Communication

Photovoltaic Devices www.advmat.de

Earth-Abundant Chalcogenide Photovoltaic Devices with

over 5% Efficiency Based on a Cu2BaSn(S,Se)4 Absorber
Donghyeop Shin, Tong Zhu, Xuan Huang, Oki Gunawan, Volker Blum,
and David B. Mitzi*

0.66%[1] power conversion efficiency (PCE)

In recent years, Cu2ZnSn(S,Se)4 (CZTSSe) materials have enabled important to the current record device with 12.6%[2]
progress in associated thin-film photovoltaic (PV) technology, while avoiding PCE. However, despite active research
scarce and/or toxic metals; however, cationic disorder and associated band on CZTSSe-based devices and similari-
tailing fundamentally limit device performance. Cu2BaSnS4 (CBTS) has ties between CZTSSe and CIGSSe, the
efficiency of CZTSSe-based solar cells
recently been proposed as a prospective alternative large bandgap (~2 eV),
still remains significantly lower than the
environmentally friendly PV material, with ~2% power conversion efficiency record for CIGSSe devices (22.6%).[2,3]
(PCE) already demonstrated in corresponding devices. In this study, a Recently, antisite disorder and associated
two-step process (i.e., precursor sputter deposition followed by successive band tailing in kesterite-based CZTSSe
sulfurization/selenization) yields high-quality nominally pinhole-free films materials have been recognized as key
performance-limiting factors in CZTSSe
with large (>1 µm) grains of selenium-incorporated (x = 3) Cu2BaSnS4−xSex
devices, due at least in part to the very
(CBTSSe) for high-efficiency PV devices. By incorporating Se in the sulfide similar size, coordination, and chemical
film, absorber layers with 1.55 eV bandgap, ideal for single-junction PV, have valence of Cu and Zn atoms.[4,5] As a
been achieved within the CBTSSe trigonal structural family. The abrupt transi- result, recent research has focused on
tion in quantum efficiency data for wavelengths above the absorption edge, minimization of cationic disordering in
coupled with a strong sharp photoluminescence feature, confirms the relative CZTSSe-related materials.[6–9]
In order to overcome the funda-
absence of band tailing in CBTSSe compared to CZTSSe. For the first time,
mental disordering issues in kesterite-
by combining bandgap tuning with an air-annealing step, a CBTSSe-based based CZTSSe, one focal point has been
PV device with 5.2% PCE (total area 0.425 cm2) is reported, >2.5× better than to employ long-term (hours to days)
the previous champion pure sulfide device. These results suggest substantial anneals below the nominal ordering tem-
promise for the emerging Se-rich Cu2BaSnS4–xSex family for high-efficiency perature for Cu/Zn atoms within the
CZTSSe structure.[10–14] While such treat-
and earth-abundant PV.
ments have enabled limited reduction of
the disordering, as determined by low-
temperature photoluminescence (PL),[11]
The kesterite-based Cu2ZnSn(S,Se)4 (CZTSSe) family has near-resonant Raman spectroscopy[13] and bandgap determina-
emerged as an attractive replacement for CdTe and Cu(In,Ga) tion using electroreflectance spectroscopy,[14] substantial band
(S,Se)2 (CIGSSe) absorbers because the constituent metals are tailing, and disorder remain in the system even after these
nontoxic and earth-abundant. Decades-long device optimiza- treatments, and no significant reduction in open-circuit voltage
tion efforts for CZTSSe-based solar cells have led to remark- (Voc) has been detected.[10] Given that the disordering may be
able device performance advances, from the first device of related to the similarity in ionic size and preferred coordina-
tion of Cu+, Zn2+, and Sn4+ ions,[4,5] another approach involves
trying to devise zinc-blende-related structures with more sub-
Dr. D. Shin, T. Zhu, X. Huang, Prof. V. Blum, Prof. D. B. Mitzi stantial ionic size mismatch among constituents, such as
Department of Mechanical Engineering and Materials Science Ag2ZnSn(S,Se)4[6,7] and Cu2MSn(S,Se)4 (M = Cd[8], Mn,[15] Fe,[16]
Duke University
Durham, NC 27708, USA
and Co.[17]) However, there are several limitations to realizing
E-mail: [email protected] this goal, including toxicity of the components (M = Cd), semi-
Dr. D. Shin, T. Zhu, X. Huang, Prof. V. Blum, Prof. D. B. Mitzi conductor type inversion (i.e., n-type behavior) for the Ag-based
Department of Chemistry compositions[6,7] and multiple-charge states for selected transi-
Duke University tion metals (M = Mn, Fe, and Co), which can introduce deep
Durham, NC 27708, USA trap/recombination centers. Among them, only Ag2ZnSnSe4
Dr. O. Gunawan compounds have recently demonstrated a PCE of ≈5%.[6,7]
IBM T. J. Watson Research Center
P.O. Box 218, Yorktown Heights, NY 10598, USA Most recently, Cu2BaSnS4−xSex (CBTSSe) materials with
a trigonal structure (space group P31) and composed of only
DOI: 10.1002/adma.201606945 earth-abundant metals (i.e., the abundances of Cu (47 mg kg−1),

Adv. Mater. 2017, 1606945 1606945  (1 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Ba (650 mg kg−1), and Sn (2.5 mg kg−1) in the earth’s crust are and air-annealing is further found to promote substantially
>25× larger than that of In (0.1 mg kg−1)[18]) have been pro- improved short-circuit currents and device performance. We
posed and demonstrated as emerging PV absorbers to address report that using air-annealed Cu2BaSnS4−xSex films with
these issues, employing ionic size mismatch coupled with coor- x = 3 leads to a record PCE of 5.2%, i.e., >2.5 times better than
dination discrimination to prevent detrimental formation of the previous champion devices[19,21] for this family of earth-
antisite disorder and associated band tailing.[9,19–21] Moreover, abundant metal chalcogenides.
these compounds exhibit a tunable bandgap in the range of A previous experimental study[19] targeting Cu2BaSnS4−xSex
1.5–2.0 eV, spanning relevant values for single- or multiple- materials indicated that a phase transition from trigonal (P31)
junction photovoltaic or photocatalytic applications.[9,19–21] to orthorhombic (Ama2) structure occurs when the sample sele-
According to advanced theoretical studies on trigonal CBTSSe nium composition exceeds x = 3. Based on this study and using
compounds, the dominant point defect in this system is the experimental lattice parameters coupled with all-electron[22]
copper vacancy (VCu) as a shallow acceptor, similar to the situ- density-functional-theory (DFT)-derived (HSE06[23] functional)
ation in high-performance CIGSSe, with other acceptor and atomic coordinates, theoretical bandgap calculations for the
donor defects having higher formation energies and therefore trigonal Cu2BaSnS4−xSex system (0 ≤ x ≤ 3) agree with measured
reduced probability of occurrence.[20] Specifically, the forma- 1.5–2.0 eV values (Figure 1). Notably, the x = 3 trigonal system
tion energies of Cu-Ba and Cu–Sn antisite defects are 1.0 eV has significantly lower bandgap than the orthorhombic x = 4
higher than that of VCu, due to large ionic size/charge mis- (fully selenized) end member. The valence band maximum
match among constituents and distinct coordination environ- (VBM) for each composition is technically off-Γ (a detailed anal-
ment of the large Ba2+ cation. These results imply that this ysis in Figure S1 in the Supporting Information reveals a very
type of cationic disordering and associated band tailing in slightly off-Γ VBM even for the x = 4 Ama2 compound). How-
CBTSSe should be energetically less favorable compared to ever, the difference between the direct (Γ to Γ) and indirect (i.e.,
the analogous situation for CZTSSe, pointing to a substantial quasi-direct; off-Γ to Γ) bandgap also decreases from 17 (x = 0)
potential for earth-abundant PV devices. Indeed, the absence to 12 meV (x = 3), small enough values to be considered insig-
of PL peak red-shifting (associated with band tailing induced nificant with respect to achieving a sharp onset to absorption
by lattice disorder[5]) relative to the bandgap in Cu2BaSnS4 at the bandgap (important for high-efficiency PV). In addition,
(CBTS), coupled with a sharp cut-off in the external quantum carrier effective masses, m*, in units of the free electron mass,
efficiency (EQE) and narrow PL peak width, strongly suggests m0, and obtained by numerical parabolic fittings to DFT-HSE06-
a low degree of disorder and band tailing for this system rela- calculated band structures including spin-orbit coupling (SOC),
tive to kesterite-based CZTSSe.[19] As a proof-of concept, Shin are shown in Table S1 in the Supporting Information, with
et al. demonstrated the first 1.62%-efficient CBTS-based solar more detailed analysis in Figure S1 in the Supporting Informa-
cell (0.425 cm2 device area) using a well-known device struc- tion. The calculated electron and hole effective masses largely
ture analogous to CIGSSe and CZTSSe devices.[19] Additionally, follow the same compositional trends as for the bandgap, i.e.,
a more recent CBTS-based device with a CdS/CdS:O bilayer they show the lowest values at x = 3 (P31) and increase slightly
yielded a slightly higher PCE of 2.03% (0.08 cm2 device area).[21] for the Ama2 structure (x = 4). The predicted electron effective
The current study reports a major device efficiency improve- masses are low (i.e., me* = 0.15–0.23m0, depending on direc-
ment, by extending the PV device results to include selenium tion and composition), with the corresponding hole effective
within the trigonal system, in order to lower the bandgap to masses perpendicular to the c axis, here labeled mh*(||a) (par-
a value more appropriate for the solar spectrum. PL and Hall allel to the reciprocal space a axis), falling in a slightly higher
effect data highlight favorable properties for PV application range. Only for mh*(||c), do the hole effective mass values

Figure 1.  a) Evolution of experimentally measured (reflectance data on bulk samples,[19] black circles) and theoretically calculated (DFT-HSE06 with
spin-orbit coupling, using the experimental unit cell parameters,[19] grey circles) bandgap values.[19] Note that the x = 3 sample adopts the P31 rather
than Ama2 structure. b) DFT-HSE06-predicted carrier effective masses obtained by numerical parabolic fitting of calculated band structures for the
lowest conduction band (me*: downward triangles) and highest valence band (mh*: upward triangles) in different directions, i.e., parallel to the
reciprocal-space a (grey) or c (black) axis. Details are provided in Table S1 and Figure S1 in the Supporting Information.

Adv. Mater. 2017, 1606945 1606945  (2 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

cover a higher range of between 1.68m0 (x = 1) and 1.23m0 the diffraction pattern has been Pawley fit to the same trigonal
(x = 3). Overall, across the full composition range (0 ≤ x ≤ 4), in structure type as for the pure sulfide analog, yielding the lattice
terms of bandgap and carrier effective masses, Cu2BaSnS4−xSex constants a = 6.588(7) Å and c = 16.440(1) Å, consistent with
(x = 3) with space group P31 appears to be most suitable for use the XRD data of bulk Cu2BaSnS4−xSex (x = 3) samples prepared
as an absorber within a single junction solar cell, leading to the by solid state reaction (Figure 2a),[19] as well as the energy dis-
question of how to stabilize this composition. persive X-ray spectroscopy (EDS) compositional analysis (also
In recent work,[24] Shin et al. developed a two-step process yielding x = 3 composition; see Figure S2 in the Supporting
(i.e., precursor sputter deposition/post-annealing) to produce Information). Beyond the crystal structure and composition,
high-quality trigonal Cu2BaSnS4−xSex (x ≤ 2.4) films with large top-view and cross-sectional scanning electron microscopy
grains, and applied these films within a photoelectrochemical (SEM) images (Figure 2b–e) reveal that both pure sulfide and
(PEC) device (i.e., for hydrogen evolution). As Se atoms incor- mixed sulfur/selenium films are continuous without pinhole
porate into the Cu2BaSnS4−xSex films, the bandgap decreases defects on the film surface and with cross-sectional grains
from 1.98 (x = 0.4) to 1.62 eV (x = 2.4), providing an initial extending substantially across the film thickness. When a large
unoptimized PEC device with more than 6 mA cm−2 at −0.4 V/ amount of Se is incorporated into the films (i.e., for x = 3), the
reversible hydrogen electrode with Pt catalyst.[24] In the present average grain size increases substantially to >1 µm (Figure 2c),
work, the Se content in analogous sputtered (see the Experi- which can reduce nonradiative grain boundary recombination.
mental section) trigonal Cu2BaSnS4−xSex films was increased Despite the increase in grain size for Cu2BaSnS4−xSex (x = 3)
to further reduce the bandgap to values within the optimal films, some voids are observed in the film near the Mo sub-
range for single-junction PV (1.0–1.6 eV). In order to enhance strate interface (Figure 2e), which may serve to enhance recom-
Se content, the initially sulfurized films were annealed in the bination or create regions of compromised electrical contact to
presence of excess selenium at 570 °C (versus 560 °C for the the device (e.g., increasing series resistance). Ultimately, this
previous study that yielded lower Se content[23]). Resulting void formation should be addressable by fine-tuning of chem-
Cu2BaSnS4−xSex x = 3 films exhibit a bandgap of 1.55 eV, close ical composition and deposition conditions to further improve
to an optimal value for maximum theoretical PV efficiency for a PV device performance.
single p/n junction, as calculated by Shockley and Queisser.[25] Photoluminescence (PL) measurements were carried out
To determine the structural quality of the sputter-deposited to verify the bandgap, Eg, of Cu2BaSnS4 and Cu2BaSnS4−xSex
baseline Se-free and Se-incorporated Cu2BaSnS4–xSex, X-ray (x = 3) films and to examine radiative recombination (note
diffraction (XRD) measurements were carried out, confirming that it has been previously shown that the PL peak falls within
the single-phase and well-crystallized nature of the films on ~10 meV[19] of the bandgap value for Cu2BaSnS4 and therefore
Mo-coated glass substrates (Figure 2a). The XRD pattern of the the PL peak provides a good estimate of Eg; this conjecture will
pure sulfide Cu2BaSnS4 film matches the trigonal P31 struc- be shown to hold in the high-Se system as well, i.e., see later
ture reported previously,[19] with lattice constants a = 6.343(1) discussion of device quantum efficiency, which provides our
Å and c = 15.830(2) Å. By incorporating Se into the sulfurized primary absorber Eg value). Using these measurements, our
film, the XRD peaks for the Cu2BaSnS4−xSex film shift toward first determination of Eg of the pure Cu2BaSnS4 film is found
lower 2θ angles. To estimate Se-incorporated film composition, to be 2.0 eV, consistent with our previous report.[19] As Se atoms

Figure 2.  a) X-ray diffraction (XRD) data for Cu2BaSnS4 and Cu2BaSnS4−xSex (x = 3; composition estimated by energy-dispersive X-ray spectroscopy
measurement; see Figure S2 in the Supporting Information) films deposited on Mo-coated glass and Cu2BaSnS4−xSex (x = 3) bulk powder prepared by
solid-state reaction.[19] A reference XRD pattern for Cu2BaSnS4 with a trigonal structure (PDF no. 03-065-7569) is also given with Miller indices noted
for higher intensity peaks. The Mo film peak remains centered at 2θ = 40.5°, thereby serving as a reference. The solid grey line shows the (014) peak
shift with incorporation of Se. XRD patterns over a wider 2θ range of 10–70° appear in Figure S3 in the Supporting Information. b,c) SEM top-view
and d,e) cross-sectional images of Cu2BaSnS4 and Cu2BaSnS4−xSex (x = 3) films deposited on Mo-coated glass. f) Room-temperature normalized
photoluminescence (PL) spectra for Cu2BaSnS4 and Cu2BaSnS4−xSex (x = 3) films obtained with 442 nm laser excitation. The PL spectra with absolute
intensities are also given in Figure S4 in the Supporting Information.

Adv. Mater. 2017, 1606945 1606945  (3 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

incorporate into the pure sulfide films, Eg decreases from film subjected to air-annealing at 300–400 °C for several min-
2.0 (x = 0) to 1.54 eV (x = 3), as shown in Figure 2f. The utes, the air-annealed films yielded 50× higher PL intensity
decrease in PL peak energy follows the change in composition compared to that observed for films annealed in a nitrogen-
as determined by XRD and EDS data. It is useful to note that filled glove box, leading to substantial improvement in device
according to a previous study on Cu2BaSnS4−xSex (0 ≤ x ≤ 4) performance.[29,30] To investigate the impact of air-annealing on
bulk samples, when the sample composition exceeds x > 3, a the PL feature for Cu2BaSnS4−xSex (x = 3) films, a film was cut
phase transition from trigonal (P31) to orthorhombic (Ama2) into two pieces, one for air and one for inert atmosphere (i.e.,
structure occurs giving rise to a higher bandgap value (1.72 eV a nitrogen-filled glove box) annealing. The annealing tempera-
for pure selenide[19]). This observation implies that the bandgap ture and dwell time were fixed at 200 °C and 2 min, respec-
for the x = 3 film falls near the lowest achievable value before tively, for each sample. As seen in Figure S6 in the Supporting
the system shifts to the orthorhombic structure. Furthermore, Information, the PL intensity for the air-annealed film is >2×
the strong and sharp PL features also indicate effective radiative more intense than that for the control film, consistent with
recombination and may reflect a less defective lattice (i.e., less reported observations of the beneficial effect of air-annealing on
cationic disorder) compared to analogous CZTSSe systems.[5] CIGSSe/CZTSSe device performance (although the effect may
A slight broadening of the PL peak for the Se-incorporated not be as large for CBTSSe as for CZTSSe,[29] at least for this
film may reflect some modest inhomogeneity in the S:Se ratio level of film quality).
throughout the film thickness. For further optimization of annealing conditions, the air-
To understand the electrical properties of the Cu2BaSnS4 annealing dwell time was adjusted from 2 to 6 min, with the
and Cu2BaSnS4−xSex (x = 3) materials, AC-field Hall measure- annealing temperature fixed at 200 °C. In this case, the air-
ments were performed at room temperature on films deposited annealing step was applied only one time to determine the
on glass. As shown in Table S2 in the Supporting Informa- optimal dwell time for air-annealing at this temperature. As
tion, both films exhibit p-type semiconductor behavior, with Se depicted in Figure S7 and Table S4 in the Supporting Informa-
incorporation in Cu2BaSnS4−xSex (x = 3) leading to an increase tion, the device efficiency gradually improves with increased
in carrier density (p ≈ 5 × 1014 cm−3) relative to the pure sulfide annealing time. Notably, an annealing time of >2.5 min sig-
film, i.e., yielding a value similar to that for high-performance nificantly enhances the fill factor among the PV parameters,
CdTe PV absorbers (≈1014 cm−3),[26] but substantially lower perhaps due to a decrease in the indium tin oxide (ITO) sheet
than reported values for high-efficiency CIGSSe and CZTSSe resistance from >80 (2 and 2.5 min) to ~60 Ω −1 (2.75 and
devices (≈1 × 1016 cm−3).[6,27] In addition, the Cu2BaSnS4−xSex 3 min). However, a much longer air-annealing step negatively
(x = 3) film Hall mobility exceeds that for x = 0, consistent with impacts the PCE, possibly due to Cd and Cu interdiffusion at
DFT-based predictions for the hole effective masses (Figure 1) the heterojunction and associated compensation between donor
and also reflective of the enhanced grain size within these (e.g., CdCu) and acceptor (e.g., VCu) defects, as has been noted for
films. Interestingly, the Hall mobility of the Se-incorporated CIGSSe and CZTSSe devices.[28,31] By applying the optimized
film (1.5 cm2 V−1 s−1) substantially exceeds that of analogous annealing conditions (i.e., 200 °C for 3 min in air), a champion
CZTSSe (<1 cm2 V−1 s−1), measured using the identical AC-field device yielded Voc, short-circuit current density (Jsc), fill factor
Hall measurement method,[6] even despite the DFT-calculated (FF), and total area (0.425(4) cm2) PCE of 611 mV, 17.4 mA cm−2,
valence band edge of Cu2BaSnS4−xSex[19] being less dispersive 48.9%, and 5.2(1)%, respectively, representing the highest
compared to that for the CZTSSe band structure. Such elec- performance achieved among this first demonstration of trig-
trical properties make Se-rich trigonal Cu2BaSnS4−xSex an onal Se-incorporated Cu2BaSnS4−xSex thin-film PV devices
attractive candidate for use within PV devices. In addition, the (Figure 3c). Average performance parameters for the four
results show that control over hole doping concentration for devices shown in the inset of Figure 3c are Voc = 592 ± 19 mV,
the Cu2BaSnS4−xSex system may be of interest for optimizing Jsc = 17.42 ± 0.45 mA cm−2, FF = 47.29 ± 1.53%, and
device performance. PCE = 4.87 ± 0.32% (Table S5, Supporting Information), indi-
To assess PV performance for devices with Cu2BaSnS4−xSex cating process homogeneity across the Cu2BaSnS4−xSex film.
(x = 3) films as the absorber, a similar substrate device structure Given the reported PCE of champion pure-sulfide Cu2BaSnS4-
was employed to that used for high-performance CIGSSe and based PV devices (as high as 2.03%,[19,21]) the current device
CZTSSe solar cells (depicted in Figure S5 in the Supporting record with Se-incorporated Cu2BaSnS4−xSex (x = 3) exceeds this
Information). Illuminated current density–voltage (J–V) data for value by >2.5-fold. The main improvement in the mixed-sulfur/
a Cu2BaSnS4−xSex solar cell with and without an air-annealing selenium device derives from improvement in current density,
process are shown in Figure 3a. The initial device without air- i.e., from 4.11 (x = 0) to 17.4 (x = 3) mA cm−2 due to reduction
annealing yielded a PCE of 2.23%. After a 200 °C-anneal for in the absorber bandgap and formation of larger grains.
2 min in air, the current density was dramatically increased An abrupt decrease in EQE between 750 and 850 nm wave-
from 12.77 to 16.65 mA cm−2 (PCE = 3.77%). Regard- lengths corresponds to the bandgap cut-off for the material and
less of repeated annealing under identical conditions (i.e., the 1.55 eV bandgap derived from the EQE data (Figure 3d)
200 °C for 2 min in air), the current density showed no fur- agrees with the value determined from PL measurement. The
ther substantial improvement, yielding maximum EQE values relatively sharp cut-off in the EQE spectrum suggests that
of ≈75% in the visible spectral range (500–600 nm) (Figure 3b). band tailing may not be a substantial issue in this material. For
Several papers report that the beneficial effect of air-annealing Cu2BaSnS4−xSex (x = 3), the red shift of the PL peak (1.54 eV)
for high-efficiency CIGSSe PV derives from defect passivation relative to the bandgap (1.55 eV), as determined from the
at grain boundaries and interfaces.[28] In the case of a CZTSSe EQE inflection point, is only ≈10 meV (Figure 3d), essentially

Adv. Mater. 2017, 1606945 1606945  (4 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 3. a) Light current density–voltage (J–V) characteristics and b) external quantum efficiency (EQE) spectra of unannealed and annealed
Cu2BaSnS4−xSex (x = 3) solar cells. The same annealing condition (i.e., 200 °C for 2 min in air) was repeatedly applied to the Cu2BaSnS4−xSex device
(x = 3). Table S3 in the Supporting Information lists the photovoltaic device performance parameters from (a). c) Light (solid line) and dark (dashed
line) J–V characteristics (inset shows a set of four PV devices on glass substrates), and d) EQE spectra of the current champion air-annealed
(200 °C for 3 min) Cu2BaSnS4−xSex (x = 3) solar cell with (dashed line) and without (solid line) applied −0.5 V bias (for the EQE data). In (d), the
PL spectrum for the Cu2BaSnS4−xSex (x = 3) film (grey line; reproduced from Figure 2d) and integrated photocurrent density from the EQE curve are
also shown. The EQE-derived photocurrent density is 17.59 mA cm−2, which is in good agreement with the measured Jsc under one sun condition
(17.4 mA cm−2). The bandgap positions of Cu2BaSnS4−xSex (x = 3) determined from the EQE inflection point (black solid line) and the PL peak maximum
(grey dashed line) are noted.

identical to the value for the pure sulfide device.[19] Given the contact in Figure 2e). For further insight regarding the device
30 and 100 meV PL peak offset values[5] for Cu(In,Ga)(S,Se)2 loss mechanisms in the champion device, we also performed
and Cu2(Zn,Sn)(S,Se)4, respectively, this small 10 meV offset a one-diode parameter extraction of the light (dark) J–V data
provides further evidence of very low levels of band tailing in using the Lambert-W function fitting method (Figure S8, Sup-
trigonal Cu2BaSnS4−xSex films. Additionally, the full-width-at- porting Information),[32] yielding: ideality factor n = 3.4(1.9),
half-maximum (FWHM) of the PL peak for Cu2BaSnS4−xSex reverse saturation current J0 = 1.3 × 10−5(3.0 × 10−10) A cm−2,
with x = 3 (≈60 nm) is much smaller compared to the FWHM series resistance RS = 1.9(5.3) Ωcm2 and shunt conductance
for Cu(In,Ga)(S,Se)2 (≈100 nm) and Cu2(Zn,Sn)(S,Se)4 GSh = 6.5(1.4) mS cm−2. The shunt conductance is higher for
(≈190 nm) reported in the literature.[5] Presumably, the very the light J–V suggesting a voltage-dependent collection issue.
different coordination environments around Cu+, Ba2+, and The high series resistances (>1 Ωcm2) correspond to a signifi-
Sn4+ discourage the formation of Cu-Ba and Sn-Ba antisite cant bulk absorber contribution, as estimated from Hall data,
disordering, leading to the observed suppression of band RS,B = t/peμ = 0.8 Ωcm2 (t is absorber thickness and e is the
tailing.[9,19–21] electron charge). This suggests that reducing film thickness
To investigate collection efficiency for photogenerated car- could increase the device FF. Finally, the high light J–V ide-
riers, bias-dependent EQE spectra (Figure 3d) show substantial ality factor (>1) suggests that space charge region recombina-
increases in EQE measured at −0.5 V, compared with analogous tion may be an issue and points to further opportunity for PCE
measurements at zero bias. The photocurrent density calcu- improvement with device structure optimization.
lated from the integrated EQE spectrum measured at −0.5 V In conclusion, Cu2BaSnS4−xSex (x = 3) films prepared by a
(dashed line in Figure 3d) yields 18.35 mA cm−2, compared selenization after sulfurization process exhibit a large-grained
with 17.59 mA cm−2 under zero bias, suggesting the possi- (>1 µm) microstructure without pinhole defects and employ
bility of further device improvements through optimization of only earth-abundant metals. By incorporating Se in the sulfide
film thickness, defect density (impacting minority carrier dif- film, absorber layers with ≈1.55 eV bandgap (the lowest value
fusion length), doping density (impacting depletion width) and achieved for this family), suitable for single-junction PV
interfaces/contacts (e.g., note the film voids near the Mo back devices, were successfully achieved. Sharp cut-offs in the EQE

Adv. Mater. 2017, 1606945 1606945  (5 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

data for wavelengths above the bandgap, coupled with a sharp system. The van der Pauw and Hall effect measurements were done
photoluminescence peak centered at a wavelength closely cor- on ≈5 × 5 mm 1-µm-thick films. The Hall effect measurements were
carried out using an IBM rotating parallel dipole line AC Hall system
responding to the bandgap, confirm the relative absence of
that generates an oscillating magnetic field and performs lock-in
band tailing in the Cu2BaSnS4−xSex films. An air-annealed detection of the corresponding Hall signal. This technique is necessary
Cu2BaSnS4−xSex-based PV device with total area 0.425(4) cm2 for Hall measurements in low mobility or insulating materials.[33] The
demonstrated a 5.2(1)% PCE, which represents a >2.5-fold J–V characteristics of the devices were measured using a Keithley 2400
improvement in performance in this family. Despite the rela- sourcemeter. The illumination source was a Newport Oriel 92192
tively flat valence band edge for Cu2BaSnS4−xSex, the majority solar simulator with an AM 1.5G filter, operating at 100 mW cm−2. The
carrier (hole) Hall mobility of Cu2BaSnS4−xSex (x = 3) exceeds external quantum efficiencies of the devices were measured using an
Enlitech QE-R system, providing a primary measure of the absorber
1 cm2 V−1 s−1, with associated hole density of 5 × 1014 cm−3. bandgap. A calibrated Si solar cell was used as a reference for the J–V
With respect to future opportunities, buffer layers with wider and EQE measurements.
bandgaps (e.g., Zn(O,S)/ZnSnO in place of CdS) may enable
improved spectral response in the short wavelength region.
Increasing the carrier density to ≈1016 cm−3 could boost the
efficiency further by increasing the quasi-Fermi level separation Supporting Information
(increasing Voc) and reducing the bulk resistance (increasing FF). Supporting Information is available from the Wiley Online Library or
Further optimization of film microstructure (e.g., removal of from the author.
internal voids) and interface engineering (e.g., band alignment
at buffer/absorber or absorber/back contact) may also offer a
pathway to high PCE. Acknowledgements
This material is based upon work supported by the National Science
Foundation under Grant No. 1511737. The work was performed
Experimental Section in part at the Duke University Shared Materials Instrumentation
Thin Film Deposition and Solar Cell Fabrication: Cu-Ba-Sn-S precursor Facility (SMIF), a member of the North Carolina Research Triangle
layers were co-sputtered using Cu, Sn, and BaS targets on Mo-coated Nanotechnology Network (RTNN), which is supported by the National
glass substrates. For pure Cu2BaSnS4 films, the precursors were Science Foundation (Grant ECCS-1542015) as part of the National
annealed at 560–570 °C for 5–10 min in the presence of excess sulfur. Nanotechnology Coordinated Infrastructure (NNCI). All opinions
To incorporate a high concentration of Se atoms, sulfurized films were expressed in this paper are the authors’ and do not necessarily reflect
annealed at 570 °C for 5 min in the presence of excess selenium. the policies and views of the NSF. The manuscript was written through
The sulfurization/selenization processes took place in a nitrogen- contribution of all authors. All authors have given approval to the final
filled dry box. The final thicknesses of the Cu2BaSnS4−xSex films were version of the manuscript.
≈800–1100 nm, as determined by cross-sectional scanning electron
microscope imaging. The Cu2BaSnS4−xSex film metal elemental ratios
were in the range of Cu/(Ba+Sn) = 0.83–0.88 and Ba/Sn = 1.05–1.15, as
determined by energy-dispersive X-ray spectroscopy. Conflict of Interest
To fabricate photovoltaic devices, Cu2BaSnS4−xSex films on Mo-coated The authors declare no conflict of interest.
glass substrate were incorporated into the following well-known device
structure used for Cu(In,Ga)(S,Se)2 and Cu2ZnSn(S,Se)4 PV: ITO/i-ZnO/
CdS/absorber/Mo/glass.[19] An ≈50 nm thick CdS layer was deposited
by chemical bath deposition. 50 nm thick i-ZnO and 150 nm thick ITO Keywords
layers were deposited using radio-frequency sputtering. A 900 nm thick
Ni/Al top electrode was deposited on the top of the device using e-beam cationic disordering, Cu2BaSn(S,Se)4, earth-abundant chalcogenides,
evaporation, to improve collection of the photogenerated carriers. The photovoltaic devices, trigonal crystal structures
device total area (0.425(4) cm2; measured using the ImageJ software
package, Version 1.48v) was defined by mechanical scribing. Finally, Received: December 23, 2016
selected completed devices were annealed at 200 °C for 2–6 min in air. Revised: March 10, 2017
Effective Mass Calculations: The same crystal geometries as in ref. [19] Published online:
were used for detailed analyses of DFT-HSE06+SOC[23]-predicted
valence band maximums and conduction band minimums and the
corresponding effective mass determinations, as detailed in Figure S1
in the Supporting Information. The FHI-aims all-electron electronic [1] H. Katagiri, K. Jimbo, W. S. Maw, K. Oishi, M. Yamazaki, H. Araki,
structure code[22] was used with “tight” settings for all numerical A. Takeuchi, Thin Solid Films 2009, 517, 2455.
parameters, and a recently implemented, high-accuracy linear-scaling [2] W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T. K. Todorov,
framework for hybrid DFT calculations[22c,d] was used to obtain the band
Y. Zhu, D. B. Mitzi, Adv. Energy Mater. 2014, 4, 1301465.
structures shown in Figure S1 in the Supporting Information.
[3] P. Jackson, R. Wuerz, D. Hariskos, E. Lotter, W. Witte, M. Powalla,
Characterization Methods: The phase purity and composition for
Phys. Status Solidi RRL 2016, 10, 583.
Cu2BaSnS4−xSex films were investigated using a PANalytical Empyrean
powder X-ray diffractometer under ambient conditions using Cu [4] S. Chen, A. Walsh, X. G. Gong, S. H. Wei, Adv. Mater. 2013, 25,
Kα radiation. To verify the film microstructures, scanning electron 1522.
microscope images were taken using a FEI XL30 system. For energy- [5] T. Gokmen, O. Gunawan, T. K. Todorov, D. B. Mitzi, Appl. Phys. Lett.
dispersive X-ray spectroscopy an acceleration voltage of 15 kV was 2013, 103, 103506.
applied. To determine the photoluminescence characteristics of the [6] T. Gershon, Y. S. Lee, P. Antunez, R. Mankad, S. Singh, D. Bishop,
films, measurements were performed at room temperature (442 O. Gunawan, M. Hopstaken, R. Haight, Adv. Energy Mater. 2016, 6,
nm laser excitation) using a Horiba Jobin Yvon LabRAM ARAMIS 1502468.

Adv. Mater. 2017, 1606945 1606945  (6 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

[7] T. Gershon, K. Sardashti, O. Gunawan, R. Mankad, S. Singh, [22] a) V. Blum, R. Gehrke, F. Hanke, P. Havu, V. Havu, X. Ren,
Y. S. Lee, J. A. Ott, A. Kummel, R. Haight, Adv. Energy Mater. 2016, K. Reuter, M. Scheffler, Comput. Phys. Commun. 2009, 180, 2175;
6, 1601182. b) X. Ren, P. Rinke, V. Blum, J. Wieferink, A. Tkatchenko,
[8] Z. Su, J. M. R. Tan, X. Li, X. Zeng, S. K. Batabyal, L. H. Wong, Adv. A. Sanfilippo, K. Reuter, M. Scheffler, New J. Phys. 2012, 14, 053020;
Energy Mater. 2015, 5, 1500682. c) S. V. Levchenko, X. Ren, J. Wieferink, R. Johanni, P. Rinke,
[9] D. Shin, B. Saparov, D. B. Mitzi, Adv. Energy Mater. 2017, V. Blum, M. Scheffler, Comput. Phys. Commun. 2015, 192, 60;
DOI: 10.1002/aenm.201602366. d) A. C. Ihrig, J. Wieferink, I. Y. Zhang, M. Ropo, X. Ren, P. Rinke,
[10] T. Gershon, D. Bishop, B. McCandless, W. Wang, R. Haight, M. Scheffler, V. Blum, New J. Phys. 2015, 17, 093020.
presented at SPIE Nanosci.+Eng., San Diego, CA, USA, August [23] a) J. Heyd, G. E. Scuseria, M. Ernzerhof, J. Chem. Phys. 2003, 118,
2009. 8207; b) J. Heyd, G. E. Scuseria, M. Ernzerhof, J. Chem. Phys. 2006,
[11] G. Rey, A. Redinger, J. Sendler, T. P. Weiss, M. Thevenin, 124, 219906.
M. Guennou, B. El Adib, S. Siebentritt, Appl. Phys. Lett. 2014, 105, [24] D. Shin, E. Ngaboyamahina, Y. Zhou, J. T. Glass, D. B. Mitzi, J. Phys.
112106. Chem. Lett. 2016, 7, 4554.
[12] X. Lin, A. Ennaoui, S. Levcenko, T. Dittrich, J. Kavalakkatt, [25] W. Shockley, H. J. Queisser, J. Appl. Phys. 1961, 32, 510.
S. Kretzschmar, T. Unold, M. C. Lux-Steiner, Appl. Phys. Lett. 2015, [26] J. Sites, J. Pan, Thin Solid Films 2007, 515, 6099.
106, 013903. [27] I. Repins, S. Glynn, J. Duenow, T. J. Coutts, W. K. Metzger,
[13] J. J. Scragg, J. K. Larsen, M. Kumar, C. Persson, J. Sendler, M. A. Contreras, presented at SPIE Solar Energy+Technol.,
S. Siebentritt, C. Platzer Björkman, Phys. Status Solidi 2016, 253, San Diego, USA, 2009.
247. [28] a) D. Cahen, R. Noufi, Appl. Phys. Lett. 1989, 54, 558; b) U. Rau,
[14] C. Krämmer, C. Huber, C. Zimmermann, M. Lang, T. Schnabel, M. Schmidt, Thin Solid Films 2001, 387, 141; c) U. Rau, D. Braunger,
T. Abzieher, E. Ahlswede, H. Kalt, M. Hetterich, Appl. Phys. Lett. H.-W. Schock, Solid State Phenom. 1999, 67, 409.
2014, 105, 262104. [29] R. Haight, X. Shao, W. Wang, D. B. Mitzi, Appl. Phys. Lett. 2014,
[15] Y. Cui, R. Deng, G. Wang, D. Pan, J. Mater. Chem. 2012, 22, 104, 033902.
23136. [30] a) K. Sardashti, R. Haight, T. Gokmen, W. Wang, L. Y. Chang,
[16] X. Meng, H. Deng, J. He, L. Sun, P. Yang, J. Chu, Mater. Lett. 2015, D. B. Mitzi, A. C. Kummel, Adv. Energy Mater. 2015, 5, 1402180;
151, 61. b) J. H. Kim, S.-Y. Choi, M. Choi, T. Gershon, Y. S. Lee, W. Wang,
[17] A. Gillorin, A. Balocchi, X. Marie, P. Dufour, J. Y. Chane-Ching, B. Shin, S.-Y. Chung, Adv. Energy Mater. 2016, 6, 1501902.
J. Mater. Chem. 2011, 21, 5615. [31] a) U. Rau, D. Braunger, R. Herberholz, H. Schock, J. Guillemoles,
[18] a) V. Alekseenko, A. Alekseenko, J. Geochem. Explor. 2014, 147, L. Kronik, D. Cahen, J. Appl. Phys. 1999, 86, 497; b) M. Kemell,
245; b) G. B. Haxel, J. B. Hedrick, G. J. Orris, P. H. Stauffer, M. Ritala, M. Leskelä, Crit. Rev. Solid State Mater. Sci. 2005, 30, 1;
J. W. Hendley, in Fact Sheet 087-02, U.S. Geological Survey 2002. c) F. Liu, C. Yan, J. Huang, K. Sun, F. Zhou, J. A. Stride,
[19] D. Shin, B. Saparov, T. Zhu, W. P. Huhn, V. Blum, D. B. Mitzi, Chem. M. A. Green, X. Hao, Adv. Energy Mater. 2016, 6, 1600706; d) C. Yan,
Mater. 2016, 28, 4771. F. Liu, K. Sun, N. Song, J. A. Stride, F. Zhou, X. Hao, M. Green, Sol.
[20] F. Hong, W. Lin, W. M. Meng, Y. Yan, Phys. Chem. Chem. Phys. 2016, Energy Mater. Sol. Cells 2016, 144, 700.
18, 4828. [32] C. Zhang, J. Zhang, Y. Hao, Z. Lin, C. Zhu, J. Appl. Phys. 2011, 110,
[21] J. Ge, P. Koirala, C. R. Grice, P. J. Roland, Y. Yu, X. Tan, 064504.
R. J. Ellingson, R. W. Collins, Y. Yan, Adv. Energy Mater. 2017, 7, [33] a) O. Gunawan, Y. Virgus, K. F. Tai, Appl. Phys. Lett. 2015, 106,
1601803. 062407; b) T. Gokmen, O. Gunawan, US9041389, 2015.

Adv. Mater. 2017, 1606945 1606945  (7 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim