ION-IMPLANTED POLY-SI / C-SI JUNCTIONS AS A BACK-SURFACE FIELD IN BACK-JUNCTION BACK-CONTACTED SOLAR CELLS Udo Römer1, Agnes Merkle1, Robby Peibst1, T. Ohrdes1, B. Lim1, J. Krügener2, and R. Brendel1,3 1 Institute for Solar Energy Research Hamelin (ISFH), Am Ohrberg 1, 31860 Emmerthal, Germany Institute of Electronic Materials and Devices, Leibniz Universität Hannover, 30176 Hanover, Germany 3 Institute of Solid State Physics, Leibniz Universität Hannover, 30176 Hanover, Germany *corresponding author: Tel.: +49 (0) 5151 999-643, Fax: +49 (0) 5151 999 400, roemer@isfh.de co-authors: merkle@isfh.de, peibst@isfh.de, ohrdes@isfh.de, lim@isfh.de, kruegener@mbe.uni-hannover.de, brendel@isfh.de 2 ABSTRACT: We present back-junction back-contacted n-type silicon solar cells with a carrier selective back-surface field fabricated by ion-implanted n+ polycrystalline (poly-) Si. In terms of the two-diodes model, we find an enhanced J02-like recombination compared to reference solar cells fabricated with a conventionally doped back-surface field in monocrystalline (c-) Si. By comparing the J02-values of solar cells with different geometries we find that the J02-values are correlated with the index of the back-surface field. Thus, we conclude that this recombination component is caused by the p+n+ junction between the poly-Si and the BBr3 diffused c-Si emitter, which probably exhibits a lot of defects in the poly-Si side of the space charge region. Due to the high J02-like recombination, the best efficiency measured on a cell with poly-Si back-surface field of 21.7 % does not exceed the best efficiency of 23.35 % measured on a reference with a c-Si back-surface field. Approaches to suppress the recombination at the p+n+ junction between emitter and BSF in poly-Si BJBC cells are proposed. Keywords: Passivated Contact, Back-Junction Back-Contact, Solar Cell, Recombination 1 INTRODUCTION Back-junction back-contacted (BJBC) Si solar cells with carrier selective junctions are a promising cell structure to achieve highest efficiencies due to the absence of optical shading and an excellent interface passivation also in the metallized regions [1, 2, 3]. Polycrystalline (poly-) Si / monocrystalline (c-) Si junctions are a promising candidate for carrier selective junctions regarding both, performance and processing issues. We recently achieved recombination current densities down to 1 fA/cm2 on phosphorus ion-implanted poly-Si / c-Si junctions [4]. While our final goal is a BJBC cell utilizing poly-Si junctions for emitter and back-surface field (BSF), we evaluate, as a first step, ion implanted n+ poly-Si as BSF in a hybrid BJBC cell structure with a “conventional” boron emitter diffused in c-Si. The hybride approach was also chosen by F. Feldmann et al., who integrated n+ poly-Si as BSF in double-side contacted n-type cells with a conventional boron front-side emitter [5]. The integration of poly-Si / c-Si junctions for one polarity (n+ BSF in this study) in our baseline RISE (rear- side single evaporation) BJBC cell process [6, 7] is straight-forward and does not require modifications of the laser-ablation based patterning of the doped regions. The latter will be required when the junctions of both polarities – emitter and BSF - shall be realized as polySi/c-Si junctions. However, the integration of poly-Si / c-Si junctions for one polarity (n+ BSF in this study) should already (i) yield a significant efficiency improvement over our BJBC baseline with conventional junctions, and (ii) point out issues - process-related or related to device physics - relevant for the poly-Si / c-Si BJBC cell structure. Examples for (ii) are the necessity of an optimization of the laser parameters for the creation of local contact openings in a dielectric rear-side reflector deposited on the poly-Si, and, the influence of the quality of p+n+-junctions in poly-Si. Those aspects are also relevant for BJBC cells with poly-Si / c-Si junctions for both polarities. Therefore we compare conventional BJBC cells with cells featuring an n+ poly-Si BSF and compare the different device geometries enabling an identification of additional recombination paths. Figure 1. Illustration of the cell structures investigated: (a) n-RISE BJBC cell with n+ poly-Si BSF, (b) reference nRISE BJBC cell with conventional c-Si BSF. Both groups of cells feature a full-area Al2O3/SiNx passivation on the cell rear-side. Table 1: Comparison of IV-quantities for the best cell with poly-Si BSF and with conventional BSF, respectively. Both cells have a BSF index of 1150 µm and an emitter area fraction of 78 %. All data refer to in-house measurements on 3.96 cm2 designated area. best cell conventional BSF 23.35 best cell polySi BSF 21.7 2 Voc [mV] Jsc [mA/cm2] FF [%] pFF [%] 677.5 42.74 80.7 82.23 673.6 42.1 77 78.8 SOLAR CELL PROCESS FLOW In this study we fabricated n-type RISE BJBC [4, 6] cells from symmetric n+ poly-Si/ n c-Si/ n+ poly-Si test structures as studied in previous works [7]. The test structures have been fabricated on 156 mm × 156 mm large pseudo square n-type Cz wafers with a resistivity of 6 Ωcm for the fabrication of the solar cells. The damageetched and RCA-cleaned wafers receive a 2.4 nm thick thermal oxide [8] and a 150 nm thick intrinsic amorphous Si layer. Next we apply a phosphorus ion implantation step. After high temperature treatment for dopant activation, which possibly also yields a local break-up of the interfacial oxide [9], a forming gas anneal was performed. At this stage, we performed photo conductance decay (PCD) measurements and extracted emitter saturation current densities J0 of 1 fA/cm2 and implied open circuit voltages Voc, impl. of 742 mV [7, 10]. The upper limit for the specific interface resistance of the poly-Si / c-Si junction was determined to be 10 mΩcm2 by a method described in Ref. [8]. For completing the BJBC cells, we deposited a SiNx capping layer on the cell rear-side, which was laser-structured in order to define the emitter regions. In these regions, the n+ poly-Si (and further ~7 µm of c-Si) was removed by KOH. We processed 24 cells with an area of 2 cm × 2 cm on each wafer. The device geometry, in particular the BSF index, was varied from 450 µm to 1150 µm. We also varied the finger widths in order to realize emitter area fractions Rs (shift light IV/JscVoc) [Ωcm2] J02 [nA/cm2] Rshunt (fit JscVoc) [kΩcm2] 0.37 146 6 317 0.4 112 33 8.5 from 66 % to 82 %. The emitter junction was formed by BBr3 boron diffusion. One should note that this method does imply a doping of the edge regions, i.e., a direct touching of the p+ boron emitter with the n+ poly-Si BSF (Fig. 1 (a)). After BSG removal, an Al2O3 / SiNx stack was deposited on the rear-side. This stack served as a passivation layer on the boron emitter and as a rear-side reflector on both the c-Si emitter and the poly-Si BSF. The stack furthermore served as a protection layer in the subsequent alkaline front-side texturing step. After cleaning, the front-side was passivated by SiNx. In order to create contact openings, local laser ablation of the AlOx / SiNx stack was applied. This step is especially challenging on the rather thin poly-Si BSF regions, since damaging of the interfacial oxide has to be avoided. In Fig. 2 the lifetime ratio measured by means of dynamic infrared lifetime imaging (dyn. ILM) [11] of a poly-Si test structure is shown before and after full area laser ablation of the dielectric stack. The laser pulse power has been varied between 0.8 µJ and 2.3 µJ. It is found that for the two highest settings the lifetime is reduced. In the lower part of Fig. 2 optical microscope images for the four highest laser power settings are shown, for the lowest investigated setting no image is shown, because no ablation of the dielectric has been obtained. For the second lowest laser power setting, the ablated areas are very inhomogeneous in size and shape and thus not appropriate for the stable processing of solar cells. Therefore the middle setting was chosen for the contact openings. The metallization of the solar cells was performed by Al evaporation, followed by sputtering of SiOx and a subsequent RISE [12] contact separation. The final cell structure is illustrated in Fig. 1 (a). As reference, we also processed “conventional” n-type RISE cells with POCl3 and BBr3 diffused junctions in c-Si (Fig. 1 (b)). These reference cells also feature a full-area AlOx / SiNx passivation on the rear and a SiNx passivation on the front. 3 Figure 2: Upper picture: Lifetime ratio of a poly-Si test structure with 100 nm PECVD SiNx before divided by the lifetime after full area laser ablation as measured by means of dynamic infrared lifetime mapping. The laser pulse power was varied between 0.8 µJ and 2.3 µJ. Lower pictures: Optical microscope images of single laser spots on the same test structure. J01 [fA/cm2] RESULTS AND DISCUSSION We expected a significant improvement of the performance of the cells with n+ poly-Si BSF as compared to the cells with conventional BSF. Besides the suppression of recombination at the BSF metal contacts, the poly-Si BSF does also imply lower recombination in the passivated regions since AlOx does not provide a good passivation on n+ doped c-Si regions [13]. On test structures, we extracted a saturation current density of 1 fA/cm2 for the poly-Si BSF and of 202 fA/cm2 for the conventional BSF and AlOx passivation. When Figure 3: Jsc-Voc curves of the best cells with conventional BSF and poly-Si BSF and respective fits with the twodiodes-model. For the cell with poly-Si BSF, the single current contributions are shown (dashed red lines). exclusively considering the passivated BSF regions, we therefore expected a reduction in the pre-factor J01 of the total cell recombination current density between 30 fA/cm2 and 67 fA/cm2 for our cells with BSF area fractions between 15 % and 33 %. This would correspond to an increase in the open circuit voltage Voc between 7.5 mV and 14 mV. However, this improvement was not observed yet in this study. Rather, the efficiency of 21.7 % for the best cell with poly-Si BSF is lower than the highest efficiency of 23.35 % measured on a reference cell with conventional BSF (Tab. 1). While both cell types exhibit maximum short circuit current densities > 42 mA/cm2 and exhibit comparable series resistance values, both the open circuit voltage Voc and the pseudo fill factor pFF of the n+ poly-Si BSF cells are significantly lower than the values of the cells with conventional BSF. By fitting the Jsc-Voc curves of the solar cells with the two diodes model (Fig. 3), we found that the major reason for the worse cell performance of the poly-Si BSF cells is a strongly increased J02-like recombination. It compromises both, Voc and pFF significantly. When changing the J02-value of the fitted Jsc-Voc curve to 6 nA/cm2 as in the case of the cell with conventional BSF, the Voc-value increases to 683 mV and the pFF-value to 83.0 %. Based on the following arguments, we conclude that the J02-like recombination can be attributed to the p+n+ junction meander between the n+ poly-Si and the p+ c-Si emitter, and the other origins, in particular the junction between the n+ poly-Si and the n-type c-Si base, can be excluded: (i) The recombination behavior of the junction between the n+ poly-Si and the n-type c-Si base was studied previously on full-area test structures [7, 8] without any indications for a significant J02-like recombination. (ii) The J02 component correlates well with the BSF index, i.e., the areal density of p+n+-junction lines between the emitter and BSF region (Fig. 4). (iii) Similar J02 values are obtained for different emitter (BSF) area fractions (Fig. 4). If the J02-like recombination would be occurring at the junction between the n+ poly-Si and the n-type c-Si base, a different area fraction of the n+ poly-Si BSF would imply Figure 4: J02 as a function of the BSF index (of the density of pn-junction lines) for cells with n+ poly-Si BSF and cells with conventional c-Si BSF. Different symbols represent different emitter area fractions (circle 66 %, square 80 %, diamond 82 %, and triangle 85 %). All data refer to the best cells (in terms of efficiency) per group, respectively. On poly-Si BSF cells with slightly lower efficiency, we even obtained J02 values up to 150 nA/cm2. different J02 values. (iv) In previous work [10] we observed a strongly non-ideal recombination behavior on test structures with intentionally fabricated p+n+ junctions in poly-Si films. A plausible explanation for the poor recombination behavior with an ideality factor close to 2 is the presumably high defect density in the poly-Si, which mediates strong recombination processes in the space charge region. Trap-assisted tunneling might also contribute to a certain amount to the poor recombination behavior. A similar recombination behavior was reported for pn-junctions in amorphous Si [14]. By contrast, a touching of n+ BSF- and p+ emitter regions in monocrystalline Si, where the defect density is low, does not result in enhanced J02-like recombination [15]. Unfortunately (at least from this point of view), our n+ poly-Si BSF with a sheet resistance of 50 Ω/□ exhibits sufficient lateral conductivity to prevent a “decoupling” of the p+n+ junction from the emitter. 4 OUTLOOK Experimentally, a suppression of the recombination at the p+n+ junction between emitter and BSF should be achievable by the avoidance of a touching of highlydoped poly-Si regions of one polarity with highly doped regions of the other polarity, regardless whether the latter is also realized in poly-Si or conventionally in c-Si. Thus, an undoped or vey lightly doped region separating emitter and BSF in BJBC would be desirable. For our n-type RISE cell structure which features a step on the cell rear-side, this undoped region could be the edge of the step. For example, when using ion implantation as doping method for the c-Si boron emitter, the edge of the step should not be doped, and thus an undoped gap between emitter and BSF is formed without adding process complexity. 5 ACKNOWLEDGEMENTS The project HERCULES has received funding from the European Union’s Seventh Program for research, technological development and demonstration under grant agreement No 608498. Parts of this work were performed in the framework of HERCULES. Major contributions to this work were generated in the SIMPLIHIGH project, which is funded by the German Federal Ministry for Economic Affairs and Energy under Grant 0325478. We thank Susanne Mau, Thomas Friedrich and David Sylla for their help with sample processing. REFERENCES [1] D. D. Smith, P. Cousins, S. Westerberg, R. J.Tabajonda, G. Aniero, Y.-C. Shen, presented at the 40th IEEE PVSC (2014). [2] K. Masuko, M. Shigematsu, T. Hashiguchi, D. Fujishima, M. Kai, N. Yoshimura, T.Yamaguchi, Y. Ichihashi, T. Yamanishi, T. Takahama, M. Taguchi, E. Maruyama, S. Okamoto, presented at the 40th IEEE PVSC (2014). [3] J. Nakamura, H. Katayama, N. Koide, K. Nakamura., presented at the 40th IEEE PVSC, in press (2014). [4] R. Peibst, N.-P. Harder, A. Merkle, T. Neubert, S. Kirstein, J. Schmidt, F. Dross, P.A. Basore, and R. Brendel, 28th EU-PVSEC, 2013. [5] F. Feldmann, M. Bivour, C. Reichel, M. Hermle and S. W. Glunz, Solar Energy Materials and Solar Cells, 01/2013; DOI:10.1016/j.solmat.2013.09.017, 2013. [6] A. Merkle, H. Schulte-Huxel, S. Blankemeyer, I. Feilhaber, R. Bock, V. Steckenreiter, S. KajariSchroeder, N.-P. Harder, and R. Brendel, Prog. Photovolt., 21, 2013, DOI: 10.1002/pip.2297. [7] R. Peibst, U. Römer, Y. Larionova, H. Schulte-Huxel, T. Ohrdes, M. Häberle, B. Lim, J. Krügener, D. Stichtenoth, T. Wütherich, C. Schöllhorn, J. Graff, and R. Brendel, presented at the 40th IEEE PVSC, 2014. [8] U. Römer, R. Peibst, B. Lim, J. Krügener, E. Bugiel, T. Wietler, and R. Brendel, Solar Energy Materials and Solar Cells 2014, DOI: 10.1016/j.solmat.2014.06.003. [9] R. Peibst, U. Römer, K. R. Hofmann, B. Lim, T. F. Wietler, J. Krügener, N.-P. Harder, and R. Brendel, IEEE J. PV., vol. 4, no. 3, 2014. [10] U. Römer, R. Peibst, T. Ohrdes, B. Lim, J. Krügener, T. Wietler, and R. Brendel, submitted to IEEE J. PV., 2014. [11] K. Ramspeck, S. Reissenweber, J. Schmidt, K. Bothe, and R. Brendel, Appl. Phys. Lett. vol. 93, no. 10, p. 102, 2008 [12] P. Engelhart, A. Teppe, A. Merkle, R. Grischke, R. Meyer, N.-P. Harder, and R. Brendel, Proc. 15th PVSEC, Shanghai, China, p. 802-803, 2005. [13] B. Hoex, M. C. M. van de Sanden, J. Schmidt, R. Brendel, and W. M. M. Kessels, Phys. Status Solidi RRL, 6, 1, 4-6, 2012. [14] A. J. Harris, R. S. Walker, and R. Sneddon, J. Appl. Phys. 51(8), pp. 4287-290, 1980. [15] U. Römer, R. Peibst, T. Ohrdes, Y. Larionova, N.-P. Harder, R. Brendel, A. Grohe, D. Stichtenoth, T. Wütherich, C. Schöllhorn, H.-J. Krokoszinski, and J. Graff, Proc. of the 39th IEEE-PVSC, 2013, pp. 1280– 1284.