Efficient passivation of n-type and p-type silicon surface defects by hydrogen sulfide gas reaction

As shown in equation (1), the SRV values can be calculated from τ_eff, if the τ_bulk and wafer thickness are known. To validate the assumption that τ_bulk can be assumed to be the τ_fit estimated from (1/τ_eff-Auger term) vs Δn curves, τ_fit values are recorded with varying τ_eff obtained by different passivation processes. If the H₂S reaction passivates only the surface defects and the surface degrades under air exposure without affecting bulk lifetimes, the τ_bulk and hence τ_fit will be independent of τ_eff. Figure 1 shows the estimated τ_fit for three types of wafers, indicating a range of surface passivation quality obtained after H₂S reaction at different process conditions and during the atmospheric degradation with different τ_eff. Clearly, τ_fit is found to be independent of the surface passivation quality, and therefore determined to be representative of τ_bulk of the respective wafers. Average τ_fit of 10 000 μs, 4000 μs and 400 μs are obtained for n1-, n2-, and p-type wafers, respectively. These τ_fit values also closely match the values obtained by state-of-the-art thermal SiO₂ and Al₂O₃ surface passivation structures (symmetric passivation on both surfaces) in n2 and p-type sister wafers from the same batch, and are depicted as 'star symbols' in figure 1.

Zoom In Zoom Out Reset image size

Figure 2 shows the SRV values calculated from τ_eff at Δn = 1×10¹⁵ cm⁻³ and assuming τ_fit = τ_bulk using equation (1) for varying H₂S reaction temperature (figure 2(a)), H₂S concentration (figure 2(b)) in Ar, and time (figure 2(c)). The figure shows that SRV < 10 cm s⁻¹ is achieved at reaction temperatures between 500 °C–650 °C, H₂S concentration >1% and for all reaction times between 5–210 min for n-type Si. The lowest SRV ≈ 1.5 cm s⁻¹ on both n1-and n2-type Si is achieved at an optimum temperature of 550 °C, 3.4% H₂S concentration with 30 min reaction time. A similar trend of SRV with H₂S reaction temperature-concentration-time is observed for p-type wafers, but the SRV values of p-type wafers are higher, presumably due to much inferior bulk quality. Nonetheless, the lowest SRV = 8 cm s⁻¹ is obtained with an optimal reaction process of 550 °C, 3.4% H₂S concentration, and 30 min reaction time. The S-passivated Si SRV of 1.5 cm s⁻¹ on n2-type Cz wafer is comparable to the best surface passivation qualities using state-of-the-art materials, SiO₂ (2.4 cm s⁻¹), a-SiN_X :H (3.5 cm s⁻¹), a-Si:H (0.7 cm s⁻¹), and Al₂O₃ (1.3 cm s⁻¹), reported in literature [9]. A direct comparison of SRVs obtained by different groups and with different materials could be misleading, since these values can vary appreciably depending on the wafer type, dopant type & concentration (bulk resistivity), wafer manufacturing process (float zone or Czochralski method), and bulk qualities. Therefore, surface passivation by a SiO₂/a-SiN_X :H stack on n2-type and an Al₂O₃/a-SiN_X :H stack on p-type sister wafers from the same batch was also performed in this work. The τ_eff and τ_fit are shown as star symbols in figure 1. The calculated SRV values of 2.7 and 13 cm s⁻¹ for n2-type and p-type, respectively, by the oxide materials are slightly higher (i.e. indicating a poorer passivation) than the best SRV of S-passivated Si (1.5 and 8 cm s⁻¹ for the respective wafer types).

Zoom In Zoom Out Reset image size

A continuous decrease in SRV, up to reaction temperatures =550 °C, suggests that the hydrogen atoms are likely not responsible for the surface passivation in this process, since hydrogen desorbs from the surface in the temperature range of 400 °C–550 °C [13]. Interestingly, an increase of surface S/Si ratio in AES is reported up to a temperature of 550 °C, followed by a sharp decrease in sulfur (S) peak intensity for temperatures >625 °C [13]. This loss of S-signal was initially attributed to desorption of S from the surface [25], but it was then argued to be due to S-diffusion into the Si bulk, since no S was detected in TPD measurements [13]. Furthermore, low-energy electron diffraction studies of S adsorption on clean Si(100) surface indicated that S adsorbates can cause a transition of the reconstructed Si(100)2 × 1 surface to its original bulk-terminated Si(100)1 × 1 structure [26–28]. Thus, considering the results reported in literature, a plausible explanation for the efficient surface defect passivation (low SRV), observed in this study in the temperature range of 550 °C–600 °C, is based on a S-induced Si(100)2 × 1 surface reconstruction to the original bulk-terminated Si(100)1 × 1 surface by breaking the Si dimer bond [13, 25]. At higher temperatures (650 °C), S atoms may diffuse to the near-surface Si bulk and create additional defects generated by sub-surface S–Si bonds leading to higher SRV, especially evident for p-type Si in figure 2(a). A small dose of H₂S (<1% in Ar) in the reaction process at 550 °C for 30 min is sufficient to reduce SRV by ∼2 orders of magnitude for both n-type and p-type Si, but no surface passivation is observed without the H₂S (figure 2(b)). This suggests that the surface passivation is determined by an extremely thin, likely ∼(sub)monolayer coverage of S. Finally, a constant low SRV for 3.4% H₂S at 550 °C in the entire range of reaction duration (5–210 min) (figure 2(c)) indicates a fast process for efficient surface passivation.

In order to apply this novel S-passivation approach to any Si solar cell structure, it is imperative to demonstrate efficient surface passivation (low SRV) on textured surfaces. The front illumination surface of a Si solar cell needs to be textured for low optical reflection loss. In this study, n2-type Si wafers were etched in dilute KOH solution to create a random pyramid structure that increases the surface area and forms {111} facets on the Si(100) surface. Figure 3 shows the SRV comparison for planar and textured n2-type Si wafers after H₂S reaction at 550 °C in varying concentration (figure 3(a)) and time (figure 3(b)). Figure 3 demonstrates that both planar and textured surfaces are passivated equally well. Therefore, the same passivation mechanism is valid for both planar and textured wafers with a fast defect-terminating process, despite the increase in area of the {111} faceted surface in textured Si.

Zoom In Zoom Out Reset image size

Atmospheric stability of S-passivated planar Si wafers is evaluated after exposure in laboratory air at relative humidity of 40%–60% for an extended period of time. Figure 4 shows that the passivation quality degrades rapidly as a function of time in air, SRV increases by ∼an order of magnitude within 30 min. The degradation rate is similar for both n- and p-type wafers, despite the difference in their initial SRV. This rapid loss of surface passivation is explained by a change of S–Si surface bonds to form SiO₂ in a thin (<5 nm) S surface layer, as reported for similar H₂S reaction conditions [15]. The resulting low temperature native oxide does not passivate Si surfaces well on its own [9]. Furthermore, loss of Si surface passivation by ultrathin layers is not surprising and is also observed for Si surfaces passivated by ultrathin (<5 nm) a-Si:H [29], Al₂O₃ [30, 31] and SiO₂ [32].

Zoom In Zoom Out Reset image size

A simple mitigation strategy to eliminate the passivation degradation can be achieved by capping the S-passivated surface using a low-temperature-deposited thin a-SiN_X :H films. A set of wafers were passivated by 3.4% H₂S reaction at 550 °C for 30 min in multiple runs, and the samples were immediately loaded into a PECVD reactor for a-SiN_X :H deposition. The air exposure of the S-passivated samples during transfer between the two systems was <5 min. A 85 nm layer of a-SiN_X :H was deposited on both sides of the S-passivated samples at 300 °C to protect the S-layer from air exposure. Freshly cleaned wafers without the S-passivation were also included in each a-SiN_X :H deposition run, for comparison. The SRV's of the samples were recorded by repetitive QSSPC measurements for up to ∼2 months, with samples stored in laboratory air. Figure 5 shows the change in SRV as a function of time for n2-type and p-type wafers with and without the S-passivation layer. The initial SRV was 5 cm s⁻¹ for n2-type Si, passivated by a S + a-SiN_X :H stack, which remained stable for months with no degradation. However, a slow degradation of SRV from 8 to ∼20 cm s⁻¹ over 2 months storage in air is observed for the p-type wafer passivated by the same stack layer. This difference in SRV stability between n2-type and p-type Si is not well understood and requires further investigation. Nonetheless, a low-temperature a-SiN_X :H capping layer effectively slows down the SRV degradation rate and stabilizes SRV for months, as opposed to a rapid SRV degradation (an order of magnitude in 30 min) observed in figure 4 without the capping layer. Such low-temperature a-SiN_X :H alone, however, does not provide any significant surface passivation, with SRV > 600 cm s⁻¹ (figure 5), which was also observed in our previous work [15, 33]. We note that the a-SiN_X :H layer can also be used as an antireflection coating by controlling the thickness and the refractive index [34].

Zoom In Zoom Out Reset image size

In order to further investigate the thermal stability of S + a-SiN_X :H stack passivation, the samples were subjected to thermal firing at peak temperature ∼750 °C for ∼5 s. Figure 5 shows that SRV after thermal firing on day 90 increases from 5 cm s⁻¹ to ∼30 cm s⁻¹ for n2-type and 30 cm s⁻¹ to ∼300 cm s⁻¹ for p-type Si. Therefore, the passivation loss is evident for both types of wafers upon thermal treatment, but the SRV values are still ∼an order of magnitude lower than the samples that did not have any S-passivation and were coated with low-temperature a-SiN_X :H only. The wafers passivated by only a-SiN_X :H do not show any changes in SRV upon thermal firing, albeit at a much higher values of >600 cm s⁻¹. A careful examination of surface morphologies before and after the thermal firing step for both a-SiN_X :H and S + a-SiN_X :H passivated Si surfaces reveals an appearance of blistering and pinholes upon thermal firing. Surface blistering is likely due to the desorption of molecular hydrogen (H₂) from the unoptimized low-temperature deposited a-SiN_X :H films, with a low film density as reported in literature [35]. These pinholes can lead to pathways for atmospheric oxidation of the S-layer underneath the a-SiN_X :H and cause fast SRV degradation, as in figure 4.

Further process optimization of a-SiN_X :H capping layers does, in fact, eliminate the thermal degradation of S + a-SiN_X :H stack passivation. A set of emitter-diffused Si wafer samples was passivated by H₂S reaction to achieve an emitter saturation current density (J_0e) < 70 fA cm⁻² and τ_eff > 600 μs. The S-passivated wafer surfaces were then protected on both-sides by an (unoptimized) 30 nm thick low-temperature (300 °C) deposited a-SiN_X :H at UDel, followed by a 70 nm a-SiN_X :H layer deposited at 425 °C, optimized for industrial Si solar cell processes, at Georgia Institute of Technology. The emitter saturation current densities (J_0e) and τ_eff of these stack-passivated samples remain stable, with no visible surface blistering and pinholes, after the same thermal firing step (∼750 °C for ∼5 s). This demonstrates that an ultrathin S-layer, in combination with an optimum a-SiN_X :H process, can indeed achieve atmospheric and thermally stable excellent surface passivation for both n- and p-type Si. Details of the dopant-diffused Si surface passivation by S and the mechanism for thermal stability is, however, beyond the scope of this paper and will be reported in a future publication.

The chemical environment at the surface and surface-near bulk of the n1-type Si wafers were investigated using XPS and XES. Figure 6 shows the XPS detail regions of S 2s (a) and Si 2p (b) after the HF cleaning process prior to H₂S reaction ('clean'), after the H₂S reaction ('passivated'), and after 8 d of air exposure ('degraded'). The gray boxes represent binding energies commonly found for the various chemical environments, namely, sulfur in a S–O and sulfide chemical environment, and silicon in Si–Si and Si–O bonding environments. No Si–O bonds are present on the 'clean' Si wafer surface, indicating the effectiveness of the cleaning process and packing procedure. Note that a small amount of sulfur in a sulfide chemical environment is already present at the surface of the 'clean' wafer (at a binding energy of ∼226.5 eV), presumably due to sulfur in the cleaning process. After the H₂S passivation reaction, the S 2s spectra show sulfur mainly in a sulfide environment, with a small sulfur signal representative of a S–O environment (binding energy ∼233 eV). After exposure to air (i.e. for the 'degraded' sample), the sulfur signal in the spectrum decreases, but remains in a sulfide chemical environment. This is attributed to a loss of sulfur during air exposure of the surface, e.g. interacting with moisture in the air. Likely, H₂S is formed and desorbed, allowing Si–O bonds to form instead [15, 36]. As evidence, the Si–O bond signal in the Si 2p core levels increases (and is dominant) after exposure to air, with a decrease in Si–Si bonds. Note that, already during the passivation reaction, some Si–O bonds are formed but Si–Si bonds remain dominant. The degradation processes then reverse this balance at the Si wafer surface; similar findings with different wafer types were also reported previously [16]. The likely source of the Si–O peak in the passivated samples is argued to be due to impurities in 99.9% H₂S source gas itself, since similar time-temperature reactions performed in the same reactor with 99.999% purity Ar with no H₂S gas flow only found negligible Si–O peaks in XPS [15]. The manufacturer certificate of analysis of 99.9% H₂S indicated that as much as 500 ppm of water vapor is present in the H₂S source gas. A higher purity H₂S source gas and/or additional in-line gas purifier would be needed to minimize/eliminate the presence of a Si–O XPS peak in the passivated samples.

Zoom In Zoom Out Reset image size

Using S L_2,3 XES, the local chemical bonding environment of sulfur at the surface and in the surface-near bulk can be studied in a complementary approach to XPS. Figure 7 shows the S L_2,3 emission of the three samples, together with a SiS₂ reference powder and a magnified section of the 'passivated' spectrum' to better identify the chemical fingerprint of S–Si bonds (note that the SiS₂ powder sample was scanned under the x-ray beam to avoid beam damage). All spectra in figure 7 were normalized to the maximum count rate of the spectrum, i.e. at the S 3s-derived emission line labeled (1). The presence of this peak indicates sulfur atoms in a sulfide environment. As for XPS, the 'clean' spectrum already shows some sulfur in a sulfide chemical environment. After the passivation process, (further) S–Si bond formation can be seen through the spectral features between 152 eV to 160 eV (also shown enlarged (×3) to compare with the SiS₂ spectrum). However, after air exposure (i.e. in the 'degraded' spectrum), these features are no longer present. This indicates the removal of S–Si surface bonds after air exposure and corroborates the decrease in XPS sulfur intensity. Overall, both XPS and XES studies thus show a reduction of S–Si bonds and the formation of Si–O bonds due to air exposure, in parallel to the observed increase in SRV.

Zoom In Zoom Out Reset image size