Nanoarray heterojunction and its efficient solar cells without negative impact of photogenerated electric field

Efficient, stable and low-cost solar cells are being desired for the photovoltaic conversion of solar energy into electricity for sustainable energy production. Nanorod/nanowire arrays of narrow-bandgap semiconductors are the promising light-harvesters for photovoltaics because of their excellent optoelectrical properties. Here, the array of preferentially oriented antimony trisulfide (Sb2S3) single-crystalline nanorods is grown on polycrystalline titania (TiO2) film by a tiny-seed-assisted solution-processing strategy, offering an Sb2S3/TiO2 nanoarray heterojunction system on a large scale. It is demonstrated that the Sb2S3 nanorod growth follows a tiny-seed-governed orientation-competing-epitaxial nucleation/growth mechanism. Using a conjugated polymer hole transporting layer on the heterojunction, we achieve a power conversion efficiency of 5.70% in the stable hybrid solar cell with a preferred p-type/intrinsic/n-type architecture featuring effectively straightforward charge transport channels and no negative impact of photogenerated electric field on device performance. An architecture-dependent charge distribution model is proposed to understand the unique photovoltaic behavior. Photovoltaic devices require reliable and scalable growth methods to produce the constituent materials. Here, the authors report a tiny-seed-assisted solution processing strategy to grow Sb2S3/TiO2 nanoarray heterojunction of which the hybrid solar cell without negative impact of photogenerated electric field exhibits a power efficiency of 5.70%.

stability, non-toxicity, earth-abundant elemental composition, and easy preparation 12 . The Sb 2 S 3 -based single-junction solar cells are theoretically expected to achieve the efficiency of η = 27-33% according to Shockley-Queisser limit 13 . Sb 2 S 3 nanoparticle-sensitized solar cells have exhibited a record efficiency of η = 7.5% 14 . Moreover, the Sb 2 S 3 nanoparticle [15][16][17] or bulk 18,19 films as a photon-harvesting layer in planar heterojunction solar cells have achieved the efficiencies of 5.69−6.78% and 3.50−5.12%, respectively. Sb 2 S 3 crystals comprise the (Sb 4 S 6 ) n ribbons that grow along [001] direction via strong covalent Sb-S bonds and are packed together alone [100] and [010] directions by van der Waals forces (Inset to Fig. 1), and therefore Sb 2 S 3 often features a one-dimensional crystalline structure and highly anisotropic properties along different axes. While the one-dimensional nanostructures of Sb 2 S 3 nanorods/ nanowires have been prepared by different methods 20 , no reports on the growth of Sb 2 S 3 nanorod/nanowire arrays are available up to now.
In this work, we develop a tiny-seed-assisted repetition of spincoating and annealing (RSCA) strategy for preparing preferentially [211]-oriented and vertically aligned Sb 2 S 3 singlecrystalline nanorods on polycrystalline TiO 2 nanoparticle film, yielding an Sb 2 S 3 /TiO 2 -NHJ, for which an epitaxial nucleation/ growth mechanism gets revealed for the Sb 2 S 3 nanorod growth. After depositing a conjugated polymer layer onto the Sb 2 S 3 /TiO 2 -NHJ, a stable hybrid solar cell with p-i-n architecture is obtained and exhibits a considerable efficiency of 5.70%. It is found that, furthermore, there is no negative impact of photogenerated electric field (E ph ) on photovoltaic performance in such p-i-n solar cells, and an electric field distribution model is suggested for understanding the distinguishing photovoltaic feature.
Noticeably, we previously developed a seed-assisted RSCA method for preparing the monolayer of [221]-oriented bulk Sb 2 S 3 single-crystalline cuboids on polycrystalline TiO 2 nanoparticle film 19 , where [211]/[221]-oriented Sb 2 S 3 crystalline seeds laterally of 60−400 nm in size were deposited on the TiO 2 film prior to the RSCA cycles for Sb 2 S 3 crystal growth, and an orientationcompeting-epitaxial (OCE) nucleation/growth mechanism was proposed for the [221]-oriented Sb 2 S 3 cuboid growth. In this work, we use the tiny [211]/[221]-oriented Sb 2 S 3 crystalline seeds of laterally <10 nm in size to perform the seed-assisted RSCA process and obtain the [211]-oriented Sb 2 S 3 single-crystalline nanorod array on the polycrystalline TiO 2 nanoparticle film. Our results demonstrate that the Sb 2 S 3 nanorod growth follows a tinyseed-governed OCE nucleation/growth mechanism, where the tiny size of Sb 2 S 3 seeds is exclusively important to govern the orientation and morphology of resulting Sb 2 S 3 nanorods. Undoubtedly, in regard to the heterojunction preparation, the insights in this work make significant advances in expanding the knowledge on the seed-assisted RSCA method and the OCE nucleation/growth mechanism.

Results
Preparation strategy. For simplicity, the Sb 2 S 3 precursor solutions for Sb 2 S 3 seed growth (SG) and nanorod growth (NG) are referred to as SG and NG solutions, respectively. The SG and NG solutions have the same composition and their concentrations (C sg and C ng ) are defined by the molecular concentration of SbCl 3 therein. For the growth of the Sb 2 S 3 nanorod array, C ng must be much higher than C sg , as indicated by solution color (Bottles 1 and 2, inset to Fig. 1). As shown in Fig. 1, an Sb 2 S 3 precursor film is spin-coated from an NG solution onto the TiO 2 nanoparticle film decorated with oriented Sb 2 S 3 crystalline seeds and then annealed in N 2 atmosphere at 300°C, which is referred to as one RSCA cycle number (n); the Sb 2 S 3 /TiO 2 -NHJ is produced after a certain RSCA cycle numbers (i.e., n ≥ 1). By coating a poly [4,8-bis(5-(2ethylhexyl) thiophen-2-yl)benzo[1,2-b;4,5-b']dithiophene-2,6diyl-alt-(4-(2-ethylhexyl)-3-fluorothieno [3,4-b]thiophene-)-2-carboxylate-2-6-diyl)] (PTB7-th, inset to Fig. 1) layer, we obtain a p-i-n hybrid solar cell with the spatially separated intrinsic Sb 2 S 3 nanorods sandwiched between p-type PTB7-th hole transporting layer (HTL) and n-type TiO 2 electron transporting layer (ETL), and scanning electron microscopy (SEM) image (Inset to Fig. 1) clearly confirms the p-i-n architecture.
Characterization of nanorod array. SEM images (Fig. 2a, b) show that a large-area Sb 2 S 3 layer is grown on the TiO 2 film and consists of the Sb 2 S 3 nanorods that are of a right square prism shape, 50−150 nm in width, ca. 320 nm in length, and a number density (i.e., the number of the different nanorods per unit area) of ca. 36 µm −2 on the TiO 2 nanoparticle (ca. 10−20 nm in diameter) film of ca. 110 nm in thickness. Clearly, the Sb 2 S 3 nanorods are grown along the directions out of substrate planes (i.e., out-of-plane growth) with the growth direction tilting some angles from the normal of TiO 2 film, resulting in an Sb 2 S 3 nanorod array on the TiO 2 film. Fig. 2c shows the transmission electron microscopy (TEM) image of a single Sb 2 S 3 nanorod of ca. 50 nm in width. Both selected area electron diffraction pattern and high-resolution TEM (HRTEM) image (Insets to Fig. 2c Fig. 2d). Moreover, the TiO 2 film on fluorine-doped tin oxide (FTO) substrate exhibits the strongest XRD peak of (101) plane of anatase TiO 2 (101) planes (Fig. 2d), inferring that the (101) crystal planes of TiO 2 nanoparticles exposed on the film surface orient in the direction parallel to substrate plane 19,21 .
Growth mechanism of nanorod array. When Sb 2 S 3 seeds are absent from the TiO 2 film, only the large and irregular Sb 2 S 3 grains without a preferential orientation are obtained (Supplementary Fig. 1). Clearly, the Sb 2 S 3 crystalline seeds are very crucial for the growth of the Sb 2 S 3 nanorod array. Our results show that the TiO 2 film surface undoubtedly is decorated by the preferentially [211]/[221]-oriented Sb 2 S 3 crystalline seeds that are of a C sg -dependent size in the range of C sg = 0.05−0.4 M (Supplementary Fig. 2  In order to get insights into the growth mechanism for the Sb 2 S 3 nanorod array on TiO 2 film, the dependences of the Sb 2 S 3 nanorod growth on the conditions (C sg , C ng , and n) for the seedassisted RSCA method are investigated. We find that the formation of [211]-oriented Sb 2 S 3 nanorod array depends exclusively on the C sg concentration, and the tiny [211]/[221]oriented Sb 2 S 3 seeds derived from a dilute SG solution of C sg ≤ 0.1 M are indispensable (Supplementary Fig. 3 and Supplementary Note 2); moreover, both C ng and n in the ranges of C ng < 1.1 M and n < 3 imposes no remarkable influences on the lateral size of Sb 2 S 3 nanorods, but the lateral size of Sb 2 S 3 nanorods gets dramatically increased to make an exception when both C ng = 1.1 M and n = 3 (Supplementary Fig. 4 and Supplementary Note 3). Here, we mainly show the length and number density of Sb 2 S 3 nanorods dependent on C ng and n values. The increase in Fig. 1 Tiny-seed assisted repetition of spin-coating and annealing (RSCA) method, structures of materials, and device. The schematic illustration shows the preparation of the Sb 2 S 3 /TiO 2 -nanoarray heterojunction (Sb 2 S 3 /TiO 2 -NHJ) by a tiny-seed-assisted RSCA method on fluorine-doped tin oxide (FTO) substrate and the structure of p-i-n solar cell, where p means p-type PTB7-th, i means intrinsic Sb 2 S 3 and n means n-type TiO 2 . The insets show the photographs of seed growth (bottle 1, colorless) and nanorod growth (bottle 2, yellow) solutions, the molecular structures of PTB7-th, the crystal structure of orthorhombic Sb 2 S 3 , and the cross-sectional scanning electron microscopy (SEM) characterization of a practical p-i-n cell. Normally, the seed growth solution concentration of C sg = 0.1 M and the nanorod growth solution concentration of C ng = 0.8 M were, respectively, used for growing the Sb 2 S 3 seeds prior to the RSCA cycles and the Sb 2 S 3 nanorods in the RSCA cycles, unless otherwise specified. During the RSCA method, each of the precursor film formations by spin-coating followed by once annealing is referred to as one RSCA cycle for the nanorod growth, and the nanorod length (i.e., nanorod array thickness) is mainly controlled by the RSCA cycle number (n).
either C ng or n leads to the Sb 2 S 3 nanorod growth along c-axis (i.e., a larger array thickness) (Fig. 3a). However, for C ng < 1.1 M and n < 3, the number density is mainly affected by C ng , and an increased C ng results in a significantly reduced number density (Fig. 3b). Actually, the number density of Sb 2 S 3 nanorods in most cases is not remarkably changed with n value, strongly indicating that the Sb 2 S 3 nanorods mainly take a self-nucleated growth similar to the Sb 2 S 3 cuboid growth 19 .
The growth of Sb 2 Se 3 nanorod arrays on selenized Mo substrate by sublimation technique was proposed to follow a split growth model on the basis of the morphology transition from thin film to nanorod array when the lateral stress of the Sb 2 Se 3 grains was beyond the tolerance of the van der Waals forces between (Sb 4 Se 6 ) n ribbons during Sb 2 Se 3 vapor evaporates 22 . Besides, quasi-epitaxial growth and dual-selection growth models were proposed, respectively, to elucidate the formation of [hk1]-oriented Sb 2 S 3 compact films on TiO 2 23 and CdS 24 nanoparticle films. Unfortunately, there is no growth mechanism for the formation of Sb 2 S 3 single-crystalline nanorods on polycrystalline surfaces. We previously proposed an OCE nucleation/growth mechanism to describe the growth of the Sb 2 S 3 cuboids on polycrystalline TiO 2 film 19 . In the OCE nucleation/growth mechanism, the epitaxial relationship of Sb 2 S 3 (101)//TiO 2 (101) exists for the heterogeneous nucleation of orthorhombic Sb 2 S 3 crystal on favorably oriented anatase TiO 2 nanoparticle surface as the result of the lattice match between Sb 2 S 3 (101) and TiO 2   and Supplementary Note 5), as schematically illustrated in Fig. 3c. First, to form the [211]/[221]-oriented Sb 2 S 3 seeds, the heterogeneously epitaxial nucleation of orthorhombic Sb 2 S 3 crystals occurs on favorably oriented anatase TiO 2 nanoparticles, followed by the self-nucleated growth of the Sb 2 S 3 nuclei into the Sb 2 S 3 tiny crystalline seeds when the orientations of the Sb 2 S 3 nuclei and the resulting seeds are governed by TiO 2 (101) plane orientations. Since the TiO 2 nanoparticles are relatively much larger (10−20 nm in diameter), each of the tiny Sb 2 S 3 seeds (laterally <10 nm in size) has a high possibility to be formed on an individual TiO 2 nanoparticle. Those tiny Sb 2 S 3 seeds inevitably have more chances to nucleate [211]-oriented Sb 2 S 3 crystals, because more TiO 2 nanoparticles tending to orient with (101) face parallel to the film plane ( Fig. 2d) inevitably result in more TiO 2 nanoparticle orientations in the range of 0 o ≤ α < 45 o (Supplementary Table 1, Supplementary Fig. 6 and Supplementary Note 5). During the first RSCA process (i.e., n = 1), the Sb 2 S 3 seeds on the TiO 2 nanoparticles oriented with α ≈ 22 o (Particle 2) have the fastest rate to take a self-nucleated growth into laterally enlarged Sb 2 S 3 nanorods without changing the seed orientations, where the growth rate in c-axis direction is much larger than in aaxis and b-axis directions, but the Sb 2 S 3 crystal growths at around sites with the less-competitive/unfavorable orientations (Particles 1 and 3) are merged or suppressed (Inset to Fig. 3c). Furthermore, with increasing the RSCA cycle numbers (n > 1), the growing nanorods still take a self-nucleated growth mainly along c-axis direction into longer ones without remarkably increasing lateral size (Supplementary Fig. 4 and Supplementary Note 3) nor changing the orientation ( Supplementary Fig. 7). During the nanorod growth, the sites, where there are no seeds and/or suppressed/merged crystal growths, create the empty interspaces in between nanorods, leading to the array of [211]-oriented Sb 2 S 3 nanorods on the polycrystalline TiO 2 nanoparticle film. On the other hand, the tiny Sb 2 S 3 seeds have a weak capability to merge and suppress the far-off nucleation at around sites due to their small size, leading to the number density (ca. 36 µm −2 ) of Sb 2 S 3 nanorods much higher than that (ca. 3 µm −2 ) of the Sb 2 S 3 cuboids derived from large Sb 2 S 3 seeds (Supplementary Figs. 3, 5 and Supplementary Notes 2, 4).
Solar cell architecture and working principle. The Sb 2 S 3 nanorod array in the Sb 2 S 3 /TiO 2 -NHJ shows an absorption onset at about 737 nm corresponding to the optical band gap (E g ) of 1.68 eV (Fig. 4a), very close to the E g value of Sb 2 S 3 bulk crystal 25 . With the valence band level (E VB ) of E VB = −5.29 eV for the Sb 2 S 3 bulk crystals 25 , the conduction band level (E CB ) of the Sb 2 S 3 nanorod array is calculated to be −3.61 eV from the E g value; moreover, the PTB7-th exhibits the absorption mainly within 550 −750 nm with an absorption onset at ca. 772 nm, corresponding to E g = 1.61 eV, the highest occupied molecular orbital (HOMO) energy level of −5.20 eV and the lowest unoccupied molecular orbital (LUMO) energy level of −3.59 eV 26 . With the E CB = −4.20 eV for TiO 2 27 , the E CB = −2.30 eV and E VB = −5.30 eV for MoO 3 28 and the work-functions (W f ) of −4.40 eV for FTO and −5.10 eV for Au 29 , the combination of Sb 2 S 3 /TiO 2 -NHJ and PTB7-th renders a ternary materials system, where TiO 2 , Sb 2 S 3 and PTB7-th components form type II hybrid heterojunctions with staggered band alignments to facilitate the charge transport in solar cell (Fig. 4b). The Sb 2 S 3 nanorod has a faceted top of (001) plane and the faceted sides of (100) and (010) planes (Fig. 2c). Density functional theory (DFT) calculations reveal that the S atoms of PTB7-th interact weakly with the Sb atoms on Sb 2 S 3 (001) and (100) planes, but no evident interfacial electric field is formed at the Sb 2 S 3 /PTB7-th interfaces (Fig. 4c); moreover, the Sb 2 S 3 /PTB7-th interfaces formed by other Sb 2 S 3 planes (e.g., (010) and (110) ones) have the same interfacial interaction features ( Supplementary Fig. 8). The wavefunctions of electronic states near the conduction band minimum (CBM) and valence band maximum (VBM) of the Sb 2 S 3 /PTB7-th interfaces formed by Sb 2 S 3 (100) plane (Fig. 4d), and Sb 2 S 3 (001), (010), and (110) planes as well ( Supplementary Fig. 8), are also calculated using DFT. It is found that the CBM wavefunction only locates in the Sb 2 S 3 material, but the VBM wavefunction predominantly distributes within the PTB7-th phase and slightly overlaps the Sb 2 S 3 material, which is similar to the well-known efficient poly(3hexylthiopene)/[6,6]-phenyl-C61-butyric acid methyl ester (i.e., P3HT/PCBM) organic photovoltaic system 30 . Those unique separate electronic states inevitably favor the dissociation and injection of photogenerated electron-hole pairs and free charge carriers at the Sb 2 S 3 /PTB7-th interface upon the excitation of Sb 2 S 3 and PTB7-th, with the photogenerated electrons remaining in or being injected into Sb 2 S 3 crystal and the photogenerated holes being injected into or remaining in PTB7-th phase.
The p-i-n solar cells ( Fig. 1) are fabricated by spin-coating PTB7-th onto the Sb 2 S 3 /TiO 2 -NHJ. Optimizing experiments demonstrate that the p-i-n solar cell delivers a peak efficiency when the Sb 2 S 3 /TiO 2 -NHJ is derived either at C sg = 0.1 M (Supplementary Table 2) or at C ng = 0.8 M and n = 2 (Supplementary Table 3). In the following sections, therefore, we focus on the optimized Sb 2 S 3 /TiO 2 -NHJ with a length of 320 nm prepared at C sg = 0.1 M, C ng = 0.8 M, and n = 2, to investigate the architecturedependent performance of solar cells. In order to reveal the photovoltaic characteristics of the p-i-n architecture, we also fabricate the p-(p + i)-n counterpart solar cells for comparison, which features the polymer infiltration into the interspaces between nanorods resulting in the (p + i) blend of intrinsic Sb 2 S 3 nanorods and p-type polymer sandwiched between p-type HTL and n-type ETL (Fig. 4e). To further visualize the PTB7-th infiltration into the Sb 2 S 3 nanorod array, the SEM image with a larger view field is shown in Supplementary Fig. 9. As illustrated by Fig. 4b, in both p-i-n and p-(p + i)-n devices, free charge carriers are generated in the Sb 2 S 3 nanorods upon illumination as the result of the small exciton binding energy (E b ) in Sb 2 S 3 (E b < 10 meV) that is much smaller than the thermal energy at ambient temperature (k B T ∼ 26 meV, where k B is Boltzmann constant and T is temperature) 31 ; however, the excitons generated in the PTB7-th phase upon into the conduction band for transport. c Density functional theory (DFT) calculated interfacial interaction at the Sb 2 S 3 /PTB7-th interfaces formed by PTB7-th on Sb 2 S 3 (001) plane (above) and Sb 2 S 3 (100) plan (below), where red color denotes charge accumulation and green color stands for charge depletion, and the electron density isosurface is set to 0.0004 e Å −3 . d DFT calculated wavefunctions of electronic states near the conduction band minimum (CBM) (above) and valence band maximum (VBM) (below) of Sb 2 S 3 (100)/PTB7-th interface system, where the isosurface of wavefunction square is set to 4 × 10 −11 e Å −3 . e Structure of p-(p + i)-n solar cell (up-left: cross-sectional scanning electron microscopy image of a practical device; below-right: schematic illustration), where p means p-type PTB7-th, i means intrinsic Sb 2 S 3 and n means n-type TiO 2 . f Current-voltage curves under AM1.5 G illumination, (g) incident photon-to-current conversion efficiency (IPCE) spectra (the right-hand axis shows the integrated photocurrent from the IPCE spectra) and (h) electrochemical impedance spectroscopic spectra (scatter: experimental data measured in the dark; solid line: fitting data) of the p-i-n and p-(p + i)-n solar cells with the highest efficiency. Inset to (h) shows the equivalent circuit used for fitting, which consists of series resistance (R s ), a hole transport and extraction resistance (R t ) at selective electrode, a recombination resistance (R rec ), and two constant phase angle elements (CPE).
illumination need to diffuse to the Sb 2 S 3 /PTB7-th interface for dissociation into free electrons (e) and holes (h), with the electrons being injected into Sb 2 S 3 and the holes remaining in the polymer, because the E b in conjugated polymers is normally about 0.4 −0.5 eV 32,33 , much higher than k B T~26 meV. Due to the high electron (μ e ≈ 10 cm 2 V −1 s −1 ) and hole (μ h ≈ 2.6 cm 2 V −1 s −1 ) mobilities in the Sb 2 S 3 crystals 19 , the electrons generated in the Sb 2 S 3 nanorods are quickly transported into the TiO 2 layer under the built-in electric field (E bi ) due to the W f difference between the FTO and Au asymmetric collection electrodes 34,35 , while the holes generated in the Sb 2 S 3 nanorods are injected into the PTB7-th layer on nanorod top in the p-i-n device or the PTB7-th layers on nanorod top and amongst nanorods in the p-(p + i)-n device. As the result of quite a low hole mobility in the PTB7-th (μ h = 2.8 × 10 −4 cm 2 V −1 s −1 ) 36 , the transport of holes in the polymer phase proceeds very slowly and they finally tunnel through the MoO 3 layer acting as an optical spacer/Au diffusion barrier 37 to reach the Au electrode (Fig. 4b).
Architecture-dependent photovoltaic performance. The current-voltage (J−V) curves for the best p-i-n and p-(p + i)-n solar cells are presented in Fig. 4f, and the averaged performance of seven independent devices for each sample are summarized in Table 1. We achieve a champion efficiency of η = 5.70% in the p-i-n solar cells, which is comparable to those of the planar heterojunction solar cells based on the Sb 2 S 3 nanoparticle films prepared by atomic layer deposition (e.g., η = 5.77%) 15 and solution-processing method (e.g., η = 5.69%) 17 . Furthermore, the p-i-n solar cell exhibits good stability and retains 88% of the initial efficiency after storage in the N 2 atmosphere for 31 days (Supplementary Fig. 10). Generally, the p-i-n solar cells are much more efficient than the p-(p + i)-n ones. The averaged fill factor (FF) of the p-i-n devices is ca. 54%, much higher than that (ca. 38%) of the p-(p + i)-n ones, and the increased FF in the p-i-n devices is accompanied by a larger open-circuit voltage (V oc ), a higher short-circuit current (J sc ), a smaller series resistance (R s ), and a larger shunt resistance (R sh ) ( Table 1). Moreover, incident photon-to-current conversion efficiency (IPCE) spectra show that both p-i-n and p-(p + i)-n solar cells have a similar and broad spectral profile in 300−750 nm, and the integrated J sc from the IPCE data (i.e., 14.78 mA cm −2 for p-i-n device and 12.51 mA cm −2 for p-(p + i)-n) one) agree with the observed J sc values (Fig. 4g). While the polymer amount accumulated in the p-(p + i)-n architecture is higher than that in the p-i-n one, the IPCE of the former device in the whole absorption range is significantly lower than that of the latter one, inferring that much more photogenerated charge carriers contribute to photocurrent generation in the p-i-n device.
The electrochemical impedance spectroscopy (EIS) spectra in the form of Nyquist plots for the p-i-n and p-(p + i)-n solar cells measured in the dark are shown in Fig. 4h. The EIS data are fitted with a transmission line model 38 as depicted by a simplified equivalent circuit (Inset to Fig. 4f), and the fitting results are summarized in Supplementary Table 4. On the Nyquist plots, the small arc at high frequency is for the hole transport and extraction resistance (R t ) at the selective electrode, while the main arc at low frequency is attributed to recombination resistance (R rec ) in the devices 39 . The p-i-n solar cell has R t = 32.08 Ω cm 2 , much smaller than that (R t = 74.58 Ω cm 2 ) of the p-(p + i)-n device, inferring that the p-i-n architecture has a much easier hole the transport and extraction at the MoO 3 /Au cathode. Moreover, the R rec value (204.66 Ω cm 2 ) in the p-(p + i)-n device is lower than that (260.10 Ω cm 2 ) in the p-i-n one, suggesting much more serious charge recombination in the p-(p + i)-n device. Furthermore, the series resistance (R s ) in the p-(p + i)-n Table 1 Architecture-dependent photovoltaic performance of the p-i-n and p-(p The Sb2S3 nanorod arrays are 320 nm in length and prepared at the seed growth solution concentration of The averaged data with standard deviations as errors are the statistic results of seven individual devices. c The data for the device with the highest efficiency. device (4.42 Ω cm 2 ) is also much higher than that (1.73 Ω cm 2 ) in the p-i-n device, in agreement with the change in the R s data measured by J−V characterization (Table 1).
Photogenerated electric field effects on photovoltaic performance. Obviously, the p-i-n and p-(p + i)-n hybrid solar cells exhibit a significantly architecture-dependent performance. A similar phenomenon has also been observed in the hybrid solar cells based on CdS nanorod array and conjugated polymer MEH-PPV, that is, a partial infiltration of MEH-PPV into CdS nanorod array results in better device performance, but for which the reason is not clear yet 40 . Since materials in the p-i-n and p-(p + i)-n solar cells are the same, the observed different photovoltaic performances of them, in particular, the change in V oc , must be governed by an architecture-dependent factor other than the materials property (e.g., band energy level difference).
Here, an architecture-dependent charge distribution (ADCD) model is proposed on the basis of light intensity attenuation in the polymer phase to describe the photovoltaic behaviors in the p-i-n and p-(p + i)-n devices, as schematically illustrated in Fig. 5a. In the p-(p + i)-n solar cells, the polymer phase in between the interspaces of Sb 2 S 3 nanorods in the Sb 2 S 3 /TiO 2 -NHJ has a thickness comparable to the Sb 2 S 3 nanorod length (i.e., 320 nm) and a remarkable absorption contribution to the generation of charge carriers. Actually, the light intensity (I) is exponentially attenuated when passing a polymer film according to the relation of I = I 0 exp(−αx), where I 0 is the incident light intensity, I is the intensity at the illumination depth of x nm and α is the absorption coefficient of PTB7-th. Given α = 10 5 cm −1 for PTB7-th, the illumination intensity at the illumination depth of 320 nm is attenuated to 4% of the incident intensity (Fig. 5b). As the result of the light intensity attenuation, therefore, in the p-(p + i)-n solar cells, there is an original photogenerated hole density gradient from a higher density at the stem region of Sb 2 S 3 nanorods to a lower one at the nanorod top region, after the photogenerated excitons in PTB7-th phase dissociate at the nanorod-side-faceformed Sb 2 S 3 /PTB7-th interface with electrons being injected into the Sb 2 S 3 nanorods. However, the conductivity of a conjugated polymer in pristine form is very low, but it increases under the photo-doped state due to illumination. Hence, subjected to the light intensity attenuation, the conductivity in the PTB7-th film inevitably becomes gradually reduced along the illumination depth direction. Consequently, during the photovoltaic process, with the diffusion of the photogenerated holes in between the interspaces amongst the Sb 2 S 3 nanorods towards the MoO 3 /Au collection electrode, a reversed hole density gradient is formed from a higher hole density at the nanorod top region to a lower one at the Sb 2 S 3 nanorod stem region, to render a photogenerated electric field (E ph ) opposite to the E bi in p-(p + i)-n architecture (right, Fig. 5a). In contrast, the p-i-n solar cells are actually an E ph -free device (left, Fig. 5a), due to the absence of the polymer infiltration into the interspaces in between the Sb 2 S 3 nanorods and thereby no photogenerated hole accumulation around each of Sb 2 S 3 nanorods.
We performed the finite element simulation of the bias electrostatic filed (E v ) distribution in the p-i-n and p-(i + p)-n solar cells at a forward bias of 1 V, to mimic the case of selectively accumulated charge carriers in the devices, using the AC/DC module on commercial COMSOL Multiphysics software. The structural models for simulation (Supplementary Table 5) were built by assuming that all the nanorods of a right square prism shape are vertically grown on the substrate and the materials dimensions are given by referring to the SEM results (Figs. 1 and  4e). All the parameters of materials property and dimensions for simulation are provided in Supplementary Table 5. In the p-i-n solar cell, where the FTO/TiO 2 and MoO 3 /Au layers are defined, respectively, as substrate and electrode for a simplified simulation, the highest electrostatic potential dominantly at the electrode gets gradually reduced along normal of HTL (i.e., PTB7-th film) until the lowest potential at the top of Sb 2 S 3 nanorods, suggesting the presence of a very narrow bias E v due to the forward bias in the HTL to drive electrons to move towards the electrode (Fig. 5c). As for the p-(p + i)-n solar cell, however, the highest electrostatic potential at the electrode gradually decreases along the axis of Sb 2 S 3 nanorods until the lowest potential at the substrate, informing a bias E v across the whole device to drive electrons to reach the electrode (Fig. 5d). Clearly, the simulations inform that, as the results of the hole accumulation near the nanorod top region and the electron accumulation in the nanorod stem region, an electric field (e.g., E v ) across the whole device and opposite to the built-in field E bi is theoretically expected to reduce the collection efficiency of electrons photogenerated inside both Sb 2 S 3 nanorods and PTB7-th phase in the p-(p + i)-n solar cells, while only an electric field across the HTL negatively affects the very limited population of electrons at the top of Sb 2 S 3 nanorods and in the HTL in the p-i-n devices. The finite element simulations theoretically agree with the expectations from the ADCD model (Fig. 5a). Interestingly, further simulated results inform that once the PTB7-th infiltration into the Sb 2 S 3 nanorod array is present either partially or completely, a similar electric field (e.g., E v ) distribution across the whole device and opposite to the built-in field E bi is theoretically expected (Supplementary Fig. 11 and Supplementary Note 6).
The ADCD model in Fig. 5a is also experimentally supported by the capacitance-voltage (C−V) measurements of solar cells in the dark. The C−V results (Fig. 5e) clearly show that the p-(i + p)-n device has a much higher capacitance (C) than the p-i-n one in the whole bias voltage (V) range, indicating a higher capability of being charged by injected charge carriers in the former device. The relation C −2 ∝ V in Mott-Schottky (MS) equation expresses the relationship amongst the C, built-in potential (V bi ), and V of the solar cell, and the intersection point of the linear slope with voltage axis on the C −2 −V plot stands for the V bi across the device 41,42 . The V bi of the p-i-n device is 0.68 V (Inset to Fig. 5e), comparable to the maximum theoretical V bi (0.70 eV) in the solar cell due to the W f difference between Au (−5.10 eV) and FTO (−4.40 eV) electrodes 29 . Reasonably, the V bi obtained from the C −2 −V plots originates from the W f offset of the asymmetric electrodes 43 . However, the p-(p + i)-n device exhibits a V bi of 0.57 V, much lower than the maximum theoretical V bi value. The V bi values of the p-i-n and p-(p + i)-n devices are almost identical to their V oc data from J−V measurements ( Table 1).
The smaller V bi accompanied by a higher capacitance at V < V bi in the p-(i + p)-n cell suggests that a negative impact on the V bi results from the charge accumulation therein 44,45 . Since there is a greatly unbalance amongst the hole mobility in PTB7-th phase (μ h = 2.8 × 10 −4 cm 2 V −1 s −1 ) and the electron mobilities in TiO 2 (μ e ≈ 1.98 × 10 −3 cm 2 V −1 s −1 ) 25 and Sb 2 S 3 (μ e ≈ 10 cm 2 V −1 s −1 ), the charge accumulation predominantly correlates with the lowmobility holes in the polymer phase in between Sb 2 S 3 nanorods under forwarding bias 46 , resulting in a distribution state of a high hole density toward the MoO 3 /Au cathode and a high electron density in the TiO 2 layer. Clearly, the C−V measurements reveal that an opposite electric field opposite to the E bi across the device is generated due to the hole accumulation in the polymer phase in between the Sb 2 S 3 nanorods with a distribution of a high density toward the MoO 3 /Au cathode to impose a significant negative impact on the device V oc in the p-(i + p)-n solar cells, but it is not the case in the p-i-n devices, in agreement with the expectations from the ADCD model.
Both the finite element simulation and C−V data demonstrate that the E ph negative effect on charge transport is much more noteworthy in p-(p + i)-n devices than in p-i-n ones. From the ADCD model (Fig. 5a), the following points can be expected: (i) the p-i-n device only have nanorod-top-face-formed Sb 2 S 3 / PTB7-th interface, but the p-(p + i)-n device has additionally a much larger area of nanorod-side-face-formed Sb 2 S 3 /PTB7-th interface besides the nanorod-top-face-formed one and a heavy photogenerated hole accumulation around each of nanorods, hence, the p-(p + i)-n device should have a much shorter electron lifetime (τ e1 ) due to the charge recombination at the Sb 2 S 3 /PTB7th interface; (ii) as the TiO 2 /Sb 2 S 3 interface in the p-i-n and p-(p + i)-n solar cells are the same, the electron lifetime (τ e2 ) subjected to the charge recombination at TiO 2 /Sb 2 S 3 interface in two solar cells should be comparable to each other, but, a more serious charge recombination at the TiO 2 /Sb 2 S 3 interface for a shorter τ e2 should practically exist in the p-(p + i)-n device because the photogenerated holes in Sb 2 S 3 nanorods do not easily escape from the TiO 2 /Sb 2 S 3 interface under the E ph effect; (iii) the transit time (τ D ) for the photogenerated electrons accommodating in the TiO 2 layer to reach the FTO collection electrode in the p-(p + i)-n device should be much longer than that in the p-i-n one, because the electron diffusion rate greatly depends on the electron concentration in conduction band 47 , and the electron density injected into the TiO 2 layer in the p-(p + i)-n device is much lower due to the serious loss by interfacial charge recombination at nanorod-side-face-formed Sb 2 S 3 /PTB7-th interface. Those expectations are confirmed by dynamic spectroscopic investigates of solar cells (Supplementary Fig. 12 and Supplementary Note 7). That is, the τ e1 = 31.53 ms of the p-(p + i)-n device is much shorter than the τ e1 = 895.44 ms in the p-i-n device; both cells have comparable τ e2 values, but the τ e2 = 38.74 μs for p-(p + i)-n cell is shorter than the τ e2 = 61.22 μs for p-i-n cell; moreover, the τ D = 19.55 μs in p-(p + i)-n device is much longer than τ D = 6.25 μs in the p-i-n device (Table 1). Surprisingly, the p-i-n device has a very long electron lifetime τ e1 related to the charge recombination at Sb 2 S 3 /PTB7-th interface, which is attributed to a very low electron population at the top of Sb 2 S 3 nanorods in such device as the result of the high electron transport in the Sb 2 S 3 nanorods toward FTO collection electrode. Moreover, the very low electron population in the Sb 2 S 3 nanorod top region inevitably results in a limited negative impact of the E ph across the HTL on the transport of photogenerated electrons in the p-i-n device, as indicated by the finite element simulation (Fig. 5c).
Advantages of the p-i-n solar cells. The observed J−V characteristic difference between the p-i-n and p-(p + i)-n devices ( Fig. 4f and Table 1) gets well accounted for by the proposed ADCD mode (Fig. 5a). First, after ruling out the band energy level influence on the device V oc , it is reasonable to believe that the V oc of the Sb 2 S 3 /TiO 2 -NHJ solar cells strongly correlates with the net built-in field in device 34,35 . Since the p-i-n solar cell under illumination has no negative E ph effects as the result of lacking the photogenerated hole accumulation around Sb 2 S 3 nanorods, the device has a higher V oc resulting from the E bi across the device due to the asymmetric W f values between Au and FTO electrodes, while there are the negative E ph effects to weaken the device E bi under illumination for a reduced V oc in the p-(i + p)-n device. Moreover, a larger nanorod-side-face-formed Sb 2 S 3 / PTB7-th interface area and the negative E ph effects on charge transport are existing in the p-(i + p)-n device, which deteriorate the charge collection efficiency to reduce reduced J sc and FF by causing a great electrical coupling between the electrons inside Sb 2 S 3 nanorods and the holes in PTB7-th phase with a small R sh for serious interfacial charge recombination at the Sb 2 S 3 /PTB7-th interface and a high R s for a retarded charge transport.
Obviously, the p-i-n solar cells based on Sb 2 S 3 /TiO 2 -NHJ have at least two unique advantages: one is that no negative E ph effects on device performance thanks to lacking the photogenerated hole accumulation around Sb 2 S 3 nanorods; another is the [211]-oriented single-crystalline Sb 2 S 3 nanorods therein provide the straightforward and highly effective charge transport channels formed by the covalently bounded (Sb 4 S 6 ) n ribbons tilting ca. 53°on the substrate for the photogenerated charge carriers inside the Sb 2 S 3 layer to reach collection electrode. Indeed, those advantages make the Sb 2 S 3 / TiO 2 -NHJ have a good application potential to efficient solar cells. However, while the efficiency (η = 5.70%) of our p-i-n solar cells is comparable to the hybrid planar heterojunction solar cell based on the Sb 2 S 3 nanoparticle film (i.e., FTO/TiO 2 /Sb 2 S 3 /P3HT/Au (η = 5.77%) 15 , but it is still much lower than those of all-inorganic solar cells of the Sb 2 Se 3 nanorod array on Mo-substrate (i.e., Mo/MoSe 2 / Sb 2 Se 3 /CdS/ZnO/ZnO:Al/Ag, η = 9.2%) 22 and the InP p-n junction nanowire array on InP substrate (i.e., InP/InP(p-n)/ITO, η = 17.8%) 9 . For a simple preparation of an HTL over the Sb 2 S 3 /TiO 2 -NHJ without infiltration into the interspaces in-between nanorods, we chose the good-performing polymer PTB7-th to prepare organic HTL. The not high efficiency in our p-i-n solar cells reasonably originates from the low hole mobility in PTB7-th and the week interaction between the PTB7-th and Sb 2 S 3 as indicated by DFT calculation (Fig. 4c) for a not very efficient Sb 2 S 3 /PTB7-th interface for charge separation and injection 48 . We envision that in order to get the more efficient p-i-n solar cells based on the Sb 2 S 3 /TiO 2 -NHJ, it is necessary to replace the polymer HTL with one of high hole mobility, where the HTL material needs to be prevented from infiltrating into the interspaces in-between nanorods, for example, by planarizing the nanoarray with an isolating substance infiltrated into the interspaces among nanorods 8,9,11 .

Discussion
In summary, the Sb 2 S 3 nanorod array on polycrystalline TiO 2 film is realized by a tiny-seed-assisted RSCA method, offering a nanoarray heterojunction system (i.e., Sb 2 S 3 /TiO 2 -NHJ) with the preferentially [211]-orientated Sb 2 S 3 nanorods of lateral sizes much larger than the beneath TiO 2 nanoparticles. The Sb 2 S 3 nanorods in the Sb 2 S 3 /TiO 2 -NHJ have a bandgap of E g = 1.68 eV and provide straightforward and highly effective charge transport channels formed by covalently bounded (Sb 4 S 6 ) n ribbons. Those make the Sb 2 S 3 /TiO 2 -NHJ have a good potential for efficient solar cells. By depositing a conjugated polymer PTB7-th film as a hole transporting layer onto the top of the Sb 2 S 3 /TiO 2 -NHJ, a hybrid solar cell with a p-i-n architecture is obtained. The p-i-n solar cells exhibit a V oc strongly correlating with the net built-in field across devices and no negative E ph impact on device performance. With the two unique advantages of the straightforward and effective charge transport channels and the absence of negative E ph impact on device performance, a power conversion efficiency of 5.70% is achieved in the p-i-n solar cells for an optimized nanoarray structure (i.e., 320-nm-thick Sb 2 S 3 nanoarray prepared at C sg = 0.1 M, C ng = 0.8 M and n = 2). Due to the absence of the negative E ph impact, the p-i-n solar cells exhibit a long electron lifetime on the sub-second scale (τ e1 ≈ 0.9 s) related to the charge recombination at Sb 2 S 3 /PTB7-th interface.
The orientation and morphology of the resulting Sb 2 S 3 nanorods on the TiO 2 film are exclusively governed by the tiny [211]/[221]-oriented Sb 2 S 3 seeds of laterally <10 nm in size. The growth of the [211]-oriented Sb 2 S 3 nanorods follows a tiny-seedgoverned OCE nucleation/growth mechanism according to the epitaxial relationship of Sb 2 S 3 (101)//TiO 2 (101), where the Sb 2 S 3 seeds on the TiO 2 nanoparticles oriented with α ≈ 22°are most competitive during the followed RSCA cycles (n ≥ 1) to take a self-nucleated growth into much larger [211]-oriented Sb 2 S 3 nanorods with a much larger growth rate in c-axis direction than in a-axis and b-axis directions. During the tiny-seed-governed OCE nucleation/growth process, the sites there are no seeds and/ or suppressed/merged crystal growths create the empty interspaces in between nanorods for the formation of the Sb 2 S 3 nanorod array. Moreover, the length of nanorods (i.e., nanoarray thickness) strongly depends on both C ng and n values; the lateral size and the number density of the Sb 2 S 3 nanorods are mainly obtained at the first RSCA cycle (n = 1) and not n-dependent, while they are affected by C ng and increasing C ng leads to increased lateral size and reduced number density.
In the Sb 2 S 3 /PTB7-th hybrid system, no evident interfacial electric field is formed at the Sb 2 S 3 /PTB7-th interfaces due to the weak interaction between the organic and inorganic components; however, the wavefunctions of electronic states near the CBM and VBM of the Sb 2 S 3 /PTB7-th interface mainly locates in the Sb 2 S 3 and PTB7-th phases, respectively, resulting in the unique separate electronic states favorable to the dissociation and injection of photogenerated electron-hole pairs and free charge carriers at the Sb 2 S 3 /PTB7-th interface upon the excitation of Sb 2 S 3 and PTB7th, with the electrons remaining in or being injected into Sb 2 S 3 crystal and the holes being injected into or remaining in PTB7-th phase. The solar cells with p-i-n architecture, consisting of Sb 2 S 3 / TiO 2 -NHJ and PTB7-th, exhibits no negative E ph effects on device performance, which is well accounted for by the ADCD model proposed on the basis of light intensity attenuation in the polymer phase and verified by the finite element simulations and C −V measurements. In contrast, in the p-(p + i)-n counterpart solar cell, the negative E ph effects mainly manifest themselves to reduce the opposite built-in filed E bi across the device, to retard the charge transport and to cause the serious recombination for a seriously deteriorated overall device performance (V oc , J sc , FF, and IPCE).
Our results provide a facile solution-processing method to prepare the high-quality Sb 2 S 3 /TiO 2 nanoarray heterojunction system with preferentially orientated, straightforward, and highly efficient charge transport channels for photovoltaic and other optoelectronic applications, as well as an understanding of the heterogeneously epitaxial growth of single-crystalline nanorods on the polycrystalline surface. Moreover, the NHJ material system and the p-i-n solar cell without the negative impact of E ph on device performance offer the guides to the conceptual design of the efficient photoactive layer and device configuration in the solar cells based on nanorod/nanowire arrays, in particular, the devices configurated using a partner material with a greatly unbalanced charge transport property.

Methods
Sb 2 S 3 precursor solution. The Sb 2 S 3 precursor solution was prepared by a method similar to our previous report except for the addition of glycerin 19 . Briefly, 0.8 mmol of SbCl 3 (≥99.0%, Alfa Aesar) and 1.44 mmol of thiourea (≥99.0%, Alfa Aesar) were sequentially dissolved in 1 ml of N,N-dimethylformamide (DMF) under stirring at room temperature, resulting in a clear yellow precursor solution with 0.8 M SbCl 3 and an SbCl 3 /thiourea molar ratio of 1:1.8, which is stable for weeks in ambient conditions. The different SG and NG solutions were derived from this precursor solution by DMF dilution. The concentrations of the SG (i.e., C sg ) and NG (i.e., C ng ) solutions were defined by the molecular concentration of SbCl 3 dissolved therein.
Preparation of Sb 2 S 3 /TiO 2 -NHJ. The FTO glass substrates (25 × 25 mm 2 , ca. 14 Ω sq −1 ) patterned into stripes (20 × 4 mm 2 ) in the middle part of the substrate by laser were commercially obtained from the Advanced Election Technology Co. Ltd., Liaoning Province, China. The FTO substrates were ultrasonically cleaned with acetone, isopropanol, and deionized water for 20 min, respectively, and then further cleaned with UV-Ozone for 15 min and washed with ethanol before use. A compact and polycrystalline TiO 2 nanoparticle film of ca. 110 nm in thickness was prepared on the FTO substrate as described elsewhere 19 . Sb 2 S 3 /TiO 2 -NHJ was prepared by a tiny-seed assisted RSCA method. Typically, the Sb 2 S 3 precursor deposited onto the TiO 2 film by spin-coating (3000 rpm, 40 s) the SG solution (120 μl; normally C sg = 0.1 M, unless otherwise specified) was sequentially annealed at 150°C for 5 min and 300°C for 10 min on a hotplate in an N 2 -protected glovebox, resulting in tiny Sb 2 S 3 seeds on the TiO 2 film. A transparent Sb 2 S 3 precursor film was prepared on a tiny Sb 2 S 3 seeds-decorated TiO 2 film by spin-coating (3000 rpm, 40 s) the NG solution (120 μl; normally C ng = 0.8 M, unless otherwise specified) and thermally annealed by the route identical with that for the Sb 2 S 3 seed formation, which was referred to as one RSCA cycle. After the seed-decorated TiO 2 film had been subjected to several RSCA cycle numbers (n), the Sb 2 S 3 /TiO 2 -NHJ with a desirable length was in situ formed on the FTO substrate.
Solar cell fabrication. PTB7-th (M w~1 80,000; M w /M n~2 .5, 1-Material) was dissolved in chlorobenzene and stirred overnight at 60°C; after the solution was naturally cooled down to room temperature, the PTB7-th solution (15 mg ml −1 ) for spin-coating was obtained. To prepare p-i-n solar cells (Fig. 1), a PTB7-th film was spin-coated (2000 rpm, 60 s) onto the top of the Sb 2 S 3 /TiO 2 -NHJ from the PTB7-th solution (70 μl) and annealed at 100°C for 15 min to eliminate the solvent residue, where the spin-coating was immediately carried out once the PTB7-th solution was loaded. In order to prepare p-(p+i)-n solar cells (Fig. 4e), the PTB7-th solution droplet (70 μl) was first loaded onto the Sb 2 S 3 /TiO 2 -NHJ and kept in the chlorobenzene atmosphere overnight for the infiltration of the polymer into the interspaces amongst the Sb 2 S 3 nanorods; then, the sample was dried by rotating (2000 rpm, 60 s) and further annealed at 100°C for 15 min to remove the solvent. Finally, an 8-nm-thick molybdenum oxide (MoO 3 ) film and a 100-nm-thick gold electrode were sequentially evaporated onto the PTB7-th layer under a pressure of 5.0 × 10 −4 Pa. The optimized Sb 2 S 3 nanorod array for efficient solar cells was prepared at C sg = 0.1 M, C ng = 0.8 M and n = 2.