Introduction
Metal oxide nanowire arrays based on wide bandgap semiconductors, such as TiO2 and ZnO, are widely used in photoelectrochemical (PEC) devices, such as dye-sensitized solar cell (DSSC) [1], quantum dot sensitized solar cell (QDSSC) [2], hydrogen generation [3], etc. Compared to conventional PEC devices using mesoporous photoanodes that consist in nanocrystals, the ones formed by nanowires have much longer majority carrier diffusion length along the crystal axis, because grain boundaries and traps on the electron transmission path are in lower density [4]. However, most of the PEC devices based on nanowire arrays have lower power conversion efficiency (PCE) than those based on nanoparticle layers, resulting from smaller surface area to load light absorbers [5]. To enhance the surface area of nanowire arrays, various hierarchical nanostructures which we collectively call structure engineering are applied to PEC devices [6]. One strategy is to increase the nanowire length and nanowire count per unit area, according to the calculation that surface area is proportional to them [7]. And another strategy is to generate fine branches on existing nanowires, and the dendriform nanostructures are typically synthesized by chemical etching [8] or multi-step hydrothermal methods [9]. Here, we introduce a strategy to replace conventional smooth transparent conducting substrates with microstructured ones, and this is not conflict with the former strategies. The photoanode with rough substrate covered by nanowire arrays has a much larger surface area than smooth one, and the gain factor depends on the roughness of the substrate. The two critical size parameters related to the performance of PEC devices are the distance between photoanode and cathode (Lelectrolyte), and the thickness of light absorption layer, i.e. length of nanowires (Lnanowire). Usually, Lelectrolyte is in magnitude of 100-1000 mm in consideration of the ion transport in electrolyte, and Lnanowire depends on the electrical properties of the metal oxide materials, which is 1-10 mm for TiO2 nanowires in rutile phase and 10-100 mm for ZnO nanowires [10,11]. On account of chemical stability, ZnO seems not an ideal choice in practice, although electron concentration and mobility of ZnO are higher than those of TiO2. Thus we only focus on rutile-nanowires-based QDSSC in this work.
The permissible roughness of substrate (Rsubstrate) is limited by Lelectrolyte and Lnanowire: if Rsubstrate is larger than Lelectrolyte, the active material around depressions of substrate has little chance to be reduced; if Rsubsteate is smaller than Lnanowire, the enhancement of surface area is not significant enough to improve the performance of devices. Therefore, former researchers only tried to develop structure engineering in the scale under Lnanowire, and we turn to the scale between Lnanowire and Lelectrolyte. And our approach can be combined with other methods for synthesis of hierarchical nanostructures to fully cover the range from 1 nm to 100 mm, so that the performance of devices can be fully optimized, in theory.
The structure engineering technology of microstructured substrate can be widely used in PEC devices as mentioned above, and we applied it to QDSSCs for demonstration to verify the effect. QDSSC is deemed as an analog of DSSC, and the quantum dot sensitizers are typically CdS, CdSe, PbS, PbSe, and InP [12–16]. The light absorption coefficient of quantum dots is higher than most dyes, and their absorption spectrum can be tailored by size confinement. But the PCE of QDSSC is lower than that of DSSC with same device structure due to the charge separation and recombination processes at interfaces [17]. Although our work cannot solve the problem of photo-generated carrier transport, the efficiency of QDSSCs increased 50% by using microstructured substrate, consistent with the estimation of surface area increment.
Experimental
Synthesis of rutile nanowires
The commercial 3 mm thick white glass was processed into rough glass by Beijing Yaxingwanbo Glass Co. LTD. During the etching process, the glass was covered by plastic mask, and 4 mm × 4 mm patterns were exposed in etchant. The fluorine-doped tin oxide (FTO) coating process was accomplished by magnetron sputtering from Wuhan Geao Co. LTD. Before photoanode fabrication, the rough FTO glass was ultrasonic cleaned in ethanol for 15 min and flushed by deionized water. The rough FTO glass was covered by Teflon tape leaving the active area exposed. The precursor solution was prepared by 15 mL deionized water and 15 mL concentrated HCl with 0.66 mL isopropyl titanate. The FTO glass was dipped in the solution heated at 150°C for 12 h in hydrothermal synthesis reactor. The substrate was flushed by deionized water again, and annealed at 450°C for 30 min.
Quantum dots deposition
The quantum dots deposition process was finished under 10°C water bath. The substrate with rutile nanowires was dipped in the solution of 20 mM<FootNote>
1 mM= 1 mmoL/L
</FootNote> CdCl2, 66 mM NH4Cl, 140 mM thiocarbamide, and 230 mM ammonia, for 1 h, followed by water flushing. The substrate was transferred to the solution with 26 mM CdSO4, 40 mM N(CH2COONa)3, and 26 mM Na2SeSO3. After 5 h, the substrate was immersed in the solution with 0.1 M Zn(CH3COO2)2 and Na2S for 1 min, and this could be repeated for 2 times. The prepared samples were annealed at 100°C for 10 min in inert ambient.
Preparation of QDSSCs
The cathode was cleaned brass sheet, and the electrolyte was the solution with 1 M Na2S and 1 M S. The whole device was sandwiched by photoanode and cathode, and the electrolyte was injected in the gap.
Characterizations
The surface morphology of FTO glass was characterized by stylus profiler (DEKTAK XT). The morphology of TiO2 nanowires was characterized by scanning electron microscopy (Nova nanoSEM 430, FEI company). The J-V measurement was obtained using an Agilent B2900 Series precision source/measure unit, and the cell was illuminated by a solar simulator (Solar IV-150A, Zolix) under AM1.5 irradiation (100 mW/cm2).
Results and discussion
Figures 1(a)-1(d) depict the schematic diagrams of the four classes of QDSSCs: the one with TiO2 nanowires on smooth FTO glass, with TiO2 nanowires on rough FTO glass, with TiO2 nanoparticles on smooth FTO glass, and with TiO2 nanoparticles on rough FTO glass, respectively. Rutile nanowire array was synthesized on the substrate directly by the hydrolysis of Ti4+ in acid condition by using hydrothermal method [18]. And the TiO2 nanowire in rutile phase was several micrometers in length and hundreds nanometers in diameter. Conventional QDSSC with mesoporous TiO2 nanoparticle layer was used as control group. Commercial TiO2 nanoparticles in anatase phase were deposited on substrate by doctor blading process, and the diameter of nanoparticle was 20 nm. Either of them could form uniform films on common FTO glass, respectively. On the other hand, we processed glass with chemical etching followed by magnetron sputtering of FTO to fabricate microstructured substrate (more details are shown in Methods Section 2.1). TiO2 nanowires or nanoparticles (control group) were deposited on it to prepare QDSSCs. Due to the epitaxial growth, TiO2 nanowires covered the surface of substrate uniformly. In details, each nanowire stands vertically to its root facet, and the density of nanowires remains unchanged. However, in case of TiO2 nanoparticles, a flat film was formed on the substrate, and the thickness was not uniform because of the morphology of substrate.
Fig.1 Schematic diagrams of QDSSC photoanodes with rutile nanowires on smooth FTO glass (a), rutile nanowires on rough FTO glass (b), anatase nanoparticles on smooth FTO glass (c), and anatase nanoparticle on rough FTO glass (d). The size of nanoparticles is exaggerated for easy reading |
The morphology of rough FTO glass was characterized by profiler study as shown in Fig. 2(a). The scan area is 499.2 mm × 632.8 mm (3.159 × 105mm2). The roughness is 8.989 mm (standard deviation of height), and the maximum height difference is 36.12 mm. Compared with the projected area, namely scan area, the actual surface area proved to be 4.773 × 105mm2 – a 51.09% increase. Figure 2(b) shows the optical microscopic image of microsctructured FTO glass. The surface was composed by blocks of pyramids ranging from 10 to 100 mm, corresponding to the profiler results. Transmission spectra demonstrate a ~70% transmission for rough FTO glass (Fig. 2(c)).
Figure 3 shows the scanning electron microscopic (SEM) images of the micro-scale hierarchical photoanodes. After TiO2 deposition and quantum dot sensitizing, the surface of rough FTO glass was covered by TiO2 nanowire arrays uniformly (Fig. 3(a)). Every TiO2 nanowire is coated by a layer of CdSe quantum dots (Fig. 3(b)). TiO2 nanowires grow epitaxially no matter the surface is rough or smooth as shown in cross-sectional SEM images (Figs. 3(c)-3(d)). And the morphologies of nanowires on rough FTO glass and smooth FTO glass are exactly the same (length: ~ 5.0 mm, diameter: ~ 200 nm, density: ~ 15/mm2). So the surface area of the TiO2 nanowire arrays (Aarr) was in direct proportion to the surface area of the substrate (Asub):
Here, ANW stands for the surface area of one nanowire, and DNW stands for the counting density of TiO2 nanowires. The surface area of rough FTO glass was about 1.5 times that of smooth FTO glass, and so does the quantum dot load capacity, light absorption, short circuit current density (Jsc), and PCE, theoretically. Figure 4(a) shows the performance of a pair of QDSSCs fabricated simultaneously based on different substrate. Jsc increases from 4.411 to 6.750 mA/cm2, rising 53.03%. PCE increases from 0.775% to 1.162%, rising 49.94%. The statistics of 12 pairs of QDSSCs are more convincing. The average Jsc increases from 3.922 to 5.283 mA/cm2, rising 34.70%. The average PCE increases from 0.723% to 0.989%, rising 36.79% (Fig. 4(b)). The average Jsc and PCE both increase by about 35%, which demonstrates the enhanced surface area. Figure 4(b) shows the scatter plot of PCE vs Jsc, they are linearly dependent according to the equation, where FF is fill factor:
After simple regression of PCE and Jsc, the slope represented the average coefficient FF × Voc, which decreases from 0.24 to 0.16 V. The rough FTO glass was also applied to conventional QDSSCs based on TiO2 nanoparticles. But the performance of the devices was not improved after replacing the photoanode substrate. Figure 5(a) shows the profile curve of the rough FTO glass coated with TiO2 nanoparticle layers in different thickness. The black line represents the cross sectional line of the microstructured FTO glass surface. The left part is smooth because it was protected during etching glass process. The right part is 12.2 mm lower than the left in average because of chemical etching. The deposition of TiO2 nanoparticles was achieved by doctor blading, so the TiO2 layer was flat and the microstructure of substrate was fully covered. Based on the smooth substrate, the thickness of TiO2 layer can be controlled to about 10 and 20 mm. In the case of microstructured substrate, the depth of etching area (~10 mm) should be considered when estimating the TiO2 thickness. The J-V measurements of various samples are shown in Fig. 5(b). The performance of QDSSCs depends on the thickness of TiO2 layers, and the optimized parameter was 21.8 mm for the device with rough FTO glass, which is comparable to the one with smooth FTO glass and 20.3 mm thickness. It indicates that the performance of TiO2-nanoparticle-based QDSSCs only relies to the average thickness of TiO2 layer, with no relation to the morphology of substrate, under normal conditions. However, if the TiO2 coverage is too thin with average thickness of 12.2 mm, Voc of the device would be 0.1 V lower, demonstrating the short circuit risk. The relation between PCE (Voc, Jsc) and TiO2 thickness is plotted in Fig. 5(c), and the outliers represent the device with thin active layer on rough substrate.
Fig.5 Profile images (a) and J-V curves (b) of QDSSCs with various average photoanode thickness. The photoanode is based on TiO2 nanoparticles. (c) Photovoltaic properties (PCE, Voc, and Jsc) of the QDSSCs as a function of average photoanode thickness. The photoanode is based on TiO2 nanoparticles |
Conclusions
The device design of QDSSCs based on microstructured substrate and rutile nanowires is proved to be effective, and the PCE and Jsc of devices can be increased by 50% compared to conventional ones with smooth substrate. The charge transport space in the device is fully utilized, in scale of 10 to 100 mm, which has never been noticed before. The microstructure would not impact the charge collection efficiency in the device, but increase the surface area of photoanode significantly. Although the technology does not work in the case of TiO2 nanoparticles, the microstructure design can be combined with the reported hierarchical nanostructures to further improve the performance of QDSSCs. Furthermore, we will apply the hierarchical design to other PEC devices, such as photocatalysis or water splitting. It leads a way to low-cost high efficient light energy utilizations.