Excellent Silicon Surface Passivation Achieved by Industrial Inductively Coupled Plasma Deposited Hydrogenated Intrinsic Amorphous Silicon

We present an alternative method of depositing a high-quality passivation film for heterojunction silicon wafer solar cells, in this paper.The deposition of hydrogenated intrinsic amorphous silicon suboxide is accomplished by decomposing hydrogen, silane, and carbon dioxide in an industrial remote inductively coupled plasma platform. Through the investigation on CO 2 partial pressure and process temperature, excellent surface passivation quality and optical properties are achieved. It is found that the hydrogen content in the film is much higher than what is commonly reported in intrinsic amorphous silicon due to oxygen incorporation. The observed slow depletion of hydrogen with increasing temperature greatly enhances its process window as well. The effective lifetime of symmetrically passivated samples under the optimal condition exceeds 4.7ms on planar n-type Czochralski silicon wafers with a resistivity of 1Ωcm, which is equivalent to an effective surface recombination velocity of less than 1.7 cms and an implied open-circuit voltage (Voc) of 741mV. A comparison with several high quality passivation schemes for solar cells reveals that the developed inductively coupled plasma deposited films show excellent passivation quality. The excellent optical property and resistance to degradation make it an excellent substitute for industrial heterojunction silicon solar cell production.


Introduction
Heterojunction silicon wafer (HET) solar cells have achieved very high efficiencies of up to 24.7% on a large area (101.8 cm 2 ) [1].Its high efficiency mainly stems from an extremely high open-circuit voltage ( oc ) of up to 750 mV thanks to the excellent surface passivation.Such a cell structure has attracted attention not only due to its high conversion efficiency, but also due to its low process temperature which avoids bowing when using thin wafers [2,3].The insertion of a high-quality hydrogenated intrinsic amorphous silicon (a-Si:H(i)) thin film between the emitter (or back surface field) and the bulk material is the key to such a high cell efficiency [4][5][6][7].Therefore, optimization of this buffer layer is of utmost importance.
Plasma-enhanced chemical vapour deposition (PECVD) of a-Si:H(i) has been the key to the successful industrial implementation of this technology as a passivation scheme, since the early 1990s [5,8,9], mainly thanks to its low temperature process and good surface passivation quality [10].However, it is also reported that the optimization for such a layer is difficult because of possible epitaxial growth and the need to balance plasma species during deposition [11][12][13][14].
PECVD hydrogenated intrinsic amorphous silicon suboxide (a-SiO  :H(i)) thin films are emerging as an alternative 2 International Journal of Photoenergy to replace a-Si:H(i) as passivation layer in HET solar cells.Its excellent passivation quality and optical property has been demonstrated by several groups in recent years [15][16][17][18][19]. Notably, Mueller et al. [16,18,20,21] have shown a record of high passivation quality on highly doped wafers using direct capacitively coupled plasma (CCP).The optical loss reported was also much lower than that of a-Si:H(i), proving promising potential as a passivation layer.However, in order to achieve a state-of-the-art passivation level, the mentioned group has used very high plasma frequencies (70-110 MHz) and prolonged annealing processes, which can be difficult to implement in industrial production.Furthermore, direct CCP can introduce extra damage to wafer surface during deposition, thus limiting the film's passivation quality.
In this paper, we present an alternative method of depositing a-SiO  :H(i) thin film using a remote inductively coupled plasma (ICP) source operating at conventional frequency (13.56 MHz).It is well known that very high density plasma can be produced in an ICP system with low and narrowly distributed particle energy compared to CCP [22][23][24], and the wafer surface is less influenced by plasma condition by keeping substrate from plasma in a remote system.Therefore, the passivation quality of a-SiO  :H(i) can be further improved with reduced process time.In this preliminary study, the optimization of a-SiO  :H(i) is carried out in terms of passivation quality on highly doped material.The influence of CO 2 partial pressure and process temperature on the passivation quality is investigated.Detailed analyses, such as quasi-steady-state photoconductance (QSSPC) measurements, spectroscopic ellipsometry (SE), time-of-flight (TOF) secondary ion mass spectroscopy (SIMS), and Fourier transform infrared (FTIR) spectroscopy, are conducted to understand the passivation quality, film thickness, composition, optical bandgap, and chemical bonding configuration.In the last section, comparison with other high-quality passivation schemes is made in order to demonstrate the excellent passivation and optical qualities with good stability against ambient environment of the ICP a-SiO  :H(i) films.

Experimental Details
The a-SiO  :H(i) thin films were fabricated by decomposing silane (SiH 4 ), hydrogen (H 2 ), and carbon dioxide (CO 2 ) in a remote ICP system, the SINGULAR-HET.It is a fully automatic ICPECVD system developed by Singulus Technologies in collaboration with SERIS for industrial HET solar cell processing.The system has one infrared (IR) heating station (PS1) for preheating and four other process stations (PS2-PS5) for deposition.It has two stations (PS2 and PS5) for topside deposition and two stations (PS3 and PS4) for bottom-side deposition.Double-side deposition is accomplished by using PS2 and PS3 without breaking the vacuum.The process temperature is adjusted by changing the heater power setting in PS1 as well as the stabilization heater in each process station.The temperature is monitored by pyrometers installed after each station.The pyrometers are calibrated to Si wafers and verified using in situ thermal couple readings in the temperature range of interest.The Si wafers are loaded into the system using an automated cassette wafer handling unit.The structure and process cycle of the SINGULAR-HET is illustrated in Figures 1(a) and 1(b).
In each process station, the ICP source is driven by a radiofrequency (RF, 13.56 MHz) generator through a matching box.The reactive gases (H 2 and CO 2 ) are fed through a gas shower into the area surrounded by an inductive coil, where the plasma is generated.The deposition precursor gas (SiH 4 ) is introduced through a distribution ring outside the inductive coil near the wafers.The decomposed species travel downwards for PS2/5 onto the wafer surface to form the film.The ICP source and process station design are shown in Figure 1(c).
The a-SiO  :H(i) films were deposited onto both sides of planar 160 m thick 156 mm wide Czochralski (Cz) wafers with crystal orientation of ⟨100⟩.The choice of the -type 1 Ωcm Cz wafer as the bulk material is based on the applicability to real HET solar cell fabrication.The film thickness was about 20 nm, except for the results shown in Section 3.1.This symmetrical a-SiO  :H(i)/c-Si/a-SiO  :H(i) structure was used to monitor the passivation quality.The same films were also deposited onto one side of highly resistive Si wafers for FTIR analyses.Before deposition, all wafers underwent an alkaline based silicon etch and a standard RCA (Radio Cooperation of America) cleaning followed by a one-minute dip in 1% hydrofluoric (HF) acid solution to remove the chemical oxide.After rinse and drying, the wafers were loaded onto substrate carriers and transferred into the load lock.Plasma parameters used in this work are summarized in Table 1.It should be noted that equal flow rates for SiH 4 and H 2 were used throughout this investigation.CO 2 flow rate was monitored and shown in terms of partial pressure defined by where square brackets represent the corresponding gas flow rate in standard cubic centimetre per minute (sccm).Furthermore, the process temperature is reported using the pyrometer reading on the wafer surface after PS1.The effective minority carrier lifetime ( eff ) of the symmetrical samples was determined using the QSSPC method in appropriate modes [39].The  eff reported are as-deposited values extracted at a minority carrier injection level of 10 15 cm −3 and approximately two minutes after deposition.Thermal annealing was avoided in this study so that the process is industrially compatible.The thickness of the film was determined by fitting SE data (300-800 nm).The fitting model comprised an underlying SiO 2 layer with a fixed After [25,26].f Reference [38].
thickness of 1 nm, an a-SiO  :H(i) layer characterized by Tauc-Lorentz (TL) model [40], and an overlayer consisting of 50% bulk material and 50% voids to represent the surface roughness.The absorption coefficient () of the film was determined from the fitted distinction coefficient () using  = 4/, where  is the wavelength.The optical bandgap of the film was then determined by Tauc's method [41].In order to investigate the bonding configuration in the film, FTIR measurement (500-4000 cm −1 ) in transmission mode was carried out with a spectral resolution of 2 cm −1 .The bonding configurations of interest are summarized in Table 2.The total H content ( H ) in the film was obtained from the integrated intensity under the Si-H(Si 3 ) wagging mode peak [35], while the microstructure factor () was determined by the following equation [42]: where   represents the integrated intensity under the peak denoted by the corresponding subscript.The peak intensity was obtained by deconvoluting the Si-H(Si 3 ) stretching peak using Gaussian curves.In particular, the peak centred at 2000 cm −1 usually represents monohydride configuration, while the peak at 2100 cm −1 represents the response from higher order hydrides [43,44].
In order to quantify the atomic concentration of C and O under different   , TOF SIMS (IONTOF) was conducted on selected samples using a 25 keV Bi+ beam with a current of 1 pA on a 80 × 80 m 2 spot.Sputtering was done using a 1 keV Cs+ gun with a current of 8 nA on a 250 × 250 m 2 area.The data were presented in both absolute atomic concentration per cubic centimetre and percentage over the host Si atoms.For percentage calculation, standard atomic concentration of crystalline Si was used (5 × 10 22 cm −3 ) without considering density change in amorphous material.This is justified by the negligible density change in amorphous Si reported by Smets et al. [45] and Remeš et al. [46,47] under different  H .

Deposition Time and Film Thickness.
The investigation first focused on the deposition rate of a-SiO  :H(i) in the ICP system in order to determine the thickness of film on which further optimization will be carried out.The result is shown in Figure 2(a).
A linear relationship between thickness and deposition time is observed, demonstrating a stable and robust deposition process for all samples.The nonzero interception at 0 deposition time can be attributed to the fluctuating plasma condition and nucleus formation at plasma ignition.A deposition rate of about 7 Ås −1 can be deduced from the slope of the linear fit.The same linear growth was also reported by Pysch et al. in a different ICP system for a-Si:H(i) deposition [48].The linear behaviour can facilitate the thickness control in the subsequent optimization.Notably, the relatively high deposition rate can greatly reduce deposition time needed in real device fabrication.For a typical passivation layer of 5 nm or below, 1 to 2 s deposition is sufficient, thus largely enhancing the throughput.
The relationship between passivation quality and film thickness was then explored and presented in Figure 2(b).It can be observed that the passivation quality increases rapidly at low film thickness.The improvement of lifetime slows down after a film thickness of 10 nm and then tends to saturate after 20 nm.The saturation of  eff in terms of film thickness can be attributed to the reduction of interface defect density due to better network and surface coverage.This saturation effect was reported in the literature on different passivation materials [16,24].Further increase in film thickness can only provide marginal improvement of passivation quality, with the possible risk of inducing morphological change in the film (discussed in Section 3.3.3).Therefore, the film thickness is controlled at about 20 nm in this paper to ensure lifetime saturation.Such a thickness avoids second order effects that distort the actual passivation quality of the film while maintaining the minimal process time.At the same time, the similar film thickness ensures fair comparison on passivation quality across different plasma settings.

CO 2 Partial Pressure
3.2.1.Passivation Quality.While fixing the film thickness at about 20 nm, a scan on CO 2 partial pressure was performed in this experiment.Other plasma parameters were kept constant in this section.By varying the concentration of CO 2 in the plasma, compositional change that affected film passivation quality was expected.The lifetime results are illustrated in Figure 3.
It is shown that the increase of CO 2 partial pressure of up to about 10% improves passivation quality.Any increase beyond the optimal amount deteriorates the lifetime.The relatively low lifetime in Figure 3 is due to nonoptimized wafer temperature in this series.However, it should be noted that only a small amount of CO 2 is needed to greatly improve the passivation quality, as compared to a-Si:H(i) (denoted by 0% partial pressure in the same figure).It is also interesting to notice that the optimal CO 2 partial pressure in our ICP system is only about half of what was reported by Mueller et al. [16,18,20].The difference could be a strong indication of different plasma chemistry and reactor design.The initial improvement of lifetime with respect to CO 2 flow rate can be explained by the increase of H content ( H ) in the film.Higher CO 2 partial pressure in the plasma induces higher O content inside the resultant film.As H atoms can easily bond to Si atoms that are back bonded by O atoms [49], higher  H is expected due to O incorporation.Indeed, this argument will be supported with FTIR spectra and H content calculations.Once beyond the optimal point, lifetime starts to reduce with increasing CO 2 flow rate.Several groups [16,18,19] attribute this phenomenon to increasing defect density due to higher C impurity concentration in the film.Considering similar CO 2 partial pressure used by Mueller et al. [16,18] and Hoex et al. [19] compared to this work, higher impurity incorporation can be regarded as one of the reasons that cause lifetime degradation at higher CO 2 flow rate in this study as well.However, by examining FTIR spectra and bonding configuration under different CO 2 flow conditions, it can be concluded that the reduction of  H at higher O concentration is a key reason for the degradation beyond the optimal point.TOF SIMS analysis also reveals that the concentration of C atoms in a-SiO  :H(i) is quite small even at the highest   used in this work.

Bonding Configuration Analysis.
In order to further investigate the influence of CO 2 partial pressure on the film property, FTIR analysis was carried out.The results are shown in Figure 4.Note that for all samples, contribution from O-H bonding group is not observable, which is similar to other a-SiO  :H(i) thin films commonly reported in the literature [15,19,37].The Si-C or C-H bonding configuration is also not clearly visible in the spectra due to extremely small inclusion of C atoms in the film [50,51].
From Figure 4, it can be observed that Si-O-Si stretching peak intensity increases with   .The observed small peak for spectrum with 0 sccm CO 2 flow is probably due to thin native oxide grown on the wafer during transfer.Assuming the peak intensity is proportional to O concentration in the film [38,52], it can be concluded that the film contains higher O content with increasing CO 2 flow.It is interesting to observe that the Si-O-Si stretching peak at 1107 cm −1 becomes more significant with higher   compared to that at 1053 cm −1 .Jana et al. [37] and Zhou et al. [15] associate these peaks to Si-H(Si 3− O  ) with  = 2 and 1, respectively, according to random bonding model [53].The increase of peak intensity at 1107 cm −1 indicates an increase of Si-H(SiO 2 ) bonding concentration.The shift of overall Si-O-Si stretching mode to a higher wavenumber is also an indication of higher O content in the film [54,55].Therefore, high O content and higher order Si-O-Si bonding configuration resulted in high CO 2 flow condition.The highest O bonding configuration at 1156 cm −1 reported by the mentioned groups is not clearly observed even at over 20% CO 2 partial pressure in our case, indicating that Si-H(SiO 2 ) is the dominant bonding configuration in the   range used in this paper.Notably, two sharp peaks appear at 845 cm −1 and 890 cm −1 , which are not commonly observed in a-SiO  :H(i) spectra.This doublet is associated with the polymer chain, (SiH 2 )  , in the film.The appearance of such a doublet is an indicator for low process temperature [38].Therefore, the lifetime result in this section is much lower than optimized.The peaks at 640 cm −1 and 2000 cm −1 denote Si-H rocking/wagging and stretching mode [35], respectively.Total H content and bonding configuration can be deduced from these peaks [56].The results are shown in Figure 5.

International Journal of Photoenergy
It can be observed that  H inside the film increases with   .Compared with Figure 4, the increase of  H can be regarded as the result of increasing O content in the film at higher CO 2 flow rate.At the same time, increasing O incorporation does not significantly affect the microstructure factor.Therefore, the improvement of passivation quality is mainly due to the increasing amount of H atoms that can saturate the interface dangling bonds.Indeed, when O content is further increased beyond the optimal point,  H starts to decrease, leading to a reduction of  eff .The reduction of  H could be due to the increased H effusion rate at low temperature as a function of O content in the film.The O cluster formed at higher   induces a void-rich network and promotes the effusion of H atoms [57,58].Same fluctuation and reduction of H content with increasing O content was also observed by Janotta et al. [49]and Yun et al. [52].The direct dependence of passivation quality on  H indicates that besides impurity incorporation,  H in a-SiO  :H(i) still dominates the passivation quality, which can be adjusted by   .The same dependence on  H is also widely observed in a-Si:H(i) [10].However, higher passivation quality of a-SiO  :H(i) comes from the fact that it has much higher  H (>20 at.% in our case) compared to that in a-Si:H(i) (12-14 at.%) [16,59].

Film Composition Analysis.
In order to get more insight on the elemental composition of a-SiO  :H(i) films deposited at various   , TOF SIMS analysis was carried out to quantify the depth-resolved concentration for C and O.The results are shown in Figure 6.
Similar to the result obtained by FTIR, O concentration increases with CO 2 partial pressure (cf. Figure 6(a)).The highest O concentration achieved in this work is about 7 × 10 20 cm −3 , which agrees well with Mueller et al. [16] at a similar CO 2 partial pressure.It should be noted that the uneven distribution of O content throughout the film could be related to the specific depth resolution of SIMS technique used in this paper.The thinness of the film and short deposition duration could also cause such an observation.
Figure 6(b) shows the change of C impurity concentration in the passivation film at different   .C concentration increases at about one order of magnitude for a CO 2 partial pressure change of 20%.C content in this study is slightly higher compared to the result obtained by Mueller et al. [16].This could be due to the difference in plasma chemistry and reactor design.Nevertheless, the highest C content in the film is only about 0.1 at.%, and it is not sensitive to the change in CO 2 partial pressure.Therefore, the change of H content in the film is the main reason for the change of passivation quality, as concluded in the previous sections.

Process Temperature
3.3.1.Passivation Quality.In Section 3.2.2,we have shown that the used process temperature is below optimum.Therefore, a scan in process temperature while fixing film thickness at 20 nm and   (10%) at optimum was performed.Other plasma parameters were kept constant.The result for temperature dependent passivation quality is shown in Figure 7.
The temperature process window of a-SiO  :H(i) in ICP system seems to be very wide compared to a-Si:H(i) thin film [12].For a very broad temperature range (over 200 ∘ C), the lifetime stays above 1 ms, demonstrating a very stable process with suppression of epitaxial formation by a-SiO  :H(i) thin film [15,60].Despite the stable lifetime, a peak is apparently present near 250-300 ∘ C, with a lifetime exceeding 4 ms at about 300 ∘ C. The excellent passivation quality is mainly due to a very high H content at the optimal condition.
The initial increase of lifetime is due to better bonding structure benefited from thermal relaxation [42,61].Higher temperature releases incorporated H atoms from trapped state (Si-H 2 ), which is reflected by a reducing  (see Figure 9).These H atoms are mobilized to saturate the interface dangling bonds, thus providing chemical passivation.This phenomenon is further investigated in the following section using FTIR.Once beyond the optimal temperature, the lifetime reduces as a result of H effusion.Indeed, 300 ∘ C seems to be the optimal process temperature in our ICP system, which agrees with H effusion temperature reported in the literature [16,62].The slightly lower effusion temperature is probably due to the difference in material and deposition technique.The reduction in  H is the cause of deteriorated passivation quality due to insufficient interface dangling bond saturation.

Bonding Configuration Analysis.
In order to understand the temperature dependence of a-SiO  :H(i) passivation, FTIR analysis was carried out.The main features on FTIR transmission spectra are similar to those shown in Figure 4.The results are illustrated in Figure 8.The intensity change in Si-O-Si stretching mode peaks with respect to process temperature is not evident, indicating a roughly constant O content in the film throughout the temperature range used.This is reasonable as   is kept constant in this section.However, the shift of O bonding structure to higher order (Si-H(SiO 2 )) with temperature is more prominent in this temperature series compared to the previous sections, as evidenced from the heightening and sharpening of the peak at 1107 cm −1 .It seems that process  temperature and   has similar effect in altering the bonding configuration inside the film.The shift of Si-O-Si stretching peak position at high temperature is also observed by Yun et al. [52] and is caused by the disproportionation of the reaction chemistry of a-SiO  :H(i) into a more oxygen-rich suboxide.
In order to find the direct reason for improved lifetime with respect to temperature,  H and  are calculated following the same procedure described in Section 3.2.2.The results are illustrated in Figure 9. Temperature ( ∘ C) Figure 9:  H and  as functions of temperature.The  H displays a gradual reduction with a plain over a large span of temperature, resulting in good lifetime with a wide process window.The lines are guides for the eye.

International Journal of Photoenergy
H at low temperature is over 30 at.%, indicating a voidrich network with a dominating Si-H 2 bonding configuration.The defective network is also evident from strong (SiH 2 )  doublet peaks at 845 cm −1 and 890 cm −1 , similar to those observed in Figure 4.With increasing temperature,  H reduces in a nonlinear behaviour, showing a plain over a wide range of temperature.At the same time, the intensity of the doublet reduces, indicating a better network structure at elevated temperature.On the other hand,  reduces in a linear manner, indicating a better a-SiO  :H(i) network with less H clusters and microvoids [42,61].The combination of a high  H due to slow H depletion and an improved amorphous network results in an excellent passivation quality within the process window observed in Figure 7.It is interesting to notice that the sudden reduction of lifetime in Figure 6 coincides with the abrupt loss of H in Figure 9 at around 300 ∘ C. As 300 ∘ C is close to the commonly reported temperature at which H effusion takes place, the reduction of lifetime beyond the optimal temperature is indeed due to H effusion.Comparing the trend shown in Figures 7 and 9, it can be concluded that the passivation quality of a-SiO  :H(i) is strongly influenced by  H in the film, similar to its a-Si:H(i) counterpart.This observation is also intuitive because interface passivation in such samples utilizes H atoms to saturate the dangling bonds as well.Yet, in this study, the apparent large process window in terms of temperature is mainly benefited from the gentle and wide slope that crosses a temperature range of more than 200 ∘ C, with a constantly decreasing .The excellent lifetime at optimal condition is a result of much higher  H (over 20 at.%) compared to standard a-Si:H(i) thin film [59], as mentioned in Section 3.2.2.
It should be noted that several groups [63,64] reported improved lifetimes after an H plasma postdeposition treatment with even increasing .This seems contradictory to our finding but can be attributed to greatly increased  H upon plasma treatment, which diffuses into the interface upon annealing.The improvement of lifetime then cannot be solely explained by the bonding configuration in the amorphous structure.Indeed, the combination of both high  H and low  in our a-SiO  :H(i) is the key to such an excellent surface passivation quality, as concluded in the earlier discussion.In case of a-SiO  :H(i), the use of H plasma treatment is therefore not necessary to achieve high passivation quality due to the inherent high H inclusion of the film.

Optical Property.
The optimized a-SiO  :H(i) has demonstrated excellent passivation properties due to enhanced H content and bonding structure.However, in order to apply the film to HET solar cells, optical properties are equally important.Therefore, Tauc's method [41] was used to find the optical bandgap from SE fitting result.The results are shown in Figure 10.
The optical bandgap increases almost linearly with deposition temperature despite the fact that  H reduces, as shown in Figure 9.In the case of a-Si:H(i), optical bandgap is positively dependent on  H in the film [65].In our case, the contradiction could possibly be explained with an altered O bonding structure.As shown in Figure 8, the elevated temperature shifts the O bonding configuration in a-SiO  :H(i) thin films to a higher mode and could possibly cause an increase in the stoichiometry parameter  in the a-SiO  :H(i) film.This change can be one of the reasons for the improved optical bandgap.The optical bandgap at optimal temperature is above 2 eV, which greatly suppresses optical absorption at low wavelengths and improves  sc when applied in HET solar cells.
On the other hand, the film thickness reduces slightly (not shown) with process temperature despite the same deposition time.Since the thickness is obtained from SE fitting, one can argue that the reduction of thickness is due to increased

∼160 𝜇m
Figure 11: Measured QSSPC lifetime of optimized a-SiO  :H(i) film as a function of minority carrier injection level.The calculation of effective surface recombination velocity is based on wafer thickness of our samples.The dashed line represents intrinsic limit of the 1 Ωcm wafer assuming no extrinsic recombination [27].As a comparison, other high-quality passivation schemes from Mark and Andres [28,29], Duttagupta et al. [30], Wan et al. [31], Mueller et al. [16,18], and Ge et al. [10] are also shown in the graph.The inset shows the sample structure for ICP a-SiO  :H(i) passivation.film density at higher temperature.Indeed, the loss of H content and the reduction of microvoid volume (see Figure 9) indicate an optically compact network, therefore causing a reduction in the film thickness.The reduction of thickness also has an impact on the film morphology, altering its crystallinity and optical bandgap as well [66].

Comparison with Existing Passivation Schemes
3.4.1.Passivation Quality.In order to demonstrate the excellent surface passivation result of the optimized a-SiO  :H(i) thin film, a comparison in terms of lifetime between our film and some other high-quality passivation schemes was made.The result is shown in Figure 11.By assuming identical surface passivation on both sides of the wafer and sufficiently low surface recombination, the effective surface recombination velocity ( eff ) can be related to  eff using the following relationship: where  is the thickness of the wafer and  bulk represents the bulk lifetime of the wafer, which is a combined result of Auger, radiative, and Shockley-Read-Hall recombination.By further assuming a perfect bulk quality (i.e.,  bulk → ∞), the For comparison, the absorption coefficients of crystalline Si (Jellison 3-1-91) [32] and a-Si:H(i) [33] are shown in the same graph.ICP a-SiO  :H(i) film shows excellent optical properties compared to a-Si:H(i) throughout the whole wavelength range.
upper limit of  eff can be calculated using (3).From the same equation, it can be shown that the lower the  eff , the better the surface passivation quality.From Figure 11 it can be concluded that the a-SiO  :H(i) thin film reported in this paper has similar high-quality passivation quality compared to other schemes, despite the usage of Cz wafers in this work.It should be noted that the reported lifetimes in this work are as-deposited values.
Comparing the passivation quality of the a-SiO  :H(i) film in this work to that of Mueller et al. [16,18], which appears to be the record lifetime on low-resistivity wafers at the time of reporting, the ICP a-SiO  :H(i) films deposited in this work yield a slightly better passivation quality which may be attributed to the lower damage remote ICP deposition.The optimized a-SiO  :H(i) passivation layer features a very high lifetime of over 4.7 ms at an injection level of 10 15 cm −3 with an equivalent  eff of below 1.7 cms −1 and an implied  oc of 741 mV.This passivation quality is excellent considering the wafer type that was used.Therefore, the ICP a-SiO  :H(i) has demonstrated great potential to be used in HET solar cells.

Optical Properties.
As a passivation layer in HET solar cells, the layer should not only have excellent electronic passivation quality but also be highly transparent so as to reduce optical absorption in the layer.The better optical property also allows the use of slightly thicker passivation layer to enhance the passivation quality.Therefore, material with high bandgap with low absorption coefficient is preferred.
We have demonstrated excellent passivation quality of our ICP a-SiO  :H(i) thin film in the previous section.An ideal bandgap of about 2 eV is also illustrated.A comparison of  of our film and other materials is made in Figure 12.  (20 nm) hardly degrades even after one week of storage.As a comparison, lifetime degradation data of ICP a-Si:H(i) [24] (11.1 nm) is superposed in the same graph.The a-Si:H(i) passivation layer shows constant degradation of lifetime immediately after deposition.
Comparing to the standard a-Si:H(i) film, ICP a-SiO  :H(i) has much lower  across the whole energy range in Figure 12, showing excellent optical property.This might be due to suppressed band-to-band absorption and subband absorption, benefiting from the high bandgap and the low defect density in the film.Therefore, the ICP a-SiO  :H(i) deposited in this work has superior optical properties.

Stability.
In this section, we investigate the stability of the passivation layer by storing the samples in dark environment and monitoring the passivation quality of the films in regular intervals.The first lifetime measurement was usually conducted two minutes after deposition of the film.The subsequent measurements were made once every 24 hours, throughout a period of one week.The results are shown in Figure 13.
It can be observed that the ICP a-SiO  :H(i) film (20 nm) in the as-deposited state shows negligible degradation after one week of storage, as reflected by the almost constant  eff over time.On the other hand, the ICP a-Si:H(i) from Pysch et al. shows severe degradation after two days of storage.The relatively more stable behaviour of ICP a-SiO  :H(i) might be due to (a) slightly thicker film in this work that shields the delicate interface from the environment and (b) a less sensitive interface against degradation due to the incorporation of O atoms during deposition.Therefore, it can be concluded that the passivation film developed in the present work is robust against degradation caused by exposure to the ambient environment.It should be noted that passivated wafers are usually sent for subsequent processing (doped layers and TCO depositions) immediately after deposition.Long-time exposure to the ambient environment is rarely observed in industrial processes.Thus, the stable ICP a-SiO  :H(i) will definitely maintain its excellent passivation quality during the short time periods between the different processes.The same stable behaviour is also reported by Larionova et al. [67], however, only with a SiN  capping layer.

Conclusion
In this paper, we presented a novel passivation scheme featuring low-temperature ICPECVD a-SiO  :H(i) thin films.By tuning the CO 2 partial pressure and process temperature, the thin film was optimized in terms of passivation and optical quality.An excellent as-deposited lifetime on solargrade highly doped -type Cz wafers of over 4.7 ms with  eff less than 1.7 cm s −1 was obtained.At the same time, the film also had high bandgap and low absorption, showing excellent optical properties as well.Comparing to typically used a-Si:H(i), the ICP a-SiO  :H(i) shows superior stability against degradation after being exposed to ambient environment.We propose the ICP a-SiO  :H(i) as an alternative candidate for high-quality surface passivation layers in HET solar cell applications.

Figure 1 :
Figure 1: (a) Illustration of processing unit structure of the SINGULAR-HET.The arrows indicate the direction for carrier rotation and transfer.(b) Illustration of the process cycle.The horizontal arrow indicates processing order and the vertical arrows represent the direction of deposition in each station.(c) Illustration of process station and plasma design for plasma stations in downward deposition configuration.After[25,26].

Figure 2 :
Figure 2: (a) a-SiO  :H(i) thin film thickness as a function of deposition time; (b)  eff as a function of a-SiO  :H(i) film thickness.The inset shows the sample structure used in lifetime determination.The lines are guides for the eye.A constant deposition rate is observed for all samples in this series.Lifetime saturation is observed after 20 nm film deposition.

Figure 3 :
Figure 3: Lifetime as a function of CO 2 partial pressure while keeping other deposition conditions constant.The inset shows sample structure used in measuring lifetime.The line is a guide to the eye.A lifetime peak is observed at CO 2 partial pressure of 10%.

Figure 4 :
Figure 4: FTIR transmission spectra of a-SiO  :H(i) films under different   .Higher O content in the film is observed with increasing CO 2 partial pressure, with Si-H(SiO 2 ) being the dominant bonding configuration. H increases with O content at first due to O facilitated inclusion but reduces beyond the optimal   due to O induced effusion.Doublet at 845 cm −1 and 890 cm −1 shows evidence of low temperature processing.The lines are guides for the eye.

Figure 5 :
Figure5:  H and  as functions of   . H displays a peak at about 10% CO 2 partial pressure, while  is insensitive to   variation.The lines are guides for the eye.

Figure 6 :Figure 7 :
Figure 6: Depth resolved (a) O and (b) C concentrations obtained from TOF SIMS.The respective elemental content in a-Si host is calculated based on standard c-Si atomic concentration.Both O and C concentrations show increasing trend with respect to CO 2 partial pressure during deposition.The highest C content achievable in this work is about 0.1 at.%.

Figure 8 :
Figure8: FTIR transmission spectra of a-SiO  :H(i) films with different process temperatures.O in the film tends to be bonded in higher order configuration with increasing temperature.The diminishing peak at 640 cm −1 and 2000 cm −1 indicates the effusion of H atoms.The disappearance of doublet at 845 cm −1 and 890 cm −1 indicates the optimised temperature.The lines are guide to the eye.

Figure 10 :
Figure 10: Optical bandgap as a function of process temperature.The line is a guide for the eye.A linear dependence of bandgap on temperature can be observed.

Figure 12 :
Figure 12: Absorption coefficient as a function of photon energy.For comparison, the absorption coefficients of crystalline Si (Jellison 3-1-91)[32] and a-Si:H(i)[33] are shown in the same graph.ICP a-SiO  :H(i) film shows excellent optical properties compared to a-Si:H(i) throughout the whole wavelength range.

Figure 13 :
Figure13: Effective lifetime as a function of time after deposition.The insets show sample structures and dimensions for both materials used in the degradation study.The passivation quality of a-SiO  :H(i) film (20 nm) hardly degrades even after one week of storage.As a comparison, lifetime degradation data of ICP a-Si:H(i)[24] (11.1 nm) is superposed in the same graph.The a-Si:H(i) passivation layer shows constant degradation of lifetime immediately after deposition.

Table 1 :
Plasma parameter variation used in this paper.