TiO2 nanorod arrays hydrothermally grown on MgO-coated compact TiO2 for efficient perovskite solar cells

The morphology of electron transport layers has a significant impact on the device architecture and electronic processes of mesoscopic perovskite solar cells (PSCs). In this study, ultrathin MgO is coated on the surface of compact TiO2 (c-TiO2). The MgO-coated c-TiO2 is first used as seeds to hydrothermally grow one-dimensional (1D) TiO2 nanorod (NR) arrays for PSC devices. Rutile nanorod arrays are fabricated via a facile solvothermal method using tetrabutyl titanate (TBT) as the Ti precursor. The microstructures and morphologies, including nanorod diameter, length, and areal density, of the TiO2 NR arrays are varied by controlling the concentration of TBT from 0.3 M to 0.7 M. Furthermore, the profound effects of the MgO modification and titania nanorod morphology on the pore-filling of perovskite CH3NH3PbI3, charge separation and recombination at the perovskite/titania nanorod interface are investigated. Our results reveal that the Ti precursor concentration strongly affects the open-circuit voltage (VOC), short-circuit current density (JSC), and fill factor (FF) of the 1D TiO2 NR array-based device. Under optimized conditions with MgO coating and at 0.4 M TBT, our champion cell with 1D TiO2 NRs demonstrates a power conversion efficiency (PCE) of 17.03% with JSC 1⁄4 22.01 mA cm , VOC 1⁄4 1.06 V, and FF 1⁄4 0.73. Under the same fabrication conditions, MgO modification enhances the average PCE to 16.24% for the PSCs with the MgO coating from 13.38% for the PSCs without the MgO coating. The devices show an approximately 18% improvement in performance, which mainly results from the opencircuit voltage and fill factor enhancements. Moreover, advantageously, the MgO modification is found to reduce the current density–voltage (J–V) hysteresis with respect to the scan direction and improve the UV stability of the non-encapsulated cells. Therefore, this study presents a promising approach to fabricate efficient and stable one-dimensional TiO2 nanorod array-based perovskite solar cells.


Introduction
Organometal halide perovskite solar cells (PSCs) are becoming a signicant portion of photovoltaic technologies due to their low cost and rapidly increasing efficiency. PSCs have already surpassed organic photovoltaics (OPVs) 1 and dye-sensitized solar cells (DSSCs) 2,3 and are now, at least in terms of efficiency, in the same range as leading thin lm PVs such as cadmium telluride (CdTe) 4 and copper indium gallium selenide (CIGS) 5,6 or other types of thin lm solar cells that have existed for decades. Currently, perovskite solar cells based on zero dimensional (0D) metal oxide nanostructures have achieved excellent power conversion efficiencies (PCEs) of 22.1% (for small area devices) 7 and 17.8% (for large area devices). 8 The major components of PSC devices include electron transporting layers (ETL), hole transporting layers (HTL) and perovskite absorbing lms. Among these functional layers, ETLs are a critical component of PSCs for the improvement in PV performance. 9 Mesoporous TiO 2 is the most common ETL used for highly efficient perovskite solar cells. [10][11][12][13][14][15] The mesoporous architecture is comprised of a large number of ca. 20 nm-sized nanoparticles (NPs) with many grain boundaries, which hamper electron transport and limit the charge collection efficiency. 16,17 In addition, due to the random distribution of NPs within mesoporous metal oxide lms, pore-lling of perovskite and hole transport materials (HTM) in the interspace between NPs remains a persistent difficulty. To overcome the problems associated with mesoscopic structures, great effort is being devoted towards the application of one-dimensional (1D) ETLs in PSCs. ETLs with 1D structures provide a direct path for the transport of photo-generated electrons. These structured ETLs also allow better pore lling of the perovskite absorber than mesoscopic TiO 2 lms due to the open and regular pore structure of the former. [18][19][20] TiO 2 , WO 3 , ZnO and CdS nanorods and nanotubes have been tested as ETLs for PSCs, but only TiO 2 nanorods have consistently produced PSCs with PCEs above 15%. [21][22][23] TiO 2 nanorods (NRs) are developing as another popular conguration for ETLs in perovskite solar cells. Qiu et al. 19 rst reported the use of TiO 2 nanorods as an ETL in PSCs. Using a CH 3 NH 3 PbBr 3 absorber on 1.5 mm-thick nanorods, they reported a PCE of 4.87%. This report triggered new research on the application of TiO 2 nanorods as an ETL for perovskite solar cells. Kim et al. 20 made comparative studies on the performance of longer and shorter TiO 2 nanorods. They concluded that shorter nanorods could provide better inltration of perovskite. Using 560 nm-long nanorods, they achieved a PCE of about 9.4%. In 2014, Jiang et al. 24 fabricated a perovskite solar cell using 900 nm-long TiO 2 nanorods as the ETL and achieved an efficiency of 11.7%. Subsequently, Mali et al. 25 were able to demonstrate a TiO 2 -nanorod based perovskite solar cell with an efficiency of 13.5%, which was achieved by passivating the surface of the nanorods with ultrathin TiO 2 via atomic layer deposition in 2015. Most recently, in 2016, Li et al. achieved record efficiencies of 18.22%. 26 They utilized a UV-ozone cleaning process to eliminate organic residues on the nanorod surface to improve the TiO 2 nanorod/perovskite interface.
In this study, ultrathin MgO is coated on the surface of compact TiO 2 (c-TiO 2 ). The MgO-coated c-TiO 2 is rst used as seeds to hydrothermally grow one-dimensional (1D) TiO 2 nanorod (NR) arrays for PSC devices. Rutile nanorod arrays are fabricated via a facile solvothermal method using tetrabutyl titanate (TBT) as the Ti precursor. The microstructures and morphologies of the TiO 2 NR arrays, including nanorod diameter, length and areal density, are tuned by controlling the concentration of TBT from 0.3 M to 0.7 M. The electrical behavior and optical properties of titania nanorod ETLs are characterized and examined to understand the sources of both underperformance and outperformance in various device parameters. We demonstrate the profound effect of the MgO modication and titania nanorod morphology on the pore-lling of perovskite CH 3 NH 3 PbI 3 , charge separation and recombination at the perovskite/titania nanorod interface. We optimize the Ti precursor concentration and morphology of the hydrothermally grown TiO 2 nanorod arrays to produce hysteresis-less CH 3 NH 3 PbI 3 perovskite solar cells with a champion PCE of 17.03% and high UV light stability.

Preparation of rutile TiO 2 nanorod arrays
Rutile TiO 2 arrays were prepared via a hydrothermal method, as reported in ref. 27. Fluorine-doped tin oxide (FTO)-coated glass substrates were cleaned by ultrasonication with soap, deionized water, acetone, and isopropyl alcohol and nally treated with ultraviolet light for about 20 minutes. Then, isopropyl titanate (200 mL) and ethanol (5 mL) were mixed to prepare a clear precursor sol. The precursor sol was spin-coated on the Zn/HCletched FTO substrate at 4500 rpm, followed by annealing at 500 C to form a compact TiO 2 (c-TiO 2 ) layer. To prevent charge recombination at the methylamine lead iodide/TiO 2 interface in perovskite solar cells, MgO coated c-TiO 2 layers were fabricated before the synthesis of TiO 2 -NR arrays by modifying the procedure reported in our previous work. 14 An MgO precursor solution of Mg(CH 3 COO) 2 in deionized water was spin-coated on the c-TiO 2 layer at 4500 rpm for 30 seconds, and was then heated at 400 C for 1 h. The concentration of the magnesium salt used was 70 mM. Herein, two pieces of clean FTO glass with TiO 2 seed layers were placed at an angle of 45 against the wall of the Teon-liner with the conducting side facing down. Concentrated hydrochloric acid (36.5-38% by weight) was added to deionized water with a volume ratio of 1 : 1 and stirred for 2 minutes. Subsequently, TBT was added, and the mixture was further stirred. The amount of titanium butoxide was set as 0.03, 0.035, 0.04, 0.05, 0.06 M to obtain the desired TiO 2 -NR arrays respectively. The hydrothermal synthesis was conducted at 150 C for 4 h in an electric oven to ensure that TiO 2 -NR arrays with a uniform thickness were obtained. Aer cooling to room temperature in air, the lms were taken out from the autoclaves, rinsed with deionized water, dried in air, and annealed at 500 C for 30 min to remove any residual organic contaminants.

Solar cell device fabrication
CH 3 NH 3 PbI 3 was coated on the TiO 2 -NR arrays via a typical onestep spin-coating procedure under a nitrogen atmosphere. 28,29 Synthesized CH 3 NH 3 I (0.1975 g) powder and lead iodide PbI 2 (0.5785 g) were stirred in a mixture of g-butyrolactone (GBL) (700 mL) and dimethylsulphoxide (DMSO) (300 mL) at 60 C for 12 h. The lm was spin-coated with the precursor solution at 2000 rpm for 30 seconds and at 3500 rpm for 40 seconds. 70 mL of anhydrous diethyl ether was dropped onto the center of the sample during the last 30-40 s. The CH 3 NH 3 PbI 3 lm was obtained aer annealing at 115 C for 20 min on a hotplate, and the lm color changed from transparent to yellow and nally to dark brown. Aer the deposition of perovskite, an HTM solution was spin-coated at 3500 rpm for 30 s. The HTM solution was obtained using 1 mL 2,2,7,7-tetrakis(N,N-di-pmethoxyphenylamine)-9,9-spirobiuorene (spiro-OMeTAD)/ chlorobenzene (72.3 mg mL À1 ) solution with 17.5 mL Li-TFSI/ acetonitrile (520 mg mL À1 ) and 28.8 mL tBP. Finally, an Ag-Al electrode (70 nm) was thermally evaporated at an atmospheric pressure of 10 À7 Torr to match the work function. 29,30

Characterization
The morphology, structure and composition of the TiO 2 arrays were investigated via eld emission scanning electron microscopy (FESEM, HitachiS-4800), and X-ray diffraction (XRD, Bruker D8 Davinci instrument, Cu-Ka: l ¼ 0.15406 nm). UV-vis absorption spectra were obtained using a UV-vis spectrophotometer (HitachiU-3900). The photocurrent densityvoltage (J-V) curve was measured using a Keithley model 2440 Source Meter under the illumination of simulated AM 1.5 G solar light (100 mW cm À2 , 1 sun intensity) from a Newport solar simulator system (equipped with a 1 kW xenon arc lamp, oriel, calibrated with a standard silicon reference cell). During device photovoltaic performance characterization, a metal aperture mask with an opening of about 0.09 cm 2 was used. External quantum efficiency (EQE) measurements (74125, Oriel, USA) were also carried out for these cells. Hydrothermal solution growth of TiO 2 -NR arrays has been widely reported on TiO 2 seed layers or directly on Ti foil substrates to increases the nucleation sites for growth. 18,31,32 In our work, the growth mechanism of the TiO 2 -NR arrays on MgOcoated c-TiO 2 is similar to the traditional case due to the ultrathin thickness of MgO. When TBT was mixed with HCl solution at room temperature, TBT reacted with HCl and H 2 O forming a Ti(IV) complex. When the solution was heated at 150 C, the Ti(IV) complex transformed into TiO 2 on the surface of the c-TiO 2 layer. Cl À ions can selectively adsorb onto the (110) crystal planes suppressing further growth of the planes, resulting in anisotropic growth along the [001] orientation 18,31 Meanwhile, the growth process was signicantly inuenced by the concentration of the Ti(IV) complex in the solution. As shown in Fig. 1, the diameter and areal density of the TiO 2 NRs changed with the concentration of TBT. As the TBT concentration increased from 0.03 M to 0.07 M, the length and mean diameter (measured by Nano Measurer) of the TiO 2 -NRs increased from 250 nm and 21 nm to 1163 nm and 76 nm, respectively, while the areal density (number of nanorods per unit area) of the TiO 2 -NR arrays increased from 224 mm 2 at 0.03 M TBT to 392 mm 2 at 0.04 M, and then decreased as the TBT concentration further increased, as shown in Table 1. The TiO 2 -NR arrays became more and more vertically aligned and compact with an increase in the TBT concentration from 0.03 M to 0.07 M. It is worth noting that the TiO 2 -NR arrays grown in 0.04 M TBT precursor solution were 298 nm in length and had a small mean diameter of 22 nm and high areal density of 392 mm 2 . The microstructures and morphologies of the grown TiO 2 -NR arrays play an important role in the performance of PSCs due to the charge separation and recombination at the perovskite/ titania nanorod interface 19,26 and the sensitivity of perovskite crystallization to the interfacial structure. 33,34 A small diameter is very important in obtaining a high area density, which is conducive to increasing the surface area of the TiO 2 -NR array. Increasing the surface area leads to improved charge separation at the interface of CH 3 NH 3 PbI 3 /TiO 2 . On the other hand, with an increase in length of the TiO 2 NRs, the charge combination is expected to be intensied. Fig. 2 shows top surface and cross-sectional SEM images of the CH 3 NH 3 PbI 3 -coated TiO 2 -NR arrays prepared with different concentrations of TBT. It can be seen that the total thickness of the CH 3 NH 3 PbI 3 absorber layer on the TiO 2 -NR arrays prepared with TBT concentrations ranging from 0.03 M to 0.07 M increased from about 400 nm to 1320 nm. The pore-lling of CH 3 NH 3 PbI 3 dramatically varied together with a change in the interspace between the TiO 2 NRs. The dispersed small sized TiO 2 NRs prepared with a low TBT concentration facilitated the loading of perovskite into the TiO 2 pores or clearance between TiO 2 NRs. The perovskite capping layers in the TiO 2 -NR arrays prepared with the low TBT concentrations of 0.03 M and 0.04 M show a very smooth and compact top surface morphology, as shown in Fig. 2(a) and (b), respectively. However, when the TBT   concentration is higher, some perovskite material voids are visible on the top surface of the perovskite absorber, as shown in Fig. 2(d) and (e). Moreover, from the side views ( Fig. 2(h)-(j)), large perovskite material voids are clearly observed on the surface or vicinity of the TiO 2 NRs, which is probably caused by an Ostwaldt type ripening process during the crystal growth. 35 The pinholes or perovskite material voids in the absorbers can decrease the shunt resistances or lead to HTM inltration and cause short circuit, resulting in poor PSC device performances. Notably, the absorber sample formed under condition of 0.04 M TBT shows the optimal top-view and cross-section morphological properties. The perovskite semiconductor material almost completely covers the TiO 2 NR array scaffold in this case, as shown in Fig. 2(g). Fig. 3(a) and (b) show the UV-vis spectra of the bare and CH 3 NH 3 PbI 3 coated TiO 2 -NR arrays prepared with different concentrations of TBT, respectively. From Fig. 3(a) Fig. 1, the results of the UV-vis spectra are thought to be attributed to the increasing length and diameter of the TiO 2 nanorod arrays, which may reduce reection and transmission. 36 The real absorption spectra of perovskite CH 3 NH 3 PbI 3 is shown in Fig. 3(c). As can be seen from Fig. 3(c), with the increase in TBT concentration from 0.03 to 0.04 M, the perovskite CH 3 NH 3 PbI 3 has gradually stronger absorption in the 450-750 nm wavelength range. However, when the TBT concentration increased further from 0.04 to 0.07 M, the perovskite has weaker absorption. The perovskite lm formed with the TBT concentration of 0.04 M exhibits the strongest light absorption over the 450-750 nm wavelength range. The increased absorption of the perovskite should be the result of improved surface coverage and more uniform crystal formation in the perovskite thin lm. Moreover, due to the similar thicknesses of the CH 3 NH 3 PbI 3 capping layers under the conditions of various TBT concentrations, as shown in Fig. 2(f)-(j), the increased absorbance is mainly associated with the optimized interfaces of perovskite/TiO 2 -NRs. Fig. 4(a) and (b) show the current density-voltage (J-V) characteristics and EQE spectra of the fabricated devices based on 1D TiO 2 NRs prepared with different TBT concentrations. The J-V measurements were recorded under AM 1.5 G solar irradiance, and the photovoltaic performance parameters and shunt (R SH ) and series (R S ) resistances are summarized in Table 2. It can be seen that the PSC device based on TiO 2 NRs prepared with a TBT concentration of 0.03 M shows a PCE of 14.55%, resulting from an open circuit voltage (V OC ) of 1.05 V, current density (J SC ) of 22.0 mA cm À2 and ll factor (FF) of 0.63. When the TBT concentration was increased from 0.03 to 0.04 M, the device showed a signicant improvement in FF with the corresponding increase in PCE to 16.24%. Upon further increasing the TBT concentration from 0.04 M to 0.07 M, lower J SC , V OC and FF values, and thus worse PV performances were observed in the related devices. At 0.07 M TBT, a considerable decrease in J SC , V OC and FF was observed, resulting in the corresponding decrease in PCE to 13.52%. From Tables 1 and 2, when the TBT concentration increased from 0.04 M to 0.07 M, the length of the grown TiO 2 -NRs increased from 250 nm to 1163 nm, while the J SC , V OC , FF and PCE values of the device performance continuously decreased. These results can be attributed to the disorder in the longer nanorod arrays and reduced pore-lling (Fig. 2). Kim et al. also found that photocurrent, photovoltage and power conversion efficiency decreased with an increase in nanorod length from 0.56 mm to 1.58 mm. 20 For solar cells, the shunt resistance (R SH ) is due to p-n junction non-idealities and impurities near the junction; however, the major contributors to the series resistance (R S ) are the bulk resistance of the semiconductor material (the active layer), the contact resistance at the semiconductor-conductive interfaces and the resistance of the conductive contacts. 37 As the TBT concentration increased from 0.03 to 0.04 M, the device showed an obvious decrease in both R S and R SH . Further increasing the TBT concentration from 0.04 M to 0.07 M, R S changed slightly, while R SH continuously decreased. At 0.07 M TBT, the shunt resistance decreased by 5 times. The lowest FF of the device prepared at 0.03 M can be associated with the largest R S . 38 The key limitation in the performance of mesosuperstructured PSCs is the balance between R SH and R S resistance. 28 Both the smallest series resistance and the second largest shunt resistance of the PSC device prepared at 0.04 M are responsible for its highest FF. When the TBT concentration is 0.04 M, the formed TiO 2 NRs have a small diameter and the highest area density, which are conducive to increasing the surface area of the TiO 2 -NR arrays and enhancing pore-lling. This leads to improved charge separation at the interface of CH 3 NH 3 PbI 3 /TiO 2 , and as a result, optimal and highperformance perovskite solar cells were assembled.

Results and discussion
The EQE values of the devices based on the TiO 2 NRs prepared with various TBT concentrations are quite different in the wavelength range of 300 and 400 nm, as shown in Fig. 4(b). This difference is due to the absorption of light by the TiO 2 NRs, as shown in Fig. 3(a). Moreover, the EQE shown in Fig. 4(b) is relatively higher for the lower TBT concentration in the wavelength range of 400 and 600 nm, which indicates that the shorter nanorods utilize light more efficiently than the longer nanorods in this wavelength region. The EQE results of the devices with different TBT amounts are in good agreement with that from the J-V measurements of these cells, as shown in Fig. 4(a).
In early-reported TiO 2 NR array based PSCs, 19,20,24-26 TiO 2 NR arrays were mostly grown on TiO 2 compact-layer seeded FTO substrates via the hydrolysis of Ti precursors. In our work, ultrathin MgO coated-c-TiO 2 layers were fabricated before the synthesis of the TiO 2 -NR arrays by modifying the procedure reported in our previous work for incorporating MgO into porous TiO 2 . 14 In order to examine the inuence of MgO modication, we measured the top surface and cross-sectional SEM images of TiO 2 NR arrays directly deposited on c-TiO 2 with 0.04 M TBT, as shown in Fig. 5(a) and (b), respectively. The length and mean diameter of the TiO 2 nanorods prepared at 0.04 M TBT without MgO modication are similar to that of the   TiO 2 NRs grown at the same TBT concentration with MgO modication; however, the areal density of the former (248 mm À2 ) is smaller than that (392 mm À2 ) of the latter, as shown in Fig. 5  and 1(b) and (g), respectively. Moreover, without MgO modication, the obtained TiO 2 NR arrays are randomly and nonuniformly tilted. The small areal density and badly aligned and unevenly distributed TiO 2 NR arrays hinder inltration of perovskite, thus leading to absorbers with poor morphology which degrade the photovoltaic performance of PSCs. 20,24,39 Furthermore, it is known that there are numerous pin-holes present within TiO 2 compact layers prepared via solution methods and these pin-holes expose the FTO gains underneath the layers, which decreases the blocking capability and electron collecting efficiency of the photoanode. 40,41 With MgO modication of the c-TiO 2 , MgO insulating material particles are spread over the c-TiO 2 layer which can partially cover these pores. Furthermore, since MgO particles are scattered over the substrate, the MgO coated c-TiO 2 layer has places where TiO 2 can have electrical contact with the TiO 2 NRs and perovskite. Upon MgO modication, the c-TiO 2 layer becomes smoother, leading the growth of more vertically aligned TiO 2 -NR arrays, as evidenced by the SEM observations shown in Fig. 1 and 5.   Fig. 6(a), the optical absorption edge of the TiO 2 -NR arrays with MgO modi-cation shis to a shorter wavelength range and a slightly lower absorbance or higher transmission is observed, especially in the short wavelength region. The higher transmission can contribute to improving the photocurrent. The blue-shi in the light absorption is due to the wider energy band gap of MgO. The perovskite absorber lm based on c-TiO 2 with MgO modi-cation displays slightly stronger light absorption in the 300-800 nm range than the case without MgO modication, as shown in Fig. 6(b). The increased absorption of the perovskite is the result of efficient inltration of perovskite, improved surface coverage and more uniform crystal formation in the perovskite thin lms. Fig. 6(c) shows the J-V curves of the best-performing perovskite solar cells based on c-TiO 2 without and with MgO modi-cation at 0.04 M TBT for forward and reverse scans. The corresponding best and mean PV parameters including J SC , V OC , FF and PCE are summarized for the reverse scans in Table 3. As can be seen from Table 3, the PSC device based on c-TiO 2 without MgO modication shows an average PCE of 13.38% AE 1.43% resulting from a V OC of 0.96 V AE 0.04 V, J SC of 21.58 AE 0.80 mA cm À2 and FF of 0.67 AE 0.04. In contrast, the PSC with MgO modication has a higher average PCE of 16.24% AE 0.79%, with a V OC of 1.04 V AE 0.02 V, J SC of 21.75 AE 0.40 mA cm À2 , and FF of 0.73 AE 0.02. With MgO modication, there are obvious enhancements in the V OC and FF. Besides, there was also a slight enhancement in the J SC , as expected, from the SEM and optical absorption examination and analysis in the previous sections. The improved photocurrent is also supported by the EQE spectra shown in Fig. 6(d). The improved open-circuit voltage can be attributed to the up-shi of the conduction band edge of TiO 2 with the MgO coating. 42 Moreover, the series resistance (R S ) derived from the J-V curves has an impact on the FF. 38 The lower R S of the PSC with MgO modication induced its higher FF. As result, the PSC with MgO modied c-TiO 2 achieved a champion PCE of 17.03%.
Mesoscopic CH 3 NH 3 PbI 3 perovskite hybrid solar cells usually have signicant J-V hysteresis with respect to the forward and reverse scan directions owing to charge accumulation or dielectric polarization by ferroelectric properties. 43 From Fig. 6(c), the PSC device based on c-TiO 2 with MgO modication exhibited not only better device efficiency but also smaller J-V hysteresis with respect to the scan direction than the PSC devices without MgO modication. With MgO modication, the reduced J-V hysteresis can be attributed to the decreased recombination by eliminating the undesirable recombination pathway between TiO 2 and the spiro-OMeTAD hole conductor and the reduction in electronic trap states, which enable faster electron transport by the formation of highquality CH 3 NH 3 PbI 3 absorbers. 44 Fig. 6(e) shows the EIS results for perovskite solar cells based on c-TiO 2 without and with MgO modication tested in the dark. In dark conditions, the structure of the PSC device could be simplied as a leaking capacitor. 14 From Fig. 6(e), the size of the semicircle is related to the recombination resistance at the interface of the TiO 2 /perovskite layer and TiO 2 /hole transport layer. The bigger the diameter of the semicircle, the lower the electron recombination at the interface. The obtained EIS indicates that the recombination resistance increased by several times aer MgO modication, which led to a signicant decrease in current loss through recombination and an increase in FF.
The stability of perovskite solar cells is a major issue restricting their terrestrial application. 14, 45 We investigated the stability of unsealed PSCs based on c-TiO 2 with and without modication with under UV irradiation. The PSCs were exposed to 365 nm UV illumination at an intensity of 90 mW cm À2 , and were removed at certain time intervals to measure the J-V curves under simulated AM1.5 100 mW cm À2 irradiance. In Fig. 6(f), it can be observed that device based on 1D TiO 2 NR arrays grown on MgO coated c-TiO 2 exhibits a signicantly improved stability. Its PCE retains more than 92% of its initial value, even aer 25 min UV irradiation in air.

Conclusions
We have demonstrated a low-temperature synthesis approach for good ultrathin passivation on TiO 2 seeds and high-quality 1D TiO 2 nanorod arrays using MgO coating and hydrothermal techniques, respectively. Ultrathin MgO coated c-TiO 2 has been used as seeds to hydrothermally grow one-dimensional TiO 2 NR arrays for PSCs. Tetrabutyl titanate (TBT) was employed as the Ti precursor. The nanorod diameter and length and areal density of the TiO 2 NR arrays were adjusted by controlling the concentration of TBT from 0.3 M to 0.7 M. The effects of MgO modication and TiO 2 NR array morphology on the pore-lling of perovskite CH 3 NH 3 PbI 3 as well as charge separation and recombination at the perovskite/titania nanorod interface have been investigated. Our results show that the TBT concentration strongly affects the V OC , J SC and FF of the 1D TiO 2 NR based PSC cell. The best perovskite solar cell based on embedded 1D TiO 2 factor. Furthermore, MgO modication reduced the J-V hysteresis with respect to the scan direction and improved the UV stability of the non-encapsulated solar cells. This study demonstrates that the employment of ultrathin MgO-coated TiO 2 as seeds is a promising strategy to hydrothermally grow high-quality 1D TiO 2 nanorod arrays for efficient and stable perovskite solar cells.

Conflicts of interest
There are no conicts to declare.