Charge‐Transfer‐Controlled Growth of Organic Semiconductor Crystals on Graphene

Abstract

Controlling the growth behavior of organic semiconductors (OSCs) is essential because it determines their optoelectronic properties. In order to accomplish this, graphene templates with electronic‐state tunability are used to affect the growth of OSCs by controlling the van der Waals interaction between OSC ad‐molecules and graphene. However, in many graphene‐molecule systems, the charge transfer between an ad‐molecule and a graphene template causes another important interaction. This charge‐transfer‐induced interaction is never considered in the growth scheme of OSCs. Here, the effects of charge transfer on the formation of graphene–OSC heterostructures are investigated, using fullerene (C₆₀) as a model compound. By in situ electrical doping of a graphene template to suppress the charge transfer between C₆₀ ad‐molecules and graphene, the layer‐by‐layer growth of a C₆₀ film on graphene can be achieved. Under this condition, the graphene–C₆₀ interface is free of Fermi‐level pinning; thus, barristors fabricated on the graphene–C₆₀ interface show a nearly ideal Schottky–Mott limit with efficient modulation of the charge‐injection barrier. Moreover, the optimized C₆₀ film exhibits a high field‐effect electron mobility of 2.5 cm² V⁻¹ s⁻¹. These results provide an efficient route to engineering highly efficient optoelectronic graphene–OSC hybrid material applications.

The growth behavior of fullerene (C₆₀) thin films on graphene templates where charge transfer occurs is presented. The number of electrons transferred from graphene to C₆₀ is controlled by in situ electrical gating of graphene during C₆₀ deposition. When this electron transfer is suppressed, high‐crystalline C₆₀ thin films are achieved for highly efficient optoelectronic applications.

1. Introduction

Graphene has excellent properties, so the possibility of integrating it with both inorganic and organic semiconductors has been intensively studied. Graphene–semiconductor heterostructures provide multifunctionality and desirable properties for scalable and flexible optoelectronic applications.1, 2 The ideally sp²‐hybridized carbon atoms of graphene constitute a basal plane with no dangling bonds, so it provides an atomically clean interface with a semiconductor; this contact is extraordinary and cannot be achieved with traditional interfaces. With the introduction of these unique graphene–semiconductor interfaces, researchers have proposed various graphene–semiconductor hybrid optoelectronic devices such as field‐effect transistors (FETs), light‐emitting diodes, solar cells, photodetectors, and barristors.3, 4, 5

Graphene is inert and is composed of a single‐atom‐thick layer, so it is a useful growth template for semiconductors, especially organic semiconductors (OSCs).6, 7 The assembly of OSC thin films on graphene is mainly determined by the interactions between OSC ad‐molecules and the graphene template (e.g., van der Waals). Therefore, the graphene template can enable epitaxial growth of highly crystalline OSC thin films.8 In addition, these interactions can easily be tuned by controlling the electronic properties of graphene,9, 10 so graphene templates offer a facile and direct approach to prepare graphene–OSC heterostructures with desirable interfacial properties. However, despite the great potential of graphene–OSC heterostructures, only a few studies of OSCs' growth behavior on electronic‐states‐controlled graphene have been reported.7, 10 Therefore, to develop a reliable method to optimize the growth of OSCs on graphene templates, the complex of OSC molecules and graphene templates and possible interactions between them should be investigated.

Here, we demonstrate that an epitaxial growth of a vacuum‐deposited fullerene (C₆₀) thin film on a graphene template can be controlled by tuning charge transfer between them. The Fermi level (E _F) of the graphene template determines the amount of charge transfer between the graphene and the C₆₀ ad‐molecules, and this amount in turn affects the molecular dynamics of C₆₀ on the graphene template. By finely tuning the E _F of the graphene template, we induced layer‐by‐layer growth of highly ordered C₆₀ films on graphene. Considering that the thin film's topological and crystalline features determine the optoelectronic properties of OSCs,11 this approach advances the efficiency of organic electronic devices. The C₆₀ films grown under optimized conditions exhibited a maximum field‐effect mobility of 2.5 cm² V⁻¹ s⁻¹. Furthermore, a graphene–C₆₀ Schottky junction prepared by our method approached the Schottky–Mott limit, which is desirable for highly efficient graphene–OSC barristors and other vertical graphene–OSC hybrid optoelectronic devices.

2. Results and Discussion

2.1. Charge Transfer between Graphene and C₆₀

We first investigated the transfer of electrons from graphene to C₆₀. Analyses using ultraviolet photoelectron spectroscopy, Kelvin probe force microscopy, and Raman spectroscopy revealed that the adsorption of C₆₀ molecules induced p‐type doping of graphene (Figure S2, Supporting Information). To clarify the relationship between charge transfer and the initial electronic states of graphene, we fabricated graphene field‐effect transistors (G‐FETs) on 300 nm thick SiO₂/Si substrates and compared the transfer characteristics of the G‐FETs before and after 3 s of C₆₀ deposition at a deposition rate of 5 × 10⁻² monolayer per second (ML s⁻¹) (Figure 1 a). To eliminate the contact resistance, we used transfer‐length‐method measurements so that the change in graphene channel resistance (R _Ch) could be solely attributed to the change in charge‐carrier density (n _g, n _g > 0 for electrons and n _g < 0 for holes).

Figure 1.

Charge transfer between graphene and C₆₀. a) Schematic diagram showing G‐FET with deposited C₆₀. b) Transfer characteristic of G‐FET before (green open circle) and after C₆₀ deposition (blue closed circle). Solid lines are model fits. c) Concentration of transferred charge carrier after C₆₀ deposition Δn _CT versus initial charge carrier concentration of bare graphene n _g,bare. d) Energy band diagrams of graphene/C₆₀ when n _g,bare < n _c (left), n _g,bare = n _c (middle), and n _g,bare > n _c (right).

After C₆₀ deposition, the R _Ch was preserved as long as the gate voltage (V _G) was <−40 V. This preservation demonstrates that deposition of C₆₀ did not cause degradation of graphene, and more importantly, that no charge transfer occurred between graphene and C₆₀ in this range of V _G. However, at V _G > −40 V, the R _Ch –V _G curve shifted to the right; this change indicates that electrons were transferred from graphene to C₆₀ (Figure 1b). This shift of R _Ch –V _G curves when the magnitude of V _G is larger than a certain value was consistently observed with other samples from different batches (Figure S3, Supporting Information).

To calculate the number of transferred electrons (Δn _CT (cm⁻²)) at a certain V _G, the R _Ch –V _G curve was fitted using the constant‐mobility model.12 Then the carrier density of bare graphene before C₆₀ deposition (n _g,bare) and the carrier density of graphene–C₆₀ after C₆₀ deposition ($n_{\text{g},{\text{C}}_{\text{60}}}$) were each calculated at each V _G as

$$n_{\text{g}}=\text{sgn}\left(V_{\text{G}}-V_{\text{D}}\right)\sqrt{{\left(\frac{\text{1}}{\mu e{R}_{\text{Ch}}}\frac{L}{W}\right)}^{\text{2}}-n_{\text{res}}^2}$$ (1)

where V _D is V _G at maximum R _Ch, μ is the carrier mobility, e is the elementary charge, L is the channel length, W is the channel width, and n _res is the residual carrier concentration in graphene. Then Δn _CT was calculated as $n_{\text{g},\text{bare}}-n_{\text{g},{\text{C}}_{\text{60}}}$. Before C₆₀ deposition, the fitted values of μ and n _res of the graphene transistor were 4470 cm² V⁻¹ s⁻¹ and 2.3 × 10¹² cm⁻², respectively. When plotted versus n _g,bare (Figure 1c), extracted Δn _CT showed no charge transfer between graphene and C₆₀ when n _g,bare was less than a critical value, n _c = −4.4 × 10¹² cm⁻². As n _g,bare approached n _c, charge transfer started and gradually increased with increasing n _g,bare. The V _G‐dependent contact resistance in G‐FETs also supports our claim that the charge transfer occurred when n _g,bare > n _c (Figure S4, Supporting Information).

The observed n _g,bare‐dependent charge transfer between graphene and C₆₀ is explained as follows. The electrons in graphene are transferred to C₆₀ when the E _F of graphene is higher than the lowest unoccupied molecular orbital (LUMO) level of adjacent C₆₀. The LUMO level of isolated C₆₀ molecules is known to be −4.5 eV,13 which is similar to the E _F of undoped graphene. However, the energy levels of organic molecules change and broaden upon adsorption of C₆₀, because of the polarizability of the substrate;14, 15 thus, the LUMO level of C₆₀ adsorbates can lie below the E _F of undoped graphene that has n _g,bare > n _c; as a result, the graphene becomes p‐type doped. The absence of charge transfer when n _g,bare < n _c is attributed to the E _F of graphene being lower than the LUMO level of the C₆₀ adsorbates (Figure 1d, left). As the E _F of graphene is raised by external gating such that it reaches the LUMO level of C₆₀, electrons are transferred from graphene to C₆₀, and the E _F of graphene is pinned to the LUMO level of C₆₀. As a result of this charge transfer, an electric field is generated between the graphene and the C₆₀, so the vacuum level at the interface becomes tilted so that the E _F of the graphene and the LUMO level of the C₆₀ are aligned (Figure 1d, right).

First, the number of charges is conserved at the graphene–C₆₀ interface as

$$\frac{C_{\text{g}}}{e}\left(V_{\text{G}}-V_{\text{D},\text{bare}}\right)=n_{\text{g},\text{bare}}=n_{\text{g},{\text{C}}_{60}}+\frac{{\sigma }_{{\text{C}}_{60}}}{e}$$ (2)

where C _g is the dielectric capacitance, V _D,bare is the V _D of the G‐FET before C₆₀ deposition, and ${\sigma }_{{\text{C}}_{\text{60}}}$ is the surface charge density in a C₆₀ film. The charge redistribution at the graphene–C₆₀ interface as a function of n _g,bare can be estimated by solving

$$\text{sgn}\left(n_{\text{g},{\text{C}}_{60}}\right)\hslash v_{\text{F}}\sqrt{\pi |n_{\text{g},{\text{C}}_{60}}|}-\text{sgn}\left(n_{\text{c}}\right)\hslash v_{\text{F}}\sqrt{\pi |n_{\text{c}}|}=e\frac{{\sigma }_{{\text{C}}_{60}}}{{\epsilon }_0}d$$ (3)

where ℏ is the reduced Planck's constant, v _F is the Fermi velocity of graphene, ε₀ is the vacuum permittivity, and d is a fitting parameter that describes the spacing between graphene and C₆₀. The left‐hand side of Equation (3) is E _F − E _F,c (Figure 1d), in which E _F,c is the critical Fermi level where the charge transfer between graphene and C₆₀ occurs. The right‐hand side is the charge‐transfer‐induced shift of the vacuum level at the interface.

The R _Ch–V _G curves of G‐FETs and the Δn _CT as a function of n _g,bare were modelled using calculated $n_{\text{g},{\text{C}}_{\text{60}}}$ and d. The models successfully replicated the experimental values (Figure 1b,c). Moreover, the charge transfer modifies the density of states of C₆₀ so that the LUMO level of charged C₆₀ molecules is split into an “occupied” LUMO level (L1) that is shifted downward and an unoccupied LUMO level (L2) that is shifted upward (Figure 1d, right).16 This downshift of the LUMO level upon charge transfer can substantially stabilize C₆₀ adsorbates on graphene.17

2.2. Growth of C₆₀ Thin Films on Graphene under Charge Transfer

With the in situ electrical gating of graphene (“Experimental and Methods” in the Supporting Information), we observed changes in i) the molecular interactions and assembly of C₆₀ ad‐molecules and ii) the growth behavior of C₆₀ crystals as the E _F of graphene gradually approached the E _F,c.

First, C₆₀ ad‐molecules may interact with each other on the graphene surface, depending on the relative position of the E _F of graphene and the E _F,c. These distinctions can be well detected by Raman spectroscopy (Figure S5, Supporting Information). In both Raman spectra, the feature peaks of C₆₀, i.e., the A _g(1) mode at ≈500 cm⁻¹ and the A _g(2) mode at ≈1470 cm⁻¹, were clearly observed.18 The position of the A _g(2) peak indicates the number of intermolecular bonds to each C₆₀ molecule, where each intermolecular bond shifts the peak by −5 cm⁻¹.19 The peak position of the A _g(2) mode of C₆₀ grown on graphene with E _F < E _F,c is consistent with that reported for pristine C₆₀ molecules.18, 19 However, the A _g(2) peak of C₆₀ grown on graphene with E _F > E _F,c was red‐shifted ≈3 cm⁻¹; this change indicates that chemically bonded C₆₀ dimers or oligomers were formed. This selective formation at high E _F strongly suggests that control of the E _F of graphene during C₆₀ growth indeed determined the charge state of the C₆₀ ad‐molecules.

The charge state of C₆₀ ad‐molecules determines the formation of covalent bonds between two C₆₀ molecules.20, 21 When C₆₀ molecules have negative charges, the activation barrier for the bonding decreases. Therefore, graphene with E _F > E _F,c induced negative charges in C₆₀ ad‐molecules, resulting in the formation of intermolecular bonds between C₆₀ ad‐molecules. By contrast, on graphene with E _F < E _F,c, C₆₀ molecules were charge‐neutral and thus did not form covalently bonded C₆₀ dimers.

Δn _CT affected molecular arrangement in C₆₀ crystals, and consequently, how those crystals assembled into thin films. We used grazing incidence X‐ray diffraction (GIXD) to characterize C₆₀ thin films with different thicknesses grown on graphene, where Δn _CT was controlled. Under ambient conditions, the most stable structure of C₆₀ crystals is face‐centered cubic (fcc);22 the diffraction patterns of the fcc C₆₀ were observed in our system of C₆₀ thin films grown on graphene (Figure 2 a).

Figure 2.

Crystal structure of C₆₀ films grown on graphene. a) 2D GIXD patterns of 2.5 ML (2 nm) and thick (100 nm) C₆₀ films grown on graphene when Δn _CT = 0 cm⁻² (left), Δn _CT = 5 × 10¹¹ cm⁻² (middle), and Δn _CT = 1.3 × 10¹² cm⁻² (right) during C₆₀ deposition. b) Cross‐sectional profiles of the 2D GIXD image along the q_z for various Δn _CT. c) The mean size of the crystalline (111) domains R ₍₁₁₁₎ versus Δn _CT. d) Schematic illustrations of C₆₀ crystal growth on graphene without (upper) and with (lower) the charge transfer between them. Insets: Low‐magnification HR‐TEM images of corresponding graphene–C₆₀ samples on TEM grids. Scale bar in insets: 200 nm.

At the early growth stage (nominal thickness of 2.5 ML), irrespective of the occurrence of charge transfer, the set of reflections of (111) family and the reflections of plane (113) and plane (220) appeared; these reflections are located along the out‐of‐plane direction (q_z) and at 30° and 35° tilt from q_z, respectively. These results indicate that C₆₀ has an epitaxial relationship with graphene, with the (111) plane of C₆₀ crystals parallel to the graphene substrate;23 this epitaxy was independent of Δn _CT. However, differences were observed in the crystal domain sizes of C₆₀ thin films grown on graphene at different Δn _CT (Figure 2c). We quantified the average crystal domain size of C₆₀ thin films by using the Scherrer equation to estimate the domain sizes of crystal plane (111) (R ₍₁₁₁₎). When Δn _CT = 0 during C₆₀ growth, C₆₀ thin films had R ₍₁₁₁₎ ≈ 60 nm, which is almost three times larger than in the film grown under very high Δn _CT.

At the final growth stage, the GIXD patterns of C₆₀ films grown with and without charge transfer both showed clear ring patterns, which reveal the presence of randomly oriented C₆₀ crystals. However, the thick C₆₀ films' ordering degree was still strongly dependent on Δn _CT. On the graphene surface where Δn _CT = 0, the reflections were still sharp with a high signal‐to‐noise ratio, i.e., most of the C₆₀ crystals were oriented. As Δn _CT increased, these reflections weakened and eventually became undetectable; this change suggests that a large fraction of newly nucleated C₆₀ crystals were randomly oriented on the pre‐existing C₆₀ thin film. The growth behavior of C₆₀ crystals on graphene, as indicated by GIXD experiments, is summarized as follows (Figure 2d). A highly crystalline film of fcc C₆₀ was epitaxially formed on graphene via a layer‐by‐layer growth mode at negligible Δn _CT during C₆₀ growth. When Δn _CT > 0, despite the epitaxial relationship between graphene and C₆₀ at the early growth stage, randomly oriented nucleation occurred during vertical growth. These inferences are confirmed by low‐magnification high‐resolution transmission electron microscopy (HR‐TEM) images (insets of Figure 2d). At Δn _CT = 0, large‐area C₆₀ layers were observed; by contrast, at very high Δn _CT, small C₆₀ clusters formed. Although the GIXD results provided a hint about the crystal structure of the C₆₀ films grown on graphene over a macro area, they could not directly reveal the arrangement among C₆₀ molecules and the carbon atoms in graphene.

Therefore, C₆₀ thin films (2.5 ML) grown on graphene were imaged at high magnification using HR‐TEM. The image of C₆₀ grown on graphene at Δn _CT = 0 clearly showed an ordered hexagonal arrangement of C₆₀ molecules over a few tens of nanometers, which is the fashion of the (111) plane of a highly crystalline fcc structure (Figure 3 a, top). Moreover, the ordering in this HR‐TEM image matches that of ABA‐stacked C₆₀ layers.24 This stacking order was uniform over the analyzed areas; this consistent order implies that C₆₀ layers were preferentially stacked on each other in an ABA manner when the thin film was grown on graphene at Δn _CT = 0. The corresponding selected‐area electron diffraction (SAED) pattern of this C₆₀ thin film also showed only a single set of hexagonal patterns, i.e., the crystalline orientation of C₆₀ was uniform along the vertical direction. Notably, when Δn _CT = 0 was maintained during C₆₀ growth, the misorientation angles between the SAED patterns of C₆₀ and those of graphene were concentrated at close to 0° and 30°, which correspond to energetically stable adsorption sites of C₆₀ molecules along the armchair and zigzag directions of graphene, respectively (Figure S7f, Supporting Information).23 This result is further evidence of an epitaxial relationship between graphene and C₆₀.

Figure 3.

Epitaxial molecular arrangement of C₆₀ on graphene. a) Typical HR‐TEM images and SAED patterns of 2.5 ML C₆₀ grown on graphene when Δn _CT = 0 (top) and Δn _CT >> 0 (bottom). Scale bars in HR‐TEM images: 3 nm; in SAED patterns: 1 nm⁻¹. Insets: High‐magnification HR‐TEM images of regions with ABA (in (top)) and ABC (in (bottom)) stacking. b) Histogram plots of nearest neighbor C₆₀–C₆₀ molecule distances extracted from HR‐TEM images when the growth associated without (left) and with (right) charge transfer. c) DFT energetic simulations of C₆₀–C₆₀ double‐bonded dimer (left) and isolated C₆₀ molecules (right).

By contrast, when C₆₀ was grown on graphene under a very high Δn _CT, HR‐TEM image and the corresponding SAED patterns (Figure 3a, bottom) typically revealed polycrystalline C₆₀ thin film along the lateral direction and vertical direction. This C₆₀ film showed ABA and ABC stacking mixed within small areas. In addition, small crystalline domains were tilted from the rest with a high angle (≈30°) in this film (Figure 3a, bottom left). Notably, the areas between the tilt grains mostly exhibited an amorphous structure. On top of this amorphous region, C₆₀ molecules could not arrange well, so the results were i) randomly oriented nucleation of C₆₀ crystals and ii) the formation of additional amorphous layers, or both. The resulting richness of tilt grain boundaries could result in the observed polycrystallinity along both the lateral and vertical directions. The dominance of (111)‐plane‐oriented C₆₀ crystal domains (Figure 2a) suggests the presence of an epitaxial relationship between C₆₀ and graphene at this small thickness, so grains that have high tilt angle may be formed by stitching C₆₀ domains aligned along the armchair direction and those aligned along the zigzag direction of graphene.

HR‐TEM was also the best tool to investigate the chemically bonded dimers in C₆₀ films (Figure S5, Supporting Information). To quantize the dimer content, we analyzed numerous intermolecular distances of two nearest‐neighbor C₆₀ molecules in films grown at Δn _CT = 0 and Δn _CT > 0 (Figure 3b). In both cases, the distance distribution showed peaks centered near 0.86 and 0.95 nm, but the relative peak heights depended on Δn _CT. We could assign the 0.85 nm peak to double‐bonded C₆₀ dimers, and the 0.95 nm peak to isolated C₆₀ molecules.25 When Δn _CT = 0, more than half of the intermolecular distances were close to 0.95 nm; this consistent separation implies that a large portion of the C₆₀ molecules were still free and intact. However, at Δn _CT > 0 the fraction of free C₆₀ molecules was substantially reduced and the proportion of double‐bonded dimers increased. These results qualitatively show that charge transfer with graphene during C₆₀ growth promoted the formation of double‐bonded C₆₀ dimers.

In addition, we performed density functional theory (DFT) simulations to calculate the electronic structure of a double‐bonded C₆₀ dimer and two isolated C₆₀ molecules (Figure 3c; Figure S14, Supporting Information). Compared with isolated C₆₀ molecules, a double‐bonded C₆₀ dimer showed an ≈0.2 eV smaller bandgap, and broader LUMO and highest occupied molecular orbital (HOMO) levels.

The effects of charge transfer on C₆₀ growth behaviors are further demonstrated by morphological analysis using atomic force microscopy (AFM) (Figure 4 a), which enabled statistical analysis of average height h _i of C₆₀ islands and surface coverage θ of the thin films during the early growth stage (Figure 4b). On the surface of graphene templates on which charge transfer was suppressed, i.e., E _F < E _F,c, the initial large‐area C₆₀ islands expanded laterally, to yield a constant monolayer thickness (0.8 nm) and a large increase of surface coverage. As electron transfer from the graphene to C₆₀ ad‐molecules increased, the number of nuclei quickly increased and each of them merely grew in height; the result was an array of grains of different heights. At Δn _CT = 0, as the growth continued, continuous C₆₀ film was formed by coalescence of large‐area C₆₀ grains; by contrast, at Δn _CT > 0, C₆₀ film was formed by full coverage of small C₆₀ islands with poor inter‐grain connection. At the later growth stage (12.5 ML), the C₆₀ thin film grown at Δn _CT = 0 revealed clear terrace structure, which is evidence of lateral growth mode, whereas the film grown at Δn _CT > 0 simply showed an array of tiny crystallites.

Figure 4.

Nucleation of C₆₀ islands on graphene. a) AFM images of C₆₀ at different nominal thicknesses of 0.25 ML (left), 1.25 ML (middle), and 12.5 ML (right) grown on graphene, without (upper) and with (lower) the charge transfer. Scale bar: 400 nm. b) Height analysis for C₆₀ islands in the AFM images. Inset: Surface coverage analysis. c) Nucleation density N _i versus Δn _CT from gate‐bias (green circle) and polymer‐contact doping (blue square). d) N _i versus thermal parameter 1/(k _B T). e) Nucleation energy barrier of C₆₀ (E _Nuc) versus Δn _CT calculated from (d). Shaded areas are to guide the eye.

The charge transfer in the graphene–C₆₀ system as well as its effects on the crystal structure and morphology of C₆₀ (Figures 2, 3, 4) were elucidated using electrically gated graphene templates. The use of polymer–substrate‐doped graphene revealed similar results (Figures S1, S6, and S9, Supporting Information). This comparison emphasizes that other factors (e.g., localized traps, the wetting transparency, or contamination on the graphene surface) that might obscure the collected results might have been effectively eliminated.7 Moreover, this polymer–substrate doping method could provide a general understanding of the observed phenomena.

To quantify the dependence of C₆₀ growth on the charge transfer from the graphene template to C₆₀ ad‐molecules, numerous C₆₀ thin films with a nominal thickness of 0.25 ML were grown on graphene templates whose E _F was finely controlled by either gating or polymer–substrate doping. The plot of the nucleation density (N _i) of these films against Δn _CT at room temperature revealed correlations between the nucleation of C₆₀ and charge transfer from graphene to C₆₀ (Figure 4c; Figure S10, Supporting Information). Clearly, N _i increased as Δn _CT increased. We also directly measured the activation energy for C₆₀ nucleation (E _Nuc) as a function of Δn _CT, as N _i = C exp (E _Nuc/(k _B T)) where C is a pre‐exponential factor, k _B is the Boltzmann constant, and T is the substrate temperature.26 To this end, N _i values as a function of the substrate temperature T were collected at various fixed Δn _CT; the slopes of plots of ln(N _i) versus 1/(k _B T) at a certain Δn _CT gave the values of E _Nuc at the Δn _CT (Figure 4d). As a result, we confirmed that E _Nuc increased as Δn _CT increased (Figure 4e).

2.3. Atomistic Mechanism of C₆₀ Thin Film Growth on Graphene under Charge Transfer

The nucleation of C₆₀ on graphene involves several atomistic processes (Figure 5 a). After adsorbing to graphene, an C₆₀ ad‐molecule diffuses on the surface until the molecule forms a dimer with another ad‐molecule or attaches to a pre‐existing island (growth).27, 28 In general, E _Nuc is related to the activation energies of all of these atomistic processes. However, the energy barrier for C₆₀ diffusion is negligible on graphitic surfaces,24, 29 so nucleation and growth of C₆₀ on graphene are predominantly limited by the rate of attachment of ad‐molecules to pre‐existing islands.

Figure 5.

Mechanism of C₆₀ growth on graphene. a) Nucleation process of C₆₀ crystals on graphene surface that involves adsorption, diffusion, dimer formation, attachment, and direct impingement. b) Energy profiles of a C₆₀ ad‐molecule versus position near and on a C₆₀ island under the absence (solid line) and presence (dashed line) of the charge transfer between the ad‐molecule and graphene.

For such attachment‐limited nucleation with negligible barriers to diffusion and dimerization, E _Nuc = [2E _i + 2(i + 1)E _B]/(i + 3),27 where i is critical cluster size, E _i is cluster energy, and E _B is activation energy for the attachment.27 This equation implies that a nucleation density increases as E _B increases. This relation is explained as follows. The presence of high E _B hinders the attachment of deposited ad‐molecules to an island, so the concentration of ad‐molecules increases on the graphene surface. Thus, the probability of ad‐molecules colliding rapidly increases, and this change favors new nucleation rather than the growth of pre‐existing islands.

Therefore, the increases in N _i and E _Nuc with increasing Δn _CT are attributable to the increase in E _B with increasing Δn _CT as E _B (Δn _CT) = E _B0  + E′_B(Δn _CT) where E _B0 is the charge‐transfer‐independent attachment barrier and E′_B is the charge‐transfer‐dependent attachment barrier. When electrons in graphene are transferred to the ad‐molecules and the islands, the ad‐molecules and islands are negatively charged and the underlying graphene becomes positively charged (Figure 5b; Figure S11, Supporting Information). Consequently, repulsive Coulomb interaction occurs between the dipole from the ad‐molecule–graphene and that from the island–graphene. This long‐range repulsive interaction would introduce an additional attachment barrier E′_B. Assuming the long‐range repulsive interaction is simple electrostatic repulsive interaction, E′_B can be estimated as

$$E{\prime }_{\text{B}}^{}=Z_{\text{avg}}{e}^{2}d\mathrm{\Delta }{n}_{\text{CT}}\text{/}2{\epsilon }_0$$ (4)

where Z _avg is the average charge state of C₆₀ ad‐molecules (Equation (4) is derived in the Supporting Information). This model successfully predicts the increase in E _B with increasing Δn _CT. In this argument, we assumed that repulsive Coulomb interaction between an ad‐molecule and an island (and not that between two ad‐molecules on graphene) dominantly affects the nucleation kinetics. This assumption can be justified because the probability of collision between two C₆₀ molecules which both simultaneously have negative charges would be very small. On the contrary, a C₆₀ island contains many C₆₀ molecules, so a C₆₀ island is likely negatively charged.

The transition from a 2D to a 3D growth mode (Figures 2, 3, 4) under the charge transfer between graphene and C₆₀ can be simply explained by invoking the repulsive Coulomb interaction between an ad‐molecule and an existing island. With increasing Δn _CT, E _B increases because of the repulsive interaction; this change inhibits the lateral growth of negatively charged islands by negatively charged ad‐molecules diffusing on the graphene surface. However, irrespective of the E _F of the graphene template, the ad‐molecules from the vapor phase can land directly on the top of the existing island because they are charge‐neutral and thus not prone to the repulsive Coulomb interaction. However, after they are deposited on the top of the islands, their dynamics are again influenced by the E _F of graphene. When Δn _CT = 0, they can move relatively freely down to the graphene surface because the Ehrlich–Schwoebel barrier is much lower than the diffusion barrier on top of the C₆₀ layer.28 When electrons are transferred from graphene to C₆₀, the E _B increases and thus acts as an energy wall surrounding the edge of islands. For an ad‐molecules on the top of the island to move downward and escape from the island, they must overcome an activation energy greater than E _B. Consequently, ad‐molecules become concentrated on the top of the island, so the island rapidly grows in the vertical direction. The rapid vertical growth in turn leads to the formation of randomly oriented crystals.

2.4. Charge Transport in C₆₀ Thin Films and Graphene–C₆₀ Junctions

To quantify the advantage of our growth‐controlled C₆₀ thin films for lateral charge transport, we grew C₆₀ thin films on graphene at controlled charge‐transfer conditions, then transferred the C₆₀ films to octadecyltrichlorosilane (ODTS)‐treated SiO₂/Si substrates and then fabricated planar C₆₀ transistors (C₆₀‐FETs). The final device included an ≈50 nm thick C₆₀ channel without the underlying graphene (Figure 6 a). We measured the transfer characteristics of C₆₀‐FETs in the saturation regime with C₆₀ thin‐film channels grown at different Δn _CT, then estimated the associated electron field‐effect mobility (μ_e) and measured the on/off ratio (I _on/I _off). For a C₆₀ thin film grown at Δn _CT = 0, the I _on/I _off of the FET device was ≈10⁷ and the average μ_e was ≈1.5 cm² V⁻¹ s⁻¹. The maximum mobility of the device was ≈2.5 cm² V⁻¹ s⁻¹, which is similar to the state‐of‐the art mobility of C₆₀ transistors fabricated by the vapor deposition method (Figure 6b).5, 30 With increasing Δn _CT, the I _on/I _off and μ_e of the device substantially decreased, and eventually reached the same level of devices fabricated with polycrystalline and small‐grain C₆₀ (Figure 6c).31 The decay occurs because the high Δn _CT causes low crystallinity, low uniformity and limited grain size, and these traits suppress the lateral μ_e of C₆₀ thin films.

Figure 6.

C₆₀ field‐effect transistors and graphene–C₆₀ barristors. a) Schematic illustration of planar C₆₀‐FET. b) Transfer characteristic of C₆₀‐FET with C₆₀ film grown at Δn _CT = 0. c) Average I _on/I _off and electron mobilities μ_e of C₆₀‐FETs versus Δn _CT during C₆₀ growth. d) Schematic illustration of graphene–C₆₀ barristor. e) I _DS versus V _DS of graphene–C₆₀ barristors at various fixed V _G (from −100 to −40 V, step 10 V) for Δn _CT = 0 (left) and at V _G (from −100 to 100 V, step 10 V) for Δn _CT > 0 (right). Inset: I _DS versus V _DS at linear scale of graphene–C₆₀ barristor for Δn _CT = 0 at V _G = −40 V (filled symbols) and V _G = −30 V(open symbols). f) Temperature‐dependent saturation current of graphene–C₆₀ barristors at various V _G for Δn _CT = 0 (left, step 10 V) and Δn _CT > 0 cases (right, step 40 V). g) The Schottky barrier height (Φ_B) obtained from (f) versus ΔE _F.

Our method of growing C₆₀ thin films on graphene provides a direct way to produce controlled graphene–C₆₀ van der Waals heterostructures. In addition to its use as a growth template, graphene can function as an active layer or electrode for various flexible optoelectronic devices because of its excellent electrical conductivity and flexibility. Recently, heterostructures composed of graphene and OSCs have shown promise for use in organic photovoltaics, organic light‐emitting diodes, organic photodetectors, and vertical FETs.4, 5 The electrical characteristics of such devices depend on the charge‐injection efficiency at the graphene–OSC interface.

Such charge‐injection efficiency at the graphene–C₆₀ van der Waals heterointerface formed with our method was demonstrated by fabricating two types of graphene–C₆₀ barristors. They had the same device structure, but one had C₆₀ film grown at Δn _CT = 0, and one had C₆₀ film grown at Δn _CT = 1 × 10¹² cm⁻² (Figure 6d), so the C₆₀ layers had enormously different morphological and crystalline features. Both devices showed typical n‐type barristor behavior (Figure 6e).1 The closeness between the LUMO level of C₆₀ and Fermi level of aluminum yields Ohmic contact between the C₆₀ and the top aluminum electrode,32 so rectifications arose from the Schottky barrier (Φ_B) between the C₆₀ layer and the bottom graphene. Increasing the V _G barely affected the current in the forward regime (V _DS < 0) but boosted the current in the reverse regime (V _DS > 0). Consistent with the band diagram (Figure 1), the increase of V _G raised the Fermi level of graphene closer to the LUMO level of C₆₀, reducing the Φ_B accordingly, until alignment was achieved between them (Φ_B ≈ 0, Ohmic contact).

Although both barristors showed rectification behavior, great distinction was observed in the current levels between the two devices. The device that used the C₆₀ layer that had been grown at Δn _CT = 0 showed substantial modulation of the reversed current by the gate voltage; and the on‐state current I _on was higher in this device than in the device that used the C₆₀ layer that had been grown at Δn _CT > 0, whereas their off‐state currents I _off were similar. As a result, this device fabricated with a highly crystalline C₆₀ film (i.e., grown at Δn _CT = 0) achieved an I _on/I _off ratio of ≈10³ at V _DS = 2 V, which is nearly two orders of magnitude greater than the I _on/I _off ratio of the other device at the same V _DS (Figure 6e; Figure S12, Supporting Information).

The most important difference between the two types of barristors was the occurrence of a Schottky‐to‐Ohmic transition, which was only observed in the device that used the C₆₀ layer that had been grown at Δn _CT = 0 (Figure 6e, left). This transition occurred at V _G = −40 V, which is consistent with the critical voltage (V _G at n _c) required to induce charge transfer between graphene and C₆₀ (Figure 1). By contrast, Schottky‐to‐Ohmic transition was not observed within the wider examining V _G range for the device with the C₆₀ layer grown at Δn _CT > 0 (Figure 6e, right); this absence implies that modulation of the E _F of graphene by electrical gating was limited at the graphene–C₆₀ interface.

Fermi‐level pinning can occur when there are interfacial states in the HOMO–LUMO gap of C₆₀ layers near graphene.33 To quantitatively analyze the Fermi‐level pinning, we used the diode equation in the reverse bias saturation regime, $I_{\text{DS}}\propto T^2\exp \left(-\frac{e{\mathrm{\Phi }}_{\text{B}}}{k_{\text{B}}T}\right)$.1 The value of Φ_B at each V _G was then estimated from the plot of ln(I _DS/T ²) versus 1/(k _B T) (Figure 6f). Φ_B increased with increasing E _F at different rates in the two device types (Figure 6g). For the barristor with the C₆₀ layer grown at Δn _CT = 0, the slope S = dΦ_B/dE _F was ≈0.9, which indicates that the graphene–C₆₀ junction in this device approached the Schottky–Mott limit.14, 34 This result demonstrates an atomically clean interface between graphene and the C₆₀ thin film, which has not been previously achieved.23, 35 To achieve this clean heterointerface for the effective tuning of the Schottky barrier, C₆₀ must be deposited directly on a thermally cleaned graphene surface,36 and the electronic state of graphene must be optimized to effectively limit the charge transfer during growth to enable growth of high‐crystallinity C₆₀ film at the interface with graphene. The latter effect of charge transfer during the growth of OSCs has been neglected previously.

For the other device, S was only 0.1, which is indicative of strong Fermi‐level pinning effect at the graphene–C₆₀ interface. The C₆₀ thin film grown on graphene with charge transfer had small and poorly connected C₆₀ grains near the graphene surface (Figures 2, 3, 4), so the interface had i) a high density of C₆₀ grain boundaries, ii) large amorphous areas, and iii) other crystalline defects that would introduce numerous interfacial trap states (Figure S13c, Supporting Information). The Fermi level of graphene was pinned at those states. In addition, because the DFT results reveal a smaller bandgap of a C₆₀ dimer compared with two isolated C₆₀ molecules (Figure 3c), the presence of C₆₀ dimers would introduce shallow charge traps, which can further contribute to the observed Fermi‐level pinning at the graphene–C₆₀ interface.

To directly confirm the interfacial states between the graphene and the C₆₀, photocurrents of G‐FETs fabricated with deposited C₆₀ thin films (20 nm) were measured under light illumination at 0.62 eV (Figure S13a, Supporting Information). For comparison, C₆₀ thin films were grown on top of graphene channels under Δn _CT = 0 and Δn _CT > 0. At a high positive gate bias (V _G = 80 V), only the G‐FET with the C₆₀ thin film grown at Δn _CT > 0 showed additional photocurrent as the device was illuminated (Figure S13b, Supporting Information). The excitation energy is much smaller than the bandgap of the C₆₀ thin film and smaller than 2|E _F| of graphene at V _G = 80 V, so the interband transitions are forbidden in both the C₆₀ thin film and the graphene.37 Therefore, the photocurrent in the G‐FET with a C₆₀ thin film was merely a result of detrapped electrons from the interfacial states, which were abundant in the layer grown at high Δn _CT. In fact, we observed positive photocurrent from the G‐FET with the C₆₀ film grown at Δn _CT > 0, but observed no photoresponse from the device with C₆₀ film grown at Δn _CT = 0.

3. Conclusion

We observed that charge transfer within the graphene–C₆₀ system during the growth of C₆₀ crystals on a graphene template governed such growth and, thus governed the thin film's corresponding crystal structure and morphology. These charge‐transfer phenomena altered the electronic states of the graphene–C₆₀ system, forming negatively charged C₆₀ nuclei and ad‐molecules. Under these conditions, the growth of C₆₀ on graphene was favored in the vertical dimension because of the high attachment barrier energy, resulting thin films with small and randomly oriented crystallites. With this understanding, we proposed that the optimized graphene template for layer‐by‐layer growth of C₆₀ with large and uniformly oriented crystals is the graphene in which the charge transfer from graphene to C₆₀ is suppressed during the C₆₀ growth. Barristors fabricated with this graphene–C₆₀ van der Waals heterostructure showed efficient tunability of the charge injection barrier, approaching the Schottky–Mott limit. In addition, the lateral electron mobility μ_e in a planar C₆₀‐FET was also boosted to a maximum μ_e = 2.5 cm² V⁻¹ s⁻¹.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Nguyen N. N., Lee H. C., Yoo M. S., Lee E., Lee H., Lee S. B., Cho K., Charge‐Transfer‐Controlled Growth of Organic Semiconductor Crystals on Graphene. Adv. Sci. 2020, 7, 1902315 10.1002/advs.201902315