Introduction

Assembly is a long-term significant endeavor since it provides an alternative to synthesis in the creation of advanced materials1,2,3,4,5. Various layer-by-layer assemblies with attractive properties had been prepared by designing the constituent molecules6,7 and showed broad applications in catalysts8,9, organic solar cells10,11, flexible electronics12,13, and condensed matter physics14,15. It was widely recognized that the stacking accuracy between the adjacent layers dominates the properties of layer-by-layer assembly16,17. Although access to the accuracy control has been tackled by methods ranging from template utilization18,19, molecular composition design20,21, and deposition procedure adjustment7,22, the state-of-the-art methods are limited to pre-modulated ones. It is of great significance to develop a post-modulation strategy, i.e., a strategy that modulates the structure after the assembly has already been constructed. However, the current methods rely on the externally applied fields, greatly limiting the application scenarios23,24.

The non-covalent chemistry predicts a route to experimental post-modulation of assembly in the absence of external fields25,26. Compared to the covalently constructed assembly, the non-covalent one is advantageous in structural modulation because of the flexible weak interactions therein27,28. In particular, the low strength leads to the dynamic dissociation and formation of weak interactions29, which inversely promises post-modulation once there is some control over the dynamic process. Nevertheless, it remains a challenge to construct and modulate a layered assembly that was fully assembled by weak interactions. Fullerene and its derivatives, featured by the large conjugation areas and the uniform spherical morphology, are considered potential building blocks in full-weak-bonded assembly30,31,32. Recently, it was demonstrated that fullerene was able to act as charge reagent and underwent cyclic electron transfer with other species, termed the electron buffering effect33. Given that fullerenes and their derivatives interact with other assembly subunits mainly through the delocalized π system, it provides an approach to achieve the post-modulation on assemblies in synergy with the weak interactions.

Herein, through the synergy of fullerene derivatives and full-weak-bonded interactions, we demonstrate the post-modulation of the layer-by-layer assembly in the absence of external fields. As a proof of principle, a bilayer full-weak-bonded assembly is prepared by stacking two self-assembled monolayers (SAMs) via the π-π stacking interaction. Scanning tunnelling microscopy (STM) images show an interlayer angular mismatch of ≈12° in the as-fabricated layer-by-layer assembly. We introduce the thiophene-functionalized fullerene (TS-C60) to the layered assembly and demonstrate the post-modulation process for angular mismatch in the absence of external fields. By employing the combined microfluidics and STM methods, we find that the TS-C60 initiates and coordinates the dynamic disassembly and reassembly of the second SAM (Fig. 1), and finally achieves the precise alignment of the two SAMs. The post-modulation is rationalized by the theoretical simulation conducted utilizing the molecular mechanics calculation.

Fig. 1: Schematic post-modulation of an angularly mismatched layer-by-layer assembly with a fullerene derivative.
Fig. 1: Schematic post-modulation of an angularly mismatched layer-by-layer assembly with a fullerene derivative.
Full size image

Top panel, chemical structures and corresponding modeling of the TS-C60, TMA, and TPDA molecules. Bottom panel, schematic illustration of the angular mismatch modulation of the assembly coordinated by TS-C60.

Results

Selection and self-assembly of the assembly primitives

To fabricate a layer-by-layer assembly, we prepared the first SAM by employing 1,3,5-benzenetricarboxylic acid (TMA) molecules, and the second SAM layer by TMA and [1,1’:3’,1”-terphenyl]-4,4”-dicarboxylic acid (TPDA) as building blocks, respectively. TMA molecules were proved to be able to form diverse self-assembled structures on highly oriented pyrolytic graphite (HOPG) surface34. The carboxyl moieties facilitate the formation of the C-H···O bonds, and the directionality of the hydrogen bonds enables the customization of the assembly morphology. The aromatic nature of TMA contributes to the formation and the stability of the layer-by-layer assembly via the π-π stacking interaction. This interaction stems from the stacking of π systems between the TMA and TPDA, as well as between the TMA and the substrate. For the second layer of the assembly, TPDA was selected considering its similar carboxylic structure. Theoretically, either its self-assembled product or the co-assembled product with TMA has the potential to form porous macrocycles through the directional hydrogen bonding. Another advantage of TPDA is that its longer molecular length brings about a greater accommodation cavity than TMA.

To clarify the behavioral rules of the building blocks forming supramolecular assemblies, we first investigated the self-assembly of TMA and TPDA molecules, respectively. STM was employed to characterize the assemblies in experiments. The single-component monolayer was obtained after the dropwise addition of a saturated heptanoic acid solution of TMA to the freshly cleaved HOPG surface. At the liquid-solid interface, TMA molecules self-assembled into networks switching among the honeycomb, flower, and zigzag motifs as the sample bias changed (Supplementary Figs. 1, 2)35, exemplifying the tuneable nature of the weak interactions. Specifically, a six-folded porous structure laid parallel to the HOPG surface at -500 mV, with each TMA molecule interacting with three other TMA molecules via hydrogen bonding (Fig. 2a, b). Combined with deep learning algorithms (see Supplementary Data 2), we calculated and calibrated the lattice parameters for each assembly (Supplementary Fig. 3). Taking measured values of a1 = b1 = 2.61 ± 0.03 nm, γ1 = 60.4 ± 0.4° (Fig. 2c), lattice parameters show a good conformity with the simulated results of a1’ = b1’ = 2.62 nm, and γ1’ = 60.0°, creating a (\(\sqrt{111}\times \sqrt{111}\))R17° flower monolayer on HOPG substrate (Supplementary Figs. 4, 5, Supplementary Table 1). Although the SAM with a porous flower pattern can serve as a host structure to establish the host-guest supramolecular complex, the compatibility is considerably low on account of the limited molecular length and the poor variability of the cavities self-assembled from TMA alone.

Fig. 2: The angularly mismatched full-weak-bonded layer-by-layer assembly.
Fig. 2: The angularly mismatched full-weak-bonded layer-by-layer assembly.
Full size image

a Schematic illustration of TMA SAM. b, c Large-range (b) and high-resolution (c) STM images of porous flower pattern of TMA. The measured lattice parameters are a1 = b1 = 2.61 ± 0.03 nm, and γ1 = 60.4 ± 0.4°. d Schematic illustration of the angularly mismatched full-weak-bonded layer-by-layer assembly with TMA and TPDA as building blocks. e Large-scale STM image of the angularly mismatched full-weak-bonded layer-by-layer assembly. The second layer exhibits a compressed-flower pattern and shows a brighter contrast than the first flower layer. f High-resolution STM image of the second-layer assembly shown as compressed-flower pattern with molecular model overlaid. The lime green and yellow molecules in molecular models represent TMA and TPDA, respectively. The measured lattice parameters are a2 = 2.52 ± 0.03 nm, b2 = 2.40 ± 0.03 nm, γ2 = 64.0 ± 0.2°. g Height profile of the first-layer assembly and the layer-by-layer assembly. h FFT image of (e). The yellow and magenta spots identify the lattice of the first layer and the second layer with angular mismatch, respectively. i, j Top-view of molecular models of the full-weak-bonded layer-by-layer assembly with (i) and without (j) angular mismatch. The blue and red axes represent the b vectors of the lattices of the flower and compressed-flower pattern, respectively. White arrows: symmetry axis of HOPG. Tunneling conditions: Iset = 80 pA, Vbias = −500 mV. Source data are provided as a Source Data file.

TPDA represents a potential candidate to improve the compatibility of the host structure. It features the same functional group but a longer molecular length compared to TMA. We investigated the self-assembly behavior of TPDA at heptanoic acid/HOPG interface in response to sample bias. Nevertheless, no well-defined assembly was observed at the positive bias, while the close-packed zigzag lattice appeared at the negative bias (Supplementary Fig. 6a, b). This observation is ascribed to the fact that there are only two carboxylic moieties in TPDA capable of forming hydrogen bonds. The simulated flower pattern of TPDA shows that the macrocycles may only be sustained by intermolecular van der Waals forces, making it insufficient to support a porous network of such large size (Supplementary Fig. 6d). And the TPDA molecule presents a steric structure due to the rotatability of the C-C bonding between the benzene rings, and is, therefore, less stable on HOPG substrate than flat lying TMA.

Angle-mismatched full-weak-bonded layer-by-layer assembly

To have the compatibility of TPDA while keeping the characteristics of porous structure, we combined the advantages of TMA and TPDA and investigated the assembly behavior of their mixture. TMA and TPDA were pre-mixed before being added to the HOPG surface. Supplementary Fig. 7a, d shows that there emerged an alternative phase featuring a pair of shared TMA molecules between two adjacent macrocycle units, proving the feasibility of co-assembling two molecules into a porous monolayer. Based on this, a sequential addition was adopted to prepare a layer-by-layer structure. Experimentally, a droplet of TPDA-saturated heptanoic solution was immediately added once the flower monolayer of TMA was obtained. A structure with both flower and another porous phase was discerned. Zoom-in image of this porous phase in Fig. 2f exhibits a compressed lattice on the vertical (the a vector) and the lateral (the b vector) directions, termed here the compressed-flower pattern, with the measured lattice parameters of a2 = 2.52 ± 0.03 nm, b2 = 2.40 ± 0.03 nm, and γ2 = 64.0 ± 0.2°. Each TPDA molecule is attracted to two TMA molecules through C-H···O interaction and is stabilized by van der Waals interaction with two other TPDA molecules therein.

It is noticed that there is a difference in contrast between the flower and the compressed-flower motifs (Fig. 2d, e, Supplementary Fig. 7e), and we therefore propose the possibility of forming a layer-by-layer structure based on the principle of constant current mode. Figure 2g gives the height distribution of the two motifs and it witnesses a distinct elevation gain of the compressed-flower layer relative to the flower layer, with a roughly two-fold relationship in height. The valleys of the compressed-flower layer are always higher than the peaks of the flower layer. To further confirm the formation of the layer-by-layer assembly, the compressed-flower layer was selectively removed by a sudden increase in the tunneling current. As shown in Supplementary Fig. 8, the intact flower layer was exposed with the removal of the compressed-flower layer. Considering all of above, the layer-by-layer structure is confirmed to be constructed with the flower layer as the first layer (bottom layer) and the compressed-flower layer as the second layer (top layer). Since all molecules within the assembly are held together by weak interactions, the full-weak-bonded layer-by-layer assembly was built.

We further explored the interlayer correlations of the full-weak-bonded layer-by-layer assembly. The interested scope was localized at the boundary of the layered structure. As shown in Fig. 2e, there is an angular mismatch between two layers as the b vectors of the lattices of the first layer (b1, blue arrow) and the second layer (b2, red arrow) are not parallel. The angular mismatch is further validated by the fast Fourier transform (FFT) results that the magenta spots (for the second layer) emerged at the ≈12°-deflected position of the yellow spots (for the first layer) when the observation window moved from the second layer alone to the boundary (Fig. 2h, Supplementary Fig. 9). By counting area of more than 4.0 × 104 nm2, the angularly mismatched domain occupies 59.7% of the full-weak-bonded layer-by-layer assembly. As shown in Fig. 2i, we built the molecular model of the 12°-mismatched structure, with the manually aligned result as comparison (Fig. 2j). It is noticed that the elimination of angular mismatch increases the area for most of the bilayer cavities that are formed by the stacking of the macrocycle units from the first and the second layers. Therefore, it is worthwhile to modulate the angular mismatch in the assembly.

Angular modulation coordinated by fullerene derivative

Given the porous characteristic of the full-weak-bonded layer-by-layer assembly, fullerene and its derivatives stand out as candidates with the potential to modulate angular mismatch. We designed and employed the thiophene-functionalized fullerene TS-C60 to perform angular modulation experiment. Figure 3a gives the synthetic route of the TS-C60 molecule36. Its structure was confirmed by 1H NMR, 13C NMR, APCI-MS and UV-vis characterizations and finally obtained a yield of 43.0% (Supplementary Figs. 10–13). Experimentally, we added 2.0 μL of TS-C60-saturated heptanoic solution to the angularly mismatched layer-by-layer assembly. An undisturbed layered structure with clear boundaries was displayed by Fig. 3d. TS-C60 molecules appeared as dispersed bright spots embedded in the macrocycles, which were assigned as isolated protrusions in height profile (Fig. 3g, Supplementary Figs. 14, 15). The layer-by-layer structure was further identified, which is characterized by a two-fold relationship between the heights of the two layers.

Fig. 3: Post-modulation of the full-weak-bonded layer-by-layer assembly coordinated by TS-C60.
Fig. 3: Post-modulation of the full-weak-bonded layer-by-layer assembly coordinated by TS-C60.
Full size image

a Synthetic route of the TS-C60 molecules. b Schematic illustration of the mismatch-free full-weak-bonded layer-by-layer assembly after the addition of the TS-C60. c, d High-resolution (c) and large-area (d) STM images of the post-modulated full-weak-bonded layer-by-layer assembly at the boundary of the first and the second layers. The interlayer angular mismatch is eliminated with the coordination of TS-C60. e, f Distribution of the TS-C60 in the first layer alone (e) and in the full-weak-bonded layer-by-layer assembly (f) after 90 min of addition of TS-C60, respectively. g Height profile of the post-modulated layer-by-layer assembly. h FFT spline of the post-modulated layer-by-layer assembly. The yellow and green spots represent the first layer and the second layer without angular mismatch, respectively. White arrows: symmetry axis of HOPG. Tunneling conditions: Iset = 50 pA, Vbias = −500 mV. Source data are provided as a Source Data file.

It was found that the interlayer angular mismatch disappeared after the embedding of TS-C60. High-resolution STM image shows that the b vectors of the lattices of the two layers lie almost parallel to each other (Fig. 3b, c). The FFT pattern in Fig. 3h also witnesses the angular modulation: the magenta spots (for the second layer in 12°-mismatched structure) disappeared, while there emerged green dots (for the second layer in mismatch-free structure) adjacent to the yellow dots (for the first layer). That is, we achieved the modulation of the interlayer angular mismatch after the formation of the layer-by-layer assembly. We performed repeated experiments to show the controllability of the post-modulation method. The statistical results showed that the average interlayer angular mismatch was reduced from 11.7° to 1.7° after the introduction of TS-C60 (Supplementary Figs. 16, 17). To clarify the object of the angular modulation, we compared the orientation of the assemblies with respect to the atomic arrangement of HOPG (Supplementary Fig. 18). The interlayer angular mismatch of the layered structure maintained stability for more than 10 min in absence of TS-C60, which indicates that the angular modulation cannot occur spontaneously. Moreover, the first layer embedded with TS-C60 essentially remained unchanged in its orientation on the HOPG, demonstrating that the modulation of the interlayer angular mismatch mainly acted on the second layer. After calculation, the position change of the end of the b vector of the compressed-flower cell before and after angular modulation can reach to a distance of 5.0 Å.

To perform control experiments, we employed pristine C60 and pyridine-functionalized C60 (PN-C60) to modulate the interlayer angular mismatch. As shown in Supplementary Fig. 19a, pristine C60 was found to have a limited influence on angular modulation due to the rapid dynamic embedding-leaving process, making it more difficult to maintain long-term stability in macrocycles. In contrast, PN-C60 is able to achieve similar angular modulation to the TS-C60 (Supplementary Fig. 19b, c). We compared the occupancy rate of three fullerene derivatives in the layer-by-layer assembly. As shown in Supplementary Table 2, pristine C60, TS-C60 and PN-C60 witnessed the average occupancy rates of 0.9%, 1.3%, and 3.8%, respectively. In particular, the long-time scanning reveals a significant difference in embedding behavior of the TS-C60 in the first layer alone and in the layer-by-layer assembly. It was found that TS-C60 began to embed in the first layer within several minutes after addition (Supplementary Fig. 20a), and 90 min of continuous testing witnessed an occupancy rate of 90.1% (Fig. 3e). However, Fig. 3f and Supplementary Fig. 21 show that the occupancy rate of TS-C60 in layer-by-layer assembly was only 1.7% despite a sufficient testing time. We applied image recognition to identify the extent of modulation of each fullerene derivative (detailed calculations shown in Supplementary Fig. 22). Other than the limited modulation of pristine C60, the average modulation area per TS-C60 is estimated to be 349 nm2, whereas each PN-C60 molecule has about half of the modulation area compared to TS-C60. This finding illustrates that the fullerene derivatives featured a catalyst dosage in the post-modulation strategy.

Subsequently, the in-situ dynamic process of the post-modulation coordinated by the TS-C60 was investigated by a combined methodology of microfluidic and STM setup (Fig. 4a, b, Supplementary Fig. 23). It takes the advantages of precise and controllable features of microfluidics and high-resolution characteristic of STM. Optimized microfluidic flow rate was controlled at 1.0 μL min−1 (Supplementary Fig. 24). By slowly injecting the TS-C60 heptanoic acid solution, the dynamic process of angular modulation was visualized in Fig. 4c–f. Within 221 s, the angular mismatch between the first and the second layers decreased from 11.3° to 9.7°. In the following 180 s, the value decreased again by 3.4°, and the final angular mismatch further lowered to 2.6°.

Fig. 4: Disassembly and reassembly of the full-weak-bonded layer-by-layer assembly.
Fig. 4: Disassembly and reassembly of the full-weak-bonded layer-by-layer assembly.
Full size image

a Device diagram of the combined microfluidic and STM methodology. b Schematic illustration of utilizing the combined methodology for in-situ investigation of the dynamic process of the post-modulation. c–f Dynamic process of angular modulation coordinated with TS-C60. The interlayer mismatch angles are 11.3° (c), 9.7° (d), 6.3° (e) and 2.6° (f), respectively. Blue and red arrows: the b vectors of the lattices of the first-layer assembly (blue) and the second-layer assembly (red), respectively. g–n The modulation process of the TS-C60 for the full-weak-bonded layer-by-layer assembly. STM images of the layered assembly with large-area (g–j) and fine (k–n) post-modulation coordinated by TS-C60 at 0 s (g, k), 50 s (h, l), 100 s (i, m), and 150 s (j, n). The flow rate was 1.0 μL min−1. White arrows: symmetry axis of HOPG. Tunneling conditions: Iset = 100 pA, Vbias = −500 mV.

During the experiments, fullerene derivatives were found to induce large-area as well as fine assembly and disassembly of the second-layer structure. Figure 4g–j exhibits the dynamic process in which the second-layer assembly underwent repeated disassembly and reassembly, as evidenced by several changes in the boundary line. Eventually, a large-area disassembly of the second layer was presented after 150 s, exposing the underlying first layer. The fine-tuning manifests itself in modest adjustments to the boundary line (Fig. 4k–n). In Fig. 4k–m, successive scanning witnessed the connection and fracture of the boundary lattice, and the outward expansion of the second-layer assembly, shown as the formation of a new row of boundary cells (Fig. 4n). In the zoom-in STM images, fullerene derivative-induced lattice extension becomes more pronounced. There captured a gradual completion process of the macrocycles around the TS-C60 at the boundary of the layer-by-layer assembly (Supplementary Fig. 25e, f). As marked by yellow rectangles, the second-layer assembly extended within 26 s after a TS-C60 molecule embedded into a macrocycle of the first layer. The lattice extension carried well along the mismatch-free orientation even after a few cell units, showing the long-range influence of TS-C60 (Supplementary Fig. 26). It was found that when some TS-C60 molecules left the previously embedded cavities, the layered assembly therein remained intact, reminiscent of the case that the catalyst left the final product in a catalytic reaction.

Post-modulation process with the synergy of multiple factors

In our post-modulation strategy, the TS-C60 with steric structure exhibits a significant distinctiveness. We set up control experiments by employing the planar pentacene and hexabenzocoronene as coordinators. They were added to perform co-assembly with the angularly mismatched layer-by-layer structure. Highly ordered domains were captured with pentacene covering the honeycomb assembly of TMA in a pairwise manner (Fig. 5a), and hexabenzocoronene formed a SAM superimposed on the flower assembly of TMA (Fig. 5c). Yet identically, no TPDA molecules were found in any of the structures, as evidenced by their high-resolution STM images (Fig. 5b, d). This is unusual because the simulated results from molecular model indicate that, both pentacene and hexabenzocoronene are able to embed in the cavities in the second layer. The only difference lies in their embedding behavior in the first layer, where pentacene is kept out of the cavities due to its longer molecular length (Supplementary Fig. 27). In this case, the original layer-by-layer assembly should still be maintained. However, experimental results demonstrate that planar molecules failed to keep the original structure, which suggests that it is necessary for coordinators to have simultaneous interactions with both layers of the layered structure. In this regard, the height of the TS-C60 allows it to penetrate the layered assembly, thus creating C-H···π interactions with each layer, which serves to stabilize the overall assembly. For TS-C60, the introduction of thiophene branched chain further stabilizes its embedding in macrocycles. Calculation of the molecular electrostatic potential for TS-C60 shows that the thiophene moiety is of electron-withdrawing effect, which gives the spherical portion of the TS-C60 a more pronounced electron-deficient character compared to pristine C60 (Supplementary Fig. 19d, e). It favors the charge transfer and thus enhances the interaction between the macrocycles and TS-C60 in the host-guest structure32,37. The rapid embedding-leaving process of pristine C60 in cavities is therefore avoided and the introduction of thiophene branched chain contributes to the retention and functionality of TS-C60 in the layered assembly.

Fig. 5: Energetics of the post-modulation.
Fig. 5: Energetics of the post-modulation.
Full size image

a–d Large-scale and high-resolution STM images of control experiments with planar molecules in place of the fullerene derivatives. Co-assemblies of pentacene (a, b) and hexabenzocoronene (c, d) with angularly mismatched full-weak-bonded layer-by-layer assembly, respectively. Insets: structures of pentacene and hexabenzocoronene. White arrows: symmetry axis of HOPG. Tunneling conditions: Iset = 100 pA, Vbias = −500 mV. e Clustered Moiré pattern of the angularly mismatched full-weak-bonded layer-by-layer assembly. f Periodic patchy Moiré pattern of the full-weak-bonded layer-by-layer assembly after angular modulation. g Relative energy diagram of spontaneous path and fullerene derivative-induced path for angular modulation. h Upper panel, distribution of overlap area suitable for TS-C60 embedding before (blue stripes) and after (orange stripes) angular modulation. The statistical step is 0.04 nm2. Lower panel, representative overlap areas and corresponding stabilization energy of TS-C60. i, j I-V characteristics (i) and differential conductance (dI/dV) versus sample bias (j) of the bare HOPG (cyan), TMA SAM (orange) and layer-by-layer assembly before (blue) and after (red) angular modulation. STS conditions: Iset = 100 pA, Vbias = −1000 mV. Source data are provided as a Source Data file.

The nonplanar structure of TPDA also has an influence on the post-modulation process. To provide a comparison, experiments were performed using isophthalic acid (IPA) with similar carboxyl moiety distribution but a planar structure. Highly ordered domains were obtained after heptanoic acid solution of IPA being added to the first layer self-assembled by TMA (Supplementary Fig. 28). Rather than forming a porous structure, such domains exhibit a zigzag motif co-assembled by TMA and IPA. This phenomenon is related to the non-flat lying conformation on the substrate due to the uneven distribution of carboxylic acid moieties in IPA and the lack of van der Waals forces among the C-H bonds resulting from its shorter molecular length to support the formation of flower assembly38. In contrast, although TPDA exhibits a non-planar configuration and reduces the overlap of the π system of the benzene ring between two layers, the rotatability of C-C bond connecting benzene rings is able to compensate for the reduction by weakening the distortion of molecule. It improves the compatibility of TPDA and facilitates the formation and the stability of the layer-by-layer assembly (Supplementary Figs. 29–31). Moreover, a smaller interlayer spacing contributes to obtaining sufficient force to sustain the stability of the overall structure compared to that between the first-layer assembly and the substrate (Supplementary Fig. 27d).

The combination of these two non-planar molecules facilitates the embedding process of TS-C60 to some extent. Due to the dynamic adjustment of the orientation of TS-C60 in the bilayer cavity, there exist chances for the thiophane moiety of TS-C60 to form a T-shaped structure with the C-H of the TPDA or a parallel stacking with benzene rings of TMA or TPDA. It provides additional C-H···π interactions to the whole system. At the same time, the additional steric hindrance brought by the introduction of branched chains and the non-planar features of the building blocks should also be considered. Therefore, the post-modulation is the result of the mutual coordination and maintenance of equilibrium among various factors.

Post-modulation in the energy perspective

To better demonstrate the post-modulation process, we calculated the free energy change of the system. In the process of angular modulation, the induction of fullerene derivatives is of a paramount importance. We compared the free energy changes of the spontaneous path and fullerene-induced path in angular modulation. As shown in Fig. 5g, spontaneous path gives a positive Gibbs free energy indicating that the process cannot proceed. In contrast, the fullerene-induced path has a reduction in energy, which illustrates the characterization of fullerene derivatives as catalysts. That is, the introduced fullerene derivative has ability to select the pathway of assembly to obtain products that would otherwise be energetically unfavorable in the conventional path.

With the decrease of the interlayer angular mismatch, the relationship between the first and second layers of the assembly also changes accordingly. The mismatch of lattice parameters of flower and compressed-flower motifs as well as the interlayer angular mismatch led to the formation of the Moiré patterns (Supplementary Fig. 32). The Moiré patterns of 12°-mismatched and post-modulated assemblies were simplified and reproduced in Fig. 5e and f, respectively. We counted the overlap area of the bilayer cavities (created from the stacking of the two kinds of macrocyclic units of the flower and compressed-flower lattices) at different positions, and colored them according to their values. It was evaluated that the overlap area should be no less than 1.0 nm2 if the TS-C60 is allowed to embed in and form a bilayer host-guest complex. For 12°-mismatched assembly, the sites with the required overlap area occurred intermittently in a clustered fashion (Fig. 5e), while they connected into periodic patches in the post-modulated one (Fig. 5f). That is, as the result of the post-modulation, the alteration in the Moiré pattern leads to the change in the embedding sites of the TS-C60. The distribution of the overlap area before and after the angular modulation is summarized in Fig. 5h. The proportion of the cavities suitable for TS-C60 embedding increases from 22.2% to 43.4% when angular mismatch is aligned. The alteration of Moiré patterns explains the nearly nine-fold difference in occupancy rate of TS-C60 in first layer alone and the layer-by-layer structure. Specifically, the probability of TS-C60 embedding in each macrocyclic unit is identical due to the unobstructed surface of the first layer alone. In contrast, when the first and the second layers are stacked, a large number of cavities are filtered out because of the obscuration of the macrocyclic units. This yields a low occupancy rate of TS-C60 in the layer-by-layer assembly.

To further illustrate the post-modulation process from an energetic perspective, we performed structure optimization and energy calculations for the related assemblies in experiments (Supplementary Fig. 33). It shows that there is an energy advantage of 25.24 kcal mol−1 for the 12°-mismatched assembly than that for the mismatch-free one (Supplementary Table 3), which convinced the fact that the former accounts for the majority of the domain. Whereas, after the introduction of TS-C60, the absorption energy of the whole system decreases because of the additional stability brought about by the host-guest interaction (Supplementary Table 4). This result agrees with the variation of the free energy and shows the feasibility of the post-modulation process. We further calculated the stabilization energies of TS-C60 in the cavities with different overlap areas (Supplementary Fig. 34). As shown in Fig. 5h, the maximum overlap area of 1.6 nm2 occurs when the centers of the macrocyclic units of the first and the second layer are aligned. In this case, the whole cavity space is available to accommodate the TS-C60 because of the minimal steric hindrance. The stabilization energy therein is maintained at an appealing level of −1.88 eV to attracting TS-C60 molecules. We also found that the most favorable position in energy (with the stabilization energy of −2.00 eV) is not located at the orthocenter of the macrocycle but occurs when the centers are slightly staggered39,40,41. Further misalignment of the centers of the macrocyclic units leads to a reduction in overlap area and a rapid elevation of the stabilization energy, culminating in the formation of an energetically unfavorable compressed-flower-TS-C60 monolayer host-guest structure (Supplementary Table 5).

To explore the influence of interlayer mismatch angle on the electronic properties of layer-by-layer assemblies, we performed scanning tunneling spectroscopy (STS) tests before and after the angular modulation (Supplementary Fig. 35). The current-voltage (I-V) characteristics in Fig. 5i shows an asymmetric behavior for the first layer alone and layer-by-layer structure with angular mismatch. They give a larger tunneling current at negative sample bias, suggesting that in both cases the highest occupied molecular orbital (HOMO) of the assemblies is closer to the Fermi level of the substrate42. It identifies the electron-donating properties of the macrocyclic units in porous assembly. In contrast, as shown in the dI/dV spectra in Fig. 5j, the HOMO and the lowest unoccupied molecular orbital (LUMO) lie at −0.91 and 0.84 V, respectively, giving a diminishing trend on asymmetry feature after angular modulation coordinated with TS-C60. Moreover, the embedding of TS-C60 and its effect of angular modulation lead to an increase in HOMO-LUMO gap of the layer-by-layer assembly, suggesting the potential applications of fullerene derivative-induced post-modulation strategy in the fields of organic electronics, optoelectronic devices, catalytic materials, and so on.

In light of the above consideration, we therefore propose a comprehensive framework for the post-modulation process in Fig. 6 with the combined experimental and theoretical results. Due to the energy preference, the second layer is more inclined to appear in a 12° angular mismatch from the first layer. When TS-C60 molecules are added, the cavities with large overlap areas are firstly occupied owing to the low stabilization energies. Finally, the interlayer angular mismatch is aligned through the localized disassembly and the reassembly via the synergy of the host-guest interaction and hydrogen bonding within the building blocks. The post-modulation effect of TS-C60 is then transmitted via the weak interactions, giving a long-range influence on the layer-by-layer assembly to construct a mismatch-free domain.

Fig. 6: Schematic illustration of TS-C60 as a coordinator to post-modulate the full-weak-bonded layer-by-layer assembly and induce the extension of the structure.
Fig. 6: Schematic illustration of TS-C60 as a coordinator to post-modulate the full-weak-bonded layer-by-layer assembly and induce the extension of the structure.
Full size image

TMA in the first layer: dark slate grey. TMA and TPDA in dissociated state: royal blue and pink. TMA and TPDA in 12°-mismatched pattern: sky blue and sandy brown. TMA and TPDA in post-modulated pattern: light blue and yellow.

Discussion

In summary, we have developed a strategy for the post-modulation of layer-by-layer assembly coordinated by fullerene derivatives with an accuracy of 5.0 Å. Interestingly, there were few fullerene derivatives in final assembly, demonstrating the role of coordinators for the fullerene derivatives in the dynamic disassembly and reassembly process, rather than traditionally being building blocks for co-assembly. The long-range influence of fullerene derivatives allows the epitaxial growth of layered assembly along the mismatch-free directions. The combined experimental and theoretical findings show the alteration in Moiré distribution of the cavities and elucidate the effects of steric configuration and stabilization energy on the post-modulation process. The STS experiment also demonstrates the potential of post-modulation strategies in regulating the properties of assemblies. Our work provides a method for the fabrication of atomically well-defined assemblies and offers the possibility for the development of molecular devices.

Methods

Materials

Fullerene (C60, 99.9%), thiophenesulfonamide (98%), and [1,1’:3’,1”-terphenyl]-4,4”-dicarboxylic acid (TPDA, 98%) were purchased from Bide Pharmatech Co., Ltd; 1,3-diiodo-5,5-dimethylhydantoin (DIH, 97%) and chlorobenzene (CB, 99%) were from Energy Chemical; heptanoic acid (>98%), 1,3,5-benzenetricarboxylic acid (TMA, 99%), pentacene (99%), and hexabenzocoronene (97%) were purchased from Tokyo Chemical Industry, J&K Scientific Limited, Solarbio, and Shanghai Yien Chemical Technology Co., Ltd. respectively. Reagents were not further processed before use.

Synthesis of thiophene-functionalized fullerene

The strategy for synthesizing thiophene-functionalized fullerene (TS-C60) molecules was derived from Nagamachi et al.36, as shown in Fig. 3. C60 (36.0 mg, 0.05 mmol), thiophenesulfonamide (38.0 mg, 0.10 mmol), and DIH (38.0 mg, 0.10 mmol) were added to a 50.0 mL three-necked full-weak-bonded layer-by-layer assembly equipped with a magnetic stirrer. After completely dissolving in 10.0 mL of CB by sonication, the resulting solution was placed in an ice bath at a pre-set temperature of 0 oC and stirred under N2 conditions. The reaction was carefully observed using thin-layer chromatography (TLC) and stopped at the indicated time. The reaction mixture was filtered through a silica gel plug to remove insoluble material. After solvent evaporation in a vacuum, the residue was separated on a silica gel column using carbon disulfide as an eluent.

Preparation of full-weak-bonded layer-by-layer assembly

Saturated solutions of TMA, TPDA, and TS-C60 in heptanoic acid were prepared by solving excess dosage of corresponding solids in the solvent, and ultrasonic agitation was applied before STM measurement to ensure the solutions were saturated. Pentacene and hexabenzocoronene were dissolved separately in heptanoic acid to prepare a solution with a concentration of 8.0 × 10−4 M. Fresh surfaces were prepared by mechanically cleaving HOPG before each experiment. To form a flower pattern of TMA, a 7.0 μL droplet of saturated TMA (heptanoic acid) solution was added along the STM tip on the HOPG surface. This dosing method allowed the STM tip to immerse naturally into the liquid film at the solid-liquid interface. The scanning process was started immediately after the deposition. Once the flower pattern of TMA was obtained, a 5.0 μL droplet of saturated TPDA (heptanoic acid) solution was carefully added to a homemade liquid cell without stopping the scanning process. A 2.0 μL droplet of saturated TS-C60 (heptanoic acid) solution was added after imaging the full-weak-bonded layer-by-layer assembly, proceeding with the same procedure.

STM characterizations

STM experiments were performed under atmospheric conditions using a Nanoscope instrument (Bruker, Santa Barbara, CA) in constant-current mode at room temperature. Mechanically cut Pt/Ir (80/20) wire with a diameter of 0.25 mm was used as the STM tip. Freshly cleaved highly oriented pyrolytic graphite (HOPG, grade ZYB) was used as the substrate for STM measurement. The bias in tunneling conditions refers to the sample bias.

Molecular mechanics simulations

Molecular mechanic simulations were conducted on Materials Studio 2018 using Forcite packages from Accelrys. A supercell cell of double-layer graphite containing 150\(\times\)150 unit cells graphite lattice (369.0 × 369.0 Å2) was constrained to fix the positions of all carbon atoms. A vacuum slab of 50 Å was built to avoid the influence of other parameters43. The geometry optimizations were performed using the pcff force field with a conjugate gradient algorithm44. The force convergence parameter was set to 10−2 kcal mol−1 Å−1, and the Spline function with cutoff distance of 14 Å and spline width of 3 Å for van der Waals and non-bond electrostatic terms were applied. The Ewald technique with an accuracy of 0.01 kcal mol−1 was used to describe electrostatic interaction. Molecular dynamics (MD) simulations were carried out in the canonical ensemble (N, V, T) using a Nose thermostat at 298 K40. The time step was set to 1 fs, and the total duration time was 100 ps, with an output frame for every 100 fs. Further details of the simulation set-up and discussion are given in Supplementary Notes 27, 28.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.