2.5 Characterization of the graphene/Si Schottky junctions analyzed by SKPM and CAFM
Scanning Kelvin probe microscopy (SKPM) and conducting atomic force microscopy (CAFM) were used to investigate the surface potential, photovoltaic effect, and rectification effect of the graphene/Si Schottky junctions and to better understand the rapid separation and transfer of charge carriers in the depletion layer of the junction under light illumination.[49-50] Figure 6ashows the excellent quality of a two-dimensional AFM surface topographic image from the graphene/Si Schottky junction. Figures 6b-6cconfirm the surface potential distribution of dark and light states, respectively. The observed surface potential decrease under the light compared to the dark states is believed due to the extensive built-in electric field between graphene and Si.[49] The built-in electric field rapidly separates the photogenerated carriers, transfers holes to the graphene, and results in a decrease in the surface potential and a downshift in the Fermi level.[49,51] Figure 6d illustrates the change between graphene and Si under light from the energy band relationship. Under illumination, the electron-hole pair rapidly separates under the built-in electric field, extending the life of the photogenerated carriers.[49,52] Figures 6e-6g , describe the graphene morphology, the current diagram under dark and light states under zero bias, respectively. It shows that the current changes significantly under light, as shown in Figure 6g , meaning that the graphene/Si Schottky junction has a noticeable photovoltaic effect.[53] Figure 6hrepresents the current value corresponding to the white dotted line inFigures 6f- 6g . To determine the rectification efficiency of the Schottky junction, the current values at varying bias voltages (−3 to +4 V) were measured over a 10×10 μm2 area under dark states, and the current maps are displayed in Figures 6i-6k . These maps were then utilized to calculate the average rectification rate of the Schottky junction (Figure 6l ). Average currents of −0.05 nA at −3 V and 3.55 nA at +3 V means that the average rectification ratio is approximately 77 at ± 3 V. Hence, our graphene/Si Schottky junction possesses not only a strong photovoltaic effect but also an excellent rectification effect.
2.6 Photoelectric response data was obtained using a near-IR photodetector fabricated using our graphene/Si Schottky junction.
The photo-response of the as-fabricated photodetector is also evaluated using a pulsed photon source and the results are shown in Figure 7a . The switching behavior of the photodetector was measured with a 1550 nm pulsed laser at a reversed bias of -1 V and the laser intensity of 30 mW/cm2. The rising/falling edges of the photocurrents under the pulsed photon irradiation correspond to the photo-induced voltages or photocurrents, stemming from the infrared thermionic emission.[54-55] Our photodetector demonstrated excellent performance characteristics (e.g. responsivity of 1.1 AW-1 and a detectivity of 1.6×1014cmHz1/2W-1), comparable to that of the conventionally CVD synthesized and then transferred graphene.[54-56] Figure 7b displays how the photocurrent changes while the intensity of the 1550 nm laser irradiation is varied. The photocurrent steadily increases as the photon intensity is raised from 5 to 30 mW/cm2, and then gradually decreases as the photon intensity subsequently decreases back to 5 mW/cm2. The photocurrent remains unchanged at all times during the irradiation with a fixed photon intensity, suggesting an excellent repeatability as summarized in Supplementary Figure S7 . To examine the response speed of the device, pulsed photon signals at different pulsing frequencies (from 500 to 2000 Hz) were used to irradiate the junction and the results are shown in Figure 7c . The relative balance between the maximum and the minimum ((Imax - Imin) / Imax) was found to merely decline by no more than 10% even at 2,000 Hz, as shown in Supplementary Figure S8 . The effect of the residual photocurrent storage in the device is practically negligible, as demonstrated in Figure 7d , where the photocurrent measurement was repeated on the same device during a three-week time span. By expanding the photo-response curve at the time axis, the rise time (tr) and fall time (tf) for the device are estimated to be 60 μs and 63 μs, as shown in Figures 7e . These results are also comparable to the photo-response recorded from the conventionally CVD synthesized and then transferred graphene.[56]To obtain the specific detectivity, the dark current noise must be obtained. Generally, it can be measured directly using the Fast Fourier Transform (FFT) algorithm method. Using a preamplifier with a sampling rate of 10 kHz to measure the dark current and the data were analyzed with FFT software. As shown in Figure 7g , based on the dark current of the device at 0 V, the measured dark current noise is 3.2 fA Hz-1/2 at 1 Hz, which is again comparable to that obtained from a CVD-synthesized and then transferred photodetector.[55,57-58]
Besides the excellent photodetection demonstration, we also fabricated two other devices from our layer-tunable and transfer-free graphene films. A field-effect transistor (FET) was successfully fabricated from our monolayer graphene directly grown on a SiO2substrate and its performance characteristics are shown in SupplementaryFigure S9 . A heating device was successfully fabricated from our monolayer graphene directly grown on a transparent glass substrate and its performance characteristics are shown in SupplementaryFigures S10-S12 . Like our graphene-Si Schottky junction photodetector, our field-effect transistor and heating device also demonstrate excellent performance properties as compared to similar devices fabricated through other methods but without the need for any graphene layer transfer.[59-61]
3. Conclusion
A novel versatile strategy for the direct synthesis of layer-tunable and transfer-free graphene on arbitrary substrates via ion implantation technology has been demonstrated. The stochastic nature of the ion-implantation in terms of carbon atom distribution as well as uncertainties during their post-implantation diffusion and precipitation into graphene structure are eliminated by using smart Janus bi-metal substrates with drastically different carbon solubility. Through this approach, high-quality of mono- and bi-layer graphene films have been synthesized directly on a variety of device-relevant substrates (Si, SiO2, glass, sapphire etc.) without the need for any graphene layer transfer. The layer-tunability is precisely controlled by the equivalent fluences (atoms/cm2) during the carbon implantation. The graphene/Si-based near-infrared photodetectors, graphene/SiO2-based field-effect transistors, and graphene/glass-based heating devices have been successfully fabricated without the graphene layer transfer, and the comparable performance characteristics of our devices to the current state of the art approaches using the graphene layer transfer have been demonstrated. Since ion implantation is a mature technology widely used in the current integrated circuit manufacturing, our work has a potential to promote layer-tunable graphene-based devices in a wide range of applications.
4. Experimental Section
Graphene Synthesis : A 100 nm thickness of Ni layer was evaporated on various substrates (Si, SiO2, glass, sapphire etc.) using an electron beam evaporator. These substrates were implanted with 70 keV carbon ions at room temperature to fluences of 4×1015 ions/cm2 and 8×1015 ions/cm2 to achieve uniform mono- and bi-layer graphene structures, respectively. Following the C-ion implantation, 150 nm Cu film was placed on top of the Ni layer at room temperature by magnetron sputtering. Sandwiched samples were then annealed at 1000 ℃ for 10 min under 10-5 mbar. After annealing, the furnace was cooled to room temperature under 10-5 mbar. Further experimental details are described in the Supplemental Material.
Graphene Transfer: After synthesis, the Cu-Ni alloys are easily removed using a thermal release tape, leaving the graphene on the arbitrary substrate for detailed characterization and direct device fabrication. Thus, there is no additional graphene transfer step needed.
Characterization: SIMS (Cameca IMS-4F, Paris, France) was performed to obtain the elemental depth profiles using a 7.5 keV Cs+ beam oriented 70.2° from the sample normal and the detector was positioned 140.4° from the incident beam. Raman scattering (SENTERRA) at the wavelength of 532 nm was used to determine the thickness, quality, and uniformity of the graphene films. Crystallographic information and the number of graphene layers were determined by TEM (JEOL JEM-ARM 300F). AFM (Multimode 8) was used to determine the uniformity and thickness of the graphene films and STM (Multimode 8) was used to assess the atomic structure of the graphene films under ambient conditions. The structure of back-gated GFETs was examined by SEM (HITAGHT S-3400N). The microstructure and electrical quantity were measured by AFM, SKPM, and CAFM (Oxford Instruments, Cypher S). By using an Agilent semiconductor parameter analyzer (B1500A), photoelectric parameters were measured at ambient conditions, by combining with a semiconductor characterization system (Keithley 4200).
Devices Manufacturing: The photodetectors, FETs, and heating devices were fabricated using monolayer graphene directly synthesized on Si, SiO2, and glass substrates, respectively. Electrodes were prepared by silver (100 nm) electron beam evaporation.
Theoretical calculation: Molecular dynamics (MD) simulations were performed using the ReaxFF,[62-64] which has been confirmed to provide accurate descriptions of bond breaking and bond forming for Ni−C systems. The initial models of the Ni-C system were built by randomly replacing Ni atoms with C atoms. The size of the initial system is 50 × 50 × 3.47 Å. Then these initial structures were optimized to reach a stable configuration for 20 ps in the NVT ensemble. Finally, the optimized model was heated from 300K to 1800K at 10K/ps, and then the system was annealed to 300K with a constant cooling rate of -1K/ps. All the MD simulations were conducted in the canonical NPT ensemble using the Lammps MD package.[65]