TiO2 Nanotubes: Synthesis and Applications

Reviews P. Schmuki et al. DOI: 10.1002/anie.201001374 Nanoscience TiO2 Nanotubes: Synthesis and Applications Poulomi Roy, Steffen Berger, and Patrik Schmuki* Keywords: anodization · electrochemistry · nanotubes · semiconductors · titanium dioxide Angewandte Chemie 2904 www.angewandte.org  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes TiO is one of the most studied compounds in materials science. Owing to some outstanding properties it is used for instance in photocatalysis, dye-sensitized solar cells, and biomedical devices. In 1999, first reports showed the feasibility to grow highly ordered arrays of TiO2 nanotubes by a simple but optimized electrochemical anodization of a titanium metal sheet. This finding stimulated intense research activities that focused on growth, modification, properties, and applications of these one-dimensional nanostructures. This review attempts to cover all these aspects, including underlying principles and key functional features of TiO2, in a comprehensive way and also indicates potential future directions of the field. 2 1.1. Introduction From the Contents 1. 1. Introduction 2905 2. Electrochemical Anodization and Self-Organization 2907 3. Properties of TiO2 and TiO2 Nanotubes 2917 4. Modification of Tube Properties 2920 5. Oxide Nanotube Layers on Other Transition Metals and Alloys 2926 6. Applications 2927 1.1. Why TiO2 Nanotubes? Ever since Iijima discovered carbon nanotubes,[1] this combination of extreme molecular geometry and exciting properties has not only inspired the field of nanotechnology but also triggered enormous efforts in physics, chemistry, and materials science. These one-dimensional (1D) nanostructures provide unique electronic properties, such as high electron mobility or quantum confinement effects, a very high specific surface area, and even show a very high mechanical strength.[2–4] Although carbon is still the most explored nanotube material, a considerable range of other materials, which are mainly transition metal oxides and sulfides, have been synthesized in a 1D or virtually 1D geometry (nanowires, nanorods, nanofibers, or nanotubes) and have also shown fascinating new properties and features.[2, 5–8] While carbon nanotubes are mainly explored for their use in microelectronic technology, inorganic nanotubes (especially metal sulfides or oxides) are mostly fabricated to exploit other material-specific properties, and the focus of interest is on biomedical, photochemical, electrical, and environmental applications.[5–18] Among all transition-metal oxides, TiO2 is the most extensively studied material (with more than 40 000 publications over the past 10 years), which makes TiO2 to one of the most investigated compounds in materials science. Bulk TiO2 is known to be a very useful non-toxic, environmentally friendly, corrosion-resistant material: it is frequently used in paint, white pigments, and sun-blockers. The key functional features are, however, exceptional biocompatibility (use in medicine) and even more the almost unique ionic and electronic properties of this oxide. TiO2 in all its crystal forms is a wide-bandgap semiconductor (Eg  3 eV) with suitable band-edge positions that enable its use in solar cells and for photocatalytic reactions. Photogenerated electron– hole pairs can be used for splitting water into oxygen and hydrogen (potentially the fuel of the future), or can be used for the remediation of hazardous wastes, such as contaminated ground waters, or the control of toxic air contaminants.[19–21] Historically, some milestones were certainly the reports by Fujishima and Honda in 1972 on water splitting on a TiO2 electrode,[22] and the works by Gerischer and Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 Tributsch,[23] Dare-Edwards et al.,[24] and Grtzel and ORegan[25] that introduced the use of the material for solar energy conversion in the 80s and 90s.[26] Over the past 20 years, the palette of potential applications has been widened towards devices with increasingly sophisticated photovoltaic, electrochromic, antifogging, or self-cleaning properties, biomedical coatings, sensors, or smart-surface coatings.[27–39] For many of these applications, it is crucial to maximize the specific surface area (which is, for example, obvious for any catalytic reaction) to achieve a maximum overall efficiency, and therefore nanoparticulated forms of TiO2 are widely used. However, other nanosize geometries, and in particular nanotubes or nanorods, may allow for a much higher control of the chemical or physical behavior. By diminishing dimensions to the nanoscale, not only the specific surface area increases significantly but also the electronic properties may change considerably (owing for example to quantum size effects, strong contribution of surface reconstruction, or surface curvature). These effects may also contribute to drastically improve the reaction/interaction between a device and the surrounding media, thereby making the system more effective (kinetics),[40–43] or even allow for entirely novel reaction pathways. Synthesis of 1D TiO2 nanostructures may be achieved by various routes including sol–gel methods, template-assisted methods, hydro/solvothermal approaches, and by electrochemical means.[8, 10, 14, 15, 36, 37, 44–58] A brief overview of various synthesis techniques is given in Section 1.2. The present review will, however, focus on perhaps the most spectacular and surprising 1D structure: self-organized TiO2 nanotube layers. These layers can be formed by a simple [*] P. Roy, S. Berger, P. Schmuki Department of Materials Science, WW4-LKO University of Erlangen-Nuremberg Martensstrasse 7, 91058 Erlangen (Germany) Fax: (+ 49) 9131-852-7582 E-mail: schmuki@ww.uni-erlangen.de  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 2905 Reviews P. Schmuki et al. electrochemical oxidation reaction of a metallic titanium substrate under a specific set of environmental conditions (Figure 1 a). This type of self-aligned oxide nanotube layers has attracted considerable interest over the past 10 years. Currently, the publication rate shows an almost exponential trend, with more than 1000 papers being published over the last 3 years. This rate can be ascribed to the fact that these structures represent a unique combination of the highly functional features of TiO2 with a regular and controllable nanoscale geometry (length, tube diameter, and self-ordering can be adjusted over large length scales). The synthesis is carried out by a low-cost parallel process: conventional electrochemical anodization. It is furthermore remarkable that the self-ordering anodization approach is not limited to titanium but can be applied to a large range of other transition metals or alloys to form highly aligned oxide nanotube or pore structures, which will be addressed more extensively in Section 5. In the following sections we will discuss the synthesis of these nanotubes and means to influence the degree of selforganization, tube length, diameter, and crystal structure of the tubes. We will introduce specific nanoscale morphologies that can be achieved (free-standing membranes, nanobamboo, tube stacks, tube-to-pore transitions), show methods to influence electronic and ionic properties (bandgap engineering, doping), and also biointerface properties. Finally, we give examples of using these nanotube structures in various devices and applications. Figure 1. The electrochemical anodization process and possible anodic morphologies: a) I) metal electropolishing, II) formation of compact anodic oxides, III) self-ordered oxides (nanotubes or nanopores), IV) rapid (disorganized) oxide nanotube formation, V) ordered nanoporous layers. Examples of morphologies of obtained structures: b) Classical highly organized nanoporous alumina (taken with permission from Ref. [116]), c) highly ordered TiO2 nanotubes (in top and side view) with dimpled structure (right) on metal surface when tubes are removed, d) disordered TiO2 nanotubes growing in bundles. RBA = rapid-breakdown anodization. 1.2. Electrochemical Synthesis versus Other Approaches This review will focus exclusively on electrochemical formation of self-organized TiO2 nanotubes in later sections, Poulomi Roy studied chemistry at Vidyasagar University, India, and obtained her PhD on Inorganic Nanomaterials from the Indian Institute of Technology, Kharagpur in 2007. She joined the group of Prof. Patrik Schmuki at University of Erlangen-Nuremberg, Germany in 2008 as a postdoctoral fellow. Her research interests include the synthesis of semiconductor nanomaterials and their application in various energy-related and biomedical fields. Steffen Berger studied materials science at the Friedrich Alexander University ErlangenNuremberg, Germany, and obtained his Diplom in 2005. He completed his PhD in 2009 in the group of Prof. Patrik Schmuki at the Chair for Surface Science and Corrosion, where he is continuing his research as a postdoctoral fellow. His research interests include the synthesis of self-organized nanomaterials by electrochemical methods with the direction of application in energy conversion and storage devices. 2906 www.angewandte.org Patrik Schmuki studied physical chemistry at the University of Basel and obtained his PhD from the ETH Zurich in 1992. From 1994 to 1997 he worked at Brookhaven National Laboratory, USA, and the Institute for Microstructural Sciences of the National Research Council of Canada. From 1997 until 2000 he was an Associate Professor for Microstructuring Materials at EPFL, Switzerland, and since 2000 he has been Full Professor and Head of the Institute for Surface Science at the Materials Science Department of the University of ErlangenNuremberg. His current research interests cover electrochemistry and materials science at the nanoscale, with a particular focus on functional materials and the control of self-assembly processes.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes but it should be noted that prior and parallel to the anodization approach, other methods were/are studied to synthesize TiO2 nanotubes. The first targeted effort to produce titania nanotubes was probably the work by Hoyer,[10] who used an electrochemical deposition method into an ordered alumina template. Later methods mainly involved other template-assisted methods, sol–gel techniques, and hydro/solvothermal methods with or without templates[14, 44, 47–50, 52, 56–59] and atomic layer deposition (ALD) into the template.[60–63] Many of the processes are acidcatalysed hydrolysis reactions of titanium alkoxide precursors followed by condensation reactions (that is, a gel-type polymeric TiOTi chain is developed, which then hydrolyzes and thus results in TiO2 precipitates). For example, TiO2 sol was sucked into the pores of the alumina template, then a heat treatment was applied, and afterwards the alumina template was selectively dissolved out.[52] TiO2 nanoparticles or nanorods are also prepared by using micelle templates of appropriate surfactants above their critical micelle concentration (the surfactant molecules aggregate and disperse in a liquid to give so-called spherical or rod-like micelles, which are used as template for TiO2 preparation). In this approach, nanotube formation is mostly carried out using water containing reverse micelles with a cylindrical exterior surface. The titanium precursor can then react at the micelle surface, and after removal of the surfactant (burn-off), a nanotube structure is obtained.[8] Usually TiCl4 or any other titanium alkoxide solutions are employed as the titanium precursor. Using certain H2O:micelle ratios[45, 46] allows the crystallite size to be varied. Another well-established solution-based method to synthesize 1D TiO2 nanostructures is the hydro/solvothermal method, which was first reported by Kasuga et al.[14, 47] In this method, bulk TiO2 powder is treated with NaOH solution in a Teflon-lined autoclave at 100–150 8C for several hours, followed by an acid treatment in HCl.[14, 47, 48] The general formation of the tube geometry is based on exfoliation of TiO2 crystal planes in the alkali environment and stabilization as TiONa+. This step is followed by a rolling of the nanolayer sheets into tubes during cooling or in the HCl treatment. Some debate exists in literature on the exact cause for the rolling-up, and several experimental factors are crucial to indeed obtain a tubular material. Such tubular structures consist of multishells, with an inner shell diameter of about 5 nm, a shell spacing of less than 1 nm, and an average tube diameter of about 10 nm. The amount of synthesized nanotubes and also their length (typically several tens to hundred nanometers) and size distribution depend on the specific reaction conditions.[48] In ALD, surfaces of templates (such as porous alumina) can be coated conformably one atomic layer after the other by using alternating cycles of exposure to a titania precursor (such as TiCl4 or Ti(OiPr)4) followed by purging and hydrolysis.[60–63] While alumina-template approaches allow nanotubular or rod structures to be formed that are vertically aligned to the substrate, they involve a sometimes critical template-removal step. For other transition-metal oxides, such as WO3, ZnO, ZrO2, V2O5, Nb2O5, Ta2O5, MnO, and CuO, nanorod or Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 nanotubular structures can also be fabricated using precursor and template techniques.[64] Applications target such areas as catalytic, optical, electrical, and mechanical devices and sensors. For example, CuO, V2O5, and MnO are very often used as intercalation compounds in lithium ion batteries;[65] in several cases nanotubes show enhanced electrochemical properties with both better lithium intercalation capacity and extraction rate than bulk or powdered materials. WO3 and Nb2O5 are the commonly used oxides in electrochromic devices, and gas sensors.[66–72] All these titanium precursor solution/template-based processes result in single tubes or loose agglomerates of tubes or bundles that are dispersed in a solution, and often a wide distribution of tube lengths is obtained. To make use of the structures in electrically contacted devices, the tubes are usually compacted into layers on an electrode surface. However, this process leads to an arbitrary orientation of the nanotubes on the electrode, which unfortunately eliminates many advantages of the one-dimensional directionality (for example providing a 1D electron path to the electrode). In contrast, the electrochemical anodization approach discussed herein is self-organizing: it leads to an array of oxide nanotubes aligned perpendicular to the substrate surface (such as in Figure 1) and to a well-defined (and controllable) tube length. The tubes are attached to the metal surface and are thus already electrically connected and easy to handle. The use of an electrochemical anodization method allows virtually any shape of titanium (and other metal) surfaces to be coated with a dense and defined nanotube layer and is thus an extremely versatile parallel (and thus easy to scale up) structuring process. Regarding anodic treatments, photoelectrochemical etch channels were also reported[56–58] and the growth of lessorganized (usually non-substrate-adherent) TiO2 nanotube bundles has been introduced by Masuda et al.[73] Later this process was referred to as rapid-breakdown anodization (RBA)[74] and extended to other electrolytes[73–75] and metals[74, 76] or metal alloys.[77] Solar-cell applications were usually targeted.[75, 78, 79] The main advantage of the technique is that comparably long nanotubes (several 100 mm)[76] can be grown within very short times (some seconds to minutes; an example is shown in Figure 1 d). The main disadvantage is that the tubes are not well-defined regarding length distribution, not well-organized over larger surface areas, and they are hardly connected to the substrate. The anodic formation of several 10 mm thick ordered nanoporous TiO2 layers (structure V in Figure 1 a) with channels several nanometers to several tens of nanometers in diameter has been recently reported.[80–83] These nanoscopic morphologies adhere well to the substrate, but to date they have not been explored extensively in view of achievable geometries or applications. 2. Electrochemical Anodization and Self-Organization Self-organized oxide tube arrays or pore arrays can be obtained by an anodization process of a suitable metal. When  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2907 Reviews P. Schmuki et al. metals are exposed to a sufficiently anodic voltage in an electrochemical configuration (as in Figure 1), an oxidation reaction M!Mn+ + n e will be initiated. Depending mainly on the electrolyte and the particular anodization parameters, essentially three possibilities for reactions exist (Figure 1): I) The Mn+ ions are solvatized in the electrolyte; that is, the metal is continuously dissolved (and corrosion, or electropolishing of the metal, is observed); II) the Mn+ ions formed react with O2 (provided by H2O in the electrolyte) and form a compact oxide (MO) layer if MO is not soluble in the electrolyte; III) under some electrochemical conditions, competition between solvatization and oxide formation is established (leading to porous MO). Under even more specific experimental conditions, a situation is established where selforganization during the growth of porous oxide takes place; furthermore, under some specific conditions, disorganized rapid growth of TiO2 nanotube bundles (IV in Figure 1 a, and Figure 1 d) or formation of thick self-organized mesoporous layers (V in Figure 1 a) can be observed. The best investigated system, in which almost perfect selforganization of pores in oxide can be established, is the anodic growth of porous Al2O3 on aluminum. Over 50 years ago,[84] it was realized that upon anodic treatment of aluminum in neutral to alkaline electrolytes, a flat, compact oxide would grow, but when anodized in acidic electrolytes relatively regular porous oxide structures could be grown to considerable thicknesses (> 100 mm). Such thick robust porous oxide layers on aluminum were used for many years in corrosion protection and coloration of aluminium surfaces.[85, 86] Around 1990, the first reports by Uosaki et al.[87] and later by Martin [88] appeared that explored the use of porous alumina as a template for the synthesis of functional nanomaterials. In 1995, in a ground-breaking paper, Masuda and Fukuda[89] reported that under specific experimental conditions, pore growth that shows a very high degree of self-organization can be achieved (see Figure 1 b). This finding triggered hundreds of papers dealing with the modification and use of porous alumina either directly (such as in filters or for photonic crystals) or indirectly (as a template for nanorod, nanowire, and nanotube synthesis by deposition of metal, semiconductor, or polymer materials.[88, 90–93] Models that try to provide a mechanistic reasoning for the occurrence of self-organization of porous alumina growth ascribe the origin to: 1) stress at the metal–oxide interface (volume expansion/electrostriction),[94, 95] 2) repulsion of electric fields,[96] or 3) establishing maximum current-flow conditions.[97, 98] As we will describe in the following chapters, many of the principles and mechanisms for self-organized porous alumina can be transferred to the formation of self-organized tubes or pores on other metals, taking into account some specific differences in the chemical character of the different metal oxides. In the light of recent findings we will also show that the distinction between porous or tubular morphology (as differentiated in Figure 1) is mechanistically not really justified. It should also be noted that the terminology pores or tubes was 2908 www.angewandte.org used in some early work interchangeably; that is, it represents pure semantics. 2.1. Self-Organized TiO2 Nanotube Arrays The first self-organized anodic oxides on titanium were reported for anodization in chromic acid electrolytes containing hydrofluoric acid by Zwilling and co-workers in 1999.[99, 100] This work showed that organized nanotube layers (although the author called the structure porous) of up to about 500 nm in thickness. The tube structure was not highly organized and the tubes showed considerable sidewall inhomogeneity. It was however recognized that small additions of fluoride ions to an electrolyte are the key to form these self-organized oxide structures. It is also noteworthy that in 1979, Kelly et al.[101] explored the influence of fluorides on the passivity of titanium and concluded that porous oxide layers formed for low fluoride concentrations, but they did not perform electron microscopy at a sufficiently high resolution to resolve the (likely) presence of self-organized TiO2 nanotube layers. After 1999, several approaches[102, 103] that used acidic fluoride-containing electrolytes essentially confirmed Zwillings findings (although the original Zwilling work was not always properly referenced). This early work established that the as-formed material is amorphous but may contain some crystallites,[103] that sidewall inhomogeneity is associated with current fluctuations,[103] and the limited thickness of the tube layers was ascribed to a oxide-growth/chemical-dissolution equilibrium (steady-state). Most crucial improvements to the geometry of the tubes were established by Macak et al.[104–106] First, it was demonstrated that the pH plays a crucial role in improving the tube layer thickness; that is, at neutral pH values, much longer tubes could be grown[104, 105] (other reports[107] turned out to be incorrect owing to false scale bars in SEM images). Second, it was demonstrated that in non-aqueous electrolytes, smooth tubes without sidewall inhomogeneity (ripples) can be grown to much higher aspect ratios and show a strongly improved ordering.[106] Using organic electrolyte systems, such as ethylene glycol, almost ideal hexagonally arranged tube layers can be grown to a thickness of several hundreds of micrometers.[108] Furthermore, an optimized two-step anodization treatment, similar to the one reported for porous anodic alumina,[109] can lead to virtually perfect hexagonally ordered arrays of TiO2 nanotubes.[110] Several other type of solvent systems, such as ionic liquids[111] or other protic solvents (CH3COOH), were explored,[112] but apart from the fact that in the latter approach comparably narrow tube diameters were observed, no further significant enhancement could be achieved. Regarding nanotubes, self-ordered surface dimples may also be encountered[113–115] (Figure 1 c). These dimples are in fact metallic surfaces from which the tube layers have been removed; that is, owing to the rounded oxide tube bottoms drilling into the metallic substrate, a scalloped interface results that becomes visible when oxide layers are lifted off by  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes voltage pulses, mechanical breakaway, or selective oxide dissolution.[116] Mzþ þ z H2 O ! MðOHÞz þ z Hþ z MðOHÞz ! MOz=2 þ H2 O 2 Anodization to form tube layers is usually carried out by applying a potential step (or ramp) at a constant voltage between 1–30 V in aqueous electrolyte or 5–150 V in nonaqueous electrolytes containing approximately 0.05 m–0.5 m (0.1–1 wt %) fluoride ions (and usually some background ion species). The growth of the oxide can then be monitored by recording the current–time characteristics (Figure 2 a). In Figure 2. a) Typical current–time (j–t) characteristics after a voltage step in the absence (a) and presence (c) of fluoride ions in the electrolyte. Either compact oxide (fluoride free) or porous/tubular metal oxide formation (containing fluoride) forms by different morphological stages (I–III). The inset shows typical linear sweep voltammograms (j–U curves) for different fluoride concentrations resulting in either electropolished metal (high fluoride concentration), compact oxide (very low fluoride concentration), or tube formation (intermediate fluoride concentration). b,c) Schematic drawing showing fieldaided transport of mobile ions through the oxide layers in the absence and presence of fluoride ions: rapid fluoride migration leads to accumulation at the metal–oxide interface. absence of fluoride ions, a compact oxide layer forms on the TiO2 surface according to Equations (1)–(3): z z H2 O þ z e ! H2 " þz OH 2 M!M þ ze z M þ H2 O ! MOz=2 þ z Hþ þ z e 2 Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 ð4Þ This is the classical anodization scheme, involving ion formation [Ti4+; Eq. (1)], reaction with O2 [created by (fieldaided) deprotonation of H2O or OH ; Eq. (2), (3)] and highfield ion migration of Ti4+ and O2 through the oxide (Figure 2 b). After initiation, the growth of an anodic oxide layer is determined by the field-aided transport of mobile ions through the oxide (Figure 2 b). Depending on the migration rate of the involved ionic species (Ti4+, O2), the growth of new oxide either proceeds at the interface between metal and oxide or at the interface between oxide and electrolyte (compact anodic TiO2 layers may in principle grow at either interfaces, but under most experimental conditions oxide grows at the metal–oxide interface). Under a constant voltage U, the field F = U/d drops constantly, thus lowering the driving force (for solid-state ion migration) with increasing film thickness d. The result is an (exponential) drop in the anodic current with time (Figure 2 a) until the field effect is lost (that is, is on the order of kT). At this point, a (practically) finite thickness is reached that mainly depends on the anodization voltage. For many transition metals (so-called valve metals), this final thickness is given by d = fU, where f is the so-called growth factor of the oxide (typically in the range of 2– 4 nm V1). Typically, the layer which is grown at the oxide/ electrolyte interface consists of less-dense oxide containing oxyhydroxides,[117, 118] while the layer at the metal–oxide interface consists of dense and stable TiO2. Often, electrochemical reactions, such as anodization processes, are characterized by current–voltage curves (Figure 2 a, inset). From these curves, threshold voltages Up for oxide formation, or the rates of electrochemical processes (for example ion-migration rates) may be determined. For oxide formation in absence of fluorides, a typical active–passive behavior can frequently be observed (Figure 2 a, inset). The presence of fluorides in the electrolyte strongly affects the anodization process, as fluorides form watersoluble [TiF6]2 species. On one hand, complexation occurs with Ti4+ ions that are ejected at the oxide–electrolyte interface [after migration through the oxide film; Equation (5)] and on the other hand by chemical attack of the formed TiO2 [Equation (6)]: Ti4þ þ 6 F ! ½TiF6 2 þ H 2 þ 2 H2 O TiO2 þ 6 F ƒ!½TiF 6  ð3Þ Simultaneously, the counter reaction, which is hydrogen evolution, takes place at the cathode [Equation (4)]: 2.2. Stages and Factors in TiO2 Nanotube Growth zþ ð2bÞ ð5Þ ð6Þ ð1Þ ð2aÞ Depending on the fluoride concentration, three very different electrochemical characteristics can be obtained[103] (Figure 2 a, inset). If the fluoride content is very low  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2909 Reviews P. Schmuki et al. ( 0.05 wt %), a characteristics as in the fluoride-ion-free case is observed; that is, after anodizing the material to a voltage above Up, a stable compact oxide layer is formed. If the fluoride concentration is high (ca. 1 wt %), no oxide formation can be observed, as all the Ti4+ formed immediately reacts with the abundant fluoride to form soluble [TiF6]2 (the reaction may then be controlled by diffusion of [TiF6]2 from the surface, and electropolishing of the sample occurs).[103] For intermediate fluoride concentrations, a competition between oxide formation and Ti4+ solvatization takes place and porous oxide or nanotube formation can be observed. A typical I–t curve for conditions that lead to nanotube formation is shown in Figure 2 a. The curve shows three stages: I) In the initial stage of anodization the curve essentially follows the fluoride-free case, and if samples are removed from the electrolyte, a compact oxide layer is present (see also Figure 3). In stage II, a current increase Figure 3. SEM images (cross-sectional views) of the initial phase of anodic TiO2 nanotube growth. First, a) a compact layer is formed, then b) local thinning and c) tube embryo formation occurs before d) stable tube growth is finally established. (Reproduced with permission from Ref. [127].) occurs, and irregular nanoscale pores are initially formed that penetrate the initial compact oxide (the current increases as the reactive area increases). In step III, the current drops again as a regular nanopore or nanotube layer forms. The penetrated compact oxide (random pore layer; initiation layer) often remains as remnants that are frequently found after anodization on the tube tops.[103, 119] Various efforts have been directed toward removing these layers,[119–121] as they interfere with any transport process into the tubes or hamper the exploitation of tube size effects.[122, 123] After self-organization occurred, the tube growth continues at steady current densities. For extended anodization, the growth may be 2910 www.angewandte.org determined by diffusional effects[37, 124] and thus agitation (and gravity effects[125]) and viscosity (influence on the diffusion constant[106]) may become important. The initiation of the growth of TiO2 nanotubular layers has been investigated in several works.[117, 126–128] A very recent study on the anodization of masked thin metallic titanium layers on silicon substrates enabled a detailed investigation of the very first growth stages;[127] a typical sequence of TiO2 nanotubular layer growth during the first 110 s is shown in Figure 3. Once the initiation phase is overcome, the tubes grow longer with time (Figure 4 a) until the etching action of fluorides (permanent thinning of the tube tops) becomes apparent,[124, 126] and a steady-state situation between tube formation at the bottom and etching at the top according to Equation (6) is established (see also Section 2.7.4). When tubes are growing in phase III (Figure 2 a), that is, under steady-current conditions, the magnitude of the current is significantly higher than when a compact oxide (at the same voltage) is formed. In terms of the driving force, this implies that an accordingly higher field must be present over the tube bottom (the barrier oxide), as the ion-migration rate is controlled by this oxide layer. A main role of fluorides is thus to maintain a thinner bottom oxide layer[129] by chemical etching of the oxide layer and immediate complexation (solvatization) of Ti4+ species arriving at the oxide–electrolyte interface. Another important factor is that fluoride ions are very small and compete with O2 migration through the TiO2 bottom oxide (Figure 2 c). In fact it has been observed that fluorides may migrate at a rate twice as high as O2 ions through oxide lattices. As a result, a fluoride-rich layer is formed at the metal–oxide interface.[118] The existence of this fluoride-rich layer was shown by Albu et al. by XPS and SEM analysis.[118, 130] In Figure 5 a, a XPS sputter-depth profile through the bottom of a nanotube layer is shown, and it demonstrates two important factors: that fluoride species are present over the entire layer, and that an approximately 20 nm thick fluoride-rich layer is present at the metal–oxide interface. This layer can also be seen by SEM, where a thin salt film is visible on the dimples of the titanium metal substrate after lifting off the TiO2 nanotubes (see Figure 5 b). The very important role of this fluoride-rich layer in the development of the further tube morphology is discussed in Sections 2.3. and 2.5. The nanotube layers, as they are grown in fluoride containing electrolytes, show a V-shaped sidewall thickness profile[118, 124] (Figure 5 c) that is due to exposure of the formed tubes to the etching fluoride-containing electrolyte (permanently etching and thinning the tube walls). The general shape and composition distribution can therefore be summarized as in Figure 5 c: The outer part of the shell (OST) is typically dense very pure TiO2, and the inner part (IST) of the tube is of a more loose nature and contains incorporated electrolyte components. Detailed SEM work showed that chemical dissolution of the inner TiO2 layer is faster and thus contributes more to the fact that the walls become thinner towards the top of nanotubes (Figure 5 d). The tube layers overall have a considerable fluoride ion content but the fluoride (or carbon) content decays strongly upon annealing.[118, 131]  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes 2.3. Factors Affecting Geometry and Composition Figure 4. a) TiO2 nanotube-layer thickness with anodization time for different electrolytes (anodization voltage for ethylene glycol electrolyte held at 60 V, and 40 V for other electrolytes): & water-based acidic,[103] ~ water-based neutral,[105] & glycerol,[106] * glycerol/H2O 50:50,[129] [364] b) Voltage dependence of the tube diameter for * ethylene glycol. different electrolytes: * water-based, ! glycerol/H2O 50:50, & glycerol, & ethylene glycol. c) Parameter study for growth rate and ordering of TiO2 nanotube layer in ethylene glycol electrolytes. Note that for conditions beyond optimized (that is, higher voltages or higher fluoride concentrations), no ordered tube layers are obtained (reproduced with permission from Ref. [108]). The inset images in (c) demonstrate the increase in ordering with parameter optimization. Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 While anodization time and etching rate define tube length, as described above, the diameter of the nanotubes is controlled linearly by the applied voltage.[124, 126, 129, 132] Some typical results are shown in Figure 4 b for some aqueous and non-aqueous electrolytes. The fact that the voltage dependence has a different slope in non-aqueous electrolytes can be ascribed to a large extent to the low conductivity of organic electrolytes and resulting IR-drop effects;[133–136] that is, the fact that the effective voltage of the electrode is Ueff = UnominalIR, where R is the resistivity of the electrolyte and I is the current. As reaction products are formed with anodization time, the conductivity of the electrolyte changes,[134] and thus variations in the diameter (larger diameters) are observed for longer anodization times. Investigations into optimized conditions for tube growth usually lead to results such as those shown in Figure 4 c, where the tube length as a function of anodization voltage and fluoride concentration was investigated. It is remarkable that the fastest growth conditions typically also represent conditions for optimized ordering. The nanotubes grown from organic electrolytes, such as ethylene glycol, DMSO, glycerol, or ionic liquids, show some significant differences in morphology and composition compared with nanotubes grown in aqueous electrolytes; owing to the low water content, very long tubes and large diameters (up to 700 nm)[137] can be grown. Most striking is, however, that in organic electrolytes, smooth tube walls can be obtained.[106] This is due to a lower water content in the electrolyte, which controls tube splitting (see Section 2.5). While a V-shape morphology is present for aqueous and non-aqueous electrolytes, in some organic electrolytes and for higher-voltage anodization, the inner tube layer (IST) shows an extremely high uptake of carbon from the electrolyte (see the table in Figure 13 c). That an inner carbon-rich contamination layer is present (for tubes formed in organic electrolytes, such as ethylene glycol) can be explained by the fact that the voltage-induced Schottky breakdown mechanism is operative for high-voltage anodization, leading to a decomposition of the organic electrolyte.[138] In aqueous electrolytes, the inner tube layer is typically more hydroxide-rich than the outer layer.[139–141] In organic electrolytes, aging of the electrolyte is often required to obtain highly defined tubes. Aging involves repeated anodization of dummy titanium sheets before the target sample is prepared. Very extensive work on aging of the electrolyte was recently performed by Lee et al.[134] Crucial factors in aging are the combined increase of the TiF62 content in the electrolyte (reduced TiO2 dissolution) and the reaching of a steady water content in the electrolyte (for example, ethylene glycol is relatively hygroscopic and tends to take up considerable amounts of water from environmental air). Under steady-state conditions, the nanotubular oxide layer thickness depends linearly on the anodization time; that is, the charge passed during anodization. However, due to etching by the electrolyte, at some point an equilibrium state between the growth of the oxide and chemical dissolution  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2911 Reviews P. Schmuki et al. Figure 5. a) XPS depth profile taken through the bottom of a lifted-off TiO2 nanotube layer. b) SEM images showing the fluoride-rich layer in the form of flakes on the titanium substrate and some left at the bottom part of nanotubes after lift-off the nanotube layer along with corresponding patches at the nanotube bottoms. c) Typical TiO2 nanotube formed in organic electrolyte and consisting of an inner (IST) and outer oxide shell (OST) and a fluoride-rich layer (FRL) (reproduced with permission from Ref. [118]). d) SEM images of TiO2 nanotubes taken at the top, from the fractures in the middle, and close to the bottom of a tube layer, illustrating the gradient in the tube-wall thickness (reproduced with permission from Ref. [110]). exists. As the etching rate, according to Equation (6), is even faster in acidic electrolytes, tube length is limited to 500– 600 nm in electrolytes at low pH. In neutral electrolyte systems, layer thicknesses of up to 2–4 mm can be obtained owing to the reduced chemical dissolution. However, owing to the hydrolysis reactions [Eqs. (2) and (3)], the pH at the tube tips is considerably lower than in the bulk electrolyte.[105, 142] Chemical dissolution can however be further decreased by anodization in glycerol or ethylene glycol based systems (reduced water content). In these electrolytes, the linear growth behavior could be significantly extended.[106] In principle, the expected length of the tubes could be calculated by converting the charge into an oxide volume (for example, by assuming 100 % current efficiency towards oxide formation and assuming the pores to be formed by purely chemical dissolution). However, the tube lengths are much larger than this estimate; that is, they would lead to a current efficiency of more than 100 %. Therefore, other tube-lengthening mechanisms, as discussed in Section 2.4 (oxide flow) must come into play. Etching of the tube at their top (in long-duration anodization experiments) leads to inhomogeneous top structures (Figure 6 a–c). The formation of needle- or grass-like morphologies (Figure 6 a–c) at the tube tops can be observed (which may be associated with the distribution of fluorides preferentially at boundary triple points, as outlined in Section 2.5) and collapsing and bundling tube tops (as the walls become too thin to support their own weight or 2912 www.angewandte.org withstand capillary forces when drying; Figure 6 a,b,d). As mentioned, there are various ways to protect tube tops against etching,[119–121] and an example is shown in Figure 6. Water content in the electrolyte affects the growth rate and the etching speed (chemical dissolution rate) of the nanotubes. In other words, the effect of water content is two-fold: it is required for oxide formation (tube bottom), but it also accelerates the dissolution of the nanotube layer (if the formed metal fluorides are water-soluble). A striking effect of the water content is that smooth tube walls are obtained for low water content, whereas side wall ripples are formed at higher contents.[106, 129, 143, 144] The reason for this effect is that for higher water contents, the fluoride-rich layer between the tubes shows a faster etching speed (chemical dissolution rate) than the growth speed of the tubes into the underlying substrate; that is, ripples at the walls of the tubes can be ascribed to the continuous etching and passivation of the cell boundary regions (see also Figure 9 b).[145] In principle, tubes can also be grown under galvanostatic conditions (that is, by keeping the anodization current constant).[146] However, this method has the drawback that the voltage may change with time and control over the tube diameter is therefore lost. Work by Taveira et al.[146] showed that during constant current anodization, the voltage indeed increases and fluctuates. Tube growth proceeds Figure 6. Typical etching effects observed with nanotube layers: a– c) Thinning of tube walls at the tops owing to etching in fluoridecontaining electrolyte leads to thinning and collapse of tubes with formation of nanograss (a,b) or perforated tube walls in the upper tube parts (c). d) Bundling at tube tops owing to a lack of mechanical support. e) Highly ordered grass-free and open tubes prepared by photoresist masking. f) Tube-wall thinning and formation of needles (grass).  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes up to a critical voltage, then a lift-off of the layers is observed and a new layer forms again. Other growth techniques are the use of alternating voltage (AV) techniques[147] or support by ultrasonic treatment;[148] the effects are interesting but overall not a very significant deviation from constant voltage results is observed. 2.4. Growth by Flow Mechanisms A remarkable feature of TiO2 nanotubes is that the tubes are longer than expected, as mentioned above and illustrated in Figure 7.[127, 155] The figure shows a cross-section of a lithographically defined nanotube area after anodic growth from a thin titanium layer. This approach allows the oxide expansion factors to be determined very well during the growth and they can be compared to the charge used to form the oxide. The expected expansion factor when metal is converted into oxide is given by the ratio of the volume of the oxide to the volume of the metal consumed and is frequently called the Pilling–Bedworth ratio (PBR).[149] The PBR of amorphous anodic TiO2[150] is 2.43, which corresponds to the dashed line in Figure 7 a (considering the recession of the metal front from its original level). However, Figure 7 shows that the tubes grew much larger than this expectation, and an expansion factor of about 3 can be observed from the crosssection. Figure 7 b shows the evolution of the oxide expansion factor (fexp) with anodization time. Early oxide growth in the initial phase occurs with a volume expansion of 2.4 (as expected for a compact anodic oxide) and proceeds, after a decrease, with an increase to a steady expansion factor of approximately 3. Only after very extended anodization is a decrease again observed, which is due to noticeable influence of tube-top etching. This unexpectedly high volume expansion is ascribed to an additional lengthening of the tubes by plastic flow. The flow concept was originally brought up for growth of porous anodic alumina in 2006 by Thompson et al.[151, 152] and was recently modeled by Hebert et al.[153, 154] For Al2O3, oxide flow is assumed to originate from the plasticity of the barrier layer generated by the substantial ionic movement in the high electric field together with compressive stresses induced by the volume expansion (PBR) and electrostrictive forces generated during growth. Another suggested origin for stress is the competition of strong anion adsorption with O2 entrance at the growing oxide lattice.[154] The result in any case is a force that pushes viscous oxide up the pore walls and thus extends the tube lengths. Overall, the increased volume expansion factors observed during the growth of TiO2 nanotubes[127, 155] along with other morphological and analytical[133, 156] findings suggest a similar mechanism for the growth of TiO2 nanotubes. Apart from length extension, another crucial implication is that the inward-migrating fluoride species will be pushed towards the cell boundaries by the flow and continuously accumulate there[118, 145, 157] (Figure 7 c). This accumulation of fluoride at the cell boundaries and even more at triple points leads to a chemical sensitization of these regions. Owing to the solubility of titanium fluorides in water, these regions are key in the formation of nanotubes rather than porous structures (outlined in Section 2.5 below). 2.5. Transition from Porous to Tubular Structures Figure 7. a) Cross-section of a titanium surface before and after anodic nanotube growth. The result indicates a much larger tube length than expected; plain volume expansion would lead to 2.4 (PBR), indicated by a dashed line (reproduced with permission from Ref. [155]). b) Dependence of the expansion factors of anodic TiO2 layers on the anodization time (reproduced with permission from Ref. [127]). c) Diagram of the flow mechanism pushing oxide and fluoride layer up the cell walls by viscous flow.[154] Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 At first sight, the self-organized oxide morphologies attained for aluminum and titanium seem to be considerably different (see Figure 1). However, bottom view images (Figure 1 c and 8 a) of TiO2 nanotube layers show that the bottom oxide is identical to porous layers. This means that when viewing the bottom or looking at cross-sections near the bottom, a compact hexagonal oxide morphology is present without a separation into individual tubes.[110] From a tilted bottom view (Figure 8 a), it is apparent that there is a continuous transition from a hexagonal porous to a tubular structure. Several studies carried out on Ti/Al alloys (that is, alloys that contain a typical tube-forming material titanium  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2913 Reviews P. Schmuki et al. Selective cell boundary etching was observed even for aluminum under specific conditions.[167–170] The effect of water concentration has also recently been studied for pure TiO2 nanotube formation, and indeed for extremely low H2O contents in a non-aqueous electrolyte a porous morphology is obtained.[171] The fact that anodization voltage also affects the pore to tube transition can be ascribed to a field effect on the fluoride ion mobility (faster or slower accumulation) and stress (by electrostriction), which in turn affects viscous oxide flow (and the fluoride layer). The separation mechanism of tube splitting by fluoride dissolution is also the main factor for the formation of sidewall ripples (see also Section 2.7).[144, 152, 171] 2.6. Self-Organization Length Scales (Tube Diameter) Figure 8. a) Formation of tubes from pore structures, illustrated by an SEM bottom view of a nanotube layer cracked off from the metal substrate at a slight angle. The bottom of a tube layer consists entirely of a close-packed hexagonal pore layer that is gradually converted into round tubes. b) Transition from a porous to tubular structure (left to right). c) Representation of the formation of a tubular morphology from an originally porous morphology by selective dissolution of fluoride-rich layers and preferential etching at triple points by H2O in the electrolyte. and a typical pore-forming material Al) were very useful in establishing why pores formed for some metals and anodization conditions but tubes formed under other conditions.[158, 159] For these alloys, oxide layers with distinct porous or tubular morphologies, or even transitional states, could be adjusted depending on electrochemical conditions (in Figure 8 b this continuous transition is shown for a Ti-54Al alloy).[158, 159] All these findings can be explained phenomenologically by the fact that during growth of the oxide (supported by plastic flow), an accumulation of fluorides takes place at cell boundaries (Figure 8 c). These regions are sensitized, as the fluoride-rich layer is prone to chemical dissolution: Selective chemical dissolution (in aqueous electrolyte) of the fluoriderich layer etches out the cell boundaries and thus leads to individual tube shapes. Under comparable anodization conditions, an ordered porous oxide is obtained for alumina but a tubular morphology is found for titania, which can be ascribed to the high solubility of TiF or TiOF compounds in aqueous media, whereas AlF species are much less soluble.[160] In other words, separation into tubes and the etching rate (and thus the point in oxide height where pore cells are split into separate tubes) is determined mainly by the fluoride concentration and the water content in the electrolyte (fluoride solubility). In this context, it is noteworthy that similar transition phenomena were reported on other valve metal alloys[36, 124, 161, 162] and also on pure valve metals, such as zirconium and hafnium.[163–166] In each case it was found that a key parameter affecting the morphology is the concentration of H2O in the electrolyte. By controlling the H2O concentration, the oxide morphology of anodic ZrO2 and HfO2 could be completely switched from nanoporous to nanotubular. 2914 www.angewandte.org The main point in tackling theoretically the reason of selforganization in porous or tubular oxide growth is the question of why a non-smooth surface is stable, as self-scalloping/ roughening is not favored in terms of surface energy. Although considerable experimental data exist that identify crucial parameters for self-ordering of porous or tubular structures, a comprehensive fully theoretical model that directly translates into quantifiable experimental data still does not exist. From work on Al, Ti, and TiAl, and also other alloys, it is clear that once tube or pore formation is possible (experimental conditions established), the applied voltage linearly determines the length scale of self-ordering (see also Figure 4). The difference in the various materials can be explained by the different high-field growth factors of the oxides.[124, 126, 172] Very elaborate theoretical work has been performed for self-ordering of porous alumina. To explain self-ordering, perturbation methods, and in particular longand short-wave stability analyses were used, but only in a few cases also associated with specific physical phenomena:[173, 174] to explain why a specific wavelength (tube or pore spacing) becomes stabilized over a flat surface. In light of the contemporary experimental findings, approaches that base stable self-ordering either on the specific ion-flux conditions[174b] (electromigration) or on the stress generated during oxide growth[174a] seem most adequate. In ion-flux models, a crucial component is adjustment of ion flux through the oxide and flux across the interfaces versus the interface shape, which under certain conditions can stabilize a short-wave instability. In stress-based models, a key question is whether purely mechanical (volume expansion) or electrostrictive effects dominate. While the length scale (wavelength) for pure Pilling–Bedworth volume expansion is far from accounting for the observed self-ordering length (100 nm range), scales estimated for observed electrostrictive forces are very likely to be in the required range (l  100 nm).[174b] In other words, the compressive electrostrictive stresses occurring in the oxide film, and in particular at the film surface, could be minimized by an increase of the surface area, that is, scalloping of the oxide film. However, the fact that optimized self-organization is typically observed for an optimized set of voltage combined with a specific current situation[110, 175] (see also Figure 4 c) and some preliminary stress measurements[176]  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes supports ion flux models that possibly need to be linked with approaches describing the final steady-state (growth) ordering by saturation effects. 2.7. Advanced Geometries 2.7.1. Tube Stacks, Bamboo, Nanolace, and Branched Tubes Various modifications in the tube geometry can be achieved by changing the anodization voltage during the tube growth process, and some examples are shown in Figure 9. Applying voltage steps can be used for example to generate tube stacks,[126, 144, 177, 178] or stratification layers thinned down sufficiently to continue growth under lower field conditions (the tube diameter will in this case adjust to the lower field and thus tube branching may occur). If tubewall separation (tube splitting) is faster than the etching process through the tube bottom, then the second tube layer will initiate in the space between the tubes. Stratification layers[36, 178] can be generated when stepping first to a lower voltage while holding the lower voltage for a time that is sufficient to establish diffusion and field conditions corresponding to the lower applied voltage, and then stepping back to the original high voltage. This process has of course to be carried out before tube bottoms are penetrated and a second tube layer grows. In this case the result is the formation of a new initiation layer between the tubes. There are two reasons for this: 1) When the voltage is stepped up, the thinned tube bottom immediately thickens in all directions as new high-field oxide is forming, or 2) because tube separation has reached the bottom but no secondary tube formation has yet been initiated. By exploring this behavior, regular tube structures with bamboo-like features can be fabricated by pulsing between two appropriate voltages.[178, 179] The distance between the bamboo rings can be altered by adjusting the holding times at different voltages. Bamboo-type nanotubes, with higher surface areas compared to normal TiO2 nanotubes, are promising in the field of dye-sensitized solar cells.[179] The oxide quantity formed during initiation process is different[119, 121, 128, 180] and more etch-resistant in fluoride electrolytes; therefore, after extended exposure to a fluoride containing environment, the tubes are selectively etched out, leaving behind the stratification layers that form a large uniform nanolace structure, which can be converted into large anatase networks.[178] 2.7.2. Double-Walled Tubes Figure 9. Advanced TiO2 nanotube morphologies: a) Bamboo nanotubes fabricated by alternating voltage anodization (reproduced with permission from Ref. [178]), b) smooth to bamboo-tube transition induced by variation of H2O content in the electrolyte (reproduced with permission from Ref. [144]), c) nanolace (reproduced with permission from Ref. [178]), d) branched nanotubes by voltage stepping (reproduced with permission from Ref. [178]), e) double-walled nanotubes, f) amphiphilic double-layer tubes (reproduced with permission from Ref. [177]). (bamboo).[178, 179] In tube stacks, a second tube layer can be grown underneath the top layer, if desired with a different diameter. Key to whether a second tube layer grows through the bottom of the top layer or starts in the V-shaped open space between the tubes depends critically on the experimental conditions.[36, 178] If the voltage is lowered during anodization, tube growth will be stopped or drastically slowed down, as the driving field across the bottom oxide is too low to maintain growth. However, owing to the permanent etching of the tube bottom in the fluoride environment, at some point the oxide is Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 As mentioned in Section 2.3., in some organic electrolytes formed at sufficiently high voltages, the nanotubes contain an inner carbon-rich shell (Figure 9 e) originating from electrolyte decomposition. Upon annealing, the carbon-rich layer can be thermally decomposed, leading to a remaining nanoparticulated inner tube wall.[118] This effect can be used to increase the inner surface area of the tube layers substantially. The exact structure of these double-walled tubes upon annealing is very sensitive to the annealing conditions, and in particular the temperature ramping speed and holding time, which allows for the fabrication of very robust membranes or entirely separated inner-shell/outer-shell structures.[118] 2.7.3. Amphiphilic Tubes Recently, it was reported that growing a first layer of nanotubes, treating them with an organic hydrophobic monolayer (octadecylphosphonic acid; ODPA), and re-growing tubes again in an organic electrolyte allows amphiphilic tube stacks to be fabricated[177] in which the lower tube layer grows through the bottom oxide of the first layer (Figure 9 f). Owing to the hydrophobic coating of the top layer and the naturally hydrophilic character of the freshly grown tubes, the  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2915 Reviews P. Schmuki et al. double-layer stacks have a highly amphiphilic behavior (see Section 6.5 for applications). Furthermore, the hydrophobic nature of the top layer strongly aids directing the initiation of the secondary layer. The tube walls are to a certain extent protected against the etching effect from the fluorides, but the voltage-induced chain scission effect (see Section 6.1) leads to preferential removal of the monolayer at the tube bottom. Furthermore, the V-shaped openings between the tubes are even more protected by monolayers owing to suppression of preferential wetting of the space in between the tubes (see Section 4.5).[181] typically some needles (often called nanograss or even nanorods[185]) remain on the tube top (see Figure 6 a,b). To counteract the formation of nanograss, various approaches have been reported[119–121] that often rely on protecting tube tops by some sort of etching resistant layers. These involve thermal rutile layers, sacrificial tube layers, or photoresist layers: these layers not only provide enhanced protection against etching of the tube tops but also suppress undesired local convection effects from the electrolyte. 2.7.4. High-Order, Defined Tube Tops A very spectacular effect, two-size-scale self-organization, was first observed when growing tubes on a complex biomedical titanium-based alloy.[186, 187] In this case, tubes of two distinct different diameters were formed under some anodization conditions that were highly ordered and arranged. This effect was later also reported for a range of more simple binary[124, 188–190] and ternary[191, 192] alloys. The origin of this highly unusual effect is still not entirely clear. While this effect was originally attributed to alloy composition, it seems however more related to geometry stabilization effects under certain anodization conditions.[36] This is most evident from very recent findings on two-level size-scale stabilization on pure titanium.[137, 193] The highest level of ordering in TiO2 nanotube layers is achieved in non-aqueous electrolytes, such as glycerol or ethylene glycol, and in fact by using ethylene glycol electrolytes under optimized conditions (electrochemical and electrolyte),[108, 118] very closely packed tubes with a hexagonal symmetry can be obtained (see Figure 4 c). The key for a high degree of ordering is to grow under optimized electrochemical conditions, which are typically conditions under which maximum current is achieved but oxide dielectric breakdown (sparking) is avoided. This is entirely in line with observations by Ono et al. that were made for highly ordered alumina growth that also is optimized at maximum current but avoiding burning events.[175, 182] Even further improved ordering of the tubes can be achieved using pre-structured surfaces that provide geometrical guiding. This was demonstrated[110] for example by tube formation followed by a removal of the tube layer and a second anodization. The organized dimpled titanium metal surface after removal of the first tube layer acts as initiation sites for tube growth in the second anodization step, which results in an optimized hexagonal arrangement. This procedure was originally established for aluminum by Masuda et al.[116] and has been used frequently since then to obtain highly ordered alumina arrays. A commonly undesirable feature is the presence of nonideal tube tops (as shown in Figure 6), and three different origins should be distinguished: 1) Initiation layer remnants that may cover the tube tops while highly ordered layers are growing underneath. These initiation layer remnants are etched off in the fluoride-containing electrolytes with time, and their creation can be minimized by using a voltage sweep approach[105] to the desired final anodization voltage rather than using direct voltage steps. In some cases these layers can be removed by adequate polarization/sonication techniques[183] or can be minimized by using sacrificial layers.[119–121] 2) Collapsing tubes: the tube tops are exposed to the electrolyte for the longest time and therefore are etched by fluorides the most. If the tube tops are considerably thinned down, the tube walls become comparably flexible and tend to stick together in irregular bundles (see Figure 6 d) owing to capillary effects while drying samples by gas streams or heating. This effect can to a large extent be avoided by supercritical drying[184] or keeping tube tops sufficiently rigid.[119–121] 3) Grass or needles: if etching thins down the tube tops even more, the tube walls start becoming perforated and disintegrate. The etching is not entirely uniform and 2916 www.angewandte.org 2.7.5. Self-Organization on Two Size Scales 2.7.6. Membranes The defined geometry (length and diameter) make nanotube layers very interesting for membrane-type applications, such as filtration or microphotoreactors (see Section 6.1). The strategy to produce such membranes typically consists of the formation of a tube layer, its separation from the substrate, followed by opening the tube bottoms. A typical free-standing membrane is shown in Figure 10. In the first report on the fabrication of such free-standing flow-through membranes,[123] nanotubes of 50–100 mm length were grown, then the underlying titanium metal was selectively dissolved in Br2/CH3OH, and then the oxide bottom of the tubes was etched open by HF vapor. Flow-through and photocatalytic activity was demonstrated for such membranes. Thereafter, several other approaches were reported, which mostly involve modifying the lift-off and opening procedure using for example sonication or voltage pulses.[194–196] However, the main problem remains the lift-off process (particularly if the membrane is thinner than 50 mm), which is increasingly difficult to handle without causing membrane rupture, and this becomes an even bigger challenge if larger membrane areas are desired. Furthermore, opening the tube bottoms uniformly over a large area by a chemical etching process (using HF, oxalic acid, or any other etchant) is difficult because as soon as tube bottoms are etched open at one location, the etchant creeps up the tube walls by capillary forces. This problem is most severe for thin membranes. A fundamentally novel approach was reported recently that eliminates these difficulties.[121] Albu et al. evaporated an aluminum film on a thin titanium foil, anodized from the titanium side, and grew tubes through the titanium into the underlying aluminum (using to advantage the fact that  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes Figure 10. A TiO2 nanotube membrane produced by anodic tube formation, lift-off, and bottom-opening processes. a) Flow through a membrane, b) optical image of a robust membrane (reproduced with permission from Ref. [118]) c, d) SEM images of a membrane open on both sides: c) top view, d) bottom view. ordered porous alumina can also be formed in fluoridecontaining electrolytes).[158] The remaining aluminum metal and oxide were then selectively chemically dissolved. Another highly beneficial feature introduced by Albu et al. is that the etched area was defined by a photolithographic process to form a grid structure. The remaining titanium metal frame allows for a high mechanical flexibility and excellent electrical contact to the enclosed nanotube packs. This strategy allowed large-scale membranes to be reliably created for the first time. 3. Properties of TiO2 and TiO2 Nanotubes The main reason why TiO2 nanotubes are currently attracting such a high interest is that some of the intrinsic properties of TiO2 provide the basis for many outstanding functional features. In the following, we will briefly discuss some of the most relevant material properties of TiO2 ; for more detailed material properties, the reader is referred to some excellent reviews.[16, 22, 28, 40, 197, 198] However, in the following sections, we will address several specific features relevant to TiO2 nanotubes. 3.1. Crystal Structure TiO2 exists naturally mainly in three crystalline phases: anatase, rutile, and brookite. Other than these, a synthetic layered phase, called TiO2(B),[199] and some high-pressure polymorphs also exist.[200] Moreover, TiO2 structures formed Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 at room temperature (sol–gel techniques, or anodic oxides) often are obtained in an amorphous form. Among the different polymorphs, rutile is generally considered to be the thermodynamically most stable bulk phase, while at the nanoscale (< 20 nm), anatase is considered to be stable, although there are some arguments in literature.[200, 201] Regarding nanoscale materials, it is important to realize that surface energy and stress have a significant effect on phase stability. The surface energy depends on the number of uncoordinated titanium cation sites. Fourfold-coordinated centers have larger surface energy than those with fivefold coordination, and the surface energy increases with the number of uncoordinated positions. In case of anatase crystals, the surface energy is lower than for rutile. However, the stress energy shows the opposite behavior. Both energies have some compensational effect; therefore, it is not surprising that a somewhat unclear picture exists regarding the most stable phase in nanoscale materials.[201, 202] Moreover, the presence of some ionic species in the lattice, such as Cl or F , also influences the stability of particular polymorphs.[180, 203–205] For example, rutile is more stable in presence of chloride ions.[203–205] In the case of anodic oxidation of titanium, the amorphous oxide layer can be converted by extended anodization times into rutile in fluoride-containing electrolytes, whereas in fluoride-free electrolytes, anatase can be obtained.[180] In anodic layer formation, the conversion from an amorphous TiO2 film into a crystalline anatase phase and dehydration effects are reported to depend not only on time but also on the applied voltage.[101, 206–209] Apart for stoichiometric TiO2, various suboxide phases exist that have very interesting properties. Substoichiometric compositions of titania may be described by the general formula TinO2n1 (4  n  10), but some defined phases are most remarkable, such as Ti4O7, Ti5O9, Ti6O11, Ti7O13, Ti8O15, Ti9O17, which are usually called Magnli phases.[210, 211] Such phases are made up of two-dimensional chains of titania octahedra; for example, Ti4O7 can be considered as three TiO2 octahedra and one TiO octahedron with oxygen vacancies at the edge of the unit cells. Such Magnli-phase titanium oxides are highly conductive (almost as high as graphite) and can be formed under reductive high-temperature treatment in H2[210] or C2H2.[212, 213] In fact, many properties of TiO2 strongly depend on bulk or surface structural defects, and in particular on the formation or presence of bulk or surface Ti3+ states or O2 vacancies.[40] On the surface of TiO2, unsaturated titanium cations, such as Ti3+, Ti2+, and Ti+, may for example be generated during vacuum or inert-gas annealing (TiO2 surfaces at low oxygen partial pressures tend to split off O2 or H2O from terminal oxide or hydroxide groups and form bridged oxide and Ti3+ states). Such effects can also be obtained under inert gas sputtering (for example, with Ar+) or e-beam exposure etc. Usually, Ti3+ states are found after such treatments, which are crucial for electronic and optical properties and also the surface reactivity. Under UV irradiation and in the presence of O2, adsorbed oxygen can be reduced to O2 and thereby inhibit the generation of Ti3+ sites. During long-time annealing, Ti3+ atoms may diffuse from the surface into the bulk of TiO2.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2917 Reviews P. Schmuki et al. TiO2 nanotubes, after electrochemical formation, are amorphous, and some reports[103, 118, 214] indicate the presence of nanocrystallites in the tube wall, particularly if anodization is carried out at higher voltages. For example, in acidic electrolyte (H2SO4), depending on the anodization conditions (sparking, potentistatic, or galvanostatic), the as formed oxide film may consist of anatase, a mixture of anatase and rutile, or rutile crystallites.[129, 215–217] Reports that suggest the presence of crystallites in as-formed oxide nanotubes and which are based on HRTEM investigations need to be examined very critically, as amorphous TiO2 seems to be particularly prone to electron-beam-induced crystallization.[214] Intended conversion of the tubes into crystalline material (anatase/rutile) by annealing is discussed in Section 4.1. In general, amorphous material (such as tubes and powders) can be converted into anatase (ca. 300–500 8C) or rutile (> 550 8C) by a thermal treatment in air. It should be noted however that surprising size effects on these transitions have been observed for nanotubes[201] and that several alloying elements affect the transition temperature.[188, 189, 218] There are some reports that show that doping with Nb, Al, Ni, Ga, Ta, and W retards the growth of anatase and rutile crystallites and delays the anatase-to-rutile phase transformation.[219–221] On the other hand, Mn, Fe, Cu, and Zn are generally believed to promote the phase transformation. For TiO2 nanotubes, niobium addition in particular was reported to lead to a significant temperature shift for anatase conversion.[188] The process is believed to be due to the inhibition of nucleation of anatase crystallites by the niobium dopants.[222, 223] In the case of zirconium[189] and tungsten,[224] slightly elevated temperatures are also needed to convert the tubes into anatase. conductivity of anatase and rutile TiO2 films reveal that the conductivity of undoped anatase and rutile is more or less similar (range of 104–107 W cm). When thermally reduced, the conductivity for anatase and rutile is significantly different (in the range of 101 W cm for anatase and 102 W cm for rutile); for anatase the conductivity becomes independent of the temperature.[227, 232] A wide range of studies investigate the optical properties of TiO2 nanotubes, mainly in photoelectrochemical arrangements in an electrolyte, or by measuring optical absorption coefficients in reflectivity or transmission mode in air. For asformed tubular layers, a photocurrent behavior (Figure 11 a) is often obtained with a band (mobility) gap of approximately 3.1–3.3 eV for the amorphous tubes.[233] After conversion to anatase, the photocurrent strongly increases and a gap of 3.2 eV results. If annealing is performed to form rutile, the gap is correspondingly narrower and is about 3.0 eV.[233] There are two major observations for tubes in comparison to 3.2. Optical and Electrical Properties In TiO2, the relevant energy levels that form the band edges and thus define the bandgap are considered to be the Ti 3d states and O 2p levels. The lowest empty energy levels are Ti dxy and thus they are representative of the conduction band (CB) edge, whereas full O 2p states (in particular, nonbonding pp states) define the valence band (VB). Both anatase and rutile show this general distribution of states.[225, 226] TiO2 has an indirect optical bandgap for anatase of 3.2 eV, for rutile of 3.0 eV, and amorphous material is reported to have a mobility gap of about 3.2–3.5 eV.[225, 227–230] Crucial for the optical and electrical properties is the presence of defects that provide additional states in the bandgap near the CB or VB. Particularly important in TiO2 are oxygen vacancies and Ti3+ states, which to a large extent dominate optical and electrical properties of the material. Typically Ti3+ states are situated about 0.2–0.8 eV below the conduction band (see also Figure 14 a).[197, 231] Electrons may be trapped in the Ti3+ sites or the holes may be trapped in oxygen states, leading to radicals covalently linked to the TiO2 surfaces. Such shallow defects can release trapped charge carriers easily to the nearby conduction or valence bands by thermal excitation. The Ti3+ content seems to entirely dominate the conductivity of TiO2 crystals. Investigations on the typical 2918 www.angewandte.org Figure 11. Photocurrent response from TiO2 nanotubes:[235] a) IPCE versus wavelength spectra of different TiO2 nanotube layers (amorphous and annealed) in 0.1 mol dm3 Na2SO4 : & amorphous (NTs or compact oxide), * 2.4 mm (annealed), * 2.4 mm (annealed) with CH3OH. Inset: Incident photon-to-current conversion efficiency (IPCE) dependence at 350 nm on the applied voltage and different tube lengths in 0.1 mol dm3 Na2SO4 in presence (^) and absence (^) of 2 mol CH3OH. b) Photoresponse spectra of TiO2 nanotubes (thickness ca. 1.8 mm) fabricated in aqueous Na2SO4 solution (&) and in ethylene glycol (*), showing altered sub-bandgap response in carbon containing electrolytes.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes compact layers: 1) To obtain a contribution to the photoresponse from the tube walls, the tubes need to be annealed; the amorphous material contains such a high density of recombination centers that there is virtually no photoconductivity in the tube walls; 2) for tubes annealed to a crystalline material, the doping level is typically very high (1018–1019 cm3[234]); that is, if in a photoelectrochemical configuration a voltage is applied at relatively moderate bias (ca. 0.3 V anodic to the flat band potential), total carrier depletion of the walls is reached[235] (using capacitance data and the Schottky approach).[234, 236–238] In other words, the space charge layer follows the wall contours only up to this threshold voltage (and may aid charge separation); at higher voltages the depletion layer cannot extend any further. This is the key reason why for tubular shapes a deviation from the classical Grtner model[239] for the potential dependence of the photocurrent is observed.[235] In general, the determined density of bulk states (1018– 19 10 cm3) is much higher for nanotubes than for conventional nanoparticles. As a result, charge-carrier transport in TiO2 nanotubes is much slower than in comparable nanoparticulate systems, as the bulk states strongly affect the trapping/ detrapping-dominated majority-carrier transport. However, the electron diffusion length under UV illumination is up to 30 times higher than for comparable nanoparticle layers owing to lower surface recombination.[235] The overall photocurrent in tubes and particles is determined to a large extent by surface recombination effects, as is evident from measurements carried out in presence of a hole scavenger (Figure 11 a). When methanol is added to the electrolyte, the diffusion length for photogenerated electrons is drastically enhanced owing to a reduced recombination probability with surface-trapped holes.[235] For tubes, the electron diffusion length depends extremely on the type of tubes used, whereas rough-tube walls (first-generation tubes formed in aqueous electrolytes) lead to a diffusion length of Ln  1–2 mm; for second-generation (organic electrolyte) tubes, Ln  25 mm can be obtained. Another interesting point is that photocurrent spectra for different types of nanotubes show a slightly different subbandgap response. For nanotubes prepared under double wall conditions,[118] that is, in some organic electrolytes at higher voltage, a tail in the photoresponse is observed owing to carbon remnants (Figure 11 b; see also Section 4.2 on carbon doping). Another optically interesting feature of TiO2 is its comparably high refractive index (n = 2.5 for anatase to 2.9 for rutile)[240] in comparison with other materials that are used to manufacture 3D optical structures (SiO2, n = 1.5; Al2O3, n = 1.7).[241] This high refractive index should allow to create very interesting photonic materials that for example are predicted to show a negative-refractive-index photonic crystal structure.[242] In practice, however, to fully experimentally realize such structures, some additional control over geometry and order of the tube arrays is needed. Optical nanosize effects in terms of bandgap widening have not been observed for TiO2 nanotubes, although the tube walls in some cases come close to the dimensions (5–10 nm) where effects could be expected.[243–245] The size, however, is Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 sufficient to observe phonon confinement in tube walls,[222, 223, 246] which is dependent on crystallite size in the wall and depends on whether 1D or 2D confinement is established. Most remarkable, however, is an unexpected phase stabilization upon annealing.[201] While annealing for nanotubes at moderate temperatures (> 200 8C, < 550 8C) usually converts nanotubes with a diameter of more than 30 nm to anatase, for smaller diameters, rutile can be stabilized. The electrical conductivity of self-organized TiO2 nanotubes on titanium metal has been estimated by two-point measurements between a metallic contact on the top of the tubes and the back metal substrate.[247] Some results as a function of annealing temperature are shown in Figure 12 a. Although two-point measurements may not yield entirely reliable values (in absolute terms), the behavior shown is very characteristic for all TiO2 nanotube layers. For low temperatures (< 200 8C), the resistivity increases with temperature, owing to evaporation of surface water. At about 300 8C, conversion of the amorphous material to anatase occurs, and a significantly higher conductivity is obtained. At temperatures higher than 500 8C, the anatase material is increasingly Figure 12. Two-point conductivity measurements for TiO2 nanotube layers a) annealed for 2.5 h at different temperatures in air[247] and b) I–U characteristics for anatase tubes before (&) and after (*) carbon doping in C2H2 and conversion into semimetallic tubes (~; hightemperature C2H2 treatment; reproduced with permission from Ref. [212]).  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2919 Reviews P. Schmuki et al. converted into more resistive rutile, which leads to a considerably lower conductivity. Using reducing and carbonizing annealing conditions, the conductivity strongly increases,[247] and reduction to titanium oxocarbides can even provide a virtually metallic behavior.[212] 3.3. Reactivity TiO2 surfaces have been extensively studied regarding gas-phase adsorption and catalytic effects on various reactions (for a comprehensive overview, see reference [40]), such as CO oxidation,[248, 249] selective reduction of NOx[250] and O2 and water decomposition.[251–253] Rutile TiO2 (110) in particular is the preferred substrate for UHV studies owing to the comparable ease with which these single crystal surfaces can be properly prepared. Overall, it has been shown that surface defects are the crucial reactive sites for many reactions. For example, on defective (110) rutile surfaces OH-surface groups can be formed after H2O adsorption,[254–256] which is proposed to be due to a reaction of bridged oxide with an adsorbed water to form hydroxides. For H2 adsorption, oxygen vacancies were reported to act as electron donors,[256, 257] and TiH can be formed. Such hydrogen atoms may recombine or diffuse into the bulk of the TiO2 and dope the material; an according increase of conductivity is observed[258, 259] (see also Figure 14 a). Studies on rutile surfaces with very low defect densities showed that defects (oxygen vacancies) were crucial for CO adsorption.[256, 260] Similarly, NO oxidation[256, 261] and SO2 reactions[256, 257] strongly depend on the presence of oxygen vacancies or presence of Ti3+ on the surface. Whereas NH3 only weakly interacts with defects,[240, 256] H2S may be dissociated.[256, 262] Ti3+-rich surfaces show considerable reactivity regarding deoxygenation reactions of organic molecules, such as alcohols.[256, 263, 264] Regarding O2 adsorption, significant adsorption and O2 formation[256, 257] was only observed on defective surfaces, and no oxygen adsorption could be detected on intact surfaces.[256, 265] Some reports exist regarding O2 dissociation at elevated temperatures, and this fact is in line with the observation that Ti3+ states can be easily annealed out (oxidized) using high-temperature O2 exposure.[256, 266] These examples may illustrate why reactions on nanoscale TiO2 particles and nanotubes are highly promising. Nanoscale materials can have an extremely high density of defect sites (lattice corner and kink sites), may possess extreme surface curvature, and may stabilize unusual crystal orientations or crystal phases. For TiO2 nanotubes, proper UHV surface studies were mainly carried out by Burghaus et al.[248, 249, 251–253] They investigated CO, CO2, H2, O2, and alkane activity on different crystal structures of TiO2 nanotubes. Currently, reports on unexpected O2 adsorption on anatase TiO2 nanotubes are highly interesting.[251] TiO2 nanotubes are the first example of an intrinsic TiO2 nanostructred material for which active O2 adsorption could be observed; this can usually only be observed for samples where active defects were introduced, for example by ion bombardment or for some powders. 2920 www.angewandte.org In solution, the most important reactive features of TiO2 are: 1) its solubility in some complexing agents (HF, organic acids), 2) the possibility of modifying the surface with organic monolayers by surface hydroxy group reactions (discussed in Section 4.5), 3) the feasibility of inducing apatite formation (see biomaterials in Section 6.5), and 4) its electrochemical properties. Electrochemically, anodically formed amorphous TiO2 layers behave like a highly defective n-type semiconductor, and annealed to anatase or rutile like a highly doped n-type semiconductor.[233, 234, 237, 238, 267, 268] A very important electrochemical feature is the ability to reduce Ti4+ in the TiO2 lattice to Ti3+ if a sufficiently negative potential is applied. The reduction reaction is accompanied by insertion of small cations (Y+ = H+, Li+ from the solution) into the TiO2 lattice [Eq. (7)]: Yþ TiIV O2 þ e ƒ!TiIII YO2 ð7Þ There is a color change associated with formation of the reduced TiIII material (usually dark blue to brown) that can be exploited for suitable electrochromic devices.[269–273] For TiO2 and other transition-metal oxides, the electrochemical reduction and re-oxidation are reversible with a switching time that is determined by the solid-state diffusion/migration of Y+ into and out of the lattice. Therefore, nanotube surfaces with nanoscale wall thicknesses allow comparably fast switching in the millisecond to second range (see also Sections 4.4 and 6.3). Under anodic polarization, a current blocking situation is established in n-type TiO2 electrodes. The material can thus be exploited for photoelectrochemical reactions, such as for photocatalysis (Section 6.1) or in solar cells (Section 6.2). In all these applications, the directionality of the nanotube layers may be beneficially exploited. TiO2 nanotubes have some beneficial effects as a substrate for noble metal catalysts in electrocatalysis. For example, it has been shown that for TiO2 nanotubes decorated with gold, a more facile O2 reduction reaction can be observed,[274] or that a highly accelerating effect for methanol oxidation catalysts can be obtained.[212, 275–278] Such electrocatalytic effects of TiO2 nanotubes have also been explored for glucose sensing.[279] All of these applications rely on the fact that owing to carrier depletion conditions in the TiO2 substrate, a high overpotential for the oxygen evolution reaction is provided. A particularly beneficial effect of the tubular substrate as a support for nanoparticles is not only the large surface area, but may be even more importantly the compartization of the electrode surface. The nanotubular arrangement can contribute to avoid catalyst aggregation and thus lead to an increase in reactivity and specific reaction rates. The loadings of expensive noble metal catalysts can be therefore considerably reduced. 4. Modification of Tube Properties Modification of the TiO2 nanotubes is mainly carried out by 1) heat treatments, 2) introducing other elements, or 3) by tube-wall decoration. The aim is to make the material suitable  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes (or even more suitable) for various applications that rely on specific electrical, optical, or chemical properties. In view of electronic properties, annealing to a crystalline structure mainly changes the conductivity and lifetime of charge carriers, while active doping or bandgap engineering by introducing other elements targets decreasing the optical bandgap, thus enabling a visible-light photoresponse. Particle decoration is often used to increase surface catalytic effects and to create solid-state junctions. Some of the modification approaches are borrowed from techniques that are already established for bulk or nanoparticulate TiO2 ; others are specific for anodic nanotubes. For anodic nanotubes, the fact that the synthesis route for the nanotube arrays is an electrochemical oxidation of a metal piece can be taken advantage of, and therefore some specific doping principles become feasible (for example using an alloy of TiX for anodization, where X is another metal that may act in its oxidized form as a doping species). In many conventional applications, nanoscale TiO2 is either used as isolated nanoparticles (for example, in a suspension) or as an electrode consisting of TiO2 nanoparticles that are compacted or sintered on a back contact. The latter case significantly differs from isolated particles in a solution, as the charge transfer for electrodes through the entire layer to the back contact is crucial. This, of course, also applies to nanotubular oxide electrodes, where charge transfer in the tube walls along the tubes to underlying titanium or the back contact is very important. Therefore, intense efforts focus on modifying the electrical properties of TiO2 nanotubes. 4.1. Annealing As mentioned above, the as-prepared TiO2 nanotubes are amorphous in nature but can be annealed (under oxidizing conditions in air or O2) to anatase or rutile.[104, 118, 233, 247, 267] Examples of typical structural and compositional characterization before and after annealing are shown in Figure 13. Detailed XRD investigations show that significant conversion of nanotube layers into anatase begins at around 280 8C. With an increasing annealing temperature, anatase formation is enhanced.[280] Above about 500 8C, the rutile phase starts to appear, and with a further increase in the annealing temperature the quantity of rutile increases; for example 12 h annealing in air results in 4 % rutile at 500 8C, 17 % at 600 8C, 30 % at 700 8C, and 46 % at 800 8C. At 900 8C and relatively short-time annealing (2 h) and with a ramping speed of 1 8C s1, the TiO2 nanotube layer is completely converted into a rutile structure. The annealing duration and ramping speed also has a significant effect on the crystal structure of TiO2 nanotubes and furthermore affects the composition of the nanotubes. As-formed nanotubes contain a significant amount of fluoride species[104] (Figure 13 b,c), and if formed in most organic electrolytes (see Figure 13 c), a significant amount of carbon is contained, being decomposition products of the electrolyte.[137] These species can be driven out to a large extent by annealing.[118] Typically, the as-formed tubes also contain some surface hydroxide groups (evident for Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 Figure 13. Annealing of TiO2 nanotube layers: a) XRD pattern of TiO2 nanotube layers as-formed and after annealing at different temperatures in air (reproduced with permission from Ref. [280]). b) XPS spectra of O 1s, C 1s, and F 1s peaks before (c) and after annealing (a). The arrow denotes an OH shoulder peak; BE = binding energy. c) Composition of tubes (according to the EDX analaysis shown at the bottom) after formation in different electrolytes (the degree of natural carbon contamination is about 2–3 atom %). Inset: conditions for double wall formation; O yes, X no (reproduced with permission from Ref. [137]). d, e) SEM and TEM images of TiO2 nanotubes showing cracks at the tube wall after annealing at 450 8C (reproduced with permission from Ref. [35]). f) Rutile layer formed underneath the TiO2 nanotube layer after annealing at 650 8C for 3 h.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2921 Reviews P. Schmuki et al. example in the O 1s XPS peak); this amount is also considerably decreased when the material is annealed.[140, 281] Two undesired effects that can occur in annealing are shown in Figure 13 d–f. When annealing tubes at temperatures above 450 8C, some cracks can occur in the tube walls (Figure 13 d,e), mostly in the OST layer (compare with Figure 5); these cracks were considered to slow down electron transport.[35] Furthermore, by annealing in O2-containing atmospheres, thin rutile layers underneath the nanotube bottoms are formed (Figure 13 f). This formation is due to direct thermal oxidation of the titanium metal substrate to rutile TiO2 during annealing. The higher the temperature and the more oxidizing the conditions, the thicker the solid layer. While the layer is typically in the range of 20–100 nm at 300– 450 8C, at higher temperatures and in an O2 environment, it can reach micrometers in thickness. Such a rutile layer under the nanotube bottom can detrimentally affect various applications that use electrode configurations, such as solar cells,[35] owing to the inferior electronic properties of rutile layers compared to anatase. Annealing can also affect the nanotube morphology. Usually, for extended annealing, TiO2 nanotubes are stable up to 650 8C, but at higher temperatures the tubes start collapsing.[118] For tubes formed in ethylene glycol, the morphology strongly depends on the ramping rate. Single-walled, doublewalled, or fused membrane structures may be obtained.[118] Depending on the annealing ramping rate, the TiO2 nanotube walls consist of crystallites of the range of few nm to 200 nm.[118] As shown above (Figure 12 a), conductivity along the tubes is affected to a large extent by the heat treatments, and therefore appropriate annealing is essential for many applications.[35, 247] A crucial point is the formation of Ti3+ species during heat treatments. Heat treatment in absence of oxidizing species or in vacuum usually leads to loss of O2 from the material and formation of Ti3+ (at least in the near-surface region). The reduced material that is formed shows visible light absorption and enhanced conductivity. This effect can be regarded as introducing a doping band or a high density of localized donor states (Ti3+ being an electron donor species located close to the conduction band of TiO2 ; see Figure 14 a). By usual thermal treatments, approximately 1 % of the lattice atoms may be reduced to Ti3+. A very noticeable effect of annealing temperature is also observed in the detailed tube wall morphology after annealing. Annealing in different gas atmospheres, such as air, N2, O2, or a N2/H2 mixture, leads to a different anatase/rutile ratios of TiO2.[282] 4.2. Doping Over the past 10–15 years, considerable effort has gone into doping or bandgap engineering of TiO2 by introducing a secondary electronically active species into the lattice. The main thrust comes from photocatalytic or photovoltaic applications: To exploit the solar spectrum much better, the onset energy for light absorption needs to be decreased. The intrinsic bandgap of 3.2 eV allows the material to absorb light only in the UV range, which means that only about 7 % of the solar spectrum can be absorbed. 2922 www.angewandte.org Figure 14. a) Distribution of energy levels introduced into the bandgap of TiO2 (ca. 3 eV) according to various DOS calculations.[283, 286–294] Solid lines are generally assumed to act as electronically active donor/ acceptor states. Dashed lines represent more unclear contributions (for example, traps or uncertain position of the energy level relative to the conduction band (CB) or valence band (VB)). s = substitutional, i = interstitial. b) XPS spectra of various nitrogen-doping approaches for TiO2. c) Illustration of contradicting findings to claims of a successful solution doping approach for TiO2 nanotubes. a 2 h at 1 V (left panel) and 20 V (right panel); c 20 h at 1 V (left panel) and 20 V (right panel) (reproduced with permission from Ref. [416]). After various early findings,[283, 284] the first report on successful band-structure modification was by Asahi et al. in 2001.[285] The authors used nitrogen doping and interpreted the results in terms of a valence band rise owing to a mixing of the introduced p states of nitrogen with the O 2p states (Figure 14). In the following years, a wide range of other elements were explored, including other nonmetals, such as carbon or sulfur.[286–291] At the same time, a variety of transition metals, such as V, Cr, Mn, and Fe[283, 292–294] were investigated, and among the transition metals, chromium and vanadium in particular were reported to be successful in activating a response to visible light. The most typical methods to prepare doped TiO2 nanostructures are: 1) treating the final or growing TiO2 nanomaterials in a solution or melt of the doping species;  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes 2) thermal treatments or synthesis in gas atmospheres of the doping species; 3) production of the nanomaterials by cosputtering or sputtering in an atmosphere of doping species; 4) high-energy ion implantation; and 5) the use of a substrate of a suitable alloy or the incorporation of active electrolyte species for TiO2 structures that grow from the metals by electrochemical oxidation. This method will be discussed in more detail in the section on growth of oxide tubes on alloys (Section 5). Approach (1) is often successfully used in precipitation reactions, that is, hydrothermal or sol–gel processes while growing the crystallites. A post-synthesis processing seems to be successful only in some specific cases and only if the TiO2 crystallites are in the range of a few nanometers.[295–297] Approach (2) is frequently used for nitrogen or carbon doping using predominantly treatments in NH3, CO, or acetylene.[298–301] Approach (3) is the classic approach used by Asahi for the first successful nitrogen doping.[301] Approach (4) is indeed a very reliable way to properly incorporate nitrogen-containing species into the TiO2 lattice at lower to medium doping levels.[302, 303] The drawbacks are that maximum (reasonable) fluences of high energy accelerators, operating at 50–1000 keV, are limited to about 1018 ions cm2, and an implantation depth into the substrate limited to several micrometers, with a somewhat inhomogeneous dopant distribution. The most successful and most studied approach is currently nitrogen doping. Carbon doping recently became subject to a well-justified dispute,[291] where the effectiveness of the approach was mainly questioned. For nitrogen and carbon, states close to the valence band are typically considered to be responsible for optical gap narrowing (Figure 14 a). There has been discussion regarding the mechanistic nature of nitrogen doping in view of true bandgap engineering. Considering that non-metal doping typically leads to some 2 % of nitrogen in the structure, and although this corresponds to a comparably very high doping concentration for a classical semiconductor, it is problematic to assume that this concentration is sufficient to raise the bulk valence band level by more than 0.5 eV. Therefore, it may be more appropriate to describe the situation of nitrogen-doped material as a high density of localized states. Another challenge in nitrogen doping is that the various methods to synthesize nitrogen-doped material lead to different states of nitrogen in the TiO2 bulk or on the TiO2 surface. This may be best illustrated by considering the XPS N 1s peak after doping treatments (Figure 14 b). Proper ion implantation of N+ and annealing leads to a peak at about 396 eV[302] that is in line with results from sputtering TiO2 in N environments[285] or peak positions obtained for titanium nitrides.[304] Wet treatments in for example amine-based solutions typically lead to peaks above 400 eV; these materials were also found to be active under visible-light illumination.[297] Peak positions above 400 eV can in many cases be interpreted as a surface doping or sensitization, with for example an NC compound. Nitrogen doping reactions that result in an XPS peak of about 400 eV are very ambiguous. This peak position is also found for molecular N2 adsorbed on Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2. Although several groups reported on successful nitrogen doping with this peak position, most results show neither visible photocurrent nor convincing visible photocatalytic activity. One of the main difficulties in obtaining reliable information of effective doping is that many results that have been reported are based on plain absorption or reflectivity measurements. In many cases, although absorption spectra show strong alternation for any sort of treated sample, the corresponding photocurrent spectrum does not show any significant response (absorption that is due to, for example, contamination of the TiO2 substrate with a visible-lightabsorbing material does not mean that electronic coupling with TiO2 occurs). This point may indicate how critical many claims of successful doping that are based purely on absorption measurements rather than an optoelectronic or electrical characterization should be regarded. Moreover, a visible photoresponse does not automatically imply activation for visible-light photocatalysis. Ion implantation has been reported to lead to most effective doping of TiO2 nanotubes[302, 303] with nitrogen, and to some extent with chromium.[305] After ion implantation, amorphization of TiO2 nanotubes occurs, and reannealing is needed to make the nanotubes defect-free. Simpler are thermal treatments in NH3[306] or modification by urea pyrolysis.[307] While treatment in NH3 indeed leads to the signature of TiN formation at 396 eV in XPS spectra, urea treatments seem to lead mainly to some surface sensitization with CN=C and CNH2 groups. However, such surface-modified nanotubes show a significant photoresponse in the visible range compared to non-modified nanotubes. A particularly straightforward approach to dope anodic oxides is the use of an alloy as a substrate. For example, nitrogen doping of TiO2 nanotubes can also be carried out by using TiN containing titanium alloy substrates.[205] Such substrates can be made with pure titanium and TiN powder by arc melting. Other elements were also explored for alloy doping, such as tungsten and niobium.[224, 308] Regarding bandgap engineering, a most interesting element is tungsten, as the introduction of WO3 into the TiO2 structure can be expected to lead to conduction band lowering (Figure 14 a). For tubes, considerable effort went into solution-based doping; however, efforts are questionable, as XPS peaks located at 400 eV (adsorbed species) were mostly obtained for nitrogen; visible absorption can be ascribed to carbon contamination for tubes prepared in organic electrolytes, which originates from decomposition of the organic electrolyte under the applied voltage.[118, 138] Some reports claim an increase in nitrogen doping by extended anodization:[309] such results however could not be convincingly confirmed (Figure 14 c). Highly successful carbon doping and conversion of the tubes can be achieved using acetylene.[212, 299] 4.3. Conversion of Tubes (Titanates, Semimetallic Phases) TiO2 nanotubes can comparably easily be converted into their perovskite oxide. Perovskite materials, such as lead titanate (PbTiO3), barium titanate (BaTiO3), strontium titanate (SrTiO3), and lead zirconium titanate (PbZrTiO3), show  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2923 Reviews P. Schmuki et al. a variety of interesting piezoelectric or ferroelectric properties.[310] There are several chemical approaches to fabricate MTiO3 (M = Sr, Ba, Pb) directly from the TiO2 bulk powder, for example by the sol–gel method or hydrothermal/solvothemal process, or template-assisted processes.[173, 174, 311–313] However, a simple way to convert anodically grown TiO2 nanotube arrays into their titanates is to treat them hydrothermally in presence of corresponding precursor solution. For example, fabrication of BaTiO3, SrTiO3, or mixed BaxSr(1x)TiO3 perovskites by a hydrothermal treatment of TiO2 nanotube arrays has been reported.[314–318] Although comparably successful, there are concerns that the highpressure autoclave environment affects the bonding of titanate tubes to the surface, and therefore other approaches to convert self-organized TiO2 nanotubes into PbTiO3 were explored.[317] For example, lead was electrodeposited into the TiO2 nanotube arrays, followed by an appropriate heat treatment. By this approach, highly ordered piezoelectric PbTiO3 nanocellular layers with uniform structure and defined dimension over large surface areas were achieved. Titanium zirconate tube formation was reported by anodization of an appropriate alloy followed by adequate heat treatment.[124, 172, 189] Moreover, alkaline hydrothermal treatment of TiO2 nanotubes can convert the material into protonated titanates, which are also promising for applications in catalysis, photocatalysis, electrocatalysis, lithium batteries, hydrogen storage, and solar-cell technologies.[15, 65, 319] A highly promising approach of converting the semiconducting TiO2 nanotubes into semimetallic TiOxCy has been reported recently.[212] By a carbothermal reduction treatment in acetylene at temperatures above 800 8C and by using short times to prevent collapse of the tubes, nanotube layers were attained that showed stable metallic conductivity. The conductivity of the TiOxCy tubes is in the range of graphite; optically the material shows a metal-type of photoresponse, and also capacitance measurements of TiOxCy nanotube electrodes shows a semimetallic behavior. The material possess a high overpotential for oxygen evolution and thus is promising for a wide range of electrochemical applications, such as a catalyst support in the field of fuel cells and other applications that require very high electron conductivity. 4.4. Filling and Decoration An approach related to doping is the modification of TiO2 surfaces with nanoparticles (metals, semiconductors, polymers). In all of these cases, essentially three beneficial effects are expected: 1) heterojunction formation that either changes the band bending (metal clusters on semiconductor) or provides suitable energy levels for charge injection, such as in dye-sensitized solar cells (DSSCs); 2) catalytic effects for charge-transfer reactions, such as reaction of O2 in photocatalytic particles; 3) surface plasmon effects, leading for example to field enhancement in the vicinity of metal particles and thus allowing more efficient charge transfer. 2924 www.angewandte.org For TiO2 nanotubes, several approaches for decoration or filling with different foreign materials (metals or metal oxides) have been reported (Figure 15). Electrodeposition reactions into TiO2 nanotubes essentially provide a very versatile way to fill or decorate oxide nanotubes. In the case of Figure 15. Examples of TiO2 nanotube filling with a) electrodeposited copper (reproduced with permission from Ref. [320]), b) decoration with NiO nanoparticles (reproduced with permission from Ref. [323]), c) Fe3O4 nanoparticles,[324] and d) TiO2 nanoparticles (reproduced with permission from Ref. [322]). TiO2 nanotube layers, several factors hamper an easy, direct electrodeposition: TiO2 is an n-type semiconductor, thus a cathodic potential needed for electrodeposition of metals represents forward bias; as a result, the tube walls have such a high conductivity that deposition occurs preferentially on top of the layers rather than within the tubes.[320] This effect is even amplified if the cathodic potential is negative to the Ti4+!Ti3+ potential, as the formed Ti3+ species acts as additional dopant. Therefore, careful consideration of the experimental conditions is essential to achieve successful metal filling. A successful strategy for complete filling is to first carry out reductive self-doping of only the nanotube bottoms by selective Ti3+ formation, which then allows electrodeposition to start at the tube bottom and then fill the tubes from bottom to top.[320] This approach was used to electrodeposit Cu in the tubes, and this kind of copper-filled nanotubes may be used as one step for establishing the p–n heterojunctions (Cu2O-TiO2) in solid-state solar energy devices. Oxide nanoparticle decoration of for example WO3[321] or TiO2[322] can be obtained by slow hydrolysis of precursors, such as TiCl4 or WCl5. In DSSCs, the TiO2 nanotubes decorated with TiO2 nanoparticles show higher solar cell efficiency in comparison to neat TiO2 nanotubes. The TiO2 nanoparticles 2–3 nm in size can be deposited inside as well as outside of the tube wall by hydrolysis of a TiCl4 solution,  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes which significantly increases the surface area and thereby lated electrons from a HOMO–LUMO transition in the improves the solar-cell efficiency.[322] While the beneficial adsorbed molecule can be injected to the conduction band of effect is surface area increase in the case of TiO2 nanoparticle the TiO2, as will be discussed in more detail in Section 6.2. decoration, with WO3 nanoparticles, junction formation between TiO2 and the misaligned bands of WO3 can also be expected. 4.5. Monolayers An approach to TiO2 nanotube filling with nickel oxide nanoparticles by using a sufficiently slow precipitation TiO2 and TiO2 nanotube surfaces can be modified reaction of Ni(OH)2 followed by a suitable thermal treatment relatively easily by covalent attachment of organic monohas recently been presented.[323] Such nanotubes show siglayers.[138, 177, 333–336] The attachment typically occurs at hydroxnificant photoelectrochemical activity under visible light, ide-terminated surface sites by a condensation reaction with possibly by charge injection from NiO states to the conphosphonates, silanes, or carboxylates (splitting off small duction band of TiO2. molecules, such as H2O or HCl) and leading to a covalent A very simple but highly successful approach involves bond between a substrate oxygen and the anchor group filling TiO2 nanotubes with a suspension of magnetic Fe3O4 (Figure 16 a). Organic monolayers are attached onto TiO2 nanoparticles.[324] By placing a permanent magnet under a surfaces mainly because of one of the following intentions: tube layer, a ferrofluid swiftly enters the tubes and the TiO2 1) to change the surface wettability, 2) to modify the biocomnanotubes can be very homogeneously filled with magnetic patibility (drug delivery, bioactivation), 3) to obtain chemical nanoparticles. Such magnetic tubes can be used in guiding and or biochemical sensors, and 4) to attach an electron injection release applications.[324] system (DSSCs). In general, the quality of the monolayer (packing density, Decoration of TiO2 nanotubes by noble metal nanoattachment strength) is in the order phosphonate > silane > particles (such as Au, Ag, and Pt) also can be carried out to carboxylate. It should, however, be noted that if chargeenhance their photocatalytic activity.[275, 325, 326] Silver nanotransfer reactions across the attached molecule are important, particles can be deposited on the tube wall by photocatalytisuch as in charge-injection sensors[337] or DSSCs (see Seccally reducing Ag+ on a TiO2 surface by UV illumination.[325] Other metal nanoparticles are preferably deposited by UHV evaporation or chemical reduction techniques.[274, 325] Ag/TiO2 nanotubes show a significantly higher photocatalytic activity compared with Au/ TiO2 nanotubes.[325] Silver-decorated tubes significantly enhanced the performance of DSSCs.[327] TiO2 nanotubes filled with zeolites were also very promising and show an interesting photocatalytic activity.[328] TiO2 nanotubes can also be decorated by narrow-bandgap semiconductors, such as CdS, CdSe, PbS quantum dots.[329–332] These quantum dots can be deposited on the nanotube wall electrochemically, by sequential chemical bath deposition methods, or by chemical treatment in presence of cadmium precursors. Such CdS/CdSe quantuam dots with a bandgap of 2–2.4 eV can absorb light in the visible range and inject the excited electron quickly into TiO2, thus performing as a photoelectrochemical solar cell. To date, such photoelectrochemical solar cell Figure 16. a) Examples of monolayer attachment on TiO2 (from left to right): surface condensation reaction; alkane phosphonate attachment (creating hydrophobic surface); APTES attachment shows efficiencies of about 4 %.[330] Another very useful principle is (bifunctional linker for other molecules, (e.g., proteins etc.); APTES coupled to ferrocene (electrochemically switchable group); and carboxylate coupling (DSSCs). b) The effect of different to attach organic dyes (or any other monolayers on wetting of TiO2 nanotube surfaces and c) the effect of the tube diameter on ODPAmolecule that has suitable HOMO– coated NTs d . & as reported in Ref. [343], * as reported in Ref. [417]. d) SEM image of microNT LUMO orbitals) to TiO2 surfaces. scopic wetting of TiO2 nanotube surfaces, showing that wetting takes place preferentially between The principle is that optically stimu- tubes (reproduced with permission from Ref. [181]). Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2925 Reviews P. Schmuki et al. tion 6.2), the electronic properties of the anchoring groups are crucial. For example, in DSSCs charge transfer from a dye molecule to the TiO2 conduction band is significantly faster for COO groups than for silanes. As Ti/TiO2 is the most used biomedical implant material worldwide, it is highly interesting to modify the surface with additional biorelevant molecules. To link active organic entities (mainly enzymes, proteins, or DNA) to TiO2 surfaces, bifunctional molecules such as APTES[138, 144, 177, 281] that carry for example a terminal NH2 group are attached. These or similar linkage methods enable further covalent attachment of virtually any organic species.[338–341] Of interest in this context is the fact that some linker molecules, such as APTES with a silane and an opposite NH2 linker group, show a different reactivity on amorphous, anatase, and rutile polycrystalline surfaces;[281] even the degree of reverse attachment (coordination of the NH2 group) varies with TiO2 crystal structure. A specific feature of almost all monolayers attached on TiO2 is that they can be cut by photocatalytic reactions[177, 335, 342, 343] or by voltageinduced reactions.[138] This principle was used for various drug and other payload release processes from dispersed TiO2 nanotube bundles or from nanotube electrode surfaces.[144, 177, 324, 344] The organic modification of nanotubes combined with a photocatalytic reaction was used to create surfaces that could be adjusted to have essentially any desired wettability property[335, 343] (that is, a water contact angle from 08 to 1708).[335, 343] Nanotube layers as such (amorphous or crystalline) are super-hydrophilic; only when treated with a suitable monolayer do they become superhydrophobic.[335, 343] Upon UV irradiation, chain scission occurs, which makes the surface increasingly hydrophilic with the duration of illumination. In this approach, the achievable superhydrophobicity depends on the tube diameter[343] (Figure 16 c), and overall wettability behavior is in accord with the Cassie–Baxter model.[345] To fill super-hydrophobic tubes with a liquid (with for example an electrolyte), organic solvents are needed, which is the key to filling hydrophobic tubes or to fabricate amphiphilic tubes (see Section 6.5).[177] Of interest in this context is, however, a very recent observation that on the microscopic level, all TiO2 nanotube layers (non-modified and modified) show preferential wetting between the tubes rather than inside the tube.[181] This observation is in line with those for dry anatase tubes: the inside of the tubes is not easily filled by aqueous electrolytes.[346] Another elegant way to adjust the wettability of nanotube layers is by applying mixed monolayers with a different degree of polarity or even actively switchable polarity. Such mixed monolayers of N-(3-triethoxysilyl)propylferrocenecarboxamide and perfluorotriethoxysilane were used to demonstrate electrical redox switching of attached ferrocene molecules and thus to induce alterations of the wettability on TiO2 nanotube layers accordingly.[347] 5. Oxide Nanotube Layers on Other Transition Metals and Alloys The principle used to grow oxide nanotubes on titanium by using a dilute fluoride electrolyte can be transferred to a considerable range of other metals and alloys (Figure 17). Depending on the exact electrochemical conditions, organized nanotubular or nanoporous structures were reported for several other valve metals, such as Zr,[164, 348–352] Hf,[163, 166] W,[74, 353–356] Nb,[357, 358] Ta[359–363] Fe,[364–366] and Mg.[367] To obtain highly organized high-aspect-ratio structures for each case, some optimization of the electrochemical conditions specific to the element or alloy is needed, and this has not yet been achieved for all of the elements, which is also apparent from the SEM top and cross-sectional images shown in Figure 17. The optimized conditions are different for each case and the ease by which optimal self-organizing conditions Figure 17. Cross-sectional and top-view SEM images of ordered oxide nanotube or nanopore layers electrochemically grown on different valve metals and metal alloys. 2926 www.angewandte.org  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes and high aspect ratios can be established varies significantly for the different elements. For zirconium and hafnium, highaspect-ratio smooth tubes can be achieved very easily under a wide range of parameters. Other extremes may be non-valve metals, such as iron, where high-aspect-ratio oxide structures were reported only very recently,[364–366] or WO3 structures, for which still no highly ordered and only comparably short tube structures were reported.[353, 355, 368–370] The main reasons for this different behavior can be ascribed to following factors: 1) the solubility of a formed oxide structure in the anodizing electrolyte affects the aspect ratio, 2) the solubility of the fluoride species (or any other sensitizer) at cell boundaries affects tube or pore morphology and the observation of sidewall ripples, 3) different Pilling–Bedworth ratios that affect stress that is generated when the oxide is formed, thus the adhesion of the tube layer to the substrate or the tube length (overshoot by plastic flow) and the self-organization length scale are also affected. Many of the nanostructures that are formed have very interesting properties. For example, WO3 nanotubes/pores show excellent ion intercalation properties (electrochromic devices, charge storage)[66, 69, 70, 224, 321, 355, 371–373] or enhanced photoelectrochemical properties.[374, 375] For ZrO2 nanotubes, bright visible luminescence[376] has been reported. An even wider field, in view of varying properties, is provided by using alloyed metal substrates for anodization, for example to tune ionic, electronic, or optical properties of the tube layers. However, to obtain ideal tube layers and defect-free homogeneous ordering of tubes over a large surface area, the alloys should be single-phase (to avoid different etching/anodization rates on different phases), and growth is achieved more easily if the elements involved are valve metals. Accordingly, self-organized oxide nanotube/pore layers of a reasonable quality have been reported on binary alloys, such as TiNb,[188, 308] TiZr,[124, 126, 172, 189] TiTa,[190, 361, 377] TiW,[224] TiMo,[378] TiAl,[158, 159] and on ternary[191, 192] and more complex alloy systems.[379, 186] Alloy anodization thus enables the growth of mixed anodic oxides with tailored and improved properties for a wide range of applications.[190, 192, 224, 308, 378, 379] For example, nanotubular layers grown on TiNb and TiW show not only the feature that geometry can be adjusted over a wider range,[224, 308, 380] but also show highly enhanced intercalation properties.[308] It is also interesting that in the case of titanium alloys, small amounts of the alloying element can drastically affect the properties while the unique nanotubular morphology is completely retained. For example, TiW (0.2 atom %) and TiMo (7 wt %) alloys show a strongly enhanced electrochromic response and improved photocatalytic properties.[224, 371, 378, 381] On all of the pure metals and alloys that have been explored to date, the nanotubular oxide layers are amorphous, and in each case they can be easily transformed to a crystalline structure by an adequate annealing treatment.[189, 224, 308, 368, 376, 378] After alloy anodization, the composition of the oxide in general is consistent with the ratio in the alloy. In the case of anodic oxide layers on TiAl alloys, the film is composed of (TiO2)n and (Al2O3)m, where m and n are the respective fractions of the base alloy TinAlm.[382] In some cases, minor amounts of mixed oxides may also be present in Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 the anodic oxides. Full conversion to a mixed oxide was only reported for TiZr.[189] It is also interesting to note that for most elements and alloys, ordered tubular and nanoporous morphologies have been observed, and transitions from one to another can be achieved (for example, TiAl, TiNb, and TiTa).[159, 308, 361, 383] However, owing to the difference in chemistry (solubility of the oxide in fluoride and solubility of the respective metal fluorides in different solutions), most elements appear to have a preferred morphology, that is, a morphology that dominates under most anodization conditions. For example, on zirconium and hafnium, nanotubular anodic oxides are usually observed, whereas on aluminum, niobium, and tantalum, nanoporous morphologies are usually obtained. For tantalum, extreme conditions are required to obtain a tubular morphology.[363] In the case of alloys, the morphology follows the specific oxide morphology of the major alloying element. A highly interesting phenomenon that was observed for several alloys but is still is not well understood is selforganization at two length scales,[37, 186, 187, 190, 191] by which two distinct tube diameters are formed during anodization (a large center tube surrounded by smaller tubes, repeated over the entire anodized area). The simultaneous ordered formation and stabilization of two tube diameters is currently ascribed to availability of current at the different tips,[37] but the phenomenon seems to be far from satisfactorily explained. In general, it can be said that all investigated organized oxide structures grown by anodization in fluoride-containing electrolytes on different metals or alloys seem to follow the same growth principles and key factors: The diameter of the tube is determined by the anodization voltage, etching of the tubes (and thus the achievable length of the tubes) depends on the chemical resistance of the oxide against fluoride etching (in a particular electrolyte), and water plays the key role for providing the oxygen source for tube growth, splits pores into tubes, and is responsible for sidewall ripple formation. Obviously, anodization techniques in fluoride-containing electrolytes allow the fabrication of nanostructured oxide layers on an extremely wide range of alloys that enable the tailored fabrication of mixed nanostructured oxides with virtually endless possibilities to create enhanced properties, and therefore have also a very high and widely unexplored technological potential. 6. Applications Applications of TiO2 nanostructures usually involve the exploitation of some unique feature of TiO2 (electronic, ionic, or biocompatibility properties) and a significant enhancement of some reaction or transport rates that is obtained by using small scale dimensions (large surface area, short diffusion path, or size confinement effects). Using nanotubular assemblies also provides a preferred dimensionality to the system. Tubes grown on a metal substrate are vertically aligned to a back contact; that is, a direction for charge transfer towards the electrode is established. Nanotube layers provide well-  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2927 Reviews P. Schmuki et al. defined top openings and are thus suitable for size-selective applications (filters) or for templating secondary material. The inside volume is well-defined and regular, thus applications, such as nano test tubes[384] are promising. In the following sections, we give an overview of current efforts towards TiO2 nanotube applications. 6.1. Photocatalysis and Dark Photocatalysis TiO2 is the most photocatalytically active material for the decomposition of organic materials (for example, it is used for degradation of organic pollutants).[28, 38, 197, 198, 385–387] The reason for this high activity are the band-edge positions relative to typical environments (such as water). The basic principles involved in the photocatalytic mechanism are shown in Figure 18. UV light promotes electrons from the valence band to the conduction band; holes and electrons will be separated (under the field of the Schottky junction with the environment), reach the semiconductor–environment interface, and react with appropriate redox species in the environment. Several highly reactive species are generated for example from the surrounding water by charge exchange at the valence band (H2O + h+!OHC) and at the conduction band (O2 + e !O2). These radicals and peroxo ions are able to virtually oxidize all organic materials to CO2 and H2O. Furthermore, at the valence band, direct h+ transfer to adsorbed species to initiate decomposition may also be considered. Alternatively, apart from decomposition of organics, h+ and e can react with H2O to form H2 and O2 ; that is, direct splitting of water can be achieved. At the conduction band, the situation is such that the redox potentials for O2 !O2 and H+!1/2H2 are very close. In other words, H2 generation and O2 formation are typically competing. At the valence band, O2 can be formed from water by various pathways, including radicals that can (if not otherwise used) react finally to O2. The reaction rates of the photocatalytic processes on pure TiO2 are typically limited by the charge-transfer process to a suitable redox species. Therefore, at the valence band, catalysts such as platinum are used to promote for example H2 evolution, and hole-capture agents such as CH3OH are often used to promote the overall reaction rate. However, as discussed in previous sections, although TiO2 has very suitable band-edge positions for high photocatalytic activity, the bandgap of about 3 eV allows only UV light to be efficiently used. To achieve visible-light-driven processes, a large number of bandgap-engineering (doping) approaches, as discussed in Section 4.2, have been explored with TiO2. The method that has been most investigated for photocatalysis is N-doping, as it is perceived to be the most efficient measure to stimulate a visible photocurrent response. However, the success of N-doping to achieve visible photocatalytic activity is somewhat questionable if the photocatalytic reaction is indeed valence-band-dominated (that is, h+ transfer to the electrolyte is the dominating reaction). As N-doping raises the valence band edge (Figure 14), the h+ leaves the semiconductor at a lower energy (less anodic redox potential), which is possibly not sufficient to decompose H2O to OHC or 2928 www.angewandte.org Figure 18. Photocatalytic activity of TiO2 nanotubes: a) the mechanism (photoinduced formation of electron–hole pairs and reaction with surroundings). b) Dark photocatalysis: sufficient voltage-induced band-bending creates valence band holes (h+) that can react with the environment, in analogy to photon-induced hole (h+) generation.[138] c) Photocatalytic activity of various nanotubes measured as decomposition rates of organic compounds (AO7; C0 = 2.5  105 m) and including a comparison to the dark photocatalysis mechanism. to achieve direct decomposition of the organic material. The fact that in some cases a visible photocatalytic effect (mostly very mild) was found for N-doped material could then be attributed to O2 generation by the conduction band. In other words, in many cases of photocatalytic degradation of a specific organic molecule, the active reaction path depends on the molecules energy states that may prefer to interact (overlap) with either valence-band or conduction-band levels.[198] Nevertheless, all of the photocatalytic applications have in common the fact that a higher overall reaction rate is achieved using high-surface-area geometries. Thus, reactions are commonly performed using suspended nanoparticles or nanoparticulate electrodes. Ordered nanotube arrangements offer various advantages over nanoparticulate assemblies, as  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes their defined geometry provides very defined retention times in nanoscopic photoreactors.[123, 388] Moreover, the 1D geometry may allow a fast carrier transport and thus less unwanted recombination losses. A first investigation to this end[388] has shown that TiO2 nanotubes can indeed have a higher photocatalytic reactivity than a comparable nanoparticulate layer. Although various factors may be responsible for this effect (optimized reaction geometry for charge transfer, UV absorption characteristics over the tube, solution diffusion effects), the work triggered intense follow-up investigations. It was shown that particle decoration with silver or gold led to a significantly increased photocatalytic activity,[325] and that also applying an external anodic voltage drastically enhances the photocatalytic activity.[138, 389] These findings suggest that in the investigated cases, a valence-band mechanism dominates, and the observed accelerating effects have a common origin in increased band bending, either by junction formation or by the applied voltage.[38] To increase the photocatalytic activity, various mixedoxide tube layers were used, such as TiMo or TiW.[381, 390] A compilation of the photocatalytic results is shown in Figure 18 c. Clearly, a highly beneficial effect is observed for tungsten and molybdenum that cannot be explained by a better charge transport in the tubes but must be ascribed to modification of the surface state distribution at the nanotubes.[38, 381, 390] Visible photocatalysis was shown for TiW oxide tubes[390] and to some degree for carbon-doped TiO2 nanotubes regarding water splitting,[298, 21, 391, 392] although there is considerable dispute on its effectiveness,[291] and for TaON nanotubes.[393] A very spectacular possibility to use TiO2 nanotubes in photocatalytic applications is to produce free-standing flowthrough membranes,[123, 193] as they allow extremely defined photocatalytic interactions (highly defined interaction times) combined with a filtration capability. Another intriguing specific feature of TiO2 photocatalysis is the ability to induce chain scission in attached organic monolayers (see Section 4.5). This feature can be used to create extremely well-defined wettability on surfaces[335, 343] or to liberate terminal payload molecules from the surface upon UV illumination. This was used for example to construct various drug-release systems.[138, 177, 333–336] Of interest and common to all approaches is the fact that chain scission (photocatalytic or voltage-induced) occurs after the anchoring group (a silane or phosphonate). The reaction may thus be based on a direct h+-induced chain scission as the diffusion range (lifetime) of OHC radicals is typically comparably long (0.1–10 mm).[22] This hypothesis is further supported by the fact that a release of intact activemolecules, such as dyes, from TiO2 nanotubes can be observed after chain scission; [144, 177, 324] if OHC formation was the main mechanism, the dye would be decomposed before it is able to leave the tubes. A particularly interesting concept in the field of photocatalysis is the combination of the TiO2 nanotube geometry with molecule-selective binding units, for example zeolites.[328, 394] Considering that zeolite structures can be tailored for selective absorption of a considerable range of organic molecules, they can for example be used to concentrate dilute Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 pollutants. If the zeolites are filled in TiO2 nanotube layers or membranes, the pollutant molecules in the zeolite can be accumulated and then be photocatalytically decomposed. Thus the zeolite is regenerated and a significantly enhanced photocatalytic activity be achieved. It is crucial that the lifetime of a photocatalytically generated OHC radical is on the order of 10 ms, which corresponds to a penetration range of several hundred nanometers to several micrometers from the generated surface into the electrolyte; this length scale fits extremely well to inner dimensions of TiO2 nanotubes. In view of biomedical uses (drug delivery, fighting cancer), X-ray-induced photocatalysis was used to liberate drugs from the surfaces of TiO2 nanotubes[344] or to directly kill cancer cells.[395] X-ray catalysis is crucial for any in vivo application as X-ray-induced reactions would allow intervention-free therapy (stimulation can take place directly through the human body; the energy of the X-rays then determines the penetration depth of the activating irradiation). Furthermore, it was recently demonstrated that electron-beam-induced photocatalytic reactions can occur in the vacuum of a scanning electron microscope using ionic liquids as a non-volatile solvent.[181, 396] Regarding photocatalytic reactions, it should also be mentioned that a similar reaction (Figure 18 b) can be triggered in the absence of light on anatase TiO2 and TiO2 nanotube surfaces if sufficiently doped.[138] If a voltage is applied to the material that causes anodic Schottky barrier breakdown,[397, 398] that is, valence-band holes are created that react with environment in a similar manner as photogenerated holes (OHC radical formation, destruction of organic and organic monolayers). TiO2 nanotubes, when annealed, have a suitable doping concentration (ca. 1018–1019 cm3) and by applying voltage (> 3 V), ionization of the valence band occurs and holes are generated that leave the TiO2 to the environment (at E = Evs). This dark photocatalysis approach may be particularly useful in environments where the use of UV light is hampered, for example in MEMS (microelectromechanical systems) devices or a lab-on-a-chip that require a photocatalytic reaction or a self-cleaning step in the dark. 6.2. Solar Cells Another very attractive application of TiO2 is its use in solar cells. Dye sensitization of TiO2 was extensively investigated in about 1970. After the work of Gerischer and Tributsch in 1968,[23] the first report on ruthenium bypyridyl sensitization on TiO2 appeared in 1980,[24] and in 1991, Grtzel and ORegan[25] used this principle to fabricate a full DSSC in which the photon-absorber layer was made of TiO2 nanoparticulates that were dye coated (Figure 19 a). This assembly reached an efficiency of about 11 %. The principle of DSSC[35, 399] involves a dye that absorbs light in the visible range and thereby excites electrons from the HOMO to the LUMO level, followed by a rapid injection of the excited electron into the conduction band of TiO2. The electrons travel through the TiO2 layer to the back contact; meanwhile, the oxidized dye on the surface is regenerated by an I/I3 electrolyte. Losses by recombination may happen principally  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2929 Reviews P. Schmuki et al. Figure 19. a) A TiO2 nanoparticle/tube-based dye-sensitized solar cell. b) Comparison of results for solar-cell performances of TiO2 nanotube layers used in DSSCs (non-modified by TiCl4).[322] d) Estimation of electron-diffusion length (Ln) TiO2 nanotubes from the experimental values of Dn and tc by taking the quasi Fermi level (QFL) into account (reproduced with permission from Ref. [403]). in three ways: 1) After electron excitation from the HOMO to the LUMO level of the dye, de-excitation of the electron occurs in a radiative or non-radiative path; 2) after injection of the electron in the conduction band of TiO2, it may recombine with the oxidized dye; and 3) the electron may recombine with the I3 ion in the electrolyte. In general, the rate-limiting factor is the dynamic competition between the electron transport through the TiO2 and the interfacial recombination of electrons.[400] An oftenaddressed critical issue for losses in nanoparticulate DSSCs is carrier recombination at grain boundaries (owing to the presence of trapping states) and long carrier diffusion paths (random walk) through the TiO2 network. Therefore, replacing the TiO2 nanoparticulate photoanode with a TiO2 nanotubular layer in particular reduces recombination probabilities and provides a directed (ideally one-dimensional) electron traveling path.[35] Since the first efforts to dyesensitize TiO2 nanotube arrays,[401] where efficiencies of only 2930 www.angewandte.org 0.036 % were reached, approximately 5 % solar cell conversion efficiency has meanwhile been achieved for tubularbased systems.[402] However, it must be clearly distinguished between cases where pure TiO2 nanotubes are used[119, 179, 280, 403] and cases where the nanotubes were additionally treated with TiCl4 (decoration with TiO2 nanoparticles) or mixed with TiO2 nanoparticles.[322, 404, 405] Of course, the observed effects (efficiency, dye loading, transport times, reaction kinetics) can only be unambiguously ascribed to the nanotubes in cases where plain TiO2 nanotubes were used; in mixed cases, effects may even be dominated by the added TiO2 nanoparticles. For pure TiO2 nanotube layers, the record efficiency stands at 5.2 %,[402] for mixtures with nanoparticles at 7 %.[404] Some critical factors were recently discussed and reviewed.[35, 280] A very interesting study has shown that a much higher electron diffusion length can be achieved in nanotubes than in nanoparticles;[403] in fact, from combined electron lifetime/ diffusion measurements, it was deduced that electron diffusion lengths of 100 mm (Figure 19 c) should be achievable. However experimentally, 15–20 mm long nanotubes show the maximum solar cell efficiency (see also Figure 19 b), which is mainly ascribed to a loss of surface adhesion for longer tube layers or disordered tube tops that increasingly appear with longer anodization times (see Section 2.7).[402] Nanotube layers annealed between 350 and 450 8C in air form anatase tubes, which proved to be more efficient than the rutile phase, as the electron transport is faster in anatase TiO2 than in rutile.[227, 280] A temperature of 450 8C and complete anatase conversion is reported to be the optimum annealing condition to achieve high-efficiency solar cells.[35, 280] As discussed in several works,[179, 280, 322, 403] the overall efficiency-limiting factor in TiO2 nanotube-based solar cells is specific dye loading; that is, the specific surface area of the tubes (BET  30 m2 g1) is considerably smaller than comparable nanoparticulate layers (BET  80–100 m2g1). Several strategies to increase the specific surface area in tube systems have been explored, such as the above-mentioned TiCl4 treatment[322, 404–406] or modifying the tube walls by creating bamboo-type structures or double-walled nanotubes.[179, 280] The geometry of the tube tops is very crucial (see for example Ref. [402]), and several approaches have been reported for removing inhomogeneity (surface nanograss) from the tube tops.[119, 184, 402] To further increase the efficiency, Tsuchiya et al.[327] introduced a new concept by showing that using silver-decorated TiO2 nanotubes, an enhancement of the conversion efficiency for DSSCs can be obtained. This effect may either be due to plasmon enhancement or junction formation. Some variations in TiO2 nanotubes, such as that formed by rapid-breakdown anodization (RBA), show promising improvement regarding the photoconversion efficiency.[75, 79] Recently, an anodic self-organized TiO2 mesosponge/nanochannel layer was reported that has a significantly higher specific surface area than tubular layers, and it seems to be capable of outperforming nanotube layers in the field of TiO2-based solar cells and other applications.[80–82]  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes 6.3. Electrochromic Devices Various transition-metal oxides, such as MnO2, WO3, Nb2O5, MoO3, and TiO2, provide an excellent host lattice for ion intercalation devices, for example, lithium ion batteries or electrochromic devices.[29, 65, 407, 408] These devices rely on a reversible uptake of small ions such as H+ and Li+ into interstitial positions of the metal oxide upon applying and releasing an electric field. In the case of transition metal oxides, ion uptake and release are frequently combined with a change in the redox state of the material and a resulting change in the electronic and optical properties of the material. For example, lithium ion intercalation into TiO2, accompanied with reduction of Ti4+ at the lattice to Ti3+, changes the apparent bandgap of the material from the UV to the visible range (with Eg  2.2 eV; see also Figure 15), which leads to a blue coloration of the material [Eq. (8)].[273, 409] Mixed TiW, TiMo, and TiNb oxide nanotubes turned out to be very efficient (see Figure 20 b). Of interest is the fact that for most nanotube systems, an optimized condition of the secondary element exists by which an optimum effect can be achieved. TiNb oxide nanotubes are particularly spectacular in scientific terms,[188, 308] for which it could be shown that the addition of niobium to TiO2 leads to a widening of the anatase lattice.[308] This not only drastically accelerates ion intercalation, lowers the threshold voltage, and makes the material much more durable in repeated switching experiments; it also allows for the reversible intercalation of much larger sodium ions, which cannot usually be intercalated into anatase under low-voltage conditions and at room temperature. Other work has shown that TiO2 nanotube layers can be lifted off from the metallic titanium substrate and be transferred onto conducting glass[411] or by complete anodization of titanium layers on conducting glass[412] to construct transparent effective electrochromic devices (Figure 20 a). 6.4. Cell Interaction and Biomedical Coatings A very important application of titanium surface modification is in biomedical applications. About 40 % of todays The currently most active electrochromic material is WO3. As a nanoparticle layer, it shows the lowest threshold voltage for intercalation and color switching, the highest color contrast, and is cyclable over 1000 times in commercial devices with a comparably small deterioration of the effect. In aquous environments and for pH values of 12–13, H+ is the dominating intercalation species, even in presence of high Li+ concentrations owing to the much lower radius of the solvated ion. The rate-limiting step for ion intercalation is a solid-state diffusion/migration process of the small ion into the host lattice and therefore the switching kinetics is typically comparably slow (seconds). To achieve short diffusion paths, nanoscale materials are highly desired, and most commercial devices consist of layers of compacted nanoparticles. At reasonable applied voltages (1– 2 V) and moderate times (seconds), a compact layer can typically be intercalated to a depth of approximately 5–10 nm. This length scale is ideal for penetration of TiO2 nanotube walls (with wall thicknesses in the range of 5–30 nm). Moreover, the vertical alignment makes the nanotube geometry ideal for maximizing the optical contrast (Figure 20 a). After the first report on electrochromic switching using TiO2 nanotubes,[410] a large body of work followed that optimized contrast, switching time, threshold voltage, and cyclability by using other oxide nanotubes/porous oxides, such as WO3, or by using doped (mixed-oxide) TiO2 nanotubes.[224, 308, 321, 355, 371, 378] Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 Figure 20. a) Electrochromic device based on TiO2 nanotubes fabricated by complete anodization of sputter-deposited thin titanium film, making it transparent, on conducting glass (reproduced with permission from Ref. [405]). b) Comparison of color-switching transients for Ti, TiNb, and TiW oxide nanotubes for H+ and Li+ intercalation and voltage dependence of contrast DR and threshold voltage UT for the different nanotube layers. Open symbols 0.1 m H2SO4, filled symbols 1 m LiClO4 in PC (propylene carbonate).  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2931 Reviews P. Schmuki et al. biomedical implant materials are based on titanium or titanium alloys.[39] TiO2 nanotubular surfaces are ideal for studying and applying size effects with living matter or biologically relevant species. Not only can the diameter of the tube surfaces be accurately adjusted to virtually any value between 10–250 nm,[132] but owing to the self-organizing nature, entire, even complex shaped surfaces (such as dental-implant screws or hip implants) can be coated easily with such nanotube layers.[413] First work on size-dependent cell interactions in 2007[122] showed that mesenchymal stem cells react in a very pronounced way to the diameter of nanotubes. Diameters of about 15 nm strongly promote cell adhesion, proliferation, and differentiation, and tube diameters of about 100 nm were found to be detrimental, as they induced programmed cell death (apoptosis). The work resulted in further studies[414] with partially conflicting outcomes, and a range of questions were brought forward, such as the role of TiO2 crystallinity, the remaining fluoride concentration, or the role of the cell type and surface pretreatment. However, subsequent work[415–419] showed clearly that the size effect, that is, a cell-stimulating influence for tube layers with a diameter of 15 nm, is of a virtually universal nature (Figure 21). Not only mesenchymal stem cells but also hematopoietic stem cells, endothelial cells, and also osteoblasts and osteoclasts show this size-selective response. In fact, the effect of size dominates over tube crystal structure (amorphous/anatase/rutile), fluoride content, and to a large extent over the wetting properties.[415–417, 419] Even tubes made from other valve metals, such as ZrO2, showed similar size effects.[415] In general, the most straightforward explanation for this effect is that integrin clustering in the cell membrane leads to a focal adhesion complex with a size of about 10 nm in diameter, and this in turn leads to an excellent geometrical fit onto/into the tube openings of about 15 nm.[122] In view of a rapid ingrowth of biomedical implants in bone, a key factor is fast kinetics of hydroxyapatite (HAp) formation on implant surfaces from body fluid. A number of studies[420–423] have shown that HAp formation can strongly be accelerated on TiO2 nanotube surfaces compared with flat TiO2 surfaces, and also in this case a strong size effect can be observed.[420] Moreover, the 3D structure is optimal for embedding precursors for HAp formation that additionally promote HAp nucleation.[421] In vivo experiments with adult domestic pigs showed that nanotube surfaces can indeed enhance collagen type I and BMP-2[413] expression and that higher implant bone contacts can be established if implants are coated with a nanotube layer.[413] However, the negative (apoptotic) effect of nanotubes may also be exploited for biomedical surfaces when cell proliferation is not desired.[39] 6.5. Drug Delivery and the Release of Other Payloads The geometry of the TiO2 nanotube arrays suggests that the material may be used as a drug-delivery capsule if nanotube layers are separated (singled-out) and stabilized, or it may be used as a drug-eluding coating on biomedical 2932 www.angewandte.org Figure 21. a, b) Influence of TiO2 nanotube diameter on activity of cells: a) Comparison of cell activity for mesenchymal stem cells (pink), primary human osteoblasts (blue), osteoclast (green), and endothelial cells (gray) for different TiO2 nanotube diameters; relative cell counts are shown 3 days after seeding. b) Fluorescence micrograph for GFPlabeled mesenchymal stem cells on TiO2 nanotubular surfaces with 15 nm and 100 nm diameters, respectively (showing drastically higher population on 15 nm tube diameters). c) TiO2 nanotube for guided drug release: Representation of magnetically loaded TiO2 nanotubes with attached drug (F). Release is triggered by photocatalytic chain scission upon UV irradiation. Inset: an example where a blue fluorescent molecule is released from magnetically actuated nanotubes (reproduced with permission from Ref. [324]). implant materials. A potential in vivo capsule is shown in Figure 21 c. It takes advantage of the fact that long molecules attached to a TiO2 surface can be released photocatalytically. Shrestha et al.[324] showed that TiO2 nanotubes can be filled with magnetic Fe3O4 particles and thus be magnetically guided to desired locations. Such tubes can then be easily coated with drugs that are attached by suitable linker molecules. Drug release is not limited to UV reactions but can also be triggered electrically (voltage induced catalysis)[138] or more importantly by X-rays,[344] which allows in vivo treatments through living tissue. Such magnetic TiO2 nanotubes can be used directly for photocatalytic reactions with cells or tissue, for example for the site-selective killing of cancer cells.[324] An even more elaborate payload filling and release mechanism was introduced by Song et al.,[144, 177] who created amphiphilic tube layers, which involves tubes that provide a hydrophobic cap (monolayer) that does not allow water (body fluid) to enter into the tubes unless opened by a photocatalytic interaction. Once the hydrophobic layer was removed, body fluids could enter and wash out hydrophilic drugs loaded within the tubes.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes Conflicting reports exist regarding drug-eluding coatings. Some reports show elution times over weeks for loading with paclitaxel, sirolimus, and BSA that are seemingly based on sampling errors.[424] Other reports could not be confirmed; in fact, when reproducing experiments of Peng et al.[424] with identical tubes and drugs, elusion time constants of minutes[425–427] rather than days or weeks were found. Most recent work demonstrates that to achieve slow release, capping of drug-loaded tubular or porous systems with biopolymer, such as poly(lactic acid), is required. Modified in this way, tubular or even better mesoporous anodic layers on titanium-based implants or stents may represent efficient and promising drug-release systems.[426] For applications on stents, a considerable drawback is the somewhat limited mechanical flexibility of the nanotube layers compared to other mesoporous oxide layers.[80, 81, 426] A way to increase the mechanical properties of tubular systems is their modification to more stable titanium oxycarbide nanotube layers.[212, 213] Payload release systems are not limited to in vivo applications, but may prove extremely valuable for MEMS or specifically lab-on-a-chip applications, where release of reactants may be triggered by optical means (photocatalytic chain scission) or electrical means (voltage-induced). 6.6. Other Applications and Aspects In terms of gas sensing, TiO2 layers have shown to have a high sensitivity to CO, H2, and NOx gases, and in particular as nanoparticulated films.[428–430] For TiO2 nanotube layers, this sensitivity has also been demonstrated.[283, 431] In terms of insolution sensing, it has also been shown that for gold nanoparticles supported on TiO2 nanotube layers, a strongly enhanced reaction rate with O2 in aqueous solution can be found,[274] which is of great interest for oxygen sensors. Such considerable support effects were also observed for methanol oxidation reactions, not only for supporting Pt/Pd on TiO2 nanotubes,[275] but even more for semimetallic tubes that showed an efficiency enhancement of 700 %.[212] These findings are very promising for applications in methanol fuel cells (that is, as a substitute for carbon-based supports). An approach that ideally takes advantage of the nanotubular geometry is as so-called nano test tubes.[384] For example, color reactions can be carried out in the extremely small tube volume of 4  1014 mL, but owing to the very high aspect ratio, an extremely high sensitivity for detection is established; that is, used as a photometric nanocuvette, a high observation length and thus high specific light-absorption path is provided. A significant advantage of using TiO2 is its transparency for large range of light and moreover its selfcleaning properties when using static devices (MEMS or labon-a-chip). Fringes occurring in reflection spectra may also be used for interference applications as a gas- or liquid-based sensor, which can be optimized in performance by tailoring the roughness of the tube sidewalls.[144] Owing to their sensitivity to H2O2 concentration variations, TiO2 nanotubes decorated with coadsorbed horseradish peroxidase (HRP) and thionine chloride were also used as biosensors.[432] Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 To date, the fact that considerable photochromic effects have been reported for silver-loaded TiO2 nanotubes [325] or that, in fact, oxide tube layers may be reduced to metallic nanostructures has remained virtually unexploited.[172] In view of tube growth, recent studies on metallurgical aspects[187] or the growth of nanotubes in ionic liquids are very interesting.[111] Novel ordered structures may be obtained by further modifying the metallic substrate. Apart from using meshes or metallic grids to improve the surface area,[433] entirely novel concepts use for example roll-bonded substrates. Tsuchiya et al.[187] showed strategies to achieve TiAl stripes that could be selectively porosified or by which tubes could be formed. The size limit of this approach to produce alternating tube and pore layers lies in the range of 1–100 mm. It may also be noteworthy that studies on ordered surface dimple formation[113–115] are of course directly related to nanotube formation, as outlined in Section 2.7.4. Concepts to create nanosize oxide composition modulation in the tube walls that create for example highly defined junctions and 3D superlattices are also very spectacular.[246] A point that clearly should be addressed but has not yet been extensively studied is the mechanical stability of the tube layers.[213] Although general adhesion is reported to be good, the presence of the fluoride-rich layer affects the adhesion of nanotube layers, for example when bending the substrate.[80, 81] This adhesion can be improved by carbonization of the tubes,[213] but even better adhesion for anodic layers can be obtained using titania mesosponges[80, 81] or nanochannel layers.[83] While progress on nanotubes on Ti, alloys, and other valve metals is constantly advancing and new exciting findings are continuously being reported, in many aspects, an even more promising form of self-organized anodic titania has been reported, namely a mesosponge structure[80, 81] that can be grown to more than 100 mm in thickness with the main advantages that it has a much higher specific surface area than TiO2 nanotubes and a significantly stronger adherence to the metallic substrate surface. The authors would like to acknowledge the DFG for financial support and Prof. Dr. S. Virtanen, S. P. Albu, W. Wei, C. Das, I. Paramasivam, H. Jha, S. Bauer, H. Hilderbrand, R. Hahn, N. Shrestha, H. Tsuchiya, and K. Yasuda for their contributions. Received: March 8, 2010 [1] S. Iijima, Nature 1991, 354, 56. [2] C. M. Lieber, Solid State Commun. 1998, 107, 607. [3] C. N. R. Rao, A. Mller, A. K. Cheetham, The Chemistry of Nanomaterials: Synthesis, Properties and Applications, WileyVCH, Weinheim, Germany, 2006. [4] C. Weisbuch, B. Vinter, Quantum Semiconductor Structures: Fundamentals and Applications, Elsevier, Amsterdam, 1991. [5] C. N. R. Rao, M. Nath, Dalton Trans. 2003, 1. [6] Y. Feldman, E. Wasserman, D. J. Srolovitz, R. Tenne, Science 1995, 267, 222. [7] R. Tenne, L. Margulis, M. Genut, G. Hodes, Nature 1992, 360, 444.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2933 Reviews P. Schmuki et al. [8] Y. Xia, P. Yang, Y. Sun, Y. Wu, B. Mayers, B. Gates, Y. Yin, F. Kim, H. Yan, Adv. Mater. 2003, 15, 353. [9] H. Nakamura, Matsui, J. Am. Chem. Soc. 1995, 117, 2651. [10] P. Hoyer, Langmuir 1996, 12, 1411. [11] L. Pu, X. Bao, J. Zou, D. Feng, Angew. Chem. 2001, 113, 1538; Angew. Chem. Int. Ed. 2001, 40, 1490. [12] M. Remškar, Adv. Mater. 2004, 16, 1497. [13] J. L. G. Fierro, Metal Oxides: Chemistry and Applications, CRC Press, Boca Raton, 2006. [14] T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino, K. Niihara, Langmuir 1998, 14, 3160. [15] D. V. Bavykin, J. M. Friedrich, F. C. Walsh, Adv. Mater. 2006, 18, 2807. [16] P. Schmuki, S. Virtanen, Electrochemistry at the Nanoscale, Springer, New York, 2009. [17] F. Krumeich, H. J. Muhr, M. Niedeberger, F. Brieri, B. Schnyder, R. Nesper, J. Am. Chem. Soc. 1999, 121, 8324. [18] C. R. Sides, N. Li, C. J. Patrissi, B. Scrosati, C. R. Martin, MRS Bull. 2002, 27, 604. [19] A. Fujishima, T. N. Rao, D. A. Tryk, J. Photochem. Photobiol. C 2000, 1, 1. [20] A. Kudo, Y. Miseki, Chem. Soc. Rev. 2009, 38, 253. [21] S. K. Mohapatra, M. Misra, V. K. Mahajan, K. S. Roja, J. Phys. Chem. C 2007, 111, 8677. [22] A. Fujishima, K. Honda, Nature 1972, 238, 37. [23] H. Gerischer, H. Tributsch, Ber. Bunsen-Ges. 1968, 72, 437. [24] M. P. Dare-Edwards, J. B. Goodenough, A. Hamnett, K. R. Seddon, R. D. Wrigh, Faraday Discuss. Chem. Soc. 1980, 70, 285. [25] B. ORegan, M. Grtzel, Nature 1991, 353, 737. [26] M. Grtzel, Nature 2001, 414, 338. [27] J. Huusko, V. Lantto, H. Torvela, Sens. Actuators B 1993, 16, 245. [28] A. L. Linsebigler, G. Lu, J. T. Yates, Chem. Rev. 1995, 95, 735. [29] S. Y. Huang, L. Kavan, I. Exnar, M. Graetzel, J. Electrochem. Soc. 1995, 142, L142. [30] R. Wang, K. Hashimoto, A. Fujishima, M. Chikuni, E. Kojima, A. Kitamura, M. Shimohigoshi, T. Watanabe, Nature 1997, 388, 431. [31] K. Satake, A. Katayama, H. Ohkoshi, T. Nakahara, T. Takeuchi, Sens. Actuators B 1994, 20, 111. [32] H. Gerischer, Electrochim. Acta 1995, 40, 1277. [33] M. Wagemaker, D. Lutzenkirchen-Hecht, A. A. van Well, R. Frahm, J. Phys. Chem. B 2004, 108, 12456. [34] A. R. Armstrong, G. Armstrong, J. Canales, R. Garcia, P. G. Bruce, Adv. Mater. 2005, 17, 862. [35] P. Roy, D. Kim, K. Lee, E. Spiecker, P. Schmuki, Nanoscale 2010, 2, 45. [36] A. Ghicov, P. Schmuki, Chem. Commun. 2009, 2791. [37] J. M. Macak, H. Tsuchiya, A. Ghicov, K. Yasuda, R. Hahn, S. Bauer, P. Schmuki, Curr. Opin. Solid State Mater. Sci. 2007, 11, 3. [38] I. Paramasivam, H. Jha, P. Schmuki, unpublished results. [39] S. Bauer, P. Schmuki, J. Park, K. von der Mark, C. von Wilmowsky, K. A. Schlegel, F. W. Neukam, unpublished results. [40] U. Diebold, Surf. Sci. Rep. 2003, 48, 53. [41] A. Hagfeldt, M. Graetzel, Chem. Rev. 1995, 95, 49. [42] A. Mills, S. Le Hunte, J. Photochem. Photobiol. A 1997, 108, 1. [43] S. Eustis, M. A. El-Sayed, Chem. Soc. Rev. 2006, 35, 209. [44] M. S. Sander, M. J. Cote, W. Gu, B. M. Kile, C. P. Tripp, Adv. Mater. 2004, 16, 2052. [45] G. L. Li, G. H. Wang, Nanostruct. Mater. 1999, 11, 663. [46] D. Zhang, L. Qi, J. Ma, H. Cheng, J. Mater. Chem. 2002, 12, 3677. [47] T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino, K. Niihara, Adv. Mater. 1999, 11, 1307. 2934 www.angewandte.org [48] a) G. H. Du, Q. Chen, R. C. Che, Z. Y. Yuan, L. M. Peng, Appl. Phys. Lett. 2001, 79, 3702 – 3704; b) Y. Suzuki, S. Yoshikawa, J. Mater. Res. 2004, 19, 982. [49] M. Adachi, Y. Murata, S. Yoshikawa, Chem. Lett. 2000, 29, 942. [50] a) M. Adachi, Y. Murata, I. Okada, S. Yoshikawa, J. Electrochem. Soc. 2003, 150, G488. [51] H. H. Ou, S. L. Lo, Sep. Purif. Technol. 2007, 58, 179. [52] B. B. Lakshmi, P. K. Dorhout, C. R. Martin, Chem. Mater. 1997, 9, 857. [53] H. Imai, Y. Takei, K. Shimizu, M. Matsuda, H. Hirashima, J. Mater. Chem. 1999, 9, 2971. [54] S. Uchida, R. Chiba, M. Tomiha, N. Masaki, M. Shirai, Electrochemistry 2002, 70, 418. [55] K. S. Raja, M. Misra, K. Paramguru, Electrochim. Acta 2005, 51, 154. [56] Y. Nakato, H. Akanuma, J. Shimizu, Y. Magori, J. Electroanal. Chem. 1995, 396, 35. [57] T. Sugiura, T. Yoshida, H. Minoura, Electrochem. Solid-State Lett. 1998, 1, 175. [58] T. Kisumi, A. Tsujiko, K. Murakoshi, Y. Nakato, J. Electroanal. Chem. 2003, 545, 99. [59] M. Zhang, Y. Bando, K. Wada, J. Mater. Sci. Lett. 2001, 20, 167. [60] H. Shin, D. K. Jeong, J. Lee, M. M. Sung, J. Kim, Adv. Mater. 2004, 16, 1197. [61] M. Knez, K. Nielsch, L. Niinist, Adv. Mater. 2007, 19, 3425. [62] Ref. [44]. [63] M. Leskel, M. Ritala, Angew. Chem. 2003, 115, 5706; Angew. Chem. Int. Ed. 2003, 42, 5548. [64] J. Bisquert in Synthesis Properties and Applications of Oxide Nanomaterials (Eds.: J. A. Rodriguez, M. Fernandez-Garcia), Wiley, New York, 2007, p. 451. [65] A. S. Aric, P. Bruce, B. Scrosati, J. M. Tarascon, W. van Schalkwijk, Nat. Mater. 2005, 4, 366. [66] S. K. Deb, Philos. Mag. 1973, 27, 801. [67] M. A. B. Gomes, L. O. D. S. Bulhoes, S. C. De Castro, A. J. Damiao, J. Electrochem. Soc. 1990, 137, 3067. [68] S. K. Deb, Sol. Energy Mater. Sol. Cells 2008, 92, 245. [69] B. W. Faughnan, R. S. Crandall, P. M. Heyman, RCA Rev. 1975, 36, 177. [70] B. Reichman, A. J. Bard, J. Electrochem. Soc. 1980, 127, 241. [71] P. Jun, L. Duren, Chem. Senses 1998, 565. [72] S. H. Lee, R. Deshpande, P. A. Parilla, K. M. Jones, B. To, A. H. Mahan, A. C. Dillon, Adv. Mater. 2006, 18, 763. [73] K. Nakayama et al. in 208th ECS Meeting, California, Abstract 819 and 843, 2005. [74] R. Hahn, J. M. Macak, P. Schmuki, Electrochem. Commun. 2007, 9, 947. [75] R. Hahn, T. Stergiopoulus, J. M. Macak, D. Tsoukleris, A. G. Kontos, S. P. Albu, D. Kim, A. Ghicov, J. Kunze, P. Falaras, P. Schmuki, Phys. Status Solidi RRL 2007, 1, 135. [76] H. Jha, R. Hahn, P. Schmuki, unpublished results. [77] H. Jha, R. Hahn, P. Schmuki, Electrochim. Acta 2010, 55, 8883. [78] N. F. Fahim, T. Sekino, Chem. Mater. 2009, 21, 1967. [79] K. I. Ishibashi, R. T. Yamaguchi, Y. Kimura, M. Niwano, J. Electrochem. Soc. 2008, 155, K10. [80] D. Kim, K. Lee, P. Roy, B. I. Birajdar, E. Spiecker, P. Schmuki, Angew. Chem. 2009, 121, 9490; Angew. Chem. Int. Ed. 2009, 48, 9326. [81] K. Lee, D. Kim, P. Roy, I. Paramasivam, B. I. Birajdar, E. Spiecker, P. Schmuki, J. Am. Chem. Soc. 2010, 132, 1478. [82] D. Kim, P. Roy, K. Lee, P. Schmuki, Electrochem. Commun. 2010, 12, 574. [83] P. Roy, T. Dey, K. Lee, D. Kim, B. Fabry, P. Schmuki, J. Am. Chem. Soc. 2010, 132, 7893. [84] F. Keller, M. S. Hunter, D. L. Robinson, J. Electrochem. Soc. 1953, 100, 411.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes [85] J. P. O Sullivan, G. C. Wood, Proc. R. Soc. London Ser. A 1970, 317, 511. [86] G. Thompson, G. C. Wood, Anodic films on aluminum. Treatise on materials science and technology, Vol. 23, Academic Press, New York, 1983. [87] K. Uosaki, K. Okazaki, H. Kita, H. Takahashi, Anal. Chem. 1990, 62, 652. [88] C. R. Martin, Science 1994, 266, 1961. [89] H. Masuda, K. Fukuda, Science 1995, 268, 1466. [90] K. Nielsch, F. Mller, A. P. Li, U. Gsele, Adv. Mater. 2000, 12, 582. [91] M. Steinhart, J. H. Wendorff, A. Greiner, R. B. Wehrspohn, K. Nielsch, J. Schilling, J. Choi, U. Gsele, Science 2002, 296, 1997. [92] D. AlMawlawi, N. Coombs, M. Moskovits, J. Appl. Phys. 1991, 70, 4421. [93] M. Zhang, Y. Bando, K. Wada, J. Mater. Res. 2000, 15, 387. [94] J. F. Vanhumbeeck, J. Proost, Electrochim. Acta 2008, 53, 6165. [95] J. P. OSullivan, G. C. Wood, Proc. R. Soc. London Ser. A 1970, 317, 511. [96] O. Jessensky, F. Mller, U. Gsele, Appl. Phys. Lett. 1998, 72, 1173. [97] S. Ono, M. Saito, H. Asoh, Electrochem. Solid-State Lett. 2004, 7, B21 . [98] S. Ono, M. Saito, M. Ishiguro, H. Asoh, J. Electrochem. Soc. 2004, 151, B473. [99] V. Zwilling, E. Darque-Ceretti, A. Boutry-Forveille, D. David, M. Y. Perrin, M. Aucouturier, Surf. Interface Anal. 1999, 27, 629. [100] V. Zwilling, M. Aucouturier, E. Darque-Ceretti, Electrochim. Acta 1999, 45, 921. [101] J. J. Kelly, Electrochim. Acta 1979, 24, 1273. [102] D. Gong, C. A. Grimes, O. K. Varghese, Z. Chen, E. C. Dickey, J. Mater. Res. 2001, 16, 3331. [103] R. Beranek, H. Hildebrand, P. Schmuki, Electrochem. SolidState Lett. 2003, 6, B12. [104] J. M. Macak, K. Sirotna, P. Schmuki, Electrochim. Acta 2005, 50, 3679. [105] J. M. Macak, H. Tsuchiya, P. Schmuki, Angew. Chem. 2005, 117, 2136; Angew. Chem. Int. Ed. 2005, 44, 2100. [106] J. M. Macak, H. Tsuchiya, L. Taveira, S. Aldabergerova, P. Schmuki, Angew. Chem. 2005, 117, 7629; Angew. Chem. Int. Ed. 2005, 44, 7463. [107] a) O. K. Varghese, M. Paulose, K. Shankar, G. K. Mor, C. A. Grimes, J. Nanosci. Nanotechnol. 2005, 5, 1158; b) M. Paulose, K. Shankar, O. K. Varghese, G. K. Mor, B. Hardin, C. A. Grimes, Nanotechnol. 2010, 21, 389801; c) M. Paulose, G. K. Mor, O. K. Varghese, K. Shankar, C. A. Grimes, J. Photochem. Photobiol. A, 2010, 215, 229. [108] S. P. Albu, A. Ghicov, J. M. Macak, P. Schmuki, Phys. Status Solidi RRL 2007, 1, R65. [109] H. Masuda, M. Satoh, Jpn. J. Appl. Phys. Part 1 1996, 35, L126. [110] J. M. Macak, S. P. Albu, P. Schmuki, Phys. Status Solidi RRL 2007, 1, 181. [111] I. Paramasivam, J. M. Macak, T. Selvam, P. Schmuki, Electrochim. Acta 2008, 54, 643. [112] H. Tsuchiya, J. M. Macak, L. Taveira, E. Balaur, A. Ghicov, K. Sirotna, P. Schmuki, Electrochem. Commun. 2005, 7, 576. [113] H. El-Sayed, S. Singh, M. T. Greiner, P. Kruse, Nano Lett. 2006, 6, 2995. [114] a) V. Vignal, J. C. Roux, S. Flandrois, A. Fevrier, Corros. Sci. 2000, 42, 1041; b) H. Tsuchiya, T. Suzumura, S. Fujimoto, 218th Meeting of the Electrochemical Society. [115] S. Singh, W. R. T. Barden, P. Kruse, ACS Nano 2008, 2, 2453. [116] H. Masuda, H. Yamada, M. Satoh, H. Asoh, M. Nakao, T. Tamamura, Appl. Phys. Lett. 1997, 71, 2770. [117] L. V. Taveira, J. M. Macak, H. Tsuchiya, L. F. P. Dick, P. Schmuki, J. Electrochem. Soc. 2005, 152, B405. Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 [118] S. P. Albu, A. Ghicov, S. Aldabergenova, P. Drechsel, D. LeClere, G. E. Thompson, J. M. Macak, P. Schmuki, Adv. Mater. 2008, 20, 4135. [119] D. Kim, A. Ghicov, P. Schmuki, Electrochem. Commun. 2008, 10, 1835. [120] Y. Y. Song, R. Lynch, D. Kim, P. Roy, P. Schmuki, Electrochem. Solid-State Lett. 2009, 12, C17. [121] S. P. Albu, P. Schmuki, Phys. Status Solidi RRL 2010, 4, 151. [122] J. Park, S. Bauer, K. Von Der Mark, P. Schmuki, Nano Lett. 2007, 7, 1686. [123] S. P. Albu, A. Ghicov, J. M. Macak, R. Hahn, P. Schmuki, Nano Lett. 2007, 7, 1286. [124] K. Yasuda, P. Schmuki, Electrochim. Acta 2007, 52, 4053. [125] D. Kim, F. Schmidt-Stein, R. Hahn, P. Schmuki, Electrochem. Commun. 2008, 10, 1082. [126] K. Yasuda, J. M. Macak, S. Berger, A. Ghicov, P. Schmuki, J. Electrochem. Soc. 2007, 154, C472. [127] S. Berger, J. Kunze, P. Schmuki, D. LeClere, A. Valota, P. Skeldon, G. E. Thompson, Electrochim. Acta 2009, 54, 5942. [128] Y. V. Bhargava, Q. A. S Nguyen, T. M. Devine, J. Electrochem. Soc. 2009, 156, E62. [129] J. M. Macak, H. Hildebrand, U. Marten-Jahns, P. Schmuki, J. Electroanal. Chem. 2008, 621, 254. [130] S. P. Albu, A. Ghicov, P. Schmuki, unpublished results. [131] J. M. Macak, S. Aldabergerova, A. Ghicov, P. Schmuki, Phys. Status Solidi A 2006, 203, R67. [132] S. Bauer, S. Kleber, P. Schmuki, Electrochem. Commun. 2006, 8, 1321. [133] D. J. LeClere, . A. Velota, P. Skeldon, G. E. Thompson, S. Berger, J. Kunze, P. Schmuki, H. Habazaki, S. Nagata, J. Electrochem. Soc. 2008, 155, C487. [134] K. Lee, J. Lee, H. Kim, Y. Lee, D. Kim, P. Schmuki, Y. Tak, J. Korean Phys. Soc. 2009, 54, 1027. [135] A. Valota, D.J LeClere, T. Hashimoto, P. Skeldon, G. E. Thompson, S. Berger, J. Kunze, P. Schmuki, Nanotechnology 2008, 19, 355701. [136] J. M. Macak, P. Schmuki, Electrochim. Acta 2006, 52, 1258. [137] S. P. Albu, P. Schmuki, Phys. Status Solidi RRL 2010, 4, 215. [138] Y. Y. Song, P. Roy, I. Paramasivam, P. Schmuki, Angew. Chem. 2010, 122, 361; Angew. Chem. Int. Ed. 2010, 49, 351. [139] S. P. ALbu, P. SChmuki, unpublished results. [140] J. Kunze, A. Ghicov, H. Hildebrand, J. M. Macak, L. Traveira, P. Schmuki, Z. Phys. Chem. 2005, 219, 1561. [141] J. Lausmaa, J. Electron Spectrosc. Relat. Phenom. 1996, 81, 343. [142] F. Thbault, B. Vuillemin, R. Oltra, J. Kunze, A. Seyeux, P. Schmuki, Electrochem. Solid-State Lett. 2009, 12, C5 . [143] K. Von Der Mark, J. Park, S. Bauer, P. Schmuki, Cell Tissue Res. 2010, 339, 131. [144] Y. Y. Song, P. Schmuki, Electrochem. Commun. 2010, 12, 579. [145] S. Berger, J. Kunze, P. Schmuki, A. T. Valota, D. J. Leclere, P. Skeldon, G. E. Thompson, J. Electrochem. Soc. 2010, 157, C18. [146] L. V. Taveira, J. M. Macak, K. Sirotna, L. F. P. Dick, P. Schmuki, J. Electrochem. Soc. 2006, 153, B137. [147] W. Chanmanee, A. Watcharenwong, C. R. Chenthamarakshan, P. Kajitvichyanukul, N. R. de Tacconi, K. Rajeshwar, J. Am. Chem. Soc. 2008, 130, 965. [148] S. Banerjee, S. K Mohapatra, M. Misra, I. B Mishra, Nanotechnology 2009, 20, 075502. [149] N. B. Pilling, R. E. Bedworth, J. Inst. Met. 1923, 29, 529. [150] H. Habazaki, M. Uozumi, H. Konno, K. Shimizu, S. Nagata, K. Takayama, Y. Oda, P. Skeldon, G. E. Thompson, J. Electrochem. Soc. 2005, 152, B263. [151] P. Skeldon, G. E. Thompson, S. J. Garcia-Vergara, L. IglesiasRubianes, C. E. Blanco-Pinzon, Electrochem. Solid-State Lett. 2006, 9, B47. [152] S. J. Garcia-Vergara, P. Skeldon, G. E. Thompson, H. Habazaki, Electrochim. Acta 2006, 52, 681.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2935 Reviews P. Schmuki et al. [153] K. R. Hebert, J. E. Houser, J. Electrochem. Soc. 2009, 156, C275. [154] J. E. Houser, K. R. Hebert, Nat. Mater. 2009, 8, 415. [155] S. Berger, J. M. Macak, J. Kunze, P. Schmuki, Electrochem. Solid-State Lett. 2008, 11, C37. [156] Q. A. S. Nguyena, Y. V. Bhargava, V. R. Radmilovic, T. M. Devine, Electrochim. Acta 2009, 54, 4340. [157] A. Valota, D. J. LeClere, P. Skeldon, M. Curioni, T. Hashimoto, S. Berger, J. Kunze, P. Schmuki, G. E. Thompson, Electrochim. Acta 2009, 54, 4321. [158] H. Tsuchiya, S. Berger, J. M. Macak, A. Ghicov, P. Schmuki, Electrochem. Commun. 2007, 9, 2397. [159] S. Berger, H. Tsuchiya, P. Schmuki, Chem. Mater. 2008, 20, 3245. [160] R. Blachnik, J. D’Ans, E. Lax, C. Synowietz, Taschenbuch fr Chemiker und Physiker, Volume 3, Springer, Berlin, 1998. [161] W. Wei, S. Berger, N. Shrestha, P. Schmuki, unpublished results. [162] S. P. Albu, P. Schmuki, unpublished results. [163] S. Berger, F. Jakubka, P. Schmuki, Electrochem. Solid-State Lett. 2009, 12, K45. [164] S. Berger, F. Jakubka, P. Schmuki, Electrochem. Commun. 2008, 10, 1916. [165] H. Tsuchiya, J. M. MacAk, A. Ghicov, L. Taveira, P. Schmuki, Corros. Sci. 2005, 47, 3324. [166] H. Tsuchiya, P. Schmuki, Electrochem. Commun. 2005, 7, 49. [167] Motoo Kawasaki, Hidemi Nawafune, Farthest Advances in Surface Finishing of Aluminum Visualized by Electron Microscopy Observation, The Surface Finishing Society of Japan, Tokyo, 2000, pp. 99 – 100 (in Japanese). [168] D. J. Arrowsmith, A. W. Clifford, D. A. Moth, J. Mater. Sci. Lett. 1986, 5, 921. [169] S. Ono, K. Takeda, N. Masuko, 2nd International Symposium on Aluminium Surface Science and Technology, ASST 2000, UMIST, Manchester, UK. [170] S.-Z. Chu, K. Wada, S. Inoue, M. Isogai, A. Yasumori, Adv. Mater. 2007, 19, 2115. [171] W. Wei, S. Berger, C. Hauser, K. Mayer, M. Yang, P. Schmuki, Electrochem. Commun. 2010, 12, 1184. [172] K. Yasuda, P. Schmuki, Electrochem. Commun. 2007, 9, 615. [173] a) J. J. Urban, W. S. Yun, Q. Gu, H. Park, J. Am. Chem. Soc. 2002, 124, 1186; b) B. A. Hernandez, K. S. Chang, E. R. Fisher, P. K. Dorhout, Chem. Mater. 2002, 14, 480. [174] a) K. G. Singh, A. A. Golovin, I. S. Aranson, Phys. Rev. B 2006, 73, 205422; b) L. G. Stanton, A. A. Golovin, Phys. Rev. B 2009, 79, 035414. [175] S. Ono, M. Saito, H. Asoh, Electrochim. Acta 2005, 51, 827. [176] Q. Van Overmeere, F. Blaffart, J. Proost, Electrochem. Commun. 2010, 12, 1174. [177] Y. Y. Song, F. Schmidt-Stein, S. Bauer, P. Schmuki, J. Am. Chem. Soc. 2009, 131, 4230. [178] S. P. Albu, D. Kim, P. Schmuki, Angew. Chem. 2008, 120, 1942; Angew. Chem. Int. Ed. 2008, 47, 1916. [179] D. Kim, A. Ghicov, S. P. Albu, P. Schmuki, J. Am. Chem. Soc. 2008, 130, 16454. [180] J. Kunze, A. Seyeux, P. Schmuki, Electrochem. Solid-State Lett. 2008, 11, K11. [181] P. Roy, T. Dey, P. Schmuki, Electrochem. Solid-State Lett. 2010, 13, E11. [182] H. Asoh, S. Ono in Electrocrystallization in Nanotechnology (Ed.: G. T. Staikov), Wiley-VCH, Weinheim, 2007, p. 138. [183] Y.-C. Nah, N. K. Shrestha, D. Kim, S. P. Albu, I. Paramasivam, P. Schmuki, Electrochem. Solid-State Lett. 2010, 13, K73. [184] K. Zhu, T. B. Vinzant, N. R. Neale, A. J. Frank, Nano Lett. 2007, 7, 3739. [185] J. H. Lim, J. Choi, Small 2007, 3, 1504. [186] H. Tsuchiya, J. M. Macak, A. Ghlcov, P. Schmuki, Small 2006, 2, 888. 2936 www.angewandte.org [187] H. Tsuchiya, T. Akaki, D. Terada, N. Tsuji, Y. Minamino, P. Schmuki, S. Fujimoto, Electrochim. Acta 2009, 54, 5155. [188] A. Ghicov, S. Aldabergenova, H. Tsuchiya, P. Schmuki, Angew. Chem. 2006, 118, 7150 – 7153; Angew. Chem. Int. Ed. 2006, 45, 6993. [189] K. Yasuda, P. Schmuki, Adv. Mater. 2007, 19, 1757. [190] H. Tsuchiya, T. Akaki, J. Nakata, D. Terada, N. Tsuji, Y. Koizumi, Y. Minamino, P. Schmuki, S. Fujimoto, Corros. Sci. 2009, 51, 1528. [191] X. J. Feng, J. M. Macak, S. P. Albu, P. Schmuki, Acta Biomater. 2008, 4, 318. [192] J. M. Macak, H. Tsuchiya, L. Taveira, A. Ghicov, P. Schmuki, J. Biomed. Mater. Res. Part A 2005, 75, 928. [193] S. P. Albu, A. Ghicov, S. Berger, H. Jha, P. Schmuki, Electrochem. Commun. 2010, 12, 1352. [194] C. J. Lin, W. Y. Yu, Y. T. Lu, S. H. Chien, Chem. Commun. 2008, 6031. [195] K. Kant, D. Losic, Phys. Status Solidi RRL 2009, 3, 139. [196] J. Wang, Z. Lin, Chem. Mater. 2008, 20, 1257. [197] T. L. Thompson, J. T. Yates, Jr., Chem. Rev. 2006, 106, 4428. [198] M. R. Hoffmann, S. T. Martin, W. Choi, D. W. Bahneman, Chem. Rev. 1995, 95, 69. [199] R. Marchand, L. Broham, M. Tournoux, Mater. Res. Bull. 1980, 15, 1129. [200] J. Muscat, V. Swany, N. M. Harrison, Phys. Rev. B 2002, 65, 224112. [201] S. Bauer, H. Tsuchiya, A. Pittrof, P. Schmuki, unpublished results. [202] M. Lazzeri, A. Vittadini, A. Selloni, Phys. Rev. B 2001, 63, 155409. [203] W. Huang, X. Tang, Y. Wang, Y. Kolypin, A. Gedankan, Chem. Commun. 2000, 1415. [204] H. Yin, Y. Wada, T. Kitamura, S. Kambe, A. Murasawa, H. Mori, T. Sakata, S. Yangida, J. Mater. Chem. 2001, 11, 1694. [205] S. T. Aruna, S. Tirosh, A. Zaban, J. Mater. Chem. 2000, 10, 2388. [206] T. Ohtsuka, J. Guo, N. Sato, J. Electrochem. Soc. 1986, 133, 2473. [207] A. Aladjem, D. Brandon, J. Yahalom, J. Zahavi, Electrochim. Acta 1970, 15, 663. [208] T. Ohtsuka, T. Otsuki, Corros. Sci. 2003, 45, 1793. [209] S. Yamaguchi, J. Electrochem. Soc. 1961, 108, 302. [210] J. R. Smith, F. C. Walsh, R. L. Clarke, J. Appl. Electrochem. 1998, 28, 1021. [211] M. Zweynert, H. Doring, J. Garche, K. Enghardt, K. Wiesener, Chem. Ing. Tech. 1998, 7, 827. [212] R. Hahn, F. Schmidt-Stein, J. Salonen, S. Thiemann, Y. Y. Song, J. Kunze, V. P. Lehto, P. Schmuki, Angew. Chem. 2009, 121, 7372; Angew. Chem. Int. Ed. 2009, 48, 7236. [213] F. Schmidt-Stein, S. Thiemann, S. Berger, R. Hahn, P. Schmuki, Acta Mater. 2010, 58, 6317. [214] S. P. Albu, H. Tsuchiya, S. Fujimoto, P. Schmuki, Europ. J. Inorg. Chem. 2010, 27, 4351. [215] J. C. Marchenoir, J. P. Loup, J. Masson, Thin Solid Films 1980, 66, 357. [216] J.-L. Delplancke, A. Garnier, Y. Massiani, R. Winand, Electrochim. Acta 1994, 39, 1281. [217] J. C. Marchenoir, J. Gautron, J. P. L. Metaux, Corrosion-Industrie 1977, 83. [218] J. Arbiol, J. Cerda, G. Dezanneau, A. Cicera, F. Peiro, A. Cornet, J. R. Morante, J. Appl. Phys. 2002, 92, 853. [219] V. Guidi, M. C. Carotta, M. Ferroni, G. Martinelli, M. Sacerdoti, J. Phys. Chem. B 2003, 107, 120. [220] L. E. Depero, L. Sangaletti, B. Allieri, E. Bontempi, A. Marino, M. Zocchi, J. Cryst. Growth 1999, 198, 516. [221] L. E. Depero, A. Marino, B. Allieri, E. Bontempi, L. Sangaletti, C. Casale, M. Notaro, J. Mater. Res. 2000, 15, 2080.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes [222] V. Likodimos, T. Stergiopoulos, P. Falaras, J. Kunze, P, Schmuki, J. Phys. Chem. C 2008, 112, 12687. [223] S. B. Aldabergenova, A. Ghicov, S. Albu, J. M. Macak, P. Schmuki, J. Non-Cryst. Solids 2008, 354, 2190 – 2194. [224] Y. C. Nah, A. Ghicov, D. Kim, S. Berger, P. Schmuki, J. Am. Chem. Soc. 2008, 130, 16154. [225] R. Asahi, Y. Taga, W. Mannstadt, A. J. Freeman, Phys. Rev. B 2000, 61, 7459. [226] Z. Y. Wu, G. Ouvrared, P. Gressier, C. R. Natoli, Phys. Rev. B 1997, 55, 10382. [227] H. Tang, K. Prasad, R. Sanjinbs, P. E. Schmid, F. Levy, J. Appl. Phys. 1994, 75, 2042. [228] J. Pascual, J. Camassel, H. Mathieu, Phys. Rev. B 1978, 18, 5606. [229] K. Eufinger, D. Poelman, H. Poelman, R. De Gryse, G. B. Marin, Appl. Surf. Sci. 2007, 254, 148. [230] V. M. Naik, D. Haddad, R. Naik, J. Benci, G. W. Auner, Mater. Res. Soc. Symp. Proc. 2002, 755, 413. [231] A. K. Ghosh, F. G. Wakim, P. R. Addiss, Phys. Rev. Lett. 1969, 22–23, 979. [232] L. Forro, O. Chauvet, D. Emin, L. Zuppiroli, H. Berger, F. Levy, J. Appl. Phys. 1994, 75, 633. [233] R. Beranek, H. Tsuchiya, T. Sugishima, J. M. Macak, L. Taveira, S. Fujimoto, H. Kisch, P. Schmuki, Appl. Phys. Lett. 2005, 87, 243114. [234] A. G. Munoz, Q. Chen, P. Schmuki, J. Solid State Electrochem. 2007, 11, 1077. [235] R. P. Lynch, A. Ghicov, P. Schmuki, J. Electrochem. Soc. 2010, 157, G76. [236] S. R. Morrison, Electrochemistry at Semiconductor and Oxidized Metal Electrodes, Plenum Press, New York, 1980, p. 401. [237] H. Tsuchiya, J. M. Macak, A. Ghicov, A. S. Rder, L. Taveira, P. Schmuki, Corros. Sci. 2007, 49, 203. [238] L. V. Taveira, A. A. Sagus, J. M. Macak, P. Schmuki, J. Electrochem. Soc. 2008, 155, C293. [239] W. W. Gartner, Phys. Rev. 1959, 116, 84. [240] U. Diebold, T. E. Madey, J. Vac. Sci. Technol. A 1992, 10, 2327. [241] D. R. Lide, Handbook of Chemistry and Physics, 84th ed., CRC press LLC, Florida, 2003. [242] V. V. Sergentu, I. M. Tiginyanu, V. V. Ursaki, M. Enachi, S. P. Albu, P. Schmuki, Phys. Status Solidi RRL 2008, 2, 242. [243] B. J. Aronson, C. F. Blanford, A. Stein, Chem. Mater. 1997, 9, 2842. [244] W. Dong, H. Bongard, B. Tesche, F. Marlow, AdV. Mater. 2002, 14, 1457. [245] B. T. Holland, C. Blanford, A. Stein, Science 1998, 281, 538. [246] W. Wei, H. Jha, G. Yang, R. Hahn, I. Paramasivam, S. Berger, E. Spiecker, P. Schmuki, Adv. Mater. 2010, 22, 4770. [247] A. Tighineanu, T. Ruff, S. P. Albu, R. Hahn, P. Schmuki, Chem. Phys. Lett. 2010, 494, 260. [248] M. Kunat, U. Burghaus, Surf. Sci. 2003, 544, 170. [249] S. Funk, U. Burghaus, Catal. Lett. 2007, 118, 118. [250] V. I. P rvulescu, P. Grange, B. Delmon, Catal. Today 1998, 46, 233. [251] S. Funk, B. Hokkanen, U. Burghaus, A. Ghicov, P. Schmuki, Nano Lett. 2007, 7, 1091. [252] M. Kunat, U. Burghaus, C. Woll, Phys. Chem. Chem. Phys. 2004, 6, 4203. [253] B. Hokkanen, S. Funk, U. Burghaus, A. Ghicov, P. Schmuki, Surf. Sci. 2007, 601, 4620. [254] K. E. Smith, V. E. Henrich, Phys. Rev. B 1986, 32, 5384. [255] K. E. Smith, J. L. Mackay, V. E. Henrich, Phys. Rev. E 1987, 35, 582. [256] A. Kiennemann, H. Idriss, R. Kieffer, P. Chaumette, D. Durand, Ind. Eng. Chem. Res. 1991, 30, 1130. [257] W. Gopel, G. Rocker, R. Feierabend, Phys. Rev. E 1983, 28, 3427. [258] J. Robertson, P. W. Peacock, Thin Solid Films 2003, 445, 155. Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 [259] P. W. Peacock, J. Robertson, Appl. Phys. Lett. 2003, 83, 2025. [260] H. Kobayashi, M. Yamaguchi, Surf. Sci. 1989, 214, 466. [261] G. Lu, A. Linsebigler, J. T., Jr. Yates, J. Phys. Chem. 1994, 98, 11733. [262] K. E. Smith, V. E Henrich, Surf. Sci. 1989, 217, 445. [263] K. S. Kim, M. A. Barteau, W. E. Farneth, Langmuir 1988, 4, 533. [264] K. S. Kim, M. A. Barteau, Surf. Sci. 1989, 233, 13. [265] J. M. Pan, B. L. Maschhoff, U. Diebold, T. E. Madey, J. Vac. Sci. Technol. A 1992, 10, 2470. [266] R. L. Kurtz, R. Stockbauer, T. E. Madey, E. Roman, J. L. de Segovia, Surf. Sci. 1989, 218, 178. [267] A. Ghicov, H. Tsuchiya, J. M. Macak, P. Schmuki, Phys. Status Solidi A 2006, 203, R28. [268] J. W. Schultze, M. M. Lohrengel, Electrochim. Acta 2000, 45, 2499. [269] L. Kavan, M. Graetzel, S.E. Gilbert, C. Klemenz, H. J. Scheel, J. Am. Chem. Soc. 1996, 118, 6716. [270] J. P. Randin, J. Electron. Mater. 1978, 7, 47. [271] I. F. Chang, Nonemissive Electrooptic Disp. [Proc. Brown Boveri Symp.], 4th, 1976, 155 – 196. [272] T. Oi, Annu. Rev. Mater. Sci. 1986, 16, 185. [273] D. C. Cronemeyer, Phys. Rev. 1952, 87, 876. [274] J. M. Macak, F. Schmidt-Stein, P. Schmuki, Electrochem. Commun. 2007, 9, 1783. [275] J. M. Macak, P. J. Barczuk, H. Tsuchiya, M. Z. Nowakowska, A. Ghicov, M. Chojak, S. Bauer, S. Virtanen, P. J. Kulesza, P. Schmuki, Electrochem. Commun. 2005, 7, 1417. [276] J. Qu, X. Zhang, Y. Wang, C. Xie, Electrochim. Acta 2005, 50, 3576. [277] M. Wang, D.-J. Guo, H.-L. Li, J. Solid State Chem. 2005, 178, 1996. [278] F. Hu, F. Ding, S. Song, P. K. Shen, J. Power Sources 2006, 163, 415. [279] I. Tanahashi, H. Iwagishi, G. Chang, Mater. Lett. 2008, 62, 2714. [280] A. Ghicov, S. P. Albu, R. Hahn, D. Kim, T. Stergiopoulos, J. Kunze, C.-A. Schiller, P. Falaras, P. Schmuki, Chem. Asian J. 2009, 4, 520. [281] Y. Y. Song, H. Hildebrand, P. Schmuki, Surf. Sci. 2010, 604, 346. [282] Unpublished results. [283] M. Anpo, Catal. Surv. Jpn. 1997, 1, 169. [284] S. Sato, Chem. Phys. Lett. 1986, 123, 126. [285] R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki, Y. Taga, Science 2001, 293, 269. [286] T. Umebayashi, T. Yamaki, S. Tanaka, K. Asai, Chem. Lett. 2003, 32, 330. [287] T. Ohno, T. Mitsui, M. Matsumura, Chem. Lett. 2003, 32, 364. [288] J. C. Yu, W. Ho, J. Yu, H. Yip, P. K. Wong, J. Zhao, Environ. Sci. Technol. 2005, 39, 1175. [289] S. Sakthivel, H. Kisch, Angew. Chem. 2003, 115, 5057; Angew. Chem. Int. Ed. 2003, 42, 4908. [290] T. Umebayashi, T. Yamaki, H. Itoh, K. Asai, Appl. Phys. Lett. 2002, 81, 454. [291] A. B. Murphy, Sol. Energy Mater. Sol. Cells 2008, 92, 363. [292] A. K. Ghosh, H. P. Maruska, J. Electrochem. Soc. 1977, 124, 1516. [293] M. Anpo, Pure Appl. Chem. 2000, 72, 1787. [294] H. Yamashita, M. Harada, J. Misaka, M. Takeuchi, K. Ikeue, M. Anpo, J. Photochem. Photobiol. A 2002, 148, 257. [295] X. Chen, C. Burda, J. Phys. Chem. B 2004, 108, 15446. [296] S. M. Prokes, J. L. Gole, X. Chen, C. Burda, W. E. Carlos, Adv. Funct. Mater. 2005, 15, 161. [297] X. Qiu, C. Burda, Chem. Phys. 2007, 339, 1. [298] J. H. Park, S. Kim, A. J. Bard, Nano Lett. 2006, 6, 24. [299] R. Hahn, A. Ghicov, J. Salonen, V.-P. Lehto, P. Schmuki, Nanotechnology 2007, 18, 105604.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2937 Reviews P. Schmuki et al. [300] E. Barborini, A. M. Conti, I. Kholmanov, P. Piseri, A. Podesta, P. Milani, C. Cepek, O. Sakho, R. Macovez, M. Sancrotti, Adv. Mater. 2005, 17, 1842. [301] T. Morikawa, R. Asahi, T. Ohwaki, K. Aoki, Y. Taga, Jpn. J. Appl. Phys. 2001, 40, L561. [302] A. Ghicov, J. M. Macak, H. Tsuchiya, J. Kunze, V. Haeublein, L. Frey, P. Schmuki, Nano Lett. 2006, 6, 1080. [303] A. Ghicov, J. M. Macak, H. Tsuchiya, J. Kunze, V. Haeublein, S. Kleber, P. Schmuki, Chem. Phys. Lett. 2006, 419, 426. [304] D. Kim, S. Fujimoto, P. Schmuki, H. Tsuchiya, Electrochem. Commun. 2008, 10, 910. [305] A. Ghicov, B. Schmidt, J. Kunze, P. Schmuki, Chem. Phys. Lett. 2007, 433, 323. [306] R. P. Vitiello, J. M. Macak, A. Ghicov, H. Tsuchiya, L. F. P. Dick, P. Schmuki, Electrochem. Commun. 2006, 8, 544. [307] R. Beraneka, J. M. Macak, M. Grtner, K. Meyer, P. Schmuki, Electrochim. Acta 2009, 54, 2640. [308] A. Ghicov, M. Yamamoto, P. Schmuki, Angew. Chem. 2008, 120, 8052 – 8055; Angew. Chem. Int. Ed. 2008, 47, 7934. [309] O. K. Varghese, M. Paulose, T. J. LaTempa, C. A. Grimes, Nano Lett. 2009, 9, 731. [310] F. Jona, G. Shirane, Ferroelectric Crystals, MacMillan, New York, 1976. [311] S. Euphrasie, S. Daviero-Minaud, P. Pernod, Mater. Sci. Eng. B 2003, 104, 180. [312] Y. Hu, H. Gu, J. You, K. Zheng, J. Wang, Key Eng. Mater. 2007, 2157, 336. [313] Y. Mao, S. Banerjeea, S. S. Wong, Chem. Commun. 2003, 408. [314] N. P. Padture, X. Wei, J. Am. Ceram. Soc. 2003, 86, 2215. [315] X. Wei, A. L. Vasiliev, N. P. Padture, J. Mater. Res. 2005, 20, 2140. [316] Y. Xin, J. Jiang, K. Huo, T. Hu, P. K. Chu, ACS Nano 2009, 3, 3228. [317] J. M. Macak, C. Zollfrank, B. J. Rodriguez, H. Tsuchiya, M. Alexe, P. Greil, P. Schmuki, Adv. Mater. 2009, 21, 3121. [318] Y. Yang, X. Wang, C. Zhong, C. Sun, L. Li, Appl. Phys. Lett. 2008, 92, 122907. [319] Q. Chen, G. H. Du, S. Zhang, L.-M. Peng, Acta Crystallogr. Sect. B 2002, 58, 587. [320] J. M. Macak, B. G. Gong, M. Hueppe, P. Schmuki, Adv. Mater. 2007, 19, 3027. [321] A. Benoit, I. Paramasivam, Y. C. Nah, P. Roy, P. Schmuki, Electrochem. Commun. 2009, 11, 728. [322] P. Roy, D. Kim, I. Paramasivam, P. Schmuki, Electrochem. Commun. 2009, 11, 1001. [323] N. K. Shrestha, M. Yang, Y. C. Nah, I. Paramasivam, P. Schmuki, Electrochem. Commun. 2010, 12, 254. [324] N. K. Shrestha, J. M. Macak, F. Schmidt-Stein, R. Hahn, C. T. Mierke, B. Fabry, P. Schmuki, Angew. Chem. 2009, 121, 987; Angew. Chem. Int. Ed. 2009, 48, 969. [325] I. Paramasivam, J. M. Macak, A. Ghicov, P. Schmuki, Chem. Phys. Lett. 2007, 445, 233. [326] I. Paramasivam, J. M. Macak, P. Schmuki, Electrochem. Commun. 2008, 10, 71. [327] H. Tsuchiya, N. Tokuoka, S. Honda, Y. Shinkai, Y. Shimizu, S. Fujimoto, J. Phys. Conf. Series 2009, 165, 012037. [328] I. Paramasivam, A. Avhale, A. Inayat, A. Bosmann, P. Schmuki, W. Schwieger, Nanotechnology 2009, 20, 225607. [329] A. Kongkanand, K. Tvrdy, K. Takechi, M. Kuno, P. V. Kamat, J. Am. Chem. Soc. 2008, 130, 4007. [330] W. T. Sun, Y. Yu, H. Y. Pan, X. F. Gao, Q. Chen, L.-M Peng, J. Am. Chem. Soc. 2008, 130, 1124. [331] H. Gerischer, M. Luebke, J. Electroanal. Chem. 1986, 204, 225. [332] R. Plass, S. Pelet, J. Krueger, M. Gratzel, U. Bach, J. Phys. Chem. B 2002, 106, 7578. [333] R. Helmy, A. Y. Fadeev, Langmuir 2002, 18, 8924. 2938 www.angewandte.org [334] B. M. Silverman, K. A. Wieghaus, J. Schwartz, Langmuir 2005, 21, 225. [335] E. Balaur, J. M. Macak, L. Taveira, P. Schmuki, Electrochem. Commun. 2005, 7, 1066. [336] A. Mishra, M. K. R. Fischer, P. Buerle, Angew. Chem. 2009, 121, 2510; Angew. Chem. Int. Ed. 2009, 48, 2474. [337] W. Lu, G. Wang, Y. Jin, X. Yao, J. Hu, J. Li, Appl. Phys. Lett. 2006, 89, 263902. [338] A. Nanci, J. D. Wuest, L. Peru, P. Brunet, V. Sharma, S. Zalzal, M. D. McKee, J. Biomed. Mater. Res. 1998, 40, 324. [339] X. Liu, P. K. Chub, C. Dinga, Mater. Sci. Eng. 2004, 47, 49. [340] B. D. Ratner in Biomaterials Science: An Interdisciplinary Endeavour (Eds.: B. D. Ratner, A. S. Hoffmann, F. J. Schoen, J. E. Lemons),Academic Press, San Diego, 1996, pp. 37 – 130. [341] M. Morra, Cells Mater. 2006, 12, 1. [342] J. M. White, J. Szanyi, M. A. Henderson, J. Phys. Chem. B 2003, 107, 9029. [343] E. Balaur, J. M. Macak, H. Tsuchiya, P. Schmuki, J. Mater. Chem. 2005, 15, 4488. [344] F. Schmidt-Stein, R. Hahn, J. F. Gnichwitz, Y. Y. Song, N. K. Shrestha, A. Hirsch, P. Schmuki, Electrochem. Commun. 2009, 11, 2077. [345] A. B. D. Cassie, S. Baxter, Trans. Faraday Soc. 1944, 40, 546. [346] D. Kim, J. M. Macak, F. Schmidt-Stein, P. Schmuki, Nanotechnology 2008, 19, 305710. [347] F. Schmidt-Stein, J. F. Gnichwitz, J. Salonen, R. Hahn, A. Hirsch, P. Schmuki, Electrochem. Commun. 2009, 11, 2000. [348] S. Berger, J. Faltenbacher, S. Bauer, P. Schmuki, Phys. Status Solidi RRL 2008, 2, 102. [349] H. Tsuchiya, P. Schmuki, Electrochem. Commun. 2004, 6, 1131. [350] H. Tsuchiya, J. M. Macak, L. Taveira, P. Schmuki, Chem. Phys. Lett. 2005, 410, 188. [351] W. J. Lee, W. H. Smyrl, Electrochem. Solid-State Lett. 2005, 8, B7. [352] Y. Shin, S. Lee, Nanotechnology 2009, 20, 105301. [353] H. Tsuchiya, J. M. Macak, I. Sieber, L. Taveira, A. Ghicov, K. Sirotna, P. Schmuki, Electrochem. Commun. 2005, 7, 295. [354] N. R. de Tacconi, C. R. Chenthamarakshan, G. Yogeeswaran, A. Watcharenwong, R. S. de Zoysa, N. A. Basit, K. Rajeshwar, J. Phys. Chem. B 2006, 110, 25347. [355] M. Yang, N. K. Shrestha, P. Schmuki, Electrochem. Commun. 2009, 11, 1908. [356] A. Watcharenwong, W. Chanmanee, N. R. de Tacconi, C. R. Chenthamarakshan, P. Kajitvichyanukul, K. Rajeshwar, J. Electroanal. Chem. 2008, 612, 112. [357] I. Sieber, H. Hildebrand, A. Friedrich, P. Schmuki, Electrochem. Commun. 2005, 7, 97. [358] S. Ono, T. Nagasaka, H. Shimazaki, H. Asoh, Electrochem. Soc. Interface 2006, 123, 2004. [359] I. Sieber, B. Kannan, P. Schmuki, Electrochem. Solid-State Lett. 2005, 8, J10. [360] I. V. Sieber, P. Schmuki, J. Electrochem. Soc. 2005, 152, C639. [361] W. Wei, J. M. Macak, P. Schmuki, Electrochem. Commun. 2008, 10, 428. [362] W. Wei, J. M. Macak, N. K. Shrestha, P. Schmuki, J. Electrochem. Soc. 2009, 156, K104. [363] H. A. El-Sayed, V. I. Birss, Nano Lett. 2009, 9, 1350. [364] S. P. Albu, A. Ghicov, P. Schmuki, Phys. Status Solidi RRL 2009, 3, 64. [365] T. D. Burleigh, P. Schmuki, S. Virtanen, J. Electrochem. Soc. 2009, 156, C45. [366] S. K. Mohapatra, S. E. John, S. Banerjee, M. Misra, Chem. Mater. 2009, 21, 3048. [367] C. M. Turhan, R. Lynch, H. Jha, P. Schmuki, S. Virtanen, Electrochem. Commun. 2010, 12, 796. [368] S. Berger, H. Tsuchiya, A. Ghicov, P. Schmuki, Appl. Phys. Lett. 2006, 88, 203119.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 TiO2 Nanotubes [369] H. D. Zheng, A. Z. Sadek, K. Latham, K. Kalantar-Zadeh, Electrochem. Commun. 2009, 11, 768. [370] A. Z. Sadek, H. D. Zheng, M. Breedon, Langmuir 2009, 25, 9545. [371] Y. C. Nah, A. Ghicov, D. Kim, P. Schmuki, Electrochem. Commun. 2008, 10, 1777. [372] S. Hotchandani, I. Bedja, R. Fessenden, P. Kamat, Langmuir 1994, 10, 17. [373] M. Deepa, N. Sharma, P. Varshney, S. P. Varma, S. A. Agnihotry, J. Mater. Sci. 2000, 35, 5313. [374] W. Lee, D. Kim, K. Lee, P. Roy, P. Schmuki, Electrochim. Acta 2010, 56, 828. [375] C. Santato, M. Odziemkowski, M. Ulmann, J. Augustynski, J. Am. Chem. Soc. 2001, 123, 10639. [376] R. Hahn, S. Berger, P. Schmuki, J. Solid State Electrochem. 2010, 14, 285. [377] H. Tsuchiya, J. Nakata, S. Fujimoto, S. Berger, P. Schmuki, ECS Trans. 2008, 16, 359. [378] N. K. Shrestha, Y. C. Nah, H. Tsuchiya, P. Schmuki, Chem. Commun. 2009, 15, 2008. [379] H. Tsuchiya, J. M. Macak, A. Ghicov, Y. C. Tang, S. Fujimoto, M. Niinomi, T. Noda, P. Schmuki, Electrochim. Acta 2006, 52, 94. [380] S. H. Jang, H. C. Choe, Y. M. Ko, Thin Solid Films 2009, 517, 5038. [381] P. Agarwal, I. Paramasivam, N. K. Shrestha, P. Schmuki, Chem. Asian J. 2010, 5, 66. [382] A. N. Kamkin, L. A. Fishgoit, A. D. Davydov, Russ. J. Electrochem. 2003, 39, 665. [383] Y. Song, X. Zhu, X. Wang, J. Che, Y. Du, J. Appl. Electrochem. 2001, 31, 1273. [384] Y.-Y. Song, F. Schmidt-Stein, S. Berger, P. Schmuki, Small 2010, 6, 1180. [385] A. Fujishima, X. Zhang, D. A. Tryk, Surf. Sci. Rep. 2008, 63, 515. [386] Y. Xie, Electrochim. Acta 2006, 51, 3399. [387] Y. K. Lai, L. Sun, Y. C. Chen, J. Electrochem. Soc. 2006, 153, D123. [388] J. M. Macak, M. Zlamal, J. Krysa, P. Schmuki, Small 2007, 3, 300. [389] M. Zlamal, J. M. Macak, P. Schmuki, J. Krysa, Electrochem. Commun. 2007, 9, 2822. [390] I. Paramasivam, Y.-C. Nah, C. Das, N. K. Shrestha, P. Schmuki, Chem. Eur. J. 2010, 16, 8993. [391] S. K. Mohapatra, M. Misra, V. K. Mahajan, K. S. Raja, J. Catal. 2007, 246, 362. [392] C. J. Lin, Y. T. Lu, C. H. Hsieh, Appl. Phys. Lett. 2009, 94, 113102. [393] S. Banerjee, S. K. Mahapatra, M. Misra, Chem. Commun. 2009, 7137. [394] C. Aprile, A. Corma, H. Garcia, Phys. Chem. Chem. Phys. 2008, 10, 769. [395] M. Kalbacova, J. M. Macak, F. Schmidt-Stein, C. T. Mierke, P. Schmuki, Phys. Status Solidi RRL 2008, 2, 194. [396] P. Roy, R. Lynch, P. Schmuki, Electrochem. Commun. 2009, 11, 1567. [397] P. Schmuki, L. E. Erickson, Phys. Rev. Lett. 2000, 85, 2985. [398] P. Schmuki, L. E. Erickson, D. J. Lockwood, Phys. Rev. Lett. 1998, 80, 4060. [399] M. Grtzel, J. Photochem. Photobiol. C 2003, 4, 145. [400] M. Grtzel, Inorg. Chem. 2005, 44, 6841. [401] J. M. Macak, H. Tsuchiya, A. Ghicov, P. Schmuki, Electrochem. Commun. 2005, 7, 1133. Angew. Chem. Int. Ed. 2011, 50, 2904 – 2939 [402] P. Roy, S. P. Albu, P. Schmuki, Electrochem. Commun. 2010, 12, 949. [403] J. R. Jennings, A. Ghicov, L. M. Peter, P. Schmuki, A. B. Walker, J. Am. Chem. Soc. 2008, 130, 13364. [404] J. H. Park, T.-W. Lee, M. G. Kang, Chem. Commun. 2008, 2867. [405] G. K. Mor, K. Shankar, M. Paulose, O. K. Varghese, C. A. Grimes, Nano Lett. 2006, 6, 215. [406] M. K. Nazeeruddin, A. Kay, I. Rodicio, R. Humpbry-Baker, E. Mller, P. Liska, N. Vlachopoulos, M. Grtzel, J. Am. Chem. Soc. 1993, 115, 6382. [407] T. Ohzuku, K. Sawai, T. Hirai, J. Power Sources 1987, 19, 287. [408] C. R. Granqvist, Nat. Mater. 2006, 5, 89. [409] D. C. Cronemeyer, Phys. Rev. 1959, 113, 1222. [410] A. Ghicov, H. Tsuchiya, R. Hahn, J. M. Macak, A. G. Mu oz, P. Schmuki, Electrochem. Commun. 2006, 8, 528. [411] A. Ghicov, S. P. Albu, J. M. Macak, P. Schmuki, Small 2008, 4, 1063. [412] S. Berger, A. Ghicov, Y. C. Nah, P. Schmuki, Langmuir 2009, 25, 4841. [413] C. Von Wilmowsky, S. Bauer, R. Lutz, M. Meisel, F. W. Neukam, T. Toyoshima, P. Schmuki, E. Nkenke, K. A. Schlegel, J. Biomed. Mater. Res. Part B 2009, 89, 165. [414] S. Oh, K. S. Brammer, Y. S. J. Li, D. Teng, A. J. Engler, S. Chien, S. Jin, Proc. Natl. Acad. Sci. USA 2009, 106, 2130. [415] S. Bauer, J. Park, J. Faltenbacher, S. Berger, K. von der Mark, P. Schmuki, Integr. Biol. 2009, 1, 525. [416] J. Park, S. Bauer, P. Schmuki, K. von der Mark, Nano Lett. 2009, 9, 3157. [417] S. Bauer, J. Park, K. von der Mark, P. Schmuki, Acta Biomater. 2008, 4, 1576. [418] K. von der Mark, S. Bauer, J. Park, P. Schmuki, Proc. Natl. Acad. Sci. USA 2009, 106, E60. [419] J. Park, S. Bauer, K. A. Schlegel, F. W. Neukam, K. von der Mark, P. Schmuki, Small 2009, 5, 666. [420] H. Tsuchiya, J. M. Macak, L. Mller, J. Kunze, F. Mller, P. Greil, S. Virtanen, P. Schmuki, J. Biomed. Mater. Res. Part A 2006, 77, 534. [421] A. Kodama, S. Bauer, A. Komatsu, H. Asoh, S. Ono, P. Schmuki, Acta Biomater. 2009, 5, 2322. [422] J. Kunze, L. Mller, J. M. Macak, P. Greil, P. Schmuki, F. A. Mller, Electrochim. Acta 2008, 53, 6995. [423] A. Kar, K. S. Raja, M. Misra, Surf. Coat. Technol. 2006, 201, 3723. [424] L. Peng, A. D. Mendelsohn, T. J. LaTempa, S. Yoriya, C. A. Grimes, T. A. Desai, Nano Lett. 2009, 9, 1932. [425] K. C. Popat, M. Eltgroth, T. J. LaTempa, C. A. Grimes, T. A. Desai, Small 2007, 3, 1878. [426] T. Dey, P. Roy, B. Fabry, P. Schmuki, unpublished results. [427] G. E. Aninwene II, C. Yao, T. J. Webster, Int. J. Nanomed. 2008, 3, 257. [428] I. D. Kim, A. Rothschild, B. H. Lee, D. Y. Kim, S. M. Jo, H. L. Tuller, Nano Lett. 2006, 6, 2009. [429] M.-H. Seo, M. Yuasa, T. Kida, J.-S. Huh, K. Shimanoe, N. Yamazoe, Sens. Actuators B 2009, 137, 513. [430] M.-H. Seo, M. Yuasa, T. Kida, J.-S. Huh, N. Yamazoe, K. Shimanoe, Procedia Chem. 2009, 1, 192. [431] O. K. Varghese, D. Gong, M. Paulose, K. G. Ong, C. A. Grimes, Sens. Actuators B 2003, 93, 338. [432] P. Xiao, B. B. Garcia, Q. Guo, D. Liu, G. Cao, Electrochem. Commun. 2007, 9, 2441. [433] K. Y. Chun, B. W. Park, Y. M. Sung, D. J. Kwak, Y. T. Hyun, M. W. Park, Thin Solid Films 2009, 517, 4196.  2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 2939