High-performance graphene photodetector using interfacial gating

1. INTRODUCTION

Graphene-based photodetectors have aroused considerable interest, and various types of device configurations and mechanisms have been developed [1–8]. Current prototype devices have shown outstanding performance with individual functionalities aiming for different applications, that is, ultrafast or ultrasensitive detection. On the fast-detecting side, benefiting from high mobility and ultrafast carrier dynamics, the intrinsic graphene-based photodiode has shown photoresponse on a femtosecond (fs) timescale [2]. On the ultrasensitive side, by employing the photogating mechanism, the hybrid graphene photoconductor has exhibited ultrahigh gain up to ${10}^{10}$ [8]. However, there is a huge gap between the two mechanisms, like two sides of a coin: fs detection has a responsivity of only milliamps per watt, while picowatt detection responds only in milliseconds to seconds, and numerous applications such as optical positioning, remote sensing, and biomedical imaging desire both speed and sensitivity. The gap between the binary performances is limited by the current mechanisms employed. Fast detecting relies on the high carrier mobility of intrinsic graphene and suffers from its gapless nature and low efficiency of electron–hole pair disassociation, while the bottleneck of photogating is the slow charge transfer and/or charge trapping process on a time scale of milliseconds or even seconds [3,5,7–17], which is indeed necessary for the charges in the channel to recirculate between source and drain, to give rise to ultrahigh gain.

In this work, by adopting a new concept of the interfacial gating effect from a lightly doped silicon (Si)/silicon dioxide (${\mathrm{SiO}}_2$) interface, we successfully bridge the gap between ultrafast response and ultrasensitivity in a graphene-based photodetector. This device architecture separates the photoexcited electron–hole pairs with an intrinsic self-built electric field at the $\mathrm{Si}/{\mathrm{SiO}}_2$ interface, and in turn the accumulated charges at the interface gate the graphene and introduce a high gain in photoresponse by taking advantage of the high mobility of graphene. This charge-transfer-free strategy with fast accumulation of the photoexcited carrier at the interface ensures a fast response of the photocurrent at the graphene channel. Moreover, the current device structure does not require a complicated fabrication process and is fully compatible with silicon technology.

2. DEVICE FABRICATION AND METHODS

A. Device Fabrication

Monolayer graphene samples are mechanically exfoliated from highly oriented pyrolytic graphite and deposited on lightly p-doped Si substrate that is terminated with 300 nm of ${\mathrm{SiO}}_2$. Source and drain electrodes (5 nm Ni adhesion layer, followed by a 50 nm Au capping layer) are defined using electron beam lithography (FEI, FP2031/12 INSPECT F50) and deposited by thermal evaporation (TPRE-Z20-IV). More than 10 devices are fabricated, and all of them show very good photoresponse behavior. In addition, other control devices on 300 nm ${\mathrm{SiO}}_2/\text{heavily}$ doped Si, and lightly doped Si covered by ${\mathrm{SiO}}_2$ with different thicknesses or 30 nm ${\mathrm{Al}}_2{\mathrm{O}}_3$ are also fabricated. The ${\mathrm{Al}}_2{\mathrm{O}}_3$ layer is deposited using atomic layer deposition (Sunaletmr-100).

B. Photoresponse Measurement

Electrical and photoresponse characteristics of the devices are measured using a Keithley 2612 analyzer under dark and illuminated conditions. Light is switched on and off using an optical chopper or acoustic optical modulator (R21080-1DS) at different frequencies. The light source is an ${\mathrm{Ar}}^{+}$ laser with a wavelength of 514 nm. A supercontinuum light source (SuperK Compact nanosecond kilohertz) is employed to attain the spectral photocurrent response. In all the photocurrent measurements, the laser and supercontinuum light are focused on the sample with a $50\times$ objective ($\mathrm{NA}=0.5$), and the spot size of light is $\sim 1\mathrm{\mu m}$, much smaller than the graphene channel length. In the power-dependent experiment, optical attenuators are introduced to change the input power. A digital storage oscilloscope (Tektronix TDS 1012, 100 MHz 1 GS/s) is used to measure the transient response of photocurrent.

3. RESULTS AND DISCUSSION

A. Characterization of Graphene Photodetector

Figure 1(a) shows the schematic diagram and a representative optical microscopy image of our device. A lightly p-doped silicon wafer (1–10 Ω cm) and a thermally grown 300 nm thick ${\mathrm{SiO}}_2$ layer are employed as the gate electrode and dielectric, respectively. We have compared the Si substrates with different doping concentrations, and the abovementioned one provides the best device performance. The mechanically exfoliated monolayer graphene is characterized by optical contrast and Raman spectroscopy (see Fig. S1a in Supplement 1) [18]. The G band and 2D band (with a FWHM of $27.7{\mathrm{cm}}^{-1}$) are located at 1580.3 and $2675.2{\mathrm{cm}}^{-1}$, respectively, and the ratio of $I_{2D}/I_G$ is $\sim 2.3$. Figure S1b in Supplement 1 shows a transfer characteristic of the device at an applied bias voltage of $V_D=10\mathrm{mV}$, measured in the dark, and suggests that the graphene is slightly p-doped because of interactions with the substrates and the absorbed water/oxygen molecules in air. The estimated hole mobility ($\mu$) is $\sim 5000{\mathrm{cm}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$. All these features indicate the high quality of the monolayer graphene and the device.

 

Fig. 1. Graphene photodetector based on interfacial gating. (a) Schematic diagram and optical image of the graphene photodetector on lightly p-doped $\text{silicon}/{\mathrm{SiO}}_2$ substrate; (b), (c) energy band diagrams of the lightly p-doped $\text{silicon}/{\mathrm{SiO}}_2$ interface with positive localized states ($q{\varphi }_0$) and its effect on graphene, respectively. The accumulation of photogenerated electrons (blue points) at the interface results in additional negative voltage under light illumination, lowers the Fermi level ($E_{f(\mathrm{Gr})}$) to its new position ($E_{f(\mathrm{Gr})}^{\prime }$), and results in a light-induced p-type doping in graphene.

Download Full Size | PPT Slide | PDF

B. Mechanism of Photodetector based on Interfacial Gating Effect

The working principle of our graphene photodetector can be understood through the energy band diagram of the oxide–silicon interface and its effect on graphene, as shown in Figs. 1(b) and 1(c). The localized interface states such as positive charge states ($q{\varphi }_0$) with energies within the silicon bandgap exist at the oxide–silicon interface, and induce a negative depletion layer (−) in the silicon near the interface and the formation of a built-in electric field (E) [19]. Due to the presence of the built-in electric field, the photogenerated electron–hole pairs in lightly p-doped Si will be separated: the holes (red points) in the valence band of Si diffuse toward the bulk Si, while the electrons (blue points) accumulate at the ${\mathrm{SiO}}_2/\mathrm{Si}$ interface [Fig. 1(b)]. This leads to the appearance of a negative voltage at the interface, which can be negligible in heavily doped silicon due to the very short lifetime of the photogenerated carriers [20]. As a result, the additional negative voltage could effectively gate the graphene channel through capacitive coupling, and lower the Fermi level ($E_{f(\mathrm{Gr})}$) to its new position ($E_{f(\mathrm{Gr})}^{\prime }$), as shown in Fig. 1(c). Therefore, an increase in hole density and high positive photocurrents in graphene are achieved.

C. Photoresponse of Graphene Photodetectors

The photoresponse characteristics at $V_D=1\mathrm{V}$ and zero gate voltage ($V_G=0\mathrm{V}$) are recorded with a laser focused on the device at wavelength of 514 nm (spot size $\sim 1\mathrm{\mu m}$). Figure 2(a) shows the photoresponse of the channel current under varying laser power, where positive photocurrents are observed when light is switched on. The dependence of the photocurrent as a function of light power is shown in Fig. 2(b). This photocurrent (at microamp scales) is large enough for direct measurement without any amplifier, even at a very low light power ($\sim 0.6\mathrm{nW}$). It should be noted that the photocurrent is saturated with the increase in light power. This is because the accumulation of photogenerated electrons at the ${\mathrm{SiO}}_2/\text{silicon}$ interface leads to a reversed electric field balance to the equilibrium built-in field. Correspondingly, less photoinduced electron–hole pairs will be separated as the net built-in field becomes weaker under higher illumination power. The responsivity of the device under varying light power is calculated and shown in Fig. 2(c), which is defined as $R=I_{\mathrm{ph}}/P$ where $I_{\mathrm{ph}}$ and $P$ are the photocurrent and incident light power, respectively. The device shows a remarkable responsivity up to $\sim 1000{\mathrm{AW}}^{-1}$ at an incident light power of $\sim 0.6\mathrm{nW}$, which is among the highest values previously reported for monolayer graphene photodetectors [1,2,5,21–24]. Based on this value of $R$, we also estimate external quantum efficiencies $\mathrm{EQE}{\eta }_e=R[h\upsilon /e]=2.42\times {10}^5\%$ and a specific detectivity $D^{*}={(A\mathrm{\Delta }f)}^{0.5}R/i_n=1.1\times {10}^{10}$ Jones ($1\text{Jones}=1\mathrm{cm}{\mathrm{Hz}}^{1/2}{\mathrm{W}}^{-1}$) at $V_D=1\mathrm{V}$ and $V_G=0\mathrm{V}$, where $e$ is electron charge, $h$ is Planck’s constant, $\upsilon$ is frequency of light, $A$ is the effective area of the device, $\mathrm{\Delta }f$ is the electrical bandwidth, $R$ is the responsivity, and $i_n$ is the noise current (the dark current waveform of the device is shown in Fig. S2 of Supplement 1). Figure 2(d) shows the spectral photocurrent response of the device at $\sim 0.05\mathrm{\mu W}$ light power from the visible to near-IR, which is obtained by a supercontinuum light source with a tunable filter. It can be seen that the excitation wavelength dependence of the photocurrent is almost flat in the visible regime and drops abruptly beyond $\sim 1050\mathrm{nm}$. This is a predictable outcome of the photoresponse mechanism proposed above, i.e., lightly doped silicon is indeed the light absorbing medium, with a photon response range from $\sim 200$ to 1100 nm. This also implies that by replacing silicon with other semiconductors, the photoresponse of the device can be further extended to a longer wavelength, e.g., HgCdTe with an adjustable bandgap (from 0.7 to 25 μm) for mid-IR photodetection [25].

 

Fig. 2. Photoresponse characteristics as a function of light power and wavelength. (a) Photoswitching characteristics of the graphene photodetector under varying light power; (b), (c) photocurrent and responsivity at $V_D=1\mathrm{V}$ and $V_G=0\mathrm{V}$ of the device as a function of the light power. The laser wavelength is 514 nm. (d) The spectral photocurrent response of the device at $\sim 0.05\mathrm{\mu W}$ light power from 450 to 1150 nm.

Download Full Size | PPT Slide | PDF

Next, we characterize the transfer characteristics of the device at $V_D=10\mathrm{mV}$ under varying light power (with a wavelength of 514 nm), as shown in Fig. 3(a). The transfer curves along with the Dirac points shift toward more positive gate voltage with increasing light power. The inset is an enlarged view of the region in the dashed rectangle in Fig. 3(a), showing the increasing light current with increasing light power. This could be understood as a result of the additional negative gating effect caused by accumulated photogenerated electrons at the lightly doped $\text{silicon}/{\mathrm{SiO}}_2$ interface. In other words, a higher gate voltage is needed to obtain the charge neutrality point (Dirac point) in the graphene device. From these curves, we extract the shift of the Dirac point ($\mathrm{\Delta }{V}_G$) as a function of light power in Fig. 3(b) (red points), which reaches a saturated value of $\sim 0.35\mathrm{V}$ at a power of $\sim 10\mathrm{\mu W}$. This corresponds to a modulated charge carrier density of $\mathrm{\Delta }n=C_g\mathrm{\Delta }{V}_G/e=\sim 2.52\times {10}^{10}{\mathrm{cm}}^{-2}$, where $C_g$ is the dielectric capacitance ($1.15\times {10}^{-8}\mathrm{F}{\mathrm{cm}}^{-2}$). Although the change in the light-induced carrier density in graphene is less significant, the resulting photocurrent is rather considerable because of the high carrier mobility of graphene. The channel current change ($\mathrm{\Delta }I$) of the device under light illumination is defined by [11]

 

Fig. 3. Gate- and bias-modulated photoresponse. (a) $I\text{-}{V}_G$ characteristics of the device under varying light power; the inset is an enlarged view of the region in the dashed rectangle, showing the increase in the light current under illumination with increased light power. (b) Horizontal shift of the Dirac point ($\mathrm{\Delta }{V}_G$) and the modulated charge carrier density ($\mathrm{\Delta }n$) as a function of light power, (c) the extracted gate dependence of photocurrents ($I_{\text{light}}-I_{\text{dark}}$) from the curves in (a). The red circles represent the photocurrents of the device at individual gate voltages under a fixed light power of 3.66 μW. (d) Photocurrents at $V_G=0\mathrm{V}$ of the device as a function of $V_D$ under varying light power. The wavelength is 514 nm.

Download Full Size | PPT Slide | PDF

According to the photodetection mechanism described in Fig. 1, the interfacial accumulated carriers will also diffuse in bulk Si in the lateral direction. To confirm this, we show the spatial dependence of the photocurrent as a function of the light position moving away from the graphene channel, as shown in Fig. S4 of Supplement 1. We find that the photocurrent still exists, even when the light spot is not on the graphene channel, but on the ${\mathrm{SiO}}_2$ substrate, similar to a photodetector based on a graphene/Si junction [27]. The measured photocurrent is reduced when the light is directed away from the device under the same power, based on the diffusion equation $L_n=\sqrt{D_n{\tau }_n}$ where $L_n$ is the diffusion length of excess carriers, $D_n$ is diffusion coefficient, and ${\tau }_n$ is the lifetime of the carriers. In the case of lightly doped silicon with $N_A=\sim {10}^{15}{\mathrm{cm}}^{-3}$, ${\tau }_n=\sim 200\mathrm{\mu s}$, and $D_n=\sim 35{\mathrm{cm}}^2{\mathrm{s}}^{-1}$, the calculated diffusion length $L_n$ is $\sim 830\mathrm{\mu m}$, which is consistent with the experimental results shown in Fig. S4 of Supplement 1. On the other hand, we fabricated a control device with graphene lying on a heavily p-doped $\mathrm{Si}/{\mathrm{SiO}}_2$ substrate (resistivity $\sim {10}^{-3}\mathrm{\Omega }\mathrm{cm}$). No obvious photocurrent could be resolved when the light ($\sim 1.8\mathrm{\mu W}$) was switched on or off (see Fig. S5 in Supplement 1). The results clearly demonstrate that photocurrents in our device do not result from intrinsic photogenerated carriers in graphene, but rather originate from the modulated charge carriers in graphene due to the interfacial gating effect. More important, it is demonstrated that the lightly doped $\text{silicon}/{\mathrm{SiO}}_2$ interface plays a critical role in the photosensitive behavior of our device. We have also studied the effects of dielectric thickness and materials (e.g., ${\mathrm{Al}}_2{\mathrm{O}}_3$) on the device performance (see Figs. S6 and S7 in Supplement 1). The results show that the thickness of the dielectric layer has no obvious influence on the device performance, because it does not affect the accumulated photoexcited charges at the interface. The photoresponse of the device on 30 nm ${\mathrm{Al}}_2{\mathrm{O}}_3/\text{lightly}$ doped Si substrate is also observed, and the responsivity is much lower compared to that of the devices on ${\mathrm{SiO}}_2/\text{lightly}$ doped Si substrate.

D. Fast Response Time

Figure 4(a) shows the transient response of the device when light (514 nm, $\sim 0.05\mathrm{\mu W}$) is switched on or off by an acoustic optical modulator with a frequency of 10 kHz. The rise (${\tau }_{\text{on}}$) and fall (${\tau }_{\text{off}}$) time are calculated to be $\sim 400$ and $\sim 760\mathrm{ns}$, respectively, based on curve fits of the transients with an exponential function. Such an ultrafast response speed is superior to other graphene-based photoconductors with a photogating mechanism [3,5–8,10–17]. More interesting, the response time of our device increases very slowly with decreasing light power, as shown in Fig. 4(b). On the other hand, although ultrahigh responsivity has been achieved in graphene-based hybrid structures and/or heterostructures, a significant increase in response time with decreasing light power is commonly observed [24]. This behavior was previously observed in phototransistors based on organic/inorganic composites [28], and could be associated with the decay of the transfer rate of electrons and/or holes from the light-absorbing materials to the conducting materials, especially in the case of a weak light signal. The high-speed response of our device is attributed to the fast separation of the electron–hole pairs assisted by the built-in electric field at the lightly doped $\text{silicon}/{\mathrm{SiO}}_2$ interface. Specifically, the holes are quickly driven into the bulk Si before they recombine with the accumulated electrons, wherein there does not exist the charge transfer process found in common graphene-based hybrid photodetectors [3,5–17]. With the aim of further investigating the high-speed photodetection of the device, we also show the time dependence of the photoresponse at a high modulation frequency of 0.5 MHz under varying light power, as shown in Figs. 4(c) and 4(d). It is demonstrated that our device can resolve weak signals at the nanowatt level under high frequency operation, which is promising for high-speed weak signal detection. (Experimental results from an additional device created on 300 nm ${\mathrm{SiO}}_2/\text{lightly}$ doped Si substrate are also shown in Fig. S8 in Supplement 1).

 

Fig. 4. Transient response of the device. (a) The transient response of the device switched on or off by an acoustic optical modulator with a frequency of 10 kHz. $P=\sim 0.05\mathrm{\mu W}$, $V_D=1\mathrm{V}$, and $V_G=0\mathrm{V}$. (b) The response time as a function of light power for our device and other graphene-based photogating devices (e.g., graphene/graphene quantum dots (G/GQDs) [10], graphene double layer structure (G/G) [5], graphene/SiC (G/SiC) [24], graphene/R6G (G/R6G)[16], graphene/Bi2Te3 (G/Bi2Te3) [7], graphene/Perovskite (G/Perovskite) [12], G/quantum dots (G/QDs) [3], graphene/MoS2 (G/MoS2) [14], graphene/carbon nanotubes (G/CNTs) [6], and graphene/Si (G/Si) [17]) reported in the literature; (c), (d) photoswitching characteristics of the device at 0.5 MHz modulation frequency under varying light power. The light wavelength is 514 nm.

Download Full Size | PPT Slide | PDF

4. CONCLUSION

In summary, by taking advantage of the interfacial gating effect from a lightly doped $\text{silicon}/{\mathrm{SiO}}_2$ interface, we demonstrate a simple approach to graphene photodetection with high responsivity and fast response. The proposed graphene photodetector exhibits a high responsivity of $\sim 1000{\mathrm{AW}}^{-1}$ for a weak signal of $<1\mathrm{nW}$ and a spectral response that extends from the visible to near-IR. More important, the photoresponse time of our device has been pushed to $\sim 400\mathrm{ns}$ and degrades quite slowly with decreasing light power, which is superior to other graphene-based photogating devices. Moreover, compared with previous graphene-based devices with top gated $p\text{-}n$ junctions [29,30], integrated with optical structures (e.g., plasmonic architecture [22], optical cavity [23], and waveguide [31]) and combined with light-absorbing materials (e.g., 2D van der Waals crystals [4,7,10,14,15], quantum dots [3,8,11], nanowire/tube [6,9]), our device possesses the advantage of a simple fabrication process and is fully compatible with the silicon technology. This work therefore not only opens up a route to graphene-based high-performance optoelectronic device, but also provides the potential to access an even wider spectral range by combing graphene with another oxide–semiconductor system.