The Effect of Annealing Treatment and Atom Layer Deposition to Au/Pt Nanoparticles-Decorated TiO2 Nanorods as Photocatalysts

The wide band gap of TiO2 hinders the utilization of visible light in high-performance photocatalysis. Herein, vertically aligned Ti nanopillar arrays (NPAs) were grown by the glancing angle deposition method (GLAD) and then thermally oxidized into TiO2 NPAs. The metallic nanoparticles (NPs) were fabricated by successive ion layer adsorption and reaction (SILAR) method. And we covered ultrathin TiO2 layer on Au/Pt NPs decorated NPA using atomic layer deposition (ALD) method and did annealing process in the end. The photoelectrochemical (PEC) performance and dye degradation have been studied. We find the dye degradation efficiency of best combination reaches up to 1.5 times higher than that of original Au/Pt-TiO2 sample under visible light irradiation. The TiO2 ALD layer effectively protects the nanostructure from corrosion and helps the transmission of electrons to the electrolyte. By controlling the annealing temperature we could achieve a matched band gap due to change in noble metal particle size. Our work demonstrates that rational design of composite nanostructures enhances the usage of broader wavelength range light and optimizes photocatalytic degradation of organic pollutants in practical applications.


Introduction
Composite nanomaterials have attracted researchers' attention due to their multifunctional applications in various fields, such as semiconductor photocatalysts for pollutants degradation. TiO 2 , due to its nontoxic nature, long-term photostability, low cost, and relatively efficient photocatalytic activity, has been investigated in this context for many years [1]. However, it is just effective and operative in the UV region because of its wide band gap (~3.0 eV) [2] and the UV region only accounts for about~4% in the solar radiation [3]. Moreover, the photogenerated electron-hole pairs have a quite short recombination time that also hinders further utilization of TiO 2 . Previously, a lot of work has been done to address these issues with apporaches such as noble-metal decoration [4], ion doping [5][6][7] or alteration of the unit morphology [8]. Nowadays the quantum size effect has been focused on as another hot topic [9,10]. TiO 2 nanostructures tend to aggregate easily, which effects the efficiency in practical experiments. Here we use the glancing angle deposition (GLAD) technique to fabricate TiO 2 nanopillar arrays (NPAs) vertically on substrates which are conveniently recycled and guarantee enough reaction surface area during the whole process.
Nanostructured noble metals could have a collective coherent oscillation of surface electrons which enables noble metals nanoparticles (NPs) to present surface plasmon resonance (SPR) effects [11,12]. These kinds of nanostructures show enhanced performance in a broad light range including the visible as well as the UV region [13][14][15], so TiO 2 is usually combined with other semiconductors to form composites which possess certain functions such as photocatalytic properties. Li et al. [16] have designed a CdS-Au-TiO 2 sandwich nanorod array capable of producing a photocurrent of 4.07 mA·cm −2 at 0 V (vs. Ag|AgCl) under full solar spectrum irradiation. It also showed a maximum solar-to-chemical energy conversion efficiency of 2.8%, which is much higher than that of pure TiO 2 . Herein Au NPs serve as plasmonic photosensitizers and form a Schottky barrier with TiO 2 . CdS and TiO 2 are band gap matched, which enhances the excited separation of electrons and holes. As for the synthesis of noble metal particles on nanostructures, there are many ways to achieve this such as chemical vapour deposition [17], sol-gel synthesis [18], hydrothermal methods [19] and successive ion layer adsorption and reaction (SILAR) [20]. Here we have co-decorated both Au and Pt NPs on the surface of nanorods through the SILAR method because of the easy preparation in room temperature.
Corrosion always happens on nanostructures during the degradation process [21]. In order to protect nanostructures and suppress the recombination, a metal oxide thin layer such as Al 2 O 3 [22], SnO 2 [23] and ZrO 2 [24] is always applied through atomic layer deposition (ALD) or other methods [25,26]. George et al. have employed Al 2 O 3 and TiO 2 ALD when fabricating an ultrathin barrier film on a Cu substrate to prevent water corrosion [27]. As for catalysis, Lu et al. [28] have applied a TiO 2 overcoat on Au catalysts by ALD. They found higher activity in CO oxidation and investigated the reaction mechanism. This technique could control the thickness of layers at the nanoscale by setting the number of deposition cycles and self-limiting the reaction of dosed precursor gases.
Herein, we report co-decorated Au/Pt NPs on TiO 2 NPAs with a wrapped up TiO 2 ALD layer. TiO 2 nanopillar arrays (NPAs) were first deposited on SiO 2 , F-doped SnO 2 (FTO) and Si substrates through the GLAD method. Later metallic NPs (Au and Pt) were deposited on the NPAs using the SILAR method [29]. Then, we coated a TiO 2 ALD layer on the nanostructured material. In order to improve the stability of the photocatalyst, we annealed the sample at different temperatures. The whole process is shown in Figure 1. We have also studied the effect of annealing on the photocatalytic performance of our fabricated nanostructures. visible as well as the UV region [13][14][15], so TiO2 is usually combined with other semiconductors to form composites which possess certain functions such as photocatalytic properties. Li et al. [16] have designed a CdS-Au-TiO2 sandwich nanorod array capable of producing a photocurrent of 4.07 mA·cm −2 at 0 V (vs. Ag|AgCl) under full solar spectrum irradiation. It also showed a maximum solarto-chemical energy conversion efficiency of 2.8%, which is much higher than that of pure TiO2. Herein Au NPs serve as plasmonic photosensitizers and form a Schottky barrier with TiO2. CdS and TiO2 are band gap matched, which enhances the excited separation of electrons and holes. As for the synthesis of noble metal particles on nanostructures, there are many ways to achieve this such as chemical vapour deposition [17], sol-gel synthesis [18], hydrothermal methods [19] and successive ion layer adsorption and reaction (SILAR) [20]. Here we have co-decorated both Au and Pt NPs on the surface of nanorods through the SILAR method because of the easy preparation in room temperature. Corrosion always happens on nanostructures during the degradation process [21]. In order to protect nanostructures and suppress the recombination, a metal oxide thin layer such as Al2O3 [22], SnO2 [23] and ZrO2 [24] is always applied through atomic layer deposition (ALD) or other methods [25,26]. George et al. have employed Al2O3 and TiO2 ALD when fabricating an ultrathin barrier film on a Cu substrate to prevent water corrosion [27]. As for catalysis, Lu et al. [28] have applied a TiO2 overcoat on Au catalysts by ALD. They found higher activity in CO oxidation and investigated the reaction mechanism. This technique could control the thickness of layers at the nanoscale by setting the number of deposition cycles and self-limiting the reaction of dosed precursor gases.
Herein, we report co-decorated Au/Pt NPs on TiO2 NPAs with a wrapped up TiO2 ALD layer. TiO2 nanopillar arrays (NPAs) were first deposited on SiO2, F-doped SnO2 (FTO) and Si substrates through the GLAD method. Later metallic NPs (Au and Pt) were deposited on the NPAs using the SILAR method [29]. Then, we coated a TiO2 ALD layer on the nanostructured material. In order to improve the stability of the photocatalyst, we annealed the sample at different temperatures. The whole process is shown in Figure 1. We have also studied the effect of annealing on the photocatalytic performance of our fabricated nanostructures.     In order to improve the degree of crystallinity, we tried anneal the samples after ALD coating. Here we choose three different temperature (200 °C, 300 °C and 400 °C, respectively). The TEM images in Figure 3 show TiO2 without annealing (Figure 3a [30]. The generation of metal-semiconductor nanojunctions can result in the formation of a Schottky junction, which can dramatically affect the optoelectronic properties of the material. When Au and Pt contact to TiO2, the charge carriers would redistribute. Electrons move from the ntype semiconductor (higher Fermi level) to the metal (lower Fermi level) until reaching equilibrium. The Schottky barrier which could efficiently capture excited electrons. These facts would all be favourable to interfacial charge transfer between each component and improve the photocatalytic performance of the whole system eventually [31]. Our experimental results show that the interface of ALD thin film becomes clearer with higher crystallinity. This phenomenon indicates that the degree of crystallinity of our nanostructures improves with annealing treatment [32]. Here the number of ALD cycles is indicated before the term "ALD" in the sample name.

Characterization of Materials
To confirm the valence of metallic NPs, the 2ALD/Au/Pt-TiO2 NPAs without annealing sample was investigated by the XPS technique ( Figure 4). The intense doublet of Pt (70.2 and 73.6 eV) and Au (83.7 and 87.4 eV) correspond to metallic Pt 0 , Au 0 [33]. These results show that the Au and Pt NPs were both successfully fabricated on TiO2 NPAs. Furthermore the outer layer is also TiO2. The O 1s XPS spectrum (Figure 4c) can be fitted by two peaks: a main peak showing at a binding energy of 529.8 eV which is attributed to lattice 'O' and another one located at 531.4 eV that refers to the hydroxyls or water adsorbed on the surface of the nanostructure. The Ti 2p spectrum includes doublets of Ti 2p1/2 and Ti 2p3/2 at binding energies of 466.1 and 460.3 eV which confirm the existence of Ti 4+ cations in TiO2 [34]. There are two peak located around 464.6 and 458.8 eV indicating Ti-O-C 2p1/2 and Ti-O-C 2p3/2 bonds which are due to the ALD reaction process which indicates intermediate products formed after the reaction between tetrakis(dimethylamido)titanium(IV) and H2O [35]. In order to improve the degree of crystallinity, we tried anneal the samples after ALD coating. Here we choose three different temperature (200 • C, 300 • C and 400 • C, respectively). The TEM images in Figure 3 show TiO 2 without annealing (Figure 3a (111), respectively [30]. The generation of metal-semiconductor nanojunctions can result in the formation of a Schottky junction, which can dramatically affect the optoelectronic properties of the material. When Au and Pt contact to TiO 2 , the charge carriers would redistribute. Electrons move from the n-type semiconductor (higher Fermi level) to the metal (lower Fermi level) until reaching equilibrium. The Schottky barrier which could efficiently capture excited electrons. These facts would all be favourable to interfacial charge transfer between each component and improve the photocatalytic performance of the whole system eventually [31]. Our experimental results show that the interface of ALD thin film becomes clearer with higher crystallinity. This phenomenon indicates that the degree of crystallinity of our nanostructures improves with annealing treatment [32]. Here the number of ALD cycles is indicated before the term "ALD" in the sample name.
To confirm the valence of metallic NPs, the 2ALD/Au/Pt-TiO 2 NPAs without annealing sample was investigated by the XPS technique ( Figure 4). The intense doublet of Pt (70.2 and 73.6 eV) and Au (83.7 and 87.4 eV) correspond to metallic Pt 0 , Au 0 [33]. These results show that the Au and Pt NPs were both successfully fabricated on TiO 2 NPAs. Furthermore the outer layer is also TiO 2 . The O 1s XPS spectrum (Figure 4c Figure 5a,b, we can see that the currents between light on/off are all about 15 µA cm −2 , which doesn't change much even under 200 • C heat treatment. However, when annealed at 300 • C and 400 • C for 1 h, the photocurrent generally increased. Especially, 300 • C annealing with two ALD layer cycles and 400 • C with three layers reach to 40 and 35 µA cm −2 (Figure 5c,d). They are enlarged about 1.5~2 times compared to the previous one without annealing treatment.
Molecules 2018, 23, x 5 of 11 generally increased. Especially, 300 °C annealing with two ALD layer cycles and 400 °C with three layers reach to 40 and 35 μA cm −2 (Figure 5c,d). They are enlarged about 1.5~2 times compared to the previous one without annealing treatment. We also investigated the photocatalytic performance and degradation of Methylene Blue (MB) dye under visible light for 1 h. The Au/Pt-TiO2 NPs sample shows a degradation efficiency of 49.3%. Furthermore we analyzed the degradation efficiency of our composite nanostructures with different ALD cycle numbers and annealing temperature, as mentioned in Table 1. Our experiments show that after ALD coating and heat treatment most samples showed an improvement in the photocatalytic degradation. The sample without ALD shows lower efficiency because the metal NPs are easily photocorroded without the protection of the outer coating layer [36]. Compared with the ALD-coated samples, the excited electrons could not only transfer through TiO2 NPAs, but also ALD TiO2 which surrounds the metal NPs. Specifically the 2ALD/Au/Pt-TiO2 NPAs sample (with annealing at 300 °C) shows the highest degradation efficiency which is enhanced by about 150% compared with the blank sample. However, on the contrary the efficiency of composite nanostructures annealed at 400 °C goes down. For samples with the same annealing temperature, two cycles and three cycles usually show better performance. Through our experiemntal data calculations, annealing mainly influences and enhances the degradation efficiency of the MB pollutant. Usually demethylation would happen on MB when degraded by TiO2, thus the absorption peak of the dye solution will cause a blue shift [37].  We also investigated the photocatalytic performance and degradation of Methylene Blue (MB) dye under visible light for 1 h. The Au/Pt-TiO 2 NPs sample shows a degradation efficiency of 49.3%. Furthermore we analyzed the degradation efficiency of our composite nanostructures with different ALD cycle numbers and annealing temperature, as mentioned in Table 1. Our experiments show that after ALD coating and heat treatment most samples showed an improvement in the photocatalytic degradation. The sample without ALD shows lower efficiency because the metal NPs are easily photocorroded without the protection of the outer coating layer [36]. Compared with the ALD-coated samples, the excited electrons could not only transfer through TiO 2 NPAs, but also ALD TiO 2 which surrounds the metal NPs. Specifically the 2ALD/Au/Pt-TiO 2 NPAs sample (with annealing at 300 • C) shows the highest degradation efficiency which is enhanced by about 150% compared with the blank sample. However, on the contrary the efficiency of composite nanostructures annealed at 400 • C goes down. For samples with the same annealing temperature, two cycles and three cycles usually show better performance. Through our experiemntal data calculations, annealing mainly influences and enhances the degradation efficiency of the MB pollutant. Usually demethylation would happen on MB when degraded by TiO 2 , thus the absorption peak of the dye solution will cause a blue shift [37]. In order to study the effect of annealing treatment in depth, diffuse reflectance UV-vis spectra of five different products (TiO 2 NPAs, Au/Pt-TiO 2 NPAs and Au/Pt-TiO 2 NPAs with annealing for 200 • C, 300 • C, 400 • C) are presented in Figure S1, which are also calculated with the Kubelka−Munk function. The equation of the band gap energy (E g ) is αE photon = K E photon − E g 1/2 , where α, K, E g and E photon are respectively the absorption coefficient, a constant, the band gap energy and the discrete photon energy. Here we plot and calculate the intercept of an extrapolated linear fit [20].
TiO 2 NPAs exhibit a UV absorption peak at 300~400 nm. The metal decoration of TiO 2 NPAs indicates a remarkable improvement in the light absorption at about 480 nm due to the surface plasmon absorption of the decorated Au NPs [38], but we didn't find any obvious difference in the UV adsorption after the annealing treatments, thus we turned our attention to the size of the metal NPs.
Here we compare the statistical NPs radius according to the TEM images. The histograms are plotted in Figure 3c,f,i. From the average diameter statistics, the size of 2ALD/Au/Pt-TiO 2 NPAs without annealing and annealing under 300 • C and 400 • C increases with higher heat treatment from 7.5 nm to 11.0 nm, but we could find that there is an extremum in the photoccatalytic properties, which means there probably exists an optimal NP diameter size.

Photocatalytic Mechanism
After visible light irradiation, strong SPR is excited by Au NPs [39]. During its dephasing, hot electrons are injected via Schottky junction into the conduction band (CB) for reduction reactions, leaving holes in the noble metal nanoparticles where oxidation reactions occur [40][41][42]. So in this system hot electrons shift from the surface of the Au NPs to the TiO 2 conduction band [43]. The work function of Pt (~5.40 eV in vacuum [44]) is higher than that of Au (~4.78 eV in vacuum) which leads to the Fermi level of Pt being lower than that of Au. Thus electrons would move from Au to Pt via TiO 2 . Here Pt NPs play a role as cocatalyst.
Here electrons can both decompose the dye and interact with electron acceptors (O 2 ) to create superoxide radicals (O 2 •− ). At same time, the positively charged Au NPs would combine with OH − into highly oxidizing species such as hydroxyl radicals, OH• or oxidize the MB molecule directly. The whole reaction is presented in Figure 6a. It is necessary to be aware that a close Au/TiO 2 Schottky contact is due to the last-step annealing process. The treatment is helpful to produce an efficient visible-light photocatalyst [45]. Furthermore heterojunction arrays, which is the light trapping caused from the one-dimensional array nanostructure with matched bandgap, also improve the light absorption [46]. When the size reaches the nanoscale, the Fermi level of noble metal NPs usually improves with decreased size because of the quantum size effect. The mechanism is shown in Figure 6b. A suitable diameter could be achieved to match the band gap. Only this could convert electrons efficiently and suppress the combination of electrons and holes [47,48]. The energy level of Au NPs must stay higher than that of oxygen gas and lower than the CB of TiO 2 . When hot electrons are excited by visible light from Au NPs, just the rational energy level could let electrons move into the CB of TiO 2  if the particles become too big, they will also cover too much of the TiO 2 's reaction surface which in turn hinders the photocatalytic properties.

Synthesis of TiO2 NPAs
Firstly, Ti NPAs were fabricated via e-beam using the GLAD technique on several substrates (i.e., quartz, silicon and FTO). The chamber need to be evacuated to 1 × 10 −8 Torr before deposition. All substrates were cleaned using ultrasonication in acetone, ethanol and deionized (DI) water, respectively. The data of deposition was similar to that used in previous work [49]. After deposition, the thin films were oxidized in a tube furnace at 400 °C for 2 h.

Metallic Nanoparticle Decoration on TiO2 NPAs
Au and Pt NPs were reacted via the SILAR method with some modifications as previously published [49]. We dipped TiO2 successively into HAuCl4 (or HPt2Cl6) and reductant (NaBH4) solutions in order to form NPs. The substrates were immersed into a certain concentration of HAuCl4 (or HPt2Cl6) solution, DI water, 1 mg/mL NaBH4 solution, DI water respectively for 60 s. This complete procedure is regarded as one cycle. The whole process was repeated until the desired cycle number was fabricated. Au/Pt-TiO2 NPAs samples were consecutively deposited by Au and Pt NPs individually five times.

ALD Deposition
TiO2 NPAs were placed in a custom-built reactor, and the substrate temperature was kept at 150 °C during the ALD process [50]. When pumped into the chamber, the precursor, tetrakis(dimethylamido)titanium(IV), was controlled at 110 °C and water was at 40 °C. A soaking step was inserted in order to confirm the sufficient diffusion of the reacting molecules on the surface of the nanostructures [51]. The precursor was first pumped using 200 ms and soaked for 5 s, after which the chamber was evacuated for 20 s. Later, H2O (gas) was pulsed for 5 ms and kept for 3 s to let them react, with a 20 s purge step at the end. This whole procedure can be regarded as one ALD cycle. The process is repeated until the desired thickness of TiO2 is achieved.

Synthesis of TiO 2 NPAs
Firstly, Ti NPAs were fabricated via e-beam using the GLAD technique on several substrates (i.e., quartz, silicon and FTO). The chamber need to be evacuated to 1 × 10 −8 Torr before deposition. All substrates were cleaned using ultrasonication in acetone, ethanol and deionized (DI) water, respectively. The data of deposition was similar to that used in previous work [49]. After deposition, the thin films were oxidized in a tube furnace at 400 • C for 2 h.

Metallic Nanoparticle Decoration on TiO 2 NPAs
Au and Pt NPs were reacted via the SILAR method with some modifications as previously published [49]. We dipped TiO 2 successively into HAuCl 4 (or HPt 2 Cl 6 ) and reductant (NaBH 4 ) solutions in order to form NPs. The substrates were immersed into a certain concentration of HAuCl 4 (or HPt 2 Cl 6 ) solution, DI water, 1 mg/mL NaBH 4 solution, DI water respectively for 60 s. This complete procedure is regarded as one cycle. The whole process was repeated until the desired cycle number was fabricated. Au/Pt-TiO 2 NPAs samples were consecutively deposited by Au and Pt NPs individually five times.

ALD Deposition
TiO 2 NPAs were placed in a custom-built reactor, and the substrate temperature was kept at 150 • C during the ALD process [50]. When pumped into the chamber, the precursor, tetrakis(dimethylamido)titanium(IV), was controlled at 110 • C and water was at 40 • C. A soaking step was inserted in order to confirm the sufficient diffusion of the reacting molecules on the surface of the nanostructures [51]. The precursor was first pumped using 200 ms and soaked for 5 s, after which the chamber was evacuated for 20 s. Later, H 2 O (gas) was pulsed for 5 ms and kept for 3 s to let them react, with a 20 s purge step at the end. This whole procedure can be regarded as one ALD cycle. The process is repeated until the desired thickness of TiO 2 is achieved.

Characterization
The chemical nature, morphology and nanostructure of all samples were characterized by field-emission scanning electron microscopy (SEM, FE-SEM, JEOL-7001F, JEOL, Tokyo, Japan) and high-resolution transmission electron microscope (HRTEM, JEOL-2011, JEOL). The surface composition was determined by X-ray photoemission spectroscopy (XPS) measurements performed on a PHI 5300 instrument (Perkin Elmer, Waltham, MA, USA). Here the binding energy was calibrated with the C1s peak (284.6 eV). The light absorption of materials were tested via a Perkin Elmer Lambda 35 UV-Vis spectrometer (Perkin Elmer) in the wavelength range between 375 nm and 900 nm.

Property Measurements
The photocurrent density was measured by an electrochemistry workstation (CHI 660D, Chenhua Instrument, Shanghai, China). Here the film and substrates were regarded as the working electrode, while a Ag/AgCl electrode (saturated KCl) is used as a reference. Pt slices with specific shapes were applied as counter electrodes. A Xe lamp (300 W) was used as a light source with ultraviolet filter to intercept the UV light (wavelength shorter than 420 nm). Photocurrent densities with increased potential from −0.1~+0.6 V bias vs Ag/AgCl electrode were tested via a light on-off process. Here we selected intervals of 5 s for the switching time. The photocatalytic degradation performance was examined by using Methylene Blue (MB) under visible light. The samples on a quartz substrate (15 mm × 15 mm) were immersed in a 25 mL beaker containing 5 mL of MB (5 µM) solution. The samples remained in solution for half an hour before degradation to achieve an adsorption/desorption equilibrium. At the end of 1 h irradiation, we tested the concentration of MB by UV-Vis spectroscopy at its characteristic wavelength of 502 nm. Both remaining percentage (C/C 0 ) and degradation efficiency (1 − C/C 0 ) × 100% were recorded. PEC experiments with light on-off at specific voltages were also performed to assess the photoelectrochemical properties. 0.1 M Na 2 SO 4 aqueous solution was used as the electrolyte.

Conclusions
In conclusion, we have successfully prepared TiO 2 ALD/Au/Pt NPs-decorated TiO 2 nanostructures with an annealing treatment. When compared with nanostructures without ALD and annealing treatment, the process shows a great improvement no matter whether in the degradation of dyes or photocurrent production under visible light. The TiO 2 ALD layers effectively protect the nanostructures from corrosion and make electrons accessible to the electrolyte. The annealing treatment could adjust the noble metal NPs into a band gap-matched nanoscale. The rational design of nanocomposites helps make the photocatalytic performance of TiO 2 in the visible wavelength region more efficient. Thus this method offers an efficient way to achieve the maximum usage of solar energy for photocatalytic degradation of organic pollutants in practical applications.

Conflicts of Interest:
The authors declare no conflict of interest.