Erschienen in: Energy Procedia ; 84 (2015). - S. 47-55 http://dx.doi.org/10.1016/j.egypro.2015.12.294 Available online at www.sciencedirect.com ScienceDirect Energy Procedia 84 (2015 ) 47 55 E-MRS Spring Meeting 2015 Symposium C - Advanced inorganic materials and structures for photovoltaics Void formation on PERC solar cells and their impact on the electrical cell parameters verified by luminescence and scanning acoustic microscope measurements Renate Horbelt a *, Giso Hahn a, Reinhart Job b, Barbara Terheiden a a University of Konstanz, Department of Physics, 78457 Konstanz, Germany b Münster University of Applied Sciences, Dep. of El. Engineering and Computer Science, 48565 Steinfurt, Germany Abstract Ideally formed local Al-contacts of passivated emitter and rear contact solar (PERC) cells feature an eutectic and an uniform local back surface field (LBSF). Under certain conditions the eutectic is missing after the co-firing process, referring to the wellknown voids. So far light beam induced current (LBIC) measurements are used to obtain information concerning the passivation quality of the LBSF in local contacts in general. In addition, the destructive technique of scanning electron microscopy (SEM) is established for distinguishing whether a void features a sufficiently thick BSF-layer or a very thin/no BSF-layer. However, both methods are very time consuming. This paper shows a non-destructive and fast characterization of solar cells by applying electroluminescence (EL) and photoluminescence (PL) measurements to investigate the effect on the electrical parameters after locating the voids by scanning acoustic microscopy (SAM). For filled contacts EL and PL measurements correlate well with the resulting values for series resistance (R S ) and dark saturation current density (j 0 ): the formed LBSF leads to a good surface passivation (high PL signal intensity, low value for j 0 ) and the eutectic layer ensures a good electrical contact (high EL signal intensity, low value for R S ). Voids with a sufficiently thick LBSF show a high PL signal intensity whereas the intensity is significantly reduced for a very thin or completely missing LBSF. Increased values for R S can be explained by the missing eutectic layer. In addition, the electrical connection of the LBSF to the paste can be derived from the value of R S. 2015 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license 2015 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of The European Materials Research Society (E-MRS). Peer-review under responsibility of The European Materials Research Society (E-MRS) Keywords: scanning acoustic microscopy; luminescence measurements; local Al-contact; void Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-0-326874 1876-6102 2015 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 under responsibility of The European Materials Research Society (E-MRS) doi:10.1016/j.egypro.2015.12.294
48 Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 1. Introduction Passivated emitter and rear contact (PERC) solar cells first introduced by Blakers et al. [1] have been well established in the silicon photovoltaic industry and feature significantly higher cell efficiencies compared to standard Al-back surface field solar cells [2]. However, one difficulty of this solar cell concept is the so-called void formation, first investigated in more detail by Urrejola et al. [3]. The application of scanning acoustic microscopy (SAM) by Dressler et al. [4, 5] enabled a fast and spatially resolved detection of voids on large area without destroying the solar cell. They combined electroluminescence (EL) and SAM measurements demonstrating that not every void leads to a decrease of EL signal intensity. Thus not every void affects the electrical cell parameters in a negative way. To prevent a negative impact on electrical cell parameters, a low surface recombination velocity (SRV) within the void seems to be the crucial factor, requiring a well-formed local back surface field (LBSF). It was demonstrated that voids show in general a thinner LBSF than filled contacts applying the same paste and firing conditions to the wafer. In the worst case a LBSF is completely missing [6]. Within this work EL, photoluminescence (PL) and SAM measurements are combined in order to investigate the impact of voids on the electrical cell parameters (series resistance (R S ) and dark saturation current density (j 0 )) in more detail. EL gives an insight in the electrical coupling of the local contact whereas PL measurements allow a better investigation of passivation quality of the LBSF. 2. Solar cell processing sequence For PERC solar cell processing Czochralski (Cz) Si wafers (2-3 cm, starting thickness ~110 μm, 125x125 mm 2 ) are used. An alkaline and single side texturization step is followed by a wet chemical cleaning. A homogeneous n + emitter POCl 3 diffusion in a quartz tube furnace is carried out, leading to a sheet resistance R sheet of 45 /sq. For emitter removal on the rear side a full area protection layer is deposited on the textured front side by inkjet printing. P-glass is removed in HF solution followed by a removal of the emitter on the rear side in a chemical polish etch, leading to a planar surface. The front side protection layer is removed in an alkaline etching solution. A wet chemical etching step is carried out, subsequently followed by a thermal oxidation in a tube furnace. Hence, both sides of the solar cell feature a thermally grown oxide as a passivation layer. In addition, a ~70 nm thick silicon nitride layer (SiN x :H) is deposited on the front side, serving as anti-reflection coating. The thin oxide layer on the rear side is covered by a SiN x :H layer with a thickness of 120 nm, both representing the dielectric passivation stack. This stack is locally opened by laser ablation (picosecond laser wavelength of 532 nm, line contacts, opening width 80 μm, pitch 1 mm). For front and rear side contact formation commercially available Ag and Al pastes are used. After co-firing of the contacts in a belt furnace edge isolation is carried out by using a dicing saw. It has to be pointed out, that the front side fingers are arranged perpendicular to the local contact openings of the rear side in order to separate the signals of luminescence and SAM measurements of front and rear side. 3. Local Al-contact formation and definition of voids The firing profile for local contact formation by applying screen-printing technology is subdivided into three main steps: ramp-up, peak firing temperature, cool down. In the beginning temperature is ramped up to about 600 C. During this time period (~30-40 sec, depending on the belt speed) all organic components are burned out. The temperature is increased further, and at 660 C Al begins to melt. The Al-particles are surrounded by a thin oxide shell maintaining the shape of the particles [7]. In the local contact opening a liquid Al/Si mixture is present, and the dissolved Si is transferred into the Al-paste matrix [8, 9]. In a second step the peak temperature is reached (<10 sec). At peak temperature the maximum amount of Si is dissolved and distributed into the Al-paste matrix [10, 11]. Finally, during cool down Si diffuses back into the local contact opening and forms the eutectic layer, ideally leading to a filled contact. Fig. 1 on the left shows a cross-sectional view of such a filled contact, taken with scanning electron microscopy (SEM). Under certain process conditions the eutectic layer is missing and so-called voids are formed (see Fig. 1 on the right). In both images a local BSF layer, which is formed according to the Al/Si binary phase diagram [12], appears
Renate Horbelt et al. / Energy Procedia 84 (2015) 47 55 as bright region. This is attributed to the higher doping concentration of the local BSF layer. The ionization energy of Si is correlated to its doping concentration [13]. Filled contacts always have a LBSF, voids show a thinner LBSF or no LBSF [6]. Up to now, several authors studied the dependence of void formation on paste composition, firing conditions and contact geometry [14-17]. However, the mechanism and origin of void formation is not fully understood. Fig. 1. Cross-sectional view of a filled contact (left). The local contact is filled with an eutectic. The bright region indicates a uniform local back surface field (LBSF). A local contact without eutectic is defined as void (right). One characteristic regarding local contact formation is schematically shown in Fig. 2 on the left. The dielectric opening width d1 after laser ablation is 80 μm and marked in the sketch. Due to Si dissolution during contact formation, the width increases to up to 100 μm as determined by SEM imaging, denoted as d2 in the sketch. This means that the eutectic is partially covered by the remaining dielectric layer with a length of d3. In case of a filled contact a good conductivity is realized by the eutectic and the extension of the dielectric layer into the contact opening - here called overlap - does not have negative impact on the electrical cell parameters. Fig. 2. Schematic cross-sectional view of a filled contact (left), indicating the increase of contact width after contact formation. The image in the centre shows a void (in the following entitled as Void A ) where the eutectic residues are partially covered by the remaining dielectric stack. The eutectic residues of the void on the right (in the following entitled as Void B ) are completely covered by the dielectric stack. A current flow perpendicular to the paper is possible (symbolized by black crosses). Voids might be affected by this overlap of the dielectric layer, as shown in Fig. 2 (centre and right). The overlap (~10 μm per side) is marked in red. Generally, voids feature residues of the eutectic at both rims of the contact. If the eutectic residue is completely covered by the dielectric layer, no direct electrical connection between eutectic residue and paste is possible. Thus the eutectic is electrically decoupled from the paste, the resistivity of the contact is increased and the direction of current flow is perpendicular to the paper, indicated by the symbol (white point with black cross). The assumption of a current flow within the eutectic residues was suggested by Chen et al. in 2012 [18]. However, they assumed voids with and without this overlap of the dielectric layer. Based on the results of Horbelt et al. in 2014 [6], an overlap of the dielectric layer is unavoidable. If the length of the eutectic residue is larger than the length of the overlap, or eutectic residues exist in the bottom of the contact (see Fig. 2 centre). The electrical cell parameters of a solar cell showing voids of such kind, labelled Void A might be less affected than those with voids as shown in Fig. 2 (right), labelled Void B, in which a current flow is only possible along the eutectic residue. 49
50 Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 4. Characterization tools 4.1. Scanning acoustic microscopy The application of SAM for characterization of solar cells allows a fast and spatially resolved detection of voids on large cell area without destroying the solar cell under investigation. Fig. 3 illustrates the working principle of the measurement set-up, exemplified on a PERC solar cell with an area of 125x125 mm 2. Fig. 3. Sketch of the scanning acoustic microscopy (SAM) working principle shown on the left, the signal analysis in the center, leading to a grey-scale image shown on the right. The solar cell is mounted on a chuck submerged by deionized water serving as a coupling medium for the ultrasonic signal which is emitted by the transducer. The measurement frequency of the applied transducer is 150 MHz, its max. resolution is 10 μm and its max. scanning speed 2000 mm/s. The working principle is based on the so called pulse-echo-mode : the ultrasonic signal is emitted and detected by one and the same transducer. The interaction of ultrasonic waves with different materials of the solar cell (e.g. Al paste, dielectric stacks, Si, etc.) and surface morphologies (e.g. areas of local rear contacts after contact formation) leads to a scattering and reflection of the incoming ultrasonic signal. Only the reflected part is detected by the transducer and converted back into an electrical signal. The sketch in the center of Fig. 3 symbolizes the attenuation of the signal amplitude by passing through the solar cell and interacting with different materials. Analyzing amplitude, phase and time of flight leads to a pixel by pixel image of the scanned area (see Fig. 3 right). The setup to fix the solar cell on the chuck is located on the four edges of the cell. The thin vertical dark lines are the front side fingers, the thick horizontal ones are the front side busbars. Voids are depicted as thin dark horizontal lines in the grey-scale image. Filled local Al contacts appear as bright area in the grey-scale image (right hand side of the cell). A more detailed description is given in [5.] The time for measurement is about 15-20 min for a large area solar cell, depending on the chosen spatial resolution. 4.2. Luminescence measurements The coupled determination of dark saturation current density and series resistance (C-DCR) was introduced by Glatthaar et al. [19] in 2010. This method allows a spatially resolved characterization of the electrical solar cell parameters within several minutes of measurement time. It is based on the principle of PL images calibrated to local junction voltages at the pn-junction:, (1)
Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 51 with a local series resistance R S,i, an externally applied voltage V appl, a local junction voltage V i, the thermal voltage V T, the local dark saturation current density j 0,i and the photocurrent density j p. According to Glatthaar et al. [19], the application of this approach allows neglecting the injection dependence of charge carriers lifetime. Additionally, the optical solar cell properties are eliminated [19]. Based on the method of Glatthaar et al. an evaluation procedure was developed at University of Konstanz [20], including EL and PL measurements for the determination of R S and j 0. For EL measurements a current density of 30.6 ma/cm 2 is applied to the solar cell (forward bias). A laser (808 nm) is used for PL measurements, working under different operation conditions (low illumination at 5E16 photons/cm 2 s, high illumination at 4E17 photons/cm 2 s, short circuit and open circuit condition), see Table 1. Table 1. Operation conditions for luminescence measurements. Measurement principle Illumination [photons/cm 2 s] Current density [ma/cm 2 ] Operating condition EL - 30.6 short circuit PL 5E16 - open circuit PL 4E17 - short circuit PL 4E17 - low extraction PL 4E17 - high extraction Images of R S and j 0 are obtained, revealing the spatially resolved impact of voids on the electrical cell parameters. 5. Results and discussion of solar cell characterization A detailed characterization of a solar cell was carried out by EL, PL and SAM measurements, the corresponding spatially resolved images are shown in Fig. 4. The imaged cell area is chosen to a size of 31x13 mm 2 to allow for taking an image that shows the different kinds of local contacts (filled and voids) and can be collected at the necessary resolution in a reasonable time. The image at the top of Fig. 4 reveals SAM results. The dark stripe running horizontally at the top indicates the front side busbar and the thin vertical lines the front side fingers, serially numbered in blue to facilitate the identification of the area under investigation. The thin horizontal lines, once again serially numbered, are the local rear side contacts. Dark lines, e.g. rear contact no. 2, 4, 5, 7, are voids whereas all other local contacts are filled with an eutectic layer, as explained in the section above. The other four images in Fig. 4 show the corresponding results of the EL (center left), PL (center right (5E16 photons/cm 2 s, open circuit), calculated R S (bottom left) and j 0 (bottom right). A strong variation in signal intensity is detectable for EL and R S. In the following R S and j 0 of filled contacts and voids are discussed.
52 Renate Horbelt et al. / Energy Procedia 84 (2015) 47 55 Fig. 4. Images of detailed solar cell characterization by SAM measurements (top), EL and PL (center) and the corresponding images of RS and j0 (bottom). The cell area is 31x13 mm2. The discussion starts with a filled contact as identified by SAM. According to the SAM measurement, local contact no. 9 is a filled contact within the scanned cell area. The line scans for EL, PL as well as RS and j0 include the contact region between front side finger no. 2 and 7. Fig. 5 shows the corresponding results. Fig. 5. Results of the line scan in x-direction in contact no. 9 (filled contact according to SAM). The numbers marked in blue indicate the corresponding front side finger number.
Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 53 The calculated average values include all variations of signal intensity, including the significant drops at the front side fingers explained by the high recombination and shadowing effect of luminescence light in this cell area. This leads to an average EL signal intensity of 2753 counts/s for filled contact no. 9. The average PL signal intensity is on a significantly lower level (340 counts/s), but constant in the contact area under investigation. The drop along the front side fingers is caused by a higher recombination and shadowing effect of luminescence light. The calculated average value for R S is 0.56 cm 2. The low level is explained by a good electrical connection due to the eutectic layer in the local contact. The eutectic consists of ~88% Al and ~12% Si, hence a good electric conductivity is obtained. The calculated average value of the dark saturation current density j 0 (0.48 pa/cm 2 ) indicates a good passivation of the local contact. This passivation is achieved by a local back surface field (LBSF) of several micrometers thickness. Based on the assumption that a filled local contact is the best case in terms of electrical connection and passivation, all these values stand for the best values achievable within this solar cell. On the contrary to the aforementioned results, voids show a different behavior, exemplified on local rear contact no. 7. Two line scans were carried out, one in the region of front side finger no. 2-7, comparable to void A. The second line scan is carried out between front side finger no. 11-14, comparable to void B. Both results are shown in Fig. 6. Fig. 6. Results of the line scan in x-direction in contact no. 7. Two sections are investigated. Void A includes the section of front side finger no. 2-7, Void B the section front side finger no. 11-14. Compared to the filled contact no. 9 void A shows a lower average EL signal intensity (2474 counts/s). Obviously the missing eutectic material hinders the injection of charge carriers into the Al-paste particles, leading to a reduction in EL signal intensity. However, the average value of R S is only marginally increased (0.58 cm 2 ) compared to R S of the filled contact. These two values indicate that the electrical connection of the void in this area is not as good as for the filled contact. On the other hand, the only very small increase of R S indicates a remaining connection of the void to the Al paste (comparable to the void in Fig. 2 center). The PL signal intensity (319 counts/s) of void A is slightly decreased compared to that of the filled contact. This reveals a slightly reduced passivation quality of the LBSF, hence most probably a thinner LBSF. No significant variation of the average value of j 0 (0.49 pa/cm 2 ) is detected. Focusing on void B, only a slight reduction of the average EL signal intensity down to 2335 counts/s is detectable. However, a noticeable increase of the average value of R S to 0.91 cm 2 leads to the conclusion that in contrast to void A this section of the contact is decoupled from the Al paste (comparable to the void in Fig. 2 right). The average PL signal intensity (367 counts/s) is even slightly higher than for the filled contact. This means that the contact surface has a comparably low surface recombination velocity. This is reflected in a constant value of j 0 (0.47 pa/cm 2 ).
54 Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 6. Conclusion Combining scanning acoustic microscope and luminescence measurements allows investigating the impact of different kind of voids on electrical cell parameters in detail. High values for electroluminescence and photoluminescence signal intensities were achieved for filled contacts as well as for voids. This indicates a good electrical coupling of both contact types to the Al paste and a high passivation quality of the LBSF. In general, the EL signal intensity of voids is lower than that of filled contacts, attributed to the missing eutectic layer. High values for series resistance are detected only for voids. This reveals no electrical coupling of such voids to the Al paste and affecting the electrical cell parameters (FF, j SC ) in a negative way. However, some sections of voids show a series resistance almost as low as that of a filled contact these sections are still connected to the paste. In the areas under investigation j 0 shows no significant difference for voids and filled contacts. Acknowledgements Part of this work was financially supported by the German Federal Ministry for the Environment, Nature Conversation and Nuclear Safety (FKZ 0325426 and FKZ 325581) and by the European Commission under FP7, contract number 256695. The authors would like to thank A. Herguth, J. Engelhardt, P. Keller, L. Mahlstaedt, F. Mutter and S. Riegel for their contribution to this work. References [1] Blakers AW, Wang A, Milne AM, Zhao J, Green MA. 22.8% efficient silicon solar cell. Appl Phys Lett 1989;55(13):1363-5. [2] Dullweber T, Hannebauer H, Baumann U, Falcon T, Bothe K, Steckemetz S, Brendel R. Fine-line printed 5 busbar PERC solar cells with conversion efficiencies beyond 21%. Proc. 29 th EU PVSEC, Amsterdam 2014, p. 621-6. [3] Urrejola E, Peter K, Plagwitz H, Schubert G. Silicon diffusion in aluminum for rear passivated solar cells. Appl Phys Lett 2011;98:153508. [4] Dressler K, Dauwe S, Droste T, Rossa J, Meidel R, Schünemann K, Ramspeck K, Gassenbauer Y, Metz A. Characterisation of rear local contacts including BSF formation using RAMAN and scanning acoustic microscopy. Proc. 27 th EU PVSEC, Frankfurt 2012, p. 755-8. [5] Dressler K, Rauer M, Kaloudis M, Dauwe S, Herguth A, Hahn G. Nondestructive characterization of voids in rear local contacts of PERCtype solar cells. IEEE Journal of Photovoltaics 2015;5(1):70-6. [6] Horbelt R, Herguth A, Hahn G, Job R, Terheiden B. Temperature dependence of void formation in PERC cells and their spatially resolved detection by combining scanning acoustic microscopy and electroluminescence measurements. Proc. 29 th EU PVSEC, Amsterdam 2014, p. 427-32. [7] Huster H. Investigation of the alloying process of screen printed aluminium pastes for the BSF formation on silicon solar cells. Proc. 20 th EU PVSEC, Barcelona 2005, p. 1466-9. [8] Uruena A, John J, Beaucarne G, Choulat P, Eyebeb P, Agostinelli G et al. Local Al-alloyed contacts for next generation Si solar cells. Proc. 24 th EU PVSC, Hamburg 2009, p. 1483-6. [9] Grasso FS, Gautero L, Rentsch J, Preu R, Lanzafame R. Characterization of Aluminium screen-printed local contacts. Proc. 2 nd Workshop on Metallization for Crystalline Silicon Solar Cells, Konstanz 2010, p. 15-21. http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-126776. [10] Lauermann T, Fröhlich B, Hahn G, Terheiden B. Diffusion-based model of local Al back surface field formation for industrial passivated emitter and rear cell solar cells. Progr Photovolt: Res Appl 2015;23:10-8. [11] Urrejola E, Peter K, Plagwitz H, Schubert G. Silicon diffusion in aluminum for rear passivated solar cells. Appl Phys Lett 2011;98:153508. [12] Murray JL, McAlister AJ. The Al-Si system. Bulletin of alloy phase diagrams 1984;5(1):74-84. [13] Sealy CP, Castell MR, Wilshaw PR. Mechanism for secondary electron dopant contrast in the SEM. Journal of Electron Microscopy 2000;49(2):311-21. [14] Meemongkolkiat V, Nakayashiki K, Kim DS, Kim S, Shaikh A, Kuebelbeck A, Stockum W, Rohatgi A. Investigation of modified screenprinting Al pastes for local back surface field formation. Proc. 4 th WC PEC (2006), p. 1338-41. [15] Bähr M, Heinrich G, Doll O, Köhler I, Maier C, Lawerenz A. Differences of rear-contact area formation between laser ablation and etching paste for PERC solar cells. Proc. 26 th EU PVSEC, Hamburg 2011, p. 1203-9. [16] Lauermann T, Zuschlag A, Scholz S, Hahn G, Terheiden B. The influence of contact geometry and sub-contact passivation on the performance of screen-printed Al 2O 3 passivated solar cells. Proc. 26 th EU PVSEC, Hamburg 2011, p. 1137-43. [17] Müller J, Gatz S, Bothe K, Brendel R. Optimizing the geometry of local aluminum-alloyed contacts to fully screen-printed silicon solar cells. Proc. 38 th IEEE PVSC (2012), p. 2223-8. [18] Chen D, Yang Y, Li Z, Liang Z, Feng Z, Verlinden P, Shen H. Analysis of morphologies and distribution of Al-doped local back surface field for screen printed i-perc solar cells. Proc. 27 th EU PVSEC, Frankfurt 2012, p. 1303-6. [19] Glatthaar M, Haunschild J, Kasemann M, Giesecke J, Warta W, Rein S. Spatially resolved determination of dark saturation current and series resistance of silicon solar cells. Phys Stat Sol RRL 2010;4(1):13 5. [20] Steuer B. Aufbau eines Photolumineszenz-Messplatzes zur Charakterisierung von Wafern und Solarzellen aus kristallinem Silizium.
Renate Horbelt et al. / Energy Procedia 84 ( 2015 ) 47 55 55 Diploma Thesis. University of Konstanz (2011). http://nbn-resolving.de/urn:nbn:de:bsz:352-162013