Atomic-Scale Tuning of Graphene/Cubic SiC Schottky Junction for Stable Low-Bias Photoelectrochemical Solar-to-Fuel Conversion.

Engineering tunable graphene-semiconductor interfaces while simultaneously preserving the superior properties of graphene is critical to graphene-based devices for electronic, optoelectronic, biomedical, and photoelectrochemical applications. Here, we demonstrate this challenge can be surmounted by constructing an interesting atomic Schottky junction via epitaxial growth of high-quality and uniform graphene on cubic SiC (3C-SiC). By tailoring the graphene layers, the junction structure described herein exhibits an atomic-scale tunable Schottky junction with an inherent built-in electric field, making it a perfect prototype to systematically comprehend interfacial electronic properties and transport mechanisms. As a proof-of-concept study, the atomic-scale-tuned Schottky junction is demonstrated to promote both the separation and transport of charge carriers in a typical photoelectrochemical system for solar-to-fuel conversion under low bias. Simultaneously, the as-grown monolayer graphene with an extremely high conductivity protects the surface of 3C-SiC from photocorrosion and energetically delivers charge carriers to the loaded cocatalyst, achieving a synergetic enhancement of the catalytic stability and efficiency.

Graphene, a single layer of graphite with closely packed conjugated hexagonal lattices, is one of the thinnest materials in the universe. Due to its unique geometric and electronic structures, graphene possesses many intriguing properties, including excellent conductivity, high transparency, great stability, flexibility, and elasticity. These exciting properties endow graphene great potential for applications in electronic, optoelectronic, biomedical, and photoelectrochemical areas.[1,2] Unfortunately, free-standing graphene is rather limited for these applications. For instance, the lack of intrinsic band gap (zero band gap) impedes its implementation into conventional electronics due to substantial leakage currents in field-effect devices and also makes it a poor material for solar light harvesting or hot carrier generation.[3−6] In light of this fact, the large-scale application of graphene is usually associated with the integration of graphene into current semiconductor technology.[7] One of the key factors that determines the performances of graphene-based devices is the quality of the graphene–semiconductor interface, where a Schottky junction is usually formed. Apart from quality, the tunability of such a Schottky junction is also of particular importance to control and gain mechanistic insight into interfacial electronic properties and transport mechanisms for applications in photodetectors, sensors, and solar cells.[7] Besides quality and tunability, another aspect that matters a lot over the graphene–semiconductor junction is the electronic structure of integrated graphene, which, in many cases, is extremely sensitive to its bonding structure and chemical environment. This concern can be exemplified by the unique 100% sp2-hybridized states of graphene, which will be broken if graphene is strongly bonded to adjacent substrates, thus disturbing the π electrons and lowering the intrinsic in-plane conductivity.[8] Particularly, for graphene layers that are produced through chemical exfoliation of graphite or chemical vapor deposition on metals, their superior properties are deteriorated during the required transfer process to semiconductor substrates, often with residues of processing chemicals. Therefore, establishing a highly qualified tunable graphene–semiconductor junction interface, while simultaneously preserving the superior properties of graphene without chemical–structural modification, is the key challenge for the engineering of graphene-based devices.

Multilayer graphene, consisting of two or more layers of graphene, is of particular interest due to its unique interlayer interaction and tunable electronic band structure.[9,10] This is because the addition of each individual graphene layer will modify the integral electronic structure and induce charge transfer density modulations at the interface, thus producing a different material with peculiar properties.[8,11,12] Herein, we report the construction of an atomic-scale tunable graphene/3C-SiC Schottky junction by tailoring the number of graphene layers via an epitaxial growth technique. High-quality and uniform graphene can be in situ grown on 3C-SiC, and the number of graphene layers can be precisely controlled, thus avoiding a conventional graphene transfer process. Meanwhile, the superior properties of graphene, including the high conductivity, transparency, and stability, are well-preserved along with the graphene growth. Most importantly, the formation of an atomic Schottky junction at the graphene/3C-SiC interface, with a tunable barrier and the built-in electric field, enables us to systematically control and comprehend the interfacial electronic properties and transport mechanisms of graphene-based devices. Benefiting from these merits, the atomically sharp graphene/3C-SiC Schottky junction is demonstrated to tackle three great challenges of photoelectrochemical catalysis: the limited light absorption, the undesirable charge recombination, and instability of photoelectrodes.[13−15] As a proof-of-concept study, the junction structure described herein also exhibits a synergetic enhancement of both the stability and the efficiency of photoelectrochemical solar-to-fuel conversion.

Results and Discussion

The epitaxial growth of graphene layers on 3C-SiC was performed by the thermal decomposition (sublimation) process of 3C-SiC(111) substrates (Figure a,b and Figure S1, Supporting Information).[16] A considerable advantage of this technique is that graphene can be in situ produced, without etchant chemicals, on a large area with extremely high quality and uniformity, thus ensuring the high quality of the graphene–semiconductor junction structure. No additional transfer process is required to fabricate electronic devices as graphene is directly grown on the 3C-SiC semiconductor in a controlled manner. Through adjusting the growth temperature and time, large-area monolayer (1L), bilayer (2L), and four-layer (4L) graphene were demonstrated to be readily in situ grown on 3C-SiC substrates in this work. As revealed by cross-sectional observation using high-resolution transmission electron microscope (HRTEM), growth of graphene on 3C-SiC was first associated with the formation of a buffer layer, which is constituted of carbon atoms with a honeycomb structure and is strongly bonded to Si atoms of the 3C-SiC surface (Figure c).[17−19] By adjusting the growth condition (see Methods), the 1L, 2L, and 4L graphene layers can be controllably grown on top of the buffer layer of 3C-SiC (Figure c). Our previous work has demonstrated that 4L graphene shows an ABCA (rhombohedral) stacking sequence.[20] It is reported that 3C-SiC(111) with the ABC stacking along the [111] direction is favorable for producing rhombohedral-stacked multilayer graphene.[11] Note that all graphene samples presented here contain the buffer layer, which is commonly observed in graphene samples grown on Si-face SiC substrates and serves as the precursor for the subsequent growth of monolayer and multilayer graphene. However, the buffer layer does not possess typical electronic properties of graphene due to the strong chemical bonding.[21,22] The cross-sectional HRTEM image shows that the graphene interlayer spacing is around 0.336 nm in the 4L graphene, close to the reported spacing of multilayer graphene.[23]

Figure 1

Synthesis and characterization of graphene/3C-SiC samples. (a) Schematic illustration of the preparation process. (b) Optical image and crystal structure of the yellowish single-crystal 3C-SiC substrate. (c) Cross-sectional HRTEM images of the buffer layer, monolayer (1L), bilayer (2L), and four-layer (4L) graphene samples grown on 3C-SiC. (d) Schematic illustration of low-energy electron diffraction (LEED) measurements. (e) Low-energy electron microscopy (field of view = 5 μm) and (f) LEED images of the as-grown graphene/3C-SiC samples. Two reciprocal unit vectors are presented in the LEED patterns, which indicate one each for 3C-SiC (red/short) and graphene (blue/long). The insets in (e) show the electron reflectivity curves collected from the labeled regions. The number of graphene layers is determined by the number of dips in the electron reflectivity curves.

Low-energy electron microscopy (LEEM) was then employed to specifically characterize the uniformity and the number of graphene layers grown on 3C-SiC. Together with LEEM, low-energy electron diffraction (LEED) could further collect the information on the buffer layer, the quality of graphene layers, as well as the detailed surface structure of these single-crystalline materials according to the diffraction spots of electrons (Figure d). According to LEEM images, the buffer layer, the monolayer, and multilayer graphene were controllably grown on 3C-SiC (Figure e). Corresponding graphene layers were determined by the number of dips in electron reflectivity curves shown in the insets of Figure e.[24] For 3C-SiC with a buffer layer (BL/3C-SiC), monolayer graphene (1L/3C-SiC), bilayer graphene (2L/3C-SiC), and four-layer graphene (4L/3C-SiC), the coverage ratio is 80, 91, 87, and 70%, respectively. The LEED results confirmed the presence of (6√3 × 6√3) R30° diffraction spots of the buffer layer, which did not exhibit any graphene electronic properties due to a covalent chemical bonding between the buffer layer and 3C-SiC.[16,25] Along with graphene growth, sharp and bright (1 × 1) graphene diffraction spots gradually dominated the LEED patterns for 1L, 2L, and 4L graphene samples, confirming a high quality of the as-grown crystalline graphene (Figure e). The Raman results corroborated that the buffer layer and high-quality graphene with different layers had been successfully grown on 3C-SiC substrates (Figure S2, Supporting Information).

Successful growth of high-quality graphene layers on 3C-SiC allowed us to systematically study their interfacial and electronic structures. Density functional theory (DFT) was first adopted to investigate interfacial charge exchange before and after the contact between 3C-SiC and graphene layers. Buffer layer was found to be partially bonded with Si atoms of the 3C-SiC substrate via part of its π electrons, which explained why the buffer layer was not electronically active and could not be considered as graphene (Figure a).[22] Meanwhile, the charge density difference calculation revealed a redistribution of charge density, as reflected by a significant electron flow from 3C-SiC to the buffer layer that led to a formation of a built-in electric field (Figure a).[4,26] The charge transfer modulations induced by the buffer layer at the interface was evidenced by X-ray photoelectron spectroscopy depth profiling analysis (Figure S3, Supporting Information). In addition to the prominent geometric interaction, the buffer layer was found to electronically couple with 3C-SiC according to its density of states (DOS) that is strongly hybridized with that of 3C-SiC within the band gap (Figure S4a, Supporting Information). These strong interactions between 3C-SiC and the buffer layer arising from their atomically sharp discontinuity are beneficial for rapid thermal equilibration, as well as charge transfer.[7,27] Interestingly, DFT calculations further show that the built-in electric field formed between the buffer layer and 3C-SiC can be further extended into the outermost graphene layer in both 1L/3C-SiC and 2L/3C-SiC (Figure b,c and Figure S4b, Supporting Information). This is due to the further donation of electrons from the 3C-SiC/buffer layer interface to the outermost graphene layer. However, without the buffer layer, interfacial exchange between 3C-SiC and graphene layers would be barely possible. This unambiguously indicates that the trapping of electrons by the buffer layer is irreplaceable for the overall charge transfer from 3C-SiC to graphene layers to reach the thermal equilibrium condition.[28] When the number of graphene layers was increased to four, the polarization effect and corresponding built-in electric field between the two outermost graphene layers became extremely weak, indicating a negligible charge transfer to the fourth graphene layer (Figure d).[29] Thus, DFT calculations clearly show that the buffer layer with a short-range effect is critical to mediate electron transfer to the first and second but not to the fourth graphene layer.

Figure 2

Electronic structure of graphene/3C-SiC samples. (a) Charge density difference for buffer layer/3C-SiC. Yellow and blue isosurfaces (0.004 au) represent charge accumulation and depletion in the space with respect to the isolated buffer layer and 3C-SiC. Charge density difference between the two outermost atomic layers of (b) 1L/3C-SiC, (c) 2L/3C-SiC, and (d) 4L/3C-SiC. The yellow isosurface represents charge accumulation in the spatial graphene region in the presence of the 3C-SiC substrate. (e) Electrostatic potential diagrams of graphene and the 3C-SiC surface obtained from DFT calculations. Energy band structures for (f) 1L/3C-SiC, (g) 2L/3C-SiC, and (h) 4L/3C-SiC. (i) Schottky barrier height of the graphene/3C-SiC junction and the difference of electrostatic potential energy (VElectrostatic) between the two outermost atomic graphene as a function of the total number of graphene layers. (j) Schematic illustration of the excitation and charge transport process under solar light irradiation.

The electron exchange between 3C-SiC and graphene would result in the formation of an atomic Schottky junction with a built-in electric field between 3C-SiC and graphene layers.[30] To prove this point, we calculated the work functions of 3C-SiC and free-standing graphene through DFT. According to the potential lineups and the distance between the vacuum level and the Fermi level, free-standing graphene displays an average work function of 4.52 eV (Figure e), much higher than that of the 3C-SiC(111) surface (4.11 eV). An energy band diagram for free-standing graphene and 3C-SiC prior to thermal equilibrium was proposed, where EF was the Fermi energy, EC the conduction band edge, and EV the valence band edge (Figure S5a, Supporting Information). Theoretically, due to the work function of 3C-SiC being lower than that of graphene, electrons are expected to flow from 3C-SiC to graphene until their Fermi levels are aligned at equilibrium upon contact, leading to the formation of a Schottky junction with a built-in electric field directed from 3C-SiC to graphene, as obtained by our DFT calculations (Figure S5b, Supporting Information).

To check our interpretation of a Schottky junction formation, angle-resolved photoelectron spectroscopy (ARPES) was then employed to measure the energy band structure and Fermi levels of graphene samples. For the free-standing graphene, the EF of graphene is expected to be located to its Dirac point (ED). Due to electron transfer to graphene, it gave rise to a high electron concentration in the graphene and lifted the EF above the Dirac point.[20,30,31] As seen in Figure f,g, the ARPES results reveal that the EF is located above the Dirac point. Relative to the Dirac point, an energy difference (EF – ED) of 0.42 and 0.34 eV was observed for 1L and 2L graphene, respectively. For the 4L graphene, EF is almost at the Dirac point with an EF – ED value falling into the ARPES resolution limit of 0.05 eV (Figure h). The decreased EF – ED value of the 4L graphene is consistent with the drastically decreased electron transfer to the fourth graphene layer according to the DFT calculation, which reflects a much lowered charge carrier concentration on the 4L graphene. Therefore, by increasing the number of graphene layers, the electron concentration in outer-layer graphene is decreasing and EF is gradually shifted toward to the Dirac point, which indicates an increased work function. Meanwhile, the width of the flat band at EF is measured to be ∼0.07 Å–1, which is quite close to the calculated value of the flat band region of 0.06 Å–1 for rhombohedral 4L graphene with a ABCA stacking sequence (Figure S6, Supporting Information).[20,32]

To determine the work function of the as-grown graphene (ΦG) of different layers, ultraviolet photoemission spectroscopy (UPS) was employed. According to the UPS results shown in Figure S7a, the work functions of 1L, 2L, and 4L graphene were calculated to be 4.13, 4.27, and 4.48 eV, respectively (Figure f–h). Thus, the work function of graphene increases with increasing the number of the graphene layers. Notably, the work function (4.48 eV) of 4L graphene is very close to the intrinsic value (4.52 eV) of free-standing graphene, consistent with the ARPES result showing that its EF is almost located at the Dirac point (Figure h and Figure S7b, Supporting Information). This can be explained by the following. As shown by the DFT calculations above, the buffer layer plays a key role in transferring electrons to the first and second graphene layer but not to the fourth graphene layer. The transferred charges should be distributed among the graphene layers with a reduced charge density toward the outer layers, while a constant Fermi level should be kept for the whole system. The fourth graphene layer is in an almost neutral state due to the negligible charge transfer from its underneath layers, which is the very reason why its Fermi level is close to the intrinsic Fermi level at the Dirac point. To quantify the Schottky junction, we calculated the Schottky barrier height (φS) by the equation of ΦS = ΦG – χSiC, in which χSiC is the electron affinity of 3C-SiC (χSiC = 4.00 eV).[33] Based on the measured work functions, the Schottky barrier heights of the 1L/3C-SiC, 2L/3C-SiC, and 4L/3C-SiC junctions were calculated to be 0.13, 0.27, and 0.48 eV (Figure f–h), respectively. This demonstrates that the Schottky barrier height increased with the increase of stacking graphene layers.

As mentioned above, another important parameter of the Schottky junction is the built-in electric field, . We define the absolute value of between the two outermost atomic layers by the equation , in which V is the energy difference of the electrostatic potential between the two outermost atomic layers. According to the calculated electrostatic potential energy diagram, the value of V between 3C-SiC and the buffer layer reached the maximum due to their strong chemical interaction (Figure i and Figure S8a, Supporting Information). The value of V between buffer layer and 1L graphene decreased to 2.02 eV, and the value further decreased to 0.59 eV for V between 1L and 2L graphene (Figure S8b,c, Supporting Information). For 4L/3C-SiC, the energy difference of electrostatic potential V between the third and fourth graphene layers was only 0.17 eV (Figure S8d, Supporting Information), which was negligible compared with the V between buffer layer and 1L graphene. The weakened built-in electric field along with multilayer graphene stacking could also be reflected in the cross-sectional HRTEM image of 4L/3C-SiC. Between the buffer layer and 1L graphene, a slight expansion of the lattice spacing up to 0.362 nm is observed due to the strong interaction between the buffer layer and the 3C-SiC substrate (Figure S9, Supporting Information). However, above the 1L graphene, interlayer lattice spacing (0.336 nm) is close to the theoretical value (0.340 nm), which is consistent with the weak interfacial built-in electric field. As a result, along with multilayer graphene growth, the built-in electric field between graphene layers continuously decreased until it almost vanished at the fourth graphene layer, whereas the Schottky barrier height of the graphene/3C-SiC junctions gradually increased (Figure i).

In principle, the graphene/3C-SiC structures with a tunable Schottky junction and electric field will allow us to systematically understand the carrier separation and interfacial charge transport behaviors, both of which are critical to the efficiency of photoelectrochemical (PEC) cells for solar-to-fuel conversion. This assumption is first supported by the small band gap (Eg = 2.36 eV) of 3C-SiC for sufficient visible sunlight absorption in comparison with that of 6H-SiC (Eg = 3.03 eV) and 4H-SiC (Eg = 3.26 eV). Also, 3C-SiC possesses suitable energy band positions that straddle the redox potential of water splitting (Figure S10, Supporting Information). Meanwhile, high transparency (∼97%) of graphene ensures the transmission of almost all solar light, and high stability of graphene may also serve as a protective material to prevent the SiC surface from corrosion or destruction. Suppose that 3C-SiC is excited by solar light, electrons will be excited to the EC, leaving holes in the EV. The photogenerated holes can be readily swept into graphene, whereas electrons are injected toward the backside of 3C-SiC by the built-in electric field (Figure j). For holes in graphene, their in-plane mobility can be considered as barrierless due to the high conductivity of graphene. Meanwhile, the directional built-in electric field may also enable an accelerated out-of-plane transport route for photogenerated holes, which is usually considered as sluggish due to the van der Waals forces between graphene layers (Figure j). The out-of-plane transport of photogenerated holes is of considerable importance as holes are apt to diffuse out of graphene for a surface chemical reaction. It is also to be noted that the in situ growth of graphene layers without any transfer process preserved the geometric and electronic structures of graphene, including its high conductivity, and the high conductivity of graphene in our case can potentially act as an ideal platform to enable reliable cocatalyst loading for more efficient interfacial charge transfer.

To test our hypothesis, we used the as-grown graphene/3C-SiC samples to construct photoanodes by directly depositing ohmic contacts. A 300 nm thick Al was deposited on the backside of the single-crystal 3C-SiC, and a layer of Au (200 nm)/Ti (5 nm) was deposited on the backside of graphene/3C-SiC samples. Both of them form ohmic contact, showing a linear I–V curve (Figure a and Figure S11, Supporting Information).[34] In a two-compartment PEC cell separated by the Nafion membrane, the photoanodes can be used for water oxidation in combination with CO2 on a commercial metal cathode (Cu cathode) for the production of solar fuels (Figure b).[35] A 0.5 M pH 7.5 KHCO3 solution was used as the electrolyte in our case. The anode compartment was continuously bubbled with Ar gas, while the cathode compartment was bubbled with CO2 of high purity. The PEC water oxidation on the as-prepared photoanodes was first evaluated, and the photoanodes showed a negligible dark current in the linear sweep voltammetry scan. Under simulated 1 sun irradiation (AM1.5G, 100 mW/cm2), 3C-SiC showed an onset potential of around 0.20 V versus the reversible hydrogen electrode (VRHE) and a sluggish photocurrent increase around 0.40 VRHE in the J–V curve (Figure c). After the growth of the buffer layer, BL/3C-SiC showed an onset potential similar to that of 3C-SiC with a slightly increased photocurrent (Figure c). As the buffer layer strongly interacted with 3C-SiC through chemical bonds, it is electronically inactive in terms of the high conductivity of graphene, which could explain why the BL/3C-SiC photoanode only showed a slightly increased photocurrent.[22] Interestingly, the 1L/3C-SiC photoanode exhibited an almost zero onset potential of water oxidation, and a precipitous photocurrent increase occurred at as low as 0.20 VRHE, a 0.20 V cathodic shift relative to the 3C-SiC or BL/3C-SiC photoanode (Figure c). Meanwhile, the saturated photocurrent density around 450 μA/cm2 with a plateau appeared as low as 0.60 VRHE for the 1L/3C-SiC photoanode. However, further increase of the graphene layers to 2L and 4L resulted in a photocurrent drop with increased onset potential (Figure c). According to the ARPES results and measured work functions, 2L/3C-SiC and 4L/3C-SiC possessed a Schottky barrier height higher than that of 1L/3C-SiC, which would theoretically facilitate the carrier separation. However, the aforementioned calculations demonstrated that the further increase of the graphene layers resulted in a decreased electric field between the outermost graphene layers. This decreased electric field would hinder the out-of-plane hole transport across the multiple graphene layer, which could explain the drop in photocurrent response for the 2L/3C-SiC and 4L/3C-SiC photoanodes.

Figure 3

Photoelectrochemical performance of the graphene/3C-SiC photoanodes. (a) Schematic illustration of the photoanode structure and a representative I-V curve of the Ohmic contact. (b) Schematic illustration of a two-compartment PEC cell separated by the Nafion membrane for water oxidation and CO2 reduction. (c) Current density–voltage (J–V) curves of the graphene/3C-SiC photoanodes in the PEC cell under illumination and (d) corresponding applied bias photon-to-current conversion efficiency. (e) Open-circuit potentials of the graphene/3C-SiC photoanodes in the dark and light. (f) J–V curves of the graphene/3C-SiC photoanodes loaded with FeOOH cocatalyst under illumination. Inset shows the scanning electron microscopy image of the FeOOH nanorods. Light source: simulated solar light (AM1.5G, 100 mW/cm2). Electrolyte: 0.5 M pH 7.5 KHCO3 solution. Counter electrode: Cu.

The reduced onset potential of the 1L/3C-SiC photoanode suggested a reduced overpotential for water oxidation, as further evidenced by applied bias photon-to-current efficiency (ABPE) (Figure d). The maximum ABPE around 0.26% was achieved at an extremely low bias of 0.08 V for the 1L/3C-SiC photoanode under simulated solar light, whereas it required a higher applied bias of 0.21 V to reach a maximum ABPE of 0.1% for the 3C-SiC photoanode. The increased photocurrent response of 1L/3C-SiC was consistent with its highest photovoltage (1.08 V), almost 2 times that of 3C-SiC (0.54 V) (Figure e). Further increase of the graphene layers to 2L and 4L resulted in a decreased photovoltage, agreeing well with the lowered photocurrent. To observe the shift of the flat band potential of 3C-SiC, we measured the Mott–Schottky plots of the as-prepared graphene/3C-SiC junctions at a frequency of 1 kHz. For 3C-SiC, the positive slope is consistent with that of typical n-type semiconductors, and according to the intercept of the linear region, the flat band potential is around −0.20 VRHE (Figure S12a, Supporting Information). When the buffer layer was incorporated, the flat band potential of 3C-SiC was slightly shifted from −0.20 to −0.21 VRHE. Interestingly, the incorporation of 1L, 2L, and 4L graphene on 3C-SiC showed an identical and more negatively shifted flat band potential at −0.24 VRHE (Figure S12a, Supporting Information). The Mott–Schottky plots display comparable slopes due to the identical doping concentration around 7.5 × 1015 cm–3 in 3C-SiC. The identical flat band potential of 1L/3C-SiC, 2L/3C-SiC, and 4L/SiC suggests that this negative flat band shift is possibly due to the passivation of surface defects by graphene layers.[36] Some surface states cause a potential drop within the Helmholtz layer that can be minimized through surface passivation.[37]

We also measured the flat band potential of 3C-SiC at a higher frequency. The flat band potential measuring a higher frequency of 5 kHz was estimated to be −0.21 VRHE, close to that measured under 1 kHz. Moreover, a similar cathodic shift of the flat band potential was also observed after the incorporation of 1L graphene (Figure S12b, Supporting Information). In comparison with 3C-SiC and BL/3C-SiC, the higher flat band potential of 1L/3C-SiC is consistent with the tendency of its higher photovoltage. However, based on the flat potential alone, we cannot explain the higher photovoltage of 1L/3C-SiC and the lowered photovoltage of 2L/3C-SiC and 4L/3C-SiC. This, on the other hand, indicates that the large photovoltage generated by 1L/3C-SiC not only is due to more negative flat band potential but also is closely related to the desirable built-in electric field. It is worth noting that, for the open-circuit potential (OCP) measurements, O2 gas was continuously bubbled into the electrolyte (0.5 M pH 7.5 KHCO3), and each dark/light OCP was obtained with constant stirring. As expected, the dark OCP values are close to ∼1.23 VRHE. Under light excitation, the photogenerated carriers result in the photovoltage under the open-circuit condition. For the 1L/3C-SiC, the monolayer graphene not only suppressed the surface recombination by passivating the surface defects of 3C-SiC but also formed an optimal Schottky junction at the interface of 1L/3C-SiC, thus enhancing the separation of photogenerated carriers. Therefore, the 1L/3C-SiC photoanode delivered the highest photovoltage. For 3C-SiC photoanodes with thicker graphene layers, the reduced interlayer electric field hinders the out-of-plane hole transport toward the electrolyte, thus giving rise to a lower photovoltage.

To study the origin of the enhanced PEC water oxidation performance of 1L/3C-SiC, the electrochemical impedance spectroscopy (EIS) was used to investigate charge transfer resistance. The Nyquist plots of 3C-SiC and graphene/3C-SiC junctions were fitted using the equivalent circuits, as shown in the inset of Figure S13. Under light excitation, the equivalent circuit consists of series resistance (Rs), charge transfer resistance from the bulk to the photoanode surface (Rct), and the charge transfer resistance from the photoanode to the electrolyte (Rct,trap). The value of Rct decreased from 177 to 115 Ω cm2 when 3C-SiC was grown with 1L graphene, which demonstrates that the 1L/3C-SiC Schottky junction reduced the resistance for charge transfer from 3C-SiC to 1L graphene (Table S1, Supporting Information). Meanwhile, Rct,trap of 1L/3C-SiC (649 Ω cm2) was also decreased compared with that of 3C-SiC (1719 Ω cm2), suggesting a much smaller charge transfer resistance from the 1L/3C-SiC photoanode to electrolyte for water oxidation. For 2L/3C-SiC and 4L/3C-SiC, their much larger Rct and Rct,trap values were consistent with the difficult charge transfer ability across the multiple graphene layers, as we demonstrated above (Table S1, Supporting Information).

Therefore, it was concluded that the formation of an atomic Schottky junction between graphene and 3C-SiC played an indispensable role in improving water oxidation efficiency by accelerating charge separation. However, the addition of graphene layers would reduce the out-of-plane charge transport toward an electrolyte for water oxidation due to the decreased built-in electric field between the graphene layers. Thus, tunability of the atomic Schottky junction enables a trade-off balance between Schottky barrier height and the electric field distributed across graphene layers, which can be balanced and optimized to simultaneously enhance electron–hole separation and the carrier transport across graphene layers. As the as-prepared 3C-SiC in our study is n-type conductive, a complementary evaluation of the hole transport properties (particularly in the out-of-plane direction across the graphene layers) through Hall effect measurements will be obtained when the semi-insulating 3C-SiC substrate can be prepared in the future, which is one of our research goals for the design of other 2D/semiconductor heterostructures.

As we discussed above, the preserved ultrahigh in-plane conductivity of graphene could act as an ideal platform to enable efficient charge transfer into the loaded cocatalysts for an improved PEC efficiency. This notion is demonstrated by the deposition of the traditional oxygen evolution cocatalyst, FeOOH. Through a facile hydrothermal method, vertical FeOOH nanorods with a thickness around 200 nm could be deposited on the 1L/3C-SiC photoanode according to the scanning electron microscopy and TEM images (Figure S14a–c, Supporting Information). Combined X-ray diffraction (XRD), TEM, and X-ray photoelectron spectroscopy (XPS) studies revealed the deposition of as-grown FeOOH nanorods on the 1L/3C-SiC photoanode (Figure S14d–f, Supporting Information). According to the J–V curve, loading the layer of FeOOH nanorods on pristine 3C-SiC or BL/3C-SiC photoanode (FeOOH/BL/3C-SiC) with a buffer layer did not significantly improve the photocurrent density (Figure f). This is probably due to the sluggish hole transfer from the electronically inactive buffer layer to FeOOH. For the 1L/3C-SiC with a layer of highly conductive monolayer graphene, loading of FeOOH (FeOOH/1L/3C-SiC) could result in an over 1-fold increase of the photocurrent up to 1.13 mA/cm2 (Figure f). According to the Mott–Schottky plots, the loading of FeOOH caused only a slight cathodic shift of the flat band potential of 1L/3C-SiC (from −0.24 to −0.25 VRHE), suggesting that the electronic structure, including the Schottky barrier at the junction, was not influenced by FeOOH loading (Figure S12c, Supporting Information). Therefore, the FeOOH mainly functioned as an efficient oxygen evolution reaction cocatalyst to enhance the kinetics of water oxidation on the 1L/3C-SiC photoanode according to the sharper photocurrent increase at relatively lower applied potential (Figure f). As further revealed by fitted impedance arc radius of FeOOH/1L/3C-SiC, Rct and Rct,trap values decreased to 51 and 179 Ω cm2, respectively (Table S1, Supporting Information). This suggests FeOOH on the conductive graphene layer not only facilitated bulk charge carrier separation but also decreased the charge transfer barrier for water oxidation at the electrode interface. Using CoOOH as another oxygen evolution reaction cocatalyst, we further demonstrated that the PEC performance of 3C-SiC could be “universally” improved with the incorporation of the graphene interlayer (Figure S15, Supporting Information). It has to be pointed out that the maximum photocurrent density of 3C-SiC with cocatalysts is still much lower than the theoretical value calculated from its band gap. This is mainly related to the fact that the indirect band gap of 3C-SiC results in a much larger light penetration depth than the sum of the width of space charge region (W) and the carrier diffusion length (LD). Consequently, most photogenerated carriers are distributed in a neutral region and recombine there, and only a small part of photogenerated carriers within the region of (W + LD) can contribute to the photocurrent (see details in Figure S10d, Supporting Information). Moreover, our previous work has reported the presence of defects such as stacking faults and double positioning boundaries in the 3C-SiC bulk that can act as bulk charge recombination centers.[38]

Online O2 detection via gas chromatography (GC) revealed that the generated photocurrent on 3C-SiC was not solely associated with O2 production under a bias of 0.6 VRHE due to self-oxidation by holes (3C-SiC + 4H2O + 8h+ → SiO2 + CO2 + 8H+) (Figure a,b).[39,40] Interestingly, the growth of buffer layer and 1L graphene not only enhanced the water oxidation efficiency of 3C-SiC but also improved the faradaic efficiency of the photoanode from 72% (3C-SiC) to 96% (1L/3C-SiC), as shown in Figure a,b. The enhanced faradaic efficiency and the improved stability of 1L/3C-SiC according to the long-term J–t curve were due to the antioxidation behavior of graphene that prevented the 3C-SiC photoanode from photocorrosion (Figure c).[40] For the FeOOH-loaded 1L/3C-SiC photoanode, stability could still be maintained (Figure c). As expected from the fact that graphene could protect 3C-SiC from photocorrosion, the 2L/3C-SiC and 4L/3C-SiC with multilayer graphene also displayed good stability (∼98% faradaic efficiency). However, the amount of generated O2 on the 2L/3C-SiC and 4L/3C-SiC photoanodes was low due to their poor photocurrent response (Figure b).

Figure 4

Photoelectrochemical CO2 reduction in combination with water oxidation on the graphene/3C-SiC photoanodes. (a) Detected O2 production over the photoanodes at a bias of 0.6 VRHE. (b) Steady-state photocurrent and faradaic efficiency for water oxidation over the photoanodes at a bias of 0.6 VRHE. (c) Comparison of the J–t behavior of the 3C-SiC, 1L/3C-SiC, and FeOOH-coated 1L/3C-SiC photoanodes at a bias of 0.6 VRHE. (d) Generation of CH4 on Cu cathode. (e) Gas chromatograms of CH4 evolved from isotopic 13CO2 as carbon source using FeOOH/1L/3C-SiC as the photoanode and Cu as the cathode (results using 12CO2 as carbon source are given as a comparison). (f) Schematic illustration of the proof-of-concept PEC cell for a selective conversion of CO2 into solar fuels using the FeOOH/1L/3C-SiC photoanode. Light source: simulated solar light (AM1.5G, 100 mW/cm2). Electrolyte: 0.5 M pH 7.5 KHCO3 solution. Error bars represent standard deviation based on five independent experiments.

In the PEC system with 3C-SiC as the photoanode and Cu as the cathode, the reduction products in the cathode compartment under a saturated CO2 atmosphere were predominantly H2 (63.0%) with CH4 (13.6%) as the additional product under a bias of 0.6 VRHE (Table S2, Supporting Information). Interestingly, when 1L/3C-SiC was used as the photoanode, CH4 (60.5%) became as the dominant product under the same bias. Selectivity of CO2 reduction to CH4 could be further increased to 82.8% when FeOOH-loaded 1L/3C-SiC was used as the photoanode (Table S2, Supporting Information). The faradaic efficiency of 82.8% for CH4 in the present work is the highest ever in photoelectrochemical CO2 reduction in comparison with previous works on CO2 reduction with CH4 as the product and among the highest in the literature for electrochemical CO2 reduction (Table S3, Supporting Information). The observed high selectivity for the rapid CH4 formation in our PEC system indicated that multiple-electron CO2 reduction process (CO2 + 8H+ + 8e– → CH4 + 2H2O) would be favored using the 1L/3C-SiC and FeOOH/1L/3C-SiC photoanodes that could achieve large photovoltages and photocurrents (Figure d and Figure S16, Supporting Information).[41,42] When the FeOOH/1L/3C-SiC photoanode was used with the Pt as the counter electrode, dominant reduction product was H2 (91.2%) with CO (7.5%) as the byproduct, suggesting that the tuned graphene/3C-SiC photoanode can also be used for solar-to-hydrogen conversion. The selectivity of the CO2 reduction on different metal anodes has been well-studied.[43] This work, however, focuses on the atomic tuning of the graphene/3C-SiC Schottky junction for promoting the charge separation and transport. As a proof-of-concept, such an atomically tuned junction shows a promising PEC performance for solar-to-fuel conversion. We further measured potential between the working electrode (EWE) and counter electrode (ECE) (|EWE – ECE|) during CO2 reduction. Under a bias of +0.6 VRHE, FeOOH/1L/3C-SiC photoanode-driven PEC with Cu as the cathode has a |EWE – ECE| of 1.43 V. For comparison, in the PEC CO2 reduction system using CuO as the photocathode and Pt as the anode, the |EWE – ECE| was reported to be around 1.80 V at a low potential of +0.07 VRHE. Therefore, photoanode-driven CO2 reduction based on the FeOOH/1L/3C-SiC photoanode is very promising as it requires less external energy input.[41,44] When isotopic 13CO2 was used as the carbon source, 13CH4 with a mass/charge ratio (m/z) of 17 was detected as the only product in the FeOOH/1L/3C-SiC photoanode||Cu cathode PEC system (Figure e). Meanwhile, the identical abundance of the corresponding fragments of 13CH4 and 12CH4 further indicated that photocorrosion of 3C-SiC was suppressed by protective graphene layers.[45,46] In addition to the Cu cathode, when a commercial Zn cathode or a Bi cathode was coupled with the FeOOH/1L/3C-SiC photoanode, CO and formic acid emerged as the respective selective CO2 reduction products (Figure f and Table S2, Supporting Information). This proof-of-concept PEC cell revealed the versatility of FeOOH/1L/3C-SiC as the efficient photoanode to drive a selective conversion of CO2 into desirable solar fuels.

Conclusions

In conclusion, we present an interesting atomic Schottky junction between epitaxial graphene layers and 3C-SiC substrates. By tailoring the number of graphene layers, this junction structure exhibits an atomic-scale-tunable Schottky junction with an inherent built-in electric field distributed across up to the fourth graphene layer. As a proof-of-concept study, the atomic-scale-tuned Schottky junction is demonstrated to promote both the separation and transport of charge carriers under solar light, thus enhancing the overall photoelectrochemical performance of the solar-to-fuel conversion, including increased photocurrent and photovoltage together with a reduced onset potential. Simultaneously, the as-grown monolayer graphene with an extremely high conductivity protects the surface of 3C-SiC from photocorrosion and energetically delivers charge carriers to the loaded cocatalyst, resulting in a synergetic enhancement of both the stability and the efficiency of this system. The knowledge gained in this study not only highlights the paramount importance of sophisticated interfacial structure design for construction of feasible renewable energy systems but also illuminates the engineering of graphene-based devices.

Methods

Preparation of the 3C-SiC Substrate

Through a sublimation process, 3C-SiC(111) with a thickness around 1 mm was grown on a 4° off-axis n-type 4H-SiC (SiCrystal).[38,47] To obtain single-crystal 3C-SiC with a thickness around 300 μm, the 4H-SiC substrate and transition interface between the two polytypes were carefully polished away. After that, the 3C-SiC(111) substrate was thoroughly cleaned with acetone, ethanol, H2O/NH3/H2O2 (5:1:1), H2O/HCl/H2O2 (6:1:1), and hydrofluoric acid. Prior to graphene growth, the obtained clean 3C-SiC(111) substrate was checked by optical microscopy and atomic force microscopy to ensure there were no macroscopic defects. Doping concentration (ND) of the n-type 3C-SiC was measured to be around 7.5 × 1015 cm–3.

Preparation of the Graphene/3C-SiC Junction

An inductively heated furnace was used to grow graphene on the 3C-SiC substrate under 1800 °C in an 850 mbar argon atmosphere for different times.[16] The ramping rate was set at 25 °C/min. For the preparation of BL/3C-SiC, 1L/3C-SiC, and 2L/3C-SiC, the 3C-SiC was annealed at 1800 °C for 1, 15, and 30 min, respectively. The thicker graphene of more than two layers can be grown by tuning the growth temperature and pressure. As it was very challenging to grow thick (>3L) graphene on the Si-face SiC substrate, we focused on how to grow graphene with four layers. However, we believe the effect of 1L, 2L, and 4L graphene on charge transfer as we discussed above should be applicable to the 3L graphene. In this work, four-layer graphene, as an example, was grown at a higher temperature of 2000 °C in an 850 mbar argon atmosphere for 30 min. For FeOOH nanorod deposition, graphene/3C-SiC samples were immersed in a precursor solution containing 30 mM FeCl3 and 45 mM urea.[48] The urea was used as the progressive OH– releasing agent. The above mixture was then kept for 30 min at 100 °C to enable rapid FeOOH deposition. After the deposition, the coated photoanodes were rinsed with DI water and then nitrogen-dried. Sheet resistance of the used FTO glass to support the Cu tape and graphene/3C-SiC samples is around 8 ohms per square. For CoOOH nanoparticle deposition, 3C-SiC was used as the working electrode, Ag/AgCl as the reference electrode, and Pt mesh as the counter electrode in a typical three-electrode cell. Next, 0.6 VRHE was applied in 0.01 mol/L CoCl2 aqueous solution under simulated solar light irradiation (AM1.5G, 100 mW/cm2) with a deposition time of 30 min.[49] Because the 1L/3C-SiC photoanode possessed a photocurrent density much higher than that of the 3C-SiC photoanode under the same bias, the CoOOH deposition time was shortened accordingly to ensure the same accumulated charge passing through the photoanodes. This could ensure the comparable amount of CoOOH deposited on the 1L/3C-SiC photoanode.

DFT Theoretical Calculations

Theoretical calculations were performed using DFT with the exchange-correlation energy functional implemented in the Vienna Ab initio Simulation Package, as described by the generalized gradient approximation with the Perdew–Burke–Ernzerhof exchange–correlation function.[50] To describe electron–ion interactions, the projector-augmented wave method was applied with a plane-wave cutoff energy of 400 eV.[50,51] For geometric optimizations and DOS calculations, the energy and force converged to 10–5 eV atom–1 and 0.02 eV Å–1, and the corresponding K-point was 3 × 2 × 1. Si-terminated 3C-SiC modified with graphene heterojunctions was established according to the suitable lattice match, evidenced by 30° offset in two lattice hexagons. Multilayer graphene was added on the basis of graphene growth. The simulative supercells (3 × 3 × 1) with a vacuum thickness of 20 Å were denoted as Si-terminated 3C-SiC modified with graphene heterojunctions. The charge density difference was calculated as Δρ = ρ(graphene/3C-SiC) – ρ(3C-SiC) – ρ(graphene). The ρ(graphene/3C-SiC) was denoted as the density of heterojunctions, whereas ρ(3C-SiC) and ρ(graphene) were denoted as the densities of the two subsystems. The electrostatic potential and work function calculation were performed with CASTEP, in which the plane-wave pseudopotential with an energy cutoff of 310 eV was employed for all calculation systems, and the corresponding energy and force convergence were 10–5 eV atom–1 and 0.02 eV Å–1, respectively.

Characterization of Graphene/3C-SiC Junctions

Atomic graphene layers were first determined by cross-sectional HRTEM (FEI Tecnai G2 TF 20 UT) operated at 200 keV. The number of multiple graphene layers, quality, and the uniformity of the graphene layers were determined by LEEM and μ-LEED in ultrahigh vacuum, which were carried out using a SPELEEM instrument at beamline I311 in the MAX IV Laboratory synchrotron radiation laboratory, Lund, Sweden. The μ-LEED patterns were obtained from the selected areas (range: 500–1000 nm), matching the size of the corresponding graphene domain. Room temperature ARPES data were recorded in an ultrahigh vacuum system at Linköping University, Sweden, using a Phoibos 100 analyzer from Specs, equipped with a two-dimensional detector and HeI radiation (hν = 21.22 eV) from a non-monochromatized resonance lamp. Energy and angular resolutions were 50 meV and 0.3°, respectively. The position of the Fermi level was determined using spectra recorded from a clean Ta foil in electrical contact with the samples. Work function data were derived from the position of the cutoff of the secondary electron emission. A negative bias around −9.526 V was applied to the samples relative to ground. This results in an increase of the kinetic energy of the emitted electrons, which is necessary in order to obtain well-defined and steep edges in the cutoff spectra. Surface morphology was determined by scanning electron microscopy (LEO 1550 Gemini, 10 kV). XRD measurements were conducted using a Philips MRD with Cu Kα1 (λ = 1.54 Å).

Photoelectrochemical Water Oxidation and CO2 Reduction

The designed photoelectrochemical reactor was made of a photoanode and a cathode compartment, which were separated by a Nafion-115 proton exchange membrane. We used 0.5 M pH 7.5 KHCO3 solution as the electrolyte, and the photoanode compartment was bubbled by high-purity Ar and cathode compartment by high-purity CO2. Graphene/3C-SiC, selected metal, and Ag/AgCl were used, respectively, as the photoanode, cathode, and reference electrode. The photoanode and cathode compartment were connected to a gas circulation system for online sampling and analysis through a gas chromatograph system (GC: Agilent HP 6890). The GC system has a Carboxen-1000 packing column to separate and analyze gaseous mixtures of O2, CO, CH4, C2H4, and H2. Liquid product of formic acid was determined by an ionic chromatograph (Thermo Scientific Dionex, ICS-900). Both the electrolyte and the circulation system were purged with CO2 for 60 min before the photoelectrochemical catalysis. Particularly, in the cathode compartment, electrolyte was pre-electrolyzed at 0.025 mA using Pt mesh for 120 min under CO2 purging to remove the metal ion impurities and to avoid possible cathode poisoning. The whole photoelectrochemical process was conducted under atmospheric pressure of CO2. The bias was applied using a using a potentiostat/EIS (Princeton Applied Research, VersaSTAT 3). All potentials in this work are with respect to Ag/AgCl and converted to potentials versus reversible hydrogen electrode (RHE) according to the equation: VRHE = VAg/AgCl + 0.197 + 0.059 × pH. Faradaic efficiencies for the photoelectrochemical reactions were calculated based on the utilized charge and the number of holes or electrons required for the formation of certain products. For an isotopic labeling experiment, 13CH4 was collected and analyzed via a gas chromatograph–mass spectrometer (Pfeiffer OmniStar) using 13CO2 (13C 99%, Sigma-Aldrich) as the carbon source. For open-circuit potential measurements, O2 gas was continuously bubbled into the electrolyte (0.5 M pH 7.5 KHCO3), and each dark/light open-circuit potential was obtained with constant stirring.