Introduction

Solid oxide electrolysers have demonstrated the tremendous advantages of electrochemical conversion of CO2 into fuels with high efficiencies using renewable electrical energy1,2,3. Oxide-ion-conducting solid oxide electrolysers can directly electrolyse CO2 into CO and pure oxygen under external applied potentials. At the cathode, CO2 molecules are electrochemically split into CO while the generated O2− ions are transported through the electrolyte membrane to the anode compartment where pure O2 gas is formed and released4,5.

The conventional Ni/YSZ electrode has exhibited excellent steam-electrolysis performance under a reducing atmosphere; however, the Ni-cermets are not redox stable and require a significant concentration of reducing gas flowing over the Ni metal to avoid the oxidation of Ni to NiO6,7. Similarly, the flowing of CO to the composite electrode is also necessary during electrolysis of CO2 because the oxidation of Ni occurs at high temperatures8. Furthermore, the catalytic activity of Ni metal toward the splitting of CO2 is relatively high; carbon deposition most likely occurs and results in the degradation of cell performance. Some researchers have demonstrated that the deposition of carbon is likely caused by reactions that occur over the catalyst and favour to occur only when CO is present in the chemical reaction system8,9,10. Cu/YSZ and Cu/SDC have also been considered as potential electrodes because of their ability to adsorb CO2, resistance to carbon fouling and low cost; however, the catalytic activity of Cu toward CO2 splitting is inferior to that of Ni11,12. Cu similarly oxidises during the direct electrolysis of CO2 or H2O in the absence of a reducing gas flowing over the composite cathode at high temperatures.

High-temperature electrolysis based on the redox-stable LSCM or LSTO cathode has been reported for the direct electrolysis of H2O, CO2 or H2O/CO2 and promising electrode performances have also been observed13,14,15,16,17. However, the LSCM, as a p-type conductor, is not well adapted to strong reducing potentials. LSCM exhibits lower conductivity than other cathode materials and undergoes adverse chemical changes under such extreme conditions; these changes lead to a large electrode polarisation resistance and to degradation of the electrolysis performance18,19. Direct electrolysis of CO2 with a current efficiency as high as approximately 50–60% has been achieved with LSCM cathodes, as reported in our previous work; however, rapid electrode degradation occurs at an external load of 2 V at 800°C14. By contrast, LSTO, which exhibits n-type conductivity upon reduction, has been considered a breakthrough in the development of redox-stable anode materials for SOFCs and has also been utilised for direct high-temperature electrolysis20. However, this material's insufficient catalytic activity still restricts the electrode polarisation and current efficiency during high-temperature electrolysis15. To improve the performance of LSTO, metal catalysts have been introduced, which has resulted in a substantially enhanced electrode performance21,22. However, control of the morphology of such catalysts on the substrate is difficult and operation over a long period of time would risk catalyst agglomeration and, therefore, performance degradation. As documented in our previous work16,23,24, steam electrolysis based on Ni- or Fe-impregnated LSCM or LSTO is stable during the first hour of operation at a load of 1.5–2 V, whereas an approximately 20–30% degradation in cell performance is observed when the operating time exceeds 5 h because of the agglomeration of metal nanoparticles on the surface of the composite cathode skeleton.

An alternative method is to incorporate the catalyst as a dopant within a host lattice during the synthesis of the catalyst in air, which is then exsolved at the surface in the form of catalytically active metallic nanoparticles or microparticles under reducing conditions25. Upon re-oxidation, the dopant can be re-incorporated into the host lattice, yielding a regenerative catalyst. In this case, any possible agglomeration of exsolved metallic nanoparticles on the substrate surface can be avoided by periodically exposing the material to oxidising conditions. Fluorite oxide, NbTi0.5Ni0.5O4 (NTNO), which is a single-phase rutile-type structure with the tetragonal (P42/mnm) space group, was first reported as a potential anode material for solid oxide fuel cells by Irvine et al.26,27. Nanoparticles of metallic Ni can be exsolved at the surface of the highly-conducting Nb1.33Ti0.67O4 substrate after high-temperature reduction under a reducing gas. The Nb1.33Ti0.67O4 is a mixed conductor because of the edge sharing of NbO6 octahedra along the c-axis, which result in Nb–Nb metal bond overlap with an inter-metallic distance of 3 Å. This high electronic conductivity is facilitated by the edge-sharing octahedra along the c-axis28. In our previous work, the nanosized Ni/Nb1.33Ti0.67O4 composite exhibited promising electrode polarisations and current efficiencies for the direct steam electrolysis29.

Copper-based catalysts exhibit lower selectivity toward H2 than toward CO or CO2 because of their high over-potential of hydrogen evolution42; however, their excellent CO2/monoxide adsorption properties and resistance to carbon fouling make copper-based catalysts appealing to many researchers30,31,32,33. In contrast to copper, Ni metal exhibits high catalytic activity but probably tends to be prone to high coke formation; such a severe deactivation poses a serious obstacle to the future application of Ni-metal catalysts. Coke is favour to occur when CO and hydrocarbons are present for their chemisorption34. Some researchers have taken a compromise approach by mechanically mixing Ni and Cu particles to form a Ni + Cu composite and therefore utilise the respective catalysis characteristics of both nickel and copper35,36,37,38. The performance trends reported for electrodes prepared by impregnation and those prepared by mechanical alloying are similar; mechanical alloying is therefore a promising route to prepare electrode materials. Recently, a novel NiCu alloy catalyst was reported and its catalytic behaviour can be continuously tuned by tailoring the molar ratio between Cu and Ni in the alloy39,40. Xie et al. observed that the 50%Ni-50%Cu alloy exhibits excellent catalytic performance for CH4 dissociation through theoretical first-principles calculations41. Bujalski and co-workers demonstrated the tailored catalytic activity of the 50%Ni-50%Cu alloy for hydrogen production from the partial oxidation of methane and methanol42. The NiCu alloy catalyst exhibits the advantages of both metallic nickel and metallic copper. Furthermore, the synergistic effect makes the NiCu alloy with continuously tuned compositions an outstanding catalyst in many scientific fields43,44,45.

Here, we report the reversible in situ growth of NixCu1-x (x = 0, 0.25, 0.50, 0.75 and 1.0) alloy catalysts to anchor and decorate the redox-reversible Nb1.33Ti0.67O4 ceramic substrate with the aim of tailoring the electrocatalytic activity of the composite materials through direct exsolution of metal particles from the crystal lattice of the ceramic oxide. The anchored interface is expected to enhance the electrocatalytic kinetics and the high-temperature stability of such nanoparticles. The electrical properties of the ceramic composite are investigated and further correlated to the electrochemical performance of their composite electrodes. Direct electrolysis of CO2 is then performed using the NixCu1-x/Nb1.33Ti0.67O4 composite cathode in a solid-oxide electrolyser at high temperatures.

Methods

All the powders were purchased from SINOPHARM Chemical Reagent Co., Ltd (China) unless specified otherwise. The synthesis of NbTi0.5(NixCu1-x)0.5O4 (x = 1.0, NTNO; x = 0.75, NTN3C1O; x = 0.5, NTN1C1O; x = 0.25, NTN1C3O; x = 0, NTCO) was carried out by a solid-state reaction using an appropriate mixture of Nb2O5, TiO2, CuO and NiO powders29. The mixture was ball-milled in acetone, dried, pressed into pellets at room temperature and treated at 1300, 1150, 1100, 1050 and 1000°C for the NTNO, NTN3C1O, NTN1C1O, NTN1C3O and NTCO, respectively, for 20 h in air. The (La0.8Sr0.2)0.95MnO3-δ (LSM) was synthesised via the combustion method using La2O3, SrCO3 and C4H6MnO4·4H2O as precursors; the final heat treatment was conducted at 1100°C for 3 h in air. However, the Ce0.8Sm0.2O2-δ (SDC) powders were prepared using the aforementioned method with Sm2O3 and Ce(NO3)4·6H2O precursors followed by a treatment at 800°C for 3 h in air52,53. The NTNO, NTN3C1O, NTN1C1O, NTN1C3O and NTCO samples were sintered and then reduced (5% H2/Ar) at 1300, 1250, 1150, 1050 and 1000°C (3°C min−1), respectively, for 20 h to achieve the reduced form of the composite Nb1.33Ti0.67O4 + NixCu1-x. The in situ growth of NixCu1-x alloy nanocatalysts when all of the oxide forms were reduced in reducing atmosphere (5% H2/Ar) at 1050°C (3°C min−1) for approximately 20 h. The reduction-oxidation cycles of the composite materials were conducted for another three cycles under reducing or oxidising atmospheres (5% H2/Ar or static air).

The X-ray diffraction (XRD, 2θ = 3°·min−1, 2θ = 10–90°, D/MAX2500V, Rigaku Corporation, Japan) analyses were performed to characterise the structural changes of the materials after the redox cycles. The microstructure of the NbTi0.5(NixCu1-x)0.5O4 and NixCu1-x/Nb1.33Ti0.67O4 were investigated by scanning electron microscopy (SEM, SU8020, HITACHI, Ltd., Japan) coupled with energy-dispersive X-ray spectroscopy (EDS). The valence states of the elements in the NbTi0.5(NixCu1-x)0.5O4 and NixCu1-x/Nb1.33Ti0.67O4 were analysed using an ESCALAB250 spectrometer equipped with an Al Kα (1486.6 eV) radiation source. The thermogravimetric analysis (TGA, STA449F3, Germany) of the oxidised NTNO and NTN1C1O samples was conducted from room temperature to 1000°C and then back to room temperature under a reducing atmosphere; the heating rate was 4°C min−1 and the flow rate of 5% H2/Ar (99.99% purity). Transmission electron microscopy (TEM) analysis with selected-area diffraction was performed to observe the oxidised and reduced NTN1C1O powders; the analyses were performed on a JEOL 2100F field-emission transmission electron microscope operated at 200 kV.

Approximately 2.0 g of NTNO, NTN3C1O, NTN1C1O, NTN1C3O or NTCO powder was compacted into a bar at a pressure of 6 MPa (20.0 mm × 7.0 mm × 2.4 mm) and subsequently sintered at 1300, 1150, 1100, 1050 or 1000°C for 10 h, respectively. The relative densities of all the samples were similar to each other, at approximately 80%. The bars were reduced completely at 1300, 1150, 1100, 1050 or 1000°C, respectively, for 20 h in a reducing atmosphere (5% H2/Ar) for the conductivity tests. The conductivity of the reduced samples, NixCu1-x/Nb1.33Ti0.67O4, were tested in 5% H2/Ar (99.99% purity) from room temperature to 800°C using the DC four-terminal method; the conductivity was recorded at a step of 0.5°C using an online system. The dependence of the conductivity on the oxygen partial pressure was tested at 800°C under an oxygen partial pressure that ranged from 10−2 to 10−19 atm; the oxygen partial pressure was adjusted by variation of the flow rate of the dry 5% H2/Ar. The oxygen partial pressure (lgpO2) and the conductivities were recorded using an online oxygen sensor (1231, ZrO2-based oxygen sensor, Noveltech, Australia) and an online multimeter (Keithley 2000 digital multimeter, Keithley Instruments, Inc., USA), respectively.

We prepared the 1 mm-thick YSZ disc electrolyte supports by dry-pressing the YSZ powder into green disks with a diameter of 20 mm and subsequently sintering the disks at 1500°C (2°C·min−1) for 10 h in air. The two surfaces of the obtained YSZ electrolyte support were mechanically polished and then ultrasonically cleaned in distilled water. The NbTi0.5(NixCu1-x)0.5O4-SDC composite electrode slurries were prepared by milling the SDC powder with NbTi0.5(NixCu1-x)0.5O4 powder at a 35:65 weight ratio in alpha-terpineol with cellulose as an additive54,55. The electrode slurries were then coated onto the electrolyte in symmetric positions in an area of 1 cm2 to assemble the symmetric cells; the cells were subsequently subjected to a heat treatment at 1000°C (3°C min−1) for 3 h in air. The current-collection layer was prepared using silver paste (SS-8060, Xinluyi, Shanghai, China), which was printed onto both electrode surfaces. The external circuit was constructed using silver electrical wire (0.1 mm in diameter), which was then connected to both current collectors using silver paste (DAD87, Shanghai Research Institute for Synthetic Resins) followed by firing at 550°C (3°C min−1) for 30 min in air. The AC impedance of the symmetric solid-oxide cells was tested at 800°C with the two electrodes exposed to different hydrogen partial pressures at the open-circuit voltage (OCV) using an electrochemical station (IM6, Zahner, Germany) with a frequency range of 4 M–0.1 Hz and a current strength of 10 mA in two-electrode mode. The electrode polarisation resistance was calculated by modelling the spectra using the Zview software. The gas flow rate was controlled at 20 ml min−1 using a mass flow meter (D08-3F, Sevenstar, China) to mix different ratios of H2 (99.99%) and N2 (99.99%).

The solid-oxide electrolysers were constructed using NbTi0.5(NixCu1-x)0.5O4 as the cathode and LSM as the anode, with YSZ as the electrolyte. The two surfaces of the electrolyte were coated with NbTi0.5(NixCu1-x)0.5O4-SDC slurry and LSM-SDC slurry in symmetric positions with an area of 1 cm2 and were subsequently subjected to the same heat treatment used for the symmetric solid-oxide cells, as previously described. The single solid-oxide electrolyser was sealed in a home-made testing jig using ceramic paste (JD-767, Jiudian, Dongguan, China) for the electrochemical measurements. The electrochemical measurements, including the AC impedance and current–voltage (I–V) plots of the solid oxide electrolyser, were performed with the cathode electrode exposed to flowing pure CO2 (99.99% in purity) and the oxygen electrode in static air at 800°C. The electrolysis of CO2 was performed at different applied voltages with CO2 fed into the cathode at 800°C. The output gas from the cathode was analysed using an online gas chromatograph (GC9790II, Fuli, Zhejiang, China) to determine the concentration of CO.

Results and Discussion

Fig. 1 (a) shows the XRD patterns of the single-phase rutile-type NbTi0.5(NixCu1-x)0.5O4 (x = 1, 0.75, 0.5, 0.25, 0), which confirms the homogeneous solid solution, i.e., even the composition of Ni and Cu changes in a wide range. The structures of the NbTi0.5(NixCu1-x)0.5O4 are tetragonal with the P42/mnm space group29. Fig. 2 shows the parameters for NbTi0.5(NixCu1-x)0.5O4 (x = 1, 0.75, 0.5, 0.25, 0) calculated from the XRD data. We observed that the cell parameters a and c increase from 4.6944(8) Å and 3.0229(5) Å to 4.6974(8) Å and 3.0297(4) Å when the proportion of nickel is decreased from 1 to 0.5, respectively. However, a remarkable decrease of cell parameter a occurs, whereas the cell parameter c sharply increases from 3.0297(4) Å to 3.0378(7) Å and then decreases to 3.0316(2) Å when the Ni proportion, x, is decreased to 0. Notably, the cell volumes are distributed as a pyramid with the cell volume of x = 0 and 1 at the bottom, which may be related to the larger ionic radii of Ni2+ (0.69 Å) and Cu2+ (0.73 Å) compared with that of Ti4+ (0.605 Å) with the same coordination numbers. The incorporation of Cu2+ may play a leading role in the expansion of the cell parameters, especially the c-axis parameter, possibly because of an increase in lattice strength with Cu doping. Fig. 1 (b) shows the XRD patterns of reduced NbTi0.5(NixCu1-x)0.5O4 (x = 1, 0.75, 0.5, 0.25, 0), which confirms that the reduced NbTi0.5(NixCu1-x)0.5O4 are a mixture of two phases: NixCu1-x + Nb1.33Ti0.67O4. As shown in Fig. 1 (b), the NbTi0.5(NixCu1-x)0.5O4 changes into Nb1.33Ti0.67O4 (PDF No. 053-0293) and metallic NixCu1-x alloy upon high-temperature reduction in 5% H2/Ar (99.99%). This result confirms the phase change of NbTi0.5(NixCu1-x)0.5O4 to highly electrically conducting Nb1.33Ti0.67O4 and pure catalytically active metallic NixCu1-x alloy.

Figure 1
figure 1

XRD patterns of NbTi0.5(NixCu1-x)0.5O4 (a: patterns of the oxidised form; b: patterns of the reduced form; c: patterns of oxidised form after three redox cycles; d: patterns of the reduced form after three redox cycles).

Figure 2
figure 2

Cell parameters of the oxidised NbTi0.5NixCu1-xO4 powder samples with x = 0, 0.25, 0.5, 0.75 and 1.0.

To investigate the reversibility of the exsolution of the metallic NixCu1-x, the composite powders were further treated for another three redox cycles at 1050°C under a reducing atmosphere (5% H2/Ar) and a oxidising atmosphere (static air) for 20 h. Fig. 1 (c) shows the XRD patterns of the composite after oxidation, which again confirms the successful integration of NixCu1-x alloys into the Nb1.33Ti0.67O4 substrate and the formation of a single-phase NbTi0.5(NixCu1-x)0.5O4 solid solution. No phase impurities are observed in the case of NbTi0.5(NixCu1-x)0.5O4, thereby firmly verifying the superior redox reversibility of the NbTi0.5(NixCu1-x)0.5O4 ceramics. As shown in Fig. 1 (d), the corresponding XRD patterns confirm that the reduced NbTi0.5(NixCu1-x)0.5O4 (x = 1, 0.75, 0.5, 0.25, 0) are still composed of conducting ceramic Nb1.33Ti0.67O4 and metallic alloy powders, which further demonstrates the excellent reversible exsolution of the metallic NixCu1-x from the lattice of NbTi0.5(NixCu1-x)0.5O4 through a reversible phase transition. The three successive redox treatment cycles of NbTi0.5(NixCu1-x)0.5O4 solid solutions suggest that the parent material NbTi0.5(NixCu1-x)0.5O4 and the NixCu1-x + Nb1.33Ti0.67O4 composite materials exhibit excellent redox reversibility. The metallic NixCu1-x alloys can be repeatedly exsolved from the NbTi0.5(NixCu1-x)0.5O4 after the reduction and again integrated back into the Nb1.33Ti0.67O4 substrate to form a homogeneous phase.

To further validate the elemental valence change in redox cycles of the NbTi0.5(NixCu1-x)0.5O4 solid solution, we performed XPS on the oxidised and reduced NTN3C1O (NbTi0.5(Ni0.75Cu0.25)0.5O4 and Ni0.75Cu0.25 + Nb1.33Ti0.67O4) samples. All XPS patterns were fitted using a Shirley-type background subtraction method. The background functions for different the spectra of different elements were fitted using 80% Gaussian and 20% Lorentzian functions46. As shown in Fig. 3 (a), only Nb5+ is observed in the oxidised NTN3C1O sample; however, the Nb5+ is reduced to Nb4+ in the Nb1.33Ti0.67O4 substrate when the sample is treated under a reducing atmosphere at 1200°C for 20 h, as seen in Fig. 3 (b)28,47. The low valence of the Nb4+ ions favours the formation of Nb-Nb metal bonds and the electronic conductivity is thus highly facilitated, resulting in enhanced electrical conductivity of Nb1.33Ti0.67O4. By contrast, element Ti, as exhibited in Fig. 3 (c), is completely in the Ti4+ state in the oxidised NTN3C1O sample; however, the low-valence Ti3+ is also observed in the reduced sample, Nb1.33Ti0.67O4 + Ni0.5Cu0.5, as shown in Fig. 3 (d), due to the chemical reduction of Ti4+ to Ti3+ after the heat treatment under a reducing atmosphere. We reasonably speculated that the Nb1.33Ti0.67O4 is actually oxygen-deficient Nb1.33Ti0.67O4-δ according to the charge balance29. The Ti3+ (2p1/2) and Ti3+ (2p3/2) peaks are observed at 462.80 eV and 457.80 eV, respectively, whereas the Ti4+ (2p1/2) and Ti4+ (2p3/2) show peaks at 464.80 eV and 458.10 eV, respectively29. Furthermore, the low-valence Ti3+ represents an approximately 0.174 mole ratio of all of the Ti according to this result, which indicates an oxygen deficiency of δ = 0.058 and a chemical formula of Nb1.33Ti0.67O3.942. The Ni and Cu in the NTN1C1O are in the form of Ni2+ and Cu2+, as shown in Fig. 3 (e) and Fig. 3 (g), respectively. In Fig. 3 (f), the Ni0 (2p1/2) peaks are observed at 873.86 and 869.40 eV, whereas the Ni0 (2p3/2) has three peaks at 861.0, 869.40 and 852.0 eV16,29,48. A similar change in the chemical state of elemental Cu is also observed in the NTN1C1O sample after reduction at high temperatures, as shown in Fig. 3 (h). The Cu0 (2p1/2) and Cu0 (2p3/2) peaks are observed at 951.88 and 932.20 eV, respectively. According to reports in the literature, the electronic potential energy of alloys of transition metals can be sensitively changed49,50. In this work, some peaks of binding energy change by approximately 0.2 to 0.3 eV in the case of Ni0 and Cu0 in the Ni0.5Cu0.5 alloy. The Ni0 and Cu0 exist in the form of metallic alloy nanoparticles in the reduced sample, which further confirms the exsolution of metallic Ni0.5Cu0.5 alloy on the ceramic substrate. The XPS results further confirm the reversible transformation between the oxidised NbTi0.5(NixCu1-x)0.5O4 solid solution and the reduced Nb1.33Ti0.67O4 ceramic electronic conductor loaded with metallic NixCu1-x alloys.

Figure 3
figure 3

XPS results for Nb(a), Ti(c), Ni(e) and Cu(g) in the oxidised NbTi0.5(Ni0.75Cu0.25)0.5O4 and for Nb(b), Ti(d), Ni(f) and Cu(h) in the reduced NbTi0.5(Ni0.75Cu0.25)0.5O4.

TGA of the oxidised NTN1C1O and NTNO samples was conducted from room temperature to 1000°C at a heating rate of 4°C·min−1 in a reducing atmosphere of 5% H2/Ar. To ensure sufficient reduction of the powder sample, the predetermined temperature at 1000°C is stabilised for as long as 3 h and then drifts to room temperature at the same rate, as previously stated. Fig. 4 (a) presents the percentage weight change of the oxidised NTNO powder as a function of temperature when heated under a 5% H2/Ar atmosphere. The weight loss reaches 7.35% for the NTNO sample because of the loss of oxygen under such reducing conditions, which is in a good agreement with the chemical change of the NbTi0.5Ni0.5O4 solid solution to the reduced Ni + Nb1.33Ti0.67O4 composite material. In comparison, the weight loss of the studied NTN1C1O powder reached approximately 7.59%, as seen in Fig. 4 (b), which is reasonably consistent with the theoretical value of 7.61% corresponding to the oxygen loss according to the chemical reaction presented in equation (1). Notably, the onset temperature of weight loss for the NTNO sample is approximately 750°C, as determined from the results in Fig. 4 (a), where the in situ exsolution of Ni metal is anticipated. Nevertheless, the weight loss of NTN1C1O starts at temperatures as low as approximately 600°C to grow the metallic alloy. This growth may occur because the incorporation of copper reduces the chemical reaction barrier of metal exsolution and thus facilitates the reversible phase change to grow the metallic NiCu alloys. As evident in both Fig. 4 (a) and Fig. 4 (b), a small weight loss always occurred during the process of cooling. This weight loss is likely to exceed 7.40 and 7.61% for the NTNO and NTN1C1O, respectively, if the samples experience longer reduction times at higher temperatures. Ultimately, the obtained reduced sample may be oxygen-deficient Nb1.33Ti0.67O4-δ + Ni or oxygen-deficient Nb1.33Ti0.67O4-δ + Ni0.5Cu0.5. Despite the different proportions, our conclusion through reasoning is that these results coincide with the aforementioned results, as shown in Fig 3.

High-resolution transmission electron microscopy (HR-TEM) analysis of the oxidised NTN1C1O reveals lattice spacings of 2.546 Å (101) and 3.322 Å (110), as shown in Figs. 5 (a) and (b), consistent with the separation spacing determined by the XRD analysis. As presented in Figs. 5 (c) and 5 (d), the corresponding lattice spacing of the parent material, Nb1.33Ti0.67O4, can be obtained as 2.537 Å (101) and 3.363 Å (110). Moreover, the interplanar spacing is 2.238 Å (111). These spacings can be demonstrated by the results of the XRD analysis, shown in Figs. 1 and 2. Furthermore, the reduction of the NTN1C1O leads to the growth of Ni0.5Cu0.5 alloy nanoparticles on the Nb1.33Ti0.67O4 surfaces in Fig. 5 (d). The analysis results for the nickel copper alloy indicates a lattice spacing of 1.781 Å (200), consistent with the standard data for the alloy Ni1Cu1 data51. These TEM results further demonstrate and validate the reversible exsolution of metallic particles on the ceramic substrate through control of the phase transformation between the NbTi0.5(NixCu1-x)O4 solid solution and NixCu1-x/Nb1.33Ti0.67O4, as confirmed by the aforementioned XRD, XPS and TGA analyses. More importantly, the TEM results reveal that the alloy particles are exsolved to in situ grow and anchor to the surface of the highly electronically conducting Nb1.33Ti0.67O4 in the form of heterojunctions, which avoids any possible agglomeration of exsolved metallic nanoparticles on the substrate surface at high temperatures. The metal alloy nanoparticles can be re-incorporated or re-exsolved into/from the host material, yielding a catalyst that can be regenerated through periodic exposure of the material to oxidising/reducing conditions.

Figure 4
figure 4

TGA thermograms of the oxidised form of NbTi0.5Ni0.5O4 (a) and NbTi0.5Ni0.25Cu0.25O4 (b) in 5% H2/Ar; the heating rate was 5°C min−1.

Figure 5
figure 5

TEM results for the oxidised form of NbTi0.5Ni0.25Cu0.25O4 (a, b) and for the reduced form of NbTi0.5Ni0.25Cu0.25O4 (c, d).

Fig. 6 shows the SEM and energy-dispersive X-ray spectroscopy (EDS) maps taken from the oxidised and reduced NbTi0.5(Ni0.5Cu0.5)0.5O4 pellets. The sintered samples were reduced in 5% H2/Ar at 900°C for 17 h. As shown in Fig. 6 (a), the Ni and Cu are homogeneously dispersed in the oxidised NTN1C1O sample and the other elements are also well distributed in the bulk. Although the oxidised sintered NTN1C1O pellet is not highly dense, clear images are still observed for the sample with low sinterability and high porosity. In comparison, uniform exsolved nanoparticles anchor on the surface of the reduced NbTi0.5(Ni0.5Cu0.5)0.5O4 pellet, as shown in Fig. 6 (b), which is further confirmed by the EDS maps of the Ni and Cu. In addition, the Nb, Ti and O, which are components of the new phase of Nb1.33Ti0.67O4, as previously discussed, are distributed evenly in the sample. The SEM and EDS results presented in Figure 6 (b) of the reduced NTN1C1O suggest the exsolution of Ni0.5Cu0.5 alloy nanoparticles to anchor on the substrate surfaces. The presence of Ni0.5Cu0.5 alloy nanoparticles is further confirmed by the SEM and EDS results in addition to the XRD, XPS and TEM results, as previously mentioned. These results indicate that the reversible exsolution of NiCu alloy nanoparticles from the parent NbTi0.5(NixCu1-x)0.5O4 material to grow in situ and anchor the metal alloy nanocatalyst on the Nb1.33Ti0.67O4 surface is successful. The dispersed NixCu1-x nanoparticles anchoring on the Nb1.33Ti0.67O4 surface are expected to prohibit the agglomeration and improve the electrocatalytic activity of the composite cathode for high-temperature CO2 electrolysis.

Figure 6
figure 6

SEM and EDS results for the oxidised form of NbTi0.5Ni0.25Cu0.25O4 (a) and for the reduced form of NbTi0.5Ni0.25Cu0.25O4 (b).

The dependence of the conductivity of the samples on the temperature and oxygen partial pressure was systematically investigated, as shown in Fig. 7. According to the literature, Nb1.33Ti0.67O4 is a mixed conductor because of the edge sharing of NbO6 octahedra along the c-axis, which results in Nb-Nb metal bond overlap with an inter-metallic distance of 3 Å29. The reduced NbTi0.5(NixCu1-x)0.5O4 ceramic samples (Nb1.33Ti0.67O4 + NixCu1-x) exhibit greater conductivity, as shown in Fig. 7 (a), because the metallic NixCu1-x nanoparticles dispersed on the substrate improve the electrical conductivity under a reducing atmosphere (5% H2/Ar, 99.99% purity). The reduced NbTi0.5Cu0.5O4 (Nb1.33Ti0.67O4 + Cu) sample exhibits the highest conductivity, which reaches approximately 500 S·cm−1 at 730°C. With increasing nickel composition, the conductivity of the reduced samples gradually decreases to 204 S·cm−1 with the maximum Ni content in the reduced NbTi0.5Ni0.5O4 (Nb1.33Ti0.67O4 + Cu) at the same temperature. This phenomenon is possibly due to the Cu metal exhibiting conductivity superior to that of Ni metal at high temperatures. The conductivities measured under a reducing atmosphere is mainly attributed to the electronically conducting ceramic Nb1.33Ti0.67O4, in agreement with results reported in the literature26,27; however, the percolating metal network is not formed due to the presence of dispersed NixCu1-x alloys. As shown in Fig. 7 (b), the conductivities of Nb1.33Ti0.67O4 + NixCu1-x are generally independent of the oxygen partial pressure at partial pressures less than 10−4 atm; however, the conductivity significantly decreases when the samples are oxidised at high oxygen partial pressure due to the oxidation of Nb1.33Ti0.67O4 + NixCu1-x to the NbTi0.5(NixCu1-x)0.5O4 solid solution.

Figure 7
figure 7

(a) The dependence of conductivity on temperature of the reduced form of NbTi0.5(NixCu1-x)0.5O4 (x = 0, 0.25, 0.5 and 1.0) samples; (b) the dependence of the conductivity on the oxygen partial pressure of the reduced form of NbTi0.5(NixCu1-x)0.5O4 (x = 0, 0.25, 0.5 and 1.0) samples at 800°C.

Fig. 8 shows the AC impedance plots of the symmetric cells tested under a series of different hydrogen partial pressures (0, 2, 5, 20, 40, 50, 60, 80 and 100% H2) with nitrogen to adjust the hydrogen concentrations at 800°C. The impedance plots show two intercepts with the real axis, where the series resistant of the cell (Rs) corresponds to the first intercept and the difference between the two intercepts is a measure of the electrode polarisation resistance (Rp). The Zview software (IM6, Zahner, Germany) was employed to calculate Rs and Rp, as reported in our previous work46,48. The ionic resistance of the 1 mm YSZ electrolyte is the primary contributor to the Rs, which is approximately 2 Ω and is generally stable over a wide range of hydrogen partial pressures. In our previous work, we reported that a symmetric solid-oxide cell with a composite electrode based on Nb1.33Ti0.67O4 exhibits poor performance, where the electrode polarisation resistance is approximately 300 Ω cm229. As shown in Figs. 8 (a), (b) and (c), the Rp of the symmetric cell based on NTNO-SDC decreases from 18.65 to 3.02 Ω cm2 over the hydrogen partial pressure ranging of 0 to 100%. Similarly, the Rp values of the symmetric cells based on NTN3C1O, NTN1C1O, NTN1C3O and NTCO are shown in Figs. 8 (d, e, f), Figs. 8 (g, h, i), Figs. 8 (g, k, l) and Figs. 8 (m, n, o), respectively; the Rp decreases from 34.72 to 6.08 Ω cm2, 45.61 to 14.26 Ω cm2, 57.04 to 7.64 Ω cm2 and 39.85 Ω cm2 to 10.81 Ω cm2, respectively. All of the Rp values change as functions of the hydrogen partial pressure, which suggests that the stronger reducing atmosphere improves the electrode polarisations. A stronger reducing atmosphere favours the growth of metallic NixCu1-x alloy particles on the surface of the electrode skeleton and thus enhances the electrocatalytic activity of the composite electrodes. However, the polarisation resistance gradually decreases with increasing nickel content when compared with all of the proportions, which is attributed to the catalytic activity of nickel metal being greater than that of copper. The electrocatalytic activity of the composite electrode can be continuously tuned according to the strong dependence of the electrode activity on the alloy compositions under reducing atmospheres at high temperatures.

Figure 8
figure 8

AC impendence plots for the symmetrical cells based on NbTi0.5(NixCu1-x)0.5O4 electrodes with x = 1 (a, b, c), x = 0.75 (d, e, f), x = 0.5, (g, h, i), x = 0.25 (j, k, l) and x = 0 (m, n, o) tested at 800°C under different hydrogen partial pressures.

To investigate the sealing of the single solid-oxide electrolyser based on the NTNxC1-xO composite cathode, we recorded the open-circuit voltage (OCV) with the cathode and anode exposed to 5% H2/Ar and static air, respectively and examined the separation between the anodic and cathodic gases. The OCVs of the cells reach approximately 1.0 V, which indicates good separation between the anodic and cathodic gases. Fig. 9 shows the I–V curve at 800°C with 100% CO2 introduced to the cathode and with the anode exposed to static air. The relationship between the current and the applied voltage is far from linear, with the maximum current densities reaching 169.71, 113.17, 80.46, 69.14 and 56.57 mA cm−2 at 1.6 V for the electrolysers based on NTNO, NTN3C1O, NTN1C1O, NTN1C3O and NTCO, respectively. Numerous investigators have demonstrated that the electrolysis current depends on the availability of the three-phase interfaces. However, the electrolysis current also increases rapidly with increasing nickel content because of the high catalytic activity of nickel. Thus, a high content of nickel catalyst improves the cathode performance.

Figure 9
figure 9

I–V curves of the solid oxide electrolysers based on NbTi0.5(NixCu1-x)0.5O4 (x = 0, 0.25, 0.5, 0.75 and 1.0) cathodes for CO2 electrolysis at 800°C.

In situ AC impedance spectroscopy was carried out under different applied voltages to investigate the changes to Rs and Rp, where Rs and Rp correspond to the series resistance and polarisation resistance of the electrolyser, respectively. As shown in Fig. 10, all of the series resistances, Rs, are stable; however, the Rp values sharply decrease as a function of the applied voltage in the range of 1.2 to 1.4 V. The Rp is significantly enhanced by the application of higher external potentials because the passing current density not only activates the electrode, but also electrochemically reduces the cathode to enhance the mixed conductivity and electrocatalytic activity. We observed that the Rp is only 2.17 Ω·cm2 at 1.4 V in CO2 in the case of the NTNO cathode, whereas the Rp is approximately 2.85 Ω·cm2 at low voltage of 1.1 V as shown in Fig. 10 (a). However, as seen in Figs. 10 (b), (c) and (d), the Rp values are 6.21, 9.94 and 6.22 Ω·cm2 for the electrolysers with NTN3C1O, NTN1C3O and NTCO cathodes at 1.4 V, respectively. Here, the exsolution of metallic nanoparticles may significantly improve the electrocatalytic activity of the composite cathodes and the passing current may further activate the composite electrodes and improve the electrode polarisation. To a certain extent, we conclude that the nickel content in NiCu alloys and the crystal structure of NiCu alloys may have a coordinated action for CO2 electrolysis; if this is not the case, the Rp should increase with decreasing proportion of nickel because nickel exhibits superior catalytic performance for CO2 decomposition.

Figure 10
figure 10

AC impendence plots of the solid oxide electrolysers based on NbTi0.5(NixCu1-x)0.5O4 (a: x = 1; b: x = 0.75; c: x = 0.25; d: x = 0) composite electrodes for CO2 electrolysis with the oxygen electrode exposed in static air at 800°C.

Fig. 11 shows the rate of CO production and the current efficiency of the electrolysers versus the nickel composition based on NTNxC1-xO composite cathodes for direct CO2 electrolysis under applied voltages of 1.2, 1.3 and 1.4 V, respectively. As shown in Fig. 11 (a), the CO production rates gradually improve with increasing applied voltage for all of the electrolysers and exhibit the highest rates at a final applied voltage of 1.4 V because high voltages favour CO2 splitting at the cathode. The CO production rate of the electrolyser with the NTN3C1O (x = 0.75) composite cathode is 0.1629 ml·min−1·cm−2 at 1.4 V, which is approximately three times greater than the rate of 0.04113 ml·min−1·cm−2 for the cell at the same applied voltage of 1.2 V. However, we note that the productionrate of CO is strongly dependent on the nickel composition in the parent material, i.e., the NbTi0.5(NixCu1-x)0.5O4 solid solution, which definitely determines the nickel concentration of the metallic NixCu1-x alloy nanocatalysts. The electrolyser produces CO at a rate of 0.07663 ml·min−1·cm−2 in the case of the NTCO composite cathode and was significantly enhanced to as high as 0.3866 ml·min−1·cm−2 in the case of the NTNO cathode at a voltage of 1.4 V. These results again verify the superior catalytic activity of the nickel catalyst toward electrochemical CO2 splitting. Similar behaviour of the CO production rate versus the nickel composition was also observed at applied voltages of 1.2 and 1.3 V. Clearly, the CO production improves slowly when x < 0.5 and then increases rapidly in the interval of 0.5 ≤ x ≤ 1. This behaviour can be attributed to the excellent catalytic activity of the NixCu1-x alloy when the nickel concentration is high.

Figure 11
figure 11

(a) CO production rate for the cells based on NbTi0.5(NixCu1-x)0.5O4 (x = 0, 0.25, 0.5, 0.75 and 1.0) composite cathode versus x of NbTi0.5(NixCu1-x)0.5O4; (b) the relationship between the current efficiency of the CO2 electrolysis and x in the NbTi0.5(NixCu1-x)0.5O4 (x = 0, 0.25, 0.5, 0.75 and 1.0) composite cathodes.

Fig. 11 (b) shows the current efficiencies of the electrolysers versus the nickel composition for the case of NTNxC1-xO composite cathodes during direct CO2 electrolysis under different applied voltages. The current efficiencies gradually improve with increasing copper content in the composition range of 0 ≤ x ≤ 0.75 because of the adsorption of CO2 onto copper. However, the current efficiencies begin to decrease when the copper content is in the rage of 0.75 ≤ x ≤ 1, although they are still greater than the current efficiencies of the cathode with x = 0. Copper with excellent CO2/CO adsorption properties is expected to facilitate the electrochemical conversion of CO2 to CO and to therefore improve the current efficiencies. The optimum composition of Ni0.75Cu0.25 alloy catalyst combines the advantages of both metallic nickel and metallic copper to maximise current efficiencies for the direct CO2 electrolysis. The highest current efficiency of 74.29% are obtained in the case of the NTN3C1O cathode at 1.4 V. Similarly, the current efficiencies are accordingly improved versus the applied voltage because the applied voltages electrochemically reduce/activate the composite cathodes to enhance the electrocatalytic activity56. The in situ growth and anchoring of NixCu1-x alloy nanocatalysts on the Nb1.33Ti0.67O4 surface combines the advantages of metallic nickel and copper and produces a synergistic effect to maximise the current efficiencies for the direct electrolysis of CO2. The coupling of metal catalysts with the electronically conducting ceramic substrate offers possibilities for the direct CO2 electrolysis at high temperatures.

Conclusion

In this work, the in situ growth of NixCu1-x (x = 0, 0.25, 0.50, 0.75 and 1.0) alloy catalysts was achieved to anchor on the redox-reversible Nb1.33Ti0.67O4 ceramic substrate in the form of heterojunctions. The exsolution/dissolution of the metal particles is completely reversible during redox cycling. The electrocatalytic activity of the composite electrode based on NixCu1-x/Nb1.33Ti0.67O4 can be continuously tuned by tailoring the alloy composition. The coupling of metal catalyst with the electronically conducting ceramic substrate achieves direct CO2 electrolysis at high temperatures, whereas the combined advantages of metallic nickel and copper produces a synergistic effect to maximise the current efficiencies. Electrocatalytic activity of the composite materials was achieved through direct exsolution of metal particles from the crystal lattice of the ceramic oxides under a reducing atmosphere at higher temperatures. The current work expands our understanding of advanced ceramic cathodes with excellent electrocatalytic activity in the field of high-temperature electrolysis.