Advances in surface passivation of c-Si solar cells

In order to avoid an unacceptably large efficiency loss when moving towards thinner silicon materials, the near-term challenge in the c-Si PV industry is to implement an effective passivation method for both cell surfaces. This paper discussed several suitable passivation schemes available. While the efficiency potential of industrially produced thin film poly-Si cells on foreign substrates cannot yet reliably be predicted, it is clear that wafer-based c-Si solar cells will allow to maintain (or even improve) today’s efficiency levels while at the same time reducing the consumption of (expensive) crystalline silicon by up to 50 %. Given the trend towards these Si materials, the most promising surface passivation methods are identified to date. The key issues to be considered are cost-effectiveness, added complexity, additional benefits, reliability, and efficiency potential. The efficiency increase for best cells is around 0.5–0.6 %abs and the current efficiency potential already demonstrated for all technologies is around 19.0 %. Average efficiencies in industrial mass production for selected technologies are 18.5–18.6 % for Cz and 17.1 % for mc-Si.


Introduction
The global photovoltaic (PV) market is booming over the years. Solar cell producers have to concentrate on the cost per watt peak (Wp) of their product, especially as the competition between the manufacturers becomes more and more important. Therefore, most cell manufacturers try to optimize their solar cell processes concerning higher efficiencies while not significantly increasing costs. There are four different ways to improve c-Si solar cell efficiency: • Increased light trapping effect by improving surface structuring and texturization. • Redistributing the emitter profile on the front surface.
• Upgrading or changing metallization processes to get thinner contacts with excellent electrical properties. • Optimizing the passivation layer on both surfaces to reduce the recombination losses.
This study focused on the present status of passivation of p-type c-Si solar cells with a special emphasis on passivation by AL 2 O 3 due to its superiority over other passivation methods available.
The surface of c-Si represents the largest possible disturbance of the symmetry of the crystal lattice and hence, due to non-saturated ('dangling') bonds, a large density of defects ('surface states') within the bandgap exists at the surface of the crystal. These dangling bond defects can also be divided into intrinsic and extrinsic defects. There are typically additional processing-related extrinsic surface defects, for instance, due to dislocations or chemical residues and metallic depositions on the surface. These defects are highly prone to different recombination mechanisms, namely surface recombination, auger recombination and Shockley-Read-Hall (SRH) recombination (occurs via defects level within the bandgap) [1,2]. The recombination processes reduce the minority carrier life time and are responsible for lower conversion efficiency of Si solar cells. One way to keep the recombination loss at tolerable minimum is to passivate the silicon surfaces electronically [3,4].
High-efficiency silicon solar cells require both surfaces (front and rear) to be well passivated. In this regard, electrical surface passivation of crystalline silicon (c-Si) solar cells has emerged as a key issue involving both conversion efficiency and fabrication cost. A low thermal budget (B400°C) passivation technique that remains stable against ultraviolet photons of sunlight for longer period ([20 years) is considered to be the more industrially convenient passivation technique.
In the following paragraphs, we will discuss the advances in surface passivation for high-efficiency c-Si solar cells.

c-Si solar cell surface passivation techniques
Surface passivation has become an essential factor for translating high-efficiency crystalline silicon solar cell concepts into industrial production schemes. In photovoltaics, a widespread method to determine the surface recombination is to measure the effective charge carrier lifetime from the photoconductance of symmetrically passivated silicon wafers in transient or quasi-static mode [5]. An excellent electrical interface quality is essential for many devices relying on the bulk electronic properties of semiconductors [6]. Electrical losses at a semiconductor interface or surface should be minimized in photonic devices based on group III-V or group IV semiconductors when radiative recombination should be the dominant process [7,8]. Moreover, electronic losses at the crystalline Si (c-Si) surface have become increasingly important in the field of c-Si solar cells due to the trend toward thinner c-Si wafers used as base material. Consequently, the reduction of recombination losses at semiconductor interfaces is a prime concern for numerous semiconductor applications.
Recombination losses at a semiconductor interface or surface can be reduced by two different passivation strategies. As the recombination rate is directly proportional to the interface defect density, the first strategy is based on the reduction in the number of defect states at the interface. The interface defect density can be reduced significantly by the passivation of under coordinated atoms (dangling bonds) by, e.g., atomic H or by a thin dielectric or semiconductor film. This strategy is commonly referred to as chemical passivation. For example, the midgap interface defect density of c-Si can be as low as 1 9 10 9 eV -1 cm -2 after the growth of a high-quality thermal SiO 2 film and a subsequent alneal in a H 2 atmosphere, e.g., a FGA [9].
The second strategy to reach surface passivation is based on a significant reduction of the electron or hole concentration at the semiconductor interface by means of a builtin electric field. As recombination processes require both electrons and holes, the highest recombination rate is obtained when the electron and hole concentration at the interface are approximately equal in magnitude (assuming identical capture cross sections for electrons and holes). In other cases the recombination rate scales with the minority carrier concentration at the surface. In the so-called fieldeffect passivation, the electron or hole concentration at the semiconductor interface is altered by electrostatic shielding of the charge carriers through an internal electric field present at the interface. This internal electric field can either be obtained by a doping profile below the interface or by the presence of fixed electrical charges at the semiconductor interface. Consequently, the application areas of field-effect passivation are limited, but the effect can be employed successfully in devices such as light emitting diodes and solar cells [10].
There are two fundamentally different types of surfaces in a c-Si solar cell: metallized and non-metallized surfaces. Metal silicon interfaces feature very high SRVs and hence need to be carefully designed to avoid excessively large recombination losses. Similarly, to ensure a good blue response of the cell, the illuminated non-metallized surface regions need to be well passivated and not too heavily doped to avoid the formation of a dead layer. In the case of laboratory c-Si cells, the importance of the passivation of both cell surfaces is well recognized.

Fundamental theory of surface recombination
The surfaces or interfaces of a silicon substrate represent a severe discontinuity in its crystalline structure. The large numbers of partially bonded silicon atoms give rise to many dangling bonds, and therefore, a large density of defect levels are found within the bandgap near the semiconductor surface. Even if the silicon surface is not bare, say due to a native oxide, the presence of silicon-oxygen bonds can strain the crystal structure at the surface, which again introduces many defect states.
For a single defect at the surface, the rate of surface recombination, U s is given by [11] where n s and p s are the concentrations of electrons and holes at the surface, and S p0 and S n0 are related to the density of surface states per unit area, N ts , the capture cross-sections, r n and r p , and thermal velocity, m th for the specific defect S n0 r n m th N ts and S p0 r p m th N ts ð2Þ In reality, defect levels are so numerous that they can be considered to be continuously distributed throughout the bandgap, and both their density and capture cross-sections will be dependent on their energy level. Similar to the definition of the recombination lifetime sðDnÞ ¼ Dn

UðDnÞ
, the surface recombination velocity, S is defined as where Dn s is the excess carrier density at the surface. Hence, the surface recombination velocity can be related to the fundamental properties of the surface defects through (3). For the case of flatband conditions at the surface and negligible carrier trapping, the excess densities of electrons and holes are equal Dn s ¼ Dp s ð Þ , giving [3] SðDn s Þ ¼ n 0 þ p 0 þ Dn s n 0 þn 1 þDn s S p0 It is the surface recombination velocity, S, that is typically used for quantifying surface recombination processes. In general, the surface passivation is controlled by the reduction of interface defect states (chemical passivation) and by shielding of minority charge carriers from the surface (fieldeffect passivation). This is illustrated in Eq. 1. The driving force of the surface recombination process is the term ðn s p s À n 2 i Þ which describes the deviation of the system from the thermal equilibrium under illumination. The relative importance of either chemical passivation or field-effect passivation depends on the passivation scheme used.

Front surface passivation
Front surface passivation was attained with thermal SiO 2 and TiO 2 fabricated by atmospheric chemical vapor deposition. Thermal growth of silicon oxide is the most effective c-Si surface passivation technique for solar cell, but this technique is not suitable for low-cost industrial processes as it requires very high temperature (above 1,000°C) treatment. High temperature degrades the bulk life time significantly as well as the stability of passivated surface [12]. Moreover, SiO 2 has very low refractive index which is not suitable for good antireflection performance. On the other hand, TiO 2 does not provide any electronic surface passivation. In the past two decades, various research efforts have been devoted to the development of a more industrially convenient solution, in particular, on passivation layers deposited at low temperature (B400°C) onto the c-Si surface. SiN x front side passivation using plasma-enhanced chemical vapor deposition (PECVD) has been evolved as an alternative to high-temperature oxidation of silicon [13][14][15]. Record low effective surface recombination velocity (SRV) of 4 cm/s has been obtained on 1 X cm p-Si wafers for SiN x using remote PECVD or high-frequency (13.56 MHz) direct PECVD [14]. SiN x passivation for front surface of c-Si solar cells is deemed to be superior over other passivation techniques such as SiO 2 , TiO 2 , etc. due to its (1) field effect passivation provided by positive interface charges, (2) properties of capture crosssections of dominant defects, (3) adjustable refractive index as antireflection coating (ARC) and (4) hydrogenated passivation of bulk defects [3,12,16,17].

Rear surface passivation
Usage of thinner wafers and an effective reduction of surface recombination losses are increasingly important for low-cost highly efficient silicon solar cells. Hence the recent trend in silicon-wafer-based PV industries is toward thinner wafers (\200 lm). But the use of thinner wafers increases the rear surface recombination [12] and consequently necessitates a well-passivated rear surface to keep the rear surface recombination loss in acceptable minimum. Moreover, the increasing demands for optical quality require a higher internal reflection as well. Therefore, all designs for high-efficiency silicon solar cells today use a dielectric passivation layer on the rear to meet the requirements of passivation quality and optical properties. Keeping in mind the above attributes a good number of rear surface passivation techniques were developed over the years; among them atomic layer deposition (ALD)-assisted Al 2 O 3 , wet oxidation process using pyrogenic steam, passivation by stack of a-Si:H/SiO 2 or Al 2 O 3 /SiN x and passivation by phosphorus-doped a-SiC x N y :H(n) alloys are the most notable rear side passivation techniques for crystalline silicon solar cells. Recent research results on these rear surface passivation techniques are aggregated below.

Rear passivation by Al 2 O 3
In the case of mc-Si wafers, thermal processes above 900°C typically lead to a significant degradation of bulk lifetime [18]. Al 2 O 3 rear passivation by ALD has been evolved as an alternative of low-temperature surface passivation for high-efficiency silicon solar cells which have comparable properties as that of the alnealed SiO 2 [19][20][21][22][23][24]. The excellent level of surface passivation by Al 2 O 3 films is reflected by SRVs \5 cm/s for low-resistivity p-type c-Si, after alnealing at 400°C in N 2 [22,[25][26][27]. A detailed study on ALD deposited Al 2 O 3 has been reported by Schmidt et al. [28].
Recently, it was demonstrated that an exceptional high level of c-Si surface passivation can be achieved for SiO 2 synthesized at low temperatures when combined with a very thin Al 2 O 3 capping film [29]. The passivation was virtually independent of the SiO 2 thickness and showed an exceptional firing stability. C-V measurements revealed that the SiO 2 /Al 2 O 3 stacks yielded a high level of chemical passivation after alnealing due to a low interface defect density. Preliminary results using an alternative method to synthesize the low-temperature SiO 2 films, and (similar) results for thermally grown SiO 2 [30], have suggested that the effective hydrogenation under influence of the Al 2 O 3 capping film may be general. Moreover, preliminary data on the long-term stability showed a low S eff (\4 cm/s) for a stack measured after 250 days. The excellent efficiency values (up to 21.3-21.5 %), the high voltages (680 mV), the excellent long-wavelength quantum efficiency, and the high short-circuit currents of these cells (40 mA/cm 2 ) are a proof for the low rear-SRV and excellent internal rearsurface reflection [31]. Moreover, Al 2 O 3 passivation overcomes the so-called 'parasitic shunting' effects which is very common for SiN x passivated rear [32]. Low-temperature-synthesized SiO 2 /Al 2 O 3 passivation schemes can provide a realistic alternative for thermally grown SiO 2 in next-generation solar cells.
The key differentiator of Al 2 O 3 in comparison with other passivation schemes is its high fixed negative charge density Q f (10 12 -10 13 cm -2 ) located at the Al 2 O 3 /Si interface which produces effective field-effect passivation by shielding electrons from the interface [33][34][35]. The origin of the negative Q f in Al 2 O 3 deposited on c-Si has been attributed to intrinsic and extrinsic defects in Al 2 O 3 . Matsunaga et al. [36] calculated the energetics of intrinsic vacancies and interstitials in Al 2 O 3 from first principles. These calculations showed that each intrinsic point defect is most stable in their fully ionized form. Hence, Al vacancies and O interstitials exhibit a negative charge and Al interstitials and O vacancies exhibit a positive charge in good agreement with the ionic nature of the Al 2 O 3 [36]. Extrinsic H has also been proposed as origin for the negative fixed charges in Al 2 O 3 . Peacock and Robertson [37] calculated that interstitial H acts as a deep trap site for electrons in Al 2 O 3 . H is, for example, a common constituent in Al 2 O 3 synthesized by ALD because H-containing precursors such as Al(CH 3 ) 3 and H 2 O are used in the deposition process. Hence, Al vacancies, O interstitials, and interstitial H are proposed as the origin of the negative Q f in Al 2 O 3 . Based on the ionic nature of Al 2 O 3 , Lucovsky [38] postulated that Al 2 O 3 consists of tetrahedrally coordinated Al in AlO 4-units and octahedrally coordinated Al 3? in a ratio of 3:1 to assure charge neutrality. Kimoto et al. [39] demonstrated that both tetrahedrally and octahedrally coordinated Al are present in Al 2 O 3 grown by thermal ALD on H-terminated c-Si. However, tetrahedrally coordinated Al was found to be dominant at the interface. This dominance was attributed to the fact that Si in the interfacial SiO x film also has a tetrahedral coordination [39]. Consequently, the interfacial SiO x film could fulfill an important role in the origin of the negative Q f that is found in Al 2 O 3 films grown on c-Si by inducing a high-density of negatively charged Al vacancies close to the interface. This hypothesis is in good agreement with the location of the negative Q f extracted from thickness-dependent C-V measurements by various authors [40][41][42].
The Al 2 O 3 passivation mechanism has been elucidated by contactless characterization of c-Si/Al 2 O 3 interfaces by optical second-harmonic generation (SHG). SHG has revealed a negative fixed charge density in 10 11 cm -2 that increased to as-deposited Al 2 O 3 on the order of 10 12 -10 13 cm -2 upon alnealing, causing effective field-effect passivation. In addition, multiple photon-induced charge trapping dynamics suggest a reduction in recombination channels after alneal and indicate a c-Si/Al 2 O 3 conduction band offset of 2.02 ± 0.04 eV [43]. This makes Al 2 O 3 especially suited for the passivation of p-type silicon and p ? emitters, while Al 2 O 3 is expected to be less suitable for n-type (due to parasitic shunting) and n? type silicon (yet this still remains to be proven). The thermal stability of Al 2 O 3 during firing and the low-temperature deposition of the films are compatible with industrial solar cell processes [44]. Al 2 O 3 is transparent (bandgap *9 eV) and stable under UV illumination [45,46], but the refractive index of *1.64 makes it less suitable for single layer ARC on the front side as compared with a-SiN x . As a back reflector, however, Al 2 O 3 is superior. Also the fact that very thin Al 2 O 3 films (down to 5 nm) can be used is another benefit of Al 2 O 3 and enables flexibility in the design of surface passivation schemes and stacks [47]. As a consequence of these (unique) properties, Al 2 O 3 is now considered an important candidate for surface passivation in industrial solar cells, either for front and/or rear side passivation. A remaining challenge in this respect is the development of deposition tools for Al 2 O 3 films compatible with the requirements, especially in terms of throughput, of the solar cell industry [48,49].

Wet oxidized rear passivation
A wet oxidation process using pyrogenic steam at 800°C has been applied on 250 lm thick boron-doped (0.5 X cm) high-quality float zone (FZ) silicon to study its suitability for rear side passivation at Fraunhofer ISE [50,51]. In the long-wavelength region of incident photons, this wet oxidized rear passivation has high internal reflectance which leads to an absolute gain of 6 % in both J sc and V oc compared with Al-back surface field (BSF) cells though there is about 3 % absolute loss in FF due to increase in series resistance for point-contacts [52].
Wet oxides grown from purified steam have been demonstrated to provide a high level of surface passivation that at least is comparable to wet oxides grown by the standard process, where the water vapor is introduced by the pyrolytic generation from high-purity gases (H 2 , O 2 ) [53]. The rear side of laboratory type laser-fired contact (LFC) solar cells has been passivated by a wet oxide grown from purified steam at a temperature of 850°C. As a reference also a standard wet oxide grown at the same temperature as well as a dry oxide grown at 1,050°C has been applied for the passivated emitter rear cell (PERC) type solar cells. For all cells, conversion efficiencies well above 20 % could be reached. The average open-circuit voltages for the different rear side passivation layers lie in a very close range of 648.4 mV for the wet oxide grown from purified steam to 650.6 mV for the dry oxidation. The rear SRV that was extracted from the measured quantum efficiency for all oxides is in the range of 80-100 cm/s. Thus, the wet oxides grown from purified steam seem to be well suited for the rear side passivation of PERC-type solar cells. The one-sun parameters of the PERC-type solar cells featuring the rear side passivation by the different dry and wet thermal oxides are summarized in Table 1.
In addition, this wet oxidation process has about one order of magnitude faster growth rate than the traditional high-temperature (1,050°C) dry oxidation process and can close the gap between laboratory and industrial application by overcoming the longer high thermal budget of dry oxidation [54].
Rear passivation by stack of a-Si:H/SiO 2 A PECVD double layer of amorphous silicon-rich oxynitride and amorphous silicon nitride (SiN x ) can provide a nearly perfect passivation after subsequent rapid thermal process (RTP) and light soaking. The resulting effective minority carriers' lifetime (s eff ) is close to the modeled maximum on p-type as well as on n-type c-Si. Restrictions on the RTP of passivated surfaces, typical of other common passivation schemes (e.g., amorphous Si), are relieved by this double layer. Harsher thermal treatments can be adopted while still obtaining salient passivation. It is shown that the strong quality of surface passivation is ensured by a mechanism that emits electrons from shallow donor states in the passivation layer system and therefore creates a positive field effect [55].
Very low SRVs \6 and \11 cm/s were obtained for SiO x /a-SiN x :H stacks synthesized by PECVD on lowresistivity n-and p-type c-Si, respectively. The stacks induced a constant effective lifetime under low illumination, comparable to Al 2 O 3 on p-type Si. Compared with single layer a-SiN x :H, a lower positive fixed charge density was revealed by SHG measurements, while field-effect passivation was absent for a reference stack comprising thermally grown SiO 2 . The results indicate that hydrogenation of interface states played a key role in the passivation and remained effective up to alnealing temperatures [800°C [56]. The best achieved energy conversion efficiency of 21.7 % for c-Si solar cells (that was confirmed by the Fraunhofer ISE CalLab) was reported in [57,58].

Passivation by phosphorus-doped a-SiC x N y :H(n) alloys
Amorphous silicon nitride (a-SiN x :H) deposited by PEC-VD was considered as the most feasible alternative to thermally grown silicon oxide on c-Si surface due to its ability to serve as both passivation and anti-reflection coating. Recently, a study conducted on amorphous silicon carbide (a-SiC x :H) proved that it could be the best alternative to a-SiN x :H since its optical characteristics can be tuned by the carbon content of the film [14,59].
Phosphorus-doped amorphous silicon carbide (a-SiC x : H(n)) demonstrated the ability to passivate p-type c-Si substrates and highly doped n-type emitters in solar cells [60][61][62][63]. Hydrogenated amorphous silicon carbide (a-SiC:H) provides excellent electronic surface passivation for crystalline silicon solar cells. The hydrogen and carbon content of the passivation layers control the surface passivation depending on hydrogen bonding and alnealing temperature [64,65]. The carbon content C c of the amorphous layers varies depending on the methane-to-silane gas flow ratio during deposition. The electronic passivation quality exhibits best thermal stability for an optimum C c = 2.3 at. %. Alnealing this sample under forming gas atmosphere up to TFG = 550°C enables excellent effective minority carrier lifetimes of 1.2 ms [66]. Hydrogen effusion measurements relate this result to an increase in H-content with rising C c and to a simultaneous shift of the effusion peaks to higher temperatures [65,67]. Higher carbon content reduces the diffusion of atomic hydrogen out of the amorphous layers. The Si-H bonding configurations in the amorphous layers, analyzed from infrared absorption spectroscopy, reveal that a-SiC:H layers with lower carbon content have a higher density. Increasing C c induces voids and microvoids in the amorphous structure, favoring the diffusion of molecular hydrogen out of the a-SiC:H layers [68,69]. Very silicon-rich films yielded effective SRVs at 1 sunillumination as low as 3 and 2 cm s -1 on 1 X cm p-and n-type crystalline silicon substrates, respectively [70]. An efficiency of 20.2 % was reported for PERC following the approach in [34,63]. The reported values had been obtained without forming gas alneal (FGA) and the stability in surface passivation had been verified for a 2-month period. Such stacks have the additional benefit of being more dielectric than silicon-rich films, thus avoiding the degradation of field effect passivation due to shunting effects [32]. This result demonstrates that a-SiC x N y : H(n) films can be useful for designing high-efficiency c-Si solar cells.

Passivation by Al 2 O 3 /SiN x -layer stack
Despite the merits of Al 2 O 3 , it was shown that the samples which are passivated by a single layer of Al 2 O 3 without the protecting SiN x layer show a largely destroyed Al 2 O 3 layer after metallization etch-back [71]. Investigations on lifetime samples showed a 2.5-fold increase in effective lifetime and an efficiency gain of 0.7 % absolute (max. efficiency 18.6 %) for surfaces passivated by an Al 2 O 3 / SiN x stack compared with fully metallized Al-BSF rear sides. An increase in the infrared spectrum of the internal quantum efficiency is determined as the source of this gain. However, this cell exhibited a reduction in FF due to an increase in series resistance locally. For Al 2 O 3 and Al 2 O 3 / a-SiN x :H stacks on 2 X cm n-type c-Si, ultralow SRVs of S eff \ 3 cm/s were obtained and the passivation proved sufficiently stable (S eff \ 14 cm/s) against a high-temperature ''firing'' process ([800°C) used for screen printed c-Si solar cells [72].

Discussion and outlook
Surface recombination is very effectively suppressed by means of silicon dioxide (SiO 2 ) grown in a high-temperature (C900°C) oxidation process. Very low SRVs are in particular realized at the lightly doped rear surface, where the combination of a thermally grown SiO 2 layer with an evaporated film of Al give-after an additional alnealing treatment at 400°C (the so-called 'alneal')-SRVs below 20 cm/s on un-metallized low-resistivity (*1 X cm) p-type silicon wafers. In addition, the SiO 2 /Al stack at the cell rear acts as an excellent reflector for near bandgap photons, significantly improving the light-trapping properties and hence the short-circuit current of the cell.
One of the main reasons why high-temperature oxidation has not been implemented into the majority of industrial cell processes up to now is the high sensitivity of the silicon bulk lifetime to high-temperature processes. In particular in the case of multi-crystalline silicon wafers, thermal processes above 900°C typically lead to a significant degradation of the bulk lifetime. Hence, low-temperature surface passivation alternatives are required for future industrial high-efficiency silicon solar cells, which should have comparable properties as the alnealed SiO 2 .
One intensively investigated low-temperature surface passivation alternative to thermal oxide is silicon nitride (SiN x ) grown by PECVD at *400°C, which has proven to give comparably low SRVs as thermal SiO 2 on low-resistivity p-type silicon [14,73]. However, when applied to the rear of passivated emitter and rear cell (PERC)-type solar cells the short-circuit current density is strongly reduced compared with the SiO 2 -passivated cell rear [74]. This effect has been attributed to the large density of fixed positive charges within the SiN x layer, inducing an inversion layer in the crystalline silicon underneath the SiN x . The coupling of this inversion layer to the base contact leads to a significant loss in the short-circuit current density. This detrimental effect is known as 'parasitic shunting' [32]. Another alternative low-temperature passivation scheme resulting in comparable SRVs as alnealed SiO 2 is intrinsic hydrogenated amorphous silicon (a-Si) deposited by PECVD in the temperature range between 200 and 250°C [75]. Despite the fact that no parasitic shunting occurs in the case of a-Si passivated cell rear, new problems arise from the high sensitivity of the a-Si passivation to thermal processes. More recently, it was shown that thin films of aluminum oxide (Al 2 O 3 ) grown by ALD provide an excellent level of surface passivation on lowly doped (*1 X cm) p-and n-type silicon wafers [25,28,76] as well as on highly boron-doped p? emitters [75]. Figure 1 shows a comparison of the injection dependent effective SRVs measured on 1-2 X cm p-type silicon wafers passivated by (1) SiN x deposited by remote-PEC-VD [78], (2) intrinsic a-Si deposited in a parallel-plate PECVD reactor [75], and (3) Al 2 O 3 deposited by means of plasma-assisted ALD [79]. The measurements are shown in the injection range between 10 12 and 10 15 cm -3 , which is the most relevant range for the rear of one-sun silicon solar cells. Within this injection range it becomes obvious from Fig. 1 that the Al 2 O 3 passivation outperforms the welloptimized SiN x as well as the a-Si passivation. Importantly, the Al 2 O 3 passivation results in a very weak injection-level dependence in the injection range between 10 12 and 10 15 cm -3 , and the effective SRV stays well below 10 cm/s even at low injection densities \10 13 cm -3 . The strong injection-level dependence of the effective SRV measured on SiN x -passivated p-Si surfaces can be attributed to recombination within the space charge region induced by the high fixed positive charge density within the SiN x layer [73]. Figure 2 shows the measured effective SRV as a function of the corona charge density deposited onto dielectric passivation layers of SiN x and Al 2 O 3 . In the flatband case when the deposited Corona charge density equals the fixed charge density within the dielectric layer, the recombination rate shows a maximum. From this maximum the charge density within the layer can be deduced. As can be seen from Fig. 2, the fixed charge density within the SiN x layer is positive and amounts to Q f = ?2 9 10 12 elementary charges/cm 2 , whereas the fixed charge density within the Al 2 O 3 layer is negative and amounts to Q f = -1.3 9 10 13 cm -2 . The fixed negative charge density within the Al 2 O 3 layer induces an accumulation layer at the p-type silicon surface in contrast to the SiN x layer inducing and inversion layer. As a consequence, the Al 2 O 3 provides a very effective field-effect passivation without the additional contribution of recombination in the space charge region, which is the dominant recombination channel in the case of the SiN x passivation of p-type silicon surfaces at low injection densities [73,77]. In addition, due to the formation of an accumulation layer instead of an inversion layer at the p-type silicon surface, the abovementioned parasitic shunting effect at the solar cell rear is not expected for an Al 2 O 3 -rear-passivated cell. In combination with its very high transparency for near-bandgap photons, ALD-deposited Al 2 O 3 should hence be an optimal choice for a dielectric layer at the silicon solar cell rear.
In addition to the outstanding surface passivation provided by ALD-deposited Al 2 O 3 , the deposition process itself is also beneficial from an application point of view. In contrast to the conventionally applied PECVD, ALD consists of two self-limiting half-reactions, which implies several important advantages: (1) ALD gives highly conformal coatings, which allows to deposit and passivate, e.g., deep trenches or even pores in silicon, (2) pin-hole and particle-free deposition is achieved, (3) as ALD is a selflimiting process, uniform films can be deposited over large areas with mono-layer growth control, and (4) very low impurity concentrations of deposited films and hence very high film quality is achieved.
The main disadvantage of ALD for photovoltaic applications is its relatively low deposition rate. However, this disadvantage can be overcome by depositing ultrathin (2-30 nm) ALD-Al 2 O 3 films and capping them with a thicker film of, e.g., PECVD-SiO x , SiN x or SiC x [80]. Apart of the advantageous optical properties of these stacks, the passivation quality of such ALD/PECVD stacks can even be superior to that of single layers of Al 2 O 3 , which could be attributed to the hydrogenation of interface states at the Al 2 O 3 /Si interface during deposition of the hydrogen-rich PECVD layer [28]. Combination of ALD and PECVD might hence be a key technology for future industrial highefficiency solar cells. Fig. 1 Comparison of the injection-dependent effective SRVs S eff (Dn) measured on 1-2 X cm p-type floatzone silicon wafers passivated by (1) SiN x deposited by remote-PECVD [78], (2) intrinsic a-Si deposited in a parallel-plate PECVD reactor [75] and (3) Al 2 O 3 deposited by means of plasma-assisted ALD [79]. The lines are guides to the eye  Table 2.

Conclusions
Solar cells made from crystalline silicon have lower conversion efficiency and hence optimization of each process step is very important. Increasing the efficiency of crystalline silicon solar cells relies on the understanding and optimization of each individual processing step as well as of the interplay between the material properties and the processing conditions. Our focus was to review the advances in existing surface passivation techniques in an industrial process line as well as in the research laboratories over the world. Every c-Si solar cell fabricated to date features one or more of these surface passivation methods. With regard to solar cell applications, it is important to make sure that the surface passivation is long-term stable ([20 years) and stable against the UV photons of sunlight.
In order to increase the efficiency of silicon-wafer-based solar cells in production well above 20 %, it is indispensable to improve the currently applied level of surface passivation at the front as well as at the rear of the cells. In the past, low surface recombination velocities (\20 cm/s) have been achieved on low-resistivity (*1 X cm) p-type silicon using plasma-enhanced chemical-vapor-deposited (PECVD) silicon nitride (SiN x ) as well as amorphous silicon (a-Si). However, the high density of fixed positive charges within the PECVD-SiN x layer induces an inversion layer at the rear of p-type Si cells, producing a detrimental parasitic shunting, which reduces the short-circuit current density by up to 3 mA/cm 2 . The passivation quality of a-Si, on the other hand, is very temperature sensitive. More recently it has been shown that atomic-layer-deposited (ALD) aluminum oxide (Al 2 O 3 ) provides an outstanding level of surface passivation, which can be attributed to its extremely high negative fixed charge density in combination with the very gentle deposition technique ALD, leading to low interface state densities. The application of these ALD-Al 2 O 3 layers to the rear of p-type solar cells shows that this new passivation scheme is indeed suitable for high efficiencies and that due to the large negative fixed charge density no parasitic shunting occurs. In a direct comparison with other passivation schemes, it is found that Al 2 O 3 even outperforms optimized thermally grown SiO 2 and opens the possibility of achieving very large opencircuit voltages up to V oc = 740 mV.
However, whenever a particular process is required to be optimized, the amount of extra steps should be kept to an absolute minimum and ideally the general cell line concept should not be changed drastically to make the approach cost-effective and easy to implement.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.