1. Introduction

Gallium Nitride (GaN), with a wide and direct bandgap, is one of the most widely used semiconductor compounds for ultraviolet (UV) detection [1–3]. The performance of traditional GaN-based photodetectors has been partially limited by the photocarrier collection efficiency due to large defect density and low conductivity of GaN. Meanwhile, the photosensitive area of GaN photodetectors is usually limited to ~0.1 mm² because of the low carrier mobility and short carrier lifetime, a disadvantage in applications such as imaging. To overcome these limitations, new types of GaN photodetectors are needed [4–6].

Graphene, a two-dimensional (2D) material, has attracted extensive attention due to its superior properties such as high carrier mobility and universal optical transmittance of about 97.7% for 300-1000 nm wavelength [7]. The 2D nature is also compatible with photonic integrated circuits [8]. A variety of prototype graphene photodetectors with different mechanisms, including photoconduction, photovoltaic, photo-thermoelectric, bolometric and plasma-wave-assisted mechanisms, have been demonstrated [9–16]. Responsivity of 3 mA/W was obtained in 200 μm × 400 μm GaN-based, metal-semiconductor-metal (MSM) type of UV photodetectors using graphene as transparent electrodes [4]. Another Schottky diode type of detectors made of graphene/un-doped GaN was reported [5]. The responsivity of these photodetectors is not high. Recently, a new type of phototransistors based on graphene hybrid structures has been developed [13–16]. The structures are basically composed of a graphene film and another light-absorbing semiconductor film. The light-absorbing film absorbs incident photons and converts them to electron-hole pairs. Under the built-in electric field at the graphene-semiconductor junction, one type of charge carriers will migrate into graphene and the other type remains in the semiconductor. If the carriers in the semiconductor have a long lifetime, the high mobility in graphene will lead to high carrier collection efficiency by the source-drain electrodes. Furthermore, the carriers remaining in the semiconductor will change the chemical potential of the semiconductor and act as a gate to alter the charge density in graphene, resulting in further current gain. These effects give a high ratio of photocurrent to light power, namely, the responsivity. In the visible to near-infrared spectral range, ultrahigh responsivity of 10⁷ A/W was obtained with a slow response time in graphene/Si photodetectors [14]. By employing a graphene/SiO₂/lightly-doped-Si architecture with an interfacial gating mechanism, both high responsivity of ~10³ A/W and fast response time of ~400 ns were achieved [15]. However for ultraviolet detection, such a hybrid phototransistor structure has yet to been applied and studied.

Here we report a hybrid graphene/GaN phototransistor as a UV detector which shows the responsivity of 361 mA/W and the response time of ~5 ms, both of which are much higher than traditional GaN photodetectors. Moreover, the remarkable performance is obtained in a 15.2 mm² active area, also much larger than previously reported GaN-based photodetectors. These results are promising for developing large area ultraviolet photodetectors with high responsivity and high speed that are ideal characteristics for applications like imaging.

2. Materials and methods

The schematic structures of our hybrid graphene/GaN phototransistors are shown in Fig. 1(a) and 1(b). The 2 μm thick unintentionally doped (un-doped) GaN film was grown on the surface of an AlN buffer layer on a 1μm thick sapphire substrate. Then a 200 nm thick Mg-doped p-GaN layer was grown on the un-doped GaN. Hall effect measurements of the p-GaN showed a hole concentration of 8 × 10¹⁷ cm⁻³ and a Hall mobility of 2.4 cm²/(V·s). The Ni /Au (5/5 nm) bi-layer recipe was used for ohmic contact with p-GaN [17], which was deposited by magnetron sputtering. Then a Ti/Au (30/300 nm) metal layer was deposited to form the contact pads and treated by rapid thermal annealing at 500 °C for 600 s in O₂. Monolayer graphene grown on 25-μm-thick Cu foil by chemical vapor deposition (CVD) was transferred, so graphene made contacts with both the metal pads and GaN [18]. A heterojunction is thus formed at the interface between graphene and GaN. The sample structure as shown in Fig. 1(a) will be referred as the plane structure. The graphene, bridging the two electrodes, behaves as a conducting channel. The channel length (L) and width (W) were about 3.8 mm and 4 mm, respectively. A similar device structure without graphene was also fabricated as the same time for comparison.

 

Fig. 1 (a) The plane structure with graphene. (b) The mesa structure with graphene. (c) Raman spectrum of the graphene sheet.

Download Full Size | PDF

Another type of structure was fabricated by photolithography and inductively coupled plasma (ICP) reactive ion etching. First, part of p-GaN was etched by ICP. The step height is ~280 nm, reaching the un-doped GaN layer. Then, a Ti/Al/Ti/Au (10/10/30/300 nm) and a Ti/Au (30/300 nm) layer were deposited on the un-doped GaN and p-type GaN by magnetron sputtering, respectively. Rapid thermal annealing at 850 °C for 30 s in N₂ was used for good contact [19]. Graphene was then transferred, making contacts with the metal pads, p-GaN and un-doped GaN. Figure 1(b) illustrates the cross-section of the structure, which we refer as the mesa structure. Similarly, a mesa structure without graphene was fabricated for comparison.

Figure 1(c) shows the typical Raman spectrum of graphene with an excitation wavelength of 633 nm (Renishaw). The intensity ratio of the 2D peak to the G peak is approximately 4:1, indicating monolayer graphene. The absence of the D peak suggests that our graphene has high quality with few defects [20].

All the electrical and optical properties of these structures were measured at room temperature in air. The optoelectronic performance was measured by Keithley 6517B. And a 325 nm He-Cd laser was applied as the illumination light source. The incident laser power on the device was calibrated by an S130VC optical power meter by THORLABS. The temporal response was characterized by a Rohde-Schwarz oscilloscope.

3. Results and discussion

The graphene film serves as a high mobility charge transport channel, and GaN absorbs the UV light. The energy band diagram of the hybrid graphene/GaN phototransistors of the plane structure is shown in Fig. 2(a). Charge transfer occurs at the graphene/GaN interface when they are in contact. The work function of the p-GaN is around 7.41 eV according to the doping concentration of Mg [21], and the work function of graphene is about 4.6 eV [5]. Therefore, the energy bands are bent, and a built-in field is established near the surface of p-GaN, which points from graphene to p-GaN. Upon UV light illumination, GaN absorbs the photons whose energy exceeds its bandgap and generates electron-hole pairs. Photon-induced carriers are then separated and transferred by the built-in field at the graphene/GaN interface. Electrons are driven into graphene, and collected by the electrodes to generate an output signal.

 

Fig. 2 (a) Energy band diagram of the plane structure with graphene. (b) Schematics of the photocarrier movement and the energy band in the mesa structure with graphene. (c), (d) Photocurrent versus voltage for the plane structure without and with graphene at various levels of incident light power, 100% is 5.83 mW. (e), (f) Photocurrent versus voltage for the mesa structure without and with graphene at various levels of incident light power. The insets show the dark current. For simplicity, Gr. denotes graphene.

Download Full Size | PDF

The optical absorption happens mainly in the p-GaN film, whose absorption coefficient α is ~10⁵ cm⁻¹ at 325 nm [22]. The net photocurrent I_photo can be obtained by subtracting the dark current from the total current as plotted in Fig. 2(c) and 2(d). The distance between the source and drain electrodes is about 3.8 mm. The UV light spot of ~1 mm in diameter was moved from one electrode to another to test the uniformity of the photoresponse over such a relatively large device area of 15.2 mm² and the results showed little difference.

Higher responsivity can be achieved by applying a larger source-drain voltage [9]. It is evident in Fig. 3(a) that the plane structure with graphene generates larger photocurrent (about 35 times) than that without graphene upon UV irradiation (0.02 mW at 10 V). The responsivity R is defined as I_photo/P_I, where P_I is the power of the incident light. The responsivity of the plane structure without graphene was around 6.7 mA/W at 0.02 mW. The responsivity with graphene was ~229 mA/W, much better than the graphene/GaN MSM UV sensors [4].

 

Fig. 3 Photoresponse characters of devices for the light at wavelength of 325 nm. (a) Photocurrent and responsivity versus illumination power at the applied voltage of 10 V of plane structure without (w/o) and with (w) Gr.. (b) Photocurrent and responsivity versus illumination power at the applied voltage of 10 V of mesa structures without and with Gr..

Download Full Size | PDF

Gain of the device with graphene originates from the high carrier mobility of graphene. The photocarriers in graphene move faster and can circulate multiple times through the circuit during the lifetime of the carriers remaining in p-GaN, giving a large photocurrent. The photocarriers in p-GaN can also lead to a gating effect, where the presence of these carriers changes the carrier concentration in the graphene sheet [14–16]. Without graphene, all the photogenerated carriers would remain in p-GaN with a much lower mobility, thus giving a much lower photocurrent. Therefore, the responsivity of the plane structure with graphene is much higher than that without graphene.

However, the above mobility-based enhancement effect has limitations that can be estimated as follows. For every absorbed photon, the photocurrent I = βηeμE/L_circuit, where β is the carrier number amplification factor, η is the internal quantum efficiency, μ is the electron mobility in the circuit, E is the electric field determined by the source-drain voltage, L_circuit is the length of the circuit. During the carrier lifetime, high μE/L_circuit ratio will make the electron travel round the circuit many times (per second). Taken typical values β = 1, η = 1, μ = 10³ cm²/Vs, E = 1 V/cm (in circuit wire), and L_circuit = 10 cm, we get a current of 1.6 × 10⁻¹⁷A and a responsivity of ~27 A/W. So without a large carrier density amplification factor β, the responsivity will not get much larger. Nevertheless, since the carrier mobility in graphene is 2~3 orders higher than that in GaN, the responsivity enhancement is quite significant.

The responsivity decreases with increasing light power as shown in Fig. 3(a) and 3(b). This is a consequence of the reduction of the built-in field with increasing number of photogenerated carriers, which accumulate and form an electric field opposite to the built-in field [9]. As the incident light power increases, the net built-in field becomes weaker and cannot separate the photogenerated carriers as efficiently.

Next, we studied how the variation of GaN substrates affects the performance of the hybrid phototransistors. Notably, while the doping type, doping level and defect density of GaN substrates greatly alter the performance of the resulting photodetectors without graphene, they show little effects on the performance of the resulting phototransistors with graphene. These loose requirements on GaN can help to lower the device cost and get higher yield. The mesa structure is such an example as shown in Fig. 2(b). On the p-GaN side, the graphene/p-GaN interface is the same as that in the plane structure. On the un-doped GaN side, the graphene/un-doped GaN interface also has a built-in field (pointing from un-doped GaN to graphene) due to the band bending. As shown in Fig. 2(e) and 2(f), the photocurrent of mesa structure with graphene was 7.2 μA at 10 V with the incident power of 0.02mW (0.4% power level), ~700 fold larger than that without graphene. Correspondingly, the responsivity of mesa structure with graphene is 361.3 mA/W at 0.02 mW, while that without graphene is 0.5 mA/W. If the incident light is on the p-GaN side, the photogenerated electrons transfer to graphene and the holes stay in GaN. If the incident light is on the un-doped GaN side, the photogenerated holes transfer to graphene and the electrons stay in GaN. Either case due to the high charge mobility and fast carrier transport in the graphene channel, the time for the photocarriers to reach the electrodes can be much shorter than the recombination time. The photocarriers in graphene can circulate multiple times through the circuit during the lifetime of the carriers remaining in GaN, giving a large photocurrent, just as in the plane structure. So, with graphene, the two structures give comparable responsivities. Without graphene, the responsivity of the mesa structure is even lower than that of the plane structure because the mobility in un-dope GaN is much lower than that in p-GaN.

To quantify the efficiency of light conversion to current, we extracted the external quantum efficiency (EQE), that is, the ratio of the number of photogenerated charge carriers to the number of incident photons. The EQE can be determined by (I_photo/q)/(P_I/hν), where q is the charge constant, h is the Planck constant and ν is the frequency of the incident light [23]. The EQE of the plane structure with graphene is ~88% at 10 V under 0.02mW incident light, which is 34 times than that without graphene (2.57%). The EQE of the mesa structure with graphene is similar (~138%), which is about 720 times than that without graphene. These high EQE values indicate gains by the multiple circulation effect due to the high carrier mobility of graphene.

The response time of the phototransistors was measured as shown in Fig. 4. Figure 4(a) plots the photoresponse of the plane structure with graphene when switching on/off the UV light. The rising time (τ_r) is measured from 10% to 90% of the signal peak value. The decay time (τ_f) is from 90% to 10%. The τ_r and τ_f of the plane structure with graphene are 5.05 ms and 5.11 ms, respectively. The response is faster than the non-polar GaN based UV photodetector, whose response time is hundreds of ms [6]. Major factors contributing to the high speed of the device include the high carrier mobility in graphene and the quick separation of the photogenerated carriers by the built-in electric field at the interface. Figure 4(b) plots the response time of the mesa structure with graphene. The τ_r and τ_f are 3.2 and 1.1 ms, respectively.

 

Fig. 4 The response time under 5.83mW incident light power. (a) τ_r is 5.05 ms and τ_f is 5.11 ms in the plane structure with graphene. (b) τ_r is 3.2 ms and τ_f is 1.1 ms in the mesa structure with graphene.

Download Full Size | PDF

Another widely used parameter for photodetectors is specific detectivity [24], which represents the sensitivity of the device and is defined by $D^{*}=\sqrt{A\Delta f}/NEP=R•\sqrt{A\Delta f}/i_{noise}$, where R is the responsivity, A is the active area of the photodetector, Δf is the working frequency, NEP is the amount of light equivalent to the noise level of the device which is the light level required to obtain a signal-to-noise ratio of unity, i_noise is the noise current under the same conditions. A noise current is associated with current fluctuation and could be directly calculated from the dark current. Thus, D* is expressed by$D^{*}=R\sqrt{A}/\sqrt{2q{i}_d}$, where q is electron charge, and i_d is the dark current of the device. Because of the high responsivity, D* of the plane structure with graphene is 1.5 × 10¹⁰ Jones (cmHz^1/2W⁻¹) at 10 V, ten times that without graphene. D* of the mesa structure with graphene is 2.3 × 10¹⁰ Jones at 10 V.

Table 1 compares the photodetection performance (including the responsivity, photosensitive area and response time) of this work with some previously reported GaN-based UV detectors. It can be seen that the overall performance of our devices exhibits significant improvements.

Table 1. Performance comparison of GaN-based photodetectors. “$■$” indicates traditional GaN-based photodetectors, and “$■$” indicates graphene/GaN photodetectors.

View Table

4. Conclusion

In summary, hybrid graphene/GaN UV phototransistors of relatively large 15.2 mm² photosensitive area were fabricated and studied. The responsivity of the plane structure with graphene is 229 mA/W, 35 times higher than the plane structure without graphene. The EQE and D* are 87.5% and 1.5 × 10¹⁰ Jones at 10V, respectively. The responsivity of the mesa structure with graphene is 361 mA/W, about 700 times higher than that without graphene. The response time of these phototransistors is ~5 ms, much faster than previously reported GaN-based UV photodetectors. The excellent performance does not have special requirements on GaN parameters such as the doping type, doping level or defect density. So our work provides a feasible way to develop large area UV photodetectors with high responsivity and high speed.