Ultraviolet light in the solar-blind band (200–280 nm) hardly radiates to the Earth’s surface due to the strong atmospheric absorption. Therefore, solar-blind ultraviolet (SBUV) detectors have attracted tremendous interest in the scope of fire monitoring, high-voltage equipment corona discharge monitoring, or biological imaging [1–5]. AlGaN material, with a tunable wide bandgap and high absorption coefficient from 3.4 to 6.2 eV, is one of the most suitable materials for SBUV detectors [6]. Moreover, AlGaN SBUV detectors have the advantage of small-size, low-power-consumption, and easy-integration over the most frequently used photomultiplier tubes that feature high vacuum, high voltage, and huge volume [7–9]. Although considerable progress has been made on AlGaN SBUV detectors in recent years, the performance still cannot meet the requirements of real applications because of the poor crystal quality and $p$-doping efficiency deterioration with increasing Al content [10,11]. In order to obtain optimal device performance on the platform with the current AlGaN material quality and $p$-doping efficiency, the structures have been extensively studied and finely optimized.

AlGaN-based detectors generally exhibit five structures, including photoconductor [12], photovoltaic metal-semiconductor-metal (MSM) [13], Schottky diode [14], $p\text{-}n$ ($p\text{-}i\text{-}n$, $p\text{-}i\text{-}n\text{-}i\text{-}n$ or $p\text{-}i\text{-}p\text{-}i\text{-}n$) diode [15–17], bipolar transistor, and tripolar transistor [18,19]. The photoconductor can have high gain due to the carrier trapping effect of the surface or internal defects, but the persistent photoconductivity phenomenon that can sustain several tens of seconds or even longer strongly limits its application [20,21]. The photovoltaic MSM structure can avoid the $p$-doping of AlGaN, while it usually has low responsivity and slow response speed due to the symmetrical back-to-back Schottky barriers and long carrier drift length, respectively [22,23]. Besides, the photoconductor and photovoltaic MSM structure generally cannot respond under zero bias, which means that an external power supply is required, leading to extra noise and energy consumption. Although the application of asymmetric electrodes provides zero-bias response through the asymmetric Schottky barriers, the external quantum efficiency (EQE) is too low [24,25]. Furthermore, the horizontal distributed electrodes make it difficult to fabricate high-pixel focal plane arrays, also limiting their application. Similarly, the tripolar transistor also possesses these disadvantages despite of its high current gain [26,27]. For a long time, researchers have pinned their hopes on the vertical high-gain avalanche photodiodes; however, the premature breakdown, high work voltage, and complex driving circuit crucially restrict their application [8,28,29].

Therefore, there is an urgent need to exploit vertical high-gain structures for AlGaN SBUV detectors, especially at zero bias. However, traditional vertical structures such as the Schottky diode, $p\text{-}n$ diode, and bipolar transistor generally have no current gain at zero bias. The reported highest EQE and the corresponding responsivity of AlGaN SBUV Schottky diode at zero bias, to our best knowledge, are 53% and 0.115 A/W at 270 nm [30], respectively. There is no current gain in the device, obviously. As for the AlGaN SBUV $p\text{-}n$ diode, by adopting techniques, including the intrinsic absorption layer, high-quality AlN template, and polarization-graded $p\text{-}\mathrm{AlGaN}$ layer, EQE and responsivity as high as 92% and 0.211 A/W at 289 nm are reached at zero bias [15]. However, the current gain is also not obtained. In terms of the AlGaN SBUV bipolar transistor, most of the reports focus on the visible-blind ultraviolet band [18,31–33], and the only SBUV detector reported shows a low EQE and responsivity of just 1.2% and 0.003 A/W at 280 nm [34], which may result from the low $p$-doping efficiency of Al-rich AlGaN. Consequently, alternative gain mechanism needs to be introduced to vertical structures to realize high EQE and responsivity in AlGaN SBUV detectors.

In this work, we have proposed a strategy of introducing a photoconductive gain mechanism into the vertical AlGaN SBUV detector and have realized it in a $p\text{-}i\text{-}n$ photodiode through inserting the multiple-quantum-well (MQW) into the depletion region. The MQW can generate current gain due to its strong hole confinement effect and the quantum confinement Stark effect (QCSE), inducing long carrier lifetime and unipolar carrier transport, which are the key factors for photoconductive gain. Experimentally, an AlGaN SBUV detector with zero-bias responsivity of 0.425 A/W at 233 nm is achieved, corresponding to an EQE of about 226%, which indicates a significant internal current gain. Compared with the structure without MQW, the gain is estimated to be about ${10}^3$ in magnitude. The MQW can also reduce the dark current because of the large band offset in the MQW, indicating a better signal-to-noise ratio (SNR). This investigation paves the way for fabricating high-gain AlGaN SBUV detectors with vertical structures at low or even zero bias.

As noted, there are two gain mechanisms in the AlGaN SBUV detector, i.e., photoconductive and avalanche gains. As for the avalanche gain, it is necessary to reach the critical electric field, which is high for Al-rich AlGaN materials. This leads to a high work voltage and complex quenching circuit. However, as for the photoconductive gain, the high voltage and complex circuit are not essential. In order to effectively introduce photoconductive gain in vertical structures, it is required to clearly understand the key factors for determining the gain.

The photoconductive gain can be calculated through Eq. (1): $$G=\tau ·(1/t_{rn}+1/t_{rp}),$$(1)where $G$ is the photoconductive gain, $\tau$ is the carrier lifetime, and $t_{rn}$ and $t_{rp}$ are the electron and hole transit times, respectively. Here, the electron and hole transit times $t_{rn}$ and $t_{rp}$ are determined by Eq. (2): $$t_{rn}=L/{\mu }_{n}E,t_{rp}=L/{\mu }_{p}E,$$(2)where $L$ is the drift length, $E$ is the internal electric field, and ${\mu }_n$ and ${\mu }_p$ represent the electron and hole mobility, respectively. In a vertical photovoltaic detector, the drift length is generally short and the built-in electric field is strong in the carrier drift region. Accordingly, the photoconductive gain can be improved by prolonging the carrier lifetime in a vertical photovoltaic detector. Besides, there is another gain mechanism in the photoconduction mode. If one type of carrier drifts much faster than the other and the transit time of the faster carriers is shorter than the lifetime, the carriers will travel through the electrodes several times to keep electrical neutrality, leading to unipolar carrier transport multiplication. Therefore, trapping one type of carrier in the transport region can also increase the photoconductive gain.

Based on the previous analyses, structures that can prolong the carrier lifetime and trap one type of carriers shall be designed into the vertical AlGaN SBUV detectors. Wurtzite AlGaN materials possess strong spontaneous and piezoelectronic polarizations, which can induce a strong built-in electric field in their MQW structures. Figure 1(a) shows an AlGaN single quantum well along the [0001] direction. As can be seen, the built-in electric field in the well layer is along the [000–1] direction, while that in the barrier layer is opposite. The carriers with opposite charge polarities in the well layer can be separated by the built-in electric field and then concentrate at different regions, which is known as the QCSE. The QCSE causes the wave function overlap of electrons and holes to decrease, resulting in the suppressed recombination probability. Correspondingly, the carrier lifetime can be obviously prolonged by utilizing MQW, resulting in photoconductive gain. On the other hand, because the effective mass of holes is much higher than that of electrons, and because the intervalley scattering of electrons is much less than that of holes [8], the holes are easier to be confined by the quantum wells than electrons. As a result, the unipolar carrier transport will be achieved and the photoconductive gain will be increased. Consequently, it is believed that the photoconductive gain can be generated in a vertical AlGaN SBUV $p\text{-}i\text{-}n$ photodiode by inserting MQW into their intrinsic depletion region.

Figure 1(b) illustrates our basic device structure design. An MQW structure is introduced into the depletion region of a vertical back-illuminated $p\text{-}i\text{-}n$ photodiode. To induce photoconductive gain, the MQW has to be inserted between the $i$-layer and $p$-layer. Upon illumination, the photogenerated electron-hole pairs will be mainly produced in the $i$-layer and then swept to the opposite directions by the built-in electric field. The electrons with high mobility are quickly swept out to the $n$-layer and extracted by the anode. Meanwhile, the holes will pass through the MQW region. The MQW will reduce the hole mobility due to the scattering effect and trap the hole due to its quantum confinement effect. Furthermore, the recombination rate is diminished by the MQW because of the QCSE, resulting in longer carrier lifetime. In this case, abundant holes will remain in the depletion region. To maintain electrical neutrality, electrons will again drift into the depletion region. Because the mobility of electrons is inherently much higher than that of holes and the intervalley scattering of electrons is less frequent than that of holes, the electrons can traverse the detector multiple times during the lifetime to generate current gain. Further, the MQW can also reduce the dark current of the device and thus enhance the SNR.

Numerical simulations are conducted by APSYS to verify the effects of the MQW on the carrier transport properties in the AlGaN SBUV $p\text{-}i\text{-}n$ photodiode. Figure 2(a) exhibits the AlGaN SBUV $p\text{-}i\text{-}n$ photodiode structure with MQW in the depletion region, labeled as device ${\mathrm{S}}_{\mathrm{A}}$. For comparison, the device with the same structure except for the MQW, which is replaced by $i\text{-}\mathrm{AlGaN}$, is constructed, as shown in Fig. 2(b), labeled as device ${\mathrm{S}}_{\mathrm{B}}$. In the simulations, the barrier of the MQW is AlN of 8 nm and the well is ${\mathrm{Al}}_{0.20}{\mathrm{Ga}}_{0.80}\mathrm{N}$ of 2 nm. The barrier and well alternately repeat for 10 times, and the MQW ends with an AlN barrier, resulting in a total thickness of 108 nm and an average optical-Al-content of 50%. For the reference device ${\mathrm{S}}_{\mathrm{B}}$, to ensure the same carrier drift length and optical absorption, the thickness and Al content of the $i\text{-}\mathrm{AlGaN}$ layer are set as 108 nm and 50%, respectively. The carrier transport properties are simulated by self-consistently solving the Schrodinger equation, Poisson equation, continuity equation, and drift-diffusion equation with proper boundary conditions. The hole and electron mobilities are set as 5 and $1000{\mathrm{cm}}^2/(\mathrm{V}·\mathrm{s})$ [35,36], respectively. The conduction and valance band offset ratio for AlGaN heterojunctions and MQW is set as 50/50, leading to both conduction and valance band offsets of 1.1 eV in the MQWs [37]. The polarization level is set to 30%. The optical absorption coefficients for AlGaN layers are extracted from Refs. [38,39].

The devices are grown on single-polished $c$-sapphire substrates by a high temperature metal organic chemical vapor deposition system. The trimethyl aluminum (TMAl), trimethyl gallium (TMGa), and ammonia (${\mathrm{NH}}_3$) are used as Al, Ga, and N precursors, respectively. The hydrogen (${\mathrm{H}}_2$) is used as the carrier gas. Besides, the silicane (${\mathrm{SiH}}_4$) and bis-cyclopentadienyl magnesium (${\mathrm{Cp}}_2\mathrm{Mg}$) are used as the $n$- and $p$-type doping sources, respectively. First, an AlN template of about 2 μm is grown on the sapphire substrate at 1300°C to improve the quality of upper-device function layers. Then, the temperature is decreased to 1180°C and the transition layers as well as a 500 nm undoped ${\mathrm{Al}}_{0.60}{\mathrm{Ga}}_{0.40}\mathrm{N}$ buffer layer are grown on the AlN template to release the stress. Next, the designed device structures as shown in Fig. 2 are grown. For device ${\mathrm{S}}_{\mathrm{A}}$, a 300 nm $n\text{-}{\mathrm{Al}}_{0.60}{\mathrm{Ga}}_{0.40}\mathrm{N}$ electron transport layer, a 300 nm undoped ${\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ absorption layer, 10 periods of $\mathrm{AlN}(8\mathrm{nm})/{\mathrm{Al}}_{0.20}{\mathrm{Ga}}_{0.80}\mathrm{N}(2\mathrm{nm})$ MQW, and a 120 nm $p\text{-}{\mathrm{Al}}_{0.50}{\mathrm{Ga}}_{0.50}\mathrm{N}$ hole transport layer are sequentially grown. Finally, the temperature is decreased to 1080°C, and a 50 nm heavily doped $p\text{-}\mathrm{GaN}$ contact layer is grown. For the reference device ${\mathrm{S}}_{\mathrm{B}}$, all growth and structure parameters are the same as that of ${\mathrm{S}}_{\mathrm{A}}$ expect for the MQW. An $i\text{-}{\mathrm{Al}}_{0.50}{\mathrm{Ga}}_{0.50}\mathrm{N}$ layer is grown to replace the MQW. Through adjusting the Al/Ga source ratio, the designed Al content can be realized. When the growth is finished, an ex situ thermal annealing at 900°C in ${\mathrm{N}}_2$ is made to activate the Mg acceptors.

Standard semiconductor device fabrication processes are used to prepare the AlGaN SBUV detectors. First, a ${\mathrm{SiO}}_2$ mask layer is deposited on the wafers by plasma-enhanced chemical vapor deposition, and the device pattern is transported to the mask layer by standard photolithography and reactive ion etching (RIE) processes. Then, the wafers are etched by an inductive coupling plasma etching system of about 600 nm to expose the $n\text{-}{\mathrm{Al}}_{0.60}{\mathrm{Ga}}_{0.40}\mathrm{N}$ layer. After dissolving the residual ${\mathrm{SiO}}_2$ mask by buffered oxide etching solution, the wafers are immersed in 10% NaOH solution at 80°C for 10 min to passivate the etching defects on the mesa sidewall. Next, the $n$-electrode pattern is transported to the wafer by photolithography and Ti 30 nm/Al 70 nm/Ni 50 nm/Au 150 nm electrode is deposited on the photolithograph-completed wafers by e-beam and thermal evaporation. After lift-off, the $n$-electrode is rapidly annealed at 600°C for 30 s in ${\mathrm{N}}_2$ to form Ohmic contact. By the same processes, the $p$-electrode of Ni 50 nm/Au 150 nm is deposited on the mesa. After lift-off, the $p$-electrode is annealed at 550°C for 5 min in ${\mathrm{N}}_2$ to form Ohmic contact. Finally, a ${\mathrm{SiO}}_2$ passivation layer is deposited on the wafers to passivate the dangling bonds and thus to reduce leakage current. The ${\mathrm{SiO}}_2$ layer on electrodes is then removed by RIE for performance test.

A high-resolution X-ray diffractometer (Brucker D8 DISCOVER) is adopted to estimate the Al content and crystal quality of the epilayers. A high-resolution scanning transmission electron microscope (HR-STEM) is employed to investigate the cross section of the devices, especially the MQW region. An optical microscope (OM, Nikon ECLIPSE LV15ONA) is used to observe the mesa and electrodes of the detectors. The current-voltage (I-V) curves are measured by the PDA FS-Pro 380 semiconductor analyzer. The spectral responses are measured by a DSR 100 system that is equipped with a xenon lamp, chopper, monochromator, Keithley 6487, SR830 lock-in amplifier, and standard Si detector. All response spectra are calibrated by the standard Si detector by considering the effective photosensitive area [40].

The I-V curves of the two devices in dark and on illumination are calculated as shown in Fig. 3(a). The device ${\mathrm{S}}_{\mathrm{A}}$ exhibits a significantly lower dark current than device ${\mathrm{S}}_{\mathrm{B}}$, demonstrating the MQW can effectively suppress the detector dark current. The decrease of dark current mainly results from the scattering effect of the band offset between the AlN barriers and AlGaN wells and the interface polarization charges. Besides, due to the strong QCSE, the generation-recombination current will also be suppressed. For the device ${\mathrm{S}}_{\mathrm{A}}$, within 40 V bias, the dark current slightly increases with the bias; for the device ${\mathrm{S}}_{\mathrm{B}}$, the dark current abruptly rises up at 10 V bias. It can be deduced that the MQW can promote the detector SNR in a large bias range as expected. On illumination, the photocurrent of device ${\mathrm{S}}_{\mathrm{A}}$ is obviously greater than that of ${\mathrm{S}}_{\mathrm{B}}$, especially at low bias, which agrees well with our prediction. Accordingly, it demonstrates that device ${\mathrm{S}}_{\mathrm{A}}$ has a higher response than ${\mathrm{S}}_{\mathrm{B}}$. The difference can be more intuitively reflected by the spectral responses, which are calculated by Eq. (3): $$R(\lambda )=\frac{I_p(\lambda )-I_d(\lambda )}{P_{\mathrm{in}}(\lambda )},$$(3)where $R$ is the responsivity, $I_p$ and $I_d$ are the photo and dark current, respectively and $P_{\mathrm{in}}$ is the incident light power. Figure 3(b) shows the response spectra of devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$ at 5 V. Both devices possess the peak response at about 230 nm. However, device ${\mathrm{S}}_{\mathrm{A}}$ has nearly four times higher peak response than ${\mathrm{S}}_{\mathrm{B}}$, directly confirming the positive effects of the MQW structure on the detector performance.

The energy band, current density, and carrier distributions are extracted to investigate the mechanism of the response gain. Figures 4(a) and 4(b) show the energy bands in the depletion regions for devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$ at 5 V bias. The upper panels are in dark; the lower ones are on illumination. As is seen, the energy band bends upward along the [0001] direction for both devices, indicating the built-in electric field is also along the [0001] direction. However, there also exist differences between devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$. First, the energy band of the MQW in device ${\mathrm{S}}_{\mathrm{A}}$ is serrated and the well layer energy band bends downward along the [0001] direction as predicted; the energy band for device ${\mathrm{S}}_{\mathrm{B}}$ of the counterpart region for ${\mathrm{S}}_{\mathrm{A}}$ is flat. Second, the energy band for the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ region tilts more than that of the MQW in device ${\mathrm{S}}_{\mathrm{A}}$; it is just the opposite in device ${\mathrm{S}}_{\mathrm{B}}$. This means that the built-in electric field in the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ absorption region is weaker than that in the MQW region for device ${\mathrm{S}}_{\mathrm{A}}$, and the case in ${\mathrm{S}}_{\mathrm{B}}$ is exactly reverse. This phenomenon also originates from the polarization effect in AlGaN hetero-structure. In device ${\mathrm{S}}_{\mathrm{A}}$, the AlN layer on the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer will induce positive polarization charges at the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}/\mathrm{AlN}$ interface, creating an electric field along the [000–1] direction in the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer and along the [0001] direction in the AlN barrier. In device ${\mathrm{S}}_{\mathrm{B}}$, the $i\text{-}{\mathrm{Al}}_{0.50}{\mathrm{Ga}}_{0.50}\mathrm{N}$ layer on the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer will induce negative polarization charges at the ${\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}/{\mathrm{Al}}_{0.50}{\mathrm{Ga}}_{0.50}\mathrm{N}$ interface, creating an electric field along the [0001] direction in the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer and along the [000–1] direction in the $i\text{-}{\mathrm{Al}}_{0.50}{\mathrm{Ga}}_{0.50}\mathrm{N}$ layer. The differences will lead to two effects in device ${\mathrm{S}}_{\mathrm{A}}\text{:\hspace{0.17em}}1$) the photogenerated electrons in the $i\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer are effectively prevented from diffusing to the MQWs and $p\text{-}\mathrm{AlGaN}$ layers; 2) the photogenerated holes are quickly swept into the MQW region and trapped by the MQW, resulting in net positive charges in the MQW region. In this case, electrons have to be drawn into the MQW from the cathode electrode to keep electrical neutrality. Because of the fast mobility, minimal intervalley electron scattering, strong built-in field, and prolonged lifetime, electrons will move fast and can travel multiple times through the devices, resulting in photoconductive gain.

Figures 4(c) and 4(d) show the electron and hole current profiles for devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$ in dark and on illumination at 5 V bias, respectively. As the electrons are swept to the $n\text{-}{\mathrm{Al}}_{0.60}{\mathrm{Ga}}_{0.40}\mathrm{N}$ region by the built-in electrical field, the current at the left boundary of the depletion region is mainly contributed by electrons. Similarly, the current at the right boundary of the depletion region is mainly contributed by holes. When observing the dark electron and hole currents at the left and right boundaries of the depletion region in Figs. 4(c) and 4(d), respectively, it is found that both dark electron and hole currents of device ${\mathrm{S}}_{\mathrm{A}}$ are much lower than that of ${\mathrm{S}}_{\mathrm{B}}$, which may result from the trapping and scattering effects of the MQW as expected. However, on illumination, the hole current of device ${\mathrm{S}}_{\mathrm{A}}$ is smaller than that of ${\mathrm{S}}_{\mathrm{B}}$, while the electron current is just the opposite. In Fig. 3(a), it is shown that the photocurrent of device ${\mathrm{S}}_{\mathrm{A}}$ is higher than that of ${\mathrm{S}}_{\mathrm{B}}$. Therefore, it can be deduced that the photocurrent of device ${\mathrm{S}}_{\mathrm{A}}$ is mainly contributed by electrons, while that of ${\mathrm{S}}_{\mathrm{B}}$ is equally contributed by electrons and holes. This is evidence that the photoconductive gain has occurred in device ${\mathrm{S}}_{\mathrm{A}}$ as predicted. Further, Figs. 4(e) and 4(f) display the electron and hole concentration distributions in devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$ in dark and on illumination. As shown in Fig. 4(f), the hole concentration in the MQW region of device ${\mathrm{S}}_{\mathrm{A}}$ increases significantly on illumination, while that of ${\mathrm{S}}_{\mathrm{B}}$ hardly changes, indicating the hole trapping effect of the MQW. Besides, although the photogenerated electrons are swept to the $n\text{-}{\mathrm{Al}}_{0.60}{\mathrm{Ga}}_{0.40}\mathrm{N}$ side, the electron concentration in the MQW region is still improved, as shown in Fig. 4(e), implying the hole attracting effect and the prolonged carrier lifetime. Accordingly, it can be concluded that the MQW structure can introduce photoconductive gain into the vertical photodiode by prolonging the carrier lifetime via QCSE and making the electrons travel multiple times via electric field and electric neutrality effects.

Devices are experimentally fabricated to verify our conceptions. Figure 5(a) illustrates the schematic structure of device ${\mathrm{S}}_{\mathrm{A}}$. The reference device ${\mathrm{S}}_{\mathrm{B}}$ has the same configuration expect for the MQW. Figure 5(b) displays the cross-sectional STEM image of the layer structure for device ${\mathrm{S}}_{\mathrm{A}}$. As is seen, the layers are grown as designed. To confirm the structure of the MQW, the HR-STEM image is measured as shown in Fig. 5(c). Clearly stacked structures are displayed. The narrow bright contrasts of about 2 nm are the wells, and the wide dark contrasts of about 8 nm are the barriers, respectively, demonstrating that the MQW region is grown as expected. Figure 5(d) shows the plane-view OM image for the fabricated devices. The diameter is 200 μm, corresponding to an effective photosensitive area of about $\mathrm{31,416}{\mathrm{\mu m}}^2$, which is the used value in calibrating the responsivity. The devices are back-illuminated when working.

The I-V curves and spectral responses of devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$ are measured to verify the detection performance. Figure 6(a) shows the I-V curves in dark and on illumination. The dark currents of both devices are small, which are at the magnitude of pA below 10 V bias. It may mainly come from the improved crystal quality by the AlN template, which can reduce the threading dislocation density and thus suppress the defect-assistant generation-recombination current. As the bias increases from 10 to 40 V, the dark current of device ${\mathrm{S}}_{\mathrm{A}}$ just slightly becomes larger, while that of ${\mathrm{S}}_{\mathrm{B}}$ obviously increases. It agrees well with the simulated results in Fig. 3(a). The carrier trapping and scattering effects by the MQW in device ${\mathrm{S}}_{\mathrm{A}}$ shall be responsible for the suppression of dark current at relatively high biases. More importantly, when they are illuminated, the photocurrent of device ${\mathrm{S}}_{\mathrm{A}}$ is much larger than that of ${\mathrm{S}}_{\mathrm{B}}$ within all the measured biases, which also agrees well with the simulations, demonstrating the existence of photoconductive gain in device ${\mathrm{S}}_{\mathrm{A}}$. Certainly, the gain originates from the MQW because it can increase the carrier lifetime and induce unipolar carrier transport multiplication. It is worth noting that the photocurrent of device ${\mathrm{S}}_{\mathrm{A}}$ at zero bias is also significantly promoted when it is compared with that of ${\mathrm{S}}_{\mathrm{B}}$, indicating that there exists photoconductive gain at even zero bias, which is important for improving the detector SNR. Comparing the photocurrent of devices ${\mathrm{S}}_{\mathrm{A}}$ and ${\mathrm{S}}_{\mathrm{B}}$, it can be estimated that the gain can reach ${10}^3$ in magnitude at zero bias.

Figure 6(b) shows the spectral response of device ${\mathrm{S}}_{\mathrm{A}}$ at zero bias. The peak responsivity is 0.425 A/W at 233 nm, corresponding to an EQE of 226% calculated by Eq. (4): $$\eta (\lambda )=\frac{1240·R(\lambda )}{\lambda },$$(4)where $\eta$ is the EQE, $R$ is in the unit of A/W, and $\lambda$ is the response wavelength in units of nm. The EQE much higher than 100% confirms the fact that there exists current gain in device ${\mathrm{S}}_{\mathrm{A}}$. Besides, the long-wavelength cutoff edge is shorter than 250 nm, ensuring the SBUV detection characteristic. Moreover, full-width at half-maximum of the response spectrum is just 13 nm, exhibiting excellent wavelength-selective characteristic. The inset in Fig. 6(b) shows the log-scale spectral response. As is seen, the responsivity ratio of peak wavelength to visible-blind ultraviolet band (365–400 nm) is about ${10}^3$, indicating a good SNR of device ${\mathrm{S}}_{\mathrm{A}}$. It is worth noting that the sapphire substrate is single-polished. Therefore, we can speculate the responsivity can be much higher once the substrate is double-polished. Besides, we have ever attempted to measure the response spectrum of device ${\mathrm{S}}_{\mathrm{B}}$. Unfortunately, the photocurrent signal is below the equipment sensitivity, and we cannot obtain the response characteristic. Actually, the I-V curves have indicated the low performance of device ${\mathrm{S}}_{\mathrm{B}}$. This phenomenon further demonstrates the MQW structure has significant improvement on the detection performance. Figure 6(c) shows the bias dependence of the peak responsivity of device ${\mathrm{S}}_{\mathrm{A}}$; the inset is the corresponding EQE. The peak responsivity and EQE only slightly increase with the bias, which implies the carrier drift velocity and depletion region expansion are not the key factors limiting the performance. It further demonstrates that the carrier lifetime increase and the unipolar carrier transport multiplication caused by the MQW are primarily responsible for the performance improvement.

Figure 6(d) summarizes the zero-bias peak responsivity and its corresponding wavelength of this work and other reported self-powered AlGaN SBUV detectors. As is seen, although significant effort has been spent on improving the self-power performance over the past decades, the highest reported responsivity is just 0.211 A/W at 289 nm, corresponding to an EQE of 92%. The EQE below 100% indicates the current gain has not been realized in self-power work mode. Otherwise, because of the restriction by the crystal quality and $p$-doping of Al-rich nitrides, most of the reported AlGaN SBUV detectors work at wavelengths from 260 to 290 nm. Hence, even shorter wavelength detection capability can rarely be achieved. In our work, to our best knowledge, by introducing photoconductive gain into a vertical $p\text{-}i\text{-}n$ photodiode via MQW structure, the highest zero-bias responsivity (0.425 A/W), shortest peak wavelength (233 nm), and highest EQE (226%) of self-powered AlGaN SBUV detector until now have been realized.

In summary, AlGaN SBUV detectors are investigated in this work. A photoconductive gain mechanism introduced into the vertical $p\text{-}i\text{-}n$ photodiode is proposed to improve the performance of a self-powered AlGaN SBUV detector. Furthermore, the conception is realized through inserting MQW into the depletion region of an AlGaN $p\text{-}i\text{-}n$ photodiode. Experimentally, the MQW-assisted photodiode exhibits a peak responsivity of about 0.425 A/W at 233 nm at zero bias, which is the highest responsivity as well as shortest peak wavelength AlGaN self-powered SBUV detector reported up to now. The corresponding zero-bias EQE of 226% demonstrates the internal photocurrent gain. When compared with the structure without MQW, the gain is estimated to be about ${10}^3$ in magnitude. The MQW can bring about current gain mainly due to its strong hole confinement effect and the QCSE, resulting in a longer carrier lifetime and unipolar carrier transport, which are the key factors for generating photoconductive gain. The investigations have provided an alternative and effective approach to obtain high current gain in vertical structure AlGaN SBUV detectors at lower biases or even zero bias.