Plasma in liquids induced modification of Cu surfaces

Copper oxide surfaces are commonly used as the catalyst for the CO2 reduction reaction towards hydrocarbons. However, the lifetime of these catalyst surfaces is limited. In this paper, a method of production of copper oxides through in-liquid plasma is explored, which may be a suitable reactivation method in such applications. The influence of the plasma, ignited in distilled water, with copper and its oxides is monitored in − situ using infrared spectroscopy and ex-situ using scanning electron spectroscop and x-ray photoelectron spectroscopy of the samples. It is shown that the interaction of the plasma with the samples causes a reduction of the copper oxide on a fast time scale and an oxidation on a longer time scale. The formation of preferentially oriented copper nanocubes is observed.


Introduction
Copper oxide surfaces are valuable catalysts for the electrolysis of CO 2 towards hydrocarbons [1].The nanostructure of copper oxide and the oxidation state of copper define the final product.For example, a surface that primarily contains CuO produces longer chain hydrocarbons in comparison to Cu 2 O, which leads to the formation of methane predominantly.Both copper and copper oxide can be used to catalyze the CO 2 reduction reaction.The size and morphology of copper nanostructures has shown to influence their catalytic activity and selectivity [2,3].It has been shown, that oxidized copper catalysts have an increased selectivity towards C 2 H 4 in comparison to pure copper [4].Furthermore, it has been shown that the Faradaic efficiency for C 2 H 4 of copper oxide nanostructures is influenced by the oxygen content of these structures [5].During CO 2 reduction, the copper and copper oxide catalyst undergoes structural changes, including the reduction of Cu x O [4,6,7].These structural changes result in a decrease in activity and selectivity for the electrolysis of CO 2 [6].This leads to a limited lifetime of these copper and copper oxide catalysts during operation, which demands a method of oxidation or reoxidation, that is able to restore the initial oxidation state.This paper aims to explore such a method to reactivate these copper oxide surfaces in operando by combining an in-liquid plasma with the electrolysis system.
Copper oxide surfaces can be prepared by many methods ranging from electrochemical synthesis to oxide growth by thermal treatment of copper in an oxygen atmosphere.As a result, copper oxide with different nanostructures is formed, ranging from nanocubes and nanorods to nanosheets [8][9][10].This very specific structural shape makes copper oxide catalysts very selective towards individual products.At the same time, the use of these catalysts also requires good control not only of the oxidation state but also of the shape of the copper oxide crystals at the surface.The formation of copper oxide and its oxidation states is known to be temperature dependent [11].At low temperatures, Cu(OH) species are formed at the metal-liquid interface and from this Cu 2 O may condense in from of nanocrystals.At higher temperatures above 250 • C, this Cu 2 O is then transformed into CuO, which is very stable.
A standard method for Cu 2 O nanoparticle production corresponds to water electrolysis [10] by applying a low voltage to an electrode submerged in water.The electrochemical reactions at a sacrificial copper anode lead to the growth of Cu 2 O nanostructures [12].The character of the nanostructures can be adjusted in this system by applying different voltages (including applying no voltage at all) and different electrolytes, pH values or temperature [8,9].The creation of a copper oxide surface can be simplified by the sequence of the following reactions: When in contact with water molecules, oxidation takes place, and copper hydroxide is formed.The electron transfer may also cause a subsequent reduction to Cu 2 O nanostructures.These nanostructures are not necessarily stable under reducing conditions and dissolve.It has been shown that a metastable Cu(OH) phase forms at the surface of the Cu 2 O prior to reduction [13].The treatment conditions now determine to which extent and in which direction this reaction sequence propagates.
A variant of standard electrolysis is plasma-enhanced electrolysis at high voltages, where the breakdown of the electrolyte occurs, causing the generation of plasma.This technique is called glow-discharge electrolysis or plasma electrolysis [14,15].One advantage of this method is a higher yield of reactive species compared to regular electrolysis [15].The plasma also activates the liquid medium; in the simplest case of using water, so-called plasma-activated water (PAW) is generated.A primary constituent of PAW is hydrogen peroxide but also other reactive species [16].The conversion of liquids by the plasma and the modification of adjacent surfaces is coined solution plasma processing [17] and has been used successfully to create nanoparticles from different metals [18].One may speculate that in-liquid plasma treatment of copper is a suitable method for Cu x O nanoparticle production.
The nature of the plasma in a liquid depends very much on the excitation scheme, which can range from DC plasmas up to nanosecond pulsed plasmas.The plasma generation follows a sequence of, first, water evaporation around an electrode before, second, plasma ignition occurs in the vapour-filled bubble.At very short nanosecond rise times of the voltage, plasma ignition occurs directly inside the liquid [19].Since nanosecond plasmas in liquids can be very well controlled, we focus on this type of plasma generation.We studied these nanosecond plasmas in the past and analysed plasma formation and plasma-induced chemistry: Shadowgraphy measurements, emission spectroscopy and colorimetry analysis of the plasma-treated liquid have been used to study the plasma-inliquid system.The analysis of the Stark-broadened H alpha line in the emission spectra yield electron densities in the order of 10 25 m −3 at 10 ns after ignition [19].An electrode tip temperature of 7000 K was deduced from the blackbody radiation of the electrode.This corresponds to the boiling temperature of the electrode material, tungsten.Due to the short pulse lengths, rapid boiling and cooling of the electrode occurs, which leads to the formation of tungsten nanoparticles that eventually dissolve in the liquid.During the expansion of the plasma, high pressures in the GPa range at the electrode tip were observed, as deduced from monitoring the sound speed of the emitted pressure waves in the liquid.The application of the high voltage with a short rise time leads to ignition of the plasma directly inside the liquid [19].The dissipated energy into the pressure times volume was estimated as 3.2 × 10 −5 J.In previous works, an ignition mechanism for both positive and negative voltages applied to this system has been postulated [20].In the present work, only positive voltages were applied.In this case, the plasma ignition is presumed to take place via field ionization, where electrons from water molecules surrounding the electrode tip tunnel into the electrode, forming a positive space charge in this small area.The discharge may be compared to a streamer discharge under these conditions.The very fast, intense plasma leads mainly to the dissociation of water into OH in the first nanoseconds of plasma generation.These OH radicals then recombine to H 2 O 2 in the plasma afterglow during the consecutive adiabatic expansion and cooling of the gas trapped inside the bubble.This H 2 O 2 then dissolves in the liquid.It was found, that the production of H 2 O 2 is dependent on the applied voltage.A production rate of up to 3 nmol/l H 2 O 2 per 10 ns plasma pulse has been measured [21].
One could envision two reaction scenarios if such a plasma is brought into contact with an adjacent copper surface immersed in the liquid.On the one hand, oxidation of copper in reactions with reactive oxygen species (such as H 2 O 2 ) in PAW may occur.On the other hand, a reduction of the copper surface due to reactions with solvated electrons or hydrogen species, like atomic hydrogen or H 2 , in the liquid may also occur.This leads to the research question of controlling the oxidation and reduction competition in plasma in liquid systems to tailor the copper oxide interface.
In this paper, we study this interaction scheme using in situ infrared spectroscopy to monitor the oxidation or reduction of a copper surface in a plasma-in-liquid system.The plasmatreated surface is analysed ex-situ using scanning electron microscopy and x-ray photoelectron spectroscopy (XPS).

Plasma generation
Figure 1 shows a schematic of the experimental setup used for copper oxide treatment by plasmas in liquids.The nanosecond pulsed plasma is generated by a pulse generator with a rising time of 2-3 ns and a pulse width of approximately 12 ns.A frequency of 10 Hz is applied at varying voltage.The driven electrode consists of a 50 µm tungsten wire and the grounded electrode is made from stainless steel.Both electrodes are mounted into a water-filled PMMA-made plasma chamber with an approximate distance of 1.5 cm between the electrodes.Distilled water with an electrical conductivity of 1 µS cm −1 and a pH of approximately 5.5 is used as the liquid.The total liquid volume is 25 ml.The plasma chamber and HV generator are placed inside a Faraday cage.The side wall of the chamber contains a substrate holder made from PMMA, where copper-coated silicon wafers are mounted as samples.The distance between the samples and the plasma is approximately 1 cm.The backside of the samples can be accessed by an infrared beam from an FTIR spectrometer.The chamber can also be operated using only a PMMA insert instead of the substrate holder to produce PAW through more extended plasma operation.Thereby, the impact of PAW on the copper surface is investigated separately from direct plasma exposure.

Sample preparation
Samples consisting of silicon wafers coated with a thin copper film with a typical thickness of 15 nm are used as a substrate.Alternatively, optical cavity substrates (OCS) consisting of silicon wafers with a 1000 nm oxide layer on both sides are used [22].These OCS substrates may have a threefold impact on the experiments: (i) the OCS enhances the sensitivity of the infrared diagnostics at specific wavenumbers; (ii) the OCS as a substrate changes the electrical conductivity of the system and allows the evaluation of any changes by electrolysis; (iii) the thick SiO 2 layer may serve as a large reservoir for oxygen atoms also inducing oxidation of the copper atoms deposited on top of the OCS.
The deposition of the copper coating is performed using a high power impulse magnetron sputtering (HiPIMS) system, where a copper target is installed.The silicon wafer is placed underneath the target at a distance of 7 cm.The coating process is performed in Argon at a pressure of 0.5 Pa at 740 V, 109 A, 1 kW, 35 Hz and a pulse time of 50 µs for 16 s.This results in a layer thickness of 15 nm.To determine the deposition rate of the coating procedure in the HiPIMS setup, a reference sample was coated for five min using the above-mentioned parameters.The thickness of the deposited copper layer was measured using a profilometer, yielding a deposition rate of 0.9 nm s −1 .

In situ-IR surface diagnostic
The surface is observed via FTIR (Bruker IFS 55 Equinox) in reflection mode by sending an IR beam from the backside through the substrate towards the very thin copper film (see figures 1(a) and (b)) on top.The exposure to air and the subsequent immersion of the copper film into water will inadvertently create a copper oxide surface that may contain copper oxide CuO and cuprous oxide Cu 2 O.The FTIR spectra are acquired in situ between the surface plasma treatments.The FTIR spectrometer and plasma can not be operated simultaneously due to the electrical noise from the discharge disturbing the spectrometer.Consequently, in operando, monitoring of the sample is not possible.
Silicon and oxidized silicon wafers are chosen as substrates.The oxide layer thickness of the OCS of 1000 nm is designed to lead to an enhancement of the sensitivity at 2000 cm −1 at an angle of incidence of 70 • by a factor of 10.Here, the angle of incidence is only 45 • , and the wavenumber range of interest is around 600 cm −1 , where the sensitivity enhancement of a silicon wafer and an OCS are somewhat similar.
The most prominent absorption lines of copper oxides can be found for CuO at 480 cm −1 and 527 cm −1 [23,24], for Cu 2 O at 617 cm −1 [25], as well as at 628 cm −1 for copper hydroxide Cu(OH) 2 [26].These line identifications are based on experiments using copper oxidation in a furnace [23,24] or electrochemical deposition methods [27,28] (some sources identify line positions also at 1100 cm −1 [29] upon copper annealing on glass plates, but these lines are presumably signals from the underlying SiO 2 substrates).In addition, absorption bands associated with the SiO 2 substrate and the CO 2 background in the beam path must be regarded during spectra interpretation.All line positions are listed in table 1.
Based on these line positions, we expect changes in the copper oxide peaks at 527 cm −1 , at 617 cm −1 , and at 628 cm −1 .Bands below 500 cm −1 suffer from the poor signal-to-noise ratio and will not be interpreted.The reactions of the plasma species with the underlying SiO 2 layer are expected to become visible at the line positions 505 cm −1 , and 880 cm −1 and around 1050 cm −1 , with the latter one being the strongest.These refer to the rocking and the symmetric and asymmetric stretching modes of SiO 2 .
The sensitivity of the IR measurement is evaluated for the two types of substrates being used, either OCS with a thick oxide layer of 1000 nm or regular silicon substrates.In the optical modelling of these samples, we assume a 5 nm film at the top side of the wafer and the reflection of an IR beam at an angle of incidence of 45 • from the backside.The ratio R/R 0 of the reflectivity R assuming a film with absorption versus the reflectivity R 0 for a film with no absorption is being calculated by a multilayer optical model for light that is polarized parallel and perpendicular to the plane of incidence.Since we do not polarize the IR light intentionally, but it has to pass several mirror reflections at an angle of incidence of 45 • , the light is expected to be mainly polarized perpendicular to the plane of incidence.As test absorption, we chose a series of Gaussian peaks spaced by 100 cm −1 with a width ∆k ≃ 30 cm −1 defined for the imaginary part of the refractive index.The variation of the real part n is calculated using the Kramers-Kronig relation and choosing n ∞ = 2.This choice is motivated by tabulated data for the refractive index of Cu 2 O [35]. Figure 2 shows the results R/R 0 for the response of an OCS with an oxide layer of 1000 nm on both sides (a) and of a standard silicon with a natural oxide thickness of 1 nm (b).The model for the optical constants n and k for the series of oscillators at different wavenumbers is shown as a solid line in figure 2(c).The literature data for Cu 2 O [35] are shown as dashed lines.
One can see that the change in R/R 0 varies with the central wavenumber of the oscillators.The variation of R/R 0 is more complex for the OCS substrates than for the silicon substrates due to the overlap with the multiple interferences within the thick oxide layer in the OCS.The signal enhancement of the OCS favours the parallel component of the light.However, it is also relatively insensitive at 500 cm −1 and at 1100 cm −1 .The standard silicon wafer's sensitivity over the analysed wavenumber range does not vary too much.
At the wavenumber of the CuO x species, the sensitivity of the OCS is slightly less than that of the silicon wafer, but both substrates are sensitive in the range of the CuO x absorptions.A 5 nm Cu 2 O film leads to a signal change of 0.5%.This is above the noise level in this wavenumber range of 0.2%.
It is also interesting to note that the minima in R/R 0 are typically at 13 cm −1 -19 cm −1 lower wavenumbers than the maximum in the imaginary part of the refractive index k: An absorption in k at 600 cm −1 leads to a minimum in R/R 0 at 587 cm −1 for the Si substrate and at 581 cm −1 for the SiO 2 substrate.This is due to the strong modulation of the real part of the refractive index.This shift in k can be seen in the data below.Such a shift is not seen for weaker absorption signals, such as for the CO 2 signature originating from the ambient in the beam path.

Ex-situ-XPS and SEM
The surface modifications are investigated ex-situ using XPS and scanning electron spectroscopy (SEM).Samples are prepared as described above.For the SEM measurements, samples with a 50 nm copper layer are used to increase the conductivity of the sample to avoid charging effects in the SEM.For both surface analysis methods, several locations on the same sample are examined for statistics to analyze any changes before and after plasma or PAW treatment.In addition to the changes induced by the plasma, any modification of the samples by exposure for a few hours to distilled water or to PAW only is investigated.
XPS is used to analyse the composition and bonding structure of the samples.The XPS analysis of the samples follows the procedure from Biesinger [36].Quantification of the four constituents of the samples, namely pure copper, copper hydroxide, cupric oxide and cuprous oxide, was derived from the Cu LMM Auger peaks of the spectra.The spectra were first charge corrected, a linear background was applied and the data was then compared to the sum of the properly weighted reference spectra, taken from literature [36].The processed spectra and their associated Cu 2p peaks can be found in the appendix, as well as a more detailed explanation of the data processing.

Plasma-induced copper oxidation and reduction
The sample preparation using the HiPIMS setup and the subsequent mounting in the plasma-in-liquid system cannot prevent the initial oxidation of the sample due to the exposure to the ambient air and distilled water before plasma exposure.In the following, we analyse three cases: the oxidation in ambient air followed by exposure to water for a long time or for 30 min or for only a few minutes.
At first, we regard the plasma-induced reduction of a freshly deposited Cu film mounted within the water-filled setup for at least a day before the measurements were taken.This interaction inside the liquid is expected to cause a thorough oxidation of the topmost layers.Figure 3 shows the measurement signal R/R 0 for a varying number of pulses for a plasma operated at +20 kV in front of a 15 nm CuO x film on top of a silicon substrate (a) and an OCS (b).The noise level is approximately R/R 0 ≃ 0.2%.The plasma fluence is expressed as a number of pulses, as indicated at each spectrum.The IR peaks are assigned in the figure .The experiments clearly show the removal of Cu 2 O or of Cu(OH) 2 rather than the formation of copper oxide on both samples.The quantification of the experiment reveals a removal of typically 5 nm copper oxide surface.Such a removal could correspond to a net removal of the complete top layers or to the conversion of copper oxide into metallic copper.Both cases cannot easily be distinguished.
In the case of the OCS substrate, one also sees a significant change in signal at the position of the SiO 2 rocking and stretching mode.This could be a reduction of the underlying SiO 2 layer as well, but also just related to the optical signal due to the change of the copper oxide film reflectivity: Since SiO 2 has very strong absorption lines and the data correspond to a ratio between the spectra R and the background R 0 , any small changes become instantly visible at the SiO 2 peak positions.These changes are minimal for the regular Si wafer, as expected.
The loss of Cu 2 O or of Cu(OH) 2 may also correspond to the result of an oxidation reaction if one assumes that Cu 2 O is converted into CuO.However, at the wavenumbers of CuO groups at 527 cm −1 , no significant signal change can be observed, although the signal-to-noise ratio at wavenumbers below 550 cm −1 may mask those changes.As discussed below, the XPS analysis of the plasma-treated samples shows no creation or removal of CuO groups upon plasma treatment.
The dynamic of the copper oxide removal shows a delay at the very beginning.This is analysed using an optical model, where we assume that Cu(OH) 2 is converted into CuO 2 before CuO 2 is eventually dissolved or converted into metallic Cu.This reaction sequence is motivated by the XPS analysis of the plasma-treated samples, as discussed below.For this, the change in R/R 0 is converted into a change in film thickness of Cu 2 O as ∆d Cu2O .We assume an initial thickness for Cu(OH) Figure 4 shows the resulting changes in film thickness with time.At first, one can see that the dynamic is identical on the OCS and the Si substrate.The substrate does not influence the reduction reaction, although the conductivity of the OCS and the Si wafer are very different.Secondly, one sees a delay in removing Cu 2 O for 100 plasma pulses.The data Second, we analyze the plasma-induced treatment of the CuO x films for freshly deposited Cu films mounted in the plasma-in-liquid system for half an hour before plasma exposure.In this case, the initial oxide originates from exposure to air and to the liquid.Figure 5 shows the measurement signal R/R 0 for the treatment of a 15 nm film after varying time spans for a plasma operated at +24 kV for a given number of pulses for a silicon wafer (a) and an OCS (b).The peak identification is assigned in the figure .One can see that the oxide removal on the OCS is similar to the case for a long time of water exposure.However, the oxide removal using a Si wafer as substrate takes much longer than the data shown in figure 3.This can be most easily explained by the assumption that the oxide for a copper film on a silicon wafer is only initially formed at the interface to the natural oxide on the Si wafer.At the onset of plasma exposure, the copper film is slowly removed.When the etching reaches the Si-Cu-O interface, a loss in Cu 2 O or Cu(OH) 2 groups becomes visible.In the case of an OCS as substrate, the large concentration of oxygen in the OCS allows complete oxidation of the sample prior to plasma exposure.
Third, we regard the plasma-induced reduction of the CuO x films for freshly deposited Cu films exposed to the plasma-inliquid system directly after preparation in the HiPIMS setup.In this case, the initial oxide may only originate from exposure to air and a very short time to the liquid.Figure 6 shows the measurement signal R/R 0 for the treatment of a 15 nm film after varying time spans for a plasma operated at +24 kV for a given number of pulses (a) and for exposure to simply the distilled water environment without striking a plasma (b).The peak identification is assigned in the figure .One can see that the signature of oxide removal is very pronounced for the plasma-exposed film but very small for exposure to water only.The removal of the Cu 2 O or Cu(OH) 2 generated by exposure of a freshly deposited copper film to the ambient and a short time to water is very fast compared to the reduction of a film exposed to water for a long time.A small removal rate of copper oxide by exposure to water only is visible.This might be due to loosely bonded copper oxide groups.One may state that the etch resistance of the samples depends very much on the preparation history.The shorter the exposure to the liquid before plasma treatment, the higher the removal rate.
By applying the voltage to the electrode, the current in the system may also affect the copper layer by regular electrolysis.To compare the plasma-induced changes to those caused by regular electrolysis, we operated the experiment at first at +10 kV, which lies below the threshold for plasma generation, followed by +24 kV and striking a plasma, as shown in figure 7. Since the sample is electrically not connected, the experimental circumstances are most similar to bipolar electrolysis [37,38], where a floating third electrode (in our case the copper sample) can undergo electrolytic reactions because the potential difference between the surface and the electrolyte varies at both ends of the sample facing either the high-voltage electrode or the grounded counter electrode.
One can see that no copper oxide is being formed or removed at +10 kV.At +24 kV and for treatment times longer than 120 s (1200 plasma pulses), the removal of Cu 2 O or Cu(OH) 2 can be observed.One may argue that the plasmainduced reduction occurs via the creation of solvated electrons by the plasma and a subsequent reaction at the copper oxide surface.Aditionally, the plasma electrolysis method is known to produce a higher amount of reactive species than conventional electrolysis [15].An increased amount of atomic hydrogen may also be the cause of the surface reduction.At the same time, the pressure waves from the plasma support the transport of reactive species near the electrode tip to the sample.The polarity of the plasma electrode does not influence such a transfer of the solvated electrons or other charged species to the sample because it occurs in the off time of the plasma, where the electric fields are absent.
According to literature, the pH value and temperature of the liquid environment can influence the formation of the CuO x species [39].Based on previous experiments [20], where an electron density of up to 10 25 m −3 in a 50 µm radius plasma region had been deduced from Stark broadening measurements, it can be estimated, that the solvated electrons cause a decrease in local pH value from 7 to 6.5 in a 1 ml volume surrounding the electrode tip for a treatment time of 20 s at 10 Hz.At pH values below 7, as present in these experiments, any formation of CuO is suppressed [39].Lower pH values also favor the reduction to metallic copper [40].
Finally, the plasma-in-liquid-induced treatment of the copper oxide surface is analysed using carbonated water.In figure 8, one can see that the reduction of oxides is much more pronounced compared to plasma treatment in distilled water.The pH values of distilled and carbonated mineral water are somewhat similar, but in the case of carbonated mineral water, other ions such as Na + , K + , Mg + 2 , Ca + 2 , Cl − , SO 2− 4 , and HCO − 3 also contribute to the reduction reaction and enhance the OH − density in the liquid.This leads to a much higher reactivity because various copper electrolyte ions complexes can be produced.Furthermore, CO 2 dissolved in water reacts first to H 2 CO 3 and then further to HCO − 3 and H + [41].These water splitting products are very reactive, and for example, H + ions can remove oxygen from the Cu x O surface, leaving ionic copper behind, which dissolves into the liquid.In addition, H + formed by water dissociation by the plasma also enhances the reduction rate of the surface.This is in agreement with our results.

XPS analysis of the sample surfaces
The sample surfaces for the different treatments are analyzed by XPS (for details of the analysis, see the appendix).Basically, four samples are compared: (i) an initial asdeposited surface, (ii) a copper surface exposed to distilled water for 4 h as a reference, (iii) a copper surface exposed to the plasma in a liquid system performed at 20 kV and 10 Hz for 3 min, and (iv) a copper surface exposed to PAW only.The The concentration of Cu 2 O after contact with distilled water was determined as 46,7%, and the plasma treatment caused an increase to 63,8% (see appendix).Such a difference can easily originate from the very different information depths of XPS versus FTIR.XPS samples information from the very surface, where the plasma treatment causes an increase of the Cu 2 O groups.When at the same time, however, the complete film is slightly etched, and the absolute amount of Cu 2 O at the very surface, but also in the bulk of the film, is removed, a net loss is only observed in the FTIR spectra, which samples the complete film.Summarizing, although the total amount of copper oxides decreases by the plasma treatment, the concentration of Cu 2 O at the very surface increases.This hints at a plasma-induced Cu 2 O nanocrystal growth.
• PAW treated sample: The PAW-treated sample shows a similar reduction, although it appears rather a conversion of Cu(OH) 2 and of Cu 2 O to metallic copper.

Surface structures
The XPS analysis shows the presence of copper and of various copper oxides at the very surface.It is reasonable to assume that these concentrations are not homogeneously distributed but rather laterally separated so that, for example, copper oxide nanostructures are situated on a metallic copper surface.This is analysed by SEM for samples exposed to distilled water and PAW.The PAW treatment is a proxy for the plasma-in-liquid surface treatment because the growth of copper oxide nanostructure takes a long time; any analysis of the surface treated by a few plasma pulses only would not create a nanostructure large enough for SEM analysis.Figure 10 shows an SEM image of a 50 nm copper surface treated with PAW for 16 h.The PAW was produced at +20 kV at 10 Hz for 30 min.One can see that the surface primarily consists of metallic copper (indicated by the dark areas in the SEM image) with smaller copper oxide islands.These islands have a diameter up to 100 µm.Such an oxidation pattern is typical, where the oxide formation starts at defects and step edges at the copper surface.
In figure 11, these oxide islands are investigated more closely.For this, the samples are exposed to distilled water (figure 11(a)) and to PAW (figure 11(b)).PAW sample solutions have been prepared by operating a +20 kV plasma for 20 min and treating a Cu wafer with PAW for four hours.
One can see that irregular Cu 2 O nanocubes are already formed by treating the samples with distilled water (see figure 11(a)).On these long time scales, copper nanocubes grow along the preferred crystalline orientation from dissolved Cu(OH) 2 in the liquid to considerable sizes, starting at nucleation seeds.One could argue that the very bright areas correspond to Cu 2 O nanocrystals, the grey areas to Cu(OH) 2 surface layer, and the very dark regions either to metallic copper or the underlying silicon substrate.It is interesting to note that the treatment in PAW leads to the formation of more regular nanocubes (see figure 11(b)), which are also preferentially oriented.It seems that the polycrystalline copper layer serves as a template to cause such a preferred orientation of the nanocubes.The crystalline nature of the substrate may still be visible when regarding the parallel lines originating from step edges in the image (see the lines marked in the top left corner in figure 11(b)).These step edges are aligned with the orientation of the copper nanocrystals, which suggests an epitaxial growth mode.By providing a more reactive environment such as PAW, the development of well-ordered Cu x O nanocubes is more pronounced because the competition between the oxidation and reduction reaction enhances the selectivity of copper  oxide nanocube formation, leading to a finely patterned surface.The shape of the nanocubes resembles cuboctahedrons, which indicates that the growth rate of the Cu 2 O (111) and (100) facets are rather similar [42].

Discussion
The interaction of the plasma-in-liquids with a copper oxide surface is a competition between reduction and oxidation.The oxidation occurs mainly in the off time in the liquid itself but depends on the composition of the initial sample.The individual possible reaction steps are illustrated in figure 12 and discussed in the following: During the initial preparation of the samples in the HiPIMS system, copper is deposited onto a Si wafer or on an OCS.The Cu atoms start to react either with the natural oxide on a Si wafer or the oxygen in the thick oxide layer on the OCS.As a result, one would expect a small Cu x O layer at the interface between the Cu film and the Si wafer.In the case of the OCS, a freshly deposited Cu layer may undergo more thorough oxidation because the supply of oxygen atoms from the SiO 2 layer may be large enough to oxidize a thin copper film of a few nanometers.
When these samples are exposed to the ambient, oxygen and water vapour may react with the surface creating a copper oxide layer on top.De Los Santos Valladares et al [11].investigated the oxidation process of 100 nm copper surfaces on SiO 2 substrates through thermal annealing and proposed an initial oxidation mechanism during which oxygen from the ambient is adsorbed to the surface, causing a thin oxide layer to form.They assume the presence of oxidation sites on the surface, from which the oxide islands expand until a completely covered oxide surface is reached.Subsequently, the copper oxide grows into the bulk at a decreasing rate over time as being limited by oxygen diffusion in the material.The temperature of the sample controls this thermal oxidation rate.In our experiments, the temperature was kept at room temperature, and it is reasonable to assume that the oxidation is incomplete and some metallic Cu remains.This agrees with the XPS results shown in figure 9.However, it should be noted that the copper films used in this work ranged in thickness from 15 to 50 nm, and the oxidation process for very thin layers may differ from the thermal oxidation of thicker films or bulk material.
When brought into contact with the liquid, a copper hydroxide interface layer Cu(OH) 2 forms by the reactions of Cu with OH species from the dissociation of H 2 O 2 or from residual OH − ions in water.This oxidation reaction is expected to be accelerated when the plasma-in-liquid system is operated since the plasma is a source of H 2 O 2 .The oxidation reaction occurs as: This Cu(OH) 2 dissolves in the liquid corresponding to a net removal of copper oxide groups.This dissolved Cu(OH) 2 may also re-deposit in the form of Cu 2 O nanocrystals following: ( The pathway of redeposition is supported by the fact that the formation of micrometer-sized nanocrystals is observed, although the copper source is a 50 nm thin film.If such nanocrystals would grow directly from bulk copper, one would expect crystal sizes in the range of 50-100 nm only.This is not observed.The driving force for this conversion of dissolved Cu(OH) 2 into Cu 2 O crystals could either be a reduction due to the presence of solvated electrons and hydrogen species from the plasma or just driven by classical crystal growth dynamics.In the electrochemical synthesis of Cu 2 O nanocrystals, it is driven by a current passing the anode since the oxidation state of the copper atom in Cu(OH) 2 to Cu 2 O changes from +2 to +1.The Cu 2 O growth is a very slow process in our system, which favors an epitaxial growth mode resulting in oriented Cu 2 O crystals, as observed in the SEM images.This epitaxial growth is especially pronounced in the case of PAW.This could be explained by a competition between oxidation and etching in PAW.As a result, the more etch-resistant structures will survive, resulting in a very ordered pattern of nanocubes at the surface.In the case of a less reactive environment, such as regular water, more disordered structures evolve.
This shows that the initial surface oxide layer either grows in the form of Cu 2 O nanocrystals, and at the same time, it is etched due to the dissolving copper into Cu(OH) 2 .Consequently, depending on the nature of the substrate (either Si or OCS) and on the time of water exposure, the final sample consists of a fully or partially oxidised copper film.
The reduction reaction from dissolved Cu(OH) 2 to Cu 2 O and also from Cu 2 O to Cu could be promoted by the solvated electrons, H and H 2 generated by the plasma in the liquid.The number of solvated electrons is estimated as follows: As stated above, a plasma treatment of 20 s at 10 Hz yields an initial electron density of up to 10 25 m −3 in the vicinity of the electrode tip based on previous experiments [20].The radius of the initial plasma region is 25 −6 m.The concentration of electrons N e created by the plasma can be estimated using: where n e is the initial electron density, V the initial plasma volume, f the frequency (10 Hz in this work), t the treatment time and mol the molar mass.The created electrons are assumed to distribute across the entire volume of 25 ml.After plasma expansion, these electrons become solvated in the liquid.The shortest treatment time applied in this work is 10 s, and the longest is 20 min.These correspond to estimated electron concentrations of 3.27 * 10 −10 mol/m −3 and 1.31 * 10 −8 mol/m −3 , respectively.We assume that a volume of 1 ml surrounding the electrode tip is being affected by these solvated electrons.We can calculate 1.5 • 10 15 of electrons generated within this 1 ml volume.The number of electrons corresponds to one electron per surface atom, and a significant electrochemical impact on the film surface is to be expected.This plasma-induced reduction causes the transformation of surface Cu x O into metallic Cu according to the XPS measurement.No significant formation of CuO can be observed.This could be most easily explained by the assumption that CuO formation is a temperature and pH-dependent process [11,39] and conditions necessary for its formation are not present in these experiments.Additionally, the net etching of the top layers would cause a removal of the little CuO possibly formed at the surface.

Conclusion
The effect of in-liquid plasma treatment and PAW treatment of Cu x O thin films was investigated.It was found that both oxidation and reduction co-occur on the surface.Contrary to initial expectations, plasma treatment was found to reduce the Cu(OH) 2 species to Cu 2 O and further to metallic copper.This is surprising, considering the lifetime of the oxidising species is longer than the reducing species.The solvated electrons, H and H 2 produced by the plasma are most likely the cause of the reduction reaction.Furthermore, the use of carbonated water was found to increase the reduction of the Cu 2 O.This suggests, the plasma-in-liquid system may not be suitable for reoxidation under the experimental conditions used in this work.However, PAW treatment was found to oxidise the samples slowly on long timescales in the range of hours.Additionally under PAW influence, well-ordered Cu x O nanocrystals could be seen.Therefore, PAW treatment may be a viable method for creation and possibly reactivation of copper oxide catalysts.In the future, it will be essential to evaluate the performance of these surfaces in electrolysis cells for conversion of CO 2 to hydrocarbons.
To make the in-liquid-plasma a suitable device for reactivation of the catalytic surfaces, the oxidation of the sample needs to take place on short timescales.This is already the case and longer treatment times are only used to grow larger Cu 2 O nanocrystals that can easily be visualised by SEM.Nevertheless, further investigations are needed to determine the optimal timing in between the PAW having an oxidizing effect and the solvated electrons and hydrogen species having a reducing effect.One option might be the treatment of the surfaces separated from the plasma only by PAW using a flow-type connection.   is compared to the sum of the properly weighted reference spectra of the four constituents.It has to be noted that the reference spectra in [36] are defined using a linear background in a fixed energy range.Therefore, the same background and energy range has to be applied to the measured spectra as well.
A.1.1.Analysis of samples from plasma treatment.The Cu2p spectra display shake-up satellite peaks before plasma treatment in (see figure 13(a)) that are absent after treatment (see figure 13(b)), indicating a removal of oxides.The plasma treatment was performed at 20 kV and 10 Hz for 3 min on a sample with 50 nm Cu layer thickness.

A.1.2. Analysis of samples from water and PAW treatment.
XPS measurements on a PAW-treated sample are performed using Cu samples treated with distilled water for 4 h as a reference.The PAW was produced at +20 kV during a treatment time of 30 min.The fraction of the different constituents in the samples Cu:Cu(OH) 2 :Cu 2 O:CuO was calculated from the analysis of the spectra shown in figure 16 to be 16% : 37% : 47% : 0% for the treatment in distilled water and 23% : 34% : 43% : 2% for the treatment in PAW. Figure 15 shows both samples to be oxidised, indicated by the presence of the shake-up satellite peaks.In comparison to the surface composition of the initial sample, as shown in figure 14(a), it is evident from the Auger peak analysis, that PAW and distilled water seem to oxidise the surface.

Figure 1 .
Figure 1.Experimental setup: (a) plasma setup including ns pulser, (b) sketch of the FTIR measurement scheme.

Figure 2 .
Figure 2. Optical modeling of the response R/R 0 of the IR measurement for an OCS substrate (a) and a silicon wafer (b).Model for the optical constants n and k for oscillators at different wavenumbers and a width of ∆k = 30 cm −1 (c).The optical constants for Cu 2 O from the literature [35] are shown as dashed lines.

Figure 3 .
Figure 3. IR spectra for the plasma-induced reduction of a 15 nm CuOx film on top of a silicon substrate (a) and an OCS (b).The samples were oxidised in the liquid for at least a day.The voltage is +20 kV, and the labels indicate the plasma fluence expressed as a number of plasma pulses.

Figure 4 .
Figure 4. Change in the Cu 2 O signal expressed as absorbing film thickness from the optical model for the film on the OCS and on the silicon substrate.The dashed line denotes a simple rate equation model for conversion of Cu(OH) 2 into Cu 2 O followed by Cu 2 O removal.

Figure 5 .
Figure 5. IR spectra for the plasma-induced reduction of a 15 nm CuOx film on top of a silicon substrate (a) and on top of an OCS (b).The exposure to the liquid was only 30 min before plasma exposure.The voltage is +24 kV, and the labels indicate the plasma fluence expressed as a number of plasma pulses.

Figure 6 .
Figure 6.IR spectra for the plasma-induced reduction of copper film on top of a 15 nm Si substrate by plasma exposure (a) and by exposure to water only (b).

Figure 7 .
Figure 7.IR spectra for the plasma-induced reduction of copper film on top of a 10 nm OCS by electrolysis at +10 kV followed by plasma electrolysis at +24 kV at 10 Hz.

Figure 8 .
Figure 8. IR spectra for the plasma-induced reduction of copper film on top of a 15 nm OCS at 20 kV with carbonated water used as the liquid.Number of pulses as indicated.

Figure 9 .
Figure 9. Quantification of surface concentrations by XPS for different treatments as indicated.

Figure 10 .
Figure 10.SEM analysis of a 50 nm Cu film after treatment with plasma-activated water generated at +20 KV at 10 Hz for 30 min and treating the sample for 16 h.

11 .
SEM analysis of a 50 nm Cu film after treatment with (a) distilled water and (b) plasma-activated water generated at +20 kV at 10 Hz for 20 min treated for 4 h.The white lines in (b) mark step edges that seem to be decorated by the oxidation process.These steps edges cover large parts of the sample.

Figure 12 .
Figure 12.Schematic reactions at the copper surface in the plasma-in-liquid system.

Figure 13 .
Figure 13.Cu2p 3/2 spectra measured by XPS of a 50 nm sample film before (a) and after (b) plasma treatment at 20 kV and 10 Hz for 3 min.Spectra were shifted for legibility.

Figure 14 .
Figure 14.Cu LMM spectra measured by XPS of a 50 nm sample film before (a) and after (b) plasma treatment at 20 kV and 10 Hz for 3 min.

Figure 15 .
Figure 15.Cu2p 3/2 spectra measured by XPS of induced surface changes caused by (a) PAW and (b) distilled water on a 50 nm surface.Treatment time h overnight.Spectra were shifted for legibility.

Figure 16 .
Figure 16.Cu LMM spectra measured by XPS of induced surface changes caused by (a) PAW and (b) distilled water on a 50 nm surface.Treatment time 15 h overnight.

Table 1 .
Line positions of SiO 2 , CO 2 , CuO, Cu(OH) 2 and Cu 2 O IR absorption bands.Preparation, or origin of the bands in brackets).
2 and for Cu 2 O of d Cu(OH)2,0 and d Cu2O and two rate constants k 1 for converting Cu(OH) 2 into Cu 2 O and k 2 for etching Cu 2 O or for converting Cu 2 O into Cu.This yields: