Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells

Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells Volume 1 photovoltaic solar energy 1 18 – chalcopyrite thin film materials and solar cells

T Unold and CA Kaufmann, Helmholtz Zentrum für Materialien und Energie GmbH, Berlin, Germany

Artificial chalcopyrite-type crystals were first synthesized and structurally characterized by Hahn et al in the early 1950s [2] The optical and electrical properties of chalcopyrite-type crystals were investigated by Shay and Wernick at Bell labs in the 1970s, originally for the application in optoelectronic devices [1] First single-crystal homojunction devices based on CuInSe2 were realized and electroluminescence was demonstrated also at Bell labs in 1974 by short anneals of n-type crystals in Se vapor [3] The first single-crystal solar cell based on CuInSe2 as an absorber material was demonstrated in the same year, using a CuInSe2/CdS heterojunction device This device, which contained a very thick, several micron n-type CdS as emitter window layer, showed a photoconversion efficiency of 5% [4] Soon after photoconversion, efficiencies above 10% were obtained by further optimization of such device structures [5]

First real thin-film solar cells based on chalcopyrite-type absorbers were prepared by Kazmerski also using CuInSe2/CdS heterojunctions [6] These types of solar cells started to receive considerable attention when Mickelson et al demonstrated solar cells based on polycrystalline CuInSe2 absorber layers with an efficiency of 9.4% in 1981 by co-evaporation from elemental sources [7] Already in 1982, an impressive thin-film solar cell efficiency of 14.6% was reported by the same group by optimizing the co-evaporation process [8] Since then, a number of technological breakthroughs, such as the discovery of Na doping, alloying with

Ga, and replacing the thick CdS window layer by a thin CdS buffer and thick conductive ZnO window layer, have led to a current record device efficiency of 20.3% for Cu(In,Ga)Se2-based thin-film solar cells [9]

Over the last 30 years, chalcopyrite materials and solar cells have been investigated by many groups worldwide, and we will attempt to give an overview of the most salient findings and lessons learned with respect to these types of solar cells Also industry has been involved in research and commercialization of chalcopyrite solar cells early on starting with Boeing and ARCO Solar in the 1980s Since then, large-scale production facilities for chalcopyrite-based thin-film photovoltaic modules have been built and ramped up, with an estimated current capacity close to 1 GW year−1 Although m2-sized modules with record efficiencies of 17%

have been demonstrated recently [10], there is still a considerable gap between efficiencies achieved on small area devices in the laboratory and actual module efficiencies obtained in large-scale production Here we will try to address the current state of knowledge and relevant challenges for commercialization with respect to chalcopyrite-type solar cells We would like to mention that a number of excellent reviews on chalcopyrite-type materials and solar cells have been published previously, which may provide additional information that is not covered in this chapter [11–17]

1.18.2 Material Properties

1.18.2.1 Structure

The crystal structure of chalcopyrite-type semiconductors can be derived from the diamond lattice in accordance with the Grimm-Sommerfeld or 8-N rule [18] This means that chalcopyrite semiconductors, just like group IV elements silicon or germanium, exhibit tetrahedral bonding, that is, every atom has four nearest neighbors In this review, we will restrict ourselves

to the Cu-chalcopyrite semiconductors formed from group I-III-VI elements Starting from the sphalerite structure of ZnS (Figure 1), the chalcopyrite lattice is obtained by the ordered substitution of the group II element (Zn) by the group I (Cu) and group III (In or Ga) elements This leads to a doubling of the unit cell in the c-direction, the so-called tetragonal crystal structure as shown in Figure 1 Because of the different bond strength and bond lengths of the group I–VI and III–V bonds, the lattice parameter c in general is not exactly 2a, which is also called the tetragonal distortion of the unit cell [1]

Because chalcopyrite-type materials consist of at least three elements, a number of different phases are possible depending on the exact compositions and the growth conditions In Figure 2, a ternary phase diagram with the corner points Cu, In, and Se is shown Because chalcopyrites are usually synthesized at sufficient chalcogen excess (here Se) the composition of, in this example CuInSe2, the thin-film materials prepared at varying Cu/(In + Ga) composition usually conform to the tie-line spanned by Cu2Se and In2Se3 The desired chalcopyrite phase in this diagram is at the center of the tie-line As will be discussed further below, many deposition techniques use this finding by moving from the In-rich to the Cu-rich side and back to an experimentally determined ideal composition at the end of the process An equilibrium pseudobinary phase diagram composed of a mixture of In2Se3 and Cu2Se, corresponding to the tie-line in Figure 2 is shown in Figure 3 [19] Observed phases are indicated as a function of growth temperature: α denotes the chalcopyrite phase, β denotes ordered defect chalcopyrite phases, such as CuIn5Se8 or CuIn3Se5 and δ

is the sphalerite phase occurring only at high temperatures It can be seen that there is a small region between stoichiometry and the copper-poor side where single-phase chalcopyrite is obtained at 500 °C This region narrows further for lower temperatures On the Cu-rich side, CuxSe phases segregate and on the Cu-poor side a coexistence of chalcopyrite and defect chalcopyrite phases is expected It is interesting to note that at low temperatures even for stoichiometric composition CuxSe phase segregation is expected, which has to be considered in the design of growth processes for these compounds It has been found that the width of the chalcopyrite single-phase region is increased by alloying with gallium (Ga) and/or doping with Na [20]

Figure 1 Crystal structure of (ZnS) sphalerite and (CuInS) chalcopyrite Note that the unit cell is doubled in the c-direction for the chalcopyrite lattice

Se 100

0

10

20 30CuSe2

Figure 3 Pseudobinary phase diagram for In2Se3–Cu2Se tie-line shown in Figure 2 The shaded area indicates the regions in the phase diagram relevant

to multistage coevaporation of high-efficiency chalcopyrite solar cells

CuInSe2 can be readily alloyed with Ga and forms a solid solution CuIn1−xGaxSe2 over the whole composition range 0 < Ga/ (In + Ga) < 1 This means that lattice constants change continuously from the lattice constants for pure CuInSe2 to those of pure CuGaSe2, in accordance with Vegards law [21] as illustrated in Figure 4 Note also that the c/a ratio, that is the tetragonal distortion, changes with composition from a mismatch value of +1% on the CuInSe2 to −3.5% for CuGaSe2 For a Ga content of about x = 0.23, the tetragonal distortion vanishes and c/a = 2.0 Recently, it has been found that for exactly this composition ratio (or c/a ratio) the grain size is significantly enhanced in Cu(In,Ga)Se2 thin films [23]

Figure 5 Absorption coefficients of CuInSe2, CuGaSe2, and CuInS2 Data from Scheer R and Schock HW (2011) Chalcogenide Photovoltaics: Physics, Technologies, and Thin Film Devices Weinheim, Germany: Wiley-VCH [16]

gaps between 1.04 and 1.68 eV [22, 24, 25] The band gap of CuIn1−xGaxSe2 increases monotonically with Ga content with a very small bowing between the end points This functional dependence can be described by

2

EC

EV

Cu2Se Cu(In,Ga)Se

2 Cu(In,Ga)3Se5

Figure 6 Band line-up for Cu(In,Ga)Se2 and secondary phases for Cu(In,Ga)3Se5, Cu2Se

The highest efficiencies with chalcopyrite solar cells have been achieved with a Ga content between 0.2 <x < 0.35 corresponding to a band gap between 1.15 and 1.25 eV It has been found that alloying with Ga mainly reduces the electron affinity of CuIn1−xGaxSe2 leading to an upward shift of the conduction band whereas the valence band stays at the same energetic position [25–27] This fact is important to consider when constructing band diagrams of CIGSe heterojunction devices In state-of-the-art Cu(In,Ga)Se2 devices grown by the three-stage co-evaporation method, a significant Ga gradient is found in the absorber layers This means that the band gap, and in particular the conduction band minimum, varies accordingly throughout the device In addition to or instead of Ga, sulfur can be added to CuInSe2 to widen the band gap of CuIn1−xGaxSe2 However, in contrast to the effect of Ga, the addition

of sulfur both lifts the conduction band and lowers the valence band [26] For a number of industrially used deposition processes, sulfur is added to CuIn1 −xGaxSe2 in order to widen the band gap close to the heterojunction [28]

Defect chalcopyrite phases are observed on the Cu-poor side of the phase diagram of CuIn1−xGaxSe2, whereas CuxSe secondary phases appear for Cu-rich growth conditions of CuIn1−xGaxSe2 This means that not only the structural and electronic properties of these phases may be relevant for the final device performance, but also their optical properties, in particular there band gap and band line-up with respect to the main chalcopyrite phase The expected band line-up for CuIn1−xGaxSe2 is shown in Figure 6 where it can be seen that for Ga contents of x ≈ 0.3 a perfect line-up of Cu2Se and mostly a lowered valence band maximum for the defect chalcopyrite compound is expected [29] This may be one of the keys why CuIn1−xGaxSe2 exhibits maximum efficiencies within this Ga-composition ratio

1.18.2.3 Electrical Properties

Cu-based chalcopyrite materials can be made p- or n-type depending on composition and growth conditions, solely due to the presence of intrinsic defects without the need for doping by external impurities Cu(In,Ga)Se2 grown Cu-rich and under chalcogen-excess is always found to be p-type [30] N-type doping of CuInSe2 can be obtained for low Cu/In ratios Also, conversion

of n-type CuInSe2 into p-type CuInSe2 by annealing in high Se pressure has been demonstrated [31] Typical charge carrier densities

of Cu-poor Cu(In,Ga)Se2 used in solar cells grown on soda-lime glass are between 1015 cm−3 and 1017 cm−3 Since the Cu/(In + Ga) ratio in these materials is usually between 0.8 and 0.9, large amounts of defects in the percent range must be present, despite the fact that the measured charge carrier densities are much lower This can be explained by the fact that Cu-poor chalcopyrite compounds with large deviations from stoichiometry such that Cu/(In + Ga) < 1 are heavily compensated, with simultaneous formation of acceptors and donors, and a corresponding much lower net carrier density [32] Also, the net doping density is found to strongly increase with the concentration of sodium incorporated in the films This is shown in Figure 7 where the charge carrier density is shown for Cu(In,Ga)Se2 deposited on polymide foil with different NaF precursor layer thicknesses [33] The effect of sodium on the net charge carrier density has been explained with a passivation of shallow donors such as InCu antisites [34], or an increase in the acceptor density via NaIn antisite defects [35] Also a catalytic effect of Na on the passivation of VSe vacancy donors by oxygen has been discussed [36] There is also evidence that Na acts preferentially at grain boundaries (GBs) rather than in the bulk of the individual grains of the thin films [35, 37] and it has been found to significantly influence the diffusion of elements and morphology of the chalcopyrite-type thin films [38]

In principle, there are 12 intrinsic defects possible in a ternary chalcopyrite: three vacancies, three interstitials, and six antisite defects Obviously, the quaternary Cu(In,Ga)Se2 or pentenary Cu(In,Ga)(S,Se)2 systems even provides for substantially more possible defects The formation of defects in thermal equilibrium can be described by defect chemical models with the formation

of the defects depending on the specific formation energies, the formation enthalpies, and the chemical potentials [39] For charged defects, the formation enthalpy also directly depends on the position of the Fermi-level during defect formation It has been found from density functional theory (DFT) calculations under most conditions the formation energy of the copper-vacancy in CuInSe2 and CuGaSe2 is very low and can be even negative, depending on the position of the Fermi level [40–42] These calculations also show that the defect transition levels for CuInSe2 can be rather shallow and that the formation of neutral defect pairs such as 2VCu-InCu is very likely [40] The DFT calculations predict that the ionization levels of these defect pairs are pushed out of the band gap, thus rendering them electronically inactive, in particular with respect to minority carrier recombination

Figure 8 Low-temperature photoluminescence (T = 10 K) as a function of excitation intensity for (a) stoichiometric epitaxially grown CuGaSe2 (data from Unold T and Gutay L (2011) Photoluminescence of thin film solar cells In: Abou-Ras D, Kirchatz T, and Rau U (eds.) Advanced Characterization of Thin Film Solar Cells Wiley [51]) and (b) Cu-poor Cu(In,Ga)Se2

Defects in chalcopyrite thin films have been investigated by admittance [43–46], Deep Level Transient Spectroscopy (DLTS) [47], and photoluminescence measurements [48–51] However, the unambiguous assignment of ionization energies to different defects remains elusive, because of the large number of possibilities to assign a transition found with these measurements Shallow defects can be best observed by photoluminescence measurements, whereas deep defects can be observed by admittance, DLCP, or DLTS measurements As expected from the phase diagram shown in Figure 3, the defect physics is expected to differ significantly for the Cu-rich versus the Cu-poor composition range for chalcopyrite solar cells Low-temperature photoluminescence measurements of stoichiometric CuGaSe2 show excitonic transitions and shallow donor–acceptor pair transitions at low temperatures, as shown in Figure 8(a) for a an epitaxial thin film grown by metal-organic chemical vapor deposition (MOCVD) [51, 52] Photoluminescence measurements on Cu-poor grown Cu(In,Ga)Se2 films which lead to the highest conversion efficiencies in devices do not show distinct, narrow defect transitions either at low temperatures or at room temperature as shown in Figure 8(b) As is typical for this material only one or two very broad defect transitions without excitonic signatures are observed, with very strong shifts of the peak energy with excitation intensity [32, 51, 53] This has been interpreted as a signature of significant potential fluctuations present in these films because of the high degree of compensation as discussed above

Photoconductivity, admittance, and capacitance voltage measurements have shown metastable effects, such as persistent conductivity after light soaking [54–56] These metastabilities have been discussed in conjunction with metastabilities in the current-voltage characteristics, such as the increase in open-circuit voltage upon illumination with AM1.5 light, which has been observed for chalcopyrite solar cells and modules [57, 58] The charge carrier density determined by capacitance profiling is found to increase after light soaking [59] Also application of a reverse bias has been shown to lead to an increase in the charge carrier density and defect profile as measured by capacitance techniques [20] There are a number of different models explaining these effects [56], among them a DX-center model [60], photodoping of CdS [61], the presence of a p+-layer close to the heterointerface [62], and an amphoteric defect model [63]

(a) (b) (c)

Mobilities of charge carriers in chalcopyrite materials have been determined for single crystals and to a much lesser extent for polycrystalline thin films The mobilities are found to differ strongly for the single-crystal and polycrystalline materials, as can be understood from the dominant role of GBs in the latter materials As also found for other semiconductor materials, the mobilities generally depend strongly on the measurement temperature and material’s composition Hole mobilities for CuInSe2 up to

60 cm2 V−1 s−1 [48] at room temperature have been measured for single crystals, and much lower values around 1 cm2 V−1s−1 have been obtained for polycrystalline thin films [50, 64, 65] These values have been inferred mostly from Hall measurements, which are difficult to interpret for polycrystalline and strongly compensated materials Hole mobility values smaller than 1 cm2

V−1 s−1 have been inferred from time-of-flight techniques [66], which has been explained by trapping in a large density of subgap states Electron mobilities of up to 900 cm2 V−1s−1 have been reported for CuInSe2 single crystals by Hall measurements [31] The determination of electron mobilities in polycrystalline thin films, which are usually p-type remains elusive

1.18.2.3.1 Surfaces and grain boundaries

GBs generally are expected to cause increased recombination of minority carriers because of the high density of broken or incorrect bonds occurring at such planes between two differently oriented crystals In addition, potential barriers at GBs can significantly reduce the effective mobility of carriers and cause carrier depletion in the bulk of the grains In contrast to other semiconductor materials such

as silicon and GaAs, GBs in chalcopyrites seem to play a remarkably small role for recombination of minority carriers and therefore device performance [67] Electron back scatter diffraction (EBSD) investigations on different chalcopyrite thin-film materials have shown that most of the GBs are highly symmetric twin boundaries, for which much lower defect densities are expected to be present when compared with randomly oriented GBs [68] In fact, relatively little dependence of the device performance on the grain size has been found, as long as the grain size reaches a certain size of several hundred nanometers In Figure 9(a), we show EBSD pattern quality maps of several Cu(In,Ga)Se2 solar cells that have been deposited with different Ga content It can be seen that the grain size varies significantly between x = 0 (CuInSe2) and x = 1 (CuGaSe2) As is shown in Figure 9(b), the efficiency of these devices increased for grain sizes of up to 0.6 µm, but does not correlate with the grain sizes for larger grains [23]

Different models have been proposed to explain GB physics in chalcopyrites as summarized in Figure 10 Downward band bending at the GB (Figure 10(a)), as it is commonly assumed for silicon [69] or CdTe [70], has also been proposed for CIGSe [71]

Precursor deposition Crystallization Optional removal of

secondary phases

Heater Back contact

Such a band bending at GBs may aid in the current collection of polycrystalline chalcopyrite solar cells and reduce recombination, as the minority carriers are attracted and the majority carriers are repelled However, numerical simulations have shown that, on the one hand, significant band bending is needed to effectively induce type inversion at the GB and that, on the other hand, such a configuration leads to significant open-circuit voltage losses in the devices [72, 73] The same simulations predict some improve­ment in device efficiency for upward bending of the GBs (Figure 10(b)) A significant reduction in the GB recombination activity, however, is predicted by two-dimensional (2D) device simulation for the case of a valence band barrier (Figure 10(c)) at the GBs [72] In this case, the minority carriers are unaffected while the holes are kept away from the zone of increased recombination at the

GB Such a valence band barrier can be caused by copper depletion because the states at the valence band maximum are due to antibonding states of Se-p and Cu-d hybrid orbitals [74] Reduction of the copper content at the GB or also at the surface is thus expected to lower the energy of the valence band maximum For an extended region of Cu depletion, an ordered defect compound may be formed (β phase), which also exhibits a lowered valence band maximum as discussed above Copper depletion has been predicted for either the anion- or the cation-terminated polar surfaces of CuInSe2, which may be stable because of the low formation energy of Cu-vacancies or InCu antisites [74] Note that in order to reduce recombination in such a GB scenario the region of lowered valence band maximum (VBM) must be at least about 3 nm wide in order to prevent tunneling of holes into the recombination region [73]

Experimental evidence on the composition, electronic structure, and recombination activity of GBs in chalcopyrites is very diverse, and it has not been possible to confirm a single GB model [67, 75, 76] This may be in part due to the fact that many of the experimental techniques are surface sensitive, and thus may be strongly influenced by surface, impurity, or oxidation effects [77] Kelvin probe force microscopy (KPFM), scanning tunneling microscopy (STM), and Hall measurements all indicate a considerable large distribution of barrier heights at GB from 0 to 300 mV [67, 77] It is important to note that upward bending and downward bending have been found for different grain boundaries on single samples With respect to recombination, it has been shown by cathodoluminescence [78] and electron beam-induced current (EBIC) measurements [79] that high symmetry twin GBs show much reduced recombination activity However, similarly as for the case of band bending detected by KPFM measurements, vastly differing behavior for different GBs on single samples has been found, which so far could not be correlated to specific structural models of GBs

1.18.3 Deposition Methods

The main criteria for the choice between different deposition processes and systems may be different for laboratory research and industrial production and scale up In the former case, mostly a high level of control over composition and high conversion efficiencies are aimed at In the latter case, apart from efficiency, low cost, high yield, reproducibility, and process tolerance have also

to be considered Not every deposition technique that performs well in the laboratory can be scaled up to homogeneously and reproducible cover square meter substrates at low cost In this section, we will give an overview of the most important fabrication routines and highlight advantages and disadvantages of the different approaches Deposition processes for chalcopyrite thin films can be generally divided into two main categories: ‘single-step’ and ‘sequential’ deposition processes A schematic for these two principal types of deposition methods is shown in Figure 11

Figure 11 Single-step (bottom) and sequential (top) deposition methods for chalcopyrite-type thin-film solar cells

In the last 30 years, many modifications of this process have been developed, which can be categorized by their number of stages in which elements are deposited at different rates Figure 12 displays the variations of the co-evaporation process as they are currently generally referred to (1) the single layer process, which consists of only one evaporation rate for each element and remains Cu-deficient throughout the whole process; (2) the bilayer process, which utilizes Cu-rich growth in the initial stage of the deposition process and ends up with an overall Cu-poor composition by reduction of Cu-supply in the last stage of the process; and (3) the three-stage approach

1.18.3.1.1 Single-stage process

For this most simple growth recipe, either complete compounds [80] or single elements are evaporated either by evaporation boats

or by Knudsen-type effusion cells in a high vacuum system For low cost and increased control of composition at varying deposition conditions, single-element sources are preferable over compound source material Best efficiencies for Cu(In,Ga)Se2 have always been found for Cu-poor material with Cu/In + Ga < 0.9, that is, within the single-phase region expected from the thermodynamic equilibrium phase diagrams Therefore, single-stage deposition requires precise control of the evaporation rates, which can be achieved by advanced evaporation source design or by means of optical process control such as atomic absorption spectroscopy (AAS) or electron impact emission spectroscopy (EIES) [81] Relatively high device efficiencies of up to 16% have been obtained by this deposition method [82] However, the average grain size for these films is rather small below 500 nm It has been found that the grain size of chalcopyrite thin films depends not only on growth temperature and final composition, but also on whether the thin film has been Cu-rich at some point during the deposition process On the other hand, as can also be seen from the equilibrium phase diagram, films grown in the Cu-rich regime contain significant amounts of CuxSe secondary phases that have been found to

be detrimental to device performance Therefore, the secondary phases either have to be etched by, for example, KCN or have to be converted into the chalcopyrite phase by use of several deposition stages with varying elemental ratios, as discussed in the following sections

1.18.3.1.3 Three-stage or multistage process

The highest conversion efficiencies to date have been achieved using CIGSe thin-film absorber layers that were deposited using co-evaporation procedures based on the three-stage process [9, 84] This process was originally implemented in the following way [85] During the first stage, In, Ga, and Se are evaporated at a substrate temperature of about 330 °C, which leads to the formation of

a very small grained (In,Ga)2Se2 compound layer During the second stage, In and Ga evaporation rates are turned to zero and only

Cu is evaporated together with Se, while at the same time the temperature is increased to the final deposition temperature between

500 and 600 °C The Cu-Se evaporation continues until the composition of the compound becomes Cu-rich, that is Cu/(In + Ga) > 1, which leads to a rapid recrystallization of the compound and to the segregation of CuxSe phases The Cu-Se

Figure 12 Different rate profiles used in co-evaporation of Cu(In,Ga)Se deposition

Rel intensity (arb units)

deposition is continued until the overall composition is about Cu/(In + Ga) ≈ 1.15, after which the Cu deposition is turned off, and

a final stage consisting of the evaporation of In, Ga, and Se proceeds until the semiconducting layer has the desired final composition, typically Cu/(In + Ga) < 0.9 Devices grown with this three-stage process usually show a depth-dependent composi­tional grading with respect to the Ga content of the thin film, which is a natural consequence of the three-stage deposition sequence [86, 87] A Ga gradient, as it is typical for such a device, is displayed in Figure 13 While the positive gradient toward the back interface of the device is a standard feature for the vast majority of processes, the slight increase toward the absorber surface and its effect on the later device performance is highly dependent on the individual process parameters

Several growth models for this optimized process were published [85, 88–90] They consider the diffusion of Cu or Cu-Se compounds into thin layers of In2Se3 or (In,Ga)2Se3 and can – in principle – be transferred to other fabrication methods According

to these growth models, a Cu-rich growth phase ensures the simultaneous presence of Cu-Se phases along with the chalcopyrite compound This enhances the ‘recrystallization’ of the growing films Although the precise details regarding recrystallization are still a matter of debate, the literature agrees on the fact that after recrystallization the electronic quality of a chalcopyrite thin-film absorber is much enhanced [87, 91–93] Aspects such as grain size, crystal structure, and atomic defects are all under discussion to be affected The Cu-poor/Cu-rich transition of the growing thin film at the end of the second stage of the three-stage or multistage co-evaporation process can be easily monitored via pyrometry, laser light scattering (LLS) or by the heater’s power input [94], because the material properties change drastically at the stoichiometry point Most importantly, the segregation of CuxSe phases on the surfaces leads to a change in the surface morphology, which can be detected by LLS, and to a change in the emissivity, which can

be detected by pyrometry or thermometry In that sense, the important features and final composition of the films can be very precisely controlled by monitoring the stoichiometry point, which is to be followed by a relatively short third stage While the application of LLS was originally developed using a He:Ne laser as a light source and the measurement of the diffusely scattered portion, a charge-coupled device (CCD) camera used as detector together with a white light source can provide spatially resolved information [95] Further in situ methods are available and are applicable also to co-evaporation methods other than the three-stage process [96, 97]

The three-stage process has also been applied to sulfide compounds [98], but for these materials conversion efficiencies have not exceeded 12.6 % so far [99]

A possible implementation of the three-stage process in commercial inline systems is shown in Figure 12(d) Here, the rates of the individual evaporation sources are fixed in time, and the time-dependence of the deposition rates used in static laboratory systems has to be translated into different locations of the sources within the inline system [100] Note that in this case the use of line sources is beneficial in order to ensure homogenous evaporation of the typically 0.3–0.6 m wide substrates [101] With a certified conversion efficiency of 19.3% the in-line adaptation of the three-stage co-evaporation process has proven the potential for its application in large-scale fabrication of high-efficiency thin-film modules [102]

1.18.3.2 Sequential Deposition

Sequential deposition generally involves at least two distinct steps: (1) the deposition of a precursor layer (usually metal) and (2) the crystallization step that transforms the precursor into a chalcopyrite absorber layer, usually performed by heating and chalcogenization [15] As for the single-step deposited absorber thin films, further treatment before proceeding with the solar cell fabrication may be required While single-step deposition is commonly performed in vacuum chambers, sequential deposition methods can be generally classified into vacuum and nonvacuum techniques [103]

Formation of chalcopyrite compounds by sequential processing has been studied using thermodynamic and reaction kinetic approaches, as well as by in situ observations during the chalcopyrite formation by energy-dispersive X-ray diffraction (EDXRD) and Raman techniques [104–106]

During the second step of these processing techniques, the chalcopyrite-type material is formed either by annealing, reactive annealing, rapid thermal processing (RTP) or by rapid thermal annealing in a reactive chalcogen atmosphere If the precursor already contains the chalcogen, simple heating in an inert gas atmosphere could in principle be used to form the absorber layer However, if the system is not sealed, chalcogen loss during this process is very likely leading to undesirable defects and defect phases in the absorber layer Metallic precursors can be used to react at high temperatures (400–500 °C) in H2S or H2Se for the formation of Cu(In,Ga)Se2 or Cu(In,Ga)S2 [117, 118] Alternatively, elemental selenium or sulfur can also be used The annealing step in this process can last from tens of minutes up to several hours The advantage of this method is a high level of control over the reaction, whereas the disadvantage is that it is very slow and has to be compatible with temperatures the substrate withstands for extended periods of time

Much faster reactions are obtained by RTP, again either without or in the presence of the Se or S [28, 119–121] By this process, high-quality chalcopyrite absorbers can be obtained from metal layers within several minutes The drawback of this method is that it

is much harder to control as the reactions leading to precursor phases, secondary phases, and finally the desired chalcopyrite phase occur within a very short time span, which depends, for example, on temperature, temperature ramp, pressure and chalcogen pressure Reactions occurring during the sulfurization step of CuInS2 have been observed in situ by EDXRD [122–125] as shown in Figure 14 From such measurements, the reaction sequence

Cu; Culn2 → Cu11ln9 → CuS; lnS; CulnS2 → Culn5S8; CulnS2; Cu2 xS → CulnS2; Cu2 xS ½2 has been determined for the reaction of Cu/In precursors with elemental sulfur in a rapid thermal anneal process [123] The experiments were performed for sulfur partial pressures of 1 mbar and 1 K s−1 heating ramps Note that the presence

of secondary phases and reaction paths strongly depend on these processing parameters as also has been found for the Cu-In-Ga-Se system [122]

For both the long-time thermal anneal and the rapid thermal anneal chalcogenization, the formation of MoSex or MoSx at the back contact, the formation of large voids close to the back contact, and adhesion of the absorber layer to the substrate in general pose major challenges The formation of thin MoSex or MoSx at the back contact interface to the absorber is desirable as it improves the electrical performance; however, thick MoSex or MoSx layers lead to a very large series resistance that deteriorates device performance Adhesion at the back contact has to be maximized by the proper choice of substrates and adjustment of the morphology and microstructure of the molybdenum back contact

Figure 14 Energy dispersive X-ray diffraction (EDXRD) performed in situ during the RTP of CuInS2 Figure from Rodriguez-Alvarez H, Mainz R, Marsen

B, et al (2010) Recrystallization of Cu-In-S thin films studied in situ by energy-dispersive X-ray diffraction Journal of Applied Crystallography 43: 1053–1061 [93]

1.18.3.3 General Considerations

The following issues have to be considered in deposition of chalcogenides:

• Maximum process temperature T: Depending on the thermal stability of the kind of substrate in use (see also Section 1.18.4), different maximum substrate temperatures are used for the deposition process Processes, which exert only little thermal stress on the substrate during thin-film deposition using maximum temperatures between 330 and 450 °C, have successfully been developed [82, 126] and it is has been shown that despite the diffusion limitation at low growth temperatures, multistep co-evaporation growth processes can produce absorbers of remarkably high electronic quality [127] The use of high process temperatures, on the other hand, speeds up diffusion during film deposition and helps to decrease compositional inhomogeneity within the complete thin film [128, 129] Pulsed laser-assisted co-evaporation has been evaluated to compensate for low process temperatures [130]

• Ga content GGI (=[Ga]/([Ga] + [In])): As described in Section 1.18.2, the Ga content of the material determines the band gap of the deposited material The use of high deposition temperatures has proven beneficial when using high Ga contents [129] Alternative routines to improve growth with high GGI, as, for example, ionization of Ga, have also been applied [131]

• Cu content CGI (=[Cu]/([Ga] + [In])) during and at the end of the deposition: The amount of Cu present during thin film preparation is important because recrystallization may be enhanced in the presence of secondary Cu-Se/S phases and also due to the fact that Cu-deficient phases may play a critical role in the energy band line up at the absorber surface, that is, at the absorber/ buffer interface (see Section 1.18.2) Therefore, it is necessary to consider not only the final composition of the absorber thin film, but also the path in the phase diagram along which the synthesis of the thin film proceeds [87, 132]

• Chalcogen flux during layer deposition: The Se partial pressure in the deposition chamber must be high enough to impede the re-evaporation of Se and In2Se from the surface of the growing thin film, in particular at high deposition temperatures [133] Structural and optoelectronic characteristics of the deposited thin film are affected by the selenium/metal ratio of the deposition rates [134, 135] Activation of evaporated Se species has been studied in terms of improved material yield and also to support Se integration in the absorber growth process when low process temperatures are used [136, 137]

• Amount of Na present and supply method: As mentioned in Section 1.18.2, the incorporation of Na in CIGSe has a beneficial impact on the final quality of the absorber layer, although no such effect has been found for Cu-rich deposited sulfide compounds A variety of methods are used in order to supply Na On standard float glass without a diffusion barrier coating

Na will diffuse through the Mo back contact into the deposited thin-film material at the elevated process temperatures If a barrier layer is applied or a Na-free substrate is used, Na needs to be supplied externally This can be achieved by co-evaporation of NaF during the CIGSe deposition process [138], by a NaF posttreatment of the CIGSe thin film after deposition [139] or by use of a NaF precursor layer that is evaporated onto the Mo back contact (see Section 1.18.4) prior to CIGSe deposition [140], where it could have also a direct impact on the CIGSe/Mo interface conditioning [141] Alternatively, on flexible substrates, Na has also successfully been supplied by a soda-lime glass layer that is located beneath the Mo back contact [142] or by adding Na to the back contact material [143] The most successful method to date is the NaF posttreatment [127] The presence of a certain element, such as Na, may have a catalytic effect on the formation of particular material phases Thus, Na is, for example, assumed

to promote the formation of oxides within CIGSe thin films [144]

Most of these process parameters are interdependent and have a direct effect on the structural and electronic characteristics of the growing CIGSe thin film Understanding material formation and the codependence of material characteristics and resulting device properties are critical for achieving high conversion efficiencies and reproducibility Since co-evaporation so far has produced the highest efficiency devices, offers reasonable process control, and gives access to a wide range of process parameters, it is the model system mostly used to study these topics

1.18.4 Device Structure

Chalcopyrite solar cells are heterojunction devices consisting of a large number of layers with different functional properties A schematic of a typical device is shown in Figure 15 Unlike amorphous or microcrystalline silicon devices, which are commonly fabricated in the superstrate configuration (sun light enters through the glass substrate), chalcopyrite solar cells are so far generally made in the substrate configuration (sunlight enters from the top) This is due to the fact that molybdenum has been found to be a very stable contact in this configuration and because the high temperatures used in the chalcopyrite absorber growth (>500 °C) lead

to undesirable diffusion and intermixing if the window and buffer layers are deposited prior to the absorber

1.18.4.1 Substrate

1.18.4.1.1 Glass

Currently, the most commonly used substrate in CIGSe thin-film solar cells and modules is conventional soda-lime glass This is due to the fact that this glass is readily available through its use in the architectural industry, has favorable thermal expansion