I. INTRODUCTION

Ultraviolet (UV) detection is widely used in various application fields including industry, military, medical application, etc^[1-4]. Gallium nitride (GaN) and aluminum gallium nitride (AlGaN) have large energy band gaps ( > 3.4 eV) suitable for UV detection and their strong physical properties allow harsh environment applications^[5-13].

In comparison with conventional p-n junction-based photodiodes, phototransistors have higher photoresponsivity characteristics due to the amplification capability^[14]. In AlGaN/GaN heterostructures, a two-dimensional electron gas (2DEG) channel is formed at the interface due to the built-in electric field induced by strong piezoelectric and spontaneous polarizations^[15]. The confined 2DEG channel has high electron mobility (~2000 cm²/Vs) that can achieve fast response speed. Most of all, a dual absorption process in AlGaN/GaN structure can enhance the photoresponsivity in a broad wavelength range. In this study, we investigated the dual absorption characteristics of AlGaN/GaN heterojunction phototransistors fabricated with different recess depths for the detection area.

II. DEVICE STRUCTURE AND FABRICATION

The epitaxial structure consisted of a 40 nm Al_0.23Ga_0.77N layer, a 500 nm i-GaN layer, a 1 μm GaN buffer layer on a Si(111) substrate. After solvent and acid cleaning processes, the surface was passivated with a 60 nm $SiN_{x}$ film using hot-wire chemical vapor deposition in order to protect the surface during high temperature ohmic process. The ohmic contacts were formed by a recessed ohmic process^[16] in which the $SiN_{x}$ film was removed by SF₆ based plasma etching and the exposed AlGaN layer was partially recessed by Cl₂-BCl₃ based plasma etching. The ohmic metal stack was Ti/Al/Ni/Au that was annealed at 820℃ for 30sec in N₂ ambient. Mesa isolation was carried out by Cl₂-BCl₃ based plasma etching. The pre-passivation SiNx film was then removed by SF₆ based plasma etching. The contact resistance and sheet resistance were 0.49 Ω·mm and 317 Ω/sq., respectively. Three samples were prepared with different AlGaN thicknesses using a low-power Cl₂-BCl₃ based plasma etching process; (i) a 40 nm thick AlGaN layer on GaN (i.e. without recess), (ii) a 20 nm thick AlGaN layer on GaN (i.e., with recess), and (iii) fully-recessed GaN (i.e. complete removal of AlGaN layer). The etch rate was 0.1 nm/sec and the etch depth was confirmed by atomic force microscopy measurement. The cross-sectional schematics of three phototransistors are shown in Fig. 1. Finally, a Ti/Au metal stack was evaporated for pad electrodes. The distance between two ohmic electrodes was 10 μm and the channel width was 50 μm (i.e., detection area = 10 × 50 μm²).

Fig. 1. Cross-sectional schematics of AlGaN/GaN phototransistors with different recess depths; (a) 40 nm AlGaN, (b) 20 nm recessed-AlGaN, and (c) fully-recessed GaN.

III. RESULT AND DISCUSSION

Fabricated phototransistors were characterized using a testing setup shown in Fig. 2. The UV light source was a Deuterium lamp and a monochromator was used to select the wavelength of light. The dark current and photocurrent characteristics measured for each phototransistor are compared in Fig. 3 where the photocurrent characteristics were measured at the wavelength of 340 nm that is transparent for Al_0.23GaN_0.77N but can be absorbed in GaN layer. The AlGaN/GaN phototransistor with a 20 nm AlGaN layer exhibited lower dark current characteristics in comparison with the phototransistor with a 40 nm AlGaN layer. This is due to the reduced 2DEG density with a thinner AlGaN layer. When the AlGaN layer is completely removed, no 2DEG channel is formed. Therefore, the highly resistive GaN channel was responsible for the extremely lower dark current characteristics of the fully-recessed GaN phototransistor. In terms of the photocurrent characteristics, very small photocurrent was measured for the fully-recessed GaN phototransistor whereas significantly higher photocurrents were measured for AlGaN/GaN phototransistors. Since the AlGaN layer is transparent for the wavelength of 340 nm, the light absorption occurs only in GaN region for all three samples. The significantly different behavior between AlGaN/GaN and GaN phototransistors can be explained as follows. The energy band diagrams for AlGaN/GaN heterostructure and GaN are compared in Fig. 4(a) and Fig. 4(b), respectively. When the electron-hole pairs are created in the GaN region of AlGaN/GaN heterostructure, the electrons move to the 2DEG channel due to the built-in electric field and then travel to the ohmic region with an applied bias voltage while the holes are accumulated in the deeper GaN region acting as positive back gating effects that further enhance the 2DEG channel density. Depending on the absorption depth, electron-hole pairs can also be generated in the deeper neutral GaN region, but they are immediately recombined with little contribution to the photocurrent^[7]. On the other hand, the fully-recessed GaN structure does not have a 2DEG channel and the energy band bending at the surface makes the electrons generated in the surface depletion region tend to move toward the deeper neutral GaN region rather than traveling toward the ohmic region. Most of the electrons that moved to the neutral GaN region are recombined before reaching the ohmic region. Therefore, much lower photocurrent was measured for the fully-recessed GaN phototransistor.

Fig. 2. Block diagram for phototransistor characterization.

Fig. 3. Dark current and photocurrent as a function of bias voltage. The wavelength of incident light was 340 nm. (a) AlGaN/GaN phototransistors with and without recess and (b) fully-recessed GaN phototransistor.

Fig. 4. Energy band diagrams for (a) AlGaN/GaN heterostructure, (b) GaN.

The photocurrent characteristics as a function of wavelength were measured from 260 to 420 nm with the bias voltage of 5 V. The measured dark current and photocurrent characteristics are shown in Fig. 5. More than two orders of magnitude higher photocurrents were measured for the AlGaN/GaN phototransistors in comparison with the GaN phototransistor in the entire wavelength range. It should be noted that the photocurrent profiles as a function of wavelength are different between AlGaN/GaN and GaN phototransistors. The photocurrent decreased monotonically below ~300 nm for the fully-recessed GaN phototransistor whereas it increased below ~300 nm for the AlGaN/GaN phototransistors; a broader wavelength range can be detected with high photocurrents by the AlGaN/GaN phototransistors. The different behavior is attributed to the additional light absorption in AlGaN layer. The cut-off wavelength corresponding to Al_0.23Ga_0.77N is 317 nm^[17]. Light with a wavelength of < 317 nm can be absorbed in not only the top AlGaN layer but also the GaN layer underneath the AlGaN layer as long as the absorption depth is thicker than the AlGaN thickness. Owing to the energy band bending profile in AlGaN/GaN heterostructure (see Fig. 5(a)), the electrons created by photon absorption in both AlGaN and GaN regions move to the 2DEG channel and travel to the ohmic region. On the other hand, the holes created in AlGaN and GaN regions move away from the 2DEG channel to the surface and substrate, respectively, acting as the positive bias effects. Consequently, the 2DEG density is further enhanced at wavelengths below ~300 nm enhancing the photocurrent of AlGaN/GaN phototransistors.

Fig. 5. Dark current and photocurrent as a function of wavelength. The bias voltage was 5 V (a) AlGaN/GaN phototransistors with and without recess, (b) fully-recessed GaN phototransistor.

The photoresponsivity is expressed by ($I_{photo}$-$I_{dark}$)/$P_{in}$ where $I_{photo}$ is the photocurrent, $I_{dark}$ is the dark current, and $P_{in}$ is the incident light power. The incident light power was measured using a calibrated UV-enhanced Si photodiode. The calculated photoresponsivity characteristics for three samples are compared in Fig. 6. The dual absorption mechanism in AlGaN/GaN heterostructure is responsible for the enhanced photoresponsivity at the short wavelength regime for the AlGaN/GaN phototransistors whereas the limited photocurrent resulted in extremely lower photores-ponsivity for the fully-recessed GaN phototransistor. The recessed-AlGaN/GaN phototransistor exhibited a maximum photoresponsivity of 1.6 ×10⁷ A/W at 375 nm with a comparable photoresponsivity at 260 nm. In comparison with the unrecessed-AlGaN/GaN photo-transistor, the recessed-AlGaN/GaN phototransistor exhibited ~30% improvement in photoresponsivity in the entire wavelength range (see Fig. 6(A)). This is attributed to the thin AlGaN barrier effects; higher channel modulation with the change in surface potential (similar effects as the higher transconductance characteristics for AlGaN/GaN heterojunction field-effect transistors with a thin AlGaN layer). It is suggested that further improvement can be achieved by structural optimization.

Fig. 6. Comparison of photoresponsivity characteristics as a function of wavelength. (a) AlGaN/GaN phototransistors with and without recess and (b) fully-recessed GaN phototransistor.

IV. CONCLUSIONS

We investigated the characteristics of AlGaN/GaN heterojunction phototransistors in comparison with GaN phototransistor fabricated by complete removal of the AlGaN layer. Superior photoresponsivity in conjunction with a broader detection wavelength range was achieved by AlGaN/GaN heterojunction phototransistor. When the AlGaN layer was partially recessed, the photo-responsivity was noticeably enhanced in the entire wavelength range due to the improved modulation effects. A maximum photoresponsivity of 1.6 × 10⁷ A/W was achieved at 375 nm with a comparable photoresponsivity at 260 nm. Further improvement in photoresponsivity can be achieved by careful optimization of Al mole fraction and AlGaN thickness for the target wavelength. It is concluded that a single AlGaN/GaN heterostructure phototransistor can be used as a broad range UV detector covering from UV-A to UV-C.