You are on page 1of 9

Article

pubs.acs.org/cm

From Nano- to Micrometer Scale: The Role of Antisolvent Treatment


on High Performance Perovskite Solar Cells
S. Paek,† P. Schouwink,† E. Nefeli Athanasopoulou,‡ K. T. Cho,† G. Grancini,† Y. Lee,† Y. Zhang,†
F. Stellacci,‡ Mohammad Khaja Nazeeruddin,*,† and P. Gao*,†

Group for Molecular Engineering of Functional Materials, EPFL Valais Wallis, Sion, CH-1951 Switzerland

Supramolecular Nano-Materials and Interfaces Laboratory, EPFL, Lausanne, CH 1015 Switzerland
*
S Supporting Information
Downloaded via INDIAN ASSOCIATION CULTIVATION SCI on January 27, 2023 at 19:55:29 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: To date, antisolvent treatment has become one of the most important means to fabricate high efficiency
perovskite solar cells (PSCs); however, the few reported antisolvents have not been analyzed on a uniform platform, and there is
hitherto no clear reasoning in the choice of antisolvents toward high performance PSCs. Here, we study the role of the
antisolvents in the nucleation kinetics of perovskite solutions and their residual influence on perovskite crystal growth, film
formation, and device performance. Through X-ray diffraction analysis on the complicated double mixed perovskite, we
qualitatively evaluate the impact of thermal annealing and antisolvent treatment (A.S.T.) on the phase composition and
microstructure of the films. By using miscible antisolvents with high boiling point instead of immiscible low boiling point
solvents, we obtain homogeneous and almost pinhole-free perovskite films. When using trifluorotoluene (TFT) to replace
toluene and chlorobenzene as a novel antisolvent, we achieve a power conversion efficiency (PCE) of 20.3% under optimized
device fabrication conditions with a composite perovskite as active layer. The conclusions from this study should assist in
establishing reproducible fabrication processes and finding better antisolvent candidates for perovskite solar cells.

1. INTRODUCTION (A.S.T.),10,11 solvent vapor annealing,12,13 and vacuum treat-


It has been calculated that the solar energy harvested by 0.4% of ment14 have been the most widely used post-treatment
the world’s surface area would be sufficient to meet the global methods. Enlightened by organic photovoltaic (OPV) device
energy demands, assuming an average solar energy conversion engineering,15 the A.S.T. is at the origin of several reported
efficiency of 15% can be popularized.1 Very recently, the world record efficiencies.11,16−20
dominating photovoltaic community based on crystalline, Normally, the perovskite precursors are dissolved in high
inorganic semiconductors was agitated by the fast development boiling point (BP) polar aprotic solvents such as dimethyl
of perovskite solar cells (PSCs).2,3 The surge in the perovskite sulfoxide (DMSO), dimethylformamide (DMF), dimethylace-
based photovoltaic technology is evidenced by the abnormally tamide (DMAc), N-methyl-2-pyrrolidone (NMP), and γ-
rapid update of the record efficiencies in three years.4 As a butyrolactone (GBL). Antisolvents assisting perovskite film
latecomer, despite the short time spent on the optimization of formation include a larger scope of solvents that cannot
the perovskite devices, the huge amount of efforts dedicated by dissolve any precursor of the perovskite recipe and the
scientists from all over the world led to rapid improvements in perovskite itself. It is known that the working mechanism of
the device engineering methods. It has been recognized that the an antisolvent is to speed up heterogeneous nucleation via the
morphology of the perovskite layer is one of the most crucial creation of an instantaneous local supersaturation on the
parameters that determine the final performance of a PSC
device.5 Among the many different methods introduced to Received: December 18, 2016
improve the surface morphology of perovskite film, gas- Revised: March 25, 2017
blowing,6 amino halide additive,7−9 antisolvent treatment Published: March 26, 2017

© 2017 American Chemical Society 3490 DOI: 10.1021/acs.chemmater.6b05353


Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

spinning substrate.10 During the treatment of antisolvents, due to differences in boiling point (BP), miscibility, and
complicated interactions are happening simultaneously under dielectric constants. Antisolvents with both low boiling point
the influence of several physicochemical properties of both and poor miscibility with solution solvents will lead to lower
solvents (Table S1). Xiao et al. screened 12 different kinds of efficiency and reproducibility comparing the antisolvents with
solvents for the treatment of spinning methylammonium lead higher boiling point and good miscibility with DMF/DMSO.
iodide (MAPbI3) perovskite films.10 However, some of these By using high boiling point and miscible antisolvents under
solvents can dissolve the ammonium halide salt and destroy the optimized conditions, we could always obtain compact
final perovskite film. Zheng et al. evaluated three different perovskite films with full coverage on the substrate providing
antisolvents and various interval dropping times on the final decent power conversion efficiencies (PCE) above 18%. When
morphology of methylammonium lead bromide (MAPbBr3).21 using TFT to replace Tol and CB as a novel antisolvent, we
To date, the most popular antisolvents for perovskite film achieve a PCE of 20.3% under optimized conditions with
formation are toluene (Tol),16 chlorobenzene (CB),18 and composite perovskite as active layer, measured under one sun
diethyl ether (ether).20 Interestingly, these three solvents have illumination.
dramatically different physicochemical properties but still
perform efficiently with different perovskite films. However, 2. RESULTS AND DISCUSSION
in these reports, no thorough study on the interaction between 2.1. Kinetic Study of Solvent−Solvent Interaction
different antisolvents as shower solvent and solution solvents as Effect on Perovskite Phase Formation. One of the major
bath solvent was conducted. More importantly, these popular variables during the antisolvent treatment is the interaction
antisolvents have not been evaluated on the same type of between the solution solvent and the antisolvent. We designed
perovskite film to make a fair comparison and elucidate the an experiment to monitor the kinetics of this process by
limitations of each solvent. A complete understanding of these recording the increase in absorbance at 550 nm as a function of
issues is crucially important for advancing our ability of time, when the freshly spin-coated perovskite intermediate films
controlling perovskite film morphology and improving solar cell from DMSO/DMF (1:4) were treated with the six kinds of
performance. antisolvents (Figure 2a−f).22 Depending on the mutual
To shed light on this perspective, in this study, a series of six miscibility and dielectric constants, the six antisolvents
antisolvents including trifluorotoluene (TFT), toluene (Tol), exhibited dramatically different kinetic behavior in terms of
chlorobenzene (CB), p-xylene (Xyl), diethyl ether (ether), and the increment and the time constant in the relative absorbance
dichloromethane (DCM) with different dielectric constants (ε, change. The kinetic traces can be fitted with mono-, bi-, and
solvating abilities) and other physicochemical properties are triexponential functions, and the corresponding time constants
chosen to treat the spinning composite perovskite films can be deduced (Table S2). Generally, the data shows a steep
deposited from a mixed DMSO and DMF solution (1:4) rise of the absorbance upon the addition of the antisolvents
(Figure 1, Table S1). The kinetics of the treatment process was (1−42 s), which is indicative of an accelerated heterogeneous
nucleation and growth process to different extents accompany-
ing the fast solvent (mobile) extraction. This fast component
(τ1) is followed by an approximate plateau that indicates the
completion of nucleation and local crystal growth, when the
solvent extraction reaches a state of equilibrium. The small
additional increase that occurs on a slow component of about
50−1000 s (τ2 and/or τ3) contributes only little to the total
change in absorbance and is attributed to the slow diffusion of
remaining solution solvents (complexed) out of the film and
slow growth of large crystals leading to enhanced light-
scattering. Figure 2b,c,f shows that the fast increase of
perovskite absorption at 550 nm is practically complete within
a few seconds (<15 s) with the DMSO/DMF miscible solvents
TFT, Tol, CB, and DCM. The fast increase in absorbance
(optical density) indicates the rapid formation of perovskite
crystals. On the other hand, for the DMSO/DMF immiscible
solvents (xylene and ether), the fast components are slower
(30−40 s) and the absorbance at the plateau stage is much
lower. Xylene showed the lowest absorbance due to the fact
that it is immiscible with all of the solution solvents. This
Figure 1. Schematic molecular structures and miscibility behavior of behavior is reasonable, as the interaction, and therefore the
bath solvents (DMSO and DMF) and shower solvents (TFT, exchange rate between the solution solvents and antisolvents, is
chlorobenzene, toluene, p-xylene, DCM, and ether).
much weaker or slower when they are immiscible.
The perovskite films are annealed at 100 °C after the kinetic
simulated by observing the change in absorbance at 550 nm. absorbance study and before UV−vis absorption and photo-
The morphology and composition of the resulting intermediate luminescence (PL) are measured. The latter two are shown in
and perovskite films were studied by optical microscopy (OM), Figure 2g. The relative absorbance at 550 nm of the films
scanning electron microscopy (SEM), X-ray diffraction (XRD), before and after thermal annealing is extracted and plotted in
and atomic force microscopy (AFM). Our findings reveal that Figure 2h. The time nodes for absorbance in the dynamic traces
the antisolvent has a substantial effect on the perovskite crystal are chosen according to the optimized device procedure (vide
growth kinetics and film morphology on a micrometer scale infra). It is interesting to see that, although the absorbance
3491 DOI: 10.1021/acs.chemmater.6b05353
Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

Figure 2. Kinetic traces at 550 nm showing the change in absorbance during the treatment of antisolvents (a) TFT, (b) toluene, (c) chlorobenzene,
(d) xylene, (e) ether, and (f) DCM as a function of time. (Orange spots indicate the absorbance at the time node before thermal annealing and after
A.S.T. in the real device fabrication step, the delay time, vide infra. The x-axis is shown as logarithmic scale.) (g) Steady-state absorption spectra of
perovskite film prepared under different antisolvents treatment after annealing (absorbance at 550 nm is highlighted as the brown spots). (h) Bar
chart showing the change in absorbance at 550 nm before and after annealing.

Figure 3. Schematic processing scheme for perovskite films with and without antisolvent treatment (A.S.T.). (a) The speed-time profile for the spin
coating process with and without the various A.S.T. Optical micrographs (OM) illustrating oriented growth of needle shaped crystals with distinct
size variation before thermal annealing as a function of various antisolvents (b−g) and without A.S.T. (h). Scale bar, 300 μm.

recorded before annealing is fluctuating among different 2.2. Perovskite Film Formation with Antisolvent
antisolvent treatments, the absorbance reached almost the Treatment. The effects of different antisolvent treatment are
same level after thermal annealing, except in the case of ether; characterized in making device scale thin layers. Figure 3 shows
this exception owed to a poor film morphology (vide infra). a schematic representation of the device fabrication processes
Also, there is no prominent change in the position of PL involving different kinds of antisolvent treatment. DMF and
spectra. DMSO as cosolvents (4:1) containing 1.15 M PbI2, 1.10 M
3492 DOI: 10.1021/acs.chemmater.6b05353
Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

Figure 4. XRD patterns of perovskite films treated with different antisolvents and without antisolvent (w/o A.S.) treatment before (a) and after (b)
thermal annealing at 100 °C for 90 min. In (d), the fwhm of Bragg peaks (before annealing) representing two different directions (c) is shown, as
well as the ratio of integral intensities. The peak widths of samples after annealing and their integral intensities are given in (e) and (f). *, δ, #, ●, and
(hkl) denote diffraction peaks corresponding to the MAI-FAI-PbI2-PbBr2-DMSO (Pb3I(Br)8 polymorphs) and nonperovskite phases of FAPbI3,
PbI2, FTO, and perovskite, respectively.

FAI, 0.2 M PbBr2, and 0.2 M MABr are prepared. After time (Ttreat). After the entire process, all the antisolvent treated
initializing a high-speed spin-coating step at 6000 rpm, films showed orange color while the film without A.S.T was
antisolvents were drop-casted on the spinning film in one yellow. Before thermal annealing, the films exhibited dramat-
shot with different, optimized delay time to reach the highest ically different textures as seen by optical microscopy (Figure
PCE (Figure 3a). The total spin span (Tspin) at 6000 rpm is 30 3b−g). Low boiling point (<50 °C) antisolvent (ether and
s for processes with and without A.S.T. and can be divided into DCM) treatment leads to a nonuniform central area in the film
two stages as spreading time (Tspread) and antisolvent treating (Figure S1). This could be due to the fact that there is no
3493 DOI: 10.1021/acs.chemmater.6b05353
Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

thermal contact between the vacuum holder of the spin-coater (H, C, and N) render this impossible at this time. Though an
and the substrate and the evaporative cooling leds to quicker unambiguous quantitative investigation of the microstructure is
evaporation in the central region of the substrate (particularly not possible without accurate intensities, some qualitative
with low bp solvents). After treatment by miscible antisolvents observations do allow us to conclude on different aspects.
(TFT, toluene, chlorobenzene, and DCM), needle shaped Therefore, the following discussion attempts to describe results
crystals grew in the form of a radial pattern or starburst with in a rather qualitative way.
TFT and DCM having the highest needle density. The lengths Given the lack of an accurate structural model (and space
of the needles were in the range of 200 to 500 μm (Figure 3b− group symmetry), a general evaluation of these diffractions was
d,g). On the other hand, the films treated with immiscible made based on the reduced cubic lattice parameter from the
solvents (p-xylene and ether) showed much shorter (∼40 nm) annealed samples by Le Bail fits, using Topas 5.30 These fits
needle crystals (Figure 3e,f). The needle crystals after p-xylene were performed in cubic, tetragonal, and trigonal (rhombohe-
treatment are more homogeneously distributed and denser than dral) space groups but yielded insignificantly different R values
those from ether treatment. This phenomenon indicates that and were thus inconclusive regarding the space group
nucleation kinetics is dominating compared to crystal growth. symmetry and associated metric. The determined volume per
The differences in crystal size after the treatments by two formula unit of perovskite under study is, nevertheless, a
distinct kinds (miscible and immiscible) of solvent are in reliable quantity and amounts to 249.9 Å3 at 298 K, the
agreement with the absorbance maxima observed in the kinetic deviation from this value is within e.s.d.’s for all A.S.T. samples
study, as the crystal growth speed is much slower with the latter after annealing. This value is perfectly compatible with a slight
(Figure 2). Just after spin-coating, the composite perovskite change of unit cell dimensions compared to the respective end
film without any A.S.T. treatment showed a wet surface covered members and in agreement with 241 Å3 at 170 K reported
with sparsely distributed short yellow needle-like and cubic recently for the single crystal of (FAPbI3)0.85(MAPbBr3)0.1531
dark crystals on top indicating a very deficient and (Table S3). Supported by the lack of occurring superstructure
inhomogeneous heterogeneous nucleation kinetic10 (Figure reflections, this evidences that the identical discrete chemical
3h). composition is formed regardless of which solvent is applied.
2.3. X-ray Diffraction Study. The impact of thermal Importantly, this also proves that the variation in photovoltaic
annealing and A.S.T. on the phase composition and micro- quantities arises entirely due to the interaction with the solvent,
structure of the films was investigated by powder X-ray not from the chemical composition.
diffraction (XRD). Figure 4 shows wide angle XRD patterns of From the XRD patterns before and after annealing (Figure
the composite perovskite film deposited on top of TiO2-coated 4a,b), we observe signals at approximately 14.1, 20.0, and 24.5°
FTO glass substrates by means of different antisolvents before (2θ, Cu radiation). Though they relate to different lattice
(a) and after (b) thermal annealing. During thermal treatment, planes in different symmetries, they correspond to orientations
solvent molecules are driven out of the precursors in order to of the perovskite octahedral framework along octahedral
form the final perovskite film. It is seen that, upon thermal corners (14.1°), octahedral edges (20.0°), and octahedral
annealing for 1.5 h at 100 °C, transformation from the faces (24.5°), respectively, regardless of the precise structure or
precursors to the perovskite phase is nearly complete in all tilt pattern of the octahedral framework (Figure 4c). Since the
cases. Minor impurities arising from the diffraction of the unit cell parameters are identical for all materials, we do not
photoinactive δ-phase of FAPbI3 and PbI2 can be found at 11.6° expect the integrated intensities to change due to changes in
and 12.7°, respectively. These impurities are inherent to the the crystal chemistry. Therefore, intensity changes should arise
vast majority of reported FA based perovskites assembled in due to different orientations of the crystallites on the substrate.
such films, evidencing difficulties of complete conversion into In Figure 4a, with the exception of the strong diffractions
the photoactive phase of the perovskite. The literature on the from the FTO substrate, the remaining Bragg signals can be
respective end members of the 4-component phase diagram, assigned to a highly crystalline intermediate phase formed by
FA, MA, I, and Br, is also ambiguous. Single crystal studies on MABr-FAI-PbI2-PbBr2-DMSO.11 A similar intermediate phase
the room temperature phase α-FAPbI3 have reported it as was recently solved as a Pb3I(Br)8 polymorph with two cations
trigonal in space group P3m1,23,24 volume per formula unit (V/ and intercalated DMSO molecules32,33 (Figure S3). For the
f.u.) = 256.3 Å3. Neutron powder diffraction studies have films without A.S.T., the prominent peak at 9.2° is the
established a cubic Pm-3m phase with a V/f.u. of 257.5 Å3.25 characteristic diffraction from the intercalated Pb3I(Br)8
Cubic Pm-3m FAPbBr3 has a much lower V/f.u. at 214.8 Å3,26 polymorph marked by asterisk in Figure 4a, whose phase
and tetragonal I4/mcm MAPbI3 has been reported at 243 Å3, at yield decreases after treatments with immiscible antisolvents (p-
298 K;27 very recently, the space group I4/mcm has been xylene and ether) and almost disappears if the antisolvent is
corrected to a subgroup, I422.28 Moreover, the substitution FA miscible with DMSO/DMF (TFT, toluene, chlorobenzene, and
by MA or vice versa follow’s Vegard’s law,29 meaning that DCM).
presently a simultaneous substitution on both the A-cation and Figure 4d summarizes the line widths at fwhm (full width at
anion site are compatible with the volume reported herein, half-maximum) and peak intensities resulting from profile fits
since a pure A-site substitution FA by MA of 15% of the organic on data collected before thermal annealing. Two immediate
molecule would, according to Vegard, result in a unit cell observations are striking. While the peak width of signals
volume of approximately 255 Å3 (Table S3). representing the lattice direction along octahedral edges is
We have herein attempted to solve and refine the specific approximately constant throughout all samples, both films
perovskite structure based on our powder diffraction data but treated with the immiscible antisolvents xylene and ether show
without success. The inherent problems of preferred pronounced broadening for the diffractions that relate to
orientation on deposited thin films, significant structural corners, suggesting a smaller domain length in this direction,
disorder, low intensity laboratory data, and the coexistence of which is in good agreement with OM images in Figure 3.
very strong scatterers (Pb, Br, and I) next to very weak ones Second, the intensity ratio between edge- and corner-related
3494 DOI: 10.1021/acs.chemmater.6b05353
Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

reflections is significantly lower for TFT and toluene treated During thermal annealing, the solvent molecules are driven
samples compared to the rest (the intensity of the 14° out of the films, while the perovskite phase proceeds its growth.
reflection of the w/o A.S. data set was not sufficient to be As shown in Figure 4b, after thermal annealing at 100 °C for 90
fitted). This alludes to grains becoming attached on the min, all the intermediate phases have converted to the typical
substrate preferentially along the octahedral edges rather than composite perovskite regardless of the A.S.T. XRD data
corners, upon A.S. treatment with TFT and toluene, and before collected from the films after thermal annealing show similar
thermal annealing. Interestingly, these are the two solvents crystallinity of A.S. treated samples with respect to the
performing best in photovoltaic measurements (Table 1). untreated one as suggested by the line widths at fwhm (Figure
4e). Though diffraction line broadening is less easily discussed
Table 1. Summary of Photovoltaic Parameters Derived from without accurate crystal structures, we can conclude, again
J−V Measurements of Composite Perovskite Based based on the prerequisite of identical chemical compositions,
Champion Devices that the antisolvent treatment does have an impact on the
coherent domain size responsible for Bragg diffraction. First, it
VOC JSC series resistance
antisolvent (V) (mA cm−2) FF η (%)a (Rs) (Ω/cm2) can be seen that, independent of the antisolvent, the crystal
w/o A.S.T. 0.98 19.6 0.74 14.4 95.8
growth direction along octahedral edges (20.0°, 2θ) produces a
trifluorotoluene 1.10 23.3 0.79 20.3 40.4
larger domain length. We note that no reported photovoltaic
toluene 1.10 23.3 0.77 19.7 46.0 perovskite phase can be found, where the 14.1° Bragg signal is
chlorobenzene 1.09 23.0 0.75 18.8 68.0 composed of a doublet and where the signals at 20.0° and 24.5°
p-xylene 1.08 21.9 0.75 17.8 52.0 are both singlets, which allows us to eliminate systematic errors
ether 1.05 22.0 0.76 17.6 47.1 due to neglected peak convolutions. In space group symmetries
a
Performance of organic−inorganic cells was measured with 0.16 cm2
I4/mcm and P3m1, the 14.1° signal is in fact composed of a
working area. doublet (110, 002, and 011 and −111, respectively). However,
only in the case of P3m1, the 20.0° Bragg signal is a single peak,
but the 001 and −111 reflections have an intensity ratio of 5/

Figure 5. Typical AFM images of perovskite films in an area of 10 × 10 μm. The films were deposited on a FTO/c-TiO2/meso-TiO2 substrate
treated with five different antisolvents, TFT (a), toluene (b), chlorobenzene (c), xylene (d), ether (e), and a film without treatment (f). Line
segments from each scan (marked with redline) (g) and the statistics of RMS from five different areas of each film (h) (Table S4).

3495 DOI: 10.1021/acs.chemmater.6b05353


Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

95; they are hence approximately a singlet in observed A.S.T. The results for the optimized devices are summarized in
experimental data. A further peak located around 31.8° (2θ) Figure 6a. The cross section of the devices was verified by SEM
was inspected regarding its fwhm, corresponding to a
crystallographic direction which is tilted by 15° with respect
to the octahedral edges. Indeed, the width of this signal comes
to lie at similar values as that corresponding to edges.
The intensity variations from the films after thermal
annealing are shown in Figure 4f. The lack of real systematics
in relative intensities between peaks manifests that the influence
of antisolvent on the orientations of the crystallites can be
eliminated by thermal annealing. However, a further general
qualitative observation on the peak shapes suggests a decrease
in domain size upon thermal annealing (Figure 4d,e),
manifested in the slightly broader edge-related and significantly
broader corner-related Bragg peaks post-annealing. These
changes most likely arise from solid−solvent interaction and
potential domain cracking due to solvent diffusion processes
and exchange reactions, which take place during the thermally
induced crystal growth. As stated throughout this discussion,
we do not attempt to quantify genuine potentially anisotropic
size effects, since they can currently not be separated from
stress and strain related broadening, inhibiting more
sophisticated analysis. Since anisotropic growth has been
reported before, we do assume that size effects dominate the
line broadening.
2.4. Atomic Force Microscopy Study of Film Top- Figure 6. Current (J)−voltage (V) curves of the optimized cells (a).
ography. Due to the unstable nature of the intermediate Statistical deviation of the PCEs for 25 different solar cells (b) from
phase, it is difficult to study further the topography of the films different antisolvents treatment under AM 1.5 conditions (100 mW/
on a nanometer scale just after A.S.T. and before annealing cm2).
without destroying the sample. We show SEM images of the
respective annealed perovskite films with different A.S.T. in and is shown in Figure S4. All the photovoltaic devices have a
Figure S2. DCM showed poor texture and low reproducibility structure of FTO/compact TiO2/meso-porous TiO2/perov-
of alike films, owed to the low boiling point and high density, so skite/Spiro-OMeTAD/Au (80 nm) as shown in Figure S5.
in this work, no further optimization of the processing Device fabrication and measurement details are given in the
conditions was carried out. It is worth noting that, after Supporting Information. The cross-section determined by SEM
thermal annealing, all the films treated with high boiling point shows that all films treated with miscible solvents (TFT,
antisolvents formed perovskite films with a similar grain density toluene, and chlorobenzene) exhibit a similar capping perov-
and coverage on a nanometer scale, as can be seen from Figure skite thickness around 485 nm. This capping layer thickness is
S2a−d. However, the films treated by ether or without A.S.T. increased to 510, 495, and 567 nm upon treatment with the
are full of pin-holes (Figure S2e,f). This emphasizes again the immiscible xylene, ether, or without A.S.T. The photocurrent
necessity and applicability of A.S.T. for the film formation of J−V curves of the best devices of a respective A.S.T. as well as
composite perovskite. We further characterized the films after of the untreated device are shown in Figure 6a, with the
different A.S.T. by atomic force microscopy (AFM) (Figure 5). corresponding metrics reported in Table 1. Optimized PCEs
We calculated the root mean-squared (RMS) roughness of the were obtained with different delay times after antisolvent
perovskite films treated with TFT (a), toluene (b), dripping: 20 s (trifluorotoluene, 110 μL), 15 s (chlorobenzene,
chlorobenzene (c), p-xylene (d), ether (e), and a film without 100 μL), 15 s (toluene, 600 μL), 15 s (xylene, 900 μL), and 10
treatment (f) to be 45.1 ± 4.3, 39.9 ± 2.8, 40.2 ± 1.5, 35.9 ± s (diethyl ether, 200 μL), respectively, before the end of the
1.9, 85.5 ± 15.5, and 105.4 ± 40.5 nm, respectively, based on procedure. As shown in Table 1, the champion PCEs under
five different areas of 10 × 10 μm from each film. The optimized conditions for perovskites with TFT, toluene,
roughness of films treated with high boiling point solvents chlorobenzene, p-xylene, ether, and without treatment are
(TFT, toluene, chlorobenzene, and p-xylene) is greatly reduced 20.3%, 19.7%, 18.8%, 17.8%, 17.6%, and 14.4%, respectively. It
compared to the film treated with a low boiling point solvent is noteworthy that the VOC of solar cells after antisolvent
(ether) or without A.S.T., as is evident from both, AFM line treatment are generally higher than that of the reference cell.
segments (Figure 5g) and RMS statistics (Figure 5h). It is This can be understood by the AFM images in Figure 5 and
noteworthy that, due to the existence of pinholes, both the SEM images in Figure S2. We can see that the number of holes
ether-treated and untreated films showed a large surface is significantly higher without antisolvent treatment, which will
inhomogeneity. In addition, the RMS statistics representative increase the contact between the hole transporting materials
of five different areas of each film exhibit a narrower with TiO2 and enhance charge recombination inside the device.
distribution in the films treated by high boiling point solvents, Figure 6b demonstrates the statistical distribution of power
regardless of the miscibility. conversion efficiency (PCE) for 25 different cells with different
2.5. Photovoltaic Property Evaluation of Different A.S.T., indicating an outstanding reproducibility for each
Antisolvent Treatments. Perovskite solar cells were respective antisolvent condition. The PCEs we achieved with
fabricated using the perovskite films prepared from the different toluene and chlorobenzene are on par with the reported values
3496 DOI: 10.1021/acs.chemmater.6b05353
Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

in the literature with similar device structure and perovskite fitted time constant values, calculated XRD data, reduced
composition.16,18,19,34 TFT, xylene, and ether are used as cubic unit cell volume per formula unit, cross-sections of
antisolvents for the first time herein to treat composite full devices, photo images of precipitation experiments,
perovskite films. Among them, TFT has the highest ε and RMS statistics, dependence of dielectric constant with
gave the highest PCE of 20.3%, outperforming toluene and ratio of DMF, and hysteresis study (PDF)
chlorobenzene. Although p-xylene and ether have ε values
comparable to other solvents, their immiscibility with DMSO
and/or DMF leads to different nucleation and growth
behaviors. However, the PCEs from the devices treated by
■ AUTHOR INFORMATION
Corresponding Authors
the two immiscible solvents gave reasonably good PCEs of *E-mail: peng.gao@epfl.ch.
17.8% and 17.6%, respectively. Importantly, all devices with *E-mail: mdkhaja.nazeeruddin@epfl.ch.
A.S.T. gave PCEs superior to those of nontreated devices ORCID
(14.4%). The difference can be reflected in the contact quality S. Paek: 0000-0002-2671-2909
with Spiro-OMeTAD, which determines the series resistance
G. Grancini: 0000-0001-8704-4222
(Rs) of the total device. Devices with TFT treatment exhibited
the lowest Rs, while the devices without A.S.T. showed the F. Stellacci: 0000-0003-4635-6080
highest Rs, which can well explain the difference in FF and JSC, Mohammad Khaja Nazeeruddin: 0000-0001-5955-4786
highlighting the importance of choosing appropriate anti- P. Gao: 0000-0002-4963-2282
solvents in achieving high quality composite perovskite films. A Author Contributions
hysteresis study was performed, and the results are shown in S.P. designed the experiment, fabricated the devices, and made
Figure S7. No significant hysteresis was observed from the the corresponding characterization. S.P. and P.G. designed and
devices involving all the selected antisolvent, indicating trivial performed the kinetic studies. E.N.A. conducted and analyzed
influence of solvent types. the AFM data. G.G. and K.T.C. carried out the PL and SEM
Through this work, we have unraveled the kinetic procession characterization. P.G. and P.S. analyzed the XRD data. P.G.
of the solvent−solvent interaction during perovskite thin film wrote the initial draft of the manuscript. M.K.N. directed the
formation and thermodynamically induced crystal domain research for this work. All the authors contributed to the
transition and, in addition, presented a new effective antisolvent discussion and writing of the final paper.
to treat the perovskite film resulting in devices with PCEs above Notes
20%. The critical parameters that determine the choice of The authors declare no competing financial interest.


antisolvent are the boiling point, miscibility, and dielectric
constant. As shown in Figure S5, only miscible solvents with
ACKNOWLEDGMENTS
dielectric constants higher than 5 promoted the formation of
perovskite crystal during bulk solvent−solvent extraction. S.P. designed the experiment, fabricated the devices, and made
the corresponding characterization. S.P. and P.G. designed and
3. CONCLUSION performed the kinetic studies. E.N.A. and F.S. conducted and
analyzed the AFM data. G.G. and K.T.C. carried out the PL and
In summary, we have systematically studied the role of
SEM characterization. P.G. and P.S. analyzed the XRD data.
antisolvent treatment during composite perovskite film
Y.L. and Y.Z. helped with the device fabrication. P.G. wrote the
crystallization before and after thermal annealing by revealing
initial draft of the manuscript.M.K.N. directed the research for
the correlations between boiling point, miscibility, dielectric
this work. All the authors contributed to the discussion and
constant, perovskite grain orientation, microstructure, and
writing of the final paper.


nanostructure. We found that a suitable A.S.T. serves not
only to speed up heterogeneous nucleation in the early stages
of film deposition but also to influence the initial crystal growth,
REFERENCES
leading to a pinhole free homogeneous film morphology during (1) Lewis, N. S.; Nocera, D. G. Powering the Planet: Chemical
the thermal annealing. By using miscible antisolvents with high Challenges in Solar Energy Utilization. Proc. Natl. Acad. Sci. U. S. A.
boiling point (TFT, toluene, and chlorobenzene), we balanced 2006, 103, 15729−15735.
(2) Gao, P.; Grätzel, M.; Nazeeruddin, M. K. Advanced Concepts in
nucleation, film smoothness, and surface coverage, which in Photovoltaics; Nozik, A. J., Conibeer, G., Beard, M. C., Eds.; RSC
turn led to better device performance. TFT as a new antisolvent Energy and Environment Series; Royal Society of Chemistry:
candidate showed promising PCE of 20.3% under optimized Cambridge, 2014.
conditions. The study presented here demonstrates the criteria (3) Gao, P.; Grätzel, M.; Nazeeruddin, M. K. Organohalide Lead
for selection of a good antisolvent and the possibility of further Perovskites for Photovoltaic Applications. Energy Environ. Sci. 2014, 7,
increasing the quality of perovskite films by choosing a different 2448.
antisolvent, which is crucial for achieving higher efficiency (4) National Renewable Energy Laboratory Best Research-Cell
polycrystalline thin film solar cells. Efficiencies; http://www.nrel.gov/ncpv/images/efficiency_chart.jpg.


(5) Xue, Q.; Sun, C.; Hu, Z.; Huang, F.; Yip, H.-L.; Cao, Y. Recent
Advances in Perovskite Solar Cells: Morphology Control and
ASSOCIATED CONTENT Interfacial Engineering. Huaxue Xuebao 2015, 73, 179−192.
* Supporting Information
S (6) Huang, F.; Dkhissi, Y.; Huang, W.; Xiao, M.; Benesperi, I.;
The Supporting Information is available free of charge on the Rubanov, S.; Zhu, Y.; Lin, X.; Jiang, L.; Zhou, Y.; et al. Gas-Assisted
ACS Publications website at DOI: 10.1021/acs.chemma- Preparation of Lead Iodide Perovskite Films Consisting of a
ter.6b05353. Monolayer of Single Crystalline Grains for High Efficiency Planar
Solar Cells. Nano Energy 2014, 10, 10−18.
General methods, central area of the film, table of solvent (7) Zhao, Y.; Zhu, K. CH3NH3Cl-Assisted One-Step Solution
properties, top surface and cross sectional SEM images, Growth of CH 3NH3PbI3: Structure, Charge-Carrier Dynamics, and

3497 DOI: 10.1021/acs.chemmater.6b05353


Chem. Mater. 2017, 29, 3490−3498
Chemistry of Materials Article

Photovoltaic Properties of Perovskite Solar Cells. J. Phys. Chem. C (26) Wang, L.; Wang, K.; Zou, B. Pressure-Induced Structural and
2014, 118, 9412−9418. Optical Properties of Organometal Halide Perovskite-Based For-
(8) Zuo, C.; Ding, L. An 80.11% FF Record Achieved for Perovskite mamidinium Lead Bromide. J. Phys. Chem. Lett. 2016, 7, 2556−2562.
Solar Cells by Using the NH 4 Cl Additive. Nanoscale 2014, 6, 9935− (27) Baikie, T.; Fang, Y.; Kadro, J. M.; Schreyer, M.; Wei, F.;
9938. Mhaisalkar, S. G.; Gratzel, M.; White, T. J. Synthesis and Crystal
(9) Bi, D.; Gao, P.; Scopelliti, R.; Oveisi, E.; Luo, J.; Grätzel, M.; Chemistry of the Hybrid Perovskite (CH3NH3)PbI3 for Solid-State
Hagfeldt, A.; Nazeeruddin, M. K. High-Performance Perovskite Solar Sensitised Solar Cell Applications. J. Mater. Chem. A 2013, 1, 5628−
Cells with Enhanced Environmental Stability Based on Amphiphile- 5641.
Modified CH 3 NH 3 PbI 3. Adv. Mater. 2016, 28, 2910−2915. (28) Arakcheeva, A.; Chernyshov, D.; Spina, M.; Forró, L.; Horváth,
(10) Xiao, M.; Huang, F.; Huang, W.; Dkhissi, Y.; Zhu, Y.; Etheridge, E. CH 3 NH 3 PbI 3: Precise Structural Consequences of Water
J.; Gray-Weale, A.; Bach, U.; Cheng, Y.-B.; Spiccia, L. A Fast Absorption at Ambient Conditions. Acta Crystallogr., Sect. B: Struct.
Deposition-Crystallization Procedure for Highly Efficient Lead Iodide Sci., Cryst. Eng. Mater. 2016, 72, 716−722.
(29) Weber, O. J.; Charles, B.; Weller, M. T. Phase Behaviour and
Perovskite Thin-Film Solar Cells. Angew. Chem., Int. Ed. 2014, 53,
Composition in the Formamidinium−methylammonium Hybrid Lead
9898−9903.
Iodide Perovskite Solid Solution. J. Mater. Chem. A 2016, 4, 15375−
(11) Jeon, N. J.; Noh, J. H.; Kim, Y. C.; Yang, W. S.; Ryu, S.; Seok, S.,
15382.
Il. Solvent Engineering for High-Performance Inorganic−organic (30) http://www.topas-academic.net.
Hybrid Perovskite Solar Cells. Nat. Mater. 2014, 13, 897−903. (31) Xie, L.-Q.; Chen, L.; Nan, Z.-A.; Lin, H.-X.; Wang, T.; Zhan, D.;
(12) Xiao, Z.; Dong, Q.; Bi, C.; Shao, Y.; Yuan, Y.; Huang, J. Solvent Yan, J.-W.; Mao, B.-W.; Tian, Z.-Q. Understanding the Cubic Phase
Annealing of Perovskite-Induced Crystal Growth for Photovoltaic- Stabilization and Crystallization Kinetics in Mixed Cations and Halides
Device Efficiency Enhancement. Adv. Mater. 2014, 26, 6503−6509. Perovskite Single Crystals. J. Am. Chem. Soc. 2017, 139, 3320−3323.
(13) Lee, Y. H.; Luo, J.; Humphry-Baker, R.; Gao, P.; Grätzel, M.; (32) Rong, Y.; Tang, Z.; Zhao, Y.; Zhong, X.; Venkatesan, S.;
Nazeeruddin, M. K. Unraveling the Reasons for Efficiency Loss in Graham, H.; Patton, M.; Jing, Y.; Guloy, A. M.; Yao, Y. Solvent
Perovskite Solar Cells. Adv. Funct. Mater. 2015, 25, 3925−3933. Engineering towards Controlled Grain Growth in Perovskite Planar
(14) Xie, F. X.; Zhang, D.; Su, H.; Ren, X.; Wong, K. S.; Grätzel, M.; Heterojunction Solar Cells. Nanoscale 2015, 7, 10595−10599.
Choy, W. C. H. Vacuum-Assisted Thermal Annealing of CH 3 NH 3 (33) Cao, J.; Jing, X.; Yan, J.; Hu, C.; Chen, R.; Yin, J.; Li, J.; Zheng,
PbI 3 for Highly Stable and Efficient Perovskite Solar Cells. ACS Nano N. Identifying the Molecular Structures of Intermediates for
2015, 9, 639−646. Optimizing the Fabrication of High-Quality Perovskite Films. J. Am.
(15) Xiao, Z.; Yuan, Y.; Yang, B.; Vanderslice, J.; Chen, J.; Dyck, O.; Chem. Soc. 2016, 138, 9919−9926.
Duscher, G.; Huang, J. Universal Formation of Compositionally (34) Bi, D.; Xu, B.; Gao, P.; Sun, L.; Grätzel, M.; Hagfeldt, A. Facile
Graded Bulk Heterojunction for Efficiency Enhancement in Organic Synthesized Organic Hole Transporting Material for Perovskite Solar
Photovoltaics. Adv. Mater. 2014, 26, 3068−3075. Cell with Efficiency of 19.8%. Nano Energy 2016, 23, 138−144.
(16) Jeon, N. J.; Noh, J. H.; Yang, W. S.; Kim, Y. C.; Ryu, S.; Seo, J.;
Seok, S., Il. Compositional Engineering of Perovskite Materials for
High-Performance Solar Cells. Nature 2015, 517, 476−480.
(17) Yang, W. S.; Noh, J. H.; Jeon, N. J.; Kim, Y. C.; Ryu, S.; Seo, J.;
Seok, S., Il. High-Performance Photovoltaic Perovskite Layers
Fabricated through Intramolecular Exchange. Science (Washington,
DC, U. S.) 2015, 348, 1234−1237.
(18) Bi, D.; Tress, W.; Dar, M. I.; Gao, P.; Luo, J.; Renevier, C.;
Schenk, K.; Abate, A.; Giordano, F.; Correa Baena, J.-P.; et al. Efficient
Luminescent Solar Cells Based on Tailored Mixed-Cation Perovskites.
Sci. Adv. 2016, 2, e1501170−e1501170.
(19) Kim, Y. C.; Jeon, N. J.; Noh, J. H.; Yang, W. S.; Seo, J.; Yun, J.
S.; Ho-Baillie, A.; Huang, S.; Green, M. A.; Seidel, J.; et al. Beneficial
Effects of PbI 2 Incorporated in Organo-Lead Halide Perovskite Solar
Cells. Adv. Energy Mater. 2016, 6, 1502104.
(20) Son, D.-Y.; Lee, J.-W.; Choi, Y. J.; Jang, I.-H.; Lee, S.; Yoo, P. J.;
Shin, H.; Ahn, N.; Choi, M.; Kim, D.; et al. Self-Formed Grain
Boundary Healing Layer for Highly Efficient CH3 NH3 PbI3
Perovskite Solar Cells. Nat. Energy 2016, 1, 16081−16088.
(21) Zheng, X.; Chen, B.; Wu, C.; Priya, S. Room Temperature
Fabrication of CH3NH3PbBr3 by Anti-Solvent Assisted Crystalliza-
tion Approach for Perovskite Solar Cells with Fast Response and Small
J−V Hysteresis. Nano Energy 2015, 17, 269−278.
(22) See Experiment Section in Supporting Information.
(23) Han, Q.; Bae, S.-H.; Sun, P.; Hsieh, Y.-T.; Yang, Y. M.; Rim, Y.
S.; Zhao, H.; Chen, Q.; Shi, W.; Li, G.; et al. Single Crystal
Formamidinium Lead Iodide (FAPbI 3): Insight into the Structural,
Optical, and Electrical Properties. Adv. Mater. 2016, 28, 2253−2258.
(24) Stoumpos, C. C.; Malliakas, C. D.; Kanatzidis, M. G.
Semiconducting Tin and Lead Iodide Perovskites with Organic
Cations: Phase Transitions, High Mobilities, and Near-Infrared
Photoluminescent Properties. Inorg. Chem. 2013, 52, 9019−9038.
(25) Weller, M. T.; Weber, O. J.; Frost, J. M.; Walsh, A. Cubic
Perovskite Structure of Black Formamidinium Lead Iodide, α-
[HC(NH 2) 2 ]PbI 3, at 298 K. J. Phys. Chem. Lett. 2015, 6, 3209−
3212.

3498 DOI: 10.1021/acs.chemmater.6b05353


Chem. Mater. 2017, 29, 3490−3498

You might also like