Recombination Reduction on Lead Halide Perovskite Solar Cells based on Low Temperature Synthesized Hierarchical TiO2 Nanorods

Intense research on the electron transport material (ETM) has been pursued to improve the efficiency of perovskite solar cells (PSCs) and decrease their cost. More important, the role of the ETM layer is not fully understood, and research on new device architectures is still needed. Here, we report the use of three-dimensional (3D) TiO2 with hierarchical architecture based on rutile nanorods (NR), as photoanode material for PSCs. The proposed hierarchical nanorod films (HNR) were synthesized by a two-steps low temperature (180 °C) hydrothermal method, and consist of TiO2 nanorods trunks with optimal length of 540 nm and TiO2 nanobranches with lengths of 45 nm. Different device configurations were fabricated with TiO2 structures (compact layer, NR and HNR) and CH3NH3PbI3, using different synthetic routes, as active material. The PSCs based on HNR-CH3NH3PbI3 reached the highest power conversion efficiency compared to the PSCs with other TiO2 structures. This result can be mainly ascribed to lower charge recombination as we determine by impedance spectroscopy. Furthermore, we have observed that CH3NH3PbI3 perovskite deposited by the two-step route shows higher efficiency, surface coverage and infiltration within the structure of 3D HNR than one-step CH3NH3PbI3-xClx perovskite.


Introduction
The recent application of organic-inorganic perovskites on solar cells by Kojima et al. 1 has open an attractive field for the easy preparation of solution-based processing solar cells at low temperature (<100 °C) for the development of suitable alternatives for energy conversion. [2][3][4][5] Moreover, due to the excellent optoelectronic properties and long diffusion length of electrons and holes, hybrid organic-inorganic lead halide perovskite materials have emerged as light harvesters for solar cells. [6][7][8][9][10] In the last few years, solution-processed hybrid perovskite solar cells (PSCs) have increased the reported efficiencies from 3.8% to a certified power conversion efficiency of 20.1%. 11 The most commonly reported PSC structure consists of a perovskite layer deposited on a compact TiO 2 layer with an additional mesoporous layer of TiO 2 . The mesoporous layer has been systematically used on dye sensitized solar cells (DSCs) and it is proposed to promote a large surface area and good loading of the absorber. 12,13 However, its role on PSC is still under debate, pointing to imposing some nucleation dynamics of perovskite growth, 14 helping in obtaining continuous perovskite layers and influencing the selective properties of the electron transporting material (ETM). 15 Nevertheless, the existence of numerous boundaries among the nanoparticles of the porous layer increases the density of trapping sites and the probability for electron recombination. One-dimensional (1D) TiO 2 nanostructures (nanowires, nanotubes or nanorods) have been demonstrated to provide a lower charge recombination rate at the grain boundaries and a superior pathway along the long axis of 1D nanostructures for electron transport in DSCs. [16][17][18][19] Recently, substantial efforts were drawn towards the fabrication of novel three-dimensional (3D) hierarchical architecture 20-23 with high surface area, fast electron transport and higher nucleation sites for the deposition of perovskite. 3D hierarchical assemblies have been prepared by chemical vapor deposition, 21 pulsed-laser deposition 22 and multistage electrospinning and hydrothermal methods. 23 Wang et al. 21 realized 3D TiO 2 nanostructures by surface-reaction-limited pulsed chemical vapor deposition (SPCVD) with superior photovoltaic performance compared to nanowires and nanoparticle systems owing to large surface area and charge transport properties. Unfortunately, this method needs a temperature of 600 °C during the entire growth process, even higher than those methods used to fabricate compact or nanostructured TiO 2 layers, where a high sintering temperature (~500 °C) is usually required to crystallize the as-deposited amorphous films. It has been also reported a positive influence of the use of TiO 2 NRs in the stability of perovskite solar modules. 24 In this paper, we report a low-temperature synthesized TiO 2 nanorods (NRs) and 3D hierarchical nanorods (HNRs) as ETM for PSCs. 3D HNRs were fabricated via twosteps hydrothermal methods leading to PSCs with maximum power conversion efficiency (PCE max ) of 10.52% under AM 1.5G illumination, which was higher than PSCs based on NRs and compact TiO 2 films. The good performance of 3D HNR PSC was possible due to the higher light harvesting of NR and HNR structures and to the reduction of recombination losses in HNR respect NR detected by impedance spectroscopy. Hence, we developed low-temperature process for ETM layers that give better performance than those using sintering steps, opening its application on PSCs.

Results
TiO 2 NR photoanodes were synthesized directly on the FTO substrates without any template or compact layer (CL), via hydrothermal method, as shown in Fig. 1a,b. Fig. 1 shows the top-view and cross sectional FESEM images of bare TiO 2 NRs with uniform length and width, nearly rectangular cross section and orderly distributed on the entire surface of the FTO substrate. 3D HNRs consist of a NR backbone that branches out into a network of smaller NRs (Fig. 1c,d). For its fabrication, the as prepared TiO 2 NR film was used as seed to grow branched TiO 2 NRs by a second-step hydrothermal modification method. Fig. 1c,d demonstrates that each TiO 2 NR is enclosed by TiO 2 nanobranches (diameter of ~13 nm and length of ~45 nm). The crystal structures of TiO 2 NR and 3D HNR were both consistent with a tetragonal rutile phase (PDF#21-1276), similar to previous results 25,26 (Fig. S1). To fabricate efficient 3D HNR PSCs, preliminary studies were carried out with four different TiO 2 NR lengths (from 380 to 1100 nm, controlled by growth time) and different thicknesses of the spin-coated perovskite layer (obtained by varying spin rates). It is worth remarking that all the films were synthesized using similar growth conditions: 180 °C and 51 mL of precursor solution and just reaction time was changed (see Experimental Section at Supporting Information). Fig. S2 shows FESEM images of the evolution of NR length as the growth time increases. After keeping the reaction for 110, 120, 125 and 130 min, the lengths of the TiO 2 NR were determined to be ~380, ~540, ~700 and ~1100 nm, respectively, see

c) d)
Devices based on the configuration FTO/TiO 2 NR/CH 3 NH 3 PbI 3-x Cl x / Spiro-OMeTAD/Au were prepared (Fig 2a). In these devices CH 3 NH 3 PbI 3-x Cl x was spincoated onto the TiO 2 NR through a one-step solution deposition method at different spin rates (from 1000 to 4000 r.p.m.) (see Experimental Section in Supporting Information). Subsequently, a spin-coated Spiro-OMeTAD layer was used as selective contact and hole transport material (HTM). Finally, gold was evaporated onto the Spiro-OMeTAD to form an ohmic contact.   Table  S1. The average PCE value (PCE avg ) was obtained from the measurement of ten different NR PSCs. PCE max of 9.1% and PCE avg of 8.92% have been obtained for TiO 2 NR length of 540 nm, with short circuit photocurrent density (J SC ) = 17.49 mA cm -2 , open-circuit voltage (V OC ) = 822.63 mV and fill factor (FF) = 0.63. Furthermore, the photovoltaic parameters were strongly dependent of the TiO 2 NR length. In general, it was found that above 600 nm NR length, the value of PCE decreases as shown in Fig.  2c and Table S1. Previous studies using 1D TiO 2 nanostructures, 25,27 suggested that this is due to charge recombination at the larger TiO 2 NR-perovskite interface. 28,29 To demonstrate the benefits of using a 3D hierarchical architecture as photoanode, PSCs were fabricated with a compact layer of TiO 2 (flat), bare TiO 2 NR and 3D HNR. For this study, flat device is suited as the control device. 3D HNR films were synthesized, as explained in the experimental section, using the optimum TiO 2 NRs (540 nm length) as seeds. We have also tested the influence of the perovskite growth method in the device configuration comparing mixed-halide CH 3 NH 3 PbI 3-x Cl x perovskite (Sing-MAICl), where a single deposition step is used, and lead iodide CH 3 NH 3 PbI 3 perovskite (Seq-MAI), where two steps are utilized, see Supporting Information for further experimental details. The different PSCs configurations are shown in Fig. 3. Fig. 3a,b,c show the device configuration for Sing-MAICl PSCs, where CH 3 NH 3 PbI 3-x Cl x was deposited by one-step technique at 2000 r.p.m., as described above. For Seq-MAI PSCs (Fig. 3g,h,i) a two-steps sequential deposition method 5 was implemented. Briefly, a PbI 2 layer was spin-coated on TiO 2 photoanodes, followed by a dipping treatment in a solution of CH 3 NH 3 I. The same perovskite deposition procedure has been utilized for all three different substrates, with no further optimization analysis at each particular substrate morphology.
Cross sectional FESEM images of FTO/TiO 2 NR (HNRs)/Sing-MAICl (Seq-MAI) were taken to investigate the infiltration of the perovskite within the structures of the TiO 2 NR and 3D HNR (Fig. 3e,f and 3k,l). We can observe the good surface coverage and infiltration of either Sing-MAICl or Seq-MAI perovskite layers within TiO 2 NR and 3D HNR photoanodes. It is important to remark that uniform coverage and adequate infiltration have been observed to be critical factors to obtain excellent photovoltaic parameters in PSCs. 30,31 Moreover, Fig. S4 shows the top-view FESEM images of the Sing-MAICl and Seq-MAI perovskite layers. Sing-MAICl layer on Flat, NR and HNR is smooth, homogeneous and perfectly covers the TiO 2 photoanode as shown in Fig.  S4a; whereas, Seq-MAI consists of cuboids (Fig. S4b) with a grain size ~ 405 nm, in agreement with the value reported for the concentration of CH 3 NH 3 I used. 32 In fact, sequential deposition that produces excellent results for PSCs when TiO 2 scaffold is employed, 5 does not produce good results when perovskite is deposited directly on top of compact TiO 2 layer. Large crystal (Fig. S4b) do not cover completely FTO layer, producing the direct contact between selective contacts (compact TiO 2 and spiro-OMeTAD). The effects of TiO 2 morphology on the performance of the PSCs fabricated from Flat, NR and HNR photoanodes were investigated. A summary of the photovoltaic performance and averaged parameters is shown in Fig. 4 and Table 1. NR and HNR present higher J SC than flat samples, due to a thicker perovskite layer, see Fig. 3. Except for the Sing-MAICl planar architecture, higher values are obtained for devices based on Seq-MAI, mostly because of the higher values on V OC . The device with HNR Seq-MAI reached a promising maximum PCE of 10.5% resulting from J SC = 17.33 mA cm -2 , V OC = 946.96 mV and FF = 0.64. Also, HNR Sing-MAICl reports the best PCE (9.44%) among the PSCs based on MAICl configuration, indicating that 3D hierarchical architecture is superior to bare NR and Flat films, for the analyzed growth conditions. 21

a) b) c) g) h) i)
photon-to-current conversion efficiency (IPCE) above ~ 480 nm is more pronounced for PSCs based on NR (Fig. 4c,d) than on HNR, although the Flat device shows the highest decrement in this wavelength range. The integrated short-circuit current density J SC,INT calculated from IPCE is consistent with the Jsc values obtained from a solar simulator, see Table 1.    More interesting is the dependence of V OC on TiO 2 morphology and perovskite type. An increase in V OC with the use of 3D HNR as photoanode instead of NR or Flat films is in accordance with previous results in the literature. 21,23 V OC is strongly influenced by the recombination rate. 33 In order to study the effect of recombination on the measured devices, impedance spectroscopy under 1 Sun illumination has been carried out. Fig. 5 shows the recombination resistance, R rec , obtained from impedance analysis. 34 There is a clear correlation between higher R rec and higher V OC . From Fig. 5a it is observed that for samples using Sing-MAICl, R rec follows the trend Flat>HNR>NR as the V OC does, see Table 1. Indicating that the recombination rate increases as NR>HNR>Flat. In contrast for Seq-MAI, see Fig. 5b, R rec follows the trend HNR>NR>Flat also in good agreement with the trend observed in the obtained V OC , see Table 1. Taking into account that the recombination rate is inversely proportional to the recombination resistance, 33 the differences between Flat samples prepared by Sing-MAICl and Seq-MAI can be ascribed to the higher recombination in Flat Seq-MAI due to the non-complete covering of the compact TiO 2 surface, See Fig. S4b, as it has been already commented. Moreover, these results also point to a superior performance of HNR in comparison with NR, and also to flat samples in the case of Seq-MAI. We hypothesize that this effect could be ascribed to a lower density of surface traps in 3D HNR, which can lead to a lower net charge recombination at the TiO 2 -perovskite interface, despite the larger effective surface of HNRs in comparison with NRs.
The superior performance of Seq-MAI in comparison with Sing-MAICl, except for the already analyzed case of Flat Seq-MAI, is also due to the lower recombination rate in Seq-MAI. The lower recombination observed for Seq-MAI in the case of NRs, see Fig. 5c, and in the case of HNRs, see Fig. S5, is the responsible of the higher V OC observed for these samples, see Table 1. The growth method used for the synthesis of perovskite layer also has an important effect in the device recombination and consequently on the final performance.

Conclusions
In summary, we have demonstrated an enhanced power conversion efficiency through the use of 3D hierarchical structures based on nanorods with optimized TiO 2 NR length and Seq-MAI perovskite layer. The superior efficiency 3D HNR as photoanode is based on one hand on the higher light harvesting properties of NR and HNR in the studied conditions, compared to flat devices. However light harvesting could be increased for flat samples with other growth conditions, but we have preferred to prepared all the samples with the same deposition procedure independently of the substrate in order to avoid introducing additional growth considerations. On the other hand the superior performance of HNR is based on the lower recombination rate obtained from HNR, especially for HNR Seq-MAI, where recombination rate is lower than in NR samples prepared with both depositions and even lower than in flat samples prepared with sequential method. Lower recombination rate causes a significant increase V OC compared to the other TiO 2 structures. Further research is necessary to identify the exact recombination pathway in each configuration and the way in which HNR hinders it, despite the higher surface area. The low temperature hydrothermal routes have the additional advantage of low-cost and the versatility to be implemented in other substrates, including flexible ones. More work is still needed to optimize the use of these low-cost ETM for the preparation of PSCs with record values as those reported in the literature. Nevertheless, the results here reported highlight the importance of the charge selective contacts in the ultimate performance of PSCs and the relationship with the recombination processes. Results also shows that not just material but structure plays an important role in the selective contacts of PSCs. Finally, they show that advanced selective contacts with reduced recombination grown at low temperature are possible, thus paving the way for future flexible opto-electronic applications.