Cobalt-Doped Black TiO2 Nanotube Array as a Stable Anode for Oxygen Evolution and Electrochemical Wastewater Treatment

  • Yang Yang
    Yang Yang
    Division of Engineering and Applied Science, Linde-Robinson Laboratory, California Institute of Technology, Pasadena, California 91125, United States
    More by Yang Yang
  • Li Cheng Kao
    Li Cheng Kao
    Department of Geosciences, National Taiwan University, P.O. Box 13-318, Taipei 106, Taiwan
    More by Li Cheng Kao
  • Yuanyue Liu
    Yuanyue Liu
    Department of Mechanical Engineering and Texas Materials Institute, University of Texas at Austin, Austin, Texas 78712, United States
    More by Yuanyue Liu
  • Ke Sun
    Ke Sun
    Divisions of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, United States
    More by Ke Sun
  • Hongtao Yu
    Hongtao Yu
    School of Environmental Science and Technology, Dalian University of Technology, Dalian 116024, People’s Republic of China
    More by Hongtao Yu
  • Jinghua Guo
    Jinghua Guo
    Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States
    Department of Chemistry and Biochemistry, University of California, Santa Cruz, California 95064, United States
    More by Jinghua Guo
  • Sofia Ya Hsuan Liou
    Sofia Ya Hsuan Liou
    Department of Geosciences, National Taiwan University, P.O. Box 13-318, Taipei 106, Taiwan
  • , and 
  • Michael R. Hoffmann*
    Michael R. Hoffmann
    Division of Engineering and Applied Science, Linde-Robinson Laboratory, California Institute of Technology, Pasadena, California 91125, United States
    *E-mail for M.R.H.: [email protected]
Cite this: ACS Catal. 2018, 8, 5, 4278–4287
Publication Date (Web):April 10, 2018
Copyright © 2018 American Chemical Society
ACS AuthorChoicewith CC-BYlicense
Article Views
PDF (3 MB)
Supporting Info (1)»


TiO2 has long been recognized as a stable and reusable photocatalyst for water splitting and pollution control. However, it is an inefficient anode material in the absence of photoactivation due to its low electron conductivity. To overcome this limitation, a series of conductive TiO2 nanotube array electrodes have been developed. Even though nanotube arrays are effective for electrochemical oxidation initially, deactivation is often observed within a few hours. To overcome the problem of deactivation, we have synthesized cobalt-doped Black-TiO2 nanotube array (Co-Black NTA) electrodes that are stable for more than 200 h of continuous operation in a NaClO4 electrolyte at 10 mA cm–2. Using X-ray photoelectron spectroscopy, X-ray absorption spectroscopy, electron paramagnetic resonance spectroscopy, and DFT simulations, we are able to show that bulk oxygen vacancies (Ov) are the primary source of the enhanced conductivity of Co-Black. Cobalt doping both creates and stabilizes surficial oxygen vacancies, Ov, and thus prevents surface passivation. The Co-Black electrodes outperform dimensionally stable IrO2 anodes (DSA) in the electrolytic oxidation of organic-rich wastewater. Increasing the loading of Co leads to the formation of a CoOx film on top of Co-Black electrode. The CoOx/Co-Black composite electrode was found to have a lower OER overpotential (352 mV) in comparison to a DSA IrO2 (434 mV) electrode and a stability that is greater than 200 h in a 1.0 M KOH electrolyte at a current density of 10 mA cm–2.


Jump To

TiO2 has long been recognized as a stable and reusable photocatalyst for water splitting and pollution. (1) However, potential applications of TiO2 are most often correlated with its oxygen evolution reaction (OER) potential. (1,2) In spite of success in laboratory-scale research, many challenges remain with respect to the use of TiO2 in solar fuel production and environmental remediation due to (1) its high band-gap energy of 3.2 eV (i.e., 385 nm) and (2) high attenuation coefficients for the penetration of incident UV photons in turbid water (i.e., both freshwater and wastewater). Some of these intrinsic barriers can be overcome when TiO2 is used in electrochemical systems applications without the need for direct photoactivation of TiO2. (3,4) However, the low mobile carrier conductivity of n-type TiO2 impedes its use as an electrocatalyst. The transport of electrons across the electrolyte/TiO2 interface requires a large anodic potential (>3 VAg/AgCl) to overcome a high Schottky barrier. (4)
The electronic properties of TiO2 can be tuned by self-doping. It is known that H2 reduction could introduce Ti3+, oxygen vacancies, and surface disorders to TiO2. The resultant Black TiO2 has a narrower band gap. (5−7) Black TiO2 catalysts on both anatase and rutile structures exhibited high activity in visible-light-driven water oxidation. (8,9) However, their activity during dark electrolysis has been less explored. Reducing TiO2 in pure H2 at temperatures over 1000 °C promotes the bulk phase transition from anatase to substoichiometric Magnéli phase Ti4O7. (10) Magnéli phase TiO2 has conductivities that approach 11 orders of magnitude greater than that of TiO2. (11) Thus, it has been shown to be a stable anode material for water oxidation and wastewater treatment. (12,13)
Given that the preparation of Ti4O7 requires highly reducing conditions, it is often difficult to prepare conductive, stoichiometric TiO2 using conventional techniques. Recently, conductive TiO2 nanotube arrays (NTAs) that are either blue or black in appearance have been reported. Black NTAs were prepared by reducing anatase NTA under an H2 or H2/Ar atmosphere at 450–550 °C. (14,15) Blue NTAs were obtained by cathodization of anatase NTA in aqueous electrolyte. (16−20) Conductive NTA electrodes have satisfactory electrochemical oxidation activity. (20,21) The synthetic preparation procedures for Blue and Black NTA are less stringent than those of the Magnéli phases, TixO2x–1.
Conductive NTAs supported on titanium plates have a major advantage over particulate electrocatalysts, since they can be utilized directly as electrodes without the need for additional adhesive substrates or organic binders. However, deactivation of both Blue and Black NTAs has been observed after a few hours of electrocatalysis due to the surface passivation. (20,21) Therefore, electrodes made from these materials are currently impractical for engineering applications.
In this study, we report on the effect of doping trace amounts of cobalt onto Black TiO2 NTA (Co-Black NTA) that results in a lowering of the OER overpotential and increases electrode stability. Even though bulk CoOx is reported to be unstable for the OER at pH <12 in phosphate-free electrolyte solutions, (22,23) we observe that CoOx is immobilized and stable on Black NTAs even at circumneutral pH. We show that Co-Black electrodes outperform an IrO2-based dimensionally stable anode (DSA) for oxidative electrochemical wastewater treatment. We also found that the increase of Co loading forms a CoOx film on top of Co-Black substrate. The resultant CoOx/Co-Black composite electrode exhibits high OER activity (overpotential of 352 mV vs 434 mV for IrO2 DSA) and stability (>200 h) in 1 M KOH electrolyte at 10 mA cm–2.

Experimental Section

Jump To

Electrode Preparation

The overall approach to synthesize NTA electrodes is schematically illustrated in Figure 1a. Amorphous TiO2 NTA (Am-NTA) was prepared by anodization of a Ti plate (6 cm2, 0.5 mm) at 42 V in ethylene glycol (EG) electrolyte with 0.25 wt % NH4F and 2 wt % H2O for 6 h. (24) After anodization, Am-NTA was subjected to a second anodization in 5 wt % H3PO4/EG electrolyte at 42 V for 1 h to enhance its mechanical stability. (20,25)

Figure 1

Figure 1. (a) NTA electrode preparation procedures. (b, c) Cyclic voltammograms of NTA electrodes in 100 mM KPi buffer at pH 7.2.

Cobalt-loaded Am-NTA (Co/Am-NTA) was prepared by dipping Am-NTA into 250 mM Co(NO3)2/ethanol solution. The sample was dipped into the coating solution for 1 min, pulled up at the rate of 10 mm/min, and finally dried at room temperature for 2 min. The dip-coating processes were repeated three times. The Co loading was determined by ICP-MS (Agilent 8800) as 0.54 ± 0.12 μmol/cm2. Lowering the Co(NO3)2 concentration in the dip-coating solutions to 50 and 25 mM produced Co(0.25)- and Co(0.17)-Black NTA with Co loadings of 0.25 and 0.17 μmol/cm2, respectively.
Annealing Am-NTA and Co/Am-NTA in air at 450 °C for 1 h yieldeds NTA and Co-NTA, respectively. Black NTA and Co-doped Black NTA (Co-Black) were obtained by annealing Am-NTA and Co/Am-NTA in a stream of 5% H2/Ar at 450 °C for 30 min and then naturally cooling to room temperature. The Blue NTA was prepared by applying a cathodic current of 5 mA cm–2 on the NTA electrode for 10 min in 0.1 M potassium phosphate buffer solution (KPi). All of the thermally treated NTAs were determined to be in the anatase phase by XRD (Figure S1). Five control samples were prepared. (1) A cobalt-doped TiO2 (Co-TiO2) film electrode was prepared by spray-coating a mixture of 250 mM titanium–glycolate complex and 25 mM Co(NO3)2 onto a Ti plate, followed by annealing in air at 450 °C for 1 h. (26) The final mass loading of Co-TiO2 was 6 mg/cm2. The molar Co loading was 0.6 μmol/cm2. (2) Amorphous cobalt hydroxide was loaded onto a Ti plate (Co(OH)x/Ti) by the electrochemical deposition method described previously, (27) with the modification of replacing the glassy-carbon substrate with a titanium plate. (3) Cobalt oxide was coated onto the Ti plate (CoOx/Ti) by a drop-casting method using 250 mM Co(NO3) in ethanol as precursor, followed by annealing under 5% H2/Ar at 450 °C for 30 min. (4) IrO2 DSA anodes were prepared by spray-coating an IrCl3/isopropyl alcohol solution onto a hot Ti plate (6 cm2, 0.5 mm) at 300 °C, followed by annealing at 450 °C for 1 h. (26) The mass loading of IrO2 was 0.5 mg cm–2. (5) Commercial IrO2 DSA (C-DSA) was purchased from Nanopac, Korea.

Catalyst Characterization

Cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were measured using a Biologic VSP-300 potentiostat. Uncompensated resistance (Ru) was measured by EIS. All of the presented anodic potentials were corrected by Ru. EIS was also used to determine Mott–Schottky plots. These details are described in Text S2 of the Supporting Information. For two-point solid-state measurements, nanotube tops were contacted by sputter-evaporated 10 nm thick Au using a Cressington 208HR sputter coater. Then resistance was obtained from the IV curves collected by potentiostat.
After CV analyses, NTA electrodes were characterized by scanning electron microscopy (SEM, ZEISS 1550VP) coupled with energy dispersive X-ray spectroscopy (EDS), scanning transmission electron microscopy (STEM, FEI TF30ST) equipped with a high angle annular dark field (HAADF) detector, X-ray photoelectron spectroscopy (XPS) with Surface Science M-Probe ESCA/XPS, and X-ray diffractometry (PANalytical X’Pert Pro). Electron paramagnetic resonance spectra were collected on a Bruker EMX X-band CW-EPR spectrometer at room temperature and 10 K. Powder samples were collected by scraping NTA from the Ti metal substrate.

X-ray Absorption Spectroscopy

X-ray absorption spectroscopy (XAS) of Ti L-edge and O K-edge were measured on Beamline 8.0.1 at the Advanced Light Source (ALS), Lawrence Berkeley National Laboratory. The resolutions of the measurements were 0.4 eV at the O K-edge and 0.3 eV at the Ti L-edge. Spectra were recorded in total electron yield mode (TEY) and were obtained by measurement of the sample drain photocurrent under irradiation with monochromatic light. The XAS at Co K-edge was conducted at BL17C1 at the National Synchrotron Radiation Research Center, Taiwan. The photon energy calibrations of the XAS spectra at the Ti L-edge and O K-edge were conducted on the basis of the reference anatase TiO2 films. The incident radiation flux was monitored by the photocurrent produced in a gold mesh in the beam path. All spectra were normalized to the incident flux recorded with Au mesh.

Cobalt Leaching and Anodic Stability Testing

The Co-Black NTA electrodes were subjected to cathodization (−5 mA/cm2, 10 min), CV (−0.5 to +3 VRHE, 10 mV s–1), and four continuous stages of constant current electrolysis (10 mA cm–2) at an interval of 3 h. At the end of each step, a water sample was collected, acidified by HNO3, and then analyzed by ICP-MS (Agilent 8800) to determine the Co concentration. The electrolyte was replaced before the next test. The anodic stability test was performed following the constant current Co leaching test. Current density was maintained at 10 mA cm–2. Electrolyte was replaced every 12 h.

OER, CER, and Water Treatment Tests

OER activities were evaluated using Tafel plots, which were collected by potentiostat with a three-electrode configuration. Stainless steel and saturated Ag/AgCl were used as counter and reference electrodes, respectively. Data were collected at constant current mode (0.01–10 mA cm–2). Each current step was maintained for 5 min to measure the steady-state anodic potential.
Hydroxyl radical production and direct electron transfer activities were measured using benzoic acid and oxalate as probe molecules, respectively. Benzoic acid was measured by HPLC (Agilent 1100) equipped with a Zorbax XDB column. Oxalate was analyzed by ion chromatography (ICS 2000, Dionex). Free chlorine concentrations ([FC]) were measured using the DPD (N,N-diethyl-p-phenylenediamine) reagent (Hach method 10102). The current efficiency was estimated using the equation(1)where V is electrolyte volume (25 mL), F is the Faraday constant (96485 C mol–1), and I is the current (A). Chemical oxygen demand (COD) in wastewater was determined by the dichromate digestion method (Hach Method 8000). NH4+ was quantified by ion chromatography.
Latrine wastewater was collected from a recycling electrochemical toilet system located on the Caltech campus (Pasadena, CA). The latrine wastewater has a pH of around 8.5 with a conductivity of 13.5 mS cm–1. It contains 500 mg L–1 COD and 18.5 mM NH4+. During the electrolysis of latrine wastewater, the Co-Black and IrO2 electrodes were operated at 10 mA cm–2 with cell voltages of 4.6 and 4 V, respectively. The electrode area/wastewater volume ratio was set as 60 m–1.

Results and Discussion

Jump To

The electrochemical activity of electrodes was determined by cyclic voltammetry (CV) in potassium dihydrogen phosphate (KPi) buffer solutions. The anodic branch current response is due to the OER. Both NTA and Co-NTA have low OER activity, as indicated by the low current densities of Figure 1b. The Co-TiO2 film electrode has higher current response in comparison to NTA and Co-NTA, but its performance is inferior to that of conductive NTA electrodes (Figure 1b vs Figure 1c). An increase in current response is observed on both the Blue and Black NTA electrodes at onset potentials of 2.7 and 1.7 V, respectively (Figure 1c). Co-Black NTA has the highest OER activity due to its lower onset potential and its higher current density. There have been previous reports that Co-doped TiO2 nanotubes could be used as photoelectrochemical water-splitting catalysts. (28−31) However, their photocurrent densities are less than 5 mA/cm2. In a recent report, Co3O4/TiO2 nanotubes were used for dark water electrolysis. (32) However, it appeared that TiO2 functioned only as an inert substrate for Co3O4 with resulting current densities <1 mA/cm2 at an applied potential of 2.3 VRHE. These results are in agreement with the inert performance observed on a Co-TiO2 film electrode with 0.6 μmol/cm2 Co loading prepared in this study. In contrast, Co-Black NTA with less Co loading (0.54 μmol/cm2) exhibits 200 times higher current density in comparison with a Co-TiO2 film electrode at 2.3 VRHE. This finding highlights the promotional role of conductive NTA substrate on the OER activity of Co-TiO2 system.
The anodic stability of the NTA electrodes at circumneutral pH was determined by applying a constant current of 10 mA cm–2 in a KClO4 electrolyte solution. Deactivation of the electrode was determined from the sharp increase in anodic potential (Figure 2a). The Blue and Black NTAs have lifetimes of 2.3 and 4 h, respectively, which are in line with previous studies. (20,21) A substantially longer lifetime and apparent stability (>200 h) were observed for the Co-Black NTA electrode (Figure 2b). Lowering the Co loading to 0.25 and 0.17 μmol/cm2 does not affect the stability of Co-Black NTA within the investigated 100 h electrolysis (Figure S2a). The results suggest that a trace amount of the Co dopant is sufficient to significantly enhance the stability of Black NTA.

Figure 2

Figure 2. (a, b) Comparison of the anodic stability of the NTA electrodes in 30 mM KClO4 at a current density of 10 mA cm–2. (c) Effect of the composition of the electrolytes on anodic stability (insert: comparison of stability of Co-Black NTA and Black NTA in 0.5 M H2SO4). (d) Effect of electrolyte composition on cobalt leaching from the NTA electrode material.

Figure 2c shows that the Co-Black electrode is also stable in 1.0 M KOH. However, deactivation occurs after 45 h in 0.5 M H2SO4. Nevertheless, Co-Black electrode still has a longer lifetime than Black NTA (insert of Figure 2c). We further investigated the leaching of Co in different electrolytes at cathodic (−5 mA cm–2), fluctuating (CV), and anodic (10 mA cm–2) currents (Figure 2d). Complete dissolution of Co from Co-Black was observed after sequential cathodization, CV, and 12 h electrolysis in H2SO4. In contrast, only 2.1% of Co was lost under similar conditions in the KOH electrolyte. Retarded Co leaching is also observed in the 0.1 M KPi electrolyte; this is probably due to the formation of less soluble CoPi. (33,34) In addition, the leaching of Co from Co-Black is undetectable (<0.03%) by ICP-MS after 9 h of electrolysis in KClO4 in the absence of phosphate. Analyses by SEM-EDS and STEM-HAADF indicate that Co is well dispersed in Co-rich spots on Co-Black throughout the tube length. CoOx particles were not found (Figures S3–S6).
Two-point solid-state conductivity measurements show that the resistances of Blue, Black, and Co-Black NTA electrodes are 5 orders of magnitude lower than that of pristine NTA electrodes (Figure S7). Knowing that the above measurements may be interfered by the contact resistance of Au coating, we further probed the semimetallic properties of conductive NTA by EIS using liquid electrolyte contacts (Figure S8). The Blue, Black, and Co-Black NTA all have good conductivity, but their OER activities vary. This implies that, in addition to bulk conductivity considerations, surface specific characteristics need to be considered.
Both electrochemical and thermal reduction should promote the formation of Ti3+ sites and adjacent oxygen vacancies (Ovs) on TiO2. However, aside from Ti4+, X-ray photoelectron spectroscopy (XPS) cannot confirm the existence of Ti3+ on Blue and Black NTA, probably due to the reoxidation of Ti3+ to Ti4+ in ambient air (Figure S9). Nevertheless, the oxidation of Ti3+ is essentially the filling of Ov by adsorbed oxygen species, which can be detected by XPS (Figure S10). The relative concentration of absorbed oxygen species is associated with the abundance of nascent Ov. The XPS O 1s spectrum reveals that the concentration of oxygen vacancies (Ov) decreases in the order Black (22%) > Blue (17%) > pristine NTA (12%). XPS analyses also found that Co doping results in a significant reduction in the Ti oxidation state in Co-Black NTA and creates more surficial Ov (25%) (Figures S9 and S10).
The NTA samples were further investigated using X-ray absorption spectroscopy (XAS) operated in total electron yield (TEY) mode. The results obtained from TEY mode involved both surface and subsurface characteristics, as XAS has a larger sampling depth in comparison to XPS. The Ti L-edge XAS is shown in Figure 3a. The spectra that result from the transition of electrons from the Ti 2p3/2 and 2p1/2 initial states to the unoccupied 3d orbital. Peaks A–C correspond to the electronic transition from 2p3/2 to 3d, while peaks D and E can be assigned to the transition from 2p1/2 to 3d. (35) The single crystal of TiO2 has octahedral (Oh) symmetry. The Oh crystal field splits the Ti 3d band into t2g and eg degenerate orbitals. The excitation of electrons from the ground state of Ti4+ to t2g orbital can be expressed as 2p63d0 → 2p53d1, which is reflected as the energy absorption at 458 eV (peak A). Ti3+ (2p63d1) already has one electron in the t2g orbital, which reduces the number of unoccupied states. Therefore, the presence of Ti3+ leads to the reduction of t2g relative intensity and eg peak width broadening. (35) The lesser resolved peaks B and C, the reduction of the A/B intensity ratio (Figure 3c), and the shift of peak A to lower energy confirm the presence of Ti3+ in Blue and Black NTA as well. (35,36) The O K-edge XAS is shown in Figure 3b. Peaks a and b can be assigned to the electronic transition from O 1s to O 2p hybridized with the Ti 3d t2g and 3d eg states, respectively. (37,38) The peak intensity ratio of t2g to eg is lower in Blue and Black NTA (Figure 3c), which suggests the presence of more Ti3+ and Ov. (35) In agreement with the XPS analyses, Co-Black has the highest number of Ti3+ and Ov, sites as supported by the lowest peak A/B and a/b ratios in XAS (Figure 3c).

Figure 3

Figure 3. (a) Ti L-edge, (b) O K-edge XAS spectra, and (c) intensity ratio of specific peaks. (d) Co K-edge XAS spectra (insert: enlarged view of pre-edge structure).

We further investigated the oxidation state and coordination environment of Co dopant by XAS. The absorption edge in Co K-edge XAS spectra of Co-Black overlaps with that of CoO (Figure 3d); this implies that surficial Co ions of Co-Black are in valence 2+, which is in line with the XPS analyses (Figure S9), and the EPR results (vide infra). The pre-edge peaks of CoO and Co3O4 correspond to the 1s–3d electronic transitions contributed by the tetrahedrally coordinated Co2+. This feature is less pronounced for Co-Black, indicating that Co2+ is in an Oh configuration in which 1s–3d electronic transitions rarely occur. (39) The overall Co K-edge profile of Co-Black is different from that of CoO and Co3O4. This results again suggest that the coordination structure of Co2+ ions in Co-Black is clearly different from that of oxide particles. Given that no CoOx particulates could be found by both SEM and TEM, it is possible that Co2+ is atomically doped into the lattice of TiO2, adopting the same Oh configuration. This assumption needs to be verified by aberration-corrected TEM in a future study.
The coordination structure of Co2+ remains intact after 100 h of electrolysis, as indicated by the unchanged Co K-edge profiles of Co-Black before and after use (Figure 3d). Of note, XPS data indicated that Co2+ in Co-NTA was prepared by annealing the Co(NO3)2 loaded Am-NTA in air, even though the calcination of Co(NO3)2 alone produced Co3O4 (i.e., a mixed Co(III)/Co(II) material). (Figure S9). These observations combined suggest that Co2+ is effectively immobilized on NTA and its reduced valence state does not result from H2 reduction but from a strong Co–TiO2 interaction. It is noteworthy that Co2+ in CoOx is a labile OER intermediate that tends to be dissolved at neutral pH in the absence of phosphate. (23,33) Our findings suggest that the Co–TiO2 interaction is strong enough to prevent Co leaching at circumneutral pH. These results could provide a new strategy to prepare stable CoOx-based OER catalysts.
NTA samples were scraped off from Ti substrate for EPR characterization. The EPR signals of powdered NTA samples should mainly reflect their bulk characteristics. Signals were normalized by sample weight in order to perform semiquantitative comparison. The EPR signal at a g value of 2.003 could be attributed to electrons localized on Ov. (14,40) As shown above, XPS and XAS unambiguously point out that Blue NTA has more surficial Ov than pristine NTA. However, higher bulk phase Ov concentration is not observed on Blue NTA, as shown by EPR (Figure