Tuning interfacial charge transfer in atomically precise nanographene–graphene heterostructures by engineering van der Waals interactions

Nanographenes (NGs) are a group of extended polyaromatic hydrocarbons with a lateral size larger than 1 nm, which possess fascinating electronic and optical properties, including tunable and strong optical absorption, high photoluminescence (PL) quantum yields, and stimulated emission.^1–6 Integration of NGs with graphene (Gr) constitutes hybrid van der Waals heterostructures (vdWHs) combining strong light–matter interaction with outstanding electrical conductivity. As such, these vdWHs provide a novel platform for exploiting intriguing optoelectronic phenomena and developing versatile devices. For instance, Yan et al.⁷ employed vdWHs composed of exfoliated Gr and NGs as the bifunctional electrodes for photoelectrochemical water splitting, simultaneously achieving efficient oxygen and hydrogen evolution. Furthermore, Liu et al.⁸ sensitized Gr by solution-processed NGs toward highly efficient photodetection.⁹ The large interfacial electronic coupling caused by the strong π–π interaction at the NG–Gr interface drives the transfer of the photogenerated holes from NGs to Gr with high efficiency (∼10%). Moreover, the charge separation time was found to be extremely long (over 1 ns), enabling effective photogating in Gr. These two fascinating charge transfer (CT) characteristics enable high-performance photodetectors with an ultrahigh responsivity up to 4.5 × 10⁷ A/W and a specific detectivity of 4.6 × 10¹³ Jones, being competitive with the state-of-the-art Gr-based heterostructures.^10–14 In the heart of these applications, interfacial CT plays a critical role in determining the final performance, yet has remained largely uncontrollable.

Recent advances in the bottom-up synthesis of NGs enable atomically precise tuning of their geometries, edge structures, and lateral sizes,^15–20 offering a potential approach to boost the CT efficiency by regulating the interaction between the adjacent layers and engineering the band alignment at the interface. For example, NGs with larger lateral sizes are expected to adhere much stronger to Gr,^21,22 thanks to the enhanced interfacial vdW interactions [as shown in Fig. 1(a)], even when normalized to the number of carbon atoms.²² This could increase electronic coupling at NG–Gr interfaces, thus facilitating CT. On the other hand, the driving force (−∆G) for CT, i.e., the energy difference between the highest occupied molecular orbital (HOMO) in NGs and the Fermi level in Gr (e.g., for transferring holes from NGs into Gr), is expected to decrease with increasing NG sizes [see Fig. 1(b)]. For NG-sensitized oxide nanoparticles, the modification of interfacial energetics resulting from NG size variation has been shown to dominate the interfacial CT rate and efficiency.²³ Smaller NGs with higher interfacial energetics enable faster CT and thus higher CT quantum efficiency.²³ Furthermore, the reorganization energy, i.e., the energy required for charge solvation, and the resulting lattice distortions associated with CT may also affect CT in our system, given the molecular nature of NGs. Although the importance of vdW interactions in the NG–Gr vdWHs has been evidenced,⁸ the interplay between the vdW interaction, interfacial energetics, and reorganization energy, and their impact on the CT processes, has remained elusive.

To address this question, herein we investigate photoinduced non-equilibrium CT dynamics in NG–Gr vdWHs with size-dependent interfacial interactions and energetics by varying the number of carbon atoms in the polyaromatic core from 42 to 96 [see Fig. 1(c)]. By employing optical-pump THz-probe (OPTP) spectroscopy, we demonstrate that the interfacial CT direction is universal in all the studied NG–Gr vdWHs with photogenerated holes transferring from NGs to Gr [as illustrated in Fig. 1(d)]. The resulting charge separation at NG–Gr interfaces is very long-lived (beyond 1 ns), leading to a photogating effect in Gr. More importantly, the interfacial CT efficiency in NG–Gr vdWHs is found to increase with NG sizes. By disentangling different contributions to the overall efficiency, we find that the electronic coupling at the NG–Gr interface dictated by the interlayer vdW interaction, rather than interfacial energetics or reorganization energy, plays the dominant role in determining the interfacial CT efficiency. Our results shed light on the CT mechanism in NG–Gr vdWHs and provide an effective knob to control the electronic coupling strength at NG–Gr interfaces by tuning the size-dependent vdW interactions.

The nanographene (NG) molecules used in this study were synthesized via a bottom-up approach, as reported in Refs. 24–27. We first dissolved NGs in powder form into toluene to get 1 mg/ml NG dispersions. Graphene (Gr) samples supported on 1 mm thick copper substrates were purchased from Grolltex, Inc. and grown by the chemical vapor deposition (CVD) method. We transferred Gr onto fused silica substrates for THz measurements following a wet transfer manner.^28,29 First, we spin-coated Gr samples with cellulose acetate butyrate (CAB) at 4000 rpm and baked them at 180 °C for 3 min. Then, the CAB-coated Gr samples were immersed into 1 M solution of ammonium persulfate (APS) for 10 min and rinsed with deionized (DI) water five times to remove Gr on the backside of copper substrates. The copper substrates were then fully etched by soaking the samples into 0.1 M APS solution for 2 h and rinsed with DI water five times to remove the attached ions. Afterward, Gr monolayers were “fished” onto the silica substrate, and CAB was removed by acetone (by 12 h soaking) and isopropanol (1 h) treatments. We prepared the NG–Gr van der Waals heterostructures (vdWHs) by spin-coating 20 µl NG dispersions on Gr samples three times in the glovebox and heating them at 100 °C for 10 min to remove residual toluene.

For the UV–vis absorption measurements of NG dispersions, 100 µl NG dispersions were diluted to 3 ml diluents and injected into cuvettes of 1 cm light path. The measurements were carried out on a Perkin–Elmer Lambda 900 spectrometer at room temperature. The spectrum was recorded with an interval of 2 nm. The UV–vis absorption measurements of NG films and NG–Gr vdWHs supported on the fused silica substrates were performed on a Perkin–Elmer 150 mm InGaAs Int. Sphere.

THz measurements of Gr, NG, and NG–Gr vdWH samples were performed by an optical pump-terahertz probe (OPTP) setup. The fundamental laser output was generated by a regenerative Ti:sapphire amplifier system, which produces 4.5 mW, 50 fs pulses at a repetition rate of 1 kHz and a central wavelength of 800 nm. The generated pulses were then split into three branches for THz generation, sampling, and optical excitation. A single-cycle THz pulse of ∼1 ps duration was generated by pumping a 1 mm thick (110) ZnTe crystal with the 800 nm fundamental pulses via optical rectification. We photoexcited samples to generate charge carriers using 400 nm pulses by frequency doubling the 800 nm fundamental pulses via a BaB₂O₄ (BBO) crystal. The time-dependent electrical field of the THz pulse was mapped out by the electro-optic sampling method.³⁰ 

Up to five NGs with an increasing number of carbon atoms in the polyaromatic core from 42 to 96 have been utilized to study the role of NG sizes on the interfacial CT process in NG–Gr vdWHs [see Fig. 1(c) for the structure; the alkyl chains at the peripheries are not considered for the nomenclature]. The right panel of Fig. 1(c) shows the UV–vis absorbance and PL spectra of NGs. As expected, the dominant absorption peaks gradually shift to a longer wavelength with the increasing lateral size of the NG. After NG deposition, Gr appears to be slightly n-doped with the Fermi level of ∼0.2 eV above the Dirac point in C60-Gr vdWHs (see the associated discussion in the supplementary material).⁸ The NG size is expected to only mildly affect the value of the Fermi level of Gr after deposition, given that the initial Fermi levels of NGs are similar and NGs are intrinsically undoped.

We examine the CT process in the heterostructure by employing ultrafast THz spectroscopy to track the photoconductivity dynamics in Gr following optical excitations to the vdWHs [see Fig. 2(a)]. In a typical OPTP measurement, we first excite the sample with 3.1 eV laser pulses (for a duration of ∼100 fs) to generate charge carriers in the vdWHs. The photoconductivity is subsequently measured by a single-cycle THz pulse with ∼1 ps duration (see details in Sec. II). Pure Gr shows a transient photoresponse on the order of a few ps due to the efficient charge carrier heating and cooling effects.^31,32 Conversely, in the vdWHs, interfacial CT could lead to a considerable change in the charge carrier density in Gr and, thus, a large conductivity modulation amplified by its high charge mobility. By knowing the initial Fermi level of Gr (∼0.2 eV above the Dirac point) and CT-induced conductivity change, one could reliably identify both the direction and efficiency of interfacial CT in Gr-based vdWHs.^8,33,34

As a control measurement, photoexcitation of pure Gr by 3.1 eV laser pulses results in a transient reduction in the conductivity with a lifetime of sub-10 ps [Fig. 2(b), red line]. The origin of such a “negative photoconductivity” has been well-understood and attributed to the enhanced momentum scattering of hot carriers in the photoexcited Gr with high charge carrier densities.^31,35–37 Following hot carrier cooling, the photoconductivity dynamics decays back to the equilibrium within several ps. We observe positive photoconductivity for pure NG films (using C96 film as an example). That is, the photo-injection of charge carriers into a semiconducting NG film leads to a transient increase in the conductivity [Fig. 2(b), brown line]. The magnitude of the positive THz photoconductivity of the NG film is substantially smaller than the negative THz photoconductivity of pure Gr, owing to the relatively low charge mobility of NG films. For the NG–Gr vdWHs, the same excitation density yields distinctly different photoconductivity from the individual Gr and NG films [Fig. 2(b), blue line]. Following the fast ∼5 ps negative photoconductivity decay due to the hot carrier response in Gr, we observe a remarkably long-lived negative photoconductivity that persists well beyond 1 ns (limited by the range of our optical delay line). Given the initial Fermi level of Gr in the vdWH is ∼0.2 eV above the Dirac point, the long-lived negative photoconductivity can be rationalized by a photoinduced hole injection process from NG to Gr (which is equivalent to transfer of “cold” electrons at the Fermi surfaces of graphene to NGs). This leads to a lowering of the Fermi level and thus electron density in Gr, decreasing the overall conductivity. Such a hole transfer scenario is in line with our previous studies for WS₂-Gr and C60-Gr vdWHs by the above-bandgap excitations.^8,33 Here, we neglect the contribution of hot electron transfer (HET) from Gr to NG for the following reasons: First, the samples are pumped by 3.1 eV laser pulses where the Gr and NG layers are both photoexcited. However, the majority of charge carriers are populated mainly in the NG layer since the absorption of NG is much stronger compared to that of Gr (>17% in NG vs ∼2% in Gr at 3.1 eV, Fig. S1). Thus, the contribution of hot carriers in Gr is expected to be limited. Second, the authors of Refs. 33 and 38 have shown that thermalized or non-thermalized hot electrons can be injected from Gr to WS₂ following subbandgap excitations. In the former case, following optical excitations, the photogenerated hot electrons can rapidly heat the charge carriers at the Fermi surface within tens of fs, generating thermalized hot carriers with a defined electron temperature obeying the Fermi–Dirac distribution. During the HET process, only hot electrons with sufficiently high energy can be emitted over the Schottky barrier into WS₂. In the latter case, injection of nascent non-thermalized electrons has also been suggested with quantum efficiency exceeding 50% when the interfacial electronic coupling is strong and the energetics of nascent hot carriers are higher than the Schottky barrier.³⁸ In our study, the nascent non-thermalized HET scenario can be ruled out because the energetics of nascent hot carriers in Gr following 3.1 eV excitations (∼1.55 eV) is still below the required energy threshold for non-thermalized HET (over ∼2 eV for C96). In addition, to test whether thermalized hot electrons in Gr contribute to CT, we have conducted control measurements by exciting only the Gr layer in the vdWHs with 1.55 eV excitations. Varied pump fluences are used to excite the system to increase the electron temperature in Gr and, thus, the probability of thermalized HET from Gr to NG (if there is any). As shown in Fig. S6, we observed no signature of CT from Gr to NGs with 1.55 eV excitations in the employed pump fluence range, ruling out the role of thermalized HET in CT. These control measurements strongly support our assignment of hole transfer from NGs to Gr as the main CT channel following optical excitations.

After confirming that the hole transfer mechanism prevails in the NG–Gr vdWHs, we further investigate the role of the NG size in the CT efficiency. For a better comparison, we summarize the photoconductivity at ∼20 ps for different NG–Gr vdWHs by varying the absorbed photon density (N_abs), as shown in Fig. 2(c). At the delay time of 20 ps, the hot carriers in graphene have relaxed back to the equilibrium. The remaining long-lived signal can be attributed to the photoconductivity changes due to CT. The photoconductivity at a fixed N_abs or the slope of the photoconductivity −N_abs curve in Fig. 2(c), therefore, represents the relative interfacial hole transfer efficiency, as summarized in Fig. 2(f) (with the left y axis; see Fig. S5 for all raw dynamics). As we can see, the smallest NG used in this study, C42, shows essentially ∼0% CT efficiency; the CT signal goes up with increasing NG sizes. At first glance, the observed trend in the size-dependent CT rate is surprising, given that the interfacial CT driving force (see Table S4), i.e., the energy difference between the ionization potential of NG and the chemical potential in Gr, is decreasing with an increase in the NG size. The obtained results seem to point to a dominant role of electronic coupling over interfacial energetics in determining the CT efficiencies in the NG–Gr vdWHs: for vdWHs consisting of small NGs, the interfacial vdW interactions are weak, as determined by its small size and low vdW interaction normalized by the number of sp² carbon atoms.²² Our results indicate that this yields weak electronic coupling and thus negligible charge flow through the interfaces. By increasing the NG size, the vdW force increases drastically, enhancing the electronic coupling and favoring interfacial CT. To test this hypothesis, we provide further analysis by taking into account the contributions of the interfacial energetics, coupling strength, and reorganization energy to the overall CT rates.

In principle, the competition between interfacial CT (at a rate k_CT) and recombination (at a rate k_R) inside NGs dictates the hole transfer efficiency η from NGs to Gr, following $η=kCTkCT+kR$⁠. The fact that only a small fraction of charge carriers (up to ∼10% for C96; see the efficiency estimation in the following discussion) are transferred from NGs to Gr indicates the presence of ultrafast competition processes characterized by a rate k_R, with k_CT ≪ k_R. As such, η can be approximated by $η∼kCTkR$⁠. The detailed recombination/loss channels remain elusive, which calls for further investigation.

To rationalize our observation of the enhanced hole transfer efficiency η with increasing NG size, we apply Marcus theory to account for the relative contributions of the coupling strength V, driving force −ΔG, and reorganization energy λ to the overall k_CT. The rate of hole transfer from the HOMO level in NGs to the continuum accepting states in Gr reads^23,39,40

where the zero-energy reference is the ionization potential of NG (see Table S3). The driving force for CT, −ΔG, is the energy difference between the ionization potential of NG and the chemical potential μ in Gr. ℏ is the reduced Planck constant, k_B is the Boltzmann constant, and T is the temperature. $vE+ΔG+μ=2E+ΔG+μπℏvF2$ is the density of states of Gr around the Dirac point, and $fE+ΔG+μ=1e(E+ΔG+μ−μ)/kBT+1$ is the Fermi–Dirac distribution. The reorganization energy is calculated using density functional theory at the PBE0-D3(BJ)/def2-TZVP//PBE0-D3(BJ)/def2-SVP level of theory^41–45 (see Table S2 and Sec. 3 in the supplementary material for computational details). The electronic coupling strength V depends on the donor–acceptor wave function overlap and, therefore, the distance between NG and Gr, determined by the vdW attraction strength between NG and Gr (given by the C₆ coefficient of NGs²²). We assume that V is linearly proportional to the interfacial vdW attraction strength as a first-order approximation. Integrating Eq. (1), we find that the estimated k_CT trend with NG size matches well to the relative hole transfer efficiency, as shown in Fig. 2(f) (the right blue axis). This result indicates that the variation of k_R is substantially smaller than that of k_CT in our experimental systems, when increasing NG size.

Moreover, our estimation disentangles the relative weight of each contribution to the CT: by increasing the size of NG from C42 to C96, the contribution of the coupling term increases by an order of magnitude, thus dictating the overall size-dependent k_CT. k_CT is not sensitive to changes in −ΔG since its variation over hundreds of meV shows a negligible impact on k_CT. This is not surprising, given that −ΔG is substantially larger than λ, so that the transfer of holes from the HOMO of NGs to each individual accepting state (with −ΔG > λ) lies in the Marcus inverted region, with k_CT rapidly decreasing with increasing −ΔG. In addition, λ plays a minor role, resulting in a small enhancement factor of 1.2 for the largest NG used in this study. Overall, the electronic coupling strength seems to be the primary factor affecting the efficiency of hole injection from NGs to Gr. We note that the previous study on CT between NG and a metal oxide revealed a dominant role of ΔG in determining NG-size-dependent CT rates.²³ In that NG-sensitized oxide system, interfacial coupling is expected to be weak and invariant with NG sizes. In line with our recent report on the uniqueness of the NG–Gr systems:⁸ thanks to the strong π–π interaction at their interfaces, our results here indicate that the electronic coupling strength plays a paramount role in determining CT rates in the constituted all-carbon-based vdW heterostructures.

Finally, to demonstrate the relevance of the heterostructure for applications, we independently determine the absolute CT efficiency η, which is defined as η = $NCTN$⁠. Here, N_CT represents the number of transferred holes and N is the number of photogenerated charge carriers, which equates to the product of the number of photons absorbed in the NG layer (N_abs) and the photon-to-charge conversion efficiency (φ). To quantify N_CT, we employ THz time-domain spectroscopy (THz-TDS) to measure the complex photoconductivity of C96-Gr vdWHs at ∼100 ps, where the contributions from the hot carrier effect in Gr and excitons in C96 to the complex photoconductivity are expected to be null. In the THz-TDS measurements, we map out the entire THz waveforms without (as a reference) and with photoexcitations at a certain pump–probe delay. We can then convert the pump-induced THz absorption and phase shift in the time domain into the complex photoconductivity spectra in the frequency domain (see details in the supplementary material, Sec. 2). Figure 2(d) shows two exemplary complex photoconductivities at different pump fluences. The obtained photoconductivity can be described by the conductivity difference between the quasi-equilibrium states in Gr before and after hole injection. Both the states can be well-described by the Drude model. This analysis provides the charge carrier density and scattering time in Gr following hole injection (see details in the supplementary material, Sec. 2). Knowing the N_CT (from THz-TDS) and N_abs (from absorption), we can estimate the CT quantum yield η (=N_CT/N ∼ N_CT/N_abs) to be ∼10% at 100 ps for C96-Gr vdWHs. Note that the actual total η following hole transfer (taking place at sub-10 ps) is underestimated, given that the photon-to-charge conversion efficiency of C96 here is assumed to be 100% (i.e., N = N_abs) and partial back recombination has taken place within 100 ps (see details in the supplementary material).

In summary, we elucidated the origin of the interfacial CT in NG–Gr vdWHs by systematically varying the NG sizes to tune the interlayer vdW interactions. We found that vdW interactions play a dominant role in determining the interfacial CT efficiency, whereas interfacial energetics and reorganization energy have a minor influence. By increasing the size of NGs from 42 to 96 carbon atoms in the polyaromatic core to enhance the interfacial coupling strength, we find that the CT efficiency increases by one order of magnitude, despite the interfacial energy driving the CT process being unfavorably reduced. Our results provide critical insights into the CT mechanism and offer a new tool to tune the electronic coupling at NG–Gr interfaces by controlling the size-dependent vdW interactions.

See the supplementary material for measurement details and computational methods.

X.Y. and S.F. thank the fellowship support from the China Scholarship Council (CSC). M.M. acknowledges the postdoctoral support from the Alexander von Humboldt Foundation. Research in Strasbourg has been supported by the European Commission through the ERC project SUPRA2DMAT (Grant No. GA-833707), the Graphene Flagship Core 3 project (Grant No. GA-881603), the Agence Nationale de la Recherche through the Labex project CSC (Grant No. ANR-10-LABX-0026 CSC) within the Investissement d’Avenir program (Grant No. ANR-10-120 IDEX-0002-02), the International Center for Frontier Research in Chemistry (icFRC), and the Institut Universitaire de France (IUF). We acknowledge funding support from the Max Planck Society.

The authors declare no conflicts of interest.

X.Y. and S.F. contributed equally to this work.

The data that support the findings of this study are available from the corresponding author upon reasonable request.