Enhancing PEM water electrolysis efficiency by reducing the extent of Ti gas diffusion layer passivation
Proton exchange membrane water electrolysis (PEM WE) suffers from several issues, such as the high cost and low stability of the electrolyser unit components. This is especially evident for an anode polarised to a high potential and in contact with an acidic membrane. Such a combination is detrimental to the vast majority of electron-conducting materials. Nowadays Ti (possessing a protective passive layer on its surface) is used as the construction material of an anode gas diffusion layer. Since the passivation layer itself is non-/semiconducting, an excessive degree of passivation leads to high surface contact resistance and to energy losses during PEM WE operation. This problem is usually solved by coating the Ti surface with precious metals. This leads to a further increase of the already very high cell investment costs. In this work an alternative method based on appropriate Ti etching (in acid) is presented. The (surface) composition of the samples treated was investigated using SEM, X-ray fluorescence and diffraction and photoelectron spectroscopy. TiHx was found in the subsurface layer. This was responsible for preventing excessive passivation of the Ti metal. The superior performance of the etched Ti gas diffusion layer (compared to non-etched) in a PEM water electrolyser was confirmed during an (> 100 h) experiment with current densities of up to 1 A cm− 2. Using the described treatment the surface contact resistance was substantially reduced and its increase during PEM WE operation was largely suppressed. As this method is very simple and cheap, it has tremendous potential for improving PEM WE process efficiency.
KeywordsPEM water electrolysis Titanium passivation Titanium hydride Photoelectron spectroscopy Surface contact resistance Etching
Water electrolysis with a proton exchange membrane (PEM WE) is an industrially and energetically interesting process, particularly due to its high intensity, flexibility and efficiency [1, 2]. On account of these properties it is an important part of the hydrogen economy scheme [3, 4]. However, PEM WE suffers from several issues, such as the high investment cost of the electrolyser unit and stability. Practically all PEM WE problems are related to the limited stability of the material and the high cost of the construction materials. This is especially pronounced on the anode side of the cell where the anode is polarised to a high potential and is in contact with an acidic (usually Nafion®-type) membrane . Such a combination is detrimental to the vast majority of electron-conducting materials which, under such conditions, are oxidised and consequently dissolved [5, 6]. For this reason, despite tremendous efforts by researchers in recent decades, practically the only useful type of anode catalyst material is based on Ir(oxide) which possesses both sufficient activity and stability. For the same reason, only a limited number of metals can be used for the construction of other anode components (e.g. gas diffusion layer, bipolar plates) . The most common is Ti (and Ta), a representative of “valve” metals which are characterised by a high degree of thermodynamic instability due to which, in the presence of a source of oxygen such as O2 or H2O, their whole surface is covered by a passivation oxide (hydroxide) layer. This passivation layer, on the other hand, is very stable, adherent and highly compact and it effectively protects the metal from further oxidation. As a result, the actual corrosion resistance of the valve metals (even in rather corrosive environments) is very high . This, together with excellent mechanical properties, low density and biocompatibility, makes Ti (and its alloys) an indispensable material in many highly valued fields such as aerospace, the chemical industry or (bio)medicine [8, 9].
As mentioned above, the excellent stability of Ti makes it a state-of-art construction material for several anode components in the PEM water electrolyser. The presence of a passivation layer, however, also has a negative effect on the surface contact resistance since the passivation layer itself is non-/semiconducting. The resistance in the film depends on its thickness as well as on its composition. The thickness of a Ti passivation layer can vary from 2 to 3 nm (naturally formed under ambient conditions) to tens or even hundreds of nm . Of course, the ohmic contact resistance increases with the thickness of the passivation layer. The effect of the composition of the passivation layer on its resistivity was discussed, for example, in . Passive layer is generally said to be formed by TiO2 (insulator). However, usually only near the top of the passivation film is the ratio O:Ti = 2 (not considering weakly bound or adsorbed O species as will be shown later). In its bulk, the passive layer contains sub- and non-stoichiometric oxides (O:Ti < 2) and acts as a n-type semiconductor . In other words, the oxygen content as well as the average oxidation state of Ti decreases in the direction towards the metal phase. It is only the layer of actual anhydrous TiO2 that functions as passivation protection. Thus, the rising thickness of the passivation layer does not necessarily help to improve its corrosion resistance . The ultimate result of excessive passivation layer formation (and increased surface contact resistance) in the PEM water electrolyser is the loss of energy efficiency. This becomes especially pronounced during intensive operation when current density can approach units of A cm−2. Under these conditions, energy losses due to ohmic resistance in the system dominate . A common way to solve the problem of excessive Ti passivation is to use the appropriate surface treatment which commonly involves coating Ti with protective layer of precious metal (usually Pt). This, of course, further increases the already very high investment costs of the device and the dependence of the PEM WE technology on precious metals. The surface treatment can be carried out by several methods. Precious metal ion implantation by magnetron sputtering  has, for example, been reported. Alternatively, alloying Ti with precious metal (even at the level of about 0.05%) has been shown to greatly improve Ti corrosion resistance . Finally, the same procedure of Ti surface coating by precious metals as that used in dimensionally stable anode (DSA) technology (known since the 1950s and perfected for chlor-alkali industrial applications in the 1960s) can be applied . Although the approach has been successful in DSA production there is a big difference between the precious metal consumption when coating a more or less flat DSA and a porous gas diffusion layer (GDL) for the PEM water electrolyser. It is fair to say that, in the PEM WE literature, problems related to the excessive passivation of Ti components are generally (almost intentionally) ignored.
At this point it is worth taking a more detailed look at how the Ti passivation issue is handled in DSA technology. The general procedure of DSA preparation has been described elsewhere . It usually involves mechanical treatment of the surface, such as sandblasting, followed by chemical etching, several coating steps and final thermal treatment. One of the most crucial steps seems to be the chemical etching. Its purpose is to roughen the surface and remove the existing passivation layer. Ideally, titanium hydride is supposed to remain on the surface as it should protect the underlying Ti metal for a few hours . The etching can, in principle, be performed in various acids. The behaviour of Ti in aqueous acid solutions has, for decades, been studied by numerous authors, cf. [18, 19, 20, 21, 22, 23, 24, 25]. The Ti behaviour is governed by the stability of the passive layer under the given conditions. In general, the rate of Ti dissolution increases with both the rising acid concentration and the solution temperature . The etching process can also be strongly influenced by varying the electrode potential. Basically in all environments, galvanostatic anodic polarisation of the Ti electrode leads to (re)passivation of the surface with the passivation potential being dependent on the electrolyte composition. The properties of the passivation layer, such as thickness, depend largely on the current density/electrode potential applied, the electrolyte composition and, of course, the duration of the polarisation . The stability of the oxide layer is greatly reduced by the presence of fluorides in an acidic environment , in fact hydrofluoric acid is practically the only acid that dissolves Ti even in diluted solutions [26, 27]. The rate of Ti dissolution in halogen hydride acid decreases with increasing atomic mass of the halogen .
Active metal dissolution is often connected with the absorption of atomic hydrogen in the Ti metal. When the concentration of hydrogen in the metal exceeds the solubility limit under the given conditions, Ti hydrides are formed. It has been reported that the rate/extent of hydride phase formation strongly depends on the exact metal composition, type of phase present, treatment of metal during its production and shaping, rate of diffusion of hydrogen in/through the passive layer (which itself is largely dependent on TiO2 crystal lattice orientation ), hydrogen partial pressure and temperature . Covington  summarised the three conditions which must exist simultaneously in order for serious hydriding to take place:
pH either low (< 3) or high (> 12);
temperature above 80 °C, otherwise only surface hydrides will be formed;
hydrogen present/generated in the system.
The first condition seems to be related to the fact that the passivation layer is very stable at medium pH values and its presence effectively prevents transport of molecular hydrogen towards the Ti  and most likely also its dissociation to atomic form . The second condition is given by a very low hydrogen diffusion coefficient in Ti at low temperatures . The thickness of the hydride layer increases with the duration of exposition to hydriding conditions, temperature and cathodic current density (upon polarisation in the active dissolution region) and can attain several tens of micrometres [34, 35]. An excessive hydridation of Ti leads to its embrittlement . Titanium hydrides are a group of binary metallic compounds with a maximum ratio of H/Ti = 2. Their thermal dehydrogenation starts far above 400 °C . Their oxidation in air proceeds in several steps starting at about 500 °C . Interestingly, many properties of the Ti hydrides, such as electron or thermal conductivity, are fairly similar to those of Ti .
Based on the above information, it seems that the Ti hydride layer possesses reasonable thermal and chemical stability, offering Ti metal a certain protection against oxidation. The aim of this work was to test the effect of an appropriate Ti metal surface treatment connected with hydride layer formation on the surface contact resistance and performance of the PEM water electrolyser.
2.1 Chemicals and materials
Freshly prepared deionised water (conductivity < 1 µS) produced by DIWA deionisation system (WATEK, Czech Republic) was used in all experiments. The following chemicals and materials were used in the experiments: 35% HCl (Lachner, Czech Republic), propan-2-ol (99.9%, Penta, Czech Republic), acetone (99.9%, Penta, Czech Republic), titanium felt (ST/TI/20/1050/77, Bekaert), 5% Nafion® solution in propan-2-ol (D521, Ion Power), Nafion® 117 membrane (Ion Power), 40% Pt on carbon (HiSPEC4000, Johnson-Matthey), IrO2 (Premion®, Alfa Aesar) and carbon felt GDL with microporous layer (SGL 38BC, Sigracet).
2.2 Ti treatment, testing and analysis
First, the Ti felt used in the experiments was degreased by washing with acetone and propan-2-ol. Afterwards, it was immersed in deionised water and ultrasonicated for 10 min in order to remove any mechanical impurities and fully flood its porous structure. After that it was placed into 35% HCl solution (preheated under reflux to 54 °C) for a period of 5 min. The maximum applicable temperature was limited by the boiling point of 35% HCl solution which is 55 °C. In order to be able to record open circuit potential of the felt during etching, the felt was contacted by Pt wire (not in contact with the solution) which is relatively resistant against HCl vapours. The etched felt was thoroughly rinsed with deionised water in order to remove reaction products and unreacted HCl. The rinsed Ti felt was left to dry and then used for further experiments.
For electrochemical experiments a HEKA PG310 potentiostat was used. A two-electrode arrangement (with working and reference electrodes) was used to measure the open circuit potential of the Ti metal. An aqueous silver chloride reference electrode saturated with KCl was used in all the experiments and all the potential values in this work refer to its potential (+ 0.197 V vs. standard hydrogen electrode). The reference electrode was placed in a capillary filled with saturated KCl solution and separated from the solution studied by a ceramic porous frit.
The morphology of the Ti felt was studied using scanning electron microscopy (SEM, Hitachi S4700, Japan). Samples were fixed to the Al holder using conducting double sided carbon tape. The phase composition of the felts was determined from X-ray diffraction (XRD) spectra measured by a PANalytical X’Pert PRO X-ray diffractometer (PANanalytical, Netherlands) using a nickel betafilter with Cu Kα radiation operated at 40 kV and 30 mA. The corresponding average wavelength was 1.5425 Å. XRD spectra were measured for angles between 5° [2Θ] and 90° [2Θ] with a step of 0.0394° and scan rate of 3.2° [2Θ] min−1. An Axios sequential X-ray fluorescence (XRF) spectrometer (PANanalysical, Holland) was used for qualitative and quantitative element analysis. It was equipped with a Rh anode end-window X-ray tube type 4GN fitted with a 75 µm Be window. The measurement time was about 20 min. X-ray photoelectron spectra (XPS) were measured using an ESCAProbeP (Omicron Nanotechnology Ltd, UK) electron spectrometer with a monochromatic Al Kα source (5.5 mA, 15 kV). The probed spot diameter was 0.8 mm. High-resolution spectra (step size of 0.05 eV) were collected using pass energy of 30 eV. The spectra were corrected for the surface charging using the line of adventitious carbon at 284.8 eV. For the deconvolution of the spectra a commercial KolXPD software (version 1.8.0) was used . A Shirley background was used during the spectra deconvolution. The ratio of Ti 2p3/2 and 2p1/2 line intensities was set 2:1. The peak of Ti0/II (with metallic character) was fitted using an (asymmetric) Doniach Sunjic function convoluted with a Gaussian, all other peaks were fitted using the Voigt function. The spin-orbit splitting values of 6.0 and 5.70 eV were determined for 2p lines of Ti0/II and (TiIII/TiIV), respectively, during deconvolution. These values correspond very well to those reported in the literature [41, 42].
2.3 PEM water electrolysis
The membrane electrode assembly was tested in a laboratory-scale single-cell PEM WE set-up. Titanium end-plates were gold plated and possessed parallel flow channels defining an active area of 4 cm2. Catalyst coated membranes (CCM) were prepared on using a CNC airbrush by first spraying an anode catalyst ink on one side and subsequently a cathode catalyst ink on the other side of a Nafion® 117 membrane. During the spraying, the membrane was placed on a bench heated to 130 °C. The anode ink was composed of IrO2, Nafion® solution, deionised water and propan-2-ol mixed in the ratio: 1.00 g:4.23 cm3:19.2 cm3:19.2 cm3. The cathode ink was prepared by mixing Pt/C catalyst, Nafion® solution and deionised water in the ratio: 1.00 g:6.00 cm3:30 cm3. The anode catalyst layer contained 85 wt% of IrO2 and 15 wt% of Nafion®. The total Ir (IrO2) loading was 0.75 (0.875) mg cm−2. The cathode catalyst layer contained 77% of Pt/C catalyst and 23% of Nafion®. The total Pt loading was 1 mg cm−2. Titanium felt and carbon felt with microporous layer were used as the anode and cathode gas diffusion layers, respectively. After the assembly the cell was constricted by a momentum of 8 N m. A long-term electrolysis at a temperature of 60 °C was performed using a Statron 3251.0 stabilised power source (Statron, Germany). The testing procedure cycle consisted of 1 h of operation at 1.8 V, after which a load curve was measured from 1.8 V down to 1.3 V and back to 1.8 V in 0.05 V steps (the voltage in the range of 1.8–1.5 V and 1.5–1.3 V was stepped in 5-min and 30-s intervals, respectively). For the measurement of the electrochemical impedance spectra the cell was disconnected from the power source and connected to a Solartron SI 1260 frequency response analyser in combination with a Solartron SI 1287 electrochemical interface (Solartron Analytical, United Kingdom). The impedance spectra were recorded (after a 5-min stabilisation of the performance at the given voltage) in 0.1 V increments, starting from a cell voltage of 1.4 V in the frequency range of 65 kHz–0.1 Hz (voltage amplitude of 0.020 V).
3 Results and discussion
3.1 Ti felt modification and analysis
A record of open circuit potential during etching of Ti felt in 35% HCl at 54 °C is shown in Fig. 1. The passive layer breakdown is visible as a sudden drop in electrode potential from around + 0.1 V to − 0.62 V (vs. Ag/AgCl) after approximately 80 s of etching. This indicates relatively weak stability/low thickness of the passivation layer of the felt. The breakdown was accompanied by the commencement of active metal dissolution (forming Ti3+ species, as documented by the purple/blue colour of the etching solution) and intensive H2 gas evolution. Ti3+ species were later (upon standing) oxidised by O2 [26, 43] (solution colour turned green). As is shown in the SEM pictures presented in Fig. 2, the morphology of the Ti surface changed dramatically after only 5 min of etching. While the surface of the pristine felt wires (Fig. 2A) was smooth, the surface of the etched sample (Fig. 2B) was graterlike (documenting the formation of pits). The average wire diameter in the pristine felt was about 20 µm. The etching duration of 5 min was chosen using a simple optimisation procedure (results are not shown here). The main objectives during the optimisation were to achieve a visible (by SEM) change in the surface morphology and not to reduce the diameter of the wires substantially. Using the series of the felts etched for varying periods of time, the approximate rate of the diameter decrease was estimated to be about 0.55 µm min−1. At the same time, numerous wires were seriously damaged after only 20 min of etching due to which the felt started to lose its integrity.
Further characterisation of the felt prior to and after etching was performed by XRF analysis. The results, from which several important conclusions can be drawn, are summarised in Table 1. Firstly, the pristine felt contains a certain amount of Y (> 0.65%) which is located near the surface of the individual wires. This can be concluded by comparing information obtained from the Lα and Kα lines. As the energy of the Lα line is lower (than that of the Kα line), the Lα radiation detected originates at a lower depth and the corresponding information is, therefore, more surface-sensitive. The content of other impurities is quite negligible. Secondly, most of the Y is removed during etching. This again confirms the accumulation of Y near the surface of the wires. The surface Y impurities most likely originate from the Ti felt production process. It is important to bear in mind that, due to the origin of the fluorescence radiation, it practically does not allow the detection of elements lighter than boron. Also oxygen and nitrogen are detected with difficulties. Additional information about the sample microstructure and composition was obtained by X-ray diffraction. The pristine sample diffraction spectrum contains peaks of Ti and Y2O3 whose content was semi-quantitatively estimated to be < 1%. After the etching, the Y2O3 phase diffraction peaks disappeared. Instead, the patterns of various Ti hydrides are present in the diffractogram. The total Ti hydride content was about 5%. The composition of the hydride phases in this particular case corresponded to TiH1.23 and TiH. However, phases such as TiH0.85 or TiH1.5 were also detected during other experiments (Fig. 3). The fact that the composition of Ti hydrides cannot be precisely controlled is in agreement with complicated nature of hydride formation . In any case, these results confirm the effective removal of the passivation layer and formation of a reasonable amount of hydride phase at the (sub)surface of the individual Ti felt wires. The potential effect of the presence of Y2O3 on hydride phase formation was not, at this stage, investigated.
Composition of the Ti felt before and after 5 min of etching in 35% HCl at 54 °C as determined from XRF measurements using Kα and Lα radiation
Further interesting information about the state of the Ti felt surface was obtained by XPS. Ti 2p XPS spectra (and their deconvolution) are presented in Fig. 4. In order to simplify the discussion of the Ti 2p spectra shown in Fig. 4, only 2p3/2 lines are discussed here. The positions of the 2p1/2 lines can easily be calculated from the spin–orbit splitting values given in the Sect. 2.2. The Ti 2p spectrum of pristine Ti felt shown in Fig. 4A is dominated by the 2p3/2 line of TiIV (at 458.6 eV). The lines attributable to the Ti species in the lower oxidation states are barely visible, which corresponds to the predominance of TiIV in the spectrum of TiO2 in the passive layer . Etching the felt changes its surface composition dramatically. This is documented by the spectrum shown in Fig. 4B, where spectral lines of Ti0/II and TiIII species (at 454.3 and 456.2 eV) become significant. Lamartine et al. reported that the energy of the Ti 2p3/2 line of Ti hydride increases linearly (compared to the Ti metal line at 454.2 eV) with a rising amount of hydrogen (slope of about + 0.3 eV per unit of H/Ti atomic ratio) . Therefore, taking into account some degree of uncertainty in the metallic Ti peak position, it is difficult to distinguish between the metal Ti and low-hydrogen content Ti hydride species. In order to account for this uncertainty, we assign these species to Ti0/II, which relates to the potential span of the oxidation states in the range of 0 to + II. At the same time, the intensity of the spectral line due to TiIV is rather low. In order to estimate the stability of the Ti hydride layer under ambient conditions, an additional spectrum of the etched Ti felt sample aged for 7 months (air, atmospheric pressure, room temperature) was measured. This is shown in Fig. 4C. It is apparent from the intensity of individual lines that the aged surface is somewhat more oxidised compared to the freshly etched sample, but its surface is still oxidised to a lesser extent than that of the pristine felt.
The compositions of the Ti surfaces (with regard to Ti species) are summarised in the bar plot in Fig. 5. It was obtained by comparing the integral areas of the individual considered Ti species observed and determined by deconvolution of the Ti 2p spectral lines, as discussed previously in the context of Fig. 4. Before etching, 84% of the Ti species in the surface layer is present in the highest oxidation state, i.e. IV. Only 7 and 9% of the Ti exists in oxidation states 0/II and III, respectively. Contrary to this, the etched surface contains only 15% of TiIV, while the concentration of Ti0/II (41%) and TiIII (44%) is quite high. The composition after aging (20% Ti0/II, 21% TiIII and 59% of TiIV) lies somewhere between the two extreme cases. It is interesting to note that the atomic ratio of Ti0/II and TiIII is very close to unity in all cases.
Oxygen (O 1s) XPS spectra are presented in Fig. 6. The spectra of the pristine and aged samples are very similar and are dominated by the line at 530.1 eV documenting the presence of O in the lattice of TiO2 on the very surface. Less intensive lines at higher binding energy are related to more weakly bound oxygen species, such as adsorbed OH (531.7 eV) and adsorbed water (533.5 eV) . On the other hand, the highest photoelectron intensity, in the case of the etched sample, was observed for a binding energy of 530.9 eV. This can be partially explained by the lower oxidation state of the surface layer Ti species leading to a weaker bond between O and Ti. What is not visible from the O 1s spectra normalised to unit area is the differences in the overall O content in the surface layer. This becomes apparent when O 1s and Ti 2p integral signals are compared. Using the atomic sensitivity factors (1.798 for Ti 2p and 0.711 for O 1s ) the atomic ratio of O/Ti can be estimated. The estimated values for the pristine, etched and (etched +) aged sample are 3.3, 1.1 and 2.2, respectively. The value of 3.3 exceeds the expected O stoichiometric coefficient (in TiO2) of 2. However, this is reasonable as part of the O is present in the form of adsorbed OH and H2O, as discussed above. The low content of O (1/3 of the O in the pristine sample) in the etched sample is in agreement with the removal of the existing passivation layer and the formation of a new thin passivation layer on top of the Ti hydrides. Finally, ageing (for 7 months, air, room temperature) of the etched sample led to doubling of the O (in the form of O2−) content (corresponding to 2/3 of the O amount in the pristine sample).
When thinking about the change in the composition of the passivation layer in radial direction, it is first useful to define the thickness of the layer probed by XPS. The mean free path of Ti 2p (and similarly O 1s) photoelectrons in Ti and TiO2 lies within the range of 1.6–2.0 nm, therefore most of the information comes from the upper 5–6 nm . Secondly, it is reasonable to assume that the majority of the TiIV is located on the very surface of the material and its content decreases when entering the material bulk. Deeper in the material, the passive layer contains more TiIII and TiII. Even deeper, a Ti hydride subsurface layer is present (between the thin passive oxide layer and the bulk of the Ti metal). It is likely that all the interfaces are fairly diffuse. This overall image is also in agreement with previous conclusions of Lu et al. . They performed angle resolved XPS measurements of the Ti passivation layer and found out that (besides TiIV) TiIII ions are present in the passivation layer towards the side in contact with the surrounding environment, while TiII ions are located at the interface of the passivation layer and the metallic phase. In general, the concentration and distribution of the individual Ti ion types in the layer is important since it greatly influences the resistance of the layer, which has great practical implications for the surface contact resistance value.
3.2 PEM water electrolysis tests
The performance of the etched Ti felt was tested in a PEM water electrolyser. The results are summarised in Fig. 7. The load curves recorded at various times during PEM WE operation presented in Fig. 7A document an important differences between the performance of the etched and the pristine (not etched) Ti felt. From the very beginning of the electrolysis, the current densities obtained using the electrolyser with etched Ti felt were almost two times higher in the whole range of cell voltages applied. Moreover, the difference in performance increased with time. The performance of the cell with the etched felt did not change throughout the 108 h-long operation. Actually, the performance even slightly improved, but the improvement is most likely related to some phenomena unrelated to passivation, such as catalyst activation. On the other hand, the current densities substantially dropped in the cell containing unmodified felt. Such behaviour can be explained by taking the ohmic resistance of the system into account. In order to get more information about the system behaviour, the electrochemical impedance spectra of the PEM water electrolyser cell were measured in the range from 1.4 V to 1.6–1.8 V in 0.1 V increments. The upper voltage range was given by the maximum current range (2 A) allowed by the used apparatus. Examples of the electrochemical impedance spectra obtained in the case of PEM water electrolyser with the pristine and the etched Ti felts at cell voltage of 1.6 V are shown in Fig. 7B. Firstly, in a high frequency region, there is clear dominating effect of inductance which is present due to induction in the massive electrolyser body (this feature is present also if the cell is assembled without Nafion® membrane and catalytic layers on the gas diffusion layers). Secondly, there is large difference in the ohmic resistance between the pristine and the etched Ti felt. A lower ohmic resistance in the case of the etched Ti felt leads to lower ohmic losses in the system and consequently a lower potential on the cathode and, what is most important, a higher anode potential. This results in the increased rate of electrode reactions, which is in accordance with a lower activation resistance (documented by smaller “diameter” of the semicircle in the Nyquist diagram) of the etched felt. The development of ohmic resistance values determined by electrochemical impedance spectroscopy is shown in Fig. 7B. A slightly apparent scattering of the values is given by the fact that the impedance spectra were each time measured at several cell voltages and the ohmic resistance slightly increased with cell voltage. The ohmic resistance of the cell with etched felt was fairly low and remained practically constant (changes from 0.23 Ω cm2 at the time of 0.5 h to 0.24 Ω cm2 at the time of 103 h). This corresponds to the “degradation rate” of 1 × 10−4 Ω cm2 h−1. At the beginning of electrolysis (time of 6 h), the ohmic resistance of the cell with pristine Ti felt already reached a value of 0.46 Ω cm2 and then dramatically increased (to about 0.64 Ω cm2 after only 30 h of electrolysis). Using the whole duration of the experiment of about 80 h for the degradation rate calculation, the degradation rate can be estimated to be at least 23 × 10−4 Ω cm2 h−1. When comparing these values, the difference of about 0.2 Ω cm2 at the beginning of the electrolysis can be assigned to the effect of additional surface contact resistance of the original passivation layer. An additional 0.18 Ω cm2 increase can be attributed to a further consolidation of the passivation layer of the pristine felt during approximately 80 h of operation.
After the electrolysis, the cell was disassembled and the etched Ti felt was analysed by XRD and XPS. The XDR spectrum shown in Fig. 3A reveals the presence of Ti hydrides (mainly TiH1.5 and TiH) in the felt even after more than 100 h of PEM water electrolyser operation (cell voltage of up to 1.8 V, 60 °C). This shows reasonable stability of the Ti hydrides under the conditions of PEM WE and explains the stable performance of the electrolyser containing the etched Ti felt. A different sample of etched felt, whose spectra is not presented in Fig. 8, was used in electrolysis cell, therefore, the hydride phase composition differs slightly (TiH1.5 phase is present instead of TiH1.23). The deconvoluted XPS spectrum of the felt after electrolysis (Fig. 3B) documents the fact that the probed 5–6 nm surface layer is largely oxidised (6% Ti0/II, 5% TiIII and 89% of TiIV). Thus, the hydrides detected by the XRD must be present deeper under the surface of the Ti felt wires. Admittedly, in this case, the precision of the deconvolution was impaired by the low intensity of the signal which was most likely caused by contamination of the surface with traces of Nafion® and IrO2 from the anode side of the CCM. The fact that the surface contact resistance was still much smaller than in the case of the pristine felt (before and after electrolysis) suggests either the presence of a thicker/compact passive layer in the latter case or a difference in the microstructure or composition of the passive layer. We are currently performing further tests in order to evaluate the stability of the etched Ti felt over a longer timescale.
The results have shown that the above described surface modification of Ti by etching (in hydrochloric acid) represents a highly interesting procedure for increasing Ti resistance to excessive passivation which leads to a substantial increase in the surface contact resistance. The superior performance of the etched Ti gas diffusion layer in a PEM water electrolyser was confirmed during a (> 100 h) experiment using current densities of up to 1 A cm− 2. The effect was ascribed to the protective function of a Ti hydride underlayer formed during etching in acid. It was found that, by using a treatment of this type, the surface contact resistance can not only be directly reduced, but its increase during PEM WE operation can be largely supressed. The potential of the described surface modification becomes even more apparent in view of the fact that currently the only alternative is coating the Ti surface with expensive precious metals such as Pt. It is obvious that the simple surface modification procedure described has huge energy savings potential for PEM WE operation.
Financial support of this work by the Grant Agency of the Czech Republic within the framework of Project No. 15-02407J and by the Deutsche Forschungsgemeinschaft, Grant no. HA6841/2-1 and no. SU189/7-1, is gratefully acknowledged. Part of the material characterisation experiments was performed utilising instrumentation financed by the Operational Programme Prague—Competitiveness (CZ.2.16/3.1.00/24501) and the “National Program of Sustainability” (NPU I LO1613) MSMT-43760/2015.
- 4.Bertuccioli L, Chan A, Hart D, Lehner F, Madden B, Standen E (2014) Development of water electrolysis in the European Union. Fuel cells and hydrogen Joint undertaking, LausanneGoogle Scholar
- 6.Pourbaix M (1974) Atlas of electrochemical equilibria in aqueous solutions, 2nd edn., National Association of Corrosion Engineers, HoustonGoogle Scholar
- 7.Draley JE (1979) Corrosion of valve metals. In: Corrosion chemistry, ACS Symposium Series. American Chemical Society, vol 89, pp 185–234. https://doi.org/10.1021/bk-1979-0089.ch007
- 8.Veiga C, Davim J, Loureiro A (2012) Properties and applications of titanium alloys: a brief review. Rev Adv Mater Sci 32:133–148Google Scholar
- 14.Fateev V, Blach P, Grigoriev S, Kalinnikov A, Porembskiy V (2014) High pressure PEM electrolyzers and their application for renewable energy systems. European Hydrogen Energy Conference, Seville, Spain, 12–14 March 2014. http://www.ehec.info/images/EHEC2014/EHEC2014_Program.pdf
- 24.Kong D-S, Feng Y-Y (2009) Electrochemical anodic dissolution kinetics of titanium in fluoride-containing perchloric acid solutions at open-circuit potentials. J Electrochem Soc 156:C283–C291Google Scholar
- 32.Wille GW, Davis JW (1981) Hydrogen in titanium alloys. Department of EnergyGoogle Scholar
- 40.http://www.kolibrik.net/science/kolxpd/. Accessed 1 Oct 2017
- 46.Moulder JF, Stickle WF, Sobol PE, Bomben KD (1992) Handbook of X-ray photoelectron spectroscopy, A Reference Book of Standard Spectra for Identification and Interpretation of XPS Data. Perkin-Elmer Corporation, Eden Prairie, MinnesotaGoogle Scholar