Ki-Joong
Kim
abef,
Changqing
Pan
ab,
Shalu
Bansal
c,
Rajiv
Malhotra
c,
Dae-Hwan
Kim
d and
Chih-Hung
Chang
*ab
aSchool of Chemical, Biological & Environmental Engineering, Oregon State University, Corvallis, Oregon 97331, USA. E-mail: chih-hung.chang@oregonstate.edu
bOregon Process Innovation Center, Advanced Technology and Manufacturing Institute, Corvallis, Oregon 97330, USA
cDepartment of Mechanical Engineering, Oregon State University, Corvallis, Oregon 97331, USA
dConvergence Research Center for Solar Energy, Daegu Gyeongbuk Institute of Science & Technology (DGIST), Daegu 42988, South Korea
eNational Energy Technology Laboratory (NETL), U.S. Department of Energy, 626 Cochrans Mill Road, Pittsburgh, Pennsylvania 15236, USA
fAECOM, P.O. Box 618, South Park, Pennsylvania 15216, USA
First published on 4th January 2017
Low-cost materials, scalable manufacturing, and high power conversion efficiency are critical enablers for large-scale applications of photovoltaic (PV) cells. Cu2ZnSn(S,Se)4 (CZTSSe) has emerged as a promising PV material due to its low-cost earth-abundant nature and the low toxicity of its constituents. We present a compact and environmentally friendly route for preparing metal sulfide (metals are Cu, Zn, and Sn) nanoparticles (NPs) and optimize their annealing conditions to obtain uniform carbon-free CZTSSe thin films with large grain sizes. Further, the solution-stable binary NP inks synthesized in an aqueous solution with additives are shown to inhibit the formation of secondary phases during annealing. A laboratory-scale PV cell with a Al/AZO/ZnO/CdS/CZTSSe/Mo-glass structure is fabricated without anti-reflective coatings, and a 9.08% efficiency under AM1.5G illumination is demonstrated for the first time. The developed scalable, energy-efficient, and environmentally sustainable NP synthesis approach can enable integration of NP synthesis with emerging large-area deposition and annealing methods for scalable fabrication of CZTSSe PV cells.
The most efficient CZTSSe thin film PV cell to date, prepared by a hydrazine-based solution process, has demonstrated a device efficiency of over 12%.7,8 However, the high toxicity and potentially explosive nature of hydrazine limit the scalability of this approach.9 An alternative route is to employ solution-based nanoparticles (NPs). Achieving a promising PV efficiency with quaternary CZTS NPs10–17 is difficult due to the limited control of the bulk composition of the thin film. An alternative NP-based approach is selenization (i.e., annealing in the presence of selenium) of binary or ternary metal sulfide NPs, such as Cu2SnS3, Cu2−xS, ZnS, and SnSx, in different stoichiometric ratios.18,19 These NPs are typically synthesized by the hot-injection technique, which is well suited to the production of small amounts of NPs in a batch reactor (typically a few tens of milligrams) but has insufficient yield for most practical applications. Microfluidic flow reactors have been developed for continuous NP production,20–22 and can potentially produce a higher NP yield per unit spatial volume of equipment than larger batch reactors within a relatively smaller volume. However, most work on such reactors has focused on controlling NP properties and the potential for scalable NP synthesis has not been fully exploited.
Another common issue with the NP-based approach is the residual carbon produced in selenization, due to the long organic chains used to chemically bond the as-synthesized NPs.15,18,19 This carbon blocks crystallization and produces multi-layered structures with both large-grain and small-grain layers. This phenomenon significantly deteriorates the film's electronic properties, e.g. resulting in a high series resistance.10,11,23 To improve the device performance by reducing the series resistance and enhancing the absorption of long wavelength photons, it is desirable to eliminate these small-grain layers. It is worth noting that starting with binary NPs which transition through an intermediate liquid phase provides a pathway for rapid mass transport during selenization, which favours the formation of large crystals through Oswald ripening.24–26 This approach assists grain growth at low temperatures and is adopted here.
Here we report a 9.08% power conversion efficiency (PCE) CZTSSe thin-film PV cell fabricated via a precursor mixture of binary NPs, prepared at room temperature in a high-throughput continuous microreactor system (Fig. 1A). Briefly, binary NPs were synthesized by continuously pumping metal precursors and the reducing agent dissolved in water into a micro-mixer at room temperature. A gas–liquid segmented flow system was used by introducing air in order to prevent clogging inside the microchannels of the micromixer. Enabled by the facileness of this continuous aqueous route and the compactness of micromixers, a high NP yield was obtained by simply using multiple micromixers in parallel, e.g., each binary NP was obtained at a rate of over ∼320 g h−1. This ability to increase the NP yield with a relatively smaller increase in the equipment size, as compared to batch reactors, has the potential for direct integration of micromixer-based NP synthesis with large-area NP deposition (e.g., roll-to-roll printing) and annealing methods (e.g., photonic and plasma annealing) without compromising on the speed of the process chain. 5 g of CZTS is needed to fabricate 1 m2 of CZTS thin film solar cells based on a rough estimation. Our three continuous-flow reactors (CuS, SnS and ZnS) combined (i.e. unit) could produce ∼1000 g h−1. Thus, one NP production unit could potentially support a 200 m2 h−1 cell manufacturing throughput. In addition, our reactors could simply be scaled up to match higher cell production throughputs by the combination of channel-up and number-up approaches.
The as-prepared binary NPs were easily aggregated and precipitated due to their poor protection, leading to an unstable solution (Fig. S1 and S2, ESI†). Organic additives including monoethanolamine (MEA) and/or polyvinylpyrrolidone (PVP) were therefore introduced into the precursor solution to create a stable NP solution from the as-synthesized binary NPs. Further, MEA27–29 and/or PVP30 also reduce carbon in the absorber layer and mitigate the residual carbon induced retardation of crystallization mentioned above. The Cu/(Zn + Sn) and Zn/Sn ratios in the precursor ink were targeted at a Cu-poor and Zn-rich stoichiometry (Cu/(Zn + Sn) = 0.85 and Zn/Sn = 1.15) and were determined via elemental analysis (Fig. S3, ESI†). The solutions of binary NPs were mixed with PVP, MEA, and thiourea (TU) in 2-methoxyethanol to form the ink precursor (see the Experimental section in the ESI†). Transmission electron microscopy (TEM) and atomic resolution high-angle annular dark-field (HAADF)-scanning TEM (STEM) analyses show a homogeneous distribution of the four elements Cu, Zn, Sn and S (Fig. S4, ESI†). These precursor inks were spin-coated onto Mo/glass substrates and pre-annealed in air, followed by selenization, to form CZTSSe films (Fig. 1B). The process variables such as the additives, pre-annealing temperature, selenization temperature and time were optimized to obtain single-layer, crystalline, and carbon-free CZTSSe thin films with large grains. This CZTSSe thin film was used to fabricate a PV device consisting of Al/AZO/ZnO/CdS/CZTSSe/Mo-glass, and the device performance was characterized.
Organics such as binders and/or surfactants need to be removed from the precursor films before selenization via pre-annealing in air, to prevent the formation of a residual carbon-rich layer. However, common metal sulfide NPs easily react during pre-annealing resulting in oxygen inclusion into the final film. Therefore, the pre-annealing temperature must be suitably set to achieve carbon-free precursor films without too much oxidation. We examined the effect of pre-annealing temperatures in the range of 250–400 °C in air on the formation of precursor films (e.g. CZTS). Elemental analysis showed that Cu, Zn and Sn contents in the pre-annealed films remained fairly constant when compared to those in the precursor ink. However, there was a slight loss in S due to the formation of metal oxides. To reduce the oxidation of binary metal sulfide NPs at high temperatures, TU was added into the precursor ink as an additional sulfur source to act as a reducing agent. XRD patterns of pre-annealed precursor films showed kesterite CZTS (JCPDS 26-0575) as the majority phase, and minimal phases of metal oxides/sulfides (CuO, SnO2, SnS, and SnS2) at high pre-annealing temperatures (Fig. 2A). The peaks of SnS2 and SnO2 became more dominant, and the CuO phase disappeared at 400 °C. Note that the formation of intermediate phases, related to the Cu, Zn or S-containing compounds e.g. Cu2−xS, ZnS, Cu2SnS3 or even the ZnO phase, was not detected under all reaction conditions in the XRD patterns. For reference, Fig. 2C shows the XRD patterns of each binary NP type synthesized by a continuous flow precipitation method. However, it should be noted that the characteristic peaks of the Cu2SnS3 and the ZnS phase in the XRD pattern overlap with that of the pure CZTS phase,30 and therefore Raman analysis was utilized to assist with the phase identification of the precursor film annealed at 350 °C in air (Fig. 2E). The formation of CZTS was confirmed by the presence of peaks at 288, 338, and 368 cm−1.31 The Raman data also showed the characteristic peaks of the intermediates such as SnS (220 cm−1),32 CuO (295 cm−1),33 SnS2 (315 cm−1),32 Cu3SnS4 (or ZnS, 350 cm−1),34 SnO2 (438 and 601 cm−1),35 and Cu2−xS (473 cm−1).36 This result indicates the formation of CZTS together with Cu3SnS4 (or ZnS) and metal oxides as minor phases and is in good agreement with the XRD result. Therefore, we deduce that with increasing pre-annealing temperature a small amount of binary NPs react through the following oxidation reactions: (1) 2Cu2−xS + 3O2 → 2Cu2−xO + 2SO2 and (2) 2SnS + O2 → SnS2 + SnO2. The chemical conversion of SnS under heating in air is thermodynamically favorable because the enthalpy of formation of SnS is −100 kJ mol−1, while that of SnS2 and SnO2 is −167 kJ mol−1 and −581 kJ mol−1, respectively. Therefore, the SnS2 component persists and the SnO2 component becomes more dominant with an increase in the pre-annealing temperature in air. In addition, a MoS2 phase was formed at 400 °C by reaction with TU in the precursor ink. Thus, the pre-annealing temperature in air should be lower than 400 °C because MoS2 layers may reduce the ohmic contact between CZTSSe and Mo.37
Furthermore, metal oxides and tin sulfides in the precursor film pre-annealed at 350 °C in air disappeared after selenization (Fig. 2B and D) without forming any intermediates, suggesting the successful intercalation of Se ions to form CZTSSe films. CZTSSe films on Mo/glass substrates show a kesterite structure (JCPDF 52-0868) with sharpened peaks, indicating a large degree of growth in the crystal grain size and potential for high-quality PV cells. No Se-containing intermediates such as CuSe, ZnSe and SnSe2 were observed in any of the tested samples, suggesting that the formation of intermediates might be bypassed and the solid reaction pathway is fairly simple. In contrast, CZTSSe films from the precursor film pre-annealed at 400 °C in air still contained SnO2 and SnS2 components, which remained even when more aggressive selenization conditions were used. This high temperature led to an increased Mo(S,Se)2 layer (Fig. S9, ESI†). In addition to this, a poor grain growth of CZTSSe films with weak adhesion was obtained at pre-annealing temperatures below 300 °C in air (Fig. S10, ESI†). Therefore, the precursor films pre-annealed at 350 °C in air were subjected to further investigation on the formation of CZTSSe films after selenization.
Typically, device performance improves as the grain size of CZTSSe layers increases and the thickness of the Mo(S,Se)2 layers reduces,37 both being controlled by selenization conditions. The surface topography and cross-sectional scanning electron microscope (SEM) images of the CZTSSe films selenized at different temperatures and times are shown in Fig. 3. It is obvious that the precursor film pre-annealed at 350 °C in air formed a densely packed uniform layer via consecutive spin-coating. The size of grains gradually increased as the temperature increased; however, heating above 580 °C led to stress-induced cracks. The samples selenized at relatively low temperatures (500 °C and 530 °C) and at short times (10 min and 20 min) exhibited a typical bilayer structure formed by a bottom layer with small grains and a top layer with large grains, along with holes in the film. The films formed at 550 °C for 30 min produced dense, crack-free and pinhole-free single CZTSSe layers (∼1.6 μm) with micrometer-sized grains. Both the XRD and Raman data confirmed that no major crystalline secondary phases were generated under our optimized selenization conditions, at 550 °C for 30 min (Fig. 4). The composition of the CZTSSe films was Cu1.80Zn1.13Sn1(S,Se)4.13, with the concentration ratios of Cu/(Zn + Sn) = 0.85, Zn/Sn = 1.13, and Se/(S + Se) = 0.85. The ratios of Cu/(Zn + Sn) and Zn/Sn in all the CZTSSe samples are almost the same independent of selenization temperatures and times (Fig. S11, ESI†). It was also found that a longer selenization time (50 min) resulted in a higher % Se (96%) in CZTSSe films (Fig. S12, ESI†).
Fig. 4 (A) Logarithmic XRD pattern and (B) Raman spectrum of the optimized CZTSSe film selenized at 550 °C for 30 min. CZTSSe film on a Mo/glass substrate can be identified as kesterite Cu2ZnSnSe4 (JCPDF 52-0868) with the tetragonal lattice constants a = b = 5.6930 Å and c = 11.3330 Å. Peaks arising from the Mo substrate and the Mo(S,Se)2 interlayer are also noted. The main peaks at 173, 193, and 232 cm−1 in the Raman spectrum confirms that CZTSe is the major phase in the sample. The peak at 243 cm−1 is possibly due to the known Mo(S,Se)2.38 In addition, within the resolution of the Raman measurement, no noticeable peaks corresponding to Cu2Se, ZnSe or Cu2SnSe3 were observed,39 which excludes the existence of possible impurity phases after selenization. |
The CZTSSe PV device consisted of Al/AZO/ZnO/CdS/CZTSSe/Mo-glass (see fabrication details in the Experimental section). Cross-sectional SEM images of the device showed dense, uniform films (Fig. 5A). There were also few voids at the CZTSSe/Mo(S,Se)2 interface, which could increase the series resistance. These voids could be effectively prevented by rapid heating in the selenization process.40,41 Atomic resolution HAADF-STEM, energy dispersive X-ray spectroscopy (EDS) line scanning (Fig. 5B) and mappings prepared by using a focused ion beam (FIB) (Fig. 6) confirmed the compositional uniformity of Cu, Zn, Sn, S, and Se, and a single layer CZTSSe film. The CdS buffer layer (∼50 nm) was visible in the S map as well. A slight amount of S was observed in the bulk CZTSSe film, indicating the replacement of S by Se during selenization of the precursor film. Trace amounts of Sn were detected in the Mo(S,Se)2 layer, indicating that the Sn was diffused into the Mo(S,Se)2 layer; however, no evidence was found for the presence of Sn(S,Se)x phases within the detection limit of Raman and XRD (Fig. 4). One hypothesis behind this observation is that the reaction of Sn(S,Se)x phases with Mo is thermodynamically favorable,42 yielding Sn–Mo(S,Se)2. This could be the dominant reaction because it is consistent with the no observation of Cu2−x(S,Se) and/or Zn(S,Se) phases. A 210 nm thick Mo(S,Se)2 layer was formed at the interface of the CZTSSe layer during selenization at 550 °C for 30 min. Detailed EDS compositional profiling across the CZTSSe film thickness also indicated uniform and almost identical ∼30 nm thick atomic % Cu rich layers at both the top and bottom of the CZTSSe film (Fig. S13 and S14, ESI†). While this might be thought to cause secondary phase formation such as Cu2−x(S,Se) which can reduce the performance of the PV cell, the absence of S and/or Se elements suggests a pure Cu layer.43 The X-ray photoelectron spectroscopy (XPS) depth profiles of the CZTSSe PV cell confirmed that Cu is in fact not present at the interface of the CZTSSe layer (Fig. S15, ESI†). Therefore, Cu-rich layers are an artifact caused by Cu migration from CZTSSe produced during the FIB machining.44
Fig. 7A shows the current–voltage (I–V) characteristics of the fabricated PV device. The device exhibited a PCE of over 9% with an open-circuit voltage (VOC) of 447.8 mV, a short-circuit current density (JSC) of 31.5 mA cm−2, and a fill factor (FF) of 0.644 (an active area of 0.09 cm2) under AM1.5 illumination, representing the highest performance efficiency (9.08% PCE) reported to date for NP-based CZTSSe PV cells without an anti-reflective coating. A small difference in JSC between EQE measurements (32.7 mA cm−2) and I–V data (31.5 mA cm−2) was observed. This is attributed to the shadowing induced by the Al top contact, which covers about 3.1% of the unit cell surface. The device showed an external quantum efficiency up to 93% and yielded an estimated CZTSSe band gap of 1.06 eV (Fig. 7B and S16, ESI†). Since the band gap for pure CZTSe and CZTS is commonly reported to be 1.0 and 1.5 eV,45 respectively, this indicates 84% of the Se content in the CZTSSe film after selenization, which is in good agreement with the XRD result (Fig. S12, ESI†). A possible cause for low VOC values compared to those of the record CZTSSe PV cells with a PCE of over 12% (ref. 7 and 8) may be the presence of SnS2 in the Mo(S,Se)2 layer, leading to an increase in the series resistance of the device because of its high band gap (2.2–2.4 eV).46 The effect of this on the contact resistance is not understood at the present time, but it may depend on how the Fermi levels of Sn–Mo(S,Se)2 line up at the junction. There are several factors that affect the PV performance related to the absorber-back contact interface: (1) the partial delamination of the absorber on the Mo substrate does lower the contact area which increases the series resistance, (2) the delamination further creates small grain-size crystals at the interface which increase the chance of carrier recombination, and the device behaves with a low shunt resistance, and (3) voids above the Mo layer could locally increase during selenization which harms the ohmic contact at the interface. The reason for the slightly low JSC value is either a fairly thick AZO window layer (∼435 nm), which partially absorbs incident light, or Al2O3 formation at the interface between the Al top contact and AZO window layer. Therefore, future work will improve the JSC and lower the series resistance of the CZTSSe device.
Fig. 7 (A) The current–voltage (I–V) characteristics under AM1.5 illumination and (B) external quantum efficiency curve of the 9.08% CZTSSe PV device. The dashed line in (B) marks the band gap estimated using [hνln(1 − EQE)]2versus the photon energy (eV) (Fig. S16, ESI†). |
The Cu/(Zn + Sn) and Zn/Sn ratios in the precursor ink were targeted at a Cu-poor and Zn-rich stoichiometry (Cu/(Zn + Sn) = 0.85 and Zn/Sn = 1.15), as determined by the EDS analysis (Fig. S3, ESI†). Each of the three NP types were mixed with TU, MEA, and PVP in 4 mL of 2-MTE solution and then ultra-sonicated for 30 min to form a well-dispersed precursor ink, followed by another 12 h of stirring at room temperature, and then 0.5 mL of MEA was added and stirred.
Footnote |
† Electronic supplementary information (ESI) available: Characterization results (TEM, XRD, UV-Vis, SEM, EDS, TGA-DSC, HADDF-STEM-EDS mappings, XPS, I–V characteristic and band gap measurement) are contained. See DOI: 10.1039/c6se00035e |
This journal is © The Royal Society of Chemistry 2017 |