Crystallographic properties and elemental migration in two-stage prepared CuIn 1x Al x Se 2 thin films for photovoltaic applications Rémi Aninat a,1 , Guillaume Zoppi a,⇑ , Agnès Tempez b , Patrick Chapon b , Neil S. Beattie a , Robert Miles a , Ian Forbes a a Northumbria Photovoltaics Applications Centre, Northumbria University, Ellison Building, Newcastle upon Tyne NE1 8ST, UK b Horiba Jobin Yvon SAS, Avenue de la Vauve, Passage Jobin Yvon, CS 45002 – 91120 Palaiseau, France article info Article history: Received 23 December 2012 Received in revised form 22 February 2013 Accepted 3 March 2013 Available online 21 March 2013 Keywords: Thin films Semiconductors Vapour deposition X-ray diffraction abstract Two-stage fabrication of CuIn 1x Al x Se 2 thin films for photovoltaic absorbers using sputtered Cu–In–Al metallic precursors has been investigated. Precursors containing different relative amounts of Al were selenised and their structural and chemical properties characterised. X-ray diffraction (XRD) analyses revealed that the Al was only incorporated into the chalcopyrite structure for precursor composition ratios x = [Al]/([Al] + [In]) P 0.38, while chemical analysis of the cross-section indicated partial segrega- tion of Al near the back of the film. Precursor films in the range of compositions that yielded no Al incor- poration were then selenised at four different temperatures. Glow discharge optical emission spectroscopy, plasma profiling time-of-flight mass spectrometry and XRD analyses provided an insight into the diffusion processes and reactions taking place during the selenisation stage. The effect of post- selenisation annealing without additional Se was also investigated, and led to partial incorporation of the Al into the CuInSe 2 lattice but no rediffusion. Ó 2013 Elsevier B.V. All rights reserved. 1. Introduction The most efficient thin-film solar cells to date, with a record effi- ciency of 20.3 ± 0.6% [1], are based on a CuIn 1x Ga x Se 2 (CIGS) absor- ber layer. CuIn 1x Ga x Se 2 , like CuInSe 2 (CIS), is chalcopyrite, with Ga substituting In in the ratio x = [Ga]/([Ga] + [In]). This ratio can be al- tered to tune the band gap between that of CuInSe 2 (CIS), 1.0 eV, and that of CuGaSe 2 (CGS), 1.7 eV [2,3]. In an ideal device, a band gap in- crease yields, in terms of current–voltage (I–V) characteristics, an in- crease of the open-circuit voltage (V OC ) and a parallel decrease of the short-circuit current (I SC ). As a result there exists an optimal band gap for single junction devices, representing the best trade-off be- tween V OC and I SC . According to detailed balance calculations, and because of the specific bands of absorption of the solar spectrum in the atmosphere, two band gaps actually yield almost equal opti- mal efficiencies under air mass 1.5 illumination: 1.15 eV, with 32.8% efficiency, and 1.35 eV, with 33.0% efficiency [4]. In CIGS de- vices, the V OC increase with band gap becomes less pronounced for x > 0.3, due to increasing defect concentrations [5]. This x = 0.3 ratio corresponds, if the Ga distribution is uniform, to a band gap of 1.2 eV [3]. For this reason, the best performing solar devices are made with the nearest optimal (effective) band gap E G  1.15 eV [6,7]. However, being able to reach the higher band gap optimum without V OC degradation would help to reduce resistive losses in so- lar cells, and even more so in modules, where the current has to be transported over greater distances and where cell interconnects are present [2]. An alternative to CIGS is CuIn 1x Al x Se 2 (CIAS), ob- tained by replacing Ga by Al. The CIAS band gap is tunable over a much wider range than CIGS: from 1.0 eV for CIS to 2.7 eV for CuAlSe 2 (CAS). CIAS devices of efficiencies up to 16.9% have been ob- tained by Marsillac et al. by co-evaporation [8]. This latter device performed 0.4% better than an equivalent CIGS device of identical band gap (1.15 eV) built along with it, which suggests reduced losses in CIAS. Several specificities of CIAS might explain this result. Among them, the fact that less Al is required in CIAS to reach a given band gap than the amount of Ga necessary to reach the same band gap in CIGS. Furthermore, since both CuAlSe 2 and CuGaSe 2 have similar lattice parameters [9,10] then, according to Vegard’s law, CIAS films with identical band gap to CIGS can be fabricated with less lattice strain, compared to the lattice of CuInSe 2 . By reducing the lattice strain, less crystal defects (e.g. dislocations) are likely to form. CIAS could therefore find applications in single junction solar devices (i.e. to reach 1.35 eV with reduced losses) as well as in tan- dem or multi-junction devices, where the different combinations of band gaps identified as optimum can be reached. For this work, the so-called two-stage process was chosen over co-evaporation because it is easier to scale-up and yields equivalent or higher effi- ciencies for CIGS devices at the module scale [11]. In this article, we 0925-8388/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jallcom.2013.03.091 ⇑ Corresponding author. Tel.: +44 1912437013. E-mail address: guillaume.zoppi@northumbria.ac.uk (G. Zoppi). 1 Present address: Département Sciences et Analyse des Matériaux, Centre de Recherche Public – Gabriel Lippmann, 41 rue du Brill, L-4422 Belvaux, Luxembourg. Journal of Alloys and Compounds 566 (2013) 180–186 Contents lists available at SciVerse ScienceDirect Journal of Alloys and Compounds journal homepage: www.elsevier.com/locate/jalcom