Self-assembled PCBM bilayers on graphene and HOPG examined by AFM and STM

In this work we report fabrication and characterization of phenyl-C61-butyric acid methyl ester (PCBM) bilayer structures on graphene and highly oriented pyrolytic graphite (HOPG). Through careful control of the PCBM solution concentration (from 0.1 to 2 mg ml−1) and the deposition conditions, we demonstrate that PCBM molecules self-assemble into bilayer structures on graphene and HOPG substrates. Interestingly, the PCBM bilayers are formed with two distinct heights on HOPG, but only one unique representative height on graphene. At elevated annealing temperatures, edge diffusion allows neighboring vacancies to merge into a more ordered structure. This is, to the best of our knowledge, the first experimental realization of PCBM bilayer structures on graphene. This work could provide valuable insight into fabrication of new hybrid, ordered structures for applications to organic solar cells.


Introduction
In the past several decades, organic solar cells have attracted tremendous scientific and industrial interest because their power conversion efficiency has dramatically increased and reached about 14% to date [1][2][3]. In addition, organic solar cells have potential advantages compared to traditional solar cells in flexibility of chemical modification as well as lowcost mass production [4]. Typically, an organic solar cell generates electric current through photon-induced electron transfer that separates electrons from holes [5]. Function of a solar cell depends on materials serving as electron donor and electron acceptor, respectively. As light enters a solar cell, the photons induce the electrons to transfer from the excited state of the donor to the lowest unoccupied molecular orbital of the acceptor. Subsequently, the separated electrons and holes reach the cathode and anode, respectively, delivering a direct current to an outer circuit [6]. The power conversion efficiency of a solar cell depends on various properties including electron affinity of electron acceptor.
In this work, we present the self-assembled structure of PCBM, a promising acceptor material for organic solar cells, deposited on graphene and HOPG. We discover novel bilayer nanostructures of PCBM on graphene and HOPG, and investigated how thermal annealing tunes the morphology of the PCBM bilayer, by using AFM and STM. Interestingly, PCBM bilayers are formed with two typical heights on HOPG, but only one on graphene. At different annealing temperatures, edge diffusion causes neighboring vacancies to emerge into a more ordered structure. This first experimental realization of PCBM bilayer structures on graphene may pave a way to fabricate hybrid structures of organic donor/acceptor molecules and graphene for applications in organic solar cells.

Experimental methods
PCBM was purchased from NanoC Inc. (Purity: 99.5%). A solution of PCBM was prepared by stirring PCBM powder in chlorobenzene and then the solution was set on a hot plate at about 70°C for 24 h. The film samples were prepared by spincoating the solution onto freshly cleaved HOPG substrate (SPI-1 grade, purchased from SPI supplies) or graphene on Cu foils synthesized by CVD. AFM measurements were carried out on a Dimension Icon (Bruker Corporation) instrument in a dark environment. Monolithic silicon cantilevers (NCST, NANO WORLD) with a spring constant of 7.4 N m −1 , first longitudinal resonance frequencies between 120 and 205 kHz, and nominal tip radius of 8 nm were employed in soft tapping mode. Simultaneous height and phase images were acquired and reproduced across multiple samples. STM characterizations were carried out in an ultrahigh vacuum scanning tunneling microscope system (Omicron STM) with a base pressure of low 10 −9 Torr. The STM tip was a chemically etched tungsten tip.

PCBM bilayer morphology
We first investigated the self-assembled structure of PCBM deposited on a graphene/Cu substrate. The main facet of Cu underneath monolayer graphene is (111) oriented, which was determined by typical Moiré patterns of graphene (inset of figure 1(a)). Typical AFM topography images of the PCBM bilayer on graphene (figures 1(b)-(d)), showed a random distribution of PCBM islands, similar to previous SEM results [37], in contrast to highly ordered hcp or 'double row' structures such as have been previously reported [38,39].
These islands are identified as PCBM bilayers with a measured height of ∼1.37 nm (figures 1(d) and (e), blue lines), which is close to double the diameter of PCBM molecules (∼0.7 nm). The height of this domain differs significantly from the height of chlorobenzene residue and a height of PCBM monolayer [38], and so we can exclude the possibility that the connected islands are due to solvent or PCBM monolayers.
In order to characterize the large-scale morphology of such a PCBM domain, we also deposited PCBM bilayer nanostructures on a HOPG substrate, which provides much larger atomically flat terraces in comparison with the flat facets in the case of the graphene/Cu substrate. The lowest coverage (∼0.15) bilayer films on HOPG were spin-coated from 0.1 mg ml −1 PCBM in chlorobenzene solution (figure 1(h)). Atomically resolved STM images of the area outside the film domain verified that it was bare HOPG substrate (insert of figure 1(h)). The PCBM bilayer film showed continuity when crossing step edges on the HOPG substrate (figure 1(h), white double arrows). The overall differences between figures 1(b) (long stripes) and 1(f), (h), and (i) (irregular islands) are due to the differences between the underlying graphene/Cu and HOPG substrates. Spin-coating from higher concentrations of PCBM resulted in higher coverage of the substrate by PCBM bilayers, but the morphology of the bilayer remained as irregular networks with randomly distributed holes. Figure 1(k) is a plot of the coverage of PCBM on HOPG versus the concentration of the PCBM/Chlorobenzene solution.
Strikingly, we found two typical heights in the PCBM bilayer deposited on the HOPG substrate: 1.64±0.09 nm (Type I, blue line in figure 1(f)), and 1.23±0.03 nm (Type II, red lines in figures 1(f) and (i)). In some films, one type of PCBM bilayer dominated the sample (figure 1(i)), while in others the two types coexisted in a sample (figure 1(f)). Our observations of PCBM monolayers, which will be discussed in the following section, showed a very different height, and hence we exclude the possibility that the Type II structure originates from a monolayer. Combining the above observations together, we conjecture that the height difference reflects two distinct types of the PCBM dimer-HOPG substrate interaction with different arrangement angles leading to different heights.
Note that these novel PCBM bilayers were observed on graphene and HOPG surfaces rather than monolayers with hexagonal close-packed structures [28,38] or double row structures on Au surfaces [38,39]. The typical height of a PCBM monolayer is about 0.7 nm, and the height of a PCBM bilayer is 1.64 nm. It is most likely that the observed PCBM bilayer structure consists of PCBM dimers standing up on the graphene surface with some tilt angle due to the weak interaction between PCBM dimers and the graphene surface. The tilt angle between PCBM dimers and the graphene surface is about 40.6°. In a previous theoretical study, Bredas' group found a similar structure where pentacene molecules on a gold surface are tilted with an angle of about 37.7° [40].

PCBM monolayer morphology
Next, we examine self-assembled PCBM monolayers deposited on graphene/Cu and HOPG. On graphene, a monolayer was formed after a post annealing for 30 min at 170°C (figures 2(a) and (b)). On HOPG, the monolayer sample is more favorable as the stabilization time of the solution increases. The overall morphology resembles the irregular network observed in the bilayer structure (figures 2(a) and (b)). The figure indicates the morphology of a PCBM monolayer on the graphene/Cu substrate, while the inserted line profile shows the height is about 0.87 nm, which is in the range of monolayer height. In order to further examine monolayers, additional PCBM monolayer samples were studied by spin coating 0.5 mg ml −1 PCBM/Chlorobenzene solution on HOPG under the same deposition conditions as the bilayer samples. The height of a monolayer has two typical values on the HOPG substrate: one is around 0.71 nm (figures 2(c) and (d)); the other one is about 0.88 nm (figures 2(e) and (f)). These two values are relatively close to each other, and both of them are comparable to those reported in previous literature [38,39]. Thus, it is suggested that both of these two heights originate from PCBM monolayers with different orientations. Compared to the mixed structures that can be observed in the bilayers, these two typical heights always appear in different samples, which may be due to a higher energy barrier (1 kcal mol −1 -about 200°C thermal energy) between the two PCBM monolayer orientations [38]. Note that our PCBM monolayer sample is filled with randomly shaped monolayer terraces, in contrast to elbow nucleating structure found previously [39].

Discussion
To understand the observed bilayer and monolayer structures, we construct molecular models based on van der Waals interactions between neighboring fullerene moieties, hydrogen-bonding between tail functional groups, and the interactions between molecules and substrates. We first discuss the model for the monolayer structure. We propose that the two typical monolayer heights of ∼0.9 and ∼0.7 nm observed in the PCBM monolayers reflect different orientations of the PCBM relative to the substrate. The thicker monolayer (∼0.9 nm) corresponds to a vertical configuration of PCBM, with its tail perpendicular to the HOPG surface ( figure 3(a)), whereas the thinner monolayer (∼0.7 nm) corresponds to a horizontal configuration, with PCBM tails parallel to the HOPG surface and interacting in pairs ( figure 3(b)). The monolayer height of 0.7 nm is consistent with the previous measurement of the size of C 60 by Robey's group [38], while the monolayer height of 0.9 nm has never been reported. Now we turn to the small differences between the modeled and measured heights of a PCBM monolayer in both of the proposed configurations. These differences arise from the fact that the tapping mode AFM tends to underestimate the height of sample surface features, and that the underestimate amount depends on the stiffness of the measured areas [41,42]. For the horizontal configuration, the actual monolayer height should be ∼1 nm, because the equilibrium van der Waals gap between C 60 and the HOPG substrate is calculated to be 0.25-0.3 nm wide (figure 3) [38]. But due to pressure exerted by the AFM tis, the monolayer height was measured at ∼0.7 nm. In fact, this measured value is comparable to a previous AFM tapping mode image of a C 60 shuttlecocks monolayer on HOPG [41]. For the vertical configuration, the monolayer height that the model suggests is 1.45 nm, marking an even larger difference from the measured height of 0.9 nm. This large difference is likely due to softness of the monolayer caused by the PCBM tail groupsthese tails are likely much softer than the stiff C 60 cages in the horizontal configuration.
Our analysis also suggests that a higher coverage of the monolayer favors the vertical configuration. This is reasonable because this configuration can accommodate a much higher area density (1.15 molecule nm −2 ) of PCBM than that of the horizontal configuration (0.73 molecule nm −2 ). Using the energy values of neighboring fullerene moieties, the hydrogen-bonding, and the interactions between molecules and substrates from previous experimental and theoretical investigations [38,39,43], we calculated the formation energy of these two configurations and the results show an energy barrier of about 1 kcal mol −1 exists between the two configurations, which inhibits thermally-induced transition between the configurations at room temperature.
We now discuss the model for the PCBM bilayer. In most cases, we observed PCBM bilayers rather than monolayers on graphene and HOPG. As described in section 3.1, there are two typical heights in PCBM bilayers on graphene and HOPG: 1.64 nm (modeled in figure 4(a)), and 1.23 nm (modeled in figure 4(b)). It was previously reported that on a gold substrate, PCBM dimers are formed with a twin chain structure in low density, but with a double row structure in high density [38,39,44], because the affinity between C 60 cages (0.28 eV) [44][45][46] is higher than the hydrogen bonding between the PCBM tail groups (0.114 eV) [44]. However, in this work, PCBM molecules were deposited by spin coating rather than physical vapor deposition. It is known that steric hindrance of sidechain-substituted PCBM molecules would forbid a possibility of a C 60 -to-C 60 coordinated structure in the region defined by their first solvation shells [47]. Considering hydrogen bonding and dipole-dipole interactions, the energy of a side-to-side dimer (0.114 eV) is higher than side-to-C 60 dimer (0.001 eV) [44]. Hence, we suggest that the PCBM molecules form side-to-side dimers in the chlorobenzene solvent before the spin coating. After the spin coating, the side-to-side PCBM dimers are distributed on the HOPG, forming a PCBM bilayers (type I and type II in figure 4). For the thicker bilayer, we propose that the PCBM dimers form a 'double hcp' structure: one C 60 cell of the dimer lies in the lower layer, the other in the upper layer, such that neighboring PCBM dimers interact sideways (Type I, figure 4(a)). For the thinner bilayer, the PCBM monomers in each dimer are also located in the lower and upper layers (Type II, figure 4(b)), but without strong sideways interactions between PCBM dimers. The two typical PCBM bilayer heights indicates the different tilt angles between a PCBM dimer and the substrate in the two configurations. A tilt angle of 49.4°corresponds to the 1.64 nm bilayer, while a tilt angle of 60.8°the 1.23 nm bilayer, where the tilt angle is defined to be an angle between an axis connecting the centers of two C 60 in a given dimer and the direction normal to the substrate. Similar to the PCBM monolayer case, the measured heights of a PCBM bilayer are ∼0.3 nm smaller than those predicted by the model in both configurations. These differences can be similarly explained by the compression caused by the AFM tips.
We further estimate the energy per unit area for the different monolayer and bilayer configurations in order to deduce their relative stability. For Type I bilayer, the binding energy of one upper layer cell of PCBM is 2.52 eV, which arises from the sum of the binding energy of the 9 nearest C 60 molecules with 0.28 eV each; for the lower layer PCBM cells, the binding energies are 3.5 eV each, which arises from the sum of the binding energy of the 9 nearest C 60 molecules and of the binding energy between C 60 and the HOPG (0.98 eV) [43,48]. Combining the above information with hydrogen binding energies of the neighboring tail groups, we find that the total energy for each Type I PCBM dimer is 6.13 eV. For Type II bilayer, the upper-layer PCBM cell only has a binding energy of 1.12 eV (4 nearest C 60 ), while the lower-layer PCBM cell on HOPG has a binding energy of 2.1 eV, considering the binding energy of one C 60 and HOPG and the binding energy of the nearest C 60 molecules. The binding energy of each PCBM dimer is 3.33 eV. Combining the above information with the hydrogen bonding of the tail groups, we find that for the horizontal configuration the binding energy one PCBM dimer is 4.31 eV. Considering the dimer concentration density of 1.15 dimer nm −2 , we find that the energy density of Type I PCBM bilayer is 7.05 eV nm −2 , which is higher than that of Type II bilayer by 1.35 eV nm −2 and that of the horizontal monolayer by 1.57 eV nm −2 . As concentration density increases, the PCBM dimers are compressed to form more vertical configurations in order to accommodate more PCBM dimers on the HOPG surface.

Thermal effects
We now investigate annealing effects on the morphology of the PCBM bilayer nanostructures. In order to quantify the large-scale morphology changes of PCBM domains, we focused on the PCBM bilayer nanostructures on HOPG substrates, which provide large atomically flat terraces. For this experiment, we annealed the samples at 140°C and 160°C for 10 min and measured the samples by AFM immediately after annealing. Figure 5(a) shows the topography before annealing and figures 5(b) and (c) are the PCBM bilayer after 140°C and 160°C annealing, respectively. To quantify the morphology changes, we performed size distribution analysis for the holes (i.e., bare HOPG area without PCBM) as shown in figure 5(d). by using the standard nanoparticle size distribution analysis method [49]. For data analysis, we use the method introduced in 'on optimal and data-based histograms' [50]. We find a right shift of the distribution peak after annealing at 140°C and 160°C. The peaks are located at 1490 nm 2 before annealing, 1884 nm 2 for annealing at 140°C, and 2291 nm 2 for annealing at 160°C, respectively. The result indicates that the smaller holes are merged together.

Summary
In summary, we demonstrated the self-assembly of PCBM bilayer nanostructures on graphene and HOPG, by using AFM and STM, and analyzed the observed morphology by comparison to molecular models. The PCBM bilayer revealed two distinct configurations on HOPG with different heights, and only one configuration on graphene. Post thermal annealing can induce merging of the bilayer nanostructures. Our results will shed light on improvement of the energy efficiency in solar cells containing graphene and organic molecules, by increasing the donor-acceptor interface area. Area distribution histogram of holes (without PCBM area) obtained from measurements of the area of holes in AFM images of before (green) and after annealing at 140°C (dark red) and 160°C (dark blue).