Nanoporous TiO ₂ films are commonly used as working electrodes in dye-sensitized solar cells (DSSCs). So far, there have been attempts to synthesize films with various TiO ₂ nanostructures to increase the power-conversion efficiency. In this work, vertically aligned rutile TiO ₂ nanorods were grown on fluorinated tin oxide (FTO) glass by hydrothermal synthesis, followed by deposition of an anatase TiO ₂ film. This new method of anatase TiO ₂ growth avoided the use of a seed layer that is usually required in hydrothermal synthesis of TiO ₂ electrodes. The dense anatase TiO ₂ layer was designed to behave as the electron-generating layer, while the less dense rutile nanorods acted as electron-transfer pathwaysto the FTO glass. In order to facilitate the electron transfer, the rutile phase nanorods were treated with a TiCl ₄ solution so that the nanorods were coated with the anatase TiO ₂ film after heat treatment. Compared to the electrode consisting of only rutile TiO ₂ , the power-conversion efficiency of the rutile-anatase hybrid TiO ₂ electrode was found to be much higher. The total thickness of the rutile-anatase hybrid TiO ₂ structures were around 4.5-5.0 μm, and the highest power efficiency of the cell assembled with the structured TiO ₂ electrode was around 3.94％. The recent acceleration in the pace of global warming has forced many researchers to focus on the development of alternative energy technologies such as solar cells and wind turbines that can replace a substantial portion of fossil fuels [1] . Among these alternatives, solar-cell technology is under active development owing to the possibility of instant application to the power supply chain [2] . Among solar cells, dye-sensitized solar cells (DSSCs) are considered more economically viable and environmentally friendly compared to silicon-based solar cells [2 - 5] . Titanium dioxide (TiO ₂ ) has been adopted as an electrode material in DSSCs because of its promising characteristics as an n-type semiconductor. It has chemical and physical stability as well as a wide band-gap (3.2 eV) under ultraviolet light[6-8]. TiO ₂ has three types of crystallographic structures: anatase, rutile, and brookite [9 , 10] . Anatase is known, despite its metastable characteristics at ambient temperature, to possess the highest photocatalytic activity because of its large surface area compared to rutile and brookite [11 - 14] . In addition, the anatase structure has low crystal-lattice packing fraction so that it absorbs light better owing to the reduced refractive index [15] . When anatase TiO ₂ is heated above 500 ℃, it is irreversibly converted to the stable rutile structure [16 , 17] . Besides the physical and chemical properties of TiO ₂ , the geometry of the nanostructured TiO ₂ electrode is a critical factor of the performance of DSSCs because the contour surface of the TiO ₂ structure serves as a network of electron pathways [18 - 20] . Recently, various types of nanostructured TiO ₂ such as nanoparticles, nanotubes, nanowires, and nanorods have been prepared and adopted as electrode materials to improve the performances of DSSCs [21 - 25] . Tan et al. [26] synthesized TiO ₂ nanoparticles attached to TiO ₂ nanowires to increase the surface area and applied them as a DSSC electrode material. They reported that the resulting DSSC had a powerconversion efficiency of 8.6％, which is 2％ higher than the efficiency of the cell with a TiO ₂ nanoparticle electrode. Lei et al. [27] grew TiO ₂ nanotubes on Ti foil using the anodic oxidation method and prepared a cell that had a power-conversion efficiency of 8.07％. Therefore, it is believed that in addition to the requirement of a large surface area for dye absorption, a one-dimensional geometry of TiO ₂ that provides straight pathways for electron transfer is essential to improve the solar-cell performance. If TiO ₂ nanotubes are directly synthesized on fluorinated tin oxide (FTO) glass, which is not a commercially available technology at the current moment, higher cell performance might be achieved. Instead, researchers have focused on the direct growth on FTO glass of nanorods that possess characteristics similar to those of nanotubes [28 - 31] . In particular, rutile TiO ₂ has a very small lattice mismatch (＜2％) with FTO, and it is rather easy to grow rutile TiO ₂ on FTO with large aspect ratios [32 - 34] . So far, it has been reported that rutile TiO ₂ nanorods can be grown on FTO glass via hydrothermal synthesis [27 - 30] . For instance, Liu et al. [35] obtaineda DSSC power-conversion efficiency of 3％ using TiO ₂ nanorods that were grown to a length of up to 4 μm. In an advanced method, Wang et al. [36] attached TiO ₂ nanobranches to the nanorods to increase the surface area and achieved a DSSC power-conversion efficiency of 3.75％ even with nanorods that had heights of 3 μm. Nevertheless, it has been known that these forms of rutile TiO ₂ are inferior to anatase TiO ₂ in terms of DSSC power-conversion efficiency, and it is very difficult to grow anatase TiO ₂ nanorods vertically on FTO glass because of the large lattice mismatch (~19％) between anatase TiO ₂ and FTO [37 , 38] . Therefore, a TiO ₂ colloid solution must be spin-coated onto the FTO glass in order to synthesize the layer of nanostructured anatase- phase TiO ₂ [39] . In this study, instead of using the thin TiO ₂ seed layer, rutile TiO ₂ nanorods were synthesized on FTO glass, on which the dense anatase TiO ₂ nanostructures were fabricated using two different hydrothermal synthesis techniques in sequence. It was assumed that the electrons that were formed in the transparent anatase TiO ₂ layerwould be transferred to the FTO glass through the rutile TiO ₂ nanorods. In addition, the surfacesof the rutile TiO ₂ nanorods were treated with aqueous TiCl ₄ solutions that had varying concentrations of TiCl ₄ for different treatment durations to prevent electron recombination [40] . Rutile-anatase hybrid TiO ₂ nanostructures were prepared on FTO glass (TEC-8, 8 Ω/square) in two sequential hydrothermalsynthesis steps. The typical procedure is as follows: FTO glass was cleaned in a mixed solution of water, acetone, and 2-propanol (volume ratio = 1:1:1) in a sonication device for 30 min. After drying in ambient condition, the FTO glass was placed in a Teflon-lined stainless-steel vessel (100 mL). It was leaned against the vessel wall, with the conductive side facing down in a mixed solution containing 30 mL of de-ionized water, 30 mL of hydrochloric acid (36.5-38.0％, Sigma-Aldrich), and 0.5 mL of titanium butoxide (97％, Sigma-Aldrich). The stainless-steel vessel was then sealed and placed in a convective oven at 150 ℃ for 20 h. Following the procedure described above, rutile TiO ₂ nanorods were vertically synthesized on FTO glass. The resulting sample was then rinsed with de-ionized water and immersed in an aqueous solution of TiCl ₄ (0.2-0.4 M) and kept for 1-12 h at ambient temperature. For the second hydrothermal synthesis, the rutile-covered FTO glass was washed with de-ionized water and dried. Next, the glass was placed in a mixed solution containing 35 mL of H ₂ O, 5 mL of sulfuric acid (ACS reagent, 95.0-98.0％, Sigma-Aldrich), and 2 mL of titanium butoxide (97％, Sigma-Aldrich) in the Teflon-lined stainless-steel vessel. Again, the steel vessel was placed in the convective oven at 180 ℃ and kept for 12 h. The resulting sample was removed, washed with de-ionized water, and dried. The sample was then calcined at 400 ℃ for 2 h. Finally, the rutile-anatase hybrid TiO ₂ nanostructures were fabricated as a form of the anatase TiO ₂ layer covering the upper tips of the rutile TiO ₂ nanorods on FTO glass. Hereafter, each sample is referred to as [R][TiCl ₄ concentration]-[TiCl ₄ treatment time][A] The first letter R stands for rutile TiO ₂ , and the final letter A represents anatase TiO ₂ . For instance, to prepare the sample R0.2-12A, rutile TiO ₂ nanorods were grown on FTO glass and then subjected to TiCl ₄ (0.2 M) treatment for 12 h, during which anatase TiO ₂ was synthesized on the nanorods. The crystal structure of the anatase TiO ₂ nanostructures was characterized by X-ray diffraction (XRD Rigaku, 40 kV, 100 mA). The morphology of the samples was examined with a fieldemission scanning electron microscope (FE-SEM S-4700, Hitachi) and a transmission electron microscope (TEM JEM-2100F HR, JEOL Ltd.) with an accelerating voltage of 200 keV, point resolution of 0.23 nm, and STEM resolution of 0.2 nm. The rutile-anatase TiO ₂ electrodes were immersed in an ethanol solution of a ruthenium-complex (N719, Solaronix) for 24 h to complete the dye absorption. The FTO glass with 0.7 mm holes was coated with a 7 mM chloroplatinic acid (H ₂ PtCl ₆ , Aldrich) solution to serve as the counter electrode. The dye-absorbed TiO ₂ electrode and Pt counter electrode were clamped firmly together with 60 mm-thick surlyn. The electrolyte containing 0.03 M I ₂ , 0.05 M LiI, 1 M 1-methyl-3-propylimidazolium iodide, 0.1 M guanidine thiocyanate, and 0.5 M tert -butyl pyridine in acetonitrile- valeronitrile (85:15) was injected into the clamped electrodes. The Ⅰ-Ⅴ characteristics of the DSSCs were measured with a Keithley 2400 source meter using an AM 1.5 (100 mW/ cm ² ) solar simulator equipped with a 1 kW xenon arc lamp (Oriel, Newport). XRD analysis was carried out to determine the crystal structures of the prepared samples. The XRD patterns of the samples are shown in Figure 1 . SnO ₂ peaks appeared in all the patterns because FTO glass was used as the substrate. A comparison of the XRD pattern of sample R ( Figure 1(a) ) with the standard data in JCPDS 88-1175 confirmed the existence of rutile TiO ₂ , as shown by the (110) peak at 2θ = 27° and the (110) peak at 2θ = 54°. The XRD patterns of R0.2-1A ( Figure 1(b) ) and R0.4-12A ( Figure 1(c) ) show additional peaks at 2θ = 25°, 39°, 48°, 54°, and 55°, which match the diffraction peaks corresponding to the (101), (112), (200), (105), and (211) planes, respectively, of anatase TiO ₂ according to JCPDS 84-1286. Therefore, it is clear that we obtained samples containing both rutile and anatase TiO ₂ on FTO glass. As reflected by the samples’ names, sample R is the one without TiCl ₄ treatment of rutile TiO ₂ , while the other samples were treated with TiCl ₄ solution before anatase TiO ₂ was synthesized on the rutile TiO ₂ . Close examination of the XRD patterns of R and other samples revealed that the characteristic peaks of SnO ₂ shifted slightly to the right. The enlarged view of XRD patterns for SnO ₂ (101) peaks is shown in Figure 2 . The (101) peak intensities of the rutile-anatase TiO ₂ samples were much weaker than that of the rutile-only sample because anatase TiO ₂ covered the rutile TiO ₂ layers, limiting the depth of X-ray penetration. The extent of the peak shift depended on the concentration of TiCl ₄ but not on the TiCl ₄ -treatment duration. When a lower concentration of TiCl ₄ was applied, a longer exposure of the FTO glass surface to the TiCl ₄ solution was expected because the amorphous TiO ₂ deposition rate was slower than that with higher concentration of TiCl ₄ . Therefore, we believe that hydrogen chloride formed during the TiCl ₄ treatment penetrated the rutile TiO ₂ layer and attacked the exposed FTO surface, and some Sn atoms were replaced by Ti atoms that have a smaller atomic radius [41] . On the other hand, with a higher concentration of TiCl ₄ , the amorphous TiO ₂ formation took place faster and covered the FTO surface quickly, preventing further attack of hydrogen chloride. Another notable feature was that the characteristic peaks of the anatase TiO ₂ in R0.4-1A and R0.4-12A ( Figure 3(c) , (d) ) were located at positions that were slightly shifted to the left when compared to the same peaks of R0.2-1A and R0.2-12A ( Figure 3(a) , (b) ). The peak shift implies that there was a change in the lattice parameters [42] . Table 1 shows the lattice parameters calculated from the (004) and (200) peaks of anatase TiO ₂ in the samples. Even though the experimental conditions for the hydrothermal synthesis of anatase TiO ₂ on the rutile TiO ₂ layer were identical for all samples, the lattice parameters of each sample were varied. This phenomenon appears to be related to the concentration of the TiCl ₄ solution, i.e., the lattice parameter a of the samples obtained with low TiCl ₄ concentration appeared to be smaller than those of samples obtained with high TiCl ₄ concentration, while the values of the lattice parameter c of all samples were similar. In order to investigate this variation in the lattice parameters of the anatase TiO ₂ on rutile TiO ₂ , TEM analysis was carried out for rutile nanorods after TiCl ₄ treatment. Figure 4 shows the TEM images of the rutile TiO ₂ nanorods treated with TiCl ₄ at various concentrations and for different durations. When the concentration of TiCl ₄ was 0.2 M, the amorphous TiO ₂ layer was thin and uniform; and the thickness of the layer was affected by the treatment duration. On the other hand, when the concentration of TiCl ₄ was 0.4 M, the amorphous TiO ₂ layerwas much thicker than that obtained from the 0.2 M TiCl ₄ solution, and an additional bulge corresponding to the rutile TiO ₂ nanorods was observed. In addition, the thickness of the layer was proportional to the treatment time. Based on these observations,we believe that the less concentrated TiCl ₄ solution provided the thin amorphous TiO ₂ layer with the seeds for anatase TiO ₂ growth. On the other hand, the high-concentration TiCl ₄ solution provided more coarse and additional seeds apart from the surface of rutile TiO ₂ . Therefore, when a lower concentration of TiCl ₄ was used, the resulting anatase TiO ₂ tended to be packed and became denser towards the interior of the particles, creating a pressure that resulted in the smaller lattice parameter. Using a similar analysis approach, Zhang et al. [43] reported fabrication of CeO ₂ and BaTiO ₃ nanoparticle thin films in which the increased lattice parameter with decreasing particle size was explained by the negative effective pressure created by the competition between the long-range Coulomb attractive interactions and the shortrange repulsive interactions in ionic nanocrystals. Li et al. [44] also reported that the concurrent packed growth of rutile TiO ₂ physically squeezed the anatase TiO ₂ phase, resulting in a decrease in the lattice parameter. The SEM images in Figure 5 show that the rutile TiO ₂ nanorods were well grown like mowed grass, as reported by Liu et al. [35] . The concentration of the precursor and other chemicals were adjusted by trial and error until the proper density of nanorods was obtained. In the image of sample RA ( Figure 5(b) ), whose anatase TiO ₂ layer was deposited on the rutile TiO ₂ nanorods without TiCl ₄ treatment, it can be seen that the dense, powder-like anatase TiO ₂ grew and filled all the gaps between the nanorods. On the other hand, when the rutile TiO ₂ was treated with TiCl ₄ prior to the anatase TiO ₂ synthesis ( Figure 5(c-f) ), the anatase phase was confirmed to grow on the tips of the rutile nanorods, leaving the gaps between the nanorods empty [45] . It should be noted that these empty gaps played very important roles in the DSSC performance. When incident light came in through the anatase layer, this empty space allowed the light to reflect multiple times and remained for a longer period in the TiO ₂ structure to produce more electrons. According to the SEM images, the height of the rutile nanorods was strongly dependent on both the concentration of TiCl ₄ and TiCl ₄ treatment duration. The rutile nanorod layer in R0.2-1A was barely recognizable, while the thicker layer of the nanorods in R0.2-12A was clearly distinguishable and showed more empty space. This means that the height of rutile nanorods increased with longer TiCl ₄ -treatment duration. This relationship was also observed in R0.4-1A and R0.4-12A. On the other hand, treatment at a higher concentration of TiCl ₄ for the same amount of time contributed to the growth of the nanorod radius. It should be carefully considered that the rutile nanorods in sample R were shortened with TiCl ₄ treatment it has been reported that theTiCl ₄ solution has an etching effect on rutile nanorods [46 , 47] . It has also been reported that rutile nanorods grow with extended TiCl ₄ treatment [47] . These reports match our experimental results that showed the height of the nanorods decreased in the following order: R0.4-12A ＞ R0.2-12A ＞ R0.4-1A ＞ R0.2-1A. The power-conversion efficiency of a DSSC strongly depends on the geometric configuration and thickness of the TiO ₂ structure. As shown in the SEM image analysis, the six TiO ₂ structures on FTO glass showed clear differences in thickness and geometric configuration. The current density-voltage ( J - V ) responses for all solar cells assembled with the six samples are plotted in Figure 6 . The photovoltaic characteristics of the DSSCs were summarized in Table 2 . The fill factor ( FF ) and open circuit voltage ( V_OC ) showed small variations among DSSCs assembled with the samples. Even though the values of total thickness of the TiO ₂ nanostructures (4.5-5.0 μm) were very similar, the photovoltaic characteristics were clearly different. As expected, the cell prepared from the sample with only rutile nanorods on FTO glass (sample R) showed the lowest shortcircuit current density ( J_SC ) of 1.65 mA/cm ² with doubled power-conversion efficiency ( PCE ) of 0.81％. When anatase was deposited on the rutile nanorods without TiCl ₄ treatment (sample RA), J_SC increased to 4.38 mA/cm ² with doubled PCE. The cell assembled with R0.2-1A showed significantly increased values of J_SC and PCE compared to 7.76 mA/cm ² and 2.83％, respectively, for the cell assembled with RA. This implies the rutile structure under the anatase layer acted as an effective pathway for the electrons generated within the anatase layer. However, the cell with R0.4-12A that had a layer of longer rutile nanorods exhibited J_SC of 5.19 mA/cm ² and PCE of 2.08％, which are lower than those of the cell assembled with R0.2-1A. Even with the longer rutile nanorod layer, the low density of the nanorods appeared to limit the rate of electron transfer. As further evidence, the cell with R0.4-1A that had nanorods with similar height but at higher density showed much higher J_SC of 9.74 mA/cm ² and PCE of 3.45％. Finally, the cell with R0.4- 12A exhibited the highest J_SC and PCE of 10.3 mA/cm ² and 3.94％, respectively. Careful examination of the SEM image of R0.4-12A revealed that the nanorods were much thicker and taller than others with relatively high density. Therefore, it is reasonable to assume that the physical shape and size of the nanorods were the main factors that affected the performance of the DSSCs. The nanorods should have both high density and long length so that light can reflect back and forth under the anatase layer, producing electrons both inside the thick anatase layer and the anatase covering the rutile nanorods. In this study, rutile-anatase hybrid TiO ₂ nanostructures were prepared and characterized. DSSCs were assembled with the prepared TiO ₂ nanostructures and their photovoltaic characterizations were obtained. The rutile nanorods were found to act as effective pathways for the electrons produced in the anatase layer. It was confirmed that the rutile nanorods should have high length and relatively high density to increase PCE of the DSSC. The empty space between the nanorods was a very important factor of PCE because it facilitated the light reflection within the nanorod layer to producing more electrons.