Decoration of Vertically Aligned Carbon Nanotubes with Semiconductor Nanoparticles Using Atomic Layer Deposition

Vertically aligned carbon nanotubes (VACNTs or “CNT forest”) were decorated with semiconductor particles (TiO2 and ZnO) by atomic layer deposition (ALD). Both the structure and morphology of the components were systematically studied using scanning (SEM) and high resolution transmission electron microscopy (HRTEM), energy-dispersive X-ray spectroscopy (EDX), Raman spectroscopy, and X-ray diffraction (XRD) methods. Characterization results revealed that the decoration was successful in the whole bulk of VACNTs. The effect of a follow-up heat treatment was also investigated and its effect on the structure was proved. It was attested that atomic layer deposition is a suitable technique for the fabrication of semiconductor/vertically aligned carbon nanotubes composites. Regarding their technological importance, we hope that semiconductor/CNT forest nanocomposites find potential application in the near future.


Introduction
The vertically aligned carbon nanotube (VACNT), a sub-branch of the carbon nanotube family, is a relevant nanotechnology research topic nowadays. Its structure was synthetized by Li et al. for the first time in 1996 [1]. The VACNT are often called carbon nanotube forests in the literature and usually synthetized by applying the catalytic chemical vapor deposition (CCVD) technique [2]. The carbon nanotube forests differ from carbon nanotubes only in terms of orderliness, and CNT Summarizing the literature data for CNT forest nanocomposites, it can be concluded that the deposition of semiconductor materials into the intertubular region of CNT forests is rather challenging. Supposedly, due to poor wettability of the CNT surface, regular impregnation techniques are generally miscarried. In consequence, the aim of this work was to apply atomic layer deposition for the fabrication semiconductor/CNT forest nanocomposites using either Ti or Zn precursor.

Catalyst Preparation
First, iron and cobalt salts (0.888 g Fe(NO 3 ) 3 ·9H 2 O and 0.855 g Co(NO 3 ) 2 ·6H 2 O) were dissolved in 50 cm 3 of absolute ethanol with the catalyst ratio of 2:3 and this catalyst ink had a transition metal concentration of 0.11 M. The effect of aging was avoided, the solution was prepared freshly before the dip-coating method.
In the following step, the catalyst layer on aluminum sheet was prepared by applying the dip-coating technique. Before this, the substrate was washed with distilled water, ethanol, and acetone in order to remove all contaminations (motes, grease spots, etc.), then the sheet was cut into 3 × 2.5 cm sized pieces. Thereafter, the clean substrate was heat treated for 1 h at 400 • C in a static oven, and in this way-according to our former results-a thicker native oxide layer formed on the aluminum sheet which promoted the production of CNT forests during CCVD synthesis [3].
Dip-coating is an easy and well-controllable method to form a catalyst layer on the substrate. The aluminum sheet was submerged in the catalyst ink for 10 s, where the dipping and withdrawal speeds were 200 mm × min −1 . The catalyst layer was built by KSV dip coater LM (KSV Instruments Ltd., Helsinki, Finland). The aluminum sheet with catalyst layer were heat treated again at 400 • C for 1 h in order to stabilize the catalyst on the surface of the substrate.

CCVD Synthesis
The CNT forests were prepared by the CCVD technique. The cut-to-size aluminum sheets were placed into a quartz boat and put in the pre-heated tube furnace at 640 • C (the aluminum melting point is 660 • C). The reaction time was 15 min, and the carbon source was ethylene (70 cm 3 /min), the carrier gas was nitrogen (50 cm 3 /min), the reducing agent was hydrogen (100 cm 3 /min), and the gas feed contained water vapor (30 cm 3 /min) in every experiment.
During the first step of the CCVD synthesis, in the reactor, inert atmosphere was set by circulating nitrogen for 2 min. As the next step, the hydrogen gas valve was opened for 5 min, in order to reduce the catalyst particles. Thereafter, ethylene gas and water vapor were introduced into the system for 15 min. At the end of synthesis all gas flows were closed except for the nitrogen. After rinsing the reactor, the quartz tube was taken out from the furnace. The system was cooled down to room temperature, then the as-prepared samples were removed from the reactor. The structure of as-prepared vertically aligned carbon nanotubes is shown in Figure 1, with an average height of CNT forest of 15.5 µm. A typical specific surface area value for multiwalled carbon nanotubes is approx. 180 m 2 /g [33]. The estimated surface of CNT to be coated was a few cm 2 .

Atomic Layer Deposition
Atomic layer deposition of the metal oxides was carried out at 1 mbar pressure in a Beneq TFS-200-186 ALD thermal reactor (Beneq, Espoo, Finland) equipped with a cross-flow reaction chamber with maximum reaction space diameter of 200 mm and height of 3 mm for fast film processing. TiO2 was deposited by the reaction of TiCl4 and H2O at 300 °C, and ZnO layers were made with (C2H5)2Zn and H2O at 200 °C. For TiO2, 400 cycles were used, whereas for ZnO, 120 were used; one cycle was composed of 0.3 s metal precursor pulse, 3 s N2 purge, 0.3 s H2O pulse and 3 s N2 purge. (Preliminary measurements revealed the sufficiency of even shorter pulse-purge periods; however longer cycles were applied for safety). The schematics of the ALD method are shown in Figure 2. Oxide layers were also grown on glass substrates under the same conditions and the theoretical thickness of the layers was measured by profilometer (AMBIOS XP-1, Crediton, UK); 40 nm for TiO2, and 20 nm for ZnO. The growth speed for TiO2 and ZnO was found to be 0.1 nm/cycle and 0.17 nm/cycle, respectively. The non-uniformity of as-prepared layers was also investigated by profilometric measurements on glass substrates arranged evenly on the entire plate of the reaction chamber with a diameter of 200 mm. Results revealed that non-uniformity on the entire plate was lower than 5% in the relevant thickness range. (Since our sample was much smaller than the plate, we can presume that nonuniformity is negligible in the current system.)

Atomic Layer Deposition
Atomic layer deposition of the metal oxides was carried out at 1 mbar pressure in a Beneq TFS-200-186 ALD thermal reactor (Beneq, Espoo, Finland) equipped with a cross-flow reaction chamber with maximum reaction space diameter of 200 mm and height of 3 mm for fast film processing. TiO 2 was deposited by the reaction of TiCl 4 and H 2 O at 300 • C, and ZnO layers were made with (C 2 H 5 ) 2 Zn and H 2 O at 200 • C. For TiO 2 , 400 cycles were used, whereas for ZnO, 120 were used; one cycle was composed of 0.3 s metal precursor pulse, 3 s N 2 purge, 0.3 s H 2 O pulse and 3 s N 2 purge. (Preliminary measurements revealed the sufficiency of even shorter pulse-purge periods; however longer cycles were applied for safety). The schematics of the ALD method are shown in Figure 2. Oxide layers were also grown on glass substrates under the same conditions and the theoretical thickness of the layers was measured by profilometer (AMBIOS XP-1, Crediton, UK); 40 nm for TiO 2 , and 20 nm for ZnO. The growth speed for TiO 2 and ZnO was found to be 0.1 nm/cycle and 0.17 nm/cycle, respectively. The non-uniformity of as-prepared layers was also investigated by profilometric measurements on glass substrates arranged evenly on the entire plate of the reaction chamber with a diameter of 200 mm. Results revealed that non-uniformity on the entire plate was lower than 5% in the relevant thickness range. (Since our sample was much smaller than the plate, we can presume that non-uniformity is negligible in the current system.)

Atomic Layer Deposition
Atomic layer deposition of the metal oxides was carried out at 1 mbar pressure in a Beneq TFS-200-186 ALD thermal reactor (Beneq, Espoo, Finland) equipped with a cross-flow reaction chamber with maximum reaction space diameter of 200 mm and height of 3 mm for fast film processing. TiO2 was deposited by the reaction of TiCl4 and H2O at 300 °C, and ZnO layers were made with (C2H5)2Zn and H2O at 200 °C. For TiO2, 400 cycles were used, whereas for ZnO, 120 were used; one cycle was composed of 0.3 s metal precursor pulse, 3 s N2 purge, 0.3 s H2O pulse and 3 s N2 purge. (Preliminary measurements revealed the sufficiency of even shorter pulse-purge periods; however longer cycles were applied for safety). The schematics of the ALD method are shown in Figure 2. Oxide layers were also grown on glass substrates under the same conditions and the theoretical thickness of the layers was measured by profilometer (AMBIOS XP-1, Crediton, UK); 40 nm for TiO2, and 20 nm for ZnO. The growth speed for TiO2 and ZnO was found to be 0.1 nm/cycle and 0.17 nm/cycle, respectively. The non-uniformity of as-prepared layers was also investigated by profilometric measurements on glass substrates arranged evenly on the entire plate of the reaction chamber with a diameter of 200 mm. Results revealed that non-uniformity on the entire plate was lower than 5% in the relevant thickness range. (Since our sample was much smaller than the plate, we can presume that nonuniformity is negligible in the current system.)

Characterization of Samples
Transmission electron microscopy (TEM) measurements were carried out using a FEI Tecnai G2 20 X-TWIN (200 keV) type instrument (FEI, Hillsboro, OR, USA), in order to test the quality of the carbon nanotubes. The samples were removed from the Al surface, then put into an Eppendorf vial and a small amount of the CNT was suspended in 1.25 cm 3 absolute ethanol. In the next step, 2-3 droplets were dropped from this suspension onto the surface of a holey carbon grid (Lacey, CF 200, Electron Microscopy Sciences, Hatfield, USA. The TEM images were analyzed by using ImageJ software (Bethesda, MD, USA).
Scanning Electron Microscopy (SEM) analyses were performed with a Hitachi S-4700 Type II FE-SEM (5-15 keV) type instrument (Tokyo, Japan), in order to determine the orientation of the CNT forests. The samples were tilted at a 35 • angle in all cases to investigate the carbon nanotube forests from each side. SEM images were analyzed by using ImageJ software. The above-mentioned instrument was also used to perform the EDX analysis, complemented by a Röntec XFlash Detector 3001 detector (Bruker, Karlsruhe, Germany).
Raman spectra were recorded on a Jobin Yvon Labram Raman instrument with an Olympus BX41 microscope (Tokyo, Japan) using green Nd-YAG laser (λ = 532 nm).
XRD patterns were recorded on a PANanalytical X'Pert Pro MPD X-ray diffractometer (Malvern Panalytical Ltd., Malvern, UK) using Cu Kα radiation.

Results
For the fabrication of metal oxide/CNT forest composites, both ZnO and TiO 2 were deposited onto the surface of carbon nanotubes using the ALD technique. Samples were thoroughly characterized right after deposition, then both ZnO and TiO 2 coated carbon nanotube forests were heat-treated in order to observe the change in the crystal structure of the semiconductor oxides. During heat treatment, inert atmosphere was used because carbon nanotubes burn away in the presence of oxygen atmosphere at 400 • C. The heat treatment was carried out in a tube furnace for 4 h, at 400 • C in argon atmosphere.

Scanning Electron Microscopy Observations
SEM measurements revealed that the deposition of both metal oxides were successful. As can be seen in Figure 2, carbon nanotubes were decorated with nanosized particles in the whole bulk of the CNT forest. Analyzing the SEM images, it was found that the structure of the carbon nanotube forests was unmodified, and the TiO 2 and ZnO were located on the carbon nanotube surface ( Figure 3). As expected, no discernible changes were identified between the SEM images of the non-heat-treated and heat-treated samples. Therefore, the samples were investigated with EDX measurements too.
The average composition of the samples is shown in Table 1. While the pristine CNT forest was composed of nearly pure carbon, composite samples contained further elements. Besides the covering oxides (TiO 2 and ZnO), traces of iron, cobalt and chlorine could be found in certain samples. The former two elements came from the catalyst of CNT growth, whilst, in the case of the TiO 2 /CNT, the chlorine was the residue from the TiCl 4 precursor used during ALD. The presence of the aluminum substrate was also observable in these data; however, values were eliminated for better comparability. According to EDX results, more ZnO was deposited than TiO 2 .
Interestingly, the relative amount of carbon compared to metal oxides (excluding aluminum support) became significantly lower after heat treatment, which may be due to the crystallization of TiO 2 . As was already mentioned, albeit the treatment was implemented under oxygen-free circumstances, a significant deficit of carbon was found compared to both Zn and Ti content.  The average composition of the samples is shown in Table 1. While the pristine CNT forest was composed of nearly pure carbon, composite samples contained further elements. Besides the covering oxides (TiO2 and ZnO), traces of iron, cobalt and chlorine could be found in certain samples. The former two elements came from the catalyst of CNT growth, whilst, in the case of the TiO2/CNT, the chlorine was the residue from the TiCl4 precursor used during ALD. The presence of the aluminum substrate was also observable in these data; however, values were eliminated for better comparability. According to EDX results, more ZnO was deposited than TiO2.
Interestingly, the relative amount of carbon compared to metal oxides (excluding aluminum support) became significantly lower after heat treatment, which may be due to the crystallization of TiO2. As was already mentioned, albeit the treatment was implemented under oxygen-free circumstances, a significant deficit of carbon was found compared to both Zn and Ti content.

Raman Spectroscopy Results
Raman spectroscopy measurements were carried out in order to determine the Raman shift as well as to investigate the effect of TiO2 and ZnO on the Raman spectra of CNT forest.

Raman Spectroscopy Results
Raman spectroscopy measurements were carried out in order to determine the Raman shift as well as to investigate the effect of TiO 2 and ZnO on the Raman spectra of CNT forest.
On the Raman spectra of each sample (Figure 4), the D (~1340 cm −1 ) and G bands (~1580 cm −1 ) of the carbon are present, which come from the carbon nanotubes. The characteristic peaks of anatase can be seen on the spectrum of TiO 2 /CNT at 141, 400, 516 and 637 cm −1 [56]. In the case of ZnO/CNT, only two small peaks of the ZnO are visible, at 380 and 438 cm −1 [57] I D /I G ratios of D and G peaks from carbon nanotubes were calculated to provide information on their graphitization ( Table 2). The ratio of the D and G peaks intensities (I D /I G ) changed during the deposition of the metal oxides (Table 2) which makes the chemical bond probable between Ti/Zn and CNT [58,59]. It was proved earlier that there is a strong correlation between the adherence of metal oxide particles to CNT surfaces and the density of defect sites in the carbon nanotube forest [60] Surprisingly, the heat-treated composite samples show lower I D /I G ratios, thus fewer defect sites again. For example, in the case of ZnO/CNT sample I D /I G was 1.08, an even better ratio than that of a pristine CNT forest and suggests only a few defect sites in the carbon nanotube. In spite of the presence of inert atmosphere during heat treatment, the oxygen in the sample could oxidize amorphous carbon and outer layers (which always contain more defects) of CNTs, resulting in increased graphitization values. This finding is in good agreement with observations done during the interpretation of EDX results.  On the Raman spectra of each sample (Figure 4.), the D (~1340 cm −1 ) and G bands (~1580 cm −1 ) of the carbon are present, which come from the carbon nanotubes. The characteristic peaks of anatase can be seen on the spectrum of TiO2/CNT at 141, 400, 516 and 637 cm −1 [56]. In the case of ZnO/CNT, only two small peaks of the ZnO are visible, at 380 and 438 cm −1 [57] ID/IG ratios of D and G peaks from carbon nanotubes were calculated to provide information on their graphitization ( Table 2). The ratio of the D and G peaks intensities (ID/IG) changed during the deposition of the metal oxides (Table  2) which makes the chemical bond probable between Ti/Zn and CNT [58,59]. It was proved earlier that there is a strong correlation between the adherence of metal oxide particles to CNT surfaces and the density of defect sites in the carbon nanotube forest [60] Surprisingly, the heat-treated composite samples show lower ID/IG ratios, thus fewer defect sites again. For example, in the case of ZnO/CNT sample ID/IG was 1.08, an even better ratio than that of a pristine CNT forest and suggests only a few defect sites in the carbon nanotube. In spite of the presence of inert atmosphere during heat treatment, the oxygen in the sample could oxidize amorphous carbon and outer layers (which always contain

X-ray Diffraction Results
In Figure 5, the diffraction peaks of TiO 2 could not be observed during the investigation of TiO 2 /CNT forest, because it was present only in a small amount on the surface of carbon nanotubes (see also the EDX data). The ZnO/CNT forest showed the peaks for the hexagonal ZnO (ICDD: 01-080-4199), and one peak (200) of the aluminum (ICDD: 00-004-0787) was present as well. A small diffraction peak is visible on the composites around 24 • 2θ, which comes from the silicon sample holder.
In Figure 5, the diffraction peaks of TiO2 could not be observed during the investigation of TiO2/CNT forest, because it was present only in a small amount on the surface of carbon nanotubes (see also the EDX data). The ZnO/CNT forest showed the peaks for the hexagonal ZnO (ICDD: 01-080-4199), and one peak (200) of the aluminum (ICDD: 00-004-0787) was present as well. A small diffraction peak is visible on the composites around 24° 2θ, which comes from the silicon sample holder.

Transmission Electron Microscopy Observations
TEM measurements also confirmed that the metal oxides were successfully anchored onto the surface of carbon nanotubes composing forests. These measurements provided information about the quality of both carbon nanotubes and metal oxides as well.

Transmission Electron Microscopy Observations
TEM measurements also confirmed that the metal oxides were successfully anchored onto the surface of carbon nanotubes composing forests. These measurements provided information about the quality of both carbon nanotubes and metal oxides as well.
From the TEM images ( Figure 6) of composite samples, it can be concluded that the metal oxides truly enfold the surface of carbon nanotubes in both TiO 2 /CNT and ZnO/CNT composites. HR-TEM images also revealed that the number of walls in carbon nanotubes were 4-5 on average and their diameter varied between 5-6 nm, while the average diameters of the TiO 2 and ZnO particles were 25 nm and 30 nm, respectively.
In the case of ZnO/CNT a more even coverage can be observed, compared to TiO 2 /CNT after heat treatment, but both metal oxides provided appreciable decoration on the CNT forest. From HR-TEM images, the interplanar lattice spacing was also calculated for metal oxides. It was observed that the lattice spacing did not change after heat treatment in both cases, thus no significant modification of crystal phases occurred. This value of 0.37 nm corresponds to the (001) facet of anatase in sample TiO 2 /CNT. At first glance, this value seems to be changed in the case of ZnO/CNT. Before heat treatment the lattice spacing was found to be 0.28 nm for ZnO, however, after heat treatment the lattice spacing was 0.32 nm based on HRTEM image. Literature data confirmed that both values belong to different facets ((0001) and (0110)) of a regular wurtzite hexagonal-structure ZnO particle [61,62]. It was also detected that the particles directly on the surface showed lower crystallinity compared to other quasi separated crystals. From the TEM images ( Figure 6) of composite samples, it can be concluded that the metal oxides truly enfold the surface of carbon nanotubes in both TiO2/CNT and ZnO/CNT composites. HR-TEM images also revealed that the number of walls in carbon nanotubes were 4-5 on average and their diameter varied between 5-6 nm, while the average diameters of the TiO2 and ZnO particles were 25 nm and 30 nm, respectively.
In the case of ZnO/CNT a more even coverage can be observed, compared to TiO2/CNT after heat treatment, but both metal oxides provided appreciable decoration on the CNT forest. From HR-TEM images, the interplanar lattice spacing was also calculated for metal oxides. It was observed that the lattice spacing did not change after heat treatment in both cases, thus no significant modification of crystal phases occurred. This value of 0.37 nm corresponds to the (001) facet of anatase in sample TiO2/CNT. At first glance, this value seems to be changed in the case of ZnO/CNT. Before heat treatment the lattice spacing was found to be 0.28 nm for ZnO, however, after heat treatment the lattice spacing was 0.32 nm based on HRTEM image. Literature data confirmed that both values belong to different facets ((0001) and (0110)) of a regular wurtzite hexagonal-structure ZnO particle [61,62]. It was also detected that the particles directly on the surface showed lower crystallinity compared to other quasi separated crystals.

Conclusions
Surveying the relevant literature, it became obvious that the deposition of different materials into the interior region of CNT forests is a real challenge. Probably due to wettability problems, conventional impregnation techniques are generally unavailable for this purpose [63]. Therefore, in this work, we applied atomic layer deposition for the fabrication semiconductor/CNT forest

Conclusions
Surveying the relevant literature, it became obvious that the deposition of different materials into the interior region of CNT forests is a real challenge. Probably due to wettability problems, conventional impregnation techniques are generally unavailable for this purpose [63]. Therefore, in this work, we applied atomic layer deposition for the fabrication semiconductor/CNT forest nanocomposites using either Ti or Zn precursor. The resulting composite materials were characterized by various methods such as SEM, XRD, HRTEM, XRD, and Raman spectroscopy. Characterization results revealed that the decoration of CNT forest with both Ti 2 O and ZnO was successful in the whole bulk of VACNTs. Knowing the properties of applied semiconductors, a follow-up heat treatment was applied, and structural change was detected. Comparing the characterization results, it was found that treatment at 400 • C caused major alteration in the carbon nanotube but not in the semiconductor properties. Regarding the technological importance of these semiconductor/CNT forest nanocomposites, the authors hope that their work contributes to the application of these materials in the field of photocatalysis, supercapacitors or other applications soon.

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