Available online at www.sciencedirect.com ScienceDirect Energy Procedia 92 (2016 ) 822 827 6th International Conference on Silicon Photovoltaics, SiliconPV 2016 Low-cost kerfless wafers with gradient dopant concentration exceeding 19% cell efficiency in PERC production line Ralf Jonczyk a, Adam Lorenz a, *, Ali Ersen a, Jasmin Hofstetter a, Kati Hübener b, Klaus Duncker b, Jessica Scharf b, Larissa Neibergall b, Kai Petter b, Jörg Müller b, Daniel Jeong b a 1366 Technologies Inc., 6 Preston Court, Bedford, MA 01730 USA b Hanwha Q Cells, Sonnenallee 17-21,06766,Bitterfeld-Wolfen OT Thalheim,Germany Abstract Industry-leading multi-crystalline PERC processing has been applied to kerfless wafers made directly from molten silicon using Direct Wafer technology, avoiding the need for ingot production and sawing of bricks into wafers, thus providing significant cost saving potential. Efficiency averaged 18,7% and a champion efficiency of 19,1% was achieved on wafers with uniform doping using only mass production tools and processes. An additional efficiency gain was obtained using non-uniform doping, with a higher dopant concentration on the backside of the wafers that we call drift field wafers. Simulations predict an efficiency increase of up to 0,9% in efficiency. In the first test, we demonstrated increased efficiency by 0,3% compared to uniformly doped wafers, leading to an average of 19,0%, with single cells > 19,3%. Further refinement of the wafer doping properties and co-optimization with mainstream cell process technologies will help accelerate the silicon PV learning curve and drive down costs for the prevailing multi-crystalline market. 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license 2016 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer review by the scientific conference committee of SiliconPV 2016 under responsibility of PSE AG. Peer review by the scientific conference committee of SiliconPV 2016 under responsibility of PSE AG. Keywords: kerfless wafers; gradient doping; PERC; multi-crystalline; drift field, Direct Wafer, plasma texture * Corresponding author. Tel.: 781.861.1611. E-mail address: alorenz@1366tech.com 1876-6102 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer review by the scientific conference committee of SiliconPV 2016 under responsibility of PSE AG. doi:10.1016/j.egypro.2016.07.077
Ralf Jonczyk et al. / Energy Procedia 92 ( 2016 ) 822 827 823 1. Introduction The Direct Wafer production process developed by 1366 Technologies fabricates high-performance kerfless silicon wafers directly from molten silicon and offers an attractive lower-cost alternative to the wasteful ingot casting and sawing methods dominating industry today. Compatibility with standard cell processes and equipment allows more rapid progress by leveraging advancements throughout the silicon PV value chain. Production lots in excess of 10,000 wafers were processed using an industrial PERC process. A champion cell efficiency of 19.1%, certified by Fraunhofer CalLab, was achieved using a p-type baseline uniformly doped kerfless wafer and the cell process currently applied in high volume manufacturing. In recent experiments, the unique capacity to create a grown-in doping gradient with higher doping on the backside was demonstrated with Direct Wafer production technology, leading to additional efficiency improvement. Each wafer is grown individually so that the doping profile can be controlled while near melting temperature. A doping gradient is assumed to introduce a built-in drift field between the back and the front side of the wafer, potentially allowing for a more efficient current extraction [1,2,3], among other potential benefits for solar cell performance. 2. Experimental setup 2.1. Wafer production Kerfless Direct Wafer product was fabricated from three identical production furnaces with a nominal capacity of >5 MW/year each, therefore capable of producing several thousand wafers per day. Baseline wafers are 180 μm thick and uniformly boron-doped to 1.5 2.0 -cm. The drift field wafers were produced with a doping gradient between the front and back surfaces, with a lower average resistivity around 0,7 -cm. Wafers are plasma textured on the front surface prior to shipment to customers. 2.2. Cell processing Solar cells were produced using an industrial PERC process, enabling a cell efficiency of 19.2% in large scale production with standard wire sawn multi-crystalline wafers [4]. The entire processing of the kerfless wafers to completed solar cell was done on mass production tools using mass production-relevant processes. 2.3. Measurement of doping concentrations Doping concentrations at the front and backside were determined for the drift field and standard wafers by CV measurements [5] of solar cells and some flipped solar cells (where the wafers were turned upside down before processing, enabling CV measurements of the doping of the wafer backside). The average doping was determined by eddy current measurements. In addition, the doping profile throughout the wafer was determined by spreading resistance measurements on a bevelled wafer. 3. Results and discussion 3.1. Doping profiles Fig. 1 shows that for the standard wafer the average doping and the doping at front and back are almost constant. The doping concentration found for the drift field wafers differ by a factor of ~ 4 between front and backside, showing that the doping engineering was successful.
824 Ralf Jonczyk et al. / Energy Procedia 92 ( 2016 ) 822 827 Fig. 1. Doping concentrations determined on the actual wafers used by Eddy Current (average) and CV (front and backside surfaces) Fig. 2 further shows an example of a doping profile measured via spreading resistance analysis. The conditions during wafer formation can be adjusted to influence the specific shape of this profile, but the plot below is representative for the first round of cell processing experiments described in this paper. 3.2. Simulation results Fig. 2. Doping profile determined by Spreading Resistance Analysis on gradient doped Direct Wafer product. To estimate the potential efficiency increase due to the incorporation of this doping engineering, we simulated three cases of different doping profiles as shown in Fig. 3 by a 2D finite element simulation of a PERC cell (see Fig. 4).
Ralf Jonczyk et al. / Energy Procedia 92 ( 2016 ) 822 827 825 Fig. 3. Doping profiles used in 2D finite-element simulation. Fig. 4. Simulated efficiency potential of drift field wafers (green) compared to uniformly doped wafers (blue) Fig. 4 shows the potential efficiency gains predicted by the simulation relative to a uniformly doped wafer with 1,8 -cm resistivity. An efficiency increase of 0,93% is expected for the drift field profile with the lowest average resistivity of 0,45 -cm ( linear ), promising a 0,14% gain over uniformly doped wafers with the same resistivity. For the two drift field profiles at 0,7 -cm, the potential gain does not depend strongly on the profile shape. In fact, even a uniformly lower doping with 0,7 -cm would already lead to 0,58% potential efficiency gain, and the additional gain from a drift field is 0,05%. However, uniformly doped wafers with resistivities <1 -cm are not used in production, because high doping at the front surface of the wafer leads to unacceptably early prebreakdown in reverse direction [3]. For drift field wafers, this limit will be lower, because the doping at the p-n junction is lower than the average doping, enabling higher average doping concentrations.
826 Ralf Jonczyk et al. / Energy Procedia 92 ( 2016 ) 822 827 3.3. Measured Cell Efficiency and breakdown voltage The cell results in Table I confirm increased efficiency from drift field wafers obtained in the experiment, with a benefit of 0,3% compared to the standard wafers. An average efficiency of 19,0% was achieved for drift field wafers, with single cells above 19,3%. The efficiency increase is mainly due to considerably higher fill factors (+1.4% on average) and due to higher V oc (+5 mv on average), both likely increased due to the higher average base doping. Table 1. Cell efficiency results, lot averages for Direct Wafer product. Cell type V oc (mv) I sc (A) FF (%) (%) Average cm Uniformly Doped Wafers 638 9,12 78,2 18,7 1,70 Gradient Doping Wafers 643 9,08 79,6 19,0 0,65 Fig. 5. shows the dependence of the reverse current I rev at -13V as a function of the average base doping for solar cells made from drift field Direct Wafer product and from uniformly doped wafers (both Direct Wafer product and wire sawn multi-crystalline reference wafers). One potentially significant difference between the uniformly doped Direct Wafer product and multi-crystalline reference, is the front surface morphology. Crystal-defect-related etch pits in iso-textured multi-crystalline cells decrease the breakdown voltage [6, 7], whereas in the case of Direct Wafer product, no such etch pits are expected due to the plasma texture. The comparison between the breakdown voltage of uniformly doped, plasma-textured Direct Wafer cells and iso-textured multi-crystalline reference cells in Fig. 4 shows only a small difference between the two groups, which may be texture related. For gradient doping wafers, the I rev increases above 2A only for cells with a resistivity below 0,5 -cm, which is in clear contrast to uniformly doped wafers, where this happens at around 0.8 -cm for both uniformly doped Direct Wafer product and multicrystalline references. Fig. 5. Dependence of the reverse current I rev at 13V as a function of the average base doping of Direct Wafer drift field solar cells (blue), uniformly doped Direct Wafer solar cells (red), and uniformly doped multi-crystalline reference wafers (green).
Ralf Jonczyk et al. / Energy Procedia 92 ( 2016 ) 822 827 827 4. Conclusions The unique possibility to introduce a doping gradient from back to front in a kerfless silicon wafer was shown to increase the solar cell efficiency by ~ 0,3% absolute, mostly due to gains in fill factor and open circuit voltage. Simulations predict an even higher potential gain and show that a big part of it stems from the increased average doping, consistent with experimental observations. A doping concentration this high cannot be used for uniformly doped wafers, because of reverse breakdown criteria. For wafers with doping gradient, however, it is possible because of the reduced doping in the region of the p-n junction, which determines the breakdown voltage. Therefore this doping engineering technique is a very promising feature unique to kerfless wafers, and we believe that in future experiments we can demonstrate an even higher efficiency gain as indicated by the simulations. References [1] Wolf, M., Proc. IEEE 51 (1963) 674 693 [2] Elnahwy, W. et al. An optimum base drift field for a p-n junction solar cell. J. Phys. D: Appl. Phys. 23 (1990) 112-117 [3] Weber, K.J. et al. The influence of drift fields in thin silicon solar cells. Solar Energy Materials and Solar Cells 45 (1997) 151-160 [4] Engelhart, S. et al. 3 Years Of High Quality Mc-Si Q.ANTUM Production Experience Approaches For Efficient Cell And Module Development. Proceedings of the 31st EUPVSEC 2015, 2.BP.1.3 [5] Schütze, M. et al. Extended Analysis of Capacitance Voltage Curves for the Determination of Bulk Dopant Concentrations of Textured Silicon Solar Cells. IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 11, p 3759 [6] Wagner, M. et al. Shunts, diode breakdown and high reverse currents in multicrystalline silicon solar cells. Proceedings of the 24th EUPVSEC, 2009; 2028 2031. [7] Breitenstein, O. et al. Understanding junction breakdown in multicrystalline solar cells. Journal of Applied Physics 109 (2011), Nr. 7, S. 71101.