Infrared Photodetectors Based on Reduced Graphene Oxide and Graphene Nanoribbons

Graphene is a single atomic layer two-dimensional (2D) system, comprising carbon atoms arranged in a hexagonal honeycomb lattice. It has high Fermi velocity (ca. 1/300 of the speed of light), a linear energy dispersion relation (vanishing effective mass), and an extremely high electrical mobility (approaching ca. 200000 cm² V⁻¹ s⁻¹ for a free sheet) of both electrons and holes.1–7 In addition, it also has remarkable photonic properties, in terms of photon absorption over a wide wavelength range (visible–IR), and a strong interband transition compared to other materials.8, 9 There has been a substantial amount of interest in few-layer graphenes even though studies of monolayer graphene have been reported more extensively in the literature. The electronic structure and properties of multilayer graphene depend on the number of layers and their stacking. Furthermore, absorbance of multilayers at energies above ca. 0.5 eV is additive, resulting in strong graphene–light interaction. Large-area single- or few-layer graphene FETs have been explored as ultrafast photodetectors.10, 11 However, difficulty in synthesizing high-quality large-scale graphene and difficulties in device prototyping inhibit their application in optoelectronics devices.

In order to overcome these problems, a chemical approach offers a route for the deposition of graphene from solution, allowing devices to be fabricated independent of substrate. As a robust yet flexible membrane, graphenes provide enormous possibilities for modification or functionalization.12 For large-scale fabrication of graphene-based devices, exfoliation of graphite into individual graphene sheets in large quantities is required. It is therefore convenient to use graphene oxide (GO) as a precursor for cost-effective, large-scale production of graphene-based materials. GO can also be transformed to reduced graphene oxide (RGO) on a large scale.13 In addition, controlled reduction of GO by chemical or thermal means allows the tunability of optoelectronic properties.14–16 Thin films prepared from solution-processed GO offer ease of material processing, low cost of fabrication, mechanical flexibility, and compatibility with various substrates, making them attractive candidates for large-area devices. GO-based thin films have already been used as transparent and flexible materials for electronic devices.17–22

Graphene nanoribbons (GNRs) possessing a bandgap have been proposed theoretically to be useful as phototransistors.23, 24 The zero bandgap in graphene is due to the identical environment of the two carbon atoms in the graphene unit cell. The potential on the two atoms should be different in order to create a gap. However, these atoms are only 0.14 nm apart, so techniques such as lithographic modification are not possible. One way to accomplish this would be to reduce the dimensionality of graphene from 2D to 1D by cutting graphene into narrow ribbons, GNRs. If the GNR is narrow enough, 1D confined states are formed, similar to those of a particle in a long, narrow quantum box. The gap opened is inversely proportional to the width of the GNR. GNRs have been made by cutting graphene with lithographic techniques,25–28 and by direct synthesis.29 For a substantial gap (0.5 eV), which can allow room temperature operation of devices, GNRs with a constant width of 2–3 nm are required. Current lithographic techniques cannot deliver such performance. Furthermore, problems with ribbon width reproducibility, edge roughness control, and production scale limit the usability of the existing techniques in technology. We have employed chemical routes to synthesize both RGO and GNRs with the specific objective of developing IR photodetectors.

In the investigation presented here, RGO sheets dispersible in N,N-dimethylformamide (DMF) were synthesized using the method of Park et al.30Figure1a shows a transmission electron microscopy (TEM) image of RGO and the corresponding electron diffraction (ED) pattern (shown in the inset). The Raman spectrum and atomic force spectroscopy (AFM) image are shown in Figures 1b and c, respectively. From the AFM image and corresponding height profile, it is evident that the RGO sample consists of 1–2 layers. The ED pattern shows the presence of hexagonal sixfold symmetry in the sample. The Raman spectrum of RGO shows the characteristic D-band (1320 cm⁻¹) and G-band (1590 cm⁻¹). The 2D-band is absent, as reported in the literature.

Figure 1. a) TEM image of RGO and the corresponding electron diffraction pattern (inset). b) The Raman spectrum of RGO. c) The AFM image and the corresponding height profile.

For the synthesis of GNRs, the method reported by Higginbotham et al.31 was employed. Figures2a and b show the TEM image and Raman spectrum of the GNR, respectively, prepared by oxidative unzipping of multiwalled nanotubes (MWNTs). The ribbons were dispersed ultrasonically in ethanol and deposited on carbon grids for observation. The image shows several folded ribbons that are fully unzipped (Figure 2a). The high-resolution TEM (HRTEM) image of the ribbons showed that they were made up of several layers. Raman spectra of the ribbons obtained after oxidation shows broad D- and G-bands and a weak 2D-band. The Raman spectrum of acid-treated MWNTs also shows these bands, but in the case of GNRs the defect-related D-band is much more intense. We also observe an intense band at 1620 cm⁻¹ due to the defect-related G′-band in the case of nanoribbons.

Figure 2. a) TEM image of GNRs and the HRTEM image (inset). b) Raman spectrum of GNRs.

Figure3a shows typical current–voltage (I–V) characteristics of RGO in the dark (squares) and under illumination (circles) at 1550 nm with 80 mW cm⁻². The metal –RGO–metal contact was confirmed by sweeping the voltage from –2 V to +2 V as demonstrated in the inset in the figure (data in the dark). Figure 3b represents the photoresponse for one of the devices, where the photocurrent is plotted as a function of time with different IR intensities on the detector. The photocurrent was measured under a bias voltage V_bias = 2 V. The IR source was turned on and off to demonstrate the reproducibility of the data with time. The photoresponse of the detector at 80 mW cm⁻² is shown in Figure 3c, which confirms that the device behaves well under continuous cycling. Figure 3d shows the performance of one of our RGO detectors, which shows that it can even sense the IR emitted from a human body.

Figure 3. a) Typical I–V characteristics of RGO in the dark (squares) and under illumination (circles) at 1550 nm at 80 mW cm⁻². The metal– RGO–metal contact is demonstrated in the inset (data in dark). b) Photocurrent as a function of time with different IR intensities at 2 V for RGO. c) The photoresponse of the device at 80 mW cm⁻² to show the reproducibility of data with time at 2 V. d) Photocurrent as a function of time with human body IR at 2 V for RGO.

Typical I–V characteristics of the photodetector based on GNRs in the dark (squares) and under illumination (circles) are shown in Figure4a at 1550 nm with 80 mW cm⁻². The metal –GNR–metal contact was confirmed by sweeping the voltage from –2 V to +2 V as demonstrated in the inset (recorded in the dark). Figure 4b shows the photocurrent as a function of time when IR with different intensities was incident on the detector under a bias voltage V_bias = 2 V. The photoresponse of the detector at 80 mW cm⁻² is shown as an inset to confirm the stability of the device. It can be seen clearly that the electrical conductivity of RGO and GNRs increases with IR laser radiation (Figures 3a and 4a).

Figure 4. a) Typical I–V characteristics of GNRs in the dark (squares) and under illumination (circles) at 1550 nm with 80 mW cm⁻². The metal–GNR–metal contact is demonstrated in the inset (recorded in the dark). b) Photocurrent as a function of time with different IR intensities at 2 V for GNRs. The photoresponse of the device at 80 mW cm⁻² is shown in the inset.

Graphene can absorb photons over a wide wavelength range, from the visible to the IR, and has a strong interband transition compared to other materials.8, 9 On absorption of light, electron–hole pairs are generated10 owing to a Schottky-like barrier at the metal –graphene contact, as shown in Figure5. The solid line shows the potential variation within the graphene channel and the dashed lines represent the Fermi levels, E_F, of the two electrodes. The electron–hole pairs generated in graphene would normally recombine on a time scale of tens of picoseconds, depending on the quality and carrier concentration of the graphene.32–34 When an external field is applied, the pairs become separated and a photocurrent is generated. A similar phenomenon can occur in the presence of an internal field formed by photoexcitation.35–37 In the investigation reported here, we used RGO and GNRs on SiO₂, where the mobility is significantly lower (a few thousand to tens of thousands of cm² V⁻¹ s⁻¹, depending on the nature and purity of the insulator). An external voltage is therefore needed to separate the photogenerated electron –hole pairs before they recombine, consequently showing detector action. In suspended, exfoliated graphene, where interactions with the substrate are eliminated, mobilities of about 200000 cm² V⁻¹ s⁻¹ have been observed. Even though we have to apply an external voltage, the advantage of our device is the enhancement of absorption of light by additive absorbance of multilayers at energies above ca. 0.5 eV. Moreover, at 1.55 μm, absorption is estimated to be enhanced by about 25% on a silicon substrate, while in freestanding bi- and trilayer graphene in air, absorption lies at about 4.6% and 6.9%, respectively, of normally incident light.

Figure 5. Representative diagram for photodetector mechanism.

The electrical conductivities of both RGO and GNRs increase with IR laser radiation intensity (Figures 3b and 4b). When the intensity of the laser light is higher, more photons are absorbed by the RGO and GNR films and generate more excitons, resulting in a larger photocurrent. From the absorption spectra of graphene, we see that graphene absorbs in the mid- to near-IR spectral range from 0.2 eV (wavelength 6199 nm) to 1.2 eV (wavelength 1033 nm).38 For photon energies above 0.5 eV (wavelength between 1033 and 2479 nm), graphene exhibits a spectrally flat optical absorbance of 2.3% ± 0.2% and double in the case of a bilayer,38, 39 which is consistent with previous reports.9, 40, 41 It has been reported recently that the optical absorbance spectrum of few-layer graphene has a strong dependence on the stacking sequence of the graphene monolayers.42 From these optical spectra of graphene, the wavelength (1550 nm) of the near-IR laser used for the experiment is within the flat absorbance spectra (wavelength between 1033 and 2479 nm). Therefore, it appears that the demonstrated graphene photoelectric effect was triggered by the absorption of the near-IR laser light (wavelength 1550 nm) used in the present investigation. Furthermore, thermal heating of graphene caused by IR laser absorption is known to decrease the conductivity of graphene, according to the literature.38, 43–45 Hence, thermal effects were excluded as the cause of the increased conductivity of graphene when illuminated by IR laser.

The detector current responsivity (R_λ), defined as the photocurrent generated per unit power of the incident light on the effective area of a photoconductor,46 and the external quantum efficiency (EQE), defined as the number of electrons detected per incident photon, are crucial parameters for photoconductors. Large values of R_λ and EQE correspond to high sensitivity. R_λ and EQE can be calculated as R_λ = I_λ/(P_λ S)47, 48 and EQE = hcR_λ/(eλ),49, 50 where, I_λ is the photocurrent(I_illumination – I_dark), P_λ the light intensity, S the effective illuminated area, h Planck's constant, c the velocity of light, e the electronic charge, and λ the exciting wavelength. According to our experimental results, R_λ and EQE of RGO are 4 mA W⁻¹ and 0.3%, respectively, whereas for GNR these values are significantly higher, being 1 A W⁻¹ and 80%, respectively, for an incident wavelength of 1550 nm at 2 V. All these results clearly demonstrate that RGO and GNRs are both promising candidates for use as high-selectivity, high-sensitivity, and high-speed nanometer-scale photodetectors and photoelectronic switches.

Response time is another important parameter of photodetectors. Figure6a shows the time response of photocurrent growth and decay (in the inset) for RGO in response to the IR illumination being turned on and off. The circles are the experimental points and the solid lines are a fit to the above equations. The dynamic response of our device to the IR source can be well described by I (t) = I_dark +A [exp(t/τ₁)] + B [exp(t/τ₂)] and I (t) = I_dark+ A [exp(−t/τ₁)] + B [exp(−t/τ₂)] for growth and decay, respectively, where τ is the time constant and t is the time when IR was switched on or off, I_dark is the dark current, and A and B are scaling constants. From these fits, the time constants for growth and decay were estimated. The current of the device rises within 2 s, the time resolution of measurement, on illumination. The rapid photocurrent rise is followed by a slower component, in which the photocurrent increases in about 26 s, before saturating. The photocurrent decay starts with a fast decay component, during which the photocurrent drops in the first 2–5 s after the excitation is turned off. This fast decay process follows a first-order exponential relaxation function with an estimated time constant of 2 s. A further slow photocurrent decay process lasts about 29 s, during which the current decreases to reach its initial dark-state value.

Figure 6. Time response of photocurrent growth and decay (in insets) for RGO (a) and GNRs (b). The empty circles are the experimental points and the solid lines are fits to the exponential equations.

Figure 6b shows the time response of photocurrent growth and decay (in inset) for GNRs in response to the IR illumination being turned on and off. The current of the device rises within 8 s when the illumination is turned on. The rapid photocurrent rise is followed by a slower component, in which the photocurrent increases in about 87 s, before saturating. The photocurrent decay starts with a fast decay component over a duration of 25 s after the excitation is turned off. Again, this fast decay process assumes a first-order exponential relaxation function with an estimated time constant of 23 s. Subsequently, a slow photocurrent decay process lasts for about 96 s until ultimately the initial state is reached. Reproducibility of all results reported here was confirmed by fabricating three devices using RGO and GNR from different synthesis batches. The devices were stable under ambient conditions with no degradation of device performance for over 3 months.

In conclusion, IR detection is demonstrated by both RGO and GNRs in terms of time-resolved photocurrent and photoresponse. The responsivity and external quantum efficiency of RGO are 4 mA W⁻¹ and 0.3%, respectively, whereas for GNR these values are significantly higher (ca. 1000 times), being 1 A W⁻¹ and 80%, respectively, for an incident wavelength of 1550 nm at 2 V. All these results clearly demonstrate that RGO and GNRs, but especially the later, can be effectively used as high-selectivity, high-sensitivity, and high-speed nanometer-scale photodetectors and photoelectronic switches, suggesting a strong impact in many future applications of graphene, including in nanoelectronics.

Experimental Section

Preparation: An aqueous GO suspension (4 mL H₂O, 3 mg GO/mL) was generated by sonication (1 h) of GO followed by addition of DMF (volume ratio DMF/H₂O = 9, resulting concentration = 0.3 mg GO/mL), resulting in a quite stable light-brown suspension of GO sheets.30 Chemical reduction of the suspension of GO sheets was carried out with hydrazine monohydrate (1 μL/3 mg GO) for 12 h at 80 °C with stirring by a poly(tetrafluoroethylene) (Teflon)-coated magnetic bar, creating a homogeneous black suspension.

Six-layer GNRs were obtained by oxidation followed by unzipping of double-walled and multiwalled (4–8 walls) carbon nanotubes (MWNTs).31 MWNTs were prepared by thermal decomposition of a CH₄ + Ar mixture over Fe_0.1Mg_0.9O catalyst at 750 °C. MWNTs were washed with HCl several times to remove the excess catalyst particles. These MWNTs (150 mg) were then suspended in a mixture of 9:1 H₂SO₄:H₃PO₄ (30 mL) by sonication and allowed to stand for 12 h. KMnO₄ (750 mg) was then added to this mixture while it was being stirred gently. The mixture was further stirred for 2 h at 85 °C and then poured into 400 mL of ice containing 5 mL of 30% H₂O₂. The solution was filtered and washed with water several times followed by coagulation in 20% HCl (30 mL) and further washing with water. Washing and coagulation cycles were repeated several times to remove catalyst and other impurities. The GNRs were treated with hydrazine hydrate in a dichloromethane/water mixture and heated at 60 °C for 12 h. The product was washed, dried, and heated at 250 °C in Ar for 4 h to remove functional groups.

Characterization: Raman spectra were recorded with LabRAM HR high-resolution Raman spectrometer (Horiba Jobin Yvon) using a He–Ne Laser (632.8 nm). TEM images were obtained with a JEOL JEM 3010 instrument. AFM measurements were performed using a NanoMan instrument.

Device fabrication: Two devices were fabricated by depositing electrodes on pre-cleaned 300 nm SiO₂ on Si substrate by evaporating Cr (5 nm)/Au (200 nm) using electron-beam evaporation. The distance between the two electrodes was 115 μm and the width of electrodes was 2 mm. RGO and GNRs were dispersed in DMF (ca. 3 mg mL⁻¹) by sonication for 30 min. The dispersion of RGO and GNRs was then dropcast between the electrodes using a micropipet, on the devices. The devices were dried overnight in vacuum to remove the residual solvent. Current –voltage characteristics of the devices were measured in the dark and IR in air using a Keithley 6430 meter. The IR source consisted of a semiconductor laser diode with peak wavelength of 1550 nm. The photocurrent was calculated by subtracting the dark current from the current under IR source illumination.

Acknowledgements

The authors thank Barun Das and Urmimala Maitra for assistance in preparing RGO and GNRs.