Effects of Anodic Aluminum Oxide Substrate Pore Geometry on the Gas-Phase Photocatalytic Activity of ZnO/Al2O3 Composites Prepared by Atomic Layer Deposition

We report on the photocatalytic activity of ZnO layers deposited by atomic layer deposition on a porous anodic aluminum oxide substrate with hexagonal pore symmetry and varied pore dimensions. ZnO/Al2O3 composites were prepared with pore diameters in the range 93–134 nm and interpore distance in the range 185–286 nm, and their photocatalytic activity was measured for gas-phase photocatalytic oxidation of acetaldehyde at varying UV illumination intensities (0.08–3.94 mW cm−2). The results show that substrates with narrower pore diameters (<115 nm, in the case of this study) have a detrimental effect on the photocatalyst performance, despite their higher effective surface. The results are explained on the basis of limited mass transfer inside the porous structure and can be used as a guideline in the purposeful design of photocatalysts with a nanoporous or nanotubular structure.


Introduction
Porous anodic aluminum oxide (AAO) has a porous structure, leading to a highly developed surface, making it a suitable support in heterogeneous catalysis [1,2]. It is prepared by inexpensive electrochemical routes, which allow for a fine control of the porous structure by adjusting anodization parameters, or varying electrolyte compositions [3]. AAO is also optically transparent in the UV to NIR wavelengths, and it is possible to prepare AAOs with a highly symmetrical 2D or 3D pore arrangement, which can also exhibit photonic crystal properties [4,5] and can be especially advantageous for supported photocatalysts [6]. Photocatalysts, such as TiO 2 and ZnO, supported on AAO have been widely studied and find applications in energy conversion and contaminated water/air remediation [7][8][9].
In this work, we present results on the effects of AAO morphology on the photocatalytic activity of supported ZnO. AAO substrates were prepared with varied pore diameters and interpore distance by adjusting the anodization potential, and ZnO was deposited by atomic layer deposition (ALD). The performance of the as-prepared ZnO/Al 2 O 3 photocatalysts was studied by measuring the photocatalytic oxidation (PCO) of acetaldehyde (CH 3 CHO) in gas phase, and UV-intensity-dependent experiments were carried out to estimate the effects of AAO morphology on the intrinsic photocatalytic performance of ZnO/Al 2 O 3 .

Preparation of AAO Substrates
Al foil (100 µm thick, 99.999% purity, Alfa Aesar, Ward Hill, MA, USA) was annealed for 5 h at 450 • C in order to minimize mechanical stresses and cut to 50 mm squares. The foils were subsequently surface-treated in 4 wt.% NaOH for 1 min at 60-70 • C, followed by a neutralization in 10% H 2 SO 4 for 30 s. A negative dry photoresist was used to completely protect the back side and form a 46 mm diameter circle (16.6 cm 2 ) of exposed aluminum on the front side for selective anodization. The exposed area was polished in 20 wt.% HClO 4 :C 2 H 5 OH at 5 • C for 1.5 min. Anodization was carried out at a temperature of 10 • C in an electrolyte of 5 wt.% H 3 PO 4 . Samples were prepared at four anodization potentials: 60, 80, 100, and 120 V. Two sets of samples were prepared simultaneously at each condition and used for characterization and photocatalytic experiments.

ALD Deposition of ZnO
ALD deposition of ZnO onto AAO was carried out by thermal ALD at 200 • C, employing a Beneq TFS-200 ALD system. Diethylzinc (DEZ, Zn(C 2 H 5 ) 2 ) and deionized water (H 2 O) were used as Zn precursor and oxidant, and N 2 as a carrier and purging gas at a flow of 300 sccm. The precursor pulse duration for DEZ and H 2 O was 0.3 s with 3 s N 2 purging time. A total of 50 ALD cycles (DEZ/purging/H 2 O/purging) were repeated to obtain a ZnO layer with a thickness of about 8 nm, confirmed by ellipsometry on reference Si substrates placed in the ALD reactor. Using this procedure, a conformal ZnO coverage can be achieved with a coverage up to several micrometers within the AAO pores, as confirmed in our previous study [10].
Immediately prior to the photocatalytic experiments, the samples were heat-treated at 500 • C for 1 h (5 • C min −1 heating and cooling rate) in order to improve the crystallinity of the ZnO layer and remove any organic contaminants surface-adsorbed during storage.

Characterization of ZnO/Al 2 O 3 Composites
The morphology of the ZnO/Al 2 O 3 photocatalysts was studied by scanning electron microscopy (SEM-Carl Zeiss Merlin microscope, 3 kV accelerating voltage). Average pore diameters and interpore distance were obtained from SEM images using the ImageJ software [11]. The geometric areas of no less than 100 pore openings were measured for at least two images for each sample (obtained near the center and the edge of the anodized alumina surface). These data were then converted into an equivalent circle diameter using Equation (1): where A is the area of a pore opening. Geometric surface area conversion factors (A e f f ) were also obtained from SEM image data by estimating the pore density per geometric SEM image area and correcting it with the average exposed area inside the pores (based on an AAO thickness of 2 µm and estimated average pore diameter), that is, Equation (2): where S SEM is the geometric area of a SEM image (in µm 2 estimated by calibrating it with the scale marker), n is the total number of pores, d is the AAO layer thickness (2 µm), and A is the area covered by pore openings. Crystallography information was obtained in the range 20-80 • via grazing incidence Xray diffraction (GI-XRD) at 0.5 • angle of incidence and using CuKα radiation (λ = 1.5418 Å) on a Siemens D5000 instrument. Rietveld refinement was applied on diffractograms using the PowderCell software [12] in order to obtain the mean crystallite size and preferential orientation, using the March-Dollase (MD) texturing model [13].
UV-VIS diffuse reflectance spectra were obtained in the range 300 < λ < 800 nm on a PerkinElmer Lambda 900 spectrophotometer, equipped with a 150 mm Spectralon-coated integrating sphere.

Photocatalytic Experiments
Gas-phase PCO experiments were carried in a photocatalytic reactor setup, which has been described in detail elsewhere [14]. Briefly, a flow-mode reaction cell was used to expose the samples to a 100 mL min −1 gas flow containing 5 ppm CH 3 CHO in synthetic air (premixed from 90 ppm CH 3 CHO in N 2 and 20%:80% O 2 /N 2 mixture, Air Liquide, 99.99% purity). Illumination was provided by an array of three high-power UV LEDs, λ = 365 ± 5 nm (P8D136, Seoul Semiconductors, South Korea). UV illumination intensity (I UV ) was in the range 0.08-3.94 mW cm −2 at the photocatalyst surface level, as measured by a calibrated thermopile radiometer (Ophir, Israel). The CH 3 CHO concentration profile (C CH 3 CHO ) was monitored in situ by a chemoresistive gas sensor (HS-130AS, Sencera Co., Taiwan). Each experiment consisted of 6-11 steps of 20 min UV illumination periods, bracketed by 15 min dark periods in order to monitor the baseline signal and correct for sensor drift. The CH 3 CHO PCO rate constant, r CH 3 CHO , was derived from C CH 3 CHO using Equation (3): where S is the geometric photocatalytic area (16.6 cm 2 ), and F is the CH 3 CHO mass flow (0.002 nmol min −1 for a flow of 100 mL min −1 5 ppm CH 3 CHO). An example of C CH 3 CHO and corresponding r CH 3 CHO profiles during an experimental sequence is shown in Figure 1.
Symmetry 2021, 13, x FOR PEER REVIEW 3 of 9 Å) on a Siemens D5000 instrument. Rietveld refinement was applied on diffractograms using the PowderCell software [12] in order to obtain the mean crystallite size and preferential orientation, using the March-Dollase (MD) texturing model [13]. UV-VIS diffuse reflectance spectra were obtained in the range 300 < λ < 800 nm on a PerkinElmer Lambda 900 spectrophotometer, equipped with a 150 mm Spectralon-coated integrating sphere.

Photocatalytic Experiments
Gas-phase PCO experiments were carried in a photocatalytic reactor setup, which has been described in detail elsewhere [14]. Briefly, a flow-mode reaction cell was used to expose the samples to a 100 mL min −1 gas flow containing 5 ppm CH3CHO in synthetic air (premixed from 90 ppm CH3CHO in N2 and 20%:80% O2/N2 mixture, Air Liquide, 99.99% purity). Illumination was provided by an array of three high-power UV LEDs, λ = 365 ± 5 nm (P8D136, Seoul Semiconductors, South Korea). UV illumination intensity ( ) was in the range 0.08-3.94 mW cm −2 at the photocatalyst surface level, as measured by a calibrated thermopile radiometer (Ophir, Israel). The CH3CHO concentration profile ( ) was monitored in situ by a chemoresistive gas sensor (HS-130AS, Sencera Co., Taiwan). Each experiment consisted of 6-11 steps of 20 min UV illumination periods, bracketed by 15 min dark periods in order to monitor the baseline signal and correct for sensor drift. The CH3CHO PCO rate constant, , was derived from using Equation (3): ] ( 3) where is the geometric photocatalytic area (16.6 cm 2 ), and is the CH3CHO mass flow (0.002 nmol min −1 for a flow of 100 mL min −1 5 ppm CH3CHO). An example of and corresponding profiles during an experimental sequence is shown in Figure  1.

Characterization of ZnO/Al2O3
Figure 2 shows top-view and cross-section SEM images of ZnO/Al2O3 with AAO anodized at different potentials. The AAO layer is about 2 μm thick, with a similar morphology in all samples. Increased anodization voltage increases both the average pore diameter (from 93 to 134 nm) and the average interpore distance (from 185 to 286 nm). Histograms of the AAO porediameter distributions at different anodization voltages are shown in Figure 3. Increasing the anodization potential, the uniformity of the pore-size distribution is altered and the mean diameter shifts towards wider pores. This observation is typical for AAO anodization processes, where the optimal self-ordering condition is achieved in a narrow potential range, depending on the electrolyte used and the operating temperature [15].  The AAO layer is about 2 µm thick, with a similar morphology in all samples. Increased anodization voltage increases both the average pore diameter (from 93 to 134 nm) and the average interpore distance (from 185 to 286 nm). Histograms of the AAO porediameter distributions at different anodization voltages are shown in Figure 3. Increasing the anodization potential, the uniformity of the pore-size distribution is altered and the mean diameter shifts towards wider pores. This observation is typical for AAO anodization processes, where the optimal self-ordering condition is achieved in a narrow potential range, depending on the electrolyte used and the operating temperature [15].

Characterization of ZnO/Al2O3
Figure 2 shows top-view and cross-section SEM images of ZnO/Al2O3 with AAO anodized at different potentials. The AAO layer is about 2 μm thick, with a similar morphology in all samples. Increased anodization voltage increases both the average pore diameter (from 93 to 134 nm) and the average interpore distance (from 185 to 286 nm). Histograms of the AAO porediameter distributions at different anodization voltages are shown in Figure 3. Increasing the anodization potential, the uniformity of the pore-size distribution is altered and the mean diameter shifts towards wider pores. This observation is typical for AAO anodization processes, where the optimal self-ordering condition is achieved in a narrow potential range, depending on the electrolyte used and the operating temperature [15].     Given the thickness of the ALD layer is only 8 nm, the larger mean crystallite size may indicate either an extended in-plane recrystallization of ZnO within the ALD layer or the formation of elongated crystals, typical for ZnO. The <100> texturing observed in the GI-XRD diffractogram is not typical for ALD-deposited ZnO, which usually grows in the <002> direction. It has however been observed by other authors employing the GI-XRD technique [16,17].
UV-VIS reflectance spectra for the uncoated AAO substrates, prepared at different anodization potentials, are shown in Figure 5a. The spectra show interference fringes, which are particularly pronounced in the 300-800 nm range for all AAO substrates, except the one prepared at an anodization potential of 120 V. The shift of the interference pattern towards longer wavelengths upon increasing the anodization potential is due to the increased effective refractive index of the AAO layer. The presence of ZnO in the ZnO/Al2O3 composites is indicated by the drop in reflectance at λ < 400 nm (Figure 5b), which can be used to calculate the optical bandgap ( ) of the ALD-deposited ZnO. The reflectance data were converted using the Kubelka-Munk equation:  Given the thickness of the ALD layer is only 8 nm, the larger mean crystallite size may indicate either an extended in-plane recrystallization of ZnO within the ALD layer or the formation of elongated crystals, typical for ZnO. The <100> texturing observed in the GI-XRD diffractogram is not typical for ALD-deposited ZnO, which usually grows in the <002> direction. It has however been observed by other authors employing the GI-XRD technique [16,17].
UV-VIS reflectance spectra for the uncoated AAO substrates, prepared at different anodization potentials, are shown in Figure 5a. The spectra show interference fringes, which are particularly pronounced in the 300-800 nm range for all AAO substrates, except the one prepared at an anodization potential of 120 V. The shift of the interference pattern towards longer wavelengths upon increasing the anodization potential is due to the increased effective refractive index of the AAO layer. The presence of ZnO in the ZnO/Al 2 O 3 composites is indicated by the drop in reflectance at λ < 400 nm (Figure 5b), which can be used to calculate the optical bandgap (E g ) of the ALD-deposited ZnO.  Given the thickness of the ALD layer is only 8 nm, the larger mean crystallite size may indicate either an extended in-plane recrystallization of ZnO within the ALD layer or the formation of elongated crystals, typical for ZnO. The <100> texturing observed in the GI-XRD diffractogram is not typical for ALD-deposited ZnO, which usually grows in the <002> direction. It has however been observed by other authors employing the GI-XRD technique [16,17].
UV-VIS reflectance spectra for the uncoated AAO substrates, prepared at different anodization potentials, are shown in Figure 5a. The spectra show interference fringes, which are particularly pronounced in the 300-800 nm range for all AAO substrates, except the one prepared at an anodization potential of 120 V. The shift of the interference pattern towards longer wavelengths upon increasing the anodization potential is due to the increased effective refractive index of the AAO layer. The presence of ZnO in the ZnO/Al2O3 composites is indicated by the drop in reflectance at λ < 400 nm (Figure 5b), which can be used to calculate the optical bandgap ( ) of the ALD-deposited ZnO. The reflectance data were converted using the Kubelka-Munk equation: The reflectance data were converted using the Kubelka-Munk equation: where F(R) is the Kubelka-Munk function, which can be approximated to the effective absorption coefficient (F(R) ∝ α). Then, E g for ZnO, a direct bandgap semiconductor, can be estimated by Equation (5): The cross section of (αhυ) 2 versus the photon energy, as shown in Figure 5c, gives the E g values listed in Table 1 for all four ZnO/Al 2 O 3 composites. The average E g of the ZnO in all ZnO/Al 2 O 3 photocatalysts was found to be about 3.25 eV, which is lower than the value for bulk ZnO (3.3 eV), but in agreement with literature data for ultrathin ALD ZnO layers [18].  Table 1 lists r CH 3 CHO rates for ZnO/Al 2 O 3 as a function of I UV . The r CH 3 CHO dependence of I UV was modelled by least square fitting of the data to Equation (6), that is:

Photocatalytic Activity
where A is the I UV -independent rate constant, and b is the I UV reaction order of the PCO process. Theoretically, the PCO rate increases linearly at low intensities (b ≈ 1), with a square root order at intermediate intensities (b ≈ 0.5), and eventually becomes independent of I UV at high intensities (b ≈ 0), where r CH 3 CHO is limited by the number of available reaction sites [19]. The results shown in Table 1 reveal that ZnO/Al 2 O 3 photocatalysts with AAO substrates anodized at >60 V perform markedly better despite their lower effective surface area. However, there is a shift from square root to a linear PCO reaction rate dependence on I UV for samples with larger pores and a sparser pore structure, which we tentatively attribute to limited UV light penetration and mass transfer limitations. The optimal apparent performance at low-to-medium I UV (0.1-1 mW cm −2 ) was observed for ZnO/Al 2 O 3 prepared at 80 V, as reflected by the highest I UV -independent rate constant (A) for this sample. A similar effect of optimal pore size related to maximum PCO efficiency was observed Symmetry 2021, 13, 1456 7 of 9 experimentally [20] and modelled theoretically by Liu et al. [21] for TiO 2 -nanotube-based photocatalysts, which have a similar geometry, and attributed to O 2 diffusion limitations.
It is reasonable to assume that the same effect can explain the observed reactivity in our case. Figure 6a shows the UV dependence of r CH 3 CHO rates for all samples and the resulting fits from Equation (6). It can be noticed that the ZnO photocatalysts, deposited on an AAO layer with larger average pore diameters (obtained at 100 and 120 V anodization potentials), exhibit an almost linear increase in the CH 3 CHO PCO rates (b > 0.8) in the studied I UV range, while the rates of the ZnO/Al 2 O 3 samples with a narrower pore structure levels off (b~0.8) in the same UV intensity illumination range. The difference is even more pronounced when the r CH 3 CHO values are corrected with geometric-to-effective area correction factors (A e f f ) obtained from SEM images by Equation (2) for each AAO substrate (Figure 6b), where it is clearly seen that catalysts based on substrates with wider AAO pores have higher intrinsic activity. These observations can be explained through diffusion limitations in the narrower pore structure [21,22]. Assuming that the ZnO coverage is reasonably uniform within the pores, as shown in our previous study [10], improving reactant diffusion and light penetration would lead to a higher number of catalytic sites utilized and hence a higher photocatalytic activity. However, any further increase in pore diameter is expected to have a detrimental effect on PCO rates due to the decrease in effective surface area (reducing the number of active sites). Thus, it is possible to purposefully design photocatalysts with optimal activity, which are based on AAOs and easy to fabricate and reproduce.
strates anodized at >60 V perform markedly better despite their lower effective surface area. However, there is a shift from square root to a linear PCO reaction rate dependence on for samples with larger pores and a sparser pore structure, which we tentatively attribute to limited UV light penetration and mass transfer limitations. The optimal apparent performance at low-to-medium (0.1-1 mW cm −2 ) was observed for ZnO/Al2O3 prepared at 80 V, as reflected by the highest -independent rate constant ( ) for this sample. A similar effect of optimal pore size related to maximum PCO efficiency was observed experimentally [20] and modelled theoretically by Liu et al. [21] for TiO2-nanotubebased photocatalysts, which have a similar geometry, and attributed to O2 diffusion limitations.
It is reasonable to assume that the same effect can explain the observed reactivity in our case. Figure 6a shows the UV dependence of rates for all samples and the resulting fits from Equation (6). It can be noticed that the ZnO photocatalysts, deposited on an AAO layer with larger average pore diameters (obtained at 100 and 120 V anodization potentials), exhibit an almost linear increase in the CH3CHO PCO rates ( 0.8) in the studied range, while the rates of the ZnO/Al2O3 samples with a narrower pore structure levels off ( ~ 0.8) in the same UV intensity illumination range. The difference is even more pronounced when the values are corrected with geometric-to-effective area correction factors ( ) obtained from SEM images by Equation (2) for each AAO substrate (Figure 6b), where it is clearly seen that catalysts based on substrates with wider AAO pores have higher intrinsic activity. These observations can be explained through diffusion limitations in the narrower pore structure [21,22]. Assuming that the ZnO coverage is reasonably uniform within the pores, as shown in our previous study [10], improving reactant diffusion and light penetration would lead to a higher number of catalytic sites utilized and hence a higher photocatalytic activity. However, any further increase in pore diameter is expected to have a detrimental effect on PCO rates due to the decrease in effective surface area (reducing the number of active sites). Thus, it is possible to purposefully design photocatalysts with optimal activity, which are based on AAOs and easy to fabricate and reproduce.

Conclusions
Photocatalytic ZnO/Al2O3 was prepared by ALD deposition of thin ZnO layers on a three-dimensional hexagonal periodic AAO support scaffold structure. The effect of AAO geometry on the gas-phase photocatalytic oxidation rates of CH3CHO was studied as a

Conclusions
Photocatalytic ZnO/Al 2 O 3 was prepared by ALD deposition of thin ZnO layers on a three-dimensional hexagonal periodic AAO support scaffold structure. The effect of AAO geometry on the gas-phase photocatalytic oxidation rates of CH 3 CHO was studied as a function of UV illumination intensity. The results show that increasing the pore diameter (93 to 134 nm) and distance (185 to 286 nm) shifts the UV intensity dependence of the PCO reaction rate from a square root to a linear dependence, thus indicating better utilization of UV light energy in the studied region (0.1-3.9 mW cm −2 ). Optimal photocatalytic activity was found for ZnO/Al 2 O 3 photocatalysts with AAO supports obtained at an intermediate anodization voltage (80-100 V). Specifically, the ZnO/Al 2 O 3 photocatalyst with a 115 nm pore diameter exhibited the highest absolute reaction rate at a low-to-medium illumination intensity, 0.1-1 mW cm −2 , which, however, became saturated at UV illumination intensities >2 mW cm −2 . In this region, the ZnO/Al 2 O 3 photocatalysts with larger pores exhibited higher reactivity to CH 3 CHO photo-oxidation, regardless of their lower overall effective area. The study demonstrates the feasibility of AAO-based substrates to be used as a testbed for investigating the intrinsic fundamental effects of a porous substrate structure on the photocatalytic activity of ALD-deposited supported photocatalysts, and results obtained in this manner may be used as guidelines for the purposeful design of photocatalysts with a porous structure and tubular geometry.