Preparation and Surface Functionalization of a Tunable Porous System Featuring Stacked Spheres in Cylindrical Pores

A geometrically tunable nanoporous system featuring enhanced active surface area by stacking of spheres in cylindrical pores is fabricated. Highly ordered arrays of straight, constricted pores are obtained by anodization of metallic aluminum. Polystyrene (PS) spheres are assembled inside the pores by flowing their suspension through the porous membrane, whereas the construction serves as a filter. After surface functionalization with a noble metal catalyst, these model electrocatalysis systems exhibit functional properties (capacitance in electrochemical impedance spectroscopy) that mirror their geometric parameters. A systematic investigation of the system's geometry as it depends on the surface chemistry of the pores, on the one hand, and the physical parameters of the infiltration procedure, on the other hand, shows that mechanical stacking prevails over surface chemical interactions to determine the stacking density. The highest values of surface area are obtained when PS spheres are put in contact with HfO2 followed by ZnO according to adsorption measurements. Surface derivatization with organic layers does not improve stacking any further. However, choosing the proper concentration of PS spheres and flow rate are crucial for obtaining densely packed sphere assemblies without clogging of the pore entrance.


Introduction
The control of geometry and morphology of nanostructured surfaces is of pivotal importance in catalysis, drug delivery, and environmental applications. [1,2]The structural engineering and DOI: 10.1002/admi.202300436control of geometrical parameters in porous structures featuring wellcontrolled sizes and shapes enable for adjusting properties and optimizing performance in applications. [3,4,5]The geometric aspect is of particular relevance in catalysis and electrocatalysis.First, chemical reactions occur at the surface, so its geometric area determines chemical (or electrochemical) turnover.Second, the transport mechanisms, and therefore kinetics, of the reagents and products (and charge carriers) within the fluid phase and potentially within the solid may place an upper boundary for their delivery to the surface.Therefore, the geometric nanostructuring of catalyst beds and electrodes as a strategy to improve the performance of chemical and energy conversion systems must compromise between two contradicting requirements. [2,6]On the one hand, maximizing the surface area of the solid phase is achieved by generating pores which increase the volumetric density of reactive sites at the phase boundary.On the other hand, the minimization of transport distances between the bulk of the fluid phase and each site of the interface favors low porosity. [7,8]These two conditions can be balanced by generating and studying systems in which pore structures are highly homogeneous and systematically variable.
One porous system that features outstanding versatility is anodic aluminum oxide (AAO) membranes.Not only does anodized aluminum offer engineering functionalities such as excellent hardness, corrosion, and abrasion resistance, [9,10] ordered porous systems based on AAO provide cylindrical, straight nanopores made with low-cost techniques for fundamental research. [11]The possibility to tune the pore diameter, pore length, and density of pores in wide ranges and in self-ordered conditions by proper choice of the anodization parameters (electrolyte bath, applied voltage, anodization duration) has been exploited for the preparation of diverse nanostructured materials as model systems for a variety of chemical and physical applications affected by geometry. [12]15][16] In line with these considerations, a number of reports on catalyst surfaces and electrodes based on functionalized AAO have documented how the performance parameters of catalytic and electrocatalytic turnover depend on the pore geometry (diameter and depth) as well as on the surface functionalization itself.[19][20][21] However, the smooth pore walls of anodized alumina may not be the best model for real catalyst supports, which often exhibit significant roughness on the sub-100 nm length scale.Hence, we present here the preparation of a novel, geometrically tunable 3D substrate system featuring inner roughness within the cylindrical pores.A well-defined change in pore diameter obtained upon anodization serves to assemble polystyrene spheres in a filtration procedure.The assembly depends on both the chemical identity of the pore walls (after corresponding functionalization with either metal oxides or organic monolayers) and the physical parameters for their assembly (concentration of the suspension and flow rate).The electrochemical activity quantifiers of an electrode obtained after noble metal deposition into this model substrate exhibit effects that follow its geometric surface area.

Preparation of the Modulated Porous System
In the first preparative step (Step 1 in Figure 1), we use anodization of aluminum to generate a filter system that exhibits long, straight pores with a large diameter in their first segment complemented with a narrower second segment.Scanning electron microscopy (SEM) images of the array in cross-sectional, top, and bottom views, Figure 2, demonstrate the high degree of order, the straight geometry of the cylindrical pores, and the low polydispersity of their diameters.The porous anodic structure has a homogeneous net length of around 25 μm following the procedures described in the experimental section (Figure 2a).The pores clearly exhibit an abrupt change of diameter (Figure 2b) between the first segment with a diameter D 1 ≈ 320 nm and length L 1 ≈ 17.5 μm and the second one with D 2 ≈ 260 nm and L 2 ≈ 8 μm (Figure 2c,d).Each of these values is tunable experimentally upon adjustments of the anodization parameters: the length depends on the anodization duration linearly (with an approximate growth rate of 2.1 μm h −1 ), whereas the diameter is set by the duration of the thermal pore widening treatments which in our conditions (10% H 3 PO 4 , 45 °C) achieve isotropic etching of the amorphous alumina at a rate of ≈0.67 nm min −1 .

Assembly of the PS Spheres Inside the Ordered Porous Systems
The subsequent processing step (Step 2 in Figure 1) shall result in a dense packing of spheres inside the wide pore section, so that the spacings between spheres later serve to separate metal particles from one another.The constriction created by the narrow segment in our anodic aluminum oxide (AAO) template enables us to trap polystyrene spheres via a procedure akin to filtration.To this goal, a suspension of polystyrene (PS) spheres with diameter D s = 130 nm is pumped through the porous AAO template, whereby the dilution, flow rate, and total flow duration represent adjustable parameters.To perform this optimization in a quantitative manner, Figure 3a,b, we add a layer of ZnO by atomic layer deposition (ALD) in order to retain the arrangement of the particles within the pores and reduce charging in cross-section SEM analysis, then use energy-dispersive X-ray microanalysis (EDX) to quantify the carbon mass percentage within a rectangular area of the cross-section encompassing the full wide pore segment.The conditions that optimize the sphere stacking (i.e., that maximize the C content) are a dilution of 500 μL PS suspension (reference see the Experimental Section) in 10 mL ethanol and 60 s of flow deposition with a flow rate of 2.50 mL min −1 .The SEM images, Figure 3c,d, demonstrate the dense and homogeneous filling of pores obtained in these optimized conditions.Interestingly, the highest flow rates but not the highest suspension concentrations yield the best degree of order.The reason for this effect lies with a quite sensitive balance between an insufficient agglomeration tendency (which lets individual spheres traverse both pore segments without clogging at the constriction) and a too high agglomeration tendency (which lets spheres agglomerate on the top face of the template already, Figure S1, Supporting Information).

Effect of the Pore Wall Chemistry on the Sphere Assembly
Beyond the processing parameters during the filtering assembly, the stacking of spheres could also be affected by their interaction with the pore walls.To investigate this, we compare the behavior of porous templates submitted to various surface functionalization preliminarily, Figure 4a.Treatment of the alumina pores with the methylation agent HMDS (hexamethyldisilazane), which is a classic method for rendering oxides hydrophobic, clearly results in a much lower density of PS spheres in all processing conditions explored.We interpret this observation as a demonstration that the assembly is not mainly driven by the chemical interactions between the hydrophobic PS spheres and pore walls.Instead, spheres stack simply based on mechanical effects (clogging), in agreement with the strong dependence on flow and concentration parameters demonstrated in the previous section.
Do electrostatic effects influence the assembly of spheres in the porous system?Let us address this question by comparing alumina with HfO 2 and ZnO surfaces.A reliable determination of their  -potentials is enabled by streaming current experiments conducted on silicon wafers coated with a 10 nm thick layer of HfO 2 or ZnO, respectively, Figure 5.
Based on these streaming current measurements, the isoelectric points (IEPs) of HfO 2 and ZnO thin films are located at pH 4-5 and pH 8-10, respectively.Both IEPs are in excellent agreement with the data reported by Kosmulski et al. [22] HfO 2 exhibits a  -potential of −37(±6) mV at neutral pH, whereas the ZnO thin film degrades at pH <9 in an aqueous solution.However, a positive  -potential at neutral pH can be expected for ZnO. [23,24]Given the negative -potential of our PS spheres in ethanol, −31(±1) mV, electrostatics may contribute to the significant differences in particle loading shown in Figure 4a   between HfO 2 and ZnO.The oppositely charged ZnO and PS interfaces lead to electrostatic attraction, which increases the chance of disordered pore clogging instead of a dense, mechanically driven, stacking of spheres (cf. Figure 4d).In contrast to this, the negative surface charge of HfO 2 tends to maintain the spheres suspended throughout their travel through the pores until they encounter the stack.This mechanism favors high particle loadings in the pores (cf. Figure 4e).In other words, the data are consistent with our model of mainly mechanically driven stacking.Adhesion of spheres to the walls, be it caused by electrostatics or a hydrophobic effect, is not conducive to a high-density assembly.Table 1 summarizes the experimental IEPs.

Characterization of Functional Samples
In the closing steps of our preparative procedure (Steps 3-5 in Figure 1), the PS spheres are coated with ZnO as an electrically conductive layer by ALD, the only method able to coat such highly porous systems with complex geometries in a conformal and homogeneous manner.The spheres can be removed via thermal oxidation, upon which a purely inorganic sample is obtained.Figure 4c,d,e demonstrates that PS is removed without damage to the ZnO overcoating.Finally, the electrocatalyst Pt is added by ALD (nominal thickness 6 nm).The final samples obtained after the full procedure established in this study exhibit the same morphology as in the previous steps (SEM in Figure 6a,c,e but with all materials desired in their crystalline form, as shown by X-ray diffraction (XRD, Figure 6b,d,f).All XRD patterns after the thermal step evidence the peaks for the Table 1.IEP values were determined experimentally by streaming current measurements for the ALD-coated samples used in this work, compared with bulk materials from the literature.
Oxide IEP (bulk mat.)IEP thin film ZnO 8.8-10.0 [22]-10 HfO 2 5.0 [22] 4-5 hexagonal (hcp) phase of ZnO (zincite, COD 9004180).The samples featuring ZnO on ZnO exhibit a crystalline layer even as deposited.However, samples with only one ZnO layer on another metal oxide are originally amorphous, and ZnO crystallizes upon annealing. Thesobservations are in line with the ALD literature, where ZnO transitioning to a crystalline form upon reaching a critical film thickness during growth is documented.In case it is crystalline before thermal treatment, annealing generates a preferential orientation along the (100) direction (peak at 31.5°).Furthermore, the final samples also reveal the peaks associated with the cubic (ccp) structure of Pt (COD 9013417), deposited in crystalline form after the thermal step.HfO 2 is deposited in amorphous form and crystallizes upon thermal treatment (monoclinic, COD 1528988).Note that anodic alumina is amorphous throughout and gives rise to both broad signals below 30°.For the reference sample with a ZnO layer but without PS spheres within the pores, weak peaks associated with the ZnO layer are observed, whereas in contrast with those, pronounced peaks appear for Pt.The XRD patterns of all nanostructured samples are consistent with the crystalline phases expected on them.
The chemical composition of all samples (Figure 7) follows expectations with the presence of the elements Al, Zn, Pt, O, and optionally Hf.EDX line profiles recorded for Al, Zn, and Hf along the cross-section exhibit the sharp step functions expected due to the presence of the pore diameter constriction.Along each pore segment, the signals are constant, evidencing that the deposition of the spheres within the pores and of the metal oxides on the surfaces are homogeneous.The Pt profiles also feature a quite sharp step on three of the four samples.The HfO 2 sample, however, is less perfect in that the Pt signal trails and reaches close to zero at about half of the spheres-infiltrated segment.This aspect is probably associated with the slow diffusion of ALD precursors through the meanders of the complex pore geometry generated by the densely packed spheres, and it will be important for the functional performance presented below.

Quantification of Specific Surface Area Via Sorption Experiments
A total of six samples are chosen for sorption analysis.One set of samples exhibits the unmodified Al 2 O 3 surface while the other two variants are coated with ZnO and HfO 2 .Of each variant, we also compare one fully prepared to the stage of Step 3 in Figure 1 (including ZnO-coated spheres) and one in the state prior to the addition of the spheres (no spheres, smooth cylindrical pores).Krypton (77 K) adsorption expressed as a function of relative pressure is shown for HfO 2 and ZnO in Figure 8a,b.The adsorbed amount is stated per unit of macroscopic sample area for a more straightforward comparison.Clearly, the amount adsorbed on the spheres-containing samples is significantly higher than with empty, smooth pores.This reflects an increase in effective surface area, assuming a similar state of the krypton adsorbate.This increase is quantified using the BET method for all three systems (Al 2 O 3 , HfO 2 , ZnO) in Table 2.It is most appreciable in the HfO 2 -based system, in line with the observations presented above by SEM and EDX.

Electrochemical Characterization
The final aspect of our study is dedicated to functionality.Taking the electrocatalytic hydrogen evolution as an example, we want to demonstrate catalytic activity and the effect of the increased specific surface area.This is achieved by quantifying the capacitive and resistive components of the samples' electrical response in electrochemical impedance spectroscopy, which can be correlated with the effective surface areas determined by sorption experiments.The data are collected in two quite different conditions in order to prove the validity of our conclusions: 1) at buffered pH 7 (phosphate buffer) under a constant applied voltage of −0.4 V with respect to the Ag/AgCl reference electrode, corresponding to a potential approximately 0.22 V less negative than the equilibrium potential at this pH under 1 bar H 2 , and (2) at pH 3 under -0.8 V (to Ag/AgCl), corresponding to a potential approximately 0.41 V more negative than the equilibrium under 1 bar H 2 .The Nyquist plots obtained for those systems are shown in Figure 9a,b.The impedance spectra for the nanostructured samples include a depressed semicircle in the high-frequency region representing the electron-transfer-limited process, and a straight line at the lower-frequency region corresponding to  diffusion-limited conditions.The electrodes based on Al 2 O 3 and HfO 2 surfaces (prior to PS infiltration) yield a truncated semicircle at pH 7, indicating fast surface turnover.This is consistent with the high quality of nanostructuring through PS spheres assembly within the porous system (quantified by high EDX carbon content).The comparison between the sample with spheres assembled on ZnO surface and the reference without spheres shows that surface area increase by spheres is associated with a semicircle of smaller diameter, consistent with the higher surface area.
A quantitative treatment of the EIS data is afforded by fitting the datasets to a classic equivalent circuit model shown in Figure 9c [25][26][27] The physical-chemical interpretation of the electrical element centers around the crucial double layer capacity C dl in parallel to the charge transfer resistance R ct to Faradaic electron transfer across the interface.A Warburg element W is placed in series with it to model mass and charge transport along the pore.Additionally, the Ohmic resistance to electron transport along the ZnO layer places a resistance element R ohm in series.This whole set of elements is placed in parallel with a capacitor C por representing the entire oxide matrix behaving as a dielectric.
The values determined from the fits are presented in Table 3 for three samples prepared as presented above with HfO 2 , ZnO, and Al 2 O 3 (from the AAO walls) as surfaces before PS spheres infiltration, with a sample featuring smooth pores (ZnO-coated pore wall), unfunctionalized with spheres, as a reference (labeled "ref" on the graphs).Standard deviations determined from the fit quality are very small and therefore not included in the table.Both datasets yield similar trends despite the slightly distinct values obtained as a result of the widely different experimental conditions.As expected, C por remains on the same order of magnitude for all samples, as well as W and R ohm , for both pH values evaluated.The most significant differences appear in the crucial C dl values.The largest double-layer capacitance is obtained for the sample preliminarily functionalized with HfO 2 .This is in line with our observations reported above concerning the quality of spheres ordering inside the pores (quantified above as EDX carbon content and specific surface area via Kr sorption).Accordingly, the sample preliminarily treated with HfO 2 and the untreated one (bare alumina) have both a higher density of spheres and a higher roughness (indicated by higher C dl ) than the sample treated with ZnO before infiltration.The reference sample which has the same chemical identity as all others but without the added roughness inside the pores caused by the spheres has the lowest C dl of all.This trend is observed for both pH values evaluated for the electrodes in this study.
Overall, high C dl correlates with low R ct as expected based on surface area.One notable deviation from this trend is observed with the HfO 2 -coated sample.Here, R ct is high despite the large surface area.The most likely reason for this behavior is the imperfect coating with Pt demonstrated by EDX line profiles (Figure 7).In summary, one example of functional performance provided by electrocatalytic proficiency for hydrogen evolution on Pt demonstrates that all geometric and chemical features of the samples are mirrored in the functional performance quantifiers.Most importantly, the surface area increase achieved with stacked PS spheres translates into electrocatalytic performance as expected.This model system can now be applied to a variety of case studies in catalysis and electrocatalysis.

Conclusions
This work establishes a preparative method for generating a tunable porous system based on anodized alumina acting as a template for a polystyrene spheres assembly achieved with high degree of control in a simple flow procedure.Our systematic exploration of the relevant experimental parameters during assembly toward achieving a high-density packing of the PS spheres inside the pores reveals that the functionalization of the pore wall surface influences the packing quality, with uncharged oxide surfaces delivering the best results.Engineering the surface chemistry so as to enhance its affinity for the polystyrene of the spheres does not provide the desired effect.Both electrostatic attraction and hydrophobic interactions cause a less ordered, less dense packing.However, the physical parameters of suspension and flow provide a much more convenient set of tools to adjust and optimize the structure of the confined spheres stack.The geometric quality of the stack can be assessed and quantified in several distinct ways.The absolute amount of organic material can be determined by EDX.Highly sensitive Kr 77 K adsorption allows one to assess the corresponding enhancement in specific surface area and we find excellent agreement with the semi-quantitative trends observed by EDX.Furthermore, when functionalized with Pt particles as an example of application, this model system yields physical-chemical performance parameters that also reflect its geometric properties.In electrochemical impedance spectroscopy, the nanostructured Pt electrode exhibits higher capacitance values when the electroactive surface area is enhanced by densely packed spheres.All three probes of the surface and its geometry, namely EDX, adsorption, and impedance spectroscopy, point to a HfO 2 coating of the smooth, cylindrical pores as the best suited to enable a dense sphere packing and maximal surface area enhancement under high flow and moderate concentration conditions.
This model system can now be applied to a broad variety of electrocatalysis and catalysis topics.Enhancing a catalyst support's geometric surface area is always advantageous to increase catalytic throughput.However, transport along long pathways in highly porous systems can become slower than turnover at the interface, in a manner which has mostly been difficult to pinpoint experimentally in an accurate and definitive manner.With the system presented here, the length and diameter of pores but also their internal roughness can all be adjusted independently and varied systematically to create samples featuring tailored residence times of the reactive mixture, tailored numbers of surface reactive sites, and tailored diffusion distances.This represents a novel experimental platform for testing and quantifying theoretical simulations, their hypotheses, and their predictions.

Experimental Section
Materials: Reagents for the preparation of the samples were purchased from Sigma-Aldrich, Alfa Aesar, Strem, or VWR and used as received.Millipore Direct-Q system was employed for the purification of water before use.Aluminum plates (99.99%) for the anodization process and Si (100) wafers for the atomic layer deposition (ALD) were supplied by Smart Membranes and Silicon Materials Inc, respectively.The sacrificial polystyrene (PS) spheres used in this study were prepared following the protocol reported in the literature. [28,29]nodization of aluminum into ordered pores: The generation of the tunable ordered porous system follows the standard two-step anodization procedure described in the literature. [30]A homemade setup was used to anodize four circular aluminum plates of 2.2 cm diameter, held by Orings underneath a polyvinyl chloride beaker.A Cu plate was placed under them and acted as an electrical contact.The assembled beakers were cooled to 0 °C overnight.Subsequently, electropolishing was performed in a solution of 1:3 v/v HClO 4 /EtOH for 7 min by applying +20V.The samples were rinsed and cooled down again in the anodization beakers for 2 h before the first anodization at 195 V (max.0.20 A) for 24 h in 1 wt.%H 3 PO 4 .After this first anodization, ordered Al 2 O 3 hemispherical indentations were achieved by submitting the samples to a treatment with chromic acid overnight at 45 °C.With a second anodization, a first segment was grown for 8 h in 1 wt.%H 3 PO 4 under the same experimental conditions as previously.This step yielded a system with well-ordered Al 2 O 3 pores.Subsequently, the pores were widened in 10% phosphoric acid solution for 45 min at 45 °C.Afterward, a short cooling was performed again, and a third anodization was carried out to grow up a second segment for 4 h.Finally, both segments were submitted to a final pore widening during 45 min at 45 °C in 10% phosphoric acid solution.The next step was to remove the metallic Al substrate on the backside of the Al 2 O 3 porous membrane in a CuCl 2 solution (0.7 M CuCl 2 in 10% HCl).Finally, the resulting samples were treated with 10% phosphoric acid for 60 min at 45°C for opening the Al 2 O 3 barrier layer and closing the pores.
Surface Functionalization of Ordered Porous System: Subsequently, the system of segmented pores with two distinct diameters was coated with ≈12 nm of either ZnO or HfO 2 by atomic layer deposition (ALD) to provide various surface properties (from acidic to basic) by using an ALD set-up with N 2 as carrier gas.The deposition of ZnO was carried out at 90°C with diethylzinc (DEZ) and H 2 O both kept in stainless steel bottles maintained at room temperature.The pulse, exposure, and purge durations of the precursors, DEZ and H 2 O, were 0.15, 40, and 70, and 0.3, 40, and 70 s, respectively.55 ALD cycles gave rise to ≈12 nm of ZnO.For the deposition of HfO 2 tetrakis(ethylmethylamido)hafnium and H 2 O maintained were used in steel bottles at 65 °C and room temperature, respectively, with pulses, exposures, and purges of 0.55, 40, and 70 s for both precursors.Two consecutive subcycles were pulsed for the TDMA-Hf precursor.Native silicon wafers were added to the reaction chamber for the determination of the layer thicknesses by spectroscopic ellipsometry.Molecular surface functionalization of the AAO walls by APTES, (3-aminopropyl)triethoxysilane, was performed by immersion in a 5% solution in toluene for 24 h under continuous stirring at room temperature. [31]For the functionalization with HMDS, hexamethyldisilazane, 100 μL of liquid HMDS was added to a closed, evacuated Schlenk flask containing the AAO sample at room temperature for 24 h.The APTES-and HMDS-modified AAO porous systems were dried before further use.
Assembly of Polystyrene (PS) Spheres: Polystyrene (PS) spheres with a diameter of 130 nm were synthesized as aqueous suspensions by a published procedure. [31]After dilution in EtOH in ratios of 250, 500, and 1 mL to 10 mL of EtOH, they were assembled in porous systems with and without surface functionalization by flowing in a microfluidic system.A peristaltic pump (Model 400 A) was used to generate a flow rate set to values from 1.65 mL min −1 to 2.5 mL min −1 for a duration from 15 to 60 s.After the stacking of PS spheres was achieved within the porous system, a new layer of ZnO (≈12 nm) was deposited by ALD following the conditions described previously.Subsequently to the quantification of carbon content on the samples by EDX, the removal of the PS spheres was performed via thermal decomposition at 450 °C in the air for 5 h in an oven.Finally, electrical contact was defined on one face of the samples by sputter-coating 10 nm of gold in a Torr CRD 622 operating in DC mode.
Isoelectric Point Determination: The electrokinetic measurements were performed by the streaming current method in a setup similar to the ones described previously. [32,33]An asymmetric cell for this setup was manufactured out of PTFE (Auer Kunststofftechnik GmbH) with a nominal channel height of 0.3 mm.The measurement cell was symmetrically calibrated with a PTFE piece, and the asymmetrically measured data were evaluated according to the procedure reported by Walker et al. [34] All streaming current measurements were performed at constant ionic strength (10 mM, KCl, Bi-oUltra, Merck) between pH 3 and pH 11 (HCl/KOH, Titrisol, Merck).Calibration of the streaming current data was obtained on thermally oxidized silicon wafers (CrysTec GmbH).A good agreement with the data reported by Scales et al. was found. [35]The reported  -potentials were calculated by the Helmholtz-Smoluchowski approximation.The isoelectric point (IEP) determination for the different coatings (HfO 2 and ZnO) was performed on silicon wafers onto which the oxides were preliminarily deposited by ALD as described above.
Catalyst Deposition: The samples dedicated to electrochemical analysis were then provided with a catalyst layer by ALD.Trimethyl(methylcyclopentadienyl)platinum (MeCp)PtMe 3 and ozone were used as precursors for the deposition of platinum (Pt).A generator model BMT 803N produced the ozone from O 2 .The deposition of Pt was carried out at 220 °C while (MeCp)PtMe 3 was kept in a steel bottle at 50 °C and delivered with pulse, exposure, and purge durations of 0.5, 30, and 90 s.Two consecutive subcycles with pulse and exposures of 0.5 and 30 s, respectively were performed before the purge with N 2 for 90 s for the metal precursor, and four microcycles were used with ozone in ALD.Approximately 6 nm were deposited with 100 cycles.
Electrochemical Studies: The final nanostructured samples were lasercut (GCC Laser Pro Spirit LS Laser) into small disks of 3.5 mm in diameter and glued on square double-sided conductive copper foils in contact with the gold current collector.An electrically insulating polyimide (Kapton) adhesive tape provided a circular window of 2 mm as the area in contact with the electrolyte for the electrochemical analysis.A three-electrode electrochemical cell was built with Pt mesh as the counter-electrode and Ag/AgCl as the reference electrode.The electrochemical measurements of EIS were done at room temperature with a Gamry Interface 1000 potentiostat in a phosphate electrolyte based on 0.1 M NaH 2 PO 4 adjusted to pH 7 or pH 3. The Gamry Echem Analyst software was used to generate the fits to Nyquist plots obtained on all samples.
Material Characterization: The morphology and geometry of the nanostructured samples were characterized by scanning electron microscopy (SEM) and the quantification of the elements by energydispersive X-ray microanalysis (EDX) using a JEOL 3010 microscope.The crystalline structure was determined by powder X-ray diffraction measurements using a Bruker D8 Advance diffractometer with Cu K  radiation ( = 1.5405Å).The determinations of the layer thicknesses were performed on Si (100) wafers with a Sentech spectroscopic ellipsometer SENpro equipped with a tungsten halogen lamp.Atomic force microscopy (AFM) was utilized to determine the surface topography of planar ALD-deposited oxide films in PeakForce Tapping mode (Bruker ICON, Nanoscope V controller) with ScanAsyst Air cantilevers (Bruker) under ambient conditions.The cantilevers have a nominal resonance frequency of 70 kHz and a tip radius <5 nm.AFM images of 1 × 1 μm 2 were recorded to determine the average roughness after a plane fit of 1st order, Figure S2, Supporting Information.
Adsorption Experiments: Adsorption isotherms were measured using an Autosorb iQ automatic volumetric adsorption analyzer (Anton Parr QuantaTec, formerly Quantachrome instruments, Boynton Beach, FL).In order to assess the very small surface area of the macroporous alumina samples, krypton (purity 99.9999%, obtained from Air Liquide) was used as adsorptive.The sorption measurements were performed at liquid nitrogen temperature (77 K).Samples were outgassed at 120 °C for 12 h under turbomolecular pump vacuum.
Surface areas were calculated by applying the BET method to krypton 77 K adsorption data in the relative pressure range p/p 0 of 0.05 to 0.1, assuming the customary p 0 for supercooled krypton of 2.63 Torr (i.e. at 77 K krypton is ca 38 K below its triple point). [36]Due to this extremely low saturation pressure, the number of molecules in the free space of the sample cell is significantly reduced (to 1/300th as compared to nitrogen or argon at their respective boiling temperatures: this leads to the high sensitivity of krypton adsorption at 77 K (enabling the assessment of absolute surface areas as low 0.01 m 2 ), which is needed here.An effective Kr cross-sectional area of 0.16 nm 2 , determined by utilizing nonporous alumina materials with known specific surface areas (i.e., benchmark data), was used for the surface area determination. [37]The Kr BET surface areas obtained in this way on the porous membrane samples were normalized to the respective macroscopic geometric sample areas.
Statistical Analysis: The determination of sphere loading density by EDX (Figure 3,4) was performed on cross-sections of porous samples, whereby the full sample depth was probed and averaged over time.The datapoints represent values obtained from the instrument's Jeol software by integration of the peaks without further processing.The uncertainty is on the order of 0.5%.Each data point of the zeta potentials vs. pH results from three measurements and is represented as a mean and corresponding standard deviation.Each zeta potential has been determined from the slope of current vs. pressure by a line fit.The fit is based on 10 different pressures, at each pressure value, the pressure and current, respectively, have been averaged from 10 measurements.No data points have been excluded as statistical outliers.However, for ZnO no data were acquired in the pH-range below pH 8 as under these conditions the film dissolved.The above data treatment is in line with the IUPAC report on electrokinetic measurements, by Delgado et al. [38] and specialized studies on the reproducibility of streaming potential measurements. [39,40]The specific surface area values determined by krypton adsorption result from up to three repeats and the uncertainties cited in Table 2 are associated with the maximum deviations from the corresponding mean BET areas.The quantities determined by electrochemical impedance spectroscopy (Table 3) are derived from the model described in Figure 9 via a fitting procedure performed by the Gamry Analyst software.The uncertainties resulting from the fits being very small, we chose to present in Table 3 values with as many significant digits as reasonable based on at least three repeats on nominally identical samples.

Figure 1 .
Figure 1.Schematic representation for nanostructuration of the samples for this study.Green lines: pore wall functionalization (metal oxide or molecular self-assembled layer); orange disks: sacrificial polystyrene spheres; pink lines: electrically conducting ZnO layer; cyan: Pt electrocatalyst.

Figure 2 .
Figure 2. SEM images of the tuneable system of segmented pores after the first steps of preparation.The sample is shown in cross-section a,b) exhibiting an abrupt diameter modulation, whereas the face view of its upper surface c) and of the pore openings obtained at the end of anodization d) show distinct diameters.

Figure 3 .
Figure 3. Assembly of polystyrene particles of D s = 130 nm inside the ordered segmented porous system.Systematic study of the stacking density as a function of: a) flow rate, b) flow duration, and a,b) suspension dilution.SEM images of assembled PS sphere systems of the highest density in c) cross-sectional view and d) top view, showing the arrangement of spheres along the pores.

Figure 4 .
Figure 4. a) Quantification of the carbon mass percentage (%C) by EDX for various surface-functionalized porous systems.SEM images of the surface functionalization of the ordered porous system, b) coated pores with a layer of metal oxide (HfO 2 ).Arrangement of the polystyrene spheres template coated with ZnO layer by ALD, after thermal treatment at 450 °C: c) as arranged in anodic aluminum oxide (Al 2 O 3 ) pores directly and in pores coated with d) ZnO and e) HfO 2 .

Figure 5 .
Figure 5.  -potential of HfO 2 and ZnO coated on silicon wafers as a function of the pH at an overall ionic strength of 10 mM in KCl.

Figure 6 .
Figure 6.Cross-section SEM images and X-ray diffraction (XRD) patterns of the final electrodes of the three main types described, that is: a,b) AAO template ALD-coated with ZnO layer before PS spheres infiltration, c,d) AAO template ALD-coated with HfO 2 , e,f) AAO template without ALD-coated, g,h)AAO template ALD-coated with ZnO without PS spheres.XRD patterns are always compared to the corresponding datasets (presented in the lower part of the graph) recorded after Step 4 of the preparative procedure (Figure 1).

Figure 7 .
Figure 7. a) EDX spectra recorded from the cross-section of the samples prepared.EDX profiles along the cross-section for the various porous systems: b) ALD-coated with ZnO without PS.With spheres: c) AAO template ALD-coated with ZnO, d) AAO template ALD-coated with HfO 2 , and e) AAO template.

Figure 8 .
Figure 8. Krypton (77 K) adsorption isotherms for samples coated with a) HfO 2 and b) ZnO, with and without infiltrated spheres.

Figure 9 .
Figure 9. EIS datasets recorded in a phosphate electrolyte prepared from 0.1 M NaH 2 PO 4 electrolyte adjusted to a) pH 7 and b) pH 3 for nanostructured porous electrodes coated with various metal oxides [fitted data: black solid line; measured data: line + symbol].c) The equivalent circuit model for the nanostructured porous electrodes.

Table 2 .
Kr BET surface areas per unit of macroscopic sample area, in m 2 m -2 .

Table 3 .
Values of the elements of the equivalent circuit from the fitted model to impedance spectroscopy data at pH 7.0 and pH 3.0.