Cu2ZnSnSe4 absorbers processed from solution deposited metal salt precursors under different selenization conditions

Authors

  • Carolin M. Fella,

    Corresponding author
    1. Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Thin Films and Photovoltaics, Ueberlandstrasse 129, 8600 Duebendorf, Switzerland
    • Phone: +41 58 765 61 01, Fax: +41 58 765 40 42
    Search for more papers by this author
  • Alexander R. Uhl,

    1. Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Thin Films and Photovoltaics, Ueberlandstrasse 129, 8600 Duebendorf, Switzerland
    Search for more papers by this author
  • Yaroslav E. Romanyuk,

    1. Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Thin Films and Photovoltaics, Ueberlandstrasse 129, 8600 Duebendorf, Switzerland
    Search for more papers by this author
  • Ayodhya N. Tiwari

    1. Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Thin Films and Photovoltaics, Ueberlandstrasse 129, 8600 Duebendorf, Switzerland
    Search for more papers by this author

Abstract

Cu2ZnSnSe4 (CZTSe) thin film solar cells are fabricated by a simple, non-vacuum deposition of metal salts dissolved in non-hazardous solvents followed by selenization in Se atmosphere. Despite a residual carbon-rich layer between the back contact and the CZTSe absorber layer, solar cells with up to 4.28% conversion efficiency are obtained for Cu-poor and Zn-rich CZTSe absorbers. A frequently reported problem, the loss of tin, is investigated with respect to the influence of the selenization conditions such as substrate temperature and selenium partial pressure. EDX point measurements directly confirm that the remaining decomposed layer consists of a mixture of binary ZnSe and Cu2−xSe phases if the substrate temperature is too high and not sufficient Se is supplied.

original image

1 Introduction

Kesterite compounds Cu2ZnSnS4 (CZTS) and Cu2ZnSnSe4 (CZTSe) are emerging as promising absorber materials for thin film solar cells, because they consist of earth abundant, non-toxic elements and their optoelectronic properties suggest great potential for cost-effective high efficiency solar cells. Highest solar cell conversion efficiencies of 8.4% 1 and 9.15% 2 were reported for thermally evaporated sulfide CZTS and selenide CZTSe absorbers, respectively. Non-vacuum deposition methods offer several advantages for the implementation in low cost industrial production, where simple and safe processing routes involving non-hazardous precursors and gases are desired. A recent review by Todorov and Mitzi 3 highlights numerous non-vacuum techniques based on solutions or nanoparticle dispersions, which can enable the true low-cost potential of Cu2ZnSn(S,Se)4 (CZTSSe) based absorbers. The record efficiency of 10.1% for a mixed sulfide-selenide CZTSSe material has been achieved with a hybrid solution-particle approach by Barkhouse et al. 4. For this hybrid approach hydrazine solvent is used, which is difficult to handle due to its hazardous and highly combustible nature. Bypassing the disadvantages of handling and safety issues, dimethyl sulfoxide was used to dissolve Cu-, Zn- and Sn-salts alongside with thiourea to grow CZTS absorber layers which were partially selenized to CZTSSe yielding up to 4.1% efficiency 5.

Regardless of the employed metal precursor deposition method, a high temperature annealing in sulfur/selenium atmosphere is essential to form a well-crystallized and phase pure CZTS(Se) material, as well as to decompose and evaporate undesirable residual precursor constituents. The high temperature sulfurization/selenization can, however, alter metal ratios between Cu, Zn, and Sn in the CZTS(Se) absorber as compared to initial precursor ratios. The absorber layer composition has a strong influence on solar cell performance 6. It was shown by Katagiri et al. 7 that only Cu-poor (Cu/(Zn + Sn) = 0.8–0.9) and Zn-rich (Zn/Sn = 1.1–1.2) CZTS(Se) absorbers result in highest efficiencies. Therefore, it is very important to understand whether initial metal ratios are retained in the sulfurized/selenized layer or any loss occurs because of the formation of volatile phases that originate during the chemical conversion. Another reason for change in metal ratios is the thermal decomposition of kesterites at elevated substrate temperatures of >350 °C at pSe = 4 × 10−6 Torr for CZTSe 8, leading to the formation of volatile metal compounds 9–11. The tin content is especially sensitive to elevated temperatures which can be explained by the kesterite decomposition reaction yielding volatile SnSe and Se 6, 12, 13.

In this work, we investigate the conversion process of a metal precursor layer into the kesterite compound with an emphasis to understand parameters of controlling the overall composition of the selenized layer. A simple solution approach employing metal salts dissolved in non-toxic alcoholic solvents is used to deposit the metal salt precursors. A similar methodology was applied by Kaelin et al. 14 to produce Cu(In,Ga)Se2 (CIGS) solar cell absorbers resulting in a highest efficiency of 6.7%. The main challenge is to optimize the selenization parameters, such as substrate temperature and selenium partial pressure in order to obtain the tailored metal ratios, and critical Sn content in the crystalline CZTSe layer without phase separation. Sn content in CZTSe can be controlled during the selenization of Cu–Zn–Sn containing precursors by controlling process parameters, such as substrate temperature and selenium partial pressure. As a proof of concept we demonstrate a 4.28% efficient solar cell based on a Cu-poor and Zn-rich – pure selenide absorber.

2 Experimental

2.1 Film preparation

Solutions of copper (II) nitrate hemipentahydrate (99.99+ %, Sigma–Aldrich), zinc (II) nitrate hexahydrate (98%, Sigma–Aldrich) and tin (IV) chloride hydrate (98%, Alfa Aesar) were prepared in a mixture of ethanol and 1,2-propanediol. The metal ratios in the precursor solution were fixed to Cu/(Zn + Sn) = 0.65 and Zn/Sn = 1.48 with total molarity of 2.29 mol/L. A second solution of higher viscosity with 10 wt% ethyl cellulose (4 mPa s, Sigma–Aldrich) dissolved in 1-pentanol (99+%, Alfa Aesar) was prepared. The two solutions were mixed to obtain the final solution with a suitable rheology for knife coating. The deposition of the precursor paste was done on a 1 mm thick soda lime glass (SLG) coated with a 600 nm thick, DC sputtered molybdenum layer. The thickness of the wet metal precursor film was approximately 30–40 µm resulting in about 1 µm thick CZTSe films. Prior to selenization, the film was first dried under a lamp at 200 °C and then on a hot plate at 230 °C to evaporate the solvents and partially burn the ethyl cellulose. The chalcogenization took place in a two-zone tube furnace, where elemental selenium vapors were transported by nitrogen carrier gas from the selenium zone to the substrate zone. The pressure in the furnace was set to either 5 or 10 mbar by partially closing the valve between pump and furnace, whereas the N2 flow was kept constant at 25 sccm/min. The temperature of the substrate and the selenium zone was varied between ∼370–660 and ∼360–380 °C, respectively. A thermocouple was situated at a distance of 3 cm from the sample. The temperature variation of 360–380 °C for the selenium zone would correspond to the variation in selenium vapor pressure of 1.5–2.7 mbar under equilibrium conditions 15. However, it was not possible to accurately determine the actual selenium partial pressure above the substrate in our case, because an open selenization reactor with significant background pressure of the carrier gas was used. Therefore, we rely on directly measurable quantities, such as selenium zone temperature and amount of evaporated selenium, to evaluate the effective selenium flux. A 3 mm thick SLG was used to avoid extensive glass softening at a substrate temperature of 660 °C. The selenization duration after reaching the maximum substrate temperature was 26 min.

The resulting CZTSe layers were characterized by scanning electron microscopy (SEM) combined with energy dispersive X-ray (EDX) spectroscopy to deduce elemental concentrations. The acceleration voltage was 20 kV if not stated differently. X-ray diffraction (XRD) was performed with Ni-filtered Cu Kα radiation. Raman measurements were performed using a HeNe laser excitation source (633 nm).

2.2 Solar cell fabrication and characterization

CZTSe absorbers were coated with a CdS buffer layer (about 70 nm thick) by chemical bath deposition and an i-ZnO/ZnO:Al window bi-layer (about 1.2 µm thick) deposited by RF magnetron sputtering. Finally, a mechanical scribing step defined the cell area to 0.09 cm2. Current density–Voltage (JV) measurements were performed under simulated AM1.5 global illumination. Quantum efficiency (QE) measurements were done for the UV–Visible and the IR wavelength regions using Si and Ge reference photodiodes, respectively.

3 Results and discussion

The selenization temperature is one of the main parameters determining the final elemental composition of selenized CZTSe layers. High selenization temperatures are beneficial for recrystallization, as well as to partially burn the ethyl cellulose binder in our process route. The evolution of composition (by EDX) as a function of the selenization temperature is presented in Fig. 1a. The selenization was carried out at a chamber pressure of 5 mbar. To ensure comparable results, the amount of evaporated selenium for each selenization run was kept constant at around 150 mg by adjusting the selenium zone temperature near 360 °C. We observe a significant Sn loss for temperatures exceeding 600 °C, while other elemental compositions remain nearly constant.

Figure 1.

(online color at: www.pss-a.com) (a) Elemental concentrations as obtained from EDX for different substrate temperatures. (b) XRD patterns of the samples described in (a). The CZTSe phase starts to form at substrate temperatures as low as 370 °C, and decomposes above 600 °C leaving binary phases Cu2−xSe and ZnSe. (c,d) SEM images of CZTSe and of the decomposed sample with 1 at% Sn (over-all compositional EDX measurement).

X-ray diffraction patterns of Sn containing layers show characteristic reflections of CZTSe phase (PDF reference 01-070-8930) starting at selenization temperatures as low as 370 °C, as can be seen in Fig. 1b. We, however, cannot exclude the presence of secondary phases, in particular ZnSe and Cu2SnSe3, because they have almost identical diffraction patterns as CZTSe 16, 17. As soon as the substrate temperature exceeds 600 °C, the Sn content is drastically reduced, yielding a mixture of binary Cu2−xSe (PDF reference 6-680) and ZnSe (PDF reference 37-1463) phases instead of the CZTSe phase. Figures 1c and d show a presumably “single phase” CZTSe layer and a decomposed layer with 1 at% Sn (EDX measurement) exhibiting Cu2−xSe and ZnSe phases. Point EDX measurements at a low acceleration voltage of 7 kV limits the analysis volume to ≤1 µm3 and is a direct experimental evidence of the co-existence of individual Cu2−xSe and ZnSe grains.

Another way to control incorporation of Sn is by means of the selenium partial pressure, Fig. 2. This can be investigated by increasing the selenium zone temperature, resulting in a higher selenium vapor pressure, whilst maintaining a constant substrate temperature of 610 °C and a total chamber pressure of 5 mbar. At the lowest selenium zone temperature of 360 °C – corresponding to a selenium evaporation of 170 mg, a significant loss of Sn is observed, which is in accordance to the previously described result in Fig. 1a. As the selenium zone temperature is elevated from 360 to 380 °C, corresponding to the increase in selenium evaporation from 170 to 3410 mg, the Sn content reaches its targeted value of 11 at%.

Figure 2.

(online color at: www.pss-a.com) Elemental concentrations as obtained from EDX for variation in selenium zone temperatures at constant substrate temperature of 610 °C. The growth of MoSe2 is indicated on the right axis.

High selenium vapor pressure promotes the selenization of the molybdenum back contact into MoSe2, see Fig. 2. The right axis is a ratio of XRD intensity counts between (100) MoSe2 over (110) Mo. There is a rapid increase in MoSe2 content for Se zone temperature of 370 °C and higher which is in agreement with Ref. 18. Although the interface MoSe2 layer is needed to provide a necessary quasi-ohmic contact to CIGS 19, and even a 300 nm thick MoSe2 layer is reported for a 9.7% efficient CZTSSe solar cell by Todorov et al. 20, there is a certain risk that the conductivity of the Mo contact might be deteriorated and could result in an increase in resistive losses.

Following our deposition route, CZTSe phase forms gradually during the selenization of Cu, Zn, and Sn metal salts embedded in an organic matrix. Although no dedicated in situ XRD measurements have been conducted, we anticipate that the formation of the CZTSe phase can involve intermediate binary and ternary phases 21, 22. When interrupting the selenization process after several minutes during heating-up at temperatures of 340 and 370 °C ex situ XRD analysis showed the presence of binary CuSe which converts into Cu2−xSe phase, suggesting that kinetically Cu species are the first to react with selenium vapor but thermodynamically CZTSe already forms at 370 °C (see Fig. 1b):

equation image((1))

Once CZTSe is formed, it can decompose according to the chemical equilibrium reaction proposed by Redinger et al. 12

equation image((2))

The decomposition reaction becomes significant under our selenization conditions at temperatures higher than 600 °C. According to the Le Chatelier principle, the reaction equilibrium (2) can be shifted to the left – the CZTSe side – by suppressing the formation of volatile species, SnSe or Se. The interplay between the critical temperature and partial pressure of volatile products is described on the example of CZTS decomposition by Scragg et al. 11, and we assume identical considerations for CZTSe. Since no additional SnSe is provided during selenization in our case, the decomposition reaction can be suppressed by using either low substrate temperature (Fig. 1a) or high substrate temperatures in combination with increased Se partial pressure via elevated Se zone temperature (Fig. 2).

The latter option is beneficial for the recrystallization of CZTSe layers, but it also results in thick MoSe2 layers. In order to find a compromise between the MoSe2 thickness and the CZTSe grain growth, we increased the total chamber pressure to 10 mbar by partially closing the valve between the reactor and the pump (while the N2 flow remained constant). At the same time, the Se zone temperature was increased to evaporate the same amount of selenium as at 5 mbar. These measures result in an elevated effective Se pressure, and hence the tailored Sn content can be preserved even for samples selenized at temperatures >600 °C, which is depicted as open squares in Fig. 1a. In an open reactor, neither the total chamber pressure nor the addition of inert carrier gas should theoretically shift the decomposition equilibrium (2). However, the higher N2 background pressure may retard CZTSe decomposition by slowing down desorption of gaseous SnSe.

In order to prove that the Sn loss in our experiments was taking place via the evaporation of SnSe with a high vapor pressure (∼1.3 × 10−2 mbar at 600 °C 23) and not via any volatile organometallic tin compounds or tin salts, a control heating experiment without selenium atmosphere was performed. Indeed we could not detect any Sn loss, confirming the assumption of the SnSe evaporation. The Sn loss via metallic tin could also be excluded due to its low vapor pressure (∼1.1 × 10−9 mbar at 600 °C 24).

By using the optimized selenization conditions, such as total chamber pressure of 10 mbar and a substrate temperature of 470 °C, crack free absorber layers with overall metal ratios of Cu/(Zn + Sn) = 0.88 and Zn/Sn = 1.17 (EDX data) were obtained and processed to solar cells. The lower Zn content in the selenized absorber as compared to initial precursor is probably due to the evaporation of metallic Zn having the highest vapor pressure of the metal elements (∼22 mbar at 600 °C 24). The Raman spectrum of the absorber layer (Fig. 3) exhibits peaks at 192, 170, and 232 cm−1, in agreement with those reported in Ref. 25. Again, the presence of secondary phases cannot be excluded under the given sensitivity of the Raman measurement.

Figure 3.

Raman spectrum for a Cu-poor and Zn-rich CZTSe layer.

The SEM cross-section image in Fig. 4a shows a ∼1 µm thick CZTSe absorber layer with small grains, rather rough surface, and a ∼2.5 µm thick carbon-rich layer between Mo and CZTSe absorber. This carbon layer stems from the incomplete decomposition of ethyl cellulose binder and organic solvents and consists primarily of carbon with metal impurities 14, 26. Despite the presence of the residual carbon layer, a conversion efficiency of up to 4.28% on a total area of 0.09 cm2 was obtained (80% of the cells exhibited efficiencies ≥3.8% on a 2.5 × 2.5 cm2 substrate). Figure 4b shows the corresponding JV curves under dark and illuminated conditions and Fig. 4c illustrates the external quantum efficiency EQE(λ) of the best cell. The carrier collection is low especially in the long wavelength region. A small minority carrier lifetime and/or too narrow space charge region, which was experimentally verified by EQE(λ) measurements with and without applied bias voltage (not shown), can be suggested as the main reasons for the low red response. An extrapolation of the QE data results in a band gap of about 0.9 eV.

Figure 4.

(online color at: www.pss-a.com) (a) SEM cross-section image of a selenized CZTSe absorber. A porous carbon-rich layer is present between the CZTSe absorber and the Mo back contact. (b) Current density–voltage and (c) EQE curves.

4 Conclusions

We would like to highlight the efficiency of 4.28% for a selenide CZTSe kesterite absorber processed from simple metal salt precursors embedded in an organic matrix despite a thick remaining carbon-rich layer between Mo and CZTSe absorber. We demonstrate that the selenization conditions, such as substrate temperature and selenium partial pressure, are critical for tailoring the film stoichiometry and especially, the final Sn content. Exceeding a substrate temperature of 600 °C for a relatively low Se partial pressure results in severe Sn loss via volatile SnSe. The remaining layer consists of a mixture of binary ZnSe and Cu2−xSe phases, directly confirmed by EDX point measurements. In order to tailor the Sn content in the CZTSe absorber layer, either reduced substrate temperatures of <600 °C or higher selenium pressure have to be employed. Densely packed and crack free crystalline Cu-poor and Zn-rich absorber films were obtained by using a substrate temperature of 470 °C and a total chamber pressure of 10 mbar. Our XRD and Raman measurements, cannot rule out, however, a possible presence of secondary phases in the CZTSe absorber.

For further efficiency improvements the remaining carbon-rich layer should be reduced or avoided, which can be realized by excluding the ethyl cellulose binder material, as demonstrated recently by Uhl et al. 26, 27. Surface sensitive and depth dependent measurements could be conducted to probe possible compositional variations within the absorber layers. Finally, a future work is planned to assess the reaction mechanism and kinetics of the CZTSe formation during precursor selenization.

Acknowledgements

Y.E.R. acknowledges the support of the Swiss National Science Foundation, Project Nr. PZ00P2_126435/1.

Biographical Information

Carolin M. Fella received her Diploma degree in Physics from the Friedrich-Alexander University Erlangen-Nuremberg, Germany. From 2006–2007 she studied economics at the University of Cantabria, Spain. In 2009 she joined the Laboratory for Thin Films and Photovoltaics, at Empa-Swiss Federal Laboratories for Material Science and Technology in Dübendorf, Switzerland, where she did her Diploma Thesis on non-vacuum In2S3 buffer layers for CIGS solar cells. Since 2010 she has been working towards her PhD degree on non-vacuum deposited Cu2ZnSnSe4 absorbers.

original image

Biographical Information

Alexander R. Uhl received his Diploma degree in nanostructural engineering from the Julius-Maximilian University, Würzburg, Germany in 2009. After graduate stays at the University of British Columbia (UBC), Vancouver, Canada, and the University of Uppsala, Sweden, he joined the Laboratory for Thin Films and Photovoltaics, at Empa. Since 2009 he has been working towards the PhD degree in science with the focus on non-vacuum deposition of chalcogenide absorber materials.

original image

Biographical Information

Yaroslav E. Romanyuk received his PhD from the Swiss Federal Institute of Technology, Lausanne in 2005. After his postdoctoral stay at the University of California, Berkeley, he joined Empa as a group leader in the Laboratory for Thin Films and Photovoltaics to research alternative absorbers, TCO and contact materials deposited by vacuum and low-cost, non-vacuum techniques. He holds several patents and has co-authored more than 50 research articles.

original image

Biographical Information

Ayodhya N. Tiwari is the head of the Laboratory for Thin Films and Photovoltaics, Empa, and Titular Professor at ETH Zürich, Switzerland. He has more than 30 years of R&D experience in various photovoltaic technologies. He is a co-author of more than 200 research publications and about 240 conference presentations including numerous invited papers and talks. Important contributions of Tiwari's group include: development of highest record efficiency flexible CIGS and CdTe solar cells; monolithic interconnected flexible solar modules; more than 19% efficiency CIGS and 15.5% efficiency CdTe solar cells on glass with processes suitable for in-line production; simple and safe non-vacuum deposition processes for CIGS and kesterite solar cells.

original image

Ancillary