MOF-derived Co2+-doped TiO2 nanoparticles as photoanodes for dye-sensitized solar cells

Facile synthesis and application of nano-sized semiconductor metal oxides for optoelectronic devices have always affected fabrication challenges since it involves multi-step synthesis processes. In this regard, semiconductor oxides derived directly from metal–organic frameworks (MOFs) routes have gained a great deal of scientific interest owing to their high specific surface area, regular and tunable pore structures. Exploring the application potential of these MOF-derived semiconductor oxides systems for clean energy conversion and storage devices is currently a hot topic of research. In this study, titanium-based MIL-125(Ti) MOFs were used as a precursor to synthesize cobalt-doped TiO2-based dye-sensitized solar cells (DSSCs) for the first time. The thermal decomposition of the MOF precursor under an air atmosphere at 400 °C resulted in mesoporous anatase-type TiO2 nanoparticles (NPs) of uniform morphology, large surface area with narrow pore distribution. The Co2+ doping in TiO2 leads to enhanced light absorption in the visible region. When used as photoanode in DSSCs, a good power conversion efficiency (PCE) of 6.86% with good photocurrent density (Jsc) of 13.96 mA cm−2 was obtained with the lowest recombination resistance and the longest electron lifetime, which is better than the performance of the pristine TiO2-based photoanode.


Results and discussion
MIL-125(Ti) MOF was fabricated through a solvothermal process reported in previous literature 48 . Figure 1a shows the powder X-ray diffraction (PXRD) patterns of MIL-125(Ti), cobalt incorporated MIL-125(Ti), and the calcined MOF derived pristine and doped TiO 2 . Before the calcination process, the diffraction peaks obtained at 2θ = 6.7°, 9.5°, 9.7°, 11.5°, and 16.6° respectively; which are all in good agreement with those reported earlier for MIL-125 (Ti) 50,51 . After calcination, the samples showed crystalline peaks with the complete transformation of MOF precursor to pure metal oxide phase 52,53 . The observed diffraction peaks of calcined samples at 25.53°, 38.54°, 48.10°, 55.04°, 68.8° and 75.2° were indexed to (101), (004), (200), (105), (211), and (215) cubic crystal planes of anatase TiO 2 according to the JCPDS card number 21-1272. The obtained XRD patterns of all cobalt doped samples almost coincide with that of pure anatase TiO 2 without any peaks related to rutile or to metallic cobalt or cobalt oxide, endorsing that the anatase phase is not effected upon Co doping. A slight shift of the most intense (101) peak towards the lower 2θ values was identified for all doped samples, which is ascribed to be the expansion of the TiO 2 unit cell due to the incorporation of Co 2+ . The ionic radius of Co 2+ (Co 2+ = 0.79 Å) is larger than that of Ti (Ti 4+ = 0.745 Å) leading to an expansion TiO 2 crystal with an increase of the lattice parameters. Also, a change in FWHM of (101) peak was observed after doping, which indicating the modification in the local structure around Ti 4+ after Co doping. This confirms the successful incorporation of Co in TiO 2 lattice. Apart from these, a small peak can be observed around 30.6° for the doped samples with the gradual increase in intensity (highlighted part in Fig. 1a). It is also inferred that the spectra obtained are devoid of any impurity during the solvothermal synthesis procedure. The obtained low-intensity peaks are attributed to the small size of the TiO 2 nanoparticles formed.
The incorporation of cobalt into the TiO 2 crystal lattice can influence the bandgap. This variation in bandgap after the cobalt incorporation can be clearly identified using diffuse reflectance spectroscopy, which is given in Fig. 1b. The full coverage of the UV-vis spectral range can be evident. In comparison to the pristine sample, Co 2+ ion-doped TiO 2 samples exhibited additional peaks in the visible region (450-800) in the sub-bandgap region. Doping has a direct influence on the optical properties of TiO 2 , which extended light absorption towards the visible region due to the formation of sub-bandgap states between the conduction and valence bands with oxygen vacancy sites 54 . The UV-vis diffuse reflection spectrum of pristine TiO 2 is flat in the visible (400-800 nm) region. However, the reflection decreases drastically upon doping with cobalt 55 . The bandgap energy is determined from diffuse reflectance spectra following the theory of P. Kubelka and F. Munk presented in 1931 (Fig. 1c) 56 . Accordingly, a gradual decrease of the optical bandgap from ~ 3.3 to ~ 2.9 eV was obtained. The UV-Visible www.nature.com/scientificreports/ light response of the prepared samples was also studied by optical absorption spectroscopy of the samples in methanolic solution as in supplementary Figure S1. All of the doped samples showed much enhancement absorption compared to the pristine sample, which indicates the enhanced optical response of the samples after cobalt doping. N2 adsorption-desorption isotherms were measured to investigate the surface area (Fig. 1d), and pore size distribution (Fig. 1e) of all samples. The samples showed type IV isotherm curves. The BET surface area and the pore volume of undoped TiO 2 were obtained as 118 m 2 g −1 and 0.32 cm 3 g −1 , respectively. A narrow pore size distribution around 5-6 nm was observed for TiO 2 from the BJH analysis. All the doped samples showed slightly less surface area compared to undoped samples; 10-Co-TiO 2 (92 m 2 g −1 and 0.30 cm 3 g −1 ) 30-Co-TiO 2 (90 m 2 g −1 and 0.33 cm 3 g −1 ) and 50-Co-TiO 2 (84 m 2 g −1 and 0.29 cm 3 g −1 ). Thus, the samples exhibit a relatively uniform mesoporous structure.
The morphological features of the sample before and after cobalt doping were characterized by SEM and TEM analysis. Figure 2a-d shows the SEM images of the doped and undoped samples. The undoped sample (Fig. 2a) depicted disc-like morphologies consisting of numerous nanoparticles with a diameter of 200-300 nm. Upon doping, the material showed drastic changes for both size and morphologies with very rough surfaces consisting of orderly arranged numerous nanoparticles of spherical shape with average sizes below 30 nm (Fig. 2b,c). At a higher doping concentration of cobalt, a more porous structured rough surface of the sample was perceived (Fig. 2d). The elemental identification of the sample was confirmed by EDS spectroscopy, and the spectra are given in supplementary Figure S2.
Further insights into the morphological and crystalline properties of samples were obtained from HR TEM analysis given in Fig. 3a-f. The pristine TiO 2 particle was found to be of an average size of 0.5 µm (Fig. 3a,b). The SAED pattern (Fig. 3c) shows clear, distinct rings confirming the polycrystalline nature. The obtained rings in SAED pattern were indexed (101), (200), and (201) planes of tetragonal anatase TiO 2 by calculating d-spacing. The result obtained is corroborated with the planes obtained in the XRD analysis of phase pure anatase TiO 2 . For undoped TiO 2 , the thickness of the particle observed was too high; as a result of this, the lattice fringes were not identified in HRTEM. However, the doped sample displayed (Fig. 3d) particle size of ~ 25 nm with clear lattice fringes as shown in Fig. 3e. HRTEM images displayed clear lattice fringes with d-spacing of 3.5 Å (inset figure of Fig. 3e) that matches (101) plane of tetragonal anatase phase of TiO 2 ascertaining the preferable crystal growth plane is (101), that is corroborated with the maximum intense peak obtained in XRD 57 . Based on the FESEM and HR-TEM studies, the crystal growth mechanism can be explained. The formation of the precursor MOF (MIL125 (Ti)) occurred with eight TiO 6 octahedra (oh) units attached by terephthalate linkers which results in the formation of a three-dimensional (3D) microporous network. When the solvothermal reaction proceeds, the solvent ions readily enter into the microporous channel and quickly react with MIL-125 (Ti) to form TiO 2 nuclei and trigger the crystal growth process. During the crystal growth, the solvent ions prefer to adsorb on the (001) crystal faces of anatase TiO 2 . Since the [001] growth direction is inhibited, other growth directions such as [100] are enhanced, resulting in nano-disk morphologies of TiO 2 . A density difference occurs between MIL-125 (Ti) and TiO 2 that can readily generate cracks in the MIL-125 (Ti) crystal. Thus the MOF precursor www.nature.com/scientificreports/ breaks into pieces. The presence of Co 2+ ions affected the crystal growth pattern and resulted in a particle-like TiO 2 NCs 58 . Upon increasing the concentration of dopants the agglomeration of NCs occurs which gives more open morphological structures made of small particle like NCs. Raman spectroscopic analysis was carried out to get more insight into the crystallinity, dopant-induced structural variations and also to understand the molecular vibration properties of all TiO 2 samples, and the spectra are shown in Fig. 4. From group theory, anatase TiO 2 has tetragonal structure and has six Raman active modes, A 1g + 2B 1g + 3E g , respectively 59 . E g peak arises by O-Ti-O symmetric stretching vibration in TiO 2 , whereas B1g by O-Ti-O symmetric bending vibration and A 1g appears due to O-Ti-O anti-symmetric bending vibration. The peaks centered at 144, 398, 518, and 641 cm −1 are characteristic to E g (1), B 1g , A 1g, and Eg modes, respectively. No Raman modes corresponding to the rutile phase were observed. The obtained spectra of the as-synthesized samples also showed the archetypal tetragonal form of phase pure anatase TiO 2 , which corroborates the obtained outcome from XRD. The intense E g (1) mode indicates the long-range order in the crystal. Upon doping with cobalt, the E g (1) peaks of all doped samples show a blue shift and a gradual decrease in intensity. This peak broadening, blue shifting as well as weakening of peak intensity is owing to the non-stoichiometry created owing to oxygen vacancy and phonon confinement effect due to the nanoscale size effect of the crystallites. The breakdown of long-range translational crystal symmetry occurs by the incorporation of Co 2+ . As a result, oxygen vacancies were introduced, which results in the weakening of the Raman signals. Thus, the Raman results obtained were found to be dependable with the XRD data.
The thermal stability of the samples was studied using TGA analysis, and the resultant data are shown in supplementary Figure S3. Maximum weight loss was obtained for the MIL-TiO 2 sample before calcination, which is possibly due to the existence of residual carbon present in the sample 60,61 . All the calcined samples (i.e., pristine and doped TiO 2 ) showed significantly less weight loss as well as high thermal stability. Also, it is observed that there is almost no weight loss above 250 °C, signifying that all the organic moieties were removed after the annealing process. It is noteworthy to mention that a slight weight loss ~ 2 wt.% observed below 250 •C is due to water adsorption on the sample surface.
The chemical composition, probability of secondary phases, oxygen vacancy, and oxidation state of surface elements were revealed by X-ray photoelectron spectroscopy (XPS) of the synthesized materials. Figure 5a-f shows the XPS spectrum of TiO 2 , and 30-Co-TiO 2 , respectively. Figure 5a is the survey spectrum of both the samples, which showed the characteristic peaks of Co, Ti, and O confirms the chemical composition in materials. The status of surface elements was further examined by the high-resolution XPS spectra; Fig. 5b presents the XPS spectrum of Ti 2p with Ti 2p 3/2 and Ti 2p 1/2 located at 458.25 and 463.97 eV, respectively, which matches the characteristic peaks of Ti 4+ -O bonds 62 .
From Fig. 5d of cobalt core-level spectra, peaks at binding energies of 780.64 eV of Co 2p 3/2 are observed, which corresponds to the oxidation state of + 2 of cobalt in TiO 2 . Another peak observed around 796.59 eV corresponds to Co 2p 1/2 . The spin-orbit splitting value of 15.95 eV, as well as the peak broadening, clearly supports   Figure 6 elucidates the J-V curves of the fabricated DSSCs. Table 1  MOF route-derived TiO 2 showed enhanced PCE of 6.06% with improved J sc and V oc of 10.83 mA/cm 2 and 0.799 V, respectively, however, the FF is reduced slightly to 70.05%Upon doping with 10 mg of cobalt,    www.nature.com/scientificreports/ TiO 2 sample. From the obtained results, it is understood that the FF of DSSCs fabricated with doped samples were decreased compared to the reference and pristine TiO 2 based DSSCs. This can be explained as follows. The FF of a DSSC device mainly depends on the morphology and thickness of the photoanode film. Increased thickness of photoanode results in a high fill factor. However, the fill factor drops if the recombination of the charge carriers is more. Here, all the fabricated photoanodes in our study, including the reference photoanode, has the average thickness of 10-13 µm. Therefore, the difference in fill factor of DSSCs between the doped and reference photoanodes is due to the morphological difference, which can be clearly understood from the SEM and TEM characterization. Also, the BET studies showed the highly porous nature of the doped samples, which causes the excess of electrolyte diffusion into the photoanode film resulting in a high recombination rate of photogenerated electrons and subsequent reduction in FF 64,65 . The fabricated DSSC cells were further analyzed to understand the electron lifetime and recombination characteristics with the help of the electrochemical impedance spectroscopic technique at a frequency range from 10 −1 to 10 5 Hz with an alternative current amplitude of 10 mV. The obtained Nyquist and Bode phase plots of pure and doped samples under AM 1.5 illumination were shown in Fig. 7a,b, respectively. In the Nyquist plot (Fig. 7a), two semicircles were obtained for all the samples. In which the small semicircle in the high-frequency region corresponds to the redox as well as electron transfer reaction at the Pt counter electrode-electrolyte interface 24 . The other large semicircle at the low-frequency region is related to the accumulation/transport of the photoinjected electrons within the photoanode film and the electron transfer across the TiO 2 /dye/electrolyte interface. This is considered as the characteristic peaks, which give a clear idea of the recombination behavior of electrons through the semiconductor oxide layer. The EIS curve is fitted using the equivalent circuit and the corresponding extracted parameters were given in Table 1. The sheet resistance (Rs) values obtained for all the devices were found to be similar and is ~ 20 Ω/cm 2 . The charge transfer resistance at the Pt CE interface obtained was below 1 Ω/cm 2 for all the devices. The DSSCs fabricated using 30-Co-TiO 2 based samples exhibited the lowest recombination resistance (11.89 Ω/cm 2 ) compared to the reference (14.72 Ω/cm2) and undoped TiO 2 (12.97 Ω/cm 2 ) based DSSCs. The 10-Co-TiO 2 and 50-Co-TiO 2 based DSSC devices showed slightly higher resistance, 12.53 and 13.13 Ω/cm2, respectively. The result indicates the much-accelerated electron transfer in the device through reduction of charge transfer resistance resulting in an increment in Jsc. Figure 7b displays the Bode phase plots obtained from the impedance results. From Bode plots, the lifetime of the photo-injected electrons (τe) in semiconductor oxide layer can be calculated from the peak frequency (ƒmax) in the large semicircle region relate to the equation: τe = 1/2πƒmax; where ƒmax is the maximum characteristic peak   66,67 . However, at higher doping concentrations, the fabricated DSSCs revealed high recombination resistance at the semiconductor oxide/electrolyte interface with lesser τe (5.2 ms) which obstruct the cell performance by critically affecting the charge transport. On the other hand, at a lower doping concentration (10 mg of Co), the obtained τe was less (15.3 ms) as compared to 16.1 ms (for 10-Co-TiO 2 ). Thus the impedance spectra are directly relating to the I-V results. Also, it is evident that doping can influence the morphology and also surface chemistry of the semiconductor film. In this work, maximum PCE was achieved for the doped sample with 30 mg Co 2+ , which was substantially ascribed to the improved surface roughness of the TiO 2 after the successful integration of cobalt ions. Moreover, the additional cobalt ions in photoanode films can prevent effective photoinduced electron-hole pairs separation by acting as recombination centers, as evident from the EIS results. The particle-like morphologies induce better interconnection that bids appropriate deployment of incident light and fast electron transport. As the doping concentration increased to 50 mg, the resulted sample shows drastic changes in morphology with lots of voids and pores, resulting in the formation of secondary impurity phases as well the incorporation of Co ions at the interstitial position. Due to this, the device shows high recombination resistance with smaller τe. Based on the above observations, the deduced electron transport over the cobalt doped TiO 2 photoanode is given in Fig. 8. Here the Co 2+ doping in TiO 2 significantly shifted the conduction band positively, which results in an enhanced driving force of electron injection and thus increases the electron  www.nature.com/scientificreports/ injection efficiency from the LUMO of dye to the conduction band of TiO 2 . Thus, improved light absorption and repressed charge recombination is obtained, as obvious from the UV-vis and impedance spectroscopic measurements 68 . Thus, a significant improvement in Jsc was achieved. Doping could also protect the photoanode against photooxidation. Here the photon energy is effectively transferred to the Co 2+ ions capable of quickly localizing the excitation and suppression of undesired reactions on the photoanode surface.

Conclusion
In summary, we have successfully synthesized Co 2+ ion-doped TiO 2 nanocrystals via the MOF route. The structure and photovoltaic performance of the doped crystals were studied systematically using various characterization techniques. It was observed that the MOF-derived TiO 2 samples have more surface roughness and more conductivity. The highest PCE of 6.86% in DSSC was achieved by 30 mg Co 2+ ion-doped TiO 2 in comparison to commercially available P25 Degussa (5.76%). The enhanced PCE was accredited to numerous factors, such as dopant-induced morphological and optical bandgap changes that resulted in improved light-harvesting efficiency with less recombination of photoinjected electrons. Better surface stability is attained after cobalt incorporation, which results in less dye aggregation on the surface with subsequent suppression in the recombination loss. The formation of several photoinduced charge carriers by charge transfer transition of Co 2+ ions mainly contributed to the improved photocurrent, which paved the way for better PV performance. To further improve the PCE, modification of photoanode by the deposition of scattering layer and the study on its thickness is in progress.

Materials and methods
N, N-dimethyl formamide (99.8%), DMF, cobalt acetate tetrahydrate (99.99%), terephthalic acid (98%), titanium (IV) butoxide (97%), and anhydrous ethanol (98%) were purchased from Sigma Aldrich. All the reagents were used intrinsically without additional purification, and deionized water was used for all the synthesis processes. All the solvothermal reaction was carried out in a 100mL autoclave provided by Amar Equipments Pvt. Ltd.
Characterization. The synthesized catalyst samples were analyzed by powder X-ray diffraction (PXRD) technique using a Panalytical-Empyrean X-ray diffractometer, Netherlands with Cu-Kα radiation (λ = 1.54 Å) at 45 kV and 40 mA to scan the diffraction angles from 10° to 80°. The phase identification was carried out using standard patterns reported in the International Centre for Diffraction Data (ICDD) database. Thermogravimetric analysis (TGA) was carried out in a PerkinElmer, Simultaneous Thermal Analyzer (STA) from room temperature to 900 °C at 10 C/min under N 2 atmosphere. FESEM and HR-TEM analysis were done using Tecnai G2 F20 instrument from FEI, Netherlands operating at 200KV. For TEM study, the samples were dispersed homogenously in isopropanol by ultrasonication and drop-casted onto a copper grid, and dried overnight before loading. Surface area and porosity measurements were analyzed using standard adsorption equipment AUTOSORB-1Q-MP-XR, Quantachrome Instruments, USA using N 2 gas with 99.995% purity. The sample was activated under a vacuum at 150 °C for 3 h followed by nitrogen adsorption−desorption isotherm measurement at 77K. The surface composition and chemical states were measured with an Omicron Nanotechnology (Oxford Instruments) X-ray photoelectron spectroscope (XPS) equipped with monochromatic Al Kα radiation, and the obtained spectra were fitted by means of a Gaussian function. The peak correction was done with reference to the standard carbon 1s peak obtained at 285.07 eV. I-V analysis for the DSSC was recorded using the CHI660e from CH Instruments under one sun irradiation by PET Photo Emission Tech with SS50AAA solar simulator. Electrochemical Impedance spectroscopy (EIS) was measured by a Biologic potentiostat-galvanostat (SP-150) under an open circuit (V oc ) under illumination. The frequency ranges from 0.1 Hz to 10 5 Hz with 10 mV alternative (AC) signal.
Synthesis:. MIL-125 synthesis. Ti-based MOFs (i.e., MIL-125) were synthesized by the solvothermal method. To a mixture solution of terephthalic acid (3 g), anhydrous methanol (6 mL), and anhydrous DMF (54 mL), butyl titanate (4.5 mM) was added slowly with continuous stirring. The whole mixture is loaded into a 100 mL teflon-lined stainless-steel autoclave and allowed to heat to 150°C for 24 hr. After, a white suspension of MIL-125 was obtained, separated by centrifugation, and washed with methanol (3 x 50 mL). The sample is subsequently dried in a vacuum oven at 80 °C for 12 h.
TiO 2 and cobalt doped TiO 2 synthesis. The as-synthesized MIL-125 sample was annealed at 400 °C at a rate of 2°C min −1 for 5 h to result in the pure titania powder stored in the sample vial for further uses. To prepare the doped TiO 2 samples, CoCl 2 ·6H 2 O (10, 30, and 50 mg) was added to the latter reaction mixture, and the same synthesis procedure was followed. After calcination, the doped samples were obtained with light to dark green depending on the amount of cobalt precursor.
Photoanode preparation. To Prepare the DSSC photoanode, the synthesized samples (0.5g) were dispersed into 4mL ethanol by ultrasonication for 15 minutes. Then add 1.5g terpineol and 0.10g ethyl cellulose and sonicate the solution to form a homogenous viscous paste. The prepared paste was deposited on cleaned fluorine-doped tin oxide (FTO) glass substrate (transparent electrode with a typical square sheet resistance of 10 Ω/sq and overall transmittance about 85% in the visible range) by doctor blade method to get a thickness of ~13-15 μm using 3M scotch tape into an area of 0.25 cm 2 (with proper masking). After drying in the air, the coated substrates were heated in a furnace (450°C for 2h). After reaching room temperature, the substrates were immersed in a 0.5 mM solution of N719 dye (Sigma-Aldrich) for 30 minutes. www.nature.com/scientificreports/ Counter electrode preparation. To prepare the Pt-counter electrodes, one or two drops of 5 mM H 2 PtCl 6 solution in isopropanol were drop cast onto the FTO and subsequently sintered at 400°C for 20 min. The transparent electrode obtained was used as such as the counter electrode without further heat treatment or modification.
Electrolyte preparation and assembling of DSSC. The photoanode and counter electrode were sandwiched together with surlyn film of 30 μm thickness as a spacer. Further, one or two drops of redox couple electrolyte consisting of 0.06 M 1-butyl-3-methyl imidazolium iodide, 0.03 M of iodine solution (I 2 ), 0.10 M guanidinium thiocyanate, 0.5 M 4-tert-butylpyridine dissolved in a mixture of acetonitrile and valeronitrile (volume ratio, 85:15), were added and introduced into the DSSC with a syringe.