1. Introduction

In silicon-photovoltaic (PV) devices, the realization of thin-film (≤50 µm) crystalline silicon (c-Si) solar cells have been significantly investigated in attempts to reduce material costs. In previous works employing conventional pyramidal texturing, J. H. Petermann et al. reported ~19%-efficiency from the 43 µm-thick Si substrates [1]. L. Wang et al. have also presented a 16.8%-efficiency using 18 µm-thin Si substrates [2]. However, also note that these conventional texturing normally removes at least several microns of silicon during anisotropic etching [3]. In our work, metal-assisted chemical etching (MaCE) effectively forms surface nanostructures without notable consumption of silicon due to selective etching behavior underneath metal nanoparticles. To compensate the reduction of light-absorber, nanostructured (NS) Si arrays featuring superior light-trapping are adopted to increase the absorbance of incident photons [4–6]. The optimum correlation length (L_c) for black silicon surface is ~100 nm which is a compromise between ultraviolet-visible (UV-VIS) anti-reflection properties and near-infrared (NIR) optical path enhancement [7]. Nevertheless, practical conversion efficiencies in NS-Si solar cells have been much lower than their theoretically estimated values [8]. The increased recombination of photogenerated carriers at the nanostructured frontside causes a degradation in cell efficiencies [9]. Additional wet-etching with chemical solutions can effectively reduce the surface area enlargement caused by nanostructuring [10–13]; however, blue responses of the quantum efficiency (QE) at short wavelengths deteriorate the photon-to-electron conversion efficiency. As an alternative, researchers have suggested to form a solar cells which have a nanocone structure at the backside for NIR optical path enhancement with a period in the range of 1000 nm and anti-reflection coatings (ARCs) on the planar frontside. Based upon this configuration, light absorption could exceed 90% in the visible range (from 500 to 900 nm) using a c-Si layer thinner than 50 μm [14].

We extend this idea to create planar black silicon structures that use classical SiNx-antireflections coating on the frontside to optimize the VIS absorption without deteriorating the electronic properties by having a nanostructured surface; concomitantly, use an integrated grating structure at the backside of the solar cell for the optical path enhancement which needs to compensate insufficient NIR absorption due to the 50 μm thickness of the solar cell. Currently, tunnel-oxide passivated contacts (TOPC) are typically adopted for backside contact [15–17]; however, TOPC approaches that utilize plasma-enhanced chemical vapor deposition (PECVD) are not useful for the nanostructured backside we prefer because amorphous Si (by PECVD) is unable to conformally coat the nanostructured morphology.

Herein, we suggest, using an atomic layer-deposited (ALD), ultrathin (~4 nm) ZnO film as the tunnel oxide for the nanostructured backside solar cells that employ a ~50-μm-thin c-Si absorber. In general, direct metal contact onto the NS Si seriously hinders the flow of charge carriers because of the large amount of unpassivated Si dangling bonds. However, our new TOPC approach increases the photon-to-electron conversion efficiencies by ~24%, in which the ZnO film successfully passivates the defect states and improves hole flows via surface band bending. This might be a first technological step towards a planar black silicon technology.

2. Experimental method

To produce 50 μm-thick substrate, Czochralski-grown (CZ) p-type (100) Si wafers (Resistivity of 9.4 ± 0.7 Ω cm) were etched using potassium hydroxide solution (46%) at 80 °C for 90 min. Those thin c-Si substrates were prepared after standard RCA (Radio Corporation of America) cleaning following piranha (H₂SO₄:H₂O₂) and HF (HF:H₂O) solutions. In order to form back-side (B) NS Si arrays, silver nanoparticles (NPs) were precipitated uniformly for 10 s using galvanic displacement reaction employing an aqueous solution of HF (4.8 M) and AgNO₃ (0.005 M) while the front-side (F) planar Si was insulated with Microstop. Then, the NS Si was allowed to form at room temperature using a mixed solution of HF (4.8 M) and H₂O₂ (0.44 M). The diameter of the NS Si was between 40 and 60 nm and the lengths of nanostructures were adjusted by varying the etching time. Residual Ag NPs were then removed using concentrated nitric acid (30 wt%) for 30 min, and the NS Si arrays were finally rinsed with de-ionized water. To fabricate B-NS Si solar cells, the emitter layer and the back-surface field (BSF) were formed by phosphorus and boron diffusion via a spin-on-dopant (SOD) method. Phosphorous and boron silicate precursors (P509 and B155, Filmtronics) were spun onto wafers, then n⁺-emitter and p⁺-back surface field (BSF) formed using the mixture of N₂ and O₂ via thermal diffusion of gaseous phosphorous at 860°C for 5 min. The phosphorus and boron glasses that remained after SOD diffusion were removed using a diluted HF solution. In the B-NS Si arrays, thin ZnO layers produced via atomic layer deposition (ALD) process using (C₂H₅)₂Zn as a metal precursor and H₂O as an oxidant. Finally, the front contact (Ag) on n⁺-emitter layer and back contact (Al) electrodes on p⁺-BSF were formed by electron-beam evaporation. The 90 nm-SiN_x via plasma-enhanced chemical vapor deposition (PECVD) using SiH₄ and NH₃ as reactants at 250 °C was used for antireflection coating on F-planar Si. Finally, 50 μm-thin B-NS Si solar cell was prepared as shown in optical and cross-sectional SEM images of Fig. 1(b).

 

Fig. 1 (a) Flow diagram and (b) cross-sectional SEM image for 50 μm-thin B-NS Si solar cells. The inset of Fig. 1(b) is an optical image showing the active area of 0.9 cm²

Download Full Size | PDF

The nanostructured Si surface was evaluated by observation at variations of the effective carrier lifetime (τ_eff) measured by quasi-steady-state photoconductance (QSSPC, WCT-120). Using an IR-pass filter (RG 850, passing wavelengths longer than 850 nm), a light source induces photogenerated carriers in the Si substrates. For the planar Si cleaned by HF solutions, average τ_eff of 104.2 μs has been extracted over ten wafers.

The interface structure of ZnO/nanostructured Si was characterized using field-emission transmission electron microscope (FE-TEM, JEOL JEM-2100F) equipped with a 200-kV field-emission gun. The electron energy-loss spectroscopy (EELS) in a high-resolution TEM was utilized with a high-energy resolution of 0.5 eV. The structure of silicon oxide in the ZnO/NS Si interface was characterized by Fourier transform infrared (FT-IR) spectroscopy. Optical reflection/transmission measurements were performed between the wavelengths of 400 and 1000 nm using a UV−Vis/NIR spectrophotometer (Lambda 750, Perkin Elmer) equipped with a 60 mm integrating sphere (Labsphere) to account for total light (diffuse and specular) reflected from the samples.

3. Results and discussion

As we have stated, minimizing the nanostructured surface area is helpful for obtaining favorable electrical performance; moreover, we have found that the short heights of the NS Si arrays (~500 nm) are sufficient to boost light absorption above 90% in the wavelengths from 400 to 800 nm [Black curve in Fig. 2(a)]. Compared to the nanostructured frontside, backside nanostructuring significantly increases light absorption at the NIR region (≥850 nm) due to the optical path enhancement. Assuming the perfect carrier collection of 100%, a short-circuit current density (J_SC) can be calculated using the following equation [18]:

 

Fig. 2 (a) Comparison of the spectral absorption and external quantum efficiency (EQE) as a function of the wavelength (from 400 to 1100 nm). Inset shows frontside and backside NS Si solar cells with a nanostructure height of 500 nm. (b) Internal quantum efficiency (IQE).

Download Full Size | PDF

Although the nanostructured frontside showed reasonable optical absorption in short wavelengths, the external quantum efficiency (EQE) response was observed to greatly degrade in frontside NS solar cells [Black dots in Fig. 2(a)]. Chen et al. previously reported a similar result using their Si nanowire solar cells [19]. Two major factors are Auger and surface recombinations. Auger recombination seriously occurs in the frontside nanostructures in contact to the emitter region degenerately doped; concomitantly, surface recombination is also increased by the surface area enlarged by nanostructuring. The spectral dependence of internal quantum efficiency (IQE) was shown in Fig. 2(b) to compare the charge carrier collection efficiency between F- and B-NS solar cells. In short wavelengths from 400 to 600 nm, ~10% decrease in IQE values was observed in F-NS solar cells in comparison to B-NS, indicating the increase in carrier recombination for high-energy photons. Note that the nanostructured surface likely forms a more heavily doped and thicker emitter compared to a planar morphology under the same doping conditions, because doping elements readily diffuse into silicon using a three-dimensional surface topology.

To specifically analyze the effect of the NS surface morphology on the recombination of photogenerated carriers, the degradation behavior of the minority carrier lifetime, caused by nanostructuring, was evaluated as a function of the minority carrier density. According to the parametrization by Richer et al., the intrinsic lifetime (τ_intr) including radiative and Auger recombinations is calculated without using surface recombination [20]. In Fig. 3(a), the effective carrier lifetime (τ_eff) affected by surface nanostructuring is plotted with τ_in. At the typical carrier concentration (~1 × 10¹⁵ cm⁻³) for cell operation, τ_eff of 388.3 μs (for planar Si) is observed to seriously decrease to 68.1 μs (for NS Si) due to a high defect density induced by unpassivated Si dangling bonds.

 

Fig. 3 (a) Lifetimes of minority carriers as a function of the minority carrier density (cm⁻³). Experimental (for planar and NS Si) and calculated (for intrinsic) data. (b) Calculated and measured short circuit current (J_SC) of frontside and backside NS Si solar cells with and without surface passivation.

Download Full Size | PDF

In contrast, backside NS solar cells can effectively minimize Auger and surface recombinations because an antireflective SiN_x coating is applied on the planar frontside and NS Si arrays are formed on the backside [see a red box of Fig. 2(a)]. Compared to the frontside NS solar cells, the EQE for the backside NS cells [Red dots in Fig. 2(b)] is increased by improving the green and red responses to sunlight. The theoretical efficiency of c-Si solar cells can be achieved if the surface recombination velocity (SRV) of the backside decreases to ≤100 cm/s (from a current value of 386 cm/s) [21].

The measured values of J_SC were smaller in both of front and backsides than their calculated values [Fig. 3(b)]. In backside NS cells, the measured J_SC value (30.8 mA/cm²) with ZnO-passivated contacts was found to be decreased in their calculated J_SC (37.73 mA/cm²). However, our group reported that the J_SC value (22.8 mA/cm²) measured for the nanostructured frontside with ALD Al₂O₃ passivation is much smaller than their calculated value (36.71 mA/cm²) [22]. This result implies that the backside NS cells utilizing thin-film c-Si could attain higher light-absorption as well as reduction in electrical loss in comparison to their frontside counterparts.

The p⁺-back-surface field (p⁺-BSF) structure, which requires the degenerate boron-doping of the backside, is critically necessary for decreasing surface recombination at the backside of p-Si solar cells [Fig. 4(a)]. Electrons (minority carriers) in p-Si likely moves to the p⁺-region because of the stronger field intensity caused by excessive holes; this creates a built-in potential at the p⁺/p interface (referred to as the p⁺-BSF). Under illumination, this potential gradient effectively delivers photogenerated holes to the Al contact due to reduced recombination with minority electrons as well as to the increased hole conductivity caused by the p⁺-layer.

 

Fig. 4 (a) Energy diagram in the Al (1 μm)/p⁺-Si (0.45-μm-thick)/p-Si (50-μm-thick) structure; orange and blue arrows denote electron flow and hole flow, respectively. The red arrow indicates holes trapped in surface states, denoted by a red rectangle. (b) Secondary ion mass spectrometry depth profile for boron annealed at 860 °C. (c) Analysis of the band diagram for the Al/ZnO/p⁺-Si structure. This energy diagram was estimated based on the following assumptions: (1) work function of Al is 4.2 eV, (2) bandgap of amorphous ZnO is 3.9 eV [for calculated optical bandgaps, see Fig. 5(a)], and (3) boron doping concentration at the p⁺-Si surface is ~1 × 10²⁰ cm⁻³ [Fig. 4(b)]. The inset shows the hole flow with the downward band bending of silicon. (d) Variation of surface potential as a function of dielectric contants (k) of ZnO. In the inset, hole carrier density (cm⁻³) decreases as D value increases from 0.5 to 1 nm; here, D deonotes a nanoscale distance from the ZnO/p⁺-Si interface into p⁺-region.

Download Full Size | PDF

Backside nanostructuring, however, generates surface states (E_ss) within the bandgap of silicon, which seriously prevent a hole flow [23–25]. To form passivated contacts on the NS Si surface, ALD−ZnO was coated prior to the formation of the Al back-electrode. Ultrathin ZnO layers were deposited at 200 °C using diethylzinc [DEZ, (C₂H₅)₂Zn] (as a metal precursor) and H₂O (as an oxygen source). A DEZ pulse was followed by an H₂O pulse to ensure reaction with DEZ chemisorbed on the surface. Figure 4(c) shows how the collection of hole carriers using the Al/ZnO/p⁺-Si contact can be improved [26,27]. An electric field is developed from p⁺-Si to Al; this is caused by the work function difference (1 eV) between p⁺-Si and Al, which induces dielectric dipoles inside the ZnO layer (which has a dielectric constant (k) of 8.75). The negative-ends of dipoles likely repel the negatively-charged boron dopants (as ionized acceptor states) that are close to the ZnO/p⁺-Si interface. The decrease in the boron concentrations nearby the ZnO/p⁺-Si interface decreases the work function of p⁺-Si, leading to the downward band bending. In Fig. 4(d), a surface potential (ψ_s) of ~103 meV (caused by downward band bending) is estimated at the normal k value of 8.75 (for ZnO). As a result of band bending, holes can be safely transported into the ZnO layer, and the probability for recombination of holes and electrons at the surface states (E_ss) confined nearby the Fermi level is diminished.

Figure 5(b) shows the backside NS Si solar cells integrated with ZnO-passivated contacts. 1.4-, 4.3-, and 7.2-nm-thin ZnO films were deposited on the NS Si surfaces using ALD cycles of 10, 30, and 50, respectively [see Figs. 5(d)-5(f)]. Electron-beam-deposited Al revealed poor step-coverage onto the high-aspect NS array; however, 4.3-nm-thin ZnO was able to conformally surround the NS Si array [Fig. 5(c)]. Ultrathin (1.4-nm-thin) ZnO was observed to form a non-uniform, sparse, island-type structures, as was reported previously [28]. Compared to τ_eff (68.1 μs) of as-prepared NS, the τ_eff values were observed to increase to 71.5, 89.6, and 90.3 μs for 1.4-, 4.3-, and 7.2-nm-ZnO deposited NSs, respectively.

 

Fig. 5 (a) Optical transmittance comparing various thicknesses of ALD-ZnO layers. The inset shows that optical bandgaps are 3.93 and 3.97 eV for ZnO thicknesses of 4.3 (red) and 7.2 nm (green), respectively. (b) A cell structure showing the ZnO-passivated contacts integrated with backside NS. Cross-sectional TEM images showing (c) the Al/ZnO/NS Si contact, and (d−f) various thicknesses (1.4, 4.3, 7.2 nm) of ALD ZnO deposited on NS surfaces.

Download Full Size | PDF

The poor contact morphology between Al and the NS Si (without ZnO) showed non-linear I-V characteristics, as shown in the inset of Fig. 6(a). In contrast, the Al/ZnO/p⁺-Si structure revealed an Ohmic behavior [Fig. 6(a)], which was caused by the improved contact resistivity. However cell performances were variable with different ZnO thicknesses (Table 1). A 1.4-nm-thin ZnO sample showed a slight increase in the cell efficiency compared to the unpassivated sample; this was due to the island-like morphology of ZnO that resulted from the partially-passivated surface. The highest photovoltaic conversion efficiency (13.1%) was recorded with the 4.3-nm-thin ZnO sample. The open-circuit voltage (V_OC), short-circuit current (J_SC), and fill factor (FF) values of this device were 566 mV, 30.8 mA/cm², and 74.9%, respectively. However, increasing the ZnO thickness further (to 7.2 nm), decreased the V_OC, J_SC and FF values by 32 mV, 2.6 mA/cm², and 4.4%, respectively. This occurred because the ZnO layer was too thick, which prevented direct tunneling from p⁺-Si to Al. As a result, the degraded hole collection leads to an increase in the Shockley-Read Hall (SRH) recombination at the mid-gap states within the silicon bandgap [29], which acts as the dominant recombination center by trapping charge carriers.

 

Fig. 6 (a) Dark I-V curves of the Al/ZnO/p⁺-Si structure with various ZnO thicknesses (1.4, 4.3, and 7.2 nm). Inset shows I-V curves without a ZnO layer. (b) J−V curves under illumination for (1) B−NS Si solar cells (black dots), (2) B−NS Si solar cells passivated by 1.4-nm-thin ZnO films (blue dots), (3) 4.3-nm-thin ZnO (red dots), and (4) 7.3-nm-thin ZnO (green dots). The inset shows the dark J-V curves.

Download Full Size | PDF

Table 1. Average photovoltaic performances over 10 devices of B−NS Si solar cells with various thicknesses of ALD−ZnO: (1) 0, (2) 1.4, (3) 4.3, and (4) 7.2 nm, respectively. (The champion data are shown by parentheses)

View Table

The behavior of SRH recombination can also be explained by the recombination current of the dark J-V characteristics [see inset of Fig. 6(b)], which shows the trap-assisted recombination process. The recombination current at the dark state under a reverse bias corresponds to the leakage current in c-Si solar cells. The reverse saturation current density (J_R) at 1.5 V was shown to increase by at least an order of magnitude as the ZnO thickness increases from 4.3 to 7.2 nm. The J_SC is calculated by the equation;

Here, J_ph is photocurrent, R_S is series resistance, and R_sh is shunt resistance; the R_sh is analyzed by a smaller (forward) bias range (0 to −0.4 V), and the R_S is analyzed by a larger bias range (−0.7 to −1.5 V) [30,31]. Note that the R_sh and R_S values in our 50-μm-thin samples should all be constant, irrespective of the ZnO thicknessess (1.4~7.2 nm). This result strongly supports the idea that a dramatic increase in J_R is a major factor that reduces J_SC. In addition, the increase in J_R can also decrease V_OC according to the following relation [32];

The interfacial bonding between silicon and oxygen was analyzed by FT-IR [Fourier transform infrared spectroscopy, Fig. 7(a)]. Weak bands between 1100 and 1150 cm⁻¹ indicate the Si−O−Si stretching mode, reflecting the formation of silicate species [33,34]. The spectra for 4.3- and 7.2-nm-thin ZnO films are shown to be much remarkable compared to that of 1.4-nm-thin film, which implies that the defective NS surface is likely passivated by stronger Si−O coordination as the ZnO thickens. Furthermore, nanoscale chemical information for the ALD ZnO (4.3-nm-thin)/NS Si interface has been obtained using the energy-loss near-edge structure (ELNES) in the electron energy loss spectra (EELS). Here, the O-K (520−560 eV) and Zn-L₁₂ (1050−1250 eV) edges were mainly investigated, as shown in Fig. 7(b) [35,36]. The c-Si surface, in which is roughly etched by Ag-assisted etching, is oxidized during the ALD processes. This results in the formation of silicon suboxide (SiO_x) at the ZnO/Si interfacial region. Compared to pristine ZnO (region IV), the interfacial region (region II) reveals a notable decrease in both of the oxygen and zinc peak intensities. Due to the disappearance of Zn peaks, we believe that the Zn−O bonds in the pristine ZnO were replaced by Si−O bonds at the interfacial region; however, their suboxide nature (SiO_x) also diminishes the O-σ* (~545 eV) intensity, which indicates the robustness of sp³ hybridization in solid oxide bonding [37]. This Si−O interface feature might be beneficial for reducing carrier recombinations via chemical passivation between Al and the NS Si contact.

 

Fig. 7 (a) FTIR spectra of NS Si passivated using ZnO thicknesses (1.4, 4.3, 7.2 nm). (b) EEL spectra spatially-resolved with a spectral interval of 1 nm. Inset shows a magnified high-resolution transmission electron microscopy (HRTEM) image for a 4.3-nm-thin ZnO layer on an NS Si surface. The scale bar in the inset represents 2 nm.

Download Full Size | PDF

4. Summary

The backside nanostructured Si solar cells showed enhanced quantum efficiency (QE) in the wavelength range from 550 nm to 1000 nm compared to their frontside counterparts. This difference was attributed to the fact that both the Auger and surface recombinations were reduced. Interestingly, the use of ALD−ZnO between the NS Si surfaces and the Al electrode formed passivated tunneling contacts, resulting in a ~24% improvement in the power conversion efficiency.