The knowledge about the internal structure of the columns (fibres) in bulk columnar mesophases depends mostly on X-ray techniques, which suffice for many purposes [7,10–11]. However, information in real space, as provided by scanning tunnelling microscopy (STM), offers unparalleled advantages to the synthesis chemist who strives to functionalize fibrils that are one-dimensional structures with only a few nanometres in diameter. The control of supramolecular self-assembly to achieve functional nanostructures depends on careful design at the molecular level, and elucidation of their internal structure is important in aiding the design and to increase the sophistication of the building units.

Many low-molecular-weight, wedge-shaped amphiphilic molecules are known to form columnar mesophases [3,6]. X-ray diffraction and scattering techniques have widely been used to decipher their internal molecular arrangements, which generally suggest a stacking of mesogenic “discs” leading to column formation [7,12]. We investigate the self-assembled fibril structures of two custom-designed amphiphilic gelator molecules: N,N′-bis[3,4-bis(decyloxy)benzoyl]hydrazine (2CHd-10) and [4-(decyloxy)benzoyl]hydrazine (1CHn-10) on the graphite (0001) surface (Fig. 1 and Experimental section). As the alkyl chain length is known to influence column formation in the bulk, the length of alkyl chains for both molecules is kept identical such that the focus of the study is solely on the geometry/symmetry aspect [6]. In Fig. 1, a wedge shape is shown superimposed on the molecular structure, where the amide moieties are at the tip of the wedge and the alkoxy chains at the tail. In general, the molecular geometry of the mesogens is decisive for the generation of columnar mesophases in the bulk, i.e., the mesogens should be wedge-shaped. The wedges can form a disc with their tips all directed to the centre; for example, six such wedge pieces may lead to a hexagonal columnar mesophase. Further stacking of discs in a face-to-face configuration leads to columns [6,12–13].

To evaluate the possibilities of self-assembly of 2CHd-10 and 1CHn-10 molecules on HOPG, we note that their amide functionalities can efficiently stabilize structures through intermolecular hydrogen bonds. On the other hand, the alkyl chains are expected to promote self-assembly into extended structures through interchain van der Waals interactions as well as adsorption on HOPG due to their epitaxial match with the C–C bonds of graphite [16]. The structures produced, may, however, generally depend on a complex interplay of many weak interactions. The molecules are prototypes for symmetric and asymmetric hydrazine species, where 2CHd-10 represents two 2CHn-10 molecules linked together such that the amide functionality is not a head group but a central part, which is expected and found to have considerable influence on the self-assembly behaviour. We observe that the molecules self-assemble into one-dimensional structures at the solution/HOPG interface, which are distinctly different from those of the “disc-stacking” pattern in the bulk. The structures of elementary fibrils are explained by dimer or tetramer precursors followed by fibril formation through van der Waals interactions between alkyl chains. The molecular geometry plays a crucial role in deciding the type of oligomer precursor: self-assembly is based on dimer building blocks for 2CHd-10, but tetramer building blocks for 1CHn-10. It appears that the large-scale morphologies result as a direct consequence of the type of oligomer precursors they form.

As a substrate, highly oriented pyrolytic graphite (HOPG) is favoured in STM studies due to its high electrical conductivity, atomic flatness, chemical inertness and here also for its hydrophobic nature. Hydrophilic substrates could hinder the self-assembling ability of the molecules by strongly interacting with their amide functionalities and forcing them to lay flat on the surface. To study adsorbate architecture on an electrically conducting substrate in real space, scanning tunnelling microscopy (STM) is the experimental technique of choice. Although STM has been highly successful in atomic/submolecular probing of planar structures [17–18] and their dynamics [19], when it comes to one-dimensional structures, it has not shown the same level of efficacy [20]. Particularly for high-resolution imaging in an ambient/solution environment, data acquisition becomes an exhausting and time-consuming process. Difficulties arise from the requirement of having single-digit nanometre-wide isolated strands and locating them on a millimetre area substrate; thermal drift; movement/perturbation induced by tip motion and tip contamination [21]; the nonplanar nature of components within individual fibril units; and the presence of dangling alkyl chains. High-resolution STM imaging of 1-D structures has been successful in studying films of strands [22–24] and innate graphitic structures [25–27], but less so with isolated organic strands. Some reports of STM imaging to obtain high-quality images of strands include those of polypropylene [28], molecular chains of magnetic molecules [29], silicon nanowires [30], and DNA/biomolecules [31–32].

Intrigued by the crystallite morphology of 2CHd-10, we investigated its chain length variants, namely 2CHd-6 and 2CHd-14, by AFM imaging, which vindicates their fibrillar-crystallite nature, with 2CHd-14 seeming to produce the longest fibres amongst the three, as demonstrated in Fig. 3. On the other hand, 1CHn-10 deviates from the wedge-shape due to having only one alkoxy chain, but optical microscopy and AFM images reveal that 1CHn-10 is capable of forming fibril assemblies extending up to several tens of micrometres, as evident from Fig. 2. Note that 1CHn-10, however, offers an additional hydrogen-bonding site (believed to enhance column ordering in the bulk). An explanation for this seemingly contrasting behaviour of 1CHn-10 and 2CHd-10 critically depends on the knowledge of the respective elementary fibril structures.

To investigate the structure formation of 2CHd-10 on HOPG at a molecular scale, a drop of a dilute solution of 2CHd-10 (≈0.24 wt %) dissolved in 1,2,4-trichlorobenzene (C₆H₃Cl₃) was deposited on HOPG and the liquid/solid interface searched for the thinnest fibrils by STM. The use of dilute solutions for STM studies is prompted by the requisite of locating isolated elementary strands on the substrate. Fibre bundles are oriented mostly randomly but isolated elementary fibrils follow low-index graphite surface directions. Generally, the length of isolated fibrils exceeds the scan range of the STM (≈800 nm). Fig. 4 is a typical STM image showing a bundle consisting of two elementary fibrils, while Fig. 4 is a close-up of the left fibril taken to reveal the internal structure and its dimensions.

Having a width of 5.6 nm, the fibril consists of bright blobs arranged side-by-side in a zigzag pattern that slightly varies along the fibril. Bright blobs can sometimes be resolved into two elliptical features that are about 1.4 nm long (arrows in Fig. 4). Assuming that electron-rich delocalized π clouds of the aromatic rings dominate the image contrast [37], the bright blobs are interpreted as hydrogen-bonded dimers, as shown in Fig. 4. Note that the distance between adjacent bright blobs is 1.5 and 2.5 nm in directions perpendicular and parallel to the fibril axis, which is much larger than the interstack distance of 0.35 nm observed in mesophase columns in the bulk [10]. This means that the fibrils cannot be explained by a stacked structure stabilized by π* orbital overlap of adjacent aromatic rings. The measured heights (brightness) of individual bright blobs in a zigzag vary slightly, which could be a convolution of electronic and topographic effects implying the three-dimensional nature of the structure hidden in the topographic image.

A structural model is proposed for the 2CHd-10 fibril based on dimer precursors involving mainly hydrogen bonding along the circumference and van der Waals bonding between interdigitated dangling alkyl chains along the fibril axis, as shown in Fig. 5 as a “net” of the fibril. Note that the periodicity of the structure along the fibril axis, predicted by this model and the molecular dimensions given in Fig. 1, is in reasonable agreement with the periodicity of 2.5 nm observed in experiments (see Fig. 4). Fig. 5 is a 3-D model in which individual ellipsoids represent 2CHd-10 molecules. The ellipse representing the bright STM contrast feature defines the symmetry axis of the molecule while the fibril axis is defined by the direction of the alkyl chains, as illustrated in Fig. 4. The tilt of α = 9° between the molecular axis and the fibril axis is determined by aligning the alkyl chains while reducing their bending with respect to the aromatic rings to a minimum.

One could construct a perfectly planar molecular layer of surface-filling molecules based on the described construction principles. However, zigzag structures result from defects introduced by dimers flipped by 180° around the fibril axis (shaded blue and green in Fig. 4). As evident from Fig. 4 and Fig. 5, such a flipped molecule can form only one hydrogen bond with the neighbouring dimer and is tilted in the opposite direction yielding a step in the molecular contour of the hydrogen-bonded units. The loss of one hydrogen bond at defect sites is partially compensated by additional interdimer hydrogen bonds (see Fig. 5). Due to a perfect interdigitation of alkyl chains, the fibril has only a weak interaction with the substrate. We speculate that defects introduce internal stress resulting in a small bending of the initially planar sheet. A fibril fragment may result if the specific zigzag structure facilitates a hydrogen-bond closure from open hydrogen bonds, as indicated (unbonded H atoms at the top of the net and O atoms at the bottom) in Fig. 5.

Fibril fragments can grow with different diameters depending on the number of molecules in the sheet, while the detailed zigzag structure determines whether a closure is possible or not. Once a closed fragment is formed, the fibril can easily grow along its axis by the attachment of more dimer units. Planar fragments that are unable to close may still grow axially leading to fibrillar crystallites, as the axial growth mechanism is basically the same as that for a closed net, i.e., through van der Waals interactions. It can also be conjectured that a closure is most plausible for nets with a small diameter, whereas large nets may lie flat on the surface and grow as crystallites. It is worth noting that the model described here displays striking similarities with the bulk mesophase fibrils in its basic constitution. First, from X-ray data for the bulk mesophase fibrils, the number of molecules per column (a disc) cross section is also found to be two, i.e., a dimer [6]. Second, the periphery of the fibril cross section in the columnar hexagonal disordered mesophase consists of six dimer units just as for the six-membered dimer fibril cross section shown in Fig. 5.

Next, we investigated the structure formation of 1CHn-10 on HOPG at the molecular scale. Fig. 6 shows an STM image taken after a drop of a dilute solution of 1CHn-10 in 1,2,4-trichlorobenzene (≈0.29 wt %) had been deposited on HOPG. Again fibrils with a length covering the entire STM scan range are observed. Unlike for 2CHd-10, we observe single-strand and three-strand fibrils (top view) as shown in Fig. 6. As revealed by Fig. 6, the periodicity of bright blobs along the fibril bundle as well as the lateral distance between strands is 2.1 nm. The individual bright blobs of about 1.5 nm appear as rather elusive features in single-strand fibrils. We assume hydrogen-bonded tetramers, as shown in Fig. 6, to be the building blocks for the strand structure. The tiny single-strand fibrils yielding only unstable STM contrast may simply be linear arrangements of tetramers between overlapping alkyl chains.

The three-strand geometry appears as a much more rigid structure, and a plausible repeating unit for it is illustrated in Fig. 6 in which four tetramer building blocks form a ring stabilized by van der Waals interactions between dangling alkyl chains. This ring is a highly symmetric unit in which the tetramer aryl cores (assuming the tetramers to be planar) appear pairwise parallel (tetramers 1║3 and 2║4), with 2 and 4 displaced from 1 and 3 by half the periodicity along the fibril axis. In a projection perpendicular to tetramer 2, tetramer 4 appears precisely below tetramer 2, and 1 and 3 appear symmetrically at the sides of 2 where the connecting lines 1–2 and 2–3 enclose an angle of 120°. In such a ring unit, 8 of the 16 available alkyl chains are van der Waals bonded to each other while four are dangling at one side of the ring and four at the other side (two alkyl chains each from 2 and 4 are not shown in Fig. 6).

Unlike 2CHd-10, here no condition is to be met for the closure of 1CHn-10 tetramers to form a ring. Due to the symmetry of the ring unit, the dangling alkyl chains are at the right positions to connect to alkyl chains of a following ring in the very same manner the tetramers in a ring unit are bonded internally, i.e., through van der Waals interactions. This interaction between subsequent tetramer rings leads to their growth into a linear chain. Hence, a string of ring units yields a fibril with a well-defined diameter and a saturation of all possible van der Waals bonds between alkyl chains. Assuming the strand of tetramers numbered 4 is lying flat on the HOPG surface, the fibril structure appears in the above-mentioned projection, and tetramers 1, 2 and 3 appear as bright blobs in the STM image of Fig. 6 ordered in a linear herringbone arrangement exhibiting the 120° angle. A 3-D space-filling model for the fibril is shown in Fig. 6, which is a periodic structure of the ring unit of Fig. 6 at a periodicity of 2.1 nm.

To visualise the structure of the 1CHn-10 fibril more clearly, a simplified model is shown in Fig. 7. The “net” in Fig. 7 (fibril dimensions are not drawn to scale for the 2-D representation) shows the interdigitation of alkyl chains between neighbouring tetramers in which the tetramer building blocks are represented by squares. The structure constituted by blocks 1, 2, 3 and 4 represents a repeating unit of the fibril. The aliphatic chains of the subsequent tetramers interact through van der Waals forces between the interdigitating chains. Thus, the capability of 1CHn-10 to achieve a fibril structure is based on its tendency to form tetramers, which is a crucial step in the process. The closure of the “net” is facilitated by the coupling between tetramers 3 and 3′ and tetramers 4 and 4′ and similarly all equivalent tetramers along the fibril, naturally defining the unique diameter of the fibril.

File 1

This article is part of the Thematic Series "Self-assembly at solid surfaces".