Germanium Alloyed Kesterite Solar Cells with Low Voltage Deficits

Article pubs.acs.org/cm

Germanium Alloyed Kesterite Solar Cells with Low Voltage Deficits A. D. Collord and H. W. Hillhouse* University of Washington, Department of Chemical Engineering, Seattle, Washington 98195, United States S Supporting Information *

ABSTRACT: It has previously been shown that substitution of germanium for tin in Cu2ZnSn(S,Se)4 provides a means to change the band gap. Here, we show that Ge substitution can also be used to improve carrier collection and decrease the open-circuit voltage deficit that has hindered kesterites. Using a simple molecular ink, we spray coat continuous composition gradients of Cu2Zn(Sn,Ge)(S,Se)4 spanning 0−90% Ge/(Ge + Sn). By mapping the absolute intensity photoluminescence and making devices from these gradients, we are able to clearly resolve changes in material properties and device performance with composition. We find that Ge can be used to increase the band gap as high as 1.3 eV (50% Ge/(Ge + Sn)) without any loss in optoelectronic material quality, but beyond 50% Ge/(Ge + Sn), the device efficiency decreases rapidly. We show evidence that the degradation results from both a deep defect located about 0.8 eV above the valence band and unfavorable band alignment. As the band gap increases, the defect moves toward midgap and becomes a stronger recombination site. After the completed devices were aged for 1 month in the ambient laboratory environment, the deep defect heals, and the Voc improves by as much as 227 mV. We demonstrate an 11.0% efficient spraycoated CZTGSSe device, without antireflective coating, that achieves 63% of the theoretical Voc as compared to the 58% for the current record device.

■

Survey notes that “the amount of germanium potentially recoverable from coal fly ash is essentially unlimited”, but demand for germanium has been too weak to prompt largescale production from coal ash.9 Here, we present a composition-spread study looking at the effects of Ge-alloying on the properties of sulfoselenide kesterites. Using spray-coated composition gradients, we show that there are significant losses in the optoelectronic properties at Ge concentrations higher than about 50% Ge. We use absolute intensity photoluminescence (AIPL)10 to demonstrate that the degradation at high Ge content results from both unfavorable band alignment and a deep defect near 0.8 eV. However, we also show that the deep defect can be passivated after aging in ambient conditions, leading to improvements in the Voc as large as 227 mV (an 84% increase). After aging, devices from spray-coated layers reach efficiencies as high as 11%, which is the highest reported value for any Gecontaining kesterite.

INTRODUCTION Thin film solar cells based on the Cu2ZnSnS4, Cu2ZnSnSe4, and the mixed sulfoselenide Cu2ZnSn(S,Se)4 are promising candidates for terawatt-scale photovoltaic deployment. The constituent elements Cu, Zn, and Sn are abundant primary metals, and thus should not be constrained by issues of availability. To date, lab-scale devices using these materials have achieved power conversion efficiencies of 12.6%.1 The record devices, and most of the highest efficiency devices, have been those containing primarily selenium (with less than 20% S/(S + Se). The best reported pure sulfide device is only 9.1% efficient.2 This may be due to superior defect chemistry of the selenium-containing compounds.3,4 However, higher selenium content materials have smaller band gaps,3,5 which is less desirable from a balance-of-systems perspective. One method for increasing the band gap of kesterite materials is to alloy with germanium, forming Cu2Zn(SnxGe1−x)Se4. By doing this, the band gap can be continuously varied from about 1.0 to 1.5 eV,6,7 the optimum range for photovoltaic devices. Analogous to gallium-alloying on the indium site in CuInSe2,8 germanium-alloying on the tin site in Cu2ZnSn(S,Se)4 may also provide a means to reduce the defect density and create a back-surface field.7 However, the Gecontaining kesterites have been much less studied than their tin-containing relatives. This may be partly due to a perception that germanium is a less viable material for earth-abundant PV. Germanium is not a primary metal like tin, but rather a byproduct of zinc production.9 However, both Ge and Sn have similar abundance in the earth’s crust. The U.S. Geological © 2016 American Chemical Society

■

EXPERIMENTAL SECTION

Molecular precursor solutions were prepared by mixing CuCl, ZnCl2, SnCl4, GeCl4, and thiourea in dimethyl formamide (DMF). The precursors were mixed targeting a composition of Cu/(Zn + Sn + Ge) = 0.75, Zn/(Ge + Sn) = 1.05, and Ge/(Ge + Sn) of either 1 or 0 (pure tin or pure germanium). A continuous gradient was formed via the Received: December 11, 2015 Revised: February 29, 2016 Published: February 29, 2016 2067

DOI: 10.1021/acs.chemmater.5b04806 Chem. Mater. 2016, 28, 2067−2073

Article

Chemistry of Materials combinatorial mixing of the inks to produce films with increasing Ge/ (Ge + Sn) along the length of the spray line. Films were deposited using a custom-built deposition system based on a Sono-Tek ultrasonic spray coater fitted in a modified glovebox and fed with an array of independently computer-controlled syringe pumps. Mixing due to Taylor dispersion in the tubing is accounted for in the pump program such that the as-deposited ink has the target composition. The films were sprayed onto Mo-coated soda-lime glass at about 300 °C. After deposition, the films were annealed in a tube furnace for 20 min at 550 °C with excess selenium, similar to previous reports.11 The absolute intensity photoluminescence (AIPL) was mapped using a calibrated confocal PL instrument with 785 nm laser excitation. AIPL maps were made from individual spectra collected from a 110 × 110 μm area with a 1.2 mm × 1.2 mm mesh. The selenized films were mapped under vacuum to prevent any degradation due to O2 or H2O exposure. The AIPL data were analyzed using a full spectrum peak fitting algorithm detailed in a previous publication.10 An example AIPL peak fit is shown in Figure S1. The films were processed into completed devices using standard techniques including CBD of CdS (40 nm), RF sputtering of i-ZnO (40 nm) then ITO (250 nm), and thermal evaporation of Ni/Al grids (50 nm/1 μm). Each spray gradient produces approximately 84 devices, each with approximately 0.11 cm2 area. Because of small variations in device area and grid shading, all photocurrent densities are based on the measured nonshaded device area for each device using a calibrated optical microscope. All devices were tested under simulated AM1.5G illumination using a 300 W Xe arc lamp with an AM1.5 filter and calibrated with a certified silicon reference solar cell.

Figure 1. (a) Photograph of molecular inks containing dissolved copper and germanium species in either DMF or DMSO. (b) The same inks shown inverted 12 h after the addition of thiourea. The DMSO ink has formed a rigid gel. (c) Photograph of an ink made using GeBr2. The vial is shown inverted to show that the ink has gelled.

Ge:Cl2:dioxane also avoids the addition of dioxane (which is necessary to stabilize the GeCl2), but because it is a fuming liquid precursor, additional safety measures are required to handle it. These considerations may ultimately be important for process design, but the current experiments suggest that highquality material may be grown from DMF-based inks using either Cu(II) plus Sn(II) or Cu(I) plus Sn(IV) precursors (see Figure S2). Example solutions after adding the copper precursor and after aging the completed solution for 1 week are shown in Figure 1. In the present study, we used the proper oxidation state precursors Cu(I)Cl, Zn(II)Cl2, Sn(IV)Cl4, and Ge(IV)Cl4 in DMF. Composition and Structure of Lateral Composition Graded Absorber Layer. As a first step to understanding how Ge-alloying impacts the Sn-based kesterites, we deposited a gradient where Ge/(Ge + Sn) was continuously varied from 0 to 90%. Following deposition, the gradient was annealed in a Se atmosphere in a graphite enclosure to form the mixed sulfoselenide CZTGSSe. Issues with Sn volatility during high temperature annealing are well-established for kesterite materials.18,19 Consistent with theoretical predictions,20 we find that more Ge is lost than Sn. As shown in Figure 2, the Ge/(Ge + Sn) ratio decreases after selenization while the Cu/ (Zn + Sn + Ge) and the Zn/(Sn + Ge) ratios increase. Longer (or multiple) selenization steps lead to even more Ge loss, as

■

RESULTS AND DISCUSSION Ink Formulation. Several previous publications have demonstrated means to produce thin films of CZTGSSe, but these approaches have been primarily limited to nanocrystal7,12−14 or hydrazine15 based techniques. Here, we present a means to deposit films of CZTGSSe using a simple, nontoxic molecular ink. These inks allow for easy compositional tuning, have near 100% materials utilization, and can be easily scaled. We have previously reported formulation of CZTS molecular inks by dissolving Cu(II)(OAc)2·H2O, ZnCl2, SnCl2·2H2O, and thiourea in dimethyl sulfoxide.16 The formulation of these inks relies on the redox reaction 2Cu(II) + Sn(II) → 2Cu(I) + Sn(IV) to generate species with the proper oxidation state.17 However, attempts to replicate this chemistry using GeBr2, GeI2, or GeCl2:dioxane instead of SnCl2·2H2O have proven unsuccessful. Although the germanium halides all have reasonable solubility in DMSO, the completed inks are not stable. Over time, they turn dark red-brown, gel, and form needle-like precipitates; examples of this are shown in Figure 1b,c. Similar behavior is also observed when the proper oxidation state precursors (no redox reaction required) Ge(IV)Cl4 and Cu(I)Cl are used. A summary of observations is included as Table S1. By substituting dimethyl formamide (DMF) for DMSO, a stable solution can be made using either GeCl2:dioxane or GeBr2. GeI2 forms a translucent yellow solution, but precipitates are clearly present. It is also possible to form a stable solution using the proper oxidation state precursors Ge(IV)Cl4 and CuCl in DMF, as shown in Figure 1b. Using precursors with the proper oxidation state is appealing in that it eliminates the need for the redox reaction to occur between Cu(II) and Sn(II). The redox reaction takes time to complete; may proceed to varying degrees depending on the concentration, time, temperature, etc.; and cannot proceed to completion when formulating off-stoichiometric inks (e.g., when Cu/Sn ≠ 2). The use of Ge(IV)Cl4 instead of

Figure 2. Composition of the gradient before and after selenization. Values were determined using EDS. 2068

DOI: 10.1021/acs.chemmater.5b04806 Chem. Mater. 2016, 28, 2067−2073

Article

Chemistry of Materials

Figure 3. (a) Change in the photoluminescence peak as a function of Ge/(Ge + Sn). (b) Map of the PL peak position showing the onset of the low energy peak. The Sn concentration increases from left to right along the length of the gradient. (c) Increase in the absolute intensity of the peak near 0.8 eV as a function of Ge/(Ge + Sn) ratio. (d) Map of the PL peak position. In both (b) and (d), the germanium concentration increases from right to left.

shown in Figure S3. As a result of these losses, the maximum Ge/(Ge + Sn) value observed in the final film following selenization is only about 80%, and the band gap varies from about 1.0 to 1.4 eV. XRD data collected on the completed devices (SLG/Mo/ CZTGSSe/CdS/ZnO/ITO) confirm that the Ge is indeed being alloyed into the absorber and forming a Cu2Zn(SnxGe1−x)(SySe1−y)4 phase. As shown in Figure S4, there is a clear peak shift with increasing Ge content, and we see the splitting of the (312) and (116) peaks into distinct peaks, as expected. The spectra are slightly shifted to higher angle due to residual sulfur, but all of the peaks can be indexed to the expected phases (including CdS, ZnO, ITO). No evidence of secondary phases was detected, but they cannot be ruled out due to overlap. Optoelectronic Quality of Absorber Layers. Absolute intensity photoluminescence (AIPL) analysis of the gradients reveals significant changes in the radiative recombination pathway with Ge content. Representative spectra from different composition regions on the gradient are shown in Figure 3a. The pure-Sn samples have a peak centered at about 1.04 eV. As the Ge concentration is increased, the band gap increases, and thus the peak shifts to higher energy. However, near 39% Ge, a PL peak begins to emerge near 0.8 eV. As the Ge concentration increases further, the high energy peak decreases until it is below the detection limit, and the broad low energy peak near 0.8 eV increases in intensity. The quasi-Fermi level splitting (QFLS) determined by fitting the AIPL data10 is shown in Figure 3b. Relatively high QFLS is observed for samples with Ge/(Ge + Sn) < 0.4, but at higher Ge content, the QFLS decreases. The exact value cannot be detemined due to the emergence of the PL peak at 0.8 eV. However, we note that the absolute intensity of the 0.8 eV peak emission increases with Ge content and is shown in Figure 3c. By using longer (or additional) selenization times, we have found that the lowenergy peak can be eliminated and replaced by a high energy peak, as summarized in Figure S5. However, as will be shown below, aging at low temperature achieves the same effect without degrading other device performance metrics. Photovoltaic Device Performance. Gradients were processed into completed devices using standard techniques, including CBD of CdS, RF sputtering of i-ZnO/ITO, and then thermal evaporation of Ni/Al grids. A photograph of one of the gradients is shown in Figure S1. Current−voltage (J−V) and external quantum efficiency (EQE) were measured for each device. The current−voltage parameters are summarized in Figure 4. We find that the open-circuit voltage (Voc) of the devices increases with increasing Ge content up until about

Figure 4. Current−voltage parameters before and after the sample was aged in the ambient lab environment. All values are normalized to the theoretical limit for the band gap. The band gap was determined from the long-wavelength EQE decay (see Figure 5).

50% Ge (the un-normalized is shown in Figure S6). However, these gains are entirely offset by the increase in band gap. As shown in Figure 4, the Voc normalized to the theoretical limit for the band gap (Voc/VocSQ) remains approximately constant at 60% up until about 50% Ge. After 50% Ge, the Voc begins rapidly decreasing as does the overall device efficiency. The peak in the efficiency that we observe near 25% Ge (Figure 4) results primarily from changes in the carrier collection (Jsc). The short-circuit current normalized to the theoretical limit for the band gap (Jsc/JscSQ), reaches a maximum near 25% Ge. To determine where the gains in current collection come from, we measured the external quantum efficiency (EQE) along the length of the gradient. The spectra are shown in Figure 5a. From the EQE, we can clearly see the shift in the absorption onset with increasing Ge content. The band gaps were obtained by extrapolating (ln(1 − EQE))2 versus energy for the long-wavelength EQE data (results in Figure 5b). Also shown in Figure 5b is the band gap calculated using the composition data, the theoretically predicted bowing parameter,6 and a constant offset to account for residual sulfur. We find 2069

DOI: 10.1021/acs.chemmater.5b04806 Chem. Mater. 2016, 28, 2067−2073

Article

Chemistry of Materials

Figure 6. Photoluminescence spectra before (red) and after (black) aging in ambient for 1 month. Figure 5. (a) External quantum efficiency of devices with varying Ge/ (Ge + Sn). EQE of the devices with the highest efficiencies (inset). (b) Comparison of the band gap determined from the EQE versus from bulk composition.

that there is good agreement between these two values, further indicating that the Ge is fully alloyed into the lattice. Comparing the EQE of devices with different Ge concentrations, we see that the decrease in Jsc/JscSQ at Ge > 30% results from losses at all wavelengths. The EQE of the 81% Ge device reaches only a maximum of about 50%, while those with 50%. The low energy peak may be eliminated by using longer selenization times or aging the sample in ambient for about 1 month. Thus far, longer selenizations have not led to improved device performance, but after aging, we see significant improvements in the devices. Following aging, we demonstrate an 11.0% efficient spray-coated CZTGSSe that achieves 63% of the theoretical Voc, the best of any reported high efficiency kesterite device.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemmater.5b04806. Photograph of a germanium gradient; example of an AIPL full spectrum fit; quasi-Fermi level splitting from a gradient with different ink formulations; Voc versus Ge/ (Ge + Sn) ratio; change in the PL peaks after each processing step; and plot used for the extrapolation of the band gap from EQE (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 206-685-5257. Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Funding

This work was supported by the DOE SunShot Program DEEE0005321 and NSF Award CBET-1133671. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Wang, W.; Winkler, M. T.; Gunawan, O.; Gokmen, T.; Todorov, T. K.; Zhu, Y.; Mitzi, D. B. Device Characteristics of CZTSSe ThinFilm Solar Cells with 12.6% Efficiency. Adv. Energy Mater. 2014, 4, 1301465. (2) Fukano, T.; Tajima, S.; Ito, T. Enhancement of Conversion Efficiency of Cu2ZnSnS4 Thin Film Solar Cells by Improvement of Sulfurization Conditions. Appl. Phys. Express 2013, 6, 062301. (3) Chen, S.; Walsh, A.; Yang, J.-H.; Gong, X. G.; Sun, L.; Yang, P.X.; Chu, J.-H.; Wei, S.-H. Compositional dependence of structural and electronic properties of Cu2ZnSn(S,Se)4 alloys for thin film solar cells. Phys. Rev. B: Condens. Matter Mater. Phys. 2011, 83, 125201. (4) Chen, S.; Wang, L.-W.; Walsh, A.; Gong, X. G.; Wei, S.-H. Abundance of [CuZn + SnZn] and [2CuZn + SnZn] defect clusters in kesterite solar cells. Appl. Phys. Lett. 2012, 101, 223901−4. (5) Haight, R.; Barkhouse, A.; Gunawan, O.; Shin, B.; Copel, M.; Hopstaken, M.; Mitzi, D. B. Band alignment at the Cu2ZnSn(SxSe1‑x)4/ CdS interface. Appl. Phys. Lett. 2011, 98, 253502−253503. (6) Shu, Q.; Yang, J.-H.; Chen, S.; Huang, B.; Xiang, H.; Gong, X.-G.; Wei, S.-H. Cu2Zn(Sn,Ge)Se4 and Cu2Zn(Sn,Si)Se4 alloys as photovoltaic materials: Structural and electronic properties. Phys. Rev. B: Condens. Matter Mater. Phys. 2013, 87, 115208. 2072

DOI: 10.1021/acs.chemmater.5b04806 Chem. Mater. 2016, 28, 2067−2073

Article

Chemistry of Materials (28) Bar, M.; Schubert, B. A.; Marsen, B.; Wilks, R. G.; Pookpanratana, S.; Blum, M.; Krause, S.; Unold, T.; Yang, W.; Weinhardt, L.; Heske, C.; Schock, H. W. Cliff-like conduction band offset and KCN-induced recombination barrier enhancement at the CdS/Cu2ZnSnS4 thin-film solar cell heterojunction. Appl. Phys. Lett. 2011, 99, 222105. (29) Scheer, R.; Schock, H. W. Chalcogenide Photovoltaics: Physics, Technologies, and Thin Film Devices; John Wiley & Sons: Chichester, U.K., 2011. (30) Halim, M. A.; Islam, M. M.; Xianjia, L.; Chong, X.; Sakurai, T.; Sakai, N.; Kato, T.; Sugimoto, H.; Tampo, H.; Shibata, H.; Niki, S.; Akimoto, K. Study of recombination process in Cu2ZnSnS4 thin film using two-wavelength excited photoluminescence. In Proceedings of the 40th IEEE Photovoltaic Specialists Conference, Denver, CO, June 8−13, 2014; IEEE: New York, 2014; pp 2334−2337. (31) Miller, D. W.; Warren, C. W.; Gunawan, O.; Gokmen, T.; Mitzi, D. B.; Cohen, J. D. Electronically active defects in the Cu2ZnSn(Se,S)4 alloys as revealed by transient photocapacitance spectroscopy. Appl. Phys. Lett. 2012, 101, 142106. (32) Chen, S. Y.; Yang, J. H.; Gong, X. G.; Walsh, A.; Wei, S. H. Intrinsic point defects and complexes in the quaternary kesterite semiconductor Cu2ZnSnS4. Phys. Rev. B: Condens. Matter Mater. Phys. 2010, 81, 245204−245214. (33) Chen, S.; Walsh, A.; Gong, X.-G.; Wei, S.-H. Classification of Lattice Defects in the Kesterite Cu2ZnSnS4 and Cu2ZnSnSe4 EarthAbundant Solar Cell Absorbers. Adv. Mater. 2013, 25, 1522−1539.

2073

DOI: 10.1021/acs.chemmater.5b04806 Chem. Mater. 2016, 28, 2067−2073