Nada
Benhaddou
ab,
Safae
Aazou
ab,
Robert
Fonoll-Rubio
c,
Yudania
Sánchez
c,
Sergio
Giraldo
*c,
Maxim
Guc
c,
Lorenzo
Calvo-Barrio
de,
Victor
Izquierdo-Roca
c,
Mohammed
Abd-Lefdil
a,
Zouheir
Sekkat
abf and
Edgardo
Saucedo
c
aFaculty of Sciences, Mohammed V University, Rabat, Morocco
bOptics & Photonics Center, Moroccan Foundation for Advanced Science, Innovation & Research (MAScIR), Rabat, Morocco
cCatalonia Institute for Energy Research (IREC), Jardins de les Dones de Negre 1, 08930 Sant Adrià de Besòs, Barcelona, Spain. E-mail: sgiraldo@irec.cat
dCentres Científics i Tecnològics de la Universitat de Barcelona (CCiTUB), Lluís Solé i Sabarís 1-3, 08028 Barcelona, Spain
eIN2UB, Departament d’Electrònica, Universitat de Barcelona, Martí i Franquès 1, 08028 Barcelona, Spain
fDepartment of Applied Physics, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan
First published on 11th March 2020
Among the thin film chalcogenide photovoltaic community there is an increasing interest in the study of cationic and anionic substitution in the different absorber materials, including CdTe, chalcopyrites – Cu(In,Ga)(S,Se)2 and kesterites – Cu2ZnSn(S,Se)4. In the last case, cationic substitution has been revealed as a key factor to solve or palliate to some extent part of the fundamental problems of the kesterite technology. Among the different possibilities, the partial or total substitution of Sn by Ge is one of the most promising options, with proved excellent results from very small up to almost complete replacement. In view of the relevance of Ge in kesterite, this work presents the complete analysis of the reaction formation of the Cu2ZnGeSe4 (CZGeSe) compound using a sequential process based on the sputtering of elemental stacked layers followed by reactive annealing under Se atmosphere, by implementing a break-off experiment. An unusual solid–liquid–vapor extended growth mechanism is observed, thanks to the previous formation of a eutectic GeSe9 liquid phase that melts at temperatures above 212 °C. Driven by this liquid phase, it is demonstrated that CZGeSe formation mechanisms follow a strict sequence, starting from more simple molecules (binary compounds), then evolving to the ternary one, and finally to the quaternary alloy which is formed through the reaction of Cu2GeSe3 and ZnSe solid phases. The relevance of the study is supported by the solar cells prepared with these absorbers, demonstrating conversion efficiency at the level of the best reports in the literature. Finally, possible strategies to manage this singular formation pathway are discussed.
Considering these issues associated with Sn, its substitution with other group IV elements from the periodic table, such as Ge, might be an interesting solution to overcome the Sn drawbacks thanks to its low propensity toward +II oxidation state. There are several examples in the literature where the partial substitution of Sn by Ge has demonstrated an increment in the cell efficiency as a result of the improved crystallinity and the enlarged grain size, as well as the boost in VOC comparing to the world record devices.8,12Table 1 summarizes a selection of some of the most successful attempts to introduce Ge in Cu2ZnSnSe4. As can be seen, from very small quantities (less than 0.005%), up to approximately 40% of Sn substituted by Ge, large improvements on solar cell devices parameters, especially in the VOC deficit, have been reported. There is then a general consensus about the positive effect of Ge on kesterite, with several identified beneficial effects in the absorber properties, including improved morphology,9 control of doping by interaction with Na,13 improved carriers’ lifetime,14,15 annihilation/minimization of deep defects,8 shallower acceptor levels,16 possible accumulations at the back contact that could be useful for graded bandgap concepts,17etc. Following this, it is unquestionable the relevance that Ge can have in the near future of kesterite to further progress in the comprehension and maturity of these family of materials.
Ref. | Method | Ge/(Ge + Sn) | Eff. (%) | V OC (mV) | J sc (mA cm−2) | FF (%) | E g (eV) | V OC-deficita (mV) |
---|---|---|---|---|---|---|---|---|
a With respect to the Shockley–Queisser limit, and estimated by the authors with the available data in the different references. | ||||||||
Hages et al.15 | Nanocrystal inks printing | 0.30 | 9.4 | 460 | 31.9 | 63.8 | 1.19 | 0.48 |
Collord et al.18 | Molecular precursor solutions | ∼0.25 | 11.0 | 583 | 33.6 | 55.9 | ∼1.15 | 0.31 |
Giraldo et al.12 | Sputtering of metals | <0.005 | 10.1 | 453 | 33.6 | 66.8 | 1.04 | 0.31 |
Giraldo et al.19 | Sputtering of metals | <0.005 | 10.6 | 473 | 33.6 | 66.7 | 1.05 | 0.30 |
Giraldo et al.9 | Sputtering of metals | <0.005 | 11.8 | 463 | 38.3 | 66.3 | 1.04 | 0.31 |
Kim et al.19 | Co-evaporation | 0.39 | 10.03 | 543 | 29.5 | 62.7 | 1.19 | 0.39 |
Kim et al.14 | Co-evaporation | 0.22 | 12.32 | 527 | 32.2 | 72.7 | 1.11 | 0.35 |
Choubrac et al.20 | Evaporation | 1.0 | 7.6 | 558 | 22.8 | 58 | 1.36 | 0.58 |
Sahayaraj et al.21 | Evaporation | 1.0 | 5.4 | 744 | 16 | 46 | 1.4 | 0.39 |
This work | Sputtering of metals | 1.0 | 6.5 | 556 | 19.6 | 60 | 1.4 | 0.58 |
Inspired by these works and very recently, few groups tried to fully substitute Sn by Ge, and the current record efficiency achieved for pure Ge kesterite (CZGeSe) is 7.6%.20 Similar absorber compound witnessed an open circuit voltage of 744 mV recognized as the highest VOC obtained with a device efficiency of 5.5%.21,22 The investigation on Ge-substituted kesterite could be additionally interesting in order to broaden the application of kesterites, allowing to go towards wider bandgap materials for PV tandem solar cells or semi-transparent devices. Nevertheless, this material has been much less studied than its Sn-based counterpart, and although recently first important studies about the fundamental properties of CZGeSe have been published,23–31 there are still several issues that need to be addressed for further progress.
In particular, important progresses in CZTSSe grown by physical vapor deposition methods have been possible thanks to the deep investigation on the formation pathways. Depending on the conditions used during the synthesis process, either binary compounds or a combination of ternary and binary ones can drive the formation of kesterite.9,28,32–34 It has been suggested that most probably this last option is preferred in order to achieve more homogeneous absorber with less secondary phases due to the simpler pathway involving less intermediate species.9,33 In the case of the Ge substituted compound, only marginal information is available about the characteristics of the system during the synthesis of the compounds. Giraldo et al.12 by using reactive annealing of metallic stacks under elemental Se atmosphere, observed the large impact of the inclusion of even very small quantities of Ge into the CZTSe matrix, proposing that a Se-rich GexSey liquid phase is formed acting as a crystallization flux. The same author proposed a change from a reaction pathway involving binary compounds for the pure Sn-kesterite, to a combination of ternary and binary for the Ge-containing one,9 but for very small Ge quantities. Brammertz et al.28 by using reactive annealing of metallic stacks under H2S atmosphere, observed that for the pure Ge compound depending on the metallic stack order, the formation reaction proceeds at different speeds, but mainly through the reaction between Cu9Se5, Cu3Ge and ZnSe, obtaining different morphologies.
In view of the lack of information about the formation pathways for the CZGeSe compound, the goal of our study is to get a deep insight into the CZGeSe synthesis characteristics. By interrupting and analyzing the annealing process at different steps, a deep understanding of the reaction evolution at different times and temperatures is well established. We observe that the CZGeSe absorber is formed by an unusual vapor–liquid–solid extended reaction promoted by the formation of GeSe9 liquid compound at the very beginning of the CZGeSe formation, inducing a dendritic growth that afterwards evolves towards well-formed micro-crystals at higher temperature annealing steps. A complete analysis of the synthesis and crystallization steps is presented, combining morphological, structural and compositional characterization. The importance of this study is supported by the demonstration of devices with efficiencies about 6.5%, thanks to the understanding of the CZGeSe formation mechanisms.
Fig. 2 Cross-sectional SEM images and XRF results obtained for the samples T1 to T6 (layer thickness and Se content). |
This effect is even magnified for the sample obtained in the break-off point T2. As can be clearly seen in the SEM image at the top part of Fig. 2, the thickness increases up to 2–3 times, the surface becomes even flatter, the dendrites at the back are better distinguishable, and the XRF analysis indicates that the sample becomes extremely Se-rich (up to 70 at% of Se). This result can only be explained by the presence of a Se-rich liquid phase, as has been previously suggested for samples containing small Ge amounts.9,12 In the case of the Ge-pure compound, the liquid phase is the dominant one during a relevant part of the synthesis process. To better understand the nature of the liquid phase, Fig. 3a shows a schematic representation of the Ge–Se phase diagram in the Se-rich region.36
Fig. 3 Schematic representation of the Ge–Se phase diagram in the Se-rich region (a). Detail of the dendrites formed at the back region corresponding to the break-off point T2. |
As can be seen, a eutectic point is formed in the highly Se-rich zone, with a composition of approximately Ge0.08Se0.92 and a melting point of 212 °C, well below to the stablished temperature for the synthesis stage (330 °C). This special regime can have two very important consequences on the synthesis step, namely:
1. Synthesis assisted by a superheated liquid:
Considering that the temperature during the synthesis process is 330 °C, this implies that the Ge-Se liquid phase is superheated (the melting point is 212 °C). It is well known from the metallurgical field that fast cooled superheated/supersaturated liquids lead easily to the formation of dendritic growth,37 as the ones observed also in Fig. 3b. This is another conclusive evidence of the formation of a liquid phase in the system. In Fig. 3a the cooling down process of the superheated liquid is also schematized, showing that a solid phase with composition between Ge0.14Se0.86 and Ge0.08Se0.92 is obtained, in good agreement with the XRF measurement which gives a phase with an average composition of Ge0.11Se0.89.
2. Extended solid–liquid–vapor (SLV) type growth:
The presence of a liquid phase in the system suggests that an extended SLV-like type growth is occurring for the synthesis of CZGeSe. In a first stage (break-off points T1 and T2), selenium gas from the atmosphere is absorbed in the precursor forming the eutectic Se-rich Ge–Se phase which melts at temperatures well below 330 °C. Following this, the eutectic compounds start to decompose and the solid phase dissolved in the liquid precipitates, while Se2(g) is released and the liquid phase is consumed, as can be seen in the SEM cross-section and the XRF measurements from break-off points T2 to T3. This shows that the SLV growth is not only useful for the growth of nano-sized structures, but also for the effective crystallization of thin films in an extended version of this process.
The combined XRD and Raman phase analysis that will be presented later will give additional insights about the compounds formed in these intermediate points of the CZGeSe synthesis, to demonstrate the peculiarities of this particular kesterite system.
Following with Fig. 2, at the break-off point T3, most of the Ge–Se liquid phase was released and the layer recovered the expected thickness and Se content. Note that the layer is slightly delaminated from the substrate; this is probably due to a combination of factors including de-wetting problems of the Ge–Se liquid phase and the Glass/Mo substrate, as well as relatively fast Cu out-diffusion as will be shown later. The surface of the film is still flat enough to indicate that the liquid phase is present in the system, but the dendritic growth is not observed anymore; instead, a nanocrystalline layer appears suggesting that a new crystalline phase has been completely formed. Finally, once the temperature is increased in the second stage (crystallization stage, from break-off points T4 to T6), the evidences of a liquid phase presence disappear, and the thin film becomes more and more crystalline. Following this logic, we can infer that the synthesis process finishes during the T1–T3 stage, and that the T4–T6 is a pure crystallization step.
To extend the analysis to the cations present in the system, Fig. 4 shows the evolution of the Se/Ge ratio and the different cationic ratios through the complete T1–T6 process. As expected, and in good agreement with Fig. 2 and 3, the Se/Ge ratio strongly increases for the break-off point T2 and then decreases to stabilize during the crystallization process. This behavior is markedly different to what is commonly observed in the pure-Sn compound, therefore we can consider that Ge acts like a catalyst, capturing Se from the atmosphere through the formation of the eutectic Ge–Se compound (hereinafter called GeSe9 for simplification) and drastically changing the characteristics of the synthesis process. Concerning the cations composition relationships, Cu/Zn ratio is quite stable during the whole process suggesting that these elements only experience diffusion processes during the annealing and, as expected, are not exchanged with the atmosphere. Nevertheless, Cu/Ge and Zn/Ge ratios increase after the break-off point T3, suggesting that mainly at the end of the synthesis processes, and beginning of the crystallization, some Ge is exchanged with the atmosphere, most probably due to the decomposition and evaporation of the eutectic GeSe9 liquid phase.
Complementary information can be obtained from SEM/EDX combined surface analysis. Fig. 5 shows the SEM top view of samples T1, T2, T3 and T4, and the corresponding EDX mapping, where Cu, Zn, Ge and Se intensity maps are shown. For T1 a smooth surface with a homogeneous distribution of all the elements is observed, suggesting that at early stages of the Ge–Se liquid phase formation there is not yet enough energy in the system to provoke large diffusion of the different elements. The surface morphology drastically changes for T2, where large (200–500 μm) “wave-like” structures are observed, corroborating the presence of a liquid phase. The formation of these morphologies suggests that the Ge–Se liquid phase does not wet completely the metallic surface below. In addition, in the frontiers of these “wave-like” structures, Cu, Zn and Ge can clearly be seen, while they are completely depleted of Se. It is also confirmed that these “wave-like” structures are only composed of Ge and Se. At T3, these structures cover less surface area, being flatter and more extended, consisting of the remnant Ge–Se liquid phase not evaporated yet. Below them, Cu, Zn, Ge and Se are clearly detected suggesting that most probably, the formation of CZGeSe is already occurring at this stage, or at least more complex selenide binary or ternary compounds are formed. Finally, by just increasing the temperature from T3 to T4, first crystalline structures start to be observed at the surface, and the elemental distribution becomes more homogenous, suggesting that all the Ge–Se liquid phase has been released from the system. Note that in some specific areas, an increased Zn concentration is detected, most probably related to the formation of an excess of ZnSe, due to the Zn-rich composition of the precursor that is also commonly observed in the pure-Sn compound.38
As a summary of this first morphology/composition characterization, the pure Ge kesterite formation seems to proceed very differently than the pure Sn one.9,12 Interestingly, Ge acts like a “catalyst” in the system, trapping Se from the atmosphere and forming a superheated liquid phase with approximately 10% of Ge and 90% of Se, which promotes a S–L–G type growth for the formation of CZGeSe. Now, it becomes very relevant to understand which phases are present at the different stages and their role during the kesterite formation.
For this purpose, in-depth compositional analysis, using Auger spectroscopy, and detailed phase analysis, using XRD and multi-wavelength Raman spectroscopy, were combined. Fig. 6 shows the in-depth Auger profile for Cu, Zn, Ge and Se (Mo is shown in grey scale as reference to identify the surface and back region), in a colored schema for a better visualization of their inter-diffusion (raw profiles are presented in Fig. S1 of the ESI†). In Fig. 6, each element intensity was normalized to the highest value. In the following, these in-depth composition profiles will be analyzed together with the XRD and Raman spectroscopy characterization presented in Fig. 7 (with 442 nm excitation wavelength, while spectra obtained with 785 nm and 532 nm excitation wavelengths are reported in Fig. S2 of the ESI†).
Fig. 6 In-depth Auger spectroscopy compositional profiles, showing the distribution of Cu, Zn, Ge and Se, from T1 to T6. The individual signals were normalized to the highest intensity value. |
Fig. 7 XRD diffractograms (a), and surface Raman spectra with 442 nm excitation wavelength (b), of samples T1 to T6. |
Comparing then Fig. 6 and 7, the following information can be drawn in terms of elemental distribution and phase formation and reaction during the whole reactive annealing process:
Confirmed phases: Ge–Se (∼10 at% Ge–90 at% Se), Cu–Zn brasses, metallic Zn, Cu–Se, ZnSe?
Considering this, the main characteristics of the T0–T1 step are:
• Fast Se in-diffusion and incorporation from the atmosphere through the formation of the GeSe9 liquid phase.
• Ge acts as a “catalyst” to trap Se from the atmosphere.
• Cu starts out-diffusing from the back region towards the surface.
The main reactions occurring can be summarized as follow:
Confirmed phases: Ge–Se (∼10 at% Ge–90 at% Se), Cu–Zn brasses, Cu–Se, Cu2GeSe3, ZnSe?
The main characteristics of the T1–T2 step are:
• Even faster Se in-diffusion and incorporation from the atmosphere towards the layers through the formation of large amounts of GeSe9 liquid phase.
• Increased Cu out-diffusion, splitting its distribution in two regions, the back region where the Cu–Zn alloy is unreacted, and the middle region where Cu diffuses by reaction with GeSe9.
The main reactions occurring during this step can be summarized as follows:
The main characteristics of the T2–T3 step are:
• A clear Se out-diffusion (and release as Se2(g)) due to the reaction between the Ge–Se liquid phase and CuSe to form the ternary compound.
• Full splitting of the Cu content in two regions, the one containing the ternary compound and the one with still unreacted Cu–Zn alloys and CuSe.
The main reactions occurring during this last part of the low temperature dwell time can be summarized as follow:
Confirmed phases: Cu2ZnGeSe4, ZnSe.
The main characteristics of the T3–T6 step are:
• Fast diffusion of all the elements, and homogeneous in-depth distribution.
• All the reactions are finished in the short T3–T4 step (ramping from low to high temperature), and T4–T6 step is mainly for the crystallization of the absorber and probably re-distribution of remaining secondary phases (mainly or exclusively ZnSe).
The main reactions occurring during this annealing time can be summarized as follow:
All this demonstrates the peculiarities of the formation of CZGeSe, with a mechanism that involves vapor, liquid and solid phases. To summarize this, Fig. 8 shows the proposed schema illustrating the formation pathways of CZGeSe, where the key steps are highlighted. In fact, the system evolves during the annealing process from simple elemental species towards more complex binary compounds first, then ternary, and finally the quaternary kesterite. Basically and based on the schematic representations of Fig. 8, to form the CZGeSe kesterite by this sequential process, it is needed:
1. In-diffusion of Se to form the binary GeSe9 liquid phase that is not only an intermediate specie to form kesterite, but also act as a catalyst.
2. Partial selenization of brasses into binary CuSe(s) and ZnSe(s) that promotes the out-diffusion of Cu (either as metal or most probably as CuSe).
3. Reaction of CuSe(s) with GeSe9(l) to form the ternary Cu2GeSe3(s) with the release of Se2(g).
4. Reaction of Cu2GeSe3(s) and ZnSe(s) to form quaternary Cu2ZnGeSe4(s).
This makes not only unique the synthesis route of this material among all the kesterite family, but also contributes to clearly identify the most relevant and challenging aspects for future improvements. In particular, as very positive aspects we can mention the presence of a GeSe9(l) phase that catalyzes the Se incorporation and drives the formation reaction. The apparent non-solubility of ZnSe into the GeSe9(l) phase and the relatively fast out-diffusion of Cu (probably as CuSe) lead to the formation of the Cu2GeSe3(s) ternary phase as the final intermediate compound for the formation of kesterite. All this synergy between a system that evolves smoothly from more simple up to more complex species involving liquid, solid and vapor phases, enhances the probability of completing all the reactions minimizing the risk of presence of unreacted secondary phases at the end of the synthesis process. In fact, the only secondary phase probably present is ZnSe.
Conversely, among the challenging characteristics of the synthesis of this material, the formation of large amounts of GeSe9(l) liquid phase can be also problematic. Back in Fig. 5, it was shown that the liquid phase seems to not completely wet the surface, forming non-homogenous “wave-like” structures that could introduce non-homogeneities in the final absorber, or even contribute to a partial peel-off from the substrate. Additionally, the relatively fast Cu-out diffusion observed during the first stage of the synthesis can also contribute to the formation of voids at the back contact, as has been demonstrated in other related technologies.9,32,39
Finally, to support the relevance of this study, solar cells were prepared with absorbers synthesized with the presented thermal routine, which was optimized prior to the experiments presented in this work. Fig. 9 (left) shows the illuminated J–V curve (AM1.5) of one of the champion cells, demonstrating a 6.5% efficiency device with 556 mV of VOC as it is also summarized in Table 1 (neither anti-reflection coating nor metallic grid were used in these devices). This is one of the highest efficiencies reported so far for the pure Ge compound, with a VOC and a FF almost equal to the current record.20 The external quantum efficiency (EQE) spectrum shows a maximum of roughly 75%, but with a flat aspect between 500–700 nm, which may suggest a relatively good quality CZGeSe/CdS junction and CZGeSe bulk material. From 700 nm, the EQE starts to decrease, most probably due to recombination problems at the back contact that could be related to the presence of voids in this region. The improvement of the de-wetting problems as well as the fast Cu out-diffusion are two specific issues that could contribute to boost, at least partially, the conversion efficiency of this material in the future.
Fig. 9 Illuminated J–V curve (AM1.5) with the corresponding optoelectronic parameters (left side), and external quantum efficiency (EQE) spectrum with the estimated bandgap (right side). |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9tc06728k |
This journal is © The Royal Society of Chemistry 2020 |