Graphene (G), an atomically thick two-dimensional nanocarbon material, has attracted increasing attention in the development of next-generation photovoltaic devices due to its excellent electrical, optical, and chemical stable properties.^[1–6] In particular, solar cells that combine crystalline silicon (Si) with graphene synthesized by the chemical vapor deposition (CVD) method have shown promising potential because of the simple and low-cost fabrication process.^[7–9] In a typical G/Si heterojunction solar cell, the graphene film not only participates in the formation of the heterojunction, but also serves as the window electrode and hole collector.^[9–11] When the device is illuminated, the incident light passes through the graphene and creates electron–hole pairs in the Si substrate.^[12] The generated electron–hole pairs are then separated by the built-in electric field. Afterwards, the separated holes diffuse across the graphene film in the lateral direction to be collected. Hence, good transparency and conductivity of the graphene film are of high importance to obtain the expected solar cell with high power conversion efficiency (PCE).

However, due to the polycrystalline feature of CVD-graphene and cracks or residues formed in the polymer-assisted transfer processes, the obtained graphene films commonly show relatively high sheet resistance in the range of several hundreds to thousands of Ω/sq.^[13–18] When applied in the Si solar cells, the high sheet resistance of graphene not only causes severely charged carrier recombination, but also brings about serious power loss. For instance, Ho et al.^[15] reported that the relatively low conductivity of graphene (990 Ω/sq) leads to a low fill factor (FF) of the G/Si solar cell, which results in a low PCE of 3.3%. Similarly, a G/Si solar cell with a PCE of 3.47% was also reported based on graphene with sheet resistance of 1500 Ω/sq.^[16] Chemical doping of graphene can yield enhanced FF and open circuit voltage of the G/Si solar cells.^[11,14,19] However, the doped solar cells face the problem of fast degradation,^[9] which greatly influences their developments.

Recently, many studies have focused on the assembly of graphene with carbon nanotube (CNT), which yields new types of nanocarbon materials with diverse structure characteristics and different properties.^[20–24] On the one hand, high strength one-dimensional CNTs can reinforce the atomically thin graphene film, which is conductive to the polymer-free transfer process.^[21] On the other hand, the hybrid films of graphene and CNTs have shown better conductivity compared to an individual graphene or CNT film, making them more adoptable for flexible transparent conductive electrodes.^[18,22,23] Apart from the hybrid films, the native oxides on Si have also been proved to greatly affect the characteristics of as fabricated heterojunction solar cells by controlling the magnitude of the thermionic emission dark current.^[14,25] However, the optimal oxides for different heterojunctions require different thicknesses.^[26–28] Our previous report has also shown that, in a PEDOT:PSS-CNT/Si solar cell, the optimal oxide thickness between PEDOT:PSS/Si heterojunction is thinner compared to that of CNTs/Si, which brings about the difficulty in the design of an oxide layer to fulfill the needs for both PEDOT:PSS and CNT based Si solar cells.^[29] Considering the similar sp² hybridization structure and properties of graphene and CNTs, their optimal oxide thicknesses are almost the same. These features provide a lot of convenience for us to construct the high efficiency Si solar cells based on graphene and CNTs.

In this work, we improved the G/Si solar cell efficiency through hybridizing graphene with CNTs and investigated the performance of the hybrid films with diverse assembly configurations in Si solar cells. Specifically, three types of hybrid films based on graphene and CNT network were assembled, including G/CNT and CNT/G hybrid films, which were fabricated by simple stacking CNT network on graphene, as well as a coplanar nanostructure (G-CNT film) synthesized by a chemical vapor deposition (CVD) method. We found that the solar cell performance depends on the nanocarbon/Si heterojunction. The G-CNT/Si solar cell, in which there are both G/Si and CNT/Si contact heterojunctions, exhibits better J–V characteristics than the ones with a single type of contact. Its high efficiency (PCE = 7.9%) is attributed to the synergistic effect of graphene and CNTs. After applying a PMMA antireflection coating, the PCE of the as-fabricated PMMA/G-CNT/Si solar cell can be further improved to 9.1%. These findings illustrate the dependence of the Si solar cell performance on different nanocarbon materials and provide a potential way to further enhance the efficiency of this type of solar cells.

Figure 1 shows the schematic fabrication processes of the three different kinds of hybrid films based on graphene and CNTs. The different fabrication methods determined the resulting configurations of the hybrid films. For the G/CNT or CNT/G hybrid films, the high-strength CNT network synthesized via a developed CVD method was coated on the graphene (grown on Cu foil) directly. Owing to the freestanding feature of the CNT network and the strong Van der Waals force between graphene and CNTs, a freestanding hybrid film could be formed after the Cu foil was dissolved.^[20] For the fabrication of the G-CNT hybrid film, the CNT network was first coated on bare Cu foil, followed by growth of graphene. After the Cu foil was removed, a freestanding coplanar G-CNT hybrid film could be obtained, as shown in Fig. 1(b).

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) Schematic fabrication processes of graphene and CNT based hybrid films with different configurations: (a) CNT/G and G/CNT hybrid films, (b) coplanar G-CNT hybrid film.

In the Raman spectra of different nanocarbon materials, as shown in Fig. 2(a), a high quality CNT network exhibits typical radial breathing mode (RBM), D band, G band, and 2D band, with the intensity ratio of D to G bands smaller than 0.05. For the as-grown graphene, no D band or RBM band, a G band (∼ 1590 cm⁻¹) with a single Lorentzian feature, and a strong 2D band (∼ 2690 cm⁻¹) are observed, which indicates the single-layer feature of graphene.^[33] On the other hand, the Raman spectra of different hybrid films are almost the same. The existence of RBM and the overview Raman spectra illustrate the coexistence of CNTs and graphene. Moreover, very weak D bands, in both the CNT/G (G/CNT) and G-CNT hybrid films, indicate that the fabrication processes did not bring obvious defects.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) (a) Raman spectra of graphene, CNT network, G/CNT, and G-CNT films. (b) Optical transmittance and sheet resistance of typical graphene, CNT network, G/CNT, and G-CNT films on glass substrates. Due to the same nanostructures of the G/CNT and CNT/G hybrid films, their Raman spectra, optical transmittances, and sheet resistances are nearly the same.

Figure 2(b) shows the optical transmittances of the as-grown CVD graphene film, CNT network, G/CNT (CNT/G), and G-CNT hybrid films with the corresponding sheet resistances. The transmittance at 550 nm of the as-grown single-layer graphene and the CNT network is 99.6% (780 Ω/sq) and 89.2% (270 Ω/sq), respectively. The assembled hybrid films, however, show a slight decrease in transmittance compared to the CNT network due to overlaying of graphene. Typically, the G/CNT or CNT/G hybrid film exhibits a transmittance of 87.1%, which is almost the same as that of the G-CNT hybrid film (85.8%). The hybrid films have lower sheet resistance than an individual graphene film and CNT network. The sheet resistance of the obtained G/CNT (CNT/G) and G-CNT hybrid films is 210 Ω/sq and 190 Ω/sq, respectively. This decrease is mainly attributed to the additional charge transport channel in the graphene and CNT network assemblies.^[23]

From the surface morphology of each hybrid film on a SiO₂ (300 nm)/Si substrate characterized by AFM (Fig. 3), it is evident that all three free-standing hybrid films have complete structures after the transfer processes, which proves their high mechanical strength. At the same time, obvious difference can also be observed between the G/CNT hybrid film and the other two hybrid films. For a G/CNT hybrid film (Fig. 3(b)), the upper graphene film acts as a soft and smooth “blanket” which covers the CNT network, hence giving rise to no noticeable CNTs’ morphology. For the CNT/G and G-CNT hybrid films (Figs. 3(a) and 3(c)), the sharp and clear height variations of the CNT network are very obvious, which are distinct compared to the G/CNT hybrid film. Moreover, owing to the strengthened effect of a continuous CNT network, the graphene counterparts in the hybrid films are almost integral.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) AFM images of (a) CNT/G, (b) G/CNT, and (c) G-CNT films on Si substrates.

In order to investigate the performance of different nanocarbon films applied in Si based solar cells, the films were transferred onto n-type single crystalline Si substrates to form heterojunctions (junction area is 9 mm² for all devices). The structure of the as-designed nanocarbon film/Si heterojunction solar cell is depicted in Fig. 4(a). The nanocarbon films served as the anode electrode and were surrounded by silver paint, whereas the cathode was constructed with EGaIn. Depending on the heterojunctions formed with Si, the resulted devices are denoted as G/Si, CNT/Si, G/CNT/Si, CNT/G/Si, and G-CNT/Si solar cells. The J–V curves of different types of solar cells tested under 1 sun condition (P_input = 100 mW·cm⁻², AM 1.5 illumination) are shown in Fig. 4(b). Briefly, the solar cells with hybrid film window electrodes perform better than the G/Si and CNT/Si solar cells. Specifically, the G/Si solar cell exhibits a short circuit current density (J_sc) of 23.6 mA·cm⁻², an open circuit voltage (V_oc) of 400 mV, and a fill factor (FF) of 34.9%, which results in a PCE of 3.3% (Table 1). This result is consistent with the results in the earlier reports.^[15,16] Besides, the CNT/G/Si solar cell shows an enhanced J_sc of 24.9 mA·cm⁻², V_oc of 427 mV, FF of 52.7%, and PCE of 5.6%. A similar result is obtained in the CNT/Si (PCE = 5.4%) and G/CNT/Si (6.5%) solar cells. Among all these devices, the G-CNT/Si solar cell possesses the highest PCE (7.9%), which shows a J_sc of 25.5 mA·cm⁻², V_oc of 480 mV, and FF of 64.5%.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (color online) (a) Structure diagram of a nanocarbon film/Si solar cell. (b) Representative J–V characteristics of different solar cells based on graphene, CNT network, G/CNT hybrid, CNT/G hybrid, and G-CNT hybrid films.

Table 1.

Table 1.

Table 1. Comparison of device parameters of different Si based heterojunction solar cells with graphene, CNT network, G/CNT, CNT/G, and G-CNT hybrid films as the top electrodes. . Device Jsc/mA·cm−2 Voc/mV FF/% PCE/% Rs/Ω n G/Si 23.6 400 34.9 3.3 47 2.5 CNT/Si 24.9 458 47.4 5.4 29 2.1 G/CNT/Si 24.4 452 58.9 6.5 25 2.0 CNT/G/Si 24.9 427 52.7 5.6 31 2.2 G-CNT/Si 25.5 480 64.5 7.9 21 1.8

Table 1. Comparison of device parameters of different Si based heterojunction solar cells with graphene, CNT network, G/CNT, CNT/G, and G-CNT hybrid films as the top electrodes. .

Furthermore, the series resistance (R_s) and ideality factor (n) are calculated under the dark measurements with the following equation:^[14] 1 where q is the elementary electric charge, and k is the Boltzmann constant. The calculated R_s and n are listed in Table 1. Also noteworthy is that the G/Si solar cell has the highest R_s (47 Ω) and n (2.5) while the G-CNT/Si solar cell shows the smallest R_s (21 Ω) and n (1.8).

In fact, the majority of reported graphene grown by the CVD method shows a sheet resistance of about 1000 Ω/sq, which is much larger than its theoretical value (30 Ω/sq).^[34] Unlike the freestanding CNT network, the graphene grown on copper catalyst substrate needs to be protected in the course of transfer. The commonly used method is spin coating a thin PMMA layer on the graphene to serve as a protective sacrifice layer.^[30] However, cracks and polymer residues are often formed in this process (Fig. 5(a)), which severely degrade the original properties of the graphene film.^[17] Moreover, the grain boundaries in polycrystalline graphene also reduce the conductivity of the graphene film.^[18] In the presence of defects and residues (Fig. 5(b)), the photon-generated charge carriers will be scattered and recombined, resulting in a poor performance of the G/Si solar cell. This is consistent with its high R_s and n calculated above (Table 1). When the CNT network is hybridized with graphene, the CNT network can provide more pathways for the charge carriers, which leads to the improved conductivity. The enhanced conductivity of the hybrid film results in less power loss of the solar cell, which brings about higher FF and V_oc compared to those of the G/Si solar cells. This improvement is reflected in the reduced R_s and n of the hybrid nanocarbon film/Si solar cells.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (color online) (a) Schematic diagram of the formation of cracks and residues in the transfer process. (b) Optical microscope image of as-fabricated G/Si solar cell. A dash-line crack of graphene can be observed by the different contrast of graphene and SiO2. (c) Mechanism graph of the photovoltaic process. The separated holes are collected by both CNTs and graphene individuals.

Interestingly, the PCE of the G-CNT/Si solar cell is higher than that of the G/CNT/Si and CNT/G/Si solar cells. As the sheet resistance and optical transmittance of these hybrid films are almost the same, the only difference is the heterojunction formed with Si. For the G/CNT or CNT/G hybrid films, there only exists one type of heterojunction (G/Si heterojunction or CNT/Si heterojunction). However, because of the hybridization method employed for the hybrid films, the graphene and CNTs contact the Si together in the G-CNT/Si solar cell, jointly forming two types of heterojunctions, as shown in Fig. 5(c). In particular, the G/Si heterojunction is established to be a Schottky junction, as many works reported.^[11,14] But there still exist some divergences on the heterojunction type of CNT/Si due to a mixture of semiconducting and metallic CNTs in a typical CVD synthesized CNT film. For instance, Jia et al.^[35] reported that the CNT/Si contact is a Schottky heterojunction. Jung et al.^[26] studied the temperature dependence of a CNT/Si solar cell and suggested that the CNT/Si contact is a p–n heterojunction. As reported earlier,^[29] our CNT/Si solar cell is similar to the latter as a whole. Accordingly, the measured PCE of the G-CNT/Si solar cell (7.9%) is almost equal to the sum of those of a CNT/Si (5.4%) and a G/Si solar cell (3.3%), and nearly twice that of the solar cell with a similar structure reported earlier.^[36] By analyzing the dependence of the solar cell performance on three kinds of nanocarbon film/Si heterojunctions, it is evident that the coexistence of G/Si and CNT/Si heterojunctions in the G-CNT/Si solar cell gives rise to a larger contact area compared to the CNT/Si or G/CNT/Si solar cells, resulting in the enhanced collecting ability of holes. Besides, once the photo-generated holes are extracted to the G-CNT hybrid film, the charge carriers can be transported more efficiently by a continuous high conductive CNT network with Y-type CNT interconnection.^[29] Therefore, the G-CNT hybrid architecture finally results in much less scattering events from the defects and impurities in the graphene film. These features can be concluded as the synergistic effect of graphene and CNT network in the as-designed G-CNT/Si solar cell.

The stability of the G-CNT/Si solar cell is further studied (Fig. 6(a)). Over 15-day storage at room temperature and ambient atmosphere, the solar cell retained 94% PCE of the original value, which shows very stable performance. To further enhance its efficiency, a thin PMMA layer, which served as the antireflection coating, was spin coated on its surface. As shown in Fig. 6(b), the resulted solar cell demonstrated an enhanced PCE of 9.1%. This improvement is mainly attributed to the increased short circuit current (30 mA·cm⁻²), almost 16% higher than the pristine value.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. (color online) (a) Time dependence of PCE of a G-CNT/Si solar cell. (b) J–V characteristics of G-CNT/Si solar cell coated with a moderate layer of PMMA (PMMA/G-CNT/Si for short).

We have analyzed the characteristics of doping-free G/Si solar cells and found that the relatively poor performance results from the high resistance of the CVD-derived graphene film. Through the introduction of a continuous and conductive CNT network, we designed three types of hybrids with different configurations, which all showed much better conductivity than graphene. As a result, a Si solar cell with a high and stable PCE of 7.9% without chemical doping was fabricated based on a coplanar G-CNT hybrid film. The PCE was improved by about 46% and 139% compared to those of the individual CNT/Si (5.4%) and G/Si (3.3%) solar cells fabricated at the same condition. By investigating the characteristics of different nanocarbon film/Si solar cells, we revealed the dependence of their performance on the heterojunctions formed in the devices. On the one hand, the efficiency improvement of the G-CNT/Si solar cell is attributed to the same optimal interfacial oxide thickness of G/Si heterojunction and CNT/Si heterojunction, which makes it feasible to obtain optimal heterojunction quality simultaneously. On the other hand, the coexistence of G/Si and CNT/Si heterojunctions in the G-CNT/Si photovoltaic device exhibits a synergistic effect, resulting in more carrier transport paths and more efficient carrier collection. Finally, the device efficiency was further enhanced to 9.1% after applying a PMMA antireflection coating. Our work also reveals new aspects of the device architecture and the potential for Si based heterojunction photovoltaic devices.