22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy DETECTION AND ISOLATION OF LOCALISED SHUNTS IN INDUSTRIAL SILICON SOLAR CELLS USING PL IMAGING Y. Augarten1, M. D. Abbott1, T. Trupke1, R. Bardos1, H. P. Hartmann2, R. Gupta3 , J. Bauer3 and O. Breitenstein3 1 The University of New South Wales, Centre of Excellence for Advanced Silicon Photovoltaics and Photonics, Sydney, 2052, NSW, Australia. Tel: +61 2 9385 4054 Email: y.augarten@student.unsw.edu.au, 2 Deutsche Cell GmbH, Berthelsdorfer Strasse 111A, 09599 Freiberg, Germany 3 Max Planck Institute for Microstructure Physics, Halle, Germany ABSTRACT: This paper discusses an improved shunt isolation process to improve the efficiency and reverse current of shunted solar cells using Photoluminescence imaging. The method is applied to twenty shunted multicrystalline commercial solar cells that were rejected due to a low shunt resistance and the associated high reverse current. The shunts in eight out of twenty cells were successfully electrically isolated, and the performance of these cells was recovered to acceptable standards. Keywords: Photoluminescence, Shunts, Characterisation contact fingers to minimize the induced shunting. In order to identify suitable parameters for this combination of processing, a first experiment was performed on a non-shunted industrial mc-Si solar cell. Three stages were used: 1 INTRODUCTION Efficiency and yield loss due to regions of low shunt resistance in industrial silicon solar cells is a major challenge facing PV manufacturers. Shunted cells significantly reduce average efficiencies, result in modules with poor low light performance, and reduce the total yield. This paper presents an update on the development of a technique to mitigate the influence of localised shunted regions by effectively isolating such regions from finished solar cells. The technique was proposed in [1] and some proof of concept experimental results were reported in [2]. Electroluminescence (EL) imaging [3] and Photoluminescence (PL) imaging [4] were used as fast experimental methods to localise shunted regions present in industrial screen printed solar cells. Subsequently a laser scribed groove was used to electrically isolate the localised shunts from the active cell area. When applied to a severely shunted industrial cell a substantial improvement in the efficiency was achieved (from 9.6% to 13.3%). However, the final efficiency of the cell was still below the average efficiency expected for similar cells and the shunt resistance of the cell was too low to pass hot-spot tests. It was determined that shunting introduced by the laser scribe, particularly where the scribe crossed a metal contact finger, was limiting the potential for improvement via isolation of the shunt. Create laser scribed isolation boxes with two different laser wavelengths (1064nm and 532nnm) and with either a single scribe or a double pass. 2. Apply Silver Etch to laser scribed regions and non laser scribed regions. 3. Laser Scribe isolation boxes within regions of silver etching. At the 532nm laser wavelength the laser was defocused when two passes were scribed, in order to reduce the energy density. After individual processing stages two luminescence images were performed as a means to test the effectiveness of the various processes in isolating the area within the scribed squares. PL images on finished solar cells carried out with the cell at open circuit conditions (referred to here as PLOC) are normally unaffected by series resistance effects and therefore provide information about the local minority carrier lifetime and local shunts. PL images carried out with simultaneous current extraction (referred to as PLSC) provide information about local variations of the series resistance and can thus be used to assess whether the area within the above mentioned laser scribed isolation boxes is successfully isolated from the main part of the cell. Finally, dark lock-in thermography (DLIT) measurements [5] were used to measure the shunts introduced by the different isolation processes (i.e. by steps 1-3 above). The present paper provides an update on the efforts to improve the shunt isolation process. A combination of chemical etching of front grid fingers and laser scribing is investigated as a more gentle process to achieve better electrical isolation and reduce the shunting induced by the laser scribing process. This improved technique is applied to a set of twenty shunted industrial solar cells in order to assess the potential of inline shunt detection and subsequent isolation for industrial applications. All twenty shunted cells investigated initially failed hot-spot tests that are routinely carried out on every cell by the manufacturer. 2 1. 3 RESULTS Figures 1-3 show PLOC and PLSC images measured after each of the processing steps 1-3 described above. In step 1 four sets of six isolation boxes were scribed with different laser wavelengths and either a single or a double laser pass across the scribed area. For proper isolation settings with low shunting and good isolation, the area within the laser scribed isolation boxes should IMPROVED SHUNT ISOLATION PROCESS The improved shunt isolation process uses PL Imaging to localise the shunted regions, followed by a combination of Laser Ablation and Wet Chemical Etching of silver 1220 22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy and etched regions appear brighter in the PLSC image than after the laser processing. The etching step thus appears to have reduced the shunting previously present, most likely by removing silver that had been driven into the junction by the laser. The direct comparison of the four etched boxes with the non-etched boxes (e.g. for the 1064nm double pass boxes) indicates a significant improvement of the isolation process. Finally the green laser was used to scribe additional isolation boxes with different laser settings into the previously etched regions at the top (Fig.3). After the final scribing step, some isolation boxes are clearly visible as bright squares in the PLSC image (Fig 3b). The settings used to scribe these boxes are the most effective, producing good isolation with low induced shunting. The low introduced shunting is further confirmed by the DLIT image shown in Fig.4. Only the 1064nm laser scribed regions that have not subsequently been etched show up as strong shunts. As demonstrated in [2] the shunting appears in that case where the laser passes the metal fingers. The 532nm laser scribes that have not been etched do not introduce significant shunting (Fig.4), however also do not yield sufficient isolation. appear bright in a PLSC image. (a) (b) Figure 1: Six isolation boxes were written with different wavelengths and either one or two laser passes. (a) PLOC image (b) PLSC image measured with 3A current extraction. The PLOC and PLSC imaging results from Figs. 1-3 and the DLIT data were used to identify optimised isolation settings: (1) Etching followed by a shallow 532nm laser groove and (2) 1064nm laser scribing with 2 passes followed by etching. (a) (b) Figure 2: Silver etching was applied to nine areas at the top of the cell and also to four previously laser written areas (the left four in each set of six). (a) PLOC image (b) PLSC image measured with 3A current extraction. Figure 4: DLIT image measured with 0.5V forward bias, showing shunts introduced after step 3. 4 (a) (b) Figure 3: Isolation boxes written with the 532nm laser with different laser settings into the previously etched regions. (a) PLOC image (b) PLSC image measured with 3A current extraction. APPLICATION TO SOLAR CELLS The optimised shunt isolation process with the best settings identified in the above experiments was applied to shunted industrial solar cells that were rejected by the manufacturer due to a low shunt resistance and the associated high reverse current. Twenty mildly to strongly shunted cells were characterised with PL imaging. PL images have previously been demonstrated to reveal the position of shunts as somewhat blurry regions of reduced luminescence intensity [1] (see e.g. Fig.8. This blurring effect is caused by lateral currents flowing from the surrounding cell area into the shunt, that drag down the local voltage and thereby the luminescence intensity in those areas. A quantitative analysis of these blurring effects will be presented at this conference [6]. The ability to detect shunts and to remove their influence from the solar cell via isolation methods depends strongly on the shape, the position and the distribution of the shunts. In that case the area within the laser scribed isolation boxes should appear bright in a PLSC image. Fig.1 shows that this is not the case for any of the laser settings used here. The PLOC image indicates that the 1064nm laser scribe with a double pass introduces significant shunting, dragging down the local voltage inside the isolation box, as has been described in [2]. Fig.2 shows images of the cell after silver etching. The nine bright patches at the top of the cell that are visible in the PLSC image show that decent isolation is achieved just by removing the metal fingers, leaving the etched regions connected to the main part of the cell only via the emitter. Silver etching was also applied to the four laser scribed areas on the left hand side of each set of six laser scribed boxes from step 1. These laser scribed 1221 22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy observed. A number of point like shunts are visible near the left hand edge. In addition the edge region appears dark in parts of the cell which indicates poor edge isolation. The edge isolation can be fixed with a single laser scribe without etching, because no metal fingers need to be isolated. The point like shunts were isolated by scribing with a 532nm laser. In this case the isolation resulted in an improvement in the implied fill factor from 65.0% to 81.4% and in efficiency from 12.4% to 14.4%. Figure 5: An example of a rejected cell. The cell has a large number of material induced distributed shunts. Isolation of the shunted area would remove approximately 25% of the cell area, so it is unlikely the efficiency of the cell would improve. The shunt isolation removes the shunted regions electrically from the main part of the cell. While this reduces (in some cases quite dramatically, as shown below) the influence of the shunt on the fill factor and open circuit voltage it also results in the isolated area no longer being able to contribute to the current. The shunt isolation therefore works best for cells with a small number of small sized moderate to strong shunts. i.e. shunted regions with a small total area. The cell in Fig.5 is affected by a large number of distributed point like shunts, caused most likely by conductive SiC filaments penetrating the wafer that are often found in cast mc-si wafers [7]. Isolation of the shunts in that cell would remove too much active cell area and an efficiency improvement therefore cannot be expected. Isolation was not attempted on cells with a shunted area similar to Fig.5. Fig.6a shows a PL image of a shunted cell that has a strong shunt in the bottom right corner in the vicinity of a grid finger. The inset of Fig.6 shows the same image but plotted on a different color scale, allowing the position of the shunt to be identified accurately. The optimised isolation settings were applied to the area around that shunt. Fig.6b shows a PL image measured after the isolation, which highlights the fact that the influence of the shunt is now confined to a much smaller isolated area. The crack that is visible at the left hand edge was introduced during cell handling and had no measurable influence on the cell performance. Fig.7 shows Suns-VOC measurements [8] performed on that cell before and after the shunt isolation, demonstrating a dramatic improvement in the implied fill factor by 20.3% relative from 69% to 83%. The improvement in cell efficiency, the latter measured under industrial conditions, was somewhat more moderate (a relative improvement of 11.8% from 13.5% to 15.1%). The lower relative improvement of the efficiency compared to the fill factor as determined from Suns-VOC is caused by the series resistance, which is not accounted for in Suns-VOC and which starts dominating the fill factor once the shunt has been isolated. The excellent implied fill factor of 83% shows that the isolation process has completely eliminated the influence of the shunt on the one-Sun cell performance. Reverse bias measurements at -10V revealed that the isolation reduced the reverse current from 15A to 0.7A. Another example for successful shunt isolation is shown in Fig.8. In this case two separate shunt mechanisms are (a) (b) Figure 6: PL image of a strongly shunted cell (SSh03) with a strong local shunt near a grid finger in the lower right. (a) before shunt isolation, (b) after isolation with optimised isolation settings. The images are plotted on the same scale. Ssh03 - Suns Voc before and after processing Before processing After processing 0 Current (A) -1 -2 FFbefore = 68.6 FFafter = 82.7 -3 -4 -5 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Voltage (V) Figure 7: Suns-Voc curves of shunted cell Ssh03 before and after shunt isolation. The fill factor had a relative improvement by 20.3%, while the relative improvement in efficiency was only 15.1%. This demonstrates that the shunt no longer limits the cell’s performance at one Sun, and has been successfully isolated. (a) (b) Figure 8: PL image of a shunted cell Ssh13 with local point like shunts and insufficient edge isolation (insets). (a) before shunt isolation, (b) after isolation with optimised isolation settings. The Cell in Fig.9. required both etching and scribing. The entire cell is significantly affected by a large horizontal shunt along the bottom busbar, evident in Fig.9. A thick 1222 22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy Fill Factors are obtained from Suns-Voc data and thus not affected by series resistance effects. line of silver paste visible on the surface indicates that some screen printing paste was inadvertently deposited in that area and then fired through the SiN to shunt the junction. Following silver etching and laser scribing, efficiency was improved from 9.8% to 14.3%. The isolated region is clearly visible at the bottom of Fig.9b. The reverse current at -10 Volts was reduced by the shunt isolation from >15 A to 1.1 A. The Suns -Voc curves (not presented here) show a relative improvement in the implied Fill Factor of 67% after processing. Various different shunts were observed in twenty shunted cells that were randomly picked from the shunted bin at Deutsche Cell, including poor edge isolation and also local material and processing induced shunts. Table 1 summarises the results on cells that were successfully processed at UNSW to remove the influence of the shunts, thereby recovering the cell efficiency. Importantly, all cells that are listed in Table 1 failed the shunt criteria of the manufacturer prior to processing at UNSW, whereas after the shunt isolation process they met these criteria. 6 (a) A three step process using PL imaging was to localise and isolate shunted regions in completed industrial solar cells was successfully applied. The process was implemented on twenty multicrystalline solar cells that were rejected from the Deutsche Cell GmbH production line due to low values of shunt resistance. In these first experiments, the reverse current and one-Sun efficiency in eight of the cells were recovered to acceptable levels. From the limited statistical data available in this study, we estimate that approximately half of the shunted cells typically found in industrial production can be recovered using the processes described in this work. Images can be taken in 1 second, and in many cases shunt isolation can be accomplished using processes similar to those already used for edge isolation in modern production lines, which increases the potential for scalability. Acknowledgement The Centre of Excellence for Advanced Silicon Photovoltaics and Photonics is supported under the Australian Research Council’s Centres of Excellence Scheme. (b) Figure 9. PL image of a cell with a badly shunted edge caused by a thick strip of silver paste near the bottom region that is thought to have caused the shunt when it was fired into the junction. (a) before and (b) after shunt isolation with optimised settings. Cracks were introduced during processing, which do not appear to have significantly affected the final results The images are plotted on the same scale. 5 SUMMARY OF RESULTS Cell Ssh03 Ssh09 Ssh11 Ssh12 Ssh13 Ssh14 Ssh17 Ssh20 Cell Ssh03 Ssh09 Ssh11 Ssh12 Ssh13 Ssh14 Ssh17 Ssh20 Initial η (%) 13.46 13.20 9.81 14.33 12.37 12.78 12.13 14.18 Final η (%) 15 14.29 14.23 15.15 14.43 14.81 14.42 14.86 Before Processing Initial Implied Irev at FF 10V (A) 15 68.6 14.6 70.6 1.07 48.7 0.3 75.9 <20 65 < 20 65.5 <20 63.8 8.6 76.5 Shunt Criteria met No No No No No No No No After Processing Final Irev Implied at -10V FF (A) 0.7 82.70 2.2 80.80 1.07 81.70 0.3 82.70 1.6 81.40 2.5 81.00 1.08 81.00 1.7 82.00 Shunt Criteria met Yes Yes Yes Yes Yes Yes Yes Yes CONCLUSIONS REFERENCES 1. T. Trupke, R.A. Bardos, M.D. Abbott, K. Fisher, J. Bauer, and O. Breitenstein, International Workshop on Science and Technology of Crystalline Silicon Solar Cells, Sendai, Japan, 2-3 October (2006). 2. M.D. Abbott, T. Trupke, H.P. Hartmann, R. Gupta, and O. Breitenstein, Progress in Photovoltaics (2007) in press. 3. T. Fuyuki, H. Kondo, Y. Kaji, T. Yamazaki, Y. Takahashi, and Y. Uraoka, 31st IEEE Photovoltaic Specialists Conference, Orlando, USA (2005). 4. T. Trupke, R.A. Bardos, M.C. Schubert, and W. Warta, Applied Physics Letters 89 (2006) 044107. 5. O. Breitenstein and M. Langenkamp, Lock-In Thermography - Basics and use for functional diagnostics of electronic components. 2003, Heidelberg: Springer. 6. M. Kasemann, Y. Augarten, T. Trupke, R.A. Bardos, and W. Warta, this conference. 7. J.P. Rakotoniaina, O. Breitenstein, M. Werner, M.H. Al-Rifai, T. Buonassisi, M.D. Picket, M. Ghosh, A. Mueller, and N.L. Quang, 20th EPVSC, Barcelona (2005). 8. R.A. Sinton and A. Cuevas, 16th EPVSC, Glasgow, United Kingdom (2000) 1152. Table 1: Efficiencies and reverse currents of eight of the eleven shunted cells before and after processing (three cells were damaged during processing, and are not shown). The efficiencies are measured by Deutsche Cell, under industrial measurement conditions. The implied 1223