Al2O3 stacks for the simultaneous passivation of n+ and p+ doped silicon surfaces by atomic layer deposition

Solar Energy Materials & Solar Cells 143 (2015) 450–456

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells journal homepage: www.elsevier.com/locate/solmat

“Zero-charge” SiO2/Al2O3 stacks for the simultaneous passivation of n þ and p þ doped silicon surfaces by atomic layer deposition B.W.H. van de Loo a,n, H.C.M. Knoops a, G. Dingemans b, G.J.M. Janssen c, M.W.P.E. Lamers c, I.G. Romijn c, A.W. Weeber c, W.M.M. Kessels a a

Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands ASM, Kapeldreef 75, B-3001 Leuven, Belgium c ECN Solar Energy, P.O. Box 1, NL-1755 ZG Petten, The Netherlands b

art ic l e i nf o

a b s t r a c t

Article history: Received 13 May 2015 Received in revised form 23 July 2015 Accepted 27 July 2015

To achieve high conversion efficiencies, advanced silicon solar cell architectures such as interdigitated back contact solar cells demand that defects at both the n þ and p þ doped Si surfaces are passivated simultaneously by a single passivation scheme. In this work, corona charging experiments show that the fixed charge density Qf is a key parameter governing the passivation of both surface types. Alternatively, Qf can be controlled from strongly negative to even positive values by carefully tuning the SiO2 interlayer thickness in SiO2/Al2O3 stacks prepared by atomic layer deposition (ALD). This control in Qf allows for a superior passivation of n þ Si surfaces by SiO2/Al2O3 stacks compared to a single layer Al2O3. For instance, for SiO2 interlayer thicknesses of  3–14 nm, the recombination parameter of an n þ Si surface having a high surface doping concentration Ns of 2  1020 cm  3 was reduced from J0n þ ¼ 81 fA/cm2 to J0n þ ¼50 fA/cm2. Simulations predict that the SiO2/Al2O3 stacks outperform Al2O3 passivation layers particularly on n þ Si surfaces having a moderate Ns in the range of 1018–1020 cm  3. On p þ Si surfaces, J0p þ ≤ 54 fA/cm2 was achieved for all ALD SiO2 interlayer thicknesses investigated (i.e., 1–14 nm). The SiO2/Al2O3 stacks presented in this work are compatible with SiNx capping and subsequent high-temperature firing steps, which are typically used in solar cell processing. Furthermore, the results were successfully reproduced in an industrial ALD batch reactor using a low-temperature process. This makes ALD SiO2/Al2O3 stacks a promising candidate for the simultaneous passivation of n þ and p þ Si surfaces in solar cells. & 2015 Elsevier B.V. All rights reserved.

Keywords: Surface passivation Atomic layer deposition (ALD) SiO2 Al2O3 Emitter Back surface field (BSF)

1. Introduction A high level of surface passivation is a prerequisite to achieve crystalline silicon solar cells with high conversion efficiencies. Less than a decade ago, it was reported that Al2O3 films prepared by atomic layer deposition (ALD) provide superior passivation of p and p þ -type Si surfaces, which was technologically challenging at that time [1–3]. The excellent passivation by Al2O3 can be related to a very low interface defect density Dit (o1011 cm  2) on Si, which is essential for chemical passivation. Furthermore, a high negative (fixed) charge density Qf in of the order of 1012–1013 cm  2 (depending on synthesis method) is present at the interface [4], which reduces the minority carrier (i.e., electron) concentration near the Si surface, providing field-effect passivation. Even an ultrathin film of Al2O3 of less than 2 nm was found to be sufficient to n

Corresponding author. Tel.: +31 40 247 4095. E-mail address: [email protected] (B.W.H. van de Loo).

http://dx.doi.org/10.1016/j.solmat.2015.07.040 0927-0248/& 2015 Elsevier B.V. All rights reserved.

passivate the Si surface, when combined with a-SiNx:H (in short SiNx) as anti-reflection coating (ARC) and/or capping layer [5,6]. The high levels of surface passivation provided by Al2O3 allow for solar cells with high conversion efficiency, including p-type concepts (e.g, PERC or Al–LBSF cells [7,8]) as well as n-type concepts such as PERL cells [9]. Due to superior uniformity and passivation performance, ALD of Al2O3 is currently piloted in industry [10]. Despite the excellent results on p-type Si, on heavily doped n þ Si surfaces the passivation of Al2O3 is compromised, as the minority carrier (i.e., hole) concentration at the surface is increased by the negative Qf of the Al2O3 [11,12]. On lowly doped n-type Si, Al2O3 generally passivates the surface well, as it induces strong near-surface inversion. However, inversion layers are associated with a lifetime reduction at low injection levels (Δn o1015 cm  3), which is within the operating regime of solar cells [13–15]. Secondly, besides affecting the surface passivation, inversion layers potentially act as undesired conduction pathways to metal contacts. Such a pathway (sometimes referred to as parasitic shunting) affects the performance of solar cells via the fill-factor, short-

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

circuit current and open-circuit voltage [16]. All of these unfavorable effects occur not only when n or n þ -type Si surfaces are passivated by the negative charge dielectric Al2O3, but they also take place when p or p þ type Si surfaces are passivated by a dielectric containing a positive charge density, such as SiNx. In conventional diffused-junction Si solar cells, the n þ and p þ doped surfaces are located at each side of the cell, and hence can be passivated independently by a dielectric having the right charge polarity. However, in more advanced solar cell architectures such as interdigitated back-contact (IBC) cells, or in even more advanced axial Si nanowire cells [17], the n þ and p þ doped Si surfaces are adjacent and preferably passivated simultaneously by one film or film stack. Hence, the aim of this work is to achieve the passivation of both n þ and p þ Si surfaces via a low-temperature and industrially viable deposition process. To this end, we propose the use of a passivation scheme without significant Qf, while providing very low Dit. Rather than relying on a fixed charge density to reduce one of the two charge carrier types at the surface, such a “zero-charge” passivation approach could rely on high surface doping concentrations of the n þ and p þ Si regions, which reduces the minority carrier concentration at the surface and in this way supresses surface recombination. In literature, there are several ways reported to manipulate the natively present fixed charge density of dielectrics and dielectric stacks, and some of them have been studied to address the simultaneous passivation of n þ and p þ Si surfaces [18–22]. For instance, to reduce the negative Qf of Al2O3, the films can be deposited at high deposition temperatures (i.e., 300–400 °C com e.g., 200 °C), although this adversely affects the level of chemical passivation [23]. Secondly, the negative Qf of thin Al2O3 films can be reduced when it is combined with SiNx capping and a subsequent high-temperature firing step Z 800 °C [21,25]. Such Al2O3/SiNx stacks show promising results on both n þ and p þ Si surfaces with various sheet resistances, irrespective of whether the Al2O3 films are deposited by PE-CVD [18], or ALD [21,22]. Nonetheless, these stacks exhibit a significant negative Qf of  (1– 2) × 1012 cm  2 [18,21,25], which is likely still affecting the passivation of n þ Si surfaces. Alternatively, it is known already for a long time that an interfacial SiO2 layer strongly reduces the fixed charge density of Al2O3 layers [26,27]. In fact, various interlayers can be used to reduce the Qf of Al2O3 layers to virtually zero. This includes SiO2 layers prepared by ALD [13,28] and by plasmaenhanced chemical vapor deposition (PE-CVD) [20], thermallygrown SiO2 [29], but also other materials such as HfO2 [30]. Moreover, when SiO2 is combined with other capping layers, such as in SiO2/SiNx or SiO2/Al2O3/SiNx stacks, very low charge densities can be obtained [7,29,31], although this strongly depends on the SiO2 thickness and process conditions [7,29]. To achieve passivation of n þ and p þ Si surfaces, for example wet-chemical or PE-CVD SiO2 films in combination with PE-CVD SiNx capping layers [19,22], and PE-CVD SiO2/Al2O3 stacks with varying SiO2 film thicknesses have been studied [20]. SiO2/Al2O3 stacks which are fully prepared by ALD are very promising candidates as “zero-charge” passivation schemes for n þ and p þ Si surfaces. It has been reported by Dingemans et al. [13] and Terlinden et al. [28] that those stacks exhibit excellent chemical passivation levels (e.g., Dit values o 1011 eV  1 cm  2 at mid gap [32]), while their Qf can effectively be tuned from strongly negative up to zero or even positive values by carefully tuning the ALD SiO2 thickness. It is hypothesized that for these stacks the SiO2 interlayer acts as a (trap-assisted) electron tunnel barrier, preventing charge-injection from the c-Si base into the electron trap sites in the Al2O3 [13,28]. The Al2O3 plays a key role in the reduction of Si/SiO2 interface defect states, which are effectively passivated by the diffusion of hydrogen from the Al2O3 film to the Si/SiO2 interface during annealing [33]. ALD SiO2/Al2O3 stacks have

451

the advantage that they can be prepared at relatively low temperatures (e.g., 200 °C), are conformal even over high-aspect ratio structures and can be prepared in a single deposition run [13,28,34]. The ALD stacks can potentially be used in high-volume manufacturing for the simultaneous passivation of both sides of IBC cells or bifacial cells. Moreover, very promising results containing the passivation of both sides of PERC cells have recently been reported for stacks having 5 nm ALD SiO2 films [35]. In previous work dealing with the passivation of n þ and p þ Si surfaces, the influence of the fixed charge density on the passivation was not fully addressed yet. In this work, first the influence of the charge density on the passivation of these surfaces is investigated via corona charging experiments. The results serve as a benchmark for the next part of the paper, where we will carefully tune the SiO2 interlayer thickness to control of the fixed charge density of ALD SiO2/Al2O3 stacks and will study its applicability as passivation scheme of both n þ and p þ Si surfaces. Next, the role of surface doping concentration on the passivation properties will be discussed. Finally, we will study the compatibility of ALD SiO2/Al2O3 films as “zero-charge” passivation scheme with typical Si solar cell processing, such as the compatibility with SiNx capping layers and high-temperature processing steps.

2. Experimental Symmetrical n þ /p/n þ lifetime samples were made by POCl3 diffusions in an industry-scale tube furnace process (Tempress systems TS81003) on double-side chemically polished, p-type Czochralski-grown (Cz) Si (100) wafers. In another tube of the furnace, symmetrical p þ /n/p þ lifetime samples were fabricated using BBr3 as precursor, on Cz n-type wafers with a random-pyramid textured surface. After diffusion, the surface doping concentration of the p þ /n/p þ sample was increased by a short wet etch [36]. The resulting (active) doping profiles were determined from electrochemical capacitance–voltage (ECV) profiling (see Fig. 1) using a WEP wafer profiler CVP21 table-top unit. Note, that the doping profile of the boron-doped samples can be affected by the texture of the surface, and a heavier doping level can be expected at the pyramid tips [37]. The sheet resistance and the homogeneity of the doped regions were measured using a Signatone four-point probe, in combination with a Keithley 2400 Source Measurement Unit. After glass removal, the samples received a short dip in diluted HF (1%, 1 min). Subsequently, SiO2/Al2O3 stacks and Al2O3 films were deposited on both sides of the samples by plasma-enhanced ALD in an Oxford Instruments OpALTM reactor at 200 °C in a single deposition run. H2Si(N(C2H5)2)2 and Al(CH3)3 were used as metalorganic precursors and O2 plasma as oxidant. More details on the ALD processes can be found elsewhere [34]. A schematic display of the passivated lifetime samples is given in Fig. 2. To activate the passivation, the stacks were either annealed for 10 min at 400 °C in N2, or capped with 70 nm SiNx followed by an industrial firing step (  800 °C for 1 s). The SiNx was prepared by in-line PE-CVD reactor (MAiA, Roth and Rau), equipped with a microwave plasma source. As deposited, the SiNx has a refractive index of n ¼ 2.05 at 633 nm. A lifetime tester (Sinton WCT 100) was used for quasi-steady state photoconductance decay (QSS-PC) and transient-PC measurements. Using QSS-PC the recombination parameter J0 was derived at high injection levels of Δn  (0.4–6.2)  1016 cm  3 using the method of Kane and Swanson [38], which allows the extraction of J0 from lifetime measurements via

1 1 1 (N + Δn)Δn − = + J0 · d 2 τeff τAuger τ SRH qni W

(1)

452

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

a

b

Fig. 1. Doping profiles as determined by electrochemical capacitance–voltage (ECV) profiling of (a) n þ /p/n þ samples of Rs ¼ 111 72 Ω/sq, and (b) p þ /n/p þ lifetime samples, having Rs ¼55 72 Ω/sq. Note, that with ECV only the active dopant concentration is measured. The sheet resistance was measured by four-point probe measurements.

Fig. 2. Schematic representation of the passivated n þ /p/n þ and p þ /n/p þ lifetime samples used in this work.

with τeff the effective lifetime, τAuger the (bulk) Auger lifetime, τSRH the Shockley–Read–Hall (SRH) lifetime of the bulk, ni the intrinsic carrier concentration, q the elementary charge, W the wafer thickness and Nd the base doping concentration. Importantly, it was verified that the method of Kane and Swanson could be used for all samples in this work, as straight lines were observed for the Auger-corrected inverse lifetime plots. Note, that J0 values of the p þ /n/p þ doped samples were divided by a factor of 1.7 to account for the increase in surface area due to the surface texture of these samples. Corona charging measurements were carried out by depositing (positive) corona charges on both sides of the lifetime samples which were passivated by Al2O3. The charges were deposited by applying a DC-bias ( 711 kV) between a tungsten needle and a grounded plate to ionize air molecules. After charge deposition step lifetime measurements were performed. Additionally, the deposited density of corona-charges Qcorona was derived from Kelvin probe potential measurements, using a Trek electrostatic voltmeter. Note that the deposited charge density on the Al2O3 surface was found to be stable for days, even when placing the charged surfaces directly onto a grounded metal.

3. Results 3.1. The influence of fixed charges on the passivation of n þ and p þ Si Before investigating the passivation quality of SiO2/Al2O3 stacks, it is illustrative to demonstrate the effect of fixed charges on the passivation of both n þ and p þ doped Si surfaces. To this end, the symmetrically doped lifetime samples were passivated by

a single-layer of Al2O3, after which lifetime measurements and positive corona charging on both sides were carried out in a stepwise fashion. In this way, the net charge density Qnet was varied from being strongly negative (i.e., Qnet ¼  5  1012 cm  2, which corresponds to Qf of the (ALD) Al2O3 films [39]) to positive values of þ 5  1012 cm  2. In Fig. 3a, the results of n þ /p/n þ samples are given. Prior to deposition, the level of surface passivation provided by the Al2O3 results in J0n þ ¼79 fA/cm2. When changing the net charge density via corona charging towards a more positive Qf, J0n þ decreases. The decrease in J0n þ can be understood as the more positive fixed charge will reduce the minority carrier (i.e., hole) concentration at the surface. A similar trend was observed in Ref. [11] for corona charging experiments on highly n-doped surfaces passivated by Al2O3. However, note that also opposite trends are observed in case of lowly doped n-type Si surfaces passivated by Al2O3 [39]. There, positive corona charging first results in an increase of surface recombination. For such lowly doped surfaces, an inversion layer is formed near the Si surface by the negative Qf of Al2O3, after which the surface is gradually moved to depletion and consequently accumulation by positive corona charging. Therefore, we can conclude that the highly-doped n þ Si surfaces used in this work were not in inversion, not even for the highest negative fixed charge densities. For high positive Qnet ¼4.9  1012 cm  2, the reduction of J0n þ saturated at J0n þ ¼ 50 fA/cm2, behavior which is observed more often for such high charge doses [11,40]. Simulations using the free-ware program EDNA [41], using Fermi–Dirac statistics, the Auger parameterization from Richter et al. [42] and the band-gap narrowing model of Schenk [43], indicate that the Auger limit of the n þ region is J0n þ ,Auger ¼16 fA/cm2. This significant difference ΔJ0 of 34 fA/cm2 can indicate the presence of other recombination processes in the n þ doped region, potentially due to SRH recombination via inactive phosphorus precipitates [44,45]. On the p þ /n/p þ doped samples in Fig. 3b, the opposite behavior in Fig. 3a is obtained when depositing positive corona charges, and a strong increase in J0p þ can be observed. On the p þ Si surface a negative Qf provides strong field-effect passivation, enabling very low J0p þ ¼ 32 fA/cm2 for Al2O3. EDNA simulations indicate that this value within error equal to the (lower) limit set by Auger recombination, i.e., J0p þ ,Auger ¼ 35 fA/cm2. Interestingly, the recombination parameter of the p þ /n/p þ sample is much more sensitive to the total fixed charge density than the n þ /p/n þ sample, as is evident from the drastic increase in J0p þ to  490 fA/cm2 at Qnet ¼5.7  1012 cm  2. This higher sensitivity to Qf of J0p þ compared to J0n þ can be explained by two

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

453

Fig. 3. Results for corona charging of Al2O3 passivated lifetime samples having Qf ¼  5  1012 cm  2 (a)–(b) and for samples passivated by SiO2/Al2O3 stacks with varying SiO2 thickness (d)–(e) (note the difference in scale of the y-axes). The J0 values were determined by QSS-PC, where J0p þ was corrected for the increased surface area due to surface texture. Lines serve as guide to the eye. In (c) and (f), the average J0 for both the n þ and p þ regions are calculated for several different area fractions An þ :Ap þ .

possible effects. First, the defects at the Si/Al2O3 interface (which essentially is a Si/SiO2/Al2O3 interface, as interfacial SiO2 is always present for ALD Al2O3 films on Si [4,13,46,47]) can have a higher capture cross-section for electrons than for holes (e.g., sn E102  sh [48,49]). As the capturing of holes is limiting the recombination at n þ Si surfaces, a lower capture cross-section aids in lower surface recombination for n þ Si than for p þ Si. Secondly, the doping profiles used (see Fig. 1) have an influence on the surface recombination, which has not been discussed so far. The exact interplay between the highly doped region and the surface recombination rate is rather complex, and will be treated in detail in a separate publication. Nonetheless, we can state here that the relatively low surface doping concentration of the boron-doped regions as compared to the phosphorus doped surfaces (i.e., Ns ¼9  1019 cm  3 versus 2  1020 cm  3, respectively) plays a key role in the increased sensitivity of J0p þ to surface passivation and charges. A route to further reduce the recombination at p þ Si surfaces in a case without fixed charges is therefore to increase the (boron-) surface doping concentration (provided that the interface properties such as Qf and Dit are independent of surface doping level), as will be further discussed in Section 3.3. The boron surface doping concentration could for instance be increased by etching of the boron-depletion region (BDR) which is typically present near the surface [36]. Finally, it is interesting to look at a case where n þ and p þ Si surfaces should be passivated by a single passivation scheme. Considering the fact that in an actual solar cell, the surface recombination also scales with the surface areas of the n þ and p þ region, i.e., Ap þ and An þ respectively, we have evaluated several

different cases using the interpolation of the corona charging results of Fig. 3a and b. In case of equal area fractions An þ ¼Ap þ (1:1), which, for instance, would be the case when both front and back-side of a bifacial solar cell are passivated at once, a value of Qnet ¼  4  1012 cm  2 would result in the lowest average J0. For ptype based IBC cells, it is common that An þ »Ap þ at the back side to prevent electrical shading, (or conversely, Ap þ »An þ for n-type Si based IBC cells [50]). As can be seen from Fig. 3c, the optimum charge density in terms of J0 shifts from Qnet ¼  4  1012 cm  2 for An þ ¼Ap þ (1:1), towards Qnet ¼0 for the fraction An þ ¼ 20  Ap þ (1:20). To summarize, it can be concluded that the fixed charge density is an important parameter when optimizing the simultaneous passivation of n þ and p þ doped Si surfaces. The optimum charge density will in practice vary from case to case, and is dependent on the doping profile and the area fractions of the n þ and p þ doped regions. In particular on p þ Si surfaces, the passivation shows a strong dependence on the charge density, most likely due to a moderate level of chemical passivation and a relative low surface doping concentration (i.e., below o1020 cm  3). 3.2.

Passivation of n þ and p þ Si surfaces by ALD SiO2/Al2O3 stacks

It is clear that the control in fixed charge density and polarity provides an additional parameter to optimize the passivation of both n þ and p þ Si surfaces. Next, we will investigate the possibility to optimize this passivation by precisely tuning the ALD SiO2 interlayer thickness of the SiO2/Al2O3 stack between 0 and 14.4 nm. It has been hypothesized, that an ALD SiO2 interlayer of

454

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

stacks also improves the chemical passivation compared to singlelayer Al2O3, as is also observed in literature by a slight decrease in Dit [34]. Further evidence for this improved chemical passivation can be found on the n þ Si surface, where the J0n þ values obtained from SiO2/Al2O3 stacks (for 3.6–14.4 nm ALD SiO2) are lower than what would be expected on the basis of corona charging of Al2O3 passivated samples with Qnet ¼0. To simulate the (average) recombination parameter in case both surface types would be passivated by SiO2/Al2O3 stacks, the results of n þ and p þ Si are combined in Fig. 3f. The results show an optimal ALD SiO2 thickness of 2–5 nm in case An þ ¼Ap þ (1:1), having a significantly lower average J0 as compared to single layer Al2O3 film (J0 ¼ 44 versus 55 fA/cm2, respectively). For An þ 4Ap þ , the optimal SiO2 thickness becomes less critical and  3–12 nm SiO2 is preferred. Fig. 4. Results from Atlas simulations for Gaussian n þ doping profiles, having the peak doping at the surface and having a depth of 0.1 mm. Values for J0,surface are given for three different fixed charge densities in the dielectric (i.e., Qf ¼ (  5,  3, 0)  1012 cm  2). As input for the simulations, the fundamental surface recombination velocities for electrons and holes were Sn0 ¼ 6500 cm/s and Sp0 ¼ 65 cm/s, respectively. In the simulations, Fermi–Dirac statistics were used, in combination with the band-gap narrowing model of Schenk [43], the Auger model of Dziewior and Schmid [55], and the Klaassen mobility model [56]. The bulk SRH lifetimes were τp0 ¼τn0 ¼ 1 ms.

2–4 nm thickness prevents the charge-injection from the Si into trap sites in the Al2O3 layers [28]. Therefore, the negative Qf of SiO2/Al2O3 stacks rapidly decreases towards virtually zero for these SiO2 thicknesses. It should be noted that also some positive charge formation in the SiO2 film can occur (this is a “bulk” charge, instead of “interface charge” as is the case fore Al2O3 films on Si). However, this charge density is relatively low (  1011 cm  2) for ALD SiO2 films and is therefore only partly contributing to the nullification of Qf the SiO2/Al2O3 stacks [28]. Yet, due to these positive bulk charges in of the SiO2 film, the overall charge polarity of the SiO2/Al2O3 stacks becomes slightly positive for increasing ALD SiO2 thicknesses (i.e., Qf changes from  5  1012 cm  2 for Al2O3 only to Qf ¼(4 72)  1011 cm  2 for SiO2/Al2O3 stacks with 12–16 nm thick SiO2 interlayer [28]). In Fig. 3d, J0n þ of the n þ /p/n þ samples passivated by single layer Al2O3 or SiO2/Al2O3 stacks are shown for varying SiO2 interlayer thickness. For all cases where ALD SiO2 was present (i.e., 1–14.4 nm), the recombination parameter J0n þ was reduced when compared to single-layer Al2O3 films. For instance, SiO2/Al2O3 stacks having an ultra-thin ALD SiO2 interlayer of 3.6 nm thickness, results in the significant improvement in J0n þ as compared to single layer Al2O3 from 81 to 50 fA/cm2. This improvement is in line with the expected transition from negative to positive charges, as a negative Qf was undesirable for the passivation of n þ Si surfaces, as also the corona charging experiments have pointed out. Interestingly, no further decrease in J0n þ was observed for SiO2 films with thicknesses 43.6 nm. The lowest values of J0n þ correspond well to the final J0n þ values after positive corona charging of 50 fA/cm2. This further supports the hypothesis that recombination in the n þ doped region itself is limiting J0n þ under these circumstances. On p þ /n/p þ samples, J0p þ gradually increases with increasing ALD SiO2 thickness from 32 fA/cm2 for single-layer Al2O3 to 52 fA/cm2 for 14.4 nm ALD SiO2 (Fig. 3e). This increase in J0p þ can be explained by the reduction in negative Qf, and accompanying loss in field-effect passivation. It is notable that the maximal J0p þ values obtained for SiO2/Al2O3 stacks (i.e., (J0p þ ≤ 54 fA/cm2 for all ALD SiO2 thicknesses investigated) are significantly lower than the charged samples passivated by Al2O3, having J0p þ ¼95 fA/cm2 for Qnet 0 cm  2. This could indicate that the introduction of ALD SiO2

3.3. Role of surface doping concentration Besides being dependent upon the relative areas of both regions, the optimal SiO2 thickness in ALD SiO2/Al2O3 stacks to passivate both n þ and p þ surfaces is also dependent upon the doping concentrations of both surfaces. Note that in this work, corona charging indicated that the n þ doped surfaces are depleted. Apparently, the high negative Qf of the Al2O3 is insufficient to induce inversion on these highly doped surfaces, having a phosphorus surface concentration of Ns  2  1020 cm  3. Using the device simulation package Atlas [51], the J0n þ values belonging to Gaussian-shaped n þ doping profiles having different surface doping concentrations and a fixed depth of 0.1 mm were calculated. Note that the commonly reported effective surface recombination velocity Seff typically varies with surface doping concentration [52]. However, Seff was not required as input parameter for the Atlas simulations. Instead, the simulations are based on the assumption that the interface input parameters Qf, Sn0 and Sp0 (being the fundamental surface recombination velocities for electrons and holes respectively) are independent of the surface doping concentration. For Sn0 and Qf of ALD Al2O3 films this was recently experimentally proven for boron-doped surfaces having Ns in the range of 9  1015–3  1019 cm  3 [53]. The simulations allow for the identification of the different contributions to the total recombination of the highly-doped region J0n þ , that is the contribution of Auger, surface, defect and radiative recombination

J0n + = J0, Auger + J0, surface + J0, SRH + J0, radiative

(2)

The results for J0,surface are presented in Fig. 4. It can be seen that for moderate surface doping concentrations (Ns  1017 cm  3), low J0,surface values are obtained for all Qf values investigated. However, in case of a negative Qf, the near surface region is strongly inverted which is why Al2O3 single layers will likely passivate very well. Nevertheless, as it is mentioned in Section 1, inversion conditions are undesirable in solar cells due to inversion layer recombination and potential parasitic shunting [15]. For higher n þ Si surface doping concentrations within the range of 1018–1020 cm  3, J0,surface is severely increased for dielectrics having a strong negative Qf (such as for Al2O3), which is in line with the experimental observations by others [11,12,21]. This strong surface recombination occurs when the n þ surface transitions from depletion to inversion, and the concentrations of electrons and holes at the surface (ns and ps, respectively) satisfy the condition spps ≈ snns [54]. Interestingly, the simulations predict that for this particular surface doping range excellent passivation can be maintained in case of Qf ¼ 0. Therefore, a strong gain in passivation for SiO2/Al2O3 stacks compared to single layer Al2O3 can be expected in the range of surface doping 1018–1020 cm  3. This may

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

be very relevant when SiO2/Al2O3 stacks are used to passivate a lightly doped n þ Si front surface of e.g., an IBC cell. Although moderate doping concentrations are often desirable to ensure a low Auger recombination (i.e., a low J0,Auger), it is insightful to consider the effects of very high surface doping concentrations Ns 41020 cm  3 on J0,surface. Interestingly, the Atlas simulations show a strongly reduced J0,surface for all fixed charge densities investigated. This effect can be understood as the high surface doping levels reduce the minority carrier (i.e., hole) concentration at the surface. For these very low J0,surfaces values, it can be expected from Eq. (2) that surface recombination will become negligible compared to the other recombination mechanisms in the highly doped region, such as Auger and SRH recombination. The results from the simulations can explain the excellent passivation results on highly-doped n þ Si surfaces by PE-CVD Al2O3/SiNx stacks, which are obtained despite of their negative Qf of (1–2)  1012 cm  2 [18], which now can (partly) be attributed to the very high phosphorus surface doping concentration used (Ns (4–6)  1020 cm  3). Overall, it can be expected that SiO2/Al2O3 stacks will outperform Al2O3 passivation schemes in a wide range of phosphorus surface doping concentrations from  5  1017 to  1020 cm  3 due to the absence of a negative Qf, which is the range of practical interest for most solar cell applications. 3.4. Aspects related to application of ALD SiO2/Al2O3 stacks in solar cells In solar cells, typically thin (  2–5 nm) Al2O3 passivation layers are combined with SiNx capping layers, which serve as antireflection coating when on the front side or as dielectric mirror when on the rear side of the solar cell. Furthermore, to “fire” the (printed) metal contacts through the passivation layers, a very short (  1 s) high temperature anneal at 800 °C (firing step) is typically performed. To study the compatibility of the ALD passivation layers of this work with SiNx capping and subsequent firing, various passivation schemes are compared on n þ /p/n þ lifetime samples in Fig. 5. The passivation schemes include single layer Al2O3 and SiO2/Al2O3 stacks with or without SiNx capping, and in addition to this, also results from a batch ALD reactor are shown. Interestingly, the passivation of n þ Si surfaces by thin Al2O3 films capped by SiNx in combination with a firing step (Fig. 5b) leads to a reduced J0n þ compared to the thicker reference Al2O3 films without SiNx capping and firing (Fig. 5a). This can possibly be

Fig. 5. (a) The recombination parameter of n þ /p/n þ lifetime samples passivated by either ALD Al2O3 films or SiO2/Al2O3 stacks, annealed at 400 °C. The horizontal axis displays the thickness of the individual ALD layers. In (b) the ALD based passivation layers are capped by 70 nm of SiNx, which was deposited in an in-line PE-CVD reactor. These stacks are subsequently fired at  800 °C for  1 s. In (c), the results of (b) were reproduced by depositing the SiO2 and Al2O3 layers in a batch ALD reactor used for high-volume manufacturing, before capping with SiNx and subsequent firing.

455

explained by a reduction in terms of Dit in combination with a reduction of negative Qf, as was observed by Richter et al. for similar passivation schemes after firing [21]. Here, we show that even though Al2O3 films capped by SiNx can provide excellent passivation of the n þ Si surfaces, its passivation could be even further improved by having a thin (3 nm) ALD SiO2 interlayer. Moreover, the ALD SiO2/Al2O3 films capped by SiNx of Fig. 5b yield similar results compared to their reference samples of Fig. 5a. Hence, these initial findings indicate that the application of a SiNx capping layer and a high temperature firing step do not significantly affect the passivation level offered by the SiO2/Al2O3 stacks. The results of Fig. 5b, which were obtained using a single-wafer ALD reactor, could be successfully reproduced by depositing the Al2O3 and SiO2 layers in a batch ALD reactor (see Fig. 5c). This reactor was designed specifically for high-volume manufacturing. Note that in the batch ALD reactor, the passivation layers were deposited on both sides of the lifetime samples at the same time. Moreover, the SiO2/Al2O3 stacks were deposited in a single deposition run without unloading the samples in between. In the batch ALD reactor, O3 was used as oxidant in the ALD processes, whereas O2 plasma was used in the single wafer reactor. Despite the differences in reactor geometry and oxidants used, both ALD processes yielded very similar passivation results as can be seen when comparing Fig. 5b with 5c. This demonstrates the robustness and scalability of the ALD processes.

4. Conclusions In this work, the simultaneous passivation of n þ and p þ Si surfaces has been investigated. Corona charging experiments on lifetime samples clearly demonstrate that the fixed charge density is a key parameter controlling the passivation of both surface types. It was shown, that by precisely controlling the Qf in SiO2/Al2O3 stacks via carefully tuning the amount of ALD SiO2 cycles, a strongly improved passivation on highly-doped n þ Si compared to single layer Al2O3 films could be achieved. Simultaneously, due to a high level of chemical passivation, a low J0p þ on p þ Si could be maintained. Simulations predict that SiO2/Al2O3 stacks have the potential to outperform Al2O3 in a wide range of phosphorus surface doping concentrations, most prominently in the range of 1018–1020 cm  3. In general, at highly-doped surfaces the condition for maximal surface recombination (i.e., sn  ns E ps  sh) is less likely to be met for “zero-charge” passivation schemes. Initial feasibility studies on the compatibility of the SiO2/Al2O3 stacks with conventional Si solar cell steps, such as capping by a SiNx layer and a subsequent high temperature (firing) step, yielded similar results in terms of J0n þ . Moreover, after successful upscaling of the ALD process from a single-wafer ALD reactor to an industrial batch ALD reactor, similar passivation results were achieved. In conclusion, this work reveals the opportunities for ALD based passivation schemes to passivate n þ and p þ Si surfaces either separately or even simultaneously in a single deposition run. These results are particular relevant for the passivation of the back side of IBC solar cells, but are also of interest for the passivation of both sides of a diffused-junction solar cell at once. Further research is currently ongoing to demonstrate the concept of a “zero-charge” passivation scheme on solar cell level.

456

B.W.H. van de Loo et al. / Solar Energy Materials & Solar Cells 143 (2015) 450–456

Acknowledgments The authors would like to thank D. S. Saynova (ECN) and C. A. A. van Helvoirt, Dr. J. Melskens, T. L. I. Frankort and R. J. van Gils (TU/e) for their valuable contributions to this work. We are grateful for financial support from the Dutch Ministry of Economic Affairs, via the Top-consortia Knowledge and Innovation (TKI) Programs “Advanced Nanolayers” (TKIZ01001) and “DutchNess” (TKIZ01007).

References [1] B. Hoex, S.B.S. Heil, E. Langereis, M.C.M. Van De Sanden, W.M.M. Kessels, Ultralow surface recombination of c-Si substrates passivated by plasmaassisted atomic layer deposited Al2O3, Appl. Phys. Lett. 89 (2006) 042112. [2] B. Hoex, et al., Excellent passivation of highly doped p-type Si surfaces by the negative-charge-dielectric Al2O3, Appl. Phys. Lett. 91 (2007) 112107. [3] G. Agostinelli, et al., Very low surface recombination velocities on p-type silicon wafers passivated with a dielectric with fixed negative charge, Sol. Energy Mater. Sol. Cells 90 (2006) 3438–3443. [4] G. Dingemans, W.M.M. Kessels, Status and prospects of Al2O3-based surface passivation schemes for silicon solar cells, J. Vac. Sci. Technol. A 30 (2012) 040802. [5] W. Liang, K.J. Weber, A.F. Thomson, Effective SiNx:H capping layers on 1 nm Al2O3 for pþ surface passivation, IEEE J. Photovolt. 4 (2014) 1405–1412. [6] A. Richter, J. Benick, M. Hermle, Boron emitter passivation with Al2O3 and Al2O3/SiNx stacks using ALD Al2O3, Photovolt. IEEE J. 3 (2013) 236–245. [7] S. Duttagupta, et al., Dielectric charge tailoring in PECVD SiO2/SiNx stacks and application at the rear of Al local back surface field Si wafer solar cells, IEEE J. Photovolt. 5 (2015) 1014–1019. [8] J. Schmidt, et al., Surface passivation of high-efficiency silicon solar cells by atomic-layer-deposited Al2O3, Prog. Photovolt. Res. Appl. 16 (2008) 461–466. [9] J. Benick, et al., High efficiency n-type Si solar cells on Al2O3-passivated boron emitters, Appl. Phys. Lett. 92 (2008) 253504. [10] J.A. van Delft, D. Garcia-Alonso, W.M.M. Kessels, Atomic layer deposition for photovoltaics: applications and prospects for solar cell manufacturing, Semicond. Sci. Technol. 27 (2012) 074002. [11] S. Bordihn, G. Dingemans, V. Mertens and W.M.M. Kessels, Passivation of n þ -type Si surfaces by low temperature processed SiO2 /Al2O3 stacksIEEE J. Photovolt. Volume 3, Issue 3, 2013, pp. 925-929. [12] B. Hoex, M.C.M. van de Sanden, J. Schmidt, R. Brendel, W.M.M. Kessels, Surface passivation of phosphorus-diffused n þ -type emitters by plasma-assisted atomiclayer deposited Al2O3, Phys. Status Solidi – Rapid Res. Lett. 6 (2012) 4–6. [13] G. Dingemans, N.M. Terlinden, M. A Verheijen, M.C.M. van de Sanden, W.M.M. Kessels, Controlling the fixed charge and passivation properties of Si(100)/Al2O3 interfaces using ultrathin SiO2 interlayers synthesized by atomic layer deposition, J. Appl. Phys. 110 (2011) 093715. [14] G. Dingemans, M.M. Mandoc, S. Bordihn, M.C.M. van de Sanden, W.M.M. Kessels, Effective passivation of Si surfaces by plasma deposited SiOx/a-SiNx:H stacks, Appl. Phys. Lett. 98 (2011) 222102. [15] I. Dirnstorfer, D.K. Simon, P.M. Jordan, T. Mikolajick, Near surface inversion layer recombination in Al2O3 passivated n-type silicon, J. Appl. Phys. 116 (2014) 044112. [16] S. Dauwe, L. Mittelstädt, A. Metz, R. Hezel, Experimental evidence of parasitic shunting in silicon nitride rear surface passivated solar cells, Prog. Photovolt. Res. Appl. 10 (2002) 271–278. [17] T. Shoji, K. Kakushima, Y. Kataoka, A. Nishiyama, N. Sugii, H. Wakabayashi, K. Tsutsui, K. Natori, H. Iwai, Effect of surface potential control and interface states for silicon nanowire solar cells. in: Proceedings of the 29th EU-PVSEC, pp. 844–847, 2014. 10.4229/EUPVSEC20142014-2AV.2.12. [18] S. Duttagupta, et al., Progress in surface passivation of heavily doped n-type and p-type silicon by plasma-deposited AlOx/SiNx dielectric stacks, IEEE J. Photovolt. 3 (2013) 1163–1169. [19] S. Duttagupta, B. Hoex, A.G. Aberle, Progress with industrially-feasible excellent surface passivation of heavily-doped p-type and n-type crystalline silicon by PECVD SiOx/SiNx with optimised anti-reflective performance. in: Proceedings of the 28th European PVSEC, pp. 993–996, 2013. 10.4229/28thEUPV SEC2013-2CO.4.6. [20] F. Lin, S. Duttagupta, A.G. Aberle, B. Hoex, Excellent passivation of nþ and p þ silicon by PECVD SiOx/AlOx stacks. in: Proceedings of the 27th European Photovoltaic Solar Energy Conference Exhibition, pp. 1251–1254, 2012. 10. 4229/27thEUPVSEC2012-2AV.6.21. [21] A. Richter, J. Benick, A. Kimmerle, M. Hermle, S.W. Glunz, Passivation of phosphorus diffused silicon surfaces with Al2O3: influence of surface doping concentration and thermal activation treatments, J. Appl. Phys. 116 (2014) 243501. [22] D.S. Saynova, et al. Dielectric passivation schemes for high efficiency n-Type c-Si Solar Cells. in: Proceedings of the 28th EU-PVSEC, pp. 1188–1193, 2013 10. 4229/28thEUPVSEC2013-2BV.1.46.

[23] G. Dingemans, W.M.M. Kessels, Aluminum oxide and other ALD materials for Si surface passivation, ECS Trans. 41 (2011) 293–301. [25] D. Schuldis, et al., Properties of the c-Si/Al2O3 interface of ultrathin atomic layer deposited Al2O3 layers capped by SiNx for c-Si surface passivation, Appl. Phys. Lett. 105 (2014) 231601. [26] J.A. Aboaf, D.R. Kerr, E. Bassous, Charge in SiO2–Al2O3 double layers on silicon, J. Electrochem. Soc. 120 (1973) 1103. [27] J. Clemens, E.F. Labuda, C.N. Berglund, Aluminum oxide/silicon dioxide, double insulator MOS structure, Bell Syst. Tech. 54 (1975). [28] N.M. Terlinden, G. Dingemans, V. Vandalon, R.H.E.C. Bosch, W.M.M. Kessels, Influence of the SiO2 interlayer thickness on the density and polarity of charges in Si/SiO2/Al2O3 stacks as studied by optical second-harmonic generation, J. Appl. Phys. 115 (2014) 033708. [29] S. Mack, et al., Silicon surface passivation by thin thermal oxide/PE-CVD layer stack systems, IEEE J. Photovolt. 1 (2011) 135–145. [30] D.K. Simon, P.M. Jordan, I. Dirnstorfer, F. Benner, C. Richter, T. Mikolajick, Symmetrical Al2O3-based passivation layers for p- and n-type silicon, Sol. Energy Mater. Sol. Cells 131 (2014) 72–76. [31] V.D. Mihailetchi, Y. Komatsu, L.J. Geerligs, Nitric acid pretreatment for the passivation of boron emitters for n-type base silicon solar cells, Appl. Phys. Lett. 92 (2008) 063510. [32] G. Dingemans, et al., Influence of the oxidant on the chemical and field-effect passivation of Si by ALD Al2O3, Electrochem. Solid-State Lett. 14 (2011) H1. [33] G. Dingemans, W. Beyer, M.C.M. van de Sanden, W.M.M. Kessels, Hydrogen induced passivation of Si interfaces by Al2O3 films and SiO2/Al2O3 stacks, Appl. Phys. Lett. 97 (2010) 152106. [34] G. Dingemans, C.A.A. van Helvoirt, D. Pierreux, W. Keuning, W.M.M. Kessels, Plasma-assisted ALD for the conformal deposition of SiO2: process, material and electronic properties, J. Electrochem. Soc. 159 (2012) H277–H285. [35] S. Bordihn, Surface Passivation by Al2O3-Based Film Stacks for Si Solar Cells, Eindhoven University of Technology, The Netherlands, 2014 (Ph.D. thesis). [36] B.W.H. van de Loo, et al., Advanced front-surface passivation schemes for industrial n-type silicon solar cells, Photovolt. Int. 24 (2014) 43–50. [37] F.J. Ma, et al., Two-dimensional numerical simulation of boron diffusion for pyramidally textured silicon, J. Appl. Phys., 2014, pp. 184103 1-8, issue 116, http://dx.doi.org/10.1063/1.4901242. [38] D.E. Kane, R.M. Swanson, Measurement of the emitter saturation current by a contactless photoconductivity decay method, in: Proceeding of IEEE Photovoltaic Special Conference 18, 69, pp. 578–583 IEEE, Las Vegas, 1985. [39] G. Dingemans, M.C.M. van de Sanden, W.M.M. Kessels, Influence of the deposition temperature on the c-Si surface passivation by Al2O3 films synthesized by ALD and PECVD, Electrochem. Solid-State Lett. 13 (2010) H76. [40] K.J. Weber, et al., Surface passivation using dielectric films: how much charge is enough? in: Proceedings of the 24th EU-PVSEC 2009, pp. 534–537. [41] K.R. McIntosh, P.P. Altermatt, P.P. Altermatt, A freeware 1D emitter model for silicon solar cells, Conference Record of IEEE Photovoltaic Specialists Conference 1 (2010) 2188–2193. [42] A. Richter, F. Werner, A. Cuevas, J. Schmidt, S.W. Glunz, Improved parameterization of auger recombination in silicon, Energy Procedia 27 (2012) 88–94. [43] A. Schenk, Finite-temperature full random-phase approximation model of band gap narrowing for silicon device simulation, J. Appl. Phys. 84 (1998) 3684. [44] P. Ostoja, S. Guerri, P. Negrini, S. Solmi, The effects of phosphorus precipitation on the open-circuit voltage in N þ /P silicon solar cells., Sol. Cells 11 (1984) 1–12. [45] B. Min, et al., Heavily doped Si:P emitters of crystalline Si solar cells: recombination due to phosphorus precipitation, Phys. Status Solidi – Rapid Res. Lett. 8 (2014) 680–684. [46] B. Hoex, M. Bosman, N. Nandakumar, W.M.M. Kessels, Silicon surface passivation by aluminium oxide studied with electron energy loss spectroscopy, Phys. status solidi – Rapid Res. Lett. 7 (2013) 937–941. [47] A. Stesmans, V.V. Afanas’ev, Si dangling-bond-type defects at the interface of (100)Si with ultrathin layers of SiOx, Al2O3, and ZrO2, Appl. Phys. Lett. 80 (2002) 1957. [48] P. Saint-Cast, et al., Variation of the layer thickness to study the electrical property of PECVD Al2O3/c-Si interface, Energy Procedia 8 (2011) 642–647. [49] A.G. Aberle, S. Glunz, W. Warta, Impact of illumination level and oxide parameters on Shockley–Read–Hall recombination at the Si–SiO2 interface, J. Appl. Phys. 71 (1992) 4422. [50] C. Reichel, F. Granek, M. Hermle, S.W. Glunz, Investigation of electrical shading effects in back-contacted back-junction silicon solar cells using the twodimensional charge collection probability and the reciprocity theorem, J. Appl. Phys. 109 (2011) 024507. [51] Silvaco Inc., Atlas Device Simulation Framework, St. Clara, CA, USA, 2015. [52] K.R. McIntosh, L.E. Black, On effective surface recombination parameters, J. Appl. Phys. 116 (2014) 014503. [53] L.E. Black, T. Allen, K.R. McIntosh, A. Cuevas, Effect of boron concentration on recombination at the p-Si–Al2O3 interface, J. Appl. Phys. 115 (2014) 093707. [54] A.G. Aberle, S. Glunz, W. Warta, Field effect passivation of high efficiency silicon solar cells, Sol. Energy Mater. Sol. Cells 29 (1993) 175–182. [55] J. Dziewior, W. Schmid, Auger coefficients for highly doped and highly excited silicon, Appl. Phys. Lett. 31 (1977) 346–348. [56] D.B.M. Klaassen, A unified mobility model for device simulation—I. Model equations and concentration dependence, Solid State Electron. 35 (1992) 953–959.