Modulation-Doped Growth of Mosaic Graphene with Single Crystalline p-n Junctions for Efficient Photocurrent Generation Kai Yan 1,, Di Wu 1,, Hailin Peng 1, *, Li Jin 2, Qiang Fu 2, Xinhe Bao 2 and Zhongfan Liu 1, * 1 Center for Nanochemistry, Beijing National Laboratory for Molecular Sciences (BNLMS), State Key Laboratory for Structural Chemistry of Unstable and Stable Species, College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, P.R. China. 2 State Key Laboratory of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian 116023, China. * To whom correspondence should be addressed. E-mail: hlpeng@pku.edu.cn, zfliu@pku.edu.cn These authors contributed equally to this work. 1
Supplementary Figures Supplementary Figure S1. Growth protocol of modulation doped graphene hybrid. Nucleation was accomplished in 1 min, followed by 3 min cooling and purging. Acetonitrile vapor was introduced afterward for 3 min to achieve full coverage. 2
Supplementary Figure S2. Characterizations of N-doped graphene grown from acetonitrile. a. Optical microscope (OM) image of transferred monolayer N-doped graphene, exhibiting excellent uniformity. b. Typical Raman spectrum of N-doped graphene, exhibiting strong D and D* band, as well as weakened and broadened 2D band. These features are contrasting with N-doped portions in mosaic graphene, indicating lower quality of direct grown N-doped graphene. c. XPS spectrum of N-doped graphene. The characterization was performed on four layers of N-doped graphene transferred onto SiO 2 /Si substrate, in order to enhance the signal. The peaks at about 284, 400, and 532 ev correspond to C 1s of sp 2 C, N 1s of the doped N, and O 1s of the absorbed oxygen and oxygen in SiO 2 substrate, respectively. The atomic percentage of nitrogen is estimated to be 0.9 atom % by taking the peak area ratio of the N 1s at 400 ev to the C 1s at 284 ev after considering the atomic sensitivity factors. d. The N 1s peak has two components at 399.7 and 402.3 ev, corresponding to pyridinic-n and graphitic-n, respectively, which can be assigned to pyridinic-n and graphitic-n, respectively 15. The ratio of the peak area indicates that 34% N atoms were pyridinic-n while the rest 66% were graphitic-n. e. KPFM image of N-doped graphene (inset) and its histogram of work function. Scale bar: 0.5 µm. The KPFM image exhibited ideal smoothness, indicating the uniform distribution of 3
dopant in N-dope graphene. The histogram of the image was fitted with a single Gaussian peak, with a distribution of 11 mev extracted. Since the same recipe was also applied to the growth of mosaic graphene, the doping uniformity in its doped portion can be guaranteed. f. Transport properties of N-doped graphene in vacuum of 10-6 mbar. Negative Dirac point indicates effective electron doping arose from nitrogen. Inset is the optical image of device. Scale bar: 5 µm. The transfer characteristic curve shows the charge neutrality point (Dirac point, V D ) is about -20 V, indicating effective electron doping (~2 10 12 cm -2 ) derived from nitrogen. Carrier mobility estimated from the transfer curve is less than 100 cm 2 /V s, which is much lower than that of mosaic graphene. 4
Supplementary Figure S3. Typical scanning electron microscope (SEM) images of the two-step growth procedure demonstrating the optimization of mosaic graphene growth. a. SEM image of discrete intrinsic graphene grains with well controlled density and gaps. b. SEM image of discrete mosaic graphene grains. One key issue in the optimization of mosaic graphene arose from the suppression of N-doped graphene nucleation in order to reduce grain boundaries. In other words, N-doped graphene should predominantly grow from existing intrinsic graphene rootstocks rather than form separately. To achieve this, the space between intrinsic graphene roots should be carefully adjusted, i.e., the gap should match the distance of spontaneous N-doped graphene grains. By tuning the growth time of intrinsic graphene, the gap between grains was controlled to ~4 µm (panel a), which is just below the ~5 µm nucleation distance studied above. As a result, most N-doped graphene grew from the edge of these roots without much self-nucleation, yielding discrete mosaic graphene (panel b). 5
Supplementary Figure S4. Growth of periodic mosaic graphene superlattice. Growth of periodic mosaic graphene superlattice. a. SEM image of predefined PMMA seeds of <1 µm, fabricated via e-beam lithography. b. SEM image of periodic intrinsic graphene grains grown from methane with reduced partial pressure 19,36. c. SEM image of mosaic graphene superlattice after coalescence of N-doped graphene. More information of periodic mosaic graphene superlattice would be discussed elsewhere. 6
Supplementary Figure S5. Raman Mapping of another domain of graphene hybrid. Besides the clear contrast in the 2D band, the FWHM (Γ) and frequency (ω) of the G and 2D bands also showed a strong difference between intrinsic and N-doped regions of graphene hybrid. Scale bar: 5 µm. Despite of features in Raman spectrums, i.e., D band mentioned in the main text, the frequency and FWHM of G band and 2D band also provide abundant information on doping. Since the FWHM of Raman bands is usually related to defect concentration, the difference in these features cannot give direct evidence for doping. However, the shift of G band and 2D band in figure S5c and figure S5f, strongly indicate an electron doping level of about 2-3 10 12 cm -2, according to a gating experiment done by A. Das 25. This is in good accordance with results extracted from LEEM and electrical transport studies. 7
Supplementary Figure S6. Charactrizations of the discrete mosaic graphene grain. a-c. SEM, optical and AFM images of the identical mosaic graphene grain, respectively. No contrast existed both in the optical and AFM image, although intrinsic and N-doped portions can be clearly identified in SEM image. The white dashed lines outline intrinsic portion of the grain. d. Raman D band mapping of the grain. The distribution of D band intensity corresponds well with the SEM image. 8
Supplementary Figure S7. Output characteristics of modulation doped graphene p-n junction. a. Linear I DS -V DS curve in a wide bias voltage range. The output characteristic curves of graphene p-n junctions exhibit a non-rectifying behavior, which is different from that of the traditional semiconductor p-n junctions. During the applied bias is increasing from -10 V to 10 V, the output current response linearly. b. Gate-dependent output characteristics presented double dips, which is in accordance with the transfer characteristics. The white dashed lines depict double Dirac point of p-n junction. The absence of inflection precludes rectification with graphene diode. However, applications such as quadruple frequency amplifier are still possible. 9
Supplementary Figure S8. Transfer characteristics of modulation doped graphene p-n junction. The resistance of graphene p-n junction in vacuum down to 10-7 mbar behaved typical curve with two charge neutrality points, separating the whole curve into three regimes labeled n-n +, p-n and p + -p. The charge neutrality points of the two portions are V D-i = 0 and V D-n = -30 V, representing typical properties of intrinsic and n-type doped graphene, respectively. Inset is the SEM image of the device. Scale bar: 5 µm. According to this study, the graphene i-n junction can be tuned to p + -p, p-n or n-n + junction by applying proper gate voltage as shown in the curve. For instance, when V D-i < Vg < V D-n, the i-n junction turns to a p-n junction because of electrostatic charge induced p-doping. However, graphene is usually p-doped by adsorbed H 2 O/O 2 in ambient or contamination induced during device fabrication. Thus, the graphene junctions without applied gate voltage are usually p-n or p+-p junctions rather than i-n junctions. However, it is easily tuned by gate voltage as mentioned above. Consequently, it is reasonable to name the structure as p-n junction. 10
Supplementary Figure S9. Spontaneous nucleation of graphene grains. a and b. Nucleation process of intrinsic and N-doped graphene, respectively. The SEM images were captured on copper. Insets showed the statistics of grain size. Before combining the two stages of modulation doping grafted growth together, experiments with a single component were carried out in order to investigate the spontaneous nucleation density with certain partial pressure of carbon source. The growth was halted before complete coalescence. For intrinsic graphene grains with ~5 Pa methane, the distance between adjacent grains is 10-20 µm, with a typical grain size of ~8 µm (panel a). However, for acetonitrile with reduced partial pressure of ~3 Pa, the nucleation distance was dramatically reduced to ~5 µm (panel b). It is straightforward to speculate that density of grain boundaries in pure N-doped graphene from ACN is substantially high. Consequently, carrier transportation would be seriously deteriorated by these boundaries. 11
Supplementary Figure S10. Gate-dependent photocurrent mapping along the channel of the identical device in Fig. 4c. a. SEM image of the device with graphene p-n junction in the middle of the channel. N-doped graphene is darker than intrinsic graphene in the SEM image. The laser spot moves across the channel along the red dashed line with steps of 0.5 µm. b. Map of photocurrent versus carrier density and laser position. Three distinctive regions corresponding to source, drain and p-n junction areas can be easily recognized. c. Raman 2D mapping the device. d. Photocurrent versus carrier density curves when the laser is located at source, drain and graphene p-n junction, respectively. The photocurrent at the source electrode decreased slowly at negative gate bias, and quickly reached zero and reversed when the electrostatic n-doping approach 1.8 10 12 cm -1. After a rapid increase, the photocurrent reached the maximum with a magnitude of ~75 na and then decreased slightly. For the photocurrent at the drain, the trend is also identical with only one polarity reversal. The maximum magnitude of photocurrent at the drain is ~80 na. It has been reported that the energy level of graphene under the contact is pinned by metal while that in the channel could be tuned through the field effect. The current reverse when the energy in the two regions are equal, which occurs only once during the sweeping of gate voltage 7. However, the trend of photocurrent at the p-n junction was significantly different as the curve exhibited a single peak with twice polarity reversals. This characteristic is attributed to photothermoelectric (PTE) effect that hot carriers excited by photon eventually result in thermoelectricity 10, 37. The PTE effect induced maximum current of ~125 na at the p-n junction even exceeds that at the electrodes. 12
Supplementary Methods Detailed procedure for mosaic graphene growth 25 µm thick copper foil (Alfa Aesar, #13382) was treated with diluted acetic acid for 10 min and then washed with water before loaded into tube furnace (Lindberg Blue M HTF55667C, 1-inch quartz tube). Diagram of the temperature profile adopted was presented in Fig. S1a. Hydrogen with partial pressure of ~10 Pa was used as background gas, in order to create reductive environment. The growth process is as following: 1. After rapid heating up to 1000 o C, copper foil was annealed for 20 min to reduce surface oxides and increase grain size. 2. Methane with partial pressure of ~5 Pa was introduced for 1 min to produce intrinsic graphene rootstocks. 3. Then the vessel was purged with large flow of argon (~100 Pa) for 3 minutes, during which the temperature gradually dropped to 950 o C. 4. A metering valve (Swagelok SS-SS4) was used to adjust the flow of acetonitrile (ACN). In our case, ACN with partial pressure of ~3 Pa was found to be an optimal setup. 30 s of injection is enough for discrete modulation doped domain with edge width of ~2 µm and 3 min is sufficient for full coverage. 5. After shutting the ACN valve, the sample was quickly cooled to 700 o C in 2 min with pure hydrogen and to RT in 20 min. 6. Mosaic graphene was transferred from copper to the desired substrate (Si/SiO2, TEM grid, etc.) using PMMA as carrier. FeCl 3 solution was used as the etchant to remove copper. After transfer, the sample is ready for SEM (HITACHI, S-4800), AFM (DI, Nanoscope IIIa), Raman spectrum (HORIBA, LabRAM HR800,) and TEM (FEI, T20) characterizations. Mobility Extraction Field-effect mobilities were extracted from the transconductance, g m, as follows, g L di L C WV dv WC V m ds µ = =, where C ox is the gate capacitance per unit area. ox ds gs ox ds 13
Supplementary Reference: 36 Wu, W. et al. Growth of Single Crystal Graphene Arrays by Locally Controlling Nucleation on Polycrystalline Cu Using Chemical Vapor Deposition. Adv Mater 23, 4898 4903 (2011). 37 Song, J. C. W., Rudner, M. S., Marcus, C. M. & Levitov, L. S. Hot Carrier Transport and Photocurrent Response in Graphene. Nano Lett 11, 4688 4692 (2011). 14