Surface chemistry has played a prominent role in organic synthesis for centuries,
most notably in the form of heterogeneous catalysis. The discovery of platinum-catalyzed
combustion of gaseous mixtures by Davy[1 ] was only the beginning of a rich history of surface chemistry that would prove to
have a tremendous impact on chemistry at large. Although the chemical transformations
themselves take place at the surface, the way in which heterogeneous catalysis is
employed is most often three-dimensional, with the catalyst dispersed in the reaction
medium or the medium being passed through a catalyst bed. Furthermore, the rugged
and often ill-defined surface of the catalyst can hardly be considered two-dimensional.
With the advent of on-surface synthesis, however, reactions have started to become
truly two-dimensional. Here, molecules are deposited onto an atomically flat surface
under typically ultrahigh vacuum (UHV) conditions. The reactants can move laterally
over the surface but have no freedom to ‘jump off’ the surface: this would imply desorption,
which, under UHV conditions, is permanent. Reactions between molecules therefore take
place in-plane. The rules of the game of on-surface synthesis therefore allow for
a distinct type of chemistry that facilitates the controlled synthesis of new types
of nanostructures. Another powerful aspect of this technique is its synergy with conventional
chemical methods: The precursor molecules for on-surface reactions can be prepared
by solution-based methods before being transferred to the UHV setup. Indeed, on-surface
reactions can be applied to a large proportion of molecules that can be tailored in
the chemistry lab. Therefore, on-surface synthesis can be particularly powerful when
used in conjunction with solution-phase synthesis.
Peter Jacobse was born in Emmen and raised near the city of Groningen, the Netherlands. He studied
chemistry and physics at the University of Groningen, and after completing his BSc
(2012), he moved to Utrecht to complete his studies in the field of nanomaterials
science. He obtained an MSc (Honors, cum laude) from Utrecht University in 2014 under
the supervision of Dr. Ingmar Swart. He wrote a research proposal about advancing
the study of graphene nanoribbons, for which he was awarded a grant from the Debye
Institute for Nanomaterials Science at Utrecht University, where he works to this
day.
In on-surface synthesis, precursor molecules are commonly deposited onto a surface
by evaporation from an effusion cell followed by adsorption, after which heating or
irradiation may be carried out to induce chemical reactions. These reactions may be
intramolecular, but in the quest for extended nanostructures, intermolecular coupling
reactions are particularly desirable. On-surface chemistry has proven to be a valuable
technique to obtain high-quality one- or two-dimensional materials like graphenes,
polyphenylenes, fullerenes, and graphene nanoribbons (GNRs).[2 ]
[3 ]
[4 ]
[5 ]
[6 ]
[7 ]
[8 ] Restriction of adsorbates to the surface effectively eliminates nonselective out-of-plane
reactions, facilitating the creation of atomically well-defined low-dimensional nanostructures.
Despite its virtues, on-surface synthesis could only take off as a new emerging field
with the advent of sensitive surface-probing techniques. Traditional chemical synthesis
mostly relies on spectroscopic techniques and other bulk characterization methods,
which are ill-suited for the typically obtained submonolayer coverages on single crystals.
Therefore, it was not until the development of highly sensitive methods such as scanning
probe microscopy that on-surface chemistry could really develop into a field of its
own.
The most used scanning probe technique in on-surface chemistry is scanning tunneling
microscopy. Here, a bias voltage is applied between a conductive tip and sample. When
the tip-sample distance is on the order of a (few) nanometer(s), electrons can tunnel
across the vacuum gap between tip and surface, creating a current which may be used
as a feedback parameter to control the tip height.[9 ] When scanning the tip across a surface littered with molecular adsorbates, the tunnel
current depends on both the physical height as well as the available electronic energy
levels that may contribute to the tunneling process. An STM topograph of an absorbed
species may thus be regarded as a convolution of its geometric and electronic structure.
As far as electronic structure is concerned, STM probes the low-energy states, which
in large aromatic systems, such as polycyclic aromatic hydrocarbons (PAHs) and nanographenes,
typically arises from π-orbitals. STM is a powerful technique in identifying molecules
and nanostructures on surfaces, and has the additional advantage that the bias voltage
can be used as an extra degree of freedom to facilitate a detailed electronic analysis.[10 ] However, the pure geometric structure – the framework of bonds and atoms – cannot
be identified as it is clouded by electrons from frontier orbitals, as well as non-resonant
tunneling processes.
In contrast, noncontact atomic force microscopy is a more recently established technique
– and admittedly, a much more demanding one – that is capable of imaging the chemical
structure of molecules with atomic resolution.[11 ] In nc-AFM, an AFM cantilever with high stiffness (typically 1,800 N m–1 ) is driven at its resonance frequency. Upon approaching the tip to the surface, the
resonance frequency shifts due to electrostatic, van der Waals, and Pauli interactions
with the sample. This landscape of interactions is established by the total electron
density, rather than the resonant electron density, and therefore includes all covalent
bonds. As a result, an nc-AFM scan can actually uncover the molecular framework of
bonds and atoms, and allows to peek ‘inside’ a single molecule.[12 ] The remarkable capability of nc-AFM to really reveal the molecular ‘skeleton’ with
atomic resolution is unprecedented.
The first realization of intramolecular contrast with nc-AFM came with the seminal
visualization of the individual benzene rings in the backbone of pentacene.[11 ] Soon after, a plethora of other, mostly aromatic molecules were examined, each time
revealing images of their respective backbones of bonds and rings in a way that closely
resembled the structural models from chemistry textbooks.[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ] Noteworthy is the extensive effort in elucidating the structure of asphaltene molecules.[20 ] Like snowflakes, no two of these large PAHs are alike, and therefore, their structure
cannot be identified with ensemble techniques. On the contrary, by using nc-AFM, each
single asphaltene could be identified by the atoms and rings displayed in the respective
images.
Nc-AFM is a powerful tool to uncover the structure of known or unknown molecules and
becomes even more interesting when the molecules themselves engage in chemistry. Reactions
on surfaces can be induced by means of heating,[2 ]
[3 ]
[21 ]
[22 ] irradiation,[23 ] or even by employing voltage pulses with an STM tip.[24,25 ] In each case, nc-AFM can be used to directly compare the structure of the molecular
adsorbates before and after the transformation. The unique insight that this provides
is particularly useful in reactions with multiple steps, multiple products, or products
that cannot be unambiguously identified with other methods.
A beautiful example of a reaction with multiple steps is the tip-induced debromination
and reversible Bergman cyclization of 9,10-dibromoanthracene.[26 ] By application of a voltage pulse, Schuler et al. succeeded in selectively cleaving
off the bromine atoms, yielding the 9-bromo-10-anthryl radical and subsequently the
9,10-anthrylene diradical. Both the reactant and the two individual radicals were
visualized with nc-AFM, as can be seen in Figure [1 ] (a). Surprisingly, it was found that further application of voltage pulses induced
reversible switching of the molecule into three different isomeric states. The isomers
turned out to be the diradical and two equivalent retro-Bergman products, with the
outermost ring cleaved internally to give an expanded, ten-membered ring. By using
nc-AFM, the team was able to visualize the ring-expanded molecules, clearly revealing
the cleaved bond and giving indisputable evidence of the transformation.
Figure 1 a) Products and intermediates of the Bergman cyclization, identified with nc-AFM.
Adapted by permission from Macmillan Publishers Ltd: Nature Communications , ref. 26, Copyright (2016). b) Coupling and cyclization reactions of oligo-(phenylene-1,2,-ethynylenes).
Adapted by permission from Macmillan Publishers Ltd: Nature Chemistry , ref. 27, Copyright (2016).
Other elegant examples of isomers and intramolecular transformations analyzed with
nc-AFM include the subtle differentiation of cumulene and aryne tautomers in a PAH,[28 ] cyclization of triangular dehydrobenzoannulene,[22 ] reactions of oligo-(phenylene-1,2-ethynylenes)[27 ]
[29 ] (see Figure [1, b ]), sulfur elimination from diphenanthrothiophene,[30 ] and conformational changes in tetraphenylporphyrins[31 ] and dibenzo[a ,h ]thianthrene.[32 ] The last three examples show that nc-AFM can even be used on molecules that are
not completely planar, although at the expense of losing contrast in regions that
are closer to the surface. All examples mentioned above are elegant instances of using
AFM in visualizing different products, reactants, and intermediates within single
molecules. In the rest of this article, we will review its application to intermolecular
reactions.
Figure 2 a) Synthesis of 7-AGNR from DBBA. b) Synthesis of boron-doped 7-AGNR from 9,10′′-dibromo-9′,10′-diborateranthryl.
Adapted by permission from Macmillan Publishers Ltd: Nature Communications , refs. 50, 56. Copyright (2013, 2015).
In the context of intermolecular on-surface reactions tracked with AFM, the dimerization
and polymerization of enediynes,[29 ] the formation of a metal-organic framework from 4,9-diaminoperylene-quinone-3,10-diimine,[33 ] and coupling of porphine onto a graphene sheet[34 ] are noteworthy examples. Coupling reactions also play an important role in the bottom-up
assembly of potentially extremely useful nanostructures: GNRs. GNRs are particularly
promising structures as they are based on the parent material graphene, which features
unprecedented electronic properties such as high-charge carrier mobility and low effective
mass. Their one-dimensional, quantum-confined nature allows for additional advantageous
properties such as a finite and tunable band gap, where fine-tuning of the electronic
structure can be achieved through tailoring the width, edge structure, and doping.[35 ]
[36 ]
[37 ]
[38 ]
[39 ]
[40 ]
[41 ]
[42 ]
[43 ]
[44 ]
[45 ] As such, they hold promise for use in future nanoelectronics. In their pioneering
work, Cai et al. presented the first bottom-up synthesis of atomically well-defined
GNR on surface.[46 ] The ribbons were fabricated in a two-step process from 10,10′-dibromo-9,9′-bianthryl
(DBBA) on an Au(111) surface. As shown in Figure [2 ] (a), the first step is a thermally induced debromination, giving rise to surface-stabilized
radicals that couple to form staggered polyanthrylene chains.[47 ]
[48 ]
[49 ] The subsequent annealing step takes place at higher temperature and induces cyclodehydrogenation
of the polyanthrylene into seven-atom wide nanoribbons with armchair-type edges (7-AGNR).
Indeed, STM images obtained from samples prepared with omission of the annealing step
show a characteristic pattern of protrusions as expected from the staggered anthryl
units in polyanthrylene chains. Furthermore, the 7-AGNR were unambiguously resolved
as such with nc-AFM (see Figure [2, a ]).[50 ] In the subsequent development of more types of nanoribbons, on-surface synthesis
has played a dominant role.[51 ] Not only does it facilitate the fabrication of atomically defect-free structures
that are difficult to obtain with solution-based methods, but is also inherently compatible
with scanning probe techniques that allow for detailed characterization.[4 ]
[5 ]
,
[52 ]
[53 ]
[54 ]
[55 ]
Further tailoring of the electronic properties of graphene nanoribbons can be performed
by making use of heteroatom doping. For example, nitrogen can be incorporated by substituting
phenyl groups with pyridyl groups in the precursor molecules.[57 ]
[58 ]
[59 ] Both Cloke et al. and Kawaii et al. showed that the bottom-up assembly could also
be used to produce boron-doped 7-AGNR, by making use of a precursor that can be thought
of as a functionalized analogue of DBBA.[56,60 ] This molecule has a single diboraanthracene moiety sandwiched between two bromoanthryl
extremities (Figure [2, b ]). The radical coupling takes place at the bromoanthryl groups in the same exact
way as in DBBA, and may thus be thought to be ‘blind’ for the embedded diboraanthracene.
However, when the subsequent cyclodehydrogenation step forces the ‘previously unnoticed’
boron-substituted anthryl units to enter the lattice, it is not difficult to notice
an analogy with the Trojan horse. Nc-AFM images of the product show the same structure
as the undoped 7-AGNR, but with characteristic variations in σ-bond contrast due to
the substitutional boron atoms.
The boron-doped 7-AGNR is just one of many different GNRs that were produced by cleverly
designing and modifying precursor molecules.[57 ]
,
[61 ]
[62 ]
[63 ] The preparation requires, in each case, a number of solution-phase synthesis steps.
Importantly, aryl-aryl coupling reactions similar to those used on surface may already
play an important role in the fume hood, before the molecules are ready to enter the
UHV setup. For example, DBBA itself has been produced by Kumada and Suzuki cross-coupling
reactions of anthryl bromides.[50,64 ] Suzuki coupling has also been used in the preparation of the precursors for the
13-atom-wide GNR, by attaching biphenylboronic acid and thiophenylphenylboronic acid
onto the bianthryl core.[61,65 ] In the latter case, a tetrabromo coupling partner was used, but fortunately the
coupling turned out to be selective towards the desired disubstituted product. The
Ullmann coupling on the surface, on the other hand, is a homocoupling, and does not
permit the same level of selectivity as can be achieved in solution. These results
highlight the importance to control the different possible coupling steps, so that
they can be carried out in the right sequence. Since the large majority of aryl-aryl
cross-coupling reactions rely on aryl halides, the greatest synthetic flexibility
can in principle be achieved by carrying out the on-surface coupling with the most
strongly bound halogen. This recognition led us to pursue the question whether aryl
chlorides could be used instead of aryl bromides in the synthesis of GNR.[64 ]
We started by synthesizing 10,10′-dichloro-9,9′-dibromo-bianthryl (DCBA): the dichloro
analogue of DBBA. Interestingly, when thermally treated in a similar way as with DBBA,
DCBA was found to give rise to extensive planar aromatic networks, rather than 7-AGNRs.
Nc-AFM clearly revealed these networks to consist of randomly interconnected bisanthene
units (bisanthene being the cyclodehydrogenation product of bianthryl; see Figure
[3 ]). To figure out why these disordered structures were obtained, we tried to identify
mechanistic intermediates by conducting the experiment at a range of different temperatures.
At 200 °C, we observed a large number of bisanthene molecules as well as coupled bisanthenes.
Once again, we used nc-AFM to reveal their structure, allowing us to discover that
the coupling products almost exclusively feature interconnecting bonds from the ‘middle’
C(10) position to the ‘corner’ CH(3) position, as can be seen in Figure [3 ] (where we use the numbering of the parent bianthryl molecules). What could be the
reason that the usual radical-radical coupling is blocked, and intercepted by this
alternative pathway?
The final piece of the puzzle was found by preparing a sample at the even lower temperature
of 120 °C, where the vast majority of DCBA molecules remain unaffected and only a
few undergo chemical transformations. Instead of coupling together, we found isolated
molecules of monochlorobisanthene (see Figure [3 ]), with the chlorine clearly distinguishable from the nc-AFM images. These images
proved that cyclodehydrogenation can actually preceed dechlorination, effectively
reversing the order of Ullmann coupling and cyclodehydrogenation as compared to DBBA.
This result can be ascribed to the increased aryl halide bond strength in DCBA compared
to DBBA, causing the dehalogenation temperature to surpass the cyclodehydrogenation
temperature. Since the radicals in the reaction mechanism are planar bisanthene radicals,
rather than the staggered bianthryl radicals, the natural coupling behavior is suddenly
severely impeded by steric effects. The least sterically hindered position, the corner,
then remains as the most likely candidate for coupling, and indeed, bisanthene radicals
couple radical-to-corner, or from CH(3) to C(10). In conclusion, although we established
that GNRs could not be obtained from DCBA, our analysis of mechanistic intermediates
using nc-AFM allowed us to fully reveal their behavior on a surface.
Another reaction mechanism that we were interested in was the formation of nanoribbons
from bianthryls on Cu(111). Intuitively, one may think that the Ullmann coupling and
cyclodehydrogenation should take place in the exact same way as on Au(111). However,
in STM the edges of the ribbons were found to exhibit a notable sawtooth-like appearance.
This peculiar feature was initially ignored by Simonov et al.,[66 ] but at the same time prompted Han et al. to claim the product to be the counterintuitive
3,1-chiral GNR (see Figure [4 ]).[67 ] The dispute that followed between the two groups could not be settled by STM imaging.[68 ]
[69 ]
[70 ]
[71 ]
Figure 3 Formation mechanism of oligo- and polybisanthenes from DCBA on Au(111). Reprinted
(adapted) by permission from Jacobse, P. H. et al. Angew. Chem. Int. Ed. 2016 , 55 , 13052–13055,[64 ] Copyright (2017) American Chemical Society. The AFM images show the products of
hydrogenation (passivation) of the corresponding radical intermediates.
Figure 4 Formation mechanism of 3,1-chiral graphene nanoribbons on Cu(111). The substituent
X = H, Cl, Br. Reprinted (adapted) by permission from Schulz, F. et al. J. Phys. Chem. C 2017 , 121, 2896-2904,
[72 ] Copyright (2017) American Chemical Society.
In order to solve the controversy, we started by growing GNRs from DBBA on Cu(111)
in the same way as Han and Simonov, and performed nc-AFM experiments to indisputably
reveal the product to be the 3,1-chiral GNR.[72 ] A simultaneous effort by Sánchez-Sánchez et al. corroborated these results.[73 ] The inevitable conclusion is that the Ullmann coupling, which should interconnect
the C(10) positions of bianthryls, is compromised; in contrast, the chiral nanoribbons
feature bonds between the CH(2) positions of adjacent monomers. We continued the experiments
by attempting the synthesis with both DCBA and unsubstituted bianthryl. Surprisingly,
in all three cases we obtained the 3,1-chiral nanoribbon. Evidently, the halogen at
the C(10) position does not play an important role in the reaction. Therefore, we
concluded that bianthryls could be activated directly at the CH(2) positions by the
copper surface.
We decided to delve more deeply into this problem by attempting to identify intermediates
in the mechanism. As noted by Simonov, the GNR formation from DBBA on Cu(111) is a
three-step process, with an additional intermediate as compared to the two-step process
on Au(111). Instead of going through a single polyanthrylene stage – or rather its
chiral GNR counterpart – two different staggered polymer structures appear in successive
stages with DBBA. The first polymer, with large spacing, was thought to originate
from an organometallic chain containing copper atoms in between the adjacent monomers.
Unfortunately, due to the staggered nature of the chains, the hypothetical interstitial
copper atoms are hidden away from view as far as nc-AFM is concerned. We noted that
the long-periodicity intermediate could only be obtained using DBBA and not for either
DCBA or bianthryl, suggesting that the nature of the intermediate is indeed related
to a debrominated product, and may therefore be the C(10)-coupled product. By performing
nc-AFM, we could determine the orientation of the monomers in both chains. This allowed
us to conclude that the short-periodicity chain is the CH(2)–CH(2) coupled polymer.
The long-periodicity chain appeared to be connected in the regular Ullmann-type fashion,
with extra spacing between the monomers likely originating from the copper atoms.
We completed our analysis by tip-manipulation experiments, and by looking into the
analogous 5-AGNR formation from dibromoperylene.[72 ] This reaction only involves planar products, and indeed allowed us to identify both
organometallic and covalent stages in the reaction. By combining all the information,
we could establish the mechanism by which the nanoribbons form on Cu(111). Even though
DBBA debrominates and takes the first step to assemble into the usual polyanthrylene
chain and subsequent 7-AGNR, the Ullmann coupling is still intercepted by activation
of CH(2) and covalent coupling at this position. On the other hand, activation at
the C(10) position plays no role in bianthryl nor in DCBA. In these cases, the organometallic
‘Ullmann’ intermediate is absent and coupling at the CH(2) positions produces the
covalent polymer directly. Subsequent cyclodehydrogenation produces the 3,1-chiral
nanoribbon in all three cases.
The elucidation of the formation mechanism of chiral nanoribbons on Cu(111), as well
as that of the oligo- and polybisanthenes on Au(111) are cases where uncovering the
structure of products and intermediates provides a unique insight into the processes
occurring in on-surface chemistry. As nanostructures continue to grow more sophisticated
and complex, nc-AFM will prove to be an important tool to elucidate molecular frameworks
of products and intermediates, providing a unique understanding that cannot be obtained
with any other method.
