Se-doped Nb2O5-Al2O3 composite-ceramic nanoarrays via the anodizing of Al/Nb bilayer in selenic acid

Ceramics International 49 (2023) 34712–34725 Available online 12 August 2023 0272-8842/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Se-doped Nb 2 O 5 –Al 2 O 3 composite-ceramic nanoarrays via the anodizing of Al/Nb bilayer in selenic acid Kirill Kamnev b , Maria Bendova a , b , Zdenek Pytlicek b , Jan Prasek a , b , Luk´ aˇ s Kejík b , Frank Güell c , Eduard Llobet d , Alexander Mozalev a , b , * a Department of Microelectronics, Faculty of Electrical Engineering and Communication, Brno University of Technology, Technicka 10, 616 00, Brno, Czech Republic b CEITEC – Central European Institute of Technology, Brno University of Technology, Purkynova 123, 61200, Brno, Czech Republic c ENFOCAT-IN2UB, University of Barcelona, Marti i Franques 1, 08028, Barcelona, Catalonia, Spain d MINOS, University Rovira i Virgili, Av. Paisos Catalans 26, 43007, Tarragona, Catalonia, Spain ARTICLE INFO Handling Editor: Dr P. Vincenzini Keywords: Al 2 O 3 –Nb 2 O 5 nanocomposite Anodizing Selenic acid Transition metal oxides Porous anodic alumina SERS ABSTRACT Novel arrays of Nb 2 O 5 -based ceramic nanostructures of various sizes (9–210 nm) and morphologies (dots, goblets, rods) aligned on substrates are fabricated via the anodizing of a thin Nb film through the initially formed porous anodic alumina (PAA) film in 1.5 M selenic acid (H 2 SeO 4 ) – a new aqueous electrolyte generating extraordinarily thinner PAA pores than any other solutions. Accordingly, the nanostructures formed in the selenic acid are 1.3-fold thinner and better self-ordered than their counterparts formed from the same Al/Nb precursor bilayer in a reference oxalic-acid electrolyte. The nanostructures have a dual (core/shell) composition: the inner material (the core) is stoichiometric Nb 2 O 5 , whereas the outer layer (the shell) is a few nm-thick substoichiometric NbO x mixed with Al 2 O 3 . The composite-ceramic nanoarrays grow doped with selenium species such as selenate (SeO 4 2− ) and selenide (Se 2− ) anions originating from the electrolyte and migrating inward under the high electric field. The incorporated Se species do not contribute to photoluminescence emission nor hinder the Raman signal from the nanoarrays, which makes them highly promising as Nb 2 O 5 -based SERS biosensing substrates. The planar PAA-inbuilt Se-doped Nb 2 O 5 –Al 2 O 3 nanostructured ceramic film performs like a high-k low-loss low-leakage-current dielectric promising for on-chip integration. More potential applications of the Se-doped ceramic nanoarrays developed here include biomedical antibacterial coatings, advanced superhydrophobic surfaces, gas-sensing, and catalytic layers. 1. Introduction Niobium pentoxide (Nb 2 O 5 ) and its compounds with other transition-metal oxides are attractive ceramic materials for catalysis, biomedicine, and surface protection applications [1–3]. Niobium pentoxide is also actively utilized as a building block for photovoltaic, sensing, energy storage, electrochromic, electronic, and optical devices [4–9]. For many modern applications, nanostructuring of Nb 2 O 5 -based ceramics is required to enhance the material’s characteristics, increase the surface-to-volume ratio, and possibly introduce novel properties. Commonly used nanostructuring techniques are the sol-gel, ion-beam sputtering, chemical vapor deposition, and atomic layer deposition [7, 10,11]. A method based on electrochemical anodizing of a thin layer of aluminum (Al) superimposed on a layer of niobium (Nb) has recently been developed to synthesize self-ordered arrays of Nb 2 O 5 nanostructures with controllable dimensions [12]. First, the Al layer is anodized to form a porous anodic alumina (PAA) film. The underlying Nb is then anodized through the superimposed PAA layer (the so-called PAA-assisted anodizing), forming Nb 2 O 5 nuclei within and beneath the barrier layer of the PAA film. Subsequent PAA-assisted high-voltage re-anodizing transforms the Nb 2 O 5 nuclei into longer oxide nanostructures such as columns or rods via the growth of niobium oxide inside the PAA nanopores [13]. Selective dissolution of the PAA overlayer yields arrays of upright-standing and spatially separated Nb 2 O 5 semiconductor nanostructures advantageous for gas-sensing and surface-finishing applications [2,4,14]. Precise control over the surface morphology and chemical composition of Nb 2 O 5 nanostructures is crucial for advancing existing and * Corresponding author. Department of Microelectronics, Faculty of Electrical Engineering and Communication, Brno University of Technology, Technicka 10, 616 00, Brno, Czech Republic. E-mail address: [email protected] (A. Mozalev). Contents lists available at ScienceDirect Ceramics International journal homepage: www.elsevier.com/locate/ceramint https://doi.org/10.1016/j.ceramint.2023.08.134 Received 25 April 2023; Received in revised form 30 July 2023; Accepted 11 August 2023 Ceramics International 49 (2023) 34712–34725 34713 developing new applications of niobium-oxide-based nanomaterials [15]. The fundamental and functional properties of the PAA-assisted Nb 2 O 5 ceramic nanoarrays can be controlled by simply adjusting the electrical and electrolytic formation conditions [12,14,16]. The nature of the anodizing electrolyte and the formation voltage decide the areal density and diameters of the PAA nanopores. In turn, these PAA properties govern the population density and diameters of the PAA-assisted niobium-oxide nanostructures. The relation has been experimentally proved in several anodizing electrolytes such as sulphuric, oxalic, phosphoric, and citric acids [2,4,12,13]. An aqueous solution of selenic acid (H 2 SeO 4 ) has recently been explored as a novel PAA-forming electrolyte generating substantially thinner pores than any other acid electrolyte at the same formation voltages [17]. Consequently, the porosity of the PAA films grown in selenic acid (2–3%) decreases by 3-fold compared with other anodizing electrolytes (6–10%). The uniquely low porosity of PAA formed in selenic acid appeared highly advantageous for fabricating photonic crystals and synthesizing metamaterials [18,19]. Suppose niobium can be anodized through the PAA films formed in selenic acid. In that case, the unique porous-cellular structure of the PAA films might be ’transferred’ to the anodic niobium oxide and promote the growth of extremely thin but long-distanced niobium-oxide nanocolumns or rods. Such nano-morphology could be in demand for modulating Nb 2 O 5 surface-plasmon resonance (SPR) because the effect largely depends on the relation between the width and spatial separation of the nanostructures [19]. The possibility to tune the SPR of the Nb 2 O 5 nanoarrays makes them particularly promising in the emerging niobium-oxide application as a substrate for surface-enhanced Raman spectroscopy (SERS) [20]. Moreover, due to the lowered contact area between the liquid and the solid, the extremely thin and long-distanced Nb 2 O 5 nanostructures are preferred for engineering superhydrophobic ceramic coatings [2,21]. Further, the extraordinarily thin PAA nanopores formed in selenic acid may effectively suppress the field crystallization of growing niobium oxide by limiting the space available for crystal formation. Such arrays of amorphous Nb 2 O 5 nanostructures can hardly be synthesized by other methods and are in demand for various electric/dielectric applications due to isotropy of properties, lack of interfacial defects, and absence of leakage paths through the crystalline grain boundaries, as reported elsewhere [22]. It is known that electrolyte-derived species readily incorporate into both PAA and PAA-assisted niobia nanostructures during anodization and affect the oxides’ chemical composition and functional properties [13]. For example, the incorporation of phosphorous (P), carbon (C), sulfur (S), chromium (Cr), or boron (B) from acid electrolytes may result in oxide coloration, reduced optical transparency, and wide-spectrum fluorescence [23]. On the other hand, the PAA formed in selenic acid is reportedly highly transparent and produces no unwanted fluorescence [24–26]. Therefore, it is anticipated that PAA-assisted niobium anodizing in selenic acid may result in nanostructured Nb 2 O 5 ceramics with similarly unhindered optical properties. The lack of optical-signal contamination is especially desired for the emerging application of Nb 2 O 5 in SERS due to the fundamental intersection of Raman scattering with fluorescence [27]. Furthermore, Se species are known to have potent antibacterial and anti-oncogenic properties. Incorporating Se species into PAA-assisted Nb 2 O 5 ceramic nanostructures via anodizing in selenic acid can boost the oxide’s biomedical characteristics and facilitate its application as implant coating [28,29]. High anti-oxidative properties of adsorbed or incorporated Se species can also enhance the activity of the nanostructured Nb 2 O 5 and Nb 2 O 5 /Al 2 O 3 catalysts [30–32]. Lastly, Se might bond with Nb to form a NbSe 2 compound with the nanoscale periodic morphology desired for emerging electronic applications [33]. In the present work, for the first time, Nb 2 O 5 -based ceramic nanoarrays were fabricated via the PAA-assisted anodizing and highervoltage re-anodizing of a thin Nb film in selenic acid. The influence of the formation conditions on the morphology, chemical composition, mixing with Al 2 O 3 , selenium incorporation, and several practically important functional properties of the Se-doped niobium-oxide-based nanoceramics were studied by high-resolution scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), Raman spectroscopy, photoluminescence (PL) spectroscopy, and electrical/ dielectric measurements. To prove the morphological advantages of the PAA-assisted ceramic nanoarrays formed in the selenic-acid electrolyte, they were directly compared with those synthesized in an oxalic-acid electrolyte at the same formation voltages. A mechanism for the PAAassisted selenic-acid niobium anodizing was proposed and experimentally justified. Potential applications of the Se-doped Nb 2 O 5 -based ceramic nanoarrays as advanced SERS biosensing substrates, superhydrophobic surfaces, biomedical coatings, catalytic, gas-sensing, and high-k planar nanocomposite dielectric layers for on-chip integration were experimentally and theoretically appraised. 2. Experimental section 2.1. Sample preparation A standard 4 ″ Si wafer covered by a 380-nm SiO 2 layer was used as the starting substrate. A 140-nm niobium layer followed by a 1000-nm aluminum layer was deposited on the substrate by the magnetron sputtering of Nb (99.95%) and Al (99.999%) targets to make a precursor Al/Nb bilayer. The wafer with the deposited bilayer was cut into ca. 1 cm ×1 cm pieces. The individual samples were anodized in 1.5 M H 2 SeO 4 aqueous solution at room temperature (21 ◦C). A custom-made polytetrafluoroethylene two-electrode cell securing a 0.8 cm 2 circular working area was used as the anodizing bath [34]. A 6 mm in diameter glassy carbon rod, distanced by 3 mm from the sample surface, was employed as the cathode. The PAA films were formed at a constant current density of either 5.0 or 0.5 mA cm −2 , resulting in a steady-state voltage of 42 or 12 V, respectively. Selected samples were re-anodized using a high-speed (10 V s −1 ) potentiodynamic polarization up to 120 or 80 V, followed by a potentiostatic polarization to ensure uniform oxide growth across the sample surface. The PAA overlayer was dissolved from selected samples in an aqueous solution of 0.2 М Cr 2 O 3 and 0.45 М H 3 PO 4 at 65 ◦C for 1 h [35]. Hereafter, samples with the dissolved PAA layer are named the ’PAA-free’ samples. After the anodizing/re-anodizing and PAA dissolution, the samples were thoroughly rinsed in Milli-Q® ultrapure water, dried in a nitrogen stream, and heated in an oven at 393 K for 1 h to desorb the water molecules from the oxide surface. 2.2. Sample analysis The surfaces and cross-fractures of PAA-free samples were examined in an FEI Verios 460L High-Resolution SEM operated at 15 kV accelerating voltage and 0.8 nA probe current. The chemical composition of the film surfaces was analyzed by XPS in a Kratos Axis Ultra DLD spectrometer using a monochromatic Al K α source. The X-ray emission power was 150 W with a 15 kV accelerating voltage focused to a 300 μ m ×700 μ m spot. Typical operating pressures were better than 10 −9 Torr. The emitted electrons were detected by a hemispherical analyzer at fixed pass energies of 160 eV for the survey and 20 eV for the highresolution spectra. The Kratos charge neutralizer system was used for all specimens. The depth profiling was performed using an argon-ion beam of 5 kV provided by a standard ion gun with a sample current of 750 nA scanned over a surface window of 2 mm ×2 mm for 90 and 270 s, resulting in slight surface etching. The sputtering-beam incidence angle was 45◦relative to the substrate. Spectra were analyzed using CasaXPS (ver. 2.3.17) software. GL(30) profiles, defined in CasaXPS, were used for all components besides the metallic core lines of Nb 3d, for which asymmetric profiles in the form of LA(1.2,5,12) were used. A standard Shirley background was used in all fitted spectra. Spectra from all samples were charge corrected to give the adventitious C 1s spectral K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34714 component (C–C, C–H) binding energy of 284.8 eV. The deconvolution of C 1s spectra was performed as described in previous works [13]. Quantitative analysis was performed using the narrow spectra and relative sensitivity factors from CasaXPS suitable for Kratos Axis Ultra, referenced to F 1s. Selected PAA-free samples were analyzed by PL spectroscopy using a chopped Kimmon IK Series He–Cd laser (325 nm and 40 mW). The fluorescence was dispersed with an Oriel Corner Stone 1/8 74000 monochromator, detected using a Hamamatsu H8259-02 with a socket assembly E717-500 photomultiplier, and amplified through a Stanford Research Systems SR830 DSP. Selected PAA-free samples were analyzed by Raman spectroscopy using the WITec confocal Raman imaging system alpha300 R. The samples were illuminated by a 532 nm laser with power reduced to 1 mW and focused by 100×objective (NA 0.9). 2.3. Electrical characterization Electrical measurements were conducted using test metalinsulator-metal (MIM) devices. A 100 nm-thick Nb layer followed by a 65 nm-thick Al layer (Al/Nb bilayer) were deposited by the magnetron sputtering onto a SiO 2 -coated Si wafer. A piece of 1.5 cm ×1.5 cm cut from the wafer was anodized at 0.5 mA cm −2 and then re-anodized to 70 V at a rate of 10 V s −1 using the methodology and anodizing setup described in section 2.1. Such anodizing was expected to generate a planar film where the PAA-assisted Nb 2 O 5 nanostructures grow up to the PAA-surface level. For making the top electrodes, a layer of ~100 nmthick platinum (Pt) was sputter-deposited through an array of 10 μ m × 10 μ m square openings in a thin ceramic plate used as the shadow mask. The unoxidized Nb layer that remained under the anodic film after the re-anodizing served as the bottom electrode. The assembled MIM microdevices were mounted to Cascade Microtech MPS 150 Probe Station connected to an Agilent Technologies Precision LCR Meter E4980A and a Keithley 4200A-SCS Parameter Analyzer. Unless otherwise specified, the top Pt contacts were used as a cathode in electrical tests. 3. Results and discussion 3.1. Anodizing behavior Typical voltageand current-time responses during the 5.0 and 0.5 mA cm −2 anodizing of the sputter-deposited Al/Nb bilayer in 1.5 M H 2 SeO 4 aqueous solution, followed without interruption by 120 and 80 V high-speed re-anodizing in the same electrolyte, are presented in Fig. 1. Both responses generally exhibit similar behavior and can be divided into three distinct stages. 1st stage: The voltage-time curves are typical for the porous anodizing of aluminum in acid electrolytes, including the PAA formation on aluminum foils in selenic-acid electrolytes of comparable concentration reported elsewhere [24,36]. Systematic growth of PAA film is associated with the periods when the voltage attains the steady-state values, U st , of ~42 V (hereafter the ’high-voltage’ sample) or ~12 V (the ’low-voltage’ sample). Once the anodizing front reaches the niobium underlayer, the voltage begins to increase, and the process is automatically switched into the potentiostatic mode at 45 or 13 V, respectively. 2nd stage: The behavior is determined by the ’smart’ PAA-assisted anodizing of the niobium underlayer: arrays of niobium-oxide nanostructures nucleate and grow downward into the Nb metal and upward into the PAA barrier layer [13]. The applied voltage and the length of the current decay impact the oxide growth. 3rd stage: The high-speed potentiodynamic re-anodizing up to 120 V (Fig. 1a) or 80 V (Fig. 1b) is followed without interruption by a short potentiostatic polarization with a current decay. The niobium oxide is expected to grow further and partially fill the PAA pores at this stage [13]. This process will likely compete with expanding the nanostructures’ bases, which can merge in a continuous bottom oxide layer. The merging of the bases depends on a combination of factors like the nanostructures’ dimensions, the ratio of ionic resistances between alumina and niobia, ionic transport numbers, anion incorporation, oxide crystallization behavior, etc. In the following text, the samples will be distinguished by their U st during the 1st stage (Fig. 1): The highand low-voltage samples prepared by the selenic-acid anodizing to the end of the 2nd stage will be named SE42 A and SE12 A , respectively. The highand low-voltage samples prepared by the selenic-acid anodizing followed by reanodizing to the end of the 3rd stage will be named SE42 R and SE12 R , respectively. 3.2. Film morphology Fig. 2 shows SEM images of the PAA-free SE42 A and SE42 R samples. The SE42 A surface (Fig. 2a) reveals an array of upright-standing, spatially separated nanosized protrusions, presumably of niobium oxide. The protrusions are surrounded by an unoxidized metal, likely Al. From Fig. 2c, the shape of the protrusions resembles a ’goblet’. The Fig. 1. Voltageand current-time responses during the anodic processing of Al (1000 nm)/Nb (140 nm) bilayer in 1.5 M H 2 SeO 4 aqueous electrolyte at 21 ◦C for forming Nb 2 O 5 nanoarrays: 1st stage – the galvanostatic aluminum anodizing for forming porous anodic alumina (PAA) at (a) 5 mA cm −2 (steady-state voltage U st =42 V) and (b) 0.5 mA cm −2 (U st =12 V); 2nd stage – the potentiostatic PAA-assisted niobium anodizing at (a) 45 V and (b) 13 V; 3rd stage – the potentiodynamic PAA-assisted niobium re-anodizing with a rate of 10 V s −1 to (a) 120 V and (b) 80 V followed by potentiostatic polarization at the same voltages. The samples are named relating to the selenic (SE) acid, U st - values (12 or 42 V), and the formation stages (A: anodizing till the end of 2nd stage, R: re-anodizing: till the end of 3rd stage), as specified above the plots. K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34715 nanogoblets’ apexes gradually transit to relatively thinner middle parts, often composed of several distinguished nanoroots that expand to form broader bases. The bases are separated from each other by unoxidized niobium metal. Fig. 2d–f depicts the SE42 R film morphology. From the surface view, the horizontal projections of the protrusions thicken with re-anodizing. The protrusions become substantially longer and acquire the shape of rods, each having a single root that is relatively thinner than the rod’s body. Although more expanded, the bases of the nanorods are still separated by the unoxidized Nb metal. The SEM analysis cannot reliably confirm the presence of unoxidized aluminum. SEM images of the SE12 A and SE12 R samples are shown in Fig. 3. The upright-standing sub-10 nm oxide protrusions populate the SE12 A surface (Fig. 3a). They have significantly smaller diameters and higher population density than the SE42 A nanogoblets. From Fig. 3c, the nanoprotrusions are dot-shaped. The individual nanodots seem separated by the unoxidized Nb metal. The insufficient SEM resolution makes it impossible to judge the presence of an unoxidized aluminum. The surface view in Fig. 3d shows that the oxide protrusions expand with re-anodizing. The apexes of the protrusions tend to conglomerate in small groups, likely due to van der Waals forces. From Fig. 3f, the reanodizing noticeably elongates the SE12 A oxide nanodots, resulting in an array of rod-like nanostructures. A substantial difference between SE12 R and the other films is that the bases of the nanorods merge in a continuous 30 nm-thick oxide layer that buffers the rods from the unoxidized Nb metal. Apparently, there is no unoxidized Al at the film surface. 3.3. Comparison between selenicand oxalic-acid anodizing The PAA-assisted niobium-oxide nanostructures grow inside and beneath the pores during the anodizing and re-anodizing [13]. Therefore, the PAA porous-cellular structure decides the diameters and surface distribution of the niobium-oxide nanoprotrusions. The pore center-to-center distance, D int , is known to depend on U st (the 1st stages in Fig. 1a and b) through a constant k =2.5 nm V −1 per linear relation D int =kU st [37]. Consequently, the center-to-center distance between the niobium-oxide nanostructures, D cent , should be dependent on D int and U st : Dcent =Dint =kUst (1) Since the PAA layer can be grown in the selenicand oxalic-acid electrolytes at the same U st , one may expect that the subsequently Fig. 2. SEM (a and d) vertical surface views, (b and e) 30 ◦-tilted surface views, and (c and f) 3D cross-fracture views of niobium-oxide nanoarrays derived from an Al/Nb bilayer anodized and re-anodized in 1.5 M selenic acid as defined and named in Fig. 1a. The PAA layers were selectively dissolved away before SEM observation. All undefined scale bars are 400 nm. Fig. 3. SEM (a and d) vertical surface views, (b and e) 30 ◦-tilted surface views, and (c and f) 3D cross-fracture views of niobium-oxide nanoarrays derived from an Al/Nb bilayer anodized and re-anodized in 1.5 M selenic acid as defined and named in Fig. 1b. The PAA layers were selectively dissolved away before SEM observation. All undefined scale bars are 400 nm. K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34716 formed niobium-oxide nanostructures would have similar D cent . On the other hand, the PAA-assisted selenic-acid anodizing is expected to lead to niobium-oxide nanostructures with noticeably smaller diameters, d n , due to the smaller PAA pore diameters, d p , compared with the PAAassisted oxalic-acid anodizing. Although the nanostructures formed in the selenic acid may have a D cent similar to those formed in the oxalic acid, they should possess a smaller d n and, therefore, acquire an advanced surface structuring, which is not attainable in other electrolytes. Since sensing, SPR, surface protection, self-cleaning, and biomedical properties of the oxide depend on the nanostructures’ geometry and their mutual arrangement, the new morphology of the PAAassisted nanostructures formed in the selenic acid could be especially relevant for the emerging applications of Nb 2 O 5 ceramics. For direct comparison, samples were prepared in 0.6 M oxalic acid (H 2 C 2 O 4 ) via the anodizing/re-anodizing of the same Al/Nb precursor bilayer. To this end, the current densities for PAA formation were adjusted to allow U st of 42 and 12 V – the same as for the selenic-acid anodizing. Similarly, the niobium underlayer was re-anodized in the oxalic acid via the potentiodynamic polarization at 10 V s −1 up to 120 or 80 V. The samples formed by the PAA-assisted oxalic-acid anodizing will be named OX42 A , OX42 R , OX12 A , and OX12 R , corresponding to their counterparts SE42A, SE42 R , SE12 A , and SE12 R formed in the selenic acid. Fig. S1–S4 shows a side-by-side comparison of the nanostructures formed in the two electrolytes. The SEM data were quantified using ImageJ software to assess the mean values of nanostructures’ dimensions summarized in Table 1. The D cent and d n values were plotted and best fitted with a normal distribution (Figs. S5–S7). D cent and population density, ρ N , were evaluated only for the anodized samples because the re-anodizing elongates the niobium-oxide nuclei without affecting their mutual arrangement. Moreover, the thin SE12 R and OX12 R nanorods tend to agglomerate, complicating image evaluation. From Figs. S1, S3, S6, and S7, one may see that the nanodots and nanogoblets formed in the selenic acid are substantially better selfordered than the corresponding nanostructures formed in the reference oxalic-acid electrolyte. Despite the same U st , the nanostructures formed in the selenic acid reveal about 10% smaller D cent and correspondingly slightly larger ρ N than the nanostructures formed in the oxalic acid. Similarly, a ~10% smaller D int was reported by Kikuchi et al. for the PAA formed on aluminum foil in selenic-acid electrolytes of comparable concentrations [24]. No electrolyte-dependent differences were revealed for the length of the niobium-oxide nanoprotrusions, h (Table 1, Figs. S1–S4). The SE12 A and OX12 A nanodots having h =20 nm are both elongated to the 125 nm-tall SE12 R and OX12 R nanorods by the 80 V re-anodizing. Similarly, the SE42 A and OX42 A nanogoblets having h =75 nm grow further to become the 210 nm-tall SE42 R and OX42 R nanorods after the 120 V reanodizing. The SE12 R and OX12 R nanorods are anchored to a continuous ~30 nm-thick bottom-oxide layer. In contrast, the SE42 A and OX42 A nanogoblets and the SE42 R and OX42 R nanorods reside on convexshaped bases separated by the unoxidized Nb. The bases of the OX42 R nanorods seem slightly larger than those of the SE42 R nanorods. Further morphological differences between the nanostructures formed in the selenic and oxalic acids are associated with their diameters d n . Based on the SEM analysis, we note the following major trends: (1) d n decreases by ~30% for the nanostructures formed in the selenic acid. The effect is apparently due to the relatively thinner PAA nanopores grown by selenic-acid anodizing. (2) d n increases by ~35% due to the re-anodizing in both acids. This effect could be explained by the partial growth of the rods within the inner part of the PAA cell walls, which is enabled by the upward migration of Nb cations through the PAA cell walls [13]. The cell walls may provide an effective pathway for ion transport due to defects such as vacancies and electrolyte-derived impurities, as described elsewhere [13,38]. The dimensions of PAA films formed in both acids were assumed based on the measured D cent and theoretical considerations summarized in Fig. S8. One may see that d n systematically exceeds d p in all the samples (Table 1, Fig. S9). The degree of niobium-oxide expansion into the PAA cell walls was assessed by comparing the porosity [37] of the PAA films, P =0.907d p 2 /D int 2 , with the niobium-oxide nanostructures’ projected area, A =0.907d n 2 /D cent 2 . It was revealed that A consistently exceeds P. In addition, A increases approximately by 2-fold after the re-anodizing. Another important revelation is that the A-value of the low-voltage samples (12/80 V) is about two times higher compared with the high-voltage samples (42/120 V) regardless of the electrolyte. Such a difference in A-values may arise because the anodic-oxide growth within the thinner pores of the low-voltage PAA is more obstructed. In such thin pores, the migration of Nb cations becomes more prevalent within the cell walls under the high electric field. Additionally, the effect may be due to the different ratios of the re-anodizing and steady-state voltage: for the low-voltage samples (80/12 V), the voltage ratio is 6.7, while for the high-voltage samples (120/42 V), the ratio is only 2.9. It is assumed that the higher the ratio, the more intensive the cation migration within the PAA cell walls. The most practically important finding is that the nanostructures formed in the selenic acid exhibit, on average, a 1.5-fold lower A than their counterparts formed in the oxalic acid. Since the nanostructures’ projection area (A) is the criterion to access the fundamental, structural, and functional benefits of the PAA-assisted oxide nanorods, the selenicacid anodizing is experimentally justified. The high-voltage samples look most advantageous for niobium-oxide nanoarray applications, such as SPR and self-cleaning surfaces, where the lowest possible A is desired. The structural, compositional, and application-related benefits of the niobium-oxide nanostructured ceramics formed by the PAA-assisted selenic-acid anodizing are further discussed in section 3.6. 3.4. Nanostructures’ chemical composition The examination of chemical composition and bonding states in the Table 1 Comparison of the PAA-assisted niobium-oxide nanoarrays synthesized in 1.5 M selenic acid (H 2 SeO 4 ) and 0.6 M oxalic acid (C 2 H 2 O 4 ) electrolytes under the same anodizing/re-anodizing voltages. 1.5 M H 2 SeO 4 (selenic acid) 0.6 M C 2 H 2 O 4 (oxalic acid – reference) SE42 A SE42 R SE12 A SE12 R OX42 A OX42 R OX12 A OX12 R Current density /mA cm −2 5.0 0.5 12.2 1.3 Steady-state voltage (U st ) /V 42 12 42 12 Re-anodizing voltage /V – 120 – 80 – 120 – 80 Center-to-center (D cent ) /nm 96 ±9 27 ±6 104 ±14 30 ±7 Population density ( ρ N ) /cm −2 1.2 ⋅ 10 10 1.6 ⋅ 10 11 1.1 ⋅ 10 10 1.3 ⋅ 10 11 Length (h) /nm 75 210 20 125 75 210 20 125 Nanostructure diameter (d n ) /nm 24 ±4 35 ±6 9 ±2 13 ±3 32 ±6 45 ±7 12 ±3 17 ±4 PAA pore diameter (d p ) /nm 16.2 4.2 24.2 7.2 Nanostructure projected area (A) /% 5.7 12.0 10.0 21.0 8.6 17.0 14.5 29.1 PAA porosity (P) /% 2.5 2.2 4.9 5.2 K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34717 niobium-oxide nanoarrays formed by the PAA-assisted selenic-acid anodizing was performed by XPS (Fig. 4), mainly for the PAA-free samples. In addition, a SE42 R sample with partially dissolved PAA, as sketched in the inset of Fig. 4m, was analyzed. In such a sample, about ~40 nm of the upper parts of the rods protrude outwards from the PAA surface level. This sample with the partially-dissolved PAA is hereafter named SE42 Rpd . For comparison, a compact niobium-oxide film was prepared by potentiodynamic anodization of an Nb layer in 1.5 M selenic acid at a rate of 0.1 V s −1 up to 10 V. The initial Nb layer was obtained by dissolving the Al layer from the Al/Nb precursor bilayer in 1 wt% NaOH solution. XP spectra were recorded on the as-prepared (not sputtercleaned) surfaces of all samples. In addition, the SE42 R sample was analyzed following two short Ar-ion sputter cycles to remove surface carbon contamination and a portion of the outer film material. The presence of C, Nb, O, Al, and Se was identified in the survey spectra of all samples; additionally, P and Cr were detected on the PAAfree surfaces. Narrow-scan C 1s, Nb 3d, O 1s, Al 2p, and Se 3d spectra were collected to analyze their core levels and bonding states. The experimental and fitted Nb 3d, Al 2p, and Se 3d spectra for the asreceived surfaces are shown in Fig. 4. The corresponding spectra for the Ar-ion-sputtered SE42 R film are presented in Fig. S10. Fig. 4a shows the Nb 3d spectrum of the SE42 A surface. Three doublets of appropriately constrained peaks are used to fit the spectrum (Nb 3d 5/2 and d 3/2 with the fixed peak-separation energy of 2.75 eV, the fixed intensity ratio of 3:2, and the full width at half maximum (FWHM) equal for oxide components) [13]. The highest binding-energy (BE) doublet (Nb 3d 5/2 at 206.85 eV) is due to Nb 5+ cations in the oxide [13]. The lowest BE doublet (Nb 3d 5/2 at 201.2 eV) of a comparable intensity is associated with metallic Nb 0 that separates the nanogoblets’ bases Fig. 4. Experimental and curve-fitted narrow-scan (first row) Nb 3d, (second row) Al 2p, and (third row) Se 3d XP spectra of PAA-free (a–c) SE42 A , (d–f) SE42 R , (g–i) SE12 A , (j–l) SE12 R surfaces. (m) Nb 3d, (n) Se 3d, and (n-inset) Al 2p XP surface spectra of SE42 R sample having partially-dissolved PAA layer. (o) Nb 3d and (p) Se 3d XP surface spectra of a compact niobium-oxide film prepared for comparison by anodizing niobium at 10 V in 1.5 M selenic acid. K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34718 (Fig. 2a–c). The amount of metallic Nb 0 is assessed as 53% of all Nb species. The third, lowest-intensity doublet, shifted to −1.35 eV from the Nb 5+ component (with Nb 3d 5/2 at 205.5 eV), is due to the presence of Nb 4+ cations in the film, with a Nb 5+ :Nb 4+ ratio of 97:3. Due to the high amount of Nb 5+ cations and the low amount of Nb 4+ cations in the surface composition, the outmost film layer is assessed as stoichiometric Nb 2 O 5 mixed with a minor amount of oxygen-deficient niobia. The Nb 3d spectrum of the SE42 R surface (Fig. 4d) is fitted with three doublets (with Nb 3d 5/2 at 207.1, 205.75, and 201.5 eV), which are similarly associated with Nb 5+ , Nb 4+ , and Nb 0 species. The Nb 5+ :Nb 4+ ratio is 95:5, close to that of the SE42 A surface. Therefore, the SE42 R nanorods also contain substoichiometric (oxygen-deficient) niobium pentoxide. The amount of Nb 0 at the SE42 R surface is much lower than that at the SE42 A surface (12% vs. 53% of all Nb species). The two-times sputtering of the sample surface with weak-energy Ar ions (Fig. S10) reduces the contribution of Nb 0 from 12 to 8%. This means that a portion of the outer material was sputtered away along with the carbon contamination. Sputtering also changes the Nb 5+ :Nb 4+ ratio to 93:7 by increasing the amount of Nb 4+ cations, which may be attributed to preferential oxygen sputtering. The presence of Nb 4+ cations (and lower-oxidation-state metal cations in general) at the surface layer is a common phenomenon for the oxides formed by PAA-assisted anodizing [13,38,39]. The Nb 3d spectrum of the SE12 A surface (Fig. 4g) is fitted with three doublets (Nb 3d 5/2 at 207.15, 203.6, and 201.5 eV), the highand lowBE doublets being due to Nb 5+ and Nb 0 . The amount of metallic niobium is relatively higher in this sample (70% of all Nb species). The lowestintensity doublet shifted to −3.55 eV from the Nb 5+ peak is assigned to Nb 2+ cations in the anodic oxide [13]. The Nb 5+ :Nb 2+ ratio is 85:15, indicating the presence of stoichiometric Nb 2 O 5 mixed with a relatively high amount of NbO suboxide. The spectrum of the SE12 R surface (Fig. 4j) consists of a highintensity doublet with BE of 207.15 eV for Nb 3d 5/2 associated with Nb 5+ cations and a low-intensity doublet with BE of 204.7 eV for Nb 3d 5/ 2 . The 2.45 eV distance between them indicates that the low-intensity doublet is due to Nb 3+ cations in the oxide. As the Nb 5+ :Nb 3+ ratio is low (98:2), the low-intensity peak is associated with small amounts of Nb 2 O 3 suboxide mixed with Nb 2 O 5 . No metallic niobium species are detected at the SE12 R surface. The high degree of oxide stoichiometry and the lack of metallic Nb on the surface may favor the application of such film as dielectric, which will be considered in Section 3.6.2. The Nb 3d spectrum of the SE42 Rpd surface (Fig. 4m) is fitted with two doublets (Nb 3d 5/2 at 207.4 and 206.25 eV), associated with Nb 5+ and Nb 4+ cations, respectively. The lower-intensity peak, shifted to only −1.15 eV from the Nb 5+ one, may be due to a more substoichiometric Nb 2 O 5 owing to oxygen vacancies [40]. Since only the upper 40 nm-long parts of the rods are exposed for analysis (the lower parts of the rods are still covered by the remaining portion of PAA, as shown in the inset of Fig. 4m), the decreased Nb 5+ :Nb 4+ ratio of 66:34 indicates that the apexes of the nanorods contain a substantial amount of reduced oxide. This implies that the niobium oxide becomes relatively more reduced towards the upper parts of the rods. As expected, no metallic Nb is detected because the PAA layer covers the metallic residues between the rods’ bases. For comparison, the Nb 3d spectrum of the compact niobium-oxide surface is shown in Fig. 4o. A single doublet reproduces the spectrum with Nb 3d 5/2 at 207.3 eV attributed to Nb 5+ cations. The surface layer of the compact anodic niobium oxide is therefore composed of fully stoichiometric Nb 2 O 5 . Summarizing the Nb 3d spectra analysis, the surfaces of the anodized and re-anodized nanostructures are predominantly composed of stoichiometric Nb 2 O 5 mixed with a small portion of substoichiometric Nb 2 O 5 (reduced or oxygen-deficient) or niobium suboxide (NbO or Nb 2 O 3 ). The oxides could be mixed at the molecular level or present as nanoinclusions and unevenly distributed along the nanostructures. The most reduced oxide forms in the SE12 A nanodots. Comparing the relative amount of the reduced oxide in the SE42 R and SE42 Rpd surfaces, we assume that the oxide substoichiometry increases toward the tops of the protrusions. All samples, except SE12 R , are confirmed to have metallic Nb separating the bases of the oxide nanostructures. The absence of Nb metal at the SE12 R surface implies that the bases of the oxide nanorods fully merge to form a continuous bottom-oxide layer during re-anodizing, as also seen in the cross-fracture SEM images in Fig. 3f. The Al 2p spectra of the SE42 A and SE42 R surfaces (Fig. 4b and e) are fitted with two single symmetrical peaks: The high-intensity peak is associated with Al 2 O 3 (74.5 and 74.65 eV), and the low-intensity peak is due to metallic Al (71.9 and 72.15 eV). The metallic Al originates from the network of Al residues remaining around the niobium-oxide nanostructures over the metallic Nb (Fig. 2). After two times Ar-ion sputtering of the SE42 R surface, the amount of metallic Al decreases from 18 to 13%, which indicates the removal of a thin surface layer. In the Al 2p spectrum of SE42 Rpd (the inset in Fig. 4n), only the peak associated with Al 2 O 3 is present (74.55 eV). The Al 2p spectrum of the SE12 A surface (Fig. 4h) is similarly fitted with a dominating peak assigned to Al 2 O 3 (74.45 eV) and a lowerintensity peak assigned to metallic Al (71.85 eV). However, here the metallic contribution is substantially minor compared with the SE42 A surface. This implies that aluminum remains at the SE12 A surface in a substantially smaller amount than at the SE42 A surface. The Al 2p spectrum of SE12 R (Fig. 4k) has a single Al 2 O 3 -related peak (74.45 eV), indicating the absence of metallic Al. The metallic contributions to the Nb 3d and Al 2p spectra originating from the corresponding metallic residues around the nanostructures’ bases were subtracted from the overall content of the elements to obtain an approximate surface amount of Nb and Al in the oxides. For Al, the corresponding native oxide was also considered and subtracted [41]. Provided that at.%(Al +Nb) =100%, the SE42 A and SE42 R nanostructures are estimated to have 65 and 61 at.% of Al, respectively, while the SE12 A and SE12 R counterparts show 81 and 47 at.% of Al, respectively. Therefore, the surface of the nanostructures is composed of mixed Nb 2 O 5 –Al 2 O 3 , whereas their interior is most likely alumina-free Nb 2 O 5 [13]. Similarly, a large amount of surface Al (about 75 at.%) has previously been reported for the PAA-assisted N-doped TiO 2 nanocolumns [39]. The noisy Se 3d spectra, associated with the low Se concentration, and their overlapping with the Nb 4s peaks complicates the analysis of the Se 3d spectra. The presence of multiple components is assumed and confirmed by comparing the spectra of the various samples, fitting them, and verifying the peak positions with the literature-reported values: Nb 4s peak of Nb 5+ cations, Nb 4s peak of Nb 0 atoms, Se 3d peaks of selenate anions (SeO 4 2− ), and Se 3d peaks of selenide anions (Se 2− ). A single peak of Nb 4s transition for Nb 5+ was assumed at ~60.5 eV [42]. The Nb 0 line was assumed at ~56.0 eV [42]. The fitting leads to 60.9 ±0.2 and 56.4 ±0.1 eV for the Nb 5+ and Nb 0 peaks, respectively. The area below the Nb 4s peaks was considered proportional to the area of the corresponding Nb 3d peaks (based on the relative sensitivity factors of 0.126 and 2.921 for Nb 4s and Nb 3d, respectively). However, the fitting results in the 2.0 ±0.1-fold lower-intensity for Nb 4s peaks than theoretically expected. The area of the Nb 0 peak was constrained as a multiple of the Nb 5+ peak to agree with the corresponding Nb 3d spectra. The Se 3d transitions were fitted as doublets with a constrained spin-orbit splitting of 0.86 eV [43], the ratio of the Se 3d 3/2 and 3d 5/2 components was fixed at 0.72 [43], and FWHMs were constrained to be equal within each component. The BEs of Se 3d 5/2 peaks of ~60 eV and ~54.5 eV were expected for SeO 4 2− and Se 2− anions, respectively [43, 44]. The fitted BEs of the Se 3d 5/2 peaks are 58.55–59.6 (avg. 58.8) eV for the selenate and 53.55–54.5 (avg. 54.0) eV for the selenide ions, agreeing with the literature. The Se 3d spectra are shown in the 3rd row of Fig. 4. They are all fitted with Nb 4s peaks for Nb 5+ and possibly Nb 0 , correlating with the corresponding Nb 3d spectra, and with two doublets of Se 3d peaks for K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34719 SeO 4 2− and Se 2− ions. Provided that at.%(SeO 4 2− +Se 2− ) =100%, the amounts of Se 2− anions are 61 at.% for the anodized (SE12 A and SE42 A ) and 49 at.% for the re-anodized samples (SE12 R and SE42 R ). After the two-times sputtering of the SE42 R surface, the amount of Se 2− ions increases from 49 to 64 at.%. On the other hand, a dominating amount of SeO 4 2− ions is found in the Se 3d spectrum of the SE42 Rpd sample (Fig. 4n), with only a minor contribution of Se 2− ions (~3 at.%). No selenides are assumed from the Se 3d spectrum of the compact niobium oxide (Fig. 4p), which is fitted only with one SeO 4 2− doublet (in addition to an Nb 4s peak of Nb 5+ ). The peak area of Se 3d components (i.e., without the contribution of Nb 4s) was used to calculate the amount of Se at the sample surfaces (provided at.%(Al +Nb +Se) =100%), after subtracting the metallic contribution from the Al 2p and Nb 3d spectra, as described above. The SE42 A and SE42 R surfaces have 5.3 and 3.2 at.% of Se, respectively, whereas the SE12 A and SE12 R surfaces have 3.9 and 4.9 at.% of Se, respectively. Sputtering the SE42 R surface affects the Se content slightly (2.8 at.%). The SE42 Rpd surface shows a little higher Se amount (4.3 at. %) than the PAA-free SE42 R . On the other hand, the surface of the compact niobium oxide has a substantially more considerable Se amount (9.6 at.%) relative to the nanostructured samples. In summary, the chemical composition of the outmost nanostructures’ material can be expressed as a mixture of Al 2 O 3 and niobium oxides: stoichiometric Nb 2 O 5 mixed with substoichiometric Nb 2 O 5 or with a suboxide expressed on average as NbO 2.5 ⋅xNb n + O n/2 . A portion of the O 2− anions in both oxides is replaced by SeO 4 2− and possibly Se 2− anions, as summarized in Table 2. There is 47–81 at.% of Al 2 O 3 at the sample surfaces, and about 1.5–3.0% of the O 2− anions are replaced by selenium species. The SE42 Rpd surface contains much more alumina on average (95 at.%), apparently due to the presence of the PAA matrix surrounding the tops of the rods, and 2.9% of the O 2− anions are replaced by selenium species. Contrarily to the nanostructured surfaces, the compact anodic niobium-oxide film contains neither Al 3+ nor Se 2− ions. It mainly comprises stoichiometric Nb 2 O 5 , in which 4.2% of the O 2− anions are replaced by SeO 4 2− anions. 3.5. Oxide growth model 3.5.1. Formation-morphology relationship Based on the present experimental findings and with reference to previous publications [12–14], we developed a model for the PAA-assisted growth of nanostructured niobium-oxide-based ceramic nanoarrays in the selenic-acid electrolyte, as sketched in Fig. 5. The growth steps depicted in Fig. 5 correspond to the film-formation stages in Fig. 1: (a and f) the steady-state PAA growth during the 1st stage, (b and g) the anodizing of the Nb underlayer during the 2nd stage, and (c and h) the re-anodizing of the Nb underlayer during the 3rd stage. Initially, the upper Al layer is anodized [45] at the high (42 V) or low (12 V) steady-state voltages. The SeO 4 2− electrolyte-derived anions are incorporated in the growing PAA. A small portion of the PAA at the barrier layer/aluminum interface, about 10% of the barrier-layer thickness, typically remains impurity-free [45,46]. When the PAA cells touch the Nb metal, niobium oxide begins to grow inside the PAA barrier layer, accompanied by the partial dissociation of the Al–O bonds under the high electric field, which is typical for some other valve metals [47–49]. Nanochannels of niobium oxide grow within the PAA barrier layer via the cross-migration of cations and anions. Finally, the growing oxide protrudes into the pores, and the nanochannels within the PAA barrier layer expand to form solid niobium-oxide bulges (Fig. 5b and g). The oxide also grows within the PAA cell walls due to the enhanced ionic transport through various defects in the outer cell-wall material: the incorporated electrolyte species, structural imperfections, vacancies, etc. [13]. Due to this phenomenon, the niobium-oxide nanostructures grow thicker than the PAA pores, and a mixed alumina-niobia region forms along the Nb 2 O 5 /Al 2 O 3 interface. The anodizing of the underlying Nb competes with the anodizing of the residual Al network surrounding the bottoms of the PAA cells. The Al residues may remain after completing the anodizing step or may be fully oxidized – the behavior depends upon the ratio of the ionic resistivities of anodic alumina and niobia. Re-anodizing to a higher voltage (Fig. 5c,h) results in proportionally elongating and thickening the niobium-oxide protrusions, further filling the pores and occupying more space within the PAA cell walls. New niobium oxide also keeps growing beneath the PAA cells due to the O 2− migration. The nanorods’ bases expand and deepen but do not merge in SE42 R (Fig. 5c). However, they merge into a continuous bottom-oxide layer in SE12 R (Fig. 5h). The different manners of the bottom-oxide growth in SE42 A -SE42 R and SE12 A -SE12 R pairs arise from dissimilar ionic resistivities of the two types of niobium-oxide nanoprotrusions. The higher the ionic resistivity of the anodic niobium oxide (closer to that of alumina), the more the bottom oxide expands under the pores. The differences in the niobiumoxide resistivities are mostly related to the variations in chemical compositions and dissimilar alumina-niobia mixing within the PAA barrier layer and cell walls. The second practically essential and closely related phenomenon is the presence or absence of unoxidized aluminum after completing the anodizing and re-anodizing stages. The ionic resistivities of the SE12 A and SE12 R niobium-oxide nanoprotrusions are higher than those of the SE42 A and SE42 R nanoprotrusions since only small traces of residual Al metal remain in SE12 A , while SE12 R is entirely Al-free. Oppositely, the relatively lower resistivity of the SE42 A and SE42 R niobium-oxide nanoprotrusions results in more aluminum remaining around the protrusions. Notably, the ionic resistivity of the Nb 2 O 5 nanostructures can be controlled by adjusting the anodizing process parameters and may be considered for engineering niobium-oxide nanoarrays for particular applications. It is possible to modulate the residual Al network on the sample surface to merge the nanostructures’ bases into a continuous Nb 2 O 5 bottom layer. For applications of the free-standing Nb-metalsupported Nb 2 O 5 nanostructures without any traces of Al metal, a NaOH Table 2 Surface chemical composition of the niobium-oxide nanostructures formed via the PAA-assisted anodizing/re-anodizing in 1.5 M selenic acid. Substoichiometric niobium pentoxide a NbO 2.5 ⋅xNb n+ O n/2 Al:Nb ratio Related to Se at.% of Se (Se +Al +Nb =100 at.%) % of O 2− replaced by Se species SeO 4 2− : Se 2− n x SE42 A 4 0.034 65:35 5.3 3.0 39:61 SE42 R 4 0.057 61:39 3.2 1.8 51:49 SE42 R after 1st sputtering 4 0.074 63:37 2.8 1.5 40:60 SE42 R after 2nd sputtering 4 0.081 62:38 2.8 1.5 36:64 SE42 Rpd 4 0.52 95:5 4.3 2.9 97:3 SE12 A 2 0.18 81:19 3.9 2.5 38:62 SE12 R 3 0.024 47:53 4.9 2.6 51:49 compact niobium oxide n/a n/a 0:100 9.6 4.2 100:0 a The outer material of the nanostructures was expressed as a mixture of Al 2 O 3 with niobium oxides (stoichiometric Nb 2 O 5 mixed with substoichiometric Nb 2 O 5 or with a suboxide, in average NbO 2.5 ⋅xNb n+ O n/2 ), in which some of the O 2− anions are replaced by SeO 4 2− (selenate) and Se 2− (selenide) anions. K. Kamnev et al. Ceramics International 49 (2023) 34712–34725 34720 etchant may be used to dissolve both the PAA and the remaining Al. Additionally, after dissolving the PAA and/or the remaining Al, the unoxidized Nb metal surrounding the nanorods may be anodized and transformed into a continuous compact Nb 2 O 5 layer of a desired thickness, as described elsewhere [50]. 3.5.2. Ionic transport and oxide growth Another essential feature that should be considered is the incorporation of SeO 4 2− (selenate) and Se 2− (selenide) anions in the nanoarrays. The SeO 4 2− anions are present in the anodizing electrolyte because of the complete dissociation of H 2 SeO 4 . These ions adsorb on the anode surface and migrate inward under the high electric field during the anodizing and re-anodizing (with a slower rate relative to O 2− (OH − ) anions). The following anode reactions express the niobium-oxide formation and selenate-ion incorporation: 2Nb 5+ +5O 2− → Nb 2 O 5 (2) 2Nb 5+ +5(1−y 1 )O 2− +5y 1 SeO 4 2− → Nb 2 O 5(1−y1) (SeO 4 ) 5y1 (3) where coefficient y 1 represents the fraction of O 2− anions replaced by selenate anions in the niobium oxide. The incorporation of selenide, or possibly hydrogen selenide (HSe − ) anions [51,52], in the outer nanostructures’ layer occurs after such anions form in the selenic-acid electrolyte of the given concentration and the corresponding pH due to the reduction of selenate anions on the cathode via a series of step reactions: SeO 4 2− +3H + +2e − → HSeO 3 − +H 2 O [51,52,53] (4) HSeO 3 − +5H + +4e − → Se 0 +3H 2 O [51,52,53] (5) HSeO 3 − +6H + +6e − → HSe − +3H 2 O [51] (6) Se 0 +H + +2e − → HSe − [51,52] (7) Reactions (4–7) can be summarized by the following overall cathode reaction: SeO 4 2− +9H + +8e − → HSe − +4H 2 O (8) The reduction of the negatively charged species at the cathode is enabled by the diffusion of anions from the bulk electrolyte across the Helmholtz layer, as explained elsewhere [54]. A red powder, presumably elemental Se, is observed on the cathode during the anodizing/re-anodizing. Considering the 3-mm distance between the electrodes in the anodizing setup, the formed selenide anions are expected to be transported from the cathode to the anode by diffusion within a few minutes, which happens already during the growth of the PAA overlayer. The fact that selenide anions are incorporated in the PAA-assisted niobium oxides (Fig. 4c,f,i,l) but are not present in the compact niobium oxide (Fig. 4p) confirms that the selenide anions did not diffuse toward the compact niobium oxide before its growth was finished (within ~2 min after applying the polarization). However, since the PAA-assisted niobium-oxide nanoarrays only start to grow after the PAA overlayer is fully formed, a substantial amount of selenide anions is already produced at the cathode. These selenide anions have enough time to diffuse toward the anode and enter the preformed PAA pores well before the beginning of anodizing and, especially, re-anodizing the niobium underlayer. Assuming that HSe − anions drop their protons at the anodic-oxide surface before the migration, the following anode reaction summarises the incorporation of selenide anions into the nanostructured niobium oxide: 2Nb 5+ +5(1−y 2 )O 2− +5y 2 Se 2− → Nb 2 O 5(1−y2) Se 5y2 (9) where coefficient y 2 represents the fraction of O 2− anions replaced by selenide anions in the anodic niobium oxide. The sum of y 1 and y 2 in reactions (3) and (9) corresponds to the fraction of O 2− anions in the oxide replaced by all-type Se species, similar to the values summarized in Table 2 as the percentages. Anode reactions (3) and (9) can be summarized by the following overall reaction for the PAA-assisted niobiumoxide growth in selenic acid: 2Nb 5+ +5(1−y 1 −y 2 )O 2− +5y 1 SeO 4 2− +5y 2 Se 2− → Nb 2 O 5 (1−y1) (SeO 4 ) 5y1 Se 5y2 (10) Conclusively, the experimental findings of this work suggest that the nanostructured ceramics formed via the PAA-assisted selenic-acid niobium anodizing have a dual (core/shell) composition: the inner Fig. 5. Schematic growth model during the PAA-assisted anodizing and re-anodizing of niobium in 1.5 M selenic acid for forming (left panel) SE42 A (42 V) and SE42 R (120 V) samples and (right panel) SE12 A (12 V) and SE12 R (80 V) samples: (a and f) the formation of PAA layer (1st stages in Fig. 1 just before the voltage begins to increase), (b and g) the growth of niobium-oxide nanostructures within the PAA barrier layer during the anodizing (2nd stages in Fig. 1), and (c and h) the development of niobium-oxide nanorods during the re-anodizing (3rd stages in Fig. 1). The dashed white lines differentiate the ’core/shell’ structure of the protrusions. The parallel-to-substrate section views show comparatively the cell-pore-rod relationship for (d) SE42 R and (e) SE12 R samples. K. Kamnev et al.