"LED FLASHER ARRAYS (LFA) FOR AN IMPROVED QUALITY CONTROL - PDF - PDF"
LED FLASHER ARRAYS (LFA) FOR AN IMPROVED QUALITY CONTROL IN SOLAR CELL PRODUCTION LINES R. Grischke1, J. Schmidt1, H. Albert2, A. Laux2, A. Metz3, U. Hilsenberg4, J. Gentischer5 1 Institut für Solarenergieforschung Hameln/Emmerthal (ISFH), Am Ohrberg 1, D-31860 Emmerthal, Germany, email@example.com, Fax: +49-5151-999-400, Tel: +49-5151-999-424 2 h.a.l.m. elektronik, Sandweg 30-32, D-60316 Frankfurt am Main, Germany 3 RWE Schott Solar, Carl-Zeiss Str. 4, D-63755 Alzenau, Germany 4 Vossloh-Schwabe Wustlich, Carl-Friedrich-Gauss-Str. 3, D-47475 Kamp-Lintfort, Germany 5 ACR Automation in Cleanroom, Villinger Str. 4, D-78078 Niedereschach, Germany ABSTRACT: The development of new solar cell measurement equipment based on an array of LEDs (LED-Flasher- Arrays, LFA ) is the aim of a joint R&D project. The increased lifetime of the LFA compared to conventional light sources such as Xenon flashers leads to strongly reduced service times. Further options are spatially and spectrally resolved measurements. The modular construction allows the extension for module testing. Extensive computer-based simulations were carried out for n+p silicon solar cells using the simulation tool PC1D. The relative deviations in the solar cell characteristics resulting from monochromatic instead of the AM1.5G illumination were studied as a function of the cell parameters. The main deviations occur in short-circuit current density while the deviations in open-circuit voltage and fill factor are less than 1 %. We demonstrate that for silicon solar cells with cell parameters varying over a broad range, the systematic error in the measured efficiency by illumination with a mix of three monochromatic wavelengths at a well-defined intensity ratio is in the range of ± 1.4 %. Hence, a 3- colour LFA has the potential to outperform today’s state-of-the-art flash light sources. Keywords: Characterisation, Qualification and Testing, Simulation 1. INTRODUCTION 2. NUMERICAL SIMULATIONS In the course of an R&D project including the Extensive computer-based simulations were carried partners h.a.l.m. elektronik, RWE Schott Solar, Vossloh- out for n+p silicon solar cells to determine the effect of Schwabe Wustlich, ACR and ISFH, a novel solar cell test monochromatic illumination on cell characteristics. The equipment has been developed, based on a matrix relative deviations in the open-circuit voltage (Voc), the assembly of light emitting diodes (LED-Flasher-Array, short-circuit current density (Jsc), the efficiency (η) and LFA ). The modularly designed LFA allows the the fill factor (FF) resulting from monochromatic application in large-scale production both of solar cells illumination instead of the AM1.5G illumination, were and solar modules. The increase in lifetime compared to studied as a function of the cell parameters using the conventional light sources such as continuously emitting simulation tool PC1D . lamps (e.g. halogen) or flash lamps (Xenon) leads to First, we have defined a “calibration cell” with strongly reduced service times (LED lifetime: 100,000 h, typical cell characteristics for a commercial screen- Xenon bulbs: 1,000 h). The spatial and temporal printed silicon solar cell under standard testing homogeneity of irradiance is adjusted by the individual conditions (25 °C, AM1.5G, 100 mW/cm2). The current control of the LEDs and can be easily calibrated wavelengths were varied between 430 nm and 940 nm. during production run time. Further options are spatially The individual intensity for each wavelength was resolved measurements and a multicolour assembly adjusted using the calibration cell so that the short-circuit giving the possibility of spectrally resolved current density reached the same value as under standard measurements and an improved accuracy. testing conditions with AM1.5G illumination (see Fig. 2). In order to include different types of n+p solar cells, the cell parameters were varied over a very broad range. These parameters are the series resistance Rs, the shunt resistance Rsh, the cell thickness W, the thickness of the antireflection coating drf, its refractive index n, the emitter depth xj, its peak doping concentration ND (Gaussian distribution), the surface recombination velocities at the front and rear Sf and Sr as well as the bulk recombination lifetime τb (Table I). Finally, this parameter field was reduced to a realistic range, reflecting realistic cell parameters occurring in an industrial cell production. In the beginning, each relevant cell parameter was varied separately. Subsequently, the interactions of the Figure 1: CetisPV-LF1, LED-Flasher-Array prototype different parameters were taken into account. by h.a.l.m. elektronik. varied. The point of intersection of the curves correspond 2.5 to the defined calibration cell. Both positive and negative deviations as a result of monochromatic illumination with different wavelengths allow the compensation of the 2.0 systematic errors by combining at least two different wavelengths. By adding a third wavelength, the deviation Normalised Intensity can be further reduced. 1.5 Short-circuit Current Density Jsc (mA/cm2) 32 1.0 0.5 30 0.0 28 400 500 600 700 800 900 1000 AM1.5 G 470 nm Wavelength λ (nm) 880 nm 26 880nm / 470nm (1.6 / 1) Figure 2: Intensity of monochromatic illumination 940nm/ 700nm/ 470nm normalised to the AM1.5G intensity using the defined (1.2 / 1.7 / 1) calibration cell. 24 1 10 100 1000 2.1 Single-parameter variations Bulk Lifetime τb (µs) The simulations show that series and shunt resistance as well as the emitter profile of the solar cell have Figure 4: Short-circuit current density as function of negligible effects on the deviations in the cell bulk lifetime for different monochromatic wavelengths characteristics. The variations in the antireflection and the combinations of wavelengths. coating, the cell thickness and the recombination parameters in the bulk and at the surfaces provoke the 2.2 Multiple-parameter variations main deviation in short-circuit current density and The interaction of different cell parameter variations consequently in efficiency, while the deviations in open- leads to much larger deviations. 60,000 parameter circuit voltage and fill factor are less than 1 %. combinations were simulated to include the whole Figure 3 shows the deviations in the cell parameter field of Table I. characteristics as a function of bulk lifetime for monochromatic illumination with a wavelength of 850 Table I: Simulated parameter range. nm. Especially at bulk lifetimes below 5 µs very large Parameter Unit Minimum Maximum errors occur. Series resistance Rs Ωcm2 1 x 10-3 10 2 Shunt resistance Rsh Ωcm2 1 x 103 1 x 106 Thickness of solar cell W µm 150 400 Thickness of antireflection coating drf nm 0 110 Refractive index of ARC n 1.5 2.5 0 Emitter doping concentration ND cm-3 3.5 x 1019 1 x 1020 Emitter depth xj µm 0.1 0.2 Surface recombination velocities at Sf 1 x 106 Deviation (%) cm/s 1 x 10² -2 front and rear Sr Bulk recombination τb µs 1 1000 -4 ∆Voc / Voc As already mentioned, a mix of more than one ∆Jsc / Jsc wavelength can reduce the measurement error ∆η / η considerably. An optimum was found for a combination -6 ∆FF / FF of the wavelengths 940 nm, 700 nm and 470 nm at an intensity ratio of 1.21 : 1.74 : 1. That means 30.6 % of the short-circuit current results from infrared light, -8 44.1 % from red light and the remaining 25.3 % from the 1 10 100 1000 blue component of the spectrum. The first four bars in Fig. 5 show the dependence of Bulk Lifetime τb (µs) the deviation in short-circuit current density on the Figure 3: Deviation in cell characteristics by wavelength combinations, applying the whole parameter illumination with a monochromatic wavelength of range from Table I. The last two bars correspond to the 850 nm instead of the AM1.5G spectrum. realistic parameter space. Obviously, the 3-colour LFA can decrease the deviation below the error typical for Figure 4 shows the impact of the wavelength on the Xenon flash lamps as used in commercial solar cell short-circuit current density. Again the bulk lifetime was production lines. performance (see Fig. 7). Another cell of the same batch 20 was used for calibration. 0 17.7 9.3 4.1 2.2 1.4 1.1 10 -0.25 -0.28 ∆ Isc / IscAM1.5G [%] 0 -1 -1.86 -3.49 -28.5 -4.7 -5.0 -3.1 -1.4 -2.8 ∆ Jsc [%] -10 -2 realistic Simulation -20 parameter 880/470 nm space LFA prototype -3 -30 0 0 0 0 0 88 /47 /47 /47 /47 8 80 42 00 00 0/6 0/7 0/7 -4 88 94 94 CELL 1 CELL 2 Spectrum [Wavelengths in nm] X Data Figure 7: Measured and simulated deviations in the Figure 5: Maximally expected systematic error in short- short-circuit current density of EFG silicon solar cells circuit current density at different spectra compared to from different classes. The measurements have been AM1.5G spectrum. performed using a 2-colour LFA prototype (wavelengths combination 880 / 470 nm). Although the spectrum of a conventional Xenon flasher is much more similar to the AM1.5G spectrum, The thickness of the antireflection coating and its which is also recognisable in the profile of the generation refractive index were measured by ellipsometry, the rate (Fig. 6), the 3-colour LFA spectrum gives better spatially resolved light-beam induced current method results. This is due to the optimal compensation of (LBIC) was used to evaluate the spatial distribution of positive and negative deviations. the minority-carrier diffusion length. 1022 AM1.5 G 1021 880 nm 880nm / 470nm (1.6 / 1) Generation Rate (cm-3s-1) 1020 940nm/ 700nm/ 470nm (1.2 / 1.7 / 1) Xenon Flasher 1019 1018 1017 1016 1015 0 50 100 150 200 250 300 Penetration Depth (µm) Figure 6: Generation profiles in a silicon solar cell for different spectra. 3. EXPERIMENTAL VERIFICATION To experimentally verify the simulation results, two exemplary solar cells were picked out of the EFG silicon solar cell production line at RWE Schott Solar. Both cells were measured using a 2-colour LFA prototype Figure 8: LBIC analysis of the EFG reference cell. (wavelengths combination 880 / 470 nm). CELL 1 is a high-quality solar cell, whereas CELL 2 is a cell of lower Because of the wavy nature of EFG wafer surfaces, the exact determination of the cell thickness as well as the refractive index and thickness of the antireflection coating is very erroneous. Furthermore, the diffusion length mappings of the wafers show a very inhomogeneous distribution. However, we were able to detect areas of similar diffusion lengths. Hence, in our simulations shown in Fig. 7 we implemented three areas of different bulk lifetime approximately representing our diffusion length mappings. The larger difference between measured and simulated deviation is observed for CELL 2.We attribute the large discrepancy between simulation and experiment to the inhomogeneity in the bulk lifetime of the used reference cell (see Fig. 8) and the not exactly known spectrum of the flash lamp which was used as reference light source. In order to verify the improved accuracy of the 3-colour LFA, a large-scale experiment is planned in which a 3-colour LFA and for direct comparison a Xenon Figure 8: Prototype of a 3-colour LFA (80 × 80 mm²) flash lamp will be integrated in a solar cell production with four 3-colour LED heads on printed circuit boards line. before encapsulation. 4. ADVANCED VERSIONS OF THE LFA ACKNOWLEDGEMENT While the 2-colour LFA was composed of discrete This project has been supported by the German LEDs (see Fig. 1), a 20 × 20 mm² large 3-colour LED Bundesministerium für Wirtschaft und Technologie head was developed by h.a.l.m. elektronik in cooperation (BMWi) under project No 0329911C. ISFH is supported with Vossloh-Wustlich-Schwabe (Fig. 8). Each LED by the state of Niedersachsen and is a member of the head includes 144 surface-mounted light emitting diodes Forschungsverbund Sonnenenergie. (SMD-LEDs) of the required colours to realise the target spectrum by supplying the LEDs with nearly nominal current. A newly designed set of printed circuit boards  World patent pending by ACR. (PCB) had to be developed by h.a.l.m. for current control  Clugston, D.A. and Basore, P.A., “PC1D Version 5: and regulation of these chip-on-board light heads as well 32-bit Solar Cell Simulation on Personal as for processor control. A sealing encapsulation will Computers,” proc. 26th IEEE Photovoltaic protect the bonded wires and optimise the radiation of the Specialists Conf., Anaheim, CA (IEEE, New York, light heads. A first 3-colour LFA prototype has just been 1997), p. 207. completed and will be compared with our simulation  CEI/ IEC Publication 904-9: Photovoltaic devices, results in the near future. Part 9: Solar simulator performance requirements, A further development of a 6-colour LFA would 1995. allow categorising the LFA by the International  R. Grischke, J. Schmidt, Leuchtdioden-Flasher- Electrotechnical Commission standard for solar simulator Arrays für die verbesserte Qualitätskontrolle im performance requirements . The advantages in Fertigungsprozess: Simulation der Solarzellen- temporal and spatial uniformity with a well-adapted Kenngrößen bei monochromatischer Beleuchtung, spectrum by adding three wavelengths in the range of 18. Symposium Photovoltaische Solarenergie, 550 nm, 650 nm and 850 nm will classify the LFA as a Kloster Banz, Germany 2003, p. 332. Class A simulator. 5. CONCLUSIONS For a wide range of silicon solar cells the systematic error in the measured efficiency by illumination with a mix of the monochromatic wavelengths 940 nm, 700 nm and 470 nm at an intensity ratio of 30.6 % / 44.1 % / 25.3 % is in the range of ± 1.4 %. In spite of the very different spectral distribution of the 3-colour LFA compared to conventional light sources, the expected measurement error is below that of presently used flash light sources. Hence, the 3-colour LFA has the potential to outperform today’s state-of-the-art flash light sources. A prototype of the 3-colour LFA on the basis of specially designed 3-colour LED heads has just been built and is currently under detailed investigation.