The anodic conditions and morphology parameters of the nanotubes, prepared in the electrolytes containing 0.175 M NH₄F and different volume ratios of water and glycerol, are listed in Table 1. SEM top view of sample 1, in Figure 1a, shows the porous film-like surface. The pits, with the obvious distance, indicate that the initial oxide layer has not been completely dissolved, but underneath which the TiO₂ nanotubes have already been formed. Similar phenomenon can also be seen in sample 5, in Figure 1e, and in the inset, neither gaps among the tubes nor are ridges on the side wall found. Sample 4, fabricated in the electrolyte containing more water, in Figure 1d, exhibits the thinner tube wall but the smoother layer surface (contrasted with those of samples 2 and 3).

Figure 1f shows the SEM bottom and bottom side (inset) views of sample 1. The tube bottom exhibits the hexagon-like cell morphology. Que Anh S. Nguyen and co-workers reported that the separation of the nanotubes from the substrate most likely occurs along the barrier layer/nanotube interface rather than along the titanium substrate/nanotube interface 20 . In this work, by mechanically bending the samples, the separated nanotube arrays were obtained, and the clear bottoms are shown in Figure 1f (inset). But from the bared Ti substrate, no oxide layer has been found.

As-formed TiO₂ nanotube arrays by anodization are commonly amorphous. As expected, in Figure 2, XRD patterns of samples 1 and 5 exhibit amorphous nanotube array structures. Hence, the XRD patterns of sample 4 are of considerable interest since the anatase (101) peak indicating the crystallization of the tubes. The samples 2 and 3 also exhibit slight anatase (101) peaks in the XRD patterns. At room temperature, crystallization of the tubes is influenced by some factors, such as the migration of Ti⁴⁺ outward from the substrate and O^2-/OH^- inward to form crystalline oxide in a ratio that favors crystal growth 21 , and the constraints imposed by the nanotube walls 13 22 . Sample 4 tends to form partial crystal structure since it, on one hand, prepared in the more water containing electrolyte which produced the faster ions mobility, can offer adequate ions that favor crystal growth, on the other hand, has the thinner tube walls, which impose the less constraints (in Figure 3). As to the samples 2 and 3, the slight crystallization maybe due to the internal stress relaxation of the tube wall that became thinner by dissolution after long time anodization. All XRD patterns, detected with the X-ray glancing angle at 4° show the intensive titanium diffraction peaks.

Que Anh S. Nguyen and co-workers reported that the barrier layer oxide is actually semi-crystalline in nature, while the tubes are amorphous in as-synthesized TiO₂ nanotube arrays 20 . In order to get further information about the tube microstructure, XRD trace, with glancing angle of 2°, was detected for sample 4. As shown in Figure 4, due to the low sampling depth, XRD pattern that represents the surface structure information only shows the strong anatase (101) peak and another diffraction peak at two-theta of 23°. This conforms that the crystal indeed exists in the tube walls, and the diffraction peak at two-theta of 23° maybe due to the metastable anatase phase in the tubes.

It has been suggested that nanotubes' growth results from the simultaneous oxidation and dissolution of the anodized oxide via F^- in an acidic electrolyte, described by the following reactions: 23

Ti + 2 H 2 O → TiO 2 + 4 e − + 4 H +

TiO 2 + 6 F − + 4 H + → [ TiF 6 ] 2 − + 2 H 2 O

Results from the present work indicate that the nanotube growth process could be a periodical model. Figure 5a shows the current–time curve recorded during anodization for sample 4. The current curve exhibits oscillations with asymmetric amplitude after the initial sharp slope while no steady stage as described by previous work 23 24 . Current oscillations, caused by periodical oxidation and dissolution 25 , imply that the nanotube layers grow in a periodical model. But what is the periodical unit? In Figure 5b, TEM image of sample 4 shows the clear tubular structure with the ridges. From the dashed line 1, one can observe that the ridges connect and exist in a flat place perpendicular to the tube. Between dashed lines 2 and 3, it can also be found that the tube bottoms are connected to form a layer. It is reasonable to believe that the bottom layer has a close relationship with the layered ridges. The bottom thickness, in the inset of Figure 5b, is about 60 nm and almost the same size as the distance between two ridges along a tube (Figure 1d). Herein, we present a layered growth model of TiO₂ nanotube arrays, in which the key processes can be described as follows.

(1) The first oxide layer is formed in the initial stage of anodization, and the reaction is described by Eq. 1. This rough layer endures high potential and thus some randomly distributed breakdown take place. The fast chemical dissolution of TiO₂ at these locations results in the pits formation, as described by Eq. 2.

(2) With the increase in the pits size, the electrolyte has the chance to infiltrate into the interface of oxide layer/metal and the second oxide layer is formed and then broken down again by the electric field. Thereafter, the oxide is formed and broken down layer by layer and the current curve exhibits oscillations correspondingly. The pores are formed due to the longitudinal prolongation and the lateral expansion of the pits with the increase in oxide layer number.

(3) With the pore size increasing, the internal surface area increases and the surface tension that causes the shrinking of the pore increases, too. If the adjacent pores become close adequately, the oxide among them would be pulled apart by the surface tension, eventually leading to the pore separation and the tube formation.

(4) Because the pore size increasing in the latter layer lags that in the former layer, the pore separation in subsequent layer occurs later. Therefore, some oxide between adjacent layers would be remained and thus the ridges are formed. In Figure 5c, the schematic diagram of the tube structure illustrates the periodical unit between two ridge flats and the barrier layer formed by connected bottoms. In this case, the tube structure is the result of the barrier layer moving forward Ti substrate periodically.