TSpace Research Repository tspace.library.utoronto.ca
Thin-film Sb2Se3 photovoltaics with oriented
one-dimensional ribbons and benign grain
boundaries
Ying Zhou, Liang Wang, Shiyou Chen, Sikai Qin, Xinsheng Liu,
Jie Chen, Ding-Jiang Xue, Miao Luo, Yuanzhi Cao, Yibing Cheng,
Edward H. Sargent & Jiang Tang
Version
Post-Print/Accepted Manuscript
Citation
(published version)
Zhou, Y., Wang, L., Chen, S., Qin, S., Liu, X., Chen, J., Xue, D.-J., Luo,
M., Cao, Y., Cheng, Y., Sargent, E. H., and Tang, J. (2015). Thin-film
Sb2Se3 photovoltaics with oriented one-dimensional ribbons and
benign grain boundaries. Nature Photonics, 9(6), 409–415.
doi:10.1038/nphoton.2015.78
Publisher’s Statement
The final published version of this article is available at Nature
Photonics via http://dx.doi.org/10.1038/nphoton.2015.78.
How to cite TSpace items
Always cite the published version, so the author(s) will receive recognition through services that track
citation counts, e.g. Scopus. If you need to cite the page number of the TSpace version (original manuscript or
accepted manuscript) because you cannot access the published version, then cite the TSpace version in
addition to the published version using the permanent URI (handle) found on the record page.
1
Thin-film Sb
2
Se
3
photovoltaics with oriented 1D ribbons and benign grain boundaries
Ying Zhou
1,2
, Liang Wang
1,2
, Shiyou Chen
3
, Sikai Qin
1,2
, Xinsheng Liu
1,2
, Jie Chen
1,2
,
Ding-Jiang Xue
1,2
, Miao Luo
1,2
, Yuanzhi Cao
1
, Yibing Cheng
1
, Edward H. Sargent
4
, Jiang Tang
1,2*
1
Wuhan National Laboratory for Optoelectronics (WNLO), and
2
School of Optical and Electronic
Information, Huazhong University of Science and Technology (HUST), Wuhan, 430074, China
3.
Key Laboratory for Polar Materials and Devices (MOE), East China Normal University, Shanghai
200241, China
4
Department of Electrical and Computer Engineering, University of Toronto, Toronto, Ontario, M5S
3G4, Canada
Correspondence and requests for materials should be addressed to J. T. (E-mail:
jtang@mail.hust.edu.cn)
Solar cells based on inorganic absorbers such as Si, GaAs, CdTe and Cu(In,Ga)Se2 permit high
device efficiency and stability. Because of the crystals' three dimensional (3D) structure, dangling
bonds inevitably exist at the grain boundaries (GBs), significantly degrading device performance
via recombination losses. Thus, growth of single-crystalline materials or the passivation of defects
at GBs are required to address this problem, which introduces added processing complexity and
cost. Here we report that antimony selenide (Sb2Se3) - a simple, non-toxic and low-cost material
with optimal solar bandgap of ~1.1 eV - exhibits intrinsically benign GBs because of its one
dimensional (1D) crystal structure. Using a simple and fast (~1 μ m/min) rapid thermal
evaporation process, we orient crystal growth perpendicular to the substrate, and produce
Sb2Se3 thin-film solar cells with 5.6% certified device efficiency. Our results suggest that the
family of 1D crystals including Sb2S3, SbSeI and Bi2S3 show promise in photovoltaic
applications.
The urgent need for high-efficiency, low-cost solar cells drives sustained research on new absorber
materials for thin film photovoltaics. Copper zinc tin sulfide (CZTS) and organic-inorganic metal
halide perovskites (CH
3
NH
3
PbI
3
) are promising absorber materials that have achieved impressive
certified device efficiencies
1,2
. However, the complexity of defect physics associated with CZTS
3,4
appears to limit further efficiency improvements in this system, and concerns over stability and reliance
on lead in CH
3
NH
3
PbI
3
5
creates opportunities for new, stable, Pb-free materials.
These and other widely-explored photovoltaic absorber materials – such as Si, GaAs, CdTe, InP,
Cu(In,Ga)Se
2
(CIGS) – exhibit a three dimensional (3D) crystal structure
6-8
, i.e., they are bound by
covalent and/or ionic bonds in all three spatial dimensions. The 3D crystal structure guarantees
isotropic carrier transport and permits relaxed orientation control. However, at discontinuities such as
grain boundaries (GBs), dangling bonds typically act as recombination centers, and require further
steps to be remedied
9-11
. In CdTe for example (Fig. 1a), recombination loss through dangling bonds at
GBs lowers open-circuit voltage even when the best available passivation methods are employed
12
.
We hypothesized that materials possessing one dimensional (1D) crystal structure (Fig. 1b), such as
2
ribboned compounds Sb
2
Se
3
and Bi
2
S
3
, could, if suitably oriented, offer compelling performance. In
Sb
2
Se
3
, for example, (Sb
4
Se
6
)
n
ribbons stack along the [001] direction through strong covalent Sb-Se
bonds while in the [100] and [010] direction, the (Sb
4
Se
6
)
n
ribbons are held together by van der Waals
forces. These materials have been rarely explored for photovoltaics because of concerns about poor
carrier transport between ribbons, and worries about the mechanical stability of films.
We posited that a new materials processing strategy designed to orient ribbons vertically on the
substrate could permit photogenerated carriers to travel efficiently along the covalently-bonded 1D axis.
In the orthogonal direction, the parallel-stacked ribbons would provide substantially no dangling bonds,
even at GBs, thereby minimizing recombination losses – an important advantage in photovoltaic
application because recombination losses at GBs is one of the major limiting factors for high efficiency
thin film solar cells.
We began with first-principles simulations (VASP code) to study the structural relaxation and electronic
structure of Sb
2
Se
3
surfaces. The available terminations of (100) and (010) surfaces in Sb
2
Se
3
require
no breaking of Sb-Se bonds (Fig. 1c) and thus produce no dangling bonds. Other surfaces parallel to
the [001] direction also have no dangling bonds, e.g., the (110) and (120) surfaces. Their surface energy
is therefore lower than those with dangling bonds, such as the (001), (211) and (221) surfaces.
Calculation reveals that, as expected, the surface energies for (100), (010), (110), (120), (001), (211)
and (221) surfaces, respectively, are 0.44, 0.25, 0.33, 0.32, 0.46, 0.56, 0.53 J/m
2
. This indicates that the
most abundant surfaces in Sb
2
Se
3
samples will be the (010), (110) and (120) surfaces, i.e. those having
the lowest formation energies and no breakage of covalent bonds. Density functional theory confirms
quantitatively that no extra states are introduced inside the bandgap by these terminations. A clean
bandgap is seen in the calculated density of states (DOS) of the four Sb
2
Se
3
surfaces (Fig. 1d) and the
gap is comparable to that of the bulk Sb
2
Se
3
. Furthermore, there is no appreciable change in the
calculated DOS over a wide energy range (-10 to 3 eV), indicating that there is no significant
reconstruction on these surfaces. In sum, the ribbons represent the fundamental building block of the
Sb
2
Se
3
structure, and surface reconstruction is negligible as long as this basic repeat unit is not broken.
Overall, this computational study confirms that, as long as the Sb
2
Se
3
ribbons are suitably oriented, the
GBs will be terminated by the intrinsically benign surfaces (e.g. (100), (010), (110) and (120) planes)
and minimize recombination loss. This is in striking contrast to most known photovoltaic absorbers, in
which the breakage of covalent bonds introduces defect states and recombination centers at GBs
13-15
.
A number of additional properties make Sb
2
Se
3
particularly worthy of experimental investigation for
high efficiency, low cost thin film solar cells
16
. Like CdTe, but in contrast with CIGS and CZTS,
Sb
2
Se
3
is a simple binary compound with fixed phase and stoichiometry. It has a very strong absorption
coefficient (>10
5
cm
-1
at short wavelengths) and its bandgap is ~1.1 eV
17,18
, optimal for single-junction
solar cells. The constituents of Sb
2
Se
3
are non-toxic and low in cost (Sb has similar cost to Cu), and, as
we proceed to show herein, Sb
2
Se
3
films are produced using minimal energy, enabling in principle a
low energy-payback time for a solar cell
19
. All of these considerations motivated further study of
oriented Sb
2
Se
3
films and devices herein.
Sb
2
Se
3
has a low melting point of 608 ℃ and a high saturated vapor pressure (~ 1200 Pa at 550 ℃)
20
,
3
enabling us to carry out film deposition using rapid thermal evaporation (RTE) in a tube furnace with
high ramp rate (supplementary Fig. S1). Sb
2
Se
3
powder was directly applied via evaporation under low
vacuum pressure (~ 8 mTorr), maintained by a mechanical pump. Once heated up, Sb
2
Se
3
powder
evaporated and condensed on the substrate because of temperature gradient, forming the Sb
2
Se
3
film.
The distance between the evaporating source and the substrate was kept at a low value of 0.8 cm to
enable high material usage and a fast deposition rate. The deposition rate was as high as 1 μm/min,
much greater than regular thermal evaporation (typically 0.01-0.1 μm/min) or sputtering (typically
0.01-0.05 μm/min) and comparable to confined space sublimation (CSS)
21
, a key technology that has
enabled the high manufacturing throughput and commercial success of CdTe solar cells. The RTE
process is distinct from CSS since, in the RTE performed herein, Sb
2
Se
3
melts and evaporates from the
liquid phase, in contrast with direct sublimation from the solid. Using this simple and fast technology,
phase-pure Sb
2
Se
3
film (Supplementary Fig. S2) and solar cells were readily fabricated with high
reproducibility.
We first analyzed using transmission electron microscopy (TEM) specific regions of the Sb
2
Se
3
films
made in actual solar devices. The full device stack consisted of a FTO substrate, a thin ( ~ 60 nm) CdS
layer produced through chemical bath deposition, an approximately 390 nm thick Sb
2
Se
3
absorber layer
deposited by the RTE process, and top Au electrodes from thermal evaporation. For TEM
characterization, the sample was prepared by cross-sectioning from the certified device using focused
ion beam. The cross-sectional TEM image showed that the Sb
2
Se
3
film was compact and composed of
large Sb
2
Se
3
grains whose size equals the film thickness (Fig. 2a). To check whether the grains were
single-crystalline, we applied high-resolution transmission electron microscopy (HRTEM) to analyze
three arbitrarily-selected points (I, II and III in Fig. 2a) with corresponding lattice fringes in Fig. 2b, 2c
and 2d. The distances between lattice lines was measured to be 0.320 nm and 0.318 nm, corresponding
to the separation of (21
1
) and (211) planes in orthorhombic Sb
2
Se
3
. The crystal planes extend
continuously from the top to the bottom of the active region of the device, confirming that this grain
was single-crystalline. This trend was confirmed with further study of multiple grains (Supplementary
Fig. S3), reinforcing the picture that the Sb
2
Se
3
films are made up of single crystalline grains.
Orientational control of the Sb
2
Se
3
film is predicted to be crucial to realize efficient carrier transport
and benign GBs. We investigated therefore the possibility of correlation between Sb
2
Se
3
film
orientation and photovoltaic device performance. Different substrate temperatures during the RTE
process were explored maintaining identical Sb
2
Se
3
film thickness. Devices of Class A were deposited
onto substrate kept at 300℃, and devices of Class B were deposited onto 350℃ substrate. Both
devices had an area of 0.095 cm
2
defined by the gold electrodes. Typical device performance (Fig. 3a)
measured under 100 mW/cm
2
illumination (class 3A solar simulator) reveals short-circuit current
density (Jsc) for Device A of 27.2 mA/cm
2
, fill factor (FF) of 53% and series resistance (R
S
) of 40 Ω;
while the corresponding values in Device B is 18.4 mA/cm
2
, 47% and 89 Ω. We first applied energy
dispersive X-ray spectroscopy (EDX) (Supplementary Fig. S4 and Table S1-3) and capacitance-voltage
(Supplementary Fig. S5) to analyze our devices and found out that both the Sb
2
Se
3
films deposited at
substrate temperature of 300℃ and 350℃ were slightly Se-rich (molar ratio Se: Sb is 1.51) and
4
showed very close doping density. This indicated that film composition and consequent doping
density likely did not account for the observed device efficiency difference. The large improvement
from device Class B (3.2% efficiency) to Class A (5.6% efficiency) cannot be accounted for by
stoichiometry/compositional differences, and we instead ascribe it to different film orientation. Indeed
for device Class A, the diffraction intensity associated with the (120) peak is much weaker than for the
(211) peak, while in device Class B the intensities of these two peaks are comparable with each other
(Fig. 3b).
To quantitate orientation effects, we calculated the texture coefficient of the (120) peaks from 140
devices and plotted the corresponding device efficiency versus texture coefficient in Fig. 3c. The
texture coefficient measures film orientation (Supplementary Fig. S6), with large texture coefficients
for a diffraction peak indicating preferred orientation along this direction
22
. There exists a strong
correlation between device performance and film orientation: device efficiency monotonically
decreases when the texture coefficient of (120) orientation increases. Further analysis revealed that for
all of these devices, the shunt resistance (R
Sh
) remains constant at approximately 1600 Ω, confirming
similar junction quality, while the series resistance increased monotonically with an increased value of
the (120) texture coefficient. We can explain this observation – that preferred orientation along the [120]
direction resulted in significantly increased series resistance and decreased device efficiency – by
noting that the [120] orientated grain consists of (Sb
4
Se
6
)
n
ribbons horizontally stacked in parallel with
the substrate (Fig. 3d). In contrast, [221] orientated grain consists of tilted (Sb
4
Se
6
)
n
ribbons stacked
vertically on the substrate. It should be noted that the [211] orientated grain is also composed of tilted
(Sb
4
Se
6
)
n
ribbons, but to a different angle, thus we limit our transport discussion to the [221] orientated
grain. Naturally, carrier transport in the [211] orientated grains should be much easier than in the [120]
orientated grains because in the former ones carriers travel within the covalently bonded (Sb
4
Se
6
)
n
ribbons, while in the latter ones they are required to hop between ribbons held together by van der
Waals forces. Furthermore, the GBs of the [211] orientated grains are composed of low surface energy
(hk0) planes such as (100), (010), (110), (120) planes, which are free of dangling bonds and should
cause little recombination loss. In sum, we explain lower series resistance, better fill factor, and higher
short-circuit current as resulting from improved transport and lowered recombination loss observed in
devices having preferred [211] orientation of Sb
2
Se
3
active layers.
For polycrystalline thin film solar cells, passivation of GBs to suppress strong carrier recombination is
mandatory to achieve high efficiency devices. Typical examples are the carefully engineered Cu
deficient GBs in CIGS solar cells
23,24
and a high temperature CdCl
2
treatment for CdTe solar cells
24,25
.
We studied the properties of GBs in our Sb
2
Se
3
solar cells using Kelvin probe force microscope (KPFM)
and electron-beam-induced current (EBIC) measurements. Two-dimensional topography spatial maps
and the corresponding surface potentials of Sb
2
Se
3
thin films (Fig. 4a and 4b) reveal that there is no
correlation between GBs (identifiable by notable changes in surface topography) with substantial
potential variation in KFPM image. Overall, the average roughness of Sb
2
Se
3
film is 23 nm, while the
average surface potential difference is a very low (much below kT) 9.1 mV. The surface potential
fluctuations are extremely small compared to CIGS and CZTS films (generally >100 mV in a 2.5 μm x
2.5 μm zone
9,25
). In an illustrative line scan crossing the GBs (Fig. 4c), the surface potential difference
between two grains is as low as 10 mV, indicating a lack of significant band bending and surface
defects in the Sb
2
Se
3
films.