RESEARCH ARTICLE

8.2% pure selenide kesterite thin-film solar cells fromlarge-area electrodeposited precursorsLaura Vauche1,2*, Lisa Risch1,2, Yudania Sánchez3, Mirjana Dimitrievska3, Marcel Pasquinelli2,Thomas Goislard de Monsabert1, Pierre-Philippe Grand1, Salvador Jaime-Ferrer1 andEdgardo Saucedo3

1 NEXCIS, 190 avenue Célestin Coq, 13790 Rousset, France2 IM2NP - UMR 7334, Aix Marseille Université, Domaine Universitaire de Saint Jérôme, 13397 Marseille, France3 Catalonia Institute for Energy Research (IREC), Jardins de les Dones de Negre 1, 08930 Sant Adrià del Besòs, Barcelona, Spain

ABSTRACT

Cu2ZnSnSe4 solar cell absorbers are synthesized by large-area electrodeposition of metal stack precursors followed byselenization. A champion solar cell exhibits 8.2% power conversion efficiency, a new record for Cu2ZnSnSe4 solar cellsprepared from electrodeposited metallic precursors. Significant improvements of device performance are achieved by theapplication of two etching procedures and buffer layer optimization. These results validate electrodeposition as a crediblealternative to vacuum processes (sputtering, co-evaporation) for earth-abundant thin-film solar cell fabrication at low cost.Copyright © 2015 John Wiley & Sons, Ltd.

KEYWORDS

solar cells; thin films; kesterite; electrodeposition; CZTSe

*Correspondence

Laura Vauche, NEXCIS, 190 avenue Célestin Coq, 13790 Rousset, France.E-mail: laura.vauche@im2np.fr

Received 6 January 2015; Revised 23 March 2015; Accepted 26 May 2015

1. INTRODUCTION

Cu2ZnSn(S,Se)4 kesterite materials have attracted growinginterest over the last years because of their potential for thelow-cost mass production of thin-film solar cells.

The kesterite compounds fulfill main requirements in or-der to compete with more established thin-film technologies:high optical absorption coefficients (>104 cm�1), p-typeconductivity, and direct bandgaps (1.0 eV for Cu2ZnSnSe4(CZTSe), 1.5 eV for Cu2ZnSnS4 (CZTS) [1]), close to theideal value for single-junction solar cells [2]. Promisingkesterite devices were recently demonstrated by Kim et al.with a Cu2ZnSn(S,Se)4 solar cell exhibiting 12.7% [3] re-cord power conversion efficiency.

Taking into account the increasing global energy de-mand, the deployment of environmentally friendly and af-fordable materials is of crucial importance. The traditionalthin-film absorbers Cu(In,Ga)(S,Se)2 (CIG(S,Se)) and CdTecontain the scarce elements indium, gallium, tellurium, andthe toxic cadmium. The associated cost pressure might limitthe future deployment of these technologies. In contrast,

kesterite materials offer the considerable advantage of beingcomposed of earth-abundant elements.

In order to reach high deployment, kesterite-based solarcell techniques should be easily manufacturable at highthroughput and exhibit good cost efficiency. Comparedwith evaporation or sputtering techniques that require vac-uum conditions, non-vacuum deposition methods exhibitnumerous advantages. Indeed, nanoparticles deposition,sol–gel, hydrazine-based, and electrodeposition methodsdemand lower equipment costs, are suitable for large areaand flexible substrates, and offer higher throughput, moreefficient material usage, and lower-temperature processing[4]. Until now, record power conversion efficiencies forkesterite materials were obtained via a wet chemical ap-proach including highly toxic hydrazine [3,5].

Electrodeposition is a well-established process in thesemiconductor industry and already used in pilot linemanufacturing facilities of CIG(S,Se) thin-film solar com-panies. 17.3% (cell) and 14.0% (60 × 120 cm2 module)CIG(S,Se) power conversion efficiencies were obtained atNEXCIS through a full industrial process [6]. Thus,

PROGRESS IN PHOTOVOLTAICS: RESEARCH AND APPLICATIONSProg. Photovolt: Res. Appl. (2015)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/pip.2643

Copyright © 2015 John Wiley & Sons, Ltd.

electrodeposited kesterite is a high-potential candidate forreaching terawatt scale deployment in the marketplace [7].Electrodeposition can be realized at room temperature usingnon-toxic and low-cost metallic salt solutions, which havelong bath lifetimes and a high metal-usage ratio (>90%),thus avoiding waste of resources. Electrodepositedkesterite-based solar cells reached recently efficiencies ashigh as 8.0% for CZTS [8] and CZTSe [9]. Noticeably,7.0% CZTSe [10] and 7.3% CZTS [11] devices were ob-tained from large-area electrodeposited precursors. Here,we also present a kesterite absorber fabrication process in-cluding electrodeposition of the metallic precursor at pre-industrial scale (15 × 15 cm2). CZTSe photovoltaic deviceswere fabricated by the following: (i) electrodeposition ofthin-film metal stacks; (ii) low-temperature annealing of theprecursor; (iii) high-temperature annealing in selenium envi-ronment; (iv) removal of secondary phases by chemicaletching; (v) chemical bath deposition of cadmium sulfidebuffer layer; and (vi) device completion. Process optimizationled to a record 8.2% efficient pure-selenide CZTSe device.

2. EXPERIMENTAL

The metal stack of copper, tin, and zinc was sequentiallyelectrodeposited on 15 × 15-cm2 molybdenum-coatedsoda-lime glass substrates. The metals were depositedfrom commercially available solutions at high-speed rates,180–300 nmmin�1, showing great compatibility with in-dustry high-throughput requirements. Tin was electrode-posited from a tin electrodeposition chemical bath thatcontains SnSO4, H2SO4, and organic additives. Zinc waselectrodeposited from a basic zinc electrodeposition solu-tion that contains Zn(OH)2, NaOH, and organic additives.The copper, tin, and zinc layers were electrodepositedusing direct current and current densities of 5, 10, and20mA cm�2, respectively. Typical deposition times arevery short, 30–90 s, and are adjusted in order to obtainthe desired thickness. By varying the thickness, variousmetal stoichiometries were obtained within the Cu-poorand Zn-rich composition range (Cu/Sn +Zn = 0.70–0.80and Zn/Sn = 1.00–1.50, as measured by X-ray fluores-cence), for which the best device efficiencies have been re-ported until now [12,13].

The metal stack precursors were annealed immediatelyafter electrodeposition at low temperature (200 °C) for30–120min in an air-annealing oven.

The precursors were reactively annealed in presence ofselenium and tin in a tubular furnace capable of working invacuum (10�4mbar) or in an inert gas (Ar) atmosphere[14,15]. The 15 × 15-cm2 precursors were cut into smallerpieces (up to 5 × 5 cm2) in order to fit into the graphitebox used for the annealing, with crucibles containing Sepowder (50mg, Alfa Aesar 99.999% purity) and Sn pow-der (5mg, Alfa Aesar 99.999%). The thermal treatmentapplied is a two-step process already described for theselenization of sputtered metal precursors [15]. Here, thefirst treatment was carried out at 350 °C for 30min (heating

ramp 20 °Cmin�1, total Ar pressure of 1mbar) and thesubsequent second treatment at 550 °C for 15min (heatingramp 20 °Cmin�1, total Ar pressure of 1 bar). After annealing,the furnace cooled down naturally (for ~2–3 h).

The as-annealed layers were submitted to chemicaletching (i) in acidic KMnO4 solution followed by a Na2Setching as described by López-Marino et al., [16] (ii) in(NH4)2S solution for 1min as described by Xie et al., [17](iii) combination of both etchings, or (iv) no etching. Afteretching, the CdS buffer layer was deposited by chemicalbath deposition, followed immediately by pulsed DC-magnetron sputtering deposition of undoped ZnO andIn2O3:Sn (ITO) bilayer (130–280Ω sq�1 sheet resistance)(CT100Alliance Concept, Cran-Gevrier, France). The com-pleted devices were low-temperature annealed at 200 °C for30min. For optoelectronic characterization, 3 × 3-mm2 cellswere scribed using a micro-diamond scriber (MR200, OEG,Frankfurt, Germany), thus avoiding the necessity of metallicgrid deposition onto the ITO surface.

The film thickness and composition were measured byX-ray fluorescence spectroscopy (Fischerscope XDV,Helmut Fischer AG, Hünenberg, Switzerland), which hadbeen calibrated with inductively coupled plasma atomic emis-sion spectroscopy measurements. Composition profiles wereacquired by glow discharge optical emission spectroscopy(GDOES) using a GD Profiler 2 (Horiba Jobin Yvon,Longjumeau, France). X-ray data were collected using a CuKα radiation source with Inel Equinox 3000 equipment (InelArtenay, France). Raman scattering measurements were per-formed in back scattering configuration with a LabRamHR800-UV and T64000 spectrometers (Horiba Jobin Yvon,Longjumeau, France). For the HR800-UV system, diode-pumped solid-state laser with a wavelength of 785.0 nm wasused for excitation. In this system, excitation and light collec-tion were made through an Olympus metallographic micro-scope (Shinjuku, Tokyo, Japan), with a laser spot size onthe order of 1–2μm (depending on the excitation wave-length). To avoid effects in the spectra related to potential mi-croscopic inhomogeneities, the spot was rastered over an areaof 30×30 μm2. Furthermore, the T64000 system workscoupled with an ion-Ar+ laser, and measurements were madewith 457.9-nm excitation line, with a 100-μm spot size on thesample. In all cases and to avoid the presence of thermal ef-fects in the spectra, the power excitation density was around50Wcm�2. The first-order Raman spectrum of monocrystal-line Si was measured as a reference before and after acquisi-tion of each Raman spectrum, and the spectra werecorrected with respect to the Si line at 520 cm�1. Morphol-ogies of the films were examined by scanning electron mi-croscopy (SEM) using a Zeiss Series Auriga field emissionscanning electron microscope (Zeiss, Oberkochen, Germany)at a voltage of 5 kV and a Philips XL40 FEG scanning elec-tron microscope (Philips, Eindhoven, The Netherlands) at avoltage of 10 kV. Atomic compositions were examined byenergy-dispersive X-ray (EDX) analysis with 10–20-kV ac-celeration voltage using an Oxford Instruments X-Max detec-tor (Abingdon, UK). Measurement of the optoelectronicproperties was carried out using a Sun 3000 Class AAA solar

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

simulator from Abet Technologies (uniform illumination areaof 15×15 cm2) (Milford, CT, USA) providing a AM1.5Gspectrum, and a Keithley source meter unit. The spectral re-sponse measurements were performed using a pre-calibratedBentham PVE300 system (Reading, UK), allowing us to ob-tain the external quantum efficiency (EQE) of the cells.

3. RESULTS

3.1. Electrodeposited metallic precursor

3.1.1. Metal stack electrodeposition.Scanning electron microscopy images of the precursor

metal stack surface as shown in Figure 1 reveal compactmorphologies of the surfaces. The thickness of the com-plete precursor metallic stacks varies between 0.5 and0.7 μm. The relative standard deviation (RSD) in thicknessof each individual metallic layer was kept below ±6% over15 × 15-cm2 substrates in order to obtain good precursorcomposition uniformity. Copper and zinc layers were de-posited with good lateral uniformities over 15 × 15 cm2

(RSD< 4%), which is desirable in order to avoid deviationsin composition. However, larger thickness variations of theelectrodeposited tin layer were observed (RSD≈ 5–6%).

Tin layer deposition is reported to be challenging also whenusing physical vapor deposition and results in the formationof films with a rough morphology [14]. Here, the tin layerexhibited small aggregates at the surface, as indicated bythe circle in Figure 1(b), probably because of the oxidationof Sn(II) to Sn(IV) in acidic tin plating baths [18]. As theimpact of the aggregates is not fully understood, theirappearance is limited by changing regularly (monthly) thetin electrodeposition bath. Further improvement in precur-sor morphology, uniformity and resulting device perfor-mance might be achieved optimizing the tin electroplatingchemistry. In this study, in order to improve local uniformityand the inter-mixing of the elements, a low-temperatureannealing was applied on the electrodeposited precursor.

3.1.2. Low-temperature annealing.Highest device efficiencies reported in literature from

electrodeposited stack precursors often require a low-temperature annealing of the precursors (Table I) in orderto improve the intermixing of the elements [11] and thecompactness, homogeneity, and adhesion of the resultingkesterite film to the Mo layer [19] or to increase thesmoothness of the kesterite surface [8]. In our study, themetal stack was low-temperature annealed in air for30–120min. Air annealing is convenient for industrial

Figure 1. Surface scanning electron microscopy images of electrodeposited metal layers: (a) Cu layer deposited on molybdenum onsoda-lime glass (SLG) substrate, (b) Sn layer on SLG/Mo/Cu, (c) Zn layer on SLG/Mo/Cu/Sn. (d) Surface and (e) cross-section image of

the full-stack precursor after pre-alloying at 200 °C.

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

processes. The treatment was carried out at 200 °C in order topromote Cu–Sn and Cu–Zn alloy formation while keeping aflat morphology. Higher temperature treatments (>230 °C)resulted in the formation of bumps on the precursor layer(Figure 2), probably because of the low melting point ofelemental tin (230 °C).

After the 200-°C annealing, alloy formation in the me-tallic precursor was observed. Directly after electrodeposi-tion, only elemental Cu, Sn, and Zn are detected by X-raydiffraction (XRD) analysis, while the alloys Cu6Sn5 andCu5Zn8 and the remaining elemental Sn are detected afterthe pre-alloying [Figure 3(a) and (b)]. Formation of theaforementioned brass (Cu–Zn) and bronze (Cu–Sn) alloysis also reported for sputtered Cu-poor and Zn-rich metallicstacks [20]. Here, alloy formation was also confirmed by achange in morphology, as observed in the cross-sectionalSEM images. More crystalline features appear, forming anapparent double layer after the low-temperature annealing,as shown in Figure 1(e). GDOES [Figure 3(c) and (d)]reveals the migration of zinc—initially at the surface of theprecursor—towards the copper layer. This indicates the for-mation of a Cu–Zn alloy at the back of the precursor layer.The upper part of the precursor consists of a mixture ofCu–Sn alloy and Sn. Oxide formation was not detected bymethods used in this study but cannot be excluded. In sum-mary, the low-temperature annealing transforms the initialMo/Cu/Sn/Zn precursor into a Mo/Cu-Zn/Cu-Sn stack, re-ferred to as Mo/Cu-Sn-Zn.

In contrast to fully intermixed pre-alloyed precursors re-ported in literature [10,11], the precursors observed hereexhibit a bilayer structure, which is probably due to apre-alloying temperature below tin melting point.

Another considerable advantage of this 200-°C annealingstep is the stabilization of the precursor. Indeed, after thepre-alloying, the precursor composition as analyzed byGDOES was stable for a few months, whereas withoutpre-alloying Zn tends to migrate quickly (hours to days) to-wards the copper layer to form CuZn5 alloy [Figure 3(a)]and more slowly (months) to form the more stable Cu5Zn8(Supporting Information S1).

In conclusion, a stable metallic precursor at pre-industrial scale (15 × 15 cm2) is obtained by high-speedstack electrodeposition followed by low-temperature airannealing, which is of great importance for a reproducibleand industry-compatible process.

3.2. Cu2ZnSnSe4 kesterite absorber

3.2.1. Thermal treatment.The alloyed precursors were cut into smaller parts

(below 5 × 5 cm2) for high-temperature reactive annealingin a tubular furnace. The precursors, disposed together withSn and Se powders inside a graphite box, were annealedfollowing a two-step process developed by López-Marinoet al. [15] In a 30-min step at 350 °C under low Ar pressure(1.5mbar), Sn and Se vapors were evaporated into the an-nealing atmosphere [14]. The subsequent step was carriedout at 550 °C and atmospheric Ar pressure (1 bar) for15min to promote grain growth [15]. While a seleniumsource is necessary to form the quaternary kesterite phase,elemental Sn is also introduced in order to limit Sn losses.Indeed, the high volatility of Zn, SnSe, and Se compoundsis well known to cause deviations in composition [21–23].Here, elemental Sn and Se in the initial annealing atmo-sphere quickly form volatile Sn–Se compounds and serveas a chalcogen source for CZTSe formation. High Se andSnSe partial pressures prevent the decomposition reaction

Table I. Low-temperature annealing parameters from literaturefor high-efficiency kesterite devices obtained fromelectrodeposited

Cu/Sn/Zn and Cu/Zn/Sn metal stacks.

Material andreference

Efficiency[%]

Temperature[°C]

Duration[min] Conditions

CZTS [11] 7.3 210–350 30 N2

CZTSe [10] 7.0 360 30CZTS [8] 8.0 310 40–150 Evacuated

glass ampouleCZTS [19] 5.6 350 20–60

Figure 2. Scanning electron microscopy surface images of electrodeposited precursors after pre-alloying at 230 °C.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

of CZTSe at elevated temperature and introduce Sn into aSn-deficient film [22].

After the annealing, ~230-nm MoSe2 and 1.5–2-μmCZTSe layers were obtained, as shown in Figure 4(a).Here, we investigated the influence of the absorber compo-sition on the eventual presence of secondary phases withinand at the limits of the Cu-poor, Zn-rich compositionrange. After annealing, the absorbers showed compositionsof Cu/(Zn + Sn) = 0.67–0.82 and Zn/Sn = 0.86–1.32. Theabsorber’s GDOES depth profile [Figure 4(b)] indicates alower Cu content at the back and front interface comparedwith the bulk suggesting that the composition is not uni-form across the layer and indicating the possible presenceof secondary phases.

3.2.2. Characterization of kesterite and secondaryphases.

Cu2ZnSnSe4 is present as a single phase in a very nar-row region of the phase diagram [24]. Small deviations incomposition or inhomogeneities can strongly influencethe properties of the absorber, noticeably when secondaryphases are formed. Best efficiencies are obtained in theCu-poor, Zn-rich composition range [3,12,13], probablybecause these conditions limit the formation of detrimentalCu-rich and Sn-rich phases [1] and promote the formationof the correct shallow-acceptor intrinsic defects for p-typedoping (VCu, CuZn) [25]. Indeed, the lower bandgap(0.84 eV [26]) ternary compound, Cu2SnSe3, is reportedto limit the open-circuit voltage of the solar cell [27,28].

Figure 3. X-ray diffraction (XRD) and glow discharge optical emission spectroscopy of the metallic precursor: (a) and (c) as electrode-posited and (b) and (d) after 200-°C annealing.

Figure 4. (a) Scanning electron microscopy cross-section and(b) glow discharge optical emission spectroscopy depth profileof an unetched absorber of composition Cu/Zn + Sn = 0.74 and

Zn/Sn = 1.25. CZTSe, Cu2ZnSnSe4.

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

Sn–Se (SnSe, SnSe2) secondary phases create shuntingpaths through the device [29] leading to detrimental effectson device performance [30]. Highly conductive Cu–Se(CuSe, Cu2–xSe) secondary phases cause devices to be elec-trically inactive [30]. ZnSe is reported to be harmless whenlocated at the back contact [31] but blocks the current whenat the surface [31–33]. Working in a Zn-rich compositionrange, the formation of ZnSe secondary phase is highlyexpected. Fortunately, surface ZnSe may be removed byoxidative etching, improving the short-circuit currentdensity (JSC) and series resistance of Zn-rich devices [16].

Scanning electron microscopy combined with EDX(SEM+EDX) analysis revealed the presence ofsecondary phases on the absorbers’ surfaces (Supportinginformation S2).

As expected, ZnSe was detected on the surface ofseveral absorbers. Characteristic small ZnSe grains coveredalmost completely the surface of Zn-richer absorbers asshown in Figure 5(b). Identification of ZnSe by XRD isvery challenging because of the overlap with the CZTSeXRD pattern [34]. Raman spectroscopy using resonant con-ditions (457.9 nm excitation wavelength) is helpful to iden-tify ZnSe phase at the surface of Zn-richer absorbers [35],as shown in Figure 5(a).

Crystalline grains—characteristic for Sn–Se—wereobserved on many absorbers, as shown exemplarily inFigure 5(e). These grains were also detected on the surfaceof Zn-rich absorbers (Zn/Sn> 1), which are also rich in Snbecause of high Cu deficiency [36]. In all the Sn-richabsorbers (Zn/Sn< 1) and in some Zn-rich absorbers

Figure 5. (a) Raman spectra and scanning electron microscopy images of Zn-richer absorber (b) before etching and (c) after KMnO4/H2SO4 + Na2S etching. (d) Raman spectra and scanning electron microscopy images of Sn-rich absorber (e) before etching and (f) after

Na2S and (NH4)2S etchings. CZTSe, Cu2ZnSnSe4.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

(Zn/Sn> 1), which are very deficient in copper, Sn–Sesecondary phase is detected by XRD (Supporting infor-mation S3). SnSe and SnSe2 secondary phases are alsoconfirmed at the surface by Raman spectroscopy usingpre-resonant conditions (785-nm excitation wavelength),but only in some Sn-rich absorbers. Figure 5(d) showsthe presence of SnSe2 in one of the investigated Sn-richabsorbers. In some Zn-rich absorbers, Sn–Se is only de-tected by SEM+EDX, probably because the amount isvery small and non-homogeneously distributed at thesurface.

The amount of Zn–Se or Sn–Se secondary phasesvaried greatly with the Zn/Sn and Cu/Zn +Sn ratios. Thepresence of these secondary phases on the surface is inagreement with the GDOES depth composition profile,where copper is shown to be depleted at the frontinterface.

In conclusion, after reactive annealing, several secondaryphases were identified at the surface of the absorbers. Whenpresent in large amounts, ZnSe was detected by Ramanspectroscopy and Sn–Se by XRD and Raman spectroscopy.However, on many absorbers, they were detected only bySEM+EDX analysis. This may be due to the detectionlimit of the techniques, but mainly to a non-uniform distri-bution of the secondary phases. Indeed, in the same ab-sorber, different zones are found, some of them showingZn-rich, others Sn-rich phases or even no secondary phases.The amount of secondary phases can be limited using pre-cursors of optimum composition (Cu/Zn + Sn = 0.65–0.85and Zn/Sn = 1.10–1.25).

3.2.3. Absorber surface optimization by etching.KMnO4/H2SO4 +Na2S [16] and (NH4)2S [17] solutions

were reported to selectively etch ZnSe and Sn–Se secondaryphases, respectively. Depending on the absorber composi-tion, (i) ZnSe, (ii) Sn–Se, or (iii) ZnSe+Sn–Se etchingswere applied on the absorbers surface.

(i) Acidic KMnO4 followed by Na2S etching removedefficiently the small ZnSe grains from the surfaceof the absorbers. As expected when removingZnSe from the surface, a decrease in Zn contentand an increase in Cu content were observed(Figure 6). ZnSe Raman peaks disappeared afteretching as shown in Figure 5(a). In Zn-richerabsorbers [Zn/Sn> 1.3; Figure 5(c)], voids weredetected after the etching procedure. For Zn-richfilms, small surface ZnSe aggregates are often re-ported and can easily be removed by etching. Bycontrast, Zn-richer films exhibit larger aggregatesof ZnSe that are associated with voids in the under-lying film [32], and after etching, these voids areexposed.

(ii) (NH4)2S etching removed Sn–Se secondary phasesfrom the surface as demonstrated by Raman spec-troscopy and XRD on Sn-rich absorbers: Afteretching, the secondary phases are no longer de-tected [Figure 5(d) and Supporting information

S3]. As expected when removing Sn–Se rich phasesby etching, a decrease in Sn content and an increasein Cu content are observed (Figure 6). At the ab-sorber surface, the amount of Sn–Se rich crystallinefeatures is reduced; the absorber surface becomeshollow [Figure 5(f)].

(iii) Combination of both etching procedures tends toreduce both Sn and Zn contents in the absorber,in agreement with individual results for each etchingprocedure (Figure 6).

3.2.4. Influence of secondary phases and etchingon device optoelectronic properties.

Glass/Mo/CZTSe/CdS/ZnO/ITO solar cell devices wereproduced from unetched and etched CZTSe absorbers bydepositing CdS, ZnO, and ITO layers. Table IIa shows op-toelectronic properties of several devices obtained fromCZTSe films with different compositions and etching con-ditions, and Table IIb shows the evolution after etching.For each composition, the absorber layer was cut intosmaller parts in order to apply different etching procedures,and one part was kept unetched for reference.

Devices made from unetched Sn-rich absorbers (Sn-richin Table IIa and IIb) exhibited low efficiencies (under 2%),showing the detrimental influence of Sn–Se secondaryphases. All device parameters clearly improved with theapplication of (ii) Sn–Se etching or (iii) ZnSe + Sn–Seetchings. Also, the devices with Sn-rich absorbers showhigh RSD of optoelectronic properties. These results arein agreement with non-uniformity, low shunt resistance,and low cell performance induced by the formation ofSnSe2 [29] and show that etching greatly reduces thenegative impact of Sn-rich secondary phases at the sur-face [17].

Devices with unetched Zn-rich absorbers exhibitedhigher efficiencies than those with unetched Sn-rich

Figure 6. Cu2ZnSnSe4 (CZTSe) absorbers metallic compositionmeasured by X-ray fluorescence for unetched samples and sam-ples etched by (i) KMnO4/H2SO4 + Na2S, (ii) (NH4)2S, or (iii)KMnO4/H2SO4 + Na2S + (NH4)2S. Composition deviations dueto the etching procedures are represented by arrows in the ter-

nary diagram.

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

TableIIa

.Kes

teritede

vice

optoelec

tron

icpa

rametersve

rsus

abso

rber

compo

sitio

nan

detch

ingpa

rameters.

Sam

ple

compo

sitio

nCu/

Zn+Sn

Zn/Sn

Etching

η[%

]VOC[m

V]

J SC[m

Acm

�2]

FF[%

]Rs[Ω

cm�2]

Rsh

[Ωcm

�2]

Eg[eV]

Max

Ave

RSD

[%]

Max

Ave

RSD

[%]

Max

Ave

RSD

[%]

Ave

Ave

Ave

Bes

tce

ll

Sn-ric

h0.68

0.94

Une

tche

d1.7

1.3

2326

124

17

19.9

17.0

1531

5.5

211.04

Sn-ric

h0.73

1.04

(ii)S

nSe

4.9

3.8

2139

036

08

28.7

26.4

638

2.4

391.05

Sn-ric

h0.75

1.16

(iii)Zn

Se+SnS

e5.0

3.1

3939

034

69

28.8

23.1

1937

4.7

391.05

Zn-rich

0.76

1.13

Une

tche

d2.9

2.5

1234

531

65

24.7

23.3

534

5.9

281.04

Zn-rich

0.78

1.11

(i)Zn

Se

5.1

4.3

1237

635

55

30.7

29.0

441

1.9

501.04

Zn-rich

er0.69

1.32

Une

tche

d3.6

3.2

935

634

72

27.0

25.6

436

5.2

361.05

Zn-rich

er0.70

1.33

(i)Zn

Se

1.4

1.2

1727

524

38

19.0

16.7

931

6.2

211.08

Bes

t0.73

1.10

Une

tche

d6.0

5.2

1039

337

72

31.7

30.5

445

2.2

811.06

Bes

t0.76

1.16

(ii)S

nSe

5.8

4.8

1538

236

53

32.1

29.9

744

1.4

671.04

Bes

t0.75

1.20

(iii)Zn

Se+SnS

e7.6

6.7

643

541

04

31.9

30.0

355

0.9

181

1.05

Interm

ediate

0.82

1.03

Une

tche

d4.2

3.2

2833

529

715

31.1

27.8

1038

3.3

311.02

Interm

ediate

0.81

1.02

(i)Zn

Se

4.9

4.1

1734

532

28

31.1

29.6

443

1.9

511.03

Interm

ediate

0.86

1.03

(iii)Zn

Se+SnS

e5.7

4.3

1536

832

77

31.0

29.4

345

1.3

511.02

Cu/Zn

+Snan

dZn

/Snratio

saremea

suredby

X-ra

yfluo

resc

ence

onab

sorbersaftere

tching

.Mainop

toelec

tron

icpa

rameters(pow

erco

nversion

efficien

cyη,

open

-circ

uitv

oltage

VOC,s

hort-circ

uitc

urrent

J SC)a

rerepo

rted

inthe

followingway

:bes

tvalue

(Max

)atc

ellsize(activearea

:0.087

cm2),av

erag

eva

lue(Ave

)for

thefullsa

mple(18–

36ce

lls,i.e.,1.5×2.5–

2.5×2.5cm

2),an

dits

relativ

estan

dard

deviation(RSD).Th

erepo

rted

values

forfi

llfactor

FF,

serie

s,an

dsh

untresistan

ceareav

erag

eforthefullsa

mple.

Ban

dgap

Egva

lueis

determ

ined

usingtheinflec

tionpo

intof

thequ

antum

efficien

cycu

rves

near

theab

sorptio

ned

ge,for

thebe

stce

llof

each

sample.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

absorbers (Table IIa), thereby confirming that ZnSe is lessdetrimental to device performance than Sn–Se secondaryphases. Table IIb shows that after (i) ZnSe etching, η,VOC, JSC, FF, Rs, and Rsh improved, demonstrating thebeneficial effect of etching by reducing the ZnSe-relatedcurrent blocking at the surface. However, for Zn-richerfilms, the etching provoked a drastic decrease of optoelec-tronic performances, probably because voids in the ab-sorber were exposed when ZnSe was removed.

Best device efficiencies were obtained for absorbers ofZn-rich and intermediate Zn/Sn composition (1.00<Zn/Sn< 1.25, best and intermediate in Table IIa and IIb),which were etched by a combination of etching procedures(iii) KMnO4/H2SO4 +Na2S + (NH4)2S. Devices withunetched absorbers in this optimum composition range al-ready exhibit efficiencies as high as 6.0%. After doubleetching (iii), table IIb shows that η, VOC, JSC, FF, Rs,and Rsh improved, while JSC remained high and un-changed, showing a better device improvement than thatwith single etching (i) or (ii). The best device efficiencyof 7.6% was reached.

In conclusion, selecting carefully the composition, thuslimiting the amount of secondary phases, led to 6.0% best de-vice efficiency.When the secondary phases are segregated atthe surface, they can be removed by adapted etching proce-dures. In this study, KMnO4/H2SO4+Na2S+ (NH4)2S etch-ing successfully improved the device performance. Bestdevice efficiency improved from 6.0% to 7.6% by meansof etching. Secondary phase removal from the surface ishighly beneficial, as it reduces the voltage deficit [27,37]and current losses [16] and improves shunt and series resis-tances. In addition, the possible surface passivation effecton the surface can help improve the p–n junction and limitthe recombination [16,17].

These results highlight the importance of the following:(i) absorber composition, which determines the abundanceand nature of present secondary phases; (ii) secondaryphase segregation control, by limiting their presence tothe surface so that they may be removed via chemical etch-ing; (iii) identification and characterization of the secondaryphases; and (iv) adjusted efficient chemical etching proce-dures, which may also provide passivation of the kesteriteabsorber surface.

This part also indicates the importance of kesterite sur-face treatments, which play a crucial role in the p–n

junction with the CdS layer. Interface optimization mightbe one of the keys for further improvement of kesterite so-lar cell performance.

3.3. Buffer layer optimization

Deposition of CdS using chemical bath deposition is basedon the slow release of Cd2+ ions and S2� ions in an aque-ous alkaline bath and the subsequent condensation of theseions on substrates mounted in the bath. The CdS film prop-erties such as thickness, structure, surface morphology, andstoichiometry change with the cadmium source used forthis process [38,39]. For comparison, two different CdSlayers were deposited by optimized chemical bath deposi-tions developed at IREC [40]. CdS depositions werecarried out on kesterite absorbers of optimum composition(Cu/Zn + Sn = 0.65–0.85 and Zn/Sn = 1.10–1.25) andetched with (iii) KMnO4/H2SO4 +Na2S + (NH4)2S. Themain difference between the two CdS is the nature of thecadmium precursor: cadmium sulfate (CdSO4) and cad-mium nitrate (Cd(NO3)2). In Section 3 of this paper, thebuffer was deposited from the CdSO4 precursor. However,the Cd(NO3)2 precursor allows slower CdS chemical bathdeposition and a better control of CdS thickness, makingit possible to deposit very thin layers of CdS, as has beenpreviously demonstrated [40]. Thin CdS buffer layers havebeen reported in the past to improve the performance ofCIGS devices, by allowing more light to reach the junction[41]. Here, with CdS deposited from Cd(NO3)2, an in-crease in current collection for the 400–520-nm wave-length region is observed in EQE, as shown in Figure 7.The current improvement in this region of the spectral re-sponse corresponds to a reduction of buffer-related lossesand can be due to a reduction of CdS thickness and/or a re-duction of CdS trap states.

Resulting device efficiencies and selected optoelec-tronic values are shown in Table III. All devices exhibitedgood optoelectronic parameters and high efficienciesfor electrodeposited kesterite. While very high VOC

were obtained with CdSO4, JSC increased when usingCd(NO3)2.

Improvement of the CZTSe/CdS interface using a thin-ner CdS buffer layer led to higher efficiencies. Best deviceefficiency improved from 7.6% to 8.2%.

Table IIb. Evolution of kesterite devices I–V parameters after etchings (i) KMnO4/H2SO4 + Na2S, (ii) (NH4)2S, or (iii) KMnO4/H2SO4

+ Na2S + (NH4)2S for different absorber compositions.

Sample Initial η [%] Etching

Parameter evolution after etching [%]

η VOC JSC FF RS RSh

Sn-rich <2 (ii), (iii) +160 +45 +45 +22 �35 +86Zn-rich 2–6 (i) +72 +12 +24 +21 �68 +79Zn-richer 3–4 (i) �62 �30 �35 �14 +19 �42Best and Intermediate 4–6 (iii) +32 +10 ±0 +20 �60 +93

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

3.4. Best devices optoelectronic properties

A Cu-poor, Zn-rich absorber (composition before etchingCu/Zn+Sn=0.74, Zn/Sn= 1.25), etched with (iii) KMnO4/H2SO4+Na2S+ (NH4)2S, with CdS deposited from Cd(NO3)2precursor, resulted in the best power conversion efficiencyof the presented solar cells (8.2%, Figure 8) with VOC, JSC,and FF values of 425mV, 30.86mAcm�2, and 62.7%,respectively. These values are obtained without MgF2 anti-reflective coating. To the best of our knowledge, this isthe highest reported efficiency for electrodepositedCZTSe-based solar cells (Table IV).

Noticeably, very high VOC were obtained for the etcheddevices. Some devices prepared with CdSO4 precursorreached VOC as high as 442mV for the full sample area(Best #2 in Table III) and 467mV for the 0.087-cm2 device(represented in Figure 8). To the best of our knowledge,these open-circuit voltage values are the highest reportedfor pure selenium kesterite devices. The voltage deficit—one of the limiting parameters for kesterite solar cells—isof similar magnitude as the ones reported for high efficientpure-selenium CZTSe devices in literature (Table IV).

The bandgap energies, as extracted from the EQE curve,show clear dependence on the Cu/Sn ratio (Figure 9).Higher bandgaps of about 1.05 eV were obtained for veryCu-poor (Cu/Sn< 1.6) films, which might contribute tohigher open-circuit voltages. An increase of copper concen-tration towards stoichiometry (Cu/Sn = 2) resulted in adecrease of the bandgap energy to 0.95 eV. Various hy-potheses may explain the bandgap variation with thecomposition change: formation of band-bending Cu-poor[Zn2+ +VCu] defects [25], exchange of Cu and Zn cationsleading to order/disorder in kesterite [49], and formationof lower bandgap Cu-rich secondary phases.

Figure 7. Low wavelength range of normalized external quan-tum efficiency (EQE) spectra for devices with CdS deposited

from Cd(NO3)2 and CdSO4 precursors (Table IV).

TableIII.

Kes

teritede

vice

optoelec

tron

icpa

rametersve

rsus

abso

rber

compo

sitio

nan

dCdS

prec

urso

r.

Sam

ple

Cu/

Zn+Sn

Zn/

Sn

CdS

prec

urso

rη[%

]max

η[%

]mea

nRSD

[%]

VOC[m

V]

max

VOC[m

V]

mea

nRSD

[%]

J SC[m

Acm

�2]

max

J SC[m

Acm

�2]

mea

nRSD

[%]

FF[%

]mea

nRs

[Ωcm

�2]

Rsh

[Ωcm

�2]

Eg

[eV]

Eg/q-VOC

[V]

Cu-

poorer

0.66

1.11

Cd(NO

3) 2

7.7

6.7

0.7

440

421

1030

.229

.20.5

550.5

134

1.07

0.63

5

Cu-

poorer

0.66

1.11

CdS

O4

6.9

6.2

0.5

457

442

1128

.427

.70.3

500.8

110

1.05

0.59

7

Bes

t#2

0.74

1.25

Cd(NO

3) 2

8.2

7.7

0.4

440

423

731

.330

.50.5

600.5

181

1.05

0.62

1Bes

t#2

0.74

1.25

CdS

O4

6.4

4.3

1.4

421

373

4829

.227

.01.9

422.5

541.04

0.61

8Bes

t#3

0.78

1.15

Cd(NO

3) 2

7.4

6.1

1.2

414

394

1831

.330

.01.1

510.4

871.04

0.63

1Bes

t#3

0.78

1.15

CdS

O4

7.5

6.4

0.6

446

416

1829

.528

.70.5

540.8

162

1.04

0.59

6

RSD,relativestan

dard

deviation.

BeforeCdS

depo

sitio

n,allabs

orbe

rlay

ersareetch

edby

(iii)KMnO

4/H

2SO

4+Na 2S+(NH4) 2S.C

u/Zn

+Snan

dZn

/Snratio

saremea

suredby

X-ra

yfluo

resc

ence

onab

sorbersbe

fore

etch

ingan

dco

uldha

vech

ange

daftere

tching

.

Mainop

toelec

tron

icpa

rameters(η,V

OC,J

SC)a

rerepo

rted

asbe

stva

lueat

cellsize

(activearea

:0.087

cm2),av

erag

eva

lueforthefullsa

mple(18ce

lls,i.e.,1.5×2.5cm

2),an

dits

stan

dard

deviation.

Therepo

rted

values

forfill

factor,series,an

dsh

untres

istanc

eareav

erag

eforthe

fullsa

mple.

Ban

dgap

Egva

luean

dtheas

sociated

VOCde

ficitE

g/q-VOCarede

term

ined

usingtheinflec

tionpo

into

fthe

quan

tum

efficien

cycu

rves

near

theab

sorptio

ned

ge,

forthebe

stce

llof

each

sample.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

4. CONCLUSION

Cu2ZnSnSe4 photovoltaic devices were fabricated by thefollowing: (i) large-scale, high-speed electrodeposition ofthin-film metal stacks; (ii) low-temperature annealing;(iii) high-temperature reactive annealing; (iv) removal ofsecondary phases by chemical etching; (v) chemical bathdeposition of cadmium sulfide buffer layer; and (vi) devicecompletion. Secondary phases at the absorber/buffer layerinterface are a major limitation for performance, and theirremoval is of crucial importance in order to obtain high-efficiency kesterite devices. Here, a champion CZTSe solarcell fabricated by electrodeposition was demonstrated,with 8.2% active-area power conversion efficiency. Eventhough this performance is not yet comparable with that

Figure 8. I–V characteristics of the electrodeposited champion Cu2ZnSnSe4 solar cell, exhibiting 8.2% power conversion efficiency,under dark and one sun illumination. For comparison, I–V characteristics of one of the best cells obtained with a CdSO4 precursor

and the best VOC cell were also represented.

Table IV. Optoelectronic parameters of high-efficiency pure selenium Cu2ZnSnSe4 (CZTSe) devices in literature.

Material and deposition method Efficiency [%] Bandgap [eV] VOC [mV] Eg/q - VOC [mV] JSC [mA cm�2] FF [%]

CZTSe [42] co-evaporation 11.6 1 423 578 40.6 67.3CZTSe [43] sputtering 10.4 1 394 606 39.7 66.4CZTSe [44] sputtering 9.7 1 408 592 38.9 61.4CZTSe [45] co-evaporation 9.15 377 37.4 64.9CZTSe [46] co-evaporation 8.9 1 385 615 42.6 54.3CZTSe [40] sputtering 8.2 1.02 392 628 32.4 64.4CZTSe [9] electrodeposition 8.0 1.02 390 630 35.3 58CZTSe [47] sputtering 7.5 432 30.5 56.8CZTSe [48] co-evaporation 7.5 356 35.4 60CZTSe [10] electrodeposition 7.0 1.10 369 730 32.4 58.5

Figure 9. Bandgap Eg determined by using the inflection pointof the quantum efficiency curves near the absorption edge ver-

sus composition ratio Cu/Sn.

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

of CZTSe devices fabricated by other methods such asco-evaporation or sputtering, these results highlight thepotential of electrodeposited kesterite-based process forlow-cost, earth-abundant solar cells fabrication.

ACKNOWLEDGEMENTS

The authors wish to thank Aurélie Laparre (NEXCIS) forthe low-temperature annealing study, Noémie Pascal andLaurence Pacpiak (NEXCIS) for the GDOESmeasurements,Xavier Fontané (IREC) for the Ramanmeasurements, SimónLópez-Marino and Haibing Xie (IREC) for the etchingprocedures and discussion, and Monika Arasimowicz(NEXCIS) for reviewing. The research leading to theseresults has received funding from the European Union’sSeventh Framework Program FP7/2007–2013 under grantagreement no. 284486 (SCALENANO), the People Pro-grams (Marie Curie Actions) under REA grant agreementno. 285897 (INDUCIS) and no. 316488 (KESTCELLS). E.Saucedo thanks the Government of Spain for the “Ramony Cajal” fellowship (RYC-2011-09212) and Y. Sanchez forthe PTA fellowship (PTA2012-7852-A).

REFERENCES

1. Siebentritt S, Schorr S. Kesterites— a challenging ma-terial for solar cells. Progress in Photovoltaics: Re-search and Applications 2012; 20: 512–519.

2. Shockley W, Queisser HJ. Detailed balance limit of ef-ficiency of p-n junction solar cells. Journal of AppliedPhysics 1961; 32: 510.

3. Kim J, Hiroi H, Todorov TK, Gunawan O, KuwaharaM, Gokmen T, Nair D, Hopstaken M, Shin B, LeeYS, Wang W, Sugimoto H, Mitzi DB. High efficiencyCu2ZnSn(S,Se)4 solar cells by applying a doubleIn2S3/CdS emitter. Advanced Materials 2014; 26:7427–7431.

4. Abermann S. Non-vacuum processed next generationthin film photovoltaics: towards marketable efficiencyand production of CZTS based solar cells. SolarEnergy 2013; 94: 37–70.

5. Wang W, Winkler MT, Gunawan O, Gokmen T,Todorov TK, Zhu Y, Mitzi DB. Device characteristicsof CZTSSe thin-film solar cells with 12.6% efficiency.Advanced Energy Materials 2014; 4: 1301465.

6. NEXCIS. NEXCIS achieves a new record perfor-mance @ 17.3% certified pixel measurement with itsCIGS PV technology, 2014. Can be found underhttp://www.nexcis.fr/en/medias/theme-3-news/id-82-23-october-2014-nexcis-achieves-a-new-record-per-formance-17-3-certified-pixel-measurement-with-its-cigs-pv-technology (accessed 06 December 2014).

7. (Lili) Deligianni H, Ahmed S, Romankiw LT. Thenext frontier: electrodeposition for solar cell

fabrication. Electrochemical Society Interface 2011;2011: 47–53.

8. Jiang F, Ikeda S, Harada T, Matsumura M. Pure sulfideCu2ZnSnS4 thin film solar cells fabricated by preheatingan electrodeposited metallic stack. Advanced EnergyMaterials 2014; 4: 1301381.

9. Jeon J-O, Lee KD, Seul Oh L, Seo S-W, Lee D-K,Kim H, Jeong J-H, Ko MJ, Kim B, Son HJ, Kim JY.Highly efficient copper-zinc-tin-selenide (CZTSe) so-lar cells by electrodeposition. ChemSusChem 2014;7: 1073–1077.

10. Guo L, Zhu Y, Gunawan O, Gokmen T, Deline VR,Ahmed S, Romankiw LT, Deligianni H. Electrodepos-ited Cu2ZnSnSe4 thin film solar cell with 7% power con-version efficiency. Progress in Photovoltaics: Researchand Applications 2012; 22: 58–68.

11. Ahmed S, Reuter KB, Gunawan O, Guo L, RomankiwLT, Deligianni H. A high efficiency electrodepositedCu2ZnSnS4 solar cell. Advanced Energy Materials2012; 2: 253–259.

12. Delbos S. Kësterite thin films for photovoltaics: a re-view. EPJ Photovoltaics 2012; 3: 35004.

13. Mitzi DB, Gunawan O, Todorov TK,Wang K, Guha S.The path towards a high-performance solution-processedkesterite solar cell. Solar Energy Materials and SolarCells 2011; 95: 1421–1436.

14. Fairbrother A, Fontané X, Izquierdo-roca V, Placidi M,Sylla D, Espindola-rodriguez M, López-mariño S,Pulgarín FA, Vigil-galán O, Pérez-rodríguezA, SaucedoE. Secondary phase formation in Zn-rich Cu2ZnSnSe4 -based solar cells annealed in low pressure and tempera-ture conditions. Progress in Photovoltaics: Researchand Applications 2014; 22: 479–487.

15. López-Marino S, Neuschitzer M, Sánchez Y,Fairbrother A, Espindola-Rodriguez M, López-GarcíaJ, Placidi M, Calvo-Barrio L, Pérez-Rodríguez A,Saucedo E. Earth-abundant absorber based solar cellsonto low weight stainless steel substrate. Solar EnergyMaterials and Solar Cells 2014; 130: 347–353.

16. López-Marino S, Sánchez Y, Placidi M, Fairbrother A,Espindola-Rodríguez M, Fontané X, Izquierdo-RocaV, López-García J, Calvo-Barrio L, Pérez-RodríguezA, Saucedo E. ZnSe etching of Zn-Rich Cu2ZnSnSe4:an oxidation route for improved solar-cell efficiency.Chemistry - A European Journal 2013; 19: 14814–14822.

17. Xie H, Sánchez Y, López-Marino S, Espíndola-Rodríguez M, Neuschitzer M, Sylla D, Fairbrother A,Izquierdo-Roca V, López-Marino S, Pérez-RodríguezA, Saucedo E. Impact of Sn(S,Se) secondary phasesin Cu2ZnSn(S,Se)4 solar cells: a chemical route fortheir selective removal and absorber surface passiv-ation. ACS Applied Materials & Interfaces 2014; 6:12744–12751.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

18. Colombara D, Crossay A, Vauche L, Jaime S,Arasimowicz M, Grand P-P, Dale PJ. Electrodeposi-tion of Kesterite thin films for photovoltaic applica-tions: Quo vadis? Physica Status Solidi 2015; 212:88–102.

19. Lin Y, Ikeda S, Septina W, Kawasaki Y, Harada T,Matsumura M. Mechanistic aspects of preheating ef-fects of electrodeposited metallic precursors on struc-tural and photovoltaic properties of Cu2ZnSnS4 thinfilms. Solar Energy Materials and Solar Cells 2014;120: 218–225.

20. Fairbrother A, Fontané X, Izquierdo-Roca V,Espíndola-Rodríguez M, López-Marino S, Placidi M,Calvo-Barrio L, Pérez-Rodríguez A, Saucedo E. Onthe formation mechanisms of Zn-rich Cu2ZnSnS4 filmsprepared by sulfurization of metallic stacks. SolarEnergy Materials and Solar Cells 2013; 112: 97–105.

21. Han J, Jeon S, Kim WK. Selenization mechanisms ofSn and Zn investigated using in situ high-temperatureX-ray diffraction. Thin Solid Films 2013; 546:321–325.

22. Redinger A, Berg DM, Dale PJ, Siebentritt S. The con-sequences of kesterite equilibria for efficient solarcells. Journal of the American Chemical Society2011; 133: 3320–3323.

23. Bendapudi SSK. Novel film formation pathways forCu2ZnSnSe4 for solar cell applications, University ofSouth Florida, 2011. at http://scholarcommons.usf.edu/etd/3005

24. Dudchak IV, Piskach LV. Phase equilibria in theCu2SnSe3 –SnSe2 –ZnSe system. Journal of Alloysand Compounds 2003; 351: 145–150.

25. Chen S, Walsh A, Gong X-G, Wei S-H. Classificationof lattice defects in the kesterite Cu2ZnSnS4 andCu2ZnSnSe4 earth-abundant solar cell absorbers. Ad-vanced Materials 2013; 25: 1522–1539.

26. Marcano G, Rincón C, de Chalbaud LM, Bracho DB,Pérez GS. Crystal growth and structure, electrical,and optical characterization of the semiconductorCu2SnSe3. Journal of Applied Physics 2001; 90: 1847.

27. Mousel M, Redinger A, Djemour R, Arasimowicz M,Valle N, Dale PJ, Siebentritt S. HCl and Br2-MeOHetching of Cu2ZnSnSe4 polycrystalline absorbers. ThinSolid Films 2013; 535: 83–87.

28. Siebentritt S. Why are kesterite solar cells not 20% ef-ficient? Thin Solid Films 2013; 535: 1–4.

29. Temgoua S, Bodeux R, Naghavi N, Delbos S. Effectsof SnSe2 secondary phases on the efficiency ofCu2ZnSn(Sx,Se1-x)4 based solar cells. Thin Solid Films2015; 582: 215–219.

30. Vora N, Blackburn J, Repins I, Beall C, To B, PankowJ, Teeter G, Young M, Noufi R. Phase identificationand control of thin films deposited by co-evaporation

of elemental Cu, Zn, Sn, and Se. Journal of VacuumScience & Technology A Vacuum Surfaces and Films2012; 30: 051201.

31. Hsu W-C, Repins I, Beall C, DeHart C, Teeter G, ToB, Yang Y, Noufi R. The effect of Zn excess onkesterite solar cells. Solar Energy Materials and SolarCells 2013; 113: 160–164.

32. Colombara D, Robert EVC, Crossay A, Taylor A,Guennou M, Arasimowicz M, Malaquias JCB,Djemour R, Dale PJ. Quantification of surface ZnSein Cu2ZnSnSe4-based solar cells by analysis of thespectral response. Solar Energy Materials and SolarCells 2014; 123: 220–227.

33. Timo Wätjen J, Engman J, Edoff M, Platzer-BjörkmanC. Direct evidence of current blocking by ZnSe inCu2ZnSnSe4 solar cells. Applied Physics Letters2012; 100: 173510.

34. Redinger A, Hönes K, Fontané X, Izquierdo-Roca V,Saucedo E, Valle N, Pérez-Rodríguez A, SiebentrittS. Detection of a ZnSe secondary phase incoevaporated Cu2ZnSnSe4 thin films. Applied PhysicsLetters 2011; 98: 101907.

35. Fontané X, Izquierdo-Roca V, Fairbrother A, Espíndola-Rodríguez M, Placidi M, Jawhari T, Saucedo E, Pérez-Rodríguez A. Selective detection of secondary phasesin Cu2ZnSn(S,Se)4 based absorbers by pre-resonantraman spectroscopy. Photovoltaic Specialist Conference(PVSC), 2013 IEEE 39th. 2013; 2581–2584.

36. Collord AD, Xin H, Hillhouse HW. Combinatorial ex-ploration of the effects of intrinsic and extrinsic defectsin Cu2ZnSn(S,Se)4. IEEE Journal of Photovoltaics2014; 5: 288–298.

37. Kanevce A, Repins I, Wei S-H. Impact of bulk proper-ties and local secondary phases on the Cu2(Zn,Sn)Se4solar cells open-circuit voltage. Solar Energy Mate-rials and Solar Cells 2015; 133: 119–125.

38. Khallaf H, Oladeji IO, Chai G, Chow L. Characteriza-tion of CdS thin films grown by chemical bath deposi-tion using four different cadmium sources. Thin SolidFilms 2008; 516(21): 7306–7312.

39. Ortega-Borges R, Lincot D. Mechanism of chemicalbath deposition of cadmium sulfide thin films in theammonia-thiourea system. Journal of the Electro-chemical Society 1993; 140(12): 3464–3473.

40. Neuschitzer M, Sanchez Y, López-Marino S, Xie H,Fairbrother A, Placidi M, Haass S, Izquierdo-Roca V,Perez-Rodriguez A, Saucedo E. Optimization of CdSbuffer layer for high performance Cu2ZnSnSe4 solar cellsand the effects of light soaking: elimination of crossoverand red kink. Progress in Photovoltaics: Research andApplications 2015. DOI:10.1002/pip.2589.

41. Contreras MA, Romero MJ, To B, Hasoon F, Noufi R,Ward S, Ramanathan K. Optimization of CBD CdS

8.2% kesterite thin-film solar cells by electrodepositionL. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip

process in high-efficiency Cu(In,Ga)Se2 -based solarcells. Thin Solid Films 2002; 403–404: 204–211.

42. Lee YS, Gershon T, Gunawan O, Todorov TK,Gokmen T, Virgus Y, Guha S. Cu2ZnSnSe4 thin-filmsolar cells by thermal co-evaporation with 11.6% effi-ciency and improved minority carrier diffusion length.Advanced Energy Materials 2015; 5: 1401372.

43. Oueslati S, Brammertz G, Buffière M, ElAnzeery H,Touayar O, Köble C, Bekaert J, Meuris M, PoortmansJ. Physical and electrical characterization of high-performance Cu2ZnSnSe4 based thin film solar cells.Thin Solid Films 2014; 582: 224–228.

44. Brammertz G, Buffière M, Oueslati S, ElAnzeery H,Ben Messaoud K, Sahayaraj S, Köble C, Meuris M,Poortmans J. Characterization of defects in 9.7%efficient Cu2ZnSnSe4-CdS-ZnO solar cells. AppliedPhysics Letters 2013; 103: 163904.

45. Repins I, Beall C, Vora N, DeHart C, Kuciauskas D,Dippo P, To B, Mann J, Hsu W-C, Goodrich A, NoufiR. Co-evaporated Cu2ZnSnSe4 films and devices.Solar Energy Materials and Solar Cells 2012; 101:154–159.

46. Shin B, Zhu Y, Bojarczuk NA, Jay Chey S, Guha S.Control of an interfacial MoSe2 layer in Cu2ZnSnSe4

thin film solar cells: 8.9% power conversion efficiencywith a TiN diffusion barrier. Applied Physics Letters2012; 101: 053903.

47. Brammertz G, Buffière M, Mevel Y, Ren Y, Zaghi AE,Lenaers N, Mols Y, Koeble C, Vleugels J, Meuris M,Poortmans J. Correlation between physical, electrical,and optical properties of Cu2ZnSnSe4 based solar cells.Applied Physics Letters 2013; 102: 013902.

48. Mousel M, Schwarz T, Djemour R, Weiss TP, SendlerJ, Malaquias JC, Redinger A, Cojocaru-Mirédin O,Choi P-P, Siebentritt S. Cu-rich precursors improvekesterite solar cells. Advanced Energy Materials2014; 4: 1300543.

49. Rey G, Redinger A, Sendler J, Weiss TP, Thevenin M,Guennou M, El Adib B, Siebentritt S. The band gap ofCu2ZnSnSe4: effect of order-disorder. Applied PhysicsLetters 2014; 105: 112106.

SUPPORTING INFORMATION

Additional supporting information may be found in theonline version of this article at the publisher’s website.

8.2% kesterite thin-film solar cells by electrodeposition L. Vauche et al.

Prog. Photovolt: Res. Appl. (2015) © 2015 John Wiley & Sons, Ltd.DOI: 10.1002/pip