1. Introduction

Detecting near-infrared (NIR) light is essential for future technological applications such as light detection and ranging (LiDAR)-based autonomous driving, machine vision, optical communication, and bioimaging [1]. For those applications, it is crucial that photodiodes (PDs) detect NIR light with high efficiency. Generally narrow bandgap materials are utilized as PDs such as InGaAs to improve NIR detection efficiency. Despite the high NIR response of these materials, silicon (Si) has gained attention for NIR PD due to its cost-effectiveness, stability, and well-established semiconductor fabrication [2]. But its bandgap limits the ability to detect NIR wavelengths, resulting in a steep decrease in photoresponse near the bandgap edge [3]. Therefore, significant research efforts have been focused on improving the NIR detection efficiency of Si PDs. One promising approach is to use Si heterojunctions with narrow bandgap materials, which can dramatically enhance the NIR photoresponse [2,4,5]. However, in the case of Si heterojunction, additional research appears to be necessary to ensure reliability and low-cost fabrication. Another approach is to use metal-Si Schottky junctions. This can not only improve the NIR photoresponse, but also extend the Si photodetection range from the NIR to short-wave infrared through an internal photoemission process [6–8].

Among these various approaches for enhancing NIR response, efficient light management has great potential by utilizing optical resonances in nanostructures [9]. When the structural dimensions are comparable to or smaller than the wavelength, the structure can possess optical resonances caused by light-matter interactions [10]. Since these optical resonances can effectively increase the light absorption path, various types of resonances have been reported including Fabry-Pérot, photonic crystal, and whispering gallery mode (WGM) resonances [11]. In particular, WGM has been regarded as the most attractive resonance because this enables circulation of the incident light for a very long time along a concave boundary through total internal reflection [12]. To date, diverse three-dimensional (3D) nanostructures have been investigated to induce WGM resonance, including nanospheres [10,13], nanoshells [12], nanopillars [14–16], and nanocones [17,18]. For the case of spherical resonators, the WGM resonance demonstrates photocurrent enhancement at longer wavelengths than the typical absorption edge of silicon [13]. Furthermore, stacked nanoshell structures induce multiple low-quality factor WGMs, enhancing broadband absorption for visible and NIR light [12]. Silicon nanopillar (SiNP) resonators are also promising structures for practical applications due to their outstanding optical properties as well as their uniform and reproducible fabrication. However, most research has mainly focused on only inversely tapered structures and hexagonal SiNP structures to induce WGM resonances.

In this work, we propose a highly sensitive NIR photodiode using wave-shaped sidewall SiNPs (WS-SiNPs). The WS-SiNP photodiodes were fabricated by conventional semiconductor process. In addition, control devices with straight sidewall SiNPs (S-SiNPs) and planar structures were fabricated simultaneously to compare the structural differences. Optoelectrical characteristics were evaluated by extracting spectral responsivity (R_λ), external quantum efficiency (EQE), and reflectance. Also, improved R_λ of WS-SiNP devices were compared to control PDs and commercial NIR PDs. Finally, 3D finite-differential-time-domain (FDTD) simulations were performed to analyze the advantage of WS-SiNP structure for enhanced NIR photoresponse.

2. Experiment and simulation details

2.1 Design and fabrication of WS-SiNP PDs

Figure 1(a) presents a schematic of the WS-SiNP photodiode, which consists of a radial p-n junction, a back-surface field (BSF) layer, and aluminum (Al) electrodes. The WS-SiNP arrays were fabricated using the conventional top-down process on a 200 mm silicon wafer (<100>, p-type/Boron, 8-12 Ω·cm, 700 µm thick). A 500 nm SiO₂ film was firstly deposited on the front side using plasma-enhanced chemical vapor deposition as a hard mask layer for the Si etching process. A BSF layer was formed with boron ion implantation at a dose of 4 × 10¹⁵cm⁻² and 25 keV, followed by rapid thermal annealing at 1,000 °C for 1 minute to activate the dopant. BSF layer has the advantage of reducing charge recombination losses by generating the electric field for faster transport of minority carriers [19]. The hard mask of SiO₂ layer was patterned to nanodots having a diameter of 800 nm and a pitch of 2.4 µm using i-line stepper lithography. The patterned SiO₂ layer was etched by inductively coupled plasma reactive ion etching (ICP-RIE) with a mixture of CF₄/CH₃/Ar gas (100/100/20 sccm). Subsequently, the WS-SiNP arrays were etched for 25 cycles with precise control of the WS depth (W_D) and height (W_H) by adjusting etching and passivation processes of the deep reactive ion etcher (DRIE). During the etch process, an SF₆/O₂ gas (400/40 sccm) mixture was used for 5 seconds per cycle, maintaining the chamber pressure, source, and bias power at 20 mTorr, 2400 W, and 45 W, respectively. During the passivation process, C₄F₈ gas was used for 4 seconds per cycle with 2400 W of source power and 20 mTorr of chamber pressure. Also, the S-SiNP devices were simultaneously fabricated as a control structure using ICP-RIE with well-controlled HBr/Cl₂/O₂ gas mixture. After forming SiNPs, the residual photoresist and hard masks were removed through the piranha cleaning process and diluted hydrofluoric (DHF) acid solution. Also, the etched surface was treated with sacrificial oxidation, followed by wet oxide removal right before forming a radial p-n junction. A highly doped 30 nm n-type silicon film (phosphorus, 2.5 × 10¹⁹cm⁻³) was deposited epitaxially on the front surface of the devices using ultra-high vacuum chemical vapor deposition. The radial p-n junction efficiently collects photo-generated carriers because the carrier diffusion distance of the junction is limited to the diameter of SiNPs. Thus, photogenerated carriers travel a short distance and are collected before recombining [20]. For isolation of the fabricated devices, the edges of each device were etched by ICP-RIE. As a front electrode, Ti (20 nm) and Al (200 nm) metal layers were selectively deposited around the SiNP array region using lift-off process. Also, a 100 nm Al layer was deposited on the back side of the devices as a rear electrode. Finally, the fabricated devices were separated into each chip (1 × 1 cm²) by dicing process for device characterization. The planar devices were fabricated with the same process except for the nanodot pattering and etching processes.

 

Fig. 1. (a) Schematic diagram of the WS-SiNP photodiode under NIR illumination. Inset is the enlarged image of a single SiNP composed of a radial p-n junction. W_D and W_H indicate the depth and height of a single WS pattern. (b) Image of the PDs fabricated on a 200 mm wafer. (c) Image of a diced photodiode chip with the size of 1 × 1 cm². The orange dashed line is an active region (0.19635 cm²), composed of n⁺ layer. SEM image of fabricated WS-SiNPs from (d) a bird’s eye point of view and (e) cross-sectional view. SEM image of fabricated S-SiNPs from (f) a bird’s eye point of view and (g) cross-sectional view as a control structure. D_T, D_B, H, and θ_B represent the top diameter, bottom diameter, height, and bottom end outer angle of the NP, respectively.

Download Full Size | PDF

Figure 1(b) and (c) display the fabricated wafer and diced chip of WS-SiNP PDs. All SiNPs were hexagonally arranged in the active area. Figure 1(d) and (e) show the scanning electron microscope (SEM) images of the WS-SiNP array with a spacing of 1.6 µm between NPs. WS surface has W_D of 130 nm and W_H of 211 nm. The top diameter (D_T), bottom diameter (D_B), and height (H) of a WS-SiNP are 803 nm, 904 nm, and 5.2 µm, respectively. Figure 1(f) and (g) depict the SEM images of the S-SiNP array with a spacing of 1.6 µm between NPs which is the same for the WS-SiNP device. The fabricated S-SiNP structure has a D_T of 809 nm, D_B of 899 nm, and H of 5.2 µm. The WS-SiNP and the S-SiNP have almost similar physical dimensions (D_T, D_B, H and spacing) except for the WS sidewall structure. Notably, the overall diameter of both structures gradually increases from top to bottom while retaining bottom outer angles (θ_B) of NPs about 91°.

2.2 FDTD simulation on the WS-SiNP arrays

3D finite difference time domain (FDTD) simulation was carried out using Ansys/Lumerical software. The simulated SiNPs structures are identical to the SEM images of the fabricated structures. To minimize the simulation time, the bottom monitor was fixed to the 1 µm depth of the substrate. This depth corresponds to the calculated depletion region for the doping concentration of n-type and p-type Si layers. The refractive index and absorption coefficient were properly fitted from the experimental data on Si before starting the simulation [21]. Periodic boundary conditions were used to implement the SiNP array environment. The x-plane and the y-plane of the designed structures were set to anti-symmetric and symmetric conditions, respectively. The boundary in the z-direction was surrounded by perfectly matched layer, an infinite boundary that could absorb electromagnetic waves [22]. An x-polarized external plane wave with a spectral range of 725-1,025 nm was vertically incident from the top surface to the bottom of the designed structures.

2.3 Measurements of photodiode characteristics

All optoelectrical properties were measured in a dark box to exclude the external light effects. The light source was generated by a 500 W xenon lamp and passed through a diffraction grating monochromator (74064, Newport) with a wavelength resolution of 0.25 nm. The diffracted light was filtered into monochromatic light through an optical filter and illuminated within the active area with different radiant power intensities. A calibrated silicon photodiode (918D, Newport) and a power meter (1918-R, Newport) were used for the reliable calculation of the light power intensity. The output current was measured using a precision semiconductor analyzer (4200-SCS, Keithley) and the transient optical response was measured at a time interval of 40 ms. The spectral response (R_λ and EQE) was evaluated under a reverse bias condition of 0 V with wavelengths swept in 25 nm steps from 750 nm to 1000 nm. In order to investigate the optical characteristic of SiNPs, the surface reflectance was measured with a UV-vis-NIR spectrometer (UV-3600 plus, Shimadzu) in the spectral range from 725 to 1,025 nm.

3. Results and discussion

3.1 Optoelectrical characterization

We analyzed the optoelectronic characteristics of the WS-SiNP PDs. Figure 2(a) shows the current-voltage (I-V) curves of a WS-SiNP PD under dark and illumination conditions. In this experiment, we illuminated the monochromatic wavelength of 905 nm, commonly used in the LiDAR system [23], with a light intensity of 100 µWcm⁻². In the dark state, the I-V curve presents apparent current rectification, indicating the successful formation of the radial p-n junction. Also, the dark current (I_dark) is much lower at zero-bias than under other reverse bias conditions. At zero-bias condition, I_dark is mostly dominated by thermal noise, while in reverse bias condition, the shot noise overtakes the thermal noise [24–26]. Thus, the zero-bias operation reduces the effect of shot noise, which favors stable operation even in low-intensity light. In the illumination state, the current (I_light) increases rapidly because of the photogenerated carriers. The difference between I_light and I_dark represents photocurrent (I_ph) used to extract NIR response. At zero-bias, the current dramatically increases from 30 pA to 12.7 µA, leading to a high I_light/I_dark ratio of 4.2 × 10⁵. This high I_light/I_dark ratio suggests that the device is able to convert light into a stable electrical output signal, even if it detects a small intensity of light. Another advantage of zero bias operation is that the device can operate without any external power supply [27]. Therefore, all PDs are characterized at zero bias. The inset in Fig. 2(a) compares the I_ph of WS-SiNP, S-SiNP and planar PDs. The WS-SiNP device has a higher output signal than that of S-SiNP and planar devices. Since all devices were measured under the same conditions, this result means that the geometric differences in the WS-SiNP contribute to its photoelectric characteristics, allowing for improved performance at the 905 nm wavelength compared to control devices.

 

Fig. 2. (a) Current-voltage (I-V) curves of a WS-SiNP photodiode in the dark and under 905 nm illumination with an intensity of 100 µWcm⁻². Inset is a comparison of the optical response of WS-SiNP, S-SiNP and planar PDs under the same conditions. (b) Current-time characteristic of a WS-SiNP photodiode at 0 V under 905 nm illumination with different light power densities and (c) the corresponding photocurrent as a function of light intensity; The red dashed line is a fitting curve with experimental data.

Download Full Size | PDF

Figure 2(b) shows the current-time response of the WS-SiNP PD for different optical power intensities at zero bias. The On and Off states demonstrate good stability and reversibility during the rapid transition from illumination to dark. Figure 2(c) summarizes the absolute value of I_ph according to the optical power intensity in Fig. 2(b). To analyze a tendency in I_ph as the light intensity, we fitted the data with the power law formula of I_ph = CP^α, where C is a constant, P is the optical power density, and α is the exponent value of the light response to its intensity. The extracted α value is 0.99, which means that collected carriers increase proportionally as the number of incident photons increases. In other words, it suggests low recombination loss in the WS-SiNP [28,29].

3.2 Enhanced NIR spectral response

In order to quantitatively characterize the NIR photodetection performance, we extracted spectral R_λ and EQE in the wavelength range from 750 to 1,000 nm. These two parameters are key metrics for PDs. R_λ represents the ratio of incident light power density to the I_ph and can be described as:

 

Fig. 3. (a) Comparison of spectral responsivity (R_λ) of WS-SiNP, S-SiNP, Planar, Commerical InGaAs, and Commerical Si PDs with standard deviations. (b) External quantum efficiency (EQE) and reflectance as a function of wavelength for fabricated devices.

Download Full Size | PDF

The superior NIR detection of the WS-SiNP devices also stand out in Fig. 3(b). EQE is described as the percentage of the number of carriers contributing to photocurrent per number of incident photons. The EQE can be expressed as:

3.3 Light confinement in the WS-SiNPs

Figure 4(a), (b), and (c) depict the |E|² distribution in the cross-sectional view of the WS-SiNP and the S-SiNP at the wavelengths of 800 nm, 900 nm, and 1,000 nm, respectively. To clearly compare the structural differences in optical properties, the |E|² distribution is expressed as the ratio of the maximum |E|² values observed in the WS-SiNP (|E_W|²) to the maximum values observed in the S-SiNP (|E_S|²) at each wavelength. A higher |E|² indicates higher light intensity, which means more absorption in a uniform medium [31]. In the x-z cross-section, the WS-SiNP clearly shows higher magnitude of |E|² than the S-SiNP at all wavelengths because the intensity of |E_W²|/|E_S²| in the WS- SiNP is greater than 1. Notably, WS-SiNP shows strong |E|² along the edges of the NP, whereas S-SiNP shows strong |E|² along the center of the NP. This suggests that in the WS-SiNP, photons travel primarily along the edges of the structure with an increased horizontal component due to WS sidewall boundaries, while in the S-SiNP structure, photons travel primarily along the center of the structure with a minimal horizontal component due to the straight sidewall boundaries. Therefore, the WS-SiNP structure induces enhanced horizontal component of vertically incident light compared to the S-SiNP structure due to the WS sidewall boundaries [32]. This enhanced horizontal light component causes the vertically incident light to stay longer in the structure, increasing the effective light absorption path without increasing its thickness. This improves the light absorption efficiency at NIR wavelength where the Si absorption coefficient is low [33]. Also, to investigate the resonance behavior, we analyzed the |E|² distribution in the x-y plane extracted from the z plane position of maximum field intensity in each NPs. It is noteworthy that WGM resonances of closely spaced modes are shown in the WS-SiNP, whereas Fabry-Pérot resonances are observed in the S-SiNP. The WGM resonance has been reported to confine the photons horizontally from curved surface boundaries by repeated internal reflection, which also leads to a significant increase in effective absorption path [34]. Therefore, the |E|² of the WS-SiNP is higher than that of the S-SiNP due to induced WGM resonances by the WS sidewall structure. These results show good agreement with the experimental data in Fig. 3(a) and (b). For detailed analysis of the light behavior, we checked the real part of the Poynting vector, which represents energy density flux in the electromagnetic field as shown in Fig. 4(d). The energy density and direction of the Poynting vector were extracted from the x-z plane, which is the dominant resonant modes in Fig. 4(b). The energy in WS-SiNP mainly accumulates near the edges by the WS sidewall structure and the Poynting vector has a rotational direction from the center of the WS-SiNP, which increases the horizontal component of the light. On the other hand, the energy in S-SiNP is mostly accumulated in the center and decreases rapidly during the Fabry–Pérot resonance propagation process. Therefore, these results provide evidence that stronger horizontal component of vertically incident NIR light is induced in the WS-SiNP than in the S-SiNP structure.

 

Fig. 4. Calculated |E|² intensity for the WS-SiNP and S-SiNP at wavelengths of (a) 800 nm, (b) 900 nm, and (c) 1,000 nm. Left insets are x-z cross-sections, with the location of dominant modes indicated by white dotted lines. Right insets are x-y cross-sections extracted from the white dotted lines. The color scales are normalized by the maximum |E|² observed in WS-SiNPs (|E_W|²) over the maximum |E|² observed in S-SiNPs (|E_S|²) for each wavelength. (d) Calculated Poynting vector distribution, extracted from the red (WS-SiNP) and green (S-SiNP) dashed line in Fig. 4(b). The colored arrows indicate the relative magnitude and direction of the normalized Poynting vector, and the colored background indicates energy density. (e) Schematic representation of the light trapping process in the WS-SiNP with WGM resonances. (f) Normalized E-field of the WS-SiNP as a function of spectral distribution and the z-axis position of the SiNP.

Download Full Size | PDF

The WS-SiNP structure also has the advantage over other WGM resonant structures in terms of reduced surface recombination [35]. This is because light absorption in WS-SiNP occurs slightly inward from the surface compared to inversely tapered or hexagonal NPs [14–16]. Other intriguing features are that the longer the wavelength, the lower the z-axis position of the WGM resonance peaks. Also, as the wavelength increases, the light resonates in the form of leaky modes, where the field radiates outward. We note that three physical phenomena are taking place here, as simply illuminated in Fig. 4(e): (1) Since light with a longer wavelength diffracts at a greater angle than light with a shorter wavelength, longer wavelengths are better able to travel to the bottom of the NPs. (2) Longer resonant wavelengths require longer light paths that can be accommodated in a larger cavity [36]. Accordingly, the lower z-positions result in strong superposition at longer wavelengths in Fig. 4(a)-(c). This is because the average diameter of the WS-SiNP structure increases gradually from the top end toward the bottom end (please see Fig. 1(e)). Figure 4(f) shows the normalized electric field of the WS-SiNP as a function of spectral distribution and the z-axis position. Clearly, it can be confirmed that the absorbed peak wavelength is redshifted toward the lower end of the WS-SiNP. (3) Since the effective refractive index decreases with a longer wavelength, the WGM resonance is in the form of a leaky mode at longer wavelengths when the light circulates horizontally through the WS sidewall structure [21]. This leaked light is re-absorbed by adjacent SiNPs, increasing NIR absorption [37,38]. To quantitatively investigate the strength of the resonance, we calculated the quality factor (Q). The Q-factors of WGM resonances are in the range of 50-231. This represents low-Q, which is suitable for solar cells or PDs because it allows for efficient coupling between incident light and resonant modes. As shown in Fig. 4(f), the low-Q WGM device can enhance spectral absorption at broad NIR wavelengths rather than just enhancing a specific resonant wavelength [12]. Therefore, the WS sidewall structure of SiNPs can be used for various NIR wavelength applications.

4. Conclusion

We have demonstrated a silicon-based NIR PD with WS-SiNPs. WS-SiNP PD showed an average R_λ improvement of 30% and 11% for each wavelength in the NIR spectra compared to the planar and S-SiNP devices, respectively. Also, EQE and reflectivity are highly enhanced by adopting the WS-SiNP structures. The maximum R_λ of 0.648 A/W was achieved at 905 nm, which is widely utilized in the LiDAR system, and this value is 32% higher than a commercial Si PD (S1337-33BQ, Hamamatsu). Also, our device shows R_λ as high as a commercial InGaAs PD (Standard type InGaAs, Hamamatsu) near the 1,000 nm wavelength. These excellent optical characteristics were verified by a 3D FDTD numerical simulator. The simulation results confirmed that the WS sidewall boundaries of the SiNP can not only enhance the horizontal components of the vertically incident light but also induce the low-Q WGM resonance, which can increase the light absorption path at various NIR wavelengths. Therefore, we suggest that the WS-SiNP structure could improve the NIR photoresponse of Si PDs. This finding not only provides additional perspectives into 3D nanostructures but also paves the way for Si-based optoelectronic devices to be more versatile for future technological applications.