26th European Photovoltaic Solar Energy Conference and Exhibition

Transcription

1 21.7 % EFFICIENT PERC SOLAR CELLS WITH ALO X TUNNELING LAYER D. Zielke 1, J.H.Petermann 1, F. Werner 1, B. Veith 1, R. Brendel 1,2 and J. Schmidt 1 1 Institute for Solar Energy Research Hamelin (ISFH), Am Ohrberg 1, Emmerthal, Germany 2 Institute of Solid State Physics, Leibniz University Hanover, Appelstr. 2, Hanover, Germany ABSTRACT: A plasma-assisted atomic-layer-deposited aluminum oxide (AlO x ) layer is implemented as a passivating tunnel layer between the aluminum contact grid and the phosphorus-diffused n + -emitter of a passivated emitter and rear cell (PERC). The nominal AlO x layer thickness is varied between 0 and 0.96 nm. We determine experimentally an optimal AlO x layer thickness of 0.24 nm leading to an independently confirmed efficiency of 21.7 % and an open-circuit voltage (V oc ) of 673 mv. Compared to the best V oc of a PERC solar cell without an AlO x layer we achieve an improvement in V oc of 12 mv clearly proving the excellent contact passivation effect of the AlO x layer. For PERC solar cells with an AlO x layer thicker than 0.24 nm the series resistance increases due to the decreasing tunneling probability. Keywords: contact, passivation, solar cell 1 INTRODUCTION Atomic-layer-deposited (ALD) aluminum oxide (AlO x ) was successfully applied in the past for the passivation of p- and n-type crystalline silicon wafers [1-5]. On the device level, an excellent rear surface passivation was demonstrated with ALD-AlO x by different groups, achieving maximum cell efficiencies of up to 21.4 % [6-8]. Recently, ultrathin ALD-AlO x with a nominal thickness of 0.24 nm was introduced by our group as a contact-passivating layer on the n + -emitter of a PERC solar cell [9]. The AlO x layer was successfully applied on a thin film crystalline silicon solar cell resulting in a new record-high efficiency of 19.1 % for transferred thin film cells [10,11]. In this paper, we describe the process including AlO x tunnel contact in detail and present an experimental analysis of the contact properties. 2 DEVICE FABRICATION 2.1 Process details Figure 1: PERC solar cell realized in this work. The structure of the PERC solar cell processed in this work is depicted in Fig. 1 and the corresponding process flow is shown in Fig. 2. We start with (100)-oriented boron-doped float-zone (FZ) silicon wafers with a resistivity of 0.5 cm and an initial thickness of 300 µm. After damage etching in a potassium hydroxide (KOH) solution and standard RCA-cleaning, a 100 nm thick silicon nitride (SiN x ) diffusion barrier with a refractive index of n=1.9 at =632 nm is deposited on both surfaces by plasma-enhanced chemical vapor deposition (PECVD, Roth & Rau SINA). After SiN x deposition a 2 2 cm 2 window is opened by frequency-doubled Nd:YVO 4 laser ablation (wavelength 532 nm, pulse duration 9 ps, SuperRapid, Lumera Laser GmbH). The opened side is further referred to as the front side. As next step the front side is textured with random pyramids in a KOH/isopropanol solution. After texturing of the front surface, an n + -emitter with a sheet resistance of 100±4 sq is formed in a quartz furnace at 855 C from a POCl 3 source. After the diffusion, the phosphorus silicate glass (PSG) on the front surface and the SiN x on the rear surface are removed in an HF solution. Subsequently, we passivate the rear surface with a 30 nm thick ALD-AlO x layer in a commercial ALD reactor (FlexAL, Oxford Instruments) at 200 C. The deposition of AlO x by plasma-assisted ALD consists of two self-limiting half reactions: (i) tri-methyl-aluminum exposure of the surface and (ii) oxidation of the precursor layer in a remote oxygen plasma. One ALD cycle leads to the formation of a nominally 0.12 nm thick AlO x layer [12]. The final AlO x thickness is calculated by multiplying the number of deposited ALD cycles with the growth per cycle of 0.12 nm due to the linear growth during the ALD process. To activate the passivation quality, we anneal the cells in a quartz-tube furnace in nitrogen atmosphere at 425 C for 15 min. Subsequently, a 100 nm thick PECVD-SiN x capping layer is deposited at 400 C in a microwave-remote PECVD reactor (Oxford Instruments, Plasmalab 80 + ) on the AlO x layer. Next, we locally open the AlO x /SiN x stack at the rear by ps-laser ablation using the laser mentioned above. The opening geometry consists of square points rectangularly arranged with a pitch of 2.1 mm and an opened area of 4 %. After a standard RCA cleaning with a short HF dip we deposit a 10 µm thick aluminum layer on the entire rear side by electron beam evaporation (Balzers BAK 600). Subsequently, the cell batch is split into several batches, where AlO x tunneling layers with nominal thicknesses between 0 and 0.96 nm are deposited on the n + -emitter using plasma-assisted ALD. Note that the final tunnel oxide between the n + -emitter and aluminum probably consists of an ultrathin SiO x layer and an AlO x layer, as depicted in Fig. 1. This ultrathin SiO x layer is a part of the final tunneling layer and has probably a thickness between nm [2, 13]. For simplicity, we mention only the AlO x layer with a thickness corresponding to the numbers of ALD cycles. To mention is the starting 1115

2 ALD reactor. Subsequently, a PECVD-SiN x double layer consisting of a 10 nm thick layer with a refractive index of n=2.4 at =632 nm and a 70 nm thick layer with a refractive index of n=2.05 at =632 nm is deposited at 300 C on the both surfaces. 3 RESULTS AND DISSCUSSION 3.1 J 0e results To distinguish whether the AlO x layers provide an improved passivation between the n + -emitter and the SiN x ARC, we measure the emitter saturation current density J 0e of test samples. The J 0e test samples are measured at high injection densities of ~ cm -3 and an intrinsic carrier concentration of n i =10 10 cm -3 is used [14]. We use a WCT-120 Sinton Instruments lifetime tester. The measured J 0e values as a function of the number of ALD cycles are shown in Fig. 4. Figure 2: Process flow for the fabrication of a PERC solar cell as developed in this work. ALD cycle that considerably differs from the later cycles described above. For the formation of an AlO x tunneling layer we start with oxidation of the surface using the O 2 plasma followed by the TMA exposure. After formation of an AlO x tunneling layer, an Al grid is evaporated through a shadow mask by electron-beam evaporation (Balzers, BAK 600). Finally, we deposit a PECVD-SiN x double layer anti-reflective coating (ARC) consisting of a 10 nm thick layer with a refractive index of n=2.4 at =632 nm and a 70 nm thick layer with a refractive index of n=2.05 at =632 nm on the entire cell front. 2.2 Emitter saturation current density measurements Figure 3: Emitter saturation current density test sample fabricated in this work. The structure of the symmetrical emitter saturation current density (J 0e ) test samples with an AlO x tunneling layer is depicted in Fig. 3. We start with a high-resistivity (~200 cm) boron-doped FZ-Si wafer. After damage etching in a KOH solution for 11 min both surfaces are textured with random pyramids in a KOH/iso-propanol solution. Next, an n + -emitter with a sheet resistance of ~100 /sq is formed in a quartz-tube furnace at 855 C from by a POCl 3 source. After removing the PSG by an HF solution, AlO x tunneling layers with nominal thicknesses of 0, 0.12 and 0.24 nm are deposited by plasma-assisted ALD on both surfaces using a FlexAL Figure 4: Measured J 0e values of samples with a 0, 0.12 and 0.24 nm thick AlO x layer before and after annealing at 350 C. Before annealing we achieve J 0e values of J 0e =(306±32) fa/cm² for cells without AlO x and J 0e =(389±24) fa/cm² for cells with a 0.24 nm thick AlO x layer. After annealing we achieve J 0e values of J 0e =(269±5) fa/cm² for cells without AlO x and J 0e =(266±5) fa/cm² for cells with 0.24 nm thick AlO x. We conclude that no additional passivation due to the AlO x layer is achieved on non-metalized emitter regions. 3.2 Solar cell The illuminated current-voltage (I-V) characteristics are measured under a diode array field at 100 mw/cm² at a temperature of 25 C. The solar cell results denoted by an asterisk (*) have been independently confirmed by the Fraunhofer ISE CalLab (Freiburg, Germany) under standard testing conditions. Measured solar cell results with an AlO x tunneling layer with a thickness between 0 and 0.96 nm are summarized in Table I. The results are provided for solar cells before and after annealing. Median values, annealing times and standard deviations are also provided. Annealing was performed on a hot plate in air at a temperature of 350 C. The optimal annealing time differs for different AlO x layer thicknesses. For each cell batch of the same AlO x layer thickness the annealing time was successively raised until degradation in cell efficiency occurred, to determine the optimal annealing time. The most constant cell parameter before and after annealing is the short-circuit current density J sc with amounts to over 40 ma/cm² for almost all cells. The only 1116

3 exception are cells with a 0.94 nm thick AlO x layer where a median J sc of 8.8 ma/cm² was obtained before annealing due to a large series resistance R s_dlt of 53.5 cm². Measurements of R s_dlt are performed using the double-light method [15]. Table I: Measured parameters of PERC solar cells with nominal AlO x tunneling layer thicknesses of 0, 0.12, 0.24, 0.48 and 0.92 nm. For each AlO x thickness the best results are presented before and after annealing at 350 C for 1-13 min. The median of all cells, the standard deviation (s. d.) and the annealing times are also provided. The aperture cell area is 4 cm². The cell parameters denoted by an asterisk (*) have been independently confirmed at Fraunhofer ISE CalLab (Freiburg, Germany). Number of Annealing V oc J sc FF ALD cycles time [min] (mv) (ma/cm²) (%) (%) 0 best cell best cell med. of med. of s. d. of best cell best cell med. of med. of s. d. of best cell best cell 4 673* 40.3* 79.9* 21.7* 2 med. of med. of s. d. of best cell best cell med. of med. of s. d. of best cell best cell med. of med. of st. d. of Figure 5: Median energy conversion efficiencies as a function of the number of ALD cycles before (blue squares) and after (red circles) annealing. Error bars mark the measured maximum and minimum efficiency values. Dashed lines are guides to the eye. The error bars mark the maximal and minimal measured efficiencies. The decrease in FF with increasing AlO x layer thickness is caused by an increasing series resistance and therefore a reduced tunneling probability through the AlO x layer. After optimal annealing, the FF and efficiency of cells with an AlO x layer thicker than 0.24 nm are successfully improved, which can be attributed to a decrease in R s. For cells with an AlO x layer of 0.24 nm before annealing we obtain an R s_dlt of 1.1±0.1 cm² and for cells with an 0.34 nm AlO x layer an R s_dlt of 4.1±0.9 cm². After annealing we obtain for cells with an AlO x layer of 0.24 nm an R s_dlt of 1.0±0.1 cm² and for cells with an 0.34 nm AlO x layer an R s_dlt of 1.7±0.2 cm². Figure 7 shows median open-circuit voltages V oc of all cells before and after annealing. Before annealing a low V oc of 646 mv (median of 7) with a very large scatter of ±58 mv is achieved for cells without AlO x layer. After applying an AlO x layer deposited by only one ALD cycle we achieve a significantly higher V oc of 662 mv (median of 6) and a significantly reduced scatter of ±3.4 mv. A further increase in AlO x layer thickness has no significant Figure 5 shows median solar cell conversion efficiencies for all cells before and after annealing as a function of the numbers of ALD cycles. The error bars mark the maximal and minimal measured efficiencies. The highest solar cell conversion efficiency ( ) before annealing are 21.1 % and 21.2 % for an AlO x layer deposited by one and two ALD cycles, respectively. Nevertheless, the cells with a tunneling layer deposited by two ALD cycles show a better reproducibility (i.e. a smaller error bar). The large scattering in efficiencies with an AlO x layer less than 0.24 nm for cells before annealing is most likely caused by Al spiking during the ARC-SiN x deposition at 300 C for ~6 min on the front side. Additional thermal treatment during the annealing increases the Al spiking and therefore the scattering in efficiency. For cells with an AlO x layer larger than 0.24 nm the cell efficiency before annealing decreases significantly, caused by an increasing fill factor FF, as shown in Fig. 6 as a function of the number of ALD cycles. Figure 6: Median fill factor FF as a function of the number of ALD cycles before (blue triangles) and after (red diamonds) annealing. Error bars mark the measured maximum and minimum values of each batch. Dashed lines are guides to the eye. 1117

4 Figure 7: Median cell open-circuit voltages V oc as a function of the number of ALD cycles before (blue triangles) and after (red stars) annealing. Error bars mark the measured maximum and minimum values of each batch. Dashed lines are guides to the eye. effect on the V oc due to an apparent saturation of the passivation quality under the Al contacts. By annealing at 350 C we improve the median V oc of our cells by 7 to 12 mv. This effect can be attributed to an improved contact passivation. The AlO x layers show a clear passivation effect due to significantly higher V oc values of 668 to 675 mv of all 24 cells with an AlO x layer compared to V oc values of 321 to 661 mv of all 7 cells without AlO x layer. The highest efficiency of (21.7±0.4) %, independently confirmed at Fraunhofer ISE CalLab, was achieved for a 0.24 nm thick AlO x layer with a V oc of 673 mv. 4.3 J-V curve analysis To determine the shunt resistance R sh and the diode saturation current density J 0 of all cells without AlO x layer and with 0.24 nm thick AlO x layer we fit a onediode model to the measured J-V curves shifted by J sc in to the fourth quadrant: J(V)+J sc =J 0 [exp((v+j(v) R s )/V T ) 1]+[V+J(V) R s ]/R sh, where R s is the series resistance, V T is the thermal voltage and J ph J sc is the photo-generated current density. Figure 8 shows an example of a semi-logarithmic plot of measured one-sun J-V curves and model fits for a typical cell from the batch without tunnel layer and for an 0.24 nm thick AlO x tunnel layer deposited using two ALD cycles before and after annealing at 350 C. The analysis of shifted J-V curves reveals an R sh value between 1.3 and 4.1 k cm 2 for cells with a 0.24 nm thick AlO x layer and R sh of <0.87 k cm 2 for cells without AlO x tunnel layer. The nearly 8 times lower average values of R sh for cells without AlO x layer provide evidence for Al shunts causing a reduced FF, as shown in Fig. 6. After optimal annealing, R sh decreases to values between 0.58 and 2.7 k cm 2 for cells with 0.24 nm thick AlO x and to R sh <0.67 k cm 2 for cells without an AlO x layer. It becomes evident that AlO x layers provide a large R sh, reducing Al spiking through the n + -emitter during annealing. Figure 8: Measured semi-logarithmic plot of one-sun J-V curves and model fits (black lines) for a typical cell from the batch without tunnel layer (circles) and for an 0.24 nm thick AlO x tunnel layer (squares) deposited using two ALD cycles before (blue symbols) and after annealing at 350 C (red symbols). The analysis of the shifted J-V curves shown in Fig. 8 reveal a J 0 value of (326±75) fa/cm 2 for cells with a 0.24 nm thick AlO x layer and J 0 =(410±47) fa/cm 2 without AlO x layer. After the optimal annealing, the J 0 values decrease to J 0 =(174±11) fa/cm 2 for cells with AlO x layer and to J 0 =(320±41) fa/cm 2 for cells without an AlO x layer. The J 0 values of (174±11) fa/cm 2 clearly indicate a contact passivation effect of the AlO x layer. Compared to the J 0e results from the test samples, were J 0e of samples with AlO x where similar to those samples without AlO x layer, it becomes evident that the AlO x layer effectively passivate the metal contact. 5 CONCLUSIONS We have demonstrated the applicability of ultrathin ALD-AlO x as a contact-passivating layer between the emitter and the metal grid of PERC solar cell. Best results were achieved by implementing an AlO x layer with a nominal thickness of 0.24 nm. We are well aware of the fact that underneath the AlO x an nm thick SiO x layer is presented between the n + -emitter and the AlO x layer. The best PERC solar cell fabricated with a 0.24 nm thick AlO x layer leads to an independently confirmed efficiency of 21.7 % and a V oc of 673 mv. Importantly, cells with implemented AlO x layers showed an improved reproducibility and only small variations in the cell parameters. An 0.24 nm thick AlO x layer leads to a nearly 8 times larger R sh compared to cells without an AlO x layer, preventing aluminum to spike through the n + - emitter during the SiN x deposition at 300 C and annealing at 350 C. The very low J 0 values of (174±11) fa/cm 2 determined by fitting the measured shifted J-V curves clearly demonstrated a contact passivation. Comparing the J 0 values of PERC solar cells and J 0e values of test samples, we found that the passivation unambiguously occurs between the n + -emitter and the Al contacts. Increasing the AlO x thickness above 0.24 nm led to a drastic increase in series resistance, and hence a reduced conversion efficiency caused by a decreasing tunneling probability. 1118

5 REFERENCES [1] G. Agostinelli, A. Delabie, P. Vitanov, Z. Alexieva, H. F. W. Dekkers, S. De Wolf, and G. Beaucarne, Sol. Energy Mater. Sol. Cells 90 (2006) 3438 [2] B. Hoex, S.B.S. Heil E. Langereis M.C.M. van de Sanden and W.M.M. Kessels, Appl. Phys. Lett. 89 (2006) [3] B. Hoex, J. Schmidt, P. Pohl, M.C. van de Sanden, and W. M. Kessels, J. Appl. Phys. 104 (2008) [4] G. Dingemans, M.C. van de Sanden, and W. M. Kessels, Electrochem. Solid-State Lett. 13 (2010) H76 [5] A. Richter J. Benick, M. Hermle, and S. W. Glunz, Phys. Status Solidi RRL 5 (2011) 202 [6] J. Schmidt, A. Merkle, R. Brendel, B. Hoex, M.C. van de Sanden, and W. M. Kessels, Photovoltaics Res. Appl. 16 (2008) 461 [7] P. Saint-Cast, J. Benick, D. Kania, L. Weiss, M. Hofmann, J. Rentsch, R. Preu, and S. W. Glunz, IEEE Electron Device Lett. 31 (2010) 695 [8] J. Schmidt, F. Werner, B. Veith, D. Zielke, R. Bock, V. Tiba, P. Poodt, F. Roozeboom, A. Li, A. Cuevas and R. Brendel, Proc. 25th Europ. Photovolt. Solar Energy Conf., Valencia, Spain (WIP, Munich, 2010) 1130 [9] D. Zielke, J. H. Petermann, F. Werner, B. Veith, R. Brendel, and J. Schmidt, Phys. Status Solidi RRL 5 (2011) 298 [10] J. H. Petermann, D. Zielke, J. Schmidt, F. Haase, E. G. Rojas and R. Brendel, Prog. Photovoltaics Res. Appl. (2011) (accepted for publication) [11] M. A. Green, K. Emery, Y. Hishikawa, and W. Warta, Prog. Photovoltaics Res. Appl. 19 (2011) 565 [12] J. L. van Hemmen, S. B. Heil, J. H. Klootwijk, F. Roozeboom, C. J. Hodson, M. C. van de Sanden and W. M. Kessels, J. Electrochem. Soc. 154 (2007) G165 [13] F. Werner, B. Veith, D. Zielke, L. Kühnemund, C. Tegenkamp, M. Seibt, R. Brendel and J. Schmidt, J. Appl. Phys. 109 (2011) [14] D. E. Kane and R. M. Swanson, 18th IEEE Photovolt. Speci. Conf., Las Vegas, USA (IEEE, New York, 1985), 578 [15] P. P. Altermatt, G. Heiser, A. G. Aberle, A. Wang, J. Zhao, S. J. Robinson, S. Bowden, and M. A. Green, Prog. Photovoltaics Res. Appl., 4 (1996),