IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 9, SEPTEMBER 2010 2227 Defect-Induced Breakdown in Multicrystalline Silicon Solar Cells Otwin Breitenstein, Jan Bauer, Jan-Martin Wagner, Nikolai Zakharov, Horst Blumtritt, Andriy Lotnyk, Martin Kasemann, Wolfram Kwapil, and Wilhelm Warta Abstract—We have identified at least five different kinds of local breakdown according to the temperature coefficient (TC) and slope of their characteristics and electroluminescence (EL) under a reverse bias. These are 1) early prebreakdown (negative TC, low slope), 2) edge breakdown (positive TC, low slope, no EL), 3) weak defect-induced breakdown (zero or weakly negative TC, moderate slope, 1550-nm defect luminescence), 4) strong defect- induced breakdown (zero or weakly negative TC, moderate slope, no or weak defect luminescence), and 5) avalanche breakdown at dislocation-induced etch pits (negative TC, high slope). The latter mechanism usually dominates at a high reverse bias. The defects leading to the etch pits are investigated in detail. In addition to the local breakdown sites, there is evidence of an areal reverse current between the dominant breakdown sites showing a positive TC. Defect-induced breakdown shows a zero or weakly negative TC and also leads to weak avalanche multiplication. It has been found recently that it is caused by metal-containing precipitates lying in grain boundaries. Index Terms—Avalanche breakdown, electric breakdown, photovoltaic cells, semiconductor defects. I. I NTRODUCTION L OCAL junction breakdown at low reverse biases has become an important technological problem, particularly in multicrystalline silicon solar cells. Theoretically, cells with a base doping concentration of 1 ∗ 10 16 cm −3 should break down at about −50-V reverse bias by avalanche breakdown, but in reality, breakdown often begins already below −5 V. The appearance of local hot spots at breakdown sites may lead to permanent damage of modules under certain operating conditions of these cells. If one cell of a string is shadowed, Manuscript received January 7, 2010; revised May 26, 2010; accepted June 4, 2010. Date of current version August 20, 2010. This work was supported by the German cluster “SolarFocus” under Project BMU 327650. This paper was presented at the 34th IEEE Photovoltaic Specialists Conference, Philadelphia, PA, June 7–12, 2009. The review of this paper was arranged by Editor S. Ringel. O. Breitenstein, N. Zakharov, and H. Blumtritt are with the Max Planck Institute of Microstructure Physics, 06120 Halle, Germany (e-mail: breiten@ mpi-halle.mpg.de). J. Bauer was with the Max Planck Institute of Microstructure Physics, 06120 Halle, Germany. He is now with CaliSolar Inc., 12489 Berlin, Germany. J.-M. Wagner was with the Max Planck Institute of Microstructure Physics, 06120 Halle, Germany. He is now with Christian Albrechts University of Kiel, 24143 Kiel, Germany. A. Lotnyk was with the Max Planck Institute of Microstructure Physics, 06120 Halle, Germany. He is now with the Faculty of Engineering, Christian Albrechts University of Kiel, 24143 Kiel, Germany. M. Kasemann, W. Kwapil, and W. Warta are with the Fraunhofer Institute of Solar Energy Systems, 79110 Freiburg, Germany. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TED.2010.2053866 the other cells may reverse bias this cell. The mechanisms behind local junction breakdown, particularly their relation to material defects, are not yet well understood. However, this understanding becomes increasingly important since the trend in solar cell industry goes toward enabling the production of silicon solar cells from high-impurity [upgraded metallurgical grade (UMG)] feedstock. There are clear indications that high impurity concentrations lead to breakdown at even lower volt- ages [1], [2]. This paper was extended to the contribution given at the 34th IEEE Photovoltaic Specialists Conference [3] by adding new high-resolution transmission electron microscopy (TEM) results investigating the physical nature of the line defects being responsible for the observed etch pits. One new author was added (N. Zakharov) who has contributed to these new results. II. EXPERIMENTAL The results presented in this paper are based on various phys- ical methods for investigating breakdown phenomena, most of them being imaging techniques. They were applied to a set of standard industrial acidic-etched multicrystalline cells made from closely neighbored wafers from the same brick, which were free of ohmic shunts. The material was cast by the vertical gradient freeze method from standard solar grade feedstock (not a UMG material). In addition to temperature-dependent current–voltage I –V characteristic measurements, mostly dark lock-in thermography (DLIT) under a reverse bias has been used for localizing the breakdown sites. Since DLIT images can quantitatively be scaled in units of a local current density, the results of DLIT images taken under different reverse biases and temperatures allow the creation of images of the local temper- ature coefficient (TC) and the steepness (slope) of the reverse current (TC-DLIT and slope-DLIT [4]). In these techniques, the derivative of the local current density with respect to tem- perature or voltage is approximately obtained as a difference quotient, which is then normalized to the average local current density. Thereby, the resulting TC or slope images show the val- ues of the relative TC or slope at the breakdown sites in units of “% current change per K” (or per V, respectively), independent of the absolute value of the breakdown current. Since for this procedure absolute values of reverse current and reverse voltage are used, a positive TC (slope) means an increase in absolute current strength for increasing temperature (stronger reverse bias). These magnitudes are appropriate to distinguish different types of breakdown from each other. The local avalanche multiplication factor (MF) was imaged by a special variant 0018-9383/$26.00 © 2010 IEEE