Available online at www.sciencedirect.com
ScienceDirect Energy Procedia 55 (2014) 431 – 436
4th International Conference on Silicon Photovoltaics, SiliconPV 2014
21%-efficient PERL solar cells with plated front contacts on industrial 156 mm p-type crystalline silicon wafers Taejun Kim, Jong-Keun Lim, Hae-Na-Ra Shin, Dohyeon Kyeong, Jinyoun Cho, Moonseok Kim, Jieun Lee, Hun Park, Kyumin Lee*, Won-jae Lee, and Eun-Chel Cho Hyundai Heavy Industries, Co., Ltd., 17-10 Mabuk-ro 240beon-gil, Yongin 446-912, South Korea
Abstract We have achieved 21% large-area conversion efficiency with PERL (passivated emitter, rear locally diffused) solar cells featuring plated front contacts. Industrial 156 mm p-type Czochralski single-crystalline silicon wafers were used as substrates. Our cells employ a single-side texture, a single-side emitter, and a double-layer anti-reflective coating composed of PECVD silicon nitride and thermal silicon oxide layers. The rear side is passivated with a dielectric stack (Al 2O3–SiOxNy), and the rear contacts are formed through laser-ablated openings by screen-printing and firing an aluminum electrode. The nickel–silver or nickel–copper–silver contacts are plated on the front side by light-induced plating (LIP). The champion batch shows average values of 669 mV open-circuit voltage, 40.0 mA/cm2 short-circuit current density, and 77.3% fill factor. The efficiency of our PERL cell is currently limited by the series resistance. Implementing small (1–2 μm) and uniform pyramids and thermally annealing the cells after plating show promising results for future improvements. © 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license © 2014 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/3.0/). Peer-review under responsibility of the scientific committee of the SiliconPV 2014 conference. Peer-review under responsibility of the scientific committee of the SiliconPV 2014 conference Keywords: Crystalline silicon; Solar cell; PERL; PERC; Copper metallization; Plating
1. Introduction The PERL (passivated emitter, rear locally diffused) structure holds the record for the conversion efficiency of silicon (Si) solar cells [1]. After more than a decade since the record-setting research, the photovoltaic industry is
* Corresponding author. Tel.: +82-31-289-5027; fax: +82-31-289-5020. E-mail address:
[email protected]
1876-6102 © 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). Peer-review under responsibility of the scientific committee of the SiliconPV 2014 conference doi:10.1016/j.egypro.2014.08.123
432
Taejun Kim et al. / Energy Procedia 55 (2014) 431 – 436
approaching a consensus on how to realize the PERL structure in the mass production environment [2–4]. The industrial PERL solar cells (often called “PERC”) reuse most of the elements of the conventional solar cell such as p-type Si wafers, random upright pyramids, silicon nitride (SiN x) anti-reflection coating, and screen-printed contacts. The rear side is passivated by either aluminum oxide (Al 2O3) or silicon oxide (SiO2), and is subsequently capped by silicon nitride (SiNx) or silicon oxynitride (SiOxNy) layers. The local back-surface field (LBSF), the defining feature of the original PERL structure compared to the similar PERC and PERT structures [1], is formed through laser-ablated openings in the rear dielectric stack by screen-printing and firing an aluminum (Al) electrode. Alternatively, the Al LBSF may be formed by laser processing the fired Al electrode without a prior dielectric ablation [2]. Hyundai Heavy Industries has carried out research and development on the industrial PERL solar cells over the past three years [5–9], and as a result, 280 Wp-rated 60-cell solar modules composed of over-20%-efficient screenprinted PERL cells will be released to the market in the second quarter of 2014. At the research lab of Hyundai Heavy Industries, we have developed a process for PERL solar cells with plated front contacts, for which we reported 20.5% large-area efficiency in 2012 [6]. Last year we reported a new champion cell efficiency of 21.0%, which was achieved after implementing an additional thermal oxidation process [9]. The cell efficiency was mostly limited by a relatively high series resistance, and since last year we have investigated the main causes, and obtained promising results on our efforts to address the issue. The results of our investigation are reported here. 2. PERL solar cell process and results 2.1. Cell processing sequence Figure 1 shows the structure and the processing sequence of our PERL cell with plated front contacts. We use commercial 239 cm2 boron-doped single-crystalline Si (sc-Si) wafers as substrates. After a saw-damage etch (KOH), we mask the back face with a dielectric to protect it through the texturing (KOH with additive) and the phosphorus diffusion (POCl3) processes. The back mask is removed along with the front phosphosilicate (PSG) glass during an HF dip, resulting in samples with texture and emitter on the front side only. A more industrial approach would be to use a single-side wet-chemical polishing tool after a double-side texturing and diffusion, as is standard in the mass production of screen-printed PERL cells, but we find the running cost of an inline wet tool too expensive for research purposes.
Fig. 1. (a) Schematic of our PERL cell; (b) The processing sequence
433
Taejun Kim et al. / Energy Procedia 55 (2014) 431 – 436
Once the PSG is removed, we perform a low-temperature oxidation, and then prepare a SiN x layer on top of the front thermal oxide by plasma-enhanced chemical vapor deposition (PECVD). The rear dielectric layers (Al2O3 and SiOxNy) are consecutively stacked on the back by atomic layer deposition (ALD) and PECVD. After locally ablating the rear dielectric stack with a laser, we apply the LDSE (laser-doped selective emitter) process [10] on the front to simultaneously open the front dielectric and form the selective emitter. Metallization is carried out by first screen-printing and firing the Al electrode at the back. The samples are then dipped in a mild HF solution (concentration less than 0.1% m/m) and passed through our plating sequence, which uses the light-induced plating (LIP) method to create either nickel–silver (Ni/Ag) or nickel–copper–silver (Ni/Cu/Ag) contacts on the front. The resulting finger width is 35 μm. The current–voltage (I–V) curves of the final cells are measured with a h.a.l.m. flash cell tester. We first analyze the grid-included external quantum efficiency (EQE) of a cell from the batch with a modified PV Measurement QE analyzer, which is calibrated with a stabilized cell characterized at the Fraunhofer ISE CalLab. The result of the EQE analysis is used to calibrate the solar simulator of our cell tester, to reduce the spectral mismatch error. 2.2. Cell results The I–V curve parameters of the champion batch (10 cells) are shown in Table 1. The champion cell shows a conversion efficiency of 21.0%. The maximum open-circuit voltage (VOC) obtained was 672 mV. Our shallow POCl3 emitter, very fine front fingers, and the rear dielectric stack on a polished surface allow an excellent shortcircuit current density (JSC) of 40.1 mA/cm2, despite the presence of three 1.5 mm-wide busbars. The efficiency is limited by the fill factor (FF), whose maximum value is only 78.1%. In comparison, our screen-printed PERL cells in pilot production show FF values well over 79%. Table 1. I–V characteristics of the champion batch; The maximum values of each I-V parameter are shown in bold face.
Cell No. 1 2 3 4 5 6 7 8 9 10 Average
VOC [mV] 669 667 667 668 667 669 668 671 672 670 668.8
ISC [A] 9.58 9.54 9.56 9.54 9.53 9.56 9.56 9.56 9.58 9.58 9.559
JSC [mA/cm2] 40.1 39.9 40.0 39.9 39.9 40.0 40.0 40.0 40.1 40.1 40.00
FF [%] 78.1 77.4 77.6 76.9 76.7 77.8 77.9 76.7 77.4 76.6 77.31
η [%] 21.0 20.6 20.7 20.5 20.4 20.8 20.8 20.6 20.8 20.6 20.68
3. Improving the series resistance An investigation on the low FF values of our plated-contact PERL cells revealed that there were three main causes for the higher-than-expected series resistance: (1) the deterioration of the rear Al electrode in the plating baths, (2) the uneven morphology of our Ni seed layer, and (3) the high contact resistance at the Ni–Si interface. In this section we disclose our efforts to address these issues. 3.1. Deterioration of the Al electrode in the plating baths In our lab’s plating sequence carried out on a manual R&D plating tool, the Al-metallized samples are completely immersed in the plating baths, as well as in a low-concentration HF solution. The surface of the Al electrode becomes rough through the plating sequence, which is already an indication that the Al electrode deteriorates due to an interaction with the chemicals.
434
Taejun Kim et al. / Energy Procedia 55 (2014) 431 – 436
To study the changes in the sheet and contact resistance of the Al electrode, samples representing the rear side of our PERL cell were prepared for a transfer length model (TLM) analysis. After preparing the rear-passivation dielectric stack (Al2O3 and SiOxNy) on a face, contact openings and Al electrodes were formed as an array of unequally spaced lines suitable for the TLM measurement. End-to-end line resistance and line-to-line resistance were measured on these samples first after firing, and then after dipping in the different chemical baths (lowconcentration HF, Ni plating solution, and Cu plating solution). With the values from the undipped samples serving as the reference, the relative changes in the Al sheet and contact resistance were calculated. The results are summarized in Table 2. Table 2. Relative increases in the Al sheet and contact resistance after dipping
0.1% m/m HF bath
Ni plating bath
Cu plating bath
Sheet resistance
+ 40 – 50 %
+3–5%
+ 230 – 270 %
Contact resistance
0%
0%
0%
It is the sheet resistance of the Al electrode that increases in the HF and Cu plating baths. We believe that the Al particles in the porous electrode are oxidized during the immersion. And as our cells are pin-array-contacted at the back, the resistive loss at the back electrode during the lateral current collection becomes significant. We have studied the cross sections of our cells with a scanning electron microscope, but the resolution was not high enough to detect minute thickness changes of the surface oxide layers of the Al particles. In comparison, our plated-contact cells with full Al BSF show excellent FF values near 80%, which is probably because the full-area Al–Si layer, not present in our PERL cells, assists in carrying the current laterally to the contact points at the back. We have verified that using a thicker Al electrode can alleviate the problem somewhat for our PERL cells. But it is not a preferred solution, because of the larger wafer bowing and also the non-optimal LBSF thickness. We are currently upgrading our plating tool to implement single-side plating processes. 3.2. Front texture morphology and the Ni seed layer uniformity Our conventional texturing process, used when preparing the champion batch PERL cells, resulted in 3–8 μmwide pyramids. A wide distribution in the pyramid size leads to a non-uniform ablation of the front dielectric during the LDSE process, because of the irregular interference between the incident and reflected laser beams. The irregular openings in the front dielectric results in a non-uniform Ni seed layer during the Ni plating, and consequently a high contact resistance for the front contacts. To address this issue, we have improved our texturing process to achieve smaller and more regular pyramid size. With the improved process, the majority of the pyramids show base size in the range of 1–2 μm. As the result, the LDSE dielectric opening and subsequently the Ni seed layer have become more uniform (Fig. 2). We have verified that the improved front texture leads to a 0.5–1.0%p improvement in the fill factor. But due to practical issues in the test runs, we were unable to update our efficiency record in time for this report.
Taejun Kim et al. / Energy Procedia 55 (2014) 431 – 436
435
Fig. 2. SEM images of the Ni seed layer with (a) the conventional front texture, and (b) with the improved front texture
3.3. Thermal annealing after plating It is a well-known fact that the contact resistance at the Ni–Si interface can be reduced significantly by forming a NiSi interfacial layer through a thermal process. The Ni silicidation also helps to improve the contact adhesion of the plated contacts [11, 12]. Preliminary experiments with a bench-top hot plate have shown that a thermal treatment of the plated-contact PERL cells can improve the fill factor by 0.6%p and the efficiency by 0.2%p on average. But extended heating leads to FF drops, which we understand as the encroachment of the space-charge region by the NiSi layer. Cells with deeper junctions were found to be more tolerant to the heat treatment. We are seeking to improve our POCl3 diffusion process to achieve deeper junctions without increasing the recombination at the emitter. 4. Conclusion Our PERL cell process makes 21.0% large-area conversion efficiency possible on commercial 156 mm Czochralski-grown single-crystalline silicon wafers. The plated nickel–copper–silver front contacts promise significant cost savings compared to the cells metallized by screen printing of Ag pastes. Our cells show excellent open-circuit voltage (> 670 mV) and short-circuit current density (> 40.0 mA/cm2). At the moment the fill factor is relatively low (< 78.2%) because of the high series resistance. We seek to reduce the series resistance by applying single-side plating processes to avoid the exposure of the Al electrode to the chemicals, by forming uniform-and-small texture pyramids to achieve uniform dielectric removal by laser, and also by engineering our phosphorus diffusion process to make it more compatible with the post-plating Ni silicidation. We have obtained promising results on these fronts.
Acknowledgments This work was supported by the New & Renewable energy Core Technology Program of the Korea Institute of Energy Technology Evaluation and Planning (KETEP), with the financial grant provided by the Ministry of Trade, Industry & Energy, Republic of Korea (No. 20133010011780).
References [1] Zhao J, Wang A, Green MA. High-efficiency PERL and PERT silicon solar cells on FZ and MCZ substrates. Sol Energ Mater Sol Cells 2001:65:429-435. [2] Kloeter B, Kowalzik P, Ploigt H-C, Buss D, Hussack A, Schuetze M, Stegemann E, Stolze L, Mueller JW. Current status of high-efficiency Q.ANTUM technology at Hanwha Q.Cells. Proc 28th EU PVSEC 2013:1117-1119.
436
Taejun Kim et al. / Energy Procedia 55 (2014) 431 – 436 [3] Xia Z, Gao Y, Tao L, Li X, Yang Z, Sidhu R, Dong J, Qiao H, Teng K, Zhang B, Xing G. Rear side dielectrics choices in PERC solar cells. Proc 39th IEEE PVSC 2013:1781-1783. [4] Tjahjono B, Yang MJ, Wu V, Ting J, Shen J, Tan O, Sziptalak T, Liu B, Sperlich H-P, Hengst T, Beilby B, Hsu K-C. Optimizing CELCO cell technology in one year of mass production. Proc 28th EU PVSEC 2013: 775-779. [5] Kyeong D, Cho SH, Lim JK, Lee K, Hwang MI, Lee W, Cho EC. Approaching 20%-efficient selective-emitter solar cells with copper front contacts on industrial 156 mm Cz Si wafers. Proc 27th EU PVSEC 2012:600-603. [6] Kim T, Seo JW, Shin HNR, Cho SH, Kim M, Kyeong D, Lim JK, Seo J, Lee K, Hwang MI, Lee W, Cho EC. 20.5%-efficient industrial PERL solar cells with copper front contacts on 156 mm Cz silicon wafers. PV Asia Pacific Conf 2012. [7] Shin HNR, Kim T, Kim MS, Cho J, Lee K, Lee W, Cho EC. Characterization of aluminum-metallization compatibility of dielectric capping layers for the development of rear-passivated silicon solar cells. Proc 28th EU PVSEC 2013:2037-2039. [8] Seo JW, Lee HK, Lim JK, Kim KH, Seo JM, Kyeong D, Ahn JH, Hwang MI, Cho EC. 20.2%-efficient industrial PERL cell with ALD Al2O3 backside passivation layer and screen-printed contacts. Proc 28th EU PVSEC 2013. [9] Lim JK, Kim T, Cho J, Shin HNR, Kyeong D, Kim MS, Lee J, Lee K, Lee W, Cho EC. Optimization of thermal oxide and silicon nitride layers for industrial high-efficiency solar cells on commercial-grade silicon wafers. Proc 28th EU PVSEC 2013:1077-1079. [10] Hallam B, Wenham S, Sugianto A, Mai L, Chong CM, Edwards M, Jordan D, Fath P. Record large-area p-type Cz production cell efficiency of 19.3% based on LDSE technology. IEEE J Photovolt 2011:1:43-48. [11] Mondon A, Jawaid MN, Bartsch J, Glatthaar M, Glunz SW. Microstructure analysis of the interface situation and adhesion of thermally formed nickel silicide for plated nickel-copper contacts on silicon solar cells. Sol Energ Mater Sol Cells 2013:117:209-213. [12] Russell R, Tous L, Philipsen H, Horzel H, Cornagliotti E, Ngamo M, Choulat P, Labie R, Beckers J, Bertens J, Fujii M, John J, Poortmans J, Mertens R. A simple copper metallisation process for high cell efficiencies and reliable modules. Proc 27th EU PVSEC 2012:538-543.
