Careful design of the synthetic routes toward engineered and thermodynamically favorable properties can push the limits of frequently used compounds, thus opening new venues for their feasible use in energy-related applications.

Graphene is an impressive active surface with remarkable mechanical and optoelectronic properties. When in contact with a semiconductor (SC) it can give rise to a junction that can provide ample photoexcited carrier transport features across the atomically thin depletion region. Recent developments in ease of synthesis and manipulation of graphene and its derivatives (bilayer graphene, few layer graphene, reduced graphene oxide, etc.) enabled the possibility of fabricating hybrid systems for next generation optoelectronic devices1. Thus far the junctions of graphene/SC have shown ample performances in different optoelectronic applications such as photodetectors2, light emitting diodes3, flexible transistors4 and photovoltaics (PVs)5,6,7. Essentially, in PV applications, graphene/SC junction can be beneficial for directional charge extraction and increasing the built-in potential across the depletion surface. Conventionally, the origin of the photoresponse in graphene sheets is ascribed to the electron-acoustic phonon decay and electron-electron scattering8,9. Particularly, the electron-electron scattering mechanism, is evaluated as the main decoupling phenomenon, responsible for multiple carrier excitation per photon and hot carrier transport in graphene/few layer graphene junctions8,10. In fact, utilizing the carrier transfer properties of the graphene/SC junctions through different interfacial engineering is an accessible method for increasing the performance of the light-harvesting systems. For instance, employing graphene as semitransparent illumination window in a planar silicon PV cell can promote dual functionality as antireflection coating and transparent conductive substrate with intrinsic electric field effect modulation for electron-hole separation5.

Despite several investigation on the planar geometry graphene (few layer graphene)/SC junctions, the effect of vertically aligned graphene/semiconductor light-harvesting systems is rather unexplored11,12. A motivated study, indicated that the graphene ripples, produced in a junction of graphene and hydrogenated silicon carbide (H-SiC) semiconductor, can present the alternation in barrier height values as large as 1.3 eV13. In that sense, we note that such transfer process is a trivial phenomenon in light-harvesting properties and output current of the hybrid systems, in which the graphene sheets are placed normal to the excitation direction. As a critical note, it is questionable if the embedded graphene sheets in SC can promote the carrier transport and separation in the bulk of semiconductors that intrinsically present low conductivity and photoexcited carrier transfer14,15? Here we refer to photoexcited carriers as the carriers originated in the system upon excitation from light, which results in both electron-electron scattering collisions and energy transfer to phonons with longer life-time16.

In our work, we have selected magnetite (Fe3O4), a well-known ubiquitous metal oxide (MOx) SC, and a room temperature ferroelectric material, composed of low-cost and abundant elements17,18. Similar to most of the transition MOx SCs, the conductivity in magnetite is attributed to the 3d6→3d6 low energy transitions and small polarons interactions of their valence electrons19,20,21. We experimentally demonstrate the ample optoelectronic performance of the Fe3O4 magnetron sputtered thin-film incorporating a vertically aligned scaffold of reduced graphene oxide (VA rGO) as a visible light absorber hybrid composite. Based on the topological features of the VA rGO in the samples, we have investigated two important functionalities of the systems. Firstly, we studied the dominant relaxation process of photoexcited carrier injection from Fe3O4 into the rGO scaffold as an arising factor for creation of secondary electrons in-depth of the adjacent Fe3O4 semiconductor. Given the morphology-dependent characteristic of the system, we employed intermodulation conductive force microscopy (ImCFM) as a key tool to interconnect the topological ripples and curves of the sample’s surface originated from embedded rGO scaffold to the derived current of the system and the interlayer capacitance. Secondly, we assigned the in-depth carrier multiplication process in the Fe3O4 layer to carrier rearrangement in rGO network structure, corresponding to effective charge transport into rGO (typically ascribed to carrier-carrier scattering in rGO) using surface photovoltage measurements (SPV). The SPV data of the VA rGO/Fe3O4 system indicated a strong broad-band potential inhomogeneous response of the samples because of uneven carrier transport in-depth of the system. Finally, we deemed to demonstrate the photoexcited carrier to current conversion of our samples in a PV system by interfacing the structure with cuprous oxide (Cu2O) p-type MOx semiconductor. Our results anticipated the optoelectronic performance of the VA graphene junctions as a light-harvesting system. Our data comprehensively presented the effect of photoexcited fast charge transport in graphene establishing highly effective charge collection from the adjacent photoexcited semiconductor, and providing a guideline for the further implementation of these devices.

Results and discussion

Identification of the photoexcited carrier transfer

In this work we have employed a device architecture based on an interconnected scaffold of VA rGO conformally embedded on a thin-film of Fe3O4 with nominal thickness of 100 nm. The amount of the rGO sheets used as scaffold in the samples were varied by the number of the electrodeposition cycles (details in Methods). The variation of the rGO concentration is an important parameter to modulate the carrier transport properties of the rGO network on the bulk of the SC. Three sets of samples with different concentration of the rGO scaffold were prepared, labeled as: VA rGO (30 cycle)/Fe3O4, VA rGO (50 cycle)/Fe3O4 and VA rGO (100 cycle)/Fe3O4. The structural features of the samples are schematically illustrated in Fig. 1. We note that assigning a conventional process of photoexcited carrier-carrier scattering and multiple hot-carrier generation mechanism mediated by rGO in the bulk of the Fe3O4, with low photoexcited carrier mobility, requires the system to deliver electrical current under photoexcitation. Thus, a trivial motive of this work corresponds to fabrication of the PV device based on rGO integrated hybrid system, whose main objective is the conversion of light in electron-hole pairs. For this purpose, the samples were ultimately covered with 100 nm Cu2O, a nominal p-type SC to create a heterojunction composite device. The results were compared to a bare heterojunction thin-film sample without rGO scaffold consisting of 100 nm Fe3O4 and 100 nm of Cu2O respectively. An important emphasis of the work corresponds to the properties of the rGO and the Fe3O4 interface, relating the photoexcited carrier transport of rGO to the formation of the bond polarization at the rGO/semiconductor interface shown in Fig. 1b12,22. Further in our analysis, we treat the VA rGO/semiconductor junction as a Schottky junction with variant Fermi level of rGO because of carrier redistribution in the junction (Fig. 1c). This phenomenon addressed in Fig. 1c by yellow and red region is the result of carrier transfer from semiconductor interface creating a downward energy-band shift in Fermi level of VA rGO, and is responsible for Schottky barrier height (ΦSBH) fluctuation at the interface. The ΦSBH variation signifies the carrier transfer at the interface of the VA rGO/Fe3O4 junction, resulting in the transfer of photoexcited charges in the bulk of consequent junction. Further, we include a localized barrier height mapping and a simple model describing the experimentally obtained trends.

Fig. 1: Overview of the morphology correspondence to local band structure.
figure 1

a Schematic illustration of the vertically aligned rGO scaffolds embedded in Fe3O4 thin-film. b Carrier transfer inbetween rGO sheets and that of Fe3O4 semiconductor. c The band-energy diagram of the hybrid system confers barrier variable charactristic of the SC/graphene junction. The variation of the Fermi level position in rGO energy band as a result of the carrier transfer.

Figure 2a demonstrates the morphology of the VA rGO sheets after deposition of the ~100 nm Fe3O4 thin-film overlayer. The features of the samples indicate that the sheets were attached to the substrate from the basal plane of the rGO, having nearly normal orientation towards their deposited platform. As a key insight into the rGO-mediated carrier transfer process, the amount of the network’s crisscross is varied by the number of electrodeposition cycles. In our work we have studied 30, 50 and 100 deposition cycles. Statistical measurements on the planar view of the samples, as shown in Supplementary Fig. 1, determined the number of rGO flakes deposited on the FTO. The obtained results indicated the number of 20-25, 80-95 and 430-450 rGO sheets per 100 µm2 for 30, 50, and 100 cycles of electrodeposition, respectively. Notably, the deposited rGO flakes may include few layer graphene, bilayer graphene or single layer graphene due to the nature of deposition method. The rGO flakes have an average sheet lateral size of 7-10 µm. The morphology of the fabricated heterojunctions after subsequent deposition of the Cu2O layer is shown in Fig. 2b, d (the structure of the samples in this case includes VA rGO, 100 nm Fe3O4 and 100 nm Cu2O in a bottom to top order). Clearly, the surface morphology of the samples presents an undulating nature as result of the underlying VA rGO architecture. Cross-section view of the VA rGO/Fe3O4/Cu2O sample (Fig. 2d) illustrates the vertical orientation of the rGO flakes even after deposition of the MOx films, in agreement with the previous observations. The thickness profile of the Fe3O4 and Cu2O heterojunction in thin-film geometry (not including rGO network) (Fig. 2doutset) indicated a ~100 nm thickness for each MOx layer, respectively. The estimated thickness and the atomic composition of the samples were further confirmed via Rutherford backscattering measurements (RBS) of the identical deposited films on Si wafer substrate as shown in Supplementary Fig. 2.

Fig. 2: Morphological characteristics of vertically aligned rGO Fe3O4 and Cu2O composites.
figure 2

SEM image of the a vertically aligned rGO scaffold after deposition of the 100 nm Fe3O4 film and b after subsequent Cu2O deposition. c SEM image corresponding to a detached single flake of the rGO sheets on a copper grid conformally covered with Fe3O4 and Cu2O films. Outsets: EDS mapping of the Fe and Cu from the Fe3O4/Cu2O deposited films on the rGO flake. d Tilted view of the hybrid composite structure. The outset SEM images magnfiy the conformal coverage of the rGO flakes and the substrate with a dual layer of Fe3O4 and Cu2O films. e Raman spectra corresponding to ultra-high vacuum annealing of the graphene scaffold. f X-ray diffraction pattern of the films both acquired in grazing incidence reflection mode. Asterisks (*) marks correspond to the reflection peaks from FTO substrate underneath the deposited films.

The X-ray diffraction (XRD) pattern of the iron oxide film conferred the presence of the spinel phase crystal structure of the magnetite (Fe3O4) sputtered film, without any phase impurity. The Rietveld refinement data on the deposited Fe3O4 layer indicated the lattice constant of the crystals to be ~0.89 nm, in agreement with previously reported results23. Similarly, Cu2O layer presented a pure phase cubic crystal with corresponding diffraction patterns of (1,1,0), (1,1,1) and (2,2,0) planes at 33°, 36° and 62° respectively. The XRD data verified the absence of metallic Cu and/or CuO phases in the sputter deposited film. During device fabrication, special care was taken to the UHV reduction degree of the rGO network by probing the Raman signal and the ratio of the D, G and 2D bands of the samples (Fig. 2e). A complex profile Raman spectrum can be observed for both samples with G-band maximum at 1591 cm−1. The 2D-band at 2922 cm−1 in GO Raman spectrum indicates the characteristics of disordered and twisted graphene24. After UHV annealing at 200 °C the 2D-band positions slightly shifted to 2873 cm−1 indicating the reduction of the GO scaffold25. The correlated ratio of the areas of 2D:G peak 0.8 in heat treated sample also verifies the rGO characteristic trait24,25.

Intermodulation conductive force microscopy

As mentioned before, the study of the output-derived current of the system can be a deterministic approach to distinguish whether the fast charge transport in rGO scaffold is a dominant mechanism for carrier collection within the bulk of Fe3O4. In that sense, the current distribution maps and the interlayer capacitance maps of the system were obtained using a multifrequency dynamic technique in atomic force microscopy (Fig. 3), the so-called intermodulation conductive force microscopy (ImCFM). The ImCFM signals afford an exquisite nonlinear current-voltage maps with high spatial resolutions, which we previously reported26. Herein, a small sinusoidal AC voltage is applied to the cantilever and the potential sweep is forwarded through FTO substrate making it possible to retrieve J-V sweep in each appointed scan pixel. The spatially localized J-V characteristic of the samples corresponding to the few denoted points on the surface of the VA rGO(30 cycle)/Fe3O4 surface is presented in Fig. 3a. All the sequential J-V curves demonstrate the same current-voltage trend. However, the variation in the obtained current flux of the sample can be assigned to the anomalous in-depth potential generated due to different carrier mobility dictated by the embedded rGO scaffold. The bright region in the 1 × 1 µm2 current maps correspond to the topological features of rGO network underneath the Fe3O4 film (Fig. 3b–e). Clearly, the surface ripples corresponding to the rGO flakes indicate a higher current gradient in average 40 magnitude larger than the areas where rGO ripples are not observable. This value is simply calculated by applying K-means clustering of the data based on their highest and lowest current values obtained in mapping data. For bare Fe3O4 region there is vividly small overall retrieved current (0.1 nA) for distances far away from the rGO interfaces. Substantially, one of the main aspects of the graphene/semiconductor junctions is already demonstrated as a barrier variable junction, in which the transfer of the charges from semiconductor to Fermi level of graphene results to a downward shift (opposite to direction of charge transfer) in the Fermi position of graphene and consequently alters the Schottky barrier height (ϕSBH) and built-in potential (ϕi) level12. The interfacial Schottky barrier changes of our system are responsible and have a strong dependence on the carrier rearrangement at the rGO/Fe3O4 depletion region27. We have demonstrated the spatial fluctuation of the barrier height due to electric field modulation and the in-depth surface dipole modification of the adjacent Fe3O4 semiconductor in Fig. 3k–m. The Schottky barrier height mapping of the samples (Fig. 3l) were acquired through applying a small AC voltage normally in kHz gain to the cantilever in the unsaturated current region of the electrical response. Later we consider the relation between the interlayer capacitance (C) responses of the sample with ϕi of the junction, as follows:

$$C = \sqrt {\frac{{e\;\varepsilon N}}{{2\left( {\varphi _i - V - \frac{{kT}}{e}} \right)}}}$$

Where N is the charge density, V is the applied reverse bias and ε is the dielectric constant of the semiconductor. Accordingly, we constructed the ϕSBH of the sample via mathematical determination of the square reciprocal of interlayer capacitance (\(1/C^2\)) versus applied bias (V). The intercept of acquired plot with X axis (applied potential as in Fig. 3k), results in the \(\varphi _i - kT/e\) and the slope is equal to \(2/\left( {e\;\varepsilon _sN} \right)\). The ϕSBH value maps can be eventually calculated considering the obtained ϕi through the relation:

$$\varphi _{SBH} = \varphi _i + E_c - E_F$$
Fig. 3: Interlayer capacitance mapping and localized barrier modulation.
figure 3

a A rGO(30 cycle)/Fe3O4 surface of 1 × 1 µm2 imaged in ImCFM mode. Current maps measured at applied 0, 0.5, 1, and −1 V bias (upper row, be) and corresponding capacitance maps (bottom row, gj). Corresponding J-V curves taken at the green, orange, blue and violet marks are shown in a. The green mark shows the J-V curve measured on location without rGO network presence, while the J-V curves measured on spots where rGO network is present under the SC film are shown in the blue, violet and orange marks and are shown in a. The corresponding capacitance curves are shown in f. Green mark is taken in the area without any clear illumination effect. The calculated Schottky barrier heights for the taken J-V curves are shown in k. The reconstructed Schottky barrier height map is shown in l and corresponding distribution histogram is in m. The scale bar in the image is 200 nm.

To minimize the electrostatic influence of the surface and to avoid the contributions of the carrier storage of the samples, we limited our investigation to the sample with the lowest concentration of the VA rGO (30 cycle)/Fe3O4. We observed strong ϕSBH modulation on the surface of the sample, where VA rGO network ripples were clearly featured (Fig. 3l), with similar trends as current mapping of the samples in Fig. 3c–e. The demonstrated ϕSBH mappings are referenced to the average neutrality surface of the Fe3O4 possible films, where VA rGO sheets are not present. We note that the crisscross of the VA rGO network causes intrinsic local inhomogeneity in the VA rGO/Fe3O4, which we assign to the Fermi level variation of the rGO network because of the applied bias and spatial inhomogeneity, as shown in Fig. 3l outset. We report a ϕSBH modulation value of an average −1 V to −0.5 V on the VA rGO network. The gradient of this fluctuation can be merely assigned to the carrier redistribution between Fe3O4 layer and the rGO layer underneath. Here, the derived current signal at negative forwarded bias from the FTO can also provide a direct evidence on barrier height modulation due to direct carrier transfer from the rGO scaffold (in this case we consider rGO scaffold as a positive acceptor semiconductor). We assigned the alternation of the barrier height in this case (observed variation by applied bias) due to the carrier transfer in VA rGO network, as discussed further in the text. Relative to the semiconductor interface, this tunable carrier transfer can possibly form dipole momentum in the interface, which induces the carrier transfer in the bulk of the Fe3O4. The reported ImCFM results here, provide evidence of a strong asymmetry in carrier transfer between rGO and Fe3O4 that is driven, as we will argue below, by strong correlations of the photoexcited charge injection and transport in rGO (typically occurring through non-collinear, Auger scattering and carrier multiplication, even if we are unable to identify precisely the conduction mechanism). At the same time, the electrical behavior shows the strong structural correlations between graphene and semiconductor28,29. We now discuss the possible sources of such asymmetry in rGO network and show its effect on in-depth carrier transfer within the Fe3O4 rather than intrinsic carrier transfer in pristine graphene. To confirm the obtained ImCFM results, we have further conducted localized conductive AFM measurements of the VA rGO (30 cycle)/Fe3O4 sample in different distinct points of the samples Supplementary Fig. 3. The conductive AFM measurements on the rGO incorporated samples present the similar J-V trends as the obtained current-voltage characteristic of the samples using ImCFM method (Fig. 3a). The conductive AFM results are collected on top of the rGO ripples remarking the increase in the carrier mobility of the hybrid sample in the presence of rGO scaffold. The conducted localized conductive AFM and the micro J-V characterizations are rather important points to emphasize the accuracy and the cohesiveness of the ImCFM results in this case. It should also be noted that, the result interpretations and the suggested band-energy diagram of the samples are considered by minimizing the charge relaxations of the systems. In this regard, we have considered that surface states related to defects and oxygen vacancies of the MOx surfaces were passivated in ambient environment30 and we have obtained high electron injection yield after rGO integrations. Although, it is noted that, some minor surface charging effects, and light-induced photothermal effects cannot be avoided from the interfaces.

Surface photovoltage measurements

We apply surface photovoltage (SPV) technique with variable excitation wavelength to study the effect of carrier-carrier interactions in potential relaxation cascade of rGO incorporated samples (Fig. 4a, b). In both cases the SPV spectra of the samples were recorded in forward and reverse scanning direction of the illumination wavelength, to avoid the contribution of the surface electrostatic hysteresis. In general, the SPV response of the thin-film nanomaterials is associated with the carrier redistribution and in-depth-diffusion processes31. Figure 4a elucidates the SPV behavior of the bare Fe3O4 thin-film compared to the VA rGO integrated Fe3O4 sample. The energy-band carrier redistribution of Fe3O4 film resulted in a broad upwards peak (characteristic of n-type semiconductors) in the SPV spectrum of both samples (with and without VA rGO) starting from ~620 nm. Notably, the VA rGO (30 cycle)/Fe3O4 sample demonstrated an enhanced potential compared to the Fe3O4 bare film even in the sub-energy band region (below 600 nm), which is contradictory to the optical conductivity results denoted in the next section. The possible mechanism can be associated with the excess charge storage ability of the samples after VA rGO integration. The VA rGO scaffold in this case can introduce in-depth charge trapping sites, which can serve to reduce the density of effective surface electrons and decrease the band-bending even in sub-band energy regions10. The SPV spectrum of the bare Fe3O4/Cu2O and VA rGO Fe3O4/Cu2O hybrid p-n junctions are shown in Fig. 4b. Accordingly, the energy band gap overlap of the Fe3O4 and Cu2O layer resulted in a wide SPV transition peak centered at ~638 nm. In this case, similar to the electronic behavior of the heterojunctions, the most important parameters determining the shift in SPV spectra of the heterojunction are the differences between energy levels of the conduction band (ΔEc) and valence level (ΔEv) of the each p-type and n-type semiconductors. Rather small differences between the optical energy-band level of Cu2O and Fe3O4 in this case can be considered as the main reason for the wide transition band at SPV spectra of the heterojunction samples. Also it should be noted that the effective electron affinity and the possible local interface dipoles during the experiments can limit the direct interpretation of the ΔEv + ΔEv values32. Unlike bare Fe3O4/Cu2O junction, the VA rGO Fe3O4/Cu2O sample presented a notably higher SPV values, which is a signature of the non-uniform carrier generation/transport due to intervention of VA rGO scaffold in transport properties of the films, also known as Stankiewicz effect. As a result, the un-homogeneity due to rGO network in net-charge carrier mobility in different areas of the samples can cause an anomalous surface photovoltaic effect, which normally results in strong open circuit photovoltage (Voc) values (further confirmed in the micro J-V measurements). We exemplify that the VA rGO network promotes higher potential gradient in-depth of the semiconductor, confirming the role of fast charge transport in rGO to mediate the photoexcited carrier redistribution in Fe3O4 film.

Fig. 4: Surface potential relaxation cascade of the composites.
figure 4

i Surface photovoltage spectroscopy of the (a) rGO integrated Fe3O4 thin-films compared to bare Fe3O4. b SPV of the hybrid p-n junctions prior and after VA rGO incorporation. The solid lines in the SPV spectra of the samples are refered to the baseline univariate normal distribution of the acquired data. c AFM image of the VA rGO/Fe3O4 junction. Outset presents the nominal height profiling of the sample’s ridges.

We further characterize the optical interaction of the VA rGO scaffold compared with bare FTO substrate and find that the structure has a spectrally flat absorption of 18% in the 370-1000 nm region, a substantial absorption spanning the visible till near infrared range (Supplementary Fig. 4). This absorption is attributed to momentum scattering of the intrinsic carriers and possible internal light trapping of the crisscross network morphology33,34. Subsequently, a step to evaluate the harvesting and transport properties of the system prior and after integration of the VA rGO scaffold is to probe the optical conductivity (σ) of the samples. In this regard, real and imaginary parts of the complex optical conductivity of bare Fe3O4 and VA rGO integrated samples were calculated from reflectance spectra (90° angle of incident) using extended Drude-Lorentz method and experimentally collected electrical conductivity of the films (Supplementary Fig. 5). The reflectance spectra of the samples (Supplementary Fig. 5a), further verified the internal light trapping and higher light absorption of the Fe3O4 samples after rGO incorporation. The fluctuation of the reflectance spectra and the refractive index (η) of the VA rGO (30 cycle)/Fe3O4 sample around 2.7 to 3.5 eV can be attributed to undulating nature of the sample’s rough surface, creating alternations in the refractive index of the films (Supplementary Fig. 5a, b). More importantly, real and imaginary optical conductivity of the VA rGO (30 cycle)/Fe3O4 presented larger optical conductivity in the visible range between 2.5 eV and 1.5 eV, where optical absorption of the Fe3O4 is dominated. That can relate the increased band-energy transition of the Fe3O4, to carrier multiplication dynamics taking place in VA-rGO network. Correspondingly, the real and imaginary parts of the optical conductivity resulted in an average increase in optical conductivity of around 44% and 29% respectively in the range of 2.5 eV to 1.5 eV after rGO integration (Supplementary Fig. 5c, d).

Topological features affect current-voltage behavior

An atomic force microscopy (AFM) topography map of the VA rGO/Fe3O4 surface shown in Fig. 4c reveals the undulating nature of the samples that encompasses several sharp ridges and dips. The sharp ripples frequently appear all over the surface but also sometimes encompass patches and deviations, presumably corresponding to inhomogeneity of the rGO scaffold underneath. The small ridges with the average diameter of nearly 40 nm are related to vertical layer rGO sheets, pinned from their basal plane under deposited semiconductor layer Fig. 4coset. The 1 × 1 µm2 height mapping of the VA rGO (30 cycle)/Fe3O4 sample demonstrated the average height between 240-320 nm. Here we note that the topological features of the sample, in this case, is showcasing an average illustration of the sample morphology. However, the average size distribution of the embedded rGO flakes are larger than the possible scanning area (as reported, rGO flakes present an average size distribution of 7-10 µm) yet the surface inhomogeneities are rather inaccessible for the AFM cantilever.

For the micro J-V curves qualitatively similar curve shapes are also observed in the p-n junction samples with different rGO contents, elucidated in Fig. 5. The results were compared to bare Fe3O4/Cu2O sample with the identical thickness of the MOx layers.

Fig. 5: Current-voltage behaviour of the heterojunctions.
figure 5

Micro J-V behavior of the champion devices of VA r-GO/Fe3O4/Cu2O heterojunctions campered with bare Fe3O4/Cu2O under 1 Sun illumination.

To build a reliable statistical behavior of the samples, we fabricated several small area PV cells, with nominally identical circular deposited contacts (diameter 2 mm), to probe the J-V characteristics of the samples. Our data correspond to the champion devices (details of device fabrication in experimental section) collected using a copper probe whilst being illuminated (1.5 AM) from the implemented hatch of the measuring table, from the backside of samples (FTO side). The acquired photovoltaic performance of the samples are presented in Table 1.

Table 1 Photovoltaic evaluations.

All the VA rGO integrated samples demonstrate a significant PV effect, compared to the bare Fe3O4/Cu2O sample, which indicated poor photoexcited carrier conversion characteristic. The highest recorded Voc corresponds to the VA rGO (50 cycle)/Fe3O4/Cu2O sample. Notably, the Jsc decreases by increasing the content of the rGO in the samples from (1.5 ± 0.1)×10−1 mA/cm2 for 30 cycles to (4.90 ± 0.5)×10−2 for 100 cycle heterojunction indicative of the charge entangling abilities of the rGO scaffold as previously observed by SPV results. A possible explanation can correspond to the charge trapping properties imposed by the VA rGO scaffold giving rise to the assumption that the excess rGO content in the samples can entangle the majority carriers prohibiting the delivery of the derived current. Yet, the Voc values remains relatively high due to the large depletion region and anomalous surface photovoltaic effect created by integrating VA rGO scaffold in the samples. In that sense, the VA rGO (30 cycle)/Fe3O4/Cu2O sample demonstrates a good tradeoff for carrier transport and the distinguished depletion layer modification (carrier generation). Similar results were recorded in the J-V characteristic of the bare Fe3O4 and the VA rGO integrated sample in Supplementary Fig. 6. The samples demonstrated a clear rise in the current flux after integration of the rGO scaffold in Supplementary Fig. 6a in accordance to previously noted data using conductive AFM analysis. For a clear comparison we have demonstrated the J-V behavior of the heterojunctions in the full range of scans from -1 V to 1 V and in both dark and under 1.5 AM Solar illuminations Supplementary Fig. 6b, c. In summary, our results shed light on the design and photophysics of vertically aligned graphene and semiconductor architecture for their application in light-harvesting systems. We have experimentally studied the effect of charge injection, transport and collection in photoexcited rGO and their effect on potential relaxation cascade within the adjacent semiconductor. Using the embedded rGO network within the semiconductor, we have set to broaden the optical response of a semiconductor with poor photoexcited diffusion length, beyond the conventional response limited by optical energy bandgap. By locally probing the surface current maps, we have indicated the drastic increase in derived current of the samples and assigning it to the rGO mediated topological features. The ImCFM results points out the interfacial carrier transfer between rGO flakes and the Fe3O4 by directly probing the modulation of the barrier height of the interface. Our results recorded a variation in barrier height level of the rGO/Fe3O4 junction between a maximal value of -1 V to a lower minimal value of -0.5 V. We assigned this variable barrier height properties of the rGO/Fe3O4 junction to the ballistic carrier transfer in rGO sheets and possible formation of the series of dipoles at the interface. Furthermore, the crisscross of the rGO network underneath the Fe3O4 is expected to be highly related to the output current, enabling a simple pathway to manipulate the carrier transfer over a wide spectral range. In that sense, we implied the rGO/Fe3O4 composite layer in an all-oxide heterojunction PV system, accompanied by Cu2O p-type layer, which resulted in considerable PV performance which was not granted on the device without VA rGO scaffold (highest Voc achieved in the range of 0.56 V). We indicated that the concentration of the rGO network, can optimally tune the internal potential of the junction and thus increase the Voc of the device. Overall, the vast integration possibilities suggested by vertically aligned architectures of the edge exposed graphene scaffold, promotes considerable potential growth for different applications and technologies where the transfer of the photoexcited carriers in-depth of the semiconductors are limited.


Sample preparations

The rGO network were prepared using cyclic electrodeposition of the graphene oxide (GO) sheets (single layer graphene oxide sheets, thickness (0.43–1.23 nm) from US Research Nanomaterials Inc) dispersed in distilled water (1 mg/ml). The deposition was conducted at the scan rate of 50 mV s−1 with upper and lower vortex potential of −0.5 V to 2 V respectively, having platinum foil (3×3 cm2) as counter electrode and the Ar cleaned FTO glass (sheet resistance 15 Ω/cm2) as the working electrode. The amount of the GO network crisscross is tuned by the number of deposition cycles. Subsequently, the deposited GO network were annealed in ultrahigh vacuum at 200 °C for 30 minutes to transform into rGO and also remove the residual surface oxygen-containing groups. It should be noted that the GO reduction conditions were previously investigated by probing the structural changes and the degree of reduction using SEM imaging and Raman spectroscopy of the samples. Later, without breaking the dictated vacuum, the samples were conformally covered with 100 nm Fe3O4 film using reactive magnetron sputtering from pure Fe target (J.Lesker purity 99.95%) in argon and oxygen environment. The Fe3O4 layer was deposited at applied power of 180 W and the relative Ar:O2 flow of 33:7 sccm. For the samples designated for heterojunction fabrication, a nominal Cu2O film was also sputter deposited on the VA rGO/Fe3O4 surface. Similarly, the Cu2O layer were deposited from a pure Cu target (J.Lesker purity 99.95%) in Ar and O2 atmosphere (0.31 Pa deposition pressure). All the deposition were carried out in room temperature while cohesively rotating the holder stage at the constant rate of 7 rpm. A schematic view of the heterojunction is presented in Supplementary Fig. 7.


The morphology of the samples were characterized using Magellan 400 Field emission scanning electron microscope (FE-SEM from FEI). The diffuse reflectance spectra were recorded with Cary5000 UV-vis-NIR spectrophotometer. The X-ray diffraction pattern of the samples were carried out using PANalytical Empyrean diffractometer, Cu anode (kα1 emission line at 8.04 keV) scanning in the range of 10°<2Ɵ < 80°. The current-voltage curve of the samples were recorded using a spring-legged probe in local contact with the surface and another contact channeled to FTO substrate. The samples are under controlled exposure of simulated solar light (1.5 AM) directed from the backside of the FTO glass. To avoid contribution of short circuit and shadowing effects, all electrical measurements were conducted without any deposited conductive electrical contacts, only using the spring-legged probe connected to the surface of samples (circular with 0.6 mm diameter). SPV measurements were performed by illuminating the samples via a Newport Cornerstone260 light source equipped with a monochromatic light modulator with a 5 Hz chopper. Subsequently, the photogenerated voltage measured using a 2400 Lock-In amplifier. Rutherford backscattering spectrometry (RBS) was carried out on the identically deposited Fe3O4 and Cu2O films on Si substrate via 1.8 MeV He4+ ion beam, scattering angle θ=160°, IBM geometry. XRUMP code simulation was used for data analysis. A NanoWizard ULTRA Speed AFM (JPK Instruments AG) equipped with a multi-frequency lock-in amplifier (MLA) (Intermodulation Products AB) was used to conduct the ImCFM measurements. The MLA was used to sample the voltage. A soft special RMN-12PT400B probes (Bruker) made of pure platinum, with a nominal spring constant of 0.3 N m−1 and an outer tip radius of about 20 nm, were used to withstand high local currents during the ImCFM measurements. The measurements were carried out in the AFM contact mode at a deflection signal to set point signal difference of 0.2 to 0.3 V. The ImCFM measurements were carried out at 20 °C temperature and the Ec-EF of 195 meV which reflected in our barrier height measurements as well.