1. Introduction

Near-infrared (NIR) sensing technology has been extensively applied in biological inspection, Time-of-Flight (ToF), surveillance, and fiber optic communication [1–4]. Wavelengths of 850 nm and 940 nm are mostly used because of their light source availability [5]. In particular, a wavelength of 940 nm is completely invisible to the human eye and is less affected by sunlight, so it is mainly selected for in-vehicle sensing applications. However, in such a wavelength, silicon-based image sensors have insufficient sensitivity because of the low silicon absorption. Although a thick silicon absorption layer improves NIR sensitivity, this approach degrades image recognition due to the crosstalk between pixels and is technically difficult to be implemented in deep trench isolation (DTI) [6]. Therefore, improving the NIR light sensitivity without a thick silicon layer is a significant issue. In recent years, to enhance the NIR sensitivity of silicon sensors, the thickness of silicon sensor has been gradually increased owing to the DTI technology improvement [7]. Another approach is to scatter the incident surface, such as a multiple-pyramid-shape surface, to increase the light propagation length and effective silicon thickness [8].

Applying plasmonic enhancement also has the potential to improve the NIR sensitivity of sensors. Plasmonics has been applied in the photonics research field, ranging from ultraviolet to infrared, such as fluorescence enhancement, surface enhanced Raman scattering, solar cells, color filters, and thermal emission control [9–15]. In sensor applications, the development of silicon metal Schottky junction-type photodetectors using plasmon hot carriers and enhanced electric fields by surface plasmon resonance have attracted attention in the detection of photon energies below silicon bandgap of 1.12 eV [16–22]. Thus, most of the approaches in plasmonics to photonics applications have included the utilization of strong electric field enhancement, which involves resonant coupling between photons and electrons in metal. In this study, we apply the quasi-resonant conditions of surface plasmons to improve the silicon absorption efficiency.

2. Simulation results and discussion

2.1 NIR sensitivity enhancement by metal gratings and metal DTI

We propose a plasmonic image sensor constructed by the metal gratings of surface plasmon diffraction and highly reflective metal DTI of the reflection of the diffracted light to improve the NIR absorption of silicon (Fig. 1). An enhanced electric field was generated near the metal surface due to the surface plasmon resonance. Under rigidly resonant conditions, the photon energy couples with electronic vibration and is converted to propagate on the metal surface, so the wave vector of the coupled light is orthogonal. Conversely, by slightly shifting the diffraction condition from the orthogonal condition, the light propagates to the silicon side. The diffracted light under conditions close to 90° is reflected by the DTI formed along the pixel boundary. Repetition reflection increases the effective propagation length of the silicon. A large diffraction angle and high diffraction efficiency increase sensitivity. Moreover, because it is close to the plasmon resonance, the reflectance from the sensor surface is suppressed, and the light utilization efficiency is improved. In this study, we investigated the conditions that maximize the absorption efficiency in silicon from the viewpoint of diffraction angle and diffraction efficiency for NIR at a wavelength of 940 nm.

 

Fig. 1. Schematic diagram of the plasmonic image sensor with metal gratings and highly reflective metal DTI. Surface plasmon-coupled photons are diffracted by periodic gratings, and the diffracted light is reflected by DTI.

Download Full Size | PDF

A silver grating was applied for the plasmonic diffraction (Fig. 2(a)). The period determines the diffraction angle. The synchronization mode of surface plasmon resonance was tuned to a wavelength of 940 nm by changing the metal line width with respect to the period. The metal height mainly contributed to the diffraction efficiency. The silver diffraction grating was covered with an SiO₂ layer for protection. A natural oxide film with a thickness of 2 nm was remained on the silicon surface. Silver was buried in the DTI. The pixel size was set to 6.5 µm. Figure 2(b) shows the setting monitors to simulate the ratio of the light consumption to silicon absorption, silver absorption, reflectance, backward scattering, trench absorption or forward scattering, and transmittance. Transmittance represents the light intensity ratio of the 0th-order light monitored at a silicon thickness of 3 µm. Reflection represents the ratio of the reflection to the normal angle. The backward scattering and the forward scattering were monitored on sides via the silver diffraction grating, respectively. In Fig. 2, silver trench was set in forward side. Therefore, we set a trench absorption monitor instead of the forward scattering monitor. The structural parameters of the silver diffraction gratings were optimized for the maximum silicon absorption at a wavelength of 940 nm in the region from the Si surface to a depth of 3.0 µm. A finite-difference time-domain simulation and rigorous coupled wave analysis were performed using commercially available software FullWAVE and DiffractMOD by RSOFT. The optimized period, width, and height of the silver grating were 265 nm, 230 nm, and 85 nm, respectively. The covered SiO₂ layer thickness was 205 nm. The propagation wavelength λ_Si of light with a wavelength of 940 nm in silicon was λ_Si =λ₀/n_Si ≃261 nm, which is slightly shorter than the period. The diffraction angle for a period of 265 nm was 80.6°.

 

Fig. 2. Simulation results for the plasmonic silicon image sensor constructed by silver gratings combined with silver DTI. (a) Simulation model, (b) monitors for light consumption ratio, (c) electric-field intensity distribution, (d) enlarged area of absorption distribution, (e) light consumption ratio, (f) integrated absorptions of silicon in thickness with and without plasmonic diffraction.

Download Full Size | PDF

By reflecting the diffracted light in the trench, the effective propagation length in silicon increased, and the absorption efficiency was improved. When the diffraction angle was 80.6°, the effective propagation distance was six times compared with a sensor without silver grating. Figure 2(c) shows the electric field intensity distribution of the x-z cross section when the optimized silver diffraction grating is irradiated with NIR at a wavelength of 940 nm. Surface plasmon resonance was excited, and an enhanced electric field was formed near the surface of the silver diffraction grating. Interference fringes were observed inside the silicon. This standing wave indicates that the light diffracted by silver grating interfered with the light reflected from the trench. The magnified image of the absorption distribution shows that light is well absorbed at the position where the standing wave is strengthened in silicon (Fig. 2(d)). Figure 2(e) shows the distribution of the ratio of the light consumption. Each value was simulated in a silicon thickness of 3.0 µm. The absorption ratio in silicon for a thickness of 3.0 µm was enhanced to 19.2%, which was a 5.3-fold improvement over the 3.6% absorption rate of bare silicon. The fitting curve of the dashed line in Fig. 2(f) was analyzed based on the transmitted light ratio and silicon absorption ratio using the absorption simulation result of the solid line. When plasmon diffraction was applied, an absorption of 44% was achieved at a silicon thickness of 10 µm. In the case of plasmon diffraction, the silicon absorption for a wavelength of 940 nm saturated at approximately 70%. The estimated quantum efficiency is almost same as integrated absorption because the absorption in dead layer less than around 200 nm is a few percent.

2.2 Plasmonic diffraction under the quasi-resonant condition

To consider the absorption enhancement principle using plasmonic diffraction and DTI reflection, the electric field intensity distribution of the x-z cross section without DTI is shown in Fig. 3(a). The enhanced electric field on the metal surface by surface plasmon excitation contributed to the diffraction on the silicon side. The vertical stripe intensity field distribution was due to the interference between the 0th-order and the ±1st-order light. The diffracted light propagated at an angle of approximately 80°. This diffracted light, which is a factor for a standing wave in silicon, was reflected by the metal DTI as resulted in Fig. 2(c). The simulation results of the ratio distribution of light (Fig. 3(b)) show that the forward scattering was 49.7%, which was considered to be the diffraction component and indicated a high diffraction efficiency. Figure 3(c) shows the dependence of the reflectance spectrum on the incident angle. The reflectance dip-curve dispersion represents the surface plasmon resonance of the silver grating. The broken line indicates the 90° diffraction condition of the 1st-order diffracted light. The matching condition between the surface plasmon dispersion relation and the 90° diffraction was an incident angle of 0°. The matching wavelength was 952 nm for a grating period of 265 nm. The incident wavelength of 940 nm was slightly shorter than the matching wavelength, and it diffracted at 80.6°. This quasi-resonant mechanism achieves a larger diffraction angle and diffraction efficiency than a general diffraction grating. Thus, the principle of silicon absorption enhancement with respect to NIR is that while exciting surface plasmon resonance, light is diffracted toward the silicon side with a large diffraction angle so that it efficiently propagates in the silicon in the lateral direction.

 

Fig. 3. Simulation results of the plasmonic diffraction by silver grating on silicon without DTI. (a) Electric-field intensity distribution, (b) light consumption ratio, (c) incident angle dependence of the reflection spectra.

Download Full Size | PDF

2.3 DTI materials for NIR reflection

In our proposed construction, DTI in silicon pixels plays an important role in the absorption enhancement of NIR. In recent image sensors, DTI has been used as a pixel isolation method. Polycrystalline silicon, SiO₂, and tungsten (W) are mainly used inside the DTI together with the negatively charged layer. The selection of this material is important for NIR image sensors. Figure 4 shows the reflectance dependence on the incident angle at a wavelength of 940 nm for various trench materials, such as SiO₂, silver (Ag), aluminum (Al), copper (Cu), and tungsten. Each complex refractive index at a wavelength of 940 nm was 1.47 for SiO₂ [23], 0.08 + i6.78 for Ag [24], 1.73 + i8.60 for Al [25], 0.24 + i6.02 for Cu [25], and 3.14 + i3.16 for W [25]. The solid line shows the characteristics when the trench layer thickness is 50 nm, and the broken line shows for the infinite layer thickness. The most useful trench material, SiO₂, exhibited 100% reflection under conditions of sufficient thickness and total internal reflection at an incident angle of 25° or larger. However, the reflectance drastically dropped with thin DTI thickness owing to the evanescent transmission. Thus, it was unsuitable for sensor pixel miniaturization. In particular, in the case of plasmon diffraction at 80°, the trench incident angle was 10°, and its reflectance was only 20% or less even under infinite thickness conditions. Therefore, SiO₂ is not applicable for increasing the effective propagation length using a system of plasmonic diffraction and trench reflection. Metals showed relatively high reflectance in most incident angle ranges, but W, which was often used as a trench material, had low reflectance in the NIR wavelength range. Ag, Al, and Cu are recommended as trench materials for NIR image sensors. Among them, Ag had the highest reflectance, and it exhibited an angular average reflectance of 96% even at a DTI thickness of 50 nm.

 

Fig. 4. Reflectance spectra dependence on the incident angle of DTI metals of SiO₂, Ag, Al, Cu, and W at an NIR wavelength of 940 nm.

Download Full Size | PDF

2.4 Constructive interference in the sensor pixel

Figure 5 shows the simulation results of the silicon absorption dependence on the inter-DTI distance. The absorption shows a sinusoidal curve owing to the interference of the DTI reflections. The pitch of the maximum absorption value was 268 nm, which was in good agreement with the grating period of 265 nm and interference wavelength of 265 nm by the DTI reflection for NIR at 940 nm that propagates in silicon at a diffraction angle of 80.6°. Therefore, the structural parameters are optimized under constructive interference. This finding indicates that the silver diffraction grating was optimized for a pixel size of 6.5 µm in the simulation of its structural parameters (Fig. 2). To apply this system to other pixel sizes, structural parameters should be optimized owing to the different conditions for constructive interference.

 

Fig. 5. Simulation results of silicon absorption dependence on the inter-DTI distance. Solid line represents sinusoidal curve fitting.

Download Full Size | PDF

3. Conclusions

We proposed plasmonic diffraction combined with highly reflective DTI to enhance NIR sensitivity. The metal grating period slightly detuned from the surface plasmon resonance resulted in a state with a large diffraction angle and high diffraction efficiency, which led to a significant improvement in NIR sensitivity. In addition, the detuning reduced the metal absorption loss. Silver grating with a 265 nm period, 230 nm width, and 85 nm height and silver DTI with layer thickness of 50 nm improved the sensitivity to 5.3 times at a wavelength of 940 nm for 3-µm-thick silicon image sensors compared to those without silver grating. The development of high-sensitivity NIR image sensors using plasmonic diffraction combined with highly reflective DTI could achieve small thickness sensor and small DTI thickness. Therefore, NIR CMOS image sensor with small pixel size could be realized, which might be applied to various applications, such as ToF, security, biometrics, biological inspection, and automobile.