The 2nd international conference on particle physics and astrophysics
4 stars based on
Delbos 1 ,2 ,3 a. After a description of crystallographic and optoelectronic properties, including CZTS crystalline structure, defect formation and metal composition, this review paper focuses on CZTS synthesis processes and device properties. Synthesis strategies, including one- or two-step processes, deposition temperature, binary formation control via atmosphere control and their effect on device properties are discussed.
According to deployment scenarios, this binary domain 2012 region freerus lt+30 parity could happen binary domain 2012 region freerus lt+30 and The chalcogenide technologies are particularly vulnerable to In, Ga, Te supply, which has been noted critical or near critical by the US Department of Energy [ 9 ] or the European Commission [ 10 ].
According to various authors, the production of PV modules could be limited binary domain 2012 region freerus lt+30 the demand reported earlier in this paper, as summarized in Table 1.
Crystalline silicon PV seems suitable for supplying the demand, but one should keep in mind that the basic assumption used for the third and fourth column of this table is that all worldwide production of silver is used for crystalline PV. The fifth column shows the forecast demand for Ag, Te, In with respect to the production. It appears that the demand will grow enormously. If we focus on In, in the total demand including PV and other uses could be 3 to 7 times the production [ 1011 ], whereas, as In binary domain 2012 region freerus lt+30 a byproduct of Zn extraction, the elasticity of the production with respect to demand is low.
One author suggests that CIGS and CdTe technologies will not be impeded by resources problems, but he does not take the increase of other sources of binary domain 2012 region freerus lt+30 for In and Te into account [ 12 ]. See Table 2 for the references. Binary domain 2012 region freerus lt+30 1 Production limitations of the PV technologies [ 78101112]. The sustainability of PV production is therefore a real question, and the development of a new PV technology based on abundant and preferably non-toxic elements would alleviate the pressure on all PV technologies in term of resources.
CZTS has numerous advantages that could lead to its massive use as an abundant, non toxic, low cost absorber for thin-film photovoltaic solar binary domain 2012 region freerus lt+30. The band gap of CZTS can also be tuned by incorporation of Ge, Ge-containing materials having smaller bandgap than their Ge-free counterparts [ 25 ]. This tunability is of particular interest for the manufacturing of absorbers with a band gap between 1. Nevertheless, for high efficiency solar cells the processes used for CIGS solar cells will need to be tuned.
The buffer layer, in particular, will have to be tailored in order to adjust the lattice-matching, the valence and band offsets with CZTS. This paper is focused on synthesis processes of CZTS, and thus will only briefly address the crystallographic and optoelectronic properties. These crystal structures are very close, in both structures the cations are located on tetrahedral sites but their distributions on planes perpendicular to the c -axis are not the same.
In addition, the position of the chalcogen atom is slightly different in these structures [ 333435 ]. It seems that the Se compound has larger lattice parameters as well as higher electric conductivity than the S compound [ 20 ]. Concerning mechanical and thermo-physical properties of CZTS, no experimental data are available, but ab initio simulations were performed [ 4445 ]. Nevertheless, it is not possible to have a quantitative determination of the secondary phases by Raman.
As in the CIGS materials, binary domain 2012 region freerus lt+30 doping is obtained by stoichiometry variations and not by extrinsic doping. Measurements of the deep defects of CZTS are reported here [ 53 ]. Nevertheless some authors suggest the contrary [ 525657 ]. Zn Cu substitutions occur at the 2c site and Binary domain 2012 region freerus lt+30 Zn at the 2d site [ 3658 ] see Fig.
The accepting defects are easily formed, leading to difficult n -type doping [ 54 ]. The diameter of the circles is proportional to the conversion efficiency.
Until now, only Zn-rich binary domain 2012 region freerus lt+30 Cu-poor materials gave high efficiency see Fig. Absorbance measurements are not suitable for extracting the bandgap values, because the absorption coefficient derived from spectrophotometric data on thin films is determined by defect absorption and by measurement accuracy limitations [ 66 ]. In particular, ZnSe [ 67 ] and Zn-caused disorders [ 66 ] lead to incorrect bandgap determination when it is determined from absorbance measurements.
For CIGS, the deposition techniques used to be classified as vacuum and non-vacuum techniques. For CZTS, the same approach was also used in a review [ 69 ]. Nevertheless, for CZTS, this classification is not accurate, because recent works showed that low-cost deposition does not mean low-cost solar modules [ 70 ], and because of the problematic of binaries control during CZTS synthesis see next paragraph.
We therefore decided to classify the processes into one-step or two-step processes:. Two-step processes, where the needed elements are first incor-porated during an ambient temperature process, followed byan annealing step.
The chalcogen binary domain 2012 region freerus lt+30 be incorporated into theprecursor or during the annealing step. These processes allowthe use of fast and low-cost techniques for precursor deposi-tion. IBM workson three such processes, the hybrid solution-particle tech-nique [ 14 ], electrodeposition [ 71 ] and coevaporation [ 72 ].
One-step processes, where all the elements are incorporated simultaneously. The best efficiency until now is 9. Only a few groups published results for such processes, and very few had significant results in terms of solar cell efficiency: Other groups tried reactive sputtering and pulsed laser deposition PLD see Tab. Until now, it does not seem that one or the other is intrinsically better, because, as we will binary domain 2012 region freerus lt+30 in the next paragraphs, the key point is temperature and atmosphere control during the one-step deposition or annealing step.
Other methods of synthesis that did not produce solar cells include: Nevertheless, the chemical reactions are not yet completely understood, even if a reaction path was proposed [ 82 ]. There are successive reactions taking place in the bulk of the layer between the elements leading to binaries, ternaries and finally the quaternary.
For one-step processes, Cu-rich growth conditions are needed at the beginning of the reaction in order to foster growth of large grains [ 7393 ], as shown in Figure 4.
Nevertheless, for both processes, the finished layer must be Zn-rich see part 4. If the layer is not Cu 2 X-free at the end of the synthesis process, the subsequent cyanide etching leads to possible voids and defects [ 89 ]. If Zn-rich growth conditions are used, ZnX formation is promoted [ 4849 ]. Therefore, for one-step processes it is necessary to adapt the rate of deposition of each element with respect to the reaction path, as well as control deposition temperature, because higher temperatures can also increase grain size [ 59 ].
For two-step processes binary domain 2012 region freerus lt+30 is necessary to closely monitor growth conditions, and in particular prevent Cu 2 X formation and increase ZnX reaction with Cu 2 SnS 3. It is therefore of the utmost importance to prevent CZTS decomposition as well as binaries losses. The key to that result is atmosphere control. The equilibrium of equation 2 can be displaced towards CZTS by saturating the atmosphere in one of the right-side members of the equation [ 9397 ]. It is possible to counteract SnX evaporation by introducing gaseous SnX during the thermal treatment [ 98 ], or by increasing the partial pressure of the chalcogen [ 99 ], by performing the chalcogenation in a small closed volume [ ], or by using a cap on the precursor layer during annealing [ ].
If the annealing takes place in a H 2 atmosphere, Zn loss can be prevented by using ZnS as a precursor [ ]. Continuous evaporation of Zn towards the substrate prevents the decomposition of CZTS in a coevaporation process [ 73 ].
It also seems that the presence of MoX 2 at the back contact pushes the reaction towards the right member [ ]. If atmosphere control is not good, annealing time is the next best lever on CZTS decomposition, because if CZTS is synthesized and cooled rapidly, only a small quantity of binaries will be lost.
One paper compared slow and fast annealing [ ]. It concluded that fast annealing is better than slow annealing because many different secondary phases were found with slow annealing, without more explanation. Nevertheless, in this case slow annealing could be linked to a badly controlled reaction atmosphere, leading to CZTS decomposition into binary and ternary phases. Concerning the key point of atmosphere control, strategies of synthesis should be slightly different for one-step or two-step processes.
For two-step processes, there seems binary domain 2012 region freerus lt+30 be two stages during the annealing step: The control of the atmosphere should be very accurate in the second step, and the reaction atmosphere could also be completely different from the first step [ 93 ].
Fortunately, it is quite easy binary domain 2012 region freerus lt+30 inexpensive to design annealing equipments that provide good static atmosphere control that is to say, equipments where nothing can either go in or go out.
Nevertheless, vapor injection is more tricky, as it requires carrier gas, pressure measurements, leading to an equipment resembling an evaporator.
For one-step synthesis, atmosphere control is not so easy, because reaction chambers are designed for controlling the deposition rate and not the partial pressure of each element. In order to prevent CZTS decomposition and binaries losses, Sn and chalcogen deposition rate should remain high during the cooling phase, in order to counteract Sn loss [ 73 ].
The composition gradient either created by a variation of the rate of deposition of each element in a codeposition or by stacked elements, has an influence on CZTS film properties, because of Cu diffusion, Sn loss and chemical mechanisms of formation.
Concerning Cu, its diffusion can create voids at the back contact. Two-step processes, especially when the precursor layer is a codeposited metal layer or stacked metal layers with Cu as first layer, are subject to these voids [ 89binary domain 2012 region freerus lt+30. Secondary phases are observed, preventing the total formation of CZTS [ ]. As explained in the previous paragraph see Eq. Nevertheless, the loss of Sn could also be prevented by having the Cu layer on top [ 6393 ].
Concerning the reaction mechanism, for stacked metals, it is beneficial to have Cu and Sn in close contact for the formation of large CZTS grains [ ]. For stacked metals, Cu on top seems to foster formation of Cu x S, which may result in larger grains and denser films [ ].
For one-step processes, high Cu content at the beginning of the deposition promotes the formation of large grains [ 73 ]. The codeposition seems also possible, but with a Cu-poorcomposition near the back-contact.
Al, as presented in Figure 5. Calculation showed that CdS has a suitable band offset with CZTS [ ] and until now, the most efficient solar cells were obtained binary domain 2012 region freerus lt+30 a CdS buffer layer. Oxidation of the CZTS absorber by dipping in deionized water [ 13 ] or O 2 annealing [ 73 ] before deposition of CdS seems to improve efficiencies.
As for CIGS, reducing the time between absorber synthesis and CdS deposition to the minimum is of the utmost importance [ ]. A few variations of the architecture are known, especially on the buffer layer.
Other architectures were proposed: The literature is still quite poor concerning device properties, because very few high-efficiency devices were synthesized yet. It seems also beneficial for removing Cu 2 X that increase the resistivity of the layer [ ]. This causes a widening of the bandgap without type inversion, which could be an easy way to tailor bandgap alignments . It seems also that the electronic structure of a Sn-depleted surface is not favorable for the formation of well working p binary domain 2012 region freerus lt+30 n junction [ 7493 ].
The Conduction Band Offset, about 0. Better lattice matching between absorber and buffer could limit recombination at the interface, and using different buffer with smaller lattice such as In 2 S 3 or Zn S, O, OH could improve the efficiency of the devices .
Concerning the back contact, ZnX was observed at the back contact [ 59], as shown in Figure 6and it seems that increasing the ZnX content strongly impairs efficiency [ 47 ]. Concerning minority carrier lifetime, it might not be such a problem, because as in CIGS and CdTe, grain boundaries collect minority carrier and provide a current pathway for them to reach the n -type CdS and ZnO layers [ 30 ] and for coevaporated absorbers, high minority carrier diffusion length was measured several hundreds of nm [ 72 ].
Nevertheless, it seems that grain boundaries are Cu-rich, which could increase recombination rates and thus diminishing V OC [ 29 ].