Chapter 1: Surface Self-assembly of Functional Supramolecular Networks

The use of self-assembly processes to prepare nanoscale structures lies at the very heart of supramolecular chemistry.¹  Indeed, through the use of self-assembly it is possible to prepare highly complex molecular architectures with targeted properties that can avoid lengthy, multi-step synthetic pathways. In the initial stages of research in the field of supramolecular chemistry the majority of studies were performed in the solution phase²  and more recently solid-state supramolecular chemistry has been developed combined with the wider field of crystal engineering.³  However, more recently the field of surface-based supramolecular chemistry has begun to be explored which has led to a significant interest in developing low-dimensional, predominantly two-dimensional, structures.^4–10  The developments in this area are the focus of this chapter, including the synthetic approaches to such structures and the detailed characterisation of the resulting two-dimensional arrays that can be achieved using scanning probe microscopies.

Surface-based chemistry has long been utilised for assembling arrays of molecules. Perhaps the most widely studied area of such chemistry is the development of self-assembled monolayers (SAMs) of thiolate molecules adsorbed onto Au(111) substrates.¹¹  The development of SAMs has been expansive and lies at the heart of many studies that attach molecules and, in some cases supramolecular arrays, to surfaces.¹²  Recent studies have demonstrated the use of supramolecular interactions between thiolate species to control the SAM formation.¹³  In the case of Au-thiolate SAMs and in many other instances deposition of molecules onto surfaces often leads to close-packed arrays that have well-defined arrangements; such arrays typically rely on van der Waals interactions and simple geometric preferences. However, the concepts of supramolecular chemistry open up the possibility of using stronger intermolecular interactions to create more complex and potentially porous structures. The logical extension of this approach is to use stronger interactions, ultimately leading to covalent coupling reactions to create robust low-dimensional arrays.

The emphasis of this chapter will be on studies where molecules are specifically designed in an attempt to control relative molecular organisation. In particular, the chapter will focus on the use of stronger supramolecular interactions, such as hydrogen-bonding, van der Waals interactions and lastly covalent coupling, to form extended arrays that propagate in two dimensions, parallel to the surface. Two-dimensional self-assembly of molecules on surfaces, when combined with scanning probe techniques, provides direct evidence of the potential of this approach. In some instances, examples show that such structures can be used to trap diffusing species as guests, in a similar fashion to porous architectures constructed in three-dimensional solids such as metal-organic frameworks (MOFs)¹⁴  and covalent-organic frameworks (COFs).¹⁵ 

The other main difference between solution-phase supramolecular chemistry and surface-based systems is of course the surface itself. Firstly, the surface defines a two-dimensional boundary upon which the self-assembly process is developed and secondly, the surface is far from innocent in the reaction process. Suitable surfaces often studied include Au(111),¹⁶   [Ag(Si(111))]¹⁷  and highly-oriented pyrolytic graphite (HOPG).¹⁸  However recent studies have started to explore other substrates including graphene¹⁹  and boron-nitride monolayers.²⁰  The choice of substrate is typically driven by their tendency to adopt weak interactions with organic molecules, which are commonly the focus of self-assembly studies. The surface organisation needs to be strongly influenced by intermolecular, supramolecular interactions, rather than surface effects to allow ready design of the resulting structure. Additionally, the requirements of specific scanning probe microscopies heavily influence the choice of substrate, i.e. STM requires conducting surfaces for imaging, and this is in many instances the determining feature. Recent improvements in resolution of atomic force microscopy (AFM)²¹  and the development of dynamic force microscopy (DFM)²²  have led to a wider scope for characterisation and therefore the choice of substrate is no longer restricted to (semi)conducting materials and insulating substrates are now realistic targets. Lastly, notably in the case of covalently coupled structures, the choice of substrate is extremely important with the substrate often taking an active role in promoting coupling reactions.

The chapter is subdivided into four sections, systems assembled using (i) hydrogen bonds, (ii) van der Waals interactions, (iii) covalent bonds and lastly a section (iv) of self-assembled systems with unusual ordering that illustrate the power of the approach and particularly the molecular-level characterisation of such systems.

The use of hydrogen bonds to create supramolecular structures goes back to the origins of the field.²³  Indeed it is relatively straightforward to use a simple molecule capable of self-recognition to create extended framework structures. Indeed, an early example of surface-based supramolecular chemistry—the formation of a unimolecular supramolecular structure mediated by hydrogen-bonding—was reported by Griessl et al.²⁴  The structure was formed via deposition of trimesic acid onto a HOPG substrate in UHV conditions and imaging using STM (Figure 1.1a,b) reveals precise details of the self-assembled molecular structure. The carboxylic acid–carboxylic acid hydrogen bonds, which adopt the classic R²₂(8) intermolecular arrangement,²⁵  ensure that an open structure is adopted in preference to a hypothetical, alternative close-packed arrangement, confirming that hydrogen-bonding interactions are the dominant force in producing these structures. The observed structure is the expected ‘chicken-wire’, or honeycomb arrangement leading to a porous network structure (Figure 1.1a,c). In addition to the honeycomb structure a secondary so-called ‘flower’ arrangement is also observed by STM measurements (Figure 1.1d). This alternative self-assembled structure results from the adoption of R³₃(12) supramolecular synthons formed by three carboxylic acid moieties from separate trimesic acid molecules. The ‘flower’ structure may form due to a higher molecular density on the surface, confirming that surface coverage also influences the final self-assembled structure by maximising the energy gained through adsorbate-substrate interactions.

A subsequent study of an elongated analogue of trimesic acid, 1,3,5-tris(4-carboxyphenyl)benzene, deposited on HOPG, but deposited from a range of alkyloic acids rather than by sublimation in UHV, also results in the formation of a honeycomb lattice at low temperatures, but a more densely packed phase at higher temperatures.²⁶  The phase transition and the transition temperature between honeycomb and denser structures were found to depend on the nature of the solvent and molecular concentration. The authors suggested that the co-adsorption of solvent molecules within the honeycomb structure stabilises the nominally porous structure at low temperatures, but upon elevation of the temperature the weakly bound solvent molecules desorb initiating the transition to the more densely packed and thermodynamically-favoured phase.

A simple example of a unimolecular self-assembled structure that uses hydrogen bonds is that of naphthalene-1,4,5,8-tetracarboxylic diimide (NTCDI).²⁷  NTCDI contains imide moieties at opposing ends of the rod-shaped molecule which can adopt imide–imide hydrogen bonds, again adopting an R²₂(8) intermolecular interaction, similar to that observed for carboxylic acid dimers. As a result of the divergent arrangement of the imide moieties linear chains are observed when the molecule is deposited onto a Ag/Si(111) surface (Figure 1.2a). Imaging using DFM of this molecule and the self-assembled array on a variety of surfaces^28–30  reveals sub-molecular details of the molecular arrangement. It is interesting to note that features in the DFM images appear to coincide with where hydrogen bonds would be expected to be observed (Figure 1.2b), however calculations indicate that these features do not arise as a result of the hydrogen bonds but due to some other, as yet undefined, phenomenon; the origin of such intermolecular features remains a highly debated topic.³¹ 

Moving beyond simple unimolecular systems it is possible to use the full force of supramolecular design concepts to bear upon the construction of surface-based arrays. Learning from nature, a particularly attractive approach is to use the specific hydrogen-bonding capability of DNA nucleobases. The use of DNA nucleobases in supramolecular chemistry has been a persistent theme since some of the seminal work of Seeman,³²  which has been expanded by a number of groups to great effect.³³  Indeed, Seeman and Winfree demonstrated the specific preparation of 2D crystals on surfaces using his ‘sticky-end’ approach.³⁴  Examples of such surface-based arrays include the self-assembly of striped lattices, amongst other structures, successfully produced on mica surfaces³⁴  and hexagonal arrays constructed from three-point star motifs, in turn built from seven DNA single strands (Figure 1.3).³⁵ 

In addition to the studies of self-assembly on surfaces of large DNA fragments, a particular focus for hydrogen-bonding studies has been the use of the individual DNA bases to create self-assembled structures on surfaces. The assembly of guanine and its derivatives on surfaces has received particular attention,^36–39  perhaps as a reflection of the importance of guanine quartets in anti-cancer drug design.⁴⁰  For example the adsorption of guanine onto a Au(111) surface under UHV conditions leads to the formation of guanine quartets as imaged by STM.^36,39  Interestingly in this example the quartets form through Hoogsteen-style hydrogen-bonding and are associated through further N–H–N hydrogen bonds to give rise to two-dimensional supramolecular structures. This example illustrates the complexity of using DNA bases due to the variety of potential hydrogen-bonding motifs that can be adopted, from classic Watson–Crick pair formation, Hoogsteen interactions, and reverse Watson–Crick and reverse Hoogsteen arrangements (Figure 1.4). The area has been recently reviewed⁴¹  and thus only very recent examples are highlighted here.

It is of course possible to exploit the hydrogen-bonding capability of the DNA nucleobases as appendages to more complex molecules leading to enhanced control over supramolecular structure. An example of such a strategy is given by the study of a porphyrin molecule (tetra-TP) in which the porphyrin core is functionalised in each meso-position by a phenylthymine moiety such that each thymine presents a hydrogen-bonding face exo to the porphyrin core.⁴²  The molecule self-assembles on a HOPG substrate to give rise to a two-dimensional grid structure wherein the molecules interact through R²₂(8) intermolecular thymine–thymine hydrogen bonds (Figure 1.5a). The asymmetric arrangement of the thymine groups appended to the tetra-TP molecules suggests the potential for the molecules to adopt a chiral arrangement when adsorbed onto a surface and this is observed. The almost perfectly square 2D unit cell observed for the tetra-TP structure suggests that all of the thymine groups within an individual tetra-TP molecule adopt the same orientation with respect to the porphyrin, and that individual domains contain only molecules of the same handedness. Overall the array of tetra-TP remains globally achiral by forming an equal area of mirror domains containing either right, or left-handed molecules. It is worth noting that many previous studies have shown that prochiral molecules tend to assemble into homochiral domains on surfaces containing molecules of only a single handedness.⁴³  Addition of a substituted adenine component, in this case 9-propyladenine, to the assembly of tetra-TP produced a molecular network with a combination of disordered regions interspersed with small domains of an ordered co-crystal structure containing both tetra-TP and 9-propyladenine. As anticipated Watson–Crick thymine–adenine interactions are observed but interestingly dimers of adenine are also formed, presumably in order to achieve maximum surface coverage, and forming an additional N–H–N hydrogen bond to a further tetra-TP molecule (Figure 1.5b). The complexity of the structure indicates the complexity in designing such structures, even when relatively predictable hydrogen-bonding synthons are present.

A related approach has been described by González-Rodríguez and De Feyter⁴⁴  who have designed rod-shaped molecules appended by DNA bases, either with guanine (G) and cytosine (C) or adenine (A) and uracil (U) (Figure 1.6a). The rods are designed such that opposing ends will form complementary hydrogen-bonding groups. Additionally, the authors have functionalised the rods with alkyl tails such that any residual space in the structure is occupied and no additional molecules are required to fill space. Self-assembly on an HOPG substrate results in the formation of cyclic structures, as imaged by STM (Figure 1.6b), and indicates the successful employment of the complementary hydrogen-bonding moieties in forming heteromolecular hydrogen-bonding interactions (G–C or A–U). Despite the alkyl chain appendages filling the space within the cyclic structures the G–C system is capable of adsorbing coronene with the alkyl chains assumed to vacate the space due to preferential physisorption of the polyaromatic guest onto the HOPG surface.

Moving beyond DNA systems it is possible to create related, non-natural, hydrogen bonding synthons to create extended structures. For example the triple hydrogen bonding interaction between 2,6-di(acetylamino)pyridines and imide groups has received extensive study in supramolecular chemistry^23,45  and can also be exploited to form bimolecular structures on surfaces.⁴⁶  An early example of a bimolecular network was prepared from perylene-3,4,9,10-tetracarboxylicdiimide (PTCDI), which bears two imide appendages at opposing ends of the rod, in a similar fashion to NTCDI, and melamine (Figure 1.7).⁴⁷  Co-deposition of the two building blocks onto a Ag/Si(111) surface under UHV conditions results in the formation of a honeycomb network with triple hydrogen bonds formed between the PTCDI imide moieties and each side of the melamine triangle. Network formation was achieved following annealing at ∼100 °C which provides sufficient thermal energy for molecules to detach from PTCDI islands and diffuse across the surface, resulting in the desired network topology. The network was found to be commensurate with the underlying Ag/Si(111) surface, suggesting an influence from the substrate in the formation of the network.

The large pores formed by the network, with a cross section of 2.4 nm, are capable of housing several large molecules and this was initially demonstrated by the introduction of C₆₀⁴⁷  into the framework pores (Figure 1.7). The deposition of C₆₀ by sublimation onto the hexagonal network leads to the formation of heptameric C₆₀ clusters within the pores of the network, with a compact hexagonal arrangement of fullerenes aligned parallel to the principle axes of the Ag/Si(111) surface. The arrangement and properties of the C₆₀ clusters were found to differ from close-packed fullerene on the same surface, which does not align with the principle axes of the underlying Ag/Si(111) substrate or form heptamers, thus demonstrating the templating and stabilising effects of the network.⁴⁷  Indeed the ability of the PTCDI-melamine family of arrays has been shown to host other fullerenes, including C₈₄⁴⁸  and Lu@C₈₂,⁴⁹  and a range of molecules, as discussed below.

The PTCDI-melamine network⁴⁷  can similarly be prepared on a Au(111) surface⁵⁰  leading to an analogous honeycomb arrangement to that observed on a Ag/Si(111) substrate. However, after annealing at higher temperatures, a parallelogram phase was also observed on Au(111), which has the same stoichiometric ratio as the honeycomb structure, but is more dense.⁵¹  The parallelogram phase is also able to trap guest molecules such as C₆₀, but due to the restricted cavity size C₆₀ dimers are observed, the size of the cluster being defined by the physical size of the network pores. Similar studies have demonstrated entrapment of Lu@C₈₂ by the parallelogram phase but in this case only single fullerenes are accommodated within each pore.⁴⁹  The parallelogram PTCDI-melamine phase also traps two decanethiol molecules which sit parallel to the surface under the UHV conditions used. The entrapment of a significantly different and potentially more reactive molecule confirmed that such networks have the potential for trapping a range of different species. It is interesting to note that adsorption of decanethiol onto the PTCDI-melamine arrays lead to the destruction of the honeycomb phase, but not the parallelogram phase, suggesting that the latter is preferentially stabilised by the thiol guest.⁵² 

Indeed the adsorption of thiol guests within the pores of PTCDI-melamine networks has been advanced significantly by Buck et al.^53,54  It has been demonstrated that the PTCDI-melamine network can be assembled onto a Au(111) surface from solution and used for a variety of further applications and particularly as a versatile patterning tool (Figure 1.8).^53,54  Due to the solution-based preparative procedure it is possible to apply traditional Au-thiolate SAM-synthesis in the presence of the array. The adsorption of adamantanethiol within the pores of the network is demonstrated with the thiolates sitting perpendicular to the surface, in contrast to related studies in UHV conditions.⁵²  The clusters of adsorbed thiols, or confined SAMs, can be exploited to guide the deposition of Cu atoms within the pores using underpotential deposition (UPD). It was demonstrated that Cu was successfully deposited only on areas of the surface that were not covered by the network, i.e. those areas covered by thiols,⁵³  and a subsequent study demonstrated that the PTCDI-melamine network acts as a diffusion barrier for Cu adatoms.⁵⁵ 

Thus, molecules can be adsorbed onto hydrogen-bonded arrays by sublimation or solution-phase deposition but these techniques may not always be suitable for sensitive molecules and thus a further technique has been developed using electrospray deposition.^56,57  Electrospray techniques have been used to deposit Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ clusters, a molecule of interest due to its ability to act as a single molecule magnet, onto the PTCDI-melamine array on a Au(111) substrate.⁵⁷  The softer deposition technique is required as sublimation is not possible for Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ due to its decomposition at elevated temperatures. In the absence of the hydrogen-bonded array the Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ clusters assemble generating filamentary structures on the surface (Figure 1.9). In contrast adsorption of Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ clusters onto the PTCDI-melamine array results in a low degree of ordering with some molecules accommodated within the pores of the network but with others positioned on top of the array (Figure 1.9). The entrapment of the Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ clusters in the array is not as efficient as that of other species, including fullerenes, most probably due to a mismatch of dimensions between the cluster (1.6 nm) and pore (2.5 nm). This study indicates that maximum organisation of guest molecules by porous frameworks can be achieved by idealising size match between host and guest and by maximising interactions between host and guest.

An advantage of the bimolecular approach to array formation is that each component can be altered allowing modification of the resulting structure. Thus, for the PTCDI-melamine arrays the PTCDI molecule^58,59  can be readily functionalised leading to manipulation of the framework and cavity dimensions within the array. Introducing steric bulk to the PTCDI molecules can control the available space for guest entrapment leading to networks that can trap individual molecules potentially in a regular array. Self-assembled arrays formed between melamine and PTCDI derivatives, and in some examples networks formed by the functionalised PTCDI derivatives alone, act as hosts to guest molecules. Whereas the honeycomb array formed by 1,7-dibromo-PTCDI and melamine on Ag/Si(111) accommodates C₆₀ heptamers, similar to the parent unfunctionalized PTCDI-melamine array, a thiopropyl-functionalised PTCDI—(SPr)₂-PTCDI—network formed in combination with melamine does not trap heptamers or hexamers.⁵⁸  Although some fullerenes can be observed in STM images following deposition, the C₆₀ molecules are positioned in an irregular fashion both within pores and also on top of the honeycomb array, similar to Mn₁₂O₁₂(O₂CCH₃)₁₆(H₂O)₄ clusters (see above) indicating that the thiopropyl groups act to inhibit fullerene adsorption by this network.

Self-assembly of thioadmantyl functionalised PTCDI, (SAdam)₂-PTCDI, and melamine leads to the anticipated honeycomb array.⁵⁹  It should be noted that the thioadmantyl groups are both considerably bulkier and more rigid than the thiopropyl appendages of (SPr)₂-PTCDI and therefore would be expected to provide an effective mechanism for inhibiting the space within the network pores. However, upon formation of a (SAdam)₂-PTCDI-melamine network some of the thioadamantyl groups are cleaved from the PTCDI moieties.⁵⁹  Consequently a variety of distinct pore sizes and configurations are formed which can be visualised following C₆₀ adsorption onto the network and subsequent STM imaging. Individual C₆₀ molecules are more readily imaged by STM than the underlying thioadamantyl groups. Using this approach, it is possible to identify the different orientations of the molecular clusters within each pore (Figure 1.10). The number of C₆₀ molecules adsorbed within a given pore is determined by the degree of thioadamantyl cleavage and clusters ranging from dimers to heptamers are observed. Considering the dimensions of the honeycomb pores and the size of the thioadamantyl appendages, clusters of greater than five fullerenes can only be explained by cleavage of the thioadamantyl groups, the heptamers corresponding to a pore where no thioadamantyl appendages remain.

Further constriction of pore size can be used to control the number of guest molecules, a principal demonstrated by the array formed by 1,7-bis(4-benzoic acid)-3,4,9,10-perylenetetracarboxylic diimide (Bz₂-PTCDI).⁶⁰  As with many disubstituted PTCDI species,⁴⁶  Bz₂-PTCDI assembles into a unimolecular honeycomb array, without the need to add melamine, through the adoption of a tri-molecular hydrogen-bonded junction (see below). The rigid protrusion of the phenylcarboxylate moieties into the framework cavities leads to the encapsulation of single C₆₀ molecules within each pore, which are in turn spaced in a regular fashion across the surface (Figure 1.11).

The material employed as substrate for the self-assembly process has a clear influence on the resulting framework structure and thus cannot be considered an innocent agent in the synthetic strategy. An interesting example that demonstrates the influence of the substrate on supramolecular assembly is the use of graphene or boron nitride (BN) ‘nanomesh’ monolayers grown on Rh(111) crystals⁶¹  that exhibit moiré patterns introducing distinct adsorption sites on the surface. The self-assembly of PTCDI and related di-functionalised derivatives, 1,7-dipropylthio-perylene-3,4,9,10-tetracarboxydiimide (DP-PTCDI) and 1,7-di(butyl)-coronene-3,4,9,10-tetracarboxylic acid bisimide (DB-CTCDI), have been studied on graphene and the structure of the resulting structures determined using STM (Figure 1.12).⁶¹  Two distinct junctions, formed by either two or three molecules are observed leading to different assemblies (Figure 1.12). For example, unfunctionalised PTCDI forms rows up to 25 nm in length running parallel to the principal directions of the substrate and adopting only simple dimeric hydrogen-bonding arrangements. The molecular arrangement on the graphene superstructure differs significantly from that observed for a graphite substrate, on which three-dimensional islands are formed.⁶²  The importance of a commensurate match between molecular dimensions and the moiré periodicity is demonstrated by a comparison with adsorption of PTCDI on a BN “nanomesh” monolayer (Figure 1.12). The BN monolayer is isoelectronic with graphene, and also displays a moiré pattern on a Rh(111) crystal, but with a slightly larger periodicity of 3.2 nm, which is non-commensurate with the dimensions of PTCDI. On BN [on Rh(111)] individual isolated PTCDI molecules trapped in energy minima associated with the moiré pattern are observed demonstrating the influence of the underlying substrate on supramolecular assembly. Both DP-PTCDI and DB-CTCDI form both linear rows, built using dimeric hydrogen bond junctions, and more complex arrays which include junctions where three molecules meet in a hydrogen-bonded trimeric arrangement. In the case of DP-PTCDI the ratio of dimer : trimer junctions is 75 : 25 in comparison to less than 1% of junctions being trimers for PTCDI. In the case of DB-CTCDI the three molecule junctions dominate the self-assembly process, with no linear dimers unambiguously identified. The array constructed from trimeric junctions results in a honeycomb framework aligned with the graphene monolayer superstructure and encloses the areas of bright contrast arising from the moiré pattern.

The hydrogen-bonding synthon exploited in the PTCDI-melamine arrays can be readily incorporated into other building blocks (e.g. thymine includes the same imide moiety seen in PTCDI) and by so doing many possible arrays can be targeted. Perhaps the simplest example is the self-assembled array formed between cyanuric acid and melamine which has been extensively studied.^63,64  Samorì et al.⁶⁵  have elegantly demonstrated the self-assembly of a range of diimide molecules and melamine (Figure 1.13) at the solution/solid interface on a HOPG surface following deposition from a 1,2,4-trichlorobenzene–dimethylsulfoxide solution. The formation of hexagonal, or honeycomb networks is anticipated for such systems in an analogous fashion to the PTCDI-melamine systems, and such arrays are observed, but detailed studies of concentration variation leads to the adoption of a range of structures. The nature of the structures formed by combinations of the diimide molecules (labelled 1–4 in Figure 1.13) with melamine ranges from open hexagonal networks to close-packed structures and significantly, varies with component concentration. Ultimately a phase diagram of polymorphs is generated for these systems which gives perspective to the complexity of the self-assembly processes involved (Figure 1.13).

The role of van der Waals interactions in the creation of self-assembled surface frameworks cannot be understated. Naturally such interactions are important in all such systems and play a highly significant role in the interaction between molecules and substrates. A large number of studies have reported using van der Waals interactions to create supramolecular structures and these have been reviewed on a number of occasions.^7–9,66,67  A particularly notable interaction that has been widely exploited is the strong adsorption between alkyl chains and HOPG. The role of the alkyl chain is clearly significant in the resulting assembly process, and interdigitation of alkyl chains from adjacent molecules is commonly observed.

A representative example of a self-assembled framework controlled by van der Waals interactions was investigated by Schull et al.⁶⁸  1,3,5-tris[(E)-2-(3,5-didecyloxyphenyl)-ethenyl]-benzene (TSB35), a trimeric compound with six pendent decyl chains was deposited onto an HOPG substrate. Self-assembly through interdigitation of the alkyl chains leads to formation of a porous network within which coronene and hexabenzocoronene could be accommodated (Figure 1.14). It can be clearly seen that the co-adsorption of coronene or hexabenzocoronene modifies the nature of the resulting structure, in a similar fashion to solution-phase host–guest chemistry. Indeed the interdigitation of alkyl chains has been extensively used by de Feyter and co-workers in a number of elegant studies.^{7–9,66,67,69–74}  One avenue that has been a particular focus of research has been the study of alkoxylated dehydrobenzo[12]annulenes.⁶⁹  A representative study of rhombic-shaped dehydrobenzo[12]annulenes assembled at a 1,2,4-trichlorobenzene/HOPG interface⁶⁹  demonstrates that variation in the chain length of alkyl substituents results in five distinct network structures, three porous and two nonporous structures.⁶⁹  Shorter alkyl chain substituents on the dehydrobenzo[12]annulene core favours formation of porous frameworks, whereas those with longer alkyl chains tend towards nonporous arrangements. Concentration of the building blocks in solution also plays a key role in the nature of the structures formed, thus, dilution of the solutions of dehydrobenzo[12]annulene leads to the transformation of nonporous structures into porous networks. Favouring of porous structures at lower concentrations is a result of factors related to overall surface coverage and, in solution phase studies, it is also possible that porous structures can be stabilized by the co-adsorption of solvent molecules.

In order to establish control over the self-assembly process researchers have designed building blocks that combine more than one type of intermolecular interaction, for example hydrogen-bonding and van der Waals interactions. This approach has been widely used and examples include N⁹-alkylated guanine derivatives,^37,75  and 5-alkoxyisophthalic acid derivatives on HOPG.^7,8,72  For the latter example, 5-alkoxyisophthalic acids, monolayers of close-packed arrays of interdigitating hydrogen-bonding ribbons are formed (Figure 1.15). Thus, the isophthalic acid groups form hydrogen-bonded ribbons whilst the alkyl groups interdigitate, acting as spacers between adjacent ribbons. If larger groups are appended to the alkyl chain, such as diphenylmethyl moieties, then the hydrogen-bonding observed in the ribbons is disrupted so that hexameric rings are formed (Figure 1.15).

A remarkable recent study by de Feyter⁷⁶  describes an approach to establish the range of polymorphs for a given system using building blocks that exploit van der Waals interactions or hydrogen-bonding interactions. The formation of polymorphs in two-dimensional supramolecular arrays is one of the biggest challenges in the research field as polymorphic arrangements always provide a potential barrier to the successful preparation of well-defined, ordered arrays. De Feyter et al. approached the question of polymorph evaluation by designing a mechanism to induce a solution flow across a substrate surface generating a lateral density gradient. This in situ generation of a gradient allows both discovery and separation of multiple polymorphs in a single experiment. The authors describe three separate systems: hexadecyloxy substituted dehydrobenzo[12]-annulene (DBA-OC16), hexadecyl substituted bis(dehydrobenzo[12] annulene) (bisDBA-C16) and 1,3,5-tris(4-carboxyphenyl)benzene (BTB) allowing demonstration of their approach with systems that exploit either van der Waals or hydrogen-bonding interactions. In each instance, more than one polymorph was found for each system by STM imaging of the respective samples: DBA-OC16 – 2 polymorphs; bisDBA-C16 – 4 polymorphs; BTB – 2 polymorphs (Figure 1.16). The approach allows not only identification and structural characterisation of the polymorphs observed but also quantification of surface coverage of each arrangement.

Both hydrogen-bonding and van der Waals interactions are excellent for preparing well-ordered structures due to the ability of these relatively weak interactions to be reversibly formed, allowing structural correction. However, to form robust two-dimensional structures an approach that slows the formation of covalently coupled arrays is the most desirable target. A number of approaches have been used to target such systems, including aryl–aryl coupling^77–80  and alkyne–alkyne coupling⁸¹  processes. Other approaches have been taken in recent studies⁸²  and the field has clear parallels to studies of covalent organic frameworks or COFs.¹⁵ 

One of the earliest approaches taken to the development of covalently coupled arrays was the use of the coupling between halo-aryl species on Au surfaces. Grill et al.⁷⁷  elegantly demonstrated the use of porphyrin building blocks with bromophenyl ‘arms’ protruding from the meso-positions of the porphyrin.⁷⁷  The number of bromophenyl arms can be varied such that a range of building blocks with one, two or three reactive appendages can be utilised. The reaction of mono-bromophenyl substituted porphyrin, by heating the building blocks on a Au(111) substrate, leads to dimer formation (Figure 1.17a) and trans-di-bromophenyl substituted porphyrin reacts to give chains (Figure 1.17b). Lastly tetra-bromophenyl substituted porphyrin can react in all four positions to give well-ordered two-dimensional grids of porphyrins (Figure 1.17c).

The approach was developed further by Grill, Hecht and co-workers through the construction of covalently coupled arrays built from two different molecular components. The strategy exploits the different reactivity of bromophenyl and iodo-phenyl groups attached to a porphyrin core in a trans arrangement leading to different reactivity between the different faces of the porphyrin molecules.^78,79  Deposition of the dibromo-diiodo-porphyrin (Br₂I₂TPP) building-block onto a Au(111) substrate followed by heating the sample leads to the covalent coupling between adjacent porphyrin species. The lower bond dissociation energy of the C–I bonds, in comparison to C–Br bonds, allows preferential coupling of the iodo-functionalised arms at a lower temperature, leading to the formation of one-dimensional chains of coupled porphyrin building blocks (Figure 1.18). Below 200 °C only the iodo-functionalised arms react leaving the bromo substituents intact, allowing the possibility of further reaction. Above 200 °C the bromo-functionalised arms can react leading to coupling of the one-dimensional chains forming two-dimensional porphyrin islands, mirroring the previously reported observations (see above).⁷⁷  However, following formation of the one-dimensional chains of coupled porphyrin building blocks, which are functionalised with bromophenyl appendages, reaction with a secondary species can be successfully achieved. The authors demonstrated the success of their approach by co-deposition of dibromoterfluorene which can cross-couple with the bromo-functionalised one-dimensional porphyrin array. Thus, coupling of the bromo-appendages on the porphyrin chain and the terfluorene molecules leads to decoration of the porphyrin chains with terfluorene groups and in some instances cross-linking of adjacent porphyrin chains is observed, leading to a complex two-dimensional array of coupled porphyrin chains and terfluorene moieties (Figure 1.18). The simple strategy of hierarchical coupling, exploiting dissimilar chemical reactivity and thermal activation barriers for the reactive components, leads to the formation of a bimolecular array and opens up the possibility for the creation of more complex covalently coupled molecular arrays.

It has also been demonstrated that it is possible to transfer such porphyrin-based covalently coupled arrays from a gold thin film deposited on mica to other surfaces using a C₆₀ adhesion layer.⁸³  It was demonstrated that a thin film of C₆₀, with a thickness ranging from 10 to 100 nm, can promote adhesion between a thin film of Au and a solution-deposited layer of polymethyldisolaxane (PDMS). The adhesion facilitated by the C₆₀ thin film allows removal of the gold film from the mica support by simple peeling. Following removal from the mica support the monolayers may be isolated and transferred to other substrates, e.g. SiO₂, by etching of the thin film of gold, mechanical transfer and removal of the fullerene layer by annealing and then dissolution.

The strength of the strategy of using covalent coupling to form surface-based arrays is demonstrated by a recent study by Amabilino and Raval.⁸⁴  The study reports the coupling of polyaromatic molecules, (pentacene, tetramesitylporphyrin, perylene) and porphyrins (H₂-porphyrin and Zn(ii)diphenylporphyrin), through C–H bond activation reactions on a copper surface. It is demonstrated that all of the molecules are capable of undergoing C–H activation and can be coupled on the surface chosen. The study not only demonstrated the ability to homo-couple the molecules into extended arrays, e.g. Zn(ii)diphenylporphyrin can be cross-coupled through the unsubstituted edges (i.e. not supporting phenyl groups) to afford covalently coupled one-dimensional chains of porphyrins, but also hetero-coupling between the different building blocks. A large variety of coupled structures are described (Figure 1.19) with up to three different building blocks cross-coupled. STM studies provide sufficient insight to allow characterisation of each individual component of the covalently coupled structure. In summary, this study demonstrates the ability to prepare highly complex covalently coupled structures from multiple building blocks and indicates the potential for creating bespoke covalent structures on surfaces.

One of the most exciting features of creating surface-based supramolecular frameworks is the ability to image the resulting structures at the molecular and sub-molecular^22,28  level. This allows a level of detailed understanding that is extremely challenging, if not impossible, in solid-state materials due to the nature of the characterisation that is employed – scanning probe microscopies for surface-based structures vs. diffraction techniques in the solid-state. The additional level of detail that can be achieved using STM, or AFM, has allowed characterisation of structures that would prove extremely challenging by any other technique.

One of the first examples of such a structure is that formed by terphenyl-3,3″,5,5″-tetracarboxylic acid.⁸⁵  Terphenyl-3,3″,5,5″-tetracarboxylic acid, adsorbed onto HOPG, leads to the formation of a two-dimensional hydrogen-bonded structure that utilizes R²₂(8) intermolecular carboxylic acid–carboxylic acid interactions. STM imaging allows identification of the position of each molecule in the extended structure and direct visualisation of a non-ordered structure. Indeed, the relative position of molecules within the array is random and reminiscent of dynamically-arrested systems, such as glasses (Figure 1.20). The intermolecular hydrogen-bonding leads to the formation of hexagonal junctions which are formed from three, four, five, or six molecules as a result of the dimensions of the molecule. As such the structure forms an extremely rare example of a random, entropically stabilised, rhombus tiling. It is important to note that the detailed structure of this framework, and other related frameworks, can only be appreciated with a molecular level understanding and that this can only be readily achieved by scanning probe microscopy. The degree of randomness of the rhombus tiling can be evaluated by detailed analysis of each image and subsequent studies demonstrated that both solvent used for deposition and temperature of experiment can affect how random the tiling structure is.⁸⁶  It is important to note that the rhombus tiling arrangement is only observed for molecules which have the appropriate dimensions, i.e. those which are rhombus shaped, and analogous molecules such as quaterphenyl-3,3‴,5,5‴-tetracarboxylic acid form regular two-dimensional structures.⁸⁷ 

The random rhombus tiling structure formed by terphenyl-3,3″,5,5″-tetracarboxylic acid on HOPG is able to trap guest molecules.⁸⁸  C₆₀ can be deposited onto the random array with preferential adsorption into one of the five possible pores, each pore is defined by different numbers and arrangements of molecules that define the periphery of the pore. The images reveal a strong preference for adsorption in pores of type A (Figure 1.21), which make up approximately 40% of pores in the initial framework but ca. 76% of pores that trap a C₆₀ molecule. Calculations confirm that adsorption in pore type A is favoured because of the higher proportion of phenyl edges in this configuration. Intriguingly images show that in regions of the network where C₆₀ molecules are adsorbed a second framework layer is assembled sitting over the initial layer, thus forming a bilayer structure (Figure 1.21). This was the first time that growth of a bilayer has been observed in such assemblies. The spherical C₆₀ molecule acts as a template for the formation of the second layer, acting as an anchor that allows the hydrogen-bonded structure to assemble. The importance of the trapped C₆₀ molecules to the formation of the second layer is demonstrated by the addition of coronene to bilayer samples which results in displacement of the entrapped C₆₀ by the flat, and more strongly adsorbed, coronene molecule and concomitant removal of the second supramolecular layer.

The rhombus tiling observed with terphenyl-3,3″,5,5″-tetracarboxylic acid is related to a Penrose tiling.⁸⁹  The first observation of a structure more directly related to such a tiling was reported for the self-assembled structure formed by ferrocenecarboxylic acid [Fc(COOH)] on a Au(111) substrate.⁹⁰  Penrose tilings are related to quasicrystal structures in that they exhibit long range, non-periodic order, and unusual rotational symmetry. Fc(COOH) assembles through intermolecular hydrogen-bonding to form pentagonal arrangements of molecules (Figure 1.22) with carboxylic acid – carboxylic acid O–H⋯O interactions being additionally stabilised by C–H…O hydrogen bonds between adjacent molecules. Importantly the cyclic pentamer observed for the surface-based array was found to be more stable than other potential hydrogen-bonding arrangements, such as dimeric systems, by DFT calculations. It was found that a related compound—ferroceneacetic acid [Fc(CH₂COOH)]—does not form the pentagonal arrangements, adopting a more conventional dimeric arrangement. This can be accounted for by the observation that the length of the additional methylene group, in comparison to Fc(COOH), prevents interactions between the carboxyl oxygen of one carboxylic acid and the hydrogens on an adjacent cyclopentadienyl fragment. The pentagonal arrangements observed are related to subunits of the Penrose P1 tiling.⁸⁹ 

The last example of a highly unusual surface-based structure was reported for the angular molecules 4,4″-dibromo-1,1′:3′,1″-terphenyl (B3PB) and 4,4‴-dibromo-1,1′:3′,1″:4″,1‴-quaterphenyl (B4PB).⁹¹  Both B3PB and B4PB are functionalised on their termini with bromo-aryl moieties which are instrumental in forming intermolecular halogen bonding interactions (Figure 1.23). These interactions result in the formation of extended structures which share the same topological arrangement as Serpinski triangles, which are fractal structures. The formation of the fractal arrangements is truly remarkable and it is clear that the energy balance between such unusual structures and other potential arrangements is very finely balanced. As with the rhombus and Penrose tilings it is important to note that a complete appreciation of the structures formed is only possible due to a molecular level understanding of the self-assembled arrangement. This can only be readily achieved using scanning probe microscopies and it is perhaps for this reason that such structures are found in surface supramolecular framework structures.

It can be clearly seen from the systems described above that surface-based supramolecular chemistry has a rich future. The ability to create well-ordered two-dimensional frameworks shows great promise for the development of bespoke materials. Thus far hydrogen-bonded frameworks have received a great deal of attention, as have those structures that rely on van der Waals interactions to control framework formation, but increasingly the principles of two-dimensional framework formation are being applied to covalently bonded structures. Traditional host–guest chemistry is also feasible using surface-based frameworks, opening up avenues of research for the preparation of nanoscale devices.

Although many of the approaches that are employed by supramolecular chemists in either the solution phase or solid state are generally applicable to surface-based processes, significant differences are also evident. Notably the surface does not play a passive role in the two-dimensional self-assembly process; for molecules to adsorb on the surface there is inherently an interaction between substrate and the molecule and this can lead to subtle differences, even affecting the conformations of molecules⁹²  and importantly the way in which they self-assemble. As the area of surface-based self-assembled frameworks continues to develop it is important that the role of the substrate becomes increasingly understood and ultimately exploited to control self-assembly. In no area is this more important than in the formation of covalently bonded surface frameworks for which the substrate often plays an integral role in the formation of the covalent bond.

This chapter demonstrates that a number of successful strategies have already been developed for the synthesis of surface-based framework structures and some of these approaches have led to the discovery of highly unusual structures based upon unusual tiling processes, frameworks which are unlikely to be discovered using other strategies or environments. Although it is clear that further reliable pathways to robust frameworks still need to be developed it is also clear that significant progress has already been made in the field. Now that significant success in synthetic strategies has been achieved the first few steps towards functional materials are already underway.