13 Cu(InGa)Se2 Solar Cells William N. Shafarman1 and Lars Stolt2 1 University of Delaware, Newark, DE, USA, 2 Uppsala University, Uppsala, Sweden 13.1 INTRODUCTION Cu(InGa)Se2 -based solar cells have often been touted as being among the most promising of solar cell technologies for cost-effective power generation. This is partly due to the advantages of thin ﬁlms for low-cost, high-rate semiconductor deposition over large areas using layers only a few microns thick and for fabrication of monolithically interconnected modules. Perhaps more importantly, very high efﬁciencies have been demonstrated with Cu(InGa)Se2 at both the cell and the module levels. Currently, the highest solar cell efﬁ- ciency is 18.8% with 0.5 cm2 total area fabricated by the National Renewable Energy Laboratory (NREL) . Furthermore, several companies have demonstrated large area modules with efﬁciencies >12% including a conﬁrmed 13.4% efﬁciency on a 3459 cm2 module by Showa Shell . Finally, Cu(InGa)Se2 solar cells and modules have shown excellent long-term stability  in outdoor testing. In addition to its potential advantages for large-area terrestrial applications, Cu(InGa)Se2 solar cells have shown high radia- tion resistance, compared to crystalline silicon solar cells [4, 5] and can be made very lightweight with ﬂexible substrates, so they are also promising for space applications. The history of CuInSe2 solar cells starts with the work done at Bell Laboratories in the early 1970s, even though its synthesis and characterization were ﬁrst reported by Hahn in 1953  and, along with other ternary chalcopyrite materials, it had been char- acterized by several groups . The Bell Labs group grew crystals of a wide selection of these materials and characterized their structural, electronic, and optical properties [7–9]. The ﬁrst CuInSe2 solar cells were made by evaporating n-type CdS onto p-type single crystals of CuInSe2 . These devices were initially recognized for their potential as near-infrared photodetectors since their spectral response was broader and more uniform than Si photodetectors. Optimization for solar cells increased the efﬁciency to 12% as measured under outdoor illumination “on a clear day in New Jersey” . Handbook of Photovoltaic Science and Engineering. Edited by A. Luque and S. Hegedus 2003 John Wiley & Sons, Ltd ISBN: 0-471-49196-9 568 Cu(InGa)Se2 SOLAR CELLS There has been relatively little effort devoted to devices on single-crystal CuInSe2 since this early work, in part because of the difﬁculty in growing high-quality crystals . Instead, nearly all the focus has gone to thin-ﬁlm solar cells because of their inherent advantages. The ﬁrst thin-ﬁlm CuInSe2 /CdS devices were fabricated by Kazmerski et al. using ﬁlms deposited by evaporation of CuInSe2 powder along with excess Se . However, thin-ﬁlm CuInSe2 solar cells began to receive a lot of attention when the ﬁrst high-efﬁciency, 9.4%, cells were demonstrated by Boeing . At the same time, interest in Cu2 S/CdS thin-ﬁlm solar cells waned owing to problems related to electrochemical instabilities and many of these researchers turned their focus to CuInSe2 . The Boeing devices were fabricated using CuInSe2 deposited by coevaporation, that is, evaporation from separate elemental sources , onto ceramic substrates coated with a Mo back electrode. Devices were completed with evaporated CdS or (CdZn)S deposited in two layers with undoped CdS followed by an In-doped CdS layer that served as the main current-carrying material . Throughout the 1980s, Boeing and ARCO Solar began to address the difﬁcult manufacturing issues related to scale-up, yield, and throughput leading to many advancements in CuInSe2 solar cell technology. The two groups pursued different approaches to CuInSe2 deposition, which today remain the most common deposition methods and produce the highest device and module efﬁciencies. Boeing focused on depositing Cu(InGa)Se2 by coevaporation, while ARCO Solar focused on a two-stage process of Cu and In deposition at a low temperature followed by a reactive anneal in H2 Se. The basic solar cell conﬁguration implemented by Boeing provided the basis for a series of improvements that have lead to the high-efﬁciency device technology of today. The most important of these improvements to the technology include the following: • The absorber-layer band gap was increased from 1.02 eV for CuInSe2 to 1.1–1.2 eV by the partial substitution of In with Ga, leading to a substantial increase in efﬁciency . • The 1- to 2-µm-thick doped (CdZn)S layer was replaced with a thin, ≤50 nm, undoped CdS and a conductive ZnO current-carrying layer . This increased the cell current by increasing the short wavelength (blue) response. • Soda lime glass replaced ceramic or borosilicate glass substrates. Initially, this change was made for the lower costs of the soda lime glass and its good thermal expansion match to CuInSe2 . However, it soon became clear that an increase in device performance and processing tolerance resulted primarily from the beneﬁcial indiffusion of sodium from the glass . • Advanced absorber fabrication processes were developed that incorporate band gap gradients that improve the operating voltage and current collection [19, 20]. From its earliest development, CuInSe2 was considered promising for solar cells because of its favorable electronic and optical properties including its direct band gap with high absorption coefﬁcient and inherent p-type conductivity. As science and technology developed, it also became apparent that it is a very forgiving material since (1) high- efﬁciency devices can be made with a wide tolerance to variations in Cu(InGa)Se2 composition [21, 22], (2) grain boundaries are inherently passive so even ﬁlms with grain sizes less than 1 µm can be used, and (3) device behavior is insensitive to defects at the junction caused by a lattice mismatch or impurities between the Cu(InGa)Se2 and INTRODUCTION 569 CdS. The latter enables high-efﬁciency devices to be processed despite exposure of the Cu(InGa)Se2 to air prior to junction formation. High-efﬁciency CuInSe2 -based solar cells have been fabricated by at least 10 groups around the world. While these groups employ a variety of processing technologies, all the solar cells have the same basic cell structure built around a Cu(InGa)Se2/CdS junction in a substrate conﬁguration with a Mo back contact. Figure 13.1 shows a cross-sectional schematic of a standard device. This structure utilizes a soda lime glass substrate, coated with a sputtered Mo layer as a back contact. After Cu(InGa)Se2 deposition, the junction is formed by chemical bath–deposited CdS with thickness ≤50 nm. Then a high-resistance (HR) ZnO layer and a doped high-conductivity ZnO layer are deposited, usually by sputtering or chemical vapor deposition. Either a current-collecting grid or monolithic series interconnection completes the device or module, respectively. A TEM micrograph of the same structure, shown in Figure 13.2, clearly demonstrates the polycrystalline nature of these materials and the conformal coverage of the CdS layer. Current collection grid + HR-Zno/n -ZnO (0.5 µm) CdS (0.05 µm) Cu(InGa)Se2 (2 µm) Mo (0.5 µm) Soda lime glass Figure 13.1 Schematic cross section of a typical Cu(InGa)Se2 solar cell ZnO:Al i-ZnO CdS Cu(InGa)Se2 Mo Substrate Figure 13.2 TEM cross section of a Cu(InGa)Se2 solar cell 570 Cu(InGa)Se2 SOLAR CELLS Several companies worldwide are pursuing the commercial development of Cu(InGa)Se2 -based modules. The most advanced, having demonstrated excellent reproducibility in its module manufacturing using the two-stage selenization process for Cu(InGa)(SeS)2 deposition , is Shell Solar Industries (SSI) in California, which was formerly ARCO Solar and then Siemens Solar. They are now in production with 5-, 10-, u 20-, and 40-W modules that are commercially available. In Germany, W¨ rth Solar is in pilot production using an in-line coevaporation process for Cu(InGa)Se2 deposition and has also reported large area modules with >12% efﬁciency. In the USA, several companies are in preproduction or pilot production: Energy Photovoltaics, Inc. (EPV) is using its own in-line evaporation process, International Solar Electric Technology (ISET) is developing a particle-based precursor for selenization, and Global Solar Energy (GSE) is pursuing a process for roll-to-roll coevaporation onto a ﬂexible substrate. In Japan, Showa Shell, using a two-stage selenization process, and Matsushita, using coevaporation for Cu(InGa)Se2 deposition, are also in production development stages. Despite the level of effort on developing manufacturing processes, there remains a large discrepancy in efﬁciency between the laboratory-scale solar cells and minimod- ules, and the best full-scale modules. In part, this is due to the necessity for developing completely new processes and equipment for the large-area, high-throughput deposition needed for manufacturing thin-ﬁlm photovoltaics. This is compounded by the lack of a comprehensive scientiﬁc base for Cu(InGa)Se2 materials and devices, due partly to the fact that it has not attracted a broader interest for other applications. This lack of a sci- ence base has been perhaps the biggest hindrance to the maturation of Cu(InGa)Se2 solar cell technology as most of the progress has been empirical. Still, in many areas a deeper understanding has emerged in the recent years. In this chapter we will review the current status and the understanding of thin-ﬁlm Cu(InGa)Se2 solar cells from a technology perspective. For deeper scientiﬁc discussion of some aspects, we refer to suitable references. In order of presentation, this review covers (Section 13.2) structural, optical, and electrical properties of Cu(InGa)Se2 including a discussion of the inﬂuence of Na and O impurities; (Section 13.3) methods used to deposit Cu(InGa)Se2 thin ﬁlms, the most common of which can be divided into two general types, multisource coevaporation and two-stage processes of precursor deposition followed by Se annealing; (Section 13.4) junction and device formation, which typically is done with chemical bath CdS deposition and a ZnO conduction layer; (Section 13.5) device operation with emphasis on the optical, current-collection, and recombination-loss mechanisms; (Section 13.6) module-manufacturing issues, including process and performance issues and a discussion of environmental concerns; and ﬁnally, (Section 13.7) a discussion of the outlook for CuInSe2 -based solar cells and critical issues for the future. In places where aspects of Cu(InGa)Se2 solar cells cannot be covered in full, reference will be made to other review works that serve to complement this chapter. 13.2 MATERIAL PROPERTIES The understanding of Cu(InGa)Se2 thin ﬁlms, as used in photovoltaic (PV) devices, is primarily based on studies of its base material, pure CuInSe2 . Thorough reviews on CuInSe2 can be found in References [23–25]. However, the material used for making solar cells is Cu(InGa)Se2 containing signiﬁcant amounts (of the order of 0.1%) of Na . MATERIAL PROPERTIES 571 Even though the behavior of CuInSe2 provides a good basis for the understanding of device-quality material, there are pronounced differences when Ga and Na are present in the ﬁlms. More recently, Cu(InGa)Se2 has been reviewed in the context of solar cells with an emphasis on electronic properties . In this section the structural, optical, and electrical properties of CuInSe2 are reviewed along with information about the surface and grain boundaries and the effect of the substrate. In each case, as appropriate, the effect of the alloying with CuGaSe2 to form Cu(InGa)Se2 and the impact of Na and O on the material properties will be discussed. 13.2.1 Structure and Composition CuInSe2 and CuGaSe2 have the chalcopyrite lattice structure. This is a diamond-like structure similar to the sphalerite structure but with an ordered substitution of the group I (Cu) and group III (In or Ga) elements on the group II (Zn) sites of sphalerite. This gives a tetragonal unit cell depicted in Figure 13.3 with a ratio of the tetragonal lattice parameters c/a close to 2 (see Table 13.1). The deviation from c/a = 2 is called the tetragonal distortion and stems from different strengths of the Cu–Se and the In–Se or Ga–Se bonds. The possible phases in the Cu–In–Se system are indicated in the ternary phase diagram in Figure 13.4. Thin ﬁlms of Cu–In–Se prepared under an excess supply of Se, that is, normal conditions for thin-ﬁlm growth of Cu(InGa)Se2 , have compositions that fall on, or close to, the tie-line between Cu2 Se and In2 Se3 . Chalcopyrite CuInSe2 is located on this line as well as a number of phases called ordered defect compounds (ODC), because they have a lattice structure described by the chalcopyrite structure with an ordered inser- tion of intrinsic defects. A comprehensive study of the Cu–In–Se phase diagram has been o completed by G¨ decke et al. . A detail of the Cu2 Se–In2 Se3 tie-line near CuInSe2 is described by the pseudobinary phase diagram reproduced in Figure 13.5  Here α is the chalcopyrite CuInSe2 , δ is a high-temperature (HT) phase with the sphalerite structure, and β is an ODC phase. It is interesting to note that the single phase ﬁeld for CuInSe2 at low temperatures is relatively narrow as compared to earlier beliefs, and does not contain Cu In Se Figure 13.3 The unit cell of the chalcopyrite lattice structure 572 Cu(InGa)Se2 SOLAR CELLS Table 13.1 Selected properties of CuInSe2 Property Value Units Reference a 5.78 A˚  Lattice constant c 11.62 A˚ Density 5.75 g/cm3  Melting temperature 986 C  Thermal expansion (a axis) 8.32 × 10−6 1/K  coefﬁcients at 273 K (c axis) 7.89 × 10−6 1/K Thermal conductivity 0.086  at 273 K Dielectric constant Low frequency 13.6 ± 2.4  High frequency 8.1 ± 1.4 Electrons 0.09  Effective mass [me ] Holes (heavy) 0.71  Holes (light) 0.092  Energy gap 1.02 eV  Energy gap temperature −2 × 10−4 eV/K  coefﬁcient Se CuSe2 ODC phases In2Se3 CuSe InSe CuInSe2 Cu2Se In2Se Cu In Figure 13.4 Ternary phase diagram of the Cu–In–Se system. Thin-ﬁlm composition is usually near the pseudobinary Cu2 Se–In2 Se3 tie-line the composition 25% Cu. At higher temperatures, around 500◦ C, where thin ﬁlms are grown, the phase ﬁeld widens toward the In-rich side. Typical average compositions of device-quality ﬁlms have 22 to 24 at.% Cu, which fall within the single-phase region at growth temperature. CuInSe2 can be alloyed in any proportion with CuGaSe2 , thus forming Cu(InGa)Se2 . Similarly, the binary phase In2 Se3 at the end point of the pseudobinary tie-line can be alloyed to form (InGa)2 Se3 , although it undergoes a structural change at Ga/(In + Ga) = 0.6 . In high-performance devices, Ga/(In + Ga) ratios are typically 0.2 to 0.3. One of the central characteristics of Cu(InGa)Se2 is its ability to accommodate large variations in composition without appreciable differences in optoelectronic properties. MATERIAL PROPERTIES 573 800 785 d 700 a+ Cu2Se (HT) 600 a +d a Temperature 520 [°C] 500 b 400 300 a +b a+ Cu2Se (RT) 200 134 100 15 20 25 30 Cu [at.%] Figure 13.5 Pseudobinary In2 Se3 –Cu2 Se equilibrium phase diagram for compositions around the CuInSe2 chalcopyrite phase, denoted α. The δ phase is the high-temperature sphalerite phase, and the β phase is an ordered defect phase (ODC). Cu2 Se exists as a room-temperature (RT) or o high-temperature (HT) phase. (After G¨ decke T, Haalboom T, Ernst F, Z. Metallkd. 91, 622–634 (2000) ) This tolerance is one of the cornerstones of the potential of Cu(InGa)Se2 as a material for efﬁcient low-cost PV modules. Solar cells with high performance can be fabricated with Cu/(In + Ga) ratios from 0.7 to nearly 1.0. This property can be understood from theoretical calculations that show that the defect complex 2VCu + InCu , that is, two Cu vacancies with an In on Cu antisite defect, has very low formation energy, and also that it is expected to be electrically inactive . Thus, the creation of such defect complexes can compensate for Cu-poor/In-rich compositions of CuInSe2 without adverse effects on the photovoltaic performance. Furthermore, crystallographic ordering of this defect complex is predicted , which explains the observed ODC phases Cu2 In4 Se7 , CuIn3 Se5 , CuIn5 Se8 , and so on. The chalcopyrite phase ﬁeld is increased by the addition of Ga or Na . This can be explained by a reduced tendency to form the ordered defect compounds owing to higher formation energy for GaCu (in CuGaSe2 ) than for InCu (in CuInSe2 ). This leads to desta- bilization of the 2VCu + InCu defect cluster related to the ODC phases [36, 37]. The effect of Na in the CuInSe2 structure has been calculated by Wei et al. , with the result that Na replaces InCu antisite defects, reducing the density of compensating donors. This the- oretical result is supported by measurements of epitaxial Cu(InGa)Se2 ﬁlms in which Na is found to strongly reduce the concentration of compensating donors . Together with a tendency for Na to occupy Cu vacancies, the reduced tendency to form antisite defects also suppresses the formation of the ordered defect compounds. The calculated effect of Na is therefore consistent with the experimental observations of increased compositional range in which single-phase chalcopyrite exists and increased conductivity [38, 39]. 574 Cu(InGa)Se2 SOLAR CELLS 13.2.2 Optical Properties The absorption coefﬁcient α for CuInSe2 is very high, larger than 105 /cm for 1.4 eV and higher photon energies . In many studies it was found that the fundamental absorption edge is well described by  α = A(E − Eg )2 /E (13.1) as for a typical direct band gap semiconductor. The proportionality constant A depends on the density of states associated with the photon absorption. From this relation, a band gap value of Eg = 1.02 ± 0.02 eV is obtained. The temperature dependence follows Eg (T ) = Eg (0) − a T 2 /(b + T ) (13.2) where a and b are constants that vary between different measurements. In general, dEg /dT is about −2 × 10−4 eV/K . A rather complete picture of the optical properties of CuInSe2 and other Cu-ternary chalcopyrites is given in Reference . Ellipsometric measurements of carefully pre- pared single-crystal samples were carried out and the dielectric functions were obtained together with the complex refractive index for different polarizations. From these mea- surements a band gap value for CuInSe2 of 1.04 eV was determined. A similar study was also made on bulk polycrystalline ingots of Cu(InGa)Se2 having different compositions from x ≡ Ga/(Ga + In) = 0 to 1 . Curves describing the complex refractive index, n + ik, for samples with x = 0 and 0.2 are reproduced in Figure 13.6. The complex refractive index can be used to calculate other optical param- eters like the absorption coefﬁcient α = 4πk/λ (13.3) In the same work the fundamental transitions for the different compositions were ﬁt to an equation describing the band gap for CuIn1−x Gax Se2 as Eg = 1.010 + 0.626x − 0.167x(1 − x) (13.4) In this equation the so-called bowing coefﬁcient is 0.167. A value of 0.21 was obtained by theoretical calculations as compared to values in the range of 0.11 to 0.26 determined in various experiments . 13.2.3 Electrical Properties CuInSe2 with an excess of Cu is always p-type but In-rich ﬁlms can be made p-type or n-type . By annealing in a selenium overpressure, n-type material can be converted to p-type, and conversely, by annealing in a low selenium pressure, p-type material becomes n-type . It is believed that this affects the concentration of Se vacancies, VSe , which act as compensating donors in p-type ﬁlms. Device-quality Cu(InGa)Se2 ﬁlms, grown with the excess Se available, are p-type with a carrier concentration of about 1016 /cm3 . MATERIAL PROPERTIES 575 3.2 x = 0.2 2.8 n x=0 2.4 2.0 1.5 1.0 x=0 k 0.5 x = 0.2 0.0 1 2 3 4 5 Energy [eV] Figure 13.6 Complex refractive index for CuInSe2 and CuIn1−x Gax Se2 with x = 0.2 (After Alonso M et al., Appl. Phys. A 74, 659–664 (2002) ) There is a large spread in mobility values reported for CuInSe2 . The highest values of hole mobilities have been obtained for epitaxial ﬁlms, where 200 cm2 /Vs has been measured for Cu(InGa)Se2 with about 1017 /cm3 in hole concentration . Single crystals have yielded values in the range of 15 to 150 cm2 /Vs. Electron mobilities determined from single crystals range from 90 to 900 cm2 /Vs . Conductivity and Hall effect measurements of thin-ﬁlm samples are made cross-grain, but for device operation through-the-grain values are more relevant, since individual grains may extend from the back contact to the interface of the junction. The sheet conductivities of polycrystalline p-type ﬁlms correspond to mobility values of 5 to 50 cm2 /Vs, but it is quite possible that they are limited by transport across grain boundaries. A large number of intrinsic defects are possible in the chalcopyrite structure. Accordingly, a number of electronic transitions have been observed by methods such as photoluminescence, photoconductivity, photovoltage, optical absorption, and electrical measurements (see, for example, Reference ). However, it is difﬁcult to assign transi- tions to speciﬁc defects on an experimental basis. Instead, theoretical calculations of the transition energies and formation energies provide a basis for identiﬁcation of the different intrinsic defects that are active in Cu(InGa)Se2 . Calculations of intrinsic defects in CuInSe2 and comparison with experimental data can be found in the comprehensive paper by Zhang et al. . A summary of their results is schematically shown in Figure 13.7. The defects that are considered most important in device-quality material are presented in Table 13.2. 576 Cu(InGa)Se2 SOLAR CELLS CBM D1 1.0 D2 Cui(0/+) D3 0.8 InCu(0/+) InCu(+/2+) VIn(3−/2−) D4 Energy 0.6 CuIn(2−/−) [eV] A6 D5 0.4 VIn(2−/−) A5 CuIn(−/0) A4 0.2 VIn(−/0) A3 A2 VCu(−/0) A1 0.0 VBM Theory Experiment Figure 13.7 Electronic levels of intrinsic defects in CuInSe2 . On the left side the theoretical values are presented and on the right side experimentally reported values are presented. The height of the histogram columns on the right side represents the spread in experimental data. (From Zhang S, Wei S, Zunger A, Katayama-Yoshida H, Phys. Rev. B 57, 9642–9656 (1998) ) Table 13.2 The most important intrinsic defects for device-quality CuInSe2 Defect Energy position Type VCu EV + 0.03 eV Shallow acceptor InCu EC − 0.25 eV Compensating donor VSe Compensating donor CuIn EV + 0.29 eV Recombination center The effect of Ga on the electronic and defect properties is discussed in Refer- ence . In those calculations, acceptor levels did not differ very much between CuInSe2 and CuGaSe2 , but the donor levels are deeper in the Ga-containing compound. This is con- sistent with observations of increased p-type conductivity at high Ga-concentrations . At typical device compositions, Ga/(Ga + In) < 0.3, any effect of increased Ga content on conductivity has not been veriﬁed. 13.2.4 The Surface and Grain Boundaries The surface morphology and grain structure are most commonly characterized by scan- ning electron microscopy (SEM), but transmission electron microscopy (TEM) and atomic force microscopy have also proved valuable. A typical SEM image is shown in Figure 13.8 and a TEM cross-sectional image in Figure 13.2. In general, the ﬁlms used in devices MATERIAL PROPERTIES 577 1µm Figure 13.8 Scanning electron microscopy image of a typical Cu(InGa)Se2 ﬁlm deposited on a Mo-coated glass substrate by coevaporation have grain diameters on the order of 1 µm but the grain size and morphology can vary greatly depending on fabrication method and conditions. A variety of defects including twins, dislocations, and stacking faults have been observed [49–51]. It has been shown by X-ray photoelectron spectroscopy (XPS) that the free sur- faces of CuInSe2 ﬁlms with slightly Cu-poor composition have a composition close to CuIn3 Se5 , corresponding to one of the ordered defect phases. Many attempts have been made to identify such a layer on top of the ﬁlms without success. It merely seems as if the composition gradually changes from the bulk to the surface of the ﬁlms. It was proposed by Herberholz et al.  that band bending induced by surface charges drives electromigrating Cu into the bulk leaving the surface depleted of Cu. This depletion is stopped when the composition is that of CuIn3 Se5 , since further depletion requires a struc- tural change of the material. Electromigration of Cu in CuInSe2 has been demonstrated and also correlated with type conversion of the chalcopyrite material . The band bending as well as the CuIn3 Se5 composition of CuInSe2 surfaces dis- appears when the material is exposed to atmosphere for some time as oxides form on the surface. The surface oxidation is enhanced by the presence of Na . The surface com- pounds after oxidation have been identiﬁed as In2 O3 , Ga2 O3 , SeOx , and Na2 CO3 . A review of the surface and the interface properties can be found in Reference . It has been common practice to posttreat Cu(InGa)Se2 devices in air at typically 200◦ C. When devices were fabricated using vacuum-evaporated CdS or (CdZn)S to form the junction, these anneals were often done for several hours to optimize the device performance [14, 56]. The main effect associated with oxygen is explained as passivation of selenium “surface” vacancies on the grains . The VSe at the grain boundaries can act as a recombination center. The positive charge associated with these donor-type defects reduces the effective hole concentration at the same time that the intergrain carrier transport is impeded. When oxygen substitutes for the missing selenium, these negative effects are canceled. The overall noted beneﬁcial effect of the presence of Na on the PV performance of Cu(InGa)Se2 thin ﬁlms lacks a complete explanation. In Reference  it is proposed that the catalytic effect of Na on oxidation, by enhanced dissociation of molecular oxygen into atomic oxygen, makes the passivation of VSe on grain surfaces more effective. This model is consistent with the observation that Na and O are predominantly found at the grain boundaries rather than in the bulk of the grains in CuInSe2 thin ﬁlms . 578 Cu(InGa)Se2 SOLAR CELLS 13.2.5 Substrate Effects The effects of the substrate on the properties of thin-ﬁlm polycrystalline Cu(InGa)Se2 can be classiﬁed into three categories: (1) thermal expansion, (2) chemical effects, and (3) surface inﬂuence on nucleation. It can be assumed that after growth, when the substrate and ﬁlm are still at the growth temperature, the stress in the Cu(InGa)Se2 ﬁlm is low. The cooling down from growth temperature imposes a temperature change of about 500◦ C, and if the thermal expansion of the substrate and Cu(InGa)Se2 is different stress will be built up in the ﬁlm. The thermal expansion coefﬁcient for CuInSe2 is around 9 × 10−6 /K in the temperature interval of interest, which is similar to that of soda lime glass. A CuInSe2 ﬁlm deposited on a substrate with a lower thermal expansion coefﬁcient, such as borosilicate glass, will be under increasing tensile stress during cooldown. Typically, such ﬁlms exhibit voids and microcracks . When the thermal expansion coefﬁcient of the substrate is higher than that of the ﬁlm material, like for polyimide, it will result in compressive stress in the thin-ﬁlm material, which may lead to adhesion failures. The most important effect of the soda lime glass substrate on Cu(InGa)Se2 ﬁlm growth is that it supplies sodium to the growing chalcopyrite material. It has been clearly shown that this effect is distinct from the thermal expansion match of soda lime glass . The sodium diffuses through the Mo back contact, which also means that it is important to control the properties of the Mo . The resulting microstructure of Cu(InGa)Se2 is clearly inﬂuenced by the presence of Na with larger grains and a higher degree of preferred orientation, with the (112) axis perpendicular to the substrate. An explanation for this effect when high concentrations of Na are present has been proposed by Wei et al. . There is a wide range of preferred orientation between different growth processes, in spite of similar device performance. One reason for this variation is most likely the different properties of the surfaces on which the chalcopyrite material nucleates. A com- parison between Cu(InGa)Se2 grown on normal Mo-coated substrates and directly on soda lime glass shows that a much more pronounced (112) orientation occurs on glass, in spite of no difference in the Na concentration, as measured in the ﬁlms afterwards . Further, the preferred orientation of the Cu(InGa)Se2 ﬁlm has been shown to be correlated to the orientation of the Mo ﬁlm  or an (InGa)2 Se3 precursor layer . 13.3 DEPOSITION METHODS A wide variety of thin-ﬁlm deposition methods has been used to deposit Cu(InGa)Se2 thin ﬁlms. To determine the most promising technique for the commercial manufac- ture of modules, the overriding criteria are that the deposition can be completed at low cost while maintaining high deposition or processing rate with high yield and repro- ducibility. Compositional uniformity over large areas is critical for high yield. Device considerations dictate that the Cu(InGa)Se2 layer should be at least 1 µm thick and that the relative compositions of the constituents are kept within the bounds determined by the phase diagram, as discussed in Section 13.2.1. For solar cell or module fabrication, the Cu(InGa)Se2 is most commonly deposited on a molybdenum-coated glass substrate, DEPOSITION METHODS 579 though other substrate materials including metal or plastic foils have also been used and may have processing advantages. The most promising deposition methods for the commercial manufacture of mod- ules can be divided into two general approaches that have both been used to demonstrate high device efﬁciencies and in pilot scale manufacturing. The ﬁrst approach is vacuum coevaporation in which all the constituents, Cu, In, Ga, and Se, can be simultaneously delivered to a substrate heated to 400 to 600◦ C and the Cu(InGa)Se2 ﬁlm is formed in a single growth process. This is usually achieved by thermal evaporation from elemental sources at temperatures greater than 1000◦ C for Cu, In, and Ga. The second approach is a two-step process that separates the delivery of the metals from the reaction to form device-quality ﬁlms. Typically the Cu, Ga, and In are deposited using low-cost and low- temperature methods that facilitate uniform composition. Then the ﬁlms are annealed in a Se atmosphere, also at 400 to 600◦ C. The reaction and anneal step often takes longer time than formation of ﬁlms by coevaporation due to diffusion kinetics, but is amenable to batch processing. High process rate can be achieved by moving continuously through sequential process steps or with a batch process whereby longer deposition or reaction steps can be implemented by handling many substrates in parallel. 13.3.1 Substrates Soda lime glass, which is used in conventional windows, is the most common sub- strate material used for Cu(InGa)Se2 since it is available in large quantities at low cost and has been used to make the highest efﬁciency devices. Cu(InGa)Se2 depo- sition requires a substrate temperature (TSS ) of at least 350◦ C and the highest efﬁ- ciency cells have been fabricated using ﬁlms deposited at the maximum temperature, TSS ≈ 550◦ C, which the glass substrate can withstand without softening too much . The glass is electrically insulating and smooth, which enables monolithic integration into modules. The soda lime glass has a thermal expansion coefﬁcient of 9 × 10−6 /K , which provides a good match to the Cu(InGa)Se2 ﬁlms. The glass composition typically includes various oxides such as Na2 O, K2 O, and CaO. These provide sources of alkali impurities that diffuse into the Mo and Cu(InGa)Se2 ﬁlms during processing , producing the beneﬁcial effects discussed in Section 13.2. However, a process that provides a more controllable supply of Na than diffusion from the glass substrate is preferred. This can be achieved by blocking sodium from the substrate with a diffusion barrier such as SiOx or Al2 O3 . Then sodium can be directly provided to the Cu(InGa)Se2 growth process by depositing a sodium-containing precursor layer onto the Mo ﬁlm [65, 66]. Commercially available soda lime glass may also contain signiﬁcant structural defects that can adversely impact module production . Borosilicate glass does not contain the alkali impurities and may have fewer structural imperfections but has a lower thermal expansion coefﬁcient, 4.6 × 106 /K , and is more expensive. Substrates such as metal or plastic foils have advantages over glass substrates owing to their light weight and ﬂexibility, which will be discussed in Section 13.6. Cu(InGa)Se2 devices have been demonstrated with different metal and high-temperature polyimide substrates [68, 69]. 580 Cu(InGa)Se2 SOLAR CELLS 13.3.2 Back Contact The Mo back contact, used for all high-efﬁciency devices, is typically deposited by direct current (dc) sputtering. The thickness is determined by the resistance requirements that depend on the speciﬁc cell or module conﬁguration. A ﬁlm with thickness 1 µm will typically have a sheet resistance of 0.1 to 0.2 / , a factor of 2 to 4 higher resistivity than bulk Mo. Sputter deposition of the Mo layer requires careful control of the pressure to control stress in the ﬁlm  and to prevent problems such as poor adhesion that it might cause. During Cu(InGa)Se2 deposition, a MoSe2 layer forms at the interface . Its properties are inﬂuenced by the Mo ﬁlm with less MoSe2 forming on dense Mo, sputter-deposited under low pressures . This interfacial layer does not necessarily degrade device performance. Metals other than Mo have been investigated with limited success . 13.3.3 Coevaporation of Cu(InGa)Se2 The highest efﬁciency devices have been deposited by thermal coevaporation from ele- mental sources . An illustration of a laboratory system for Cu(InGa)Se2 coevaporation is shown in Figure 13.9. The process uses line-of-sight delivery of the Cu, In, Ga, and Se from Knudsen-type effusion cells or open-boat sources to the heated substrate. While the evaporation temperatures for each metal will depend on the speciﬁc source design, typical ranges are 1300 to 1400◦ C for Cu, 1000 to 1100◦ C for In, 1150 to 1250◦ C for Ga, and 300 to 350◦ C for Se evaporation. The sticking coefﬁcients of Cu, In, and Ga are very high, so the ﬁlm composition and growth rate are determined simply by the ﬂux distribution and effusion rate from each source. The composition of the ﬁnal ﬁlm tends to follow the pseudobinary tie-line between (InGa)2 Se3 and Cu2 Se (see Figure 13.4) according to the relative concentration of Cu compared to In and Ga. The relative concentrations of In and Ga determine the band gap of the ﬁlm, according to equation (13.4), and the effusion rates can be varied over the course of a deposition to change the ﬁlm composition through its thickness. Se has a Heater and substrate Growth monitor Evaporation sources To pump Figure 13.9 Conﬁguration for multisource elemental coevaporation DEPOSITION METHODS 581 much higher vapor pressure and lower sticking coefﬁcient, so it is always evaporated in excess of that needed in the ﬁnal ﬁlm. Insufﬁcient Se can result in a loss of In and Ga in the form of In2 Se or Ga2 Se . Different deposition variations, using elemental ﬂuxes deliberately varied over time, have been explored using coevaporation. Four different sequences that have been used to fabricate devices with efﬁciencies greater than 16% are shown in Figure 13.10. In each case, the targeted ﬁnal composition is Cu-deﬁcient with Cu/(In + Ga) = 0.8 − 0.9. The total deposition time may vary from 10 to 90 min, depending on the effusion rates from the sources. So, for a ﬁlm thickness of 2 µm, typical deposition rates vary from 20 to 200 nm/min. The ﬁrst process is the simplest stationary process in which all ﬂuxes are constant throughout the deposition process . In most cases, however, the ﬂuxes are varied using what is referred to as the Boeing process in which the bulk of the ﬁlm is grown with Cu- rich overall composition so that it contains a Cux Se phase in addition to Cu(InGa)Se2 . The ﬂuxes are then adjusted to ﬁnish the deposition with In- and Ga-rich ﬂux so that the ﬁnal ﬁlm composition has the desired Cu-deﬁcient composition. One modiﬁcation of this is the second process shown in Figure 13.10. This process was ﬁrst implemented with CuInSe2 ﬁlms deposited on non-Na containing substrates at TSS = 450◦ C, producing ﬁlms with increased grain size and improved device performance. The effect of Cux Se as a ﬂux for enhanced grain growth at higher TSS was proposed by Klenk et al. . However, in devices containing Na and Ga and with TSS > 500◦ C, no difference was found in the device performance using ﬁlms with Cu-rich or uniform growth processes . The third process shown in Figure 13.10 is a sequential process in which the In and Ga are deposited separately from the Cu. This was ﬁrst proposed by Kessler et al.  with the deposition of an (InGa)x Sey compound, followed by the deposition of Cu and Se until the growing ﬁlm reaches the desired composition. The layers interdiffuse to form the Cu(InGa)Se2 ﬁlm. A modiﬁcation by Gabor et al.  allows the Cu delivery to continue until the ﬁlm has an overall Cu-rich composition. Then a third step is added to the process in which In and Ga, again in the presence of excess Se, are evaporated to bring the composition back to Cu-deﬁcient. The metals interdiffuse, forming the ternary chalcopyrite ﬁlm. This process has been used to produce the highest efﬁciency devices . The improved device performance has been attributed to a band gap gradient, which results from the Ga concentration decreasing from the Mo back contact to the ﬁlm’s free surface , and to improved crystallinity of the ﬁlms . The last process shown in Figure 13.10 is an in-line process in which the ﬂux distribution results from the substrate moving sequentially over the Cu, Ga, and In sources. This was ﬁrst simulated in a stationary evaporation system  and has subsequently been implemented by several groups in pilot manufacturing systems (see Section 13.6). A reproducible coevaporation process requires good control of the elemental ﬂuxes from each evaporation source. While the evaporation rates from each source can be con- trolled simply by the source temperature, this may not give good reproducibility, especially for the Cu source that is at the highest temperature. Open-boat sources in particular will not give reproducible evaporation rates. Consequently, direct in situ measurement of the ﬂuxes is often used to control the evaporation sources. Electron impact spectroscopy , quadrupole mass spectroscopy , and atomic absorption spectroscopy  have all 582 Cu(InGa)Se2 SOLAR CELLS TSS 550 0.3 Relative flux Cu TSS [C] 0.2 In 450 0.1 Ga 350 0 TSS 550 0.3 Cu Relative flux TSS [C] 0.2 450 In 0.1 350 Ga 0 TSS 550 0.3 Relative flux Cu TSS 0.2 450 [C] In 0.1 Ga 350 0 TSS 550 0.3 Relative flux TSS [C] Cu 450 0.2 In 0.1 350 Ga 0 Time [min] Figure 13.10 Relative metal ﬂuxes and substrate temperature for different coevaporation pro- cesses. In all cases, a constant Se ﬂux is also supplied been successfully implemented. Direct ﬂux measurement may be critical in a manufactur- ing scale process, particularly if source depletion over long run times causes the relation between source temperature and effusion rate to vary over time. In addition, the process can be monitored by in situ ﬁlm thickness measurement using a quartz crystal monitor, or optical spectroscopy or X-ray ﬂuorescence of the growing ﬁlm . The latter has also been used to measure composition. When the process includes a transition from Cu-rich to Cu-poor composition near the end of the deposition, it can be monitored by a change DEPOSITION METHODS 583 in the temperature resulting from a change in the emissivity of the ﬁlm  or by the infrared transmission . The primary advantage of elemental coevaporation for depositing Cu(InGa)Se2 ﬁlms is its considerable ﬂexibility to choose the process speciﬁcs and to control ﬁlm composition and band gap. As proof of this ﬂexibility, high-efﬁciency devices have been demonstrated using many process variations. The primary disadvantage results from the difﬁculty in control, particularly of the Cu-evaporation source, and the resulting need for improved deposition, diagnostic, and control technology. A second disadvantage is the lack of commercially available equipment for large-area thermal evaporation. 13.3.4 Two-step Processes The second common approach to Cu(InGa)Se2 ﬁlm formation, usually referred to as two- step processing or selenization, has many variations in both the precursor deposition and the Se reaction steps. This general approach was ﬁrst demonstrated by Grindle et al.  who sputtered Cu/In layers and reacted them in hydrogen sulﬁde to form CuInS2 . This was ﬁrst adapted to CuInSe2 by Chu et al. . The highest-efﬁciency Cu(InGa)Se2 cell reported using the reaction in H2 Se is 16.2%, on the basis of the active area , but there has been less effort at optimizing laboratory-scale cell efﬁciencies than with coevaporated Cu(InGa)Se2 . Showa Shell and Shell Solar have successfully scaled up this process to pilot commercial production and have demonstrated large-area module efﬁciencies as high as 13.4% . The metal precursor is used to determine the ﬁnal composition of the ﬁlm and to ensure spatial uniformity. Sputtering is an attractive process because it is easily scal- able using commercially available deposition equipment and can provide good uniformity over large areas with high deposition rates. However, other processes may have lower cost. CuInSe2 has been formed using metal precursor layers deposited by electrodeposi- tion , thermal or electron beam evaporation , screen printing , and application of nanoparticles . Precursors that include Se, such as stacked layers of Cu/In/Se  or binary selenides, have also been used as precursor materials in various sequences and combinations . Electrodeposition [95, 96] of Cu, In, Ga, and Se is effectively just another option for precursor deposition since the ﬁlms similarly require a selenium reaction step. The precursor ﬁlms are typically reacted in either H2 Se or Se vapor at 400 to 500◦ C for 30 to 60 min to form the best device quality material. Poor adhesion  and formation of a MoSe2 layer  at the Mo/CuInSe2 interface may limit the reaction time and temperature. Reaction in H2 Se has the advantage that it can be done at atmo- spheric pressure and can be precisely controlled, but the gas is highly toxic and requires special precautions for its use. The precursor ﬁlms can also be reacted in a Se vapor, which might be obtained by thermal evaporation, to form the CuInSe2 ﬁlm . A third reaction approach is rapid thermal processing (RTP) of either elemental layers, including Se, [99, 100] or amorphous evaporated Cu–In–Se layers . The reaction chemistry and kinetics for the conversion of Cu–In precursors to CuInSe2 has been characterized by X-ray diffraction of time-progressive reactions  and by in situ differential scanning calorimetry . The results of these experiments 584 Cu(InGa)Se2 SOLAR CELLS describe CuInSe2 formation as a sequence of reactions starting with the formation of Cu11 In9 and In liquid, which will contain a small concentration of dissolved Cu. These react with Se to form a series of binary compounds. The formation of CuInSe2 then follows from 2 InSe + Cu2 Se + Se → 2 CuInSe2 with complete reaction in ∼15 min at 400◦ C. The reaction path was shown to be the same for the reaction of Cu/In layers in either H2 Se or elemental Se . The addition of Ga, regardless of the precursor deposition sequence, does not readily give a ﬁlm with uniformly increased band gap. Instead, all Ga in the reacted ﬁlm accumulates near the Mo forming a CuInSe2 /CuGaSe2 structure, so the resulting device behaves like CuInSe2  and lacks the increased operating voltage and other beneﬁts of a wider band gap discussed in Section 13.5.4. Nevertheless, Ga inclusion provides improved adhesion of the CuInSe2 ﬁlm to the Mo back contact and greater device performance, possibly owing to an improved structure with fewer defects . The Ga and In can be effectively interdiffused, converting the ﬁlms to uniform band gap, by annealing in an inert atmosphere for 1 h at 600◦ C . This anneal, however, may be impractical for commercial processing, so ﬁlms in the best devices have the band gap increased by the incorporation of S near the front surface, forming a graded Cu(InGa)(SeS)2 layer [20, 107] that can give enhanced operating voltage in devices. The primary advantages of two-step processes for Cu(InGa)Se2 deposition are the ability to utilize more standard and well-established techniques for the metal deposition and reaction and anneal steps and to compensate for long reaction times with a batch processing mode or RTP of Se-containing precursors. Composition and uniformity are controlled by the precursor deposition and can be measured between the two steps. The biggest drawback to these processes is the limited ability to control composition and increase band gap, which may limit device and module performance. Other difﬁculties that must be overcome include poor adhesion and the use of hydrogen selenide, which is hazardous and costly to handle. 13.3.5 Other Deposition Approaches CuInSe2 -based ﬁlms have been deposited using a wide range of thin-ﬁlm deposition methods, in addition to those discussed above, which have been proposed as potential low-cost alternatives for manufacturing. These include reactive sputtering , hybrid sputtering in which Cu, In, and Ga are sputtered while Se is evaporated , closed- space sublimation , chemical bath deposition (CBD) , laser evaporation , and spray pyrolosis . Great effort was made to explore different thin-ﬁlm deposition techniques before coevaporation and the two-step processes above became dominant. These methods are reviewed in Reference . 13.4 JUNCTION AND DEVICE FORMATION The ﬁrst experimental device that indicated the potential for CuInSe2 in high-performance solar cells was a heterojunction between a p-type single crystal of CuInSe2 and a thin ﬁlm of n-type CdS [10, 11]. Consequently, in the early thin-ﬁlm work the junction was JUNCTION AND DEVICE FORMATION 585 formed by depositing CdS on the CuInSe2 ﬁlms . The device was further developed to contain an undoped layer of CdS, followed by CdS doped with In, both deposited by vacuum evaporation . This deﬁned the device structure (see Figure 13.1), which is basically the same as is commonly used today since the doped CdS is functionally a transparent conductor. A performance gain was achieved by alloying the CdS with ZnS to widen the band gap . Further improvement of the performance was achieved when the doped CdS layer was replaced with doped ZnO [115, 116]. The undoped CdS layer adjacent to the Cu(InGa)Se2 ﬁlm was reduced in thickness in order to maximize the optical transmission. Since ZnO has a wider band gap than CdS, more light is transmitted into the active part of the device, resulting in a current gain. A conformal and pinhole-free coating of this thin CdS layer is obtained by using chemical bath deposition to make the CdS buffer layer. 13.4.1 Chemical Bath Deposition Chemical bath deposition (CBD) of thin-ﬁlm materials can be viewed as a chemical vapor deposition (CVD) in the liquid phase instead of the gas phase. It is also referred to as solution growth. The method has been used in particular for chalcogenide materials such as PbS , CdS , and CdSe . A variety of precursor compounds or ions can be used to deposit a speciﬁc compound. Deposition of CdS buffer layers on Cu(InGa)Se2 is generally made in an alkaline aqueous solution (pH > 9) of the following three constituents: 1. a cadmium salt; for example, CdSO4 , CdCl2 , CdI2 , Cd(CH3 COO)2 2. a complexing agent; commonly NH3 (ammonia) 3. a sulfur precursor; commonly SC(NH2 )2 (thiourea). The concentrations of the various components of the solution can be varied over a range and each laboratory tends to use its own speciﬁc recipe. One example of a recipe that is being used to fabricate state-of-the-art Cu(InGa)Se2 solar cells is 1. 1.4 × 1/103 M CdI2 or CdSO4 2. 1 M NH3 3. 0.14 M SC(NH2 )2 The Cu(InGa)Se2 ﬁlm is immersed in a bath containing the solution and the deposition takes place in a few minutes at a temperature of 60 to 80◦ C. This can be done either by immersion in a room-temperature bath that subsequently is heated to the desired temperature or by preheating the solution. The reaction proceeds according to the formula Cd(NH3 )4 2+ + SC(NH2 )2 + 2 OH− → CdS + H2 NCN + 4 NH3 + 2 H2 O In practice, the chemical bath deposition is typically done in the laboratory with a very simple apparatus consisting of a hot plate with magnetic stirring, a beaker holding the solutions into which the substrate is immersed, and a thermocouple to measure bath 586 Cu(InGa)Se2 SOLAR CELLS Sample holder Plastic lid Glass container Water Sample Solution PTFE-coated magnet Magnetic stirrer Figure 13.11 Typical laboratory apparatus for chemical bath deposition of CdS temperature. A typical arrangement, incorporating a water bath for more uniform temper- ature, is shown in Figure 13.11. Scale-up of the CBD process for manufacturing will be discussed in Section 13.6.1. The growth of CdS thin ﬁlms by CBD occurs from ion by ion reaction or by clustering of colloidal particles. Depending on the bath condition, the resulting CdS lattice structure may be cubic, hexagonal, or a mixture . Under typical conditions used for Cu(InGa)Se2 solar cells, the relatively thin CdS layers grow ion by ion, resulting in dense homogeneous ﬁlms  with mixed cubic/hexagonal or predominantly hexagonal lattice structure [51, 122, 123]. The ﬁlms consist of crystallites with a grain size of the order of tens of nanometers . Compositional deviation from stoichiometry is commonly observed. In particular, ﬁlms tend to be sulfur-deﬁcient and contain substantial amounts of oxygen [124, 125]. In addition to oxygen, signiﬁcant concentrations of hydrogen, carbon, and nitrogen have also been detected in device quality ﬁlms . The concentration of these impurities has been correlated to a reduction of the optical band gap and the amount of cubic CdS in relation to hexagonal CdS . 13.4.2 Interface Effects The interface between the Cu(InGa)Se2 and the CdS is characterized by pseudoepitaxial growth of the CdS and intermixing of the chemical species. Electronic band alignment will be discussed in Section 13.5.3. Transmission electron microscopy has shown that chemical bath–deposited CdS layers on Cu(InGa)Se2 ﬁlms exhibit an epitaxial relation- ship at the interface with (112) chalcopyrite Cu(InGa)Se2 planes parallel to the (111) cubic or (002) hexagonal CdS planes [51, 123]. The lattice mismatch is very small for pure CuInSe2 with a (112) spacing of 0.334 nm as compared to a spacing of 0.336 nm for (111) cubic and (002) hexagonal CdS. In Cu(InGa)Se2 the lattice mismatch increases with the Ga content. CuIn0.7 Ga0.3 Se2 and CuIn0.5 Ga0.5 Se2 have (112) spacing of 0.331 nm JUNCTION AND DEVICE FORMATION 587 0.34 0.33 Lattice constant [nm] 0.32 CuIn1−xGaxSe2 Cd1−x ZnxS 0.31 0.0 0.2 0.4 0.6 0.8 1.0 X Figure 13.12 The lattice spacing of the (112) planes of CuIn1−x Gax Se2 and the (111) cubic or the (002) hexagonal planes of Cd1−x Znx S. Empirical data from References  ((CdZn)S) and  (CuInSe2 , CuGaSe2 , and (Cu(InGa)Se2 ) are included and 0.328 nm, respectively. Figure 13.12 displays the (112) spacing for Cu(InGa)Se2 as a function of Ga/(In + Ga) ratio together with the (111)/(002) spacing of CdS–ZnS alloys. When Cu(InGa)Se2 ﬁlms are immersed in the chemical bath for deposition of CdS, they are also subjected to chemical etching of the surface. In particular, native oxides are removed by the ammonia . Thus, the CBD process cleans the Cu(InGa)Se2 surface and enables the epitaxial growth of the CdS buffer layer. In early single-crystal work, p –n homojunction diodes were fabricated by indif- fusion of Cd or Zn into p-type CuInSe2 [131, 132] at 200 to 450◦ C. Investigations of CuInSe2 /CdS interfaces did show interdiffusion of S and Se above 150◦ C and rapid Cd diffusion into CuInSe2 above 350◦ C . More recently, intermixing of the constituents of the Cu(InGa)Se2 /CdS heterojunction has been observed even when the relatively low- temperature CBD process is used for growth of the CdS layer . Investigations of the effect of a chemical bath without the thiourea showed an accumulation of Cd on the Cu(InGa)Se2 surface, possibly as CdSe . Accumulation of Cd on the Cu(InGa)Se2 surface was also observed in the initial stage of CdS growth in the complete chemi- cal bath . The results were not conclusive on whether any interfacial compound is formed, but TEM investigations showed the presence of Cd up to 10 nm into the Cu- deﬁcient surface region of the Cu(InGa)Se2 layer . At the same time, a reduction of the Cu concentration was noted. An interpretation in which Cu+ is replaced with Cd2+ is proposed, on the basis of the very close ion radii of these ions, 0.96 and 0.97, respectively. XPS and secondary ion mass spectrometry (SIMS) proﬁles of Cu(InGa)Se2 ﬁlms and CuInSe2 single crystals exposed to chemical baths without thiourea also show evidence of indiffusion or electromigration of Cd . 13.4.3 Other Deposition Methods In the early days of Cu(InGa)Se2 research, vacuum evaporation of 2 to 3-µm-thick CdS was the standard method to fabricate the junction and 9.4% efﬁciency was obtained with 588 Cu(InGa)Se2 SOLAR CELLS pure CuInSe2 absorbers . With evaporation it is difﬁcult to nucleate and grow very thin continuous CdS layers such as those normally used in current state-of-the-art Cu(InGa)Se2 devices, and the optical transmission of the window will be limited to energies less than the CdS band gap, 2.4 eV. Substrate temperatures of 150 to 200◦ C are used to obtain good optical and electrical properties of the evaporated CdS ﬁlms. This is substantially higher than the substrate temperature used for chemical bath deposition. Improved device performance was achieved by alloying the evaporated CdS with ZnS . Mixed (CdZn)S has a wider band gap, allowing increased optical transmission, and better lattice match to Cu(InGa)Se2 than CdS. The main drawback with vacuum evaporation is poor conformal coating resulting in nonuniform and incomplete coverage of the sometimes relatively rough Cu(InGa)Se2 ﬁlms. Sputter deposition leads to more conformal coverage. The general success of sput- tering for industrial large-area deposition motivated the exploration of sputter-deposited CdS buffer layers. Using optical emission spectroscopy to control the sputtering pro- cess, Cu(InGa)Se2 devices with efﬁciencies up to 12.1% were fabricated, as compared to 12.9% for reference cells with chemical bath–deposited CdS . Both evapora- tion and sputtering are vacuum processes, which can be incorporated in-line with other vacuum processing steps and do not create any liquid wastes. Still, CBD remains the preferred process for the CdS layer owing to its advantages in forming thin conformal coatings. Atomic layer chemical vapor deposition (ALCVD) is a method that also allows accurate control of the growth of thin conformal layers . The method is being indus- trially used for deposition of another II-VI compound, ZnS. Inorganic precursors for deposition of CdS require the substrate temperature to be excessively high (>300◦ C) and work with organic precursors has been limited. The strong driving force for replacement of the environmentally nondesirable cadmium has focused the development of ALCVD on materials other than CdS. This is also valid for regular CVD, although some metal organic CVD (MOCVD) work has been reported. The full potential for chemical vapor–deposited CdS has therefore not been explored. Electrodeposition can be used to deposit CdS ﬁlms but its use has not been reported in Cu(InGa)Se2 devices. 13.4.4 Alternative Buffer Layers The cadmium content in Cu(InGa)Se2 PV modules with CBD CdS buffer layers is low. Investigations show that the cadmium in Cu(InGa)Se2 modules can be handled safely, both with respect to environmental concerns and hazards during manufacturing (see Section 13.6.5). In spite of this, it would be preferable to eliminate cadmium in new products. There are in principle two approaches to Cd-free devices: (1) ﬁnding a buffer material that replaces CdS and (2) omitting the CdS layer and depositing ZnO directly onto the Cu(InGa)Se2 ﬁlm. In practice, the two approaches tend to merge when the chem- ical bath deposition of CdS is replaced with a surface treatment of the Cu(InGa)Se2 with no or negligible ﬁlm deposition before the subsequent deposition of the ZnO. A number of approaches and materials have been tried. A selection of promising results are presented in Table 13.3. JUNCTION AND DEVICE FORMATION 589 Table 13.3 Performance of Cu(InGa)Se2 thin-ﬁlm solar cells with various buffer layers and junc- tion-formation methods alternative to chemical bath deposition of CdS Buffer Deposition method Efﬁciency VOC JSC FF Reference material [%] [mV] [mA/cm2 ] [%] None 10.5a 398 39.0 68  None 15.0b 604 36.2 69  ZnO MOCVD 13.9a 581 34.5 69  ZnO ALCVD 11.7 512 32.6 70  Zn treatment ZnCl2 solution 14.2b 558 36.3 70  Zn(O,S,OH)x Chemical bath 14.2c,d 567e 36.6e 68  ZnS Chemical bath 16.9a,b 647 35.2 74  Zn treatment + ZnS Chemical bath + ILGARf 14.2 559 35.9 71  Zn(Se,OH) Chemical bath 13.7b,d 535 36.1 71  ZnSe ALCVD 11.6a 502 35.2 65  ZnSe MOCVD 11.6 469 35.8 69  Inx Sey Coevaporation 13.0a 595 30.4 72  ZnInx Sey Coevaporation 15.1 652 30.4 76  Inx (OH,S)y Chemical bath 15.7a,b 594 35.5 75  In2 S3 ALCVD 13.5 604 30.6 73  a Active area b With antireﬂection layer c Minimodule d Conﬁrmed e Recalculated to single-cell values f Ion Layer Gas Reaction When the numbers in Table 13.3 are analyzed, one must keep in mind that the qual- ity of the Cu(InGa)Se2 layer varies signiﬁcantly between the experiments. For example, in the early results with direct ZnO , the reference cells with chemical bath–deposited CdS showed 12.4% efﬁciency, whereas the 15% efﬁciency results  are obtained from Cu(InGa)Se2 , which at best yielded an efﬁciency of 18.8%. On the other hand, an inferior junction-formation method may cause a larger degradation of cell efﬁciency at higher efﬁ- ciency levels, since its defects may be relatively more important to the cell performance. In order to evaluate the various Cd-free junction-formation methods from that respect, the efﬁciency from each experiment is displayed in Figure 13.13 together with its reference, or estimation thereof. In most cases, the Cd-free device is comparable to the CBD–CdS device within typical variations. Altogether, it appears as if there are several possibilities for obtaining high efﬁ- ciency without Cd. All the listed methods include one or more of the elements Zn, In, and S. Zn is directly included in most of the buffer materials or indirectly as ZnO transparent contact with Inx Sey , In(OH,S)x , and In2 Se3 . Indications that n-type doping with Zn occurs similarly to that with Cd have been found by the treatment of Cu(InGa)Se2 in Cd and Zn solutions , and are consistent with junction formation by solid-state diffusion into single crystals . In Figure 13.13 a slight tendency can be noted toward larger difference between Cd-free and CdS reference cells for the direct ZnO approaches. It appears as if a buffer layer between the Cu(InGa)Se2 and the ZnO is beneﬁcial. Such a layer could passivate the 590 Cu(InGa)Se2 SOLAR CELLS 20 Cd-free CBD-CdS 16 Efficiency [%] 12 8 4 InSex None In2Se3 CBD-ZnS In(OH,S) None Zn(O,S,OH) Zn(Se,OH) MOCVD-ZnSe ILGAR-ZnS ALCVD-ZnSe ZnInxSey MOCVD-ZnO Zn-treatment ALCVD-ZnO Figure 13.13 The efﬁciency of Cu(InGa)Se2 solar cells with a selection of Cd-free junction- formation methods together with corresponding values of Cu(InGa)Se2 cells with chemical bath–deposited CdS Cu(InGa)Se2 surface, which would reduce the recombination in a shallow n-type emitter, and possibly also serve to protect the junction and near-surface region during subsequent deposition of the transparent contact materials. 13.4.5 Transparent Contacts The early Cu(InGa)Se2 devices used CdS doped with In or Ga as front-contact layers in addition to the CdS buffer layer. Short wavelength light (<520 nm) was absorbed near the surface in the thick CdS layer and did not generate any photocurrent. When chemical bath deposition allowed CdS buffer layers to be thin enough such that it no longer limited the short wavelength collection in the Cu(InGa)Se2 , photocurrent could be gained by increasing the band gap of the contact layer. Since the contact layer must also have high conductivity for lateral current collection, the obvious choice is a transparent conducting oxide (TCO), a class of materials used in such devices as displays and low-emission coatings on window glass panes. There are three main materials in this class: SnO2 , In2 O3 :Sn (ITO), and ZnO. SnO2 requires relatively high deposition temperatures that restrict the potential in Cu(InGa)Se2 devices that cannot withstand temperatures greater than 200 to 250◦ C after CdS is deposited. ITO and ZnO can both be used, but the most common material is ZnO, favored by potentially lower material costs. A good overview of TCO thin-ﬁlm materials can be found in Reference . The most commonly used low-temperature deposition method for TCO ﬁlms is sputtering. ITO layers are routinely fabricated on an industrial scale using dc sputtering. Industrial practice is to use ceramic ITO targets and to sputter in an Ar:O2 mixture. Typical sputter rates range between 0.1 to 10 nm/s, depending on the application . Sputtering of doped ZnO ﬁlms is not as developed as is sputtering of ITO. Neverthe- less, it is the preferred method for depositing the transparent front contact on Cu(InGa)Se2 JUNCTION AND DEVICE FORMATION 591 devices, with and without CBD–CdS, in the majority of the R&D groups. Typically, ZnO:Al ﬁlms are deposited by radio frequency (rf) magnetron sputtering from ceramic ZnO:Al2 O3 targets with 1 or 2 weight% Al2 O3 . In large-scale manufacturing, dc sputter- ing from ceramic targets is favored since it requires simpler equipment and offers higher deposition rates . Reactive dc sputtering from Al/Zn alloy targets has also been used in the fab- rication of Cu(InGa)Se2 /CdS devices with the same performance as with rf sputtered ZnO:Al . The use of Zn/Al alloy targets allows lower costs than ceramic ZnO:Al2 O3 targets, but reactive sputtering requires very precise process control owing to the so-called hysteresis effect  so that optimal optoelectronic properties are achieved only within a very narrow process window. Deposition rates in 4 to 5 nm/s range have been achieved. Chemical vapor deposition (CVD) provides another deposition option and is used by one commercial manufacturer of Cu(InGa)Se2 modules to deposit ZnO . The reaction occurs at atmospheric pressure between water vapor and diethylzinc and the ﬁlms are doped with ﬂuorine or boron. ALCVD deposition of ZnO has also been tested . The atomic layer by atomic layer growth gives very low deposition rates, but the surface-controlled growth process gives uniform layers within a wide process window concerning reactant ﬂow. This allows large batches to be processed, resulting in a reasonable throughput in spite of the limited growth rate. As with the Mo back contact, the requirements for sheet resistance of the trans- parent contact layer will depend on the speciﬁc cell or module design. Typically, small area cells use layers with 20–30 / , while modules may require 5–10 / . In either case, the sheet resistance is usually controlled by the layer thickness. 13.4.6 Buffer Layers It is common practice to use a buffer layer of undoped high-resistivity (HR) ZnO before sputter deposition of the TCO layer. Depending on the deposition method and conditions, this layer may have a resistivity of 1–100 cm compared to the transparent contact with 10−4 –10−3 cm. Typically, 50 nm of HR ZnO is deposited by rf magnetron sputtering from an oxide target. The gain in performance by using an HR ZnO buffer layer in ordinary devices with CBD–CdS is related to the CdS thickness [156, 160, 161]. One explanation of the role of a ZnO buffer layer is given by  as resulting from locally nonuniform electronic quality of the Cu(InGa)Se2 layer that can be modeled by a parallel diode with high recombination current. The inﬂuence of these regions on the overall performance is reduced by the series resistance of the HR ZnO layer. This series resistance has a negligible effect on the performance of the dominant parts of the device area. A related explanation would attribute the local areas with poor diode characteristics caused by pinholes in the CdS layer, which create parallel diodes with a Cu(InGa)Se2 /ZnO junction. In this case improved diode quality due to the ZnO buffer would improve overall performance. Either case is consistent with the observation that a beneﬁcial effect from the ZnO buffer is not observed when the CBD–CdS layer is thick enough . 592 Cu(InGa)Se2 SOLAR CELLS Another potential reason for using an HR ZnO buffer layer is to add protection of the interface region from sputter damage induced during deposition of the TCO layer which typically requires more harsh conditions. This seems to be particularly important for some alternative Cd-free buffer layers or with dc magnetron–sputtered TCO layers . 13.4.7 Device Completion In order to contact laboratory test cells, a metal contact is deposited onto the TCO layer. It is shaped as a grid with minimum shadow area in order to allow as much light as possible into the device. Solar cell measurement standards recommend a minimum cell area of 1 cm2 , but many labs routinely use cells in the order of 0.5 cm2 . The metal grid contact can be made by ﬁrst depositing some tens of nanometers of Ni to prevent the formation of a high resistance oxide layer, and subsequently depositing a few micrometers of Al. Evaporation through an aperture mask is a suitable deposition method. After deposition of the metal grid, the total cell area is deﬁned by removing the layers on top of the Mo outside the cell area by mechanical scribing or laser patterning. Alternatively, just the layers on top of the Cu(InGa)Se2 can be removed, by photolitho- graphy and etching, since the lateral resistance of the Cu(InGa)Se2 prevents collection outside the cell area. The only signiﬁcant difference in the device layers between lab cells and modules is the thickness of the TCO. Modules normally do not have any grid that assists in current collection over the cell area, so a substantially thicker TCO layer, that is, higher sheet conductivity, is needed in order to keep resistive losses low. A TCO layer with higher sheet conductivity may also have lower optical transmission in the infrared due to increased free-carrier absorption resulting in a decreased photocurrent. 13.5 DEVICE OPERATION Cu(InGa)Se2 solar cells have achieved efﬁciencies approaching 20%, the highest of any thin-ﬁlm solar cells, largely by empirical processing improvements and in spite of rela- tively poor understanding of the underlying mechanisms and electronic defects that control the device behavior. However, a more complete picture of the device operation is emerg- ing to enable both a better understanding of the devices and identiﬁcation of pathways to further improvements. The operation of Cu(InGa)Se2 /CdS solar cells is characterized by high quantum efﬁciency (QE ) and short-circuit current. The open-circuit voltage increases with the band gap of the absorber layer and is insensitive to grain boundaries and defects at the Cu(InGa)Se2 /CdS interface. A basic device model can be constructed in which the voltage is limited by recombination through bulk trap states in the space charge region of the Cu(InGa)Se2 absorber layer. Recombination at the Cu(InGa)Se2 /CdS interface is min- imized by proper doping and band alignment or surface treatment to create an effective n-type inversion layer in the near-junction region of the absorber layer. The device operation can be described by identifying loss mechanisms. These can be divided into three categories. The ﬁrst are optical losses that limit generation of carriers and therefore the device current. The second are recombination losses that limit DEVICE OPERATION 593 the voltage. Finally, there are parasitic losses, such as series resistance, shunt conductance, and voltage-dependent current collection, which are most evident by their effect on the ﬁll factor but can also reduce JSC and VOC . 13.5.1 Light-generated Current The highest efﬁciency Cu(InGa)Se2 device has JSC = 35.2 mA/cm2  out of a possible 42.8 mA/cm2 available for a band gap of 1.12 eV under AM1.5 global illumination. Quan- tum efﬁciency is a valuable tool to characterize the losses responsible for this difference in current. The light-generated current is the integral of the product of the external quantum efﬁciency (QE ext ) and the illumination spectrum. QE ext is controlled by the band gap of the Cu(InGa)Se2 absorber layer, the CdS and ZnO window layers, and a series of loss mechanisms. These losses are illustrated in Figure 13.14 where typical QE curves at two different voltage biases, 0 V and −1 V, are shown. The QE curve at −1 V is slightly higher at longer wavelengths. The current loss under 100 mW/cm2 illumination is listed in Table 13.4 for each of these mechanisms. Losses 1 to 5 are optical and 6 is electronic. In practice, the magnitude of each of these losses will depend on the details of the device design and optical properties of the speciﬁc layers. The losses include the following: 1. Shading from a collection grid used for most devices. In an interconnected module this will be replaced by the area used for the interconnect, as discussed in Section 13.6.2. 2. Front surface reﬂection. On the highest-efﬁciency devices this is minimized with an antireﬂection layer for which an evaporated MgF2 layer with thickness ∼100 nm is commonly used. However, this is not practical in a module in which a cover glass is typically required. Energy [eV] 3.0 2.0 1.5 1.0 1.0 (1) 0.8 (2) (3) (6) (3) Quantum efficiency (4) 0.6 (5) 0.4 0.2 0.0 400 600 800 1000 1200 Wavelength [nm] Figure 13.14 Quantum efﬁciency (solid lines) at 0 V and −1 V and optical losses for a Cu(InGa)Se2 /CdS solar cell in which the Cu(InGa)Se2 has Eg = 1.12 eV 594 Cu(InGa)Se2 SOLAR CELLS Table 13.4 Current loss, J , for E > 1.12 eV due to the optical and collection losses illustrated in Figure 13.14 for a typical Cu(InGa)Se2 /CdS solar cell Region in Optical loss mechanism J Figure 13.14 [mA/cm2 ] (1) Shading from grid with 4% area coverage 1.7 (2) Reﬂection from Cu(InGa)Se2 /CdS/ZnO 3.8 (3) Absorption in ZnO 1.8 (4) Absorption in CdS 0.8 (5) Incomplete generation in Cu(InGa)Se2 1.9 (6) Incomplete collection in Cu(InGa)Se2 0.4 3. Absorption in the TCO layer. Typically, there is 1 to 3% absorption through the visible wavelengths, which increases in the near IR region, λ > 900 nm, where free-carrier absorption becomes signiﬁcant, and for λ < 400 nm near the ZnO band gap. 4. Absorption in the CdS layer. This becomes appreciable at wavelengths below ∼520 nm corresponding to the CdS band gap 2.42 eV. The loss in QE for λ < 500 nm is proportional to the CdS thickness since it is commonly assumed that electron–hole pairs generated in the CdS are not collected. Figure 13.14 shows a device with a ∼30 nm-thick CdS layer. In practice, the CdS layer is often thicker and the absorption loss greater. 5. Incomplete absorption in the Cu(InGa)Se2 layer near the Cu(InGa)Se2 band gap. Band gap gradients, resulting from composition gradients in many Cu(InGa)Se2 ﬁlms, also affect the steepness of the long-wavelength part of the QE curve. If the Cu(InGa)Se2 is made thinner than ∼1.0 µm, this loss becomes signiﬁcant  because of insufﬁcient absorption at long wavelengths. 6. Incomplete collection of photogenerated carriers in the Cu(InGa)Se2, discussed below. QE ext is then given by QE ext (λ, V ) = [1 − R(λ)][1 − AZnO (λ)][1 − ACdS (λ)] QE int (λ, V ) (13.5) where R is the total reﬂection, including the grid shading, AZnO is the absorption in the ZnO layer and ACdS is the absorption in the CdS layer. QE int , the internal quantum efﬁciency, is the ratio of photogenerated carriers collected to the photon ﬂux that arrives at the absorber layer and can be approximated by  exp[−α(λ)W (V )] QE int (λ, V ) ∼ 1 − = (13.6) αL + 1 where α is the Cu(InGa)Se2 absorption coefﬁcient, W is the space charge width in the Cu(InGa)Se2 , and L is the minority carrier diffusion length. This approximation assumes that all carriers generated in the space charge region are collected without recombination loss. Since W is a function of the applied voltage bias, QE int and total light-generated current are, in general, voltage-dependent, so the latter can be written as JL (V ). Values of W in the range 0.1–0.5 µm have been reported for typical cells at 0 V. DEVICE OPERATION 595 1.0 500 0.8 800 1000 Absorption 0.6 1100 nm 0.4 0.2 0.0 0.0 0.5 1.0 1.5 Depth [µm] Figure 13.15 Absorption of light with different wavelengths in Cu(InGa)Se2 with x = 0.2 The absorption of light with different wavelengths in Cu(InGa)Se2 with x = 0.2 is shown in Figure 13.15. At thickness d, this is given by exp(−αd) with α calculated at each wavelength using equation (13.3) and the data in Figure 13.6. If the effective collection length L + W is smaller than 0.5 to 1 µm, a signiﬁcant fraction of electrons are generated deeper into the Cu(InGa)Se2 layer, and their incomplete collection can be a signiﬁcant loss mechanism for Cu(InGa)Se2 devices [116, 165]. The effect of JL (V ) on current–voltage behavior increases with forward voltage bias and therefore has its largest effect on the ﬁll factor and VOC [166, 167]. The effect of a voltage-dependent collection on JSC is illustrated in Figure 13.14 by the increase in QE measured at −1 V applied voltage bias compared to that measured at 0 V. 13.5.2 Recombination The current–voltage (J –V ) behavior of Cu(InGa)Se2 /CdS devices can be described by a general diode equation: q J = JD − JL = JO exp (V − RS J ) + GV − JL (13.7) AkT with the diode current JO given by: b JO = JOO exp − (13.8) AkT The ideality factor A, barrier height b , and prefactor JOO depend on the speciﬁc recombi- nation mechanism that dominates JO , while the series resistance RS and shunt conductance G are losses that occur in series or parallel with the primary diode. General expres- sions for A, b , and JOO in the cases of recombination through the interface, space charge region, or bulk of the absorber layer can be found in various textbooks (see, for example ). 596 Cu(InGa)Se2 SOLAR CELLS x Eg [eV] 20 a 0 1.02 b 0.24 1.16 c 0.61 1.40 0 [mA/cm2] Current a b c −20 −40 −0.3 0.0 0.3 0.6 0.9 Voltage [V] Figure 13.16 Current–voltage curves for Cu(InGa)Se2 /CdS solar cells with different relative Ga content giving (a) Eg = 1.02, (b) 1.16, and (c) 1.4 eV Energy [eV] 3.0 2.0 1.5 1.0 1.0 0.8 Quantum efficiency 0.6 c b a 0.4 0.2 0.0 400 600 800 1000 1200 1400 Wavelength [nm] Figure 13.17 Quantum efﬁciency curves for the devices shown in Figure 13.16 To understand the speciﬁc diode behavior of Cu(InGa)Se2 /CdS solar cells, it is instructive to look at the effect of the Cu(InGa)Se2 band gap, varied by changing x ≡ Ga/(In + Ga), and temperature. Figures 13.16 and 13.17 show J –V and QE curves for 3 devices with x = 0, 0.24, and 0.61, corresponding to Eg = 1.02, 1.16, and 1.40 eV, respectively. VOC increases and the position of the long-wavelength QE edge shifts to greater energy as Eg increases. Figure 13.18 shows the temperature dependence of VOC for these devices. In each case, as T → 0, VOC → Eg /q. Thus, combining equations (13.7) DEVICE OPERATION 597 1.4 c 1.2 b 1.0 [Volts] VOC 0.8 a 0.6 0.4 0 100 200 300 T [K] Figure 13.18 Temperature dependence of VOC for the devices shown in Figure 13.16 and (13.8) and assuming G JL /VOC , the open-circuit voltage becomes Eg AkT JOO VOC = − ln (13.9) q q JL with the barrier height b = Eg . The different recombination paths are effectively connected in parallel so that VOC will be controlled by the single dominant mechanism with the highest current. The values of b and A can be used to distinguish between recombination in the bulk absorber, in the space charge region of the Cu(InGa)Se2 , or at the Cu(InGa)Se2 /CdS interface [27, 169]. Each of the curves in Figure 13.16 can be ﬁt to equation (13.7) with A = 1.5 ± 0.3. For a wide range of thin-ﬁlm solar cells, it has been demonstrated that VOC (T → 0) = QE g and 1 < A < 2 similar to the data above. Speciﬁcally, this has been shown for CuInSe2 [116, 170] and Cu(InGa)(SeS)2  devices, independent of the (CdZn)S buffer-layer band gap , and for a variety of different absorber-layer depo- sition processes . These results for b and A indicate that Cu(InGa)Se2 /CdS solar cells operate with the diode current controlled by Shockley–Read–Hall type recombi- nation in the Cu(InGa)Se2 layer . This recombination is greatest through deep trap states in the space charge region of the Cu(InGa)Se2 where there are comparable sup- plies of electrons and holes available, that is, p ≈ n. The variation in A between 1 and 2 depends on the energies of the deep defects that act as dominant trap states . As these states move toward the band edges, A → 1 and the recombination becomes closer to band-to-band bulk recombination. These observations exclude recombination in the neutral bulk region of the absorber layer, which should give A = 1. Interface recombination would give b < Eg [Cu(InGa)Se2 ] with a dependence on the (CdZn)S band gap, although A might vary from 1 to >2 . Back surface recombination at the Mo/Cu(InGa)Se2 interface will 598 Cu(InGa)Se2 SOLAR CELLS be negligible so long as the minority-carrier diffusion length is small compared to the total Cu(InGa)Se2 thickness. If L + W ≈ d, a back surface ﬁeld may be implemented, for example, by increasing the Ga content near the Mo to give a band gap gradient. In real Cu(InGa)Se2 materials with imperfect structures, trap defects will not exist at discrete energies but form defect bands or tails at the valence and conduction bands. Then the total recombination current can be determined by integrating over the defect spec- trum. Recombination through an exponential bandtail was used to explain the temperature dependence in A observed in some devices . Analysis of the temperature dependence of A was further explained by a tunneling enhancement of the recombination current, particularly at reduced temperatures . The same defects in the Cu(InGa)Se2 space charge region that control recombination were also used to explain observed metastabilities including persistent photoconductivity and open-circuit voltage decay . Admittance spectroscopy has proved to be a useful tool to characterize the distribution of electronic defects in Cu(InGa)Se2 /CdS solar cells  and the density of an acceptor state ∼0.3 eV from the valence band has been correlated to VOC . The minority-carrier lifetime is another valuable parameter to characterize Cu(InGa)Se2 /CdS devices. Transient photocur- rent  and time-resolved photoluminescence  measurements each were used to calculate lifetimes in the range of 10 to 100 ns for high-efﬁciency devices. Still, a critical problem that remains is to identify which of the calculated or measured defects discussed in Section 13.2 provides for the recombination traps that limit voltage in the devices. A good review of the characterization of electronic defects and their effect on Cu(InGa)Se2 devices is provided by Rau and Schock . In practice, analysis of J –V data is commonly used to determine the diode param- eters JO , A, and b . This requires that RS and G are negligible, or suitable corrections are made to the data, and that JL is independent of V . Failure to account for JL (V ) can lead to errors in analysis of current–voltage data  and in many cases the fundamental diode parameters cannot be reliably determined except from J –V data measured in the dark. In addition, it must be veriﬁed that there are no nonohmic effects at any contacts or junc- tions, which cause the appearance of a second diode for which equation (13.7) does not account. Such nonohmic behavior is often observed at reduced temperatures [170, 172]. Once it has been demonstrated that all these parasitic effects are negligible, or corrections have been made, then JO can be determined by a linear ﬁt to a semilogarithmic plot of J + JL versus V –RS J and A can be determined from the slope of the derivative dV /dJ versus 1/J in forward bias , or both JO and A can be obtained by a least squares ﬁt to equation (13.7). Finally, b can be determined from the temperature dependence of VOC as in Figure 13.18. It must be noted that most descriptions of transport and recombination ignore the effect of grain boundaries, implicitly assuming that grains are columnar and all transport can proceed without crossing grain boundaries. However, this is rarely, if ever, strictly true, so a comprehensive description of Cu(InGa)Se2 solar cells must account for the possibility of recombination at grain boundaries reducing current collection or voltage. The effect of grain boundaries can be expressed as an effective diffusion length, leading to the conclusion that grain-boundary recombination is small . This can occur if the grain boundaries are doped more p-type than the bulk grains so that electrons are prevented from reaching and recombining at defects in the grain boundaries . DEVICE OPERATION 599 13.5.3 The Cu(InGa)Se2 /CdS Interface It may seem surprising that recombination at the Cu(InGa)Se2 /CdS interface does not limit VOC since, in processing Cu(InGa)Se2 solar cells, no special efforts are made to match lattices or reduce interface defects and the devices are typically exposed to air between the Cu(InGa)Se2 and CdS depositions. This can be explained by type inversion of the near-junction region of the Cu(InGa)Se2 induced by the band alignment and dop- ing [169, 182–184]. In this case, the Fermi level at the interface is close to the conduction band so that electrons in the near surface region of the Cu(InGa)Se2 are effectively major- ity carriers and there is an insufﬁcient supply of holes available for recombination through the interface states. It has alternatively been proposed that doping due to Cd diffusion during the chemical bath deposition of CdS results in the formation of an n-type emitter and a p –n homojunction in the Cu(InGa)Se2 . This would require the junction to remain very close to the Cu(InGa)Se2 /CdS interface to minimize recombination of carriers generated near the interface, and would therefore be very process-speciﬁc. The Cu(InGa)Se2 /CdS band diagram shown in Figure 13.19 demonstrates that the conduction-band offset EC between the CdS and the Cu(InGa)Se2 is critical for creat- ing the type inversion in the Cu(InGa)Se2 . In this diagram, the bulk Cu(InGa)Se2 layer is p-type with Eg depending on the relative Ga concentration, the CdS layer is n-type with Eg = 2.4 eV and is totally depleted, and the bulk ZnO n+ -layer has Eg = 3.2 eV. A thin HR ZnO layer between the n+ -ZnO layer and the CdS is also assumed to be depleted. Pos- itive EC indicates a spike in the conduction band, that is, the conduction-band minimum in the CdS is at higher energy than the conduction-band minimum of the Cu(InGa)Se2 . Figure 13.19 shows the case with EC = 0.3 eV and a −0.3 eV conduction-band offset between the ZnO and the CdS . Models of current transport and recombination have considered the effect of EC [184–187]. These models show that if EC is greater than about 0.5 eV, collection of photogenerated electrons in the Cu(InGa)Se2 is impeded and ZnO CdS Cu(InGa)Se2 Eg = 3.2 eV Eg = 2.4 eV Eg = 1.2 eV EC EF EV ∆EC JREC (p = n) Figure 13.19 Band diagram of a ZnO/CdS/Cu(InGa)Se2 device at 0 V in the dark. Note that the recombination current JREC is greatest where p = n in the space charge region of the Cu(InGa)Se2 and not at the interface 600 Cu(InGa)Se2 SOLAR CELLS JSC or FF is reduced sharply. With a smaller spike, electrons can be transported across the interface assisted by thermionic emission . On the other hand, for sufﬁciently negative EC the induced type inversion of the Cu(InGa)Se2 near the interface is elimi- nated and interface state recombination will limit VOC . An ODC layer at the surface of the absorber layer increases the band gap and primarily affects the valence band , so it may enhance type inversion near the junction. However, there is no convincing evidence that this layer exists in devices, so it is not shown in Figure 13.19. Owing to its importance in the electronic behavior of Cu(InGa)Se2 /CdS devices, several efforts have been made to calculate or measure EC with varying results. Band-structure calculations gave EC = 0.3 eV . XPS and ultraviolet photoelectron spectroscopy (UPS) measurements of the valence band alignment indicate a positive EC between 0.2 and 0.7 eV [52, 190, 191]. These electron spectroscopy methods require ultrahigh vacuum conditions that necessitates that the CdS is deposited by vacuum evaporation. It is possible that the interface formation is different when CdS is grown by chemical bath deposition, for example, due to chemical interdiffusion, resulting in a different alignment of the conduction bands. Indirect measurements of the junction formed with chemical bath deposited CdS using a surface photovoltage technique gave EC = −0.1 eV . Finally, inverse photoemission spectroscopy showed that substantial chemical intermixing occurs across the interface resulting in EC = 0 . 13.5.4 Wide and Graded Band Gap Devices While the highest efﬁciency devices generally have Ga/(In + Ga) ≈ 0.1–0.3 giving Eg ≈ 1.1–1.2 eV, signiﬁcant effort has been made to develop high-efﬁciency solar cells based on wider band gap alloys. This is driven primarily by the expectation that wider band gap alloys will yield higher module efﬁciencies due to reduced losses related to the trade-off between higher voltage and lower current at maximum power. The resulting reduction in power loss, proportional to I 2 R, can be used to either (1) increase the module’s active area by reducing the spacing between interconnects or (2) decrease the optical absorption in the TCO layers since they can tolerate greater resistance. Wider band gap should give a lower coefﬁcient of temperature for the device or module output power, which will improve performance at the elevated temperatures experienced in most real terrestrial applications. Wide band gap devices could also be used as the top cell in a tandem or multijunction cell structure. The wider band gap materials that have attracted the most attention for devices are Cu(InGa)Se2 and CuInS2 . CuGaSe2 has Eg = 1.68 eV, which is well suited for the wide band gap cell in tandem structures. CuInS2 has Eg = 1.53 eV, which could be nearly optimum for a single-junction device. The highest-efﬁciency devices based on CuInS2 are deposited with Cu-rich overall composition and then the excess Cu, in the form of a Cux S second phase, is etched away before CdS deposition . Cu(InAl)Se2 solar cells have also been considered . Since CuAlSe2 has Eg = 2.7 eV, the alloy requires smaller changes in relative alloy concentration and lattice parameter from CuInSe2 than the Ga alloys to achieve comparable band gap. The highest efﬁciency devices of different alloys are listed in Table 13.5. The effects of Ga incorporation on device behavior are not fully understood. The addition of a small amount of Ga to CuInSe2 increased the open-circuit voltage even when DEVICE OPERATION 601 Table 13.5 Highest-efﬁciency devices for different alloy absorber layers Material Eg Efﬁciency VOC JSC FF Reference [eV] [%] [%] [mA/cm2 ] [%] CuInSe2 1.02 15.4 515 41.2 72.6  Cu(InGa)Se2 1.12 18.8 678 35.2 78.6  CuGaSe2 1.68 8.3 861 14.2 67.9  CuInS2 1.53 11.4 729 21.8 71.7  Cu(InAl)Se2 1.16 16.9 621 36.0 75.5  0.9 16 14 0.8 Efficiency VOC [V] [%] 12 0.7 10 0.6 8 1.1 1.2 1.3 1.4 1.5 1.6 Eg [eV] Figure 13.20 Efﬁciency ( ) and VOC (ž) as a function of Cu(InGa)Se2 band gap, varied by increasing the relative Ga content, (From Shafarman W, Klenk R, McCandless B, Proc. 25th IEEE Photovoltaic Specialist Conf., 763–768 (1996) . The dashed line has slope VOC / Eg = 1 the Ga was conﬁned to the back of the absorber and did not increase the band gap in the space charge region . The effect of increasing band gap in Cu(InGa)Se2 /CdS solar cells on VOC and efﬁciency is shown in Figure 13.20. Efﬁciency is roughly independent of band gap for Eg < 1.3 eV or Ga/(In + Ga) < 0.5 [165, 199]. With even wider band gap, VOC increases to greater than 0.8 V, but the efﬁciency decreases. This indicates poorer electronic properties of the Cu(InGa)Se2 absorber layer, which has two effects: voltage- dependent current collection , which causes the ﬁll factor to decrease, and increased recombination , which reduces VOC below that expected from equation (13.9) . The dashed line in Figure 13.20 shows a line with slope VOC / Eg = 1. Ideally, the increase in VOC would have only a slightly smaller slope due to the dependence on JL in the second term of equation (13.9). Admittance spectroscopy showed a correlation between the recombination and the density of a defect with an activation energy ∼0.3 eV, which increases with Eg . Transient photocapacitance measurements showed a defect band centered at 0.8 eV from the valence band, which moves closer to midgap for increas- ing band gap and therefore becomes more efﬁcient as a recombination trap . As the band gap becomes wider, type inversion of the absorber layer near the interface may no longer occur and interface recombination can become more signiﬁcant. Analysis of both CuGaSe2  and CuInS2  solar cells showed that the low open-circuit voltages 602 Cu(InGa)Se2 SOLAR CELLS could be caused by either space charge or interface recombination, depending on the device preparation. Band gap gradients formed by controlled incorporation of Ga or S have been proposed as a means to increase device efﬁciency by separately reducing recombination and collection losses [19, 204–206]. A gradient in the conduction band from wide at the Cu(InGa)Se2 /Mo interface to narrow near the space charge region has been used to enhance minority-carrier collection [199, 206] and to reduce back surface recombination when the diffusion length is comparable to the ﬁlm thickness . Alternatively, a gradient from wide at the Cu(InGa)Se2 /CdS interface to narrow at the edge of the space charge region could reduce recombination and increase VOC . In this case, the smaller band gap in the bulk portion of the device can still enable high optical absorption and JSC [19, 206]. The most effective implementation of a surface band gap gradient may be the incorporation of S near the front surface  since the main effect is in lowering the valence band, instead of raising the conduction band as with Ga, and there should be less impact on collection of light-generated electrons. 13.6 MANUFACTURING ISSUES The competitiveness of a PV technology will primarily be governed by its performance, stability, and costs. The best Cu(InGa)Se2 cells and modules have demonstrated efﬁciency on a par with commercial crystalline silicon products. Long-term stability appears not to be a signiﬁcant problem, as shown in ﬁeld tests of prototype modules, but low-cost production remains to be demonstrated in practice. It is evident that thin ﬁlms have the potential to be produced at very low costs. Moisture barriers of aluminum ﬁlms that are deposited on plastic foils to be used, for example, in potato chip bags cost less than 0.01 $/m2 to produce. This particular example is at the low end of production costs and more advanced functional coatings are in general substantially more expensive to manufacture. Thin ﬁlm coatings on architectural glass cost on the order of 1 $/m2 . Thus PV modules constructed from thin-ﬁlm materials have the possibility for very low manufacturing costs. Whether Cu(InGa)Se2 module production will be able to achieve this low-cost potential will depend on how well the process technology fulﬁlls the requirements for material costs, throughput, and yield. 13.6.1 Processes and Equipment Deposition processes can be either batch-type, in which a number of substrate plates are processed in parallel, or in-line, in which one substrate plate immediately follows the preceding one. In batch processing, a process step is completely ﬁnished before the next batch is started, whereas a substrate plate may enter an in-line process step before the previous substrate is ﬁnished in order to keep the line continuously running. One common view on volume production is that in-line continuous processing is a prerequisite for low costs. Fabrication of large-area thin-ﬁlm products with physical vapor deposition is often made in a continuous or quasi-continuous in-line system. However, the cost of a batch process can be equally low, provided the throughput is large enough. For manufacturing of Cu(InGa)Se2 modules, this means that the CdS chemical bath deposition MANUFACTURING ISSUES 603 can well fulﬁll low-cost production criteria even though it is normally a batch process. Similarly, growth of the Cu(InGa)Se2 layer by batch selenization does not necessarily need to be associated with higher costs than Cu(InGa)Se2 fabricated by in-line coevaporation, provided the cycle time is short enough or batch size large enough. The commercial availability of large-area deposition systems depends on the spe- ciﬁc process. Sputtering deposition is widely used for fabrication of large-area thin-ﬁlm coatings of various kinds, for example, in the glass industry. Similar processes are used in the fabrication of most Cu(InGa)Se2 modules for the Mo back contact and the TCO front contact, so the same type of equipment, available from a number of suppliers, can be used. Sputtering is also typically used for deposition of the metal precursor ﬁlms for fab- rication of the Cu(InGa)Se2 layer by a two-step process. The selenization step, however, requires speciﬁc custom-made process equipment. This could be selenization furnaces in which batches of plates with the precursor layers are exposed to a selenium-containing atmosphere or an in-line selenization chamber in which the plates are continuously transported through an environment with elemental selenium and substrate temperature control . Elemental coevaporation of the Cu(InGa)Se2 layer requires custom-made equip- ment including specially designed evaporation sources for uniform deposition of large-area substrates with accurate control. In-line evaporation using linear sources is a straightfor- ward approach that is being developed at several laboratories and companies. An example of such a piece of equipment is illustrated in Figure 13.21. The chemical bath deposition of CdS or Cd-free buffer layers is suitable for low- cost batch processing, in that it is a surface-controlled process that requires a limited solution volume. The equipment for dipping batches of Cu(InGa)Se2-coated substrate plates is relatively simple and can be custom-made. The dry buffer deposition methods under investigation have not been developed to a stage in which manufacturing is under consideration. Chemical vapor–deposited doped ZnO as an alternative to sputtering is typically done as a batch process with a relatively small number of substrate plates deposited per Raw Evaporation sources Coated plates plates Atmosphere Vacuum Vacuum Atmosphere Figure 13.21 In-line coevaporation system for Cu(InGa)Se2 with linear evaporation sources above u the substrate plates and heaters below them [Courtesy of Zentrum f¨ r Sonnenenergie- und Wasser- stoff-Forschung (ZSW)]. Reproduced by permission of Michael Powalla, ZSW Stuttgart, 2001 604 Cu(InGa)Se2 SOLAR CELLS run. Throughput will eventually become an issue. However, in-line CVD processes have been developed, for example, in the manufacture of amorphous silicon solar modules. 13.6.2 Module Fabrication Soda lime ﬂoat glass is the substrate material that so far has given the best results in terms of both performance and reproducibility. It fulﬁlls criteria on cost (3 $/m2 for 4-mm-thick glass in large volumes), smoothness, and stability, so it is well suited for commercial production. One limitation that needs to be addressed in the development of production processes is that soda lime glass starts to soften above 500◦ C. At the same time, the best PV properties of Cu(InGa)Se2 are achieved at growth temperatures above 500◦ C. Plastic deformation due to glass softening is not acceptable in a module-production process and careful optimization of the time–temperature proﬁle is needed to minimize the deformation. Flexible substrate materials are attractive both for the possibility to make a lightweight ﬂexible product with advantages for certain applications and for the possibility to deposit the thin-ﬁlm materials in roll-to-roll processes, which are potentially cost- advantageous. Such roll-to-roll processing of semiconductor thin ﬁlms was originally demonstrated with evaporation of CdS for solar cells . The substrate materials that have shown promising results are polyimide, titanium, and steel [68, 69]. The drawbacks of polyimide are low-temperature tolerance, since the best polyimide ﬁlms readily available can only withstand 400 to 450◦ C, and high thermal expansion. The main drawback of titanium and steel is their conductivity, which means that an electrically isolating layer is needed in order to allow monolithic series-interconnection of the cells. Such an isolation layer is not easy to make without local defects that will cause shunting of the cells. For these ﬂexible substrate materials, sodium has to be supplied separately. An essential cost advantage with thin-ﬁlm PV modules as compared to silicon wafer–based PV modules is the possibility of monolithic interconnection. This allows modules to be fabricated directly, instead of ﬁrst making cells followed by tabbing and stringing to make the series interconnection as required for silicon-wafer solar cells. A typical monolithic interconnection is illustrated schematically in Figure 13.22. The most common way to make the patterning is by using laser ablation for the Mo patterning (P1) and mechanical scribing for the two subsequent patterning steps (P2 and P3). The ﬁnal fabrication steps include attachment of electrical wires and buss bars. These are metal stripes that can be soldered, welded, or glued to contact areas near the edges of the substrate plates. Before lamination with a front cover glass, the thin-ﬁlm layers are removed from the outer rim of the substrate plate in order to improve the adhesion to the lamination material, which is usually ethylene vinyl acetate (EVA). Edge sealing and framing ﬁnishes the product, but can be omitted for some applications. 13.6.3 Module Performance The evolution of record efﬁciencies as reported from the certiﬁed measurement labs is displayed in Figure 13.23. The module efﬁciencies lag behind the cell efﬁciencies but follow the same basic trend. There are additional losses associated with making MANUFACTURING ISSUES 605 P1 P1 Mo Glass substrate P2 P2 HR ZnO CdS Cu(InGa)Se2 P3 P3 TCO Figure 13.22 Schematic description of the manufacturing steps to make monolithic interconnec- tions for thin-ﬁlm Cu(InGa)Se2 PV modules 19 NREL 18 Cells > 1 cm2 17 Mini-modules < 100 cm2 ÅSC NREL 16 Modules > 3000 cm2 Conversion efficiency 15 ÅSC Showa 14 Boeing NREL [%] IPE 13 Siemens Solarex Siemens 12 Siemens Siemens 11 Siemens 10 ARCO ÅSC - Ångstrom Solar Center, Uppsala University 9 IPE - Institut fur Physikalische Elektronik, Stuttgart University NREL - National Renewable Energy Laboratory, USA 8 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Year Figure 13.23 Evolution of Cu(InGa)Se2 device record efﬁciencies in the past decade. All data are taken from the Solar Cell Efﬁciency Tables periodically published in Progress in Photovoltaics 606 Cu(InGa)Se2 SOLAR CELLS series-interconnected modules instead of cells, both from series resistance and from inactive device area. In an optimized, conventional thin-ﬁlm module design, these kinds of losses correspond to about 1% unit of efﬁciency. With a more advanced design using metal grids for interconnection, interconnect losses can be made nearly negligible . Another kind of difference between modules and record cells is associated with the free- dom to use higher process temperatures for cells that are not sensitive to deformed glass. These results are not necessarily relevant to module fabrication but indicate the potential of the materials. In a product the initial efﬁciency is of little interest if it deteriorates after some time in operation. Cu(InGa)Se2 modules fabricated by ARCO Solar and later Siemens Solar have shown stable performance in ﬁeld tests over more than 12 years , as shown in Figure 13.24. On the other hand, severe degradation has been observed after exposure of cells to 85% relative humidity at 85◦ C for 1000 h , the so-called damp heat test, which is one of the certiﬁcation tests in the IEC 61 646 protocol. While this test is rather severe and may not be relevant to thin-ﬁlm modules, it shows the need for encapsulation techniques that minimize the exposure of the thin-ﬁlm materials to moisture. The outdoor module performance demonstrated in Figure 13.24 shows that Cu(InGa)Se2 PV modules have the stability and performance to compete in any power application, be it stand-alone or grid-connected. Thin-ﬁlm modules have a great advantage over silicon-wafer PV for consumer applications in which the power needed often is relatively small. The large substrate plates, which have a power of 40 WP or more, can easily be cut into smaller pieces, to essentially any power speciﬁcation. This is much less costly than making small crystalline-silicon modules in which each cell has to be cut into pieces before assembling the modules. Additionally, the patterning structure of the interconnects can be designed to ﬁt a large variety of shape and voltage requirements. Aesthetically, the solid black appearance of Cu(InGa)Se2 modules may be preferred to the nonuniform bluish appearance of the silicon-wafer modules in some building-integrated applications. 12 Modules tested outdoors at NREL 1988 module is 0.1 m2 others are 0.4 m2 Aperture efficiency 10 [%] 8 6 4 1989 1992 1995 1998 2001 Date Figure 13.24 Examples of outdoor testing results at NREL of Cu(InGa)Se2 modules showing stability over 12 years. Fluctuations in years 1992–1996 are due to changes in testing conditions. (Data courtesy of Shell Solar Industries) MANUFACTURING ISSUES 607 Finally, for space applications, Cu(InGa)Se2 thin-ﬁlm solar cells offer potential advantages since the radiation tolerance is high as compared to crystalline-silicon solar cells [4, 5]. The potential to use a lightweight plastic substrate could lead to solar cells with very high speciﬁc power, that is, power divided by mass, which is critical for some space applications (see Chapter 10 for a more complete discussion). However, Cu(InGa)Se2 space solar cell technology has not yet reached a commercial stage. 13.6.4 Production Costs Material costs have direct and indirect components, and depend on the material yield of the deposition processes. The direct material costs, that is, the cost of the feedstock, will not be reduced by an increased volume of the production, depending only on the feedstock market price and how much material is needed in the ﬁlm. The indirect costs, including preparation of sputtering targets or other source materials, will be reduced when production volumes are sufﬁciently large. The material yield, or fraction of the source material that ends up in the ﬁlm, may be less than 50% for various thin-ﬁlm processes. For sputtering, typically 30% of the target material ends up in the ﬁlms. In addition to materials, the other main production cost for thin-ﬁlm modules is the capital cost of the equipment. To ﬁrst order, any large-scale automated deposition equipment will have comparable price. Therefore, the throughput or production capacity will be very important for determining the capital cost. Costs around 20 $/m2 for each thin-ﬁlm deposition or process step may be accept- able in pilot production, but clear pathways toward costs in the range 1 to 5 $/m2 for large-volume production need to be identiﬁed. Throughput has a direct effect on cost. In an in-line process, this will depend on the substrate width and linear speed, which fundamentally depends on the deposition rate and desired thickness of the layer. If the deposition rate is relatively low, it can be compensated by having a long deposition zone in the system, for example, by having multiple targets in a sputtering system with only a relatively small increase in capital cost. All cost advantages for thin ﬁlms are lost if the production is not completed with high yield. The overall manufacturing yield can be broken down into electrical yield and mechanical yield. The electrical yield reﬂects the module reproducibility since it is the fraction of the modules produced which fulﬁll minimum performance criteria. The mechanical yield is the fraction of the substrates entering the production line that make it to the end. Mechanical losses result from broken glass substrates or malfunctioning equipment. In general, the overall yield should be well over 80%. Another manufacturing cost is the energy usage. The energy payback time for Cu(InGa)Se2 modules is expected to be fairly low; four months has been estimated by Alsema and van Engelenburg , compared to three years for crystalline-silicon modules . Production-cost analyses result in a range of projected manufacturing costs. There are predictions of 1.5 to 2 $/WP for ﬁrst-generation Cu(InGa)Se2 plants with a few MWP yearly capacity and projected costs of 0.4 to 0.6 $/Wp for large-volume manufacturing [214, 215]. 608 Cu(InGa)Se2 SOLAR CELLS 13.6.5 Environmental Concerns One of the environmental issues related to the materials in Cu(InGa)Se2 modules is the availability of less common elements. The content of the critical materials in grams per kWP has been calculated assuming 12% module efﬁciency and the result is compared with the amount reﬁned annually in Table 13.6 . The fourth column expresses how much module power could be obtained from the amount reﬁned annually and the last column shows a similar calculation based on the reserves of the various elements. Owing to uncertainties in estimates of reserves, or maximum resources, Table 13.6 just gives an indication of where, and at what level, potential problems in material supply may occur. It is clear that In is the potential bottleneck as regards primary material supply. CuInSe2 toxicity has been studied by administering it to rats . Even at high doses negligible effects were detected. A lowest observed adverse effect level (LOAEL) of 8.3 µg/kg/day for humans was derived from these studies. The other substances that constitute Cu(InGa)Se2 modules are largely nontoxic except for Cd. Many aspects of its use in PV manufacturing have been studied by Fthenakis and Moskowitz . Chemical bath deposition of CdS is the process step that presents the greatest health concerns due to the use of Cd, thiourea, and the genera- tion of waste solutions. In electrodeposition of CdTe, which also is a wet process using Cd precursors, it was found that the greatest health hazards from Cd are from dust gener- ated during feedstock preparation and from ﬁne particles near the baths . Biological monitoring at a process station showed that exposures can be maintained at a level that presents no risk to workers. Thiourea is a toxic and carcinogenic substance that also presents an exposure risk. Rinse water and dilute solutions of acids and Cd-compounds can be treated by a two-stage precipitation/ion exchange process. The Cd can be removed, and recycled, down to 1 to 10 ppb levels . Most Cu(InGa)Se2 processes use elemental Se, but the forms that are handled are solid shots or pellets that give off very little dust that could be inhaled. Elemental Se is considered to have a relatively low biological activity, but many compounds are very Table 13.6 Critical materials in Cu(InGa)Se2 modules with respect to primary supply (After Andersson B, Azar C, Holmberg J, Karlsson S, Energy 5, 407–411 (1998) ) Element Material Amount Amount Reserves/ content reﬁned reﬁned/ content [g/kWP ] [kton/y] content [TWP ] [GWP /y] Mo 42 110 2600 130 Cu 17 9000 529 000 30 000 In 23 0.13 5.7 0.1 Ga 5 0.06 12 2.2 Se 43 2 46 1.9 Cd 1.6 20 12 500 330 Zn 37 7400 200 000 4100 THE Cu(InGa)Se2 OUTLOOK 609 active and highly toxic. In particular, hydrogen selenide, a gas used in some selenization processes, is extremely toxic with an “immediately dangerous to life and health” (IDLH) value of only 2 ppm . There are also environmental concerns for the hazards during the operation of Cu(InGa)Se2 modules with one potential risk being the leaching of critical materials into rainwater. This only happens if a module is broken or crushed, so the normally well-encapsulated active layers are exposed. An experimental study of the emissions of toxic elements into rainwater from crushed CuInSe2 modules and into soil exposed to the water concluded that no acute danger to humans or the environment is likely to occur . The main hazard during the active life of the CuInSe2 modules is related to ﬁre accidents. A study of the potential risks associated with ﬁres in PV power plants shows that they are very limited . A ﬁre in a commercial-size system could result in harmful concentrations up to 300 m downwind of the ﬁre if most of the CuInSe2 materials are released. With release of 10% of the CuInSe2 materials, concentrations were not harmful even under worst-possible meteorological conditions. The study concluded that there are no immediate risks to the public from ﬁres in sites with CuInSe2 modules. Concerns for disposal of Cu(InGa)Se2 have also been tested with respect to leach- ability. Zn, Mo, and Se are eluted in the highest amounts. On the basis of landﬁll criteria, CuInSe2 modules will pass requirements in both Germany and the United States . Because of the low volume and leaching rates of critical elements from CuInSe2 modules, they will not be classiﬁed as hazardous waste according to most US regulations . The evolution of environmental regulations, disposal options, and economics makes recycling increasingly important. In large-scale use of Cu(InGa)Se2 modules, the supply of rare elements, in particular indium, but also selenium and gallium, provides a further motivation for recycling. The cost of recycling may be favorably offset if module materials can be reclaimed. In particular, if the glass sheets can be salvaged and reused, there will be a net gain associated with the recycling procedure. Thus, recycling may be an important consideration in the choice of encapsulation method. Double glass structures are functional and may reduce the release of CuInSe2 materials during ﬁres, but may increase the costs for recovering metals and reusing glass plates . 13.7 THE Cu(InGa)Se2 OUTLOOK Clearly, there has been tremendous progress in Cu(InGa)Se2 solar cells as evidenced by the high module and cell efﬁciencies fabricated by many groups, the range of deposition and device options that have been developed, and the growing base of science and engineering knowledge of these materials and processes. There is good reason to be optimistic that cell efﬁciencies greater than 20% will be achieved before long and that module performance and yield will continue to improve. Still, there is a lack of understanding of many of the critical problems associated with semiconductor processing and a need to devote time and research focus at both the laboratory scale, to address fundamental issues, and on the pilot line, to address equipment and scale-up problems and to validate processes. From their earliest development, CuInSe2 -based solar cells, along with other thin- ﬁlm PV materials including Cu2 S, CdTe, and amorphous Si, attracted an interest because of their perceived potential to be manufactured at a lower cost than Si wafer-based PV. 610 Cu(InGa)Se2 SOLAR CELLS However, after more than 25 years of research and development of CuInSe2 , manufac- turing has only recently moved past the pilot-production stage and has not demonstrated any cost advantages. A fundamental question must be asked: what needs to be done to ensure that Cu(InGa)Se2 solar cell technology reaches its potential for large-scale power generation? Part of the answer is to address the critical need for the accelerated development of new manufacturing technology including improved deposition equipment and processes based on well-developed engineering models. Also, new diagnostic and process-control tools will have to be developed. This requires fundamental materials and device knowledge to determine what properties can be measured in a cell or module fabrication process that can act as reliable predictors of ﬁnal performance. Better processes, equipment, and control based on a more solid knowledge base can directly translate to higher throughput, yield, and performance. There is also a critical need for continued improvement in the fundamental science of the materials and devices [222, 223]. Signiﬁcant improvements in efﬁciency will only come from increased VOC so the chemical and electronic nature of the defects that limit it, and their origin, must be understood. This can contribute to a comprehensive model for the growth of Cu(InGa)Se2 , relating processing parameters to defect formation, junction formation, and device limitations. In addition, a fundamental understanding of the role of sodium and the nature of the grain boundaries and free surface needs to be developed. A greater understanding of the role of the CdS layer and the chemical bath process might enable alternative materials that do not contain cadmium and have wider band gap to be utilized with greater efﬁciency and reproducibility. A second fundamental question to be asked is: what might be the breakthroughs that could lead to the next generation of thin-ﬁlm Cu(InGa)Se2 -based solar cells? Further development of wide band gap alloys to enable cells to be made with Eg ≥ 1.5 eV without any decrease in performance will have several beneﬁts for module fabrication and performance as discussed in Section 13.5. In addition, development of a cell with Eg ≈ 1.7 eV is a prerequisite for tandem cells based on the polycrystalline thin ﬁlms to be developed. A monolithic tandem cell has the potential to attain efﬁciencies of 25% or more. The CuInSe2 alloy system is ideally suited for such a structure since a CuInSe2 cell with Eg = 1.0 eV would make an ideal bottom cell with any of the alloys that increase band gap to 1.7 eV for the top cell. Even if a high-efﬁciency wide band gap cell is developed, such a structure will require the development of a transparent interconnect between the top and bottom cells and improvements in cell structure or low- temperature processes to allow the bottom cell to survive the subsequent processing of the top cell. Low-temperature processing of the Cu(InGa)Se2 layer without loss of efﬁciency in the ﬁnal solar cell can have signiﬁcant additional beneﬁts. With lower substrate tem- perature, alternative substrate materials, like a ﬂexible polymer web, can be utilized. In addition, lower TSS can reduce thermally induced stress on the substrate, allowing faster heat-up and cooldown, and decrease the heat load and stress on the entire deposition system. Similarly, there may be signiﬁcant cost and processing advantages to a cell structure that enables the use of a Cu(InGa)Se2 layer much less than 1 µm. REFERENCES 611 With all these challenges to improve the fundamental knowledge behind Cu(InGa)Se2 materials and devices and to develop new manufacturing technology and breakthrough advancements, research and development on Cu(InGa)Se2 and related materials remains exciting and promising. All of the reasons for the initial excitement over the potential for thin-ﬁlm Cu(InGa)Se2 remain valid. The high efﬁciency, demonstrated stability, and tolerance to material and process variations give great hope that it will be a major contributor to our solar electric future. REFERENCES 1. Contreras M et al., Prog. Photovolt. 7, 311–316 (1999). 2. Tanaka et al., Proc. 17th Euro. Conf. Photovoltaic Solar Energy Conversion, 989–994 (2001). 3. Wieting R, AIP Conf. Proc. 462, 3–8 (1999). 4. Burgess R et al., Proc. 20th IEEE Photovoltaic Specialist Conf., 909–912 (1988). 5. Jasenek A et al., Thin Solid Films 387, 228–230 (2001). 6. Hahn H et al., Z. Anorg. Allg. Chem. 271, 153–170 (1953). 7. Shay J, Wernick J, Ternary Chalcopyrite Semiconductors: Growth, Electronic Properties, and Application, Pergamon Press, Oxford (1974). 8. Tell B, Shay J, Kasper H, Phys. Rev. B4, 4455–4459 (1971). 9. Tell B, Shay J, Kasper H, J. Appl. Phys. 43, 2469, 2470 (1972). 10. Wagner S, Shay J, Migliorato P, Kasper H, Appl. Phys. Lett. 25, 434, 435 (1974). 11. Shay J, Wagner S, Kasper H, Appl. Phys. Lett. 27, 89–90 (1975). 12. Meakin J, Proc. SPIE Conf. 543: Photovoltaics, 108–118 (1985). 13. Kazmerski L, White F, Morgan G, Appl. Phys. Lett. 29, 268, 269 (1976). 14. Mickelsen R, Chen W, Proc. 15th IEEE Photovoltaic Specialist Conf., 800–804 (1981). 15. Mickelsen R, Chen W, Proc. 16th IEEE Photovoltaic Specialist Conf., 781–785 (1982). 16. Chen W et al., Proc. 19th IEEE Photovoltaic Specialist Conf., 1445–1447 (1987). 17. Potter R, Sol. Cells 16, 521–527 (1986). 18. Hedstr¨ m J et al., Proc. 23rd IEEE Photovoltaic Specialist Conf., 364–371 (1993). o 19. Gabor A et al., Sol. Energy Mater. Sol. Cells 4, 247–260 (1996). 20. Tarrant D, Ermer J, Proc. 23rd IEEE Photovoltaic Specialist Conf., 372–375 (1993). 21. Rocheleau R, Meakin J, Birkmire R, Proc. 19th IEEE Photovoltaic Specialist Conf., 972–976 (1987). 22. Mitchell K et al., IEEE Trans. Electron. Devices 37, 410–417 (1990). 23. Kazmerski L, Wagner S, “Cu-Ternary Chalcopyrite Solar Cells”, in Coutts T, Meakin J, Eds, Current Topics in Photovoltaics, 41–109, Academic Press, London (1985). 24. Haneman D, Crit. Rev. Solid State Mater. Sci. 14, 377–413 (1988). 25. Rockett A, Birkmire R, J. Appl. Phys. 70, R81–R97 (1991). 26. Rockett A, Bodeg˚ rd M, Granath K, Stolt L, Proc. 25th IEEE Photovoltaic Specialist Conf., a 985–987 (1996). 27. Rau U, Schock H, Appl. Phys. A 69, 131–147 (1999). 28. Suri D, Nagpal K, Chadha G, J. Appl. Cystallogr. 22, 578 (1989) (JCPDS 40-1487). 29. Ciszek T, J. Cryst. Growth. 70, 405–410 (1984). 30. Neumann H, Sol. Cells 16, 317–333 (1986). 31. Wasim S, Sol. Cells 16, 289–316 (1986). 32. o G¨ decke T, Haalboom T, Ernst F, Z. Metallkd. 91, 622–634 (2000). 33. Ye J, Yoshida T, Nakamura Y, Nittono O, Jpn. J. Appl. Phys. 35, 395–400 (1996). 34. Zhang S, Wei S, Zunger A, Phys. Rev. Lett. 78, 4059–4062 (1997). 35. Herberholz R et al., Eur. Phys. J. 6, 131–139 (1999). 36. Wei S, Zhang S, Zunger A, Appl. Phys. Lett. 72, 3199–3201 (1998). 612 Cu(InGa)Se2 SOLAR CELLS 37. Schroeder D, Rockett A, J. Appl. Phys. 82, 4982–4985 (1997). 38. Wei S, Zhang S, Zunger A, J. Appl. Phys. 85, 7214–7218 (1999). 39. Ruckh M et al., Sol. Energy Mater. Sol. Cells 41/42, 335–343 (1996). 40. Kazmerski L et al., J. Vac. Sci. Technol., A 1, 395–398 (1983). 41. o a Rinc´ n C, Gonz´ lez J, Sol. Cells 16, 357–362 (1986). 42. Alonso M et al., Phys. Rev. B 63, 075203 1–13 (2001). 43. Alonso M et al., Appl. Phys. A 74, 659–664 (2002). 44. Wei S, Zunger A, Appl. Phys. Lett. 72, 2011–2013 (1998). 45. Nouﬁ R, Axton R, Herrington C, Deb S, Appl. Phys. Lett. 45, 668–670 (1984). 46. Neumann H, Tomlinson R, Sol. Cells 28, 301–313 (1990). 47. Zhang S, Wei S, Zunger A, Katayama-Yoshida H, Phys. Rev. B 57, 9642–9656 (1998). 48. Schroeder D, Hernandez J, Berry G, Rockett A, J. Appl. Phys. 72, 749–752 (1998). 49. Kiely C, Pond R, Kenshole G, Rockett A, Philos. Mag. A 63, 2149–2173 (1991). 50. Chen J et al., Thin Solid Films 219, 183–192 (1992). 51. Wada T, Sol. Energy Mater. Sol. Cells 49, 249–260 (1997). 52. Schmid D, Ruckh M, Grunwald F, Schock H, J. Appl. Phys. 73, 2902–2909 (1993). 53. Gartsman K et al., J. Appl. Phys. 82, 4282–4285 (1997). 54. Kylner A, J. Electrochem. Soc. 146, 1816–1823 (1999). 55. Scheer R, Trends Vac. Sci. Tech. 2, 77–112 (1997). 56. Damaskinos S, Meakin J, Phillips J, Proc. 19th IEEE Photovoltaic Specialist Conf., 1299–1304 (1987). 57. Cahen D, Nouﬁ R, Appl. Phys. Lett. 54, 558–560 (1989). 58. Kronik L, Cahen D, Schock H, Adv. Mater. 10, 31–36 (1998). 59. Niles D, Al-Jassim M, Ramanathan K, J. Vac. Sci. Technol., A 17, 291–296 (1999). 60. Bodeg˚ rd M, Stolt L, Hedstr¨ m J, Proc. 12th Euro. Conf. Photovoltaic Solar Energy Conver- a o sion, 1743–1746 (1994). 61. a Bodeg˚ rd M, Granath K, Rockett A, Stolt L, Sol. Energy Mater. Sol. Cells 58, 199–208 (1999). 62. Shafarman W et al., AIP Conf. Proc. 394, 123–131(1997). 63. Contreras M et al., Thin Solid Films 361–2, 167–171 (2000). 64. Boyd D, Thompson D, Kirk-Othmer Encyclopaedia of Chemical Technology, Vol. 11, 3rd Edi- tion, 807–880, John Wiley (1980). 65. Probst V et al., Proc. 1st World Conf. Photovoltaic Solar Energy Conversion, 144–147 (1994). 66. Contreras M et al., Proc. 26th IEEE Photovoltaic Specialist Conf., 359–362 (1997). 67. Jensen C, Wieting R, Proc. 13th Euro. Conf. Photovoltaic Solar Energy Conversion, 1631–1633 (1995). 68. Basol B, Kapur V, Leidholm C, Halani A, Proc. 25th IEEE Photovoltaic Specialist Conf., 157–162 (1996). 69. Hartmann M et al., Proc. 28th IEEE Photovoltaic Specialist Conf., 638–641 (2000). 70. Vink T, Somers M, Daams J, Dirks A, J. Appl. Phys. 70, 4301–4308 (1991). 71. Wada T, Kohara N, Nishiwaki S, Negami T, Thin Solid Films 387, 118–122 (2001). 72. Matson R et al., Sol. Cells 11, 301–305 (1984). 73. Mattox D, Handbook of Physical Vapor Deposition (PVD) Processing, Noyes Publ., Park Ridge, NJ (1998). 74. Jackson S, Baron B, Rocheleau R, Russell T, Am. Inst. Chem. Eng. J. 33, 711–720 (1987). 75. Shafarman W, Zhu J, Thin Solid Films 361–2, 473–477 (2000). 76. Klenk R, Walter T, Schock H, Cahen D, Adv. Mater. 5, 114–119 (1993). 77. Kessler J et al., Proc. 12th Euro. Conf. Photovoltaic Solar Energy Conversion, 648–652 (1994). 78. Gabor A et al., Appl. Phys. Lett. 65, 198–200 (1994). 79. Hasoon F et al., Thin Solid Films 387, 1–5 (2001). 80. Stolt L et al., Appl. Phys. Lett. 62, 597–599 (1993). REFERENCES 613 o 81. Stolt L, Hedstr¨ m J, Sigurd D, J. Vac. Sci. Technol. A3, 403–407 (1985). 82. Powalla M, Voorwinden G, Dimmler B, Proc. 14th Euro. Conf. Photovoltaic Solar Energy Conversion, 1270–1273 (1997). 83. Eisgruber I et al., Proc. 28th IEEE Photovoltaic Specialist Conf., 505–508 (2000). 84. Nishitani M, Negami T, Wada T, Thin Solid Films 258, 313–316 (1995). 85. Negami T et al., Mater. Res. Soc. Symp. 426, 267–278 (1996). 86. Grindle S, Smith C, Mittleman S, Appl. Phys. Lett. 35, 24–26 (1979). 87. Chu T, Chu S, Lin S, Yue J, J. Electrochem. Soc. 131, 2182–2185 (1984). 88. Gay R et al., Proc. 12th Euro. Conf. Photovoltaic Solar Energy Conversion, 935–938 (1994). 89. Kapur V, Basol B, Tseng E, Sol. Cells 21, 65–70 (1987). 90. Basol B, Kapur V, IEEE Trans. Electron. Devices 37, 418–421 (1990). 91. Garcia F, Tomar M, Jpn. J. Appl. Phys. 22S, 535–540 (1983). 92. Schulz D et al., Proc. 27th IEEE Photovoltaic Specialist Conf., 483–486 (1997). 93. Knowles A, Oumous H, Carter M, Hill R, Semicond. Sci. Technol. 3, 1143, 1144 (1988). 94. Nadenau V et al., Prog. Photovolt. 3, 363–382 (1995). 95. Pern F et al., Sol. Cells 24, 81–90 (1988). 96. Guillemoles J et al., J. Appl. Phys. 79, 7293–7302 (1996). 97. Sato H et al., Proc. 23rd IEEE Photovoltaic Specialist Conf., 521–526 (1993). 98. Kessler J, Dittrich H, Grunwald F, Schock H, Proc. 10th Euro. Conf. Photovoltaic Solar Energy Conversion, 879–882 (1991). 99. Oumous H et al., Proc. 9th Euro. Conf. Photovoltaic Solar Energy Conversion, 153–156 (1992). 100. Karg F et al., Proc. 23rd IEEE Photovoltaic Specialist Conf., 441–446 (1993). 101. Mooney G et al., Appl. Phys. Lett. 58, 2678–2680 (1991). 102. Verma S, Orbey N, Birkmire R, Russell T, Prog. Photovolt. 4, 341–353 (1996). u 103. Wolf D, M¨ ller G, Thin Solid Films 361–2, 155–161 (2000). 104. Orbey N, Norsworthy G, Birkmire R, Russell T, Prog. Photovolt. 6, 79–86 (1998). 105. Jensen C, Tarrant D, Ermer J, Pollock G, Proc. 23rd IEEE Photovoltaic Specialist Conf., 577–580 (1993). 106. Marudachalam M et al., Appl. Phys. Lett. 67, 3978–3980 (1995). 107. Kushiya K et al., Proc. 25th IEEE Photovoltaic Specialist Conf., 989–993 (1995). 108. Thornton J, Lomasson T, Talieh H, Tseng B, Sol. Cells 24, 1–9 (1988). 109. Talieh H, Rockett A, Sol. Cells 27, 321–329 (1989). e 110. Guenoun K, Djessas K, Mass´ G, J. Appl. Phys. 84, 589–595 (1998). 111. Murali K, Thin Solid Films 167, L19–L22 (1988). 112. Galindo H et al., Thin Solid Films 170, 227–234 (1989). 113. Abernathy C et al., Appl. Phys. Lett. 45, 890 (1984). 114. Kazmerski L, Ireland P, White F, Cooper R, Proc. 13th IEEE Photovoltaic Specialist Conf., 184–189 (1978). 115. Potter R, Eberspacher C, Fabick L, Proc. 18th IEEE Photovoltaic Specialist Conf., 1659–1664 (1985). 116. Mitchell K, Liu H, Proc. 20th IEEE Photovoltaic Specialist Conf., 1461–1468 (1988). 117. Cashman R, J. Opt. Soc. Am. 36, 356 (1946). 118. Kitaev G, Uritskaya A, Mokrushin S, Sov. J. Phys. Chem. 39, 1101 (1965). 119. Kainthla R, Pandya D, Chopra K, J. Electrochem. Soc. 127, 277–283 (1980). 120. Kaur I, Pandya D, Chopra K, J. Electrochem. Soc. 127, 943–948 (1980). 121. Lincot D, Ortega-Borges R, J. Electrochem. Soc. 139, 1880–1889(1992). 122. Lincot D, Ortega-Borges R, Froment M, Philos. Mag. B 68, 185–194 (1993). 123. Nakada T, Kunioka A, Appl. Phys. Lett. 74, 2444–2446 (1999). 124. Kylner A, Rockett A, Stolt L, Solid State Phen. 51–52, 533–539 (1996). 125. Hashimoto Y et al., Sol. Energy Mater. Sol. Cells 50, 71–77 (1998). 126. Kylner A, Lindgren J, Stolt L, J. Electrochem. Soc. 143, 2662–2669 (1996). 614 Cu(InGa)Se2 SOLAR CELLS 127. Kylner A, Niemi E, Proc. 14th Euro. Conf. Photovoltaic Solar Energy Conversion, 1321–1326 (1997). 128. Cherin P, Lind E, Davis E, J. Electrochem. Soc. 117, 233–236 (1970). 129. Grzeta-Plenkovic B, Popovic S, Celustka B, Santic B, J. Appl. Crystallogr. 13, 311–315 (1980). 130. Kessler J et al., Tech. Digest PVSEC-6 , 1005–1010 (1992). 131. Yu P, Faile S, Park Y, Appl. Phys. Lett. 26, 384, 385 (1975). 132. Tell B, Wagner S, Bridenbaugh P, Appl. Phys. Lett. 28, 454, 455 (1976). 133. Kazmerski L, Jamjoum O, Ireland P, J. Vac. Sci. Technol. 21, 486–490 (1982). 134. Heske C et al., Appl. Phys. Lett. 74, 1451–1453 (1999). 135. Kylner A, J. Electrochem. Soc. 143, 1816–1823 (1999). 136. Ramanathan K et al., Proc. 2nd World Conf. Photovoltaic Solar Energy Conversion, 477–482 (1998). 137. Wang L et al., MRS Symp. 569, 127–132 (1999). a 138. Leskel¨ M, Ritala M, Thin Solid Films 409, 138–146 (2002). 139. Kessler J et al., Proc. 23rd IEEE Photovoltaic Specialist Conf., 447–452 (1993). 140. Olsen L et al., Proc. 26th IEEE Photovoltaic Specialist Conf., 363–366 (1997). 141. Sterner J, Kessler J, Bodeg˚ rd M, Stolt L, Proc. 2nd World Conf. Photovoltaic Solar Energy a Conversion, 1145–1148 (1998). 142. Bhattacharya R et al., Sol. Energy Mater. Sol. Cells 63, 367–374 (2000). 143. Kushiya K et al., Sol. Energy Mater. Solar Cells 49, 277–283 (1997). 144. Nakada T, Furumi K, Kunioka A, Trans. Electron Devices 46, 2093–2097 (1999). 145. Mufﬂer H et al., Proc. 28th IEEE Photovoltaic Specialist Conf., 610–613 (2000). 146. Ennaoui A, Blieske U, Lux-Steiner M, Prog. Photovolt. 6, 447–451 (1998). 147. Othake Y, Kushiya K, Yamada A, Konagai M, Proc. 1st World Conf. Photovoltaic Solar Energy Conversion, 218–221 (1994). 148. Engelhardt F et al., Proc. 2nd World Conf. Photovoltaic Solar Energy Conversion (1998). 149. Othake Y et al., Proc. 25th IEEE Photovoltaic Specialist Conf., 793–796 (1996). 150. Othake Y et al., Sol. Energy Mater. Sol. Cells 49, 269–275 (1997). 151. Hariskos D et al., Proc. 13th Euro. Conf. Photovoltaic Solar Energy Conversion, 1995–1998 (1995). 152. Guillemoles J et al., Jpn. J. Appl. Phys. 40, 6065–6068 (2001). 153. Gordon R, MRS Bull. 25, 52–57 (2000). 154. Lewis B, Paine D, MRS Bull. 25, 22–27 (2000). 155. Menner R, Sch¨ fﬂer R, Sprecher B, Dimmler B, Proc. 2nd World Conf. Photovoltaic Solar a Energy Conversion, 660–663 (1998). 156. Ruckh M et al., Proc. 25th IEEE Photovoltaic Specialist Conf., 825–828 (1996). 157. Westwood W, “Reactive Sputter Deposition”, in Rossnagel S, Cuomo J, Westwood W, Eds, Handbook of Plasma Processing Technology, Chap. 9, Noyes Publ., Park Ridge, NJ (1990). 158. Tarrant D, Gay R, Proc. 2001 NCPV Prog. Rev. Meeting, 113–114 (Nov. 2001). 159. Stolt L, Hedstr¨ m J, Skarp J, Proc. 1st World Conf. Photovoltaic Solar Energy Conversion, o 250–253 (1994). 160. Rau U, Schmidt M, Thin Solid Films 387, 141–146 (2001). 161. Kessler J et al., Proc. 16th Euro. Conf. Photovoltaic Solar Energy Conversion, 775–778 (2000). 162. Cooray N, Kushiya K, Fujimaki A, Okumura D, Jpn. J. Appl. Phys. 38, 6213–6218 (1999). a o 163. Lundberg O, Bodeg˚ rd M, Malmstr¨ m J, Stolt L, Prog. Photovolt. 11 (2002). 164. Fahrenbruch A, Bube R, Fundamentals of Solar Cells, 231–234, Academic Press, New York (1983). 165. Shafarman W, Klenk R, McCandless B, J. Appl. Phys. 79, 7324–7328 (1996). 166. Eron M, Rothwarf A, Appl. Phys. Lett. 44, 131 (1984). 167. Ohnesorge B et al., Appl. Phys. Lett. 73, 1224–1227 (1998). REFERENCES 615 168. Fahrenbruch A, Bube R, Fundamentals of Solar Cells, 105–161, Academic Press, New York (1983). 169. Phillips J et al., Phys. Status Solidi B 194, 31–39 (1996). 170. Roy M, Damaskinos S, Phillips J, Proc. 20th IEEE Photovoltaic Specialist Conf., 1618–1623 (1988). 171. Walter T, Menner R, K¨ ble C, Schock H, Proc. 12th Euro. Conf. Photovoltaic Solar Energy o Conversion, 1755–1758 (1994). 172. Shafarman W, Phillips J, Proc. 23rd IEEE Photovoltaic Specialist Conf., 364–369 (1993). 173. Sah C, Noyce R, Shockley W, Proc. Inst. Radio Engrs. 45, 1228–1243 (1957). 174. Eron M, Rothwarf A, J. Appl. Phys. 57, 2275–2279 (1985). 175. Walter T, Herberholz R, Schock H, Solid State Phen. 51, 301–316 (1996). 176. Rau U, Appl. Phys. Lett. 74, 111–113 (1999). 177. Meyer T et al., Eur. J. Appl. Phys. 8, 43–52 (1999). u 178. Walter T, Herberholz R, M¨ ller C, Schock H, J. Appl. Phys. 80, 4411–4420 (1996). 179. Herberholz R et al., Proc. 14th Euro. Conf. Photovoltaic Solar Energy Conversion, 1246–1249 (1997). 180. Nishitani M, Negami T, Kohara N, Wada T, J. Appl. Phys. 82, 3572–3575 (1997). 181. Sites J, Mauk P, Sol. Cells 27, 411–417 (1987). 182. Turner G, Schwartz R, Gray J, Proc. 20th IEEE Photovoltaic Specialist Conf., 1457–1460 (1988). 183. Schwartz R, Gray J, Lee Y, Proc. 22nd IEEE Photovoltaic Specialist Conf., 920–923 (1991). 184. Klenk R, Thin Solid Films 387, 135–140 (2001). 185. Niemegeers A, Burgelman M, De Vos A, Appl. Phys. Lett. 67, 843–845 (1995). 186. Liu X, Sites J, AIP Conf. Proc. 353, 444–453 (1996). 187. Minemoto T et al., Thin Solid Films 67, 83–88 (2001). 188. Schmid D, Ruckh M, Schock H, Sol. Energy Mater. Sol. Cells 41–2, 281–294 (1996). 189. Wei S, Zunger A, Appl. Phys. Lett. 63, 2549–2551 (1993). 190. Nelson A et al., Phys. Rev. B 42, 7518–7523 (1990). o 191. L¨ her T, Jaegermann W, Pettenkofer C, J. Appl. Phys. 77, 731–738 (1995). 192. Kronik L et al., Appl. Phys. Lett. 67, 1405–1407 (1995). 193. Morkel M et al., Appl. Phys. Lett. 79, 4482–4485 (2001). 194. Scheer R et al., Appl. Phys. Lett. 63, 3294–3296 (1993). 195. Paulson P et al., J. Appl. Phys. 91, 10153–10156 (2002). 196. Nadenau V, Hariskos D, Schock H, Proc. 14th Euro. Conf. Photovoltaic Solar Energy Con- version, 1250–1253 (1997). 197. Siemer K et al., Sol. Energy Mater. Sol. Cells 67, 159–166 (2001). 198. Marsillac S et al., Appl. Phys. Lett. 81, 1350–1352 (2002). 199. Shafarman W, Klenk R, McCandless B, Proc. 25th IEEE Photovoltaic Specialist Conf., 763–768 (1996). 200. Hanna G, Jasenek A, Rau U, Schock H, Thin Solid Films 387, 71–73 (2001). 201. Heath J et al., Appl. Phys. Lett. 80, 4540 (2002). 202. Nadenau V, Jasenek A, Rau U, Schock H, J. Appl. Phys. 87, 584–593 (2000). 203. Hengel I, Neisser A, Klenk R, Lux-Steiner M, Thin Solid Films 361–2, 458–462 (2000). 204. Gray J, Lee Y, Proc. 1st World Conf. Photovoltaic Solar Energy Conversion, 123–126 (1994). 205. Topic M, Smole F, Furlan J, J. Appl. Phys. 79, 8537–8540 (1996). 206. Dullweber T, Hanna G, Rau U, Schock H, Sol. Energy Mater. Sol. Cells 67, 145–150 (2001). 207. Dullweber T et al., Thin Solid Films 387, 11–13 (2001). 208. Probst V et al., Thin Solid Films 387, 262–267 (2001). 209. Russell T et al., Proc. 15th IEEE Photovoltaic Specialist Conf., 743–748 (1982). a 210. Kessler J, Wennerberg J, Bodeg˚ rd M, Stolt L, Sol. Energy Mater. Sol. Cells 67, 59–65 (2001). 616 Cu(InGa)Se2 SOLAR CELLS 211. Wennerberg J, Kessler J, Stolt L, Proc. 16th Euro. Conf. Photovoltaic Solar Energy Conver- sion, 309–312 (2000). 212. Alsema E, van Engelenburg B, Proc. 11th Euro. Conf. Photovoltaic Solar Energy Conversion, 995–998 (1993). 213. Jester T, Prog. Photovolt. 10, 99–106 (2002). 214. Woodcock J et al., Proc. 14th Euro. Conf. Photovoltaic Solar Energy Conversion, 857–860 (1993). 215. Zweibel K, Sol. Energy Mater. Sol. Cells 59, 1–18 (1999). 216. Andersson B, Azar C, Holmberg J, Karlsson S, Energy 23, 407–411 (1998). 217. Thumm W et al., Proc. 1st World Conf. Photovoltaic Solar Energy Conversion, 262–265 (1994). 218. Fthenakis V, Moskowitz P, Prog. Photovolt. 3, 295–306 (1995). 219. Steinberger H, Prog. Photovolt. 6, 99–103 (1998). 220. Moskowitz P, Fthenakis V, Sol. Cells 29, 63–71 (1990). 221. Eberspacher C, Fthenakis V, Proc. 26th IEEE Photovoltaic Specialist Conf., 1067–1072 (1997). 222. Schock H, Nouﬁ R, Prog. Photovolt. 8, 151–160 (2000). 223. Birkmire R, Proc. 26th IEEE Photovoltaic Specialist Conf., 295–300 (1997). 6 Chalcopyrite Based Solar Cells Renier Klenk, Martha Ch. Lux-Steiner Hahn-Meitner-Institut Berlin Gleinicker, Berlin 6.1 INTRODUCTION Chalcopyrite based solar modules uniquely combine advantages of thin ﬁlm technology with the efﬁciency and stability of conventional crystalline silicon cells . It is therefore believed that chalcopyrite based modules can take up a large part of the photovoltaic (PV) market growth once true mass production is started. The most important chalcopyrite compounds for photovoltaic applications are CuInSe2 , CuInS2 , and CuGaSe2 with bandgaps of 1.0, 1.5, and 1.7 eV, respectively. Together with related materials they offer high optical absorption and a wide range of lattice constants and bandgaps (Figure 6.1). The compounds can be alloyed to obtain intermediate bandgaps. Starting with single crystals , chalcopyrite based solar cells have been under investigation since 1974. The ﬁrst chalcopyrite cells had a CuInSe2 absorber and therefore the technology is most advanced for lower gap materials with a composition close to CuInSe2 . Today the efﬁciency of lab scale thin ﬁlm devices is close to 20 % , an efﬁciency comparable to the best multicrystalline silicon cells. Many scaling up and manufacturing issues have been resolved. Pilot production lines are operational and modules are commercially available. As of 2005 the market share of chalcopyrite PV modules is not yet signiﬁcant but major problems that might prevent further commercialization have not been identiﬁed. 6.2 POTENTIAL OF CHALCOPYRITE PHOTOVOLTAIC MODULES The attractive potential of chalcopyrite photovoltaic modules can be summarized by key points which we will brieﬂy illustrate here and assess in more detail in the reminder of the chapter: r High efﬁciency. r Stability. r Low cost. r Effective use of raw materials. Thin Film Solar Cells Edited by J. Poortmans and V. Arkhipov C 2006 John Wiley & Sons, Ltd 238 THIN FILM SOLAR CELLS 3 CuAlSe2 2.5 CuGaS2 Bandgap (eV) 2 CuGaSe2 1.5 CuInS2 AgInSe2 CuInTe2 1 CuInSe2 0.54 0.56 0.58 0.6 0.62 Lattice constant a Figure 6.1 Bandgaps and lattice constants of selected chalcopyrites according to data compiled in  (lines are a guide to the eye only). r Short energy payback time. r Adaptable to various applications. r Large supporting research and development community. Chalcopyrites clearly offer the highest efﬁciency potential among all thin ﬁlm technologies. The record efﬁciency for a small, lab scale cell is close to 20 %, using just a single layer antireﬂective coating and a standard metal grid but none of the complex concepts that have been used to produce record silicon cells. Submodule efﬁciencies are at almost 17 % , and square foot and larger modules range from 14 to 12 % efﬁciency [5–7]. It is conceivable that the maximum efﬁciency can be increased further. A signiﬁcant boost, exceeding even the theoretical limit for silicon, can be expected from the development of multijunction cells . The excellent performance is notable not only under standard reporting conditions but also when assessing monitoring data from outdoor installations. In contrast to amorphous silicon based cells chalcopyrite devices do not show any degra- dation under illumination. Outdoor testing indicates that achievable product lifetimes may be comparable to those of conventional photovoltaic modules. The low cost potential is roughly comparable to that of other thin ﬁlm technologies and is rooted in the use of inexpensive substrates, effective use of raw materials, high throughput, and large area deposition at low temperatures as well as monolithic interconnection. Apart from the substrate, the total thickness of a chalcopyrite cell, including all ﬁlms, is in the range of 2 to 4 μm, which implies that the raw material usage is only a tiny fraction of the material input for a silicon cell. Mass production will not be limited by the availability of raw materials. The energy payback time (EPBT) is obviously an important parameter when considering how far photovoltaics can contribute to the future energy supply. The much lower thermal budget of thin ﬁlm preparation (lower process temperatures as well as short process times) leads to a signiﬁcant beneﬁt. CHALCOPYRITE BASED SOLAR CELLS 239 As we will show in this chapter, there is considerable ﬂexibility concerning the choice of components of a chalcopyrite cell or module as well as concerning the preparation methods for these components. It is therefore possible to design products with an optimum efﬁciency/cost trade-off for various applications, with power demands ranging from mW to MW and with illumination intensity ranging from indoor, low level to high level under concentration. Chal- copyrite cells can be grown on rigid as well as ﬂexible substrates. They perform well in chal- lenging environments because they are mechanically robust, can operate in a wide temperature range, and can tolerate high radiation levels. The chalcopyrite technology is supported by a networked international research commu- nity comprising universities, research institutes, and companies with experience and long term commitment. This ought to guarantee that any problems surfacing in industrial implementation can be attacked with the necessary background knowledge and that new developments are in the pipeline for future products. The large ﬁnancial risk of implementing mass production facilities is limiting the production volume. However, the growing involvement of compa- nies developing deposition equipment and fabrication infrastructure provides a solid basis for ongoing commercialization. 6.3 TECHNOLOGY FOR THE PREPARATION OF CHALCOPYRITE SOLAR CELLS AND MODULES The solar cells consist of a number of ﬁlms which are deposited onto a rigid or ﬂexible substrate (Figure 6.2). The ﬁrst ﬁlm, typically molybdenum, serves as a nontransparent back-contact. It is covered by the actual chalcopyrite ﬁlm. This p-type ﬁlm absorbs most of the light and generates the photocurrent (absorber). The heterojunction is formed by depositing a very thin n-type buffer layer (typically CdS) and an n-type wide gap transparent front contact (usually heavily doped ZnO). Figure 6.2 Scanning electron micrograph of the cross section of a typical chalcopyrite solar cell with Cu(In,Ga)Se2 (CIGSe) absorber (substrate not shown). 240 THIN FILM SOLAR CELLS 500 °C CuInS2 S Float glass Mo Cu In Strukturierung 1 Mo Sputtering Laser Scribing Precursor Sputtering Reactive RTP ZnO Chemisches Bad Strukturierung 2 Verkapselung Chemical Mechanical Scribing TCO Mechanical Scribing Encapsulation Bath (Pattern 2) Sputtering (Pattern 3) Figure 6.3 Typical sequence of processes to prepare a chalcopyrite photovoltaic module . The sequence shown here is based on the sequential approach (see text) using rapid thermal processing (RTP). Reprinted from Thin Solid Films, 481–482, R. Klenk, J. Klaer, R. Scheer, M. Ch. Lux-Steiner, u I. Luck, N. Meyer, U. R¨ hle, 509, Copyright (2005) with permission from Elsevier. Thin ﬁlms with properties suitable for photovoltaic applications can be prepared by a variety of processes. In pilot lines the absorber is grown by multisource evaporation or by a sequential approach (reactive annealing of metal ﬁlms). To provide just one example, Figure 6.3 shows a sequence of processes necessary to produce a CuInS2 based module using the latter method. In the following paragraphs we will, for each of the ﬁlms, introduce the most common state-of-the- art methods as they are the basis of today’s production lines. Methods still under development and aiming at future signiﬁcant cost reduction are described in Section 6.6. 6.3.1 Absorber The most advanced absorber materials are based on CuInSe2 (CISe). Due to its rather low bandgap (1.0 eV) a common practice has evolved where gallium is added to obtain Cu(In,Ga)Se2 with a wider bandgap of approximately 1.15 eV. Due to reasons outlined be- low, sulfur is incorporated additionally in sequential processes to obtain Cu(In,Ga)(Se,S)2 (CIGSSe). All these absorbers require an optimized sodium concentration for optimum ﬁlm properties (see Section 126.96.36.199). Another line of development starts from CuInS2 (CIS). In contrast to the other compounds, this base material is grown under copper excess and it also has a higher bandgap (E g = 1.5 eV). Addition of further elements such as sodium or gallium is therefore not strictly necessary. The most commonly used absorber preparation methods are multisource evaporation and sequential processes. Evaporation offers a much more direct control of ﬁlm formation, including deliberate depth proﬁles for bandgap engineering as well as making optimum use of the ﬂuxing activity of Cu(S,Se) phases. This is probably the reason CHALCOPYRITE BASED SOLAR CELLS 241 why evaporation still yields the highest cell performance. In particular, unlike evaporation, the sequential processes lead to an aggregation of gallium close to the back contact which causes low gallium content in the active cell region and, consequently, a bandgap below the optimum value. The addition of H2 S to the annealing environment has helped to overcome this problem to a certain degree but it also adds to the overall complexity of the processes and the device. In evaporated ﬁlms the standard bandgap (in the order of 1.15 eV) can be achieved merely by incorporating sufﬁcient amounts of gallium. 188.8.131.52 Multisource evaporation Thin ﬁlms can be grown in a straightforward manner by coevaporating the constituent elements onto a heated substrate. The stoichiometry (concentration of VI element relative to the metals) is handled by a group VI overpressure which has to be maintained in the initial stage of cooling down the substrate. On the other hand, the molecularity (ratio of group I metal versus group III metal concentration) has to be adjusted by tight control of the metal source temperatures. Single crystal substrates with a suitable lattice constant and surface termination can be used and will result in epitaxial growth of the thin ﬁlm. The morphology and other properties of the resulting ﬁlm depend strongly on the molecu- larity. Copper rich ﬁlms exhibit larger grains. They are a mixture of chalcopyrite with a close to ideal composition and Cu-VI binary phases, typically found at the surface after cooling down the sample. The chalcopyrite grown under these conditions is characterized by lower defect density and reduced compensational doping in comparison to material grown without Cu excess (Figure 6.4). These observations suggest that the binary phases play an active role in the growth mechanism, also with regard to incorporation of the VI element . Films with an overall molecularity close to unity are often found to be inhomogeneous on a scale of several μm due to localized segregation of Cu-VI binary phases. Conductivity (Ω cm)-1 -3 10 -4 10 -5 10 0.48 0.49 0.50 0.51 0.52 0.53 Ga/(Cu+Ga) before etching Figure 6.4 Lateral conductivity of CuGaSe2 thin ﬁlms. Films grown under Cu rich conditions show higher effective doping than those grown under Cu poor conditions. The Cux Se segregations were removed by chemical etching prior to the measurement. 242 THIN FILM SOLAR CELLS Line Sources Se Cu Se In Ga Se Load Lock Load Lock Substrate Carrier Heaters Figure 6.5 Schematic view of an in-line evaporation system. Substrates are supported by a carrier and transferred into the evaporation chamber via a load lock. The substrate temperature is raised by an array of heaters. Before reaching the maximum temperature they pass the copper source and then, after reaching the maximum temperature, the indium and gallium sources. The substrate carriers are gradually cooled down prior to leaving the evaporation chamber via the load lock at the right hand side. Notwithstanding the small grain size, efﬁciencies in the order of 14 % are readily achieved with slightly Cu poor Cu(In,Ga)Se2 ﬁlms prepared by this simple coevaporation approach. On the other hand, several schemes have been developed to exploit the growth assistance that goes along with excess copper. In the bilayer approach , a Cu rich coarse grained seed ﬁlm is grown ﬁrst. The Cu rate is then diminished so that the Cu excess present in the seed ﬁlms is gradually consumed in this second stage, ending up with a slightly Cu deﬁcient ﬁlm. The idea here is to combine the superior properties of Cu rich ﬁlms with the absence of second phases in Cu poor ﬁlms. It is clear, however, that the growth mechanism cannot be sustained once the Cu VI phases have been consumed which may lead to disrupted growth and new nucleation. The advantage of this approach lies in the fact that it can be translated into an industrial inline process where the moving substrate passes ﬁrst the Cu, then the In and Ga sources (Figure 6.5). Other schemes start from Cu poor ﬁlms or even ﬁlms without any copper and add more copper in a second stage (inverted bilayer). If enough copper is delivered in this stage the ﬁlm becomes Cu rich and recrystallizes. In this case a third stage is needed to again reduce the Cu content and achieve single phase material. This scheme is generally known as the three stage process  and has resulted in the highest efﬁciency of lab scale devices so far (close to 20 %). In principle, there is again the problem of maintaining the growth mechanism after the excess copper has been consumed in the third stage. However, the three stage process enables precise control. The transition points where the ﬁlm enters and leaves the Cu rich regime are observable through changes in substrate temperature  (or power delivered to the heater to maintain a constant temperature) which is often used for in situ process control. Monitoring the intensity of light reﬂected off axis from the substrate is an alternative method for process control (laser light scattering ). The three stage process results in a certain depth proﬁle of the Ga/(Ga+In) ratio which is believed to contribute to the excellent performance of cells prepared from these absorbers (bandgap engineering). 184.108.40.206 Sequential processes Sequential (two step) processes have been developed as an alternative approach to absorber formation [14, 15]. Here, a metallic precursor is typically deposited by sputtering. Sputtering CHALCOPYRITE BASED SOLAR CELLS 243 of pure gallium is problematic due to its low melting point. Copper/gallium alloyed targets are typically used instead. The chalcopyrite is formed in the second step by exposing the precursor to a chalcogen containing atmosphere at elevated temperatures (selenisation/sulfurisation). The method is particularly attractive for production. A process well established in industry is DC sputtering and off-the-shelf equipment is readily available. It is characterized by good repro- ducibility and large area uniformity of the thicknesses of the individual ﬁlms. Consequently, the important Cu/III ratio can be tightly controlled in this ﬁrst step. The high temperature, corrosive environment which is potentially problematic with respect to equipment degradation over time is limited to the second step. Here, it is less critical because this second step mainly affects the chalcogen stoichiometry which is to a great extent self adjusting. The second step can be carried out in a tube furnace or by rapid thermal processing (RTP). Annealing in a tube furnace (using mixtures of an inert carrier gas and H2 S and/or H2 Se reac- tive components) is typically a slow batch type process where several substrates are processed simultaneously and where the substrate size is somewhat restricted by the maximum available diameter of the furnace tubes. More recently rapid thermal processing furnaces have been introduced [16, 17]. In one of the processes selenium is evaporated onto the metal precursor (stacked elemental layers) before the high temperature annealing whereas sulfur is still intro- duced as H2 S during the RTP. Chalcopyrite formation in the two step process depends largely on the thermodynamics and phase formation kinetics of the material system. These funda- mental properties have therefore been investigated in detail in order to be able to optimize the processes [18, 19]. In addition to producing low gap Cu(In,Ga)(S,Se)2 absorbers, two step processing has also been found particularly suitable for preparing CuInS2 (E g = 1.5 eV) . In this case it is possible to simply place pieces or powder of sulfur next to the substrate in the RTP furnace, thereby eliminating the need for chalcogen evaporation in a separate process as well as the use of any toxic gas. Very short annealing periods (a few minutes at top temperature) are achievable due to the growth assistance by Cux S using Cu rich (Cu/In = 1.2–1.8) precursors. Films grown under these conditions exhibit high p-type conductivity, therefore the sodium concentration (see Section 220.127.116.11) is not critical . Blocking layers and sodium precursor ﬁlms are not used. As already mentioned, it is also not necessary to incorporate additional elements for bandgap adjustment. 18.104.22.168 Sodium Recognizing the important inﬂuence of sodium on device performance had been a major breakthrough in the development. Sodium appears to inﬂuence the growth mechanism leading to superior morphology as well as higher effective p-type doping. In terms of cell perfor- mance the latter seems to be the more important aspect. The observed increase in open circuit voltage is quantitatively consistent with the measured increase in net doping in good cells where recombination in the space charge region is dominant . The fact that an increase in device performance is also achieved by diffusing sodium into an already prepared ﬁlm is also a strong indication that the electronic effects are more signiﬁcant than the morphology . Especially in the slower evaporation processes, sodium diffusing from a soda lime glass substrate through the back contact can yield the required sodium concentration in the absorber ﬁlm. If the substrate does not contain sodium, sodium salts (e.g., NaF) can be coevaporated 244 THIN FILM SOLAR CELLS or deposited as a precursor ﬁlm onto the back contact before depositing the absorber. The latter approach, in combination with a diffusion blocking layer underneath the back contact is sometimes also used with soda lime substrates to achieve a more accurate control of sodium concentration, especially in fast sequential processes. Films prepared with excess copper are characterized by large grains as well as high con- ductivity and, hence, do not require sodium doping. At the moment, this is technically relevant only for CuInS2 based modules. 6.3.2 Contacts 22.214.171.124 Diffusion barrier and back contact Sputtered molybdenum is the most frequently used back contact. The inherent stress in the ﬁlm can be adjusted over a wide range through the pressure of the working gas in the sputter process . Optimized ﬁlms adhere very well to glass or other substrates and laser or photolithographic patterning is straightforward. A possible alternative to sputtering is e-gun evaporation which so far has been used only in laboratory scale preparation but may also have cost advantages in large scale production. The actual contact to the chalcopyrite is complex and may involve a Mo–chalcogenide intermediate layer which forms during absorber preparation. At low temperatures the I (V ) curve often deviates from that of an ideal diode under forward bias. This bend-over of the curve has been attributed to blocking at the back contact of the cell. However, no voltage drop across the back contact could be found with appropriate test structures . There are other models to explain the I (V ) curve bend-over without involvement of a blocking contact . Alternative contact materials for improved optical reﬂection and novel device conﬁgurations are under investigation (Section 6.6). As already mentioned, diffusion barriers (silicon oxide, silicon nitride) deposited onto the glass substrate before the molybdenum are not strictly necessary but can be used for a more precise control of sodium doping. On the other hand, metal foil substrates often require additional coatings underneath the molybdenum for blocking of impurities as well as substrate planarization and isolation . Load Lock Sputter Chamber Se Evaporation RTP Furnace Load Lock CuGa In Substrate Heaters Figure 6.6 Schematic view of an in-line sequential system. Copper and gallium are sputtered from an alloy target. After sputtering indium, the substrate is transferred to an evaporation chamber and coated with selenium. The completed stack is annealed in the rapid thermal processing furnace. No substrate heating is required during metal and selenium deposition. CHALCOPYRITE BASED SOLAR CELLS 245 126.96.36.199 Buffer The thin (typically 50 nm) buffer layer is grown from a chemical bath . Typical solutions contain a Cd salt, thiourea as a sulfur source and ammonia in an aqueous solution. The substrate is immersed in the cold solution. The solution is then heated to 60–80 ◦ C. The thiourea hy- drolyzes and cadmium and sulfur ions recombine to form CdS. The ﬁlms grow either directly at the substrate or nanoparticles are formed in the solution and deposited onto the substrate. Depending on the deposition conditions, and due to the aqueous environment, the ﬁlm may contain signiﬁcant amounts of oxygen and hydrogen. Chemical bath deposition (CBD) of CdS is very reproducible and yields good cell performance on any chalcopyrite absorber. 188.8.131.52 Window The preferred window or TCO (transparent conductive oxide) ﬁlm consists of ZnO deposited by sputtering or metal organic chemical vapor deposition (MOCVD). This ﬁlm needs to have a high lateral conductivity in modules to avoid ohmic losses. It is therefore highly doped with aluminium or gallium (sputtered ﬁlms) or boron (MOCVD). However, depositing a ﬁlm with low lateral resistance directly onto the buffer increases the negative inﬂuence of local defects (such as pin holes ) and local ﬂuctuation of absorber properties (e.g. the bandgap ). This can be avoided by ﬁrst depositing a thin (in the order of 100 nm) ZnO ﬁlm with lower conductivity, i.e., by sputtering from an undoped target or by adding oxygen to the working gas. The window layer contributes signiﬁcantly to the module cost. Low resistivity is therefore desirable to minimize the ﬁlm thickness. In practice, the resistivity of large area ZnO thin ﬁlms is in the order of 5 × 10−4 cm and cannot be signiﬁcantly improved because higher doping reduces the electron mobility and causes poor transmission due to free carrier absorption. It has been argued that the ﬁlm properties are very close to physical limits  and that, in the long run, ZnO could be replaced by other materials with higher mobility (lower effective electron mass). Approaches to reduce the cost of ZnO preparation include new methods to fabricate ceramic targets or to use reactive sputtering from metallic targets. The ZnO ﬁlm plays an important role for module stability in accelerated lifetime testing under damp heat conditions which forms a part of the EN/IEC 61646 certiﬁcation. The lateral resistance tends to increase, giving rise to ﬁll factor losses. It is therefore mandatory to optimize ZnO preparation not only with respect to the as-grown properties but also by taking into account the degradation in damp heat. 184.108.40.206 Monolithic integration and encapsulation Manufacturing of modules adds some process steps to the cell preparation outlined above. The module is divided into cells which are connected in series by monolithic integration. The connection is made from the molybdenum back contact to the TCO during TCO deposition (Figure 6.7). A front metallization has been suggested but is normally not applied. The common scheme requires three patterning steps: an isolation scribe in the molybdenum (P1), scribing the absorber to create a gap which is later ﬁlled by TCO (P2), and an isolation scribe of the complete cell structure down to the molybdenum (P3). While the preferred tool for P1 patterning is a 246 THIN FILM SOLAR CELLS Figure 6.7 Optical microscopy image of a laser scribed line in a molybdenum ﬁlm on glass. Pulse frequency of the laser and scribing velocity have been adjusted for optimum overlap of pulses. pulsed Nd-YAG Laser (Figure 6.8), photolithographic patterning is also possible. After laser patterning the substrate is again subject to wet cleaning with rotating brushes to remove loose particles. P2 and P3 patterning are carried out by mechanical scribing. P2 patterning can be carried out before or after deposition of the undoped ZnO layer, the latter method may give a better contact. The interconnection area constitutes a loss in active area and hence photocurrent. In princi- ple, the scribe lines themselves should be as narrow as possible, the distance between P1, P2 and P3, respectively, should be minimal, whereas the distance between two interconnection areas should be maximal. In practice, a certain width of scribe line has to be maintained for sufﬁcient isolation (P1, P3) and contact resistance (P2). Hence the total width of the interconnection is in the range of 0.5 to 1 mm. In addition, the allowable distance between interconnection areas is limited by the lateral resistance of the ZnO ﬁlm. The distance can be increased when using a thicker ZnO ﬁlm. However, a thicker ﬁlm, apart from being costly, also causes photocurrent losses due do its reduced transparency. Consequently the typical distance is in the range of 5 to 10 mm which results in an area loss due to interconnections of approximately 10 %. Wide gap absorbers offer more ﬂexibility in designing the module due to their reduced current density. Typical photocurrent densities under full illumination are 42 mA/cm2 at a bandgap of 1 eV P1 P2 P3 ZnO:Al CdS/ZnO CIGSe Mo Substrate Figure 6.8 Schematic cross section of the cell interconnect in monolithic integration. This ﬁgure shows the variant where the P2 scribing is carried out after the deposition of the undoped ZnO. CHALCOPYRITE BASED SOLAR CELLS 247 Integral fluorescence intensity [a.u.] 600μm after DH before DH P2 P3 1.6 1.4 1.2 1.0 0.8 0.6 0.4 Lateral position [mm] Figure 6.9 Locally resolved sulfur ﬂuorescence (XES) intensity across an interconnect test structure before and after accelerated ageing in damp heat (DH). The position of the scribe lines is indicated (P2, P3) . (CISe) and 22 mA/cm2 at 1.5 eV (CIS), respectively. A computer simulation has been made available to optimize the module patterning for a given set of ﬁlm properties . The interconnection appears to play a certain role in module degradation in accelerated aging tests (damp heat). Locally resolved X-ray emission spectroscopy (XES) scans (Figure 6.9) on specially prepared test structures (larger scribe line distance, reduced ZnO thickness) results in a preliminary model of contact degradation. Before damp heat, a large sulfur signal is observed within the P3 scribe line which is due to sulfurisation of the molybdenum surface (which occurs during absorber formation). A smaller signal is observed within the P2 scribe due to the signal being attenuated by the ZnO ﬁlm. This latter signal increases signiﬁcantly upon damp heat treatment. The spectra suggest that this is due to the formation of ZnSO4 , i.e., damp heat causes a chemical reaction between MoS2 and ZnO thereby deteriorating the contact in the P2 scribe line. Concerning the described standard procedures there is still room for improvement because patterning unfortunately interrupts the in-line vacuum processing. Patterning is also critical with respect to the throughput (cycle time).The ﬁnal glass–glass laminate is produced using EVA foil and standard laminators as in the silicon technology. Encapsulation is critical for passing the accelerated aging tests. 6.4 CHARACTERIZATION AND MODELING In terms of modeling, the chalcopyrite based solar cell is a quite complex device compris- ing a number of polycrystalline compound semiconductor ﬁlms and several heterointerfaces. Nevertheless, tremendous progress has been achieved in measuring, modeling and understand- ing various aspects of the device. Novel characterization methods speciﬁcally adapted to the problems at hand have been developed and introduced. Examples are Kelvin probe force mi- croscopy (KPFM) to measure work functions with submicron lateral resolution (Figure 6.10), 248 THIN FILM SOLAR CELLS Figure 6.10 KPFM measurement  of a mechanically polished cross section of a ZnO:Ga/ (Zn,Mg)O/CIGSSe thin-ﬁlm heterostructure: a) topography, b) work function measured simultaneously with topography. Bright colour corresponds to high work function. c) is a plot of height, work function, and electrical ﬁeld along the path indicated in a). Reprinted from Thin Solid Films, 481–482, Th. Glatzel, H. Steigert, S. Sadewasser, R. Klenk, M.Ch. Lux-Steiner, 177 , Copyright (2005) with permission from Elsevier. inverse photoemission spectroscopy (IPES) to directly measure conduction band line ups, and X-ray emission spectroscopy (XES) for the analysis of buried interfaces. In general terms, quantitatively extracting material and device parameters from a measurement result requires a model that describes the correlation between the property assessed by the measurement and the underlying physical parameters. Due to the complexity of compound polycrystalline semi- conductor ﬁlms, the models are often an approximation. The extracted parameters have to be interpreted as effective parameters bound to the speciﬁc model. They can be useful for com- paring different samples quantitatively but can be misleading when used for calculations in a different context. Numerical modeling has emerged as a useful tool for a better understanding of the device. It is important because analytical approximations that can be safely made for other solar cells types are not valid in the chalcopyrite cell. In conclusion, the materials science of chalcopyrites, and solar cells based on them, is a wide ﬁeld under active development. Here, we have to limit the discussion to a few selected topics and the reader is referred to the literature for more in depth information (the overview given in  is a good starting point). 6.4.1 Cell concept The high optical absorption of the direct semiconductor chalcopyrites makes very thin ab- sorbers feasible. However, it also means that the incident sunlight is absorbed close to the surface. Assuming it would be possible to dope chalcopyrites in a well controlled manner it would still be challenging to reach high efﬁciency with a homojunction solar cell. Depending on surface passivation, the major part of carriers generated between the surface and the pn junction would be lost due to surface recombination. This problem is avoided by introducing the window/absorber heterojunction concept. Due to the wide bandgap of the window, the absorption is shifted away from the surface to the internal interface. Even assuming no surface CHALCOPYRITE BASED SOLAR CELLS 249 passivation and a very high surface recombination velocity, the losses are nevertheless small because only an insigniﬁcant part of the light is absorbed in the window. On the other hand, the internal heterointerface might itself cause recombination, leading not only to photocurrent loss but also to high bucking currents and consequently low open circuit voltage. But calculations show that interface recombination does not necessarily have a signiﬁcant impact on cell performance . The severity of interface recombination depends on: r The density of recombination centers at the interface. r The doping of absorber and window. r The type and density of charged interface states. r The conduction band line-up. A low density of interface states is always advantageous but may in practice be difﬁcult to achieve because the large area technology cannot be compared to the ultra clean environment required for defect free epitaxial growth. The other, more feasible approach to lowering re- combination lies in minimizing the density of either electrons or holes at the interface, which requires appropriate doping, band line-up and interface charge. In terms of bucking current (open circuit voltage) minimizing the density of either type of carrier yields comparable results. Considering the photogenerated carriers it is, however, mandatory that the electrons collected from the absorber are majority carriers at the interface (inverted interface). In conclusion, the structure should be an n+ window/p absorber heterojunction where the Fermi level (E F ) at the interface is close to the conduction band and where the Fermi level intersects midgap energy at a small distance from the interface in the absorber. The interface charge should be positive (donor like defects) to assist in establishing this structure. Likewise, the energetic position of the absorber conduction band edge (E C ) should be slightly lower than that of the window (spike) because the opposite situation (cliff) tends to push the absorber conduction band away from the Fermi level. Furthermore, in the cliff type band line-up electrons from the window side of the interface could recombine with holes from the absorber side which leads to a reduced barrier. In the actual chalcopyrite cell these design considerations must be implemented by the buffer layer and its preparation technique. Measurements of the band line-up and Fermi level position at the interface (see  for a review of chalcopyrite surface and interface properties) are not straightforward and the results show signiﬁcant variations [38–41]. The band diagram shown in Figure 6.11 is believed to be a reasonable approximation and shows the type inversion of the interface as required according to the above considerations. There are theories and supporting measurements that indicate an inherent widening of the absorber bandgap towards the surface [39, 42, 43]. It has been speculated that this effect is due to the segregation of distinct Cu poor ordered vacancy compound (OVC) but there is no conclusive evidence concerning the actual structure. If this widening is mainly due to a shift of the valence band edge (as shown in Figure 6.11) this also decreases the hole density at the interface and lowers the interface recombination even further. Obviously, this is only helpful for device performance when the transition from bulk chalcopyrite to surface bandgap widening occurs without introducing a high defect density. However, lattice matching between the bulk and surface phases may be absent in wide gap chalcopyrites . 250 THIN FILM SOLAR CELLS ZnO:Al ZnO CdS Cu(In,Ga)Se2 EG = 1.2 eV 1 EF 0 Energy (eV) EC -1 OVC -2 -3 EV -4 0.25 0.5 0.75 1 1.25 Distance from front contact (μm) Figure 6.11 Tentative calculated  band diagram of a ZnO/CdS/Cu(In,Ga)Se2 heterojunction as- suming a widening of the bandgap at the absorber surface due to an ordered vacancy compound (OVC). 6.4.2 Carrier density and transport Without deliberate doping, the majority carrier (hole) density in chalcopyrite thin ﬁlms is usually in the range of 1014 −1017 cm−3 and well suited for the application. Oxygen and sodium are the most common impurities known to inﬂuence the carrier density. The density of donors is found to be almost comparable to the acceptor density (compensation, self- compensation ). The effective mobility in polycrystalline ﬁlms at room temperature is in the range 1–100 cm2 /Vs. A complete set of temperature dependent conductivity and Hall effect data including polycrystalline as well as single crystal samples is only available for a certain (prepared with Cu excess) type of CuGaSe2 . They are summarized and discussed in . Charge trapped at grain boundaries gives rise to space charge regions extending into the grains. The resulting band bending leads to potential barriers. Hall effect measurements suggest barriers in the order of 50 to 150 meV which can be explained by a charge density in the order of 1012 cm−2 at the CGSe grain boundary . Kelvin probe force microscopy in ultra high vacuum is a relatively new characterization tool well suited to further clarify the grain boundary models. The high lateral resolution of the microscope allows work function measurements across individual grain boundaries. An average drop in the work function of 110 meV across grain boundaries has been measured in good agreement with the Hall data . The electrical ﬁelds in the vicinity of grain boundaries would sweep minority carriers into the grain boundary. However, excessive grain boundary recombination is in contradiction to the high photocurrents which are observed even in ﬁne grained ﬁlms. Alternative models of the grain boundary have been proposed, based on band structure calculations . A comparison of the impact of either type of grain boundary upon device performance has been performed by two-dimensional numerical calculations . Kelvin probe force microscopy results on the differences in the local variation of the work function with illumination (surface photovoltage) suggest that both aspects, i.e., trapped charge as well as the band structure, play a role in deﬁning the grain boundary properties . CHALCOPYRITE BASED SOLAR CELLS 251 There are a number of studies concerning time constants and pathways for radiative re- combination as deduced from steady state photoluminescence and photoluminescence decay. Radiative recombination appears to be dominant in good quality polycrystalline ﬁlms at low (8.5 K) temperature . Free carrier lifetimes in the order of some nanoseconds have been measured. Defect related lifetimes were signiﬁcantly higher, presumably due to trapping of carriers. Recent photoluminescence investigations  reveal a shallow donor (D) and two shallow acceptor levels (A1, A2) which are present throughout the Cu(In, Ga)Se2 compounds. The levels are found at slightly increasing depth when going from CuInSe2 (D = 10 meV, A1 = 40 meV, A2 = 60 meV) to CuGaSe2 . (D = 13 meV, A1 = 60 meV, A2 = 100 meV). The relative concentration of the acceptor level is correlated with Cu/(In+Ga) ratio used during preparation of the ﬁlm. For CuGaSe2 , the same set of shallow defects can explain the photoluminescence as well as the Hall measurement data . It may be tempting to assign these levels to the point defects of a ternary system whose formation energies and depth has been theoretically calculated  , but the assignment remains uncertain in view of more complex defects and impurities which might also play a dominant role. At room temperature the photoluminescence decays with a time constant of several tens of nanoseconds due to nonradiative recombination. The majority of measurements described in the literature and concerning minority carrier transport is assessing the minority carrier diffusion length. The latter can be extracted from electron beam induced current (EBIC)  and quantum efﬁciency measurements . Typical values are in the range of 1–2 μm for good quality ﬁlms. This implies that the diffusion length and the extension of the space charge region are comparable, which can cause problems in extracting both parameters from a single measurement. The ambiguity can sometimes be resolved by measuring at varied applied bias voltages and using an analytical approximation to calculate the ﬁeld zone as a function of bias . 6.4.3 Loss mechanisms Depending on absorber material and bandgap, record efﬁciencies for chalcopyrite based so- lar cells are in the range of 10 to almost 20 %. Optical  and contact related losses are small, at least in small area cells with front contact grids. They need to be considered for modules (without grids [4, 60]) and thin absorbers . In general, close to ideal photocur- rent collection is achievable whereas open circuit voltage and ﬁll factor offer room for future improvement. Figure 6.12 shows the external quantum efﬁciency of a Cu(In,Ga)S2 (E g ≈ 1.53 eV) solar cell  which is close to unity at the maximum with a single layer antire- ﬂective coating. The curve is limited by the ZnO and absorber bandgaps at low and high wavelengths, respectively. Holes generated in the CdS buffer layer are not collected which leads to a drop in quantum efﬁciency for photon energies higher than the bandgap of CdS (the interfaces on both sides are not inverted with respect to the n type CdS which leads to a high recombination probability for holes, nevertheless, partial collection has been reported ). The bucking current is responsible for losses in open circuit voltage. It also leads to losses in ﬁll factor because the diode ideality factor A is higher than unity. In high efﬁciency CIGSe cells the bucking current is due to bulk recombination in the space charge region of the absorber. The diode ideality factor (between one and two) and its temperature dependence are in agreement with analytical models describing recombination over an exponentially decaying density of 252 THIN FILM SOLAR CELLS 1 External quantum efficiency 0.8 CdS 0.6 0.4 Cu(In,Ga)S2 ZnO 0.2 0 400 500 600 700 800 Wavelength (nm) Figure 6.12 External quantum efﬁciency of a MgF2 /ZnO/CdS/Cu(In,Ga)S2 solar cell. The blue response is limited by the bandgaps of ZnO and CdS, respectively. The red response is limited by the bandgap of the absorber. states [63, 64]. Such a defect distribution could be due to band tails and is also observed in optical and admittance spectroscopy [65, 66]. Typical Urbach energies are in the range of 50 to 100 meV. In addition to the broad defect distribution there also seems to be a narrow defect dis- tribution at 250–300 meV above the valence band . It has been found by several meth- ods and in samples prepared by various methods, including single crystals . This de- fect maintains its energetic position relative to the valence band within the whole CIGSe system (for varied Ga/(Ga+In) ratios). However, its concentration seems to correlate with the gallium content, with a minimum concentration found at Ga/(Ga+In) ≈ 0.3. Losses in open circuit voltage were found to correlate with the concentration of this defect which sug- gests its signiﬁcance as a recombination path . Its concentration is found to increase upon radiation with energetic particles and to decrease with subsequent annealing at mod- erate temperatures . Other authors have identiﬁed a deeper defect at 0.8 eV above the valence band edge . This defect is approaching mid-gap position at high gallium con- tents and can thus be expected to be an effective recombination center in wide gap CIGSe absorbers. The saturation current in state-of-the-art low gap cells is mostly thermally activated and the activation energy corresponds to the bandgap of the absorber. The diode ideality factor is only mildly temperature dependent. Measurement of the open circuit voltage is as a function of temperature and extrapolation yield: Voc (T = 0 K) = E g /q. Cells with inferior efﬁciency show a stronger inﬂuence of tunneling. Depending on sample and temperature, higher and more temperature dependent diode ideality factors are observed. If the tunneling inﬂuence is not too severe, the bucking current mechanism can be described by tunneling assisted recombination via trap states in the space charge region of the absorber . In this case, the activation energy of the saturation current still corresponds to the bandgap after taking into account the temperature dependent diode ideality. Due to a common mistake in the evaluation of measurement results  this model may have also been applied to devices where the inﬂuence of tunneling is in fact much stronger. CHALCOPYRITE BASED SOLAR CELLS 253 Cells based on wide gap Cu(In,Ga)S2 absorbers exhibit open circuit voltages which are lower than expected, considering the bandgap and the good bulk properties deduced from photocurrent collection. Transport analysis  reveals that the dominant bucking current mechanism changes with illumination. The ideal border cases are recombination over a re- duced barrier at the interface without major assistance by tunneling (under illumination) and recombination in the space charge region with signiﬁcant tunnelling assistance (dark). While this change of recombination mechanism with illumination is a unique feature of Cu(Ga,In)S2 based cells , reduced thermal barrier and tunneling currents appear to be more general problems and are also observed in wide gap C(I)GSe based solar cells [75, 76]. The achieved development status is illustrated by the exemplary efﬁciencies listed in Table 6.1. The ﬁndings for the high efﬁciency cells described so far indicate that a further improvement of efﬁciency beyond 20 % ought to be feasible by minimizing the bulk defect density. However, it has also been argued that empirical device optimization has already led to an optimum with respect to band edge ﬂuctuations [77, 78]. Such ﬂuctuations (which are also observed in photoluminescence) may assist in charge separation, i.e. photocurrent collection, but are detrimental in terms of bucking current. On the other hand, even if the efﬁciency of low gap single junction cells is approaching practical limits, wide gap and tandem cells offer a large potential for ultra high efﬁciency (Section 6.6.7). Table 6.1 Efﬁciency of selected small area, laboratory style chalcopyrite based solar cells (partly with antireﬂective coating) Lab [Ref] Substrate Absorber Buffer Efﬁciency Remarks NREL  CISe 15 % E g = 1.04 eV NREL  CIGSe 19.5 % E g = 1.15 eV NREL  CIGSe 10.2 % E g = 1.64 eV NREL  glass CGSe CdS 9.5 % E g = 1.68 eV HMI  CIS 11.4 % E g = 1.5 eV HMI  CIGS 12.3 % E g = 1.53 eV HMI  CIGS 10.1 % E g = 1.65 eV NREL  steel 17.5 % ﬂexible HMI  titanium CIGSe CdS 16.2 % ﬂexible ETHZ  polyimide 14.1 % ﬂexible NREL/Aoyama glass CIGSe Zn(S,O,OH) 18.6 % Cd freea buffer by CBD Gakuin Univ.  Shell Solar/HMI  CIGSSe In2 S3 14.7 % Cd free buffer by ILGAR Shell Solar/HMI  CIGSSe (Zn,Mg)O 12.5 % Cd free buffer by sputtering (dry process) ZSW/CNRS  CIGSe In2 S3 16.4 % Cd-free buffer by ALCVD (dry process) Stuttgart Univ.  CIGSe Inx S 14.8 % Cd-free buffer by evaporation (dry process) a see  for a more complete list of Cd free devices. 254 THIN FILM SOLAR CELLS Figure 6.13 Partial view of the pilot production line for CIS modules (substrate size 120 × 60 cm2 ) at Sulfurcell in Berlin, Germany. Sulfurcell uses the process sequence shown in Figure 6.3. Reproduced with the permission of Dr. Nikolaus Meyer, Hahn-Meitner-Institut Berlin GmbH. 6.5 SCALING UP AND PRODUCTION Efﬁcient chalcopyrite based photovoltaic modules were demonstrated years ago . Despite this early proof-of-concept, the status has been limited to pilot lines (Figures 6.13 and 6.14), or announcements of such, for a long time. Consequently, chalcopyrites do still not play a signiﬁcant role in the marketplace. Nevertheless, industrial laboratories have always had a major impact on development and an increasing number of companies world wide is involved in the development of commercial products for the power market as well as niche applications. The ﬁrst commercial products, announced by Siemens Solar Industries in 1998, were small Figure 6.14 Large area in-line sputter coater (Von Ardenne Anlagentechnik, Dresden, Germany) for the deposition of molybdenum and zinc oxide at the Wuerth Solar pilot line in Marbach, Germany. Reproduced with the permission of Von Ardenne Anlagebtechnick GmbH. CHALCOPYRITE BASED SOLAR CELLS 255 modules with 5 and 10 W rated output power manufactured on a pilot line in Camarillo, CA. The biggest producers today are Shell Solar and Wuerth Solar. Shell Solar is now operating the production line in Camarillo which has been upgraded with respect to production capacity and product size. The largest Shell Solar modules are rated at 40 W, the Wuerth Solar modules go up to 80 W due to their larger panel size (60 × 120 cm2 ). Both, Shell Solar and Wuerth Solar have also published yield data and efﬁciency distribution of their (pilot) production which clearly demonstrate the feasibility of mass production of chalcopyrite based modules. Wuerth Solar has announced a new factory with a rated production volume of 15 MWp/a from 2007 onwards . Nevertheless, the transition from laboratory to large scale manufacturing has, in general, been more difﬁcult than expected. According to  the main obstacles have been: r Commercial scale equipment. r Quality control and in situ monitoring. r Uniformity. r Low open circuit voltage. r Throughput. r Stability. For some process steps the machines for large area deposition had to be custom designed based on experience with lab scale equipment which is an expensive and error prone pro- cess. With the increasing number of installations the equipment manufacturers are gaining experience in manufacturing these systems. The rapid commercialization of other thin ﬁlm technologies provides synergy effects. Glass companies experienced in very large area thin ﬁlm coating of glass (low emission coating) are using this experience to offer glass substrates already coated with diffusion blocker and molybdenum. Transparent conductive oxide sput- tering machines (Figure 6.14) can be derived from those sold for preparation of TCO coatings for ﬂat panel liquid crystal and organic displays. Equipment manufacturers are now actively participating in the establishment of pilot production lines and it is to be expected that turn-key production facilities with process and product speciﬁcation will be available in the not too distant future. Criteria for go/no go classiﬁcation after each individual process step and methods for quality control by in situ monitoring had been neglected for a long time in basic research. However, signiﬁcant progress has been made in recent years as more research laboratories have also become involved in scaling up activities. Examples include Raman spectroscopy  and photoluminescence decay  to assess the absorber quality, and X-ray ﬂuorescence  for in system stoichiometry measurements. Laser light scattering and substrate temperature monitoring have already been described in Section 6.3.1. Uniformity or low open circuit voltage appear to be no longer a general problem as indicated by the module efﬁciencies that have been achieved in practice (Table 6.2). Throughput in sequential processes has been improved by the development of the rapid thermal annealing processes. For evaporation technology a feasible approach appears to be to scale up the sources for increased width of deposition. Further throughput improvements are desirable for large scale production. Minimizing the required thickness of individual layers 256 THIN FILM SOLAR CELLS Table 6.2 Efﬁciency of selected minimodules and full size modules Company or Institute Size Efﬁciency Remarks Hahn-Meitner-Institut  5 × 5 cm2 9.7 % Se free CIS Uppsala University  5 × 5 cm2 16.6 % With grid Showa Shell  30 × 30 cm2 14.2 % Cd free Shell Solar  60 × 90 cm2 13.1 % Global Solar Energy  7085 cm2 10.1 % Not monolithically interconnected, cells on metal foil Wuerth Solar  60 × 120 cm2 12.2 % , alternative methods for patterning, eliminating the buffer layer from the module structure , and high rate reactive TCO sputtering carry potential for such improvement. Outdoor testing of modules has generally demonstrated excellent stability [98, 99]. Owing to the increasing production volume there is a growing number of installations (Figure 6.15) where the actual performance  and long term stability can be assessed (Figure 6.16:). Accelerated lifetime testing, especially the damp heat testing procedure which forms a part of the EN/IEC 61646 certiﬁcation, has, however, been cumbersome [101, 102]. Partly, this is due to transient effects which occur during stress tests. These can lead to an apparent degradation, however, the efﬁciency recovers after several days of light soaking. The exact Figure 6.15 Solar Tower, Training and Technology Center Handwerkskammer Heilbronn, Germany. c Wuerth Solar frameless CIS Fa¸ ade modules with a total nominal power of 8 kWp (STC) were installed in April 2001. The installation comprises 120 modules with the dimensions 60 cm by 120 cm and 40 modules with the dimensions 40 cm by 60 cm. Reproduced with permission of Wurth Solar GmbH & Co.KG. CHALCOPYRITE BASED SOLAR CELLS 257 Figure 6.16 Results of long term outdoors measurements of CIGSSe based module prototypes. Modules installed at the National Renewable Energy Lab (NREL) outdoor test facility (OTF) in Golden, Colorado . Reproduced with permission of Shell Solar Industries. causes for degradation are still under investigation but empirical optimization has already achieved modules which have been independently certiﬁed [6, 99]. In conclusion, technical problems have been overcome to a large extent. On the other hand, they still contribute to the ﬁnancial risk. As long as specialized deposition equipment is produced only in small numbers and turn-key facilities are not available, the technology development has to be taken step by step. This implies a considerable lead time before the production volumes are high enough to achieve competitive production costs in relation to the conventional silicon modules produced in large scale facilities. Innovative niche market products which exploit inherent features of chalcopyrite thin ﬁlm devices not readily available with silicon can be sold at higher prices. While they can offer a faster return on investment for smaller (start up) companies, the true medium and long term beneﬁt, as acknowledged by independent studies, lies in the cost reduction potential exploitable only through mass production for the power market. 6.5.1 Cost estimations An early cost study  predicted that chalcoypyrite based modules could be manufactured at 0.6 € /Wp in a plant with 60 MWp annual capacity whereas even the most cost effective multicrystalline silicon based technology would require a 500 MWp/a factory to achieve sim- ilar costs [104, 105]. Another study  comparing the direct module manufacturing costs of single and multicrystalline silicon as well as amorphous silicon, CdTe, and chalcopyrites estimated the cost to be 2.25 $/Wp for a 10 MWp/a chalcopyrite production line. This was lower than the cost for all other technologies, except for multicrystalline silicon estimated at 2.10 $/Wp. The estimation was based on the assumption of 9 % average module efﬁciency and 65 % production yield for the chalcopyrite modules, numbers that are clearly more favourable in 258 THIN FILM SOLAR CELLS today’s pilot production. It was estimated that a 100 MWp/a production capacity at improved average efﬁciency and yield could result in costs of 1 $/Wp, 15 % lower than that of mul- ticrystalline silicon. In general, the study was based on a ﬁrst generation baseline process and additional cost beneﬁts are expected from current technology updates. A more recent estima- tion for CuInS2 modules  already claims manufacturing costs of 1.5 €/Wp for a small production line (5 MWp/a). 6.5.2 Module performance It has been pointed out already that chalcopyrite cells offer a very high efﬁciency potential. Monolithic integration and large area nonuniformity cause only small losses, therefore high efﬁciencies could also be demonstrated for test structures (minimodules) and even full size modules (Table 6.2). Considering mass production, the distribution of efﬁciencies (or power output for a given module size) is much more relevant than the record efﬁciency of a single module. Wuerth Solar reports  an output of 79.9 ± 2.2 Wp for a batch of 306 modules (60 × 120 cm2 ) indicating that a very narrow distribution curve is feasible. In 2004 their production yield was more than 80 % with an average module efﬁciency slightly higher than 11 % . In 2002 Shell Solar reported  an average efﬁciency of 10.9 % for nearly 16 000 laminates produced in Camarillo (Figure 6.17). Considering the application the rated power output is an important point. However, the module has to perform well not only at the standard testing conditions applied for measur- ing the rated power but also at conditions typically encountered in operation, such as lower illumination intensity, varying spectral distribution, partial shading and elevated temperatures (performance ratio). Chalcopyrite modules have shown high power output under these con- ditions. The typical interconnection scheme with a large number of narrow cells reduces the Figure 6.17 Distribution of 1 × 4 laminates produced in by Shell Solar in 2002. Data includes some 15 785 laminates. Gaussian ﬁt has an average of 10.9 ± 0.6 %. Shown for comparison are the lower speciﬁcation limits for the product family . Reproduced with permission of Shell Solar Industries. CHALCOPYRITE BASED SOLAR CELLS 259 impact of partial shading. Shunts have been reduced resulting in better performance at low illumination intensity. Higher absorber bandgaps limit the losses with increasing module tem- perature . Consequently, it has been observed that chalcopyrite modules can outperform silicon modules in terms of the annual energy output on a kWh/kWp basis [99, 109, 111]. 6.5.3 Sustainability Sustainability has many aspects and while not all relevant issues are completely clariﬁed at this point the available data are quite promising (see below). Due to the high efﬁciency and performance ratio, the long lifetime, the low material consumption, and low thermal budget a chalcopyrite module will have a favourable energy balance. There are a number of research efforts  addressing speciﬁcally the sustainability of mass production. Like any large scale deployment, mass production of chalcopyrite PV modules will result in production related waste materials, energy consumption, and raw material depletion. Recycling of waste and modules at the end of their lifetime is mandatory for a sustainable technology. In addition, the producer may at some time be legally required to take back modules as is the case already for electronic products in Europe . 220.127.116.11 Availability of raw materials and recycling Thin ﬁlm technologies make very efﬁcient use of raw materials. While 0.5–1 kg/m2 of semi- conductor grade silicon are required for a conventional module, the material consumption per square meter for the active ﬁlms of a CISe module is given as: 7–20 g molybdenum, 1.5–4 g copper, 3–9 g indium, 7–20 g selenium, and 1–3 g zinc (depending on the exact module struc- ture and yield ). This implies that the total material input is comparable to the material used for just the grid metallization of silicon modules. Nevertheless, it has been argued that indium is a bottleneck concerning the abundance of raw materials. In 2003 it was used mainly for coatings (65 %), solders and alloys (15 %), and electrical components (10 %) . Indium based coatings are used in the production of ﬂat panel displays where indium tin oxide (ITO) is used as a transparent contact. The annual world production of indium, mainly from zinc ores, is in the order of 300 tons, which translates into chalcopyrite modules with approximately 15 GWp/a in total. On the other hand, indium is about three times more abundant in the Earth’s crust than silver, the latter having an annual production of 20 000 tons and a reserve base of 570 000 tons . These numbers imply that the availability of indium is not likely to be an ultimate limiting factor. Moreover, it has already been shown that even thinner active ﬁlms in a chalcopyrite module are feasible . The ﬂat panel industry may replace ITO by the cheaper ZnO in future and indium recycling will, in addition, contribute to higher availability and lower market prices. Indium free absorbers are under development (see Section 6.6.3). Within the last three decades the prices for indium have been varying over a wide range from below 100 $/Kg to more than 500 $/Kg. However, even the latter price would imply that indium is responsible for only approximately 2 % of the module manufacturing costs. Waste from the dry processes consists of material that is deposited onto chamber walls, shutters, substrate carriers etc. Recycling should be possible, but the amounts are probably too low to make it attractive unless mass production has started. Sputter targets would presumably 260 THIN FILM SOLAR CELLS be returned to the manufacturer to reclaim the remaining raw materials. Research has put more emphasis on recycling the waste generated by wet chemical buffer layer deposition [117, 118]. This could be partly done directly at the production site by removing reaction products from the solution then readjusting the concentrations and feeding the solution back to the process. A complete process sequence for disassembling and recycling of off spec modules and semiproducts has been tested successfully and could also be used for end-of-life modules . The module is heated to 250 ◦ C which softens the EVA encapsulation and allows a mechanical removal of the cover glass. Window and buffer layers are etched away by a mild acidic solution. The chalcopyrite ﬁlm is scraped from the back contact and ﬁnally the latter is dissolved in nitric acid. The test conﬁrms the feasibility of disassembling the module layer by layer which results in low cross contamination. It has been suggested that these relatively clean materials, such as the chalcopyrite powder, could be used directly in adapted preparation processes. It has already been shown that the substrate glass can simply be reused for new modules. 18.104.22.168 Energy payback time A comparative study on energy payback times has been assisted by the only company that produces silicon as well as chalcopyrite modules . It was concluded that the EPBT for the chalcopyrite module is 1.8 years as compared to 3.3 years for the silicon module. It should be noted that the aluminium frame is responsible for a large fraction of the materials energy content of the chalcopyrite module. The study was also based on ﬁrst generation chalcopyrite technology and signiﬁcantly lower EPBT should be feasible. 6.6 DEVELOPING FUTURE CHALCOPYRITE TECHNOLOGY The chalcopyrite module is still under active international development. Progress in fundamen- tal understanding and preparation technology will result in signiﬁcant improvements in market potential, module performance, sustainability, and minimized ecological impact of large scale production of next generation modules. We will highlight areas of ongoing development in the following paragraphs. 6.6.1 Lightweight and ﬂexible substrates Transferring the technology of chalcopyrite based solar cells from rigid glass substrates to ﬂexible, low-mass substrates  opens new market segments. Furthermore, ﬂexible substrates are a requirement for roll-to-roll processing which could lower production costs . Such substrates can be plastic or metal foils (Figure 6.18). When combined with foil substrates, the low mass, excellent radiation hardness, and relatively high (compared to other thin ﬁlm technologies) efﬁciency make chalcopyrite based technology the leading candidate for thin ﬁlm cells for space applications . Available plastic foils do not tolerate the high substrate temperatures ordinarily used for chalcopyrite preparation. Common problems at high temperatures are shrinkage, warping, outgassing, and loss of ﬂexibility. Lowering the substrate temperature is possible, but the CHALCOPYRITE BASED SOLAR CELLS 261 Figure 6.18 Prototype of a ﬂexible CIGSe solar cell for space applications on a thin titanium substrate . efﬁciencies that can be achieved are, in consequence, somewhat lower. The best results, still below 15 % efﬁciency, are achieved on polyimide foils . On the other hand, the conductivity of metal foils raises new challenges with respect to pin hole free processing and monolithic integration. Some metals also tend to diffuse into the absorber and cause deterioration of its properties. Isolating and/or diffusion blocking layers are under development to circumvent these problems. In another approach monolithic integration is abandoned altogether. Individual medium sized cells are fabricated and interconnected using a shingling scheme. Sodium doping is mandatory with any foil substrate and can be carried our through a precursor ﬁlm, coevaporation or even diffusion after ﬁlm preparation. 6.6.2 Cadmium free cells The CdS buffer layer is very thin and contamination of the environment from modules is very unlikely even in extreme conditions. Nevertheless, cadmium is a hazardous material and its elimination from the module may increase the general acceptance of the product and reduce production costs (by avoiding costly safety measures). According to European Union directives  several heavy metals (including cadmium) must be not be contained in new electrical and electronic equipment after July 2006. While photovoltaic modules do not currently fall under these regulations  they illustrate the general effort to reduce the amount of cadmium in circulation. Owing to the intensive work on cadmium free buffer layers there is a variety of possible alternatives. We will introduce a small selection in this chapter. More information can be found in a recent review article . 22.214.171.124 Wet chemical processes Chemical bath deposition is not restricted to the deposition of CdS. The most common re- placements are indium or zinc based sulﬁdes, oxides, hydroxides, or mixtures thereof. The 262 THIN FILM SOLAR CELLS achievable efﬁciency is very close to that of the standard device . However, the deposition parameters appear to be more critical and need to be adapted to the speciﬁc absorber. Accord- ingly, it can be challenging to achieve a high reproducibility of cell performance. Cadmium free cells are often found to exhibit metastability effects, e.g., signiﬁcant improvement with light soaking . It is not yet fully clear how far the long term stability of the cell is also affected by the modiﬁed buffer layer. Only one of the current pilot production efforts is known to use a zinc based buffer layer (Table 6.2). Novel chemical deposition methods such as ILGAR (ion layer gas reaction) have been developed and have been applied successfully mostly to those Cd free alternative buffer layers . There are indications that ZnO deposited by ILGAR can combine the functions of the buffer and the (normally sputtered) undoped part of the window layer . 126.96.36.199 Dry processes Wet chemical deposition is often believed to be signiﬁcantly cheaper than other deposition methods. However, in a typical processing sequence, buffer layer deposition is the only wet chemical process, which implies that the whole infrastructure has to be implemented for just this single process step. In addition, it prevents true in-line processing. While this is currently not a big issue because the deposition systems are typically not connected and scribing is also carried out in atmosphere, it may be a drawback in view of future large volume production. Dry processes are therefore under investigation. The most successful approaches have been MOCVD  and ALCVD (atomic layer chemical vapor deposition ). The former has so far produced only small samples, albeit with good efﬁciency, whereas the latter has been shown to work at least on medium sized substrates (30 × 30 cm2 ). Atomic layer chemical vapor deposition is, however, an inherently slow process and therefore only cost-effective if carried out in systems that can accommodate a large number of substrates simultaneously, in contradiction to the requirements for an in-line capable process. Buffer layers can also be evaporated which appears particularly attractive if the absorber is also prepared by evaporation . Efﬁciencies of almost 15 % can be achieved provided the buffer layer is evaporated insitu, i.e., without breaking vacuum after the absorber deposition . Evaporation, MOCVD, and ALCVD, in analogy to CBD, are ‘soft’ deposition methods and there have been assumptions that this is a major advantage. On the other hand, sputtering of a Cd free buffer layer, if feasible, is clearly a technically very attractive solution. Sputtering is, however, not a soft method and it has been speculated that one of the main functions of the buffer layer consists of protecting the absorber surface from sputter damage during window layer deposition. This assumption is supported by the typically poor device performance observed when simply omitting the buffer layer and sputtering the TCO directly onto the absorber. It was shown, however, that a chemical treatment (partial electrolyte) is, in principle, sufﬁcient to stabilize the absorber surface for window layer sputtering . Only a very thin ﬁlm is deposited under the conditions of this chemical treatment, which would not be sufﬁcient to prevent sputter damage. More recently it has been shown that an efﬁciency of more than 12 % can be achieved by a sputtered Inx S buffer layer without chemical treatment . A different novel approach to heterojunction formation originates from the question whether the window layer can be modiﬁed in such a way that a buffer layer is no longer necessary. Based on the assumption that the band line up at the chalcopyrite/TCO direct junction is causing the CHALCOPYRITE BASED SOLAR CELLS 263 poor performance, experiments have been carried out using (Zn,Mg)O rather than pure ZnO in the ﬁrst part of the TCO double layer stack for improved band alignment. Kelvin probe force microscopy measurements of the work function of solar cell cross sections support this model . In terms of cell efﬁciency (12.5 %), the proof-of-concept could be achieved , but further development is necessary to demonstrate long term stability and reproducibility in pilot production. It has been shown that the alloy can be deposited by sputtering from a single mixed target and that its conductivity is comparable to the standard undoped ZnO. Hence, sputtering of (Zn,Mg)O is a simple drop-in replacement and can be integrated easily into existing pilot lines to prepare standard (with CBD buffer) modules and to investigate modules without buffer layers in the same setup. 6.6.3 Indium free absorbers It is sometimes argued that the lack of abundancy of indium may be limiting the long term per- spectives of chalcopyrite based photovoltaics (see Section 188.8.131.52). This has triggered research on In-free absorbers. Compounds such as Cu2 ZnSn(S,Se)4 crystallize in the kesterite structure which can be derived from the chalcopyrite structure by replacing half of the In atoms by Zn and the other half by Sn atoms. It has been shown that the synthesis of crystals and thin ﬁlms can be carried out using methods well known from chalcopyrite preparation [132, 133]. Also, heterojunction solar cells can be prepared using the established contact layers (molybdenum back contact, CdS buffer etc.). On the other hand, secondary phases are more problematic in this quarternary system as compared to ternary chalcopyrite. Cell efﬁciencies achieved so far are in the range of 5 %  and indicate that research efforts would have to be stepped up signiﬁcantly to make kesterites a feasible alternative to chalcopyrite absorbers. 6.6.4 Novel back contacts Molybdenum appears to be an almost ideal contact material at least for current typical cell structures and preparation methods. Corrosion of molybdenum  is one of the few known issues and has been reported to contribute to module degradation in accelerated lifetime testing. It is conceivable that the stability could be improved by using molybdenum based alloys instead of pure molybdenum. Another disadvantage of molybdenum is its poor optical reﬂection which may become relevant in view of efforts to reduce the absorber thickness. The choice of other metals is limited due to their instability under typical absorber deposition conditions. Tungsten seems to yield a good ohmic contact but its optical properties in this respect are poor. Tantalum and niobium have slightly higher reﬂections and preliminary studies  indicate that they may be feasible in terms of contact performance. Transparent conductive oxide coated metal contacts may be an alternative solution to achieve stability, and good electrical as well as optical performance for thin cells. Ohmic TCO/chalcopyrite contacts are also required for novel structures such as bifacial and (semi) transparent cells (see below). 6.6.5 Bifacial cells and superstrate cells In a variant of the cell structure the nontransparent rear metal contact is replaced by a TCO ﬁlm . Such a bifacial cell (Figure 6.19) has advantages in certain applications. At the current 264 THIN FILM SOLAR CELLS Substrate Superstrate Bi-facial Blocking Contact Ohmic Contact TCO Chalcopyrite Glass Metal Figure 6.19 Schematic cross sectional views of chalcopyrite solar cell conﬁgurations. development state the efﬁciency for illumination through the back contact is signiﬁcantly lower because the carriers are generated outside the ﬁeld zone in proximity to the poorly passivated contact (poor blue response). Further optimization of absorber thickness, diffusion length, and contact passivation appears to be feasible. In the conventional module structure the nontransparent back contact is deposited onto the glass substrate. The module is illuminated through the encapsulation material which has to be transparent (usually a second sheet of glass). It is, in principle, possible to reverse the cell structure by starting with the deposition of the transparent contact (Superstrate conﬁgura- tion, Figure 6.19). In this case the TCO/(buffer)/chalcopyrite interface needs to be a blocking junction. The light enters the cell through the superstrate which has the advantage that the module can be encapsulated with nontransparent material of lower mass and lower cost. The efﬁciency in this structure is reduced due to junction degradation during the high temperature absorber preparation and/or inferior quality of the absorber deposited at lower temperatures. The proof-of-concept is limited to small area cells [138–140]. Heterojunctions can also be prepared on TiO2 ﬁlms . This material can be grown so that it is nanostructured (porous) and is better known from its application in dye sensitized cells. Filling the pores of TiO2 with a chalcopyrite semiconductor has been proposed as a realization of the ETA (extremely thin absorber) concept . The recombination velocity at the TiO2 /chalcopyrite interface must be kept low in view of the very large interface area in this structure. It has been found that blocking or buffer layers are useful in achieving better interface properties [143, 144]. 6.6.6 Nonvacuum processing Nonvacuum processing is being investigated because it may reduce the up front investment required to implement a certain process step. This lowers the initial barrier and ﬁnancial risk but does not necessarily imply lower overall production costs. The running costs may be actually be higher due to the mix of vacuum and nonvacuum equipment and corresponding CHALCOPYRITE BASED SOLAR CELLS 265 infrastructure requirements, larger amounts of waste generated in wet chemical processing, etc. There is no known vacuum-free alternative for the preparation of the molybdenum back contact, however, metallized glass may be bought from the glass industry and does not have to be prepared on site. Nonvacuum buffer (CBD) and window layer (MOCVD) deposition is already established. Nonvacuum absorber processes are commonly based on sequential processes similar to those described in Section 184.108.40.206 above. They comprise a precursor deposition at room temperature followed by (reactive) annealing at elevated temperature. This second stage of sequential processing, i.e., selenization in mixtures of H2 Se and inert gas is an already established nonvacuum process. Electrodeposited metal precursors were investigated early in the history of chalcopyrite solar cells  and there appears to be a renewed interest  in spite of the moderate lab scale efﬁciencies achieved. Electrodeposition of a metal/Se precursor is the basis of a process developed with strong industrial participation and so far yielding small area cells with 10 % efﬁciency . Higher efﬁciencies could be achieved by another group with selenium containing precursors, but only after manipulating the precursor composition by means of conventional evaporation . Another interesting approach is the preparation of inks or slurries which can be deposited by spraying, printing, dip coating, or doctor blading. In one approach  the metals are dissolved in acid and hydroxide nanoparticles are precipitated from this solution. The particles are dried, which results in a ﬁne powder of mixed oxides. The powder is dispersed in an aqueous solution to obtain the ink. In this and similar processes the ﬁlm composition can be controlled precisely and with relative ease by adjusting the relative amounts of material used for particle preparation. The oxide particle precursor needs an additional reduction step prior to selenization carried out in diluted hydrogen. The precursor ﬁlm can be porous because the volume increase during selenization leads to sufﬁcient densiﬁcation . In principle it would be possible to include selenium (and/or sulfur) in the precursor particles and to use nonreactive sintering, but results from various approaches were not convincing. 6.6.7 Wide gap and tandem cells Wide gap cells are interesting for two reasons: better single junction cells and top cells for tandem conﬁguration. A moderate increase of the bandgap to about 1.4 eV for a single junc- tion cell places it at the theoretical maximum of the bandgap/efﬁciency relation for the solar spectrum . Because of the reduced current density, a higher resistance of the TCO can be tolerated. At the same time the doping of the TCO can be at the upper limit because the absorber cut off wavelength is below the onset of free carrier absorption in the TCO. In consequence the TCO can be made thinner which leads to cost reductions. Compared to low gap absorbers, the loss of performance at higher module temperature is much less severe which leads to improved annual output in terms of kWh/kWp (performance ratio) especially in hot climate. Tandem cells could eventually lead to an efﬁciency exceeding even the theoretical limit for silicon cells. Low gap CIG(S)Se cells (E g ≈ 1.1 eV), such as NREL’s 19.5 % efﬁcient cell, are ideally suited as the bottom cell. The top cell absorber should have a bandgap of roughly 1.7 eV  and the front as well as the back contacts need to be highly transparent. Ideally, the top cell would be deposited directly onto the bottom cell (monolithic tandem, Figure 6.20). This, however, requires fundamental changes to the chalcopyrite technology. The top cell, after laying down the window layer, does not tolerate temperatures above 200 ◦ C. This maximum allowable temperature would have to be raised signiﬁcantly and low temperature 266 THIN FILM SOLAR CELLS Mechanically Stacked Monolithic Low gap Wide gap TCO Chalcopyrite Glass Metal EVA Blocking Contact Ohmic Contact Figure 6.20 Schematic cross sectional views of tandem conﬁgurations. processing would have to be implemented for the deposition of the top cell. It is therefore more likely that the ﬁrst tandem modules will be mechanically stacked after individual preparation of top and bottom cells, respectively. If the top module were prepared in analogy to the bifacial cell described above this would imply a laminate with three sheets of glass. If the top module had a superstrate structure, top and bottom could be laminated together face-to-face and, like in the conventional single junction module, only two sheets of glass would be necessary for an encapsulated module (Figure 6.20, right hand side). In practice the realization of highly efﬁcient wide gap cells has not yet been achieved. As pointed out previously, this may have partly historical reasons. On the other hand, signiﬁcant differences in parameters determining the cell efﬁciency have been theoretically predicted  and are observed in the experiment . It is not fully clear which of these are the most crucial and how they could be resolved. This is expected to change as the development of widegap cells and tandems is one of the focus points of international basic research. Preliminary prototypes of monolithic  as well as mechanically stacked  tandems are described in the literature. Due to the problems in realizing a highly efﬁcient wide gap chalcopyrite solar cells it has been suggested to use a different material such as amorphous silicon  or CdTe  for the top cell. REFERENCES  J. E. Jaffe, A. Zunger, Electronic structure of the ternary chalcopyrite semiconductors CuAlS2 , CuGaS2 , CuGaSe2 and CuInSe2, Phys. Rev. B, 29 (1984) 1882.  S. Wagner, J. L. Shay, P. Migliorato, H. M. Kasper, CuInSe2 /CdS heterojunction photovoltaic detectors, Appl. Phys. Lett. 25, (1974) 434. CHALCOPYRITE BASED SOLAR CELLS 267  K. Ramanathan, M. A. Contreras, C. L. Perkins, S. Asher, F. S. Hasoon, J. Keane, D. Young, M. Romero, W. Metzger, R. Nouﬁ, J. Ward, A. Duda, Improved performance in ZnO/CdS/CuGaSe2 Thin-ﬁlm solar cells, Prog. Photovolt: Res. Appl,. 11 (2003) 225. a  J. Kessler, J. Wennerberg, M. Bodeg˚ rd, L. Stolt, Highly efﬁcient Cu(In,Ga)Se2 mini-modules, Sol. En. Mat. Sol. Cells, 75 (2003) 35.  K. Kushiya, Progress in large-area Cu(InGa)Se2 -based thin-ﬁlm modules with the efﬁciency of over 13 %, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 319.  J. Palm, V. Probst, W. Stetter, R. Toelle, S. Visbeck, H. Calwer, T. Niesen, H. Vogt, O. Hernandez, M. Wendl, F. H. Karg, CIGSSe thin ﬁlm PV modules: from fundamental investigations to advanced performance and stability, Thin Solid Films, 451–452 (2004) 544.  M. Powalla, B. Dimmler, New developments in CIGS thin-ﬁlm solar cell technology, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 313. u  R. Klenk, J. Klaer, R. Scheer, M. Ch. Lux-Steiner, I. Luck, N. Meyer, U. R¨ hle, Solar cells based on CuInS2 – an overview, Thin Solid Films, 481–482 (2005) 509.  R. Klenk, T. Walter, H.-W. Schock, D. Cahen, Model for successful growth of polycrystalline ﬁlms of CuInSe2 by multisource physical vacuum evaporation, Adv, Mat., 5 (1993) 114.  R. A. Mickelsen, W. S. Chen, Apparatus for forming thin-ﬁlm heterojunction solar cells employing materials selected from the class of I-III-VI2 chalcopyrite compounds, US Patent 4 392 451 (1983).  A. M. Gabor, J. R. Tuttle, D. S. Albin, M. A. Contreras, R. Nouﬁ, A. M. Hermann, High-efﬁciency CuInx Ga1−x Se2 solar cells made from (Inx ,Ga1−x )2 Se3 precursor ﬁlms, Appl. Phys. Lett., 65 (1994) 198.  N. Kohara, T. Negami, M. Nishitani, T. Wada, Preparation of device-quality Cu(In,Ga)Se2 thin ﬁlms deposited by coevaporation with composition monitor, Jpn. J. Appl. Phys., 34 (1995) L1141.  K. Sakurai, R. Scheer, C. A. Kaufmann, A. Yamada, P. Fons, Y. Kimura, T. Baba, K. Matsubara, S. Niki, Process variants for Cu(In1−x Gax )Se2 deposition investigated by in situ spectroscopic light scattering, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 384.  B. M. Basol, V. J. Kapur, CuInSe2 thin ﬁlms and high-efﬁciency solar cells obtained by selenization of metallic layers, Conf. Rec. 21st IEEE Photovoltaic Specialists Conf., Kissimmee, USA, 21–25 May 1990, IEEE, New York (1990) 546. a  J. Klaer, J. Bruns, R. Henninger, K. Siemer, R. Klenk, K. Ellmer, D. Br¨ unig, Efﬁcient CuInS2 thin-ﬁlm solar cells prepared by a sequential process, Semicond. Sci. Technol., 13 (1998) 1456.  F. Karg, V. Probst, H. Harms, J. Rimmasch, W. Riedl, J. Kotschy, J. Holz, R. Treichler, O. Eibl, A. Mitwalsky, A. Kiendl, Novel rapid-thermal-processing for cis thin-ﬁlm solar cells, Conf. Rec. 23rd IEEE Photovoltaic Specialists Conf., Louisville, USA, 10–14 May 1993, IEEE, New York (1993) 441. a  K. Siemer, J. Klaer, I. Luck, J. Bruns, R. Klenk, D. Br¨ unig, Efﬁcient CuInS2 solar cells from a rapid thermal process (RTP), Sol. En. Mat. Sol. Cells, 67 (2001) 159.  D. Wolf, G. M¨ ller, Kinetics of CIS-formation studied in-situ by thin ﬁlm calorimetry, Thin Solid u Films, 361–362 (2000) 155.  R. Scheer, R. Klenk, J. Klaer, I. Luck, CuInS2 based thin ﬁlm photovoltaics, Sol. En., 77 (2004) 777. a  I. Luck, J. Kneisel, K. Siemer, J. Bruns, R. Scheer, R. Klenk, N. Janke, D. Br¨ unig, Inﬂuence of Na on the properties of Cu-rich prepared CuInS2 thin ﬁlms and the performance of corresponding CuInS2 /CdS/ZnO solar cells, Sol. En. Mat. Sol. Cells, 67 (2001) 151. a  M. Ruckh, D. Schmid, M. Kaiser, R. Sch¨ fﬂer, T. Walter, H. W. Schock, Inﬂuence of substrates on the electrical properties of Cu(In,Ga)Se2 thin ﬁlms, Sol. En. Mat. Sol.Cells, 41/42 (1996) 335. 268 THIN FILM SOLAR CELLS  D. Rudmann, M. Kaelin, F.-J. Haug, F. Kurdesau, H. Zogg, A. N. Tiwari, Impact of Na on struc- tural properties and interdiffusion of CuInSe2 and CuGaSe2 thin ﬁlms, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 376.  J. H. Scoﬁeld, A. Duda, D. Albin, B. L. Ballard, P. K. Predecki, Sputtered molybdenum bilayer back contact for copper indium diselenide-based polycrystalline thin-ﬁlm solar cells, Thin Solid Films, 260 (1995) 26.  W. N. Shafarman, J. E. Phillips, Direct current-voltage measurements of the Mo/CuInSe2 contact on operating solar cells, Conf. Rec. 25th IEEE Photovoltaic Specialists Conf., Washington, USA, 13–17 May 1996, IEEE, New York (1996) 917.  M. Burgelman, F. Engelhardt, J. F. Guillemoles, R. Herberholz, M. Igalson, R. Klenk, M. Lam- pert, T. Meyer, V. Nadenau, A. Niemegeers, J. Parisi, U. Rau, H. W. Schock, M. Schmitt, O. Seifert, T. Walter and S. Zott, Defects in Cu(In,Ga)Se2 semiconductors and their role in the device performance of thin-ﬁlm solar cells, Prog. Photovolt: Res. Appl., 5 (1997) 121.  F. Kessler, D. Rudmann, Technological aspects of ﬂexible CIGS solar cells and modules, Sol. En., 77 (2004) 685.  R. H. Mauch, J. Hedstrom, D. Lincot, M. Ruckh, J. Kessler, R. Klinger, L. Stolt, J. Vedel, H. W. Schock, Optimization of windows in ZnO-CdS-CuInSe2 heterojunctions, Conf. Rec. 22nd IEEE Photovoltaic Specialists Conf., Las Vegas, USA, 7–11 Oct 1991, IEEE, New York (1991) 898.  U. Rau, M. Schmidt, Electronic properties of ZnO/CdS/Cu(In,Ga)Se2 solar cells – aspects of heterojunction formation, Thin Solid Films, 387 (2001) 141.  U. Rau, P. O. Grabitz, J. H. Werner, Resistive limitations to spatially inhomogeneous losses in solar cells, Appl. Phys. Lett., 85 (2004) 6010.  K. Ellmer, Resistivity of polycrystalline zinc oxide ﬁlms: current status and physical limit, J. Phys. D: Appl. Phys., 34 (2001) 3097. a o  Ch.-H. Fischer, M. B¨ r, A. Grimm, I. K¨ tschau, I. Lauermann, J. Reichardt, S. Sokoll, M. Aggour, H.-J. Lewerenz, M.C. Lux-Steiner, L. Weinhardt, O. Fuchs, C. Heske, C. Jung, W. Gudat, T. P. Niesen, S. Visbeck, F. Karg, X-rays shed light on the ’hidden’ interfaces of solar cells, BESSY- Highlights 2003 (2004) 14.  M. Burgelman, A. Niemegeers, Calculation of CIS and CdTe module efﬁciencies, Sol. En. Mat. Sol. Cells, 51 (1998) 129.  U. Rau, H. W. Schock, Electronic properties of Cu(In,Ga)Se2 heterojunctions solar cells-recent achievements, current understanding, and future challenges, Appl. Phys. A, 69 (1999) 131.  Th. Glatzel, H. Steigert, S. Sadewasser, R. Klenk, M.Ch. Lux-Steiner, Potential distribution of Cu(In,Ga)(S,Se)2 -solar cell cross-sections measured by Kelvin probe force microscopy, Thin Solid Films, 481–482 (2005) 177.  R. Klenk, Characterisation and modelling of chalcopyrite solar cells, Thin Solid Films, 387 (2001) 135.  M. Burgelman, P. Nollet, S. Degrave, Modelling polycrystalline semiconductor solar cells, Thin Solid Films, 361–362 (2000) 527.  R. Scheer, Surface and interface properties of Cu-chalcopyrite semiconductors and devices, Trends Vac. Sci. Technol., 2 (1997) 77.  D. Schmid, M. Ruckh, H.-W. Schock, A comprehensive characterization of the interface in Mo/CIS/CdS/ZnO solar cell structures, Sol. En. Mat. Sol. Cells, 41/42 (1996) 281. u  M. Morkel, L. Weinhardt, B. Lohm¨ ller, C. Heske, E. Umbach, W. Riedl, S. Zweigart, F. Karg, Flat conduction-band alignment at the CdS/CuInSe2 thin-ﬁlm solar-cell heterojunction, Appl. Phys. Lett. 79 (2001) 4482. a  T. Schulmeyer, R. Kniese, R. Hunger, W. J¨ germann, M. Powalla, A. Klein, Inﬂuence of Cu(In,Ga)Se2 band gap on the valence band offset with CdS, Thin Solid Films, 451–452 (2004) 420. CHALCOPYRITE BASED SOLAR CELLS 269  L. Kronik, L. Burstein, M. Leibovitch, Y. Shapira, D. Gal, E. Moons, J. Beier, G. Hodes, D. Cahen, D. Hariskos, R. Klenk, H.-W. Schock, Band diagram of the polycrystalline CdS/Cu(In,Ga)Se2 heterojunction, Appl. Phys. Lett., 67 (1995) 1405.  D. Schmid, M. Ruckh, F. Grunwald, H. W. Schock, Chalcopyrite/defect chalcopyrite heterojunc- tions on the basis of CuInSe2 , J. Appl. Phys., 73 (1993) 2902. o  I..M. K¨ tschau, H.W. Schock, Depth proﬁle of the lattice constant of the Cu-poor surface layer in (Cu2 Se)1−x (In2 Se3 )x evidenced by grazing incidence X-ray diffraction, J. Phys. Chem. Solids, 64 (2003) 1559.  M. Contreras, H. Wiesner, J. Tuttle, K. Ramanathan, R. Nouﬁ, Issues on the Chalcopyrite/Defect- Chalcopyrite Junction Model for High-Efﬁciency Cu(In,Ga)Se2 Solar Cells, Sol. En. Mat. Sol. Cells, 49 (1997) 239.  S. Schuler, S. Siebentritt, S. Nishiwaki, N. Rega, J. Beckmann, S. Brehme, M.Ch. Lux-Steiner, Self-compensation of intrinsic defects in the ternary semiconductor CuGaSe2 , Phys. Rev. B, 69 (2004) 45210.  S. Siebentritt, Hole transport mechanisms in CuGaSe2 , Thin Solid Films, 481–482 (2005) 312.  S. Siebentritt, S. Schuler, Defects and transport in the wide gap chalcopyrite CuGaSe2 , J. Phys. Chem. Solids, 64 (2003) 1621.  S. Sadewasser, Th. Glatzel, S. Schuler, S. Nishiwaki, R. Kaigawa, M.Ch. Lux-Steiner, Kelvin probe force microscopy for the nano scale characterization of chalcopyrite solar cell materials and devices, Thin Solid Films, 431–432 (2003) 257.  C. Perrson, A. Zunger, Anomalous grain boundary physics in polycrystalline CuInSe2 : the existence of a hole barrier, Phys. Rev. Lett., 91 (2003) 266401.  K. Taretto, U. Rau, J. H. Werner, Numerical simulation of grain boundary effects in Cu(In,Ga)Se2 thin-ﬁlm solar cells, Thin Solid Films, 481–482 (2005) 8. o  D. Fuertes Marr´ n, S. Sadewasser, A. Meeder, Th. Glatzel, M. Ch. Lux-Steiner, Electrical activity at grain boundaries of Cu(In,Ga)Se2 thin ﬁlms, Phys. Rev. B, 71 (2005) 033306.  K. Puech, S. Zott, K. Leo, M. Ruckh, H.-W. Schock, Determination of minority carrier lifetimes in CuInSe2 thin ﬁlms, Appl. Phys. Lett., 69 (1996) 3375.  N. Rega, S. Siebentritt, A. Zajogin, S. Nishiwaki, M. Ch. Lux-Steiner, R. Kniese, M. J. Romero, Excitonic luminescence of Cu(In,Ga)Se2 , Thin Solid Films, 481–482 (2005) 286.  S. Siebentritt, I. Beckers, T. Riemann, J. Christen, A. Hoffmann, M. Dworzak, Reconciliation of luminescence and Hall measurements on the ternary semiconductor CuGaSe2 , Appl. Phys. Lett., in press.  S. B. Zhang, Su-Huai Wei, A. Zunger, H. Katayama-Yoshida, Defect physics of the CuInSe2 chalcopyrite semiconductor, Phys. Rev. B, 57 (1998) 9642.  R. Scheer, M. Wilhelm, V. Nadenau, H. W. Schock, L. Stolt, Collection functions in chalcopyrite heterojunction solar cells: a comparison of Cu(In,Ga)(S,Se)2 absorbers, Proc. 14th European Photovoltaic Solar Energy Conf., Barcelona, Spain, H. A. Ossenbrink, P. Helm, H. Ehmann (Eds.), H.S. Stephens, Bedford, UK (1997) 1299.  R. Klenk, H. W. Schock, Photocurrent collection in thin ﬁlm solar cells - calculation and characteri- zation for CuGaSe2 /(Zn,Cd)S, Proc. 12th European Photovoltaic Solar Energy Conf., Amsterdam, NL, R. Hill, W. Palz, P. Helm (Eds.), Stephen & Associates, Bedford, UK (1994) 1588. a  N. Meyer, T. Dylla, D. Fischer, M. E. Beck, A. J¨ ger-Waldau, M. Ch. Lux-Steiner, Solar cells based on HCVD grown CuGaSe2 absorbers with high open-circuit voltages, Proc. 16th European Photvoltaic Solar Energy Conf., Glasgow, UK, 1–5 May 2000, H. Scheer, B. McNelis, W. Palz, H. A. Ossenbrinck, P. Helm (Eds.), James & James, Glasgow (2000) 429.  K. Orgassa, U. Rau, Quang Nguyen, H. W. Schock, J. H. Werner, Role of the CdS buffer layer as an active optical element in Cu(Ga,In)Se2 thin-ﬁlm solar cells, Prog. Photovolt. Res. Appl., 10 (2002) 457.  J. Klaer, I. Luck, A. Boden, R. Klenk, I. Gavilanes Perez, R. Scheer, Mini-modules from a CuInS2 baseline process, Thin Solid Films, 431–432 (2003) 534. 270 THIN FILM SOLAR CELLS  R. Kaigawa, A. Neisser, R. Klenk, M. Ch. Lux-Steiner, Improved performance of thin ﬁlm solar cells based on Cu(In,Ga)S2 , Thin Solid Films, 415 (2002) 266.  J. Parisi, D. Hilburger, M. Schmitt, U. Rau, Quantum efﬁciency and admittance spectroscopy on Cu(In,Ga)Se2 solar cells, Sol. En. Mat. Sol. Cells, 50 (1998) 79. o  T. Walter, R. Menner, Ch. K¨ ble, H. W. Schock, Characterization and junction performance of highly efﬁcient Zno/CdS/CuInS2 thin ﬁlm solar cells, Proc. 12th European Photovoltaic Solar Energy Conf., Amsterdam, NL, R. Hill, W. Palz, P. Helm (Eds.), Stephen & Associates, Bedford, UK (1994) 1755.  J. E. Phillips, W. N. Shafarman, E. Shan, Evidence for amorphous like behavior in small grain thin ﬁlm polycrystalline solar cells, Proc. 1st World Conf. Photovoltaic Energy Conversion, Hawaii, IEEE, New York (1994) 303.  R. Herberholz, T. Walter, H. W. Schock, Density of states in CuIn(S,Se)2 thin ﬁlms from modulated photocurrent measurements, J. Appl. Phys., 76 (1994) 2904. a  R. Sch¨ fﬂer, H. W. Schock, Composition dependent optical properties of polycrystalline CuInSe2 thin ﬁlms, Solid State Phenomena, 51–52, Scitec Publications (1996) 347. u  T. Walter, R. Herberholz, C. M¨ ller, H. W. Schock, Determination of defect distributions from admittance measurements and application to Cu(In,Ga)Se2 based heterojunctions, J. Appl. Phys., 80 (1996) 4411.  G. Hanna, A. Jasenek, U. Rau, H. W. Schock, Inﬂuence of the Ga-content on the bulk defect densities of Cu(In,Ga)Se2 , Thin Solid Films, 387 (2001) 71.  A. Jasenek, A. Boden, K. Weinert, M. R. Balboul, H. W. Schock, U. Rau, High-energy electron and proton irradiation of Cu(In,Ga)Se2 heterojunction solar cells, Mat. Res. Soc. Symp. Proc., 668, Materials Research Society (2001) H3.2.1.  J. T. Heath, J. D. Cohen, W. N. Shafarman, D. X. Liao, A. A. Rockett, Effect of Ga content on defect states in CuIn1−x Gax Se2 photovoltaic devices, Appl. Phys. Lett., 80 (2002) 4540.  U. Rau, Tunneling-enhanced recombination in Cu(In,Ga)Se2 heterojunction solar cells, J. Appl. Phys., 74 (1999) 111. o o  U. Malm, J. Malmstr¨ m, C. Platzer-Bj¨ rkman, L. Stolt, Determination of dominant recombination paths in Cu(In,Ga)Se2 thin-ﬁlm solar cells with ALD-ZnO buffer layers, Thin Solid Films, 481–482 (2005) 208.  I. Hengel, A. Neisser, R. Klenk, M. Ch. Lux-Steiner, Current transport in CuInS2 :Ga/CdS/ZnO Solar Cells, Thin Solid Films, 361–362 (2000) 458. o  J. Reiß, J. Malmstr¨ m, A. Werner, I. Hengel, R. Klenk, M. Ch. Lux-Steiner, Current transport in CuInS2 solar cells depending on absorber preparation, Mat. Res. Soc. Symp. Proc., 668, Materials Research Society (2001) H9.4.1.  V. Nadenau, U. Rau, A. Jasenek, H. W. Schock, Electronic properties of CuGaSe2 -based hetero- junction solar cells. Part I. Transport analysis, J. Appl. Phys., 87 (2000) 584.  M. Turcu, O. Pakma, U. Rau, Interdependence of absorber composition and recombination mech- anism in Cu(In,Ga)(S,Se)2 heterojunction solar cells, J. Appl. Phys., 80 (2002) 2598.  J. H. Werner, J. Mattheis, U. Rau, Efﬁciency limitations of polycrystalline thin ﬁlm solar cells: case of Cu(In,Ga)Se2 , Thin Solid Films, 481–482 (2005) 399.  U. Rau, J. H. Werner, Radiative efﬁciency limits of solar cells with lateral band-gap ﬂuctuations, Appl. Phys. Lett., 84 (2004) 3735.  J. AbuShama, R. Nouﬁ, S. Johnston, S. Ward, X. Wu, Improved performance in CuInSe2 and surface-modiﬁed CuGaSe2 solar cells, Proc. 31st IEEE Photovoltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 299.  M. A. Green, K. Emery, D. L. King, S. Igari, W. Warta, Solar cell efﬁciency tables, Prog. Photovolt: Res. Appl., 13 (2005) 49.  B. Egaas, M. A. Contreras, R. Nouﬁ, unpublished result.  C. A. Kaufmann, A. Neisser, R. Scheer, R. Klenk, M. A. Kroon, Cu(In,Ga)Se2 thin ﬁlm solar cells on ﬂexible substrates, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, CHALCOPYRITE BASED SOLAR CELLS 271 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 3637.  e D. Br´ maud, D, Rudmann, G. Bilger, H. Zogg, A. N. Tiwari, Towards the development of ﬂex- ible CIGS solar cells on polymer ﬁlms with efﬁciency exceeding 15 %, Proc. 31st IEEE Photo- voltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 223.  M.A. Contreras, T. Nakada, M. Hongo, A.O. Pudov, J.R. Sites, ZnO/ZnS(O,OH)/Cu(In,Ga)Se2 /Mo solar cell with 18.6 % efﬁciency, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 570.  o a N.A. Allsop, A. Sch¨ nmann, H.J. Mufﬂer, M. B¨ r, M.C. Lux-Steiner, Ch.H. Fischer, Spray-ILGAR indium sulﬁde buffers for Cu(In, Ga)(S, Se)2 solar cells, Prog. Photovolt. Res. Appl., submitted.  T. Glatzel, H. Steigert, R. Klenk; M. Ch. Lux-Steiner, T. P. Niesen, S. Visbeck, Zn1−x Mgx O as a window layer in completely Cd-free Cu(In, Ga)(S, Se)2 based thin ﬁlm solar cells, Sol. En. Mat. Sol. Cells, in press.  N. Naghavi, S. Spiering, M. Powalla, B. Cavana, D. Lincot, High-efﬁciency copper indium gallium diselenide (CIGS) solar cells with indium sulﬁde buffer layers deposited by atomic layer chemical vapor deposition (ALCVD), Prog. Photovolt. Res. Appl., 11 (2003) 437.  o A. Strohm, T. Schl¨ tzer, Q. Nguyen, K. Orgassa, H. Wiesner, H.W. Schock, New approaches for the fabrication of Cd-free Cu(In,Ga)Se2 heterojunctions, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1741.  S. Siebentritt, Alternative buffers for chalcopyrite solar cells, Sol. En., 77 (2004) 767.  K. Mitchell, C. Eberspacher, J. Ermer, D. Pier, Single and tandem junction CuInSe2 cell and module technology, Conf. Rec. 20th IEEE Photovoltaic Specialists Conf., Las Vegas, USA, 26–30 Sept. 1988, IEEE, New York (1988) 1384.  M. Powalla, B. Dimmler, K.-H. Groß, CIS thin-ﬁlm solar modules - an example of remarkable progress in PV, Proc. 20th European Photovoltaic Solar Energy Conf., Barcelona, Spain, June 6–10 2005, in press.  R. W. Birkmire, Recent progress and critical issues in thin ﬁlm polycrystalline solar cells and modules, Conf. Rec. 26th IEEE Photovoltaic Specialists Conf., Anaheim, USA, 29 Sept-3 Oct 1997, IEEE, New York (1997) 295.  E. Rudigier, I. Luck, R. Scheer, Quality assessment of CuInS2 -based solar cells by Raman scatter- ing, Appl. Phys. Lett., 82 (2003) 4370.  o J. Palm, V. Probst, A. Brummer, W. Stetter, R. T¨ lle, T. P. Niesen, S. Visbeck, O. Hernandez, M. Wendl, H. Vogt, H. Calwer, B. Freienstein, F. Karg, CIS module pilot processing applying concurrent rapid selenization and sulfurization of large area thin ﬁlm precursors, Thin Solid Films, 431–432 (2003). 514  a M. Powalla, D. Hariskos, E. Lotter, M. Oertel, J. Springer, D. Stellbogen, B. Dimmler, R. Sch¨ fﬂer, Large-area CIGS modules: processes and properties, Thin Solid Films, 431–432 (2003) 523.  a o O. Lundberg, M. Bodeg˚ rd, J. Malmstr¨ m, L. Stolt, Inﬂuence of the Cu(In,Ga)Se2 thickness and Ga grading on solar cell performance, Prog. Photovolt: Res. Appl., 11 (2002) 77.  D. E. Tarrant, R. R. Gay, Process R&D for CIS-based thin-ﬁlm PV, Annual Technical Report 24 April 2002–23 April 2003, NREL(SR-520-35684, National Renewable Energy Lab, Golden (2004).  L. Mrig, S. Rummel, Outdoor stability performance of CIS and CdTe photovoltaic modules at SERI, Conf. Rec. 21st IEEE Photovoltaic Specialists Conf., Kissimmee, USA, 21–25 May 1990, IEEE, New York (1990) 1038.  M. Powalla, B. Dimmler, R. Schaefﬂer, G. Voorwinden, U. Stein, H.-D. Mohring, F. Kessler, D. Hariskos, CIGS solar modules: progress in pilot production, new developments and appli- cations, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, 272 THIN FILM SOLAR CELLS W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1663.  a H.-D. Mohring, D. Stellbogen, R. Sch¨ fﬂer, S. Oelting, R.Gegenwart, P. Konttinen, T. Carlsson, M. Cendagorta, W. Herrmann, Outdoor performance of polycrystalline thin ﬁlm PV modules in different european climates, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 2098.  o J. Malmstr¨ m, J. Wennerberg, L. Stolt, A study of the inﬂuence of the Ga content on the long-term stability of Cu(In,Ga)Se2 thin ﬁlm solar cells, Thin Solid Films, 431–432 (2003) 436.  o J. Klaer, R. Scheer, R. Klenk, A. Boden, Ch. K¨ ble, Stress behaviour of CuInS2 and Cu(In,Ga)Se2 thin ﬁlm PV-modules studied by a speciﬁc test structure, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1847.  J. M. Woodcock, H. Schade, H. Maurus, B. Dimmler, J. Springer, A. Ricaud, A study of the upscaling of thin ﬁlm solar cell manufacture towards 500 MWp per annum, Proc. 14th European Photovoltaic Solar Energy Conf., Barcelona, Spain, 30 June–4 July 1997, H. A. Ossenbrink, P. Helm, H. Ehmann (Eds.), H.S. Stephens, Bedford, UK (1997) 857.  U. Ugalde, J. Alonso, T. Bruton, J. M. Woodcock, K. Roy, K. De Clerq, A general procedure for cost evaluations: application to the MUSIC FM project, Proc. 14th European Photovoltaic Solar Energy Conf., Barcelona, Spain, 30 June–4 July 1997, H.A. Ossenbrink, P. Helm, H. Ehmann (Eds.), H.S. Stephens & Associates, Bedford, UK (1997) 897.  T. M. Bruton, MUSIC FM - Five years on fantasy or reality, Proc. PV in Europe – From PV Technology to Energy Solutions, Rome, Italy, 7–11 October 2002, WIP Munich, Germany (2002) OA6.1.  L. Frantzis, E. Jones, M. Wood, C. Lee, P. Wormser, Opportunities for Cost Reduction in Photo- voltaic Modules, Proc. 16th European Photvoltaic Solar Energy Conf., Glasgow, UK, 1–5 May 2000, H. Scheer, B. McNelis, W. Palz, H. A. Ossenbrinck, P. Helm (Eds.), James & James, Glasgow (2000) 2069.  u N. Meyer, I. Luck, U. R¨ hle, J. Klaer, R. Klenk, M. Ch. Lux-Steiner, R. Scheer, Towards thin ﬁlm module production using a simpliﬁed CIS process, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1698.  M. E. Beck, S. Wiedemann, R. Huntington, J. VanAlsburg, E. Kanto, R. Butcher, J. S. Britt, Advancements in ﬂexible CIGS module manufacturing, Proc. 31st IEEE Photovoltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 211.  B. Dimmler, M. Powalla, R. Schaefﬂer, CIS solar modules: pilot production at Wuerth Solar, Proc. 31st IEEE Photovoltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE New York (2005) 189.  D. E. Tarrant, R. R. Gay, V. Probst, F. H. Karg, CIS thin ﬁlm development and production status at shell solar, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003). 325.  u F. H. Karg, D. Kohake, T. Nierhoff, B. K¨ hne, S. Grosser, M. Ch. Lux-Steiner, Performance of grid coupled pv-arrays based on CIS solar modules, Proc. 17th European Photovoltaic Solar Energy Conf., Munich, Germany, 22–26 Oct. 2001, B. McNelis, W. Palz, H.A. Ossenbrink, P. Helm (Eds.), WIP Munich, Germany (2001) 391.  SENSE – Sustainability evaluation and recycling development for thin-ﬁlm solar cells, contract ENK5-CT-2002-00639 (http://www.sense-eu.net/).  a M.J de Wild-Scholten, K. Wambach, E.A. Alsema, A. J¨ ger-Waldau, Implications of Euro- pean environmental legislation for photovoltaic systems, Proc. 20th European Photovoltaic So- lar Energy Conf., Barcelona, Spain, 6–10 June 2005 W. Palz et al. (Eds.), WIP-Munich (2005) 3143. CHALCOPYRITE BASED SOLAR CELLS 273  F. H. Karg, Entwicklung und Fertigung von CIS-Solarmodulen bei Siemens Solar FVS The- men 2000, G. Stadermann (Eds.), Forschungsverbund Sonnenenergie, Berlin, Germany (2000) 145.  US Geological Survey, Mineral Commodity Summaries, January 2004.  K. Ramanathan, J.C. Keane, B. To, R.G. Dhere, and R. Nouﬁ, The potential for high performance in CIGS solar cells: a laboratory perspective, Proc. 20th European Photovoltaic Solar Energy Conf., Barcelona, Spain, June 6–10 2005, in press.  D. Hariskos, M. Powalla, N. Chevaldonnet, D. Lincot, A. Schindler, B. Dimmler, Chemical bath deposition of CdS buffer layer: prospects of increasing material yield and reducing waste, Thin Solid Films, 387 (2001) 179.  B. Malinowska, M. Rakib, G. Durand, Re-use of chemicals in CBD of CdS thin ﬁlm for CIS solar cells, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1820.  K. Kushiya, M. Ohshita, M. Tanaka, Development of recycling and reuse technologies for large-area Cu(InGa)Se2 -based thin-ﬁlm modules, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 1892.  K. Knapp, T. Jester, Empirical investigation of the energy payback time for photovoltaic modules, Sol. En., 71 (2001) 165.  A. Neisser, C. A. Kaufmann, R. Klenk, R. Scheer, M. A. Kroon, G. Oomen, H.-W. Schock, Prototype development - ﬂexible Cu(In,Ga)Se2 thin ﬁlm solar cells for space applications, Conf. Rec. 31st IEEE Photovoltaic Specialists Conference and Exhibition, Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 227.  See http://europa.eu.int/comm/environment/waste/weee index.htm.  K. Kushiya, M. Tachiyuki, T. Kase, Y. Nagoya, T. Miura, D. Okumura, M. Satoh, I. Sugiyama, O. Yamase, Improved FF of CIGS thin-ﬁlm mini-modules with Zn(O,S,OH)x buffer by post- deposition light soaking, Conf. Rec. 26th IEEE Photovoltaic Specialists Conf., Anaheim, USA, 29 Sept–3 Oct 1997, IEEE, New York (1997) 327. a  H. Mufﬂer, M. B¨ r, C.-H. Fischer, R. Gay, F. Karg, M. C. Lux-Steiner, ILGAR technology, VIII sulﬁdic buffer layers for Cu(In Ga)(S,Se)2 solar cells prepared by ion layer gas reaction (ILGAR), Conf. Rec. 28th IEEE Photovoltaic Specialists Conf., Anchorage, USA, 15–22 Sept 2000, IEEE, New York (2000) 610. a  M. B¨ r, C. H. Fischer, H. J. Mufﬂer, B. Leupolt, T. P. Niesen, F. Kang, M. C. Lux-Steiner, High efﬁciency chalcopyrite solar cells with ILGAR-ZnO WEL-device characteristics subject to WEL composition, Conf. Rec. 29th IEEE Photovoltaic Specialists Conf., New Orleans, USA, 19–24 May 2002, IEEE, New York (2002) 636.  S. Siebentritt, P. Walk, U. Fiedeler, I. Lauermann, K. Rahne, M.Ch. Lux-Steiner, Th.P. Niesen, F.Karg, MOCVD as a dry deposition method of ZnSe buffers for Cu(In,Ga)(S,Se)2 solar cells, Prog. Photovolt: Res. Appl., 12 (2004) 333.  S. Spiering, A. Eicke, D. Hariskos, M. Powalla, N. Naghavi, D. Lincot, Large-area Cd-free CIGS solar modules with In2 S3 buffer layer deposited by ALCVD, Thin Solid Films, 451–452 (2004) 562. e  S. Gall, N. Barreau, S. Harel, J. C. Bern` de, J. Kessler, Material analysis of PVD-grown in- dium sulphide buffer layers for Cu(In,Ga)Se2 -based solar cells, Thin Solid Films, 481–482 (2005) 138.  K. Ramanathan, F. S. Hasoon, S. Smith, A. Mascarenhas, H. Al-Thani, J. Alleman, H. S. Ullal, J. Keane, Properties of Cd and Zn partial electrolyte treated solar cells, Conf. Rec. 29th IEEE Photovoltaic Specialists Conf., New Orleans, USA, 19–24 May 2002, IEEE, New York (2002) 523.  D. Hariskos, R. Menner, S. Spiering, A. Eicke, M. Powalla, K. Ellmer, M. Oertel, B. Dimm- ler, In2 S3 buffer layer deposited by magnetron sputtering for Cu(inGa)Se2 solar cells, Proc. 274 THIN FILM SOLAR CELLS 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June 2004, W. Hoff- man, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA-Florence (2004) 1894.  Th. Glatzel, H. Steigert, R. Klenk, M. Ch. Lux-Steiner, T. P. Niesen, S. Visbeck, Zn1−x Mgx O as a window layer in completely Cd-free Cu(In,Ga)(S,Se)2 based thin ﬁlm solar cells, Technical Digest Int. PVSEC-14, 707, accepted for publication in Sol. En. Mat. Sol. Cells.  Th. Magorian Friedlmeier, N. Wieser, Th. Walter, H. Dittrich, H.-W. Schock, Heterojunctions based on Cu2 ZnSnS4 and Cu2 ZnSnSe4 thin ﬁlms, Proc. 14th European Photovoltaic Solar Energy Conf., Barcelona, Spain, 30 June–4 July 1997, H.A. Ossenbrink, P. Helm, H. Ehmann (Eds.), H.S. Stephens & Associates, Bedford, UK (1997) 1242.  H. Katagiri, N. Ishigaki, T. Ishida, K. Saito, Characterization of Cu2 ZnSnS4 thin ﬁlms prepared by vapor phase sulfurization, Jpn. J. Appl. Phys., 40 (2001) 500.  H. Katagiri, Cu2 ZnSnS4 thin ﬁlm solar cells, Thin Solid Films, 481–482 (2005) 426.  a J. Wennerberg, J. Kessler, M. Bodeg˚ rd, L. Stolt, Damp heat testing of high performance CIGS thin ﬁlm solar cells, Proc. 2nd World Conf. Photovoltaic Solar Energy Conversion, Vienna, Austria, 6–10 July 1998, J. Schmid, H. A. Ossenbrink, P. Helm, H. Ehmann, E. D. Dunlop (Eds.), Joint Research Center ISPRA (1998) 1161.  K. Orgassa, H. W. Schock, J. H. Werner, Alternative back contact materials for thin ﬁlm Cu(In,Ga)Se2 solar cells, Thin Solid Films, 431/432 (2003) 387.  T. Nakada, Y. Hirabayashi, T. Tokado, D. Ohmori, Cu(In1−x ,Gax )Se2 thin ﬁlm solar cells using transparent conducting oxide back contacts for bifcial and tandem solar cells, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 2880.  a R. Klenk, R. Mauch, R. Sch¨ fﬂer, D. Schmid, Progress in CuGaSe2 based thin ﬁlm solar cells, Conf. Rec. 22nd IEEE Photovoltaic Specialists Conf., Las Vegas, USA, 7–11 Oct 1991, IEEE, Ney York (1991) 1071.  T. Nakada, T. Kume, T. Mise, A. Kunioka, Superstrate-type Cu(In,Ga)Se2 thin ﬁlm solar cells with ZnO buffer layers, Jpn. J. Appl. Phys., 37 (1998) L499.  F.-J. Haug, D. Rudmann, H. Zogg, A. N. Tiwari, Light soaking effects in Cu(In,Ga)Se2 superstrate solar cells, Thin Solid Films, 431–432 (2003) 431.  M. Nanu, L. Reijnen, A. Goossens, J. Schoonman, CuInS2 /TiO2 heterojunction solar cells obtained by atomic layer deposition, Thin Solid Films, 431–432 (2003) 492.  o I. Kaiser, K. Ernst, Ch.-H. Fischer, C. Rost, I. Sieber, M.-Ch. Lux-Steiner, R. K¨ nenkamp, The EAT-solar cell with CuInS2 a photovoltaic cell concept using an extremely thin absorber (ETA), Sol. En. Mat. Sol. Cells, 67 ¨(2001) 89.  F. Lenzmann, M. Nanu, O. Kijatkina, A. Belaidi, Substantial improvement of the photovoltaic characteristics of TiO2 /CuInS2 interfaces by the use of recombination barrier coatings, Thin Solid Films, 451–452 (2004) 639.  C. Grasso, M. Burgelman, Theoretical study on the effect of an intermediate layer in CIS-based ETA-solar cells, Thin Solid Films, 451–452 (2004) 156.  V. K. Kapur, B. M. Basol, E. S. Tseng, Low cost thin ﬁlm chalcopyrite solar cells, Conf. Rec. 18th IEEE Photovoltaic Specialists Conf., Las Vegas, USA, 21–25 Oct 1985, IEEE, New York (1985) 1429.  J. Rechid, R. Thyen, A. Raitzig, S. Wulff, M. Mihhailova, K. Kalberlah, A. Kampmann, 9 % efﬁciency: CIGS on Cu substrate, Proc. 3rd World Conf. Photovoltaic Energy Conversion, Osaka, Japan, 11–18 May 2003, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, C. Wronski, W. C. Sinke (Eds.) (2003) 559.  J.-F. Guillemoles, O. Roussel, O. Ramdani, M. Benosman, C. Hubert, J. P. Fauvarque, A. Chomont, e N. Bodereau, P. Panheleux, P. Fanouill` re, J. Kurdi, N. Naghavi, P. P. Grand, L. Parissi, M. Ben- Farah, J. Connolly, S. Taunier, P. Mogensen, D. Lincot, O. Kerrec, The CISEL project: recent achievements, Proc. 19th European Photovoltaic Solar Energy Conf., Paris, France, 7–11 June CHALCOPYRITE BASED SOLAR CELLS 275 2004, W. Hoffman, J.-L. Bal, H. Ossenbrink, W. Palz, P. Helm (Eds.), WIP-Munich and ETA- Florence (2004) 1669.  R. N. Bhattacharya, W. Batchelor, K. Ramanathan, M. A. Contreras, T. Moriarty, The performance of CuIn1−x Gax Se2 -based photovoltaic cells prepared from low-cost precursor ﬁlms, Sol. En. Mat. Sol. Cells, 63 (2000) 367.  Vijay K. Kapur, Ashish Bansal, Phucan Le, Omar I. Asensio, Non-vacuum processing of CuIn1−x Gax Se2 solar cells on rigid and ﬂexible substrates using nanoparticle precursor inks, Thin Solid Films, 431/432 (2003) 53.  C. Eberspacher, C. Fredric, K. Pauls, J. Serra, Thin-ﬁlm CIS alloy PV materials fabricvated using non-vacuum, particles-based techniques, Thin Solid Films, 387 (2001) 18.  R. Klenk, S. Bakehe, R. Kaigawa, A. Neisser, J. Reiß, M. Ch. Lux-Steiner, Optimising the open- circuit voltage of Cu(In,Ga)S2 solar cells – design and analysis, Thin Solid Films, 451–452 (2004) 424.  T. J. Coutts, K. A. Emery, J. S. Ward, Modeled performance of polycrystalline thin-ﬁlm tandem solar cells, Prog. Photovolt. Res. Appl., 10 (2002) 192.  S.-H. Wei, S. B. Zhang, A. Zunger, Effects of Ga addition to CuInSe2 on its electronic, structural, and defect properties, Appl. Phys. Lett., 72 (1998) 3199.  S. Siebentritt, Wide gap chalcopyrites: material properties and solar cells, Thin Solid Films, 403– 404 (2002) 1.  W. N. Shafarman, P. D. Paulson, Losses in CuInSe2 -based thin ﬁlm monolithic tandem solar cells, Proc. 31st IEEE Photovoltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 231.  S. Nishiwaki, S. Siebentritt, P. Walk, M. Ch. Lux-Steiner, A stacked chalcopyrite thin-ﬁlm tandem solar cell with 1.2 V open-circuit voltage, Prog. Photovolt. Res. Appl., 11 (2003) 243.  K. Mitchell, C. Eberspacher, J. Ermer, D. Pier, Single and tandem junction CuInSe2 cell and module technology, Conf. Rec. 20th IEEE Photovoltaic Specialists Conf., Las Vegas, USA, 26–30 Sept. 1988, IEEE, New York (1988) 1387.  M. Symko-Davies, Status of high performance PV: polycrystalline thin-ﬁlm tandems, Proc. 31st IEEE Photovoltaic Specialists Conf., Lake Buena Vista, USA, January 3–7, 2005, IEEE, New York (2005) 410.