Black TiO2-Based Dual Photoanodes Boost the Efficiency of Quantum Dot-Sensitized Solar Cells to 11.7%

Quantum dot-sensitized solar cells (QDSSC) have been regarded as one of the most promising candidates for effective utilization of solar energy, but its power conversion efficiency (PCE) is still far from meeting expectations. One of the most important bottlenecks is the limited collection efficiency of photogenerated electrons in the photoanodes. Herein, we design QDSSCs with a dual-photoanode architecture, and assemble the dual photoanodes with black TiO2 nanoparticles (NPs), which were processed by a femtosecond laser in the filamentation regime, and common CdS/CdSe QD sensitizers. A maximum PCE of 11.7% with a short circuit current density of 50.3 mA/cm2 is unambiguously achieved. We reveal both experimentally and theoretically that the enhanced PCE is mainly attributed to the improved light harvesting of black TiO2 due to the black TiO2 shells formed on white TiO2 NPs.


Introduction
In recent decades, energy shortages and environmental pollution have emerged as some of the most important concerns that need to be addressed [1]. Solar cells are an effective strategy to utilize solar energy as a kind of clean energy [2]. Among the third-generation solar cells, quantum dot-sensitized solar cells (QDSSCs) have attracted considerable attention due to the stunning properties of quantum dots (QDs) such as a tunable bandgap, a large light absorption coefficient, high stability, and multiple exciton effects [3,4]. Typical QDSSCs are composed of a photoanode, an electrolyte, and a counter electrode (CE). The photoanode is the crucial part and is responsible for the collection and transfer of photogenerated electrons, and generally consists of a mesoporous, wide bandgap oxide film (electron acceptor) coated with a light harvesting material (QD sensitizer). An ideal photoanode should have high specific surface area to absorb QD sensitizers, a suitable crystal structure to obtain extensive light absorption, and excellent ability for electron injection and transfer while having a small charge recombination rate [5]. Although numerous wide bandgap semiconductors such as ZnO, SnO 2 , and ZnS have been used as photoanodes in QDSSCs, TiO 2 is still the most extensively used material due to its excellent chemical stability, non-toxicity, and low cost, etc. [6][7][8]. Limited by the wide bandgap, TiO 2 can only absorb ultraviolet light, which accounts for~5% of sunlight. Although the absorption range of TiO 2 can be expanded by combining it with narrow bandgap semiconductors such as CdS, CdSe, PbS, CdSe, etc., the power conversion efficiency (PCE) of QDSSCs is still far from the theoretical value in spite of the great efforts dedicated to optimizing the species

Preparation of Black TiO 2
Black P25, rutile TiO 2 , and anatase TiO 2 were prepared by a titanium-sapphire laser system (Spectra-Physics, Spitfire ACE), which generated linearly polarized femtosecond laser pulses with a center wavelength of 800 nm, a repetition rate of 500 Hz, and a pulse width of about 40 fs [27]. A powdered sample of TiO 2 was flattened on the bottom of the container with a glass slide, and the container was fixed on a two-dimensional movable platform equipped with electric motors. A rotatable half-wave plate and a polarizer were placed in the laser propagation path to control the laser pulse energy at 1.0 mJ. The laser beam was then reflected by two mirrors with high reflection at 800 nm to make the beam propagate vertically to the surface of the TiO 2 samples. The laser beam passed through a f = 1 m focal lens to form a single filament [28][29][30], and then hit the flattened TiO 2 sample to produce black TiO 2 . The container was moved with the electric motors so as to raster it. As an example, the pristine and processed P25 TiO 2 samples are shown in Figure 1a with their crystalline structures.

Preparation of Photoanodes
Typically, 0.8 g TiO 2 (Aladdin, Shangai, China), 0.4 g ethyl cellulose (Aladdin), and 3.245 g α-terpineol (Aladdin) were dispersed in 8.5 mL ethanol (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) under stirring. The obtained paste was spin-coated on FTO substrates (Zhuhai Kaiwo Optoelectronics Technology Co., Ltd., Zhuhai, China) and dried at 80 • C for 20 min, where the FTO substrate had been cleaned by immersing in deionized water, absolute ethanol, acetone, absolute ethanol, and deionized water in turn and washing by ultrasonic methods for 30 min, respectively. Then, the films were annealed at 450 • C for 30 min to acquire porous TiO 2 films. The prepared TiO 2 film was sensitized with CdS QDs by the successive ionic layer adsorption and reaction method using cadmium acetate (0.05 M, Sinopharm Chemical Reagent Co., Ltd.) and anhydrous sodium sulfite (0.05 M, Aladdin). Then, the chemical bath deposition method was used to deposit CdSe QDs using selenium powder (0.08 M, Tianjin Guangfu Science and Technology Seven Development Co., Ltd., Tianjin, China), Na 2 SO 3 (0.05 M, Aladdin), Cd(CH 3 COO) 2 (0.05 M, Aladdin), and trisodium nitrilotriacetic acid monohydrate (0.12 M, TCI Shanghai). Finally, the ZnS passivation layer was coated by immersing the photoanode in 0.1M Zn(AC) 2 ·2H 2 O solution (Sinopharm Chemical Reagent Co., Ltd.) and 0.1 M Na 2 S·9H 2 O (Aladdin) solution repeatedly for two cycles.

Preparation of CuS and CuS/Brass-Mesh Counter Electrode (CE)
To prepare CuS CEs, a 50 mL solution including Na 2 S 2 O 3 ·5H 2 O (1 M) and CuS·5H 2 O (1 M, Tianjin Guangfu Science and Technology Seven Development Co., Ltd.) was prepared and the pH was adjusted to 2.0 by acetic acid (Aladdin). Clean FTO substrates were put in the solution and kept at 70 • C for 3 h. Then, the substrates were dried at 130 • C for 30 min to acquire CuS CEs. As for the preparation of the CuS/brass-mesh CE, the copper mesh (Hebei Xingheng Materialtech Co., Ltd., China) was soaked in 70 • C hydrochloric acid (36%, Sinopharm Chemical Reagent Co., Ltd.) for 2 h to remove the zinc on the surface. Then, the copper mesh was washed with deionized water and dried at room temperature. The treated Cu mesh was soaked in a mixed solution of CH 4 N 2 S (0.01 M, Sinopharm Chemical Reagent Co., Ltd.) and C 2 H 8 N 2 (Ethylenediamine, 1.5 M, Tianjin Guangfu Science and Technology Seven Development Co., Ltd.) for 24 h, and then the previous cleaning and drying steps were repeated.

Assembly of QDSSCs and Dual Photoanode QDSSCs
The polysulfide electrolyte was prepared by dissolving 2.4 g Na 2 S (Aladdin) and 0.32 g S powder (Sinopharm Chemical Reagent Co., Ltd.) in a 10 mL solution with a methanol/deionized water volume ratio of 7:3. Finally, the CuS CE, polysulfide electrolyte, and different photoanodes were assembled into a sandwich structure device divided by a polymer gasket filled with polysulfide electrolyte. The CPV concept was integrated into QDSSCs with dual-photoanode architecture (D-A), where the prepared CuS/brass mesh CE mentioned in Section 2.3 was used as the counter electrode and sandwiched between two identical photoanodes. Each photoanode and counter electrode were separated by the aforementioned polymer gasket. The fabrication procedure of QDSSCs is shown in Figure 1b.

Results
To investigate the effect of phase structure on device performance, three different black TiO 2 NPs have been prepared based on pristine P25, rutile TiO 2 , and anatase TiO 2 , respectively. The X-ray diffraction (XRD) results in Figure 2a-c indicate that the main diffraction peaks do not change after the laser ablation treatment for all three samples. The peak intensities of anatase and rutile TiO 2 are weakened slightly, which may be due to the transformation of the surface of TiO 2 into amorphous TiO 2 , resulting in a decrease in crystallinity, while the decrease in P25 TiO 2 is almost invisible. Fourier transform infrared spectra (FTIR) are shown in Figure 2d-e. Similar to the XRD results, the infrared peak positions of pristine TiO 2 are the same as those of black TiO 2 . The characteristic peaks of anatase TiO 2 and rutile TiO 2 can be observed in FTIR spectra of P25 TiO 2 and black P25 TiO 2 in Figure 2f. In the Raman spectra ( Figure S1 in the Supplementary Materials), it can be seen that there are no significant shifts in the Raman peak positions for the black TiO 2 , while the bands seem slightly smaller than the pristine TiO 2 in all three cases. This indicates that there is no significant difference between black TiO 2 and white TiO 2 in structure and functional groups [17][18][19]. The structural variation was further investigated by X-ray photoelectron spectroscopy (XPS). Figure 3a shows the XPS surveys of P25 TiO 2 with and without the laser treatment, indicating that there is no obvious difference in their chemical compositions. High-resolution XPS spectra are used to further clarify the detailed variations. The N1s XPS spectrum of black P25 TiO 2 in Figure 3b indicates that laser treatment of P25 TiO 2 in air can induce N doping [10,11,13]. Figure 3c shows the O1s XPS spectra of pristine P25 TiO 2 and black P25 TiO 2 . The peaks at 530.3 eV, 532.2 eV, and 533.2 eV correspond to O Ti-O , O V , and O O-H bonds, respectively. The oxygen vacancy related peak is located at 532.2 eV, and it can be observed that the O V peak intensity in black P25 TiO 2 is higher than that in pristine P25 TiO 2 , indicating that the oxygen vacancy after the laser treatment is significantly increased. This is an important reason for the color change of P25 TiO 2 after the laser treatment. Figure 3d demonstrates the Ti2p XPS spectra of pristine P25 TiO 2 and black P25 TiO 2 . The two characteristic peaks of Ti 4+ are located at 459.25 eV and 465.1 eV, and the two charac-teristic peaks of Ti 3+ are located at 458.2 eV and 463.95 eV. The proportion of Ti 3+ in black P25 TiO 2 is higher than that of pristine P25 TiO 2 , which is consistent with the change of oxygen vacancy content in Figure 3b, confirming the aforementioned explanation [31][32][33]. Shown in Figure 4 are the UPS spectra of anatase TiO 2 and rutile TiO 2 before and after laser treatment. Both results indicate that the top of the valence band of black TiO 2 is lower than that of pristine TiO 2 , which is consistent with previous reports [14,15,18]. From the UV-vis absorption spectra ( Figure 5a) and diffuse reflectance spectra ( Figure S2b in the Supplementary Materials) of the six samples it can be seen that the main absorption of the six samples occurs in the spectral range of less than 400 nm, mainly due to the inherent bandgap of TiO 2 (3.02-3.20 eV) [34]. It should be emphasized that the change in the absorption up to 2500 nm can also be observed (see Figure S2a in the Supplementary Materials), which indicates the strong effect of the laser treatment on the optical properties of TiO 2 .
The absorption of black TiO 2 in the visible and infrared region is overall higher than that of pristine TiO 2 , and the significant enhancement in light absorption will generate more photoexcited electrons, and thus a higher photocurrent J sc . Shown in Figure 5b,c are the bandgap diagrams of anatase TiO 2 and black anatase TiO 2 , and rutile TiO 2 and black rutile TiO 2 , respectively, from which it can be seen that the bandgap of the black TiO 2 has no obvious narrowing. This is in contrast to the black TiO 2 calcined at high temperature in the reducing gas, in which the reduction of TiO 2 is comparatively complete and there is a more obvious bandgap narrowing. The laser treatment only acts on the surface of TiO 2 NPs. The test thickness of UPS is about 1-2 nm; therefore, UPS spectra mainly reflect the surface properties of NPs. Therefore, the black TiO 2 we prepared is likely to be a core-shell structure, the surface layer should be amorphous TiO 2 , and the core is untreated crystalline TiO 2 . The XPS results of anatase TiO 2 and rutile TiO 2 before and after the laser treatment are similar to that of P25 (see Figures S3 and S4 in the Supplementary Materials). The detailed analysis of the energy band variation induced by the laser treatment is further investigated by first-principle calculations and will be shown later.   The results are consistent with the aforementioned prediction that the core is untreated crystalline TiO 2 , and the difference exists on the surface area. For black P25, an amorphous layer with 1.02 nm thickness can be observed. That is to say, the black TiO 2 prepared by the laser treatment is a core-shell structure, the surface layer is amorphous TiO 2 , and the core is untreated crystalline TiO 2 , which confirms the conclusion drawn from  Then, we investigated the PCEs of QDSSCs fabricated with a single TiO 2 photoanode, and the corresponding device parameters are shown in Table 1. For J-V curves of anatase TiO 2 in Figure 7a, the PCE of anatase TiO 2 is 3.6% with a J sc = 18.3 mA/cm 2 , and for black anatase TiO 2 the PCE is 4.7% with a J sc = 22.9 mA/cm 2 . The overall PCE is increased by 30%, which is due to the improved absorption of TiO 2 after the laser treatment. The schematic diagram of the dual photoanodes is shown in Figure 8. For the PCE of D-A TiO 2 QDSSCs, by one sun irradiating from the top photoanode and with concentrated sunshine illumination from the bottom photoanode with the help of a parabolic reflector, the PCE is found to be greatly enhanced. The PCE of dual anatase TiO 2 photoanode is 7.2%, and that of black anatase TiO 2 is 9.1%. For rutile TiO 2 in Figure 7b, the PCE increased from 1.6% to 2.3% after the laser treatment, and the overall PCE increased by 43.75%. The PCE of the dual rutile TiO 2 photoanode is 3.9% and that of the black rutile TiO 2 is 5.3%. The best performance is achieved by P25 TiO 2 and is shown in Figure 7c, which is due to the synergic effect of rutile and anatase TiO 2 . The J sc increased from 16.6 mA/cm 2 to 25.0 mA/cm 2 after the laser treatment, inducing an increasing in PCE from 4.0% to 5.9%. The PCE of the dual P25 TiO 2 photoanode is 8.0% and that of the black P25 TiO 2 is 11.7% with a J sc of 50.3 mA/cm 2 . From the J-V curves, it can be confirmed that the increase in PCE is mainly due to the enhanced J sc , which suggests the improved collection efficiency of photogenerated electrons.   To further verify the above mechanism, we measured incident photon-to-electron conversion efficiency (IPCE), as shown in Figure 7d-f. For pristine TiO 2 without the laser treatment, all three kinds of devices exhibit a photo-response in the visible region, which is consistent with previous reports since CdS/CdSe QD sensitizers absorb visible light. Replacing pristine TiO 2 with black TiO 2 induces the extension of the IPCE to the nearinfrared region as shown by the red dotted curves, which is responsible for the increase in J sc . It should be noticed that the QD sensitizers used in the present study only absorb visible light, and thus the extended IPCE should be caused by the black TiO 2 , which is consistent with the absorption properties of black TiO 2 , shown in Figure S2 in the Supplementary Materials. In addition, it should be mentioned that we adopted a dual photoanode architecture, and two photoanodes are connected in parallel and share a CuS mesh CE. Meanwhile, the CPV concept was integrated into QDSSCs. Light management as an important technology to improve the conversion efficiency in solar cells; aiming to increase the photon flux received by solar cells. By the light trapping effect, the optical path length is increased, thereby improving the PCE of solar cells. The IPCE test on two different measurement devices (Crowntech QTest Station 1000 CE, America and DG-6050 Zolix, China) are different from the PCE test on two different solar simulators (SS150 Solar Simulator Zolix and Sirius-SS150A-D Zolix), where sunlight is illuminated on both sides of cell. This is because the IPCE test only allows illumination from the top side due to the device design. Therefore, the actual IPCE should be higher than the present test value in Figure 7. The reported PCE 11.7% is an average value, and a larger PCE of over 12% could be acquired. Shown in Figure 7g-i are Nyquist curves of anatase TiO 2 , rutile TiO 2 , and P25 TiO 2 photoanodes with and without the laser treatment, which were measured in the dark and under illumination, respectively. The Nyquist curves are obtained from the Electrochemical Impedance Spectroscopy (EIS) measurements, which can reveal the interfacial reactions of photoexcited electrons in the QDSSCs. The electrochemical system can be regarded as the equivalent circuit shown in the inset, where R s represents the series resistance in the high frequency region, and CPE 1 and CPE 2 are the chemical capacitances of the cathode and photoanode, respectively. R 1 and R 2 refer to the charge transfer resistance between the counter electrode and electrolyte (R 1 ) and that at the TiO 2 /QDs/electrolyte interface (R 2 ), which can be determined from the radii of the first and second circles of the Nyquist curve, respectively. As shown in Figure 7g-i, the radii of all of the first circles of the Nyquist curves are unobservable, indicating that the charge transfer resistance R 1 is negligible for all the QDSSC devices. Furthermore, the radii of the second circles of the black TiO 2 based photoanodes are remarkably smaller than those of the pristine TiO 2 , corresponding to the smaller charge transfer resistance, which indicates that the charge transport of black TiO 2 -based devices is faster than that of pristine TiO 2 -based photoanodes. This is due to oxygen vacancy doping providing a more convenient transport channel for electrons [8,35]. This greatly improves the ability of the TiO 2 thin film to separate, collect, and transport photogenerated carriers in the cell, thereby reducing the loss of the photogenerated carriers during the device operation. Among the three kinds of TiO 2 , the smallest impedance is achieved by black P25 TiO 2 , which is consistent with the J-V results. Compared with the pristine sample, the decrease in interfacial electron transfer resistance enables facilitation of electron transport, leading to a significant increase in J sc .
XPS has confirmed that the laser processing can induce oxygen vacancies accompanied by N-doping in TiO 2 , while TEM results suggest that the laser processing only affects the surface of TiO 2 NPs, inducing a core/shell structure; thus, the oxygen vacancies and N-doping should mainly exist in the surface layer. The results in Figure 2 indicate that the bandgap does not change obviously before and after the laser treatment, which is believed to originate from the measurement depth of XPS, and reflects the information on near surface layer. Therefore, the analysis of energy band variation induced by the laser processing is further investigated by first-principle calculation of pristine TiO 2 and Ndoped TiO 2 rich in oxygen vacancies induced by laser processing [28,30,[36][37][38][39]. It is found that the oxygen atoms in the anatase phase are equivalent, so any oxygen vacancies will not affect the calculation results. Shown in Figure 9 are the calculated results in the range of −10-10 eV for anatase and rutile TiO 2 before and after the laser treatment (see Figure  S9 in the Supplementary Materials for a larger energy range of −60-30 eV). For anatase TiO 2 , it can be observed from Figure 9a,b that the bandgap of TiO 2 has been significantly narrowed due to the doping of oxygen vacancies and nitrogen atoms. This is mainly due to the donor behavior of the introduced oxygen vacancies, leading to the formation of a donor level. Thus, the forbidden band width is reduced from 3.171 eV to 1.498 eV. However, the substitutional doping of O atoms by N atoms would introduce holes, which act as acceptors and produce a shallow acceptor energy level above the top of the valence band of pristine anatase TiO 2 ; therefore, there is no sharp narrowing in the bandgap. For rutile TiO 2 in Figure 9c,d, similar bandgap narrowing from 3.083 to 0.13 eV can be confirmed. This is significantly larger than that of anatase TiO 2 . Compared with the anatase TiO 2 structure, the rutile unit cell has fewer atoms (see Figure S10 in the Supplementary Materials for unit cells of the pristine TiO 2 and black TiO 2 with nitrogen-doping and oxygen vacancies). The same number of oxygen vacancies and nitrogen atoms account for a larger atomic proportion, which should be the reason for the smaller calculated bandgap than that of anatase TiO 2 . The experimental and theoretical investigations in the present work have shown that the laser processing mainly induces N-doping and oxygen vacancies near the surface layer of TiO 2 , which reduces the bandgap, inducing a broadened absorption [40].
Since black TiO 2 itself shows broad absorption, it is possible to use black TiO 2 as a photoanode, without loading CdS/CdSe sensitizers, to realize high PCE (the corresponding results are shown in Figure S11 of the Supplementary Materials). It could be confirmed that the pure black TiO 2 photoanode shows very poor PCE, suggesting that the pure black TiO 2 photoanode cannot realize effective photogenerated electron extraction, which is probably due to a failure to find a suitable electrolyte and CE that matched black TiO 2 . The polysulfide (S 2-/Sn 2-) electrolyte is commonly used in QDSSCs since it can stabilize the commonly used chalcogenide QD sensitizers and provide an acceptable photovoltaic performance [41]. However, in the absence of sulfide QDs, the polysulfide electrolyte could not directly combine with TiO 2 to achieve redox reaction. Photosensitizer can only be used in solar cells if the right electrolyte is combined with a matching pair of electrodes [26]. While the debate around the photoelectrical and photocatalytic mechanism of black TiO 2 still exists, the improvement in PCE of QDSSCs in this work could be attributed to two reasons according to our experimental and theoretical investigations: (i) the expanded absorption of black TiO 2 to the near-infrared region induces more excited photoelectrons in the photoanode, and thus increases the J sc ; and (ii) the dual photoanode design and CPV integration, where the two photoanodes are connected in parallel and share a CuS mesh CE. The CuS grown on Cu mesh as a CE allows the transmitted light from the top cell to arrive at the bottom photoanode. Thus, the dual photoanode structure can capture more light and increase current density. In addition, the CPV systems can also enable an augmented light transmission to the top photoanode through the parabolic reflector at the bottom, providing additional excitation energy for light harvesting. To our best knowledge, the present work achieves the highest PCE of CdS/CdSe QD co-sensitized QDSSCs, which is mainly due to the obvious enhancement in J sc (Table S1 in Supplementary Materials).

Conclusions
In the present study, we have developed a strategy to boost performance of QDSSCs, in which black TiO 2 produced by laser processing was used as a photoanode material. It has been shown that the laser fabrication induces N-doping and oxygen vacancies near the surface layer of TiO 2 and forms the core/shell structure, the synergic effect of which may be the origin for the extension of the absorption of black TiO 2 into the visible and infrared region. Combined with the CPV structure design, a PCE of 11.7% with a J sc of 50.3 mA/cm 2 in the QDSSCs were achieved, which is due to the expanded absorption of black TiO 2 , the dual photoanode design, and CPV integration. While great efforts have been dedicated to synthesizing novel QD sensitizers and to the complicated structure design of photoanodes, the present work has provided an alternative way to boost performance of QDSSCs, which are expected to exhibit higher performance by further optimization of cell design parameters.

Conflicts of Interest:
The authors declare no conflict of interest.