Enhanced Near‑Infrared Absorption via Two‑Step Fabrication of Structured Black Silicon and Its Device Applications

Background

Micro‑ and nanostructured black silicon has become a focal point for enhancing light absorption across the visible and NIR spectra. DRIE, a cyclic etch‑passivation technique employing SF₆ for isotropic etching and C₄F₈ for sidewall passivation, facilitates the creation of high‑aspect‑ratio structures whose geometry is governed by mask dimensions, gas flows, electrode power, and cycle duration. Subsequent ion implantation dopes the etched silicon, introducing sulfur atoms that generate sub‑band‑gap states and broaden the absorption spectrum [1–11]. These engineered surfaces have found applications in sensitive photodetectors, photovoltaic cells, biochemical sensors, displays, and optical communications [12–20]. In our earlier study, we investigated a Si‑PIN detector with front‑surface BSi, noting a trade‑off between NIR sensitivity and visible‑light responsivity due to carrier collection challenges. This prompted the current investigation of BSi on the back surface to simultaneously enhance NIR and maintain acceptable visible performance.

The present article reports on the light‑absorptance enhancement and band‑gap narrowing achieved by a DRIE–PIII two‑step process. We analyze how different etching parameters affect absorptance from 400 to 2000 nm and evaluate a Si‑PIN detector incorporating back‑surface BSi for responsivity across 400–1100 nm.

Methods

Figure 1a illustrates the uniform, periodic cylindrical arrays selected for optical simulations using FDTD software. Figure 1b shows how absorptance varies with four cylinder‑array models that share a diameter (D = 4 µm) but differ in center spacing (T₁ = 12 µm, T₂ = 10 µm, T₃ = 8 µm, T₄ = 6 µm) after optimization.

Optical simulation model (a) and simulation results (b) of microstructured silicon

Based on the simulation, three photolithography masks were designed: mask I (D = 4 µm, T = 6 µm), mask II (D = 4 µm, T = 8 µm), and mask III (D = 4 µm, T = 10 µm). A 15 × 15 cm² n‑type silicon wafer (500 µm thick, resistivity 2500–3000 Ω cm) received the NR9‑1500PY photoresist and was patterned with the circular arrays. The wafer was etched in a DRIE chamber (ICP‑100D) using SF₆ (etch) and C₄F₈ (passivation) cycles for 30, 70, or 100 repetitions. Post‑etch, residual photoresist was removed under oxygen plasma to yield clean surfaces. To further enhance NIR absorptance, mask III samples were subjected to PIII with a sulfur dose of 1.0 × 10¹⁵ cm⁻² at 800 eV.

The schematic diagram of mask size

Morphology was examined with a JSM‑7500F field‑emission SEM. Absorptance was measured at room temperature using an NIR2500 fiber‑optic spectrometer coupled to an Idea Optics integrating sphere (IS‑20‑5). Detector responsivity was determined with an OPHIR Vega optical power meter, a Scitec Instruments optical chopper (Model‑300CD), and a Keithley 2636B under dark conditions. All measurements were calibrated and repeated on 4–6 samples to ensure reliability.

Results and Discussion

Figure 3 presents SEM images of the aligned microstructured arrays for the three mask sizes. The top view deviates from perfect circles due to the interplay between DRIE etching and mask fidelity. The etching depth increases with cycle count, reaching ~1.87 µm (30 cycles), 2.35 µm (70 cycles), and 3.15 µm (100 cycles). Lateral etching, despite sidewall passivation, leads to slightly tapered pillars, confirming the need to fine‑tune lithography and cycle parameters for optimal morphology.

Typical microstructured silicon arrays made by DRIE for different mask sizes. a Mask I. b Mask II. c Mask III

Sectional views (Figure 4) confirm the progressive increase in pillar height with cycle number. Although DRIE’s passivation layer protects sidewalls, some lateral etching remains, influencing the final shape.

Sectional views of mask III samples fabricated for different cycle times. (a) 30, (b) 70, and (c) 100

Figure 5a shows absorptance of the microstructured arrays without PIII. Compared to bare C‑Si, the etched arrays enhance absorptance across 400–2000 nm, with the 100‑cycle sample achieving up to 70 % in the NIR. Multiple internal reflections within the pillars increase the optical path length, boosting absorption. However, absorptance above 1000 nm remains modest, motivating sulfur doping. After PIII, Figure 5b reveals a pronounced absorptance increase across the entire spectrum. The 100‑cycle sample’s peak absorptance rises to 83 %, and the average absorptance reaches 62 %. Sulfur implantation introduces impurity bands that overlap, narrowing the effective band‑gap and enabling sub‑band‑gap photon absorption [10,11]. This is evident from the band‑gap reduction from 1.12 eV (C‑Si) to 1.045 eV, 1.033 eV, and 1.025 eV for 30, 70, and 100 cycles, respectively (Figure 7). The analysis employed Kubelka–Munk theory and Tauc plots to extract these values. Figure 6 illustrates the multi‑reflection mechanism within the pillar arrays, explaining the enhanced absorptance.

Absorptance of C‑Si and black silicon fabricated by different cycle times before (a) and after PIII (b) and comparison of 100‑cycle samples (c)

The transmission path of incident light on the surface of microstructured silicon

Band gaps of C‑Si (a) and black silicon made by different cycle times: (b) 30, (c) 70, (d) 100

Integrating the back‑surface BSi into a Si‑PIN photodiode (Figure 8a) involved: (1) oxidizing both sides of a monocrystalline n‑type wafer; (2) diffusing boron to form the P‑layer on the front; (3) depositing Si₃N₄ and polishing the back to ~200 µm; (4) doping an N⁺ layer and growing the BSi pillars atop it; and (5) patterning metal contacts. Device 2, featuring a 2 mm‑diameter photosensitive area, exhibits a dark current comparable to commercial Si‑PIN photodiodes. Under 1060 nm illumination, its responsivity reaches 0.53 A W⁻¹, a marked improvement over the commercial S1336‑44BK (device 1). Notably, device 2’s responsivity in the 680–1100 nm range surpasses device 1 by ~60 nm, confirming the efficacy of back‑surface BSi in extending NIR sensitivity.

Detector image (a), dark current (b), I–V curve under 1060 nm wavelength illumination (c), and responsivities of two different detectors (d): device 1 from ref. [22] and device 2 based on the results of present paper. The inset of d shows the device structure

Device 2’s limited improvement in visible‑light responsivity stems from the fact that photons above the C‑Si band‑gap are absorbed in the front P‑layer, making the back‑surface BSi largely irrelevant for carrier collection. In contrast, NIR photons penetrate the front layer, are absorbed in the N layer, and the resulting carriers are efficiently collected by the N⁺ contact under reverse bias, leading to the substantial responsivity gain.

While the present study demonstrates a viable route to NIR‑enhanced photodiodes, further optimization of DRIE parameters, ion implantation energies, and device architectures could yield even higher performance and broader applicability.

Conclusions

We have successfully fabricated microstructured black silicon via a DRIE–PIII two‑step process. Three pillar geometries (mask I: D = 4 µm, T = 6 µm; mask II: D = 4 µm, T = 8 µm; mask III: D = 4 µm, T = 10 µm) yielded heights of 1.87, 2.35, and 3.15 µm, respectively. These structures achieved peak absorptance of 83 % and an average of 62 % across 400–2000 nm, attributed to increased optical path length and a narrowed band‑gap from sulfur doping. Incorporating the BSi as the back surface of a Si‑PIN photodiode produced a responsivity of 0.53 A W⁻¹ at 1060 nm and 0.31 A W⁻¹ at 1100 nm, markedly outperforming a commercial S1336‑44BK detector in the NIR. This approach offers a promising pathway for high‑sensitivity NIR photodetectors.