Solar Energy Materials & Solar Cells

Transcription

1 Solar Energy Materials & Solar Cells 95 (2011) Contents lists available at ScienceDirect Solar Energy Materials & Solar Cells journal homepage: An inter-laboratory stability study of roll-to-roll coated flexible polymer solar modules Suren A. Gevorgyan a, Andrew J. Medford a, Eva Bundgaard a, Subarna B. Sapkota b, Hans-Frieder Schleiermacher b, Birger Zimmermann b, Uli Würfel b, Amine Chafiq c, Monica Lira-Cantu c, Thomas Swonke d, Michael Wagner d, Christoph J. Brabec d, Olivier Haillant e, Eszter Voroshazi f,g, Tom Aernouts g, Roland Steim h, Jens A. Hauch h, Andreas Elschner i, Michael Pannone j, Min Xiao j, Anthony Langzettel j, Darin Laird j, Matthew T. Lloyd k, Thomas Rath l,m, Eugen Maier m,n, Gregor Trimmel l,m, Martin Hermenau o, Torben Menke o, Karl Leo o, Roland Rösch p, Marco Seeland p, Harald Hoppe p, Timothy J. Nagle q, Kerry B. Burke q, Christopher J. Fell q, Doojin Vak r, Th.Birendra Singh r, Scott E. Watkins r, Yulia Galagan s, Assaf Manor t, Eugene A. Katz t,u, Taehee Kim v, Kyungkon Kim v, Paul M. Sommeling w, Wiljan J.H. Verhees w, Sjoerd C. Veenstra w, Moritz Riede x,o, M. Greyson Christoforo x, Travis Currier y, Vishal Shrotriya y, Gregor Schwartz z, Frederik C. Krebs a,n a Risø National Laboratory for Sustainable Energy, Technical University of Denmark, Frederiksborgvej 399, DK-4000 Roskilde, Denmark b Fraunhofer Institute for Solar Energy Systems ISE, Heidenhofstrasse 2, D Freiburg, Germany c Centre d Investigació en Nanociencia i Nanotecnologia (CIN2, CSIC), Laboratory of Nanostructured Materials for Photovoltaic Energy, ETSE, Campus UAB, Edifici Q, 2nd Floor, E-08193, Bellaterra (Barcelona), Spain d Bavarian Center for Applied Energy Research, Am Weichselgarten 7, D Erlangen, Germany e Atlas Material Testing Technology GmbH, Vogelsberstrasse 22, Linsengericht-Altenhasslau, Germany f IMEC VZW, Organic Photovoltaics, Kapeldreef 75, B-3001 Leuven, Belgium g Katholieke Universiteit Leuven, ESAT, Kasteelpark Arenberg 10, B-3001 Leuven, Belgium h Konarka Technologies GmbH, Landgrabenstr. 94, Nürnberg, Germany i H.C. Starck Clevios GmbH, Chempark Leverkusen Build. B202, D Leverkusen, Germany j Plextronics, Inc., 2180 William Pitt Way, Pittsburgh, PA 15238, USA k National Renewable Energy Laboratory, Golden, CO 80401, USA l Christian Doppler Laboratory for Nanocomposite Solar Cells, Graz University of Technology, Stremayrgasse 9, 8010 Graz, Austria m Institute for Chemistry and Technology of Materials, Graz University of Technology, Stremayrgasse 9, 8010 Graz, Austria n Polymer Competence Center Leoben GmbH, Roseggerstrasse 12, 8700 Leoben, Austria o Institut fuer Angewandte Photophysik, Technische Universitaet Dresden, George-Baehr-Straße 1, Dresden, Germany p Institute of Physics, Ilmenau University of Technology, Weimarer Str. 32, D Ilmenau, Germany q CSIRO Future Manufacturing Flagship and CSIRO Energy Technology, 10 Murray Dwyer Circuit, Mayfield West, NSW 2304, Australia r CSIRO Future Manufacturing Flagship and CSIRO Materials Science and Engineering, Ian Wark Laboratory, Clayton, VIC 3168, Australia s Holst Centre, P.O. Box 8550, 5605 KN Eindhoven, The Netherlands t Department of Solar Energy and Environmental Physics, J. Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boker Campus 84990, Israel u The Ilse Katz Institute for Nanoscale Science and Technology, Ben-Gurion University of the Negev, Beersheva 84105, Israel v Solar Cell Center, Korea Institute of Science and Technology, P.O. Box 131, Cheongryang, Seoul , Republic of Korea w Energy Research Centre of the Netherlands (ECN), P.O. Box 1, 1755 ZG Petten, The Netherlands x Department of Electrical Engineering, Stanford University, Stanford, California, CA-94305, USA y Solarmer Energy, Inc., 3445 Fletcher Avenue, El Monte, CA 91731, USA z Heliatek GmbH, Treidlerstr. 3, Dresden, Germany article info Article history: Received 12 November 2010 Accepted 7 January 2011 Available online 15 February 2011 Keywords: Round robin Inter-laboratory study abstract A large number of flexible polymer solar modules comprising 16 serially connected individual cells was prepared at the experimental workshop at Risø DTU. The photoactive layer was prepared from several varieties of P3HT (Merck, Plextronics, BASF and Risø DTU) and two varieties of ZnO (nanoparticulate, thin film) were employed as electron transport layers. The devices were all tested at Risø DTU and the functional devices were subjected to an inter-laboratory study involving the performance and the stability of modules over time in the dark, under light soaking and outdoor conditions. 24 laboratories from 10 countries and across four different continents were involved in the studies. The reported n Corresponding author. address: frkr@risoe.dtu.dk (F.C. Krebs) /$ - see front matter & 2011 Elsevier B.V. All rights reserved. doi: /j.solmat

2 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Polymer solar cells Flexible modules Outdoor testing R2R manufactured OPV results allowed for analysis of the variability between different groups in performing lifetime studies as well as performing a comparison of different testing procedures. These studies constitute the first steps toward establishing standard procedures for an OPV lifetime characterization. & 2011 Elsevier B.V. All rights reserved. 1. Introduction Round robins (RR) and inter-laboratory studies (ILS) are useful methods to reach a consensus on solar cell performance. This has been employed recently for organic solar cells [1] and in the past for inorganic solar cells [2 9]. Furthermore, RR and ILS can help in establishing standard procedures for accurate quantification of device performance. In the case of organic solar cells that present a dynamic response and often pronounced degradative behavior [10] it is of importance to evaluate the stability of these devices and gain consensus on what stability means, how it is observed and how it is quantified. The organic photovoltaics (OPV) community does make use of calibration laboratories, such as NREL (US) and Fraunhofer ISE (Germany), for standardized efficiency measurements, but there is an urgent need for internationally accepted ageing and test procedures and it is desirable to develop and maintain some standard procedures for measuring and reporting the stability of OPV devices. Furthermore, flexible roll-to-roll (R2R) processed solar cells represent the most feasible scale-up route for the polymer solar technology and it is thus of interest to investigate the stability of such devices under a variety of conditions and outdoor climates. The two International Summits on Organic Photovoltaic Stability (ISOS-1 and ISOS-2) have attempted the establishment of standard protocols for lifetime tests [11] and ISOS-3 has taken this further through experimental studies involving a large group of competing laboratories. An inter-laboratory stability study provides a powerful route to simultaneously gain insight into the variance between stability measurements under a set of conditions in various laboratories and can explore the effect of a wide range of testing conditions and materials. In this report, we detail results from such an inter-laboratory study of flexible R2R processed polymer solar cells with various layer structures that were prepared in one location (Risø DTU, DK) and distributed to a multitude of laboratories where shelf life and light soak stability tests as well as outdoor exposure testing were carried out under a set of pre-determined conditions. Based on the data that were collected, we evaluate the current status of standardization between different laboratories involved with polymer solar cell research and correlate stability data from testing under simulated and actual outdoor conditions. 2. Experimental procedure and methodology 2.1. Manufacture of polymer solar cells The manufactured devices had the following structure: PET/ ITO/electron transporting layer (ETL)/active layer/pedot:pss/ag paste. The types of ETL as well as the types of Poly (3-hexylthiophene) (P3HT) polymers used in a bulk mixture with [6,6]- phenyl-c 61 -butyric acid methyl ester (PCBM) for active layers are presented in Table 1. Four different device variations were produced, as listed in Table PET/ITO The flexible substrate comprised ITO covered poly(ethylene terephthalate) (PET) (130 mm) in rolls having a roll width of 305 mm and a length of 100 m. The nominal sheet resistance was 60 O square 1. The desired striped pattern of the ITO was prepared by printing a UV-curable etch resist in areas of the ITO pattern in a full R2R process. The ITO was subsequently etched using a full R2R etching machine comprising etching baths (CuCl 2 ), stripping baths (NaOH), washing baths (demineralized water) and drying sections (hot air). The thickness of the ITO was 80 nm. The process has been described earlier [12] ZnO Zinc oxide nanoparticles (ZnO np) were prepared in acetone solution and stabilized with methoxyethoxyacetic acid (MEA). The preparation of thin film ZnO was included in this effort as it presents devices without an inflection point in the IV-curve [14]. The ZnO layer was applied using a modified slot-die coating procedure [13]. The films were coated with a wet layer thickness of 5 m at a speed of 2 m min 1. The drying temperature was 140 1C (30 s drying time) Active layer Four polymers presented in Table 1 were employed: Sepiolid P200 P3HT purchased from BASF; Plextronics P3HT provided by Plextronics; Merck P3HT provided by Merck. The Risø DTU P3HT was home made. Polymers were dissolved in chlorobenzene followed by an addition of PCBM (purchased from Solenne). A final concentration of 22 mg ml 1 P3HT and 20 mg ml 1 PCBM was employed. The films were coated with a wet layer thickness of 8 m at a speed of 2 m min 1. The drying temperature was 140 1C (30 s drying time) PEDOT:PSS Poly(3,4-ethylenedioxythiophene)poly(styrenesulfonate) PEDOT:PSS was purchased as a screen printing paste from Agfa (Orgacon EL-P 5010). It was diluted with isopropanol (1000 g PEDOT:PSS was mixed with 500 g isopropanol) and shaken vigorously for 1 h. The viscosity was around 200 mpa s. Wetting of the surface of the active layer with isopropanol prior to PEDOT:PSS coating improves the wetting and coating. PEDOT:PSS was dried at temperatures up to 140 1C with a residence time in the oven of around 5 min Printed silver electrode The silver electrode was a UV-curing type [15] that was screen printed on a flat bed screen printer with a 120 mesh screen having the desired pattern with a small outline and 16 serially connected cells as described earlier [16] Encapsulation, contacting and cutting of the modules The encapsulation was achieved by cold lamination of a barrier foil carrying an adhesive on both sides of the completed solar cell. Table 1 Active polymer and electron transport layer for various module types. Module type Active material Electron transport layer RN Risø DTU P3HT Nanoparticulate ZnO ST Sepiolid P3HT (BASF SE) Thin film ZnO PN Plextronics P3HT Nanoparticulate ZnO MN Merck P3HT Nanoparticulate ZnO

3 1400 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) The barrier foil contained UV filter cutting the UV irradiation from 390 nm. The barrier on the back side was made such that a part of the printed silver electrode was exposed. This served two purposes. Firstly, it enables reliable contacting and secondly it will allow for cells to exhibit some degradation over the course of this experiment. The modules were cut manually into individual devices and button contacts were applied, using an automated machine. All steps and machinery has been detailed earlier [15] Module geometry The geometric sizes of each module were 12 cm 8.5 cm. The active area was comprised of 16 serially connected stripes with total area of 35.5 cm 2 to be illuminated. Fig. 1 shows the sketch of the module Inflection point It has been observed previously that R2R cells produced using the nanoparticulate ZnO ink exhibit dynamic performance due to an inflection point in the I V curve, which appears after dark storage (s-shaped curves). The inflection is sometimes observed immediately after production, and sometimes appears only after several days/weeks of dark storage [14]. It was discovered that this inflection point can be removed via a photo-annealing process, in which the cells are left under illumination for min. Although this phenomenon is not entirely understood, the role of ZnO photoconductivity is known to be critical, and thus by using light sources with higher UV contents the photoannealing occurs faster. Although the encapsulation of modules contains a UV barrier (cutting from 390 nm), it still leaves a gap of absorption range of nm for ZnO, where the photodoping can take place. Elevated temperatures can accelerate the doping process as well. The inflection point was studied in detail and reported by Lilliedal et al. [14]. The groups that would undertake the further degradation studies of the cells were advised to photo-anneal modules after receiving them in order to restore their optimal performance Module selection and package distribution among groups The performance of all modules was first recorded using a R2R measurement setup [15]. The setup is known to deviate slightly from measurements outdoors, due to spectral mismatch and inaccuracies in calibration of intensity, and considerable time passed between R2R characterization and further calibrated measurements. However, the setup allows for rapid characterization of a large number of cells, and its results are still comparatively useful. The data were used to select functional modules, and estimate that the power conversion efficiency of all modules were within 70.3%. Ideally the performance of all modules would have been recorded under carefully calibrated conditions; however, the amount of time needed to do this was deemed impractical for such a large study, and thus only one cell per package (to be sent to each of the participating laboratories) was carefully measured under both indoor and outdoor test conditions. Secure contacts were attached by placing a button contact on electrodes of each device. Previous studies at Risø DTU have shown that this in most cases is sufficient to ensure contact throughout the lifetime of the cell. A total of 340 modules were distributed to a total of 24 groups. The modules were prepackaged into 9 samples for standard and 18 samples for extended packages, placed in bubble wrap envelopes and shipped to groups using ordinary UPS mail service. The packages also included the necessary instructions for testing procedures and a laser cut black cardboard mask for masking the calibrated module prior to measurement. The samples were shipped on the 19th of May 2010 and received on the 24th of May An with a description of the ILS stability studies and a request of participation was sent to the list of participants of the ISOS-3 conference. The laboratories that confirmed the participation in the studies received a package of modules for testing. The starting date of the experiments was set to the 28th of May 2010 and the studies were scheduled to be completed on the 15th of July Test procedures Fig. 1. The sketch of the module front side. The dimensions of the module are shown as well. Several standardized testing procedures that were developed based on the proceedings of ISOS-1/ISOS-2 [11] were distributed along with the cells. It was also suggested that the participating laboratories perform and report any additional tests which they considered interesting. The standards were simplified in some cases in order to meet capabilities of a wider range of OPV research groups [17]. The short summary of different procedures of testing is summarized in Table 2. The documents with the Table 2 Short summary of five stability testing procedures. Test ID ISOS-D-1 ISOS-D-2&3 ISOS-L ISOS-O-1 ISOS-O-2 Description Shelf High temperature storage Laboratory Outdoor Outdoor and damp heat simulations Light None None 0.6 to 1 sun Stored outdoor, measured indoor Stored outdoor, measured outdoor Temperature Ambient Controlled, C Controlled, C Ambient Ambient Humidity Ambient (Controlled) (Controlled) Ambient Ambient Environment Drawer Chamber/dry oven (Light soaking Ambient Ambient chamber) Logging interval 1 day to 1 week 1 day to 1 week 1 15 min 1 day to 1 week 15 min to 1 day Load None None None None None

4 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) instructions sent with packages can be found in the supporting information to this article. Three basic types of measurements were chosen, which involved: Shelf life studies performed by leaving modules in the dark in either ambient conditions (ISOS-D-1 Shelf) or in controlled temperature/humidity chambers (ISOS-D-2 high temperature storage and ISOS-D-3 damp heat). Light soaking in indoor conditions (ISOS-L laboratory simulations) performed by placing the modules in simulated sunlight under laboratory conditions. The modules were either stored under illuminated conditions and measured continuously or were periodically taken out and measured under calibrated sun simulators. Outdoor studies under real sun were performed either by storing and measuring modules under natural sunlight (ISOS- O-2), or by bringing the modules indoors (daily or weekly) and testing them under a calibrated light source (ISOS-O-1). The test IDs differ from the experimental IDs used for the customized procedures recommended to the groups in the following way: ISOS-D-1 corresponds to T1A ISOS-D-2&3 corresponds to T1B ISOS-L corresponds to T2 ISOS-O-1 corresponds to T3B ISOS-O-2 corresponds to T3A The recommended procedures can be found in the supporting information 2.5. Calibrated module testing As previously mentioned each package contained one module, which was carefully measured under both indoor and outdoor conditions at Risø DTU prior to shipping. Only the ST type modules (see Table 1) were chosen for calibration studies. The cells were masked using lasercut masks with an aperture of cm 2 in order to limit illumination to the active area. The nominal active area of the cells was 35.5 cm 2, which was used for determination of the PCE of the modules. Cells were tested under indoor conditions using a KHS solar constant 575 solar simulator which was calibrated to an intensity of 1000 W m 2, using a pyranometer mounted on the test platform (mismatch corrections were not performed). The temperature was monitored during testing using a thermocouple. No active cooling was used. A photograph of the test setup is shown in Fig. 2. Modules were placed under the simulator for approximately 5 min, while the temperature stabilized. The device temperature varied between C and was recorded at the beginning of each scan. Scans were taken from 1 to 10 V at a sweep rate of 100 mv/s and a step speed of 100 ms (scan time of 11 s), and repeated five times to ensure that the performance was stable. After the final scan, the cell was covered and a dark scan was taken. Outdoor testing was conducted by mounting the masked cells on a solar tracker at Risø DTU. The intensity of solar illumination was measured using a bolometer attached to the solar tracker, and temperature was monitored by attaching a thermocouple to the back of a reference cell. The temperature varied between 22 and 28 1C depending on wind speed, orientation and illumination. IV scans were recorded under the same conditions as during indoor testing, but were only repeated once due to lack of constant solar flux. The days of measurements were somewhat cloudy, which resulted in light reflections from clouds and Fig. 2. Measurement setup used for indoor testing including pyranometer (dome at the back) and thermocouple (white wire). The solar cell is shown with the mask mounted in front such that only the desired aperture is illuminated. fluctuations in solar flux (720 W m 2 ). The characterizations of the modules under both indoor and outdoor conditions were performed during of May 2010 and the packages were shipped to recipient laboratories on the 19th of May The groups were instructed to measure and report the performance of the calibrated modules under a calibrated light source and/or natural sunlight prior to degradation studies. 3. Results and discussion 3.1. Participating laboratories The study comprised five types of stability tests plus an accurate quantification of the performance of the calibrated module and groups had to choose which and how many of experiments they wished to perform. Two laboratories out of 24 performed different studies (not discussed here). Table 3 shows the list of 22 groups together with the geographic locations of the laboratories and the types of experiments performed. The reported data were analyzed according to the category of experiments and presented in the next sections. Measurements were carried out using all types of modules described earlier (Table 1). However, in this document only ST type modules will be discussed due to more stable and reproducible performance, while studies of other types will be reported elsewhere Performance variability of the calibrated modules As mentioned in the experimental section, each package contained a calibrated device, which was carefully measured at Risø DTU under both indoor and outdoor conditions prior to shipping and the recipient groups were asked to carry out a similar performance measurement. The purpose of this study was, firstly, to quantify the ability of groups to perform accurate performance measurement and secondly, to compare the results with previously reported ILS studies of R2R manufactured OPV modules [1]. While the solar cells employed in this study were encapsulated, the encapsulation was not complete (no edge sealing) to allow for some degradation to be observed over the 1000 h that experiments were intended to endure. A fully edge sealed version of these solar cell modules is known to present significantly better stability Measurements at Risø DTU A correlation between outdoor and indoor measurements of the calibrated modules at Risø DTU revealed that cell performance

5 1402 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Table 3 The list of the participating laboratories with the map showing their locations. # Contact Person Laboratory name (country) Latitude, longitude Experimental test IDs a Test IDs (ISOS-3 protocols) b 1 Riede, M. Stanford University (USA) N, W T3A ISOS-O-2 2 Wonders, J. and Haillant, O. Atlas Weathering Services Group (USA) 33 o 53 0 N, W T3B c ISOS-O-1 3 Rivera, J. and Haillant, O. Atlas South Florida Test Service (USA) 25 o 33 0 N, W T3B c ISOS-O-1 4 Xiao, M. Plextronics Inc. (USA) N, W T1A, T1B, T2 ISOS-D-1&3, ISOS-L 5 Shrotriya, V. Solarmer (USA) N, W T1A, T1B, T3B ISOS-D-1&2, d ISOS-O-1 6 Lloyd, M. NREL (USA) N, W T1A, T1B, T2 ISOS-D-1&3, d ISOS-L 7 Hauch, J. Konarka (Germany) 49 o 26 0 N, 11 o 4 0 E T1A, T1B, T2 ISOS-D-1&2&3, ISOS-L 8 Hermenau, M. IAPP (Germany) 51 o 1 0 N, E T1A, T1B, T2 ISOS-D-1&2, ISOS-L 9 Schwartz, G. Heliatek GmbH (Germany) 51 o 4 0 N, E T1A, T1B, T2, T3B ISOS-D-1&2&3, d ISOS-L, ISOS-O-1 10 Hoppe, H. TU Ilmenau (Germany) 50 o 41 0 N, E T1A, T2, T3A, T3B ISOS-D-1, ISOS-L, ISOS-O-1&2 11 Elschner, A. H.C. Starck (Germany) 51 o 4 0 N, E T1A, T2, T3A, T3B ISOS-D-1, ISOS-L, ISOS-O-1&2 12 Swonke, T and Brabec, C. ZAE Bayern (Germany) 49 o 32 0 N, 1 1 o 1 0 E T1A, T1B, T3B ISOS-D-1&2, d ISOS-O-1 13 Zimmermann, B. Fraunhofer ISE (Germany) 48 o 0 0 N, 7 o 50 0 E T1A, T2, T3B ISOS-D-1, ISOS-L, ISOS-O-1 14 Veenstra, S.C. ECN Solar Energy (Holland) 52 o 46 0 N, E T1A, T1B, T2, T3B ISOS-D-1&2, d ISOS-L, ISOS-O-1 15 Galagan, Y.O. Holst Center (Holland) 51 o 24 0 N, E T1A, T3B ISOS-D-1, ISOS-O-1 16 Voroshazi, E. IMEC (Belgium) 50 o 51 0 N, E T1A, T1B ISOS-D-1&2 d 17 Lira-Cantu, M. CIN2, CSIC, (Spain) 41 o 30 0 N, E T1A, T3B ISOS-D-1, ISOS-O-1 18 Rath, T. TU Graz, (Austria) 47 o 4 0 N, E T1A, T1B, T2, T3B ISOS-D-1&2, ISOS-L, ISOS-O-1 19 Kim, K. KIST (Korea) 37 o 36 0 N, E T1A, T2, T3B ISOS-D-1, ISOS-L, ISOS-O-1 20 Katz. E and Manor, A. BGU (Israel) 37 o 51 0 N, E T3A ISOS-O-2 21 Watkins, S. CSIRO (Newcastle, Australia) S, E T1A, T1B, T2, T3A, T3B ISOS-D-1&2, d ISOS-L, ISOS-O-1&2 21 Watkins, S. CSIRO(Melbourne, Australia) S, E T1A, T2, T3A, T3B ISOS-D-1, ISOS-L, ISOS-O-1&2 22 Krebs, F.C. and Gevorgyan, S.A. Risø DTU (Denmark) 55 o 41 0 N, E T1A, T1B, T2, T3B ISOS-D-1&2, d ISOS-L, ISOS-O-1 a Test IDs used during the experiments for referring to the customized procedures recommended to the groups. The procedures can be found in the supporting information different. b The corresponding recommended ISOS-3 procedures (see [17]). The tests however are customized in the studies and somewhat deviate from the recommended ISOS procedures. c ASTM standard tests: EMMAQUA &, desert weathering, subtropical weathering. The performance of the modules was only measured at Risø DTU before and after storage at ATLAS. d The used temperatures/relative humidity levels in D-2&3 tests (see Table 10) deviate from ISOS protocols [17]. inside is 3/4 of the performance observed outside in natural sunlight, as shown in Fig. 3. The ratio was independent of the sequence of indoor and outdoor measurements. The exposure times in both locations were kept in the same range. The difference was attributed to a high sensitivity of the ZnO photoconductivity to the UV-content. The outdoor spectrum is known to be richer in the high energy UV-region ( nm) than that of the used KHS 575 solar simulator. This observation was ascribed to a combination of the higher UV-content of natural sunlight and the lower device temperature, when testing outside. It is worth noting that there is some spread around the fitted line plotted in Fig. 3; this is attributed to inaccuracies in the estimation of the solar intensity outdoors due to variation in solar flux Calibrated testing across labs 19 groups reported results of measurements on calibrated devices. Table 4 shows the list of the groups together with the date of measurement and the measured PV parameters. Various light sources were used for measurements and reported by the recipient laboratories, including natural sunlight. Fig. 4 presents

6 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) photo-doping procedure prior to measurement under any of the two conditions Comparison to the previous ILS Table 6 compares values of standard deviations of these measurements with data from the previous ILS study [1]. There is a significant improvement in reproducibility of the performance of different modules in this set of samples compared to samples prepared in the previous ILS study. A number of issues concluded from the previous ILS study has been addressed during the production of these modules, which obviously improved the reproducibility. In particular: Fig. 3. Correlations between outdoor and indoor (red triangles) characterizations at Risø DTU. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article) the PV parameters plotted as Risø DTU indoor/outdoor versus the recipient laboratory. Table 5 shows values of the PV parameters averaged over Risø DTU indoor/outdoor measurements and across the recipient laboratories. Standard deviations (SD) calculated as a percentage of the average values are shown in parentheses. While the SDs of data measured at Risø DTU were essentially within 10%, the spread across the recipient laboratories was reaching 14% for I sc and FF, and 23% for PCE. If the obvious outliers observed in plots are excluded, then the SDs of recipient laboratories decrease for I sc to 11%, V oc to 4%, FF to 11% and PCE to 15%. The average of measurements at recipient laboratories is rather close to Risø DTU-indoor results and inferior to Risø DTU-outdoor. Two groups that used natural sunlight for testing reported results closer to Risø DTU-outdoor (average of two outdoor measurements: I sc ¼10.9 ma; V oc ¼8.6 V; FF¼54%; PCE¼1.45%). It rules out the factor of degradation of modules during the transportation to the recipient laboratories. When comparing parameters measured indoor vs. outdoor, it is clear that the V oc is much higher in the case of outdoor studies, which can be explained by the fact that in many cases the measurements under simulated light sources were performed at higher device temperatures, due to the light soaking prior to the measurement. However, the temperature difference cannot explain the increase of the other PV parameters. Although the scarcity of the data does not allow for making any firm conclusions, a number of reasons can be suggested, which can explain the observed pattern. In many cases, the solar simulators that provide a class A spectrum in the visible range can have lower irradiation levels in the UV range compared to natural sunlight. Due to sensitivity of the ZnO to UV, it can lead to different rates of photo-doping under simulated and real sun irradiations. As a consequence, in most cases, the indoor performance of such modules can be inferior to outdoors. Another factor that can possibly contribute to the difference is that due to different procedures of indoor and outdoor measurements, there is high probability that modules will be exposed to light for significantly longer periods of time than under indoor conditions prior to the measurement resulting in different rates of photo-doping. Since the inflection point effect is not completely understood at present, it is hard to suggest optimal solutions for avoiding this problem. A straight forward solution could be measuring the devices under both indoor and outdoor conditions or assuring rigorous A more compact geometry was chosen in this case, which eased the uniform illumination of the samples in the different laboratories. Male and female buttons were introduced for contacting electrodes. It allows easy and repeatable contacting of the electrode without damaging the module. A laser-cut mask was used for masking modules, which improved the accuracy of measurements. Detailed instructions of the measurement procedure and the dynamic behavior of the sample performance were provided in the protocol together with cells. In Table 6, a twofold decrease of deviations is observed for measurements at Risø DTU compared to previous data and in the case of recipient measurements only the FF is exceeding the previous results. The larger deviations of the FF are explained by a more pronounced inflection point effect in the modules of these studies, due to non-complete encapsulation (no edge sealing). As discussed earlier [14], the incomplete sealing makes the effect of ZnO doping and de-doping reproducible making the effect of the inflection point more pronounced in the new modules. Similarly, the correlation between Risø DTU and recipient laboratory measurements is poorer in the new studies (comparison not shown here), which is again ascribed to the reason above. The spread of reported module performance across laboratories was significantly reduced compared to previous ILS studies and reaching 23% with and 15% without outliers for PCE. The spread was mostly ascribed to the inflection point effect, which is pronounced through the sensitivity of modules towards the irradiation spectrum and the time of exposure of modules to light prior to the measurement. The same reasons caused a significantly higher performance of modules measured under natural sunlight compared to simulated light measurements Shelf life studies (ISOS-D-1) The purpose of ISOS-D-1 was to establish if it is possible to compare shelf life of modules across groups with reasonable data deviations. In such studies, the samples are usually stored in a dark ambient room environment, such as, for example, a drawer and the variability of the environmental conditions across different laboratories are minimal compared to other stability studies. Thus, the ISOS-D-1 test can give a good insight into how much the diversity of measurement procedures and equipment can contribute in the spread of stability data Variations across laboratories 19 groups shown in Table 7 reported on shelf life studies for a total of 32 ST type modules. The received data were mostly text or excel files with raw IV-curve data and sheets of time vs. photovoltaic parameters extracted from IV-curves. Measurement periodicity varied across groups. The reported PV parameters were

7 1404 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Table 4 The list of laboratories, the date of measurement and the measured photovoltaic parameters for calibrated modules. Package number Laboratory Meas. date Light source Light intensity (W m 2 ) Sample temperature Risø DTU indoor (I sc /V oc /FF/PCE) Risø DTUoutdoor Recipient 1 Riede, M. 18/06 Outdoor (Sun) /8.1/50/ /8.6/59/ /8.7/55/1.5 (Stanford) 2 Katz, E. (BGU) 08/06 Outdoor (Sun) /8.4/51/ /8.7/53/ /8.6/53/1.4 4 Voroshazi, E. 25/05 Abet 2000 AAA /8.2/49/ /8.7/52/ /8.1/58/1.5 (IMEC) Xe Arc Lamp 5 Galagan, Y.O. AM /8.5/49/ /8.8/52/ /6.9/41/0.6 (Holst Center) simulator 6 Shrotriya, V. 01/06 Oriel SS model /7.6/53/1 11.5/8.7/54/ /8.4/54/1.2 (Solarmer) Xe arc lamp 7 Lira-Cantu, M. 29/05 KHS SC /8.1/44/1 11.2/8.3/47/ /8.1/33/0.7 (CIN2, CSIC) MH lamp 9 Hermenau, M. 28/ /7.6/44/ /8.2/48/ /7.8/50/1.6 (IAPP) 10 Hoppe, H. (TU 28/05 Class B solar 9.8/8.3/50/ /8.7/49/1.4 8/8.3/54/1 Ilmenau) simulator 11 Lloyd, M. 11/06 AM /8.7/47/ /9.2/52/ /8.7/46/1.1 (NREL) simulator 12 Xiao, M. 10/06 Newport-Oriel /6.8/43/0.9 12/7.9/48/ /7.9/53/1.3 (Plextronics) Xe arc lamp 14 Rath, T. (TU 31/05 KHS SC /8.6/53/ /9.2/55/1.6 10/8.8/52/1.3 Graz) MH lamp 16 Watkins, S. (CSIRO, Melbourne) 28/05 Newport-Oriel Xe arc lamp (AM1.5 filters) /8.3/53/ /8.7/55/ /8.6/42/ Watkins, S. (CSIRO, Newcastle) 02/06 Newport-Oriel Xe arc lamp (AM1.5 filters) /8.3/53/ /8.7/55/ /8.5/46/ Kim, K. (KIST) 30/ /8.1/48/1 11.5/8.5/48/ /8.8/40/ Swonke, T Newport /8.5/50/ /8.9/49/ /8.6/53/1.6 (ZAE) 94061A Sol1A (ABA) Xe arc lamp 21 Schwartz, G. Steuernagel /8.2/48/ /8.6/52/ /4.1/32/0.4 a (Heliatek) SC1200 MH lamp 22 Elschner, A. Atlas solar cell /8.2/49/ /8.6/54/ /8.2/49/1.2 (H.C. Starck) test 575 MH lamp 23 Veenstra, S.C. 03/06 Wacom /7.8/43/ /8.7/46/1.3 9,7/8.2/43/1.0 b (ECN) WXS-300S Hauch, J /8.2/51/ /8.6/54/ /8.8/50/1.3 (Konarka) All/indoor Risø DTU 14/05 KHS SC (75) MH lamp All/outdoor Risø DTU 17/05 Real sun 950 (750) a Contact failure was observed. b Spectral mismatch corrections were performed. low at the first measurement (at t 0 ) and were gradually increasing after each measurement, due to photo-doping induced by the light source during measurements and mostly reaching the peak values (t max ) after 2 or 3 measurements (1 3 days). Since however, the dynamic nature of the inflection point was largely dependent on the duration of the light exposure of modules during measurements, it resulted in a spread of t max. Deviations of t max from an average were significantly affected also by the frequency of the measurements at the beginning of the test and were reaching up to 740 h. Across 1000 h of stability measurements however such deviations were insignificant. The diversity of the inflection point behavior affected the accurate determination of quantities, such as burn in range, initial stabilized efficiency PCE s 0 and T80 (time when device degrades to 80% of optimum performance), as suggested in the protocol of ISOS-2 [11]. As a consequence, the data were chosen to be processed and presented in the following manner. The PCE parameter is discussed here, but the same approach was applied to all photovoltaic parameters (I sc, V oc, FF and PCE). Decay curves of PCE versus time were normalized to peak values PCE max. The normalized values were manually extracted from each curve at five different time points (t 0, t 200, t 400, t 700 and t 1000 ) and summarized in the general plot. Average values and SDs were calculated for each time point. Fig. 5 presents the calculated values together with the average values intersected by a dashed line for guiding the eye and the error bars representing the standard deviations. The average of t max across laboratories was calculated and set as an additional time point for the average curve. Catastrophic failures [11] were recorded for two modules (not shown here), which are attributed to the handling of modules

8 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Indoor Indoor Recipient Laboratories Outdoor Recipient Laboratories Outdoor Risø Indoor/Outdoor Risø Indoor/Outdoor Indoor Indoor Recipient Laboratories Outdoor Recipient Laboratories Outdoor Risø Indoor/Outdoor Risø Indoor/Outdoor Fig. 4. The spread of I sc, V oc, FF and PCE for calibrated modules. Red rectangles correspond to outdoor and blue rhombs to indoor measurements performed at Risø DTU. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Table 5 List of average values with SDs as a percent of average values in parentheses. Parameter Risø outdoor Risø indoor Recipient laboratory I sc (ma) 11.6 (72.4%) 9.7 (74.6%) 10.6 (714%) a V oc (V) 8.6 (73.3%) 8.1 (75.3%) 8.1 (76%) a FF (%) 51.6 (76.7%) 48 (76.8%) 48 (714%) a PCE (%) 1.45 (78.3%) 1.1 (710%) 1.17 (723%) a a Without outliers the standard deviations would be I sc 11%, V oc 4%, FF 11% and PCE 15%. Table 6 Comparison of SD values as a percent of average values for this and previous ILS. Parameter Risø DTU (previous ILS) Risø DTU (new ILS) Recipient (previous ILS) Recipient (new ILS) I sc 8.6% 4.6% 28% 14% (11%) a V oc 8.9% 5.3% 9.2% 6% (4%) a FF 10.1% 6.8% 9.7% 14% (11%) a PCE 22% 10% 26.8% 23% (15%) a a The values in the parentheses are the standard deviations for recipient (New RR) without the outliers. during periodic measurements. The average T80 for PCE is shown as well, which is about 615 h if calculated from t 0. For the most stable modules, T80 reached more than 1000 h. Table 8 shows the average values of each parameter during the decay together with the SDs. The SDs are increasing along the degradation of devices and reaching nearly 10% for I sc, V oc and FF and reaching 13% for PCE at t One source of data-spread was the inflection point effect. As was discussed in the earlier published article [14] and shown by a number of groups in these studies, the photo-doping and de-doping of ZnO is a reproducible effect. Fig. 6 shows an example of dynamic behavior of a module stored in dark and periodically measured before and after photo-annealing for a few minutes. If the devices were not photo-annealed long enough to remove the inflection point at each measurement, it could lead to deficient photo-annealing of ZnO, and thus contribute to degradation of the module performance. Additionally, some groups reported de-lamination or cracking in devices induced by the periodic handling/mechanical stressing of modules due to measurements. This can result in catastrophic failure of the performance, as was reported for two modules. In addition, de-lamination can accelerate the gradual decay of PV parameters and therefore, contribute in the spread of stability data Intra-laboratory vs. inter-laboratory 9 laboratories measured 2 or 3 modules under the same conditions, which allowed for comparison of inherent variations among the modules versus variations across laboratories. The ratios of the standard deviations (averaged over 1000 h) for intra- and inter-laboratory measurements are shown in Table 9 calculated

9 1406 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) for nine laboratories. The results show that a large portion of data-spread across the laboratories could be easily contributed by the intrinsic deviations within modules. The standard deviations of photovoltaic parameters measured on similar modules across 18 laboratories were reaching 13% after 1000 h of shelf life. A large part of the spread was due to the inherent deviations within the modules. Two factors possibly Table 7 The list of laboratories for ISOS-D-1 studies. The starting date of the experiment, temperature and RH ranges are listed as well. Package number Laboratory Starting date Sample quantity Storage temperature Storage R.H. (%) 4 Voroshazi, E. (IMEC) 01/ Galagan, Y.O. (Holst 2 25 Center) 6 Shrotriya, V. 01/ (Solarmer) 7 Lira-Cantu, M. 29/ (CIN2, CSIC) 29/05 9 Hermenau, M. 31/05 6 (IAPP) 10 Hoppe, H. (TU 28/ Ilmenau) 11 Lloyd, M. (NREL) 11/ Xiao, M. 28/05 2 (Plextronics) 14 Rath, T. (TU Graz) 31/ Watkins, S. (CSIRO, 01/06 3 Melbourne) 16 Watkins, S. (CSIRO, Newcastle) 01/ Zimmerman, B. (ISE) 31/ Kim, K. (KIST) 30/ Swonke, T. (ZAE) 31/ Schwartz, G (Heliatek) 22 Elschner, A. (H.C. 28/ Starck) 23 Veenstra, S.C. (ECN) 03/ (72) Hauch, J. (Konarka) 3 26 Risø DTU 31/ r45 influencing the spread of data were the periodic handling of the modules and the inflection point effect Effect of storage temperature and relative Humidity (ISOS-D-2&3) It is well established that the environmental factors such as temperature and level of relative humidity (RH) have a major influence on device degradation kinetics [18 20]. Thus, it is important to choose a number of temperature/rh combinations, which can be representative for standard testing of OPV modules and can be accepted and steadily used among groups in the OPV community. The purpose of ISOS-D-2&3 was to establish capabilities of various groups in performing temperature/rh controlled tests and perhaps to get an insight into which conditions are the most optimal in terms of reproducibility among groups. 13 groups carried out ISOS-D-2&3 type experiments with a total of 38 ST type modules. Table 10 shows the list of groups that performed the ISOS-D-2&3 test together with values of temperature and RH varying in the ranges C and 0 85% correspondingly. The list shows the diversity of the conditions used by groups. In some cases RH was not recorded/reported, which put some constrains on the comparison of data. The way the data was chosen to be processed and presented here was as follows. Degradation rates of PV parameters were calculated using the linear fitting of curves at the stabilized regions. Two graphs were plotted with degradation rate versus temperature and degradation rate versus relative humidity. Fig. 7 shows the corresponding plots. The degradation pattern of I sc is similar to PCE, while stabilities of V oc and FF are less affected by temperature and humidity. Due to the scarcity of data, it is hard to draw conclusions on the behavior of the modules under different temperature/rh levels. One observation is that temperatures above 65 1C and RH values above 60% are detrimental for the module stability. Fig. 8 shows the 3-dimensional representation of the decay rates for different temperature/rh combinations. The plot shows the diversity of the conditions used for stability testing. Such diversity together with the sensitivity of devices towards temperature and RH obviously Measured Data Average T80 ~ 615 hrs Fig. 5. PV parameters versus time for 29 ST modules (orange rectangles) measured during ISOS-D-1 test. The parameters are normalized to peak values. The dashed line intersects the average values for guiding the eye and the error bars represent the SDs. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

10 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Table 8 The average of PV parameters at different times and the corresponding SDs for ISOS-D-1. Time (h) I sc (normalized) V oc (normalized) FF (normalized) PCE (normalized) Average (%) SD (%) Average (%) SD (%) Average (%) SD (%) Average (%) SD (%) Fig. 6. PCE versus time for a module stored in dark and periodically measured before and after photo-annealing for few minutes. Table 9 Intra-laboratory and inter-laboratory SD values averaged over 1000 h and the ratio inter/intra in percent calculated for nine laboratories. SD I sc V oc FF PCE Inter-laboratory Intra-laboratory Ratio 47% 58% 66% 57% makes the comparison of the stability data across different laboratories impractical and therefore, suggests more rigorous requirements on testing procedures. We suggest that only a small number of temperature/rh combinations should be used among all groups to assure the possibility of data comparison. It was shown that these modules are sensitive towards temperature and RH especially at elevated values. Therefore, careful control and reporting of temperature and RH levels during stability testing of such modules is important. Furthermore, the groups performing ISOS-D-2&3 test used a rather large diversity of temperature/rh combinations, which made it impractical to compare the module stability performance across laboratories. Thus, certain standardized sets of temperature/rh combinations have to be defined in stability measurement protocols that can be agreed upon and followed by all groups in the OPV community Light soaking test (ISOS-L) The purpose of ISOS-L test was to establish the capability of different groups of carrying out laboratory light soaking tests and to check the reproducibility of results across different laboratories. Table 11 shows the list of 14 laboratories that carried out the ISOS-L test. The table also includes the light sources used during tests. Due to the scarcity of data and large deviations among different reports, the approach used for processing ISOS-D-1 data would not be very informative in this case. The standard deviations for most of the parameters at all time points were above 20%. Therefore, original data is discussed here instead. Fig. 9 shows all curves for 18 ST type modules (data for three modules were unreadable) with colors defining types of the light sources used for light soaking. Different types of lamps were used, categorized as incandescent lamps (halogen lamps¼blue rhombs in Fig. 9), high-intensity discharge lamps (metal halide lamps, Xe arc lamps¼green triangles in Fig. 9) and sulfur plasma lamps (red circles in Fig. 9). For comparison, the black solid line shows the average of modules tested in the real sun under outdoor conditions for 1000 h. The time for the outdoor curve is adjusted so that at a given time point the modules have received approximately the same amount of energy dosage as in the indoor studies. Obviously there are large variations in the stability reports of the modules even when only one lamp type is considered. One pattern that can be seen is that modules tested under sulfur plasma lamps show significantly better stability (i.e. sulfur plasma lamps overestimate OPV stability). One possible explanation can be that the UV and IR contents in plasma light sources are much lower when compared to metal halide lamps. Fig. 10 shows an example of spectral distribution of a sulfur plasma lamp, Xe arc lamp with daylight filter (WACOM) and AM1.5G global radiation provided by ECN. The figure clearly demonstrates the large difference of the spectra especially in the UV-region. Although samples need some initial UV photo-doping to recover from the inflection point, the long term UV irradiation can be detrimental for the stability. It points to the fact that the spectrum of the light source can play an important role during stability studies. Another source of stability variations was the diversity of temperatures of the modules during the light exposure (see Table 11). In addition, of the four groups that carried out the ISOS-L test on two specimens, two groups measured modules in situ during the light soaking test, while two others periodically removed modules from the testing setup and measured them under a different calibrated light source. It allowed for comparing the two different procedures and determining the effect of the periodic handling on the device stability performance. The calculations showed that deviations within modules in the case of periodic handling were nearly 3 times higher than for the ones measured in situ. The following conclusions can be drawn from results of ISOS-L experiments. The spectral distribution of the irradiation, which is defined by the type of lamps used (or filters used), significantly affects the degradation rate of modules and consequently the comparability of the stability reports across the laboratories. An additional factor contributing to the spread of data is the diversity of the temperature of the modules during the light exposure. Finally, the comparison of the tests, where the modules were measured in situ or periodically taken out of the testing setup for characterization, showed that variations are increased threefold by periodic handling of modules. Finally, it is also evident from Fig. 9 that the data observed when using metal halide and Xe-lamps for the ageing represents the ageing observed under real sunlight quite well.

11 1408 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Table 10 The list of laboratories for ISOS-D-2&3 experiments. The starting date of experiment, storage setup and the ranges of temperature and RH are listed as well. Package number Laboratory Starting date Sample quantity Storage setup Storage temperature Storage R.H. (%) 4 Voroshazi, E. (IMEC) 01/06 1 Glove-box Oven Drawer (desiccant) 25 o5 4 Env. chamber Shrotriya, V. (Solarmer) 01/06 2 Tenney T6S temperature and humidity test chamber 9 Hermenau, M. (IAPP) 31/05 1 Oven 65 1 Oven Lloyd, M. (NREL) 11/06 1 Env. chamber Env. chamber Xiao, M. (Plextronics) 28/05 2 Thermotron Rath, T. (TU Graz) 31/05 3 Oven Watkins, S. (CSIRO, 01/06 1 Oven Newcastle) 19 Swonke, T. (ZAE) 31/05 2 Oven Schwartz, G. (Heliatek) 1 50 o o Veenstra, S.C. (ECN) 03/06 5 Oven 46.6 (70.1) 8 (72) (70.1) 3 (71) (70.2) 0.3 (70.2) 25 Hauch, J. (Konarka) 2 Weather chamber Oven Risø DTU 31/05 1 Oven Degradation Rate Degradation Rate Temperature ( C) Relative Humidity (%) Fig. 7. Degradation rate of the modules versus the storage temperature (left) and the storage RH level (right). The legends show the corresponding temperature and RH values. Degradation Rate RH (%) 3.6. Outdoor testing (ISOS-O-1&2) Temp. ( C) Fig. 8. Degradation rate of the modules versus different combinations of temperature/ RH values. Two types of outdoor experiments were explored. In one the modules were being stored outdoors at all times and measured in the same position under real sun (ISOS-O-2) and in the other one the cells were being stored outdoors and taken inside periodically (daily or weekly) and measured under a calibrated sun simulator (ISOS-O-1). In one case, an outdoor solar tracker was used to characterize modules (ISOS-O-1 at Risø DTU). There were two reasons to split the outdoor test into ISOS-O-1&2: firstly, in order to meet capabilities of a wider range of laboratories and secondly, it would allow for a comparison of the two different approaches and study the influence of the periodic handling. Five laboratories performed ISOS-O-2 type experiments and 16 chose to follow the ISOS-O-1 approach. Tables 12 and 13 show lists of the laboratories participating in ISOS-O-2 and ISOS-O-1. The tables include the types of the outdoor platforms used for studies. A number of photographs is presented in Fig. 11, which show the variability of the setups and conditions that the modules were exposed to. The energy dosage received by the modules varied significantly (735% after 1000 h of exposure) across the reports of different groups, due to the different weather conditions in different locations Results from ISOS-O Outdoor fluctuations In the ISOS-O-2 experiment, the devices were left outside during the entire study (day and night) and periodically measured

12 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Table 11 The list of laboratories for ISOS-L experiment. The starting date of experiment, the types of light sources and intensities and the ranges of temperature and RH are shown as well. Package number Laboratory Starting date Sample quantity Light source Intensity (W/m 2 ) Storage temperature (1C) Storage R.H. (%) 9 Hermenau, M. (IAPP) 31/05 1 Halogen lamp (without UV) 10 Hoppe, H. (TU Ilmenau) 28/05 1 Metal halide lamp a 11 Lloyd, M. (NREL) 11/06 2 UV soak ( nm) b 12 Xiao, M. (Plextronics) 28/05 5 Q-Sun Xe arc lamp 14 Rath, T. (TU Graz) 31/05 5 LG PSH0731WA Sulfur plasma lamp 16 Watkins, S. (CSIRO, Melbourne) 01/06 1 Newport-Oriel Xe arc lamp 16 Watkins, S. (CSIRO, Newcastle) 01/06 1 Metal halide lamp Zimmerman, B. (ISE) 31/05 5 Sulfur plasma lamp Kim, K. (KIST) 30/05 4 Xe arc lamp Schwartz, G. (Heliatek) 2 Sulfur plasma lamp o10 b 22 Elschner, A. (H.C. Starck) 28/05 5 Atlas XLS Xe arc lamp 23 Veenstra, S.C. (ECN) 03/06 5 Sulfur plasma lamp (72.5) 14 (73) 25 Hauch, J. (Konarka) 2 Atlas XLS 1000 Xe arc lamp 26 Risø DTU 31/05 4 KHS Metal halide lamp a 2 3 Measurements were performed during 1000 h of testing. b Illumination of samples was not continuous, but periodic. Fig. 9. Decay of PV parameters for 14 modules measured in different labs. Red circles correspond to sulfur plasma lamps, green triangles to metal halide and xenon lamps and blue rhombs to halogen lamps. The real sun data is shown as a black line. The measured points are connected by dashed lines for guiding the eye. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.) under the real sun. The groups mainly used stationary platforms at certain angles mostly adjusted to the latitude of the location. The results were normalized with respect to the energy dosage received by the modules, so that at a given time point an equal radiant exposure was received. The reason that the time axis was chosen instead of dosage for presenting the data was to allow easy comparison with other experiments. Furthermore, the normalization was performed in such a way that modules would receive nearly 1010 MJ/m 2 after 1000 h of exposure and axes can be easily interpreted as energy dosage as well (in MJ/m 2 units). Normally, the irradiation intensity was recorded using a pyranometer placed next to the cells. The data were either reported with I sc values normalized to the light intensity or raw data was provided together with the irradiation intensity values. There were large fluctuations in the normalized values of I sc for a number of reasons: Time delays between recorded irradiation intensity and the IV-curves. Shadowing of the modules not recorded by a pyranometer.

13 1410 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Strong winds causing flipping of modules or loosening of contacts. Periodic cloud cover during IV measurements of the modules. However, given the high frequency of measurements, the fluctuations could be filtered out. Bird droppings and dust were additional factors contributing to performance variations AM1.5G WACOM Sulphur plasma Wavelength (nm) Fig. 10. Spectral distribution of a sulfur plasma lamp (blue), Xe arc lamp (WACOM) (black) and AM1.5 global irradiation. The plot is provided by ECN. Table 12 The list of laboratories for the ISOS-O-2 experiment. The starting date of the experiment, the outdoor platform and the angle of exposure are listed as well. Package number Laboratory Starting date Sample quantity Outdoor platform 1 Riede, M. 18/06 8 Stand Still 01 (Stanford) 10 Hoppe, H. (TU 28/05 2 Stand still 451 Ilmenau) facing south 16 Watkins, S. (CSIRO, 02/06 1 Stand still 371 Melbourne) facing north 16 Watkins, S. (CSIRO, 01/06 4 Stand still 321 Newcastle) facing north 22 Elschner, A. (H.C. Starck) 07/06 2 Stand still facing south 101 Exposure angle Results of measurements The data were processed and presented using the energy dosage instead of the time. Fig. 12 shows the results of ISOS-O-2 measurements on 11 ST type modules by a total of five groups. Table 14 shows the average values and SDs calculated for four different energy dosage levels. Of about 1000 MJ/m 2 corresponded to nearly h of outdoor exposure depending on weather conditions and the angle of module exposure. Deviations of the PV parameters across the testing are within 11%, which is rather low, but not surprising considering that only five groups participated. Such low values can also be explained by the fact that modules were measured in situ and thus, the handling was minimal ISOS-O-1 In the ISOS-O-1 experiment the cells were stored outside, but the measurements were carried out inside under calibrated light sources (one group used solar tracker outside). It excluded the factor of fluctuations and miscalculations of solar irradiation intensity, yet added the factor of periodic handling (periodic dismounting from and remounting on the outdoor platform) of modules. As a result, there were reports of catastrophic failure for four modules, which have not been included in the calculations. For five modules, the energy dosage was not available and thus the average dosage values were used for estimation of the degradation rates. A source of data-spread was also the inflection point effect, since the measurement was carried out daily or weekly and thus, the peak performance of modules, which in most cases was reached within the first few hours of exposure to sunlight, could be easily missed. It could lead to normalization of PV parameters to inaccurate peak values and result in additional spread of data. 16 laboratories carried out ISOS-O-1 type experiments on 27 ST type modules. Fig. 13 shows the measured stability data of ISOS-O-1 versus the energy dosage. Table 15 shows the average and SD values for ISOS-O-1. The deviations are in some cases exceeding 20%, which is caused by the reasons described above ISOS-O-1 versus ISOS-O-2 Similar to the ISOS-L test, inherent deviations within modules measured in one laboratory were calculated for both ISOS-O-2 Table 13 The list of the laboratories for the ISOS-O-1 experiment. The starting date of experiment, the outdoor platform and the angle of exposure are listed as well. Package number Laboratory Starting date Sample quantity Outdoor platform Exposure angle 2 Katz, E. (BGU) 07/06 8 Stand still facing south Galagan, Y.O. (Holst Center) 1 6 Shrotriya, V. (Solarmer) 01/06 2 Stand still facing south Lira-Cantu, M. (CIN2, CSIC) 29/05 3 Stand still facing south a 10 Hoppe, H. (TU Ilmenau) 28/05 4 Stand still facing south 451 b 13, 24 Haillant, O. (ATLAS) Rath, T. (TU Graz) 31/05 3 Stand still Watkins, S. (CSIRO, 01/06 1 Stand still facing north 371 Melbourne) 16 Watkins, S. (CSIRO, Newcastle) 01/06 1 Stand still facing north Zimmerman, B. (ISE) 31/ Kim, K. (KIST) 30/ Swonke, T (ZAE) 31/05 2 Stand still Schwartz, G. (Heliatek) 1 22 Elschner, A. (H.C. Starck) 01/06 2 Stand still facing south Veenstra, S.C. (ECN) 03/06 2 Stand still facing south Risø DTU 31/05 1 Solar tracker a As recommended by re.jrc.ec.europa.eu/pvgis/apps/radmonth.php. b Details of studies at ATLAS are described in Section 3.10.

14 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Fig. 11. (1) HC Starck, rooftop platform: modules facing south at 101 (Germany); (2) Risø DTU, module mounted on the solar tracker (Denmark); (3) stand still platform: modules facing south at 301 (BGU, Israel); (4) horizontally oriented table platform (Stanford, USA); (5) CSIRO, rooftop platform: modules facing north at 321 (Newcastle, Australia); (6) TU Graz, rooftop platform (Austria); (7) CSIRO, rooftop platform: modules facing north at 371 (Melbourne, Australia); (8) stand still plate: modules facing south at 101 (Barcelona, Spain); and (9) rooftop testing station: modules facing south at 281 (Solarmer, USA). and ISOS-O-1. The results showed nearly twofold increase of deviations of I sc and PCE in the case of ISOS-O-1, which is again ascribed to the factor of the periodic handling. ISOS-O-2 proves to be a more accurate testing method, due to the minimal handling involved and due to the high frequency of the measurements. On the other side, the number of groups performing ISOS-O-1 was three times larger despite the fact that measurements in ISOS-O-1 require more workload due to periodic mounting and dismounting of specimens. This is probably caused by the lack of necessary facilities/equipment for performing the ISOS-O-2 test Standard outdoor testing at ATLAS ATLAS Material Testing Technology LLC carried out three types of outdoor testing according to ASTM international standards, such as Emmaqua, Desert Weathering and Sub-tropical Weathering. The table in Fig. 14 shows the types and the details of each experiment and the photos illustrate the outdoor platforms used for testing. However, due to the relatively low amperage of the small modules studied here, it was not possible to measure the IVcurves at the recipient ATLAS laboratories. Thus, after different times of exposure the modules were shipped to Risø DTU and measured under calibrated light simulators to evaluate the decay level of each module at a certain time period. All the measured values were normalized to measurements performed at Risø DTU prior to shipping to an ATLAS. Fig. 15 shows the decay of PCE after a certain time of exposure. One has to keep in mind that each point in the plot corresponds to different modules and the plot should not be regarded as presentation of decay curves. However, the results give a good insight into how the modules behave at different conditions. The time axes only consider the exposure time and not the shipping time and storage in the dark.

15 1412 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Fig. 12. PV parameters versus the energy dose for 11 modules (orange rectangles) measured during ISOS-O-2 test. The parameters are normalized to peak values. The dashed line intersects the average values for guiding the eye and the error bars represent the SDs. Table 14 The average of PV parameters at different energy dosage levels and the corresponding SDs for ISOS-O-2. Energy dose (MJ/m 2 ) I sc (normalized) V oc (normalized) FF (normalized) PCE (normalized) Average (%) SD (%) Average (%) SD (%) Average (%) SD (%) Average (%) SD (%) Fig. 13. Variation of PV parameters versus energy dose for ISOS-O-1 test (orange rectangles). The parameters are normalized to peak values. The dashed line intersects the average values for guiding the eye and the error bars represent the SDs. Clearly Emmaqua is degrading samples very fast compared to the two other tests, while no difference can be seen between sub-tropical and desert experiments. Emmaqua involves periodic spraying of devices with water and includes concentrated UV irradiance ( 5) and a relatively higher temperature due to the concentration of the visible and IR ranges. Since the data from

16 S.A. Gevorgyan et al. / Solar Energy Materials & Solar Cells 95 (2011) Miami (hot humid weather) shows no significantly faster degradation compared to the data from Phoenix (hot arid weather). The influence of water would thus seem to be limited in comparison to the effect of sunlight concentration during the Emmaqua testing. A more detailed description of the experimental setups and procedures can be found in the literature [21]. These studies are good demonstrations of the possibility of storing devices in one place and characterizing in another Comparison of different tests Table 16 summarizes results of the measurements and outlines the factors affecting the spread of data for all five categories of tests Maximum power point (MPP) versus open circuit The group from TU Graz tested the degradation of two identical devices under light soaking (ISOS-L), while one of them was kept at open circuit and the other at the maximum power point (MPP), using a Keithley 2400 SMU. Fig. 16 shows the decay of the photovoltaic parameters for both modules. Although, there is a significant difference in the dynamics of the decay, overall the degradation of both devices is in the same range. However, since, only two modules were tested, it is hard to make any firm conclusions. Keeping the device at an MPP during storage (active load) requires an expensive setup, which can be a hurdle for some groups. Thus, in order to compare stability among many groups, perhaps open circuit or a passive load (a resistor, which keeps the cell at an MPP at the t 0 point) should be suggested as a standard procedure, while active load can be used in advanced testing techniques Unreadable data Some of the data were categorized as unreadable, since it was impossible to estimate the degradation rate of the parameters. One of the main reasons was the reproducible effect of the inflection point discussed earlier. The modules require periodic photodoping prior to each measurement and if the photo-doping is not Table 15 The average of PV parameters at different energy dosage levels and the corresponding SDs for ISOS-O-1. Energy dose (MJ/m 2 ) I sc (normalized) V oc (normalized) FF (normalized) PCE (normalized) Average SD Average SD Average SD Average SD Fig. 15. Decay of all pv parameters of the modules after certain time of exposure. Each point corresponds to different module Fig. 14. The table outlining the details of three standard tests carried out by ATLAS and photos illustrating the outdoor platforms for Emmaqua (1), desert weathering (2) and sub-tropical weathering (3).