Titanium dioxide (titania; TiO₂) is a well-studied material that has been studied since at least 1916 [1] . Titania has a number of properties that make it useful for a wide variety of applications; these include using TiO₂ as the basis for energy efficient solar cells [2] [3] , as photocatalytic materials to clean air and water [4] [5] [6] , for self-cleaning coatings [7] , as components of various sensor devices [8] [9] [10] , and as a gate dielectric in MOSFET technologies [11] [12] [13] . Further, because it is a wide bandgap semiconductor, titanium dioxide is becoming increasingly important for many next-generation modern optical and electronics applications, such as transparent electronics systems, transparent thin-film transistors, and see-through active matrix displays. The success of each of these applications depends critically upon the particular crystallographic state (anatase, rutile, or brookite) of the titania being utilized (see, e.g., Park et al. [3] , Gÿorgy et al. [10] , Kim et al. [13] , and Luttrell et al. [5] ).

Over the last 100 years, a wide variety of methods have been developed to produce titania, each method optimized for the final form of titania required. Methodologies that are currently employed to produce bulk TiO₂ include solid state reactions [2] [4] [14] and sol-gel methods [9] [15] ; however, generally speaking, there are a wider variety of thin-film production techniques used for titania, including reactive sputtering [16] , spray pyrolysis [17] , sol-gel techniques [13] [18] , chemical vapor deposition [19] , and pulsed-laser deposition (PLD; [5] [6] [20] - [28] ).

While each of these growth techniques has its advantages and disadvantages, PLD is one of the most common, forming the basis of more than 100 publications over the last ten years. However, when searching this literature for PLD growth protocols for specific crystallographic phases, a clear pattern emerged: researchers tended to favor using a pure titanium target with a silicon substrate to grow anatase titania (see, e.g., Di Fonzo et al. [4] , Luca et al. [20] , and Gÿorgy et al. [21] ), whereas typical growth protocols for rutile thin films used rutile titania targets and either glass or silicon substrates (see, e.g., Gÿorgy et al. [21] , Dzibrou et al. [24] , and Long et al. [25] ). Kitazawa et al. [26] , Luttrell et al. [5] and Le Boulbar et al. [27] used c-cut Al₂O₃ as their substrate for growing rutile TiO₂, but in every instance in which researchers used Al₂O₃ as a substrate, it was switched for LaAlO₃ when they attempted to produce anatase TiO₂. According to Luca et al. [20] , Janisch et al. [29] , and references therein, growing TiO₂ on Al₂O₃ leads to either rutile, mixed-phase films or brookite films. For studies that considered multiple distinct crystallographic forms, researchers changed either the PLD target (see, e.g., Hsieh et al. [22] and Ohshima et al. [23] ) or the substrate (see, e.g., Luttrell et al. [5] , Kitazawa et al. [26] , and Le Boulbar et al. [27] ) to achieve their goal.

Further, there is no prior report of the production of anatase TiO₂ on Al₂O₃. We note, however, that some authors (see, e.g., Murugesan et al. [30] , Djerdj and Tonejc [31] ) call titania films “anatase” when anatase is the dominant phase in mixed-phase films. This may be done because it has been suggested that pure anatase cannot be grown on sapphire substrates (Luca et al. [20] ).

In this paper, we show that both pure anatase and rutile phases of TiO₂ can be individually grown using a single PLD target and a single substrate material (c-cut Al₂O₃) by carefully controlling the growth and annealing conditions. Understanding how to produce given phases of a material using a single PLD target and a single substrate is vital for both understanding the growth physics of the material and for large-scale manufacturing, since changing growth materials can make it difficult to determine correlations between growth conditions and the performance of the resulting film. Such ambiguity may be at least one reason why different research groups found vastly different growth parameters to be ideal for the same TiO₂ crystallographic phase (cf. Hsieh et al. [22] , Dzibrou et al. [24] , Long et al. [25] , and Choi et al. [28] ).

We grew thin films using custom-built pulsed-laser deposition system (shown in Figure 1) with a Coherent COMPex Pro KrF excimer laser (λ = 256 nm, pulse width = 25 ns) held at a 45˚ angle to a rotating TiO₂ mixed-phase target inside a custom-built growth chamber (total volume = 35 L) with a base pressure of 6.5 × 10⁻¹¹ Torr. The thin film growth process was typically started near pressure of ~2 × 10⁻⁹ Torr. The target was 1-in. in diameter, 0.25-in. thick, and 99.99% pure (SuperConductor Materials). To ensure even ablation, the target was rotated at ~3 rpm. In every case, we grew the film on a 10 mm × 10 mm c-cut (0001) Al₂O₃ substrate that was heated from back side with a platinum wire heater; temperature was measured with a thermocouple placed in a representative position, and the substrate was rotated parallel to, but in an opposite direction from the target at ~3 rpm during the entire deposition process.

Anatase films were annealed inside the growth chamber; the film was held at its growth temperature and pressure for one hour and then allowed to cool at a rate of ~9˚C/min. until it reached room temperature. Rutile films were annealed in a custom-built quartz-tube furnace (base pressure of ~1 × 10⁻⁴ Torr). To ensure that the films did not crack, they were inserted into the furnace at room temperature and returned to the temperature and pressure at which they were grown with a ramp rate of 9˚C/min; after 1 hour for anatase and 2 hours for rutile films, the films were returned to room temperature with a −9˚C/min cooling ramp rate.

Finally, film thicknesses were measured by imaging a cross section of each sample with a Zeiss Crossbeam 340 focused ion-beam/scanning-electron microscope.

The specific growth and annealing protocols required to produce each film are presented in Table 1; using these protocols, films ranging from tens of nm to up to 1.5 µm were created by changing only the number of laser pulses. Anatase films grew at 0.1Å per laser pulse (5Å per second); rutile films grew three times faster.

After production, each film was analyzed using a Rigaku Ultima 3 powder x-ray diffraction (XRD) system to inspect quality of the films in terms of crystallographic phase and crystallinity. The samples were measured from 20˚ to 85˚ (2θ) with a step size of 0.02˚ and an integration time of 0.6 s per step.

To measure the roughness, representative films were measured with an Asylum Research MFP-3D-BIO atomic force microscope (AFM) in tapping mode with Bruker MSNL probes (nominal tip radii of 2 nm) to characterize the surface morphology of each sample. Individual locations on each sample for a 10 µm × 10 µm scan were selected near the center of each sample to minimize edge effects. For each scan, the rate was set to 0.5 Hz, the scan angle was fixed at 90˚, the set point was held at 1 V, and the integral gain was set to 10.

To study the interface between the films and substrate, we used x-ray photoelectron spectroscopy (XPS). Each film was produced by the protocol described above, but changing the number of laser pulses to 200 for rutile films and 600 for anatase films in order to produce films that were ~6 nm thick. XPS measurements were completed using a Physical Electronics PHI 5000 Versa Probe spectrometer using a monochromatic Al Kα (hν = 1486.6 eV) x-ray source. Peaks

^aCorresponding film thickness: 100 nm; ^bCorresponding film thickness: 90 nm.

reported were charge corrected using the adventitious carbon 1s peak at 284.5 eV as a reference.

We show the x-ray diffraction patterns in Figure 2, in which the rutile and anatase patterns are on the top (red) and bottom (blue), respectively. The rutile pattern has been shifted vertically for visual clarity; patterns were identified with the JadeTM software package. The anatase phase matched with PDF#97-015-460431 and the rutile phase matched with PDF#03-065-1119; [32] each peak is labeled with its corresponding Miller index from the corresponding matching file.

Note that both plots are clean; there are no anatase peaks in the rutile pattern and vice versa. The peak labeled with a red dagger (†) in the top panel is a reflection from the Inconel sample holder and is not part of the film.

Since the XRD patterns of anatase and brookite are very similar, we must be cautious when claiming that we have pure anatase phase TiO₂. To aid in this, we note that in the anatase pattern in Figure 2 (the lower, blue curve), there is no peak near 30.81˚ (the location of the (121) reflection of brookite), and there is a peak at 62.67˚, corresponding to the (024) reflection of anatase. The combination of these two facts are sufficient to claim that our anatase is brookite-free (Di Paola et al. [33] and Hu et al. [34] .

Substrate effects

The main conditions required to produce strain-free epitaxial film growth are (a) a thermal match between the film and the substrate and (b) matched crystallographic lattice structures. Mismatches in either result in films that have residual stress and potential lattice defects that may alter the performance of the final film. [35] Therefore, the choice of substrate is critical when growing any kind

of thin film, and especially those with multiple crystallographic forms (e.g., TiO₂). To investigate how potential mismatches may affect the growth of anatase and rutile TiO₂, we generated Table 2, below. In Table 2, for every substrate material listed in column 1, we provide a room-temperature value of the lattice constant a in column 2, and then compute a temperature-adjusted lattice mismatch ε as:

ε = a f − a s a s , (1)

in which the subscripts f, and s stand for “film,” and “substrate,” respectively, and in which the individual lattice constants are computed at the growth temperature indicated in Table 1 for anatase (250˚C; column 3) and rutile (700˚C; column 4) TiO₂. The lattice mismatch values assume room-temperature TiO₂ lattice constants of 3.7852 and 4.5933 A, respectively. [36] Lastly, we considered that since it is increasingly difficult to find external funding, substrate cost may be a factor in decision-making, so we provide the approximate cost of each substrate in USD/cm² the last column of Table 2.

Anatase films

The anatase sample was both grown and annealed at 250˚C. This low growth temperature, combined with the significantly higher oxygen partial pressure resulted in a film that grew relatively slowly (having growth rate of 0.1 Å per pulse, or 5 Å per second). The sample is transparent, and has an rms roughness < 1 nm (for reference, the typical roughness of the c-cut Al₂O₃ substrates was measured to be 0.098 nm). The bottom panel of Figure 2 shows that the XRD pattern of this sample is clean, showing only anatase peaks; the large peak at 38.014˚ indicates that our anatase sample has a preferred orientation in the (004) plane.

It is apparent from Table 2 why many researchers producing anatase TiO₂ would select LaSrAlO₄ and SrTiO₃ as their substrates. However, these represent two of the three most expensive substrates on our list; the third best lattice

^aSubstrate costs were estimated from the MTI Corp. online catalog at mitxtl.com on 22 June 2018; ^bDobrovinskaya et al. [37] ; ^cReeber and Wang [38] ; ^dBlakemore [39] ; ^eHoward et al. [40] ; ^fKawamura et al. [41] ; ^gHössinger [42] ; ^hWatanabe et al. [43] ; ⁱAckermann and Sorrell [44] ; ^jSchmidbauer et al. [45] ; ^kde Ligny and Richet [46] ; ^lWood [47] ; ^mSpreadborough and Christian [48] .

match in Table 2 is Al₂O₃, which is available at a significantly lower cost. We also note that SrTiO₃ can be grown in the (100) orientation as an atomically flat buffer layer in which the entire surface is terminated at the TiO₂ layer, [49] [50] leading to a virtually perfect match between the buffer layer and the film. In fact, both SrTiO₃ and LaAlO₃ have been shown to be excellent substrates for both anatase and rutile TiO₂ (see, e.g., Kennedy and Stampe [51] ), but under conditions significantly different from ours (e.g., LaAlO₃ was used at temperatures above its phase change, occurring at ~820 K). Such excellent crystallographic matches under the conditions described by Kennedy and Stampe [51] also explain why SrTiO₃ is used as a substrate for TiO₂. However, these spectacular crystallographic matches come both at considerable cost and require a relatively complex growth protocols to obtain both phases of titania on the same substrate. For example, Hsieh et al. obtained both anatase and rutile TiO₂ on SrTiO₃ substrates, but while the anatase phase was deposited directly on the substrate, the rutile phase was grown by oxidizing titanium nitride films [22] .

One of the most interesting discoveries we made was determining how anatase titania films could be grown so easily on sapphire substrates given the very large lattice mismatch. We used XPS to study the interface physics between our c-cut sapphire substrate and film.

The high resolution Ti(2p) spectrum shown in Figure 3 (black curve) was deconvolved into four individual curves: the peaks labeled A (at 459.0 eV; red curve) and B (at 464.6 eV; green curve) represent the 2p3/2 and 2p1/2 energy levels of TiO₂, respectively. Similarly, the peaks labeled C (at 457.8 eV; magenta curve) and D (at 462.0 eV; blue curve) correspond to the 2p3/2 and 2p1/2 energy levels of Ti₂O₃. These results indicate that a thin layer of Ti₂O₃ was observed using XPS. Ti₂O₃ has hexagonal lattice structure with nearly identical to Al₂O₃ lattice parameters. When the upper layer of sapphire is the oxygen-saturated layer, titanium atoms can be deposited in such a way that the resulting structure is the

exact replication of Al₂O₃ structure. Therefore, a thin binding layer of Ti₂O₃ is formed on sapphire surface during anatase film deposition to allow matching the sapphire crystal to anatase structure. Further anatase TiO₂ growth becomes energetically more efficient compared to growth on sapphire, since the binding energy of direct Ti-Ti bond (117.6 kJ/mol (Luo [52] )) is lower than binding energy of direct Ti-Al bond (263.4 kJ/mol (Luo [52] )). Ti₂O₃ layer is thin enough so it is not visible in XRD scans presented in Figure 2, but has a large enough effect to mitigate the lattice mismatch in growing anatase films with virtually zero strain.

Rutile films

We confirm the results of Gouma and Mills [53] , who have shown that the rutile phase is preferred when films are grown and annealed at higher temperatures compared to those required for the growth of anatase films. Our rutile film grew at 0.3 Å per pulse (or 1.5 Å per second), appears to be clear and smooth, and has an rms roughness of <1 nm.

In the top panel of Figure 2, we can see that the XRD pattern of this sample only contains peaks for rutile phase of titania, except for one small peak near 38.5˚. This anomalous peak matches the XRD pattern of the Inconel sample holder used in these experiments, and are not part of the film. The sample has preferred orientation in (110) plane as evidenced by the very large peak at 27.432˚; this explains why one of the possible rutile peaks in our range (the (020) peak at 39.185˚) is not present.

Similar to the anatase samples (above), we were above to deconvolve a single high-resolution XPS spectrum (black curve) into three individual peaks. In Figure 4, the peaks labeled A (at 459.1 eV; red curve) and B (at 464.8 eV; green curve) represent the 2p1/2 and 2p3/2 energy levels of TiO₂, respectively. The peak labeled C (at 460.1 eV; blue curve) represents the 2p1/2 energy level of TiO.

The very small amplitude of the TiO peak in the deconvolved spectrum suggests that TiO is only present in the film as an deficiency defect at the crystalline interface.

From Table 2, it is clear that Al₂O₃ is one of the most optimal substrates for growing rutile titania, and the literature confirms that it is frequently used for this purpose (see, e.g., Luttrell et al. [5] , Kitazawa et al. [26] , and Le Boulbar et al. [27] ). We note for completeness that an intermediate layer of Ti₂O₃ is not present in the rutile sample, unlike anatase grown sample. This suggests that that the low lattice mismatch between rutile titania and c-cut sapphire doesn’t require an intermediate layer at their interface. However, a small amount of TiO is present in the rutile sample, this could be caused by growth and annealing oxygen pressure being slightly below an optimal value for rutile growth.

We have shown that the growth of pure rutile and pure anatase TiO₂ is possible without changing substrate materials or PLD targets, and we have done so using only c-cut Al₂O₃ as a substrate―despite the large lattice mismatch between Al₂O₃ and anatase phase of TiO₂. Growing anatase in this manner does, however, induce significantly more strain in the anatase film compared to the rutile film grown on the same substrate. XPS data showed that anatase growth on c-cut sapphire is likely possible, due to the specific PLD growth conditions, to accommodate growth of Ti₂O₃ intermediate layer that binds TiO₂ film with tetragonal lattice structure to hexagonal structure of Al₂O₃, although this conclusion requires additional experimental data to confirm it as the only way anatase can be grown this way. This intermediate layer is not found in our rutile films, however, oxygen vacancy defects were observed by XPS. They are possibly a result of lower oxygen pressure used during growth and annealing of rutile films. In any case, for applications in which titania is used for its optical properties, Al₂O₃ substrates may be a new preferred substrate, both because it could significantly decrease production costs and because of the various material properties of sapphire.

This research made use of the Texas Tech University College of Arts & Sciences Microscopy Center and the X-Ray Diffraction Service of the Department of Chemistry and Biochemistry. The authors would like to thank Dr. D. Unruh for his assistance with a number of aspects related to XRD. ABK would also like to thank Dr. B. Weeks for helpful discussions that improved this paper.

The authors declare no conflicts of interest regarding the publication of this paper.

Gordienko, A. and Kaye, A.B. (2018) Unique Pulsed-Laser Deposition Production of Anatase and Rutile TiO₂ on Al₂O₃. Crystal Structure Theory and Applications, 6, 19-31. https://doi.org/10.4236/csta.2018.72002