II. TiO₂ NR GROWN BY SPCVD TECHNIQUE

SPCVD can be considered as a derivative of the atomic layer deposition (ALD) technique that has been widely applied to the growth of conformal thin films with precisely controlled thickness down to the subnanometer level.Reference George³⁰ The SPCVD reaction system is schematically shown in Fig. 1. The reaction chamber was a stainless steel flow type tube with a diameter of 35 mm, which was located inside a tube furnace. Two precursors were connected to two separated 1/8″ tubes through solenoid valves and extended to the center of the growth chamber. Substrate was located at the center and ∼5 cm away from the precursor injection nozzle. A constant flow of 40 sccm N₂ was introduced into the chamber as the carrier gas and could be turned on/off by valve 3. The growth chamber was continuously pumped by a mechanical pump to a base pressure of ∼300–400 mTorr, and the connection was controlled by valve 4. Both valves 3 and 4 were always on unless a dwelling cycle was applied. The chamber temperature was maintained at 600 °C. TiCl₄ and H₂O vapor precursors were pulsed into the chamber for 1.5 s each and separated by N₂ purging for 60 s. One complete growth cycle included 1.5 s of H₂O pulse + 60 s of N₂ purging + 1.5 s of TiCl₄ pulsing + 60 s of N₂ purging. Because of the similar cyclewise operation process as ALD, SPCVD inherited the unique conformal coating merit of ALD and is able to achieve conformal growth of TiO₂NRs arrays inside highly confined spaces, such as anodic aluminum oxide (AAO) nanochannels. Formation of NR morphology was believed to be due to the high deposition temperature and elongated purging time.

FIG. 1.

Schematic SPCVD growth system.

The as-grown TiO₂ NR-coated AAO channel is show in Fig. 2. Figure 2(a) shows the cross section of the AAO template after 660 cycles of SPCVD growth. Scanning electron microscope (SEM) images acquired from the top, middle, and bottom sections are shown in Figs. 2(b)–2(d), respectively, revealing the uniform and dense coverage of TiO₂ NRs along the entire channel length. Closer examination shows that all NRs were rooted on the walls of the AAO channels and pointed inward [Fig. 2(e)]. The majority of the as-synthesized TiO₂ NRs had lengths and widths of between 170–210 and 25–30 nm, respectively. Transmission electron microscopy (TEM) and x-ray diffraction spectrum revealed that the NRs were single-crystalline anatase TiO₂.

FIG. 2.

(a) A cross section of the AAO template after 660 cycles of SPCVD growth. (b–d) The uniform and dense coating of TiO₂ NWs inside the AAO channels at the top, middle, and bottom sections, respectively, as indicated by the dashed yellow boxes in (a). (e) Higher magnification SEM of TiO₂NWs rooted on the walls of AAO channels.

The evolution of TiO₂ NRs inside AAO channels was first investigated by observing the morphologies after growth cycles of 85, 170, 330, 660, 900, and 1200, respectively [Fig. 3(a)]. Statistical analysis revealed a fairly uniform NR geometry distribution inside AAO channels [Fig. 3(b)]. This growth behavior was similar to the unique conformal coating feature of ALD. Length and width measurements revealed a faster growth rate in the first several hundred cycles [Fig. 3(c)]. Before the 330-cycle growth point, the average growth rate along the length was ∼0.5 nm/cycle, while subsequent rate dropped to ∼0.1 nm/cycle. Growth rate along the width direction was ∼0.03 nm/cycle. The slower growth rate was possibly the consequence of less volume left in the AAO channel for precursor transportation when the sizes of NRs became larger. The length of NRs saturated at 230–240 nm, which correlated with the average diameter of the AAO channels (∼250 nm).

FIG. 3.

(a) SEM images of typical NW morphology after different SPCVD cycles. (b) Length distribution analysis of the NWs grown by different number of SPCVD cycles. (c) Plots of NW length and thickness versus growth cycles.

A typical TiO₂ NR morphology observed by TEM is shown in Fig. 4(a), where the NR exhibited a highly uniform thickness, well-faceted surfaces, and an aspect ratio of ∼7. Uniform contrast was observed on the corresponding dark field image [Fig. 4(b)], indicating the high lattice quality and a dislocation-free structure. No foreign elements, such as Cr that may be released from the stainless steel reactor,Reference Danon, Bhattacharyya, Vijayan, Lu, Sauter, Gray, Stair and Weitz³¹ was detected, and the as-synthesized TiO₂ NRs exhibited a sharp cutoff absorption wave length at ∼400 nm.Reference Shi, Hara, Sun, Anderson and Wang³² Selected area electron diffraction (SAED) taken on the NR further confirmed the anatase structure [inset of Fig. 4(c)] with the {100} as side surfaces and the {011} as the surfaces in-plane of the image.Reference Zhang, Li, Jin, Zhang, Zhang and Zhang¹⁷ Combination of the [002] and $[0\bar 11]$ directions was identified as the NR growth direction. High-resolution transmission electron microscopy (HRTEM) clearly revealed the arrangement of the (002) and $(0\bar 11)$ planes along the NR growth direction with a lattice spacing of 0.48 and 0.35 nm, respectively [Fig. 4(c)]. Based on the SEM, SAED, and TEM observations, structure of the anatase NRs is constructed and illustrated in Fig. 4(d). The four side planes of the NR are (011), $(0\bar 1\bar 1)$, (100), and $(\bar 100)$. They are perpendicular to each other and form a rectangular cross section.

FIG. 4.

(a, b) Bright field and dark field TEM images of a single-crystalline anatase TiO₂ NR demonstrating high thickness/width uniformity and excellent crystal integrity. (c) HRTEM image of the TiO₂ NR. Inset is the corresponding SAED pattern. (d) Proposed TiO₂ NR structure shaped by three groups of atomic planes: ⁶, ⁶, and ⁶. A unit cell is represented by a cube with a, b, and c parallel to the [100], [010], and [001] directions, respectively. (e) HRTEM image of a typical NW tip faceted by the (001), (011), $(0\bar 1\bar 1)$, and $(0\bar 11)$planes.

III. 1D MORPHOLOGY FORMATION MECHANISM

Understanding the 1D anisotropic growth mechanism by SPCVD is essential for achieving morphology control and applying this capable nanomaterial synthesis technique to different material systems. The key is to understand the evolution of 1D morphology from a textured polycrystalline thin film by ALD. Early research on ALD showed that crystalline TiO₂ thin films could be grown at 600 °C using TiCl₄ and H₂O as precursors.Reference Ritala, Leskela, Nykanen, Soininen and Niinisto³³ The differences for achieving NR growth were only the elongated purging time (60 s) and highly confined submicrometer-sized spaces.

The surface restructuring and surface (or gas phase) diffusion mechanisms that were responsible for the surface roughness of ALD films were adapted to explain the evolution of NR morphology. Ideally, TiCl₄ and H₂O are anchored onto the growing surface by reactions:

(1)$${\rm{-OH(surf}}{\rm{.)}} + {\rm{TiCl}}_{\rm{4}} {\rm{(g)}} \to {\rm{-\!O\!-\!\!TiCl}}_{\rm{3}} {\rm{(surf}}{\rm{.)}} + {\rm{HCl(g),}}$$

(2)$${\rm{-O\!-TiCl}}_{\rm{3}} {\rm{(surf}}{\rm{.)}} + x{\rm{H}}_{\rm{2}} {\rm{O(g)}} \to {\rm{-O\!\!-\!\!Ti\!-\!\!(OH)}}_x {\rm{Cl}}_{{\rm{3}} - x} {\rm{(surf}}{\rm{.)}} +\,\, x{\rm{HCl(g) (}}x = 1,2,3{\rm{)}}{\rm{.}}$$

Meanwhile, the surface –OH and –Cl groups can also react with each other forming surface bridging groups:

(3)$${\rm{\!Ti\!\!-\!\!OH(surf}}{\rm{.)}} + {\rm{Ti\!\!-\!\!OH(surf}}{\rm{.)}} \to {\rm{Ti\!\!-\!\!O\!\!-\!\!\!Ti(surf}}{\rm{.)}} + {\rm{H}}_{\rm{2}} {\rm{O(g),}}$$

(4)$${\rm{Ti\!\!-\!\!OH(surf}}{\rm{.)}} + {\rm{Ti\!\!-\!\!Cl(surf}}{\rm{.)}} \to {\rm{Ti\!\!-\!\!O\!\!-\!\!Ti(surf}}{\rm{.)}} + {\rm{HCl(g)}}{\rm{.}}$$

The Ti–O–Ti bridging groups are acidic due to cation polarization. They could also reversibly react with TiCl₄ with a much slower rate compared to the –OH groups:

(5)$${\rm{Ti\!-\!O\!-\!Ti(surf}}{\rm{.)}} + {\rm{TiCl}}_{\rm{4}} {\rm{(g)}} \leftrightarrow {\rm{Ti\!-\!O\!\!-\!\!TiCl}}_{\rm{3}} {\rm{(surf}}{\rm{.)}} + {\rm{Ti\!-\!Cl(surf}}{\rm{.)}}{\rm{.}}$$

Reactions (3)–(5) are responsible for the reduced growth rate in ALD. Furthermore, HCl by-products also could make important contributions to the growth, particularly at high temperature, if they are not quickly removed from the growing surfaces:

(6)$${\rm{Ti\!\!-\!\!OH(surf}}{\rm{.)}} + {\rm{HCl(g)}} \leftrightarrow {\rm{Ti\!\!-\!\!Cl(surf}}{\rm{.)}} + {\rm{H}}_{\rm{2}} {\rm{O(g),}}$$

(7)$${\rm{Ti\!\!-\!\!O\!\!-\!\!\!Ti(surf}}{\rm{.)}} + {\rm{HCl(g)}} \leftrightarrow {\rm{Ti\!\!-\!\!OH(surf}}{\rm{.)}} + {\rm{Ti\!\!-\!\!Cl(surf}}{\rm{.)}}{\rm{.}}$$

Equations (6) and (7) are the reversed reactions of (2) and (4), respectively. Thus, reaction (6) is a backward reaction and would reduce the growth rate, while reaction (7) could reactivate the bridging groups and facilitate the growth. Different combination of reactions (1)–(7), together with the diffusion of HCl by-product, could induce different growth rate along different crystal surfaces and thus faceted anisotropic growth might be possible.

TEM observations (Fig. 4) revealed that the NRs were grown along the combination of the [002] and [011] directions and the {011} and {100} planes were the side surfaces. Two possible mechanisms were suggested—surface recombination and HCl restructuring.

IV. PHASE AND MORPHOLOGY CONTROL

At 600 °C, SPCVD can grow highly uniform anatase TiO₂ NRs over a large area, even inside highly confined submicrometer-sized spaces. By increasing the deposition temperature to 650 °C with other deposition conditions unchanged, the phase of NRs was turned into rutile.Reference Shi and Wang³⁴ After 1000 growth cycles, short NRs were resulted covering the entire Si substrate surface [Fig. 5(a)]. The NRs were typically a few hundred nanometers long and tens of nanometer wide. Most of them had a rectangular cross section [inset of Fig. 5(a)]. TEM image showed that the TiO₂ NRs had a typical sword-like shape [Fig. 5(b)].Reference Barnard and Curtiss¹⁵ The single-crystalline rutile structure was confirmed by the SAED pattern as shown in the inset of Fig. 5(b). Sharp surfaces and dislocation-free lattice of the rutile TiO₂ NRs were identified by the HRTEM image shown in Fig. 5(c). The rutile NRs were grown along the [001] direction and surrounded by four equivalent {110} surfaces. This experiment showed that simply increasing the deposition temperature could change the phase of TiO₂NRs from anatase to rutile but had negligible influence on the NR morphology.

FIG. 5.

(a) SEM image of rutile TiO₂ NRs grown on Si substrate. Inset is the magnified image of the TiO₂ NRs. (b) TEM image of a TiO₂NR. Inset is the corresponding SAED pattern confirming the rutile phase. (c) HRTEM image of the NR acquired from the rectangle region in (b).

Typically, growth of NWs can be facilitated by introducing noble metal catalysts, such as Au, to expedite the nucleation and/or deposition. TiO₂ NR morphology evolution was thus studied with the presence of Au catalysts by sputtering ∼5-nm-thick Au thin film on a silicon substrate. The deposition was conducted at 650 °C and a completely different morphology was resulted. Figure 6(a) shows a low magnification SEM image of the as-grown TiO₂ NWs on silicon substrate. The entire substrate was covered by a large quantity of NWs, where short-range alignments of the NWs can be observed. The aligned NW region typically consisted of a few rugged strips that were parallel to each other. Both the distance between neighboring strips and the width of each strip were around a few hundred nanometers. This local alignment phenomenon indicated that the parallel strips likely originated from a crystallized film underneath. Tilted image of the NW sample revealed that the aligned arrays were actually packed TiO₂ flakes with spaced NWs grown on the topside, where several NWs with rough surfaces and nonuniform widths could be clearly seen growing along the same orientation [inset of Fig. 6(a)]. These NWs shared the same thickness as the base flakes, while their widths decreased from the bottom (∼100 nm) to the tip (∼20–30 nm). The lengths of those NWs were typically 1–2 μm.

FIG. 6.

(a) SEM image of TiO₂ nanostructures grown on Au-coated Si substrate. The quasialigned nanostructure arrays are pointed by white arrows. Inset is a tilted image showing the NWs-on-flake structure. (b) TEM image of a single piece of the NWs-on-flake structure. Inset is the SAED pattern confirming the single-crystalline rutie phase. (c) HRTEM images acquired from NW.

TEM images shown in Fig. 6(b) clearly revealed the NWs-on-flake structure and the curvy side surfaces of these NWs. This configuration suggested that the NWs were most likely formed by bifurcations of the flake matrix. Each bifurcation produced two branches as the newly formed small flakes or NWs. The branching took place at different vertical positions along the flakes until the width of the new flake became smaller than 100 nm, and thus, the NW morphology was resulted. The single-crystalline structure was confirmed by the SAED pattern shown in the inset of Fig. 6(b). The flat flake surface was the (110) facet, and the NWs were grown along the $[1\bar 10]$ direction. HRTEM image revealed the perfect crystal structure of the NWs, as shown in Fig. 6(c). The uniform lattice pattern and contrast indicated that the NWs had a very uniform thickness. The lattice spacing measured from Fig. 6(c) and the lattice constants calculated from the SAED pattern further confirmed the rutile phase of the NWs. This high-quality lattice structure was found consistent over the entire flake–NW structure except the bifurcation regions.

Different from most other single crystal NW morphology made by either hydrothermal or vapor deposition approaches, which usually consisted of low index flat surfaces, the as-synthesized rutile TiO₂ NWs showed curvy side surfaces,Reference Zhang, Li, Jin, Zhang, Zhang and Zhang^17, Reference Wu, Shih, Wu, Tseng and Chen²⁴ whereas no dislocations were observed, as shown in Fig. 6(c). The uniform contrast also indicated that the thickness along the curved edge was still uniform. Such curved side surfaces consist of high index crystal facets and thus renders much higher surface energy compared to the low index facets, which, as a result, might enhance the catalytic activity of the TiO₂NWs.

Further control experiments revealed that longer purging time only formed a quasicontinuous rutile TiO₂ film when catalyzed by the same Au thin film. If no Au thin film was present and a short purging time (10 s) was used at 650 °C, only rutile TiO₂ nanoparticles were observed, which was consistent with the experimental results conducted at 600 °C. From the control experiments, it was feasible to conclude that a thin layer of Au could facilitate the growth of crystalline TiO₂ films during a SPCVD process. If the precursor was sufficient during the deposition, the films could evolve into bifurcated NWs. If the precursor supply was limited, the film-like morphology would not change. However, currently, it is still unclear about the exact role of Au during the growth. No Au was observed from the NWs by TEM after growth, suggesting that the Au did not likely act as the catalyst for a VLS growth. It might be possible that Au wetted Si at high temperature and created a large number of favorable sites on the surface for absorbing incoming molecules, thus facilitating 2D lateral growth.

The deposition condition related TiO₂ nanostructure growth is summarized in Table I to illustrate their relationships. Comparing the first two conditions clearly shows that higher deposition temperature was the main reason for the formation of rutile phase since it is thermodynamically more stable than anatase phase at higher temperature.

TABLE I.

Summary of the growth condition related TiO₂ nanomorphologies.

V. PHOTOCATALYTIC PROPERTY OF TiO₂ NW GROWN BY SPCVD

The unique capability of SPCVD for growing dense TiO₂ NRs inside highly confined spaces allowed the creation of tree-like three-dimensional (3D) NW architectures that were ideal for high-performance PEC electrodes. The 3D NW architecture is schematically shown in Fig. 7(a). Vertically aligned Si NW arrays were used as the backbones, and high density TiO₂ NRs were uniformly grown around the Si NWs. An additional thin film of anatase TiO₂ was coated on the NR–NW heterostructure by ALD to completely cover Si NW surfaces. When this heterogeneous structure was used as a PEC anode for water splitting, photocatalyzed H₂O oxidation reactions would only occur on the TiO₂ surfaces. In principle, the photogenerated electrons should quickly be transported through the Si NW backbone and reach the counter electrode with minimal losses owing to the high conductivity of Si NWs and isolation from electrolyte.Reference Hwang, Boukai and Yang^5, Reference Takabayashi, Nakamura and Nakato³⁵ The very large surface area provided by the TiO₂NR arrays together with the high-speed electron transport channels provided by the Si NWs are promising features that could lead to a significant improvement of PEC efficiency.

FIG. 7.

(a) Schematic tree-like TiO₂ NR–Si NW 3D architectures. (b) Cross section of vertical Si NW arrays covered with TiO₂ NRs. (c, d) The middle and bottom portions of Si NWs showing dense and uniform coverage of TiO₂NRs along the entire NW length.

Vertical Si NW arrays made from dry reactive ion etching and metal-assisted wet etching were both successfully applied for TiO₂ NR growth. For example, Fig. 7(b) shows the cross section of wet-etched Si NW arrays that were ∼25 μm in length. SEM images acquired along the length direction of the Si NWs forest confirmed the dense and uniform coverage of TiO₂ NRs. These TiO₂ NRs are 30 ± 6 nm in diameter and 243 ± 31 nm in length statistically. Figures 7(c) and 7(d) are two representative images indicating that the density and dimensions of TiO₂ NRs were nearly identical at the middle and bottom regions. The NRs grew out laterally from the Si NWs and filled out the interspaces. Thus, considerably large surface area and high porosity were created by such a 3D architecture. Given that the density of TiO₂ NR branches measured to be ∼20 per 1 μm Si NW length, and the average diameter and density of Si NWs at ∼100 nm and ∼10 NWs per square micrometers, respectively, the roughness factor of the TiO₂NR–Si NW heterostructure was calculated to be ∼8 per micrometer in thickness (compared to ∼3 per micrometer for bare Si NW arrays). It should be noted that this number is a rough estimation from the observed NR dimension and density. The actual values may show big variation from this number due to the deviated density or distribution of the Si NW forest. For instance, the density of Si NWs is not uniform, and at certain areas, it could be much higher than 10 NWs per square micrometers. The bundling effect at the tip region of Si NWs could significantly reduce the available surface area locally. Therefore, this calculated roughness factor just presents a reference showing the capability of surface area density improvement by the 3D NW architectures. The actual value may vary sample by sample and the best way to determine it is to measure it experimentally.

A 3D TiO₂ NR–Si NW architecture was used as the photoanode for PEC water splitting. A typical plot of photocurrent density versus bias potential is shown in Fig. 8(a). Interrupted illumination was applied to demonstrate the instant dark current and photocurrent density. The dark current density remained at a very low level (<10⁻² mA/cm²) under bias potentials between −1.2 and 0.5 V (versus saturated calomel electrode) indicating the high quality of the crystal surfaces of TiO₂. The J-E curve exhibited a fill factor of 0.54 and short-circuit current density (J _SC) of 1.14 mA/cm², which demonstrated good charge transport properties and interfacial reaction chemistries of these 3D TiO₂NR–Si NW anodes.

FIG. 8.

(a) J-E curve of TiO₂ NR on 10-μm dry-etched Si NW under interrupted illumination. (b) J-E curves of different TiO₂–Si NW configurations showing enhanced PEC performance by introducing high density TiO₂ NR branches. (c, d) SEM images of TiO₂ NR–Si NW structures after 250 and 375 cycles ALD overcoating, respectively. More overcoating cycles buried more TiO₂ NRs in the TiO₂ film.

To demonstrate the merits of the 3D NW architectures as PEC anodes, performance of three TiO₂–Si NW-based configurations were characterized and compared. They are wet-etched Si NWs with TiO₂ NRs, dry-etched Si NWs with TiO₂ NRs, and wet-etched Si NWs coated with a 375-cycle (∼37 nm) TiO₂ film. Both TiO₂ NR and Si NW samples were further overcoated with an additional 375-cycle TiO₂ film to isolate Si NWs from KOH electrolyte. All the Si NW templates were 10 μm long. A light source with intensity of 100 mW/cm² was provided by a 500 W Hg (Xe) arc lamp (Oriel, 66142, Irvine, CA). A liquid water filter (Oriel, 6123NS) was applied to eliminate the infrared light heating effect on electrolyte. Figure 8(b) shows the photocurrent density of these three samples as functions of the bias potential. The three plots followed the same trend and exhibited the same V _OC, while J _SC of these samples were dramatically different. The TiO₂ thin-film-coated Si NWs produced a J _SC of ∼0.66 mA/cm², which was comparable to the value reported by others.Reference Hwang, Boukai and Yang⁵ Both NRs-coated samples exhibited significantly higher J _SC owing to their larger surface areas. Because the density of wet-etched Si NWs was higher than that of dry-etched Si NWs, it yielded the highest J _SC (∼1.43 mA/cm²), which was more than twice of what was produced by the thin-film-coated sample.

A series of 3D NW anodes with different lengths and overcoating thicknesses were fabricated and their performances were compared. The sample conditions, J _SC, and efficiencies are summarized in Table II. As shown in Figs. 8(c) and 8(d), increasing overcoating cycles could “bury” more NRs in the TiO₂ film and decrease the overall surface area of TiO₂ crystals. Therefore, the highest efficiency (2.1%) was identified from the 20-μm sample with the 250-cycle overcoating (∼25 nm). This value was three times higher than that of TiO₂ film–Si NWs core–shell structure. This comparison suggested that the 3D NW architecture is likely superior to straight NW arrays for PEC electrode design. The efficiency could be further improved by optimizing the number of overcoating cycles and the length/density of NW backbones.

TABLE II.

Summary of photocurrent density and PEC efficiency of TiO₂–Si heterostructures.