Synthesis of Pt@TiO2@CNTs Hierarchical Structure Catalyst by Atomic Layer Deposition and Their Photocatalytic and Photoelectrochemical Activity

Pt@TiO2@CNTs hierarchical structures were prepared by first functionalizing carbon nanotubes (CNTs) with nitric acid at 140 °C. Coating of TiO2 particles on the CNTs at 300 °C was then conducted by atomic layer deposition (ALD). After the TiO2@CNTs structure was fabricated, Pt particles were deposited on the TiO2 surface as co-catalyst by plasma-enhanced ALD. The saturated deposition rates of TiO2 on a-CNTs were 1.5 Å/cycle and 0.4 Å/cycle for substrate-enhanced process and linear process, respectively. The saturated deposition rate of Pt on TiO2 was 0.39 Å/cycle. The photocatalytic activities of Pt@TiO2@CNTs hierarchical structures were higher than those without Pt co-catalyst. The particle size of Pt on TiO2@CNTs was a key factor to determine the efficiency of methylene blue (MB) degradation. The Pt@TiO2@CNTs of 2.41 ± 0.27 nm exhibited the best efficiency of MB degradation.


Introduction
TiO 2 has been extensively studied in contaminants degradation and hydrogen generation because of its chemical stability, non-toxicity, and low cost since TiO 2 electrode was discovered that photodecomposition of water could be achieved by illumination of UV light in 1972 [1], However, TiO 2 has not been widely applied to the environmental industry because the photoexcited electrons and holes inefficiently diffuse to the surface for redox reaction due to high recombination rate of electrons and holes [2][3][4]. In addition, bulk TiO 2 suffers from the property of low surface area and then limits its photocatalytic reactivity [5,6]. Although the nanosize of TiO 2 was fabricated by wet chemical process to increase the surface area of TiO 2 [4], the TiO 2 particles aggregated during phase transformation [7]. In order to improve the efficiencies of semiconductor photocatalyst, CNTs as a template to coat with metal oxides (MOs) have attracted great attention in environmental and energy applications because of several reasons. First, CNTs facilitate high specific surface area for MOs deposition and prevent MOs from agglomeration during annealing due to its 1-D structure. Second, CNTs as sensitizer can enhance the photodegradation reaction. Third, heterojunction of MOs@CNTs provides the interface to lower the recombination rate of photo-induced electron-hole pair [8][9][10]. Therefore, TiO 2 -based structure such as TiO 2 /CNTs is a good complementary approach to achieve higher quantum efficiency than that of pure TiO 2 [11].
Although the TiO 2 /CNTs heterojunction is considered as a good candidate for increasing the quantum efficiencies [11], the effect of MO@CNTs hierarchical catalysts for photodegradation is still limited. For improving the performance of hierarchical catalysts, additional loading with noble metal such as Pt is a good approach because Pt provides more reactive sites in photoreaction as a co-catalyst [12]. However, uniform and well-dispersed Pt particles with precisely controllable size on TiO 2 are hard to achieve because the deposition usually involves chemical wet processes, such as photo-deposition [13] and impregnation [14], which have some limitations. For instance, for Pt@TiO 2 @carbon composites, the loading of Pt is random through photo-deposition because it is difficult to obtain uniformly reacted surface of TiO 2 with Pt precursors due to the limitation of irradiation [15]. The agglomeration of Pt particles fabricated by impregnation method takes place when the reduction of metal ions is conducted by chemical reagents or reductive atmosphere such as H 2 and NH 3 at high temperatures [16,17]. In comparison to other techniques, atomic layer deposition (ALD) provides excellent thickness control and nearly perfect conformal coverage of complex nanostructures as a result of the surface-controlled growth [18]. Besides, ALD has been confirmed not only to prepare thin films with high aspect ratio but also to fabricate particle catalyst in nanoscale [19]. For example, Pt nanoparticles as a catalyst are well distributed on CNTs with controllable loading and size for proton exchange membrane fuel cell (PEMFC) [20,21]. The effectively decreasing the loading of Pt by ALD to meet the commercial requirement of PEMFC was achieved by the evidence of highly specific power density.
In this study, pristine CNTs were firstly functionalized by temperature-assisted acid-treatment, and then TiO 2 @CNTs structures were prepared by ALD. Pt nanoparticles were subsequently deposited on the TiO 2 @CNTs structures by plasma-enhanced ALD (PEALD) to form Pt@TiO 2 @CNTs composites. Characterization and photocatalytic degradation of methylene blue (MB) by the nanostructures were also investigated.

Fabrication of TiO 2 @CNTs Structure by ALD
The sample of a-CNTs was used as a support for coating with TiO 2 by ALD. Firstly, 5 mg of CNTs were sonicated in 15 mL N-Methyl-2-pyrrolidone for 60 min to get good dispersion on a Si wafer. The dispersed CNTs were then deposited with TiO 2 at 300 • C by ALD using TiCl 4 and H 2 O as the precursors. The temperature of precursors was maintained in room temperature. Each cycle consisted of a precursor pulse for 0.08 s and a purge with N 2 for 7 s. The number of precursor/purge cycles was from 25 to 200 cycles of ALD to achieve precise control of the particles size of TiO 2 .

Fabrication of Pt@TiO 2 @CNTs Composites by PEALD
Pt particles were then deposited on the as-prepared TiO 2 @CNTs composites by PEALD using (methylcyclopentadienyl) trimethylplatinum (MeCpPtMe 3 ) and oxygen plasma as precursors [23]. The sequence of Pt-N 2 -O 2 plasma-N 2 was 0.5-5-3-5 s, respectively. The power of induced couple plasma was 400 W and substrate temperature was 300 • C. The temperature of Pt precursor is heated to 80 • C for enough vapor pressure. The cycles of PEALD were from 10 to 100 to meet precisely controllable size and loading of Pt particles.

Photocatalytic Activity and Characterization
The surface chemistry was analyzed by Fourier-Transform Infrared Spectrometer (FTIR, Bruker, Vertex 80v and Tensor 27, Billerica, Massachusetts, USA). The TiO 2 @CNTs and Pt@TiO 2 @CNTs composites prepared by using a-CNTs as the template were examined by transmission electron microscopy (TEM, JEOL 2100F, Akishima, Tokyo, Japan). The structural analysis was performed by X-ray diffraction (XRD, Shimadzu 6000, Nakagyo-ku, Kyoto, Japan) with the Cu K α radiation. The optical properties were obtained by photoluminescence spectroscopy (PL, Horiba Jobin Yvon, Labram HR 800, Minami-Ku, Kyoto, Japan) with excitation wavelength of 325 nm (He-Cd laser, Kimon IK3301R-G Itabashi-Ku, Kyoto, Japan), laser power of 30 mW, and a spot size of 0.79 µm 2 . The photocatalytic activities of all the samples were examined by studying the MB degradation with 200 W Hg lamp as irradiation source. The initial concentration of MB is 6.26 × 10 −5 M. The distance between photoreactor and lamp was 10 cm. 2 mg of prepared samples were added to the MB solution and P25 was tested as control group. The absorption of MB solution at 664 nm was measured by a UV-Visible spectrometer (Hitachi U-3010, Chiyoda-ku, Tokyo, Japan) for every 10 min illumination until 60 min. The total organic carbon (TOC) concentration was determined with a TOC analyzer (Shimadzu TOC-5000, Nakagyo-ku, Kyoto, Japan). Photoelectrochemical cell (PEC) measurement with 1 M KOH as electrolyte was performed in a three electrode electrochemical cell. Ag/AgCl electrode and Pt foil were used as the reference electrode and counter electrode, respectively. All the samples prepared for PEC measurement were dispersed in polyethylene glycol to form slurry and subsequently printed onto fluorine-doped tin oxide (FTO) by doctor-blade technique as working electrodes [24]. The procedure was repeated to obtain the film thickness of 3 µm that was determined by alpha-step profilometer (Veeco Tektak 150, Plainview, New York, NY, USA). The photocurrent was measured by a potentiostat (Solartron 1286, Bognor Regis, West Sussex, UK) by illumination with a 200 W Hg lamp at a sweeping rate of 10 mV·s −1 from −1.0 to 0.4 V. The active area was 1 cm × 1 cm. Figure 1 shows the FTIR spectra of p-CNTs and a-CNTs. The FTIR peaks are identified for different types of organic functional groups. The small peak at 1040 cm −1 can be related to the C-O stretching vibration of either alcohol or carboxyl groups [25]. The peak at 1280 cm −1 is attributed to the nitrates NO 2 symmetric stretching bands [26]. The peak at around 1400 cm −1 is possibly associated with O-H bending deformation of carboxyl groups and the C=C stretching bands placed by the functional groups. The absorption peaks at around 1700 cm −1 is ascribed to the C=O stretching vibration of carboxyl groups [26]. A clear broad peak at around 3400 cm −1 can be referred to O-H stretching. The outcome of FTIR analysis suggests that oxidation has conducted by the acid treatment and demonstrates that there is significant difference between p-CNTs and a-CNTs. Figure 2 shows the morphology of TiO 2 deposited both p-CNTs and a-CNTs by 50 ALD cycles at 300 • C. There was more TiO 2 particles deposited on the surface of a-CNTs than that on p-CNTs. The coverage of TiO 2 on a-CNTs was more complete because of higher oxygen-containing functional on the surface of a-CNTs. This suggests that the functional groups played an important role in the density of TiO 2 particles deposited on the CNTs, resulting in that the TiO 2 particles coated on a-CNTs were more uniform and compact than that on p-CNTs. This implies the nucleation sites of TiO 2 on a-CNTs were higher than that on p-CNTs after acid treatment. However, the particle size of TiO 2 deposited on both p-CNTs and a-CNTs was not obviously different to each other. In order to increasing the amount of TiO 2 particles loaded on the CNTs for higher efficiency of MB photodegradation, the acid-treated CNTs for 6 h was choose to be the support as following samples prepared by ALD to form TiO 2 @CNTs composites.

Fabrication and Characterization of TiO2@CNTs Structures
The difference of TiO2 particle size grown by various ALD cycles at 300 °C was observed by TEM, as shown in Figure 3. It is worthy to note that the particle size distribution of TiO2 fabricated by variation of ALD cycles was uniform and close to Gaussian distribution. The deviation of average size of TiO2 is less than 8%. This shows that the TiO2 particles could be precisely and uniformly coated on the surface of a-CNTs by ALD to synthesize complete TiO2@CNTs heterojunction structures. By counting more than one hundred particles from the TEM images, the average diameters of TiO2 particles are from 3 to 13 nm and plotted in Figure 4. The particle size increases as the cycle number increases, and the relation can be divided into two regions from Figure 4. Based on the generally accepted mechanism of ALD using TiCl4 and H2O as precursors [27], the process can be separated to two half reactions for a typical self-terminating process in ALD. In the very beginning, TiCl4 reacts with the surface OH groups and then releases HCl in the first-half reaction: n(-OH)(s) + TiCl4(g) → (-O-)nTiCl4-n(s) + nHCl(g) In which (s) presents the surface. The nonstoichiometric titanium chloride (TiClx) species adsorbed on the surface may react with water and then release HCl in the second-half reaction, which results in the recovery of the surface OH groups again for the next cycle:

Fabrication and Characterization of TiO2@CNTs Structures
The difference of TiO2 particle size grown by various ALD cycles at 300 °C was observed by TEM, as shown in Figure 3. It is worthy to note that the particle size distribution of TiO2 fabricated by variation of ALD cycles was uniform and close to Gaussian distribution. The deviation of average size of TiO2 is less than 8%. This shows that the TiO2 particles could be precisely and uniformly coated on the surface of a-CNTs by ALD to synthesize complete TiO2@CNTs heterojunction structures. By counting more than one hundred particles from the TEM images, the average diameters of TiO2 particles are from 3 to 13 nm and plotted in Figure 4. The particle size increases as the cycle number increases, and the relation can be divided into two regions from Figure 4. Based on the generally accepted mechanism of ALD using TiCl4 and H2O as precursors [27], the process can be separated to two half reactions for a typical self-terminating process in ALD. In the very beginning, TiCl4 reacts with the surface OH groups and then releases HCl in the first-half reaction: In which (s) presents the surface. The nonstoichiometric titanium chloride (TiClx) species adsorbed on the surface may react with water and then release HCl in the second-half reaction, which results in the recovery of the surface OH groups again for the next cycle:

Fabrication and Characterization of TiO 2 @CNTs Structures
The difference of TiO 2 particle size grown by various ALD cycles at 300 • C was observed by TEM, as shown in Figure 3. It is worthy to note that the particle size distribution of TiO 2 fabricated by variation of ALD cycles was uniform and close to Gaussian distribution. The deviation of average size of TiO 2 is less than 8%. This shows that the TiO 2 particles could be precisely and uniformly coated on the surface of a-CNTs by ALD to synthesize complete TiO 2 @CNTs heterojunction structures. By counting more than one hundred particles from the TEM images, the average diameters of TiO 2 particles are from 3 to 13 nm and plotted in Figure 4. The particle size increases as the cycle number increases, and the relation can be divided into two regions from Figure 4. Based on the generally accepted mechanism of ALD using TiCl 4 and H 2 O as precursors [27], the process can be separated to two half reactions for a typical self-terminating process in ALD. In the very beginning, TiCl 4 reacts with the surface OH groups and then releases HCl in the first-half reaction: In which (s) presents the surface. The nonstoichiometric titanium chloride (TiCl x ) species adsorbed on the surface may react with water and then release HCl in the second-half reaction, which results in the recovery of the surface OH groups again for the next cycle: Nanomaterials 2017, 7, 97 5 of 16  For TiO2@CNTs composites, once the TiClx species has formed on the surface of CNTs, the H2O molecules then reacted with TiClx to form TiO2 and subsequently left the OH functional groups again   For TiO2@CNTs composites, once the TiClx species has formed on the surface of CNTs, the H2O molecules then reacted with TiClx to form TiO2 and subsequently left the OH functional groups again For TiO 2 @CNTs composites, once the TiCl x species has formed on the surface of CNTs, the H 2 O molecules then reacted with TiCl x to form TiO 2 and subsequently left the OH functional groups again in the surface. The surface functional groups on a-CNTs would affect the growth rate of TiO 2 particles because the reaction was determined by the interaction among surfaces of CNTs, TiCl 4 molecules and H 2 O molecules. There was a substrate-enhanced growth process with a higher growth rate when the ALD cycles were less than 50 [28]. In this region, the surface a-CNTs with fully oxygen-containing functional groups enhanced the TiO 2 particles to grow. However, that particles were probably too tiny to be observed when the cycle number was less than 25. After 50 cycles of ALD, TiO 2 was deposited on the existing particles, which resulted in a linearly lower growth rate in the subsequent cycles due to consuming out of the initial functional groups. The growth rates were calculated to be 1.5 Å/cycle and 0.4 Å/cycle for substrate-enhanced growth process and linear growth process, respectively.
The XRD patterns in Figure 5 show the phase of TiO 2 deposited on a-CNTs by variable ALD cycles. The TiO 2 particles with different ALD cycles deposited on a-CNTs exhibited a crystalline anatase phase. The intensity of anatase phase increases as the ALD cycles increase from 25 to 200. Rutile phase was not observed in the TiO 2 particles coated on a-CNT at 300 • C. However, the crystallinity of TiO 2 was hardly observed when the number of cycle was less than 25. According to Zhang's report [29], the nucleation of TiO 2 ALD was about 20 cycles. The size of TiO 2 nuclei was less than 2 nm when temperatures were 120 • C and 200 • C. In the meanwhile, the crystalline particles were also consisted of amorphous shell and crystalline core. The ratio of amorphous shell and crystalline core decreased as ALD cycles decreased [29]. The TiO 2 particles less than 25 ALD cycles were probably too weak to be observed in XRD patterns due to the initial nucleation of TiO 2 . It was in agreement with the TEM analysis, as shown in Figure 3a. Although TiO 2 deposited on a-CNTs at 300 • C was uniform from 50 to 200 cycles, the TiO 2 nanoparticles around 10 nm with higher photocatalytic property have been demonstrated [30]. Based on that, the TiO 2 particles fabricated by 100 ALD cycles with 9.56 nm were chosen for the following characterization and MB photodegradation. in the surface. The surface functional groups on a-CNTs would affect the growth rate of TiO2 particles because the reaction was determined by the interaction among surfaces of CNTs, TiCl4 molecules and H2O molecules. There was a substrate-enhanced growth process with a higher growth rate when the ALD cycles were less than 50 [28]. In this region, the surface a-CNTs with fully oxygen-containing functional groups enhanced the TiO2 particles to grow. However, that particles were probably too tiny to be observed when the cycle number was less than 25. After 50 cycles of ALD, TiO2 was deposited on the existing particles, which resulted in a linearly lower growth rate in the subsequent cycles due to consuming out of the initial functional groups. The growth rates were calculated to be 1.5 Å/cycle and 0.4 Å/cycle for substrate-enhanced growth process and linear growth process, respectively.
The XRD patterns in Figure 5 show the phase of TiO2 deposited on a-CNTs by variable ALD cycles. The TiO2 particles with different ALD cycles deposited on a-CNTs exhibited a crystalline anatase phase. The intensity of anatase phase increases as the ALD cycles increase from 25 to 200. Rutile phase was not observed in the TiO2 particles coated on a-CNT at 300 °C. However, the crystallinity of TiO2 was hardly observed when the number of cycle was less than 25. According to Zhang's report [29], the nucleation of TiO2 ALD was about 20 cycles. The size of TiO2 nuclei was less than 2 nm when temperatures were 120 °C and 200 °C. In the meanwhile, the crystalline particles were also consisted of amorphous shell and crystalline core. The ratio of amorphous shell and crystalline core decreased as ALD cycles decreased [29]. The TiO2 particles less than 25 ALD cycles were probably too weak to be observed in XRD patterns due to the initial nucleation of TiO2. It was in agreement with the TEM analysis, as shown in Figure 3a. Although TiO2 deposited on a-CNTs at 300 °C was uniform from 50 to 200 cycles, the TiO2 nanoparticles around 10 nm with higher photocatalytic property have been demonstrated [30]. Based on that, the TiO2 particles fabricated by 100 ALD cycles with 9.56 nm were chosen for the following characterization and MB photodegradation.

Fabrication and Characterization of Pt@TiO2@CNTs Structures
After the TiO2@CNTs structure was fabricated, Pt particles as co-catalyst were then deposited on the TiO2 surface at 300 °C with 25, 50, and 100 cycles by PEALD using MeCpPtMe3 and oxygen plasma as precursors. When the lengths of pulse time for the sequence MeCpPtMe3-N2-O2 plasma-N2 were 0.5-5-3-5 s, the saturated deposition rate of Pt on TiO2@CNTs was 0.39 Å/cycle. The size distributions and Pt particles by calculating more than one hundred particles from the TEM images are showed in Figure 6. The sizes of Pt particle with 25, 50, and 100 PEALD cycles are 1.47 ± 0.25 nm, 2.41 ± 0.27 nm, and 3.90 ± 0.37 nm, respectively. The particle size of Pt related to number of ALD cycle

Fabrication and Characterization of Pt@TiO 2 @CNTs Structures
After the TiO 2 @CNTs structure was fabricated, Pt particles as co-catalyst were then deposited on the TiO 2 surface at 300 • C with 25, 50, and 100 cycles by PEALD using MeCpPtMe 3 and oxygen plasma as precursors. When the lengths of pulse time for the sequence MeCpPtMe 3 -N 2 -O 2 plasma-N 2 were 0.5-5-3-5 s, the saturated deposition rate of Pt on TiO 2 @CNTs was 0.39 Å/cycle. The size distributions and Pt particles by calculating more than one hundred particles from the TEM images are showed in Figure 6. The sizes of Pt particle with 25, 50, and 100 PEALD cycles are 1.47 ± 0.25 nm, 2.41 ± 0.27 nm, and 3.90 ± 0.37 nm, respectively. The particle size of Pt related to number of ALD cycle was shown in Figure 7. From Figure 7, the particle size increases with the cycle number, and the relation only has one slope. In comparison to TiO 2 fabricated by ALD using TiCl 4 and H 2 O as precursors, there is no substrate-enhanced growth process with a higher growth rate before having consumed the initial functional groups [28]. Therefore, the growth rate of Pt particles on the TiO 2 surface is the same as that on Pt particles deposited earlier, which results the slope derived from particle sizes of Pt divided to cycle number of ALD is linear. This implies the growth behavior of Pt in PEALD between the interface of TiO 2 and Pt precursors is more inert than that of TiO 2 in ALD between the interface of a-CNTs and TiO 2 precursors. Generally speaking, the precursors of ALD reacting to surface of substrate can be divided to two steps: one is physisorption and the other is chemsorption. The adsorption can be considered reversible in physisorption, as illustrated in reactions (3) and (4) where * is the active surface of substrate for adsorption of precursor molecules. Adsorption rate, r a , refers to the amount of molecules M attached to the surface per unit time. Desorption rate, r d , denotes to the amount of molecules M detached from the surface per unit time. For physisorption, the partial pressure of the reactant is a key parameter to describe the adsorption behavior with three assumptions [28], the monolayer is achieved by the maximum amount of adsorbed molecules, all adsorption sites on the surface are assumed equal, and adjacent molecules adsorbed on the surface are assumed not to interact with each other. Once the physisorption is achieved, the precursor would react with functional groups, dangling bond, and defect on the surface of substrate as chemical bonding so called chemsorption. Therefore, the coverage of adsorbed species denoted as Q describing the adsorption process, as following: After saturation in the ALD process, the coverage is constant (dQ/dt = 0), chemsorption coverage Q eq is integrated by the condition of Langmuir isotherm to show the relationship between equilibrium constant of adsorption, K, as following: Considering time effect into Equation (5), the chemisorption coverage Q is calculated as a function of time, However, the three assumptions to derive the Equation (5) are not available in real process. For Pt in PEALD, the molecule of MeCpPtMe 3 is larger than TiCl 4 . Steric hindrance of the MeCpPtMe 3 molecules is taken into account. The initially accessible ligands of the chemisorbed species on the surface block up the space for the neighbor reactants [31]. Therefore, the initial k a of MeCpPtMe 3 in PEALD is lower than that of TiCl 4 in ALD. That presumably explains the initial growth rate of Pt is the same as the following PEALD cycles. By contrast, the molecule of TiCl 4 is smaller and easier to react to surface of a-CNTs without steric hindrance, which causes the initially higher growth rate due to higher adsorption rate. However, once the consumption of functional group on the surface of a-CNTs has been achieved, the TiCl 4 molecule would deposit on the residual TiO 2 particle with lower growth rate because molecules of TiCl 4 reacted to the existed TiO 2 particles rather than surface of a-CNTs.

Photocatalytic Efficiency of Pt@TiO2@CNTs Structure
The adsorption-desorption equilibrium of MB by the samples has been reached before irradiation. The adsorption capacities of dye molecules for all TiO2@CNTs, Pt@TiO2@CNTs samples are almost the same. This implies that the adsorption capacity was mainly affected by the CNTs as substrate. Figure 8 displays the photocatalytic activities of the TiO2@CNTs of 9.56 nm with 100 ALD cycles at 300 °C (Ti@C) and Pt@TiO2@CNTs with the particle size of 1.47 ± 0.25 nm, 2.41 ± 0.27 nm, and 3.90 ± 0.37 nm fabricated by 25, 50, and 100 PEALD cycles at 300 °C, respectively (Pt25@Ti@C,

Photocatalytic Efficiency of Pt@TiO2@CNTs Structure
The adsorption-desorption equilibrium of MB by the samples has been reached before irradiation. The adsorption capacities of dye molecules for all TiO2@CNTs, Pt@TiO2@CNTs samples are almost the same. This implies that the adsorption capacity was mainly affected by the CNTs as substrate. Figure 8 displays the photocatalytic activities of the TiO2@CNTs of 9.56 nm with 100 ALD cycles at 300 °C (Ti@C) and Pt@TiO2@CNTs with the particle size of 1.47 ± 0.25 nm, 2.41 ± 0.27 nm, and 3.90 ± 0.37 nm fabricated by 25, 50, and 100 PEALD cycles at 300 °C, respectively (Pt25@Ti@C,

Photocatalytic Efficiency of Pt@TiO 2 @CNTs Structure
The adsorption-desorption equilibrium of MB by the samples has been reached before irradiation. The adsorption capacities of dye molecules for all TiO 2 @CNTs, Pt@TiO 2 @CNTs samples are almost the same. This implies that the adsorption capacity was mainly affected by the CNTs as substrate. Figure 8 displays the photocatalytic activities of the TiO 2 @CNTs of 9.56 nm with 100 ALD cycles at 300 • C (Ti@C) and Pt@TiO 2 @CNTs with the particle size of 1.47 ± 0.25 nm, 2.41 ± 0.27 nm, and 3.90 ± 0.37 nm fabricated by 25, 50, and 100 PEALD cycles at 300 • C, respectively (Pt25@Ti@C, Pt50@Ti@C, and Pt100@Ti@C). P25 and Pt-loaded P25 with 50 PEALD cycles (Pt50@P25) were also tested for control specimens. The efficiency of MB degradation can be fitted by a pseudo first-order kinetics with different particle size of Pt [32], as follows: where C is the concentration of MB after reaction time t, C o is the initial concentration of MB, and k is the rate constant. According to the photodecomposition result, the k values for each TiO 2 @CNTs and Pt@TiO 2 @CNTs samples were calculated from the slopes, as shown in Figure 8. A blank test reveals that the MB is slowly degraded by the irradiation after 60 min. The kinetic rate constant k for the blank test is 0.0016 min −1 . The TiO 2 @CNTs composite and P25 have almost the same efficiency for MB degradation with the k values being 0.0063 and 0.0060 min −1 , respectively. The k value of Pt@TiO 2 @CNTs with Pt particles of 1.47 ± 0.25 nm by 25 cycles is 0.0087 min −1 . As the particle size of Pt increases to 2.41 ± 0.27 nm by 50 cycles, the photodegradation efficiency substantially increases with rate constant raising to 0.0158 min −1 . That would presumably is due to a large amount of Pt for MB decomposition. However, as the particle size of Pt increases to 3.90 ± 0.37 nm by 100 cycles, the rate constant reduces to 0.010 min −1 . To sum up, the samples of Pt@TiO 2 @CNTs with different Pt loading have a higher decomposition rate of MB than those without Pt loading as co-catalyst. Pt particles provide more effective reaction sites for reductive reaction and cause formation of Schottky barrier between the interface of Pt and TiO 2 . Once the interfacial charge transfer was facilitated, the recombination rate of electrons and holes would become lower and then result in improving separation of electrons and holes [33]. The Pt@TiO 2 @CNTs with Pt deposited by 50 PEALD cycles shows the higher k value than that of P25 with the same PEALD cycles of Pt. The synergistic effect of Pt loaded on TiO 2 @CNTs is more obvious than that of P25.
Pt50@Ti@C, and Pt100@Ti@C). P25 and Pt-loaded P25 with 50 PEALD cycles (Pt50@P25) were also tested for control specimens. The efficiency of MB degradation can be fitted by a pseudo first-order kinetics with different particle size of Pt [32], as follows: where C is the concentration of MB after reaction time t, Co is the initial concentration of MB, and k is the rate constant. According to the photodecomposition result, the k values for each TiO2@CNTs and Pt@TiO2@CNTs samples were calculated from the slopes, as shown in Figure 8. A blank test reveals that the MB is slowly degraded by the irradiation after 60 min. The kinetic rate constant k for the blank test is 0.0016 min −1 . The TiO2@CNTs composite and P25 have almost the same efficiency for MB degradation with the k values being 0.0063 and 0.0060 min −1 , respectively. The k value of Pt@TiO2@CNTs with Pt particles of 1.47 ± 0.25 nm by 25 cycles is 0.0087 min −1 . As the particle size of Pt increases to 2.41 ± 0.27 nm by 50 cycles, the photodegradation efficiency substantially increases with rate constant raising to 0.0158 min −1 . That would presumably is due to a large amount of Pt for MB decomposition. However, as the particle size of Pt increases to 3.90 ± 0.37 nm by 100 cycles, the rate constant reduces to 0.010 min −1 . To sum up, the samples of Pt@TiO2@CNTs with different Pt loading have a higher decomposition rate of MB than those without Pt loading as co-catalyst. Pt particles provide more effective reaction sites for reductive reaction and cause formation of Schottky barrier between the interface of Pt and TiO2. Once the interfacial charge transfer was facilitated, the recombination rate of electrons and holes would become lower and then result in improving separation of electrons and holes [33]. The Pt@TiO2@CNTs with Pt deposited by 50 PEALD cycles shows the higher k value than that of P25 with the same PEALD cycles of Pt. The synergistic effect of Pt loaded on TiO2@CNTs is more obvious than that of P25.  Figure 9 shows the change of TOC for the photocatalytic degradation of MB during UV irradiation with all samples as photocatalysts, which is consistent with the tendency shown in Figure  8. The gradual decrease of TOC represented the gradual disappearance of organic carbon when the MB solution which contained photocatalysts was exposed under UV irradiation. The k values calculated from TOC were 0.0060, 0.0118, 0.0063, 0.0087, 0.0158, or 0.0100 min −1 with P25, Pt50@P25, Pt25@Ti@C, Pt50@Ti@C, or Pt100@Ti@C after UV irradiation for 60 min.  Figure 9 shows the change of TOC for the photocatalytic degradation of MB during UV irradiation with all samples as photocatalysts, which is consistent with the tendency shown in Figure 8. The gradual decrease of TOC represented the gradual disappearance of organic carbon when the MB solution which contained photocatalysts was exposed under UV irradiation. The k values calculated from TOC were 0.0060, 0.0118, 0.0063, 0.0087, 0.0158, or 0.0100 min −1 with P25, Pt50@P25, Pt25@Ti@C, Pt50@Ti@C, or Pt100@Ti@C after UV irradiation for 60 min. Interestingly, the reason why the efficiency of Pt@TiO2@CNTs with Pt particles of 2.41 ± 0.27 nm is higher than that of 3.90 ± 0.37 nm deserves further investigation. As a whole, effective electron-hole separation is related to the loading, size, and distribution of Pt particles on TiO2 [34,35]. From the view of photocatalytic reaction, Pt particles of several nanometers can enhance activity of photodegradation. However, larger Pt particles may provide more recombination sites for photogenerated electrons and holes [35,36]. In addition, not only surface reactive site on TiO2 but also the light absorption of TiO2 are reduced by the too high Pt loading [34]. Therefore, optimum size, loading, and uniform distribution of Pt particles on TiO2 are necessary to obtain effective electron-hole separation and to achieve a higher photocatalytic activity. Pt particles with 0.2 nm different size could particularly increase the efficiency of photoreduction of CO2 to CH4 [37]. The conversion rate increases almost twice within only 0.2 nm different of Pt particles. This implies that a 1-2 nm different particle size in this study may make a big difference in the MB degradation. Based on quantum confinement effect, decreased particle size of Pt has larger energy band gap [37]. Pt particles have higher energy band separation and prevent electron transfer from TiO2 conduction band to Pt when the particles are extremely small. However, electrons can easily transfer to Pt particles with a suitable energy level lower than −4.4 eV when the particle size of Pt increases [37], as shown in Figure 10. If the size becomes too larger, the behavior of Pt particles is more like bulk Pt due to energy level narrowing, which increases the recombination rate of electrons and holes [36]. To sum up, the Pt particles with 2.41 ± 0.27 nm reconcile the quantum confinement effect with suitable energy band gap and avoid the behavior of bulk Pt as recombination center of electrons and holes. This presumably explains that the sample with Pt particles of 2.41 ± 0.27 nm shows the highest efficiency for MB degradation. Table 1 summarizes the particle sizes of Pt for the highest reaction rate of some applications from the literature. The optimum particle size of Pt in this study is 2.41 ± 0.27 nm for degradation of MB, compared to 1.9, 2, 1.75, 2, and 2.2 nm for applications of C2H4 oxidation, glycerol reforming, water splitting, and CO oxidation by using different process for deposition of Pt, respectively. The all show an optimum particle size close to 2 nm. Therefore, for the Pt@TiO2@CNTs structure with the size of 2.41 ± 0.27 nm of Pt for superior efficiency of MB degradation is probably because of its high area and appropriate diffusion length for electrons and holes.
Besides, it is well know the activity of photocatalyst can be demonstrated by the PEC analysis, as shown in Figure 11. In comparison to pure P25 and TiO2@CNTs, the open circuit potential (Eocp) of the Pt@TiO2@CNTs with different particle sizes of Pt are all −951 mV (versus Ag/AgCl) with negatively shifted 130 mV. The lower open circuit potential implies that Pt@TiO2@CNTs has less recombination sites than that of TiO2@CNTs and P25, which results in higher photocurrent density Interestingly, the reason why the efficiency of Pt@TiO 2 @CNTs with Pt particles of 2.41 ± 0.27 nm is higher than that of 3.90 ± 0.37 nm deserves further investigation. As a whole, effective electron-hole separation is related to the loading, size, and distribution of Pt particles on TiO 2 [34,35]. From the view of photocatalytic reaction, Pt particles of several nanometers can enhance activity of photodegradation. However, larger Pt particles may provide more recombination sites for photo-generated electrons and holes [35,36]. In addition, not only surface reactive site on TiO 2 but also the light absorption of TiO 2 are reduced by the too high Pt loading [34]. Therefore, optimum size, loading, and uniform distribution of Pt particles on TiO 2 are necessary to obtain effective electron-hole separation and to achieve a higher photocatalytic activity. Pt particles with 0.2 nm different size could particularly increase the efficiency of photoreduction of CO 2 to CH 4 [37]. The conversion rate increases almost twice within only 0.2 nm different of Pt particles. This implies that a 1-2 nm different particle size in this study may make a big difference in the MB degradation. Based on quantum confinement effect, decreased particle size of Pt has larger energy band gap [37]. Pt particles have higher energy band separation and prevent electron transfer from TiO 2 conduction band to Pt when the particles are extremely small. However, electrons can easily transfer to Pt particles with a suitable energy level lower than −4.4 eV when the particle size of Pt increases [37], as shown in Figure 10. If the size becomes too larger, the behavior of Pt particles is more like bulk Pt due to energy level narrowing, which increases the recombination rate of electrons and holes [36]. To sum up, the Pt particles with 2.41 ± 0.27 nm reconcile the quantum confinement effect with suitable energy band gap and avoid the behavior of bulk Pt as recombination center of electrons and holes. This presumably explains that the sample with Pt particles of 2.41 ± 0.27 nm shows the highest efficiency for MB degradation. Table 1 summarizes the particle sizes of Pt for the highest reaction rate of some applications from the literature. The optimum particle size of Pt in this study is 2.41 ± 0.27 nm for degradation of MB, compared to 1.9, 2, 1.75, 2, and 2.2 nm for applications of C 2 H 4 oxidation, glycerol reforming, water splitting, and CO oxidation by using different process for deposition of Pt, respectively. The all show an optimum particle size close to 2 nm. Therefore, for the Pt@TiO 2 @CNTs structure with the size of 2.41 ± 0.27 nm of Pt for superior efficiency of MB degradation is probably because of its high area and appropriate diffusion length for electrons and holes.
Besides, it is well know the activity of photocatalyst can be demonstrated by the PEC analysis, as shown in Figure 11. In comparison to pure P25 and TiO 2 @CNTs, the open circuit potential (E ocp ) of the Pt@TiO 2 @CNTs with different particle sizes of Pt are all −951 mV (versus Ag/AgCl) with negatively shifted 130 mV. The lower open circuit potential implies that Pt@TiO 2 @CNTs has less recombination sites than that of TiO 2 @CNTs and P25, which results in higher photocurrent density and higher photodecomposition rate of MB [38]. The photocurrent of the Pt@TiO 2 @CNTs with Pt particles of 2.41 ± 0.27 nm has higher photocurrent densities than those other samples. The result is consistent to the MB degradation and shows that the optimized particle size of Pt plays an important role for design of noble metal as co-catalyst to improve efficiency of photocatalyst.    particles of 2.41 ± 0.27 nm has higher photocurrent densities than those other samples. The result is consistent to the MB degradation and shows that the optimized particle size of Pt plays an important role for design of noble metal as co-catalyst to improve efficiency of photocatalyst.   Furthermore, the PL spectra are useful to study the efficiency of trapping, immigration, and transfer of charge carriers for understanding the recombination of electrons and holes because recombination of free charge carries result in signal of PL emissions [39]. Figure 12 shows the PL spectra of TiO 2 @CNTs and Pt@TiO 2 @CNTs with various particle size of Pt are all located at around 505 nm. For TiO 2 , the emission wavelength at around 505 nm is the charge transfer from Ti 3+ to the oxygen anion in the TiO 6 8− complex [39], and the recombination rate of electrons and holes is related to the intensity of the PL emission [40,41]. Therefore, the samples with the higher intensity of PL emission could be suggested to the existence of more trapping centers that enhance the recombination rate of electrons and holes. For Pt@TiO 2 @CNTs with Pt particles of 2.41 ± 0.27 nm, the lowest intensity proves that the proper particle size of Pt can suppresses the recombination of electrons and holes in comparison to other samples. The result is well in line with the PEC analysis and MB degradation. Figure 11. Current-voltage characteristics of P25, TiO2@CNTs (Ti@C), and Pt@TiO2@CNTs with different particle size of Pt (Pt25@Ti@C, Pt50@Ti@C, and Pt100@Ti@C).

Conclusions
The formation of TiO 2 @CNTs and Pt@TiO 2 @CNTs structures with various ALD and PEALD cycles was studied in detailed. It demonstrated that the CNTs with nitric acid treatment could generated oxygen-containing functional groups on the surface. The density of TiO 2 particles deposited by ALD is related to the amount of oxygen-containing functional groups on the surface of CNTs. The saturated growth rates of TiO 2 on a-CNTs were 1.5 Å/cycle and 0.4 Å/cycle, which result from substrate-enhanced process and linear process, respectively. Compared to TiO 2 , the saturated growth rate of Pt on TiO 2 @CNTs was linearly 0.39 Å/cycle due to the steric hindrance of Pt precursors. With increasing the ALD cycles, the as-deposited TiO 2 particles all show anatase phase and increase the intensity of crystallinity at 300 • C. The composite with Pt particles of 2.41 ± 0.27 nm has the highest efficiency MB degradation because it reconciles the quantum confinement effect with suitable energy band gap and avoid the behavior of bulk Pt as recombination center of electrons and holes. In comparison to TiO 2 @CNTs, the samples with Pt loading show the relatively negative E ocp and higher photocurrent due to the Pt particles as co-catalyst. The result of PL spectra proved the lower electrons and holes recombination rate with optimum particle size of Pt. For Pt@TiO 2 @CNTs of 2.41 ± 0.27 nm, the higher photocurrent and lowest intensity of PL emission all demonstrate lower recombination rate of electrons and holes in agreement with MB degradation. The key achievements of this study are the precise control of particle sizes of TiO 2 and Pt and quantification of the photocatalytic efficiencies of TiO 2 @CNTs coupled with Pt particles, which provide useful information for further design of Pt@TiO 2 @CNTs structure for catalytic, sensor, and photochemical applications.