Ultraviolet (UV) detection is an important technology in military and civil applications, such as missile warning, fire monitoring, public security investigation, and environmental detection [1–4]. To achieve UV detection, wide-band-gap semiconductors have to be used, and thus research efforts have been currently devoted to SiC, ZnO, Ga2O3, and AlGaN compounds [5–8]. Among the candidates, AlGaN-based III-nitride material can be an excellent choice for fabricating UV photodetectors (PDs) owing to the adjustable energy band gap (3.4–6.2 eV), high electron mobility (135–1000cm2⋅V−1⋅s−1), good thermal conductivity (1.3–3.19W⋅cm−1⋅K−1), and excellent chemical stability [9,10]. Based on AlGaN materials, different PD structures have been developed, e.g., metal-semiconductor-metal (MSM) PDs, Schottky barrier PDs, PIN PDs, and avalanche PDs [11–14]. Among them, AlGaN-based MSM PDs have attracted significant research attention due to their less complicated fabrication process, low noise, and high detectivity [15]. Conventional MSM PDs use two back-to-back metal-semiconductor contacts to form a symmetric Schottky-type metal/semiconductor structure [16]. Such design can possess a large absorption region for effectively absorbing photons. However, it is well known that the poor carrier transport efficiency is one of the major limiting factors that hinder the improvement of the photocurrent and responsivity [17]. On one hand, the electric field is generated at the Schottky contact, which is determined by the difference of work functions between the metal and the semiconductor. On the other hand, the limited difference of the work function between the metal and the semiconductor makes the space charge region at the Schottky contact region thin. Therefore, the absence of the electric field in the rest of the optical absorption region strongly hinders the carrier transport, especially when the illumination signal becomes strong. This is because the photogenerated carriers can screen the electric field, which will further reduce the carrier transport between the two Schottky contacts [18,19] and then lead to a saturated photocurrent. The other figure of merit for PDs is that a low dark current is required. For Schottky-contact-based MSM PDs, the leakage current that arises from the thermionic emission at the metal/contact interface cannot be neglected, especially when one metal/semiconductor is reversely biased. Then the image force at that metal/contact region will cause a reduced barrier height [20], which will inevitably result in the increased leakage current, i.e., the dark current for PDs. Therefore, tremendous effort shall be made to increase the difference between the dark current and the photocurrent so that the enhanced responsivity can be obtained.

An effective method to increase the responsivity is making Schottky contacts with different metals for MSM PDs. Then the different work functions for the two metals will produce the asymmetric energy band, which is very useful to promote the carrier transport. Moreover, such asymmetric energy band profiles can even generate photocurrent when the devices are not biased, i.e., a self-powered MSM PD is achieved [21–25]. One typical AlGaN-based PD with an asymmetric structure is achieved by using Ni and Au as electrodes at the two ends; the different barrier heights for the Ni/GaN and Au/GaN junctions enable more efficient charge transport and the increased responsivity [22]. However, the responsivity can be further promoted if the energy band can be made more asymmetric. For that purpose, a Pt-Ag GaN MSM PD is reported. The very large potential gradient of 1.39 eV between the Pt and Ag metals then favors the enhanced thermionic emission rate of the electrons for the Pt-Ag MSM PD [21]. A big energy barrier variation can also be obtained by using a Ni-Ti/Al MSM PD structure, and the responsivity of 0.104 A/W can be achieved at the applied bias of 0 V [23]. Other self-powered GaN MSM PDs can also be fabricated by using Ni-Cr Schottky contact pairs and Ni-Ag Schottky contact pairs according to the report in reference [23]. Another MSM PD with the asymmetric structure is achieved by varying the contact area of the Au electrodes [25].

Therefore, our previous literature review shows that, to make self-powered MSM PDs, different Schottky metal contacts shall be used so that the different work functions can produce the asymmetric energy band profiles. Nevertheless, the tilted energy band and the carrier transport are strongly limited by the work function variation for the two Schottky metals. Additionally, the depletion region width that is produced by the metal/semiconductor structure is small, and this can be easily compensated for by the photogenerated carriers. Then the carrier transport can be strongly affected, especially when the illumination signal is strong enough and the saturation photocurrent might be yielded [26]. In this work, we take the advantage of the polarization effect at the AlGaN/GaN junction to produce the electric field, i.e., the AlGaN/GaN junction here imitates the metal-based “Schottky junction.” On one hand, by properly optimizing the epitaxial growth condition, a fully coherent growth mode can be achieved for AlGaN/GaN heterojunction that is free from strain relaxation. Then the AlGaN/GaN junction can be fully polarized and the polarization-induced electric field therein will be extremely strong [27]. On the other hand, the polarization-induced electric field cannot be easily screened by free carriers, which promises a high photocurrent even if the detected optical signal intensity is increased.

Figure 1(d) illustrates the schematic energy band diagrams of the traditional metal/GaN/metal PD structure for Device R. As is shown, the built-in electric fields generated by the two Schottky contacts are of the opposite directions, and the electrons can meet both energy barriers with the same energy barrier height; this energy band profile can strongly hinder the carrier diffusion, and we can also predict no current flow if Device R is not biased. However, when an Al0.10Ga0.90N layer is used under one metal contact, then the very strong polarization-induced electric field can be obtained at the Al0.10Ga0.90N/GaN interface, which can realign the energy band downwards because of the polarization-induced positive charges at the Al0.10Ga0.90N/GaN interface [see Fig. 1(e)]. The energy band profile in Fig. 1(e) can also favor the carrier transport, even if Device N is not biased. Our results show that the energy band can be bent by more than 1 eV when a negative bias is applied onto the Schottky metal contact that is fabricated on the GaN layer, which remarkably increases the carrier transport and enhances the photocurrent.

In the first step, the epilayers of Devices R and N are grown on AlN/sapphire templates by MOCVD. The trimethyl gallium (TMGa), trimethyl aluminum (TMAl), and ammonia (NH3) are used as the Ga, Al, and N precursors, and hydrogen (H2) is used as the carrier gas. Through tuning the Al/Ga precursor ratio, the designed Al content can be 0.10 for the AlGaN layer. Next, a device mesa is fabricated by using photolithography and dry etching. The electrodes are prepared by following standard processes of photolithography, e-beam deposition, and lift-off. Finally, the interdigital electrode of 10/10/200 nm Ni/Au/Au is fabricated.

The APSYS calculations are carried out with a commercial software: Crosslight Software Inc. APSYS 2018 (http://www.crosslight.com), which can solve the continuity equation and Poisson equation by using the finite element method. The drift and diffusion processes for carriers can also be considered. The Shockley–Read–Hall (SRH) recombination lifetime is set to 10 ns [28]. The electron affinity for the Ni is set to 5.15 eV [29]. The polarization effect is considered at the Al0.10Ga0.90N/GaN heterojunction, and the polarization charge density is set to ∼3.97×1016m−2 [30]. The energy band offset for the Al0.10Ga0.90N/GaN junction is set to 50/50 [31]. The absorption coefficient for GaN in terms of the wavelength can be found in Table 1 [32]. Other parameters regarding nitride semiconductors can be found elsewhere [33].Table 1.

Absorption Coefficient of GaN Under Different Wavelengths

The dark current and the photocurrent as functions of the applied bias for the as-fabricated GaN-based MSM PDs are presented in Fig. 2(a). The bias is applied at the left electrode, such that the positive bias is biased at the left electrode in the first quadrant and the negative bias is biased at the left electrode in the second quadrant. The other electrode is grounded when the devices are measured. We have observed symmetric current-voltage characteristics for Device R, which agree with the energy band profile in Fig. 1(d). Nevertheless, the asymmetric current in terms of the applied bias is obtained for Device N, which is a signature for the unbalanced energy band profile as indicated in Fig. 1(e). The dark current for Device N is larger than that for Device R in the first quadrant, which means that the polarized Al0.10Ga0.90N/GaN interface can increase the electric field in the GaN absorption layer. This can promote the drift and diffusion for carriers. However, the dark current for the two devices shows negligible difference in the second quadrant, which indicates the similar Schottky barrier height for the Ni/Au/GaN structure. This also means that the metal deposition process for our devices is stable. When the devices are illuminated under the 266 nm laser diode, both devices generate photocurrent with the higher photocurrent level for Device N, especially in the first quadrant. The much more enhanced photocurrent for Device N in the first quadrant is mainly attributed to the polarized Al0.10Ga0.90N/GaN interface. The photocurrent for Device N is also slightly larger than that for Device R in the second quadrant, which is likely due to the hole trapping effect [22], such that we get the Al0.10Ga0.90N mesa by using dry etching, and this can produce traps on the GaN surface for trapping holes. Then the electrons will be less recombined with holes and can contribute to the drift and diffusion current. Figures 2(b) and 2(c) show the photocurrent for Devices R and N under different laser powers, and we can see that the photocurrent for Device N is always higher than that for Device R, which infers that Device N can more accurately detect the very strong optical signal without photocurrent saturation. Note that we are not able to further increase the optical signal power because of the limited laser power. To even better demonstrate the origin of the enhanced photocurrent for Device N, we calculate and present the two-dimensional electric field profiles between the two electrodes in Fig. 2(d). The positive direction for the electric field is defined to point to the right side [see the arrows in Fig. 2(d)]. When we look at the right sides of the devices, we find that, thanks to the polarized Al0.10Ga0.90N/GaN interface, the generated electric field for Device N is larger than that for Device R. When we investigate the left sides of the devices, the electric field is negative, i.e., it points to the left side for both devices. However, the electric field for the horizontal component therein is smaller for Device N (−80V/cm) than that for Device R (−2.5×103V/cm). The smaller electric field intensity for Device N at the left side arises from the screening effect by the polarization-induced positive charges at the Al0.10Ga0.90N/GaN interface. The electric field profile for Device N can even better favor the carrier transport, which is further evidenced in Fig. 2(e). We can intuitively see that the current for Device N is higher than that for Device R, especially in the region from 33 to 40 μm. It indicates that we require a strong electric field in the optical absorption region so that the photogenerated carriers can be more effectively moved between the two metal electrodes.

In order to reveal the origin of the responsivity for Device N, we calculate and show the energy band diagrams for Devices R and N under 266 nm UV light illumination in Figs. 3(b) and 3(c), respectively. Note that both devices are not biased. In Fig. 3(b), we define two metal-semiconductor junctions as J1 and J2 for Device R. Similarly, J1′ and J2′ are also defined for the Al0.10Ga0.90N/GaN junction and metal-semiconductor junction for Device N in Fig. 3(c), respectively. When the two devices are illuminated by the 266 nm UV light, electron-hole pairs are generated in the GaN absorption layer. For Device R, the same energy barrier can be observed at J1 and J2. Therefore, the photogenerated electrons will be confined in the region with lower conduction band energy. Although the holes are located at the valence band with high energy, the low hole mobility will still localize most holes in the GaN absorption region. If we look into the calculated energy band for Device N in Fig. 3(c), we can see that the energy band is strongly bent. The photogenerated electrons experience potential variation between J1′ and J2′, and the potential difference provides the kinetic energy for both electrons and holes. Then the electrons and holes can overcome the barrier height at the Ni/Al0.10Ga0.90N/GaN junction and the Ni/GaN junction, respectively. As a result, the photocurrent is produced and we detect the responsivity when the applied bias is 0 V for Device N. The calculated two-dimensional electric field profiles for Devices R and N under the 266 nm UV illumination are also given in Fig. 3(d). The electric field profiles are consistent with the energy band alignments for Devices R and N, respectively. The direction of the electric field at the position of J1 is opposite to that at the position of J2, while the electric field intensity is the same. It means that the net electric field is neutralized. Therefore, neither electrons nor holes can freely move between the two metal electrodes. On the other hand, although the directions for the electric field at positions of J1′ and J2′ oppose to each other, the electric field intensity at the position of J2′ is larger than that at the position of J1′. Therefore, the electrons and holes can be pushed to obtain movement between the two metal contacts. These presented results indicate a promising and simple method to fabricate excellent self-powered GaN-based PD by using the inherent polarization effect for III-nitride-based compounds.

A comparative chart is presented in Table 2, which summarizes the responsivity and peak wavelength for the reported GaN-based MSM PDs. It shows that the highest responsivity of 106A/W is at the peak wavelength of 250 nm according to Ref. [35]. However, the self-powered mode cannot be achieved. Our peak responsivity of 13.56 A/W with an on/off ratio of 106 at the wavelength of 346 nm and the self-powered detection capability can be simultaneously achieved. Therefore, our design can realize both the high responsivity and the self-powered detection capability, proving the advantage of the polarization effect for improving the GaN-based PD performance. Furthermore, our design can detect the solar blind wavelength range of 240 nm in the self-powered mode.Table 2.

Performances for the Reported GaN-Based and AlGaN-Based PDs in the Literature

In conclusion, we demonstrate the fabrication and the characterization of a self-powered GaN-based MSM UV PD. The self-powered GaN UV PD is realized by using the polarized Al0.10Ga0.90N/GaN interface, and the polarization-induced electric field can produce the asymmetric energy band. Then, photogenerated carriers can be transported even if the device is not biased, and peak responsivity as high as 0.005 A/W is obtained at the solar blind wavelength of 240 nm. Moreover, the potential difference between the two metal electrodes is further proven by showing the asymmetric dark current and photocurrent in terms of the applied bias for Device N. Our measurement also demonstrates a high responsivity of 13.56 A/W at the applied bias of 3 V bias for the proposed structure. The photoconductive mode is also observed by showing the increased rise time when the time-dependent photo response is measured for Device N, which can be another possible reason for the very high photocurrent and the responsivity. However, the stronger electric field can make the carriers swept out of the GaN absorption layer rapidly, and hence a faster decay time is obtained for Device N. Thus, plenty of research effort shall also be made to achieve both the high responsivity and the fast photo-response capability. Therefore, this study provides a feasible and cost-effective route for designing GaN-based self-powered MSM UV detectors with very high responsivity. We strongly believe that the reported structures are very useful for the community to make high-performance GaN-based UV detectors, and the device physics revealed here enriches the understanding of photosensitive devices.