the elongated bright areas are clear evidence

of anisotropic diffusion during the process of nucleation

induced spontaneous dewetting. In Fig. 5(b) the elongation

is not as pronounced, however, in both cases the direction of

fastest dewetting is perpendicular to the fastest growth

direction of the nanostructures that form and is parallel to

the surface corrugation that templated the wetting layer

(shown in Fig. 1). In Fig. 5, the dewetting process occurs on

a much larger length scale for 6P on Cu(110). It is highly

anisotropic (410 : 1) and produces long thin streaks in the

[1�10] direction that grow wider in tandem with the length

increase of the nanostructures. It is also important to note that

the boundary of the dewetted region is sharp in the [001]

direction and smeared out in the direction parallel to the fast

diffusion direction [1�10]. 6P deposited on Cu(110) 2 � 1–O, on

the other hand, produces a less anisotropic (roughly 4 : 1)

spontaneous nucleation dewetting bright area with sharp

boundaries in all directions. It will be shown later that the

high anisotropy observed during dewetting on Cu(110) is due

to molecules diffusing on a reconstructed layer that channels

molecules towards the nuclei. On the other hand, 6P

molecules, during dewetting on Cu(110) 2 � 1–O, diffuse on

an un-reconstructed, tilted layer, which weakly confines

molecules into anisotropic diffusion.

3.3 Temperature dependence of 6P deposited on Cu(110)

In this section, temperature dependences of the dynamic

behaviour of 6P during pre-nucleation and post-nucleation

deposition times are examined. It will be shown that by

constructing a simulation of photoemission intensity curves

and comparing it to experimental data for several deposition

temperatures, the layer filling scheme can be sub-divided into

two growth modes: layer-by-layer or hole creation. It is then

postulated that a single mechanism involving hole creation

and filling can explain both growth modes. This is confirmed

by analysing the temperature dependence of the condensation

at steps during second and third layer filling. Furthermore, by

comparing surface densities required to complete a layer, it is

postulated that at all temperatures (20�3) critical nuclei form

when 6P is deposited on Cu(110). Although this would seem to

contradict the observation of (21�3) crossed needles, analysis by

PEEM and STM of the nature of the top surface remaining

after dewetting (the molecular reconstruction) shows that

there are highly anisotropic channels that are formed by a

mixture of tilted and flat molecules. This mixed orientation,

also present in the crystal structure of (21�3) and (�629) planes, is

what encourages growth of crossed needles. Finally, the

postulate that (20�3) critical nuclei can give rise, via a

reconstruction of the surrounding area, to crossed-needles is

confirmed by PEEM images of crossed needles exhibiting

additional (20�3) straight needles. Further confirmation is

provided by the fact that nucleation energies for 6P on both

substrates are nearly identical.

Area-averaged photoemission intensity versus deposition

time plots, for 6P deposited on Cu(110) at various substrate

temperatures ranging from RT to 150 1C, are shown in Fig. 6.

The temperature dependence of the amount of material

deposited that is needed to begin next layer filling is estab-

lished from the varying positions of layer completion kinks

(between 2.4 and 3 ML) and cusps (between 3.8 and 5.3 ML)

along the normalised time axis. For the temperature range

50–125 1C the second layer is completed at 2.4 ML and

nucleation begins at 3.8 ML as described in the previous

section. At RT or 140–150 1C however, more material must

be deposited to complete the second layer and to initiate

nucleation.

As can be seen from Fig. 6, the variability of the photo-

emission intensity evolution beyond 1 ML is not straight-

forward. To determine what processes are involved a model is

required that connects individual layer densities of second,

third, fourth and higher layers to the total transmitted photo-

emission intensity. However, since a finite deposited amount

can be arranged into several configurations, each of which is

described by a sum of several layer densities, the model must

connect observed photoemission intensities to a narrow range

of possible configurations. Although these configurations are

averaged in x and y dimensions, the high sensitivity of PEEM

in threshold regime allows configurations to be resolved in the

z-dimension. This allows the presence of holes that may form

in each layer of a given configuration to be evaluated

from averaged photoemission intensities measured. However,

from photoemission intensities alone it is not possible to

distinguish low density crystalline planes which contain flat

and tilted molecules (e.g. the (21�3) plane) from morphological

holes in layers of closed structure crystalline planes such as

(20�3) where all molecules are tilted. Nevertheless, positions of

Fig. 6 Normalised area-averaged photoemission intensity versus

normalised deposition time for 6P deposited on Cu(110) 2 � 1–O, at

various temperatures. Note the variability in kink position of the

curves. Also note that differences observed in the onset of increasing

photoemission (o1 ML) do not account for the variability in kink

positions.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

layer completion kinks and cusps, which have been shown to

indicate how much deposited material is required to complete

a layer, together with the simulation of photoemission

intensity, can be used to establish which crystalline plane fills

a layer and which morphological phenomena are involved.

3.3.1 Simulation of PEEM intensity time evolution. To

construct the model several factors must be considered,

including (a) number density of molecules on the surface per

unit area, (b) number of molecular layers, (c) lateral resolution

of the PEEM instrument, (d) PEEM contrast regime and (e)

molecular inter-layer rearrangement mechanisms. The PEEM

instrument extracts photoelectrons from the substrate surface,

focuses them onto a micro-channel plate that increases the

gain of the photoelectron signal and finally produces an image

on a phosphorescent screen. PEEM has a resolution that is

insufficient to resolve individual molecular sites on the surface,

thus final images are area-averaged and hence our model

requirements are greatly simplified. The strategy is to evaluate

the average photoemission intensity for each possible

configuration that is realistically attainable via molecular

re-arrangement for a given number density of molecules per

unit area. Since there aren’t any non-intrusive experimental

techniques that can laterally resolve individual molecule

dynamic re-arrangements, this PEEM simulation is a worth-

while approach. The reference molecular density is taken to be

the density of a (20�3) full-coverage tilted layer. In view of the

fact that our aim is to interpret trends in area-averaged

photoemission intensity versus deposition time plots, such as

in Fig. 6, a simple 3D layered 3 � 3 grid will provide sufficient

resolution to compare to the experimental data when small

amounts between 1–4 ML of material are deposited. Note that

only the second layer and above are considered in this model

since molecules of the first layer are strongly bound to the

Cu(110) surface.

Shown in Fig. 7 is the multilayered N � N grid that is used

to construct the simulation. Each molecule occupies one site

(i, j) on the grid (no double occupancy is allowed) and

molecules are moved between layers L, by subtracting a

molecule from a layer (creating a hole in that layer) and

adding the molecule on to a different layer (creating a

‘‘tower’’). Occupancy of each available site is given by

Si,j,L = 1. For unoccupied sites Si,j,L = 0. Furthermore,

several constraints are required. During re-arrangement the

total number of molecules must be conserved. Molecules can

only occupy a higher layer if equivalent sites in the layers

beneath are occupied too—this prevents evaluating states with

unphysical ‘‘floating’’ molecules. Only structures such as

stepped pyramids and cuboids are considered. Arches and

other similar structures which violate the above rules are

excluded. Thus, the constraint Si,j,L Z Si,j,L+1 is applied in

the simulation during re-arrangement. In order to model the

layer filling as measured by PEEM, deposition quantities are

increased in 1/N2 ML* steps and at each deposition step the

photoemission intensity of each possible re-arrangement is

calculated. Note that for the purposes of the model ML*

refers to a (20�3) layer of tilted molecules (1 ML* = 1.4 ML).

The model will be re-normalised later to the flat-lying wetting

layer in order to compare to PEEM data.

In our case deposition is simulated as follows: using a 3 � 3

grid, at every 1/9 ML* step increase in number density of

molecules, all possible configurations of molecules on the grid,

following the rules above, are determined. As an example, all

possible configurations attainable via re-arrangement for

4 molecules on a 3 � 3 grid are given in Fig. 7(d). The various

configurations lead to different area-averaged photoemission

intensities for a given deposition amount. For each possible

configuration the normalised area-averaged photoemission

intensity of the 3 � 3 grid is calculated. To evaluate photo-

emission intensities in the photoemission attenuation regime,

the total transmission probability per site is calculated as a

product of transmission probabilities per layer Ti,j,L for each

site (i, j). The total transmission probability per site is then

averaged over the 3 � 3 grid to obtain the normalised

area-averaged photoemission intensity. More generally the

expression for photoemission intensity for every 1/N2

deposition step is given by,

IPEav ¼1

N2

XNi

XNj

YLmax

L¼1Si;j;LTi;j;L

" #

The uppermost layer that has an occupied site is Lmax (L = 1

is the flat-lying first layer that is excluded from the model).

Values evaluated for Lmax that lie outside the range of

experimentally obtained PEEM data intensity values are

excluded. These configurations are usually unrealistic struc-

tures that would produce higher photoemission intensities

than are observed and thus are not required for the simulation.

For example one of the various configurations in Fig. 7(d),

where 4 molecules are towered on top of a complete first layer

(i.e. Lmax = 5), produces a spike. Such a configuration is far

less likely to form than a shape with a complete first layer and

4 molecules in the second. In terms of the minimisation of

surface free energy principle, dewetting a surface to expose the

underlying surface in order to create a spike with a large

surface area to volume ratio leads to a large, and unfavour-

able, increase in total surface free energy.

The results of the PEEM simulation, re-normalised by

setting 1 ML* = 1.4 ML (for deposition times 41 ML), are

given in Fig. 8 (grey squares). The simulation data can be

Fig. 7 (a) One of many possible configurations of 12 molecules on the

surface distributed over 3 layers. (b) Total intensity transmission per

site (i, j) through the 3 layers. (c) The resultant area-averaged

transmitted intensity. (d) Intensity in-equivalent structural configura-

tions possible with 4 molecules deposited (the white 3 � 3 grid in this

case represents the permanent first layer). Note: the maximum layer

number (Lmax) with non-zero occupancy increases from 2 to 5 from

left to right.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

broken up into two kinds: minimum intensity data points

which correspond to layer-by-layer filling and all other data

points of higher intensity that correspond to layer filling with

hole creation. At 1 ML the transmitted intensity is maximal as

there are only holes in the second layer fully exposing the

L = 1 flat-lying wetting layer beneath. Following the lowest

intensity simulation data points, for each deposition step

between 1 ML and 3.8 ML, it is clear that the photoemission

intensity decreases linearly within a layer and exponentially

between layers. These lowest intensity simulation data points

correspond to the case where no towering occurs, and hence

represent layer-by-layer filling. The layer completion kink

observed at 2.4 ML and 3.8 ML for layer-by-layer filling

and higher intensity data points in the simulation can be

explained together since they have a common origin. For

instance, when towering occurs during second layer filling

the number of holes in the second layer increases as site

occupancy in higher layers increases. When averaged, the

differential intensity change upon towering leads to an increase

in photoemission intensity for a given deposition amount. The

reason is that creating n unoccupied sites in the second layer

via re-arrangement will increase the area-averaged trans-

mission by (1/9) per unoccupied site. However there is a

counter balance, since to conserve the total number of sites

occupied, n molecules must be towered somewhere on the

surface. In the case of Lmax = 3, the area-averaged trans-

mission will decrease by (1/9)2 per tower created in this layer.

The averaging of these differential intensity changes leads to a

small rise in photoemission intensity. The various area-

averaged intensities of the different configurations possible

for every deposition step are plotted in Fig. 8. As can be seen

in Fig. 8, there is a distinct change in the rate at which the area-

averaged photoemission intensity varies when the deposited

amount increases linearly. This produces the characteristic

layer completion kinks described previously during the

transition from only filling the second layer to only filling

the third layer at 2.4 ML (this is confirmed experimentally by

the condensation at steps in the previous section). The pre-

sence of a kink in the simulation is therefore an indication of a

transition to filling the next layer; however this is not limited to

layer-by-layer filling since it can still occur when the hole

density of a particular layer is held constant.

3.3.2 Comparison of PEEM simulation to experiment. The

main hypothesis of the simulation is that molecules attenuate

equally no matter which layer (41 ML) they are in. The inset

in Fig. 8 shows the photoemission intensity as a function of

layer distance and it clearly follows a 1/2L dependence which

indicates that molecules in each layer above the first layer

attenuate equally and therefore no change, as a function

of distance to the surface, influences the photoemission

attenuation. In addition, no change is observed in ARUPS

workfunction measurements after the wetting layer is deposited.12

From inspection of Fig. 8 there is a good fit between the

photoemission intensity curve at 115 1C (which is representa-

tive of depositions between 50–125 1C) and the simulation

data points which correspond to configurations where the

number of holes is kept to a minimum (the lowest intensity

data points for each deposition point in the timeline). There is

an excellent fit to slopes between 1–3.8 ML and kink positions

at 2.4 ML and 3.8 ML (whereupon fourth layer filling would

begin if spontaneous nucleation dewetting did not occur). This

indicates that at these deposition temperatures layer filling

proceeds layer-by-layer and thus the hole density in all layers is

zero. Conversely, at RT the photoemission curve follows a

different path (this is also representative of the 140–150 1C

depositions). This indicates that at these temperatures layer

filling does not proceed layer-by-layer. Instead, the simulation

indicates that curves proceed along a path that creates a

non-zero hole density in layers during layer filling.

Photoemission intensity variability. Higher photoemission

intensity during deposition of second and third layers at RT

and 140–150 1C, and delayed layer completion kinks and cusps

confirm that some competing processes occur that are

cancelled out at deposition temperatures in the range 50–125 1C.

These processes must be temperature activated but also must

compete to produce variability in layer filling rates. We

propose that hole filling and hole creation fulfil these criteria.

At RT molecules will diffuse slowly on the surface and will

thus slowly fill naturally occurring holes that are created by

random spatial distributions of molecules arriving from the

evaporator. Note that due to the flat-lying first layer, second

layer holes are temporarily filled by molecules in a flat-lying

orientation. Lower temperatures slow the rate at which

flat-lying molecules re-tilt to join tilted parts of the layer;

and hence result in longer waiting times before tilted molecules

can occupy the same site (this will also prevent lateral growth

of sub-critical nuclei). On the other hand, between 140–150 1C

molecular diffusion is much faster so holes are filled much

more quickly. However high temperatures lead to more hole

creation since, in the context of a fixed intermolecular binding

energy, according to Boltzmann statistics, more molecules will

have enough energy to de-bind and therefore diffuse elsewhere.

Fig. 8 PEEM data for RT and 115 1C depositions overlaid on top of

PEEM simulation. Note that straight line segments of experimental

photoemission curves indicate that the deposition rate is constant

during deposition. Inset: photoemission intensity at kink positions for

some photoemission curves corresponding to layer-by-layer growth

plotted versus layer number (ML*). A 1/2L fit is also given and plotted

up to 6 ML*.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

Higher temperatures also shorten the time that a flat-lying

molecule occupies a hole by increasing the re-tilting rate, and

thus encourages the formation of kinetically stabilised tilted

layers. In the temperature range 50–125 1C the rate of hole

filling equals the rate of hole creation and thus they cancel out.

In this case, only the rate of deposition will determine the rate

of layer filling. However the rate of deposition does not

influence the filling mechanism itself. This becomes apparent

when comparing the absolute deposition rates in the

50–125 1C temperature range: all three depositions have

differing deposition rates, with the 115 1C deposition rate

roughly a factor 3 faster than the others.

For RT and 140–150 1C depositions, the PEEM simulation

indicates that hole creation and filling can explain the differing

behaviour. A hole in the simulation represents a lower density

per unit area than for the same area filled with tilted molecules,

only if the total number of molecules is conserved by moving

excess molecules to another layer. Thus configurations B and

C, shown in Fig. 9(a), both satisfy the condition for a hole in

the model. Areas of flat-lying molecules, areas with small

packing angles o 331 surrounded by tilted regions with

packing angles of 331 or regions with configurations such as

tilted/tilted/flat/tilted/tilted/flat can all equally represent a hole

in the simulation so long as excess molecules are transferred to

other layers. Hence there is a difficulty in distinguishing

between morphological holes (Fig. 9(a) B) and structural holes

of a crystal plane (Fig. 9(a) C).

By comparing the simulation data to PEEM data in Fig. 8 it

is possible to extract a symbolic representation of filling

schemes for growths at all temperatures. In Fig. 9 two

examples are given (b) layer-by-layer growth observed in the

temperature range 50–125 1C and (c) hole/tower creation

during growth at RT. The layer density is represented

pictorially by the number of boxes per layer out of a maximum

of 9 spaces available. As can be seen layer-by-layer growth is

straightforward, however the RT growth filling is more

complex. Nevertheless the principle behind the filling mecha-

nism is simple. From the simulation it is clear that at room

temperature layer filling alternates between second and third

layers until a certain deposition amount is reached and then

layer filling alternates between third and fourth layers. This

change produces characteristic kinks in the slope of the area

averaged photoemission intensity versus deposition time plots.

Since the simulation is normalised to the density of a fully

tilted (20�3) plane, due to the good fit from 50–125 1C, the

increased deposition amount required to initiate layer comple-

tion kinks and cusps at RT suggests that layers are (20�3)

planes with morphological holes in second and third layers.

This will give rise to the higher photoemission intensity than

would be expected for the increased deposition amount.

Deposition time variability. The photoemission intensity

versus deposition time, re-plotted to emphasise variability in

kink and cusp time positions (indicated by arrows), is given in

Fig. 10. As can be seen, more material must be deposited to

initiate a change in filling at RT and 140–150 1C. The good fit

of the simulation to depositions in the temperature range

50–125 1C together with the completion of second and third

layers at 2.4 and 3.8 ML confirms that in this intermediate

temperature range layers formed have densities that are almost

identical to those required for (20�3) critical nuclei formation

on Cu(110) 2 � 1–O. The implication is of course that since

layer filling on Cu(110) is similar, nucleation processes in this

temperature range should be very similar to that of 6P

deposited on Cu(110) 2� 1–O. More importantly this suggests

that critical nuclei formed are small (20�3) crystallites and not

(21�3) crystallites as one would expect from the observation of

(21�3) crossed needles on the surface post-nucleation. As will be

shown later this conclusion is confirmed by further experi-

mental observations.

In Fig. 2, layer densities for complete coverage (�629), (20�3)

and (21�3) are given relative to the layer density of (20�3). These

densities are to be compared with the values calculated from

the layer completion kinks and cusp positions in Fig. 10 and

listed therein. The calculated surface densities in Fig. 10 are

evaluated by tentatively assuming that the surface density,

calculated from layer completion kinks and cusps, represents

the increased density of a single layer (i.e. no hole/tower

creation). In the case of the RT deposition, it is known from

static X-ray diffraction studies that (�629) crystallites form. It is

tantalising to assume that because of the close fit of the (20�3)

layer density to the 50–125 1C growth, that perhaps at RT a

meta-stable (�629) double layer forms during the pre-nucleation

deposition period. The sum of 1.28 and 1.18 = 2.46 is very

close to the sum of two layers of (�629) = 2.42 as shown in

Fig. 2. However, the higher overall density of a double (�629)

layer compared to a double (20�3) layer would mean that the

photoemission intensity should be lower due to more

attenuation. This is not observed and instead the photo-

emission intensity is in fact higher. Only a variability in the

attenuation, due to differences in attenuation as a function of

molecular orientation within a layer (a higher proportion of

flat molecules in the (�629) plane) could support this interpreta-

tion. It is nevertheless a possibility; however at 140–150 1C

kink positions cannot be attributed to a double (21�3) layer

Fig. 9 (a) Full-coverage layer-by-layer growth (A) and examples of

morphological structures equivalent to a hole in the simulation: (B) an

example of a morphological hole in a double layer of (20�3) and (C) an

example of a structural hole where the crystal structure is more open.

(b) Symbolic representation of layer-by-layer growth at 50–125 1C and

hole creation at RT, as a function of increasing deposition amounts, as

inferred from comparison of simulation to experimental PEEM data.

Note that the first layer (L= 1) is neglected. The simulation deals with

the number density per layer relative to the density of a complete (20�3)

layer (9 boxes per layer). Holes in lower layers increase the photo-

emission transmission intensity in greater proportion to holes in upper

layers. (c) At RT the filling sequence changes from alternate filling of

second and third layers to third and fourth layers at the position of the

kink. The position of the kink is indicated by a light grey box, with a

black arrow indicating the molecular re-arrangement that also occurs.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

without including hole/tower creation. This is because the

surface density, for instance at 150 1C, 1.73 + 1.36 = 3.09

is much greater than the density 2.46 of a double (21�3) layer.

As mentioned previously, from the simulation alone it is not

possible to differentiate between morphological holes in tilted

layers and lower density planes since both lead to higher

absolute photoemission intensities. However in the next

section, with the support from further PEEM data, it will be

confirmed that critical nuclei that form at all temperatures are

(20�3) crystallites and hence a filling scheme involving the

balance of hole creation versus hole filling of (20�3) layers is

correct.

3.3.3 Condensation at steps temperature dependence.

Extracting the area calculated intensity rms from PEEM

images provides another data set that corroborates the above

argument that, due to the temperature dependence of mole-

cular diffusion and molecular de-binding, hole filling and hole

creation rates compete over the whole temperature range

RT–150 1C. As shown previously in Fig. 4(g), condensation

at steps during second and third layer deposition occurs during

layer filling at 125 1C. The simulation correctly identifies

condensation at steps during deposition as layer-by-layer

filling. Condensation at steps is essentially a larger spatial

scale version of layer-by-layer filling whereby terraces are filled

from the step edges outwards rather than by homogeneously

distributed clusters on the terraces. On the other hand, for 6P

deposited on Cu(110) 2 � 1–O no re-distribution, such as

condensation at steps, is observed since the average packing

angle of the first tilted layer increases linearly with time to

incorporate new molecules into the layer. This results in a

characteristic bowing of the photoemission intensity versus

time plot during second layer deposition as seen in Fig. 3.

Shown in Fig. 11 are the characteristic intensity rms plots for

the temperature ranges RT, 50–125 1C and 140–150 1C. For

6P deposited on Cu(110), energetic and geometric weaknesses

of interaction with the underlying flat-lying wetting layer

ensure that second layer condensation at steps occurs no

matter what filling scheme is in action. From inspection it is

clear that third layer condensation at steps is absent at RT and

150 1C. The reasons for this absence are rooted in the same

underlying processes which produce a density of holes at RT

and 140–150 1C, namely the imbalance of hole creation and

hole filling rates in the layer as suggested by the PEEM

simulation. As discussed previously, the reason for the absence

of third layer condensation at steps as opposed to second layer

condensation at steps is the difference between diffusing and

assembling on a flat-lying wetting layer as opposed to on a

tilted layer. In a similar fashion to 6P deposited on Cu(110)

2 � 1–O, the comparatively larger energetic and geometric

strength of interaction of 6P molecules diffusing and

assembling on top of the tilted second layer results in a

homogenous distribution of clusters on the terraces during

third layer filling. It is this homogeneity which produces a flat

response and a minimum intensity rms in Fig. 11 during third

layer filling on Cu(110).

Upon closer inspection of Fig. 11, it is clear that the

positions of intensity rms minima (indicated by arrows) do

not coincide with the positions of second layer completion

kinks in Fig. 10 (indicated by white/shaded areas in Fig. 11),

for RT and 150 1C depositions. At RT the intensity rms

minimum lags the layer completion kink position whereas at

150 1C the intensity rms minimum preceeds the layer

completion kink position in normalised deposition time. This

can be easily explained using the same mechanisms that result

in differences in the rate of hole creation to hole filling. RT

molecular diffusion and molecular re-tilting rates are slow,

Fig. 10 Area-averaged photoemission plots versus normalised time

(flat-lying wetting layer, L = 1) at all temperatures arranged to

emphasise the variability in the required deposition amount to

complete second and third layers during the pre-nucleation deposition

period. The first layer deposition period is greyed out. The surface

densities of each layer deposited, relative to the surface density of the

(20�3) plane, are included for second and third layers. Notice how for

the 50–125 1C temperature range the surface density of the second and

third layers is in very good agreement with the surface density of the

(20�3) plane.

Fig. 11 Condensation at steps as a function of deposition tempera-

ture quantised and plotted as area-calculated intensity rms versus

normalised deposition time. In the temperature range 50–125 1C,

layer-by-layer filling of terraces by condensation at steps results in

kink positions coinciding with intensity rms minima (see for instance

Fig. 4 (b)). The intensity rms minima after second layer filling are

indicated with arrows. To facilitate comparison with layer completion

kinks, grey areas demark the boundaries of first and third layer filling.

The point at which nucleation occurs is marked by an asterisk (*).

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

therefore second layer inhomogeneities introduced by the

random arrival of molecules from the evaporator do not

anneal out; hence these inhomogeneous features persist for a

while during third layer deposition. On the other hand, at

150 1C molecular diffusion, hole creation and re-tilting rates

are much faster. These three processes combine to prevent any

significant condensation at steps from growing to a large size.

Thus any condensation at steps is quickly broken up into

smaller clusters that are homogenously distributed on the

terrace. The observation that condensation at steps is

temperature dependent is further confirmation that the

temperature dependence of PEEM photoemission intensity

data in Fig. 6 and 7(b) is real and that layer filling mechanisms

described, and hence the conclusions drawn from the PEEM

simulation, are in fact correct.

3.3.4 Dewetting induced second and third meta-stable layer

re-construction. The nucleation of 6P to form energetically

stable crossed needles, as seen in Fig. 1(a), initiates

spontaneous dewetting of molecules from areas surrounding

nucleation sites that are readily observable by PEEM. Shown

in Fig. 5(a) are highly anisotropic dewetting streaks that lie at

either side of nuclei in the [110] direction. From PEEM and

LEED measurements it is known that the final structure of 6P

deposited on Cu(110) has a flat-lying wetting layer, a single

reconstructed layer on top and 6P crossed (21�3) needles

distributed over the reconstructed layer. This means that the

tilted meta-stable double layer (second and third layers), which

from measured densities are (20�3) type layers, must re-arrange

upon dewetting to form a single reconstructed layer. In order

to investigate this several techniques such as static STM and

dynamic PEEM are employed.

The main question that must be answered is how crossed

(21�3) needles grow from the nucleation of 6P arranged in (20�3)

type layers? When 6P is deposited on Cu(110) 2 � 1–O, a (20�3)

layer nucleates to produce straight (20�3) needles (see

Fig. 1(b)). However, when 6P is deposited on Cu(110) crossed

needles with a characteristic angle of 201 are observed

(see Fig. 1(a)). This angle corresponds to the top-view of

two needles of the (21�3) plane intersecting (these are crossed

needles, see angle in Fig. 2). So how does the switch over

occurs, and what role does the dewetting and subsequent

reconstruction of the meta-stable (20�3) double layer play in

deciding the final crystal structure of the 6P nanostructures?

As will be shown the answer lies in the flat-lying wetting layer,

which till now has only played a minor role in for instance

facilitating condensation at steps during second layer filling.

STM of dewetting induced reconstruction. During layer

filling in the pre-nucleation deposition period, two kinetically

stabilised layers of tilted molecules form that according to the

PEEM simulation are in a (20�3) arrangement. At RT and

140–150 1C these layers are not perfect since there are holes in

the layers which get temporarily filled by flat-lying molecules

and excess molecules diffuse in dilute fourth, fifth layers. The

key point is that the general (20�3) orientation describes the

densest configuration (by volume) possible for 6P molecules to

assemble in and yet remain lying-down and aligned in the [1�10]

direction. As can be seen from the highly anisotropic

dewetting streaks, molecules in the meta-stable double layer

are aligned when nucleation occurs. However upon

nucleation, molecules diffuse quickly to nuclei and hence

reduce the surface density of molecules in regions surrounding

nuclei. This reduction in surface density reduces the kinetic

stabilisation of the meta-stable double layer (fewer collisions

and attachments) and thus the double (20�3) layer relaxes into a

more energetically stable structure. This structure is a single

reconstructed layer with a combination of flat-lying and titled

molecules that is similar, but not identical, to the (21�3) or (�629)

crystal plane.

An STM image, acquired post-nucleation, of the

re-constructed layer grown on Cu(110) is given in Fig. 13.

The STM image shows channels, running along the [1�10]

direction, that are formed by a structure similar to that in

Fig. 9(a) C. The layer fully covers the surface but its density is

lower than the total surface density of a double (20�3) layer and

hence produces a photoemission intensity increase during

dewetting. Dewetting areas appear as bright streaks in the

dewetting image in Fig. 5(a). The STM image in Fig. 12 clearly

shows channels of approximately a molecular width running

roughly in the [110] direction. These highly anisotropic

channels appear during the spontaneous nucleation dewetting

shown in Fig. 5(a). The result is that excess molecules leave the

reconstructed areas by diffusing along the channels until they

join a needle. If they diffuse in another direction they will be

on top of un-reconstructed regions. Since nuclei are observed

to grow, this confirms that molecules are diffusing along the

dewetted zone towards the nucleation site. Therefore if the

channels observed by STM are the result of a surface

reconstruction induced by dewetting, the diffusion of mole-

cules on the surface should also be anisotropic.

PEEM measurement of diffusion anisotropy. It is possible to

study this molecular diffusion process once the whole surface

reconstruction is completed (roughly double the time required

for nucleation to begin). A simple way to probe molecular

diffusion on the reconstructed surface is to close the shutter

of the molecular evaporator and examine the change in

Fig. 12 STM image of the second layer of 6P on Cu(110). Note that

the channels run in the same direction as the highly anisotropic

dewetting steaks and are roughly perpendicular to the [001] direction.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

photoemission intensity that accompanies the reduction in the

surface density of 6P molecules in areas between needles. Once

the shutter is closed the equilibrium surface density of mole-

cules re-adjusts to a lower value since the rate of molecules

leaving the surface to join the 3D crossed needles remains

high, the rate of molecules leaving the crossed needles to join

the surface remains low and the number of molecules arriving

at the surface from the evaporator is now zero. Eventually a

new equilibrium surface density in the areas between needles

will be reached where the number of molecules leaving the

needle to diffuse in the surrounding area will equal the number

of molecules leaving the surrounding area to join the needle.

The resultant increase in photoemission intensity (since

there is a depletion of molecules attenuating) is plotted in

Fig. 13. To test whether the molecular diffusion is anisotropic

the photoemission intensity is plotted against square-root-time

and compared to the inverse5 of the one dimensional

diffusion20 equation r(x,t) = (r0/2)(pDt)�12exp(�x2/4Dt)

where r0 is the initial surface density, D is the diffusion

coefficient, x is the position along a 1-D ordinate and all other

symbols have their usual meaning. A straight line indicates

that the system diffuses one dimensionally. Certain considera-

tions must be taken into account. For instance, the photo-

emission intensity changes are measured when there are

crossed needles on the surface. Since the whole image is used

to acquire the average intensity changes, the effect of dark

needles remaining constantly dark while the surface between

the needles undergoes rapid changes in intensity must be

accounted for. When the intensity for the one-dimensional

diffusion equation is integrated and averaged over a finite

length (e.g. the distance between needles) a distinctive concave

curve prior to the onset of the linear regime is observed as in

Fig. 13. The larger distance between crossed needles at 150 1C

produces a straighter plot.

The fact that a linear response is observed for all four

temperature ranges explored indicates that molecular diffusion

is anisotropic at all deposition temperatures from RT to

150 1C. This confirms that at all deposition temperatures a

reconstruction occurs that produces channels such as those in

Fig. 9(b). Comparisons to similar diffusion measurements of

6P diffusing on a (20�3) type wetting layer for 6P deposited on

Cu(110) 2 � 1–O (see the inset in Fig. 13) suggest that, if

hypothetically, the (20�3) type meta-stable double layer did not

undergo a reconstruction and remained intact, a pseudo 2-D

diffusion would be observed at high temperatures. This is not

observed. It is interesting to note that since the total number of

molecules of the meta-stable double layer (the second and

third layers) exceeds the number required to form a single

reconstructed layer (the post-dewetting permanent second

layer), the reconstruction cannot proceed unless excess

molecules are removed from the surface via large scale mass

transport to growing nuclei. This reduction in surface density

shifts the equilibrium away from kinetic stabilisation of the

meta-stable double layer and towards a single reconstructed

layer that is energetically stable. The energetic stability comes

from partial interaction of the reconstructed layer with

flat-lying molecules of the underlying wetting layer. During

the pre-nucleation deposition period, the surface free energy of

the first/second layer interface is large due to the weak

interaction between aromatic planes of first layer molecules

and hydrogen atoms of the 6P tilted layer. The re-construction

is therefore an attempt, by incorporating flat molecules, to

reduce the surface free energy of this interface. This is the layer

measured by STM. The exact ratio of flat lying to tilted

molecules of the reconstructed layer is most likely temperature

dependent (as is the rate at which molecules re-tilt). However

the important aspect of the reconstruction, in relation to the

formation of (�629) and (21�3) crossed needles, is that the

reconstruction has a mixture of flat and tilted molecules in a

similar fashion to (�629) and (21�3) layers (see Fig. 2). It is for

this reason that crossed needles (21�3) grow when 6P is

deposited on Cu(110) and straight (20�3) needles grow when

6P is deposited Cu(110) 2 � 1–O, even though both involve an

intermediate step where (20�3) layers form and (20�3) critical

nuclei crystallise.

3.3.5 Direct evidence of the formation of (20�3) critical

nuclei. More compelling evidence of the intermediate (20�3)

layers and (20�3) critical nuclei formation for 6P deposited on

Cu(110) can be found in PEEM images of crossed needles.

Comparison with 6P deposited on Cu(110) 2 � 1–O strongly

suggests that critical nuclei that form will also be (20�3) type.

This would appear to be counter-intuitive because (21�3)

crossed needles are observed post-nucleation. However, as

has been shown by the anisotropy of diffusion and from

STM, as soon as dewetting occurs areas surrounding critical

nuclei in the [110] direction reconstruct into a layer containing

flat and tilted molecules. The outcome of this process is that as

soon as critical nuclei (20�3) grow larger and begin to self-

assemble over reconstructed areas that now surround nuclei,

the favourable structure to assemble changes from (20�3) to

(21�3). Strong evidence that supports the formation of critical

(20�3) nuclei is provided by the occasional (20�3) needle that

happens to grow quicker than the reconstruction wave

proceeds outwards. Some examples are given in Fig. 14. This

is possible because (20�3) needles grow perpendicular to

the spontaneous nucleation fastest dewetting direction. Some

Fig. 13 Multi-plot of photoemission intensity versus square-root-

time, acquired after 6P nucleation on Cu(110) and dewetting is

complete. Measured from the time the molecular beam shutter is

closed. Inset: equivalent plot for 6P on Cu(110) 2 � 1–O.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

examples of (20�3) needles are also observed at 140–150 1C. As

can be seen in Fig. 14, at lower temperature (115 1C), curvy

needles grow because the rate of needle growth is comparable

to the rate at which the reconstruction wave advances away

from the critical (20�3) nucleus. This leads to needles growing

on top of partially reconstructed areas which will be a mixture

of (20�3) type and flat/tilted reconstructed layers; and hence

some crossed needles straighten out in the [001] direction as

they begin to grow on un-reconstructed (20�3) type regions. As

can be seen in Fig. 14, the fastest (i.e. longest) growing needle

(excluding the fast growing (20�3) needles) is always the needle

that is most likely to curve or kink. From inspection of many

PEEM images of crossed needles grown at this temperature,

the direction of curviness is also always towards the midline

(in the [001] direction) of a given crossed needle. This direction

of curviness confirms that the reconstruction wave starts at the

centre of the crossed needles (where the critical nucleus is

located). This is further confirmed by the fact that all fast

growing (20�3) needles point towards the centre of crossed

needles. This is in fact a fingerprint of the original critical

nucleus being (20�3) type. At 140–150 1C crossed needles have

no curviness, since at these higher temperatures molecular

re-arrangements, required for reconstruction, proceed much

quicker than crossed needles grow. Thus crossed needles never

cross the reconstruction wave travelling boundary and thus

remain straight. Another feature of crossed needles that

becomes more apparent at higher temperatures is the hole

that appears at the intersection point of crossed needles. It is

highly probable that this is a hole created by (20�3) critical

nuclei sacrificially Oswald ripening the crossed needles.

Since (20�3) critical nuclei have a crystalline structure

that is inherently less energetically stable than surrounding

(21�3) crossed needles (as evidenced by the re-construction

during dewetting), eventually molecules forming (20�3) critical

nuclei detach and diffuse to (21�3) needles and leave behind

a hole.

3.3.6 Nucleation densities of (20�3) critical nuclei on Cu(110)

and Cu(110) 2 � 1–O. The rate of nucleation per unit area,

when plotted in the form of an Arrhenius plot, allows the

nucleation energy to be determined, for a particular substrate/

molecule combination, from the slope of the plot. However,

the term nucleation energy is slightly misleading since many

activation energies and binding energies determine the final

rate of nucleation. The positive slope of the plots in Fig. 15

indicates that at higher temperatures fewer nucleation sites

form. It is important to note that nucleation sites are formed

when a critical surface density of molecules is reached. It is at

this point that kinetically stabilised sub-critical nuclei become

energetically stable via dynamic processes that position mole-

cules into static arrangements such as energetically stable

crystalline structures. The process of kinetic stabilisation is

controlled by the rate at which molecules join sub-critical

nuclei—which would appear to be controlled by activation

barriers associated with diffusion. However, high layer

densities required for 6P to nucleate suggest that molecules

spend more time bound to other molecules in the form of small

sub-critical nuclei. Hence, intermolecular binding energies will

therefore determine the rates at which molecules de-bind from

sub-critical nuclei which will in turn affect (a) the lifetime of

sub-critical nuclei and (b) the rate at which molecules become

available to diffuse to other subcritical nuclei. Conversely,

when the critical surface density is low, such as with nucleation

from a 2D gas phase, molecules spend a greater proportion of

time diffusing rather than binding/de-binding.

The nucleation rate data presented in Fig. 15 demonstrate

that the slopes for both 6P deposited on Cu(110) and Cu(110)

2 � 1–O in the temperature range 85 o T o 150 1C are nearly

identical at B950 meV. The nucleation rates for depositions

o85 1C are too high for 6P on Cu(110) to be accurately

assessed. The near identical nucleation energy does strongly

Fig. 14 PEEM images, treated by applying an edge-detect filter

(gradient algorithm) to enhance weak features, clearly show several

instances of straight needles (indicated by arrows) emerging from

crossed needles. Straight needles are (20�3) needles that grew out of

(20�3) critical nuclei ahead of the advancing surface reconstruction

wave. The reconstruction wave changes two meta-stable (20�3) type

layers into a single reconstructed layer. Hence the straight (20�3)

needles are generally longer than crossed needles. Notice that all

(20�3) straight needles point directly to the central intersection of

crossed needles where critical nucleation began. Many crossed needles

appear to have a missing centre which would correlate well with an

original (20�3) critical nucleus sacrificially Ostwald ripening to

surrounding (21�3) needles.

Fig. 15 Arrhenius plot of the natural log of the nucleation rate per

unit area versus 1/T. The slope of the plot is used to determine the

nucleation energy for 6P on each surface. Note the general similarity in

slope for both systems studied. This is further confirmation that the

formation of (20�3) critical nuclei is common to both.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

This journal is c the Owner Societies 2011 Phys. Chem. Chem. Phys.

confirm that critical nuclei formed on both substrates are (20�3)

6P crystallites. The molecular binding energy of 6P molecules

in bulk-type arrangement was measured5 to be 2.1 eV from

PEEM measurements of desorption of 6P molecules from

needles grown on Cu(110) 2 � 1–O. Since the nucleation

energy is a combination of binding energies and activation

barriers, the maximum total activation energy for the nuclea-

tion of 6P deposited on Cu(110) and Cu(110) 2 � 1–O surfaces

is 1.15 eV.

4. Conclusions

The comparison of 6P depositions on two substrates with

differently oriented wetting layers reveals that there are several

pre-nucleation mechanisms that determine the nucleation

process. Since the 50–125 1C deposition temperature range

of 6P deposited on Cu(110) produces (20�3) critical nuclei and

6P deposited on Cu(110) 2� 1–O produces (20�3) critical nuclei

at all temperatures, these are the best depositions to compare.

In both cases, critical nucleation occurs after filling two layers

with tilted molecules where each layer has a density approxi-

mately equal to the density of a (20�3) crystal plane. However,

6P deposition on Cu(110) is temperature dependent whereas

deposition on Cu(110) 2 � 1–O is not. The reason is that when

6P is deposited on Cu(110) 2� 1–O there is only one dominant

orientation for molecules—a tilted orientation whereby mole-

cules are bound to the layer. However for 6P deposited on

Cu(110), there are two equivalent orientations that molecules

in the second layer can adopt—tilted or flat-lying orientation.

In the meta-stable double layer, molecules can either tilt to

bind with other tilted molecules or equivalently lie flat to bind

with the flat-lying wetting layer. This mixture of orientations is

confirmed by the observation, by STM, of channels consisting

of a mixture of molecular orientations in the energetically

stable post-nucleation reconstructed layer. The binding

energies of both orientations are roughly the same since the

molecules are bound to other molecules by their aromatic

plane in both cases. This means that for 6P deposited on

Cu(110) these two molecular orientations are in continuous

competition during layer filling and hence a temperature

dependence is observed. The relative rate at which holes are

occupied with flat-lying molecules, or flat-lying molecules tilt

and join the tilted layer, is given by the difference in chemical

potential between the two states. Since both states have similar

binding energies, it is the surface density of available molecules

that determines the chemical potential for each state.

As evidenced by the rearrangement of both the third and

second layers into a single reconstructed layer, molecules in

second and third layers only exist as two tilted layers because

of the kinetic stabilisation enabled by the high surface density

of 6P molecules. This confirms that for 6P deposited on

Cu(110) the dynamic behaviours of the two tilted layers are

coupled to each other in the form of a meta-stable double

layer. On the other hand, for 6P deposited on Cu(110) 2� 1–O

the tilted layers are not dynamically coupled since the wetting

layer is energetically stabilised by the interaction with the

Cu(110) 2 � 1–O reconstruction. This means that in this case

only one tilted layer behaves dynamically.

These factors play a significant role in determining at what

total surface density critical nucleation begins. It is important

to ask the question: why does nucleation not begin earlier, say

during second or third layer filling? Given that condensation at

steps forces molecules to adopt a much higher local surface

density compared to when molecules are evenly distributed

over the surface (as with 6P deposited on Cu(110) 2 � 1–O),

surely nucleation could begin during second layer deposition?

Answering these questions leads one to examine the very

nature of the nucleation process. As described earlier, the

reconstruction offers us a glimpse into what the preferred

structure to assemble on a flat-lying wetting layer is. During

layer filling, 6P molecules will always try to reach this

energetically stable structure, but high temperatures will tend

to encourage flat-lying molecules in the second layer to

de-bind and tilt, and the presence of an ever increasing amount

of molecules will quickly fill the gap left behind with tilted

molecules. This process, given that both tilted layers are

dynamically coupled, is repeated during third layer filling via

molecules re-arranging together with molecules in the second

layer beneath. It is worth remembering that (20�3) meta-stable

layers are kinetically stabilised by a continuous filling of holes

that appear in the layers. Averaged over a long time, at

50–125 1C, there are no holes and this is illustrated in Fig. 16.

It is this competition between tilted and flat orientations for

6P deposited on Cu(110) that prevents subcritical nuclei from

Fig. 16 Schematic of the filling scheme for 50–125 1C for 6P

deposited on Cu(110). (a) Wetting layer completion (first layer). (b)

Condensation at steps of the incomplete second layer. Molecules

diffuse flat then join the sub-critical (20�3) tilted second layer that

extends over the terraces from the steps. (c) Completion of the second

layer for 50–125 1C. The interaction with the wetting layer beneath

prevents critical nucleation from occurring. (d) Partial condensation at

steps by the incomplete third layer due to coupling with the tilted

second layer to form a double meta-stable layer. (e and f) Critical

surface density reached when a third layer can spontaneously form

large enough nuclei that no longer require kinetic stabilisation and are

thus energetically stable. This leads to dewetting of the double layer of

tilted molecules surrounding the critical nucleus. (g) Sketch showing

the dewetting induced spontaneous re-construction of the double tilted

layer into channels that include flat lying molecules. These channels,

visible in Fig. 12, run into and out of the page. The effect of the

reconstruction is to force molecules to assemble into (21�3) layers

rather than continue with (20�3).

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2011

growing in size within the layer, since any gaps opened up by

molecules climbing on top of subcritical nuclei will be filled by

a neighbouring molecule that will lie flat. It is the stability of

flat lying molecules, offered ultimately by the flat-lying wetting

layer, which disrupts the growth of sub-critical (20�3) nuclei

and shortens their lifetime. It is only when complete coverage

is reached on the third layer that there is no re-arrangement

space for any gaps to be stabilised by flat-lying molecules

(the layer is fully kinetically stabilised with tilted molecules)

and this allows sub-critical (20�3) nuclei to grow,

un-interrupted, to be large enough to become an energetically

stable nuclei. At higher temperatures 140–150 1C this process

is repeated except that more molecules are required to initiate

critical nucleation. This is due to the continuous process of

hole creation and hole filling leaving a net density of holes in

the layer as molecules climb to higher layers.

Upon critical nucleation, spontaneously induced dewetting

reduces the surface density (as molecules diffuse and attach to

the nuclei) which in turn reduces the kinetic stabilisation of the

second and third layers. The interaction with the flat-lying

wetting layer once again plays a role as it promotes an

energetically stable reconstruction that includes flat and tilted

molecules (the holes are permanently filled with flat-lying

molecules). It is because of the similarity of the reconstruction

structure to the (21�3) and (�629) crystal planes that ultimately

crossed needles assemble around the initial (20�3) critical

nucleus.

To summarise, the conclusions that can be drawn from the

comparison of the nucleation of 6P on Cu(110) 2 � 1–O and

6P on Cu(110) are:

(a) On an anisotropic substrate whereby 6P molecules align

with their long axis parallel to the surface, the densest

arrangement of molecules in a layer is of a (20�3) type. There-

fore this is the preferred arrangement of 6P molecules in a

layer that is kinetically stabilised.

(b) The geometric and energetic strength of interaction

between layers decides whether large scale molecular

re-distributions, such as condensation at steps, will occur. It

will also decide whether kinetically stabilised layers are in an

oriented liquid state or, such as with 6P deposited on Cu(110),

in a more crystalline state.

(c) The creation of (20�3) critical nuclei requires a double

layer of (20�3) type form, with at least one layer being

kinetically stabilised. Once a critical density is reached

energetically stable (20�3) critical nuclei will form.

(d) The first layer of 6P molecules, oriented by the choice of

metal or metal oxide substrate, ultimately determines what

crystal structure the nanostructures will have. There may be

intermediate steps that will determine the rate of nucleation

but ultimately the dynamic interaction of the meta-stable layer

with the wetting layer will decide how the nanostructures will

continue growing.

Acknowledgements

Supported by the Austrian Science Foundation (FWF)

through the NFN: Interface Controlled And Functionalised

Organic Films.

References

1 P. Puschnig, S. Berkebile, A. J. Fleming, G. Koller, K. Emtsev,T. Seyller, J. D. Riley, C. Ambrosch-Draxl, F. P. Netzer andM. G. Ramsey, Science, 2009, 326, 702–705.

2 G. Koller, S. Berkebile, J. Ivanco, F. P. Netzer and M. G. Ramsey,Surf. Sci., 2007, 601, 5683.

3 M. G. Ramsey, D. Steinmuller and F. P. Netzer, Synth. Met., 1993,54, 209.

4 M. G. Ramsey, M. Schatzmayr, S. Stafstromand and F. P. Netzer,Europhys. Lett., 1994, 28, 85.

5 A. J. Fleming, F. P. Netzer and M. G. Ramsey, J. Phys.: Condens.Matter, 2009, 21, 445003.

6 G. Koller, S. Berkebile, M. Oehzelt, P. Puschnig, C. Ambrosch-Draxl, F. P. Netzer and M. G. Ramsey, Science, 2007, 317, 351.

7 S. Berkebile, G. Koller, P. Puschnig, C. Ambrosch-Draxl,F. P. Netzer and M. G. Ramsey, Appl. Phys. A: Mater. Sci.Process., 2009, 95, 101–105.

8 M. Oehzelt, L. Grill, S. Berkebile, G. Koller, F. P. Netzer andM. G. Ramsey, ChemPhysChem, 2007, 8, 1707.

9 S. Berkebile, G. Koller, G. Hlawacek, C. Teichert, F. P. Netzer andM. G. Ramsey, Surf. Sci., 2006, 600, 313.

10 P. Stadler, A. M. Track, M. Ullah, H. Sitter, G. J. Matt, G. Koller,T. B. Singh, H. Neugebauer, N. Serdar Sariciftci andM. G. Ramsey, Org. Electron., 2010, 11, 207–211.

11 K. Al-Shamery, H.-G. Rubahn and H. Sitter,Organic Nanostructuresfor Next Generation Devices, Springer, 2008.

12 F.-J. Meyer zu Heringdorf, M. C. Reuter and R. M. Tromp,Nature, 2001, 412, 517.

13 H. Yasufuku, M. Okumura, S. Kera, K. K. Okudaira, Y. Haradaand N. Ueno, J. Electron Spectrosc. Relat. Phenom., 2001,114–116, 1025.

14 N. M. Buckanie and F.-J. Meyer zu Heringdorf, Surf. Sci., 2007,601, 1701.

15 H. Marchetto, U. Groh, T. Schmidt, R. Fink, H.-J. Freund andE. Umbach, Chem. Phys., 2006, 325, 178.

16 H. H. Rotermund, S. Nettesheim, A. von Oertzen and G. Ertl,Surf. Sci., 1992, 275, 645.

17 H. H. Rotermund, Surf. Sci., 1993, 283, 87.18 S. Berkebile, G. Koller, P. Puschnig, C. Ambrosch-Draxl,

F. P. Netzer and M. G. Ramsey, Appl. Phys. A: Mater. Sci.Process., 2009, 95, 101–105.

19 J. Ivanco, F. P. Netzer and M. G. Ramsey, Org. Electron., 2007, 8,545–551.

20 K. A. Jackson, Kinetic Processes, Wiley-VCH, Weinheim, 2004,p. 188.

Dow

nloa

ded

by U

nive

rsita

tsbi

blio

thek

Gra

z on

07

Febr

uary

201

1Pu

blis

hed

on 2

7 Ja

nuar

y 20

11 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/C0C

P015

16D

View Online