Advances in PassDop technology: recombination and optics

Transcription

1 Available online at ScienceDirect Energy Procedia 124 (2017) th International Conference on Silicon Photovoltaics, SiliconPV 2017 Advances in PassDop technology: recombination and optics Bernd Steinhauser, Andreas Büchler, Henning Nagel, Simon Gutscher, Sven Kluska, Jan Benick, Pierre Saint-Cast, Jonas Bartsch, Martin Hermle a) a Fraunhofer Institute for Solar Energy Systems, Heidenhofstrasse 2, Freiburg, Germany Abstract The PassDop technology is a promising approach to realize passivated emitter and rear locally diffused (PERL) silicon solar cells within a reasonable process sequence. In previous studies, the feasibility of this concept was evaluated, but two topics were only briefly discussed: the recombination at the locally diffused and contacted surface and the optical layer. In this study we present an analysis of the recombination at laser processed spots of local laser diffusion from a-sin x :P. We show that initially the local recombination prefactor J 0b,met for the metallized area is very high (10 4 fa/cm 2 ) but a dedicated anneal can be used to reduce the recombination to a competitive level (600 fa/cm 2 ). The improvement was allocated to the inner area of the laser-processed spot by measuring μ-pl before and after annealing. To improve the optics we show that SiO x can be used on the rear if an etch-back is applied after the laser diffusion. Applying SiO x, results similar to the reference MgF 2 layer for J sc and FF were achieved, while providing much improved adhesion. Finally, we show large area n-type silicon solar cells featuring a PassDop rear side as well as laser contact opening in combination with nickel- and copper-plating on the front. Combining these technologies, we were able to achieve an energy conversion efficiency of 22.2 % The Authors. Published by Elsevier Ltd. Peer review by the scientific conference committee of SiliconPV 2017 under responsibility of PSE AG. Keywords: surface passivation; n-type; laser doping; nickel plating 1. Introduction n-type passivated emitter and rear locally diffused (PERL) silicon solar cells have proven to offer a highefficiency potential [1]. The PassDop approach (see Fig. 1) introduced by Suwito et al. demonstrated an industrially feasible realization of the rear side of such a PERL structure [2]. While the original PassDop layer was based on silicon carbide, we introduced an approach based on phosphorus doped amorphous silicon nitride with which we were able to reach 22 % on large area [3]. The main properties that such a PassDop scheme has to fulfill are a good surface passivation, good doping efficiency during the laser diffusion process, high internal reflectance and low recombination at the local back The Authors. Published by Elsevier Ltd. Peer review by the scientific conference committee of SiliconPV 2017 under responsibility of PSE AG /j.egypro

2 314 Bernd Steinhauser et al. / Energy Procedia 124 (2017) surface field (LBSF). While the first two were discussed in previous publications the other two topics were only discussed briefly. In case of internal reflectance an MgF 2 layer was used to enhance light trapping at the rear. However, as shown later on, this layer is not compatible with module fabrication as the layer adhesion (both of the layer on the PassDop layer and of the aluminum on the MgF 2 layer) was low. Thus for the final step of the cell concept towards making modules, an alternative was required or one would have to make sacrifices in the optical efficiency of the device. Due to its low refractive index, SiO x was thought of as a good candidate, but Suwito found that applying the laser process on this layer led to incomplete ablation preventing contact formation [4]. The focus on our work here is to test if the SiO x layer can be made compatible with the laser diffusion process, meaning reliable contact formation without sacrificing the enhanced light trapping. In addition, we investigate if the layer adhesion is high enough for module fabrication. The last topic to be discussed is the recombination at the PassDop LBSF. While most authors use the model of Fischer [5] to calculate the recombination at the LBSF, we decided to use the model proposed by Saint-Cast et al. mainly due to defining parameters which do not or only weakly depend on the LBSF size (a parameter with high uncertainty). In addition the simple structure of the model allows for proper propagation of uncertainties and thus to better judge the validity of the results. With regard to cell fabrication, the PassDop rear side was combined with laser contact opening on the front followed by nickel- and copper-plating as an industrially feasible solution for the front side metallization. However, the cells featuring NiCu-plating were limited in, which was attributed to bulk degradation [3]. Hence in this work, we aimed at combining the above improvements with the removed bulk lifetime limitation to further explore the potential of the cell concept. Fig. 1. Schematic of the PassDop process sequence for rear side passivation as well as LBSF and contact formation: 1. Cleaning of the silicon surface. 2. Deposition of the doped passivation layer. 3. Application of the laser process to open the layer and simultaneous diffusion of the dopants into the silicon to create the LBSF. 4. Contact formation, e.g. by evaporation of aluminum. 2. Recombination at the PassDop LBSF 2.1. The LPA model The LPA model can be used to describe the recombination at a locally processed area (LPA), e.g. by a laser process. In detail, the model is described in [6] and [7]. Here, we only repeat the important aspects for the application of the model on n-type surfaces, as the original paper focussed on p-type silicon. The model proposes a linear relation between the effective surface recombination velocity and surface density of the LPA :, (1) where is the effective surface recombination velocity at the passivated surface. is a proportionality factor and can be seen as a characteristic measure of the recombination at the surface due to the local processing. To determine the recombination at the actual LPA (e.g. a single laser spot) can be calculated, which is the corresponding recombination parameter for a single LPA 1 1, (2) where is the diffusion resistance (see [6]) and a is the contact radius of a circular spot or the half width of a square-shaped contact. It should be noted that this relation is only valid for point contacts and not for lines. For the latter case see [7] or [8]. Before just calculating, one should consider the case where is very high, thus. This

3 Bernd Steinhauser et al. / Energy Procedia 124 (2017) means that is mainly determined by and hence cannot be determined reliably. Graphically this is shown in Fig. 3, but will be discussed later on. If was successfully determined, one can calculate the effective surface recombination velocity at the LPA using (3) with the area of a single LPA. To calculate from, the method proposed by Kane and Swanson was applied [9] Experimental n-type shiny-etched 1 cm float-zone wafers were passivated symmetrically by the SiN PassDop process using plasma-enhanced chemical vapour deposition (PECVD) [3]. After passivation, the samples were processed singlesided by the PassDop laser process and annealed at 425 C for 25 min using forming gas (FGA) or atomic hydrogen (atom. H) atmosphere. In each state, a QSSPC measurement was performed using a Sinton WCT-120 lifetime tester [10]. The minority carrier lifetime was determined at an excess carrier density of 10. For higher minority carrier lifetimes the transient was evaluated to determine the lifetime. For lower minority carrier lifetimes the generalized method was used [11]. The optical calibration factor to determine the generation for the generalized method was determined in the state before the laser process by comparison of the transient and generalized lifetime curves. After the laser process, the samples were measured with the non-processed side facing the flash lamp. To determine the effective surface recombination velocity at the laser-processed side, the relation 1 (4) was used, where is the wafer thickness and is the lifetime of minority carriers at the surface [7]. For symmetrical samples (before the laser process), the relation was used [12]. To evaluate the recombination at the LPA the model by Saint-Cast et al. was used as described in the previous section. The uncertainties of the determined parameters were propagated by using the partial derivative method. For an uncertainty of 10% was assumed as the recombination radius could divert from the radius that was observed optically. For the laser process a Jenoptik IR70 diode-pumped solid-state laser system operating at a wavelength of 1030 nm and a repetition rate of 30 khz was used. The pulse length was 1.3 s, the beam waist on the substrate was 1. For further information on the laser process, see [13]. All experiments were performed at a laser power of 10 W. To evaluate if the change in the surface recombination rates could be allocated to the area within the laser spot or outside, we performed scanning microscopic photo-luminescence (μ-pl) measurements before and after annealing at 425 C on a hot plate. It was ensured that the same spot was measured before and after annealing Results and discussion Before discussing the actual data, it should be mentioned that determining based on equation (4) led to an increased error in the calculation of. For the symmetrical case (, Sproul showed that the relation is valid with an error of 4 % [12]. For the case with the strongest asymmetry in our data set of and 40, the error made in the calculation increased to about 7 % (determined by calculating the corresponding eigenvalues similar to Sproul). For even stronger asymmetries a higher error would be the result. In all cases will be under-estimated. However, as such a strong asymmetry does mean that the recombination at the LPA is very high, this also means that very likely the process is not suitable and therefore the precise knowledge of is not of much interest. In addition, the scattering found in the data especially at higher LPA densities was the stronger influence on the uncertainty in determining.

4 316 Bernd Steinhauser et al. / Energy Procedia 124 (2017) S eff [cm/s] After Laser FGA, 425 C atom. H, 425 C Lin. Fit Lin. Fit Lin. Fit p eff [cm³/s] Directly after Laser 425 C FGA 425 C atom. H p LPA infinity p LPA [cm³/s] LPA Density [1/cm²] Fig. 2. Determined S eff (Δn= cm -3 ) for single side laser processing and subsequent annealing (FGA or atomic H). For the calculation of the bulk lifetime, we used the model of Richter [4] a [cm] 10-3 Fig. 3. p eff and p LPA in relation to the contact radius a for the given layers. For samples closer to p LPA cm 3 /s, a significant increase in the evaluation error has to be expected. Fig. 2 shows in dependency of the LPA density directly after the laser process as well as after annealing with forming gas (FGA) and when using atomic hydrogen. Directly after the laser process, a strong dependency of on the LPA density was observed as the values were ranging from approx. 10 cm/s to over 200 cm/s. After annealing the influence of the LPA was much lower. A significant difference between FGA and annealing using atomic hydrogen was not observed. is represented by the slope of the linear regression shown in the graph. While being high initially, it was decreased significantly after annealing of the samples. State Table 1. Determined values by fitting the p LPA model to the S eff data. p eff [10-4 cm 3 /s] p LPA [10-4 cm 3 /s] S met [cm/s] J 0b,met [fa/cm 2 ] After Laser 440±20 680± ± ±2000 After Laser + FGA (425 C, 25min) 40±2 41±2 270±60 620±130 After Laser + atom. H (425 C, 25min) 41±3 43±3 260±60 600±130 Fig. 3 shows versus the contact radius. The solid lines give the contour lines for each order of magnitude for. The diagonal blue, solid line indicates the case for an infinitely large, when recombination is limited by the diffusion of the charge carriers to the LPA. The symbols show the determined values for and for the given processes. Even directly after the laser process, the symbol was found to be well below the diagonal, hence an evaluation of is feasible. In addition, the graph indicates that both as well as were decreased by one order of magnitude due to the annealing. The corresponding values for and are given in Table 1, with a of approx cm 3 /s before and cm 3 /s after annealing. was significantly higher ( 55%) than before annealing, but only marginally (3-5%) after performing the annealing. In addition the table gives the values for and that were determined from. Before annealing, would be expected in the range of 10 4 fa/cm 2, after annealing values around 600 fa/cm 2 were determined, which for a laser diffusion process is an excellent result. For a contact fraction of approx. 1.2 %, as commonly used for the solar cells, this would result in a total of 21 fa/cm 2, including 3 fa/cm 2 for the passivated surface and 10 fa/cm 2 for the c-si bulk. In addition, it should be noted that the determined after FGA at below is one order of magnitude below as determined by Suwito et al. for the original PassDop layer despite using 1 Ωcm bulk material as well [2]. As can be observed in Table 1, the relative uncertainties calculated for both as well as are significantly higher than that of. The reason for this is the uncertainty in the area of a laser spot. However, it

5 Bernd Steinhauser et al. / Energy Procedia 124 (2017) should be noted that this is only relevant if a comparable value is required (e.g. for comparing different laser diffusion methods or recombination models). In case of a device simulation which is our main motivation for the calculation the area is well defined in the simulation program. The determined then corresponds to this area and the relative uncertainty is identical to that of. Thus, as an input parameter for device simulations can be determined at much better precision. Fig. 4. μ-pl image of a PassDop spot after annealing at 425 C. The green overlay gives the visual microscopy image of the same spot. The black line indicates the line scan that was performed. Fig. 5. μ-pl line scans of a PassDop spot before and after FGA. It was ensured that the same spot was measured in both states. Fig. 4 shows the μ-pl image acquired for one PassDop laser spot after annealing. As an overlay the image acquired by visual microscopy is shown as well. As can be observed from the figure, the recombination edges matched the visual edges quite well, although in the μ-pl image, a bleeding of carriers from the passivated surface was observed. The visual image showed an inner structure which translated into a variation in the PL signal as well. However, for the interpretation of these variations, it has to be considered that the structures can be interpreted as an increased roughness and therefore the effect might be of optical nature. The black line indicates the PL signal line scan that was performed before and after annealing. The result of this comparison is shown in Fig. 5. Outside the spot, the PL signal did not change significantly after annealing. Inside the spot, an increase in the PL signal by a factor of 3-5 was observed. This means that the improvement in the LPA recombination discussed above is not a result of passivation healing (e.g. due to thermal damaging by the local temperature increase during the laser process) but instead it is the actual LBSF that is improved. Since no difference between anneal using forming gas and atomic hydrogen was found saturation of dangling bonds by hydrogen can be excluded as a source of the improvement. At this point we don t know the exact effect leading to the reduction in the recombination rates, but stress release or healing of crystal defects seem to be the most probable effects at this annealing temperature. 3. Evaluation of SiO x for enhanced light trapping 3.1. Experimental 1 Ωcm n-type shiny-etched float-zone silicon wafers were coated by the SiN PassDop process as described in the previous section. After the PassDop deposition the wafers were coated by 200 nm of SiO x using PECVD. To create the point contacts the PassDop laser process was applied. After the laser process, part of the wafers received an etch-back in a single-side inline tool using 1% HF for 120 s. To judge the adhesion of the layers, 3 μm aluminum were deposited in an inline high-throughput thermal evaporation reactor. For the adhesion tests, a control group featuring 100 nm of MgF 2 deposited by thermal evaporation instead of the SiO x layer was prepared. Instead of the longer process, the control samples were only dipped in HF (30 s, 1%) prior to Al evaporation. Otherwise the process sequence for the MgF 2 samples was similar. An influence of the length of the HF process on the stack adhesion was excluded.

6 318 Bernd Steinhauser et al. / Energy Procedia 124 (2017) Results and discussion Fig. 6. SEM images of a laser spot without (left) and with (right) SiO x etching after evaporation of Al. Fig. 6 (left) shows an SEM image of a PassDop/SiO x laser spot for the PassDop sequence as shown in Fig. 1 without any additional processing. Remnants of the layer are clearly visible at the edge of the processed area similar to that observed by Suwito. As such structures are not present when processing the layer without SiO x or if processing with MgF 2, the remnants were attributed to improper ablation of the SiO x layer. While we did not fabricate solar cells with laser spots in this stage it was shown by Suwito that these structures lead to contact shielding and hence to an increase in the series resistance. Fig. 6 (right) shows an SEM image of a PassDop/SiO x laser spot after performing the HF etch back of the SiO x layer before evaporation of aluminum. Only a small portion of the remnants are still visible at the edge of the spot, the main parts were etched back. While not visible in the SEM image, it was found that the remaining barrier at the edge is low enough (and not continuous) so that contact formation should be possible. This was investigated in a solar cell batch on 2 2 cm 2 for a different project where especially a direct comparison to cells featuring the MgF 2 optical layer was performed. It was found that in comparison to MgF 2 (52 cells) the mean difference in the fill factor was averaged over 135 cells in case of SiO x. The parameter of interest is mainly the series resistance R s. To calculate this we determined both the R s -free pseudo-fill factor by SunsVoc [14] and the actual fill factor. The difference can be attributed to R s [15]. For the SiO x cells we found that ΔFF was in the range of. The same ( ) was determined for the cells featuring the MgF 2 layer. Now the good series resistance might have been an effect of a complete etch back of the SiO x layer thus reducing the effect of light trapping. This was not observed in the averaged J sc, which was actually found to be higher than for the cells featuring the MgF 2 layer. However, the main reason for that was not increased light trapping with SiO x but rather inhomogeneities in the front side anti-reflective coating which was observed for some of the cells with MgF 2. In principle no significant differences between SiO x and MgF 2 would be expected here, which is proved by the spectral response measurement shown in Fig. 8 as well. Both the reflectance and quantum efficiencies for MgF 2 and SiO x (here 4 exemplary cells are shown) did match very well and only low variation at the rear was determined. Thus the etch back of the SiO x did allow us to achieve proper contact formation at the rear without needing to sacrifice enhanced light trapping when using this optical layer. The remaining question was the adhesion of the layer as this was the main motivation to replace MgF 2. The adhesion tests are shown in Fig. 7 for both c-si/passdop/sio x /Al and c-si/passdop/mgf 2 /Al stacks. For the latter, most of the measurement points were found to be below and especially the median was at only. In contrast, the stack with SiO x showed much improved adhesion with many measurement point above 2.5 and a median of, thus complying with the standard DIN EN

7 Bernd Steinhauser et al. / Energy Procedia 124 (2017) Peeling Force [N/mm] SiO x MgF Distance [mm] Fig. 7. Adhesion test for a c-si/passdop/siox/al stack in comparison to c-si/passdop/mgf2/al stack. EQE, R, IQE Symbols: SiO x (4 cells) EQE R IQE Lines: MgF 2 (1 cell) EQE R IQE Wavelength [nm] Fig. 8. Spectral response for PassDop cells with a SiO x optical layer after etching in comparison to the reference MgF 2 optical layer. 4. PassDop solar cells with NiCu front contacts 4.1. Experimental The solar cells were fabricated on 1 cm n-type shiny-etched float-zone silicon. A thermal oxide was applied to mask the rear side from texture formation and diffusion. Texturizing was performed in KOH. The emitter was formed by tube furnace BBr 3 diffusion followed by a drive-in step to create a deep-diffused 100 /sq boron emitter. The rear was passivated by the SiN PassDop layer [3], the front was passivated by a stack of Al 2 O 3 and a-sin x. As the SiO x technology as described in the previous section was not yet ready when fabricating the cells, a layer of 100 nm MgF 2 was applied on the rear side. Next the rear side was opened using the PassDop laser process (see section 2.2). The front and rear passivation were activated by FGA at 425 C for 25 min. The front side was opened using ps-laser contact opening. PVD aluminum was deposited onto the rear to create the rear side contact. The front side was metallized by forward-bias electroplating of nickel and copper Results and discussion The JV results for the best cell are given in Table 2. A high V oc of 680 mv was determined as well as a high J sc of 40.2 ma/cm 2. The fill factor was especially high at 81.1 % resulting in an efficiency of 22.2 %. This means that in comparison to previous results with NiCu front contacts, we were able to achieve a significant improvement in the overall efficiency. Unfortunately, we were only able to integrate part of the improvement the healing of the rear side contacts so far. PassDop cells with the SiO x optical layer on the rear and NiCu front contacts are still pending. With regard to the recombination V oc at 680 mv is high, but not yet at the limit that the cell structure allows, which in this case can be attributed to the front side passivation. Fortunately, we were able to reduce J 0e from 40 fa/cm 2 to around 20 fa/cm 2 for currently running batches, thus a significant improvement in V oc should be possible. Table 2. Calibrated JV measurements by Fraunhofer ISE CalLab for the large area PassDop cell. Cell type Laser Contact Opening +NiCu Plating ¹Aperture area Area¹ [cm 2 ] V oc [mv] J sc [ma/cm 2 ] FF [%] [%]

8 320 Bernd Steinhauser et al. / Energy Procedia 124 (2017) Conclusion We analysed the recombination at the PassDop LBSF for the SiN PassDop layer. The investigation showed that the recombination is initially high with. Annealing the LBSF results in a significant reduction in the recombination rates resulting in. At a common contact fraction of 1.23 % as used in the solar cells, this means that the complete rear side results in a. To achieve this it did not matter if an atmosphere of forming gas or atomic hydrogen was used for the anneal. To provide module-compatible light trapping at the rear side, we investigated the feasibility of SiO x as an optical layer on top of the PassDop layer. While after the laser process remnants of the layer could form a contact barrier we showed that a controlled etch back can remove this remnants without sacrificing the enhanced light trapping effect. In contrast to MgF 2, the SiO x layer was found to provide good adhesion allowing for module fabrication. Finally we fabricated PassDop cells with NiCu plated front contacts and using this approach we achieved a cell efficiency of 22.2 %. Acknowledgments The authors would like to thank A. Leimenstoll, F. Schätzle, S. Seitz, and N. Weber for sample preparation as well as E. Schäffer for measuring the solar cells. This work was funded by the German Federal Ministry for Economic Affairs and Energy under grant No B IdeAl. References [1] Benick J, Hoex B, van de Sanden, M. C. M., Kessels, W. M. M., Schultz O, Glunz SW. High efficiency n-type Si solar cells on Al2O3- passivated boron emitters. Appl. Phys. Lett. 2008;92:253504, doi: / [2] Suwito D, Jager U, Benick J, Janz S, Hermle M, Glunz SW. Industrially Feasible Rear Passivation and Contacting Scheme for High- Efficiency n-type Solar Cells Yielding a Voc of 700 mv. IEEE Trans. Electron Devices 2010;57:2032 6, doi: /ted [3] Steinhauser B, Kamp M, Brand AA, Jäger U, Bartsch J, Benick J, Hermle M. High-efficiency n-type silicon solar cells: Advances in PassDop technology and NiCu plating on boron emitter. IEEE J. Photovoltaics 2016;6:419 25, doi: /jphotov [4] Suwito D. Intrinsic and doped amorphous silicon carbide films for the surface passivation of silicon solar cells. Dissertation, Konstanz U. Konstanz. [5] Fischer B. Loss analysis of crystalline silicon solar cells using photoconductance and quantum efficiency measurements. Dissertation, Universität Konstanz. Konstanz; [6] Saint-Cast P, Nekarda J, Hofmann M, Kuehnhold S, Preu R. Recombination on Locally Processed Wafer Surfaces. Energy Procedia 2012;27:259 66, doi: /j.egypro [7] Steinhauser B. Passivating Dopant Sources for High-Efficiency n-type Silicon Solar Cells. Dissertation, Universität Konstanz. Konstanz; [8] Norouzi. H.M., Saint-Cast P, Jäger U, Steinhauser B, Benick J, Büchler A, Bitnar B, Wolf A. Development and Characterization of AlOx/SiNx:B Layer Systems for Surface Passivation and Local Laser Doping.IEEE J. Photovoltaics accepted. [9] Kane DE, Swanson RM. Measurement of the emitter saturation current by a contactless photoconductivity decay method (silicon solar cells). In: 18th IEEE Photovoltaic Specialists Conference Las Vegas; 1985, p [10] Sinton RA, Cuevas A. Contactless determination of current voltage characteristics and minority-carrier lifetimes in semiconductors from quasi-steady-state photoconductance data. Appl. Phys. Lett. 1996;69:2510 2, doi: / [11] Nagel H, Berge C, Aberle AG. Generalized analysis of quasi-steady-state and quasi-transient measurements of carrier lifetimes in semiconductors. J. Appl. Phys. 1999;86: , doi: / [12] Sproul AB. Dimensionless solution of the equation describing the effect of surface recombination on carrier decay in semiconductors. J. Appl. Phys. 1994;76:2851 4, doi: / [13] Jäger U, Suwito D, Benick J, Janz S, Preu R. A laser based process for the formation of a local back surface field for n-type silicon solar cells. Thin Solid Films 2011;519: , doi: /j.tsf [14] Sinton RA, Cuevas A. A quasi-steady-state open-circuit voltage method for solar cell characterization. In: 16th EU PVSEC; 2000, p [15] Pysch D, Mette A, Glunz SW. A review and comparison of different methods to determine the series resistance of solar cells. Sol. Energy Mater. Sol. Cells 2007;91: , doi: /j.solmat