Optimizing the Operational Parameters Used in Preparing Anodic Aluminum Oxide (AAO) Templates from Low-purity Aluminum

Anodized aluminum oxide (AAO) templates with well-defined geometric pore features and uniform nanoporous arrays were successfully obtained from recyclable low-cost, low-purity (99.16%) aluminum via a two-step anodizing process. The templates were characterized by scanning electron microscopy (SEM) and operating parameters were optimized using a Box-Behnken experimental design. The best conditions were found for an oxalic acid electrolyte concentration of 0.6 M, an anodizing potential of 46 V, and a bath temperature of 10 °C. This process resulted in an averaged regularity ratio (ravg) value of 2.66 which compares very favorably with previously reported values obtained from higher-purity aluminum (99.5%). Templates developed from low-purity aluminum are more suitable for high-volume industrial applications where there is a practical trade-off between cost and the quality of the geometric pores.

AAO geometric pore features were comprehensively described for the first time by F. Keller et al. in 1953. 1 The authors studied how the anodizing potential, temperature, and different electrolytes used in the anodizing process affected the anodic oxide coating features.Since then, multiple variations of the anodization process have been developed to create uniform and reproducible nanostructures for different applications, such as purification, catalysis and protein gating as reviewed by Huang et al. 2 Furthermore, Zaraska et al. 3 have reported metallic nanowire structures being used for nanoscale electronic components, opto-electronic devices, and high-density magnetic memories.
While the ability to fine-tune AAO geometric pore features makes them attractive for many nanostructured material applications, they are typically produced using expensive high-purity aluminum which makes them unsuitable for many high-volume, cost-sensitive applications.This paper examines the ability to fine tune the geometric pore features of AAOs produced from low purity aluminum since there is a considerable difference between the cost of high-purity and low-purity aluminum.For example, 99.999% aluminum sheets are typically five times more expensive than samesized low-purity 99.16% aluminum sheets, as quoted on Goodfellow. 4 Fine tuning and optimizing the individual pore features of low-purity AAOs facilitates the use of AAOs in a wider range of high-volume industrial applications.
In comparison to high-purity aluminum, there have been fewer studies conducted on fine-tuning the anodization parameters for lowpurity aluminum.Zaraska et al., 5 Michalska-Domańska et al., 6 and Tsyntsaru 7 are among the authors who investigated low purity aluminum.Zaraska et al. 5 produced 4cm 2 AAO templates from AA1050 (99.5%) aluminum and described their application in the production of Pd nanowires.Michalska-Domańska et al. 6 and Tsyntsaru 7 showed how a relatively high regularity of pores can be achieved with lower purity aluminum (Michalska-Domańska et al. 6 reported results for AA1050 while Tsyntsaru 7 compared results for AA1050, AA6082, and AA6060 alloys).The alloying elements in the low purity aluminum foils cause thermal stresses and for this reason Tsyntsaru 7 performed a second anodization step which produced wellordered nanopores.The author remarked on the importance of strictly controlling the operating conditions when working with low-purity aluminum foils.
The main operating conditions of the electrolytic cell in the anodizing process are listed below: i) electrolyte solution (oxalic acid) concentration (C) ii) electrolyte bath temperature (T) iii) anodizing potential (P)   In this study, AAOs were prepared from 99.16% low-purity aluminum foils.Oxalic acid was chosen as the electrolyte based on Reddy et al. 8 research of different acids which concluded the oxalic acid was the one that produced AAOs with the highest regularity ratio.The test ranges of the three main operational parameters were chosen based on conventional Mild Anodizing procedures 9 and the documented review of anodizing parameters for the production of AAOs by Sulka. 10 The temperature range also took into account the desire to use a range close to room-temperature to facilitate the lower-cost requirements in high-volume production applications.
Three settings for each of the three operational parameters were chosen and inputted to a BBD Box-Behnken Design (BBD) using Minitab 18 software.BBD is one type of Response Surface Methodology (RSM) which has been shown to produce more accurate models than the one variable at a time (OVAT) models 11 that have been commonly used in papers related to AAO research.RSM is also an efficient statistical technique as Azami and Omidkhah 12 have outlined.Of the two main RSM designs, the Central Composite design (CCD) and the BBD, the BBD was chosen for this research as it requires fewer test runs while still producing good-quality 2nd order models.For the three settings of each of the operating parameters entered into Minitab, the BBD proposed 15 test runs, each using a different combination of parameter settings.Each run was applied to individual 1 cm 2 areas of aluminum foil and the responses of the five geometrical pore features were analyzed using SEM images.The results from the analysis were used as feedback to z E-mail: lina.castro@udea.edu.coECS Advances, 2023 2 042501 the BBD which then produced 2nd order polynomial model equations to estimate each of the five geometric pore features listed below for any combination of operating parameter values.i) averaged regularity ratio (r avg ) ii) pore diameter (D p ) iii) eccentricity (Ecc) iv) aspect ratio (AR) v) interpore distance (D i ) r avg , was the pore feature of particular interest in this study because it is a measure of the interpore distance uniformity.A higher value of r avg indicates better regularity of pores which is critical for producing one dimensional nanostructures with good-quality physical features.The optimal operating parameter values that were determined by the BBD in this work to produce the maximum r avg for the 1 cm 2 samples were applied to produce larger 78 cm 2 AAOs for use in the production of one-dimensional nanostructures.The 78 cm 2 AAOs achieved porous templates with highly ordered cell configurations from the low-purity aluminum.This indicates the BBD produced a good-quality model equation for r avg .
The main aim of this work was to identify the optimal operating conditions to achieve AAOs with high r avg from 99.16% low purity aluminum.This will facilitate the use of the much lower-cost 99.16% aluminum in high-volume industrial applications.

Materials and Methodology
Commercial aluminum foil with an advertised purity of 99.16%, and 0.25 mm thickness, was purchased for use from Goodfellow.Its main elements composition was characterized by X-ray Fluorescence analysis using a Thermo OPTIM'X spectrometer and the weight percentages of each of the main elements in the foil were recorded in Table I.

Box-Behnken Design (BBD)
The sets of values used for each parameter input were as follows: i) C = 0.3 M, 0.45 M, and 0.6 M ii) T = 10 °C, 15 °C, and 20 °C iii) P = 30 V, 45 V, and 60 V Given these parameter values, the BBD returned fifteen different combinations of operating parameter values to run, as outlined in Table II.
Each test run in Table II was applied to individual 1 cm 2 areas of aluminum foil to produce fifteen AAO templates for analysis.Images of each AAO sample were taken using a Scanning Electron Microscope (SEM).MIPAR Image Analysis software pre-processed these SEM images to reduced noise and enhanced contrast.The MIPAR software then segmented the images to produce Grain Analysis images where the pore edges were defined to a high precision.
The Grain Analysis images were used by the MIPAR software tools to measure the pore geometric properties of D p , Ecc, AR, and D i .D p was measured using a pore's maximum Feret diameter, (defined in the MIPAR software as the largest line length that fits within a pore).Ecc describes how circular or elongated a pore is: a value close to 0 indicates the pore is close to circular while a value close to 1 indicates the pore is an elongated shape close to a line.AR gives similar information as Ecc-it is the ratio of the major to minor axis lengths of a pore as calculated by the MIPAR software.Its value is ⩾1, with a value close to 1 indicating the pore is circular while higher values indicate the pore is an elongated or irregular shape.D i is the distance between a pore's centroid and the next-closest pore centroid.Each of the above properties were measured for each pore in each 1 cm 2 AAO sample, and a Gaussian curve was fitted to each set of results.The value that matched the maximum point on the curve was used to give the single geometric property value for that sample.
r avg uses an expression derived from the standard regularity ratio which is a measure of the interpore distance uniformity.The standard regularity ratio is analyzed by applying a simple Fast Fourier Transform (FFT) to three intensity profiles along the three major hexagonal directions of a Grain Analysis image.However, Stȩpniowski et al. 13 found that this analysis is only appropriate for highly organized pore arrangements.For analysis of relatively poorly ordered pore arrangements associated with low-purity aluminum AAOs, the authors recommend advanced whole-space FFT processing based on the concept of regularity ratio spread.
The r avg proposed by the authors for poorly ordered pore arrangements is a modified expression of the standard regularity ratio and is expressed as follows: where I is the intensity of the radial average (see Fig. 8b), W is the width of the radial average at half of its height (see Fig. 8b), n is the number of pores on the analyzed image, S is the analyzed surface area.The r avg FFT maps and their respective radially averaged profiles (examples are shown in Fig. 8) were generated from the Grain Analysis images with ImageJ software tools (NIH, USA) which gave options to produce FFT maps and radially averaged profiles based on The parameter values of each test run in Table II, along with their respective pore feature results, were inputted into Minitab 18  software which applied the BBD to produce 2nd order polynomial models (Eq.2) for estimating the value of a geometric pore feature for any applied operating values: where Y is the response (value of the geometric pore feature) X X and i j are the independent operating parameters C, T, P, e.g.The models were reduced to their significant terms and their response surfaces shown in Fig. 7 were obtained from MATLAB R2015b, where one variable was held constant while the other two were varied.The 99.16% low-purity foil was cut using a die into circles of 1.3 cm diameter (1.33 cm 2 ) with a protruding stripe for the anode potential connection (samples are shown in Fig. 1a).To obtain the 1 cm 2 of working area with only one face of the aluminum foil exposed, 1 cm 2 circular areas of the samples were covered with electrical tape before the samples were submerged in an epoxy resin mixture.The electrical tape protected the sample area from the epoxy resin.After the resin dried, the electrical tape was carefully removed from the surface and the exposed 1 cm 2 aluminum areas of each sample were degreased by sonicating them in acetone and ethanol solvents for 5 mins.
The Electrolytic cell setup shown in Fig. 1b was used in the anodizing process and consisted of the following: i) A 120 ml capacity, double-walled reactor consisting of an inner bath containing the oxalic acid solution and an outer section to circulate water to regulate the temperature of the oxalic acid solution.ii) A VWR AD07R-20 circulating bath for circulating temperature-controlled water in the outer section of the reactor to regulate the electrolyte bath temperature.iii) A magnetic stirrer for convection and to further regulate the temperature throughout the electrolyte bath.iv) A thermometer for monitoring the temperature of the electrolyte bath and feeding the information back to the circulating bath for temperature control.v) A Bench Power Supply (BK PRECISION 9110 0-60 V/5 A DC Power Supply).This applied the anode potential to the protruding strip of the aluminum that was being anodized and the cathode potential to a 7 cm 2 rectangular piece of platinum separated 4 cm from the anode.
The two-step anodization process for producing the 1 cm 2 AAO templates consisted of the following stages for each test run listed in Table II.A graphical representation of the steps is show in Fig. 2.
i) The electropolishing was conducted for 3 mins with an electrolyte mixture of perchloric acid:ethanol in a 1:5 volume ratio.The electropolishing potential set to 8 V and the temperature of the electrolyte set to 10 °C to achieve goodquality polishing and avoid excessive abrasion of the foil.ii) The first anodizing step produced concavities at the bottom of the aluminum foil.It was conducted for four hours using an electrolyte consisting of oxalic acid.C, T and P were adjusted for each of the fifteen test runs as outlined in Table II.iii) The etching step is conducted to remove the AAO layer grown from the first anodizing step to leave only the concavities on the surface of the aluminum.A mixture of H 3 PO 4 :CrO 3 , 6:1.8 wt.% at 40 °C was used for 2 h under constant magnetic stirring.iv) The second anodizing step was conducted under the same selected conditions as the first anodizing step for 1 h, with pores developing from the concavities that were produced from the first anodizing step.
The geometric pore features of the resulting templates shown in Fig. 5 were characterized using a SEM (JEOL JSM-6490 lV, 0.3 kV-30 kV) and MIPAR Image Analysis software.

Preparation of the 78 cm 2 AAO Templates
The steps for producing the 78 cm 2 AAOs were similar to those for the 1 cm 2 AAOs but the operating parameters were fixed to those the model predicted would produce the maximum r avg (the model being developed from the 1 cm 2 AAO testing).Also, since the 78 cm 2 AAOs required through-hole pores to be used for the production of one-dimensional structures, an additional step of refluxing was performed for the dissolution of the bottom-side aluminum substrate.Furthermore, both sides of the template then went through a second etching to produce isotropic AAO templates with through-hole pores.Figure 3 shows a graphical representation of the steps involved.
The Electrolytic cell setup for electropolishing and anodizing was adjusted for the 78 cm 2 AAOs.A larger 1.5 L electrolyte bath was required for the 78 cm 2 aluminum samples and the foils was placed at the bottom of the bath and fixed with screws.A magnetic stirrer would have been ineffective at mixing the electrolyte solution with the aluminum foils at the bottom of the bath so a mechanical stirrer was used instead.A 5 cm diameter circular stainless-steel mesh was used for the cathode and placed 6 cm from the aluminum foil which was connected to the anode.
The 99.16% low-purity aluminum foil was cut into circular shapes of 15.0 cm diameter with a protrusion for the anode connection.The surface of the foil was degreased with acetone and ethanol.78 cm 2 areas of these shapes were exposed to the electrolyte at the bottom of the bath.Figures 4a and 4b show a sample foil and the Electrolytic cell setup for producing the 78 cm 2 AAO templates.The electropolishing process was performed for 2 min using an electrolyte mixture of perchloric acid:ethanol in a 1:5 volume ratio.The potential was set to 32 V and temperature to 10 °C.These operating parameters were found through experimentation to produce good-quality polishing of the 78 cm 2 foils with this setup while avoiding excessive abrasion of the foil.II).The images were taken at the same X50,000 magnification.

ECS Advances, 2023 2 042501
After polishing, the first anodizing step was conducted for 4 h using the operating parameters that were found to produce the maximum r avg value for the 1 cm 2 AAOs: the electrolyte consisted of oxalic acid with C = 0.6 M; P was set to 46 V; T was set to 10 °C.
The first etching step was conducted in a mixture of H 3 PO 4 :CrO 3 , 6%:1.8% wt. at 40 °C for 2 h.The second anodizing step was performed under the same operating conditions as the first anodizing step over a 16 h period.The dissolution of the bottom-side aluminum substrate was then achieved through refluxing the alumina in a saturated solution of iodine/methanol for 2 d.
The final step, the 2nd etching, was performed at room temperature on both sides of these 78 cm 2 alumina with H 3 PO 4 5% wt. for 30 min to produce the isotropic AAO templates from which one-dimensional nanostructures were produced.
The 78 cm 2 through-hole alumina membranes produced from the above steps were cut into circles of 1 cm 2 using a laser beam to provide samples for wider experimentation and research.

Results and Discussion
Results for the 1 cm 2 AAO templates.-TheSEM images taken of the 1 cm 2 AAO samples, produced using the input parameter sets proposed by the BBD, are shown in Fig. 5.The numbers in the topleft-hand corner of the images refer to the test run numbers in Table II.The images were collected at the same magnification of X50,000.
The MIPAR software produced Grain analysis images from the SEM images to determine the geometric pore feature responses and investigate the effects of the three operating parameters (C, T, and P) on these responses.Figure 6 shows the SEM image of AAO Test Run #4 along with its respective Grain Analysis image.
The Minitab BBD produced 2nd order polynomial model equations to estimate each of the geometric pore features for any combination of operating parameter values.Minitab ANOVA analysis was applied to determine the significant parameter terms based on a term's p-value being less-than or greater-than 0.05.The model equations were then reduced to these significant terms in Table III, along with the polynomial constant terms, to more easily identify the critical operating parameters of each model. 14he reduced model equations based on the significant terms are as follows: The models for Ecc and AR produced a combination of low pvalues (<0.05) and relatively low R 2 and R 2 adj values, suggesting the models estimates the trend due to the input parameters but that there is high variability of individual responses around the model.The following sections discuss the results obtained for each evaluated response and their quality measurements.
Averaged Regularity Ratio (r avg ) r avg was obtained using the modified expression of the regularity ratio proposed by Stȩpniowski et al. 13 An FFT map and its respective radially averaged profile were obtained for each sample.The FFT map of a perfect hexagonal arrangement consists of six distinct spots at the corners of a hexagon. 15However, the relatively poorly ordered pore arrangements from the low-purity aluminum  AAOs in this study resulted in ring-shaped FFT maps. Figure 8 shows the FFT map of sample #4 from Fig. 5 and its respective radially averaged profile.The radially averaged profile of a verywell defined hexagonal arrangement has a narrow radial spur with a very high Intensity amplitude while the spurs of less-well organized arrangements are wider and have lower amplitudes.The intensity (I) and width at half the intensity (W) of the profile were measured with OriginPro software as proposed by Stȩpniowski et al. 13 and shown below in Fig. 8b to determine r avg from Eq. 1.
The measured r avg results from each test run in Table II along with their respective combination of operating parameter values were inputted into Minitab software.Minitab then applied ANOVA tests and the BBD to determine the significant terms and coefficients for the model equation.
The results of the first ANOVA tests are shown in Table IV.The C and P 2 terms were determined to be statistically significant having p-values < 0.05 (and high F-values).The second ANOVA test shown in Table V produced the model equation (Eq. 3) from these terms along with the 1st order P term (as is typical when its 2nd order term is significant) and the 1st order T term (to retain its influence). 14r 1.406 0.1346P 0.01252T 0.542C 0.001462P 3 The primary ANOVA assumptions were verified with graphs provided in the Supplementary Material.
The second ANOVA calculated a lack-of-fit p-value of 0.352 indicating the r avg responses correlate well with the model equation.The R 2 and R 2 adj coefficients of determination were calculated to be 76.02% and 66.42% respectively, furthermore indicating a reasonable correlation between the responses and the model.
The R 2 and R 2 adj percentages of the variability that are not explained by the operational parameters C, T, and P are due to secondary operational parameters such as anodizing time being more critical in the r avg response when producing AAOs from low purity aluminum.Cheng and Ngan 16 reported how anodizing time can have a significant effect in the pore arrangement process.According to their study, better arrangements develop progressively over anodizing time resulting in higher regularity ratios.However longer anodizing times are more expensive and not suitable for high volume applications which are the focus in this study.
The model equation was used to create the response surface shown in Fig. 7a. Figure 7a1 shows r avg for P vs C while T is held constant at the mid-point of its range (15 °C).r avg can be seen to be highest when the P is close to its mid-range and C is at the high-end of its range.Figure 7a2 shows r avg for T vs C while P is held constant at the mid-point of its range (45 V). r avg can be seen to be highest when T is at the lowest point in its range and when C is at its highest point.Figure 7a3 shows r avg for P vs T while C is held constant at the mid-point of its range (0.45 M). r avg is shown to be highest when P is close to its mid-range and T is at its lowest point.
From the r avg model equation, Minitab determined the optimal operating conditions for C, T, and P, to be 0.6 M, 10 °C, and 46 V respectively, giving a maximum r avg = 1.89.The model equation along with the three graphs of Fig. 7 indicate that a higher concentrations and lower temperatures than the investigated ranges would produce higher values of r avg .However, the production process of AAOs is more difficult to control with higher concentrations because the resultant AAOs are thinner and more fragile.Also, this work focused on a temperature range closer to room temperature to address the lower-cost requirements for highvolume AAO production applications.4shows the model derived for D p response.The R 2 and R 2 adj coefficients of determination are 86.46% and 82.77% respectively, indicating the model explains a high percentage of the variability of the response.Moreover, the model's Lack of Fit of 0.76 implies the second order model fits the data well.The model's p-value < 0.05 indicates a statistically significant association between the response and the model terms.

D
72.9 0.749P 9.65T 0.3433T 4 The model equation doesn't have a term dependent on C indicating that D p did not vary significantly over the concentration range used.This result agrees with similar results obtained by Chung et al. 17 who assessed the effect of two different concentrations of Oxalic acid (0.3 M and 0.5 M) on 99.0% low purity aluminum foils.The authors found only slight variation in the diameter of pores for different combinations of the concentrations with 1 h and 3 h anodizing times.a) DF: degrees of freedom.b) Adj SS: It quantifies the amount of variation in the response data that is explained by each term in the model.c) Adj MS: measures how much variation a term explains, assuming that all other terms are in the model.d) F-value: The F-value is the test statistic used to determine whether the term is associated with the response.e) p-value: The p-value is a probability that measures the evidence against the null hypothesis.If the p-value is less than or equal to the significance level (0.05), the model explains variation in the response.f) S: Standard deviation between the data points and the model. 14here are both T and T 2 terms in the model.The T term has a negative coefficient and doesn't have a significant effect, causing only a slight increase in D p at the lower temperature range.The positive T 2 coefficient has a more significant effect, causing the D p to increase significantly at the higher end of the temperature range as can be seen in Fig. 7b.In practice, the reason for this is that the higher temperatures accelerate the chemical dissolution of the pore walls.
The D p model's linear dependence on potential has been previously recorded by Stępniowski and Bojar. 18The authors prepared AAOs with potential ranging from 20.0 to 60.0 V in a 0.3 M oxalic acid solution, and at four values of temperature: 35 °C, 40 °C, 45 °C, and 50 °C.The D p increased linearly with potential at all temperature settings.However, the authors used the OVAT experimental approach which does not convey information on how the different operating parameters interact with each other to produce the response as the BBD experimental design in this study does.
Eccentricity and aspect ratio (Ecc and AR).-The models obtained for Ecc and AR, given by Eqs. 5 and 6, show that P and T are the primary variables of influence on their responses.

Ecc
T 0.333 0.001724P 0.0482 0.001679T 5 The R 2 and R 2 ad coefficients of determination for the models are 53.89% and 41.31% respectively for Ecc, and 67.70% and 54.78%  for AR.The relatively low values suggest that the pore shape responses have a non-trivial dependence on operational parameters other than the three primary ones used in this experimentation (e.g. the purity of the aluminum foil and anodizing times).Minitab found the optimized T and P operating conditions for both Ecc and AR to be 20 °C and 30 V respectively (C not factoring into the model).20 °C was the maximum value of temperature used in the experiments and as can be seen from Figs. 7c and 7d, the models for Ecc and AR would have benefitted from a higher temperature range, with the Ecc value getting closer to the desired 0, and AR closer to 1.However, as previously mentioned, this work focused on a temperature range closer to room temperature to address the lower-cost requirements for high-volume AAO production applications.Minitab returned values of 0.68 and 1.25 for Ecc and AR, respectively at their optimal operating conditions.
Zaraska et al. 19 also found that higher temperatures corresponded with higher circularity responses.The authors used 99.9995% high purity aluminum foils to assess the shape of pores using the following circularity expression in Eq. 7 at temperatures of 1 °C, 3 °C, 10 °C, and 17 °C: where S is a surface area occupied by single pore.
The authors observed a significant increase of the pore circularity with increasing temperature, the maximum circularity obtained at 17 °C, the highest assessed temperature used in their experiments.They attributed the results to the enhanced field-assisted isotropic dissolution of aluminum oxide occurring at higher temperatures.
The optimal setting for P was found to be 30 V, the minimum value in the range used.This indicates the models for Ecc and AR would benefit from a lower P range but the priority was given to the r avg feature in this study when considering the P range.Furthermore, lower potentials take longer to reach AAO thicknesses that can be employed in some large-scale applications resulting in more costly production.
The benefits to the pore shapes associated with using lower potentials when producing AAOs from low-purity aluminum was also discussed by Kim and Lee. 20The authors suggested that the metallic impurities existing in the low-purity aluminum foil resulted in more pronounced pore-shape defects, due to the irregular directions of the electric field at higher potentials.An example of these higher potential pore-shape defects observed in our AAOs is shown in Figs.9a and 9b with pore branching and elongated pore shapes due to pore clustering.
Interpore distance (D i ).-Figure 7f and the D i model given by Eq. 8 clearly show the linear relationship between D i and potential.C and T terms don't factor into the model equation.The R 2 and R 2 ad coefficients of determination for the D i model are 99.29% and 99.24%, respectively.These very high values of quality reflect how the model fits the D i responses very well.
The linear relationship between the D i and potential has also been found in previous literature.For example, Gasco-Owens et al. 21prepared samples at room temperature, anodized in phosphoric acid at 0.1 M, 0.2 M, 1 M, and 2 M concentrations and at 60 V and 90 V potential settings.The authors showed D i to  model derived by Hwang et al. 22 for the 99.999% aluminum foils (with 0.3 M concentration and at 15 °C) is given below in Eq. 9: A comparison between the models was made in Table VI.The models were compared using four potential values.The results of D i showed a slight variation between the two models, the differences may be primarily related to the purity of the aluminum used in each work.

ECS Advances, 2023 2 042501
The lower purity aluminum foils produced shorter D i , likely caused by the metallic impurities in the aluminum foil causing metallic oxides to develop during the anodization process.These metallic impurities contribute to less uniform arrangements of pores and to clustering of the pores at 60 V as shown in Fig. 10a. Figure 10b shows two groups of D i at 60 V, one group shows the distribution between the regular pores and the other between the pores within a cluster.Figure 10c shows a bimodal distribution of the D i of the SEM image in (a), showing the distance between the regular pores and the distance between pores in a cluster.The clustering results in lower average D i values.
Results for the 78 cm 2 AAO templates.-The78 cm 2 AAO samples were obtained from two step anodizing process by applying the optimal operating conditions found for the 1 cm 2 samples: 46 V; 10 °C; 0.6 M. Figure 11a shows an image of a 78 cm 2 sample after its aluminum base was detached.Figure 11b is a X50,000 SEM image of the top-side surface before the 2nd etching was applied, showing the self-organized substrate with a hexagonal arrangement of pores.r avg was measured to be 2.66 using its FFT map (inset) and its radially averaged profile.
Figures 11c and 11d show cross-sections of the sample at different magnifications.The anodizing rate of 3.6 μm h −1 for 16 h resulted in the AAOs having a film thickness of 58.5 μm which provided adequate mechanical stability for handling and analysis.The pore channels for this particular section of the sample were measured to have a mean diameter of 60.7 nm.
Defective areas of the 78 cm 2 template can be seen from lowermagnification SEM images.Figure 12a shows damaged areas likely due to the alloying elements that caused an uneven distribution of the electrolyte current.Gaston-Garcia et al. 23 reported similar defects they suggested were due to burning caused mainly by the Fe and Si impurities.
Figure 12b shows another area of the 78 cm 2 template with precipitates and voids due to bursting of oxygen bubbles.Tsyntsaru et al. 7 reported these precipitates are caused by the alloying elements, such as Si in the aluminum foil.The oxygen bubbles can form and burst during the electrochemical process at the oxide/electrolyte interface as explained by Feng et al. 24 The presence of these defects is not significant for the suggested target applications.
Forty-five 1 cm 2 samples were laser-cut from each 78 cm 2 AAO template.Figure 13a shows a SEM image of the closed pores on the bottom-side of one of these 1 cm 2 sample.
The AAOs then went through a 2nd etching to obtain through hole templates.Figure 13b shows a noticeable change in the color of templates from grey to white after the 2nd etching.Figures 13c and  13d show SEM images of the resultant top and bottom pores respectively.Their D p distributions are shown in Figs.13e and 13f with the average diameter for the top pores measured at 52.13 ± 1.72 nm while the bottom pores were measured to have an average diameter of 75.03 ± 0.42 nm.
r avg for the 78 cm 2 AAOs was measured to be 2.66 while the r avg for the original 1 cm 2 aluminum samples used to determine the optimal conditions was measured at 1.89.The difference may be primarily due to the different stirring methods used in the setups as previously described.The regulating fluid currents caused by the 1 cm 2 stirring were parallel to the 1 cm 2 surface while they were perpendicular for the 78 cm 2 samples.The perpendicular currents are better for regulating the temperature on the surface of the aluminum because the regulating fluid goes directly into and around the pores.Sulka 10 reported how different stirring methods can have a significant effect on the AAO geometric pore features.The author found it critical for the temperature to be well regulated inside and around the pores for best results.
Michalska-Domańska et al. 6 reported a r avg of 1.5 with 99.5% Aluminum (using an electrolyte solution consisting of sulfuric acid and glycol) as being higher than previously published regularity ratios, so the r avg of 2.66 achieved in this work with 99.16% aluminum using an oxalic acid solution is noteworthy.

Conclusions
In this work, high-quality AAOs were obtained from recyclable (99.16%) low-purity aluminum foils by successfully optimizing the operational parameters through the BBD approach.The optimal conditions for highest r avg , with low-cost production in mind, were found for an oxalic acid electrolyte concentration of 0.6 M, an anodizing potential of 46 V, and a bath temperature of 10 °C.The method of stirring was also found to be a significant contributing factor.The r avg value achieved for the 78 cm 2 AAOs is higher than previously published regularity ratios even when compared to smaller-size samples using higher-purity (99.5%) aluminum.The fabrication of these high-quality, larger AAOs from low-purity aluminum is a promising low-cost solution for several industrial applications.

k0β
is the Constant term in the polynomial ki β are the coefficients of the linear terms C, T, P kii β are the coefficients of the quadratic terms C 2 , T 2 , P 2 kij β are the coefficients of the parameter-pair interactional terms, PT, PC, TC Minitab applied Analysis of Variance (ANOVA) tests to determine the statistical significance and quality of the models along with the effects and interactions of the operating parameters.These ANOVA tests returned the p-values for the models, the significant parameter terms in the model equations (based on p-value < 0.05), goodness-of-fit measurements R 2 and R 2 adj , and a Lack-of-Fit measurement.The ANOVA assumptions were validated for each model (Supplementary Material).

Figure 2 .
Figure 2. Steps involved in producing the 1 cm 2 AAO templates.

Figure 3 .
Figure 3. Steps involved in producing the 78 cm 2 AAO templates.

Figure 4 .
Figure 4. (a) Sample aluminum foil used for the 78 cm 2 AAO templates; (b) Electrolytic cell setup to produce the 78 cm 2 AAOs templates.

Figure 5 .
Figure 5. SEM images of the 1 cm 2 AAOs for each set of operating parameters listed in Table II (images (a) to (o) correspond to test runs 1 to 15 respectively in TableII).The images were taken at the same X50,000 magnification.

7 .
The response surfaces obtained for each model are shown below in Fig.The models for D p and D i have combinations of low p-values (<0.05) and high R 2 and R 2 adj values, suggesting the models explain both the trend due to the input parameters and indicating the relatively low variability of individual responses around the model.

Figure 8 .
Figure 8.(a) FFT map of sample #4 from Fig. 5; (b) Radially averaged profile of the FFT map in (a).

Figure 9 .
Figure 9. SEM images of sample produced with C = 0.6 M, P = 60 V, T = 15 °C (Test Run #5 in Table II); (a) Cross-section of sample showing pore branching; (b) surface image showing pore clustering.

Figure 10 .Figure 11 .
Figure 10.(a) SEM image of sample produced at 60 V − 15 °C − 0.3 M (Test Run #6) showing clustering of pores; (b) D i vs P plot for all samples; (c) Bimodal distribution of D i for the 60 V − 15 °C − 0.3 M sample in (a).

Figure 12 .
Figure 12.(a) Lower-magnification SEM image showing defects on the AAO surface.(b) SEM image of voids due to the bursting of oxygen bubbles and precipitates due to alloying elements.

Figure 13 .
Figure 13.(a) SEM image of the closed pores on the bottom-side of a 1 cm 2 laser-cut sample; (b) A noticeable change in the color of templates from grey to white was observed after the 2nd etching; (c) SEM images of the top-side pores after 2nd etching; (d) SEM images of the bottom-side pores after 2nd etching; (e) diameter distribution of the top-side pores in (c); (f) diameter distribution of the bottom-side pores in (d).

Table I .
Weight Percentages of the main elements in the low-purity aluminum foil used in this study.

Table III .
Summary of the BBD models obtained and their quality measurements for each geometric pore feature response evaluated.

Table IV .
First ANOVA test for r avg.
a DF b Adj SS c Adj MS d F-Value e p-Value

Table VI .
22mparison between the D i model determined by Hwang et al.22and the model derived in this work for 99.16% Al.

Table V .
Second ANOVA test on the significant and P, T terms.