Reaction mechanisms for on-surface synthesis of covalent nanostructures

Abstract

In recent years, on-surface synthesis has become an increasingly popular strategy to form covalent nanostructures. The approach has great prospects for facilitating the manufacture of a range of fascinating materials with atomic precision. However, the on-surface reactions are enigmatic to control, currently restricting its bright perspectives and there is a great need to explore how the reactions are governed. The objective of this topical review is to summarize theoretical work that has focused on comprehending on-surface synthesis protocols through studies of reaction mechanisms.

Full text

This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 130.236.82.7
This content was downloaded on 02/02/2016 at 19:36
Please note that terms and conditions apply.
Reaction mechanisms for on-surface synthesis of covalent nanostructures
View the table of contents for this issue, or go to the journal homepage for more
2016 J. Phys.: Condens. Matter 28 083002
(http://iopscience.iop.org/0953-8984/28/8/083002)
Home Search Collections Journals About Contact us My IOPscience

1 © 2016 IOP Publishing Ltd Printed in the UK
Journal of Physics: Condensed Matter
J Björk
Reaction mechanisms for on-surface synthesis of covalent nanostructures
Printed in the UK
083002
JCOMEL
© 2016 IOP Publishing Ltd
2016
28
J. Phys.: Condens. Matter
CM
0953-8984
10.1088/0953-8984/28/8/083002
8
Journal of Physics: Condensed Matter
1. Introduction
On-surface synthesis has great prospects for manufacturing
covalent nanostructures with atomic precision. In contrast to
the vast eld of heterogeneous catalysis, which has primarily
focused on the chemical transformation of rather small organic
molecules, the central research target is the synthesis of covalent
materials extended in either one or two dimensions. By coupling
molecular building blocks, aided by the reactivity on a metal
surface under ultra-high vacuum (UHV) conditions, covalent
structures can be formed following a bottom-up strategy. This
has resulted in the synthesis of various covalent materials, such
as graphene nanoribbons [1], extended graphdiyne wires [2],
porous graphene [3], and single-chirality carbon nanotubes [4].
In principle, the dimensions of a formed structure could be
controlled on the sole basis of molecular building blocks and the
type of reaction triggered between the molecules. Thus, by equip-
ping right molecular building block with proper reacting groups,
any imaginable type of covalent material should be within reach.
However, the reality is not that simple, and mastering the on-
surface reactions has shown to be a severely complicated task.
Since the insight into the mechanisms of the reactions relevant
for on-surface synthesis is rather limited, it is difcult to foresee
how a molecule will react on a surface. Therefore, we often rely
on trial-and-error, coupled to insight from organic synthesis,
and nding new covalent materials has become a quite time-
consuming activity. To improve this research eld into a direc-
tion where we can make, from a blueprint of a covalent material,
rational choices of the appropriate molecules and surfaces for
creating a material, there is a need for a greater understanding
of the underlying mechanisms of relevant on-surface reactions.
The ultimate goal would be to comprehend how the these mech-
anisms are governed by the selection of molecules and surfaces.
To reach this ambition, there is a great need for theoretical
input into mechanisms of on-surface reactions. The objective
of this topical review is to sum up the theoretical work that has
been done so far, and identify some of the research questions
that theory should focus on during the forthcoming years, for
advancing the propitious eld of on-surface synthesis.
The topical review is structured the following way: rst I
will give a brief introduction of how reactions on surfaces can
be studied by electronic structure theory. This is followed by
sectionsdevoted to separate on-surface reactions, each sum-
marizing key experimental studies followed by a more in-
depth discussion of the efforts from theoretical modeling. I
will particularly be addressing on-surface Ullmann coupling,
Reaction mechanisms for on-surface
synthesis of covalent nanostructures
JBjörk
Department of Physics, Chemistry and Biology, IFM, Linköping University, Sweden
E-mail: jonas.bjork@liu.se
Received 15 May 2015, revised 8 January 2016
Accepted for publication 14 January 2016
Published 2 February 2016
Abstract
In recent years, on-surface synthesis has become an increasingly popular strategy to form
covalent nanostructures. The approach has great prospects for facilitating the manufacture of
a range of fascinating materials with atomic precision. However, the on-surface reactions are
enigmatic to control, currently restricting its bright perspectives and there is a great need to
explore how the reactions are governed. The objective of this topical review is to summarize
theoretical work that has focused on comprehending on-surface synthesis protocols through
studies of reaction mechanisms.
Keywords: surface chemistry, density functional theory, transition state theory
(Some guresmay appear in colour only in the online journal)
Topical Review
IOP
0953-8984/16/083002+15$33.00
doi:10.1088/0953-8984/28/8/083002
J. Phys.: Condens. Matter 28 (2016) 083002 (15pp)

Topical Review
2
surface chemistry of terminal alkynes, and cyclodehydrogena-
tion reactions; for which the most extensive theoretical invest-
igations have been made; but will briey cover other types of
on-surface reactions. The topical review is concluded by an
outlook.
2. Density functional theory description of
on-surface reactions
Here, density functional theory (DFT) will be introduced,
including a brief discussion of how well DFT describes mol-
ecules on surface (in particular, the treatment of van der Waals
interactions). Then it will be described how DFT together
with transition state theory can be used to study reactions,
outlining different methods for computing reaction pathways
and transition states.
In density functional theory, the total energy of a system is
obtained as a functional of the ground state electron density,
()≡n n r
[5]
E nTnvn
nn
En
rr rrr
rr
rr
d
1
2
dd ,
sx
c
[] [] ()()
()()
[]
∫∫
=+ +
−
+
′
′
′
(1)
where the rst term is the kinetic energy of non-interacting
electrons, the second and third terms give the electron-nuclei
and electron-electron Coloumb energy, respectively, and the
nal term is the so-called exchange-correlation (XC) energy.
All these terms can be determined exactly, within the numer-
ical accuracy of the calculations, except for the XC energy,
which has to be approximated.
2.1. Treating van der Waals interactions
An important, and highly timely, aspect of modeling the
adsorption of molecules on surfaces is how to treat so-called
van der Waals (vdW), or London dispersion, forces. By con-
struction, conventionally used generalized gradient approx-
imation (GGA) and local density approximation (LDA) fail to
include these interactions, why adsorption heights are gener-
ally overestimated for weakly adsorbed systems, or molecules
with a π-conjugated core, and the reactivity of a molecule on a
surface may not be described correctly. Two schools of thought
for treating vdW interactions in DFT have been established.
The rst is based on semi-empirical corrections, known as
dispersion-corrected DFT [6], while the other solves the prob-
lem by introducing a non-local density functional commonly
denoted as the van der Waals density functional (vdWDF) [7].
Without going into detail, it has been illustrated that with their
respective most recent developments, molecular adsorption
heights are now described with a precision of 0.1 Å both by
dispersion-corrected DFT [8, 9] and vdWDF [10–14].
2.2. Studying reaction mechanisms
The approach for studying reaction paths of reactions related
to on-surface synthesis has exclusively been transition
state theory, in which it is assumed that a reaction can be
characterized by an initial state (IS), a transition state (TS)
and a nal state (FS) and the rate of crossing the transition
state at a temperature T is given by the Arrhenius equation
()ν =−AEkTexp/,
aB
(2)
where the activation energy
E
a
is dened as the energy differ-
ence between the TS and FS, and
k
B
is Boltzmann’s constant.
The prefactor A may be approximated from the vibrational
frequencies at the TS and IS, but is normally assigned the
rule-of-thumb value of 10
13
s
−1
.
The IS and FS are local minima on the potential energy
surface, obtained by conventional structural optimization by
minimizing the forces in the calculations. The TS of a reac-
tion is, however, less trivial to determine as it is characterized
by a saddle point on the potential energy landscape. There are
several methods for nding saddle points in DFT.
The most commonly applied way for nding saddle
points in on-surface synthesis is with the Nudged Elastic
Band (NEB) method. In this method several images (states)
are connected to trace a path between the IS and the FS.
The reaction path is then found by minimizing the forces
acting perpendicular to the tangents of the path (nudging).
To ensure that images are equally distributed along the path,
spring-forces acting parallel to the tangent of the path are
introduced (elastic bands). From the NEB method, typically
none of the images are found at, or even close to, the TS and
the TS energy is estimated by interpolation. This led to the
development of climbing image NEB (CI-NEB) [15]. The
only difference from regular NEB is that the spring force
acting on the highest energy is replaced with the negative of
the force parallell to the tangent of the path [15]. This way
the highest energy image, referred to as the climbing image,
moves up the energy surface along the elastic band, towards
the transition state of the path. Importantly, the path deter-
mined by the NEB methods depends on the initial interpola-
tion between IS and FS. Therefore, depending on the type of
reaction, one may need to compare the outcome of several
initial guesses of the path.
With some vigilance, NEB and CI-NEB present reli-
able ways of studying reaction paths. Particularly they nd
the number of barriers separating initial and nal states.
However, as they rely on accurate tangent description of
the reaction path (obtained by nite differences between
images), they require an adequate number of images for
converging the path. This makes the methods numerically
expensive as we need individual DFT calculations for the
separate images. Minimum mode following methods,
exemplied by the Dimer method [16, 17], are numerically
cheaper since they focus on the optimization of a transition
state, omitting the information about the complete reaction
path. The Dimer method is based on using three images (or
states): the central image and the two images constituting the
dimer. The dimer images are slightly displaced from each
other by a xed distance with the central image in the mid-
dle. The TS search algorithm involves two steps. In the rst
step the vector dened by the dimer is rotated into the lowest
curvature mode of the potential energy at the central image.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
3
Then, the central image and the dimer are translated a cer-
tain step length, moving the central image towards the TS.
The TS is found by iterating the algorithm until the forces
on the central image are converged, under the condition that
the curvature of the potential energy at the central images in
the direction of the dimer is negative. The method explores
trans ition states either by moving the dimer in different
directions from the IS (even without knowledge of the FS),
or by making use of the output from a NEB calculation as
initial guess of the TS.
For complex multidimensional reactions, often encounter ed
in on-surface synthesis, the last alternative is most appealing,
since we have the full trajectory between initial and nal state
from NEB, thereby knowing the number of barriers separat-
ing the FS from the IS and, for example, the Dimer method is
used to rene the TS found from NEB.
3. On-surface Ullmann coupling
The on-surface Ullmann coupling is without doubt the most
commonly applied strategy for creating covalent materials
through on-surface synthesis. The approach is reminiscent
of the Ullmann coupling in wet chemistry, in which aryl
halides comprise the molecular building blocks. The halo-
gens of these molecules are thermally abstracted more easily
than other atoms, making it possible to generate surface-sta-
bilized radicals [18], which can then couple into patterns of
covalently bonded molecules. By controlling the number of
halogens, and their positions within the molecules, one can in
principle control the dimensions of the formed nanostructures.
The basic concept is illustrated in gure1 for two model sys-
tems based on benzene substituted with different number of
halogen atoms.
The on-surface Ullmann coupling was demonstrated
already 1992 for the synthesis of biphenyl from iodobenzene
on Cu(1 1 1) [19], but was not demonstrated for the formation
of covalent nanostructures until 15 years later [20]. In a pio-
neering study by Grill et al it was illustrated how porphyrins
equipped with bromine can couple into larger molecules (0D),
chains (1D) or clusters (2D) depending on the number of bro-
mine atoms in the molecules [20]. The method has since then
been used numerous times [1, 3, 18, 21–31]. For example, it
can be used to form graphene nanoribbons, either comple-
mented by a cyclodehydrogenation reaction [1], or as the sole
reaction step [31]. In several cases metal-organic intermediate
structures have been observed [32–36], in which the dehalo-
genated molecules coordinate to metal adatoms thermally
generated from the substrate. The exact role of these adatoms
in the overall process is not yet clear, as the theoretical work
so far has focused on reactions on atomically at surfaces, as
will be described here.
3.1. Ideal conditions for creating ordered 2D-materials
An important step for the on-surface Ullmann coupling
is the recombination of surface-stabilized radicals. The
recombination is, in a simplied picture, associated with
two processes: radical diffusion and radical-radical cou-
pling. The coupling step is considered as a more or less
irreversible process, resulting in that self-healing, inher-
ent to supramolecular self-assembly, is generally missing
(although this problem may be solved by the formation of
metal-organic intermediates [36]). Bieri et al [18] demon-
strated that this adds some restriction on the diffusion and
coupling step, for the formation of ordered covalent 2D
materials. They dened a coupling probability between two
adjacent radicals as
ν
νν
=
+
P ,
couple
coupl
ed
iff
(3)
where
ν
couple
is the rate of coupling and
ν
diff
is the rate of dif-
fusion. Small values of P indicates that the diffusion is much
faster than the coupling (a coupling limited process), while
values close to 1 indicates that the coupling is much faster
than the diffusion (a diffusion limited process). Figure2 illus-
trates kinetic Monte Carlo (kMC) simulations of the nucle-
ation of surface-stabilized radicals for different values of P.
For P values close to 1, fractal-like networks are formed,
while smaller values of P give more ordered 2D networks.
The results of Bieri et al [18] illustrates the requirement of
a coupling-limited process for the formation of ordered 2D
materials with the on-surface Ullmann coupling, and any
other method requiring a recombination step between surface-
stabilized radicals.
3.2. Recombination mechanisms
Theoretical calculations of reaction mechanisms together with
experimental observations corroborated the prediction that a
coupling-limited process is a prerequisite for the formation of
ordered 2D materials. Figure3 illustrates calculated recom-
bination mechanisms of cyclohexa-m-phenylene radicals
(CHPR) on Cu(1 1 1) and Ag(1 1 1) [18]. CHPR is chemi-
cally bonded via its six radical sites to the metal atoms and the
calculated reaction sequence is similar for both surfaces: Two
carbon radicals (one from each molecule) are bonded to adja-
cent surface atoms in the initial state (denoted II). The reac-
tions are initiated by a multi-barrier diffusion step, leading
to the intermediate states IM2, where the two molecules are
chemically bonded with a carbon radical to the same surface
atom. On Cu(1 1 1), the coupling-step is barrier-free (IM2 to
IM3), while the nal step in the simulated path is simply a
diffusion step for one half of the dimer. On Ag(1 1 1), the cou-
pling is a two-step process between IM2 and FI, with a barrier
larger than that for diffusion.
These results indicate that the recombination process of
CHPR is diffusion limited on Cu(1 1 1) (compare the diffu-
sion barrier of 2.2 eV to the spontaneous coupling reaction),
while it is coupling limited on Ag(1 1 1) (diffusion barrier
of 0.8 eV and coupling barrier of 1.8 eV). Thus, from the dis-
cussion around equation (3) one would expect that CHPR
more likely forms ordered 2D materials on Ag(1 1 1) than
on Cu(1 1 1). Notably, experiments have shown exactly this:
On Ag(1 1 1) a porous graphene structure with relatively low
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
4
defect density was observed, resembling the simulated struc-
ture obtained for a small coupling probability in gure2(c).
Contrarily, branches of a single molecule width were found on
Cu(1 1 1) [3]. In the same study, experiments found networks
of intermediate quality on Au(1 1 1), suggesting it has proper-
ties somewhere between Ag(1 1 1) and Cu(1 1 1) in terms of
promoting a coupling or a diffusion limited process. However,
there are no theoretical studies for the recombination of two
CHPRs on this surface, which certainly would be of great
interest.
3.3. The formation of biphenyl
Before discussing the complete reaction sequence to form
biphenyl with on-surface Ullmann coupling presented in [37]
it is important to have in mind that Nguyen et al were the rst
Figure 1. Basic principle of the on-surface Ullmann coupling for the formation of (a) 1D and (b) 2D materials, by controlling the number
of halogens of the molecular precursors.
Figure 2. Kinetic Monte Carlo simulations for the recombination process between surface-stabilized radicals for different coupling
probabilities P, dened by equation(3). The smaller values of P, the smaller number of defects are found in the resulting 2D networks.
Reprinted with permission from [18]. Copyright (2010) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
5
to study the particular case of recombining phenyl radicals
on Cu(1 1 1) [38]. Both these studies [37, 38] gave the same
conclusion regarding the diffusion and coupling on Cu(1 1 1),
and for simplicity, the discussion herein is based on the results
presented in [37].
To date, the only study of the complete procedure for a
synthesis protocol making use of the on-surface Ullmann cou-
pling investigated the formation of biphenyl from bromo- and
iodobenzene on Cu(1 1 1), Ag(1 1 1) and Au(1 1 1) [37]. The
overall process is initiated by dehalogenation, followed by the
recombination (diffusion and coupling) of the surface-stabi-
lized phenyl radicals, resulting in biphenyl.
We will begin by discussing the dehalogenation of bro-
mobenzene and iodobenzene on the three surfaces. A typical
energy prole is shown in gure 4(a) and IS, TS and FS of
the dehalogenation is exemplied for bromobenzene on Au(1
1 1) in gure4(b). In the IS, the molecule is physisorbed on
the surface, while in the TS a chemical bond has begun to form
between the halogenated carbon atom and the surface as the
same time as the carbon-halogen bond is elongated. Ultimately,
in the FS the C-Br bond is completely dissociated, with the
resulting phenyl radical and bromine chemisorbed to the
surface. The picture is very similar on all the three surfaces for
both bromo- and iodobenzene. An important point is the chemi-
cal state of the resulting radicals. In the FS the phenyl radical
is chemisorbed on the surface, resulting in the un-paired spin
being quenched, why in literature one frequently use the nota-
tion ‘surface-stabilized radical’ [18]. Often, the term radical is
used quite loosely, but as a general rule, for these type of radi-
cals the un-paired spin is most likely quenched, and the notation
that we have a surface-stabilized radical is more appropriate.
In gure 4(c) the energy barriers (activation energies)
and reaction energies are shown for the six dehalogenation
reactions. Given that both reactions are highly endothermic
in gas phase, with reaction energies of 3.85 eV and 3.33 eV
for bromobenzene and iodobenzene, respectively, it is clear
that all three surfaces have catalytic effects on dehalogena-
tion. For bromobenzene, the barriers range from 1.02 eV
on Au to 0.66 eV on Cu(1 1 1). A similar trend, shifted by
roughly −0.3 eV is found for iodobenzene, with barriers of
0.71 eV on Au(1 1 1) and 0.40 eV on Cu(1 1 1) [37]. Notably,
the reaction energies follow a quite different trend, with larg-
est difference between the two molecules for Au(1 1 1) and a
rather small difference for Cu(1 1 1).
Figure 3. Calculated recombination pathways of CHPR on Cu(1 1 1) and Ag(1 1 1), as indicated. The chemical structure of CHPR is
shown in the inset. Reprinted with permission from [18]. Copyright (2010) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
6
The next step of the reaction is the recombination of two
phenyl radicals into biphenyl. The overall recombination is
associated with two processes, namely diffusion and the cou-
pling between surface-stabilized radicals. As illustrated in
gure 5 the phenyl can follow two types of diffusion. The
slide diffusion, in which the orientation of the phenyl ring is
the same in the IS and the FS, and a ip diffusion in which
the phenyl ring ips between two sites such that its orientation
changes. On all surface, except for Au(1 1 1) for which the
two diffusion modes have the same TS, the ip diffusion has
the lowest barrier. However, the ip diffusion is probably only
possible for molecules of similar size as phenyl, as it has a TS
or intermediate state where the molecule is standing upright on
the surface [37]. Such a conguration would not be possible
for a larger molecule due to the cost of breaking vdW interac-
tions, and denitely not for a molecule with more than one
radical site, as it would require the scission of chemical bonds.
When the molecules have diffused over the surface, they
will eventually come in the near proximity to another mol-
ecule such that they can react. In the case of phenyl radicals,
the closest two molecules can be without reacting is when
they are chemisorbed to the same surface atom, dened
as IS of the coupling reaction. The pathways for coupling
are depicted in gure6. In the TS the two radicals are still
chemisorbed to the surface atoms, and only differs from the
IS in that the two radical carbon atoms are closer to each
other. The energies of the states are dened with respect
of having the two molecules well separated. The activation
energy for coupling was dened as the energy of the TS
rather than the energy difference between TS and IS. The
activation energy for coupling is smallest for Cu(1 1 1) and
largest for Ag(1 1 1), while an intermediate value was found
for Au(1 1 1). For all surfaces the coupling is highly exo-
thermic, manifesting the irreversibility of the covalent bond
formation [37].
Following the argumentation that the slide diffusion is
more relevant than the ip diffusion, it gives that the phenyl
radical has the same trend as CHPR, in which the recombina-
tion is diffusion limited on Cu(1 1 1) and coupling limited
on Ag(1 1 1). For Au(1 1 1), the recombination is considered
being neither diffusion nor coupling limited, as the barrier of
diffusion is similar to that of the coupling.
4. Surface chemistry of terminal alkynes
Molecules functionalized with terminal alkynes participate
in another group of on-surface reactions. In particular, homo-
coupling between terminal alkynes has gained a lot of inter-
est since it was rst reported on a surface in 2012 [39]. This
method has enabled thermal coupling of acetylene functional
groups, creating butadiyne (diacetylene) bridges between
adjacent molecules, and is highly relevant for the realization
of materials related to graphdiyne, a carbon allotrope in which
benzene rings are interlinked by butadiyne bridges. For exam-
ple, extended graphdiyne wires have been synthesized on a
silver surface [2].
The surface chemistry of terminal alkynes has proven to be
extremely versatile, and several reactions have been reported
[40–46]. Maybe most intriguingly, a cyclotrimerization reac-
tion has been observed in a few studies [41–45]. For example,
Zhou et al demonstrated how a linear molecular building blocks
undergo a [2+2+2] cyclotrimerization on Au(1 1 1) [43]. In
another example [44], 1,3,5-tris-(4-ethynylphenyl)benzene
(Ext-TEB) molecules were observed to cyclotrimerize on Au(1
1 1). Notably, the Ext-TEB molecule was also used in the rst
report of the homo-coupling [39], but on Ag(1 1 1). The two
types of reaction products obtained for Ext-TEB on Ag(1 1 1) and
Au(1 1 1), together with the basic principles of the homo-coupling
and cyclotrimerization, are compared in gure7. Furthermore,
if instead deposited on Cu(1 1 1), Ext-TEB dehydrogenates
and forms an unusual deprotonated alkynyl hydrogen bonding
network, inhibiting the coupling between molecules [46]. The
Ext-TEB molecule gives an excellent example of the versatile
surface chemistry of terminal alkynes, where the chemoselectiv-
ity can be tuned by changing the reactivity of the surface.
The versatile surface chemistry of terminal alkynes have
been observed in several other studies. For example, the linear
4,4”-diethynyl- 1,1’:4’,1”-terphenyl molecule gives a multitude
Figure 4. (a) Typical potential energy prole for dehalogenation
reactions, such as (b) the dissociation of bromobenzene on
Au(1 1 1). (c)
E
barrier
(left) and
E
react
(right) for the dissociation of
bromobenzene and iodobenzene on the (1 1 1) facets of Au, Ag,
and Cu. Reprinted with permission from [37]. Copyright (2013)
American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
7
of side reactions on Ag(1 1 1), with coupling motifs of up to ve
monomers, while a high chemoselectivity towards the homo
coupling is obtained when depositing it with a delicately tuned
coverage on the stepped Ag(877) surface [2]. The deposition
of 1,4-diethynylbenzene on Cu(1 1 1) gives rise to a multitude
of reactions, such as homo-coupling and cyclotrimerization
[42], but does not result in the deprotonated alkynyl hydro-
gen bonding network reported for Ext-TEB on Cu(1 1 1) [46].
It should be further noted that while the Au(1 1 1) surface
has been observed to activate mainly cyclotrimerization reac-
tions in a couple of examples [43, 44], other molecules on this
surface have resulted in the homo-coupling among with other
reaction products [41]. Thus, the chemoselectivity toward
homo-coupling versus cyclotrimerization is not simply con-
trolled by the choice of surface, but rather the molecule-sur-
face combination is important.
Figure 5. Energy diagrams for (a) sliding diffusion and (b) ipping diffusion of phenyl on Au(1 1 1), Ag(1 1 1), and Au(1 1 1), where the
top and side views of the paths are depicted in the top panel for (a) Ag(1 1 1) and (b) Au(1 1 1). In both processes, phenyl diffuses between
two adjacent surface atoms rendered darker than other surface atoms. On Au(1 1 1), the ipping and sliding diffusions have identical TSs
and differ only by the relative orientation of the molecule in the IS and FS. The ipping diffusion (b) is a two-step process on Cu(1 1 1) and
Ag(1 1 1). Reprinted with permission from [37]. Copyright (2013) American Chemical Society.
Figure 6. Coupling reaction of two phenyls into biphenyl on Au(1 1 1), Ag(1 1 1) and Cu(1 1 1), depicted for Ag(1 1 1) in the top panel.
Energies are given with respect to having the two phenyl well separated from each other on respective surface. Reprinted with permission
from [37]. Copyright (2013) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
8
For a more extensive account of the work devoted to the
surface chemistry of terminal alkynes see the recent review in
[47]. Hereon I will focus on the theoretical efforts made for
understanding the underlying mechanisms of their multifac-
eted surface chemistry.
4.1. Mechanism of homo-coupling
Two independent theoretical studies of the homo-coupling
have been presented [48, 49], both reaching the conclusion
that the coupling of the terminal alkynes precedes the release
of hydrogen atoms. I will center the discussion around the
dimerization between 1,3,5-triethynyl-benzene (TEB) mol-
ecules on Ag(1 1 1), one of the systems used to initially dem-
onstrate the homo-coupling [39], and reect on how it relates
to other molecules and surfaces.
The initial coupling step between two TEB molecules is
illustrated in gure8. The barrier of coupling the two mol-
ecules (going from IS to
IntS1
trans
) is 0.90 eV [48], which
is the same barrier found for ethynyl-benzene on Ag(1 1 1),
while a slightly smaller barrier (0.79 eV) was found for
Au(1 1 1) [49]. Following the initial coupling, the formed TEB
dimer can exist in a trans (
IntS1
trans
) and cis state (
IntS1
cis
).
In
IntS1
trans
one of the carbon atoms is under-coordinated and
it has been made quite clear that the molecule will exist in the
more stable
IntS1
cis
conguration at some point before con-
tinuing the reaction [48].
The homo-coupling between two TEB monomers is nal-
ized by two subsequent dehydrogenation steps, with barri-
ers of 1.27 eV and 1.53 eV, respectively, as shown in gure9.
Particularly the second barrier is considerable having in mind
that the reaction takes place at 300 K in experiments [39].
Several explanations for this anomaly between theory and
experiments have been proposed. For example, the calcul-
ations give the potential energy landscape at 0 K and it
was shown that temperature effects, in terms of vibrational
enthalpy and entropy, indeed reduce the dehydrogenation
barrier [48]. Another possibility might be that the system
does not have time to thermally equilibrate after the highly
exothermic coupling step. As a consequence, the energy
gained in the coupling is reinvested into the dehydrogenation
steps, reminiscent of hot adsorbates that can form in dissocia-
tive adsorption [50].
An alternative path for initiating the homo-coupling would
be to directly dehydrogenate the terminal alkyne prior to the
coupling. However, calculations have shown that the barrier
is twice that to initially couple two TEB molecules [48]. A
similar barrier was found for dehydrogenating ethynyl-ben-
zene on Ag(1 1 1), while it is slightly lowered to 1.64 eV on
Au(1 1 1) [49]. In other words, initially coupling of terminal
alkynes is signicantly more probable than dehydrogenating
the molecules followed by the coupling.
4.2. Considerations of the cyclotrimerization mechanism
Zhou et al provided insight into the cyclotrimerization between
terminal alkynes by studying the formation of benzene from
three acetylene molecules [43], as shown in gure 10. The
reaction was considered to be non-concerted, where two acet-
ylene rst couple, followed by the reaction with the third mol-
ecule, resulting in benzene. The rst step is smaller than the
second one (compare 1.54 eV to 1.72 eV), suggesting that the
intermediate (INT), in which only two alkynes have coupled,
should be observable. Indeed, such intermediate state has
been observed in STM experiments [43].
One needs to bear in mind that these calculations were
done by using a planar cluster of 14 Au atoms representing
the surface [43], which will be signicantly more reactive
than the real Au(1 1 1) surface. Furthermore, the pathway’s
intermediate state (denoted INT) resembles the cis intermedi-
ate state (
IntS1
cis
) for the homo-coupling, which is a result
of using acetylene as model molecule, for which it is not
possible to differentiate the initial step of cyclotrimerization
and homo-coupling. This aspect is demonstrated in gure11,
which compares the respective mechanisms of homo-coupling
and cyclotrimerization. The study by Zhou et al still provides
valuable insights into the cyclotrimerization between alkynes,
despite that investigations on realistic surfaces and more rep-
resentative molecules are needed for a full appreciation of the
reaction.
5. Surface-assisted cyclodehydrogenation
Cyclodehydrogenation is a category of intramolecular ring-
closure reactions in which a carbon-carbon bond is formed
simultaneously with the release of two hydrogen atoms
(bonded to the carbon atoms in the reactant). In gure12 a
cyclodehydrogenation reaction is demonstrated for an illustra-
tive model system.
Surface-assisted cyclodehydrogenation has been used
together with on-surface Ullmann coupling for the on-sur-
face synthesis of different types of graphene nanoribbons on
Au(1 1 1) and Ag(1 1 1), such as the transformation of a poly-
anthrylene intermediate into a seven carbon atoms wide rib-
bon [1]. It has also been employed to form large polycyclic
Figure 7. The prefered reaction path of the Ext-TEB molecule
depends on the underlying surface: on Ag(1 1 1) the homocoupling
reaction is obtained [39], while the molecules undergo a
cyclotrimerization reaction on Au(1 1 1) [44].
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
9
Figure 8. The homo-coupling between two TEB molecules is initiated by a coupling step, with the hydrogen remaining on the molecules.
The coupled transition state can exist in a trans- and cis-isomer, where the latter is the more stable one. The two hydrogen atoms taking part
in the coupling were rendered red for clarity. Reprinted with permission from [48]. Copyright (2014) American Chemical Society.
Figure 9. Following the initial coupling of two TEB molecules (gure 8), the resulting dimer undergoes two subsequent dehydrogenation
steps, which nalizes the overall homo-coupling. The two hydrogen atoms being split-off were rendered in red for clarity. Reprinted with
permission from [48]. Copyright (2014) American Chemical Society.
Figure 10. Mechanism of the formation of benzene from three acetylene molecules through cyclotrimerization on Au(1 1 1). Reprinted
with permission from [43]. Copyright (2014) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
10
aromatics hydrocarbons (PAHs), so called nanographenes, from
cyclohexa-o-p-o-p-o-p-phenylene (CHP) on Cu(1 1 1) [51].
5.1. Nanographenes
Treier et al performed a rigorous theoretical investigation
of the reaction mechanism of the cyclodehydrogenation of
CHP into a nanographene [51], the main results of which
are illustrated in gure13. The overall reaction procedes in
a six-step process, via the formation of three supplementary
carbon-carbon bonds and abstraction of six hydrogen atoms.
The reaction is initiated by a dehydrogenation step (state 1
to 2), followed by a carbon-carbon coupling and concomitant
dehydrogenation (state 2 to 3). This is accompanied by a third
dehydrogenation (state 3 to 4) and carbon-carbon coupling
(state 4 to 5). The reaction is nalized by a combined carbon-
carbon coupling and dehydrogenation (state 5 to 6), ultimately
leading to the release of a hydrogen molecule (state 6 to 7).
It is quite interesting to note that two of the intermediate
states were observed in STM experiments [51], namely state
3 and 6. These states are followed by two of the rate-limiting
steps of the overall reaction and it is intriguing how well this
is captured by the computed pathway.
5.2. Graphene nanoribbons
The formation of graphene nanoribbons is probably the most
famous, and well-studied, example of on-surface synthesis.
The overall reaction is shown in gure14. Firstly, brominated
bianthryl units couple into polyanthrylene following an
on-surface Ullmann coupling procedure. This is followed
by the cyclodehydrogenation step, yielding the graphene
nanoribbons.
A staggering amount of work has been devoted to the seven
atoms wide armchair graphene nanoribbons initially demon-
strated by Cai et al [1]. However, only a couple of theoretical
studies have been devoted to the mechanism of the cyclodehy-
drogenation reaction [52, 53].
Blankenburg et al studied the mechanism on Ag(1 1 1), and
backed up their theoretical results by STM experiments [53].
Their reaction pathway is illustrated in gure15. In short, they
found that the reaction is initiated by the coupling between
two carbon atoms (S0 to S1), followed by the abstraction of
a hydrogen atom (S1 to S2). Then a hydrogen is tautomer-
ized, in other words migrated between two carbon atoms (S2
to S3) followed by an additional carbon-carbon coupling step
(S3 to S4), which is succeeded by the tautomerization of an
additional hydrogen atom (S4 to S5) and the abstraction of
two hydrogen atoms (S5 to S6). The reaction then procedes
stepwise in a similar fashion. For a more detailed view of the
mechanism see [53].
An interesting aspect of the mechanism presented by
Blankenburg et al is that the cyclotrimerization between two
carbon atoms has a reducing effect on the barrier of cyclotri-
merization between the adjecent anthracene units (at the same
side of the nanoribbon). This was in fact also observed STM
experiments, in which it was possible to activate the cyclotri-
merization in one side of the polyanthrylene, while keeping
the other side unreacted [53]. Another study reported a theor-
etical investigation of the cyclotrimerization on Au(1 1 1)
[52], and came to the same conclusion that the cyclotrimeriza-
tion proceeds in a domino-like fashion, where the cyclodehy-
drogenation between two pairs of anthracene units enable the
coupling between the next units. However, in this study is was
concluded that both sides of the polymer is cyclodehydrogen-
ated before moving to the next units and not the side-wise acti-
vation that was found on Ag(1 1 1) [53]. It is not clear whether
this is a difference between Ag(1 1 1) and Au(1 1 1), or if the
Figure 12. In cyclodehydrogenation two carbon atoms couple
simultaneously as detaching their hydrogen atoms, effectively
resulting in a ring-closure.
Figure 11. Comparison between the mechanism of (a) homo-coupling and (b) cyclotrimerization between terminal alkynes, with the
metal surface indicated as Me. If using acetylene as a model compound, in which the group R is represented by a hydrogen atom, the rst
intermediate state of the homo-coupling and the intermediate state of the cyclotrimerization are identical.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
11
model system used for studying the cyclodehydrogenation on
Au(1 1 1) did not capture all aspects of the reaction, and con-
stitutes an open question still awaiting an answer.
6. Other types of on-surface reactions
Ullmann coupling, coupling between terminal alkynes and
cyclodehydrogenation are probably the three most thoroughly
studied on-surface reactions from theory, although a lot of
work remains to fully comprehend them. Here I will highlight
some of the work that has been performed to elucidate mech-
anisms for other types of reactions.
Throughout this topical review different types of cycliza-
tion reactions have been accounted for. In fact, several addi-
tional examples of cyclization on surfaces exist. For example,
Figure 13. Calculated reaction mechanism for the triple cyclodehydrogenation of CHP into a nanographene on the Cu(1 1 1) surface.
Reprinted by permission from Macmillan Publishers Ltd: Nature Chem. [51], copyright (2011).
Figure 14. Schematics of the formation of graphene nanoribbons.
In a rst step, brominated bianthryl units form polyanthrylene
through on-surface Ullmann coupling. Then the polymer
cyclodehydrogenates, resulting in the graphene nanoribbon [1].
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
12
Yang et al demonstrated how acetyl-functionalized molecules
couple either in a dimerization or a cyclotrimerization reaction
[54]. Theoretical considerations of the reaction pathway of the
initial dimerization between two molecules concluded that the
reaction is initiated by the dehydrogenation of the methyl in
an acetyl group, illustrated in gure 16(a). From this point
there are two competing pathways: either the dehydrogen-
ated molecule reacts with an intact molecule (gure 16(b)), or
two dehydrogenated molecules react (gure 16(c)). The latter
alternative requires a sufcient concentration of dehydrogen-
ated molecules, but since the initial dehydrogenation is signif-
icantly larger than coupling two molecules, it was concluded
that the coupling between an intact and a dehydrogenated
molecule is more likely [54]. The initial coupling is followed
by subsequent dehydroxylation and dehydrogenation steps,
and the abstraction of oxygen was supported by XPS experi-
ments [54].
In another study, an azide-alkyne cycloaddition (gure 17)
was observed on Au(1 1 1) [55]. Two reaction pathways were
considered by DFT calculations, one occurring at an atomi-
cally at Au(1 1 1) surface, and another in the presence of an
Au adatom, as shown in gure18. Both pathways have a bar-
rier of around 0.7 eV, which was also found for the reaction in
gas phase, concluding that the Au surface has an insignicant
chemical impact on the reaction [55]. This gives an excellent,
but rare, example where the surface instead of catalyzing the
reaction, provides a support that enables the reaction by low-
ering the degrees of freedom of the molecular building blocks
into two dimensions. Further examples of cyclization reac-
tions have been reported. For example, Bergman cyclization
Figure 15. Calculated reaction mechanism for the cyclodehydrogenation of poly-anthracene into a graphene nanoribbon on the Ag(1 1 1)
surface. Reprinted with permission from [53]. Copyright (2012) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
13
was demonstrated on Cu(1 1 0) [56], but for which no reaction
pathways have been considered from theory.
As a nal example, Sun et al studied reactions between
quaterphenyl molecules on Cu(1 1 0) in STM experiments,
observing that the molecules couple via a selective C-H acti-
vation of the meta-carbon site of the terminal phenyl group [57].
The experimentally observed behavior was explained by
theor etical modeling, which concluded that the activation
energy for splitting of a hydrogen from the meta-carbon has a
signicantly smaller barrier compared to splitting-off hydro-
gen from other carbon atoms.
7. Outlook
In this topical review I have outlined some of the theoretical
work that has been performed to understand reactions respon-
sible for the formation of covalent nanostructures on surfaces,
with particular focus on the on-surface Ullmann coupling,
Figure 16. Comparison of different reaction pathways involving acetyls. (a) Dehydrogenation of the acetyl’s methyl group, (b) coupling
between an intact and a dehydrogenated molecule, and (c) coupling between two dehydrogenated molecules. Valence bond structures have
been included for reactants, intermediates and products for the different pathways, with the metal surface denoted as Me. Note that in
(b) IntS2 and FS have the same chemical structure and differ only by the adsorption site of the abstracted OH-group. Adapted with
permission from [54]. Copyright (2015) American Chemical Society.
Figure 17. Schematics of the azide-alkyne cycloaddition.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
14
surface-chemistry of terminal alkynes and surface-assisted
cyclodehydrogenation.
Regarding the on-surface Ullmann coupling, theoretical
modeling has successfully modeled the pathways for specic
systems. However, the predictive power of the current knowl-
edge is rather limited, and we need to put our forthcoming
efforts in systematic studies of on-surface reactions. A part-
icularly important milestone would the development of a
theory that could predict whether a specic molecule-surface
combination gives rise to a coupling-limited or a diffusion-
limited recombination process. This would specically guide
the formation of ordered covalent 2D materials, for which
a coupling limited process is a prerequisite. Furthermore,
to date, theoretical studies have exclusively considered pro-
cesses on atomically at surfaces. However, as seen numerous
of times in experiments, metal-organic intermediate structures
can play an important role, possibly enabling self-healing
protocols. Insight into the exact role of metal adatoms would
therefore be of immediate interest.
Molecules functionalized with terminal alkynes have dem-
onstrated a highly versatile surface chemistry. Although there
are some aspects that remain to be unraveled about the homo-
coupling reaction, the most immense theoretical effort will
be to comprehend other possible types of reactions terminal
alkynes can undergo. In particular, we need a more sophis-
ticated understanding of the cyclotrimerization, and the next
years theory is expected to make important contributions in
this regard. In the long-term perspective, theoretical modeling
should aim at nding general design rules governing the che-
moselectivity of the surface chemistry of terminal alkynes.
Cyclodehydrogenation is probably the most studied, and
well-understood, on-surface reaction from theory. Although
there are some issues to unravel about how the reaction differs
between various surfaces, the on-surface synthesis eld is in a
more urgent need for theoretical studies about other reactions.
Here we have focused on the work where theory has pro-
vided input into on-surface reaction mechanisms. However,
for the majority of the reported reactions, information of
mechanisms is missing. It is only the last years that the comp-
uter resources have become sufcient to study on-surface
reactions in a more routinely and systematic manner. During
the next years it is anticipated that theoretical insights, and
with this our appreciation, of on-surface synthesis will grow
with a spiralling rate.
References
[1] CaiJ etal 2010 Nature 466470
[2] CireraB, ZhangYQ, BjörkJ, KlyatskayaS, ChenZ,
RubenM, BarthJV and KlappenbergerF 2014 Nano Lett.
141891
[3] BieriM etal 2009 Chem. Commun. 6919
[4] Sanchez-ValenciaJR, DienelT, GröningO, ShorubalkoI,
MuellerA, JansenM, AmsharovK, RufeuxP and FaselR
2014 Nature 51261
[5] KohnW and ShamL 1965 Phys. Rev. 140A1133
[6] GrimmeS 2006 J. Comput. Chem. 271787
[7] DionM, RydbergH, SchröderE, LangrethDC and
LundqvistBI 2004 Phys. Rev. Lett. 92246401
[8] RuizV, LiuW, ZojerE, ScheferM and TkatchenkoA 2012
Phys. Rev. Lett. 108146103
[9] BürkerC, FerriN, TkatchenkoA, GerlachA, NiederhausenJ,
HosokaiT, DuhmS, ZegenhagenJ, KochN and SchreiberF
2013 Phys. Rev. B 87165443
[10] KlimešJ, BowlerDR and MichaelidesA 2010 J. Phys.:
Condens. Matter 22022201
[11] MittendorferF, GarhoferA, RedingerJ, KlimešJ, HarlJ and
KresseG 2011 Phys. Rev. B 84201401
Figure 18. Calculations of azide-alkyne cycloaddition on Au(1 1 1) for an atomically at surface ((A)–(C)) and aided by an Au adatom
((D)–(F)). Reprinted with permission from [55]. Copyright (2013) American Chemical Society.
J. Phys.: Condens. Matter 28 (2016) 083002

Topical Review
15
[12] HamadaI 2014 Phys. Rev. B 89121103
[13] BjörkJ and StafströmS 2014 ChemPhysChem 152851
[14] MatenaM, BjörkJ, WahlM, LeeTL, ZegenhagenJ,
GadeLH, JungTA, PerssonM and StöhrM 2014 Phys.
Rev. B 90125408
[15] HenkelmanG, UberuagaBP and JónssonH 2000 J. Chem.
Phys. 1139901
[16] HenkelmanG and JónssonH 1999 J. Chem. Phys. 1117010
[17] KästnerJ and SherwoodP 2008 J. Chem. Phys. 128014106
[18] BieriM etal 2010 J. Am. Chem. Soc. 13216669
[19] XiM and BentBE 1992 Surf. Sci. 27819
[20] GrillL, DyerM, LafferentzL, PerssonM, PetersMV and
HechtS 2007 Nat. Nanotechnol. 2687
[21] LafferentzL, EberhardtV, DriC, AfrichC, ComelliG, EschF,
HechtS and GrillL 2012 Nat. Chem. 4215
[22] FanQ, WangC, HanY, ZhuJ, HieringerW, KuttnerJ, HiltG
and GottfriedJM 2013 Angew. Chem. Int. Ed. 524668
[23] ChenM, XiaoJ and SteinruckH 2014 J. Phys. Chem. C
1186820
[24] Rastgoo-LahroodA, BjörkJ, HecklWM and LackingerM
2015 Chem. Commun. 5113301
[25] BasagniA, SedonaF, PignedoliCA, CattelanM, NicolasL,
CasarinM and SambiM 2015 J. Am. Chem. Soc.
1371802
[26] FanQ, WangT, LiuL, ZhaoJ, ZhuJ and GottfriedJM 2015
J. Chem. Phys. 142101906
[27] MorchuttC, BjörkJ, KrotzkyS, GutzlerR and KernK 2015
Chem. Commun. 512440
[28] SimonovKA 2015 ACS Nano 98997
[29] SmykallaL, ShukrynauP, KorbM, LangH and HietscholdM
2015 Nanoscale 74234
[30] BegganJP, BoyleNM, PryceMT and CafollaAA 2015
Nanotechnology 26365602
[31] ZhangH etal 2015 J. Am. Chem. Soc. 1374022
[32] WangW, ShiX, WangS, Van HoveMA and LinN 2011 J.
Am. Chem. Soc. 13313264
[33] Di GiovannantonioM etal 2013 ACS Nano 78190
[34] GutzlerR etal 2014 Nanoscale 62660
[35] EichhornJ, NieckarzD, OchsO, SamantaD, SchmittelM,
SzabelskiPJ and LackingerM 2014 ACS Nano
87880
[36] EichhornJ, StrunskusT, Rastgoo-LahroodA, SamantaD,
SchmittelM and LackingerM 2014 Chem. Commun. 507680
[37] BjörkJ, HankeF and StafströmS 2013 J. Am. Chem. Soc.
1355768
[38] NguyenMT, PignedoliCA and PasseroneD 2011 Phys.
Chem. Chem. Phys. 13154
[39] ZhangYQ etal 2012 Nat. Commun. 31286
[40] CireraB, ZhangYQ, KlyatskayaS, RubenM,
KlappenbergerF and BarthJV 2013 ChemCatChem 53281
[41] GaoHY, WagnerH, ZhongD, FrankeJH, StuderA and
FuchsH 2013 Angew. Chem. Int. Ed. 524024
[42] EichhornJ, HecklWM and LackingerM 2013 Chem.
Commun. 492900
[43] ZhouH, LiuJ, DuS, ZhangL, LiG, ZhangY, TangBZ and
GaoHJ 2014 J. Am. Chem. Soc. 1365567
[44] LiuJ, RufeuxP, FengX, MüllenK and FaselR 2014 Chem.
Commun. 5011200
[45] RissA etal 2014 Nano Lett. 142251
[46] ZhangYQ etal 2015 J. Phys. Chem. C 1199669
[47] KlappenbergerF, ZhangYQ, BjörkJ, KlyatskayaS, RubenM
and BarthJV 2015 Acc. Chem. Res. 482140
[48] BjörkJ, ZhangYQ, KlappenbergerF, BarthJV and
StafströmS 2014 J. Phys. Chem.
C 1183181
[49] GaoHY, FrankeJ, WagnerH, ZhongD, HeldPA, StuderA
and FuchsH 2013 J. Phys. Chem. C 11718595
[50] MeyerJ and ReuterK 2014 Angew. Chem. Int. Ed. 534721
[51] TreierM, PignedoliCA, LainoT, RiegerR, MüllenK,
PasseroneD and FaselR 2011 Nat. Chem. 361
[52] BjörkJ, StafströmS and HankeF 2011 J. Am. Chem. Soc.
13314884
[53] BlankenburgS, CaiJ, RufeuxP, JaafarR, PasseroneD,
FengX, FaselR and PignedoliCA 2012 ACS Nano 62020
[54] YangB, BjörkJ, LinH, ZhangX, ZhangH, LiY, FanJ, LiQ
and ChiL 2015 J. Am. Chem. Soc. 1374904
[55] Díaz AradoO, MönigH, WagnerH, FrankeJH,
LangewischG, HeldPA, StuderA and FuchsH 2013 ACS
Nano 78509
[56] SunQ, ZhangC, LiZ, KongH, TanQ, HuA and XuW 2013
J. Am. Chem. Soc. 1358448
[57] SunQ, ZhangC, KongH, TanQ and XuW 2014 Chem.
Commun. 5011825
J. Phys.: Condens. Matter 28 (2016) 083002