Anodic Nanostructures for Solar Cell Applications

As a versatile, straightforward, and cost-effective strategy for the synthesis of self-organized nanomaterials, electrochemical anodization is nowadays frequently used to synthesize anodic metal oxide nanostructures for various solar cell applications. This chapter mainly discusses the synthesis of various anodic TiO 2 nanostructures on foils and as membranes or powders, and their potential use as the photoanode materials based on foils, transparent conductive oxide substrates, and flexible substrates in dye-sensitized solar cell applications, acting as dye-loading frames, light-harvesting enhancement assembly, and electron transport medium. Through the control and modulation of the electrical and chemical parameters of electrochemical anodization process, such as applied voltages, currents, bath temperatures, electrolyte composition, or post-treatments, anodic nanostructures with controllable structures and geometries and unique optical, electronic, and photoelectric properties in solar cell applications can be obtained. Compared with other types of nanostructures, there are several major advantages for anodic nanostruc‐ tures to be used in solar cells. They are (1) optimized solar cell configuration to achieve efficient light utilization; (2) easy fabrication of large size nanostructures to enhance light scattering; (3) precise modulation of the electrochemical processes to realize periodic nanostructured geometry with excellent optical properties; (4) unidirectional electron transport pathways with suppressed charge recombination; and (5) large surface areas by modification of nanostructures. Due to the simple fabrication processes and unique properties, the anodic nanostructures will have a fascinating future to boost the solar cell performances.


Anodic oxides
The electrochemical anodization of anodic oxides is conducted with two-electrode or threeelectrode configurations in an electrochemical cell, with the targeted metals as the working electrode (anode) and usually Pt/carbon as the counter electrode (cathode). The electrolytes used are typically fluoride-containing organic or aqueous solutions. A voltage is applied between the two electrodes to fabricate various nanostructures on the anodic metal surface. The anodic oxides feature the self-organized configurations with regular and open structures, such as tubes, pores, channels, bundles, powders, and various tailored shapes. The formation mechanism of oxide nanostructures is of great concern. For tube arrays, the possible processes are as follows: (1) The first step is the tube initiation. The tube growth initiates at some preferable positions on the metal surface that can provide both high local electric field and narrow channel-like surface morphology. It has been indicated that the morphological instability of the initial oxide surface layer causes the formation of pores [1]. Hence the surface state of the starting metal substrate can significantly influence the surface morphology of the anodic oxides, such as pore size distribution, regularity, and shapes. (2) The second step is the tube growth. One most accepted view is that the oxidation of the metal with the assistance of electric field forms a compact thin oxide layer (barrier layer) on the surface [2]. The thin oxide layer is also partially dissolved under the assistance of electric field. The oxidation and dissolution happen simultaneously. When the oxide growth rate at the metal/oxide interface equals to the oxide dissolution at the oxide/electrolyte interface, the thickness of the compact oxide layer keeps unchanged. Then the compact layer at the oxide/metal interface moves towards the inner part of the metals and the oxide nanostructures form into the metals.
Due the advantages such as simplicity, high efficiency, and low cost of electrochemical anodization, it has been utilized to fabricate various wide band gap metal oxide nanostructures, such as titanium oxide (TiO 2 ), aluminum oxide (Al 2 O 3 ), hafnium oxide, zirconium oxide, niobium oxide, tantalum oxide, tungsten oxide, and their alloys. Among them, the most widely studied anodic oxides are Al 2 O 3 and TiO 2 . For Al 2 O 3 , usually nanoporous structures can be obtained, which can be used as filters and templates. TiO 2 nanotube structure was first reported by Zwilling in 1999 by electrochemical anodization of Ti metal in aqueous fluoride containing electrolyte [3,4]. Recent studies have indicated that the nanotube morphology is in fact converted from nanopores by dissolution of the oxides at the interpore region. Due to the unique properties, TiO 2 nanotubes have been utilized in dye-sensitized solar cells (DSSCs), perovskite solar cells, quantum dot solar cells, photocatalysis, batteries, supercapacitors, electrochromic devices, sensors, drug release, and cell differentiation applications. In this chapter, we mainly discuss the fabrication of TiO 2 nanotubes and the design and improvement of DSSCs.
The electrochemical anodization is a highly controllable technique. The growth of anodic TiO 2 nanostructures can be influenced by many key anodization parameters, such as applied voltages, currents, bath temperatures, post treatments, and kinds of electrolyte. The control of the complex oxidation formation and chemical etching of the oxides can be realized during the anodization process, to establish the balance between the oxidation and dissolution. As a result, the morphology, regularity, growth rate, size, length, wall smoothness, and single or double walled nanotubes can be artificially designed. For example, fast growth of nanotubes can be realized in an electrolyte-containing lactic acid, which is desirable for industry production [5]. The nanotubes with lengths ranged from several 100 nm to 1000 μm have been fabricated [6]. Furthermore, by in-situ doping with various non-metal or metal elements in the electrolytes containing additives, visible light response of the anodic oxides can be obtained. The control of anodic nanostructures will greatly influence their physical and chemical properties, and their performances in various devices.

Solar cell applications
The development of low-cost new-generation solar cells has attained broad attention recently. Dye-sensitized solar cell (DSSC) is a kind of photoelectrochemical cell with the advantages of low cost, simple synthesis, large area, and high stability. Since the first report by O'regan and Grätzel with 7.1% efficiency [7], the highest efficiency of above 13% has been achieved [8]. In DSSCs, typically TiO 2 nanoparticles are used as the photoanode material, which are coated on transparent conductive oxide (TCO) substrates to form porous networks. The dye sensitizers are adsorbed on the surface of nanoparticles to harvest incident light. The dyes are surrounded by liquid redox electrolytes, which can reduce the oxidized dyes and accept electrons from the counter electrode. In the photon-to-electricity conversion process, the light absorption and the charge transport are two separate processes. The electron-hole pairs are separated by the heterojunctions with different work functions. No obvious built-in electric field exists in DSSCs, and the electrons go through the TiO 2 network by diffusion. Inspired by the development of DSSCs, other types of sensitized solar cells such as quantum dot (QD)/semiconductor sensitized solar cells and perovskite solar cells have been introduced. For QD systems, by adjusting the size of QDs, the band gaps can be tuned [9]. Furthermore, by using QDs, multiple exciton generation effect exists [10]. The perovskite solar cells are a new type of solar cell devices. Due to the high absorption efficiency and broad absorption range of perovskite materials, the efficiency has been above 20% recently [11].
For solar cell applications, the anodic nanostructured materials are usually used to replace TiO 2 nanoparticles to fabricate the photoanodes. In DSSCs, the photoanodes mainly play two important roles. First, the nanostructures provide large specific internal surface areas for the anchoring of dyes. Second, the nanostructures provide the charge diffusion pathways to transport the injected charges to the outer circuit. As a result, the morphology, structure, crystal structure, and surface state of anodic nanostructures can determine the performance of DSSCs based on these photoanodes, including loading of dyes, light harvesting, charge transport, recombination, and collection, and finally power conversion efficiency.

1D TiO 2 nanotubes
The nanoparticle photoanodes in DSSCs have the randomly distributed sizes, and loosely and irregularly packed structures. As a result, the injected electrons encounter a large amount of nanoparticles (about 10 3 to 10 6 as estimated) when diffuse through the photoanodes. The real path for electrons to travel before reaching the substrate is very long. This increases the probability of electron recombination loss at the oxide/electrolyte interface. One-dimensional (1D) nanostructures, such as nanotubes, nanowires, nanofibers, and nanorods, can provide the directional electron diffusion, which shortens the electron pathways, and are proved to have better collection efficiency [12]. By electrochemical anodization, various 1D nanostructures have been developed, and one very important nanostructure is TiO 2 nanotubes. The first attempt to use TiO 2 nanotubes to replace nanoparticles in DSSCs is reported by Schmuki et al. [13], and afterwards various reports have been emerging. The electron recombination rate in TiO 2 nanotubes is found to be very slow, with much larger electron lifetime (more than 10 times) than TiO 2 nanoparticles [14]. Thus, anodic TiO 2 nanotubes are very promising for high performance solar cells.

Free-standing membranes
Usually, the anodic nanotubes are grown on the Ti metal foil substrate, which is opaque. When directly utilizing TiO 2 nanotubes on foils in DSSCs, back-side illumination cell configuration is needed [15]. That is to say, the solar light enters the solar cell from the counter electrode. As a result, the light would be reflected by counter electrode that is coated with a thin layer of catalyst, and also be absorbed by dark redox electrolyte before it can be absorbed by dyes loaded on the anodic nanostructures. It has been estimated that about approximately 30-40% light energy would be lost by using this back-side illumination configuration [16]. To realize the front-side light illumination and to fabricate transparent electrode for broad applications, the anodic TiO 2 nanotubes should be fabricated on the TCO substrates.
One strategy is the direct growth of TiO 2 nanotubes by anodizing the sputtered or thermal evaporated thick Ti metal films on TCO substrates. However, there exist two problems. One is that the synthesis of thick and high-quality metal films on TCO substrate is very difficult, with complex and expensive procedures. The other is that the anodization oxidation process would lead to the increase of the TCO substrate resistance, the weak connection between the anodic oxide films and the substrate, and even the peeling off of the oxide films from the substrate. The low electric conductivity will result in the efficiency loss of solar cells.
The more promising method is to peel the anodic oxide layer off the metal substrate to obtain free-standing membranes, and transfer them onto TCO substrate. Various strategies such as ultrasonication separation, solvent evaporation, selective metal dissolution, and chemicalassisted separation have been proposed to peel off the films, but the procedure is complex and need careful handling. Furthermore, to obtain open bottom structure for effective tube filling and flow through application, usually additional chemical etching steps of tube bottoms are needed.
To obtain high-quality free-standing membranes with simple synthesis procedures and tunable bottom morphologies, the self-detaching method has been proposed [17]. The asformed anodic oxide layer is thermally treated at a certain temperature, and anodized again in the same electrolyte. After a short time, the oxide layer can be peeled off from the substrate without any post-treatment procedure. If the as-formed oxide layer is subjected to heat treatment at a low temperature (e.g., 200°C), the detached layer is amorphous and shows a black color. After annealing, the layer becomes crystallized and transparent (Figure 1a). For high temperature-treated layer (e.g., 400°C), the tubes are crystallized in the anatase phase before detachment. As a result, the detached layer is already transparent without subsequent annealing (Figure 1b). The detachment is probably because of the difference in mechanical and chemical stability between the top anodic layer (which we try to peel off) and the newly formed layer underneath the top layer [18]. Furthermore, according to the heat treatment temperatures during the detachment, the free-standing oxide layer can have open tube bottoms (200°C, Figure 1c-f) or closed bottom ends (400°C). It should also be noted that during the detaching anodization, elevating the bath temperature can facilitate the layer detachment and promote the formation of open bottoms. This elevated temperature can reduce the electrolyte viscosity and enhance the filed-assisted chemical dissolution at the tube bottom. Also, the strategy is a green technology without the use of corrosive solution and ensures continuous production [19].

Front-side illumination
The above-mentioned detached free-standing membranes can be used in DSSCs to achieve front-side illumination solar cell configuration, improving the light utilization efficiency. The detached oxide layer is transferred and adhered onto the TCO substrate by a thin layer of TiO 2 nanoparticles, and then sintered to enhance the connectivity (Figure 2a). For this detaching and transfer method, the solar cell efficiency is much higher (about 1.75 times) than on foil, due to the improvement of illumination configuration [20]. Furthermore, it has been found that tube bottom morphology also affects the solar cell efficiency. The solar cell based on tubes with open bottoms shows a better performance than that with closed bottoms, with an efficiency improvement of 17.7%. The photographs of the two kinds of photoanodes before annealing can be seen in Figure 2b. The removal of the tube bottom and the barrier layer can induce more dye loading of the tubes and less light scattering of the bottom caps. Furthermore, the open bottoms can facilitate the diffusion of redox electrolyte and thus reduce the recombination probability of electrons with the oxidized ions in the electrolyte [21]. Other solar cell configurations utilizing detached membranes include the bottom down or bottom-up structures and combination of nanoparticle/nanotube layers, which can be designed as required. Besides the free-standing films, nanotube powders can also be used to achieve the front-side illumination configuration, which is discussed in Section 3.

Scattering effect
The widely used dyes in DSSCs commonly have high absorption efficiencies only within a narrow wavelength region, with low response for the red-and near-infrared light. Simply increasing the thickness of photoelectrode film to enhance the absorption of photons will cause the increase of resistance and recombination loss. Therefore, it is essential to develop light scattering structures to increase the photon absorption opportunity by dye molecules in the weak absorption regions. As a common optical phenomenon, light scattering effect could extend the optical traveling length of incident light, so as to improve the light-harvesting efficiency and to achieve high performances. Based on Mie theory, to achieve efficient light scattering, the size of scattering centers should be comparable to the wavelength of the incident light. As a consequence, scattering structures with various morphologies have been introduced.
In general, there are three different photoelectrode configurations to combine light scattering centers. The first one is the mixed structure, for which large particles such as scattering centers are embedded into the photoelectrode films (Figure 3a). The scattering centers in such structure could introduce multiple scattering in the light absorption layer. However, the large particles would unavoidably cause the loss of dye adsorption because of the low surface area. The second one is the double layered structure, for which the light scattering layer is placed on the top of the nanocrystalline film (Figure 3b). Adding the scattering layer on the top could ensure sufficient dye loading. However, the light scattering would be weaker than in the mixed structure. The third one is the photoelectrode composed of hierarchical nanostructures with dual functions (Figure 3c). There have been many studies on the fabrication of such hierarchical nanocrystallite aggregates. The intensive light scattering could be guaranteed without much loss of specific surface area, but the preparation process is much more complicated. As a result, to achieve prominent light scattering and thus high DSSC performance, we should overcome the drawbacks of these structures.

Conventional scattering centers
The large TiO 2 particles with diameters of about 200-400 nm are commonly used as light scattering centers, which can be simply mixed into the photoelectrode films to increase the light harvesting efficiency. In the early studies, Grätzel et al. have found that when TiO 2 nanoparticles are hydrothermally prepared at 250°C (normally below 230°C), the nanoparticle film would become translucent because of the spontaneous particle agglomeration [22]. Subsequently, the fabrication of large TiO 2 particles with high crystallinity has been paid more and more attention.
Besides nanoparticles, solid spheres fabricated by facile hydrothermal method also exhibit light scattering effect. However, the large particles or spheres usually suffer from low surface area for dye loading, which would affect the light harvesting. To overcome the weakness of low surface area, TiO 2 spheres with rough surfaces have been synthesized. Also, mesoporous spheres with dual functions have been fabricated by nanocrystallite clustering, which are dominant in light scattering without the loss of surface area for dye-uptake. The hollow spheres are an alternative candidate for light scattering centers because of their larger surface area and better infiltration of electrolyte. Recently, core-shell microspheres have gradually emerged and become very promising scattering center. The core-shell structure could not only provide large light scattering, but also confine the light inside the spheres.

TiO 2 nanotubes powders
Apart from the above structures, 1D nanostructures could also serve as light scattering centers, such as nanorods, nanofibers, nanotubes, and their aggregates. For example, the lightto-electricity conversion yield of 6.08% has been achieved by blending of large TiO 2 nanorods (800 nm) and small nanorods (20-40 nm) on the top of small nanorod films, benefit from the low charge transport resistance and high light scattering effect [23]. TiO 2 nanofibers have been readily fabricated by simple electrospinning and applied in DSSCs as photoanode films [24]. The mixture of 1D nanostructures with nanoparticles could make a significant improvement of the performance of DSSCs also because of the combined effect of strong light scattering, abundant dye-loading amounts, and the improvement of electron transport properties.
TiO 2 nanotube arrays with 1D nanostructure have been found to have prominent light scattering effect, with the combination of superior electron transport when applied in DSSCs [14,25]. Due to the formation of large crystalline grains and existence of nanotube bundles, the TiO 2 nanotubes could generate effective light scattering. Lee et al. [26] obtained TiO 2 nanotube powders and coated the powders on the top of the nanocrystalline film, and attained noticeable increment of light harvesting.
The mixed structures of nanotubes and nanoparticles have also been explored in many studies. Embedding nanotube powders inside the nanoparticle film can (1) promote the permeation of liquid electrolyte, (2) increase the electron transport in the photoanode, and (3) introduce the scattering effect of nanotubes, which would be profitable for high perform-ance DSSCs. Lin et al. have fabricated TiO 2 nanotubes by anodic growth and involved ultrasonic separation of the resulted nanotubes to obtain TiO 2 nanotube powders (Figure 4a) [27].
To reduce the electron trap states, the nanotube powders are thermally treated at a high annealing temperature (650°C) to enhance crystallinity (the details are discussed in Section 5).
The hybrid photoanodes can be formed by mixing these highly crystallized nanotube powders with TiO 2 nanoparticles (Figure 4b). By adjusting the weight ratio of the nanotube powders, the performance of DSSCs could be optimized and the highest efficiency of 6% has been achieved.

Large nanotubes
Generally, the diameter of nanotubes fabricated by conventional electrochemical anodization is about 100 nm, which is obviously smaller than the wavelength of visible light. The large diameter nanotubes of size comparable to the visible light wavelength (500 nm or above) can be synthesized under high-voltage anodization. For example, by anodization at a high voltage of 180 V in an organic electrolyte containing lactic acid (aged for 10 hours), large nanotubes with top diameter of 300 nm and bottom diameter of 500 nm have been fabricated (Figure  5a,b) [28]. The resulting nanotube membranes are transferred onto a thick TiO 2 nanocrystalline film to act as the light scattering layer (Figure 5c,d). The large diameter nanotubes show a superior scattering property, and the photoanode incorporated with large nanotubes on top is nearly opaque in visible light (Figure 5e). By introducing the large diameter nanotube membranes, the performance of DSSCs has been greatly improved, which is 19% higher than that without scattering layer, and 11% higher than that using normal nanotubes (100 nm) as the scattering layer (Figure 5f). To enhance the functionality of the photoanodes of DSSCs, the multi-stacked photoanodes have also been developed by integration of three or more various TiO 2 architectures [29,30]. For TiO 2 nanotubes, the similar multi-layered structures have been fabricated (Figure 6a), based on the study of large diameter nanotubes [31]. First, bi-layered TiO 2 nanotube membrane with top large nanotubes (~540 nm) and underlayer normal nanotubes (~130 nm) has been fabricated by two-step anodization in different kinds of electrolytes (Figure 6b). Secondly, the bi-layered membrane is transferred onto a layer of TiO 2 nanoparticles for building photoanodes. The three layers are stacked together with gradually decreased sizes from top to bottom. For this type of multi-layered photoelectrode, layers with different tailored nanostructures play different roles on the performance of the DSSCs. By optimizing the thickness and synergistic effects of each layer (Figure 6c), large amounts of dye adsorption, reduced recombination during the electron diffusion, and efficient light scattering can be simultaneously guaranteed. DSSCs based on the multi-functional photoanode shows a photoelectric efficiency as high as 6.52%.

Photonic crystal effect
As illustrated above, the improvement of light harvesting is very important to enhance the solar cell performances. Besides light scattering centers, the photonic crystals (PCs), including one-dimensional (1D), two-dimensional (2D), and three-dimensional (3D) structures, can also be used to enhance the light harvesting. PCs are the optical materials with periodically changed refractive index. By adding PCs on the top of nanocrystalline film, PCs will greatly influence the light behavior inside the photoanode, to reflect the light, which would transmit through the photoanode back to the photoanode if the light is in the band gap of PCs, reabsorbed by dye molecules. The possible mechanisms of PC effect are photon localization, special photon behavior at the band edge, and light reflection.
The PC structures can be fabricated by various strategies, such as self-assembly, lithography, and laser writing. For example, TiO 2 -SiO 2 nanoparticle alternate layers have been deposited onto TiO 2 nanoparticle mesoporous layer by spin-coating to act as 1D PCs [32]. The fabrication of such multi-layers, however, needs complex processing procedures. Using soft-lithographic technique, 2D PCs with high periodicity have been produced, to enhance the photocurrent generation in DSSCs [33]. 3D PCs usually consist of opal/inverse opal structures from PS spheres [34][35][36], which can provide a complete photonic band gap and large enhancement in light harvesting.

Anodic photonic crystals
For PC applications, the excellent regularity of anodic nanostructures is required. The porous and tube structures by electrochemical anodization have the inherent 2D periodicity. Due to the high periodicity, porous Al 2 O 3 nanostructures have been applied as 2D PCs for applications such as lasing and light-emitting diode [37,38]. However, for TiO 2 nanotubes, due to the randomness of tube initiation, the regularity is not so satisfactory. To obtain highly ordered arrays, focused ion beam (FIB) sculpting has been used to achieve the patterned metal surface prior to anodization, which can guide the subsequent nanotube growth. This technique has been largely studied by Lu et al. [39], both on Ti and Al metals, and various kinds of patterns have been produced. The regularity can well meet the requirement of PCs, but the main problem is that the fabrication procedure is complex and time-consuming, which is not suitable for large-scale production. By using the so-called two-step or multi-step anodization [40], the dips formed by the first-step anodization on the metal surface can also act as the template to guide the tube growth, improving the tube regularity to a certain extent.

TiO 2 nanotube 3D photonic crystals
Usually, the anodized TiO 2 nanotubes have the smooth tube walls (Figure 7a). By the constant voltage anodization in certain electrolytes, the random and spontaneous current oscillation can lead to the formation of small ripples on the tube walls (Figure 7b). Inspired by this phenomenon, the fabrication of regular 1D nanostructures along the tube axis has been proposed. For Al 2 O 3 , Lee et al. [41,42] tried to fabricate 3D ordered nanoporous structure by mild and hard anodization in two different electrolytes. Periodic voltage anodization (cyclic anodization or pulse voltage anodization) has also been used to fabricate branched nanostructures for PCs [43]. The key point to induce the structure change along the longitudinal direction is the abruption of the established steady growth state of anodic nanostructures. TiO 2 possesses higher refractive index (n ~ 2.7) than Al 2 O 3 (n ~ 1.7), more suitable to be used as structural color materials to realize the complete bandgap. Hence, the strategies to induce ordered periodic structure along the TiO 2 tube axis have been greatly concerned. The bamboo-type TiO 2 nanotubes have been achieved by the periodic or pulse voltage anodization [44], which have certain longitudinal periodicity. However, due to the unstable nanotube growth rate under constant voltage anodization, the periodic voltage can only fabricate the structures with short range regularity. Furthermore, the structural parameters and thus PC characteristics cannot be preciously adjusted.
To realize the strict control of regularity, the periodic structures along the longitudinal direction can be fabricated by periodic current anodization [45]. Because the applied current is directly related to the growth rate of anodic oxide, the steady current can ensure the uniformity of the oxide growth and the control of the interrupt of current can lead to the periodic structures along the axis. During the current pause, the oxide growth stops, but the chemical dissolution continues, resulting in the structural difference on the tube walls ( Figure  7c). Unlike the smooth or rippled tube wall morphologies, the tube segments can be clearly observed with the concave shaped interfaces. In the range of about 20 periods from top to bottom, the distance between segments (period length) is almost the same, revealing excellent long-range regularity (Figure 7d). As discussed above, due to the higher reflective index, TiO 2 is more suitable photonic material than Al 2 O 3 . By using periodic current-pulse sequences, the photonic features of TiO 2 nanotubes, including period length, interfacial morphology, period number, and type of period (periodic, quasiperiodic, or aperiodic), can be precisely and continuously modulated by the anodization parameters. The periodic nanotube films (after detachment) with different period lengths show different colors and transparencies (Figure 8a), and corresponding different reflection spectra (Figure 8b). Furthermore, the structural color of the nanostructured film is not static. The film with fixed period structure shows different colors with different light incident and viewing angles (Figure 8c). TiO 2 nanotube PCs have been coupled directly to nanotube layers in DSSCs by a single step [46]. Also the thin PC membrane can be placed on the top of nanocrystalline TiO 2 layers as semi-transparent photoanode for DSSCs [47]. During the design of the solar cells, we should consider the match between the bandgap of PC structures and the incident light spectrum, to maximize the solar cell light harvesting. Thus, the different PCs with different band structures are designed and coupled to DSSCs. When using N719 dye as light absorber, it is found that DSSCs coupled with 150 nm periodic structure show the best performance. The advantages of the strategy using the above 3D nanotube PCs in solar cells are (1) tunable photonic structure, (2) controllable fabrication process, (3) easy integration by direct fabrication of both light absorbing layer and PC layer tubes by a single-step anodization or by membrane transfer, (4) tight mechanical and electrical connection between the light absorbing layer and PC layer, facilitating the charge transport, (5) interconnected two layers at the interface region, facilitating the electrolyte infiltration, and (6) easy fabrication of large area, transparent, and flexible PC films.

Substrate effect
In DSSCs, the widely used model describing the diffusion process in the TiO 2 electrode is the multiple trapping model. During the electron transport, the electron undergoes the trapping and detrapping processes by the trap states. To enhance the charge collection and thus efficiency, the fast transport and slow recombination of electrons are required. It has been reported that the order of the nanomaterials can greatly influence the electron property. For disordered nanoparticles, there exist numerous particle-particle interfaces. While for nanotubes, electrons transport along the tube axis, which would lead to higher electron transport rate [48,49]. However, in anodic tubes, the transport is not as fast as expected. Various studies have been devoted to investigating the possible underlying mechanism. The widely accepted view is that there exist a large amount of trap states in tubes, as compared with nanoparticles, which suppress the electron transport [14,50,51]. The trap states, usually oxygen vacancies or Ti 3+ ions, are most likely originated from non-crystallized amorphous regions, grain boundaries, and the impurities induced during the anodization process.
The improvement of tube crystallinity by high temperature annealing is supposed to be useful for the reduction of the trap densities. However, usually the substrate effect exists during annealing of anodic oxides on metal substrate. That is to say, the metal substrate can greatly affect the crystal phase and nanostructure of the oxide film layer during the annealing process. For TiO 2 nanotubes on Ti, rutile phase can be detected at a low annealing temperature of 400-450°C. Furthermore, the nanotubes are gradually condensed at high temperatures, and finally the porous structure is fully destroyed at about 700-800°C. This is due to the fact that the Ti foil can be directly oxided to the rutile phase during high temperature annealing, forming a thin and compact oxide layer at the oxide/metal interface region, and gradually becomes thicker [52]. This rutile layer will initiate the crystal phase and structure transformation of the upper TiO 2 layer, from the bottom to the top. The substrate effect exists at different annealing conditions, and even more severe when the annealing temperature increases [53][54][55]. Thus, anodic oxide films on metal foils can only be crystallized at a relatively low annealing temperature for DSSC applications. The high temperature annealing would cause the destruction of nanostructure and increase of resistance, both of which deteriorate the solar cell performance.

Highly crystallized nanotubes
To realize the high temperature annealing and thus highly crystallized TiO 2 nanotubes for DSSCs, the substrate effect should be eliminated. This can be realized by annealing of freestanding TiO 2 nanotube membranes before the attachment of the membrane onto the TCO substrate. For membranes without metal substrate, the substrate effect does not exist, and the crystallization behavior is completely different. There have been several attempts to anneal membranes at high temperatures [56][57][58]. The initiation temperature of phase transformation and structure densification becomes much higher. The main problem is that during the high temperature annealing, the membrane is inclined to curling and cracking, due to the low quality of the membranes fabricated by various strategies. By optimizing the self-detaching anodization process as discussed in Section 2, high-quality TiO 2 nanotube membranes can be obtained. The membranes can withstand the high temperature annealing up to 700°C and the hollow, porous, and ordered structure is maintained (Figure 9a-e) [59], although the crystallites in the tube walls gradually become larger (Figure 9f). The TiO 2 nanotube membranes with high crystallinity have been used as the photoanode in DSSCs (Figure 10a). The electron transport is found to be enhanced significantly because of the reduction of the impurity and defects and thus the electron trap states. As a result, the electron diffusion length is much longer in the highly crystallized nanotube membranes (Figure 10b). Despite the lower surface area and thus lower dye loading amount, a significantly improved solar cell efficiency of 7.81% has been obtained for 700°C annealed sample. The high temperature annealing for enhancement of crystallinity can also be used to fabricate flexible DSSCs. For such type of solar cells, the high temperature annealing is also not applicable because the flexible substrate (usually PET or PEN) cannot withstand high temperatures. For membranes, the high temperature annealing is completed before the transfer of nanotubes to the flexible substrate. Thus, the conductivity and quality of flexible substrate are not influenced by the high temperature annealing process. The only problem is how to keep the membrane adhered tightly to the flexible substrate, ensuring the electron transport pathways [60].

Small size nanotubes
Anodic TiO 2 nanotubes usually have low specific surface areas (BET surface area ~20 m 2 /g), as compared with other nanomaterials. For normal hexagonal closely packed nanotube arrays, due to the compact and ordered structure, large tube size, and smooth tube wall, the internal surface area is limited. When the nanotubes are applied in DSSCs as the photoanode material, the light cannot be fully absorbed, leading to low light harvesting efficiency. To increase the surface area, the most common strategy is the decoration of small diameter nanoparticle on the hollow tube inner or outer surface by TiCl 4 treatment or immersion filling, to fabricate tube/ particle mixed structure. The synthesis of bi-layered structure consisting of both tube layer and nanoparticle layer, double-walled nanotubes, and bamboo type tubes with rings or ripples on tube outer walls are all benefit for the improvement of surface areas.
Fabrication of small diameter but large thickness tube layer is another strategy to increase the surface area. Typically the small diameter tubes are very short [61], and thus the surface area per electrode area is still very low. The attempt has been focused on the fabrication of high aspect ratio tubes with small diameters and large thicknesses, increasing the dye loading amount per electrode area. In usual anodization conditions, it has been found that the outer diameter of TiO 2 nanotubes is proportional to the anodization voltage within a certain range, but inversely proportional to the bath temperature. By anodization at a low voltage and a high temperature, the diameter can be greatly reduced. When the bath temperature increases from 20 to 50°C, the outer diameter can be decreased from 93 to 75 nm [62]. On the other hand, the high bath temperature can promote the growth of anodic nanotubes, leading to the small tubes with large thicknesses, which can provide enough surface area for dye anchoring.

Water immersion treatment
Besides the in-situ anodization strategies, post-treatment method can also be used to tune the geometry and structure of nanotubes and guarantee the sufficient surface area for dye adsorption. By simple water immersion of as-grown TiO 2 nanotubes at room temperature (about 1-3 days) or hot water immersion, the tube wall morphology can be changed in a controllable way [63]. This is only useful for as-formed tubes before annealing, which are amorphous. The water treatment leads to the formation of hybrid-walled tubes with outer wall unchanged (smooth tubes) while inner wall converted to small nanoparticles (Figure 11a,b).
The TEM image clearly shows that in the hybrid structure, the tube outer wall consists elongurated nanocrystallites along the tube axis with lengths of dozens of nanometers to several hundred nanometers (Figure 11c). The small nanoparticles existing in the inner wall have the average crystal size of about 11 nm. The hybrid structure of the tube wall after water treatment appears to be similar to the filling or decoration of tube inner part with small nanoparticles. But it is apparently different than after water treatment, the solid tube wall appears to become thinner. This can be explained by the fact that in hybrid tubes, the nanoparticles in the inner wall are, in fact, converted from the tube wall. The formation of particles consumes the tube wall. During the water immersion process, the inner wall is inclined to be converted to nanoparticles. Normally the tube wall fabricated in organic electrolyte consists of inner and outer shells. The inner shell contains a large amount of carbon, which is originated from the anodization electrolyte in the anodization process [64]. Due to the different chemical compositions of the two shells, the inner walls are more easily to be converted by water to nanoparticles. This is a simple but effective way to enhance the tube roughness. Also along with the structure change, partial phase transition of nanotubes from amorphous to anatase has been observed. But if the tubes are annealed (>200°C), the tubes are stable, and the water immersion cannot cause any structural change even for a long time. It has been reported that the TiO 2 dissolution/precipitate process in water would cause the rearrangement of the construction unit of TiO 2 , leading to the spontaneous variation of crystal structure and morphology [65,66].
As discussed above, the conversion of the inner tube wall to nanoparticles can enhance the tube roughness and lead to the full utilization of the tube hollow space, increasing the internal surface area of anodic nanotubes, while keeping the tubular morphology unchanged. That is to say, the unique properties of tubes can be maintained, such as more convenient dye adsorption and electrolyte infiltration, short electron diffusion path, and slow electron recombination. After the subsequent annealing, the hybrid nanostructures are applied in DSSCs. The BET surface area increases with increasing immersion time. Before treatment, the BET surface area is about 20.1 m 2 /g, and increases to 39.9 and 42.7 m 2 /g after 2 and 3 days immersion, respectively. The 3 days water treated sample has a BET surface area about 2.1 times larger than normal tubes. Accordingly, the photoanode dye loading amount increases by 38.9% and 57.8% as compared with normal tubes. An optimized efficiency of 6.06% is obtained, improved by about 33%. Due to the maintenance of tubular structure, the water treatments do not affect the electron recombination property. This is beneficial for the improvement of the DSSC efficiency as compared with previous report that the nanotubes decorated with nanoparticles showed decreased electron lifetimes [67,68].

Conclusion remarks
This chapter discusses about the fabrication of anodic nanostructures by electrochemical anodization and their application in dye-sensitized solar cells to enhance the power conversion efficiency. The solar cell configurations can be optimized by free-standing anodic membranes to maximize the light utilization. By using large-sized nanotubes, due to the light scattering effect, the full interaction of the incident light with the absorbing layer can enhance the light harvesting of the solar cells. For periodic structures, the interface of the period is typically voids, resulting in the required periodic modulation of the refractive index for photonic crystals, which show the tunable photonic bandgaps, a property very attractive for light harvesting. By fabrication of high-quality membranes, highly crystallized nanotube membrane can be obtained, which can provide superior electron collection properties in solar cells. The surface areas of nanotubes can be increased by using high-aspect-ratio nanotubes and tube wall modification. The solar cells equipped with the proposed anodic nanostructures are expected to show excellent device performances, valuable for practical applications.