Ultraviolet photoconductive devices with an n-GaN nanorod-graphene hybrid structure synthesized by metal-organic chemical vapor deposition

The superior photoconductive behavior of a simple, cost-effective n-GaN nanorod (NR)-graphene hybrid device structure is demonstrated for the first time. The proposed hybrid structure was synthesized on a Si (111) substrate using the high-quality graphene transfer method and the relatively low-temperature metal-organic chemical vapor deposition (MOCVD) process with a high V/III ratio to protect the graphene layer from thermal damage during the growth of n-GaN nanorods. Defect-free n-GaN NRs were grown on a highly ordered graphene monolayer on Si without forming any metal-catalyst or droplet seeds. The prominent existence of the undamaged monolayer graphene even after the growth of highly dense n-GaN NRs, as determined using Raman spectroscopy and high-resolution transmission electron microscopy (HR-TEM), facilitated the excellent transport of the generated charge carriers through the photoconductive channel. The highly matched n-GaN NR-graphene hybrid structure exhibited enhancement in the photocurrent along with increased sensitivity and photoresponsivity, which were attributed to the extremely low carrier trap density in the photoconductive channel.


Results
Growth and characteristics of monolayer graphene transferred to Si (111). The fabrication sequences involved in the process used to synthesize n-GaN NR-graphene hybrid structures are shown in Fig. 1. A typical graphene growth procedure and the corresponding temperature profile are shown in Fig. 2a. The graphene synthesis process is discussed in detail in the Methods section. Figure 2b presents a representative Raman spectrum of graphene samples that were obtained at a temperature of approximately 1050 °C and then transferred onto a Si substrate. The spectrum exhibits two primary features: a G-band at approximately 1584 cm −1 and a 2D-band at approximately 2673 cm −1 , which are typical peak positions for defect-free graphene 28,29 . The associated G/2D band and the full-width at half-maximum (FWHM) of the 2D-band maps in Fig. 2c illustrate the uniformity of the graphene films over large areas (approximately 490 μ m 2 ) that predominantly exhibit typical monolayer properties, as identified by the I G /I 2D ≤ 0.38 and the FWHM of the 2D band of ≤ 38 cm −1 . Chu et al. 29 have reported the formation of a graphene layer over SiO 2 /Si (100) with a G-band/2D-band peak intensity ratio ≤ 0.4, whereas Kwak et al. 30 have reported a ratio of approximately 0.5. In this study, Raman spectroscopy revealed further improvement in the quality and uniformity of the monolayer graphene (with a peak intensity ratio I G /I 2D ≈ 0.38) grown using the CVD process and then transferred to a Si (111) substrate. To our knowledge, this study represents the finest graphene layer transferred directly to a Si substrate reported thus far 29,30 . The electrical properties of the graphene films that were transferred onto Si substrates were evaluated using the circular transfer length method (CTLM) structure, as illustrated in Fig. 2d. Figure 2e shows the measured resistances (blue circles) as a function of the channel spacing S from 4 to 500 μ m. A geometry correction was extracted by fitting the measured resistances to the following equation [31][32][33] : Here, R T is the total resistance between the channel spacing of the contact pads, R s is the sheet resistance of the graphene films, r 1 is the radius of the inner circular contact, S is the channel spacing of the contact pads, and L T is the transfer length. From the linear fit (red line), the estimated sheet resistance and transfer length of the graphene films were R s ≈ 483 ohm/sq and L T ≈ 13.60 μ m, respectively, implying that highly crystalline monolayer graphene films were transferred onto the Si substrate.
Excellent crystallinity of n-GaN NR-graphene hybrid structure. Figure 3a,b show the surface morphologies of GaN nanorods grown on Si substrates without and with a graphene layer, respectively. The monolayer graphene clearly provided a smooth surface for the growth of denser NRs with the The total resistance measured using the CTLM method. The plot shows the total resistance as a function of channel spacing (S = 4-500 μ m) after the correction factors were applied. The measured resistances (blue circles) are determined by the CTLM contact structure between the inner and outer contacts with various channel spacing. From the linear fit (red line), the following values are obtained: sheet resistance (R s ), contact resistance (R c ), and transfer length (L T ). The error bars represent the standard deviation of the resistance in 5 different devices. prominent presence of individual NRs (Fig. 3b) and without further formation of structural defects. In addition, the graphene layer actually helped improve the material quality of the n-GaN NRs, as evident from the single-crystal X-ray diffraction (XRD) patterns (Fig. 3c,d). Figure 3c shows four XRD peaks for the GaN NRs centered along the (100), (101), (112) and (201) planes when grown without the graphene layer; however, no GaN NRs were grown along the (002) or c-plane. With the integration of a graphene layer, the XRD peaks indicated the existence of GaN NRs grown along the c-plane, i.e., along the (002) and (004) planes ( Additionally, when Raman spectroscopy was performed for an individual n-GaN NR, the Raman spectra revealed peak positions for the Si substrate and GaN nanorods at approximately 520 cm −1 and 567 cm −1 , respectively (Fig. 4a); however, the spectrum of the nanorod body exhibited weak G-band and 2D-band peaks associated with graphene (Fig. 4b). The Raman spectrum of the empty space between the nanorods, as shown in Fig. 4c,d, exhibited strong peaks attributable to the Si substrate and the G-band and 2D-band peaks associated with graphene, respectively. Prominent G-and 2D-bands of higher intensity (compared to those in Fig. 4b) were observed at peak positions of approximately 1584 cm −1 and 2673 cm −1 , respectively, in the spectrum of the graphene layer (Fig. 4d) 29 . This finding provided clear evidence that, even after the growth of n-GaN NRs on the monolayer graphene, the physical existence and material properties of the graphene layer were not affected and the graphene layer was not damaged or impaired. Even though the Si substrate peak was observed in the Raman spectrum of the nanorod body ( Fig. 4a), this peak was much weaker compared to the Si peak in Fig. 4c.
High-resolution transmission electron microscopy (HR-TEM) images showed that the n-GaN NRs were uniformly grown along the c-plane, < 0001> direction (Fig. 5a). To verify the nanorod-graphene hybrid heterostructure, a high-magnification lattice image was recorded. The n-GaN NR-graphene hybrid structure was clearly observed, and single-layer graphene was observed to exist on the Si substrate (confirmed by Raman spectroscopy in Fig. 2c); in addition, a defect-free n-GaN NR was observed to be overgrown on the graphene layer, as shown in Fig. 5b. The thickness of the graphene layer was almost uniform throughout the substrate and was determined to be less than 5 Å. The possibility of growing a well-defined 5 Å graphene layer and its prominent existence suggest that the proposed hybrid structure will exhibit high performance. A representative selected-area electron diffraction (SAED) pattern taken along the < 1120> zone axis on n-GaN NR is presented in Fig. 5c, where (0001), (0110), and (0111) diffraction spots are present 25,34 . The clear diffraction spots indicate that the SAED pattern was of a high-quality hexagonal structure of a GaN NR. No extended defects, such as misfit dislocations and stacking faults, were observed in the n-GaN NRs. A clearer lattice construction of the n-GaN NR is shown in Fig. 5d, which was acquired from inverse fast-Fourier-transform (IFFT) imaging. The interplanar spacings of the < 0001> and < 0110> planes on the NR structure, as measured from TEM images, were approximately 4.78 and 2.39 Å, respectively, as shown in Fig. 5d. Figure 6 compares the photoluminescence (PL) spectra of n-GaN NRs grown on Si (111) substrates with (red line) and without (black line) a graphene layer. Both of the PL spectra show a band-edge emission at approximately 365 nm, corresponding to the n-GaN NR structures 34 . We also observed an emission as a broad band at approximately 546 nm in the spectrum of the n-GaN NRs grown on the Si substrate without graphene. The appearance of the broad emission at approximately 546 nm is associated with vacancy and structural defects 34 . However, in the spectrum of the n-GaN NRs grown on the graphene layer on the Si substrate, this broad emission originating from structural defects was not observed; i.e., the junction of the n-GaN NR-graphene heterostructure is almost strain-free. The defect-free growth of n-GaN NRs on graphene, as revealed by HR-TEM images (Fig. 5b), confirms this observation. The FWHM of approximately 20 nm of the PL peak corresponding to the n-GaN NRs grown without a graphene layer was reduced to approximately 17 nm with the introduction of the graphene layer; this observation confirms that the graphene layer not only acted as an obstruction for forming structural defects but may have also improved the uniformity and material quality of the n-GaN NRs. The improvement of the material quality of the n-GaN NRs can also be correlated with the XRD peaks corresponding to n-GaN NRs in Fig. 3d, where the intensity of the (002) peak in the XRD pattern of the n-GaN NRs was strong when a graphene layer was introduced. Therefore, on the basis of the results of   our study, it is likely that defect-free n-GaN NR-graphene hybrid structures grown on Si substrates will be suitable for use in cost-effective, compact and efficient photoconductive device applications.
Superior photoconductivity of the n-GaN NR-graphene hybrid devices. The photocurrent behavior of the bare semi-insulating Si (111) substrate during photoconductivity measurements is shown in Fig. 7a, whereas Fig. 7b shows the photoconduction between n-GaN NRs on a Si substrate without a graphene layer. The photoconductivity of the channels based on the n-GaN NR-graphene hybrid structures on Si substrates is shown in Fig. 7c. The photocurrent I P is defined as (I max − I dark ) at a specific bias voltage 2,35 . The insets of Fig. 7a,b,c depict typical I-V characteristics of photoconductive channels with (red line) and without (dark line) a light source. Figure 7d shows the almost nine fold enhancement in the photocurrent for the n-GaN NR-graphene heterostructure when these three types of channels were compared. The inset of Fig. 7d shows the variation of the sensitivity 2,35 (I P /I dark ) with bias for the bare Si substrate and for n-GaN NRs on the same substrate without and with a graphene layer; the sensitivity was estimated to be approximately 10%, 4%, and 60%, respectively, at a bias voltage of −1.0 V. The degradation in sensitivity for the n-GaN NRs grown on the Si substrate without a graphene layer might be attributable to the presence of growth-related structural defects, as already established on the basis of the photoluminescence results (Fig. 6). The lower densities of the GaN NRs, as the field-emission scanning electron microscopy (FE-SEM) image illustrates (Fig. 3a), might be the additional cause of this low sensitivity for the same sample. With the introduction of a graphene monolayer, the absence of structural defects (Fig. 6), enhanced density (Fig. 3b) and improved material quality (Fig. 3d) of the GaN NRs contributed to the increase in sensitivity to 60%.
The same behavior of the photocurrent and sensitivity was repeated when the photoresponsivity of these three channels was compared (Fig. 8a). The photoresponsivity was estimated as I P /(Illuminated area × Power density of light) 2,35 . For our study, we used a maximum power density of 100 mW/cm 2 and maintained the same dimensions of the illuminated area for all of the channels. At a bias of 2.0 V, the absolute photoresponsivity was calculated to be 20, 21, and 106 mA/W, respectively, for the bare Si substrate, n-GaN NRs grown without graphene and n-GaN NRs with graphene. As Fig. 3b previously demonstrated, a greater density of GaN NRs resulted from the underlying graphene layer. A larger responsivity was therefore evident for the hybrid structure because of both the reduced empty space between individual NRs and the larger photoconductive area in the channel. Photoresponsivity analysis again indicated that the n-GaN NR-graphene hybrid structure was always a better option as a photoconductive channel compared with GaN NRs directly grown on a Si substrate. Hyungwoo Lee et al. 2 established that graphene could be used as a photoresponsive channel for graphene-CdS NR hybrid structures and, thus, that the photoresponsivity of the structures could be enhanced. We applied the same concept for the n-GaN NR-graphene hybrid structure under study. Graphene exhibits excellent electrical conductivity 2,14,36 even though the lifetime of photogenerated carriers is very short in monolayer graphene 2,37,38 . Thus, a single layer of graphene can serve as a highly efficient transport medium of charge carriers between n-GaN NRs and metal electrodes.
Different characterization methods have already been used to establish that the material quality of n-GaN NRs is improved when grown on monolayer graphene rather than directly on a Si substrate. To validate this fact through the electrical behavior of the channel, an optical-power-dependent study of the photocurrent was performed. When the photocurrent was measured with various optical power densities of the light source for the n-GaN NR-graphene heterostructure, the photocurrent was observed to increase with increasing incident optical power density (Fig. 8b). Various photocurrent values at a bias of 2.0 V were plotted for this channel against the optical power density (Fig. 8c) and fitted by a simple power law, I p ∞ P x 2,35,39 , where P is the incident optical power of the white light source and the exponent x determines the density of trap levels between the Fermi level and the conduction band of the n-GaN NRs in the photoconductive channel under study. These trap levels are responsible for the absorption of photogenerated carriers and result in suppression of the photocurrent in the channel. With Scientific RepoRts | 5:10808 | DOi: 10.1038/srep10808 a lower density of trap levels, even at low incident power, photogenerated carriers should participate in the photocurrent without being absorbed by the trap levels. Thus, the photocurrent I p would increase linearly with increasing optical power when the trap level density is lower. To support this fact, the exponent x should be approximately 1 in the power law. If the photoconductive channel possesses a high density of trap levels, x should be significantly smaller than 1. When the photocurrent values were fitted in Fig. 8c, the photoconductive channels showed x ≈ 0.94, which is larger than any value reported thus far for nanorod-graphene hybrid structures 2 . The evidence of minimal defects or trap states in the channel reveals why this MOCVD-grown n-GaN NR-graphene hybrid structure on Si is ideal for use in cost-effective, highly efficient photoconductive device applications. The photocurrent values, which are associated with the spectral response, were measured to be almost constant at approximately 36 mA at wavelengths up to an abrupt cut-off wavelength of 380 nm (Fig. 8d). For light of wavelengths longer than 380 nm, the photons did not possess sufficient energy to generate photocarriers to contribute to the photocurrent. Thus, the photocurrent values were measured to be approximately 10 mA at wavelengths longer than 380 nm for this n-GaN NR-graphene hybrid structure (Fig. 8d).

Discussion
This study demonstrates the superior photoconductive behavior of an n-GaN NR-graphene hybrid device structure synthesized on a Si (111) substrate using a high-quality graphene transfer method and the relatively low-temperature MOCVD process with a high V/III ratio. Raman spectroscopy and HR-TEM images confirmed the presence of monolayer graphene highly matched with n-GaN NRs in the hybrid structure. The excellent electrical conduction properties of the graphene monolayer transferred on Si (111) not only affected the growth of defect-free, highly dense n-GaN NRs but also improved the material quality. As a result, the n-GaN NR-graphene hybrid structure under study represents the superior characteristics of an ultraviolet photoconductive device with extremely low carrier trap density. Moreover, the synthesis of the hybrid structure directly on a Si substrate by the growth of n-GaN NRs without using metal-catalyst or droplet seeds added special significance. To our knowledge, such simple, cost-effective and highly efficient ultraviolet photoconductive devices fabricated using highly matched n-GaN NR-graphene hybrid structures are reported here for the first time.

Methods
CVD growth of graphene and transfer method. First, a Cu foil (25 μ m thick, 99.8%, Alfa Aesar) was electrochemically polished in H 3 PO 4 (85%) solution to obtain a flat and smooth Cu foil surface with a root-mean-square surface roughness of less than 15 nm (in the scanning area of 100 μ m × 100 μ m). After the Cu foil was loaded into a low-pressure chemical vapor deposition (LP-CVD) system, it was heated to a process temperature of 1050 °C and maintained for 60 min under 5 sccm of H 2 for in situ surface cleaning. A monolayer graphene film was then grown for 15 min under a CH 4 /H 2 mixture (10 and 5 sccm, respectively) with a reactor pressure of approximately 90 mTorr. After graphene growth, the chamber was rapidly cooled to room temperature with a flow of H 2 . Following graphene growth, the graphene film was transferred onto a Si substrate for further investigation. In all of our growth experiments, monolayer graphene was observed to grow on both sides of the Cu foil; therefore, the backside of the Cu foil was pre-treated with oxygen plasma to remove the undesired graphene film. To transfer the graphene film to a Si substrate, poly(methyl methacrylate) (PMMA) was spin-coated onto the graphene grown on Cu foil, and the PMMA/graphene/Cu foil was then cured on a hot plate at 180 °C for 1 min. The Cu foil was then etched from the PMMA/graphene/Cu foil assembly using an aqueous solution of (NH 4 ) 2 S 2 O 8 . The PMMA/graphene assembly was subsequently washed with deionized water, transferred onto a semi-insulating Si (111) substrate and dried. The PMMA was then removed using acetone.
Synthesis of highly matched n-GaN NR-graphene hybrid structure using MOCVD. n-GaN nanorods were grown on the graphene-transferred Si (111) substrate using a relatively low-temperature MOCVD process and a high V/III ratio without the formation of any metal-catalyst or droplet seeds. The precursors for gallium and nitrogen were trimethylgallium (TMGa) and ammonia (NH 3 ), respectively; silane (SiH 4 ) gas was used for n-type doping, and hydrogen was used as the carrier gas. To grow n-GaN nanorods, TMGa, NH 3 and SiH 4 were introduced into the reactor chamber for 45 min at flow rates of 0.2 sccm, 3.0 SLM and 10 sccm, respectively (V/III ratio = 15,000). To protect the graphene layer from thermal damage during the growth of the n-GaN nanorods, the growth temperature was fixed at 870 °C, which is relatively low compared to the usual temperatures of greater than 1000 °C 15 . Finally, Au/Ni metal electrodes were formed on the graphene layer to evaluate the photocurrent, photoresponsivity and cut-off wavelength.
Characterization of the hybrid structure and fabricated device. Raman spectroscopy was performed using a WiTec Alpha 300R M-Raman system; the excitation wavelength was 532 nm. The results of the Raman study were used to confirm the existence of highly ordered graphene films. The surface morphology of the n-GaN NRs was analyzed using FE-SEM. A Hitachi S-7400 system was used for the FE-SEM study; it was operated at 15 kV and at a 13° tilt-view. Single-crystal XRD measurements were performed using a Rigaku diffractometer equipped with a Cu-Kα radiation source. The morphology of the n-GaN NR-graphene heterostructures was further revealed by cross-sectional HR-TEM. A JEOL JEM Scientific RepoRts | 5:10808 | DOi: 10.1038/srep10808 2010 system operated at 200 kV was used for the HR-TEM study. Samples of the hybrid structure were prepared by being coated with platinum using a dual-beam focused ion beam (FIB, Quanta 3D FEG) technique with a beam current of 65 nA and a resolution of 7 nm at 30 kV. The optical properties of the n-GaN NRs were investigated by photoluminescence (PL) spectroscopy at room temperature using the 325 nm line of a He-Cd laser as an excitation source. For photoconductivity measurements of the fabricated ultraviolet photoconductive device, we utilized a solar simulator (McScience Lab 100) as a light source. This light source generated white light with a maximum power density of 100 mW/cm 2 . Spectral photoresponse measurements of the fabricated device were performed with a xenon arc lamp (300 W) within the range of 300-550 nm. A monochromator (Oriel Cornerstone 130) was used to provide the monochromatic light incident on the channel.