Titanium anodizing in a choline dihydrogencitrate salt–oxalic acid deep eutectic solvent: a step towards green chemistry in surface finishing of titanium and its alloys

Deep Eutectic Solvents (DESs) are “green” competitors for some conventional plating baths and electrolytes used for surface modification. Their use allows a material to be obtained with a structure different from that observed in conventional plating or finishing technologies. In this work the titanium anodizing process was investigated in a bath based on a choline dihydrogencitrate salt and oxalic acid (1 : 1 molar ratio) green solvent. Titanium anodized at the lowest voltage applied (10 V) was a deep yellow color, which turned to deep blue at 30 V. The surface morphology and topography of titanium, both anodized and untreated, were monitored by optical, scanning electron (SEM and HR-SEM) and atomic force (AFM) microscopy. Anodizing at 10 V produced a fine granular morphology of the oxide layer, while anodizing at 30 V led to the formation of a probably thicker and quite uneven oxide layer, characterized by a distinct and coarse granular morphology. The average size of the micro-nodules was higher than those at 10 V and porous structures have been also identified. According to X-ray photoelectron spectroscopy (XPS) the stoichiometric TiO2, regardless of the applied voltage during anodizing, was practically the only component of the oxide layer produced on titanium in the DES bath. At 10 V, the oxide layer was thicker (>10 nm) than the natural Ti passive layer (approx. 2.2 nm), which, apart from TiO2, also contained oxides of titanium at lower oxidation states, i.e. +2 and +3. Moreover, the XPS technique was supported by electrochemical impedance spectroscopy (EIS), especially in the context of the structure of the oxide layer and its interaction with a corrosive environment. The corrosion resistance of anodized titanium was assessed in 0.05 mol dm−3 solution of NaCl by the linear polarization resistance (LPR) technique and polarization curves. During interpretation of the impedance spectra, the layers produced by the anodizing process were described using the two-layer model. It was assumed that the inner layer formed directly on the surface of metallic titanium was responsible for the barrier properties (resistance of 2.8 MΩ cm2). The porous outer layer formed on it has a much lower corrosion resistance, i.e. 800–1300 Ω cm2.


Introduction
Titanium is a metal with very good corrosion resistance and has a very high strength-to-density ratio. These properties determine the wide use of this metal for the production of many devices operated under aggressive conditions. The corrosion resistance of titanium is due to the passive oxide layer spontaneously formed on its surface, the thickness of which can be up to 10 nm.
The morphology and structure of the oxide layer can be modied and the TiO 2 layer thickened by an electrochemical or chemical process. The titanium anodizing process can be carried out in various solutions, among which most oen these are based on: sulfuric(VI) acid, chromic(VI) acid or phosphoric(V) acid, in which oxide layers are formed of relatively high roughness. Thinner and less porous passive layers are obtained by anodizing titanium in solutions such as: citric acid, tartaric acid or boric acid, and quite recently also in a deep eutectic solvents (DES). Anodizing greatly improves the corrosion resistance of titanium. Depending on the anodizing conditions, i.e. applied voltage, bath composition, temperature and process timeoxide layers with a thickness of several to several hundred nanometers can be obtained. Layers of different thicknesses have a different color: gold, purple, blue, orange, green. It is used in the jewelery industry to obtain decorative effects and improve the aesthetics of products, and in other industries for marking and identifying products. Titanium is used in medicine as the basic metallic material for implants. This use, in addition to corrosion resistance in the environment of physiological uids, requires titanium to be biocompatible and bioactivity allowing the implant surface to connect to the bone. Titanium anodizing leads to an improvement in abrasion resistance, which is particularly important in the case of products manufactured for the needs of the aviation, military and tool industries. The TiO 2 layer produced in the anodizing process can improve the adhesion of paint coatings to the substrate. Finally, titanium oxide layers with a crystalline structure show photocatalytic and hydrophobic properties, which can be used in the production of photovoltaic cells, anodes in lithium-ion batteries and self-cleaning materials.
The interest in the use of deep eutectic solvents in surface nishing of various construction metals and their alloys is constantly growing. DES are strong competitors for some conventional plating baths and electrolytes used for surface modication. Compared to standard baths, which are oen corrosive and harmful to humans, DES are environmentally friendly, non-toxic, biodegradable, inexpensive and easy to produce. 1 Can be treated as green solvents. Their use allows to obtain a material with a structure different from that observed in conventional plating or nishing technologies. In recent years it has been show that in many areas it is possible to transfer the electroplating processes from aqueous baths to those based on DES. This is the case i.a. with Ni and composite Ni plating, 2-5 Zn 6 and Zn alloy plating 7 and especially Cr(III) plating. 8,9 Besides, it can be clearly seen in scientic literature that research is oen directed towards surface nishing of specic construction materials such as stainless steels and light metalstitanium and aluminum alloys. For these materials, the emphasis is on the use of DES-based baths mostly for two different processes: anodizing or electropolishing.
The electropolishing of stainless steel and pure metals in the most popular DES bath based on choline chloride and ethylene glycol was studied by: Abbott et al., 10 Karim et al. 11 and Protsenko et al. 12 Similar DES was used in studies of aluminum electropolishing by A. Kityk et al. 13 and T. M. Abdel-Fattah et al. 14 Recently, A. Kityk et al. 15 and W. O. Karim et al. 16 investigated the electropolishing process in baths based on DES -rst for Al-Mg alloy and second for titanium. Both achieved a signicant reduction in roughness of the tested materials. However, there are very little works that describe titanium anodizing in DES.
In contrast to the electropolishing process, obtaining an anodizing effect of titanium in DES requires a slightly different approach. For example, it is imperative that the rate of dissolving the metal should not exceed that of forming stable oxide layer. Again, these green solvents (DES) potentially offer new environment to conduct the anodizing process of titanium, because the vast majority of titanium anodizing processes in common aqueous baths unfortunately require the use of: sulfuric acid, 17 phosphoric acid, 18 oen with the addition of hydrouoric acid 19 or ammonium uoride 20,21 and still hexavalent chromium compoundschromic acid. 22 Not so long ago, C. Y. Chen et al. in their studies of the titanium anodizing process in NH 4 F-glycerol electrolyte showed that the addition of DES (1 : 1 molar ratio of succinic acid and choline chloride) to anodizing bath allows to control the structure of TiO 2 nanotubes formed on its surface. 23 This only conrms the high potential of DES liquids in the surface nishing of titanium and its alloys.
Another obvious step would be to completely transfer the titanium anodizing process to the non-aqueous DES bath. It would require the use of DES solvent, whose physicochemical properties will allow for the anodizing process in a wide range of parameters, and by selecting the current-voltage conditions one could control the thickness and structure of the forming oxide layer. The literature cited above, as well as preliminary studies, showed that the most popular DES based on choline chloride (conventionally "ChCl") and ethylene glycol or carboxylic acids, during anodic polarization interact too aggressively not only with the titanium surface, but also with materials such as aluminum or even stainless steels. 24 The presence of chloride ions in DES bath may be a very likely cause of this behavior. Therefore, in this work we have performed thorough research on the feasibility of Grade 2 titanium anodizing in a deep eutectic solvent based on choline dihydrogencitrate salt (conventionally "ChCit") and oxalic acid (conventionally "OA") in 1 : 1 molar ratio. Looking from the point of view of electroplating and anodic polarization of metals in the DES baths, ChCit is not as popular hydrogen bond acceptor as choline chloride. ChCit nds application in biotechnology as the component of i.a. enzyme-friendly DES solvents for biocatalysis, 25 DES-based starch plasticizers 26 and aqueous twophase systems with DES solvents. 27 In addition, successful attempts to copper electrodeposition can also be found in DES plating baths composed of ChCit and ethylene glycol. 28 Finally, thanks to the use of ChCit : OA solvent, instead of ChCl : OA one, we get more so, i.e. relatively viscous and with lower ion mobility, eutectic mixture containing neither ethylene nor propylene glycol.
Anodizing of Grade 2 titanium was performed under potentiostatic polarization, and the effect of the process on the metal surface was assessed. Surface morphology, as a function of anodizing voltage, was monitored by optical, scanning electron (SEM) and atomic force (AFM) microscopy. Particular attention was paid to the structure, morphology and chemical composition of the obtained oxide layer. Due to the expected low thickness of this layer, X-ray photoelectron spectroscopy (XPS) was chosen for this purpose. The XPS technique was supported by electrochemical impedance spectroscopy (EIS), especially in the context of the structure of the oxide layer and its interaction with a corrosive environment. The corrosion resistance of anodized titanium was assessed in 0.05 mol dm À3 solution of NaCl also by linear polarization resistance (LPR) technique and polarization curves.

Materials and methods
The base material were disks of technical titanium (Grade 2, purchased from WOLFTEN Sp. z o.o., Poland), 14.8 mm diameter and 2 mm thick, geometric area 4.3 cm 2 , in the as-delivered form, i.e. without additional grinding and polishing. Before anodic polarization, titanium disks were rst ultrasonically degreased in methanol at 25 C for 10 min. Titanium was then etched and activated at room temperature in an aqueous mixture of HNO 3 (10%) and HF (1%) for 10 s under vigorous stirring. Aer activation it was thoroughly washed in deionised water under ultrasonic conditions and immediately dried with compressed air to get rid of residual water. DES bath was prepared by mixing the choline dihydrogencitrate salt ($98%, Sigma-Aldrich) with oxalic acid dihydrate (ReagentPlus®, $99.0% (GC), Sigma-Aldrich), both as supplied, in 1 : 1 molar ratio at 60 C until homogeneous and transparent liquid was formed. At this temperature the conductivity of DES bath amounted to 0.32 mS cm À1 . The electrolyzer for anodizing consisted of two rectangular mixed metal oxide titanium cathodes (MMO 167 by Umicore, N-type mesh) with 6 cm Â 3 cm dimensions placed in parallel in a 250 ml thermostated electrochemical cell. Titanium disk has been placed vertically between two cathodes with 20 mm spacing on each side, and the current was supplied to anode with a 304 stainless steel wire with a diameter of 0.5 mm (the steel did not come into contact with the bath during anodizing). Plating bath was not mechanically stirred. Anodic polarization was realized at a constant voltage U ¼ 10-30 V for 10 min at 60 C. Then the samples were thoroughly rinsed in deionised water (ultrasonic assisted), rinsed in methanol (also ultrasonic assisted), vacuum-dried and stored in a desiccator under argon atmosphere until required.

Research techniques
The basic tool for investigation of the surface morphology of anodized titanium was Leica DM6 M upright materials microscope equipped with Leica Flexacam C1 digital camera. The whole set worked under the control of Leica Application Suite X (LAS X) soware. The observation of the samples was performed by using LED illumination and incident light contrast methods: Brighteld (BF), Darkeld (DF), Differential Interference Contrast (DIC) and Polarization (POL). The images were captured at approximately 500Â magnication in 12 Mpix resolution, using Differential Interference Contrast (DIC) and Zaxis stitching technique to obtain extended depth of eld.
FEI Quanta 250 SEM microscope equipped with SDD Octane Elect Plus EDS detector was used to evaluate the effect of anodic polarization. SEM images have been captured in SE (secondary electron) mode with accelerating voltage of 10 kV, under 10 À4 Pa pressure, sample tilt of ca. 60 C and dynamic focus. EDS analysis was performed at take-off angle equal to 37 within the area of ca. 100 Â 100 mm at 25 kV accelerating voltage. SEM/Ga-FIB FEI Helios NanoLab™ 600i microscope was used for high resolution imaging of the surface. Surface topography was characterized by Nanosurf FLEX-Axiom atomic force microscope (AFM). Contact mode was used (static force) with a CONTR bar, while the measurement parameters were as follows: area 4 mm Â 4 mm, 1024 points per line, 1.5 s per line, 20 nN force.
The chemical analysis of titanium surface before and aer anodic polarization in DES bath was performed by X-ray photoelectron spectroscopy (XPS) using a SPECS PHOIBOS 100 spectrometer equipped with a non-monochromatized Al source (1486.7 eV). The spectrometer operated at 250 W for high resolution spectra. Surface etching during XPS measurements was carried out by Ar + sputtering with the beam energy of 4 keV and a beam current density of 7.5 mA cm À2 . The spectra were processed and tted in SPECLAB soware using a Gaussian-Lorentzian curve prole and a Shirley baseline. The C 1s peak at 284.8 eV was used as the reference.
Linear polarization resistance (LPR) has been used to track the corrosion resistance of anodized titanium in deaerated 0.05 mol dm À3 solution of NaCl over 48 h. LPR was performed by polarizing the samples from À15 to +15 mV vs. open circuit potential (E OC ) with a scan rate of 0.125 mV s À1 . Impedance spectra were recorded at the open circuit potential (E OC ) during 48 h of exposure to NaCl solution with a resolution of 10 pts per dec, in a frequency range from 100 kHz to 10 mHz and ac signal of 10 mV (rms). The measurements were carried out in 400 ml corrosion cell (Metrohm-Autolab) using Reference 600 (Gamry) potentiostat/galvanostat/ZRA. A stainless steel rod (geometric area exposed to the solution: 4 cm 2 ) and the AgjAgCl (3 M KCl) electrode (Metrohm) mounted in a Luggin capillary were used as the counter and reference electrodes, respectively. The geometric area of working electrode exposed to NaCl solution was 1.0 cm 2 . The corrosion cell was kept in a Faraday cage during EIS experiment. The NaCl solution was deaerated for each sample by purging with 99.9% pure argon for 20 minutes. Equivalent circuit modeling, graphing, analysis of impedance spectra, and determination of polarization resistance were performed using EchemAnalyst (Gamry) soware.

Results and discussion
3.1. The effect of titanium anodizing on its surface morphology Anodizing of titanium is associated with the formation of an oxide layer, the color of which depends on its thickness, morphology and structure. Therefore, the rst tool to evaluate an anodized surface was optical microscopy. Fig. 1 shows, besides a blank reference titanium surface, only the selected anodizing voltages, i.e. 10 and 30 V, because these samples have been chosen, as the representative, for further microscopic and spectroscopic analyses. The surface appearance of a sample in the as-delivered state was characterized by signicant unevenness and a pattern of scratches resulting from the rolling of the sheet from which disc samples were cut (Fig. 1). Chemical etching did not involve signicant changes, apart from showing the grain boundaries (Fig. 1). A common feature of all anodized samples was the variable color, which depended on the anodizing voltage. At the lowest of the applied voltages (5 V), a light yellow color was observed, which became more intense aer increasing the voltage to 10 V (Fig. 1). At 15 V the color started to change to light blue to achieve a distinct blue tint at 20 V. Increasing the anodizing voltage to 30 V resulted in a clear saturation of the blue color (Fig. 1). Fig. 2a presents a SEM microphotograph of the as-delivered titanium sample aer only degreasing step. For this surface, parallel scratches derived from rolling operation were observed. In addition, the surface was characterized by signicant heterogeneity and protruding pieces of material, probably the result of plastic deformation.
A clear change in morphology was observed only aer the etching of titanium (Fig. 2b). The use of an etching solution, composed of HNO 3 and HF, activated titanium and revealed its microstructure. The surface of the grains was also etched, which in places led to the formation of micropore-like structures (Fig. 2b). As the anodizing voltage was 10 V (Fig. 2c), the said "porosity" did persist, but aer anodizing at 30 V (Fig. 2d) it ceased to be the dominant feature. However, the grain boundaries were still visible (Fig. 2d).
Due to the expected very small thickness of anodic layers, the samples were further subjected to a high-resolution (HR-SEM) and atomic force (AFM) microscopy, which are denitely better to evaluate morphology and topography of thin lms. HR-SEM and AFM analyses conrmed formation of the oxide layer and a clear change in a layer morphology when the anodizing voltage has been increased from 10 to 30 V ( Fig. 3 and 4). Moreover, the observed layer morphology conrms that anodization in proposed DES is gentle and takes place in a completely different way than when DES solvent is used only as an additive to the anodizing bath. 23 Titanium surface aer only degreasing was characterized by a parallel arrangement of ne scratches (Fig. 3a) with some irregularities revealed in the AFM microscope photo (Fig. 4a)  and surface area roughness, S a , equal to 20.8 nm. Chemical etching caused these scratches to disappear and created more wavy surface (Fig. 3b). The observed surface morphology was granular and, due to higher S a (34.5 nm), very likely that this texture reected titanium grains, as evidenced by AFM in Fig. 4b. Aer anodizing at 10 V this wavy surface was somewhat "smoothed" (Fig. 3c) but still remained granular with lower S a ¼ 10.9 nm (AFM photo in Fig. 4c). Aer increasing the anodizing voltage to 30 V, the formation of a probably thicker and quite uneven oxide layer was observed (Fig. 3d). This anodic layer was characterized by a distinct and granular morphology (Fig. 4d), but the average size of the nodules together with S a ¼ 28.3 nm were denitely higher than those at 10 V. The topographic map of this layer also revealed some porous structures, clearly visible in the upper part of Fig. 4d (white arrows) and conrms earlier conclusions from microscopic observations Fig. 3d.
The results of microscopic analyzes: optical, scanning electron and atomic force indicate the formation of an oxide layer on the surface of technical grade titanium in the process of its anodic polarization. The scanning electron microscopy (SEM) technique used in conjunction with energy dispersion spectroscopy microanalysis (EDS), however, turned out to be insufficient. Therefore, it was decided to use X-ray photoelectron spectroscopy (XPS), which is predisposed to study the chemical composition and morphology of e.g. very thin passive layers on metals.

Surface chemistry of bare and anodized titanium
Native passive layers on titanium are of low thicknessusually of a few nm. Those made articially, in the anodizing process, are denitely thickertens of nanometers. Due to the low layer thickness in combination with the sampling depth in the XPS (several nanometers) technique, the result of the XPS analysis of such layers can be largely "distorted" by the outermost layer of contamination (adventitious) carbon. Therefore, rst XPS analyzes were performed both on the surface in the as-delivered state (i.e. without Ar + sputter cleaning) and then aer short and gentle Ar + sputter clearing. This gentle ion bombardment (90 s, 1 keV, 1.5 mA cm À2 ) allowed to get rid of this outermost layer of carbon, reveal the composition of the titanium anodic layer and at the same time avoid photoreduction of titanium. Table 1 shows the surface composition of titanium degreased and etched, and additionally anodically polarized. The surfaces of all analyzed titanium samples were covered with a layer of its natural oxides of various thicknesses, and the most outer surface was made of adsorbed carbon (adventitious carbon). Carbon, present on the surfaces of the samples in signicant amounts, was a typical contamination carbon. Its quantity decreased very quickly aer Ar + sputtering, from ca. 21 at% to ca. 4 and 7 at% ( Table 1). The structure of carbon bonds (spectra not included) in both anodized samples, i.e. 10 V and 30 V was similar: C-C/CH (56% and 50%), C-O (18 and 20%), C]O (8 and 10) and COO (18 and 20%), respectively. It is worth noting that aer Ar + cleaning the share of COO/COO-R groups still remained high, and even grew to 27% (at 10 V). Most likely it was an insufficiently rinsed, adsorbed residue of oxalic acid or choline dihydrogencitrate (spectra not included). On the surface of the reference Ti sample (degreased and etched) the share of COO groups was lower (ca. 9%), and aer Ar + cleaning (under the same parameters) it dropped to zero. Slightly higher carbon content on the surface of titanium anodized at 30 V could result from a greater surface roughness, and thus lower effectiveness of the Ar + ion bombardment. Aer Ar + cleaning of the surface of anodized samples, the ratio O : Ti ¼ 2 : 1, pointing to the stoichiometric TiO 2 . Additionally, it was conrmed by the deconvolution of highresolution Ti 2p spectra (Fig. 5b and c). Estimated bond energy for the Ti 2p 3/2 and Ti 2p 1/2 doublet was, respectively, 458.64 and 464.34 eV, which is characteristic of bulk TiO 2 . In the envelope of Ti 2p spectrum neither the Ti(0) component derived from the metallic substrate nor the Ti component in the lower oxidation states, i.e. Ti 3+ or Ti 2+ , was noticed. The absence of a metallic component also meant that the thickness of the oxide layer was greater than approx. 4 nm, because as is commonly accepted  sampling depth in XPS is 3l IMFP , as in the case of Ti 2p photoelectrons in TiO 2 means thickness of 3 Â 1.2 nm ¼ 3.6 nm. 29 A more detailed morphology of the oxide layers produced on titanium by the anodic polarization method as well as for titanium native oxides layers (as reference), was modeled using QUASES "Analyze" soware. 30 Overlayer thicknesses from deposited adventitious carbon, for Ti native layer, were rst determined by analysis of the Ti LMM and O KLL regions (Fig. 6 "as received") using the "Analyze" program and Mg Ka excitation. Then, the thickness of the oxide layer was determined using the O KLL region, but aer removing the contamination carbon from the surface (Fig. 6 "Ar + sputtered"). All spectra in Fig. 6 show a good overlap of the energy loss cross-section with the extrinsic loss portion of the scans. From "Analyze", assuming "burried layer model", a carbon overlayer of 1.2 nm and an overall lm thickness of 3.4 nm were determined. At this thickness of the passive layer, not only the dominant peak was visible in the high resolution Ti 2p spectrum of TiO 2 , but also metallic Ti (from the substrate), TiO and Ti 2 O 3 in shares respectively: 25, 55.5, 1.3 and 18.2% (Fig. 5). The procedures described in the works of Biesinger et al. 31 were used in the deconvolution of the Ti 2p spectra.
For the oxide layer formed on titanium during polarization at 10 V, a similar analytical procedure was used with QUASES "Analyze" soware. In the rst step, the thickness of the contamination carbon layer was determined by analyzing the Ti 2p region (and Ti LMM, but spectrum not included). Based on the results obtained with the "Analyze" program, the thickness of adventitious carbon layer was calculated as 1.2 nm for the as received sample (Fig. 7). However, it turned out that it was impossible to accurately estimate the thickness of the oxide layer. Because a good overlap of the energy loss cross-section with the extrinsic loss portion of the scans were obtained for all layers with modeled thicknesses above 10 nm. Even if contamination carbon was removed from the surface, the   thickness of the oxide layer was still higher than the max of ca. 10 nm (with Mg Ka excitation beam).

The effect anodic layer on the corrosion resistance of titanium
3.3.1. dc corrosion study in 0.05 M NaCl solutioninstantaneous corrosion rate. The effect of anodizing process in ChCit : oxalic acid eutectic solvent on the corrosion resistance of titanium was assessed by linear polarization resistance (LPR) and polarization curves. First the non-destructive LPR technique has been applied, that enabled estimating and monitoring the instantaneous corrosion rate very quickly. The resulting polarization resistance (R p ) as a function of exposure time of the samples in 0.05 mol dm À3 solution of NaCl is presented in Fig. 8.
From LPR method the values of corrosion current density (I corr ) were calculated using Stern-Geary equation (assuming b c and b a equal to 120 mV dec À1 ). The evolution of this parameter as a function of exposure time in 0.05 mol dm À3 solution of NaCl during exposure is presented in Fig. 9.
During the LPR measurements, attention was also paid to monitor the corrosion potential (E corr ) of the anodized titanium in NaCl solution (Fig. 10). Although it is not a parameter that determines corrosion resistance, it gives some information about the state of the surface of the tested material and its interaction with the corrosive environment.
Finally, aer 48 h of exposure of the samples in 0.05 mol dm À3 solution of NaCl, polarization curves have been registered in a potentiodynamic mode (Fig. 11). The curves obtained for the samples etched in HNO 3 and HF show that chemical etching activated the surface, which caused a noticeable increase in the corrosion current density value and a shi towards negative values (by about 60 mV) of a cathodic-anodic transition potential compared to this potential for of samples "as-delivered" (Fig. 11). Chemical etching affected also the course of anodic branch. The slope of this curve (Fig. 11) was denitely smaller than all other samples, which probably indicates worse barrier properties (possibly: a thin, leaky passive layer created on the etched titanium surface).
Potentiodynamic polarization measurements have clearly shown that subjecting titanium to polarization in EDS (ChCit : oxalic acid DES solvent) increases its corrosion resistance. In Fig. 11 it is well visible, that the corrosion current densities of titanium anodized at 10 and 30 V were lower than that of non-treated titanium. It is also worth emphasizing the fact that the current densities measured in the passive range at the anodic branch were also clearly smaller, especially for the anodizing voltage of 30 V (Fig. 11). That may suggest the presence of a dense/coarser oxide layer on the surface of this sample.
3.3.2. Electrochemical impedance spectroscopy (EIS) in 0.05 M NaCl solution. Impedance spectra have been collected aer 48 h of exposure of titanium samples in deaerated 0.05 mol dm À3 NaCl solution. They have been presented in the Nyquist (Fig. 12) and in the Bode (Fig. 13) representations to better justify the choice of the physical model for tting procedure.
A cursory analysis of the Nyquist plot ( Fig. 12) seemingly suggests a similar corrosion behavior of all titanium samples. Fig. 9 Corrosion current density (I corr ) values calculated from LPR method for titanium samples with anodic layer and the reference samples exposed for 48 h in 0.05 mol dm À3 solution of NaCl.  However, the spectra also indicates the existence of a time constant in the range of medium frequencies (gure inset in Fig. 12), especially in the impedance spectrum of titanium anodized at 30 V. Attention is drawn to the very high values of the imaginary part of impedance reaching values exceeding 1 MU cm 2 . It can be assumed that this is due to an oxide layer with very good barrier properties, which translates into high corrosion resistance of the material, i.e. in the Bode representation of impedance spectra (Fig. 13) we observe signicant increase of the impedance modulus at 0.01 Hz (jZj 0.01 Hz ) aer anodization. Another change, which is undoubtedly related to the formation of anodic oxide layer, was the appearance of two, although not clearly separated, phase angle (q) maxima at ca. 400 Hz and several Hz (Fig. 13) for sample anodized at 30 V.
Moreover, the phase angle (q) did not decrease even at the lowest frequencies (Fig. 13). This would conrm the existence of oxide layer barriers, because otherwise, i.e. if the titanium underwent active corrosion, the imaginary part of impedance would decrease (Z imag in Nyquist plot - Fig. 12) and we would observe a decrease in the value of jZj in this frequency range.
There are many electric equivalent circuits in the literature to model the corrosion resistance of titanium. The rst of them is the Randles-type circuit (model I and II in Fig. 14), which only takes into account the resistance of electrolytic solution and the polarization resistance connected in parallel with capacitance (alternatively modeled with a constant phase element -CPE). 32,33 The impedance of the CPE is dened by eqn (1), where: T is a time constant parameter (U À1 cm À2 s ÀP ), u is the angular frequency of the ac signal and P is the CPE exponent.
The second-type circuits are more complex (model III-V in Fig. 14). The unquestionable advantage of the two-time constant circuits is the ability to model the morphology of oxide lm. 33,34 J. Pan et al. 34 used model III (Fig. 14) to simulate a two-layer structure of titanium oxide lm which was composed of thin inner layer with excellent barrier properties and a relatively porous outer layer.
Fitting with one-time constant circuit from Fig. 14 gave satisfactory results but only for "as-delivered" and "etched" samples. The determined electrical quantities R s and CPE had very low residual errors (0.07-0.58%). Only for R the error was greater and amounted to 9.9-12.1%, which was related to the deterioration of the lowest frequency alignmentpossibly due to diffusion through the oxide layer. The physical meaning of the elements used in this model were as follows: R sthe resistance of electrolytic (0.05 mol dm À3 NaCl) solution, R and CPEthe resistance and capacitance describing both the properties of passive oxide layer and faradaic process. This is an obvious simplication which can be a good reference point when optimizing the physical model. Fitting results summarized in Table 2 clearly indicate a much lower corrosion resistance (three times lower resistance, R) of etched sample. Apart from this observation, it is difficult to infer about the oxide layer structure because the model used does not allow it.
Some improvement in the t was obtained aer extending the model I with an additional Warburg (W s ) element, connected in series with R (model II in Fig. 14), whose impedance Fig. 12 Nyquist plot of the impedance spectra recorded after 48 hours exposure in deaerated 0.05 mol dm À3 NaCl solution of titanium samples: after degreasing ("as-delivered"), degreasing and etching ("etched") and after additional anodic polarization at 10 and 30 V at 60 C for 10 min in DES anodizing bath. Fig. 13 Bode plot of the impedance spectra recorded after 48 hours exposure in deaerated 0.05 mol dm À3 NaCl solution of titanium samples: after degreasing ("as-delivered"), degreasing and etching ("etched") and after additional anodic polarization at 10 and 30 V at 60 C for 10 min in DES anodizing bath.
(Z W s ) can be represented by eqn (2), where: R reects diffusion impedance, T reects time of diffusion of the particle through layer thickness, u is the angular frequency of the ac signal and P is the W s exponent.
There are no major differences between the pseudocapacitances determined from both equivalent circuits (model I and II) - Table 2. However, signicant differences are visible in the case of resistance values (R). The use of the W s element meant that total resistance (R in Table 2 for model I) was distributed between the R and the W s -R parameter. Resistance, R, in model II decreased in the case of both samples by about an order of magnitude, while the W s -R parameter began to dominate and assumed values similar to those calculated for the resistance R in model I ( Table 2). In turn, the W s -P parameter reached values of 0.6-0.7, which, however, differ from P ¼ 0.5 for pure diffusion through a layer.
The attempt to use models I and II for anodized samples was unsuccessful. Also model III (Fig. 14) previously used by J. Pan et al. 34 did not apply here despite numerous tting attempts. Ultimately, the model IV proved to be successful in the case of titanium samples anodized at 10 V and 30 V (Fig. 14). The physical sense of the elements used in this model was as follows: R sthe resistance of electrolytic (0.05 mol dm À3 NaCl) solution, R out and CPE outthe resistance and capacitance of the outer porous oxide layer (electrolyte resistance inside this porous/less dense/less compact layer), R in and CPE inthe resistance and capacitance of the inner barrier oxide layer. According to Table 3 the outer porous layer, R out , resistance was close to 1.2 kU cm 2 and very similar for both anodized surfaces. However, the electrical (pseudo)capacitance of the outer layer, CPE out , was much smaller (probably in the order of a dozen mF cm À2 ) for the sample anodized at 30 V. Also the exponential parameter P of CPE out for that surface was slightly higher, which could also indicate its lower heterogeneity ( Table 3). The same observation applies to the pseudocapacitance of CPE in for "30 V" sample. The biggest difference, however, was in the   resistance of the inner layer, R in . While aer anodizing at 10 V the resistance of inner layer was ca. 28 MU cm 2 , then aer anodizing at 30 V, R in was calculated of 10 13 U cm 2 (!). Of course, it is unlikely, especially since error for this element was 10 7 %, as indicated in Table 3. This prompted us to modify the equivalent circuit for this sample. R in and CPE in elements were replaced with open Warburg (W o,in ) and pure capacitor (C in ), respectively model V in Fig. 14. Ultimately, a very good t was obtained (c 2 ¼ 1.5 Â 10 À5 ) and errors ca. 0.2-5.2%. The component values calculated from this ne model are shown in Table 4.

Conclusion
It has been shown that the technical grade titanium anodizing process can be successfully transferred to the alternative green solvent DES, composed of choline dihydrogencitrate salt-oxalic acid in 1 : 1 molar ratio.
The change in voltage in the DES titanium anodizing process affects the color of the formed oxide layer in a similar degree to that of conventional water-bath anodizing.
Aer anodizing at 10 V, the titanium surface shows a negrained structure. At 30 V, a coarser, rather uneven, oxide layer is formed, with signicantly larger grains.
The as-delivered titanium surface was covered with an oxide layer, the thickness of which was estimated to be ca. 2.2 nm (plus extra 1.2 nm of adventitious carbon in the toplayer). DES anodizing produces a homogeneous TiO 2 layer whose thickness at a potential of 10 V exceeds 10 nm.
As shown by XPS tests, the stoichiometric TiO 2 , regardless of the applied voltage in the anodizing process, was practically the only component of the oxide layer produced on titanium in the DES bath.
The potentiodynamic polarization technique showed that the passive layers produced in anodizing process of titanium in the DES bath exhibit at least one order of magnitude greater corrosion resistance compared to the layers present on titanium as delivered.
Electrochemical impedance spectroscopy (EIS) allowed to describe the layers formed in the anodizing process using a twolayer model. The barrier properties correspond to the inner layer formed directly on the surface of metallic titanium, the maximum resistance of which was 2.8 MU cm 2 . The resulting porous outer layer has a much lower resistance which ranges from 830 U cm 2 to 1300 U cm 2 . The native oxide layer on the Grade 2 titanium surface showed a polarization resistance of approx. 1 MU cm 2 . The choice of the best electrical equivalent circuit allowed us to conclude that the phenomenon of diffusion (mass transport through the passive layer) has a dominant share in this resistance. Table 4 Fitting results using model V for impedance spectra recorded after 48 h exposure in deaerated 0.05 mol dm À3 NaCl solution of titanium samples after anodization at 10 and 30 V at 60 C for 10 min in DES bath composed of choline dihydrogencitrate salt and oxalic acid Sample R s (U cm 2 ) CPE out -T (U À1 cm À2 s ÀP ) CPE out -P R out (U cm 2 ) C in (F cm À2 )