To produce TiO_x nanocondensates, Ti (99.99% pure, 1.0 mm in thickness) plate immersed in de-ionized water within a glass beaker was subjected to energetic Nd-YAG-laser (Lotis, 1064 nm in wavelength, beam mode: TEM00) pulse irradiation for up to 3000 pulses inside an ablation chamber under the laser parameters compiled in Table 1. A relatively high power density of 1.4 × 10¹¹W/cm²(i.e. pulsed energy 650 mJ/pulse; pulse duration 16 ns; beam size 0.03 mm²; fluence 2.2 kJ/cm²; frequency 10 Hz under Q-switch mode) caused a larger yield of atom clusters as a colloidal solution. The solution was then settled in capped vial for up to one week in desiccators in order to study the assembly of the resultant multiple walled tubes (MWTs).

FR free run mode, QSQ-switch mode, A amorphous phase, T1TiO, T2Ti₂O₃, T3TiO₂, NCA nano chain aggregate, MWT multiple wall tube

* Based on SAED lattice parameter of the TiO₂ nanocondensates and the Birch-Murnaghan equation of state of the rutile with relevant bulk modulus and its pressure derivative 45 (cf. text)

The optical absorbance of the as-deposited nanocondensates and further developed MWTs in solution with specified dwelling times were acquired by a UV–Vis spectrophotometer (U-3900H, Hitachi) operating at an instrumental resolution of 0.1 nm in the range of 200 to 800 nm. The powders recovered from such samples were dried for microstructure observations using optical polarized microscopy and scanning electron microscopy (SEM, JEOL JSM-6700F, 10 kV, 10μA). The crystal structure of the MWTs was determined by X-ray diffraction (XRD, Bede D1, Cu Kα, 40 kV, 30 mA, at 0.05° and 3 s per step from 2θ angle for 20° up to 100°). The d-spacings measured from XRD trace were used for least-squares refinement of the lattice parameters with an error ±0.0001 nm using bulk gold reflections as a standard.

Field emission scanning transmission electron microscopy (STEM, FEI Tecnai G2 F20 at 200 kV) with selected area electron diffraction (SAED), and point-count energy dispersive X-ray (EDX) analysis at a beam size of 1 nm was used to study the structure and composition of the nanoparticles and the tubular walls. Z-contrast images and compositional line scanning profiles are acquired by high-angle annular dark-field (HAADF) detector and EDX under STEM mode. Lattice imaging coupled with Fourier transform patterns were used to study the rolling planes of the MWT and their partial epitaxy relationship with the associated TiO_x nanoparticles. The Gatan Image Filter (GIF) coupled with electron energy loss spectrum (EELS) by TEM was employed to identify the chemical bonding state of the individual TiO_x nanoparticle and that associated with the nanotubes.

Powdery sample mixed with KBr was studied by FTIR (Bruker 66v/S) for the extent of OH⁻ signature on MWT. The powdery MWTs settled on a vitreous SiO₂ substrate were studied by micro-Raman in a backscattering geometry by a Jobin–Yvon HR800 system working in the triple-subtractive mode for the estimation of internal stress for the constituting TiO₆ and/or TiO₄ polyhedra. The same sample was also examined by X-ray photoelectron spectroscopy (XPS, JEOL JPS-9010MX, Mg Kα X-ray source) calibrated with a standard of C 1 s at 284 eV for the determination of Ti²⁺, Ti³⁺, and Ti⁴⁺ peaks. Photoluminescence (PL) spectra of the powdery samples were recorded using a Jobin–Yvon spectrophotometer (Trix 320) at an excitation wavelength of 325 nm (He-Cd laser) at room temperature.

According to the combined XRD (Fig. 1) and electron microscopic evidences, the as-fabricated nanoparticles with an average particle size of ca. 20 nm are predominantly TiO, Ti₂O₃, and TiO₂ rutile with a considerable extent of nonstoichiometry when fabricated by specified laser parameters denoted as 1, 2, 3, and 4 (cf. Table 1). The phases remained unchanged despite a much higher assembly/sedimentation rate for samples 3 and 4 than 1 and 2, the latter typically formed nano chain aggregate (NCA), upon dwelling in water for a week. The nanoparticles in all samples were more or less precipitated in the bottom of the vial in the first day of dwelling. After a week of dwelling in water, additional tubular materials with micron-scale diameter were deposited from samples 3 and 4, which were fabricated at a much higher power density than samples 1 and 2. The optical property, microstructure, and composition of the phases are shown representatively in the following.

The colloidal titanium oxide solutions as formed via PLAL in free run mode (i.e. samples 1 and 2) are sky-blue, whereas those produced by a much higher power density under Q-switch mode (i.e. samples 3 and 4) are muddy-white under the naked eye (not shown). After dwelling in water for a week (Fig. 2a), more deposits were formed from the colloidal solutions of samples 3 and 4 than samples 1 and 2. The optical absorbance near UV region for samples 1 and 2 is basically similar to that of the TiO₂ nanocondensates in aqueous or solvent-protected solution 1516. A slightly higher power density cause a higher concentration of the nanocondensates and hence a considerably higher absorbance for sample 2 than sample 1. Samples 3 and 4 prepared by a power density almost 3 orders-of-magnitude higher than samples 1 and 2 showed a significant absorbance peak near 200 nm. Sample 3 also showed an absorption shoulder similar to that of samples 1 and 2 presumably due to Ti²⁺ and Ti³⁺ ions, as discussed later. Room temperature aging of the solution for up to 1 week caused progressive accumulation of the opaque materials at the bottom and along the humidified wall of the capped bottle, and hence lowering of the absorbance peak, in particular sample 4.

Optical micrographs under open and crossed polarizers for the representative deposits retrieved from sample prepared under a relatively low (i.e. sample 2) and 3 orders-of-magnitude higher power density (i.e. sample 4) showed that the titanium oxides tended to assemble as particles in the former case (Fig. 2b) but tubes in the latter case (Fig. 2c).

The corresponding SEM observations (Fig. 3) showed further the morphology and size difference between samples 2 and 4. In sample 2, the equiaxed titanium oxide nanoparticles with a mean particle size of ~20 nm were assembled as NCA or in a closely packed manner (Fig. 3a–3c). By contrast, the titanium oxide microtubes in sample 4 have a mean diameter as large as 2–3 μm and were entangled or bifurcated leaving necks at the junctions (Fig. 3d–3f). In general, the tubes/pipes in samples 3 and 4, i.e. assembled from the nanocondensates fabricated under a relatively high power density, are nearly perfect, unfolded and hierarchical branching, as compiled in Appendix 1.

The compositions of the individual nanoparticles in the less laser power-density activated case, as represented by sample 2, were characterized by STEM to be close to TiO₂, Ti₅O₉, and Ti₂O₃, as compiled in Fig. 4. TEM BFI and lattice image coupled with Fourier transform patterns and point-count EDX analysis of sample 2 further identified the equiaxed nanoparticles of Ti₂O₃ and TiO₂ rutile with {101} twin planes in Fig. 5. The co-existing TiO phase showed corrugated {100} facets and point defects, presumably Ti and/or oxygen vacancies, and occasionally formed bicrystals with (100) tilt boundary due to coalescence over (100) vicinal surfaces (Fig. 6). Dense dioxide nanoparticles, such as α-PbO₂-type and fluorite-derived type TiO₂ the same as that reported by ref. 1718, were occasionally found (not shown). The tubular material assembled from highly activated nanocondensates have corrugated lamellar wall with 0.386 nm interspacing on the average and more or less attached with crystalline nonstoichiometric titania nanoparticles (Fig. 7). The lamellar interspacing is ca. half that of the basal layer (200) interspacing (0.786 nm) of H₂Ti₃O₇4 implying the former is a disordered precursor of the latter.

The background-subtracted Ti L-edge and O K-edge EELS spectra of titanium oxide microtube (sample 4), and the nanocondensates of TiO₂, Ti₂O₃, and TiO (sample 2) retrieved from the colloidal solutions as compiled in Fig. 8, showed that the titanium ions are 4+ , 3+, and 2+ in valence as for TiO₂, Ti₂O₃, and TiO, respectively 19. The two major L₃ and L₂ edges can be attributed to the spin orbit splitting of the 2p core hole for a separation by about 5 eV having the two edges subdivided by the strong crystal-field splitting of Ti⁴⁺ from the surrounding oxygen atoms in view of the EELS study for anatase, rutile and titania-based nanotubes as well as the calculated spectra using crystal-field multiplet code for Ti⁴⁺20. The EELS O–K edge exhibited a strong peak due to the Ti 3d and O 2p hybridization which is spitted at 2.5 eV for anatase and 2.75 eV for rutile 20. By contrast, the O–K edge splitting of the present samples is minor for TiO and vague for Ti₂O₃ and microtubes (Fig. 8). Apparently, the Ti L-edge and O K-edge EELS spectra of the titanium oxide microtube are quite different from those of TiO₂ rutile but are similar to those of Ti₂O₃, indicating titanium ions are predominantly in 3+ oxidation state for the microtubes.

The valence state of the nonstoichiometric titania in samples 2 and 4 were further confirmed by XPS Ti 2p_3/2 and O 1 s peaks coupled with Gaussian fits (Fig. 9) to be predominant 3+ as opposed to 4+ and 2+ . Actually, the binding energies around 455.9, 456.7, and 458.5 eV could be identified as TiO, Ti₂O₃, and TiO₂ species. The decomposition of the O 1 s spectrum revealed the binding energies of the individual components to be 530.08 (Ti⁴⁺–O), 531.1 (Ti³⁺–O), 528.3 (Ti²⁺–O), and 532.36 eV (OH⁻), in agreement with previous work on various titanium oxide phases 212223.

The vibrational spectra shed light on the extent of OH⁻ signature and the internal stress of the constituent polyhedra of the nonstoichiometric titania. The FTIR spectra of the titanium oxides nanoparticles (sample 2) and microtubes (sample 4) in Fig. 10a showed the latter has much stronger band between 400 and 700 cm⁻¹ which can be assigned as Ti–O stretching and Ti–O–Ti bridging stretching after 24 and stronger band near 1057 cm⁻¹ which can be attributed to the TiO molecules vibration 25. The weak bands near 2850 and 2920 cm⁻¹ were from EtOH contamination during IR sample preparation for both cases. The extent of Ti³⁺–O stretching below 700 cm⁻¹26 cannot be differentiated because of the common broad band around 600 cm⁻¹ for the present two cases.

The OH⁻ signature of the nanoparticles and microtubes were testified by the IR band near 3430 cm⁻¹ analogous to the fully hydroxylated rutile surface having three distinct absorption bands around 3655, 3530, and 3400 cm⁻¹27. The hydroxyl groups were likely associated with a number of different surface sites of the nonstoichiometric titania in the present case to cause various interactions, such as hydrogen bonding between surface OH groups and molecular water with bending vibration near 1640 and 1726 cm⁻¹ after the assignment of ref. 27, and hence a broad IR band analogous to the case of hydroxyl groups linkage to TiO₂ rutile and anatase 28. The microtubes were much more OH-signified than the nanoparticles upon dwelling in water for a week. The combined results of XRD and Raman shifts indicated that the protonated microtube is structurally different from titanium hydrate, such as H₂Ti₃O₇67.

The Raman shifts of the nanoparticles in sample 2 are mainly 614 (A_1g), 443 (E_g), 248 (due to second order phonon), and 144 cm⁻¹(B_1g) with the assigned modes of TiO₆ octahedra in parenthesis (Fig. 10b) after the assignment of rutile by ref. 29 and 30. These Raman shifts change to 606, 416, 264, and 414 cm⁻¹, respectively, for the microtubes in sample 4. This indicates that the extent of TiO₆ distortion in terms of internal stress is different, according to pressure dependence of TiO₆ Raman shifts as discussed in Sect. 4. The Ti ion charges/oxygen vacancies are also considerably different in the two cases to change the vibration/stretching behavior as indicated by the presence of Ti₂O₃ with characteristic band at 344 cm⁻¹ due to Ti³⁺ occupancy in octahedral site 3132 for sample 2 but not for sample 4, although monoxide TiO is not Raman active 33 to support this point.

The PL spectra of samples 2 and 4 (Fig. 11) showed that the luminescence of the present nonstoichiometric titania nanoparticles and microtubes covers a rather broad wavelength from 400 to 700 nm with individual peaks much broader than the case of alkaline stabilized titanate nanotubes with TiO₆ octahedral units 34. The PL bands at 2.84 and 2.74 eV were probably originated from the self-trapped excitons localized on the TiO₆ octahedra 35. These types of trapped sites were found below the conduction band edge due to the conduction band splitting 35. The bands at 2.53 and 2.02 eV were attributed to oxygen vacancies 22. The 2.53 eV vacancy level, which corresponds to 1.05 eV below conduction band for a band gap of 3.6 eV, has been attributed to Ti⁴⁺ ions adjacent to oxygen vacancies by the intragap surface states. This deep electron trap has been confirmed using femtosecond photogenerated charge dynamics in TiO₂ nanoclusters 36. The band at 2.47 eV could be attributed to the TiO₆ octahedra as the basic structural unit analogous to the titanate nanotube 34. The band at 2.31 eV is assigned to Ti₂O₃37. The band at 2.16 eV can be assigned to the shallow trap level related to oxygen vacancies on the surface of TiO₂ nanotube. The PL emission identified with oxygen vacancies might have occurred with photogenerated conduction band electrons trapped by ionized oxygen vacancy levels in TiO₂ nanotube and subsequently recombined with the holes in the valence band 35.

Based on the above vibration and PL spectroscopic analyses, the TiO₆ octahedral are likely the common units for the present titania nanoparticle/microtubes and the crystalline titanium hydrate nanotubes, such as H₂Ti₃O₇ with long range ordering in the basal layer 67 or other stoichiometries with controversial structures 8910.

The PLAL by free run with a relatively low power density caused more TiO, Ti₂O₃, and TiO₂ than the amorphous phase (cf. Table 1). By contrast, Q-switch mode with a higher power density caused more oxidized crystallites, i.e. TiO₂ rather than TiO and Ti₂O₃ as in the case of sample 4, besides the predominant amorphous lamellar phase which further assembled and rolled up as microtubes upon dwelling in water. Apparently, PLAL under a higher power density would cause more thorough oxidation of the nanocondensates on the one hand and the amorphous phase on the other hand. In fact, by rapid reactive quenching with water in the liquid-plasma interface, the ablated species can be readily oxidized 38. As for phase amorphization by a dynamic condensation and very rapid cooling process, it has been demonstrated by the PLA synthesis of amorphous Al₂O₃ nanocondensates 39. The cooling rate in the similar PLA process was estimated to be close to 10⁹ K/s for 10-nm-sized Al₂O₃39 as well as α-PbO₂ type TiO₂ nanocondensates 17, 4 orders of magnitude higher than that required to quench an amorphous state for oxides 40. The PLAL process is expected to have an even higher cooling rate under the influence of water rather than air cooling, to quench the present titania nanocondensates as amorphous state under an additional factor of a rather large extent of nonstoichiometrty and associated defect clusters as addressed in next section.

The application of pressure to certain crystalline materials, i.e. so-called pressure induced amorphization, can cause them to become amorphous under certain conditions 41. An anatase-amorphous transition regime was also reported to occur for TiO₂ of very fine crystallite size upon static compression at room temperature using the diamond anvil cell technique 42. Shock-wave loading and liquid confinement typical for a PLAL process 43 would also cause the already crystallized nonstoichiometric titania to become amorphous. In this connection, PLAL of TiO₂ single crystal and Ti plate targets under the wavelength of 355 nm and maximum laser pulse energy 160 mJ/pulse in de-ionized water was reported to cause mostly amorphous phase 38. The amorphous lamellae appeared to be formed by a relatively high power density in the present PLAL process and then assembled and rolled up as microtubes upon subsequent dwelling in water. It is not clear, however, whether the amorphous lamellae are in high-density amorphous state 44.

The TiO₂ rutile nanoparticles in samples 1 to 4 have a significant internal stress up to 4–5 GPa for the lattice (Table 1) based on SAED lattice parameters and the Birch-Murnaghan equation of state of the rutile with relevant bulk modulus and its pressure derivative 45.

The Raman shifts provide another estimation of the internal stress in terms of the TiO₆ polyhedra shared by the TiO, Ti₂O₃, and TiO₂ and amorphous lamellar phase in the samples. Regarding this approach, the pressure dependence of Raman shifts has been studied for anatase 42 and rutile-type TiO₂46. The major Raman modes (E_g(1), B_1g(1), A_1g + B_1g, E_g(3)) of anatase nanocrystals are all well represented by linear increases in Raman shift with pressure up to 41 GPa 42. (The Raman modes E_g and A_1g of the synthetic rutile-type TiO₂ single crystal also have a higher wave number under an applied pressure up to 35 GPa 46). Using both calibration curves, the TiO₆ polyhedra of the present TiO_{1 − X}, Ti₂O_{3 − X}, and TiO_{2 − X} and amorphous lamellar have quite different stress state from that of the rutile lattice based on its lattice parameters. The internal compressive stress for the TiO₆ polyhedra of the TiO, Ti₂O₃, and TiO₂ nanocondensates in sample 2 is indicated by its A_1g mode but not E_g mode being at a higher wave number than the reported ambient value of anatase (513 cm⁻¹) 40 or rutile (608 cm⁻¹) 29. In other words, there is a significant distortion of the polyhedra in the present nonstoichiometric titania phases due to the combined effects of quenching a dense state and varied stoichiometry and protonation. The TiO₆ polyhedra of the microtubes in sample 4 were quite relaxed based on the A_1g and E_g modes being at a much lower wave number (606, 416 cm⁻¹) than the titanium oxide nanocondensates in sample 2 (614, 443 cm⁻¹). This indicates that the TiO₆ polyhedra in the amorphous lamellae were relaxed when the lamellae were rolled into microtubes. It should be noted that size, nonstoichiometry, and temperature would also affect the vibration mechanism of the TiO₆ polyhedra, as indicated by previous Raman shifts study of rutile 52930.

The present nanocondensates have a rather high Ti³⁺/Ti²⁺ and Ti³⁺/Ti⁴⁺ atomic ratio and OH⁻ signature according to XPS and FTIR results, respectively. In other words, the Ti³⁺ and the dopant Ti²⁺, Ti⁴⁺, and H⁺ ion species were co-incorporated in the TiO, Ti₂O₃, and TiO₂ and amorphous lamellar when fabricated by PLAL. Under such a complicated case, charge-compensating defect clusters [V_Ti″′ + Ti_Ti^•] in association with the interstitial proton in the predominant Ti₂O₃ as indicated by XPS would occur through the following equation in Kröger-Vink notation 47:

Here, H_i^• signifies single positively charged hydrogen in the interstitial octahedral and/or tetrahedral sites; Ti_Ti^• dominating single positively charged titanium at titanium sites and V_Ti″′ triple negatively charged titanium vacancies in the crystal lattice. In such a defect chemistry scheme, the Ti⁴⁺(effective ionic radii, 0.0605 nm) would replace a larger sized Ti³⁺(0.067 nm) in CN 6 48, for the volume compensation of interstitial protons. It is also possible that the volume-compensating effect, due to the undersized Ti⁴⁺ dopant in the Ti³⁺ site, forced Ti_Ti^• to enter the interstitial tetrahedral site, i.e. Ti_i^…., and hence more charge-compensating cation vacancies. Alternatively, Ti²⁺ may act as dopant by substituting Ti³⁺ and hence charge compensates the interstitial protons through the following equation:

The oxygen vacancies and titanium interstitials were suggested to be the dominant defects in TiO₂ rutile under low oxygen atmosphere 49 or doped with cation with lower valence such as Zn²⁺50. The titanium vacancies are however favored in the present Ti₂O₃ to charge compensate the proton dopant. In fact, Ti₂O₃ allows a considerable extent of nonstoichiometry with O/Ti ratio ranging from 1.48 to 1.51 in the Ti–O binary at temperatures 49. The incorporation of proton during PLAL is expected to cause an even higher O/Ti ratio by introducing more titanium vacancies through Eq. 1.

The defect clustering of Ti²⁺, Ti⁴⁺, and H⁺ co-doped Ti₂O₃ through Eqs. 1 and 2 may be effective as soon as the oxolation by forming Ti–O–H linkage, typical in a hydrothermal process 51, caused considerable polymerization and various defect clusters in the crystals as observed by TEM. The hydrolysis and polymerization process to form specific Ti–O–H species is quite effective so that the sedimentation of the nanoparticles is much faster than Au nanoparticles via the same PLAL process 12. Aside from the detailed defect clustering scheme, the co-existence of Ti²⁺, Ti³⁺, Ti⁴⁺, and H⁺ ions and charge/volume-compensating defects in the present nonstoichiometric titanium oxides would account for the Raman shift of rutile Eg band in view of the Ti/O ratio dependence of Raman shift for nanophase TiO₂ rutile and anatase 52.

The formation of tubes/pipes in samples 3 and 4 versus NCA or closely packed manner in samples 1 and 2 can be attributed to different growth mechanisms. The former case is analogous to the Au clusters being assembled as lamellae and then rolled up as Au micro- and nano-tubes 12. To roll up a planar layer, it requires either caps as for carbon nanotubes with single or multiple walls 53 or intrinsic structural anisotropy as for clay minerals chrysotile 54. Only intrinsic structural anisotropy is considered here for the protonated microtubes. We suggest that the amorphous titania lamellae stabilized via oxolation in the PLAL process were water/air interface-mediated at the water level and bubbles to become asymmetric on the opposite sides, i.e. more H⁺ signified on the water side to exert a strain field for rolling up upon dwelling in water.

The H₂Ti₃O₇ nanotube with monoclinic structure (space group C2₁/m) based on ab initio calculations and TEM examinations has been synthesized by treating TiO₂ powders in concentrated NaOH in aqueous solution and then individual trititanate layers are peeled off from the plates and scroll up into nanotubes 6755. It was found that surface tension due to an asymmetry related to H deficiency in the surface layers of the plates is the principle driving force of the cleavage, and the dimension of the nanotubes is controlled by this surface tension together with interlayer coupling energy and Coulomb force 55. However, H₂Ti₃O₇ was not formed in samples 2 and 4 based on the combined XRD, FTIR, and Raman results. The reasons of forming protonated amorphous titanian microtubes with multiple lamellar wall rather than H₂Ti₃O₇ nanotube in the present PLAL process could be the following. First, the present samples were synthesized under water rather than high concentration of Na⁺ cations and strongly basic conditions for alkali-treated processes that are essential in the synthesis of the titanium hydrate nanotubes 6. Second, dynamic high pressure–temperature conditions and very rapid heating/cooling rate via the present PLAL process may favor amorphous lamellae rather than H₂Ti₃O₇ which typically has corrugated ribbons of edge-shared TiO₆ octahedrons via a thermal equilibrium synthesis route 7.