Innovative all-silicon based a-SiN_x:O/c-Si heterostructure solar-blind photodetector with both high responsivity and fast response speed

Solar-blind photodetectors (SBPDs), sensitive to the ultraviolet (UV) light of wavelength in the range of 200–280 nm, have recently aroused the interest of researchers owing to their wide applications in both civil and military fields, such as fire prevention, air purification, ozone monitoring, missile warning, space communication, and so on.¹ Si, usually regarded as the backbone of microelectronic industry, is also deemed to be a principal material for commercial PD products.² Due to the limitation of its sensitivity to low-energy radiation because of the narrow bandgap, the application of the Si material in solar-blind UV PDs is not as prevalent as its visible or near infrared counterparts. In practical use, Si PDs are typically capped with a solar-blind bandpass UV filter to suppress the extra response from light beyond the pass-band,^3,4 thus promoting the performance of Si SBPDs. However, it is difficult as well as costly to fabricate high-quality bandpass filters with wide bandwidth and achieve optical alignment of such filters with Si PDs. In addition, the reduction in the lifetime of Si SBPDs induced by the heat due to exposure to UV irradiation also severely deteriorates the device performance, thereby restricting their further development in practical applications.

To circumvent the disadvantages of Si SBPDs, a variety of wide-bandgap (WBG) semiconductors have emerged with several advantageous properties, such as wide bandgap of remarkably higher than 3.4 eV, high thermal conductivity, enhanced radiation hardness, etc.⁵ These WBG semiconductors normally include Al_xGa_1−xN, Mg_xZn_1−xO, diamond, Ga₂O₃, etc., which are particularly suitable for applications in solar-blind UV photodetection.^6–15 In the last decades, the technology maturity of preparing high-quality WBG semiconductors has advanced to the point where fabrication of high-performance SBPDs becomes available. Through continuous efforts, impressive progress has been made in the research on these WBG SBPDs with various structures, such as photoconductors,^7,8 metal–semiconductor–metal (MSM) PDs,^9–11 p–n (p–i–n) photodiodes,^12,13 Schottky barrier photodiodes,^14,15 and so on.

From the perspective of the operational principles, WBG SBPDs with varieties of structures can be classified into two categories, namely, photoconductive-type detectors and photovoltaic-type ones.¹⁶ The photovoltaic-type SBPDs typically have fast response speed owing to the immediate separation of photogenerated electron–hole pairs under driven by the built-in electric field, whereas they take on low photoresponsivity as restricted by the maximum gain of unity.¹⁷ For instance, Xie et al. demonstrated a heterojunction structure SBPD based on cubic Mg_xZn_1−xO with a fast response time of 15 µs and spectral responsivity of 1.16 A/W at 240 nm.¹² In contrast, the photoconductive-type detectors can acquire high responsivity due to the large internal gain, which often stems from the trapping of minority carriers in long-lifetime defect states. Nevertheless, the carrier trapping/releasing by the deep-level defects slows down the response speed.¹⁸ For example, Qin et al. fabricated ε-Ga₂O₃ MSM SBPD, which exhibited a record-high responsivity of 230 A/W at 254 nm and a decay time of 24 ms.⁹ Besides, Qian et al. prepared an a-GaO_x MSM PD with a high responsivity of 70.26 A/W at 254 nm and a decay time of 20 ms.¹¹ It is evident that the high responsivity can be achieved, but at the expense of the detector response speed.

In terms of device applications, both the photoresponsivity and response speed are key figures of merit for SBPDs. To date, some strategies have been exploited to try to break the trade-off between the photoresponsivity and response speed of the SBPDs. Cui et al. investigated the effect of oxygen flux regulation on the performance of Ga₂O₃ PDs.¹⁹ By increasing the oxygen flux during the magnetron sputtering process, the response speed was improved with the decay time shortened from 1.48 s to 19.1 µs due to the reduction in concentration of oxygen vacancy (V_O) defects. However, the suppressed carrier trapping effect of V_O defects resulted in the decrease in peak responsivity from 91.88 to 0.19 A/W. In addition, Zhang et al. incorporated the nitrogen in Ga₂O₃ to reduce the density of V_O defects, which resulted in a remarkable reduction in response time down to about 100 µs.²⁰ On the contrary, the peak responsivity decreased by more than two orders of magnitude, which could be attributed to increased recombination probability of photoexcited carriers induced by introduction of nitrogen. Other than Ga₂O₃, amorphous silicon nitride (Si₃N₄) films with a wide bandgap of 5.8 eV have been recently proposed to be an available material for deep UV photodetection. Li et al. attempted to obtain Si₃N₄ photoconductive PDs with both high responsivity and fast response speed by simultaneously shortening the carrier transit distance and passivating the trap states in the photosensitive layer.²¹ Nevertheless, the responsivity and response time of PD turned out to be 1.09 mA/W and 127 ms, respectively, which have not yet achieved the goals of both high responsivity and fast response speed. As can be seen from the results attained previously, the similarity among them lies in the device structure where the carrier photogeneration/recombination process and carrier transport process occur in the same layer. In this case, while reducing the density of trap states in the photosensitive layer to promote the response speed of PDs, it will inevitably exert an adverse effect on the carrier transport process in the same layer, thus resulting in weakened photoresponsivity.

Facing the challenge of coordinating the inherent contradiction between high responsivity and fast response speed, in the present work, we propose an innovative design idea, which separates the carrier photogeneration/recombination process from the carrier transport process in the different layers. Specifically, we employ a luminescent WBG amorphous oxygen-doped silicon nitride (a-SiN_x:O) film as an absorption layer combining with a p-type crystalline Si (c-Si) substrate as a carrier transport layer to construct a photoconductive-type MSM structure SBPD. In our previous work,^22–24 a-SiN_x:O films were fabricated by plasma enhanced chemical vapor deposition (PECVD). By controlling the Si/N composition ratio, the optical bandgap E_g^opt can be tuned in the range from 2.91 to 4.54 eV,²⁵ which covers the solar-blind spectral region. The incorporation of oxygen atoms as impurity atoms in the SiN_x network creates N–Si–O bonding configuration that leads to the formation of a recombination center for photogenerated carriers. Due to the low-density interfacial trap states at the SiN_x:O/Si interface benefiting from the good passivation effect of the a-SiN_x:O film, the photogenerated carriers in the a-SiN_x:O layer can be transferred to the c-Si layer under driven by the built-in electric field in a-SiN_x:O/c-Si heterojunction without suffering from interface defect states. Meanwhile, in the c-Si transport layer, as the transit time of injected carriers is much shorter than their recombination lifetime, the device can yield a large photoconductive gain. This innovative a-SiN_x:O/c-Si bilayer SBPD exhibits prominent performance, including the high spectral responsivity of 4 × 10³ A/W at 225 nm, the fast response speed of 4.3 µs, and the specific detectivity (D*) on the order of 10¹⁴ Jones. In contrast to that of normal photoconductive-type SBPDs reported by Cui et al.,¹⁹ Zhang et al.,²⁰ and Li et al.,²¹ both high responsivity and fast response speed become available for this all-silicon based photoconductive-type SBPD. On the other hand, when compared with the WBG crystalline semiconductor counterparts, the design of all-silicon a-SiN_x:O/c-Si bilayer SBPDs reveals advantages not only in adopting economic Si-based materials but also in manufacturing processes compatible with mature CMOS technology, so that it is preferable for the development of cost-effective large-area SBPD arrays.

The a-SiN_x:O films with a thickness of ∼100 nm were deposited on the p-type c-Si or quartz substrate by means of PECVD. Prior to the deposition, the native oxide on the surface of p-type c-Si was rinsed out by using diluted hydrofluoric acid solution so that the surface of the c-Si substrate can be cleaned for subsequent deposition of the high-quality a-SiN_x:O film. The a-SiN_x:O film was deposited by using a gas mixture of silane (SiH₄), ammonia (NH₃), nitrogen (N₂), and diluted oxygen (O₂). The bandgap of the a-SiN_x:O film can be increased by varying the gas flow ratio R (R = [NH₃]/[SiH₄]) from 0.5 to 8, which has been described in our previous work.²⁵ Herein, the flow ratio R was chosen to be 8 for the film optical bandgap to match the solar-blind UV band. Oxygen atoms were incorporated in the film as impurity atoms, which exerted no influence on the film bandgap, whereas they contributed to the formation of oxygen-related defect states. The deposition parameters of radio frequency power, reaction pressure, and substrate temperature were all kept constant at 20 W, 580 mTorr, and 250 °C, respectively. After that, the samples were annealed at 1000 °C in N₂ atmosphere for 1 h. The bonding configurations and chemical compositions of the sample were investigated by Fourier transform infrared (FTIR) and x-ray photoelectron spectroscopy (XPS) measurements, which were performed by using the Nexus 870 FTIR spectrometer and the Thermo ESCALAB 250 x-ray photoelectron spectrometer with monochromatic Al K alpha radiation source, respectively. Before XPS measurements, the top layer of the sample with a thickness of about 40 nm was got rid of by an Ar ion beam (2 keV, sputtering time ∼400 s). The optical absorbance spectrum of the film was deduced from transmission and reflectance spectra measured by the Shimadzu UV–vis–NIR spectrophotometer.

For convenience of electrical probing, the interdigital Ti(20 nm)/Au(80 nm) electrodes with ten pairs of fingers were patterned on the top of the a-SiN_x:O film by using electron beam evaporation combined with the conventional standard photolithography as well as lift-off methods. The length, width, and spacing of interdigital electrodes were 400, 10, and 10 µm, respectively. The measurements of current–voltage (I–V), capacitance–voltage (C–V), and conductance–voltage (G–V) were performed at room temperature by using an Agilent B1500 semiconductor parameter analyzer. The photoresponse spectra were measured by a monochromator (Horiba, iHR320) equipped with a 500 W xenon arc lamp as the optical excitation source. The incident power density was calibrated by using a Si reference photodiode. A 213 nm pulse laser with a pulse width of 1.2 ns at a repetition rate of 1 kHz was employed as the excitation source for transient photoresponse measurements, and a digital oscilloscope (Tektronix, TBS1120) was employed for data recording.

Figure 1(a) displays XPS spectra of Si 2p, N 1s, and O 1s photoelectron emission peaks for a-SiN_x:O film. By integrating each peak area, the atomic concentrations of Si, N, and O were determined to be around 51.0%, 46.7%, and 1.4% respectively, as shown in the inset of Fig. 1(a). A small amount of O in the film denotes that the oxygen is incorporated as an impurity element in the a-SiN_x networks. The binding energy values of three photoelectron emission peaks were calibrated by the contaminant carbon peak at 284.8 eV. The Si 2p peak is centered at around 102.23 eV, which is just intermediate between the binding energy of the Si 2p peak of stoichiometric Si₃N₄ (101.9 eV) and that of SiO₂ (103.35 eV). It implies that the N–Si–O bonding configuration is present in the a-SiN_x:O film.²⁶ FTIR spectroscopy was also employed to characterize the bonding configurations of the a-SiN_x:O film, as demonstrated in Fig. S1 of the supplementary material. The results of the FTIR spectrum also verify the presence of N–Si–O bonds in a-SiN_x:O films,²⁷ which is in coincidence with the results measured by XPS. In our preceding investigation on luminescence properties of a-SiN_x:O films, it has been confirmed that the N–Si–O bonding configuration leads to a new defect state, which acts as an efficient radiative recombination center responsible for light emission from a-SiN_x:O films.²⁸ 

On the basis of transmittance and reflectance spectra, the optical absorbance spectrum of the a-SiN_x:O film can be obtained in terms of the formula A(absorbance) = 100% − T (transmittance) – R (reflectance), as drawn in Fig. 1(b). The spectrum is on the rise below the band edge of around 258 nm, suggesting that the spectral range of absorption precisely falls in the coverage of the solar-blind UV detection. As plotted in the inset of Fig. 1(b), the dependence of (αhν)^1/2 on hν in the higher photon energy region was linearly fitted by the Tauc relation of (αhν)^1/2 = K (hν – E_g), where α is the absorption coefficient, hν is the photon energy, K is a constant, and E_g is the optical bandgap energy. As a result of fitting, the optical bandgap is determined to be 4.8 eV, corresponding to the absorption band edge of 258 nm.

For exploring its absorption features in the solar-blind UV region, the a-SiN_x:O film was first prepared on the quartz substrate to build an a-SiN_x:O/quartz MSM PD, whose I–V characteristics are presented in Fig. 2(a). The dark current of the PD is so small that could be hardly measured when it approaches the measurement limit of the instrument. Under illumination by the UV lamp of 254 nm, the PD exhibits a distinct photocurrent with symmetrical shape as the bias voltage varies, while it is totally blind to the light of 365 nm. Unfortunately, although the bias voltage is elevated to 20 V, the photocurrent can only reach around 40 pA, probably due to the highly resistive nature of the a-SiN_x:O film. The results indicate that the a-SiN_x:O film is able to act as a good absorption layer for the solar-blind UV region, while it is not suitable to be employed individually to construct an effective PD due to its low charge carrier concentration.

In this study, a-SiN_x:O/c-Si bilayer SBPD was designed by depositing a layer of the WBG a-SiN_x:O film on the p-type c-Si substrate. Thereinto, the a-SiN_x:O film functions as a UV absorption layer, while the c-Si substrate acts as a carrier transport layer. By the adoption of this design idea, the carrier photogeneration/recombination process can be separated from the carrier transport process in the different layers. The structure sketch of the a-SiN_x:O/c-Si bilayer SBPD is described in the inset of Fig. 2(b). On the top surface of the a-SiN_x:O film, the Ti/Au electrodes take the configuration of interdigitated fingers, which consists of ten pairs with each finger 400 µm in length and 10 µm in width. To facilitate the transit of the carriers across the c-Si transport layer in a short time, the spacing of the interdigital electrodes is kept at a short distance of 10 µm, as shown by a microscope photograph of the electrode.

Figure 2(b) presents the I–V characteristics of the a-SiN_x:O/c-Si bilayer SBPD that is quite different from the a-SiN_x:O/quartz one. When the bias voltage varies from −2 to 2 V, the I–V curves show a symmetrical shape both in a dark environment and under irradiation by the UV lamp. In the dark, the PD exhibits dark current on the order of 10⁻⁹ A at a bias voltage of 1 V. Once the a-SiN_x:O/c-Si SBPD is irradiated by the UV lamp of 254 nm, the photocurrent rises abruptly by about 2 orders of magnitude relative to the dark current. In contrast, the photocurrent of the a-SiN_x:O/c-Si SBPD is approximately 5–6 orders of magnitude higher than that of the a-SiN_x:O/quartz PD. The above evidence indicates that the current in the a-SiN_x:O/c-Si SBPD preferentially passes through two back-to-back SiN_x:O/Si heterojunctions connected on a coplanar surface of c-Si rather than through the top a-SiN_x:O film. The validity of this inference is based on the following two points, i.e., much lower resistivity of Si than a-SiN_x:O and good passivation of a-SiN_x:O on the c-Si surface. From C–V and G–V measurements as shown in Fig. S2 of the supplementary material, the interfacial state density at the a-SiN_x:O/c-Si interface was deduced to be around 5.5 × 10¹¹ eV⁻¹ cm⁻², signifying a good passivation effect of a-SiN_x:O on the surface of c-Si. In this case, the current can flow across the a-SiN_x:O/c-Si interface without suffering from interface defect states. For this reason, the symmetrical I–V characteristics displayed in Fig. 2(b) should be ascribed to the symmetry of two a-SiN_x:O/c-Si heterojunctions between the electrodes. In addition, for the sake of characterizing the switch on/off current ratio, the photo-to-dark current ratio (PDCR) has been calculated by the relation: PDCR = (I_illum − I_dark)/I_dark, where I_illum is the current under the illumination of the UV lamp, while I_dark is the current in the dark. One can see that PDCR is capable of reaching ∼10², indicative of high photosensitivity for a-SiN_x:O/c-Si bilayer SBPDs.

Figure 3 shows the spectral response of the a-SiN_x:O/c-Si heterostructure SBPD when applied by various bias voltages, which are plotted in a semi-log scale. The responsivity of the PD is derived by R = (I_illum − I_dark)/P_λS, where I_illum and I_dark are currents under the UV illumination and in the dark, respectively, P_λ represents the intensity of incident monochromatic light, and S denotes the illuminated area of PD. Each photoresponse curve shows a peak with its center located at around 225 nm. With the applied voltage increasing from 5 to 15 V, the responsivity in the solar-blind spectral range is raised gradually. As guided by the dashed line in Fig. 3, the cut-off wavelength of responsivity (at −20 dB) is around 258 nm, which is in line with the absorption band edge of the a-SiN_x:O layer in Fig. 1(b). As defined to be the ratio between peak responsivity at 225 nm and that at 400 nm, the UV/visible rejection ratio (R_225nm/R_400nm) is derived to be around 2 × 10², indicating a relatively high spectral selectivity to the solar-blind band. The inset of Fig. 3 shows the peak responsivity at 225 nm as a function of bias voltage. With the bias varying from 5 to 15 V, the peak responsivity increases and eventually achieves as high as about 4 × 10³ A/W at a bias of 15 V, which is an order of magnitude higher than that of Ga₂O₃ MSM photoconductive-type PDs reported by Qin et al.⁹ and Xu et al.,¹⁰ respectively. The linearity relationship of peak responsivity vs the bias voltage suggests that the spectral response of a-SiN_x:O/c-Si heterostructure SBPDs could be most likely dominated by the photoconductive gain mechanism, which will be described in detail below.

According to the spectral responsivity of the PD, the external quantum efficiency (EQE) and specific detectivity (D*) can be deduced, respectively. EQE of the PD, which is defined as the number of collected electron–hole pairs per incident photon, can be estimated by the use of the equation η = Rhc/qλ, where R is the responsivity, h is the Planck constant, c is the speed of light, q is the electron charge, and λ is the incident light wavelength.²⁹ It is found that the peak value of EQE at a bias of 15 V can be evaluated to be around 2.2 × 10⁴, which is much greater than unity, implying the presence of a large internal gain. In addition, D* is generally exploited to measure the capability of detecting small signals. The thermal-related noise and the shot noise are two main contributions to the noise current of the WBG SBPDs. Because of the large resistance of the a-SiN_x:O layer and the lower voltage across the device, the thermal-related noise exerts less effect on the device noise than the shot noise from dark current. Hence, the shot noise predominantly contributes to the total noise of the device.³⁰ D* can be determined by the equation $D*=R/2qJd$⁠, where R is the peak responsivity, q is the electron charge, and J_d is the current density in the dark condition.³¹ The evaluated D* on the order of 10¹⁴ Jones can be achieved, which signifies that the PD has great potential for energy-efficient weak signal detection.

In order to investigate the response speed, which is another significant parameter of PD, the time-dependent photocurrent response of a-SiN_x:O/c-Si heterostructure SBPDs under the chopper-modulated UV illumination was measured at 5 V bias, as shown in Fig. S3 of the supplementary material. According to the multi-cycle photoresponse of PD, the photocurrent of the device is demonstrated to be stable and reproducible during operation. However, since the response time measured by a chopped light source is much longer as restricted by the excitation profile of the mechanically chopped light, the response speed of the device is further explored by the transient response measurement under irradiation by a 213 nm pulsed laser with a pulse width of 1.2 ns, as displayed in Fig. 4. From Fig. 4(a), one can see that the photocurrent can follow the varying pulsed laser signal and exhibit rapid on/off switching. The dynamic photocurrent response of PD also shows its good stability and reproducibility. In Fig. 4(b), a magnified one cycle of transient photoresponse is plotted. The decreasing portion of time-resolved response is fitted by using a biexponential relationship: I(t) = I₀ + I₁ exp(−t/τ₁) + I₂ exp(−t/τ₂), where I₀ is the steady-state dark current contribution, I₁ and I₂ are the photocurrent components, and τ₁ and τ₂ represent the fast and slow-decay lifetime, respectively. A good fit is accomplished and yields the fast-decay time τ₁ = 4.18 µs, the slow-decay time τ₂ = 28.6 µs, and the component weight ratio I₁/I₂ = 180 for the decay characteristics at 5 V bias. By further use of the relation ⟨τ⟩ = (I₁τ₁ + I₂τ₂)/(I₁ + I₂),³² the corresponding weighted-average decay lifetime has been calculated to be 4.3 µs, suggesting that the fast-decay component is dominant. As is well known, the response speed of the PD is usually limited by RC time constant, where R is the total resistance and C is the capacitance of the depletion region in the PD. From I–V and C–V measurements, the resistance of PD was determined to be on the order of 10⁶ Ω, and the detector capacitance was estimated to be ∼30 pF, respectively. Hence, the RC time constant is evaluated to be roughly 30 µs, which corresponds to the slow-decay time τ₂ = 28.6 µs. It implies that RC time constant has little impact on the measurement of the fast response speed of the device.

Such a fast response speed of the a-SiN_x:O/c-Si SBPD should be probably attributed to the rapid recombination of photogenerated carriers, the reason for which can be described from the following two aspects. One is suppression of charge trapping/releasing processes due to the low-density interfacial trap states at the well-passivated a-SiN_x:O/c-Si interface as well as the absence of bulk defect states in the c-Si transport layer. The other more essential reason for the fast response speed is the important role of the N–Si–O recombination center in facilitating the recombination of photogenerated carriers in the a-SiN_x:O absorption layer. As revealed in our previous work, N–Si–O bonding configuration in the a-SiN_x:O film created corresponding defect states to serve as the radiative recombination center, which was determined to lie in the bandgap of the film. The radiative recombination of optically pumped carriers via the N–Si–O recombination center would result in photoluminescence (PL) from the film. The recombination processes and luminescent mechanism are displayed in Figs. S4 and S5 of the supplementary material. By means of time-resolved PL (TRPL) measurements, the carrier recombination lifetime via N–Si–O defects was deduced to be in the nanosecond range,³³ which is fast enough for the photogenerated carriers to recombine promptly in a-SiN_x:O absorption layer of our SBPD. Consequently, the device can switch off within a few microseconds after shutting off the UV light stimuli.

To figure out the origin of the large internal gain in a-SiN_x:O/c-Si heterostructure SBPDs, the sketch of the operational principle and the schematic energy band diagram of the device are illustrated in Fig. 5. The energy band diagrams of individual Au, Ti, a-SiN_x:O, and c-Si are depicted in Fig. 5(c). With reference to parameters provided by the previous literature,³⁴ the electron affinities (χ) for a-SiN_x:O and c-Si are taken as 2.0 and 4.05 eV, respectively. Accordingly, the conduction band offset ΔE_c and the valence band offset ΔE_v between a-SiN_x:O and c-Si are estimated to be 2.05 and 1.63 eV, respectively. Hence, a-SiN_x:O/c-Si heterojunction with a type I band alignment is formed, as demonstrated in Fig. 5(d). The built-in electric field in the direction from a-SiN_x:O to c-Si was thus generated. As can be seen from Figs. 5(a) and 5(b), in the area where the a-SiN_x:O layer is exposed to the UV light, the electron–hole pairs are photoexcited by the photons with energies higher than the bandgap of a-SiN_x:O. The photogenerated holes at the vicinity of the a-SiN_x:O/c-Si interface are swept swiftly out of the a-SiN_x:O absorption layer to the c-Si transport layer by the built-in electric field of heterojunction and drift under the action of external bias, while the photogenerated electrons are prone to relax down to the N–Si–O recombination centers that are situated below the conduction band in the bandgap of a-SiN_x:O, as indicated by Fig. 5(d). In this case, when the carrier transit process in the c-Si layer is much faster than the carrier recombination process in the a-SiN_x:O layer, the carriers will circulate through the transport layer for charge neutrality and be collected by the electrodes for multiple times, thus producing a large internal gain. Furthermore, the photoconductive gain can be expressed by the relation: G = τ_r/τ_t = τ_r/(l²/μV), where τ_r is the carrier recombination lifetime, τ_t is the carrier transit time, l is the spacing of the interdigitated finger, μ is the carrier mobility, and V is the bias voltage.³⁵ By taking τ_r = 4 µs (decay lifetime), l = 10 µm, μ = 500 cm²/(V · s) (hole mobility in c-Si), and V = 15 V, the photoconductive gain is derived to be ∼3.0 × 10⁴, which is of the same order of magnitude as the corresponding EQE of 2.2 × 10⁴. Therefore, it provides evidence for confirming the predominant role of the photoconductive gain mechanism in the above-mentioned high responsivity of 4 × 10³ A/W.

In addition to the investigation on a-SiN_x:O/c-Si SBPDs with symmetric back-to-back SiN_x:O/Si heterojunctions, we have also studied the performance of the SBPDs with asymmetric back-to-back SiN_x:O/Si heterojunctions, which is prone to be formed when increasing the size of electrodes. The current density–voltage (J–V) characteristics of a-SiN_x:O/c-Si heterostructure SBPDs with large interdigital electrodes of 5 × 5 mm² in size are depicted in Fig. 6(a). In the dark, the J–V curve of the device exhibits an obvious rectifying behavior, and the rectification ratio of ∼10³ can be reached within ±10 V. Once the PD is irradiated with the UV lamp of 254 nm, the current rises abruptly by 2 orders of magnitude at a bias voltage of 10 V. However, negligible variation between the photocurrent and the dark current is found at negative voltages. In contrast to the symmetrical shape of the I–V curve displayed in Fig. 2(b), the photovoltaic-like behavior in the J–V curve presented here may be ascribed to the asymmetry of two back-to-back SiN_x:O/Si heterojunctions as a result of the increase in the electrode size.

Figure 6(b) shows the semi-log plot of the spectral response of the a-SiN_x:O/c-Si heterostructure SBPD when the device is applied by zero bias. As seen from Fig. 6(b), it is of interest to find that a distinct peak emerges in the responsivity curve with its central wavelength at around 245 nm. The peak responsivity is about 9 mA/W, suggesting that the PD can operate without any external power supply. Namely, the PD is able to be driven in the self-powered mode. The observed self-powered photoresponsivity in the PD should be attributed to the asymmetric back-to-back SiN_x:O/Si heterojunction, the formation of which probably results from the inhomogenious distribution of interfacial states at the SiN_x:O/Si interface.^36,37 In this case, for the self-powered operation mode, the separation, transit, and collection of photogenerated charge carriers under UV irradiation are facilitated by remanent potential stemming from offsetting the built-in electric field in two asymmetric back-to-back SiN_x:O/Si heterojunctions, as illustrated by the inset of Fig. 6(b).

Figure 7 summarizes the relationship of responsivity vs response time for some representative WBG semiconductor SBPDs with various structures, including photodiode,^{12,14,15,21,38} photoconductor,^7,8 and MSM.^{9–11,37,39–42} As guided by the dashed arrow lines, the statistical results of the data constitute an area with a trend of sloping upward, implying the intrinsic contradiction between the high responsivity and fast response speed in these WBG semiconductor SBPDs. Particularly, for the MSM photoconductive-type PDs, most of them exhibit higher responsivities (around 10² A/W), whereas they take on longer response times (hundreds of milliseconds). However, the a-SiN_x:O/c-Si heterostructure SBPD manifests both high photoresponsivity and fast response speed, which thus can overcome the inherent contradiction mentioned above. In contrast to WBG semiconductor counterparts, this all-silicon based a-SiN_x:O/c-Si heterostructure SBPD with prominent performance is highly competitive in fabricating cost-effective, high-performance, and large-area SBPD arrays. Nevertheless, it is noteworthy that the dark current of a-SiN_x:O/c-Si heterostructure SBPD is larger when compared to that of WBG SBPDs fabricated on the sapphire substrate, whereas it is almost the same or smaller than that of WBG semiconductor/c-Si SBPDs.^36,37 The strategies for reducing the dark current of the device on the Si substrate will be further explored in future work.

In summary, with an innovative design idea of separating the carrier transport process from the carrier photogeneration/recombination process in the different layers, an all-silicon based a-SiN_x:O/c-Si heterostructure SBPD was fabricated by utilizing the a-SiN_x:O film as a solar-blind UV absorption layer and exploiting c-Si as a transport layer. Benefiting from the designed bilayer device structure and the role of the N–Si–O recombination center in the a-SiN_x:O absorption layer, this new type of SBPD can achieve prominent performance with both high responsivity (4 × 10³ A/W at 225 nm) and fast response speed (4.3 µs), as well as high specific detectivity on the order of 10¹⁴ Jones, which can overcome the inherent contradiction of high responsivity and fast response speed in the normal photoconductive-type PDs. Moreover, the SBPD with asymmetric back-to-back SiN_x:O/Si heterojunctions exhibits a peak responsivity of 9 mA/W at 245 nm at zero bias, indicative of the ability to operate in the self-powered mode. In view of adopting the cost-effective Si-based material and introducing architecture comprising two active layers, the all-silicon based a-SiN_x:O/c-Si heterostructure SBPD provides a promising new avenue for implementing the high-performance photodetection and large-area SBPD arrays.

See Fig. S1 of the supplementary material for details on the bonding configuration characterization of the a-SiN_x:O absorption layer. Figure S2 provides C–V and G–V measurements for deducing the interfacial state density at the SiN_x:O/Si interface. The time-dependent photoresponse of the a-SiN_x:O/c-Si heterostructure SBPD when exposed to the chopper-modulated UV illumination is provided in Fig. S3. In Fig. S4, the PL and PL excitation (PLE) spectra of the a-SiN_x:O film for disclosing the luminescent mechanism are presented. Finally, Fig. S5 illustrates the carrier recombination processes in the a-SiN_x:O absorption layer.

The authors acknowledge funding support from the National Key R&D Program of China (Grant No. 2018YFB2200101), the National Natural Science Foundation of China (Grant No. 61634003), the Project 333 of Jiangsu Province (Grant No. BRA2019188), and the Science and Technology Support Program of Taizhou (Grant No. TS201921). The authors also would like to thank Professor D. J. Chen, Professor J. D. Ye, and Professor H. Lu from Nanjing University for their support in responsivity and response speed measurements as well as fruitful discussions.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.