EO.2.3_Baehr_final LBSFc AL 浆烧结调节

DIFFERENCES OF REAR-CONTACT AREA FORMATION BETWEEN LASER ABLATION AND ETCHING PASTE FOR PERC SOLAR CELLS M. Bhr 1 , G. Heinrich 1,2 , O. Doll 3 , I. Khler 3 , C. Maier 1 , A. Lawerenz 1 1 CiS Forschungsinstitut fr Mikrosensorik und Photovoltaik GmbH, Konrad-Zuse-Str. 14, 99099 Erfurt, Germany 2 Technical University of Ilmenau, Department for Physics, Helmholtz Street 13, 98693 Ilmenau, Germany 3 Merck KGaA, Frankfurter Street 250, 64293 Darmstadt, Germany Corresponding author Mario Bhr Tel 49361-6631-214, Fax 49361-6631-413, mbaehrcismst.de ABSTRACT Key parameters in the manufacturing of the rear metallization for PERC solar cells are the thermal budget within the firing process, the materials used, e.g. metallization paste, and the geometry of the rear-side openings. As a further key parameter, the way of opening the dielectric passivation layers is investigated. Three different methods for locally removing a dielectric layer were analyzed, including one approach by laser processing using laser irradiation in a short and ultrashort pulse length regime and an other one by screen printing an etching paste. After locally removing the passivation stack, the rear-side metallization was carried out by application of different Al pastes and firing profiles. From SEM pictures and comparisons of the width of the local openings after structuring and after firing, different general statements were deduced as about the influence of the structuring width, the influence of the firing profile or the relation of the method of structuring with the formation of the contact area. Keywords PERC solar cells, metallization, laser processing, 1 INTRODUCTION The industrial rear side metallization of crystalline silicon solar cells is based on full-area aluminum contact. This technology is already introduced and step-by-step improved to reach higher conductivity, lower bowing, higher reflection, higher aluminum doping and thus higher passivation quality and better optical confinement resulting in higher cell efficiencies. Nevertheless, solar cells based on the passivated emitter and rear cell concept PERC are becoming more and more attractive not only in the RD but also at production level. Reasons for that are the availability of firing stable passivation layers 1 , fast and reliable techniques for opening the contact area 2 and techniques for contact formation 3 . For an easy industrial application, a process flow for the rear side was suggest by Agostinelli 4 , named i-PERC. In Fig. 1 a schematic of this process flow is shown. Texture I SiN X deposition on front Deposition passivation layer rear Local opening rear passivation Firing II III IV V VI VII Planarization rear VIII Emitter diffusion Metallization front and rear Fig. 1 Schematic of the process flow for i-PERC solar cells based on p-type material as suggested in 4 . It is well known from previous publications that the contact formation is influenced by different mechanisms that must be under control to achieve best results in cell efficiency 5,6,7 . These mechanisms are the thermal budget, the geometry of the contact openings and in our case the aluminum paste used and thus its composition. Since the aluminum is interacting with the silicon base, we would also address this issue as a key parameter for a good contact formation. We investigated the influence on contact formation for two different structuring techniques that are available for industrial mass production. These are first the laser ablation and second the etching via screen printed etching paste. Both are suited for local removing of typical passivation layers as silicon oxide SiO X , silicon nitride SiN X , aluminum oxide AlO X and stacks of these. For our experiments a stack of AlO X /SiN X was used, deposited by PECVD techniques. 2 EXPERIMENTAL Samples were prepared on typical solar grade p-type Cz silicon of the size 156 156 mm with an initial thickness of 200 m, boron doping with a resistivity of 3 – 6 Ωcm. The samples were damage etched removing about 10 m/side and cleaned. A passivation layer of amorphous AlO X and amorphous SiN X was deposited subsequently by a remote PECVD technique with thicknesses of about 25 nm and 80 nm, respectively, on both sides. The next processing steps of locally opening the passivation stack A and screen printing the Aluminum paste B on one side as well as firing the whole sample C in an inline IR belt furnace was performed. This was done according to Fig. 3. The samples finally had the structure as depicted in Fig. 2. p-type silicion Aluminum 80 nm SiNx 25 nm AlOx Local contact Fig. 2 Schematic of the cross section of the sample structure processed. ns laser lines dots fs laser line dots Etching paste lines dots Fast firing Medium firing Slow firing Alu 1 Alu 2 Fig. 3 Parameter variation for investigation of main influences. The opening of the stack was varied by the technique as well as different geometries were processed. Two Al pastes were investigated as well as three different firing profiles. These three different methods in opening a dielectric passivation layer on top of a silicon substrate were chosen since all have a different mode of operation. The opening processes are described more in detail in the following sections. To all laser methods, the layout of the laser pattern was similar On one large wafer substructures with different laser parameter sets were processed, with areas of at least 2.5 x 2.5 cm. In those substructures the openings were processed with equidistant pattern. The pitch used was between 350 and 500 m. For the etching paste, lines or dots were processed on all Cz wafer. Subsequently the wafer was broken into nine equivalent parts and further processed. 2.1 ns-laser ablation The laser parameters of the ns-laser system used are shown in Table 1. Since infrared light was used, the interaction with the dielectric layers is very low due to the transparency of SiN X and AlO X layers in the IR wavelength regime. The main part of the pulse energy is placed into the silicon, where the material can be heated up during the pulse duration of some hundred nanoseconds. Especially the removing of dielectric layers is operating by heating the Si-substrate above the melting temperature and subsequently removing the dielectric by evaporating and blowing up the silicon and the dielectric. Molten Si recrystallizes after laser interaction partially epitactically. It can be stated, that the opening of a dielectric passivation layer with a IR ns laser is always combined with a strong influence on the Si substrate. Due to the “long” interaction time between the laser pulse and the Si substrate, a deep impact zone is produced either by the melting zone or by the heat affected zone in the surrounding. There no melting takes place but silicon is partly distorted with a high amount of dislocations. Depending on the used laser power, the interaction depth can be some ten micrometer 8 . For local removing the passivation stack SiN X /AlO X from a Si substrate using a ns laser we speak more general of a strong influence on the Si substrate. In Fig. 4 such openings are shown, indicating the melting process. Fig. 4 Micrographs of typical local openings fabricated by a ns laser. Not all the samples processed showing such a strong melting process. Table 1 Laser parameters used for laser ablation. The pulse energy given is the maximum value applied only for some of the openings. For both laser systems a gaussian beam shape was used. Parameter ns laser fs laser Type Rofin Powerline 100 D Jenoptik JenLas 2dfs Pulse duration τ P 300 ns 300 fs Wavelength λ 1064 nm 1025 nm Pulse energy E P 1,6 mJ 36 J 2.2 fs-laser ablation The laser parameters of the IR fs-laser system used are also shown in Table 1. In comparison to the ns laser the same wavelength regime was used, resulting in comparable absorption between the IR laser light and the dielectric layers and Si substrate, respectively. Differences arise from the shorter pulse length of 300 fs. Within this very short time the incoupled power is much higher, heat diffusion can be neglected. The ablation mechanism is based on the very short and intense interaction with the passivation stack and the Si substrate Both, direct and indirect ablation takes place 9 , resulting in very flat interaction depth with Si of only some micron. When material is removed, a very small layer of amorphous Silicon is remaining. For local removing the passivation stack SiN X /AlO X from a Si substrate using a fs laser a moderate influence on the Si substrate can be assumed. In Fig. 5 such openings are shown, indicating the melting process. Debris Bare silicon surface Overlap Fig. 5 Micrographs of typical local openings fabricated by a fs laser. 2.3 Etching paste structuring The etching paste used to locally remove the passivation stack was isishape LBSF paste. The paste can be screen printed using common techniques and is based on phosphoric acid. The printed paste must be activated at temperatures higher than 300 C to etch off the dielectric. After processing, the substrate must be cleaned to remove remaining residues of paste and dielectric layers. Because of the selectiveness of this etching process against Si, only the 105 nm passivation stack is etched off. For local removing the passivation stack SiN X /AlO X from a Si substrate using screen printable LBSF etching paste we speak more general of a no significant influence on the Si substrate 10 . Fig. 6 Pictures from an optical microscope of a typical opening fabricated by etching paste. 2.4 Metallization The samples were metallized by means of screen printing on this side of the pieces where the passivation layer intentionally was locally opened. Two different Al pastes were used in the experiment, which are different in application and composition as described in Table 2. Table 2 Metallization pastes used. Al 1 Al 2 Type industrial full area BSF paste L-BSF Al paste The firing was performed in an IR belt furnace using three different temperature profiles as shown in Fig. 7. The belt speed were changed at same temperature settings. This results in comparable profiles with same shape and maximum temperatures, but different length and thus temperature gradients. These were for the cooling from maximum temperature to 600 C roughly 65, 60 and 50 K/s for the belt speeds of 3,3, 3,0 or 2,7 m/min respectively. 0 200 400 600 800 1020304050 Time /s T e m p er at u r e / C 3,3 m/min 3,0 m/min 2,7 m/min Fig. 7 Substrate temperatures applied to the samples with same temperature settings but different belt speeds. A maximum temperature for all settings of 800 C was measured, but different overall times and thus, temperature gradients. Data were recorded using a datalogger from DATAPAQ using the newly developed sample clamp. 2.5 Sample preparation Cross section of samples were investigated using scanning electron microscopy SEM. The preparation was done by breaking the sample, etching the sample in a selective etch that etches differently doped regions differently. Thus, using SEM the local BSF region can be directly investigated either quantitatively when measuring the thickness or qualitatively when looking at the homogeneity. On the broken and etched samples, at least three SEM pictures were taken and analyzed with respect to the homogeneity of the contact formation and BSF homogeneity and thickness. BSF thickness was measured in different areas. Further, in the SEM picture the distance d 2 was measured, what denotes the longest length in the interface to the aluminum paste matrix, cf. Fig. 8. 3 RESULTS After the structuring described in the experimental section, representative samples were investigated by optical microscope to determine the opening area, correlated with a “diameter” d 1 . d 1 is either a diameter in case of a dots or a width in case of lines compare also Fig. 8. The distribution of d 1 values for different opening methods is given in Fig. 9. 100 m At edges, dielectric partly lift-off Bare rough Si surface amorphous d 1 d 2 Fig. 8 Diameters d 1 and d 2 given for two examples. The optical microscope picture left was taken prior to metallization as top view, the SEM cross section after metallization and firing. nsDot Dot nsLine Line fsDot Dot fsLine Line EtchDot Dot EtchLine Line 0 20 40 60 80 100 120 Etching Pastefs laser Openi ng di am e t er d 1 / m ns laser Fig. 9 Distribution of different opening diameters d 1 achieved for the different opening methods and patterns dots and lines. In Fig. 10 the correlation between both the values d1 and d2 are shown in dependence to the patterning of dots or lines. The distribution should be following a linear trend, but is wildly scattered around the 1/1 line, independently from the type of processing. Later this graph is discussed more in detail. 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 d 1 after structuring /m d 2 a f ter A l -fi r i n g / m Etching paste dots Etching paste lines fs laser dots fs laser lines ns laser dots ns laser lines Fig. 10 Correlation of the relation between the diameter of the opened area after structuring abscissa d 1 and the maximum diameter of the point contact after firing ordinate d 2 . All SEM pictures further shown are representatives for the given parameter set. A further description with respect to the BSF quality and the filling is given with the parameter BSF thickness and the homogeneity of the BSF as well as the availability of voids empty spaces in the contact area. 3.1 Influence of thermal budget The influence of the firing profile on contact-area formation was investigated. For the ns laser opening of dots in combination with Al 2 the contacts are well developed for the high belt speed, see Fig. 11. The BSF is very homogeneous and has a thickness of approximately 2 m. The thickness of the BSF is strongly decreased and becomes more inhomogeneous and interrupted with lower belt speed, revealing more and more voids. In contrast, for line contacts the results is the other way around Better developed contacts are observed for lower belt speeds. This is the same for line contacts opened with fs laser and metallized with Alu 1, although the BSF thickness is very thin with values 0.3 m, cf. Fig. 12. 2 m homogeneous BSF SomeVoids Firing 1 3,3 m/s Many 1,5 m inhomogeneous Firing 2 3,0 m/s Many 1 m inhomogeneous Firing 3 2,7 m/s Fig. 11 Influence of different firing profiles on the contact formation. Structuring was made with the ns laser producing dots of 90 m. Al 2 was used. 0 m inhomogeneous BSF ManyVoids Firing 1 3,3 m/s None 0,3 m homogeneous Firing 2 3,0 m/s None 0,1 m homogeneous Firing 3 2,7 m/s Fig. 12 Influence of different firing profiles on the contact formation. Structuring was made with the fs laser producing lines of 50 m. Al 1 was used. Although, here only two parameter sets were