D-sorbitol-induced phase control of TiO2 nanoparticles and its application for dye-sensitized solar cells

Using a simple hydrothermal synthesis, the crystal structure of TiO2 nanoparticles was controlled from rutile to anatase using a sugar alcohol, D-sorbitol. Adding small amounts of D-sorbitol to an aqueous TiCl4 solution resulted in changes in the crystal phase, particle size, and surface area by affecting the hydrolysis rate of TiCl4. These changes led to improvements of the solar-to-electrical power conversion efficiency (η) of dye-sensitized solar cells (DSSC) fabricated using these nanoparticles. A postulated reaction mechanism concerning the role of D-sorbitol in the formation of rutile and anatase was proposed. Fourier-transform infrared spectroscopy, 13C NMR spectroscopy, and dynamic light scattering analyses were used to better understand the interaction between the Ti precursor and D-sorbitol. The crystal phase and size of the synthesized TiO2 nanocrystallites as well as photovoltaic performance of the DSSC were examined using X-ray diffraction, Raman spectroscopy, field-emission scanning electron microscopy, high-resolution transmission electron microscopy, and photocurrent density-applied voltage spectroscopy measurement techniques. The DSSC fabricated using the anatase TiO2 nanoparticles synthesized in the presence of D-sorbitol, exhibited an enhanced η (6%, 1.5-fold improvement) compared with the device fabricated using the rutile TiO2 synthesized without D-sorbitol.

Many solution based methods have been reported for the synthesis of TiO 2 nanoparticles, such as sol-gel 10 , solvothermal 11 , and hydrolysis 12 etc. Among these, hydrothermal synthesis method has the advantages of providing mono-dispersed particles, controlled structural morphology, and phase homogeneity etc., at relatively low temperatures 13 . In the hydrothermal synthesis, various parameters are being suggested to affect the crystallinity as well as the size of the TiO 2 product. Zheng et al. proposed a dissolution-precipitation mechanism for TiO 2 formation in which the concentration of the TiCl 4 precursor was considered to determine the crystallinity of the TiO 2 product; anatase crystallites grew larger and transformed into rutile 14 . The thermodynamic stability is reported to depend on the particle size, and anatase phase of TiO 2 is more stable than rutile phase at particle diameters below approximately 14 nm 15 . In addition, the pH of the precursor solutions was also suggested to affect the growth mechanisms and thus crystal structures of the TiO 2 nanocrystals 16 . The acidic/alkaline conditions employed in the synthesis of TiO 2 nanoparticles were observed to affect the performance of DSSC 17 . However, determining how to control the conditions necessary to yield TiO 2 nanocrystals with a definite crystal shapes and surface orientations to meet the requirements of DSSC remains a crucial problem.
The aim of the present work is to prepare TiO 2 nanocrystals with pure anatase phase using a low-temperature (< 200 °C) hydrothermal method. We investigated the role of D-sorbitol as a complexing agent on the formation of anatase TiO 2 . D-sorbitol was selected because of its non-toxic biological origin and environmentally friendly nature, low cost, and ability to assist complex formation. In the present study, we observed that reactions proceeds even in the absence of D-sorbitol however, the resultant TiO 2 product was pure rutile rather than anatase. The driving force for the anatase TiO 2 synthesis was studied from the complex species of D-sorbitol with Ti cations coupled through hydroxyl ions in the solution. The solution approach offered the possibility to control the reaction pathways on a molecular level and enabled the synthesis with a well-defined crystal polymorph and morphologies without impurities. The effects of the preparation conditions on the crystal phase of the TiO 2 nanocrystals as well as the photovoltaic performance of DSSC equipped with the prepared TiO 2 nanocrystals were studied. It was observed that the nanocrystallites of anatase TiO 2 prepared using the hydrothermal method exhibited comparable/enhanced DSSC performance compared with the commercial anatase/rutile TiO 2 . The present method is a facile single-step process, and TiO 2 nanoparticles prepared in this present work are chemically, environmentally and mechanically stable for several days, justifying their long-term uses.

Results and Discussion
Reaction mechanism. In context with the report of Gopal et al., the experimental Ti-O phase diagram indicates that anatase is more stable than rutile at room temperature and atmospheric pressure 18 . Both anatase and rutile TiO 2 consist of TiO 6 2− octahedra, which share edges and corners in different manners. In the rutile case, two opposite edges of each octahedron are linked through a corner oxygen atom, forming linear chains of octahedra. In contrast, anatase exhibits no corner sharing but instead has four edges shared per octahedron. The anatase structure can be viewed as zigzag chains of TiO 6 2− octahedral, linked to each other through edge-sharing bonding 19 . Because anatase has more edge sharing, and the interstitial spaces between octahedra are larger, it is less dense than rutile (the densities of rutile and anatase are 4.26 and 3.84 gcm −3 , respectively) 20 . It has been accepted that when the four-fold Ti precursor ([TiCl 4 ] or [Ti(ROH) 4 ]) reacts with water, the coordination number of Ti 4+ increases from four to six through its vacant d-orbitals to accept oxygen lone pairs from nucleophilic ligands 21 . These six-fold structural units undergo condensation and become the octahedra that are incorporated into the final precipitate structure. The octahedra agglomerate through corner and edge sharing during the condensation reactions 22 . During the particle agglomeration, the acidity of the reaction medium is suggested as a critical factor for the hydrolysis of TiCl 4 in aqueous solution 23 . Under highly acidic conditions, the agglomeration of rutile TiO 2 could be attributed to hydrogen bonding among the protonated nanocrystallites 24 . In addition, because of the lower surface energy of anatase compared with that of rutile, selective formation of the anatase phase is favored under weak acid conditions as polycondensation of Ti(OH) n Cl 6-n species is weak (slow) 25 . Cheng et al. also explained the difference in the crystallization of anatase and rutile TiO 2 by the hydrolysis of TiCl 4 in an aqueous solution using ligand field theory 26 , and the crystallization occurred via dehydration between partially hydrolyzed Ti(OH) n Cl 6-n complexes.
Adapting the previous studies, in the present study, the reaction mechanisms illustrated in Fig. 1 is proposed depending on the presence of D-sorbitol. When the effect of D-sorbitol on crystallization was considered, one can suppose that D-sorbitol anions substitute for the chlorine anions during the hydrolysis process to form   For more confirmation of complex formation, we conducted 13 C NMR spectroscopy measurements over pure D-sorbitol and TiCl 4 -D-sorbitol using D 2 O as a reference solution. The 13 C NMR spectra were used to study the interaction of the metal and D-sorbitol complexes. Surface morphology and structural analysis. Next, the morphology and crystal structure of the synthesized TiO 2 nanocrystals were compared depending on the usage of D-sorbitol. The FE-SEM images demonstrate the uniformity of the synthesized TiO 2 consisting of well-interconnected nanocrystallites. The average diameter decreased from ~60 nm to ~20 nm when D-sorbitol was present in the precursor solution ( Fig. 3a,b), signifying capping capability of D-sorbital for an over growth. We observed a consistent size difference in the HR-TEM and BET data (see section below). In addition, the HR-TEM images of the as-prepared TiO 2 nanocrystallites confirm their high crystallinity regardless of the presence of D-sorbitol in the precursor solutions. However, the measured lattice parameters for the TiO 2 nanocrystallites changed, implying that different crystal phases were synthesized. The lattice parameters were measured from the HR-TEM images (Fig. 3c,d) and the positions of the main diffraction peaks in the SAED patterns (Fig. 3e,f). The distances between two adjacent lattice planes, for two cases, were 0.32 nm [in good agreement with the (110) crystallographic plane of rutile (Fig. 3c,e)], and 0.36 nm [in good agreement with that of (101) for anatase TiO 2 (Fig. 3d,f)] (hence forth, called rutile TiO 2 and anatase TiO 2 , respectively) 33 . The (101) crystal faces of anatase have lower surface energy and are expected to be more stable than the other faces 34 , and our HR-TEM images also demonstrated the strongest ring pattern of (101) in SAED spectrum.
The change in the crystal structures of the synthesized TiO 2 nanocrystals was also confirmed by XRD patterns (Fig. 4a) and Raman spectrum (Fig. 4b), both of which consistently demonstrated that in the presence of D-sorbitol only anatase TiO 2 is obtained; otherwise, rutile is favored. The observed XRD peaks were well attributed to rutile TiO 2 (JCPDS no. 870710) and anatase TiO 2 (JCPDS no. 86-1156). The observation of strong XRD peaks was indicative of the good crystallinity of the as-prepared TiO 2 . In table 1, crystal size, calculated using Scherrer formula, is presented. As-prepared anatase TiO 2 nanocrystals were stable up to 700 °C and transformed into rutile as the calcination temperature increased to 800-1000 °C [ Fig. S1a-d in the electronic supplementary information (ESI)]. These results indicate that compared with anatase, rutile is the thermodynamically stable phase of TiO 2 . Raman spectroscopy also corroborated the presence the rutile and anatase phases of TiO 2 . In Fig. 4b, the Raman shifts at 143, 235, 447, and 612 cm −1 are attributed to the B 1g , two-phonon scattering, E g , and A 1g modes of the rutile phase, respectively 35 . The four Raman shift peaks at 144, 400, 514, and 638 cm −1 are attributed to the E g , B 1g , A 1g , and E g symmetries of the anatase phase, respectively 36 . In addition, as presented in Fig. S2(a-c), XPS analysis was used to investigate the chemical Ti 4+ state of both the rutile and anatase TiO 2 phases; more or less similar electronic states and chemical compositions were observed on the surface. Regardless of the preparation procedure (with or without D-sorbitol), the amounts of the adsorbed residues on the two TiO 2 surfaces were similar.
Hydrolysis rate estimation. The dynamic light scattering (DLS) technique was used to study the effect of D-sorbitol on the hydrolysis process of TiCl 4 at room temperature. The acidity (pH = 0.6) of the solutions for the DLS measurements was the same as those of the starting TiCl 4 (1 M) and D-sorbitol solution to validate the comparisons. The resolution of the DLS apparatus was 2 nm. As observed in Fig. 5a, in the absence of D-sorbitol, TiCl 4 in aqueous solution could hydrolyte to form particle agglomeration with a size distribution of ~69 nm. While in the presence of 0.05 M D-sorbitol the size distribution was ~25 nm, as observed in Fig. 5b. Moreover, Fig. 5c shows that for 0.1M D-sorbitol, the size distribution is decreased to 13 nm. With a further increase in the D-sorbitol concentration up to 0.15 M, no particle formation occurs. The systematic decrease in particle-size is an indication of an agglomeration-free reaction, supporting the conclusion that the interaction of D-sorbitol with TiCl 4 prevents the rapid hydrolysis of TiCl 4 . The formation of small-sized anatase nanocrystallites as embryos could be due to the inhibition of crystal growth by the coordination of D-sorbitol anions. Consistent results were obtained by Ambade et al. for ZnO nanorods 37 .
To better understand the hydrolysis reaction, we kept both samples (TiCl 4 solution in aqueous medium and TiCl 4 -0.1 M D-sorbitol) at room temperature for more than ten days (Fig. 5d). The TiCl 4 solution in aqueous medium started to become turbid (white) with particles as sediment after two days. These primary crystallites subsequently coalesced, and a precipitate settled slowly. However, despite ~13 nm particle-size, the TiCl 4 -0.1 M D-sorbitol solution was clear and transparent until more than one month. This conclusion was also supported by the DLS measurement, where the D-sorbitol anion could bond to Ti cations by preventing the fast hydrolysis at room temperature. It is believed that the slow hydrolysis (as the solution is clear and transparent) plays a critical role in developing small-sized particles, which eventually can help in the phase transformation process from rutile to anatase. However, to obtain anatase TiO 2 from the D-sorbitol-containing solution, an adequate temperature is necessary to initiate the nucleation process followed hydrolysis 38 .
Surface area and pore-size analysis. The specific surface area and pore-size distribution of both as-prepared TiO 2 nanostructures were characterized using nitrogen gas adsorption. A type-IV isotherm and H1-type hysteresis loop were confirmed for both TiO 2 nanostructures (Fig. 6), suggesting macroporosity in rutile and mesoporosity in anatase TiO 2 39 . The specific surface area, calculated using the standard multi-point BET method, was 14.28 m 2 g −1 for rutile TiO 2 , which was only one-third to that of the anatase TiO 2 (47.77 m 2 g −1 ). The as-prepared TiO 2 exhibited a narrow pore-size distribution centered at 60.28 nm for rutile TiO 2 and 16.79 nm for anatase TiO 2 (inset of Fig. 6). The performance of DSSC depends on the type of porosity, particle/pore size and charge transport properties of the TiO 2 photoanode 40 . Generally, smaller nanoparticles have a larger surface area but a shorter electron diffusion length, whereas larger nanoparticles have a longer electron diffusion length  but a smaller surface area 41 . Because of the multiple factors, an optimal particle-size is required to achieve high solar-to-electrical power conversion efficiency (η ). For example, Cao et al. concluded that a particle-size of 15 nm can be the best among 10-20 nm sized samples for superior DSSC application 42 .

DSSC performance.
To understand the DSSC performance depending on the preparation methods, we first measured UV-Vis absorption spectra of dye-adsorbed photoanodes (Fig. 7a). All of the photoanodes exhibited a wide absorbance in the visible region (centered at approximately 530 nm); however, the prepared anatase TiO 2 photoanode exhibited higher absorption compared with the commercial (100% anatase, for more details please see Experimental section) and rutile TiO 2 electrodes, which is consistent with the order of the dye adsorption amounts on the TiO 2 surfaces ( Table 1). The different crystallinity, smaller particle-size, and higher surface area of the prepared anatase TiO 2 could increase the dye adsorption, which is evident from the enhanced  UV-Vis absorption. The performance of the DSSC was tested under illumination of simulated AM1.5 G solar light (100 mW cm −2 ), and the J-V characteristics are presented in Fig. 7b for each individual cell. In Table 1, the crystal phase and photovoltaic performance parameters are summarized. The short-circuit current density (J SC ) of the anatase TiO 2 photoanode (12.19 mA cm −2 ) was 1.5 times greater than that of the rutile TiO 2 electrode (7.96 mA cm −2 ). In addition, the V oc of the anatase TiO 2 electrode was similar but increased by 0.02 V compared with that of the rutile electrode. Therefore, the η of the cells made of anatase TiO 2 was 1.5 times higher (η = 6%) than that for rutile TiO 2 (η = 3.8%), which is mainly attributed to the enhancement of J SC . The standard deviation of the photovoltaic parameters was calculated to validate the accuracy and reproducibility of the DSSC performance of the TiO 2 nanocrystallites (Fig. S3). The remarkable performance of the DSSC fabricated with the anatase TiO 2 electrode might originate from its crystal phase, morphology, and high electrical conductivity and mobility (Table S1) 43 . Contrary, due to the presence of several stacking faults and dislocations, electrode with rutile TiO 2 nanocrysyallites demonstrated lower conductivity and low dye intake capacity and thereby, smaller light harvesting capacity and lower DSSC performance 44 . Generally, smaller particles provide more active sites for dye adsorption and reaction in DSSC because of the larger specific area, leading to higher photo-to-electric power conversion efficiency. Moreover, upon comparison with the commercial TiO 2 electrode we observed that the crystal phase is a critical factor to achieve enhanced η for DSSC, which again indicates the importance of crystal phase control in TiO 2 synthesis. Our preparation method revealed that D-sorbitol can successfully control the crystal phase of TiO 2 to achieve high performance of DSSC.
To further explore the effects of the properties of TiO 2 photoanodes on the performance of corresponding DSSC, EIS measurements were performed. Fig. 7c presents Nyquist plots of the three cells (i.e., anatase, commercial, and rutile TiO 2 ) measured at a forward bias of V oc . Two semicircles, including a small one at higher frequency and a large one at lower frequency, are observed in the plots. The small semicircle is assigned to the charge-transfer resistance (R 1 ) and the capacitance (CPE 1 ) at the platinum counter electrode/redox electrolyte interface, whereas the larger semicircle is attributed to the recombination resistance (R 2 ) and chemical capacitance (CPE 2 ) at the TiO 2 /dye/redox electrolyte interface 45 . Therefore, the size of the second semicircle (the value of R 2 ) is very important to understand the changes in the photoanode. Large difference in the R 2 values is observed between the rutile and anatase TiO 2 photoanodes. The anatase TiO 2 photoanode exhibited a smaller R 2 value (19.7 Ω) than the rutile TiO 2 photoanode (31.0 Ω), indicating faster (hole) generation and transport as well as a slower electron-hole recombination rate. The electron lifetime (τ ) was calculated according to the equation τ = (1/2π f max ), where f max is the maximum frequency of the mid-frequency peak 46 . The τ values, estimated from Bode phase plots, were 1.59 × 10 −4 , 2.24 × 10 −4 , and 2.18 × 10 −4 ms for rutile, anatase, and commercial anatase TiO 2 , respectively (Table S2). For anatase TiO 2 , the higher τ value was due to the reduced charge transfer resistance and decreased electron recombination, enabling more efficient electron transfer with an enhancement of the device performance. In addition, the decay of V oc was used to reflect the regression of the electron density in the conduction band of the photoanodes as it is widely used as a kinetic parameter, which contains useful information about the rate constant of the electron transfer process in DSSC [47][48][49] . The τ values (Fig. 7d) were calculated by fitting the photovoltage decay plots obtained from the V oc decay measurements and by applying an equation developed by Bisquert et al. 50 . The higher τ value for the anatase TiO 2 photoanode implied a lower charge recombination rate and improved electron transfer efficiency compared with commercial and rutile TiO 2 , which is consistent with the impedance results and leads to an improvement in the DSSC performance.

Conclusion
During hydrothermal growth of TiO 2 , D-sorbitol was demonstrated to be a crystal-phase-controlling agent. As-prepared TiO 2 had a rutile crystal phase when prepared via the hydrolysis of the TiCl 4 precursor in an acidic environment, whereas pure anatase TiO 2 was obtained when D-sorbitol was added into the precursor solution. The intermediate complex formation between Ti ions and D-sorbitol molecules was recorded using FT-IR and 13 C NMR spectroscopy of anatase TiO 2 . The DLS measurements supported the conclusion that the interaction between D-sorbitol and TiCl 4 prevents its rapid hydrolysis, resulting in the systematic decrease in the TiO 2 particle-size as the concentration of D-sorbitol increased. We expect that the slow hydrolysis plays a critical role for small-size particle formation and assists in the anatase phase transformation. The photovoltaic performances of the rutile and anatase TiO 2 polymorphs were compared. Solar-to-electrical power conversion efficiency of the DSSC fabricated using the pure anatase TiO 2 electrode was 6.0%, which was 1.5 times higher than that prepared using the rutile TiO 2 (3.7%) electrode prepared under the same experimental conditions and comparable (5.8%) to commercial TiO 2 . Our study demonstrated that comparable DSSC performance achieved for anatase TiO 2 prepared using a simple hydrothermal method might arise from its phase, crystal-size, morphology, surface orientation, and high electrical conductivity and mobility.
Hydrothermal synthesis of TiO 2 nanostructures. TiO 2 nanostructures were prepared using a simple one-step hydrothermal method. In the standard experimental procedure for the synthesis of the anatase phase, 5 mL of 1 M TiCl 4 and 2.5 g D-sorbitol, were mixed in 80 mL of deionized (DI) water (Milli-Q water; 18.2 MΩ .cm). The mixture was constantly stirring for 10 min before transferring into a 100-mL Teflon-lined stainless-steel autoclave. The autoclave was sealed and maintained at 150 °C for 24 h, followed by cooling to room temperature. The resulting yellowish-white product was centrifuged at 8000 rpm for 10 min and washed several times with deionized water and ethanol (1:1 volume ratio) to remove any undesired impurities. The product was heated at 550 °C for 1 h to obtain the white powder of TiO 2 . The same experimental conditions were applied for the synthesis of rutile TiO 2 except D-sorbitol.
Characterizations. The surface morphologies of both the rutile and anatase TiO 2 nanostructures were examined using field-emission scanning electron microscopy (FE-SEM, Nova NanoSEM200-100 FEI) images. The phases of the TiO 2 photoanodes were confirmed by X-ray diffraction (XRD) spectra (XRD-6000, Shimadzu, Japan) obtained at Cu-Kα radiation (λ = 0.1542 nm). Phase analysis was additionally performed using a Raman microscope (Renishaw, inVia Raman microscope, UK) to corroborate the formation of rutile and anatase TiO 2 phases. The laser beam (λ = 532 nm) was focused using a lens to produce a spot on the photoanode. Fourier-transform infrared (FT-IR) spectroscopy was measured from 500 to 4000 cm −1 using an IR spectrometer (Nicolet iS10, Smart MIRacle, Thermo Scientific). The high resolution transmission electron microscopy (HR-TEM) and selected area electron diffraction (SAED) measurements were performed using a FEI TECNAI G2 20 S-TWIN equipped with a LaB6 cathode and a GATAN MS794 PCCD camera. The micrographs were obtained at an acceleration voltage of 200 kV. The powders of TiO 2 nanocrystals were suspended in ethanolic solutions separately and dropped onto a Formvar/carbon, 200 mesh TH, copper grids before HRTEM measurements. X-ray photoelectron spectroscopy (XPS) spectra were acquired using a PHI 5000 Versa Probe (Ulvac-PHI) using a monochromatic Al Kα X-ray source (1486.6 eV). The data were collected from a spot-size of 100 × 100 μ m 2 . The carbon 1s peak (284.6 eV) was used as a reference for internal calibration. The UV-Vis absorption spectra of dye-adsorbed TiO 2 photoanodes were recorded using a Varian Cary 5000 spectrophotometer. To quantify the amounts of dye adsorbed onto the TiO 2 photoanodes, the dye molecules were desorbed by dipping in 0.1 M NaOH solution (ethanol and water at a 1:1 ratio) for 24 h at room temperature. The specific surface area was measured using Brunauer-Emmett-Teller (BET) technique (Belsorp II, BEL Japan INC). The dynamic light scattering (DLS) technique (Photal Otsuka electronics ELSZ-1000 instrument) was used to understand the particle-size variation. 13 CNMR spectra were measured (Bruker 400-MHz FT-NMR, D 2 O) with δ and values from large to small. Fabrication and evaluation of DSSC. TiO 2 paste was prepared by mixing 1.0 g TiO 2 powder, 3.5 g α -terpineol, and 0.5 g ethyl cellulosein ethanol (3.0 mL) and acetic acid (0.2 mL) solvent and stirring for 24 h to form homogeneous slurry, separately for each TiO 2 phase. TiO 2 colloid paste was spread over the FTO substrate via a doctor blade technique with adhesive tape as a spacer. The substrate was sintered at 450 °C for 30 min in air, which resulted in an approximately 10-μ m-thick TiO 2 porous film. The dye sensitizer used in this work was cis-di(isothiocyanato)-bis-(2,2-bipyridyl-4,4-dicarboxylato)ruthenium(II)bis-tetrabutyl ammonium (so-called Scientific RepoRts | 6:20103 | DOI: 10.1038/srep20103 N-719, 0.5 mM in a mixed solvent of acetonitrile and tert-butanol in a volume ratio of 1:1), which was used as received from Solaronix. DSSC were assembled by adding an electrolyte solution (0.6 M tetrapropyl ammonium iodide, 0.1 M iodine, 0.1 M lithium iodide, and 0.5 M 4-tert-butylpyridine in acetonitrile) between the dye-sensitized TiO 2 photoanode and a platinized conducting-glass electrode. The two electrodes were clipped together, and a cyanoacrylate adhesive was used as a sealant to prevent leakage of the electrolyte solution. A solar simulator (150-W Xe lamp, Sun 2000 solar simulator, ABET 5 Technologies, USA) equipped with an A.M. 1.5G filter was used to generate simulated sunlight, and the intensity of 1 sun (100 mW cm −2 ) was calibrated with a reference silicon solar cell. The photocurrent density-applied voltage (J-V) spectra of various TiO 2 photoanodes were obtained with the aid of a Keithley 2400 source meter. The electrochemical impedance spectroscopy (EIS) measurements of the TiO 2 photoanodes were recorded using a two-electrode system by a potentiostat (IviumStat Technologies, Netherland) in the frequency ranges of 150 kHz to 0.1 Hz.