Defect-Induced Raman Scattering in Cu 2 O Nanostructures and Their Photocatalytic Performance

Advanced oxidation processes using photogenerated charges in semiconductors constitute an approach to reduce and oxidize pollutants, with an efficiency that depends on the photo physics and defect chemistry of the photocatalyst. In this study, 2D Cu 2 O coatings on flat copper metal and on 3D copper nanopillars are created via low-temperature oxidation and compared. The structures are characterized by X-ray diffraction, Raman spectroscopy, and electron microscopy. The thickest surface oxide layers on the 3D structures show outgrowth of high-aspect ratio CuO nano-needles through the Cu 2 O layer, rationalized through a field-induced copper ion diffusion mechanism. Raman scattering provides details about both the specific copper oxide phase present and the type and extent of defects, with a resolution spanning from hundreds of nano-meters to micrometers. We show that defects in Cu 2 O induce Raman activity in several of its modes that are purely IR-active or optically silent in pristine Cu 2 O. The experimental results are corroborated by linear response density functional theory (DFT) calculations for full vibrational mode analysis. The Cu-supported 2D copper oxide systems exhibit effective photocatalytic performance at quite low probe pollution concentration (10 μ M), while the 3D nanopillar structures enhance the photo-catalytic efficiency by around 30% compared to their planar counterpart under these conditions.


Introduction
Clean water is crucial for human survival and development, supporting sustainable scenarios for society.Over 844 million people lack access to clean drinking water, [1] and about 1.2 billion people live in water-scarce regions. [2]Contaminated water and poor sanitation are considered to cause 80 % of diseases in the developing world, where infrastructure and electrical energy for water purification are often lacking and millions of people dying annually from diseases such as diarrhoea and cholera, [3,4] .Quick and cost-effective water cleaning methods can significantly improve health conditions and minimize disease-related suffering in these regions.Furthermore, the increase in organic pollutants poses a serious global environmental threat that requires urgent countermeasures. [5]Some of these pollutants fall under the category of persistent organic pollutants and, although their use is well monitored, they nevertheless still pose a clear hazard due to their high stability, long lasting exposure and tendency to accumulate in food chains. [6]Others fall instead under the category of emerging pollutants which are new compounds that typically derive from agricultural and industrial activities (e. g., pharmaceuticals, pesticides, plasticisers, etc.) and, as such, are less studied and in most cases unregulated. [7]emiconductor-based photocatalysis [8][9][10][11] here offers an offgrid solution for wastewater treatment and water disinfection, and triggers chemical reactions upon light exposure.When these materials are illuminated, they generate electron-hole pairs, leading to the production of reactive oxygen species (ROS) like hydroxyl radicals.These ROS possess strong oxidizing properties, effectively breaking down and neutralizing organic pollutants and pathogens present in the wastewater.This process offers direct and robust means of wastewater treatment, especially when nanostructures are used, as they provide larger surface areas for increased reactivity and improved photocatalytic properties. [12,13]Tailoring active sites in these nanostructures further enhances their photo-catalytic properties.However, applying nanostructured materials on a large scale for water cleaning is challenging due to issues like high synthesis costs, difficulty in recovery and separation from solution reaction systems, lengthy treatment durations, and high energy requirements.To overcome these challenges, it is more convenient to form metal oxide nanostructures directly on solid and readily available substrates via simple oxidation processes.These substrates can be easily removed from the water for purification and regenerated using simple methods like heat treatment or washing, making recycling through conventional methods possible.
With these ideas in mind, abundant and easily attainable materials and methods that could be effectively applied to photo-driven water cleaning would be beneficial.Considering the availability of Cu in many communities from used electric cabling, cooking gear, water piping, and Cu foils in varieties of food packaging, this would form an easily attainable starting material in both developed and developing countries.Upon oxidation of Cu, Cu 2 O (or CuO) can be obtained and employed as a photocatalyst due to its known ability to efficiently absorb light in the visible spectrum, which constitutes a major component of the solar radiation, 42.3 % (calculated from AM1.5 reference spectra (ASTM G173-03) using 400-700 nm as limits for the visible light region).Cu 2 O, which has widely been investigated for solar cell applications and solar fuel generation, [14,15] is attractive also from a practical point of view, since its synthesis can be carried out easily at low temperatures in air by direct heating of copper. [16]Thermal oxidation of copper in air at low temperatures has been employed in the present study to form a thin layer of Cu 2 O directly onto Cu substrates having different morphologies and structures.The growth of copper oxide on copper substrates represents a wellstudied topic in literature, documenting the influence of the main experimental parameters on the resulting layer growth mechanism, [17] thicknesses, [18,19] morphology, [20] wettability, [18] adhesion [16,21] and composition [16,21,22] of the formed oxides, [20] while less is reported on the extent and type of defects beyond their mere presence via photoluminescence.The possibility of further enhancing the overall reactive surface per footprint area for these photo-catalytic substrates has also been explored here for the combination of these two materials without spoiling the above premises.This aim has been pursued by means of threedimensional (3D) nanostructured substrates based on Cu nanopillars, which were subsequently covered with thin layers of Cu 2 O by controlled thermal oxidation in air and compared with the Cu 2 O purposely created on a flat Cu substrate in the form of thin film.The growth of these high-aspect ratio Cu structures has been achieved via template-assisted electrodeposition [23][24][25] , which provides a possible cost-effective route to produce 3D nano-architectures.This type of electrochemical synthesis approach could be applied with locally available power sources and precursor salts, and is potentially scalable to cover larger substrate areas.As the oxidation step to form a thin layer of copper oxide on the flat copper substrate or the 3D copper nanopillars is carried out under low-temperature conditions, the materials can be formed or regenerated in an oven or on a hot plate within a few minutes.Moreover, an eventual recycling of these photocatalytic systems is straightforward.
In the present study, we investigate the oxidative growth of the copper oxides on planar and 3D structured Cu-metal, their crystallinity, nanostructuring, and defect-induced Raman scattering.The study includes vibration mode analysis using the symmetry of the crystal phase and the results are corroborated with DFT calculations.A critical comparison of the photocatalytic water purification properties obtained in both planarand 3D Cu substrates coated with Cu 2 O layers is presented.At low pollutant concentrations, it is anticipated that planar surfaces will be sufficient for purification purposes, while water with a higher concentration of pollutants will benefit from a higher surface area of the photocatalysts.The most important characteristics deriving from the different morphologies, as well as particular materials features of the various specimens and their related photocatalytic properties, are finally discussed.

Electrochemical growth of copper nanopillars and Cu 2 O coating
Copper nanopillars were grown via templated-assisted electrodeposition on squared Cu supports having a side of 1 cm.Prior to the electrodeposition process, the Cu substrates (Goodfellow) were ultra-sonicated for about ten minutes in acetone and subsequently rinsed with de-ionized water.The growth of these elongated Cu structures was carried out with an electrochemical cell and a procedure similar to those reported in previous studies. [24]In short, a precursor solution based on CuSO 4 • 5H 2 O (100 g • L À 1 , Merck), (NH 4 ) 2 SO 4 (20 g • L À 1 , Merck) and diethyltriamine (DETA, 80 mL • L À 1 , Fluka) was used as plating bath, and track-etched polycarbonate (PC) membranes with pores having a nominal size of 200 nm (Whatman -Cyclopore, 0.2 μm) were employed as template for the electrodeposition procedure.The electrodeposition through the pores of these membranes was performed at 50 °C using a potentiostat (VersaSTAT 4 -Princeton Applied Research) and a double-pulse galvanostatic technique. [23,24,26]This general procedure was implemented here applying first an initial nucleation pulse of 1.5 V (vs.Cu/Cu 2 + ) followed by rapid galvanostatic pulses.The galvanostatic schedule consisted of repeated cathodic pulses of 6, 90 and 0 mAcm À 2 applied for 0.25, 0.05 and 0.5 seconds, respectively.This pulse routine was repeated for 1250 cycles to enable the formation of Cu pillars with lengths approximately comprised between 4-5 μm.After a careful disassembly of the electrochemical cell, the Cu substrate, which was still attached to the PC membrane, was immersed in a beaker containing dichloromethane in order to dissolve completely the membrane template at room temperature and ultimately expose the Cu nanopillars grown through its pores.As low currents, low voltages and repetitive pulses are used, this deposition scheme can likely be simplified by a system with a battery and a simple chip pulse generator, while only a low temperature oxidation is needed for flat copper sheets, foils, wires or other purposely-roughened copper surfaces.
Oxidation of the nanopillars to form a thin outermost layer of Cu 2 O was performed via a simple heat treatment in air for a few minutes on a hot plate controlled by a thermostat.A controlled oxidation of three substrates was performed at 150 °C for different heating times of 2, 4 and 8 minutes, respectively, in order to form Cu 2 O with different thicknesses and to compare the latter with the pristine Cu nanopillars.
Additionally, a blank Cu substrate was oxidized under the same conditions in air at 150 °C for 8 minutes to provide a planar reference sample containing a well-developed surface layer of Cu 2 O in the form of thin film.

Structural and morphological characterization
The surface morphology of the various substrates with and without Cu nanopillars was investigated by scanning electron microscopy (SEM) using a Zeiss Gemini 1550 microscope with a field emission (FE) source and applying an acceleration voltage of 10 kV for the electrons.Imaging of the specimen surface was obtained by means of secondary electrons, which were collected by a dedicated in-lens detector.
X-ray diffraction (XRD) was carried out using a Siemens diffractometer (D-5000) equipped with a Cu Kα radiation source and the measurements were run in a Bragg-Brentano configuration.A rotation speed of 15 rpm was applied during the diffraction experiments.
Raman spectroscopy was carried out to investigate the composition of the oxide layers grown on the nanopillars and the planar substrates.The spectra were obtained via a Raman microscope (Renishaw inVia Reflex) utilizing a 532 nm excitation wavelength from a solid-state frequency doubled Nd:YAG laser (Renishaw).The laser beam was focused on the surface of the substrates via a builtin optical microscope and a 50× magnification objective.A constant laser power of 0.05 mW was applied to the samples during the analysis.An initial calibration of the instrument was conducted before the measurements by having a characteristic reference peak from a Si wafer at 520.6 cm À 1 .Twenty cumulative acquisitions having a measuring time of 20 s were employed for recording each spectrum.Moreover, the exposure of the sample surface to the laser beam was minimized between subsequent scans to avoid any possible degradation of their surface.

Density functional theory calculations
Density functional theory (DFT) calculations with Crystal 17 software [27,28] were performed for Cu 2 O to calculate the ground state geometry, linear response DFT to obtain the vibrational modes in order to assign the peaks of the experimentally obtained spectra, and subsequent linear response DFT using the dielectric tensor to extract the transverse optical-longitudinal (TO-LO) split.The starting geometry [29] of the material was obtained from the Inorganic Crystal Structure Database (ICSD) and this was optimized with respect to cell parameters and atomic positions using the generalized gradient approximation functional PBE [30] and two different hybrid functionals: the well-known Becke 3-parameter Lee-Yang-Parr (LYP) functional B3LYP with 20 % Hartree-Fock (HF) exchange contribution defined from the exchange-correlation energy functional LYP with the three parameters a = 0.20, b = 0.72, and c = 0.81.E X LSDA is the Vosko-Wilk-Nusair (VWN) local spin density approximation to the correlation functional, [31] E X HF the HF exchange, E X B the Becke GGA functional, [32] while E C LSDA and E C LYP represent the VWN and LYP correlation functionals, respectively. [33]The second hybrid functional used here is the PBE0 functional defined from E XC PBE0 = 0.25 E X HF + 0.75 E X PBE + E C PBE . [34]The electron distribution was modelled at the triple zeta valence level using a basis set with polarization quality. [35,36]A Monkhorst-Pack grid of 6×6×6 k-points was used for Brillouin zone sampling.For the geometry optimization, a convergence level of 10 À 10 Hartree was used since the subsequent force and Raman frequency calculations require a higher convergence than the standard level of 10 À 8 Hartree for electronic convergence.Such an approach is here applied since the forces depend on the second derivative of the nuclei positions that in turn can be at different positions with retained degenerate electronic energy at a lower convergence threshold.As the bonds in Cu 2 O have an ionic-type character, a split between transverse optical (TO) modes and longitudinal optical (LO) modes is expected, due to the additional restoring force from the oppositely charged ions upon displacement along the chain.This TO-LO-split was calculated separately using the dielectric tensor obtained from the frequency calculation.

Optical measurements and evaluation of the cleaning properties
The water cleaning properties of the bare Cu metal, the planar Cu surface covered with the Cu 2 O layer and the three variants of 3Dstructured Cu 2 O photocatalysts grown on the Cu nanopillars were studied in a photo-bleaching experiment, using methylene blue (MB) as model substance for evaluating the degradation rate of organic contaminants in water.One should note that decolouration forms a convenient measure of the initial rate of breaking bonds in an extended conjugated system, while full water purification needs further time to degrade also the colourless parts of the original compound.The photocatalysts were placed in 5 mL polystyrene (PS) cuvettes, positioned at the back, and 4.5 mL of 10 μM methylene blue solution was subsequently added.The photocatalyst was illuminated with an AM 1.5 G solar simulator with an intensity at the position of the sample of 100 mW cm À 2 , measured with a pyranometer (Thorlabs 160T).The optical measurements were conducted using an HR2000 + spectrophotometer (Ocean optics) with a deuterium and halogen light source (Micropack).The spectral range between 400 nm and 800 nm was measured and an average of 100 consecutive scans with a measurement time of 2 ms was used.The absorption of the MB solution was measured perpendicularly to the illumination from the solar simulator.There was a two-minute settling time before the experiment started, during which the illumination was on.The experiments were performed without stirring, measurements were taken every minute and the experiment continued for 135 minutes (2 h and 15 min).As a reference experiment, illumination of the dye solution without any Cu or Cu 2 O sample was also carried out.The data presented is in its raw form.

Morphological and structural properties of the coated and uncoated substrates
The surface morphologies of the various planar and 3D Cu substrates, both in their pristine and oxidized forms after different heating times at 150 °C, are presented in Figure 1.A comparison of the scanning electron microscopy (SEM) micrographs in Figure 1 demonstrates that the main effect of the oxidation of the pristine planar-and 3D substrates consisted in a roughening of their respective surfaces.From Figure 1b, it is evident that the oxidation of the pristine Cu plate (Figure 1a) caused the growth of grains (or scales), which covered uniformly all the surface of the sample.The sizes of these oxide grains in Figure 1b span from 100 to 200 nm and exhibit a dense coverage with a continuous connection among the grains.The as-grown Cu nanopillars in Figure 1c display a rather homogenous spatial distribution and their exposed surfaces are smooth, showing well-defined round contours.Their lateral dimensions match well with the nominal pore size of the template, as it is seen in Figure 1c, with diameters of around 200 nm.The effect of the template pores is also visible in the shapes, merging or mutually intersecting, following the original pore tracks in the PC membrane.A rough evaluation of the numerical density of these Cu structures per unit area, N d , yields � 4 μm À 2 .Mutual distances between individual nanopillars and their bundles are typically below 1 μm with heights of 4-5 μm extracted from cross-section micrographs, in good agreement with the conditions utilized for their electrochemical growth.In a real-world application, the precise morphology and dimensions of the nano-structuring are not of uttermost importance, as long as the overall surface area is increased.Nevertheless, it is important here to critically evaluate the effect of these different approaches and the influence of the 3D structuring enabling an increased available catalytic surface per footprint area.
The oxidation of the Cu nanopillars for different heating times (i.e., 2, 4 and 8 minutes) caused a major change in their surface morphology, as it can be observed in Figure 1d, 1e and 1 f, respectively.The oxidized structures exhibit on average larger diameters (e. g., � 250-300 nm) and the presence of small grains with sizes approximately around 20-30 nm.The rough-ening of their surface upon oxidation proceeded in a way similar to that taking place on the planar Cu substrate.The theoretical thicknesses of Cu 2 O from the different heating times have been evaluated (see Supporting Information).It is interesting to observe that the grains formed on flat surfaces grew slightly larger than those found at the extremities of the nanopillars, as it can be seen by comparing Figure 1b and 1d.Indeed, apart from such an enhanced geometrical surface area, due to these elongated 3D features, the 3D pillars should also have an influence on the heat transfer from the underlying support and its propagation/dissipation throughout the entire 3D Cu structure.Incidentally, it is worth noticing that due to a very high thermal conductivity of metallic Cu ( � 400 Wm À 1 K À 1 at 298 K), it would be possible to obtain a locally heated capillary-like surface during the photocatalytic processes with this 3D nano-architecture, which in turn would be beneficial to increase the rate of the pollutant degradation reactions.
A comparison of Figure 1c, 1d, 1e and 1f shows that the roughening of the lateral surface of the nanopillars appears to evolve in a similar fashion upon their progressive heat treatment, especially when compared with the initial morphology change of the pristine Cu structures.Nevertheless, with increasing heating times, a progressive development of irregular aggregates on top of the nanopillars is noticed.These aggregates do not appear to have followed the contours of the nanopillars, but instead mainly developed across them, bridging their tips and filling in the space between their extremities.Accordingly, the increasing time intervals for heating and related oxidation of the nanopillars caused not only an initial roughening of their exposed surfaces, but also a progressive decrease of their interspacing with an effective lateral extension, mainly localized at their tips.This is due to an enhanced grain growth, which also affects possible access to empty spaces in the structure, e. g. blocking of open pathways from their uppermost terminations.Particularly interesting are the morphologies in correspondence of the heating times of 4 (Figure 1e) and 8 (Figure 1f) minutes, since the oxidation of the nanopillars also resulted in a growth of extremely thin and sharp nano-needles.These tapered nanostructures, which grew mainly at the extremities of the oxidized pillars, have typical lateral dimensions of approximately 10 nm and lengths ranging from few tens of nm up to about 300 nm for the longest ones found in Figure 1f.The formation process can be understood by considering the findings of previous reports [17,19,[37][38][39][40][41] on the formation of metal oxide nanostructures obtained by thermal oxidation of metals.In our case, due to the high aspect ratio growth, their growth can be considered to be governed by enhanced diffusion of Cu cations through the Cu 2 O to the tips and subsequent reaction with oxygen in air to form cupric oxide (i.e., CuO), yielding a local stoichiometry different from that of the underlying grains (i.e., Cu 2 O) from which they originate.
This is schematically illustrated in Figure 2. and such a mechanism agrees with findings from previous reports, [37,38,42] in which CuO is usually formed on Cu 2 O in presence of oxygen and moderate temperatures.This type of oxide growth has previously been considered to be governed by an outward diffusion of cations upon Cu oxidation, [43][44][45] since both Cu 2 O and CuO are intrinsically cation deficient p-type semiconductors.Quantification of Cu 2 + ions mass transport via grain boundary diffusion for the growth of CuO nanowires (NWs) can be found in an earlier work, [37] where one of the key driving forces for this process is identified in the strain accumulated at the CuO/Cu 2 O interface.Other reports have described this process in an alternative -yet compatible way -via nucleation of CuO NWs with a solid-state transformation, and that their growth in a similar way exhibits a diffusion-controlled behaviour, following a parabolic law with annealing times. [38]This is in line with the classical theory of Wagner [46] in the high-temperature regime, where the thermal energy is sufficient for ionization and diffusive transport, [47] while the electric field across the interface of the growing oxide film plays a more important role at low-temperatures. [48,49]If we consider that the diffusion and drift current density J i of a cation species i along the direction x can be described by Equation 1, where D i is the diffusion coefficient of the cation species in the media, c i is their concentration per unit volume, μ i the chemical potential, q i their charge, and E the electric field.The validity of equation 1 depends on the assumption of a small electric field E so that γqE !k B T for a cation jump distance of γ. [49] The extent of a space-charge layer, however, is determined by the Debye-Hückel screening length L D = ɛɛ 0 k B T/e 2 C d where ɛ and ɛ 0 are the dielectric constants of the copper oxide and vacuum, respectively, and C d is the number of elementary charges per volume, due to charged defects at equilibrium in the intrinsic oxide.The electric field would thus only dominate when x !L D , while diffusion starts to dominate at longer distances as in the case of the more extended nano-needles.The initiation of the anisotropic growth is therefore likely to occur via a starting nucleation site by an initial interfacial field, while the continuous growth would depend on a higher diffusion coefficient along the axis of the nano-needles, instead.The morphologies shown in Figures 1e and 1f, can be rationalized with the combination a locally induced field and strain [50,51] present near the edges of the oxidized nanopillars, favouring the growth of CuO nano-needles at these low temperatures.In fact, the formation and growth of a CuO phase with high-aspect ratios at 150 °C within minutes is rather unexpected, while here these pointy nanostructures seem to have developed to a good extent, ideally forming thin branches poking out from the underlying Cu 2 O oxide, and reveal a highly preferential cation flow and reactivity in the direction of the needles.The presence of CuO nano-needles is beneficial, since they do not only enhance the overall exposed surface areas, but also strengthen the effective electric field in proximity of their tips.In addition, thin CuO nanowires have successfully been applied for disruption of cellular membranes of microorganisms and, in this way facillitated disinfection of water. [52]Semiconductor nano needle and nanowire structures have also been shown to physically puncture the cell membrane of bacteria [53,54] and, if the nano needles in our material have such an ability, they will also contribute to the decontamination and purification from bacteria in the treated water.
The structure and composition of the resulting materials were further investigated to clarify these intriguing morphological aspects.The XRD patterns obtained in correspondence of the various pristine and oxidized specimens at 150 °C are reported in Figure 3.The pristine Cu plate shows characteristic diffractions that match with the crystalline structure of copper (JCPDF #04-0836) and their relative intensities indicate a preferential orientation along the (200) and (220) crystal planes.A similar trend for the peaks related to the Cu structure is also visible in the diffraction pattern of the Cu plate after heating in air for 8 minutes, which also exhibits a feature around 36.4

°.
The latter corresponds to the strongest (111) diffraction originating from a Cu 2 O lattice structure (JCPDF #01-078-2076) and perfectly agrees with the previous result presented in Figure 1b, where small and homogenous copper oxide grains were formed on the surface of the Cu substrate.The diffraction pattern of the as-grown nanopillars confirms that these elongated structures are composed of crystalline Cu with a preferential orientation along the (111) crystallographic plane.No traces of any spurious crystalline species or compound can be detected in the same diffractogram.The XRD patterns of the oxidized Cu nanopillars with increasing heating times show a progressive appearance of the characteristic diffractions associated with the Cu 2 O structure.The first peak to become visible for Cu 2 O for the sample heated for 2 minutes is the expected strongest diffraction, i. e. originating from the (111) crystal orientation.A progressive enhancement of this diffraction is exhibited by the patterns of the other specimens that underwent heating for longer times of 4 and 8 min.It is also possible to discern in these diffractograms other weak features associated with the (110), ( 200), ( 220) and ( 222) crystal planes of Cu 2 O, with the latter diffraction detected only for the sample heated for 8 minutes.
A faint -yet still visible -signature of CuO is noticed in the XRD patterns of the specimens heated for 4 and 8 minutes.A broad and weak feature is present around 39 °, which is attributed to the very close diffractions originating from the (111) and (200) planes of CuO (JCPDF # 00-005-0661).Moreover, a shoulder around 35.5°is also seen on the left side of the main (111) peak of Cu 2 O, which is ascribed to the most intense (À 111) diffraction of CuO.These findings confirm the formation of CuO onto the nanopillars oxidized for longer times and indicate that this oxide phase was formed only on these elongated structures, since the oxidation of the planar Cu support for 8 minutes did not give any characteristic diffraction for CuO and did not generate any nano-needle on the sample surface.
A series of Raman spectra for the different pristine and oxidized specimens at 150 °C is presented in Figure 4.The spectra of both the pristine Cu plate and as-grown Cu nanopillars display a featureless profile, which reflects the fact that metallic Cu is Raman inactive.Metallic Cu is also known for its use in surface enhanced Raman scattering (SERS), especially in the form of rough substrates, and enables an increased detection of Raman-active species deposited on its surface due to enhanced localized surface plasmon resonance (LSPR).Therefore, this technique is particularly suitable to check the surface composition (or possible oxide contamination) of these Cu structures.The fact that the pristine Cu nanopillars did not show any feature in their Raman spectrum is in line with the earlier XRD results and demonstrates that these Cu structures are stable against creation of an XRD-detectable oxidation layer after their electrochemical growth and subsequent dissolution of the PC template membrane, apart from the formation of a possible thin native copper oxide.Conversely, the Raman spectrum of the oxidized plate displays a clear signature of Cu 2 O, as it can be noticed from the two most prominent peaks at 150 cm À 1 and 217 cm À 1 .Their relative intensities appear rather enhanced compared to those of the other remaining peaks.Effectively, the formation of homogenous crystalline grains of Cu 2 O (see Figure 1b), having a theoretical energy bandgap E g � 2.2 eV, is here anticipated to give rise to a near-resonant absorption at the excitation wavelength of 532 nm with resulting resonant Raman scattering (RRS).This phenomenon would then account for the observed amplification of the peaks involved in the redistribution of the electrons in the electronic excitation.
Two other characteristic spectral features are also visible around 299 cm À 1 and 636 cm À 1 , while two minor bands, with a weak and broad shapes can be found around 414 and 520 cm À 1 , respectively.Overall, these peaks and their positions agree well with previously reported experimental data for Cu 2 O [55][56][57] and show that the oxidation of both the planar and the 3D substrates yielded Cu 2 O. CuO is hardly detectable in the spectra of the specimens with longer oxidation times, being further away from resonance conditions (E g � 1.2 eV for CuO) with the excitation wavelength of 532 nm, and its overall amount is very limited as well.A clear detection of this CuO compound is further complicated here, because its characteristic spectral features are close to those of Cu 2 O and partly overlapping.In fact, the major peaks reported for CuO are normally detected around 290 cm À 1 , 338 cm À 1 and 624 cm À 1 , respectively. [58]Only a minor asymmetry and a slight shift for the peak intensity of the broad feature initially centred around 640 cm À 1 can be noticed in the spectra of the samples having longer oxidation times.For the latter, their maximum value has shifted to 630 cm À 1 , thus suggesting a weak presence of CuO, which otherwise is not directly detectable by this analysis alone.
It has been reported that CuO can be formed by thermal oxidation of Cu at temperatures below 1050 °C and that Cu undergoes a two-step oxidation where Cu 2 O is the initial phase obtained, while the CuO formation typically occurs after long thermal treatments (days). [14]This is consistent with the observation that we obtain predominantly Cu 2 O, since the thermal treatment time is short, while the creation of the highaspect ratio CuO nano-needles for the 8 min oxidized samples can be rationalized via the a electric field induced growth, as outlined above.An overview of the investigated samples together with their morphological, structural and compositional properties extracted from the previous analyses is provided in Table 1.

Determination of vibrational modes, TO-LO split and peak assignment via DFT
The DFT calculations for the lattice vibrations of Cu 2 O converged and produced only positive vibrational frequencies indicating that a true energy minimum was found.The energy bandgaps, refractive indexes and cell parameters obtained with the PBE, B3LYP and PBE0 functionals were in good agreement with previously reported data, whereas the computed figures for the dielectric constant were lower than characteristic experimental values for the hybrid DFT functionals, while B3LYP instead reproduced the experimental band gap and cell parameters to a better degree.The fundamental material constants obtained from the calculations are listed in Table 2.
We will use the results from the B3LYP calculations below for analysis of the vibrational properties, nevertheless, the latter can -without lack of generality -also be performed with any of the other functionals with the same trends, yet with shifted absolute values.The frequency calculation only showed one Raman active vibrational mode for the Cu 2 O crystal, as expected from the symmetry of the pristine material.This vibration corresponds to a T 2g vibration mode and this single active mode is clearly in contradiction with our experimental data and other previously published spectra for Cu 2 O. [61][62][63] This observation had been made before and this apparent contradiction has been explained by the fact that point defects at certain positions in the lattice break the symmetry in the crystal and can, depending on their specific location, make part or all the vibrations Raman active. [62,64]It is therefore relevant to consider all the obtained vibrational frequencies in a full analysis.In addition, as we expect that the electron distribution will be anisotropic in a CuÀ O system, the partial charges on the respective element will give an additional Coulombic force that strives to restore the average bond length and thus stiffens the LO vibrations and shifts the LO modes to higher frequencies.The latter is referred to as TO-LO split that occurs in polar bonded crystals.The TO-LO calculation indicated that the TO-LO split would occur for the two IR-active frequencies.In Figure 5a, the result of the DFT calculation with the B3LYP functional is shown in the form of simulated Raman spectrum, in which the different categories of modes are shown in different colors.The calculated peaks for relevant combination modes are also included.The peak intensities for Raman and IR modes are obtained from the DFT calculations, whereas a value of 10 % of the maximum intensity was used for both silent and combination modes.A Gaussian distribution has been added to give a resemblance to experimental Raman spectra.Figure 5b shows the movement of the atoms in the Raman active T 2g vibrational mode, while Figure 5c-e illustrate common defects and how they break up the symmetry of the crystal structure.
According to Sander et al., [64] defects both at the oxygen site and oxygen interstitial in a tetrahedral position will make the E u and A u silent mode, as well as the T 1u IR modes, Raman active.There is also a shared Cu vacancy defect (not shown here) that can activate all the vibrational modes including the T 2u vibration that remains inactive for all the other defects.Due to uncertainties of the many possible modes and the fact that they may -or may not -be activated depending on the type of defects in the material, DFT calculations are here necessary to reliably support the mode assignments.In Table 3, all the vibrational frequencies obtained by DFT are tabulated together with their assignments and the positions of the corresponding peaks in our experimental data.The T 2u (low) mode is out of the  [59] 2.310 [60] 5.34 [60] 4.2670 [29] *Refractive index at high frequency.**Dielectric constant at high frequency.
wavenumber range for our experimental measurement and therefore the assignment starts with the E u mode.The experimental peak found at ~150 cm À 1 can originate either from the TO or LO mode of T 1u .In Table 3, it is marked as T 1u (LO) for the best numerical fit.The two experimental peaks at 217 cm À 1 and 299 cm À 1 fit well with multimode peaks of our DFT calculated frequencies.The Raman active T 2g mode is assigned to the weak peak at 520 cm À 1 , which is consistent with earlier works [56,57,64] that attributed this mode to peaks inbetween 515 cm À 1 -530 cm À 1 .It should be noted that this peak is relatively far from the DFT predicted position and that Reydellet et al. [62] assigned the T 2g mode to a different peak located at 590 cm À 1 , a spectral feature that is not present in our experimental data.Symmetry arguments and the performed DFT calculations can explain the large number of peaks observed in the experimental spectra and, additionally, why there is a variation in previously reported peak positions, as various synthesis routes could generate different types of defects.
Referring back to our experimental data in Figure 4, the peak assignments from the DFT calculations and the defect activated modes of Sander et al., [64] the following can be concluded about the defects present in our samples.The only mode that is activated by a single defect is the T 2u and, since its frequency is outside the measurement range, the presence of a shared Cu vacancy cannot be proven.All the other vibrational modes are present in the experimental data and these are activated by either a vacancy at an oxygen site, a Cu atom at an oxygen site, an interstitial oxygen at a tetrahedral site, or the shared Cu vacancy.Hence, the symmetry arguments alone are not enough to distinguish them, since they all activate the observed modes.It is however notable that in the flat oxidised  sample the T 1u mode at 150 cm À 1 is much stronger compared to the 2•E u mode at 217 cm À 1 and also more prominent than what is observed for the nanostructured samples.Therefore, this intensity difference can be related to different defect characteristics.

Analysis of related water cleaning performance under illumination
The UV-VIS data from the photo-bleaching experiments of Cusupported photocatalysts are presented in Figure 6.In the graphs, the absorption spectrum for methylene blue (MB) is plotted in sequence for each minute during the experiment.
The results show that all the experiments with the different photocatalytic active surfaces were efficient in bleaching the dye molecules, in contrast to the reference (i.e., intrinsic MB photo-degradation without any photocatalyst) for which Figure 6c clearly shows that MB is still present after 135 min.Within the timeframe of these experiments, all these photocatalysts were able to degrade the MB and decoloured more than 93 % of the dye solution.A comparison of the performance of the various catalysts is presented in Figure 7 using the absorption maximum evolution with time and normalizing by the initial absorption.According to Lambert-Beers law, the absorption relates to the concentration through the relationship A = ɛ'c, where A is the absorbance, ɛ is the absorptivity of the light-absorbing species, ' is the optical path length in the  solution and c is the concentration of the species.Since the absorptivity and the optical path length in the cuvette are fixed, the absorbance is directly proportional to the concentration and A/A 0 can be substituted by c/c 0 , where A 0 and c 0 respectively represent the absorbance in correspondence of the initial concentration at the time t = 0.
The data plotted in Figure 7 shows that even though the total bleaching is similar for all of the tested photocatalysts, the bleaching rate at the initial stage is clearly higher for the photocatalysts embedding the Cu nanopillars subjected to oxidation for 4 and 8 minutes (i.e., where also some CuO nanoneedles were formed).The slope of the trend lines in Figure 7 indicates that the bleaching is 34 % faster than for the planar Cu substrate with a Cu 2 O layer (i.e., Cu plate oxidized for 8 minutes).Such a linear-type degradation behaviour in this region is consistent with the fact that the reaction is limited by the catalyst, either by its intrinsic catalytic efficiency or by the number of available active catalytic sites on its surface.After reaching approximately half of the initial concentration c 0 , the degradation rate starts decreasing, which indicates that the degradation reaction becomes limited by diffusion of the MB to the surface.Here, at a relatively low concentration of the pollutant, the flat surfaces are quite effective.The 3D-nanostructured surfaces instead have a higher number of available active sites, which is a likely cause for the observed increased catalytic activity, and clearly constitute a relevant benefit in presence of higher pollutant concentrations.A possible explanation for the increased catalytic behaviour of the two 3Dnanostructured substrates with longer oxidation times is the presence of both Cu 2 O and CuO on their outermost surface.In a previous study, [65] we concluded that Cu 2 O possesses a higher catalytic efficiency than CuO, nevertheless, that photocatalytic efficiency can also be increased by combining two semiconductor materials, e. g. by creating a junction between them.However, since Cu 2 O and CuO both are p-type semiconductors, we find it unlikely that their junction would be very efficient in facilitating the charge transfer.An approach to increase the efficiency could instead be to form a heterojunction with an energy-aligned material. [65]Reverting to Figure 6 and 7, we now focus on the Cu sample.Why does the Cu metal bleach the methylene blue?The metal absorbs light below 650 nm, however, in the metal the photons will not generate highenergy electrons in states localized at the surface, as in a semiconductor, but instead oscillating electronic states that relax into heat in the high femtosecond regime.The electrons in high energy states, which could be sustained at a higher electrochemical potential for longer timeframes, are instead the ones extracted for redox reaction and utilized in the photocatalytic water cleaning process to degrade the pollutants.Because of the electronic structure of the metal, the photons will instead generate a localized amount of heat that is either transferred locally, or, as in the majority of the cases, distributed evenly across the metal, before the rare event of a redox reaction with a pollutant molecule at the surface occurs.The energy transfer to the pollutant molecule is then minor and the degradation of the pollutant would mainly be affected by the thermal contribution.As seen in Figure 6 and 7, this is clearly not the case.It is more likely that a thin native oxide layer has formed on the surface of Cu, either while being handled in air or in the water solution.This would alter its surface properties, making it similar to the flat Cu 2 O sample, and would then result in comparable effectiveness in degrading pollutants at low concentrations.The degradation experiments with these two samples also show remarkable similarities, supporting this potential scenario.

Conclusions
The creation of Cu 2 O on flat and 3D nanostructured copper substrates via low-temperature heating under ambient conditions has here been investigated, where the Cu sample with thin (native) Cu 2 O and the oxidized samples are compared in terms of morphology and structure, as well as their resulting performance as systems for photocatalytic water purification.The growth of the formed copper oxide on the different substrates is analysed via X-ray diffraction, scanning electron microscopy and Raman spectroscopy, revealing the formation of Cu 2 O as the dominating phase, with a sparse presence of CuO nano needles created for the samples undergoing oxidation for the longest times.Particular attention was given to defect-induced Raman scattering, explaining the activation of Raman inactive modes in Cu 2 O due to defects and subsequent broken local symmetry.Theoretical Raman spectra using linear response density functional theory (DFT) calculations were utilized to provide a full vibrational analysis, where symmetry arguments were used to explain the activation of the modes experimentally observed for Cu 2 O.The thickest surface oxide layers show outgrowth of CuO nano-needles rationalized through a field-induced copper ion diffusion mechanism, which could be beneficial if the system would be directed towards anti-bacterial applications.All of the Cu-supported copper oxide systems provided an effective photocatalytic performance at these relatively low probe pollutant concentrations (10 μM) with the 3D nanopillar structures further enhancing the efficiency up to 34 % compared to their planar counterpart.Given the abundance and availability of Cu, along with the use of low-temperature oxidation in air and possible subsequent annealing for catalyst reactivation, this approach can be considered robust and cost-effective for water purification in off-grid applications, particularly in developing countries.Furthermore, this method can also be adapted for targeted pollutant removal in existing water purification systems in urbanized areas, making it a versatile and compatible option for various water treatment scenarios.

Figure 1 .
Figure 1.SEM micrographs showing the characteristic surface morphology of various types of planar and three-dimensional Cu substrates before and after oxidation in air at 150 °C for different heating times.Planar Cu sample before (a) and after (b) oxidation in air for 8 minutes.(c) Pristine as-deposited Cu nanopillars and related surface morphologies after (d) 2 minutes, (e) 4 minutes and (f) 8 minutes of oxidation of their nanopillars.Note a surface roughening of the oxidized specimens and a progressive growth of sharp nano-needles close to the extremities of the oxidized nanopillars for longer oxidation times of 4 and 8 minutes.

Figure 2 .
Figure 2. Schematic illustration of the proposed needle formation process, where Cu 2 + cations diffuse through the nano-needles to the tip that represents the most energetically favourable site to nucleate and form CuO.

Figure 3 .
Figure 3. XRD diffraction patterns obtained for the various types of planar and three-dimensional Cu substrates before and after oxidation in air at 150 °C for different heating times.Note that for the oxidized specimens the strongest diffractions marked with an asterisk are ascribed to Cu, whereas the indexed peaks are associated with the crystal structure of Cu 2 O, labelled with the symbol (^).Weak diffraction features originating from CuO are indicated by the symbol (¤).

Figure 4 .
Figure 4. Raman spectra recorded at room temperature for planar and three-dimensional Cu substrates before and after their oxidation in air at 150 °C for different heating times.The Cu 2 O peaks are marked with ^and the CuO peaks with ¤.The T 1u (low) and 2•E u modes that are discussed in the results section are shaded in light green and light blue, respectively.

Figure 5 .
Figure 5. (a) Calculated Raman spectrum via B3LYP functional for Cu 2 O including Raman modes, IR modes, silent modes, modes originating from TO-LO splitting and a selection of combination modes.(b) An illustration of the atomic movement during the Raman active T 2g vibration, calculated with a hybrid-DFT using the B3LYP functional.The nature and position of common defects in Cu 2 O include vacancies (c), substitutional atoms (d), and interstitial oxygen atoms (e).

Figure 6 .
Figure 6.VIS-absorption spectroscopy showing the photo-bleaching of methylene blue as a function of time for a) Planar Cu metal (native oxide), b) planar Cu substrate covered by a Cu 2 O layer, c) blank reference (no sample), d) Cu nanopillars oxidized for 4 min, e) Cu nanopillars oxidized for 8 min.

Figure 7 .
Figure 7.Comparison of the photocatalytic activity for the different planar and 3D-structured catalysts.The normalised intensity of the main absorption peak of methylene blue (665 nm) is plotted against time.For the initial stage, in which the reaction is not limited by diffusion, linear trend lines have been added.

Table 1 .
Overview of a series of morphological, structural and compositional features for the examined Cu-based samples.

Table 2 .
Calculated band gaps, refractive index, dielectric constant, and cell parameters for Cu 2 O calculated by one GGA and two hybrid-DFT functionals and corresponding experimental values from literature.

Table 3 .
Vibrational modes from the DFT calculation for Cu 2 O and assignments of experimental Raman peaks.
* Defect induced Raman peaks ** Out of range of measurement.