SEARCH

SEARCH BY CITATION

Keywords:

  • CZTS;
  • Cu2ZnSnS4;
  • kesterite;
  • reactive sputtering;
  • sulfides;
  • thin film solar cells

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Cu2ZnSnS4 (CZTS) is a promising thin-film absorber material that presents some interesting challenges in fabrication when compared with Cu(In,Ga)Se2. We introduce a two-step process for fabrication of CZTS films, involving reactive sputtering of a Cu-Zn-Sn-S precursor followed by rapid annealing. X-ray diffraction and Raman measurements of the sputtered precursor suggest that it is in a disordered, metastable CZTS phase, similar to the high-temperature cubic modification reported for CZTS. A few minutes of annealing at 550 °C are sufficient to produce crystalline CZTS films with grain sizes in the micrometer range. The first reported device using this approach has an AM1.5 efficiency of 4.6%, with Jsc and Voc both appearing to be limited by interface recombination. Copyright © 2012 John Wiley & Sons, Ltd.

INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Solar cells based on Cu2ZnSn(S,Se)4 (CZTSSe) are showing rapid improvements, but device efficiencies are still far behind those of the related compound Cu(In,Ga)Se2 (CIGS). When compared with CIGS, the replacement of indium (and gallium) by zinc and tin engenders two fundamental problems for fabrication. First, as no ternary metallic Cu-Zn-Sn phases exist [1], any fabrication route that relies on thermal sulfurization or selenization of a metallic film will inevitably suffer from phase separation of the metals during processing, making the formation of undesired secondary phases such as Cu2SnS3 (CTS) more likely. Second, due to the relative instability of Sn(IV) compared with In(III) (or Ga(III)), chalcogen loss (leading also to Sn loss) during heating is considerably easier for CZTSSe when compared with CIGS [2-4]. These basic aspects of CZTSSe chemistry suggest that the fabrication methods that produce highly efficient CIGS or CuInS2 may not perform as well for CZTSSe. However, they also assist us in the search for the “best” fabrication route for CZTSSe, which should involve the following: (i) simultaneous deposition of all the metal elements and chalcogens—to avoid phase separation—and (ii) supply of excess chalcogen (and ideally tin sulfide/selenide) vapor at all times during annealing—to suppress chalcogen and Sn loss [2]. If these two criteria cannot be met within a single system, then deposition and annealing must be performed in separate stages. The CZTSSe literature to-date shows that the most successful methods do indeed adhere to these principles [5-7]. Here, we introduce a sputtering-based fabrication route that is not only industrially relevant but also takes into consideration the particular chemistry of CZTSSe. We use reactive sputtering to prepare fully sulfurized Cu-Zn-Sn-S precursor films at low temperature. After that, a short anneal is sufficient to induce crystallization of the Cu2ZnSnS4 (CZTS) phase, which results in a dramatic increase in grain size. We present characterization of the precursor and annealed CZTS films, as well as of solar cells produced from them.

EXPERIMENTAL

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Substrates were prepared by direct current (DC) sputtering of Mo onto cleaned soda-lime glass slides. Cu-Zn-Sn-S precursor films were deposited onto the substrates by reactive pulsed DC magnetron co-sputtering using metallic Cu : Sn (2:1) alloy and pure Zn targets in an atmosphere of H2S. The target powers were 480 and 530 W, respectively, with a pulsing frequency of 20 kHz. The base pressure of the system (Von Ardenne CS 600) was below 10−4 Pa, and the sputtering pressure was 0.7 Pa with a H2S flow rate of 20 sccm. The substrates were heated to a temperature of approximately 120 °C during deposition.

The precursors were annealed in a tube furnace under a static argon atmosphere (0.3 atm). The base pressure was below 0.1 Pa. Sample temperature was estimated from a thermocouple embedded in the sample holder, a Ti plate. To achieve a high heating rate, the samples were transferred directly from an adjoining water-cooled zone into the preheated furnace, whereupon the sample temperature rose to 550 °C within 3 min. In this study, a dwell period of 3 min was used, during which the samples reached a maximum temperature of 557 °C. Then, the samples were withdrawn to the cold zone, where they cooled to below 200 °C within 2.5 min.

The precursor composition was measured using a LEO 440 SEM with an energy dispersive X-ray spectroscopy (EDS) system. To avoid the problem of overlapping Mo Kα and S Lα lines, measurements were made on precursors sputtered directly onto silicon wafer. Without the Mo layer, delamination of the CZTS film occurs during annealing, so comparable measurements on annealed samples are not possible. The metallic composition of the annealed films was measured by X-ray fluorescence spectroscopy (XRF, PANalytical Epsilon 5, Almelo, The Netherlands) calibrated by Rutherford backscattering spectrometry. Scanning electron microscopy (SEM) was carried out using a LEO 1550 (Carl Zeiss AG, Oberkochen, Germany). X-ray photoelectron spectroscopy (XPS) was performed in a Quantum 2000 Scanning ESCA Microprobe from Physical Electronics using mono-chromatic Al Kα radiation (1486.7 eV) and depth profiling using an argon ion gun at an energy of 0.5 keV. X-ray diffraction (XRD) was recorded using a Siemens D5000 (Siemens, Berlin, Germany). A grazing incidence angle of 1° was used to enhance the diffracted intensity from the CZTS film; the presence of the substrate Mo peak in the diffractogram indicates that the entire film depth is probed. Raman spectra with an excitation wavelength of 514 nm were recorded using a Renishaw system (Renishaw, Gloucestershire, UK). Spectra with an excitation wavelength of 325 nm were recorded with a LabRam Horiba HR800-UV Jobin-Yvon spectrometer (Horiba, Jobin-Yvon, Inc., Longjumeau, France) with duo-scan option (measured size on sample 30 × 30 µm using a 1 µm size laser spot with an excitation power of 0.4 mW. A Veeco Dektak 150 stylus profilometer (Veeco Instruments, Inc. Plainview, NY, USA) was used to measure film thicknesses.

Samples for devices were etched in a 5 wt% potassium cyanide solution for 2 min. Device finishing was carried out with a cadmium sulfide buffer layer, intrinsic and aluminum-doped zinc oxide window and transparent conductive oxide layers and a Ni/Al/Ni contact grid. Our baseline device finishing processes are described in detail elsewhere [8]. Quantum efficiency (QE) measurements, calibrated using an externally calibrated Hamamatsu Si solar cell, were performed with and without bias. Band gaps were calculated by plotting (QE)2 versus and extrapolating the linear region to the abscissa. Current–voltage (J–V) measurements were made under a halogen lamp with a cold mirror using a temperature controlled sample stage. To minimize errors due to mismatch between the halogen lamp and the AM1.5 spectrum, the intensity of the J–V lamp was adjusted to obtain the AM1.5 short circuit current density obtained from QE, after correcting for grid shading. Dark and illuminated temperature dependent J–V measurements were performed in a liquid nitrogen-cooled cryostat at temperatures ranging from 100 to 340 K. Illumination was obtained from three light emitting diodes (blue, green and red) with a total intensity adjusted to give a short circuit current density from the CZTS cell equal to that obtained from the QE measurement.

RESULTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

To determine whether the as-sputtered precursors were fully sulfurized, we used the ratio Rs = 2 S/(Cu + 2Zn + 4Sn), where Cu, Zn, Sn and S are the measured atomic fractions from EDS and the coefficients refer to the oxidation states in Cu2ZnSnS4 (i.e., Cu(I), Zn(II), Sn(IV) and S(−II)). This is a more useful way to describe the sulfur content than S alone because it accounts for the chemical incorporation of sulfur in bonding with Cu, Zn and Sn. Rs will have a value of 1 if there is sufficient sulfur to form the required oxidation states, even when the material is non-stoichiometric (as is often the case). If Rs is less than 1, the reaction with sulfur is incomplete. In the present case, Rs had a value of 1.04 ± 0.06, indicating that the precursor is indeed fully sulfurized.

The SEM images in Figure 1 show that the precursor film is dense and fine-grained, with a columnar structure. The diameter of the grains as viewed from above is around 30–70 nm. The film thickness measured using profilometry was 1.0 µm. The as-sputtered precursor film shows specular reflectivity.

image

Figure 1. (a) Top view and (b) cross section of as-sputtered Cu-Zn-Sn-S precursor film.

Download figure to PowerPoint

The film morphology undergoes dramatic changes during annealing. The annealed film exhibits large grains, which span the entire film thickness, Figure 2. Typical grain diameters when seen from above are in the range 0.2–1.5 µm, with the largest grains reaching around 2 µm in diameter. Between the large grains, smaller grains of a different phase are observable. EDS measurements show that these are richer in Zn than their surroundings and probably correspond to ZnS, expected given the Zn-rich composition of the sample (Cu/Sn = 1.92, Zn/Sn = 1.31).

image

Figure 2. (a) Top view and (b) cross section of annealed Cu2ZnSnS4 film.

Download figure to PowerPoint

X-ray photoelectron spectroscopy sputter depth profiles of precursor and annealed samples (Figures 3(a) and (b)) indicate that the film composition is rather uniform with depth. The precursor has a slight gradient in Cu and Sn contents. This gradient is no longer present in the annealed sample.

image

Figure 3. X-ray photoelectron spectroscopy sputter depth profiles of (a) the Cu-Zn-Sn-S precursor and (b) the annealed film.

Download figure to PowerPoint

Grazing incidence XRD of the precursor film yields a strong diffraction pattern with rather broad reflections (Figure 4). In the precursor, only the main reflections expected from a zinc-blende structure are seen (Figure 4), with none of the tetragonal peaks that would be expected for the kesterite CZTS phase or ternary Cu-Sn-S phases. No binary Cu-S or Sn-S phases are seen. We note that in samples made under the same conditions, but with less than the stoichiometric amount of sulfur, we see metallic phases in XRD (Cu-Zn alloy and elemental Sn), but these phases are not observed in fully sulfurized samples such as those presented here. One additional peak occurs as a shoulder below the (112) reflection (marked in the plot as †). This peak could not be definitively assigned, although we note that it is also present in reactively sputtered pure ZnS films produced in the same system, and coincides with the (100) peak of hexagonal ZnS (wurtzite). Scherrer analysis of the XRD peaks yielded a crystallite size of roughly 20 nm (calculated from a full width at half maximum of 0.5° (2θ) for the (112) peak, using the peak width for the annealed sample as a reference for instrument broadening). This result agrees well with the observed particle diameter from SEM (Figure 1). After annealing the precursor for 3 min above 550 °C, the XRD peaks sharpen considerably, and we find an excellent match to the kesterite phase, with all of the minor reflections [9].

image

Figure 4. Grazing incidence X-ray diffractograms of (a) the as-sputtered Cu-Zn-Sn-S precursor, and (b) the annealed Cu2ZnSnS4 film. Peaks marked “*” arise from the Mo substrate.

Download figure to PowerPoint

Raman spectroscopy of the precursor film measured with a 514 nm excitation wavelength shows a broad dominant band at 335 cm−1 and a second band with lower intensity at about 300 cm−1 (Figure 5(a)). Fitting also suggests the presence of additional contributions at 257, 357 and 376 cm−1. The dominant band most closely matches CZTS, with a small shift compared with the reference (338 cm−1, [10]). The band at ~300 cm−1 could be attributed to the reported cubic Cu2SnS3 (CTS) phase [11]. To help assign the Raman modes, we also prepared Cu-Sn-S precursors using the same reactive sputtering method. The Raman data for the Cu-Sn-S sample (with a Cu/Sn ratio of 1.76), are shown in Figure 5(b), and we see a broad superposition of bands between 295 and 356 cm−1 encompassing all the modes expected from the cubic, tetragonal and monoclinic Cu2SnS3 (CTS) phases [11, 12]. When we compare the Cu-Sn-S and Cu-Zn-Sn-S precursor spectra, we see that the Cu-Zn-Sn-S precursor Raman cannot simply be described as a mixture of CTS and ZnS. In both precursors, the broadening of the Raman bands is indicative of a small phonon correlation length [13]. This is not surprising given the low deposition temperature.

image

Figure 5. Raman spectra recorded with a 514 nm excitation wavelength for the following: (a) the as-sputtered Cu-Zn-Sn-S precursor and the annealed Cu2ZnSnS4 film, with the spectrum of the substrate after film removal shown as an inset, and (b) a Cu-Sn-S precursor and the corresponding annealed film. The literature peak positions for the different Cu2SnS3 phases described in the text are marked.

Download figure to PowerPoint

Upon annealing of the Cu-Zn-Sn-S precursor, the Raman bands become much narrower, and both the characteristic A1 modes of kesterite CZTS are clear (Figure 5(a)). Weaker contributions are observed at 167, 250, 269, 350, 336 and 374 cm−1, close to frequencies characteristic of E and B symmetry modes from CZTS [14, 15]. The shoulder at around 350 cm−1 may also arise partly from ZnS, which has its main vibrational mode in this spectral region [16, 17]. The same sharpening of the spectrum occurs for the annealed Cu-Sn-S sample, and we can now see clear bands at 298 and 356 cm−1, that are close to the main vibrational modes reported at 303 and 356 cm−1 for the cubic Cu2SnS3 phase [11]. It is clear that in the annealed CZTS film, any contribution from CTS must be rather weak because there is no peak visible at around 298 cm−1. The annealed CZTS film was scraped away to expose the Mo back contact. The Raman spectrum thus recorded (shown as inset to Figure 5(a)), showed peaks at 412, 385 and 287 cm−1, which reveals the presence of an interfacial MoS2 layer in the annealed sample [18], as frequently found in CZTS devices (see for example [5]). This layer was not observable in SEM images, meaning that its thickness is probably less than 10 nm. No such layer was observed for the precursor, showing that MoS2 formation occurred during annealing, not during reactive sputtering.

To determine whether ZnS was present in the samples, Raman spectra were also recorded using a 325 nm (UV) excitation wavelength, and are shown in Figure 6. A pre-resonant excitation of ZnS that takes place under these excitation conditions should cause a strong increase in the intensity of the main vibrational mode from ZnS (at 348 cm−1) [17], as well as the appearance of peaks at 696 cm−1 and 1045 cm−1 that are identified with second-order and third-order ZnS bands [14]. In the CZTS precursor, there was no increase in the contribution at around 348 cm−1, and no higher-order contributions. This strongly suggests the absence of the ZnS phase in the precursor, at least at the surface region of the film. Conversely, for the annealed CZTS film, the spectrum measured under UV excitation clearly shows the presence of ZnS, expected given the Zn rich composition and from the EDS measurements earlier. The information depth of the Raman measurement is relatively modest (around 150 nm using a 514 nm excitation wavelength, and an estimated 50 nm with a 325 nm excitation wavelength [11]). To be sure that the measurements are representative of the entire film, we have also recorded spectra on cross sections of the precursor and annealed samples. The results were identical, to the surface measurements, showing that there is no variation in phase composition with depth (see supporting information). This is consistent with the uniform composition profiles from XPS.

image

Figure 6. Raman spectra recorded with a 325 nm excitation wavelength for the Cu-Zn-Sn-S precursor and corresponding annealed film, showing the presence of a pre-resonant excitation of ZnS in the annealed sample, but not in the precursor. The literature peak positions of the main LO ZnS mode are marked for the cubic (solid line) and hexagonal (dashed line) phases [16, 17]. Peaks at higher frequencies are attributed to second and third order ZnS bands.

Download figure to PowerPoint

Devices were made from both an as-sputtered Cu-Zn-Sn-S precursor and the corresponding annealed film. J–V curves are shown in Figure 7, and the device parameters in Table 1. Despite the low temperature deposition, the precursor still yielded a functioning photovoltaic device, with rectifying behavior in the dark and a non-zero photocurrent. A high series resistance seems responsible for the very poor fill factor. QE measurements (Figure 8) reveal a sloping band edge with an onset at around 1.35 eV, significantly lower than the band gap of CZTS (1.5 eV [19]). A 3-min anneal at 550 °C greatly improved the device characteristics, resulting in a total area efficiency of 4.6% in the best cell. All device parameters are reduced compared with those of the best published pure sulfide CZTS device (Table 1) [5]. Losses in the short circuit current density are both due to a reduced overall QE level and an increased loss for longer wavelengths. The latter suggests influence from recombination at the back contact or poor minority carrier diffusion length whereas a reduced level for all wavelengths could be due to interface recombination. By comparing dark and illuminated current voltage characteristics, a slight voltage-dependent current collection can be noted, expected to contribute to losses in fill factor. The voltage dependence was confirmed from reverse biased QE measurement, where a relatively constant increase in QE was seen for wavelengths above 500 nm. The open circuit voltage is rather low given the band gap, which appears to be a common problem in CZTS devices.

image

Figure 7. Current–voltage curve for devices made from a reactively sputtered Cu-Zn-Sn-S precursor and a rapidly annealed Cu2ZnSnS4 film.

Download figure to PowerPoint

Table 1. Parameters for devices made from a Cu-Zn-Sn-S precursor and corresponding annealed Cu2ZnSnS4 film, as well as those for the best published pure sulfide device.
SampleVoc [mV]Jsc [mA cm−2]FF [%]Efficiency [%]Eg [eV]
Precursor1836.2227.90.311.35
Annealed (3 min at 550 °C)51314.660.84.61.43
Best published CZTS device [5]66119.565.88.4Not given
image

Figure 8. Quantum efficiency spectra for devices made from a reactively sputtered Cu-Zn-Sn-S precursor and a rapidly annealed Cu2ZnSnS4 film.

Download figure to PowerPoint

The evolution of the open circuit voltage from temperature dependent current–voltage measurements under illumination (IVT) for the annealed device is shown in Figure 9. In accordance with previous reports for CZTS [20], a maximum in efficiency is observed at around 170 K, owing to strongly increasing series resistance and reduced Jsc for the lowest temperatures. Extrapolation of Voc to 0 K gives an activation energy for recombination Ea = qV of ~1.0 eV. An activation energy smaller than the band gap of the absorber indicates that Voc is limited by interface recombination rather than recombination in the bulk of the absorber layer [21]. This is consistent with previous results on CZTS thin film devices, although we note that in monograin-based CZTSSe solar cells, an activation energy equal to the band gap has been reported [22]. Because both Voc and Jsc appear limited by interface recombination in the present device, interface improvements are expected to be the most critical issue in further process development.

image

Figure 9. The temperature dependence of open circuit voltage for the device made from an annealed Cu2ZnSnS4 film, showing extrapolation of the linear region to determine the activation energy for recombination.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

The nature of the reactively sputtered Cu-Zn-Sn-S precursor films is interesting. They are very uniform, have a granular morphology and yield a clear X-ray diffraction pattern, indicating that they are mainly crystalline despite the low temperature of their deposition. Assignment of the particular crystalline phases present in the reactively sputtered precursor is an important problem for which Raman measurements provide useful insight. First and foremost, it is clear that the Cu-Zn-Sn-S precursor Raman spectrum most closely resembles CZTS. However, the broadness of the peaks, indicative of a short phonon correlation length arising from a small crystallite size or high defect density, makes it difficult to judge the phase purity of the sample. The most likely secondary phases are ZnS and Cu2SnS3 (CTS). Despite the fact that the precursor was Zn-rich, the presence of ZnS was ruled out by UV Raman measurements. Large amounts of Cu2SnS3 can also be ruled out because, in the absence of ZnS, this would be inconsistent with the overall composition of the sample. We conclude from the Raman measurements that the precursor sample is primarily composed of a CZTS-like phase. On the other hand, returning to the XRD data for the precursor, only the characteristic peaks of a cubic zinc-blende structure are seen, with none of the peaks expected for the lower symmetry of CZTS (or CTS). This is not simply a side effect of peak broadening, because it is not only the smallest peaks that are absent. Cubic symmetry could arise in CZTS (or CTS) if the cation substructure was disordered, which would cause the symmetry of the parent lattice to be recovered (both CZTS and CTS are derived from the cubic zinc-blende structure by an ordered substitution of Zn2+ cations for Cu+ and Sn4+ cations [23]). This phenomenon is well-known in chalcopyrite materials such as CuInSe2, where a high-temperature tetragonal-to-cubic phase transition results from cation randomization [24]. Such a transition is also found for CZTS above 866 °C [9]. Cation disorder can also occur at lower temperatures when non-equilibrium synthesis techniques, such as sputtering, are used [25]. One very pertinent example is the observation of Cu–Au type cation ordering in reactively sputtered CuInS2 prepared below 420 °C [26]. Cation disorder has also been reported for Cu-Sn-S phases [27], and has been anticipated by computational modeling of CTS and CZTS [28, 29]. Given all of these, and considering the non-equilibrium nature of our precursor deposition process, it is reasonable to invoke cation disorder to explain the XRD data here. This hypothesis is also fully consistent with the Raman data. In all of the ZnS-derived structures, S anions are tetrahedrally coordinated by four cations. With a 2:1:1 Cu : Zn : Sn composition, the most statistically likely combination of four cations around an S anion is “Cu2ZnSn”, just as in the kesterite structure. Thus, the highest Raman intensity is obtained in the region of the CZTS sulfur “breathing” mode at around 338 cm−1 [30]. However, the lack of long-range order will restrict the phonon correlation length, giving broad bands and a peak shift [13]. This hypothesis also helps to explain why ZnS is not observed, even in the Zn-rich precursors: statistically speaking, “Zn4” tetrahedra around S anions have a low probability of arising. Much more likely is the formation of “Cu2Sn2” or “Cu3Sn” tetrahedra, which are motifs of CTS structures [29]. These could then be responsible for the weak Raman mode at around 300 cm−1. Furthermore, in the Cu-Sn-S precursors, we saw a superposition of several different phases; this observation is again consistent with the proposed disordered cation substructure. Cation disorder in chalcopyrites has a large influence on band gap: the disordered phases of CuInS2 and CuInSe2 have band gaps around 0.5 eV lower than the ordered phases [24]. This could explain why the band gap of the precursor in the present case is much lower than that of the annealed sample. At room temperature, we can describe these disordered structures as metastable; they only exist because there is not enough kinetic energy available to form the ground state, ordered kesterite phase.

Turning to the annealing step, we observe the rapid growth of CZTS grains upon heating, alongside the formation of ZnS. The final grain size spans the entire film thickness, which may be beneficial for charge carrier transport in the device. Upon annealing, a considerable sharpening of the Raman and XRD peaks occurs, consistent with a decrease in disorder and increase in grain size. The emergence of the tetragonal symmetry of CZTS is also observed. The observed rate of grain growth is approximately 200 times faster than has been reported for a normal growth process during isothermal annealing at 580 °C [31]. This may be related to the structure of the precursor. Prior to grain growth, formation of nuclei must occur, which requires that all elements of the phase concerned are intimately mixed. Because the precursor is not only well-mixed but in fact already consists of a CZTS-like phase, nucleation of true CZTS grains will not face large kinetic barriers. The normal driving force for grain growth is a reduction in grain boundary area, but, in this case, we have an additional contribution from the difference in free energy between the ordered and disordered CZTS phases. This, along with the lack of kinetic barriers, may explain the rapid rate of grain growth. A proposed model for grain growth from the reactively sputtered precursors is the spontaneous formation of small domains of kesterite ordering within the disordered precursor grains or at grain boundaries, followed by an expansion of the ordered region by consumption of the disordered material that surrounds it (similar to a classical recrystallization process, see, for example, [32]). Because the samples are Zn rich, excess Zn that cannot be incorporated in the growing grains is rejected as ZnS particles in between the larger CZTS grains, as observed by SEM–EDS and Raman measurements of the annealed sample. The presence of secondary phase particles generally inhibits grain growth by blocking the movement of grain boundaries; so-called Zener pinning [32]. Therefore, a reduction of the Zn content might be beneficial for a further increase in CZTS grain size.

To determine if rapid heating was a contributing factor in grain growth, an identical precursor sample was annealed with heating and cooling rates of approximately 5 °C min−1. A similar morphology was observed (see Supporting Information), and therefore, we can rule out a direct effect of the rapid thermal processing. The occurrence of voids at the back contact seems to be more severe in this sample, which may be because more time was allowed for vacancy diffusion and coalescence prior to CZTS grain growth [31].

Early device results are promising, given that the annealing process is not yet optimized. Further increases in band gap and Voc occur at anneal times up to 10 min, although overall device performance suffers from a loss in shunt resistance. The time evolution of device performance will be discussed in detail in a future publication. QE and IVT measurements indicate that the device performance is probably limited by interface recombination. If this can be addressed, then the device efficiency could be improved considerably.

CONCLUSIONS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Taking into account the particular chemistry of Cu2ZnSnS4 (CZTS), two-stage processing—that is, low temperature deposition of a Cu-Zn-Sn-S precursor followed by annealing—seems to be the most promising approach for film synthesis. Here, we prepared CZTS films by rapid annealing of reactively sputtered Cu-Zn-Sn-S precursors. Surprisingly, given the low deposition temperature (120 °C), Raman measurements suggested the sputtered precursors were primarily composed of a CZTS-like phase with a high-defect density. XRD showed that this phase had a cubic symmetry. These data can be explained by metastable disorder in the cation substructure of the precursor. Annealing the precursors at around 550 °C for 3 min gave a remarkable increase in grain size, resulting in grain diameters similar to the film thickness (1 µm). The rapid grain growth may be linked to the metastable nature of the precursor. ZnS particles segregated between CZTS grains during annealing, and a thin MoS2 layer formed at the back contact. Initial device results are promising, with a best cell efficiency of 4.6%. It appears from QE and temperature dependent J–V measurements that Jsc and Voc are both limited by interface recombination, suggesting that the heterojunction with CdS needs attention in order to improve the devices.

ACKNOWLEDGEMENT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

The authors are grateful to the Swedish Energy Agency and the Carl Trygger Foundation for funding this work. Tobias Törndahl is thanked for useful discussions about XRD, and Piotr Szaniawski for assistance with IVT measurements. Authors from IREC and Univ. de Barcelona acknowledge funding support from “Ministerio de Economía y Competitividad” project KEST-PV (ref. ENE2010-121541-C03-01/02)

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Supporting Information

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. ACKNOWLEDGEMENT
  9. REFERENCES
  10. Supporting Information

Supporting information may be found in the online version of this article.

FilenameFormatSizeDescription
pip_2265_supporting_information.pdfPDF document168KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.