**CuO Nanoparticles Supported on TiO2 with High Efficiency for CO2 Electrochemical Reduction to Ethanol**

#### **Jing Yuan, Jing-Jie Zhang, Man-Ping Yang, Wang-Jun Meng, Huan Wang \* and Jia-Xing Lu \***

Shanghai Key Laboratory of Green Chemistry and Chemical Processes, School of Chemistry and Molecular Engineering, East China Normal University, Shanghai 200062, China; yuanjing158yj@126.com (J.Y.); zjj18221019582@163.com (J.-J.Z.); YMP2018@126.com (M.-P.Y.); mwjynl@outlook.com (W.-J.M.)

**\*** Correspondence: hwang@chem.ecnu.edu.cn (H.W.); jxlu@chem.ecnu.edu.cn (J.-X.L.); Tel.: +86-21-5213-4935 (H.W.); +86-21-6223-3491 (J.-X.L.)

Received: 7 March 2018; Accepted: 18 April 2018; Published: 21 April 2018

**Abstract:** Non-noble metal oxides consisting of CuO and TiO2 (CuO/TiO2 catalyst) for CO2 reduction were fabricated using a simple hydrothermal method. The designed catalysts of CuO could be in situ reduced to a metallic Cu-forming Cu/TiO2 catalyst, which could efficiently catalyze CO2 reduction to multi-carbon oxygenates (ethanol, acetone, and n-propanol) with a maximum overall faradaic efficiency of 47.4% at a potential of −0.85 V vs. reversible hydrogen electrode (RHE) in 0.5 M KHCO3 solution. The catalytic activity for CO2 electroreduction strongly depends on the CuO contents of the catalysts as-prepared, resulting in different electrochemistry surface areas. The significantly improved CO2 catalytic activity of CuO/TiO2 might be due to the strong CO2 adsorption ability.

**Keywords:** electrochemical reduction; CO2; CuO; TiO2; ethanol

#### **1. Introduction**

Electrochemical reduction of CO2 (CO2ER) is a promising and feasible method with which to sustainably transform this waste stream into value-added low-carbon fuels [1–5], which has the following advantages: (1) An electrochemical system can be operated under moderate reaction conditions [5,6]; (2) The reaction process involves highly complex multiple protons and electrons transfer steps, which lead to wide product distribution, such as methanol, ethanol, acetone, and so on [7,8]; (3) The electricity could be supplied by clean and sustainable energy, such as solar, wind, and hydropower [9,10]; and (4) This reaction requires minimal chemical intake and is convenient for large-scale applications [5,8]. However, CO2ER is facing severe challenges, including poor faradaic efficiency (FE), high overpotential, and low selectivity [11,12], which urges us to design new catalysts to address the above efficiency and selectivity issues.

Over the past few decades, numerous trials have been made to explore catalysts with distinguished performance for CO2ER. Until now, multifarious catalysts including metals [5,11–13], metal oxides [14–16], and metal complexes [17,18] have been reported. Among these catalysts, Cu, as a relatively low-cost and earth abundant metal, has a unique capacity to produce hydrocarbons through a multiple protons and electrons transfer pathway; however, the traditional Cu catalysts show high overpotential [16] and low selectivity for diversiform products [5,12,19–21]. Substantial efforts have been pursued to enhance energetic efficiency for CO2ER through altering surface structures, morphologies, and the nature compositions. Recently, Cu-based catalysts have been reported to possess enhanced FEs for CO2ER; for instance, Takanabe et al. [22] used Cu-Sn alloy for the efficient and selective reduction of CO2 to CO over a wide potential range. Yu and her team [23] reported that Cu nanoparticle (NP) interspersed MoS2 nanoflowers facilitate CO2ER to hydrocarbon, such as

CH4 and C2H4 with high FE at low overpotentials. Sun et al. [24] introduced a Core/Shell Cu/SnO2 structure that shows high selectivity to generate CO with FE reaching 93% at −0.7 V (vs. the reversible hydrogen electrode (RHE)).

Normally, TiO2, as a semiconductor material, is one of the most widely used photocatalysts and electrocatalysts for CO2 reduction because of its nontoxicity, low cost, and high chemical stability [25,26]. Furthermore, TiO2 has been reported to act as a redox electron carrier to facilitate a variety of reduction reactions, including CO2 conversion [27,28], and to assist in CO2 adsorption [29,30]; thus, it may stabilize the CO2ER intermediate and reduce overpotential. Some early works could verify this point in the CO2ER, such as Ag/TiO2 [31], Cu/TiO2 [32], and Cu/TiO2/N-graphene [33].

CuO/TiO2 has been previously prepared using different synthetic methods and is used for various applications, such as photodriven reduction of CO2 [34] and hydrogen production reaction [35]. Nevertheless, to the author's knowledge, CuO/TiO2 as electrocatalysts for CO2ER have been rarely investigated. In previous works, we have demonstrated that CuO with various morphologies can be highly efficient electrocatalysts for CO2 reduction to ethanol in simple aqueous medium, but at high overpotential (−1.7 V vs. the saturated calomel electrode (SCE)) [36]. Inspired by previous studies, in this work, we systematically and carefully synthesized a series of nano-sized CuO/TiO2 catalysts for CO2ER. The prepared CuO/TiO2 catalysts were optimized by varying the amount of loaded CuO NPs. CuO/TiO2 with an intended CuO content of 60%, as an efficient electrocatalyst, exhibits the most outstanding activity (achieved total FE of 47.4% at the potential of −0.85 V vs. RHE) for CO2ER in 0.5 M KHCO3 solution at room temperature among all as-prepared CuO/TiO2 catalysts.

#### **2. Results and Discussion**

#### *2.1. Catalyst Characterizations*

A simple and mild hydrothermal synthesis method was used to prepare six different weight ratios of CuO/TiO2 catalysts, which are defined as CuO/TiO2-1, CuO/TiO2-2, CuO/TiO2-3, CuO/TiO2-4, CuO/TiO2-5, and CuO/TiO2-6, corresponding to the intended CuO content of 5 wt %, 10 wt %, 20 wt %, 40 wt %, 60 wt %, and 80 wt %, respectively. X-ray diffractometer (XRD) patterns shown in Figure 1 indicated that these catalysts consist of both CuO and TiO2, which were represented by solid lines and dashed lines, respectively. It can be clearly seen that all as-prepared catalysts with different amounts of CuO loadings present similar XRD patterns. The diffraction peaks ascribed to CuO were significantly shown with the increasing of CuO contents in the XRD patterns. Meanwhile, the actual CuO contents were analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES) and summarized in Table S1. These values are in good agreement with the intended values.

**Figure 1.** XRD patterns of (**a**) pure TiO2, (**b**) CuO/TiO2-1, (**c**) CuO/TiO2-2, (**d**) CuO/TiO2-3, (**e**) CuO/TiO2-4, (**f**) CuO/TiO2-5, (**g**) CuO/TiO2-6, and (**h**) pure CuO.

To thoroughly examine the element proportion and chemical state of CuO/TiO2 catalysts, X-ray photoelectron spectroscopy (XPS) of CuO/TiO2-5 catalyst was investigated. As expected, Cu, Ti, and O elements from CuO/TiO2 catalyst were observed in the full spectrum from Figure 2a. Furthermore, Figure 2b displays the high resolution spectrum of Cu 2p, separated into Cu 2p3/2 and Cu 2p1/2 at 933.8 eV and 953.8 eV, respectively. The distance between these Cu 2p main peaks positions is 20.0 eV, which agrees well with previous reports about CuO spectrum [37]. Moreover, additional confirmation of CuO state was seen with the broad satellite peaks at a higher binding energy than the main peaks. The main peak of Cu 2p3/2 at 933.8 eV was accompanied by two satellite peaks on the higher binding energy side at about 943.8 eV and 941.5 eV, which suggests the existence of CuO [38–41]. From this figure, we can clearly see that the main peak of Cu 2p1/2 at 953.8 eV and its satellite peak at 962.5 eV were separated by about 9.0 eV, which also confirms the presence of CuO [42]. Both Ti 2p3/2 and Ti 2p1/2 peaks at 458.7 eV and 464.4 eV, respectively, were observed (Figure 2c) with a separation of 5.7 eV, indicating that TiO2 existed in this catalyst [43]. The obtained XPS spectrum of O 1 s was presented in Figure 2d. An obvious peak appeared at 529.7 eV, which can be indexed to O2- in the CuO and TiO2. Notably, there are other three weak O 1 s peaks. One located at 530.7 eV is identified with surface hydroxyls, which is likely the by-product from the synthesis process of CuO/TiO2. Additionally, the remaining two peaks observed at 531.5 eV and 532.8 eV are confirmed to C=O and C-O [44,45]. A slight peak of C 1 s was detected in Figure 2a, which is always observed in XPS spectra of real-world solids [46,47]. The high resolution spectrum of C 1 s was shown in Figure S1, consistent with the results of O 1 s.

**Figure 2.** (**a**) The full XPS spectrum of CuO/TiO2-5 catalyst and high-resolution XPS spectra of (**b**) Cu 2p; (**c**) Ti 2p and (**d**) O 1 s.

To obtain the morphological information regarding the CuO/TiO2 catalysts, scanning electron microscope (SEM) images were firstly shown in Figure S2. One can see that the catalysts were composed of 3-dimension NPs with a certain amount of CuO nano-floc supported on the surface of TiO2. Further from the SEM images, for low CuO content of CuO/TiO2 (5 wt %, 10 wt %, and 20 wt %), CuO NPs were dotted sporadically on the surface of TiO2. Subsequently, for high CuO content of CuO/TiO2 (40 wt %, 60 wt %, and 80 wt %), it is evident that the CuO NPs completely covered on the TiO2 surface forming a massive and compact layer. Pure CuO NPs presented very porous, sponge-like structures. Moreover, transmission electron microscopy (TEM) investigation was used to gain deeper insight into the structural feature of the catalysts. TEM of CuO/TiO2-5 was shown in Figure 3a. Numerous CuO NPs were irregularly interspersed on the TiO2 surface, in accordance with the SEM results. High-resolution TEM images of CuO/TiO2-5 were displayed in Figure 3b to acquire more detailed information about the structure of the catalyst. The fast Fourier transform (FFT, inset of Figure 3b) pattern of CuO and TiO2 shows concentric rings and bright discrete diffraction spots, which are indicative of high crystallinity. From Figure 3b, distinctive lattice fringes were found in both TiO2 and CuO NPs. The lattice fringes with interplanar spacing of 0.188 nm were assigned to the (200) plane of the TiO2, expressed by orange lines (inserted in Figure 3b). Additionally, the lattice fringes with d-spacings of 0.256 nm corresponded to the plane (111) of CuO phase, represented by the yellow lines (inserted in Figure 3b).

**Figure 3.** TEM images of CuO/TiO2-5 catalyst in (**a**) low magnification and (**b**) high magnification (the insets are FFT pattern and the detailed images of white dashed circle).

#### *2.2. Catalyst Properties*

The electrocatalytic activity of CuO/TiO2-5 catalyst was carried out in the typical three-electrode system through cyclic voltammetry (CV) measurement, which was firstly tested at a scan rate of 50 mV/s in N2-saturated and CO2-saturated 0.5 M KHCO3 solution, respectively, as shown in Figure 4a. In both N2 and CO2 atmosphere, two obvious reduction peaks were observed, corresponding to the in situ reduction of CuO to Cu2O and sequentially Cu2O to Cu [48]. As expected, XRD characterization could verify the formation of Cu metal from CuO/TiO2-5 catalyst after 2 h electrolysis, depicted in Figure S3. Three significant diffraction peaks for metallic Cu (denoted by solid diamond) appeared, suggesting that CuO NPs in CuO/TiO2-5 catalyst could be in situ electroreduced to metallic Cu-forming Cu/TiO2 in the electrolysis process and subsequently serves as an effective catalyst for CO2 reduction, verified by previous reports [16,36,47]. In the more negative potential region, an obvious increase of current density (J) after −0.65 V vs. RHE was shown, relating to the hydrogen evolution reaction (HER) in the N2-saturated 0.5 M KHCO3 solution (pH = 8.63, black line), while more dramatic increase of J was observed in the CO2-saturated 0.5 M KHCO3 solution (pH = 7.21, red line). To avoid the pH effect, a CV curve was recorded in the N2-saturated solution with HCl solution (pH = 7.21, same as that of the CO2-saturated 0.5 M KHCO3 solution, blue line), which shows lower J than that in CO2-saturated 0.5 M KHCO3 solution after −0.65 V vs. RHE. It demonstrates that CO2 reduction is more favorable than HER on CuO/TiO2-5.

**Figure 4.** (**a**) CV curves of CuO/TiO2-5 catalyst over glassy carbon electrode (GCE) in N2-saturated without (black line) and with (blue line) HCl solution, and CO2-saturated (red line) 0.5 M KHCO3 solution; (**b**) FEs for different products over CuO/TiO2-5 catalyst at various potentials.

The CO2ER electrocatalytic activity on CuO/TiO2-5 catalyst was also evaluated by controlled potential electrolysis (from −0.65 V to −1.05 V vs. RHE) in CO2-saturated 0.5 M KHCO3 solution. The main liquid products of the electrolysis are ethanol, acetone, and n-propanol, which were detected by 1H NMR. The overall FEs (Figure 4b) presented a sharply incremental tendency at an applied potential range from −0.65 V to −0.85 V vs. RHE and achieved a maximum of 47.4% (37.5% for ethanol, 4.3% for acetone, and 5.6% for n-propanol) at the potential of −0.85 V vs. RHE. As the potentials shift more negatively, the overall FEs significantly decreased to 18.8% at the potential of −1.05 V vs. RHE because of the competition from HER. Furthermore, a similar variation trend for FEethanol, FEacetone, and FEn-propanol was observed and depicted in detail in Figure 4b. Moreover, Figure S4 shows the total current vs. time curve for the CuO/TiO2-5 electrode at −0.85 V vs. RHE, which exhibited an initial current of 85 mA as the CuO was reduced and, subsequently, a sharp decline current in a short time. Finally, a stable current of 16 mA in the long test appeared. Notably, the FE for ethanol was maintained at approximately 35% throughout the electrolysis. This finding suggests not only efficient but also stable activity for CO2 reduction on this electrode.

We also studied the effect of CuO contents in various CuO/TiO2 catalysts on CO2ER activities, as shown in Figure 5a. All CuO/TiO2 catalysts showed two obvious reduction peaks in the CO2-saturated 0.5 M KHCO3 solution; however, the position and size of the reduction peaks varied greatly among the catalysts, which, assigned to in situ, reduce CuO to Cu. Noteworthily, in the more negative potential region (below −0.65 V vs. RHE), all CuO/TiO2 catalysts show different catalytic abilities for CO2 reduction. The J of all CuO/TiO2 catalysts at a potential of −0.85 V vs. RHE were summarized in Table S2. CuO/TiO2-5 showed the largest J. The dramatic difference of electrocatalytic performance on various CuO/TiO2 catalysts might be closely related to the composition of these catalysts. Figure 5a shows that increasing the CuO content (from 5 wt % to 60 wt %) in the CuO/TiO2 catalysts can facilitate the reduction of CO2, leading to enhanced activity for CO2 reduction. However, on the CuO/TiO2-6 (with the intended CuO content of 80 wt %) catalyst for sequentially increasing the content of CuO, CO2 reduction was subsequently suppressed. The results clearly indicate that an optimum content of CuO in the catalysts is required to achieve the maximum activity for CO2 reduction.

Potentiostatic electrolysis (at −0.85 V vs. RHE) of CO2 on various CuO/TiO2 catalysts was performed in CO2-saturated 0.5 M KHCO3 solution to further investigate the catalytic activities of CuO/TiO2 catalysts with different CuO contents. Figure 5b summarizes the FEs of ethanol, acetone, and n-propanol achieved over all CuO/TiO2 catalysts. As expected, the CuO/TiO2-5 with the intended CuO content of 60 wt % gives the highest total FE, which reached 47.4%, when compared with other CuO/TiO2 catalysts, which is in a well agreement with CV results as shown in Figure 5a. However, the total FEs decreased along with the higher CuO content (80 wt % and even 100 wt %) of CuO/TiO2, which is ascribed to the large number of CuO NPs that accumulated on the surface of TiO2, as shown in Figure S2f,g, which probably generated the decrease of active areas for CO2ER. Yet, low CuO

content (5 wt %–40 wt %) of CuO/TiO2 also exhibited slightly poor electrocatalytic performance for CO2ER due to the low content of CuO NPs. It demonstrates further that the catalytic activity for CO2ER strongly depends on the CuO contents of the catalysts as-prepared, which results in different electrochemistry surface areas (ECSA). To confirm this finding, ECSA of a variety of CuO/TiO2 electrodes were examined by CV in a potential range in which the faradaic process did not occur with the double layer capacitance in N2 atmosphere 0.1 M HClO4 solution (Figure S5 and Table S3). As expected, CuO/TiO2-5 catalyst showed a noticeable performance improvement for the double layer capacitance (Cdl), which gives the positive correlation with ECSA. The Cdl in Table S3 indicated the ECSA increased with increasing CuO content up to 60 wt %; however, a slight decrease was observed at 80 wt %, which reconfirms the CuO content of CuO/TiO2 catalysts plays a vital role in ECSA that can affect CO2ER, which is in accordance with the results of Brunauer− Emmett−Teller (BET) specific surface area (Table S4).

**Figure 5.** (**a**) CV curves of various CuO/TiO2 catalysts over GCE in CO2-saturated 0.5 M KHCO3 solution. Scan rate, 50 mV/s; (**b**) FEs for different products over various CuO/TiO2 catalysts at −0.85 V vs. RHE in CO2-saturated 0.5 M KHCO3 aqueous solution.

To further study the improved performance of CuO/TiO2-5 catalyst, we compared the electrocatalytic abilities of CO2 reduction on CuO/TiO2-5 and CuO/C catalysts. The CuO/C was successfully synthesized and characterized by XRD (Figure S6), SEM, and TEM (Figure S7). CuO/C catalyst represents the onset potential for CO2 reduction at −0.75 V vs. RHE, corresponding to a 0.10 V more negative onset potential than CuO/TiO2-5, as shown in Figure S8. This finding suggests that TiO2 as the supporter is more beneficial to facilitating CO2ER. The J in CO2-saturated solution of CuO/TiO2-5 at the potential of −0.85 V vs. RHE was 8.307 mA/cm2, which is larger than that of CuO/C (1.402 mA/cm2). Additionally, potentiostatic electrolysis in CO2-saturated 0.5 M KHCO3 solution for the CuO/C was performed at the potential of −0.85 V vs. RHE, which shows poor active towards CO2 reduction. Only 16.3% for overall FE (12.9% for ethanol, 1.1% for acetone, and 2.3% for n-propanol) was analyzed by 1H NMR. In contrast, CuO/TiO2-5 catalyst presents excellent activity for CO2 reduction, which achieved 47.4% for overall FE. Besides, as seen from Figure S5 and Table S3, the ECSA of CuO/TiO2-5 reveals a larger value than the one of CuO/C. As is well known, CO2 adsorption on the active sites is the prerequisite for subsequent CO2 reduction reaction. The relatively larger amount of CO2 adsorption on active sites may offer more original reactants such as CO2 [49]. The CO2 adsorption abilities of CuO/TiO2-5 and CuO/C were evaluated by CO2 adsorption, which were displayed in Figure S9. CuO/TiO2-5 catalyst possessed a more remarkably improved CO2 adsorption capacity than CuO/C, which was 5.80 mg/g and 1.34 mg/g, respectively. It is reasonable to assume that the enhanced CO2 adsorption capacity can make a significant contribution to the superior performance of CO2 reduction. By comparing the electrochemical and material characterizations of CuO/TiO2 with those of CuO/C, we gained insight that TiO2 plays an important role in promoting the electrocatalytic performance of CuO/TiO2. Hence, on the basis of the above analysis, a brief illustration of CO2ER over CuO/TiO2 catalysts is stated. Large amount of CO2 is first adsorbed

on TiO2. CuO/TiO2 catalysts could be in situ electroreduced to Cu/TiO2 at less negative cathode potentials than the ones of CO2ER. Then, the adsorbed CO2 gets one electron from the achieved Cu/TiO2 electrode and is converted to CO2 −, which can be dimerized to \*C2O2 −. Notably, the C-C bond-making step on the Cu surface is a key step for CO2 reduction to ethanol and even the C3 product (acetone and n-propanol) [50]. Interestingly, the catalytic ability of CuO/TiO2-5 at −0.85 V vs. RHE expresses more obvious performance than the one of 40 wt % Cu/TiO2 (the optimal catalyst in X wt % Cu/TiO2 system) at −1.45 V vs. RHE [34], which indicates that in situ electroreduced Cu NPs have greater activity to catalyze CO2 reduction. The low-cost CuO/TiO2 catalyst is an efficient alternative to expensive materials for the application of CO2ER in industry.

#### **3. Materials and Methods**

#### *3.1. Materials and Instruments*

Cupric acetate monohydrate [Cu(Ac)2·H2O, Analytical Reagent (AR) grade], ammonium carbonate [(NH4)2CO3, AR grade], potassium bicarbonate [KHCO3, AR grade], and titanium dioxide [TiO2, AR grade] were purchased from Sinopharm Chemical Reagent Co. (SCR, Shanghai, China) with 99% purity and used as received. Nafion® 117 solution (5%) and Nafion® 117 membrane were obtained from Dupont (Wilmington, DE, USA). Carbon paper (CP, HCP010) and conductive carbon black (C, VXC72R) were purchased from Shanghai Hesen Electrical CO. (Shanghai, China).

Crystal-phase X-ray diffraction (XRD) patterns of CuO/TiO2 catalysts were recorded using an Ultima IV X-ray powder diffractometer (Kurary, Tokyo, Japan) equipped with Cu Kα radiation (*k* = 1.5406 Å). The values for actual CuO loadings of the synthesized catalysts were determined on an inductively coupled plasma atomic emission spectroscopy (ICP-AES, IRIS Intrepid II XPS, Waltham, MA, USA). The scanning electron microscope (SEM) images of the catalysts were obtained using a Hitachi S-4800 field-emission scanning electron microscope (Tokyo, Japan). Transmission electron microscopy (TEM) patterns were recorded by TECNAI G2F30 transmission electron microscope (Tokyo, Japan). The X-ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific, Waltham, MA, USA) analysis was performed on the Thermo Scientific ESCA Lab 250Xi using 200 W monochromatic Al Kα radiation. The 500 μm X-ray spot was used. The base pressure in the analysis chamber was about <sup>3</sup> × <sup>10</sup>−<sup>10</sup> mbar. Typically, the hydrocarbon C1s line at 284.8 eV from adventitious carbon was used for energy referencing. Brunauer−Emmett−Teller (BET) specific surface area was characterized by nitrogen adsorption in a BELSORP-MAX instrument (MicrotracBEL, Tokyo, Japan) after outgassing the samples for 10 h under vacuum at 573 K. CO2 adsorption was obtained in a BELSORP-MAX instrument (MicrotracBEL, Japan) under CO2 atmosphere at 298 K. Potentiostatic electrolysis and cyclic voltammetry (CV) were performed using a CHI 660 C electrochemical station (Shanghai Chenhua Instrument Co. Ltd., Shanghai, China).

#### *3.2. Materials Synthesis*

A simple and mild hydrothermal synthesis method was used to prepare six different weight ratios of CuO/TiO2 catalysts, which are defined as CuO/TiO2-1, CuO/TiO2-2, CuO/TiO2-3, CuO/TiO2-4, CuO/TiO2-5, and CuO/TiO2-6, corresponding to the CuO content of 5 wt %, 10 wt %, 20 wt %, 40 wt %, 60 wt %, and 80 wt %, respectively. Typically, due to similar fabrication procedure, for simplicity, CuO/TiO2-5 acts as the specimen to illustrate the following experiment. The detailed synthesis procedure for CuO/TiO2-5 is depicted as follows: 0.05 M of Cu(Ac)2 aqueous solution (60 mL) was mixed with 127 mg of TiO2. After stirring for several minutes, an appropriate amount of 0.05 M of (NH4)2CO3 aqueous solution was added drop-wise with stirring to control the synthetic rate. Then, the mixture was stirred gently for 3 h at the ambient temperature. Subsequently, the above mixture was centrifuged, washed with distilled water, and stored in a vacuum oven with 60 ◦C for 4 h to dry the precipitate. Finally, the collected powder was transferred to a crucible, further maintaining at 220 ◦C for 3 h to obtain the final CuO/TiO2 powder. Pure CuO catalyst was acquired by the same

method, without adding TiO2. Additionally, CuO/C catalyst with the CuO content of 60 wt % got through the similar synthetic process, but with C replacing TiO2.

#### *3.3. Electrode Preparation and Electrochemical Test*

Potentiostatic electrolysis was carried out in an H-type three-electrode cell with a piece of Nafion® 117 cation exchange membrane (H+ form) as a separator. The working electrode was manufactured using the following route: 10 mg of CuO/TiO2 catalyst was suspended in a mixture solution with 20 μL of 5 wt % Nafion® solution and 40 μL of distilled water. After sonification for 20 min, the mixture was spread on a porous CP (2 × 2 cm) by micropipette and then dried in air. The counter and reference electrodes were Pt sheet and SCE, respectively. A 0.5 M of KHCO3 aqueous solution serves as electrolyte, which was saturated with CO2 by bubbling for 30 min before the electrolysis experiment. CO2 was bubbled continuously throughout overall experiment time. CV tests were measured in a single cell system using a standard three-electrode setup at 0.5 M KHCO3 aqueous solution under CO2 and N2 atmosphere with a scan rate of 50 mV/s, respectively. In this system, 2 μL of the above prepared catalyst suspension was coated on glassy carbon electrode (GCE, diameter = 2 mm) playing as the working electrode, and a Pt mesh and an SCE acted as the counter electrode and the reference electrode, respectively. All potentials were based on RHE as reference potentials, and converted by the following equation:

$$E \text{ (vs. RFHE)} = E \text{ (vs. SCE)} + 0.059 \times \text{pH} + 0.241...$$

#### *3.4. Product Analysis*

Liquid phase products were analyzed by 1H-nuclear magnetic resonance (NMR) spectra recorded on an Ascend 400 (400 MHz, Bruker, Germany) spectrometer in D2O with Me4Si as an internal standard.

#### **4. Conclusions**

In this study, a simple synthesis strategy was employed to fabricate CuO/TiO2 material. A series of characterizations has demonstrated that CuO NPs were uniformly distributed on the surface of TiO2. Different CuO/TiO2 catalysts displayed dramatically different electrocatalytic performance for CO2ER, depending strongly on the CuO contents of the catalysts as-prepared. High FEs of CO2 reduction products (ethanol, acetone, and n-propanol) reach up to 47.4% on CuO/TiO2-5 at the potential of −0.85 V vs. RHE among all CuO/TiO2 and CuO/C catalysts. The CO2 adsorption ability test suggests that CuO/TiO2-5 exhibits more superior CO2 adsorption performance than CuO/C, indicating that CuO/TiO2-5 is more beneficial to CO2 reduction. The fact that the as-synthesized CuO/TiO2 catalysts exhibit high activity for CO2ER is attributed to the special roles of TiO2 that show excellent CO2 adsorption ability. Our work may provide some concepts for designing cheap and effective catalysts for highly efficient CO2ER.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/8/4/171/, Table S1: CuO composition of the synthesized CuO/TiO2 catalysts. Figure S1: The high-resolution XPS spectrum of C1s for CuO/TiO2-5 catalyst. Figure S2: SEM images of (a) pure TiO2, (b) CuO/TiO2-1, (c) CuO/TiO2-2, (d) CuO/TiO2-3, (e) CuO/TiO2-4, (f) CuO/TiO2-5, (g) CuO/TiO2-6, and (h) pure CuO. Figure S3: XRD patterns of CuO/TiO2-5 catalyst after 2 h electrolysis. The solid diamond represents the in situ generated Cu metal. Figure S4: CO2 reduction electrolysis data at −0.85 V vs. RHE for CuO-TiO2-5 catalyst. Table S2: The current density (J) for all the CuO/TiO2 catalysts in CO2-saturated environment at the potential of −0.85 V vs. RHE. Figure S5: Plot of I vs. potential of (A) CuO/TiO2-1, (B) CuO/TiO2-2, (C) CuO/TiO2-3, (D) CuO/TiO2-4, (E) CuO/TiO2-5, (F) CuO/TiO2-6, (G) pure CuO, and (H) CuO/C in N2-saturated 0.1 M HClO4 solution cycled between 0.35 and 0.55 V vs. SCE at scan rates in the range of 5–60 mV/s. Insets show plot of Ic vs. ν in which the linear regressions give capacitance information. Table S3: The analysis of Cdl of various CuO/TiO2 and CuO/C catalysts. Table S4: The analysis of specific surface area of CuO/TiO2 catalysts. Figure S6: XRD patterns of CuO/C and C catalysts. Figure S7: SEM (a) and TEM (b) images of CuO/C catalyst. Figure S8: CV curve of CuO/C catalyst over glassy carbon electrode in CO2-saturated (red line) 0.5 M KHCO3 solution. Figure S9: CO2 adsorption-desorption isotherms at 297 K of CuO/TiO2-5 (red square) and CuO/C (blue triangle). Filled and empty symbols represent adsorption and desorption, respectively.

**Acknowledgments:** We gratefully acknowledge the financial support from the National Natural Science Foundation of China (21673078, 21473060, 21773071).

**Author Contributions:** Jing Yuan and Jing-Jie Zhang conceived and designed the experiments; Jing Yuan and Wang-Jun Meng performed the experiments; Jing Yuan, Man-Ping Yang, and Huan Wang analyzed the data; Jing Yuan, Huan Wang, and Jia-Xing Lu wrote the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Hydrogen Production from Chemical Looping Reforming of Ethanol Using Ni/CeO2 Nanorod Oxygen Carrier**

#### **Lin Li, Bo Jiang, Dawei Tang \*, Zhouwei Zheng and Cong Zhao**

School of Energy and Power Engineering, Key Laboratory of Ocean Energy Utilization and Energy Conservation of Ministry of Education, Dalian University of Technology, Dalian 116024, China; lilinnd@dlut.edu.cn (L.L.); ericchiang@mail.dlut.edu.cn (B.J.); 953596340zzw@mail.dlut.edu.cn (Z.Z.); zc287643861@mail.dlut.edu.cn (C.Z.) **\*** Correspondence: dwtang@dlut.edu.cn; Tel.: +86-411-8470-8460

Received: 8 June 2018; Accepted: 22 June 2018; Published: 25 June 2018

**Abstract:** Chemical looping reforming (CLR) technique is a prospective option for hydrogen production. Improving oxygen mobility and sintering resistance are still the main challenges of the development of high-performance oxygen carriers (OCs) in the CLR process. This paper explores the performance of Ni/CeO2 nanorod (NR) as an OC in CLR of ethanol. Various characterization methods such as N2 adsorption-desorption, X-ray diffraction (XRD), Raman spectra, transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), H2 temperature-programmed reduction (TPR), and H2 chemisorption were utilized to study the properties of fresh OCs. The characterization results show the Ni/CeO2-NR possesses high Ni dispersion, abundant oxygen vacancies, and strong metal-support interaction. The performance of prepared OCs was tested in a packed-bed reactor. H2 selectivity of 80% was achieved by Ni/CeO2-NR in 10-cycle stability test. The small particle size and abundant oxygen vacancies contributed to the water gas shift reaction, improving the catalytic activity. The covered interfacial Ni atoms closely anchored on the underlying surface oxygen vacancies on the (111) facets of CeO2-NR, enhancing the anti-sintering capability. Moreover, the strong oxygen mobility of CeO2-NR also effectively eliminated surface coke on the Ni particle surface.

**Keywords:** chemical looping reforming; hydrogen; oxygen carrier; CeO2; nanorod

#### **1. Introduction**

Hydrogen is considered an efficient energy carrier that is environmentally benign [1]. Chemical looping reforming (CLR) is a prospective alternative for hydrogen production due to its energy efficiency and inherent CO2 capture [2,3]. The fixed-bed reactor configuration CLR process (Figure 1) is performed by alternatively switching the feed gases; the oxygen carriers (OCs) are stationary and periodically exposed to redox atmosphere [4]. The key to developing a CLR process is to screen high-performance OCs. Ni-based OCs have been widely investigated because of their ability for carbon–carbon and carbon–hydrogen bonds cleavage [5–7]. Zafar et al. [8] prepared a series of OCs including Fe, Cu, Mn, and Ni supported by SiO2 and MgAl2O4 and concluded that NiO supported on SiO2 exhibited high H2 selectivity in CLR. Löfberg et al. [9] demonstrated that Ni plays two essential roles in CLR process, i.e., the anticipated activation of reactants as well as the regulation of the oxygen supply rate from solids. Dharanipragada et al. [10] reported that the Ni-ferrites OC suffers from the loss of oxygen storage capacity due to Ni sintering in chemical looping with alcohols. Improving oxygen mobility and sintering resistance remain the major challenges of the development of Ni-based OCs in the CLR process due to the high activation energy of oxygen anion diffusion in NiO (2.23 eV in CLR) and the low Tammann temperature of Ni (691 ◦C) [11–14].

**Figure 1.** Schematic description of CLR of ethanol.

Recent research has revealed that CeO2 exhibits excellent redox property due to abundant oxygen storage capacity and the strong capability to stabilize Ni nanoparticles because of strong metal support interaction (MSI) [15,16]. CeO2 can readily release lattice oxygen under reducing conditions, thus creating oxygen vacancies which are associated with oxygen mobility [17]. The MSI between the Ni and CeO2 could tune the physiochemical properties of Ni, contributing to high catalytic reactivity and stability [18]. Jiang et al. [19] proposed that the oxygen vacancies of CeO2 could effectively eliminate surface coke deposition, activate steam, and shorten the "dead time" in the CLR process. Dou et al. [20] revealed that the surface oxygen originating from the CeO2 lattice can oxidize coke precursors, keeping the OC surface free of coke deposition. It has been reported that MSI and the mobility of lattice oxygen show strong dependence on the morphology of CeO2 [21]. Lykaki et al. [22] have demonstrated CeO2 morphology dominates reducibility and oxygen mobility, which follow the sequence: nanorod (NR) > nanopolyhedra (NP) > nanocube (NC). Moreover, the apparent activation energy of these three CeO2 shapes for the CO oxidation in a hydrogen-rich gas shows the opposite trend, implying the potential highest water gas shift (WGS) activity of NR [23].

Therefore, in this work, a Ni/CeO2-NR OC for CLR of ethanol for hydrogen production is prepared by a hydrothermal method. The physicochemical properties are investigated by N2 adsorption-desorption, X-ray diffraction (XRD), inductively coupled plasma optical emission spectroscopy (ICP-OES), Raman spectra, high resolution transmission electron microscopy (HRTEM), X-ray photoelectron spectroscopy (XPS), and H2 temperature-programmed reduction (H2-TPR). The performances of the Ni/CeO2-NR OC are tested in a packed-bed reactor and compared with the other reference bulk OC, i.e., Ni/CeO2.

#### **2. Results and discussion**

#### *2.1. Characterization of OCs*

Predominant physicochemical properties of fresh Ni/CeO2-NR and CeO2-NR are tabulated in Table 1. The introduction of Ni species has a pronounced influence on the texture of CeO2-NR. The CeO2-NR support showed higher Brunauer-Emmett-Teller (BET) surface area than that of Ni/CeO2-NR. The average pore diameter and pore volume also exhibited the same trends. It has been reported that the BET surface of CeO2-NR decreased by 13–18% after the incorporation of 7.5 wt % CuO [22]. The actual Ni content of Ni/CeO2-NR determined by ICP-OES was 9.7 wt %.


**Table 1.** Physicochemical properties of fresh Ni/CeO2-NR and CeO2-NR.

1. Determined by ICP-OES; 2. Determined by H2 chemisorption; 3. Determined by XRD from Ni (111) and CeO2 (111) plane; 4. Calculated from the HRTEM images.

The XRD profiles of fresh Ni/CeO2-NR and CeO2-NR OCs are shown in Figure 2. The reflection peaks at 28.5◦, 33.1◦, 47.5◦, 56.3◦, 59.1◦, 69.4◦, 76.7◦, and 79.1◦ could be indexed to (111), (200), (220), (311), (222), (400), (331), (420), and (422) planes of a fluorite-structure CeO2 (Space group: *Fm3m*), respectively. There were no other impurity peaks of the hexagonal structure of Ce(OH)3 or Ce(OH)CO3 detected, indicating the excellent crystalline purity of CeO2. Two peaks at 44.5◦ and 51.8◦, which could be respectively attributed to (111) and (200) facets of Ni, were observed for Ni/CeO2-NR. The mean crystal sizes of Ni and CeO2 were obtained from the Scherrer Equation (listed in Table 1). The crystal sizes of CeO2 for CeO2-NR and Ni/CeO2-NR were 14.8 and 16.1 nm, respectively, indicating that the incorporation of Ni would not significantly change the structure characteristics of CeO2 support. Similar crystallite sizes of CeO2-NR prepared by the hydrothermal method have been reported in other work [22,24].

**Figure 2.** XRD profiles of Ni/CeO2-NR and CeO2-NR.

Raman spectroscopy was employed to characterize surface and bulk defects. As shown in Figure 3, both samples present four characteristic peaks. The appreciable peak at 462.8 cm−<sup>1</sup> resulted from the first order F2g mode of the fluorite cubic structure. The Raman shift at 596.1 cm−<sup>1</sup> could be assigned to the defect-induced band (D band). The relocation of O atom from the interior of tetrahedral cationic sub-lattice to the interior of ideally empty octahedral cationic sites (Frenkel interstitial sites) would lead to the deformation of the anionic lattice of CeO2 [25]. The intensity of the D band is a gauge of the distortion of ionic lattice, which results in punctual defects and oxygen vacancies [22]. Therefore, the value of *ID*/*IF*<sup>2</sup>*<sup>g</sup>* is commensurate with the number of defect sites in ceria. The relative intensity of *ID*/*IF*<sup>2</sup>*<sup>g</sup>* decreased after the incorporation of Ni, suggesting NiO species suppress the surface oxygen vacancy of CeO2-NR. Li et al. [26] reported that Vanadium atom would bond to the surface of CeO2-NR by generating V–O–Ce species, thus covering oxygen defects and stabilizing adjacent Ce atoms. The inconspicuous peak at 258.1 cm−<sup>1</sup> results from the second order transverse acoustic mode (2TA) or doubly degenerate transverse optical mode (TO), which are Raman inactive in a perfect crystal [27]. The peak at 1179.9 cm−<sup>1</sup> may be related to the stretching mode of the short terminal Ce=O. Moreover, the Raman shift of Ni/CeO2-NR moved towards a low wavenumber in comparison with that of CeO2-NR, indicating the incorporation of Ni affects the symmetry of Ce–O bonds [21].

**Figure 3.** Visual Raman spectroscopy of Ni/CeO2-NR and CeO2-NR.

The TEM of CeO2-NR and Ni/CeO2-NR are illustrated in Figure 4. The Ni/CeO2-NR maintained the original nanorod shape after Ni incorporation. The CeO2-NR was approximately 14 nm in diameter and several hundreds nm in length. As shown in Figure 4a, the calculated interference fringe spacings (d) are 0.27 and 0.31 nm, revealing the CeO2-NR predominantly exposes the (100) and (111) facets. The preferable exposure crystal facets of CeO2-NR are consistent with those of samples derived from the CeCl3 precursor in other studies [28,29]. Theory calculation has demonstrated that the (111) is the least reactive facet, followed by (100) and (110). In addition to the preferable crystal facets, there are other aspects degerming CeO2 activity. Several 'dark pits' are shown in the box of Figure 4. Liu et al. [29] have reported that these dark pits are related to the surface reconstruction and defects; their study also concluded that these defects play a more important role in determining the CeO2 activity than the exposure planes. Sayle et al. [30] have performed a molecular dynamic modeling to simulate the synthesis of CeO2-NR and discovered that the atomistic sphere model exhibits many steps on the (111) planes of CeO2-NR. The migration of oxygen in CeO2 occurs by a vacancy hopping mechanism; therefore, the clusters of defects are conducive to oxygen transfer. If the diffusion rate of oxygen anions becomes adequately high, a consecutive oxygen flow is generated, resulting in high reducibility. The Ni particle size distribution, which is calculated from 52 particles, was inserted in Figure 4, and the average particle size was 8.1 nm (shown in Table 1). Shen et al. [31] have elucidated that the strong interfacial anchoring effect, which exists between the surface oxygen vacancies on (111) planes of CeO2-NR and the gold particles, only allows the gold particles to locally rotate or vibrate but not to migrate to form aggregates. Therefore, the strong MSI would significantly improve the Ni dispersion on the CeO2-NR surface, resulting in a small particle size of Ni.

**Figure 4.** (**a**) HRTEM image of CeO2-NR, (**b**) TEM image of CeO2-NR, (**c**) and (**d**) TEM images of Ni/CeO2-NR.

XPS was employed to investigate the valences of Ce and Ni cations. The XPS spectra of Ce 3d of Ni/CeO2-NR (shown in Figure 5a) could be deconvoluted into two spin-orbit series, i.e., 3d5/2 (*u*) and 3d3/2 (*v*). The multiplet splitting components labeled *u0, u1, u2, v, v1* and *v2* corresponds to the 3d104f0 state of Ce4+, while the *u0* and *v*<sup>0</sup> are related to 3d104f1 state of Ce3+. These two Ce species in Ni/CeO2-NR indicate the OC surface was partly reduced because of oxygen desorption and the formation of oxygen vacancies. It is widely accepted that oxygen vacancies are produced to maintain electrostatic balance once Ce3+ exists in fluorite Ce (Equation 1). The percentage of Ce3+ cations to the total Ce cations is determined by the area ratio of different Ce species in XPS spectra. The Ce3+ ratio of Ni/CeO2-NR (16.9%) was lower than that of pure CeO2-NR (24.3%) reported in Lykaki et al. [22], suggesting NiO species could inhibit the formation of surface-unsaturated Ce3+. The decrease in surface Ce3+ species of Ni/CeO2-NR was consistent in the trend of *ID*/*IF*<sup>2</sup>*<sup>g</sup>* in the Raman result. The Ni 2p XPS spectra (illustrated in Figure 5b) were characterized by two spin-orbit groups, i.e., 2p3/2 (855.2 and 856.4 eV) and 2p1/2 (873.3 eV), as well as a shake-up peak at 861.8 eV. The photoelectron peak of 2p3/2 over Ni/CeO2-NR shifted towards high-binding energy in contrast to those over pure NiO (844.4 eV, reported in Lemonidou et al. [32]) and Ni/CeO2 (854.5 eV, reported in Tang et al. [33]). This result implies there is an enhanced MSI between Ni and CeO2-NR. The Ni/Ce atom ratio of the outer surface of CeO2-NR (0.49) is higher than the nominal one (0.32), suggesting a Ni species enrichment from bulk to surface.

$$4\text{Ce}^{4+} + \text{O}^{2-} \rightarrow 4\text{Ce}^{4+} + \frac{2\text{e}^-}{\delta} + 0.5\text{O}\_2 \rightarrow 2\text{Ce}^{4+} + 2\text{Ce}^{3+} + \delta + 0.5\text{O}\_2\tag{1}$$

where *δ* represents an empty position derived from the removal of O2<sup>−</sup> from an oxygen tetrahedral site (Ce4O).

**Figure 5.** (**a**) XPS spectra of Ce 3d; (**b**) XPS spectra of Ni 2p.

H2-TPR was performed to investigate MSI. As illustrated in Figure 6, the H2-TPR patterns of CeO2-NR are comprised of two reduction peaks. An inconspicuous peak at 262 ◦C, which can be assigned to the reduction of surface-adsorbed oxygen, was observed, while a broad peak ranging from 400 ◦C to 550 ◦C could be ascribed to the reduction of Ce4+ to Ce3+ [22,34]. With regard to Ni/CeO2-NR, the TPR profiles consisted of three peaks. The small peak at 220 ◦C may have been attributable to the reduction of the NiO species, which slightly interacted with CeO2-NR supports. Zhang et al. [21] have reported that this NiO species is characterized by small radius and could incorporate into CeO2-NR surface. The second peak at 356 ◦C could be ascribed to the reduction of NiO species, with a strong interaction with CeO2-NR supports. The H2 consumption of the second peak was highest among the three peaks, indicating most of the NiO strongly interacted with the support. The third peak at 494 ◦C was also attributed to the reduction of Ce4+ to Ce3+. However, the broad reduction peak of Ni/CeO2-NR shifted to a lower temperature than that of pure CeO2-NR, suggesting the introduction of Ni improves the reducibility of CeO2-NR. It has demonstrated that the promoted reduction behavior is caused by the generated Ni-Ce-O species which improves the deformation of CeO2 [35].

**Figure 6.** H2-TPR patterns of Ni/CeO2-NR and CeO2-NR.

#### *2.2. Activity Tests of OCs*

The activity of Ni/CeO2-NR was tested in CLR of ethanol and compared with Ni/CeO2. Figure 7 displays the performance of both OCs in CLR: H2 selectivity and ethanol conversion at the fuel feed step as well as the concentration of CO, CO2 and O2 at the air feed step. The H2 selectivity and ethanol conversion of both OCs remained relatively stable during the process. With regard to Ni/CeO2-NR, the average ethanol conversion remained at 88.0%, and the average H2 selectivity was 78.9%. However, with the average ethanol conversion and H2 selectivity of conventional Ni/CeO2 at 72.9% and 60.5% respectively, the activity of Ni/CeO2-NR was superior to the Ni/CeO2. As the Ni content and other experimental conditions were the same for both OCs, the improved activity therefore resulted from the support.

**Figure 7.** Activity tests of (**a**) Ni/CeO2-NR and (**b**) Ni/CeO2.

Various factors contributed to the enhanced catalytic activity. Notably, the CeO2-NR with high specific-to-volume ratio guarantees excellent Ni dispersion, as evidenced by TEM and TPR results. Such a one-dimensional nanostructure enables uniform and small Ni nanoparticles to be finely dispersed on supports, thus generating many accessible catalytic active sites. Additionally, the high concentration oxygen vacancies of CeO2-NR, which was proven via Raman and XPS analysis, can activate and produce OH groups from steam. H2 and CO2 can be generated via the reaction between the formed OH groups as well as intermediate species, thus improving H2 selectivity. It has been proposed that the interface of the metal/CeO2 is the main site for steam reforming reaction [36].

The air feed process is highly oxygen consuming and is accompanied by the generation of CO2 and CO due to coke oxidation and partial oxidation. As illustrated in Figure 7, all the samples exhibited CO2 and CO evolution at the air feed stage. The integration areas of C-containing gas concentrations corresponded to the quantity of carbon deposition. Ni/CeO2-NR exhibited small peak areas compared with the Ni/CeO2 OC, indicating that the carbon deposition on Ni/CeO2-NR was effectively suppressed. This is a result of the abundant oxygen vacancies and its strong oxygen storage capacity, which enhanced the oxygen mobility and thereby facilitated the removal of carbon deposition. In addition, the depleted CeO2-NR could be partially replenished by the air feed step. The O2 concentration increased from 0%to 23% at the end of the air feed stage, indicating the end of coke elimination and replenishment of lattice oxygen.

#### *2.3. Stability Tests*

The durability of Ni/CeO2-NR and Ni/CeO2 was tested in a 10-cycle CLR process. As shown in Figure 8, the H2 selectivity of Ni/CeO2 slowly declined with the increase of the running cycles, while the Ni/CeO2 NR maintained its activity throughout the test. During the tests, the H2 selectivity of Ni/CeO2-NR decreased from 81.7% to 79.2%, while Ni/CeO2 exhibited a significant decrease in H2

selectivity from 61.8% to 51.4%. The deactivation of OCs predominantly caused Ni sintering and carbon deposition. Notably, the Ni/CeO2-NR OC exhibited a more durable performance than the Ni/CeO2, revealing that the CeO2-NR supported metal was more resistant to deactivation. The superior durability of the Ni/CeO2-NR sample resulted from the strong MSI between Ni and CeO2-NR support, which improved the metal-sintering resistance. The abundant oxygen vacancies of CeO2-NR are essential units for the anchoring of metal particles. The essential role of CeO2 is to disperse and stabilize Ni particles over its surface oxygen vacancies, which depend on the morphology of CeO2. As evidenced by XPS, the (111) plane of CeO2-NR plays an important role in anchoring the Ni particles. It has been proposed that the interfacial Au atoms which are located away from the particle perimeter (covered interfacial atoms) would closely interact with the underlying surface oxygen vacancies on the (111) facets of CeO2-NR [31]. Because this interfacial region was not involved in the reforming reaction, this strong interaction would effectively stabilize the Ni particles on CeO2-NR. Additionally, small size Ni particles can lower the driving force for coke diffusion and thereby help to reduce the carbon deposition. Furthermore, the strong oxygen mobility of CeO2-NR is indispensable to removing the carbon deposition at the metal surface. It has been demonstrated that the oxygen deposited in CeO2 lattice can react with the carbon species left over from the steam reforming reaction [37]. Overall, CeO2-NR support not only promotes the anti-sintering capability of OCs but also reduces the carbon deposition, thus improving the durability of OCs. A comparison of several investigations regarding CLR over different Ni-based OCs is tabulated in Table 2. The Ni/CeO2-NR in this work showed higher average fuel conversion in long-term tests despite the low Ni loading, indicating the strong sintering resistance and high reforming activity of this OC.

**Figure 8.** Stability tests of OCs.



#### **3. Materials and Methods**

#### *3.1. Preparation of OCs*

CeO2-NR was synthesized on the basis of our previous work with small modifications [38]. Firstly, 1 g of CeCl3 (99.9%, Aladdin, Shanghai, China) and 19.3 g of NaOH (97%, Aladdin) were dissolved in 50 mL deionized water and continuously stirred for 25 min. The resulting mixture was then put into a Teflon-lined autoclave and kept at 100 ◦C for 24 h. Subsequently, the sample was separated, washed, then dried at 85 ◦C for 10 h and finally calcined at 400 ◦C for another five hours. The Ni/CeO2-NR oxygen carrier was synthesized by a wet-impregnation method. Ni(NO3)2·6H2O (98%, Aladdin) of 0.2 g, equivalent to 10 wt % Ni loading, was first dispersed in 5 mL deionized water, and the prepared CeO2-NR of 0.5 g was added into the solution. The mixture was stirred under sonication for three hours at 60 ◦C and then dried at 85 ◦C for 10 h. The as-synthesized sample was then calcined at 700 ◦C for two hours.

The other reference bulk Ni/CeO2 was also synthesized by a wet-impregnation method, and the procedure has been reported in Xu et al. [39]. The Ni content of this OC was also set to 10 wt %.

#### *3.2. Characterization of OCs*

The texuture of OCs were analyzed by a Micrometric Acusorb 2100E apparatus (Ottawa, ON, Canada) at 77 K. The OCs were degassed at 573 K for three hours before tests.

ICP-OES (Nijmegen, The Netherlands) was applied to measure the elemental composition. Before tests, the sample was solved by hydrofluoric acid solution.

XRD was performed, using a Shimadzu XRD-600 instrument (Kyoto, Japan), to identify the phase composition. The scanning 2*θ* degrees ranged from 10◦ to 80◦, and the scanning rate was set to 4◦/min. A graphite-filtered Cu Kα radiation (λ = 1.5406 Å) was applied as a radiation source.

Raman spectra were employed by a Renishaw Spectroscopy (Gloucestershire, UK) with a visible 514 nm Ar-ion laser under ambient condtions. During the mesurement, the flowing gas was He, and the test temperature was maintained at 300 ◦C.

TEM was conducted, using FEI Tecnai F30 (Cleveland, OH, USA), to investigate the morphology of the Ocs. The samples were first solved in ethanol under sonication, followed by dispersal on a copper grid-supported carbon foil and dried in air.

XPS was carried out by a ThermoFisher K-Alpha system with a 150 W Al Kα source (Waltham, MA, USA). The samples were placed on the holder, and the scanning step was set to 0.15 eV.

H2-TPR was employed by a Micrometrics AutoChem 2920 instrument (Ottawa, ON, Canada). In a typical procedure, a sample of 90 mg was preheated at 450 ◦C for one hour in an Ar flow of 35 mL/min and subsequently cooled to 80 ◦C. A mixture of 10 vol % H2 in Ar flow (35 mL/min) was then inserted, and the temperature was increased from 80 ◦C to 100 ◦C at a rate of 5 ◦C/min simultaneously. H2 chemisorption was also performed by the same apparatus to analyze the Ni dispersion. The sample was first reduced at 700 ◦C in a H2/Ar flow (30 mL/min) for one hour, followed by cooling to 100 ◦C under Ar flow. Subsequently, H2 pulses were introduced until the eluted peaks of successive pulses became steady. The Ni active surface area was obtained from the H2 adsorbption volume considering the stoichiometric ratio Hadsorbed/Nisurface = 1 and a surface area of 6.5 × <sup>10</sup>−<sup>20</sup> m2 per Ni atom [40].

#### *3.3. Activity and Stability Tests*

The performance of CLR of ethanol by Ni/CeO2-NR and reference conventional bulk OCs, i.e., Ni/CeO2, were conducted in a packed-bed reactor, whose schematic was described in our previous work [41,42]. The tested OCs of 0.5 g were loaded at the center of the quartz reactor. During a fuel feed step, an ethanol solution (4 mL/h) with a steam to carbon ratio (S/C) of four was preheated to 150 ◦C and then inserted into the reactor in a N2 flow of 180 mL/min. The reacter temperature of the fuel feed step was 650 ◦C. An Agilent 7890A (Santa Clara, CA, USA) chromatography with two detectors, thermal conductivity detector (TCD) and flame ionization detector (FID), was applied to verify the effluents. TCD with a TDX-01 column was applied to detect N2, H2, CO, CO2, and CH4; the FID with a Porapak-Q column was applied to measure the concentration of C3H8O3, H2O, and CH3CHO. In an air feed step, the air flow was 100 mL/min, and the reaction temperature was also 650 ◦C. The exhausted gases were detected by another GC (Agilent 7890A) with a TCD detector. A 5A molecular sieve column was used to detect the O2, and the TDX-01 column was applied to detect CO2, CO, and N2. The durations of the fuel feed step and the air feed step were 60 and 10 min, respectively. A N2 purge process of five minutes was performed between the fuel feed step and the air feed step to eliminate the residue gas in the reactor.

The conversion and H2 selectivity were calculated as follows:

$$\text{X} = \frac{F\_{\text{in}} - F\_{\text{out}}}{F\_{\text{in}}} \times 100\% \tag{2}$$

$$S\_{H\_2} = \frac{1}{6} \times \frac{moles \text{ } H\_2 \text{ produced}}{moles \text{ } ethanol \text{ } feed \times X} \times 100\% \tag{3}$$

#### **4. Conclusions**

A Ni/CeO2-NR OC was synthesized by hydrothermal method and tested in CLR of ethanol process in this work. H2 selectivity of 80% was achieved by Ni/CeO2-NR in a 10-cycle stability test. The characterization results show that the Ni/CeO2-NR possesses high Ni dispersion, abundant oxygen vacancies, and strong MSI. The small particle size and abundant oxygen vacancies contributed to the WGS reaction, thus improving the catalytic activity. The buried interfacial Ni atoms strongly anchored on the underlying surface oxygen vacancies on the (111) facets of CeO2-NR, therefore enhancing the anti-sintering capability. Moreover, the strong oxygen mobility of CeO2-NR also effectively eliminated surface coke on the Ni particle surface.

**Author Contributions:** Experiment, L.L., Z.Z. and C.Z.; Data Curation, Z.Z. and C.Z.; Writing-Original Draft Preparation, L.L. and B.J.; Writing-Review & Editing, B.J.; Supervision, D.T. and L.L.; Project Administration, D.T. and L.L.; Funding Acquisition, L.L. and D.T.

**Funding:** This research was funded by (National Natural Science Foundation of China) grant number (51706030), (Fundamental Research Funds for Central Universities) grant number (DUT18JC11) and (China Postdoctoral Science Foundation) grant number (2017M611219).

**Acknowledgments:** We deeply appreciate the kind assistance from the Key Laboratory of Ocean Energy Utilization and Energy Conservation of Ministry of Education (China).

**Conflicts of Interest:** The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Experimental Clarification of the RWGS Reaction Effect in H2O/CO2 SOEC Co-Electrolysis Conditions**

#### **Evangelia Ioannidou 1,2, Stylianos Neophytides <sup>1</sup> and Dimitrios K. Niakolas 1,\***


Received: 31 December 2018; Accepted: 29 January 2019; Published: 2 February 2019

**Abstract:** In the present investigation, modified X-Ni/GDC electrodes (where X = Au, Mo, and Fe) are studied, in the form of half-electrolyte supported cells, for their performance in the RWGS through catalytic-kinetic measurements. The samples were tested at open circuit potential conditions in order to elucidate their catalytic activity towards the production of CO (rco), which is one of the products of the H2O/CO2 co-electrolysis reaction. Physicochemical characterization is also presented, in which the samples were examined in the form of powders and as half cells with BET, H2-TPR, Air-TPO and TGA re-oxidation measurements in the presence of H2O. In brief, it was found that the rate of the produced CO (rco) increases by increasing the operating temperature and the partial pressure of H2 in the reaction mixture. In addition, the first results revealed that Fe and Mo modification enhances the catalytic production of CO, since the 2wt% Fe-Ni/GDC and 3wt% Mo-Ni/GDC electrodes were proven to perform better compared to the other samples, in the whole studied temperature range (800–900 ◦C), reaching thermodynamic equilibrium. Furthermore, carbon formation was not detected.

**Keywords:** SOECs; RWGS reaction kinetics; Au–Mo–Fe-Ni/GDC electrodes; high temperature H2O/CO2 co-electrolysis

#### **1. Introduction**

Solid oxide electrolysis is a contemporary process for CO2 capture/recycling, which is proven as an attractive method to provide CO2 neutral synthetic hydrocarbon fuels. In particular, co-electrolysis of H2O and CO2 in a solid oxide electrolysis cell (SOEC) yields synthesis gas (CO + H2), which in turn can be used towards the formation of various types of synthetic fuels [1–3] by applying the Fischer-Tropsch process. According to thermodynamics, SOECs offer benefits for endothermic reactions, such as H2O and/or CO2 electrolysis at temperatures >700 ◦C, because a larger part of the required electrical energy can be substituted by thermal energy [4,5]. In addition, high temperature can lead to a significant decrease in the internal resistance of the cell and acceleration of the electrode reaction processes due to fast reaction kinetics [4–7].

H2O/CO2 Co-electrolysis is a much more complex process compared to pure steam or CO2 electrolysis. This is because three reactions take place simultaneously, namely H2O electrolysis, CO2 electrolysis, and the catalytic Reverse Water Gas Shift reaction (RWGS). More specifically, at the cathode of SOECs, two electrochemical reactions take place in parallel at the triple phase boundaries, i.e., H2O and CO2 electrochemical reductions. The oxygen ions (O2−), produced by these reactions, are moved to the anode, through an oxygen ion-conducting electrolyte (Yttria-Stabilized Zirconia − YSZ), where oxygen (O2) gas is formed [8,9]:

$$\text{Cathode:}\qquad \text{H}\_2\text{O} + 2\text{e}^- \rightarrow \text{H}\_2 + \text{O}^{2-} \quad \Delta\text{H}\_{800^\circ \text{C}} = 151.8 \text{ kJ/mol} \tag{1}$$

$$\text{CO}\_2 + 2\text{e}^- \rightarrow \text{CO} + \text{O}^{2-} \quad \Delta \text{H}\_{800^\circ \text{C}} = 185.9 \text{ kJ/mol} \tag{2}$$

$$\text{Anode:}\qquad \text{O}^{2-} \rightarrow \frac{1}{2}\text{O}\_2 + 2\text{e}^-\tag{3}$$

Besides the electrolysis reactions mentioned above, the reversible water gas shift (RWGS) catalytic reaction also occurs at the cathode with the co-existence of CO2 and H2O:

$$\text{Cathode:}\qquad \text{H}\_2 + \text{CO}\_2 \rightarrow \text{H}\_2\text{O} + \text{CO}\quad \Delta\text{H}\_{800} \\ \text{\textdegree C}=\text{36.8 kJ/mol}\tag{4}$$

It thus appears that CO can be produced either electrocatalytically by the reduction of CO2 or catalytically via the RWGS reaction. Questionable conclusions have been proposed in the literature about the role of the RWGS reaction for CO production. Some studies mention that CO is mainly produced electrocatalytically and RWGS has a small participation [10–15], while other studies mention that CO is absolutely produced catalytically via the RWGS reaction [16,17]. However, it has not been shown conclusively to which degree the RWGS is responsible for CO production in an SOEC.

Specifically, Mogensen and co-researchers found that the performance of co-electrolysis on an Ni/YSZ cathode varied between the highest pure H2O electrolysis and the lowest CO2 electrolysis, being much closer to the prior process [13,14]. This implies the significant contribution of the RWGS reaction in the H2O/CO2 SOEC co-electrolysis process and also confirms the co-existence of CO2 electrolysis. The above conclusion was supported by recently published results on computational modelling of the direct and indirect (with electro-generated H2) reduction of CO2 [18,19]. The direction of the Water Gas Shift (WGS) reaction could be forward or backward, depending on the operating conditions, indicating that at high temperatures (>838 ◦C), CO was produced electrochemically and the WGS was shifted in reverse towards H2 and CO2.

In contrast, Yue and Irvine [20] suggested that the RWGS reaction did not contribute critically in catalysing H2O/CO2 co-electrolysis on s (La,Sr)(Cr,Mn)O3 (LSCM) perovskite based cathode. The LSCM cathode displayed higher catalytic activity for pure CO2 electrolysis than pure H2O electrolysis, and the performance for H2O/CO2 co-electrolysis was much closer to pure CO2 electrolysis.

On the other hand, Bae et al., Hartvigsen et al., Kee et al., and Zhao et al. considered that CO production on Ni-based cathodes was mainly from the RWGS reaction, whereas the electrochemical processes are dominated by H2O electrolysis with identical performances for H2O electrolysis and H2O/CO2 co-electrolysis [16,17,21].

According to the above mentioned reports, until now, no agreement has been reached regarding the role of the RWGS reaction in the production of CO. As a result, it is crucial to quantify the degree of CO formation for each reaction. The discrepancies regarding the CO production route can be affected both by the structural characteristics of the electrode/catalyst (specific surface area, reducibility and re-oxidation behaviour, porosity, particle size, ionic and electronic conductivity), but also by the operating conditions (gas composition, temperature, etc.) [9,22]. As reported by Li's study [10], the heterogeneous thermochemical reactions occur at the external surface of the cathode and they are 20–100 times faster than the electrochemical reactions, which occur close to the electrolyte at the three phase boundaries. Furthermore, structural modifications of the cathode could enhance mass transport and promote CO production through the catalytic RWGS reaction, resulting in H2O/CO2 co-electrolysis performance close to that of H2O electrolysis.

The majority of studies on high temperature H2O/CO2 co-electrolysis utilize Ni-containing ceramic cathodes with YSZ and Gadolinia-Doped Ceria (GDC), similarly to the case of H2O electrolysis [10–14]. Ni-based materials indeed are cheap and exhibit porous structure, high electronic conductivity, appropriate catalytic activity, and a similar Thermal Expansion Coefficient (TEC) with the electrolyte. Consequently, they could act as excellent SOECs cathodes for H2O/CO2 co-electrolysis [7,22–26].

However, SOECs comprising Ni-based cathodes face some critical degradation issues, which are more pronounced with an increasing current density [3,12,14,27–33]. The main reasons for degradation have been reported to be microstructural changes that take place after prolonged co-electrolysis, resulting in passivation and blocking of the Three-Phase Boundaries (TPB) area [12,13,28,30,31,33,34]. Post-mortem microscopy investigations in the Ni/YSZ electrode have shown irreversible damages of the electrode's microstructure, such as loss of Ni-YSZ contact, decomposition of YSZ, Ni grain growth, loss of Ni percolation (loss of Ni–Ni contact), and even migration of Ni from the fuel electrode [28,30,31,33]. Taking into account the above, it is critical to develop alternative cathode materials for H2O/CO2 co-electrolysis with improved structural properties.

Previous studies reported that the electrocatalytic efficiency of SOECs can be improved by alloying transition metals with the state of the art (SoA) Ni catalyst [35–39]. Modification by means of alloying may be a promising strategy to promote the catalytic activity due to ligand and strain effects that change the electronic structure [40] of the active element. Recent studies focused on the Ni–Co alloy with Sm-doped ceria (Ni–Co/SDC) as an alternative material, with enhanced performance for H2O electrolysis in SOECs [36]. Specifically, the addition of Co increased the intrinsic catalytic activity of pure Ni and simultaneously expanded the active reaction region [36,41]. Other reports have shown that an Ni–Fe bimetallic cathode, mixed with Ba0.6La0.4CoO3 on La0.9Sr0.1Ga0.8Mg0.2O3 (LSGM) electrolyte, improved the electrochemical performance of the electrode for H2O electrolysis [38] and CO2 electrolysis [39], due to prevention of Ni aggregation. Furthermore, the enhanced performance of the Ni–Fe formulations could be attributed to the increased stability and improved catalytic activity, due to the addition of Fe, as this was concluded by combining experimental results with theoretical Density Functional Theory (DFT) calculations [41].

Our research group have studied the effect of Au and/or Mo doping on the physicochemical and catalytic properties of Ni/GDC for the reactions of catalytic CH4 dissociation and methane steam reforming [23,42–45] in the presence of H2 and H2S impurities. These modifications resulted in SOFC fuel electrodes with high sulfur and carbon tolerance as well as with improved electrocatalytic activity. Regarding the doping level of Au and Mo, the amount of 3wt% was indicated as an appropriate loading for achieving the most promising results under the examined SOFC conditions. Recently, the 3wt% Au-Ni/GDC electrode was tested under SOEC conditions for the H2O electrolysis process and exhibited improved electrocatalytic performance compared to Ni/GDC [46]. The enhanced performance of the Au-doped cathode was attributed to the creation of a surface Ni–Au solid solution, which causes weaker interplay of absorbed H2Oads and Oads species with the modified cathode. The result is a durable and resistant electrode to surface oxidation with an improved "three phase boundaries" area, especially at temperatures higher than 800 ◦C [46].

The presented investigation deals with modified X-Ni/GDC electrodes (where X = Au, Mo, Fe), in the form of half-Electrolyte Supported Cells (ESCs), for their performance in the RWGS through catalytic-kinetic measurements. Ni/GDC, 3wt% Au-Ni/GDC, 3wt% Mo-Ni/GDC, 3wt% Au–3wt% Mo-Ni/GDC, and 2wt% Fe-Ni/GDC modified cathodes were tested at Open Circuit Potential (OCP) conditions to elucidate their catalytic activity towards the production of CO (rco), which is one of the products from the H2O/CO2 co-electrolysis reaction. The latter approach is considered as an attempt to create a reference profile for the catalytic performance of the candidate electrodes, by applying the same H2O/CO2/H2 feed conditions as those under the co-electrolysis operational mode. The samples, both in the form of powders and as half cells, were physicochemically characterized, including specific redox stability measurements in the presence of H2O.

#### **2. Results and Discussion**

#### *2.1. Physicochemical Characterization*

#### 2.1.1. Specific Surface Area Values of Au–Mo–Fe-Modified Powders

The specific surface area values (SSA) for the modified-NiO/GDC samples, calcined in air at 600 and 1100 ◦C, as well as after reduction with H2 at 850 ◦C, are presented in Table 1.

The first remark is that all samples exhibit quite similar and low SSA values, which decrease further by increasing the calcination temperature from 600 ◦C to 1100 ◦C and after H2 reduction at 850 ◦C. Furthermore, Mo and Au–Mo modification slightly decrease the SSA of Ni/GDC, while the addition of Fe causes a significant increase in the SSA values up to approximately 70%. Taking into account the generally low values and the ±0.2 m<sup>2</sup> <sup>g</sup>−<sup>1</sup> accuracy limits of the measurements, it can be concluded that the reduced Mo and Au–Mo-Ni/GDC powders have similar SSAs with Ni/GDC. On the other hand, the 70% increase in the SSA of the Fe modified sample can be primarily ascribed to the formation of FeOx species, in the oxidized form of the sample, which inhibit the decrease of SSA during reduction.

**Table 1.** BET Specific Surface Area (SSABET) of commercial NiO/GDC, and X-modified NiO/GDC samples (where X = Au, Mo, and Fe). The measurements were performed on powders, calcined at <sup>600</sup> ◦C, 1100 ◦C, and after H2-reduction at 850 ◦C. Error/accuracy of SSABET <sup>=</sup> <sup>±</sup>0.2 m2 <sup>g</sup><sup>−</sup>1.


2.1.2. H2 Reducibility and air Re-Oxidation Behaviour of Au–Mo–Fe-Modified Powders

Temperature Programmed Reduction followed by Oxidation (H2-TPR, Air-TPO) measurements were performed, in order to investigate the reducibility and re-oxidation behavior of the prepared composite powders, as well as the existence of possible bulk effects due to the modification with Au, Mo, and Fe. The corresponding H2-TPR and Air-TPO TGA profiles of modified-NiO/GDC samples, calcined at 1100 ◦C, are presented in Figure 1.

**Figure 1.** *Cont.*

**Figure 1.** TGA (**A**) H2-TPR (12 vol.% H2/Ar), and (**B**) O2-TPO (33 vol.% air) profiles of X-modified NiO/GDC samples (where X= Au, Mo, and Fe). The measurements were performed on powders, calcined at 1100 ◦C. Ftotal = 60 cm3/min (STP conditions: 0 ◦C, 1 atm).

H2-TPR and air-TPO measurements showed different reduction and oxidation behavior in each of the examined modified samples. In particular, two main reduction peaks (Figure 1A) and one oxidation peak (Figure 1B) are mainly observed, which can be attributed to the reduction of NiO species to Ni and the reverse [42,47]. Moreover, the presence of Au does not seem to affect significantly the reduction of NiO. On the other hand, MoO3 and, to a certain extent, Fe inhibit the NiO reduction, implying a stronger Ni–O bond. The inhibition effect of MoO3 is also observed in the case of the ternary 3Au–3Mo-NiO/GDC sample and indicates a bulk interaction between Ni–Au–Mo, which has been thoroughly investigated in previous studies [23,42,43]. In particular, the second broad peak ("shoulder") at 580 ◦C, which is observed for the 3Mo-NiO/GDC and 3Au–3Mo-NiO/GDC samples, may be associated with the reduction of MoO3 species, which are reduced at higher temperatures in the range of 600–770 ◦C [42,47]. The presence of Fe seems to affect the main reduction peak of NiO, in a similar way as MoO3, suggesting the possible formation of a solid solution between Ni and Fe. The latter observation is currently under further clarification.

Upon re-oxidation of the samples, the unmodified Ni/GDC as well as the Au modified sample are oxidized practically at the same temperature with a similar TPO profile. The reduced state of the Fe, Mo-, and Au–Mo- modified samples proved to be more resistant to re-oxidation in 33 vol.% air/Ar (6.7 vol.% O2/Ar), since they had to reach higher temperature for complete re-oxidation.

#### 2.1.3. H2O Re-Oxidation Profiles of Au-Mo-Fe–Modified Powders

Isothermal-TGA measurements, under 15.5 vol.% H2O/Ar conditions, were carried out at 650 ◦C, 700 ◦C, 750 ◦C, and 800 ◦C and the results are depicted in Figures 2 and 3. These measurements investigate the activity of the powders for the H2O dissociation reaction and their concomitant

re-oxidation rate. H2O acts as an oxidative agent and interacts with the Ni atoms on the surface of each sample towards H2 and NiO [reaction (5)] [46].

$$\text{H}\_2\text{O} + \text{Ni} \rightarrow \text{H}\_2 + \text{NiO} \tag{5}$$

Thereafter, there is a progressive diffusion of the absorbed oxygen species (Oads) from the surface in the bulk phase of the sample, which is oxidized further [46,48–50]. Figure 2 depicts the H2O re-oxidation profiles, as an increase of the weight (Δwt%), of the pre-reduced Ni/GDC, 3wt% Au-Ni/GDC, 3wt% Mo-Ni/GDC, 3wt% Au–3wt% Mo-Ni/GDC, and 2wt% Fe-Ni/GDC samples in the temperature range between 650–800 ◦C.

**Figure 2.** Isothermal TG analysis of the pre-reduced, (**A**) Ni/GDC, (**B**) 3wt% Au-Ni/GDC, (**C**) 3wt% Mo-Ni/GDC, (**D**) 3wt% Au–3wt% Mo-Ni/GDC, and (**E**) 2wt% Fe-Ni/GDC samples in the temperature range between 650–800 ◦C. (**F**) Comparative TG profiles of the samples at 800 ◦C. Feed Conditions: 15.5 vol.% H2O/Ar, Ftotal = 100 cm3/min (STP conditions: 0 ◦C, 1 atm).

**Figure 3.** Arrhenius plots of the H2O re-oxidation rate (resulting from the slope of the isothermal TG profiles in Figure 2) for: Ni/GDC, 3wt% Au-Ni/GDC, 3wt% Mo-Ni/GDC, 3wt% Au–3wt% Mo-Ni/GDC, and 2wt% Fe-Ni/GDC Feed Conditions: 15.5 vol.% H2O/Ar, Ftotal = 110 cm3/min (STP conditions: 0 ◦C, 1 atm).

The re-oxidation profiles of the samples in Figure 2 can be separated in three sections [46]. The first section corresponds to the initial sharp increase in weight and is ascribed [46] to the complete, bulk, re-oxidation of the partially reduced CeO2 in GDC and specifically of Ce3+ to Ce4+. The second section occurs during the following 12–25 min of the reaction, where the samples, except for 2Fe-Ni/GDC, keep their reduced state without any changes in their weight. This step is associated [46] to the dissociation of bulk NiH species, which are formed during the prior reduction period, and maintains the samples reduction. NiH dissociation is considered as an activated process, because by decreasing the operating temperature from 800 ◦C to 650 ◦C, the specific "delay" period increases from 12 min to 25 min. The first two sections are quite similar for the pre-reduced powders, with identical low SSA values (Table 1). On the other hand, the initial re-oxidation behavior, in the first section, of the pre-reduced 2Fe-Ni/GDC is more intense, indicating the strong effect of H2O.

In the third section of the TG profiles, further re-oxidation takes place [reaction (5)] and the main discrepancies are detected among the samples. In the range of 650–800 ◦C, the ternary 3Au–3Mo-Ni/GDC sample is the most tolerant in bulk re-oxidation by H2O, while the binary 2Fe-Ni/GDC sample is the least tolerant. The decrease in temperature (800 → 650 ◦C) leads to inhibition of the bulk re-oxidation, as observed by the reduced slopes in the TG profiles, whereas the trend among the samples does not change. The high tolerance of the 3Au–3Mo-Ni/GDC sample against bulk re-oxidation by H2O is attributed to the synergistic interaction of nickel with gold and molybdenum [23,42].

The slope on the linear part of the third section in the TG curves represents the intrinsic dissociation rate of H2O and correspondingly the re-oxidation rate of Ni atoms on each sample [reaction (5)] [46]. Figure 3 depicts the Arrhenius plots of the oxidation rates of the samples, which were calculated from the slopes on the linear parts of the TG profiles (Figure 2). The Arrhenius plots (Figure 3) show that the binary 2Fe-Ni/GDC sample has the highest H2O re-oxidation rates, while the ternary 3Au–3Mo-Ni/GDC has the lowest. This is further confirmed by the calculated apparent activation energies (Ea,ap), which are defined from the Arrhenius plots and are reported in Table 2.


**Table 2.** Apparent activation energies (kJ mol<sup>−</sup>1) for the dissociation of H2O on commercial Ni/GDC [46], and modified Ni/GDC samples. The values correspond to the slopes of the Arrhenius plots in Figure 3.


The calculated values verify that the binary 2Fe-Ni/GDC sample exhibits the lowest Ea,ap (20 kJ/mol) for the catalytic dissociation of H2O, whereas the ternary 3Au–3Mo-Ni/GDC sample exhibits the highest Ea,ap (63 kJ/mol). The above behavior can be interpreted by the fact that Ni-Fe interaction reinforces the bond of H2Oads, resulting in a higher re-oxidation rate (lower Ea,ap). On the other hand, following the same interpretation, Ni–Au–Mo interaction weakens the bond of H2Oads, resulting in a lower re-oxidation rate (higher Ea,ap) [46].

The calculated Ea,ap values are in good agreement with our previous study [46] and other studies [51–53] in the literature, which focus on theoretical investigations. The binary 2Fe-Ni/GDC sample is the most active for the thermochemical dissociation of H2O and is less resistant against bulk re-oxidation. On the contrary, the ternary 3Au–3Mo-Ni/GDC sample exhibits the lowest activity for the thermochemical dissociation of H2O and the highest tolerance against bulk re-oxidation.

Taking into account the above, H2O, apart from being the main reactant of the co-electrolysis process, is also considered as a potential poisoning agent of the electrode. This means that the bonding strength of the adsorbed oxygen species, which result from H2O decomposition, may induce the re-oxidation of the electrode and finally the deactivation of the sample [46,53]. The H2O poisoning effect in electrolysis and co-electrolysis processes can also be correlated to the CH4 poisoning effect in SOFC applications, where degradation is enhanced by the strong adsorption bond of CH4 on the Ni surface [23,42–45,54,55]. Thus, the stronger binding of the adsorbed oxygen species, which result from the catalytic dissociation of H2Oads, can similarly cause a poisoning effect on the Ni surface, leading to faster re-oxidation and finally deactivation of the sample [46,53]. According to Besenbacher et al. [56], the surface solid solution between Au and Ni induces significant modifications in the electronic properties of Ni (Fermi level, work function, and d-band center) and affects the bonding strength of the adsorbed species on the surface of the sample.

In this respect, the Au- and Au–Mo- doping of Ni should shift the d-band center to lower energies, with respect to the Fermi level of nickel, thus inhibiting the interaction and the dissociation of H2O. Therefore, the 3Au-Ni/GDC and 3Au–3Mo-Ni/GDC samples are suggested to be the least active samples for the dissociation of H2O, having the highest Ea,app and consequently the lowest binding energies for both H2O and Oads. The binary 2Fe-Ni/GDC is suggested to be the most active sample for the dissociation of H2O, with the lowest Ea,app, resulting in Oads species with the highest binding energy and the potential to cause faster re-oxidation/poisoning of the electrode [46,53] during SOEC operation.

#### *2.2. Catalytic-Kinetic Measurements of the RWGS Reaction*

All the prepared samples, in the form of half-electrolyte supported cells, were tested at Open Circuit Potential (OCP) conditions to elucidate their catalytic activity towards the production of CO, which is one of the products from the H2O/CO2 co-electrolysis reaction. Furthermore, the effect of the modification (type of dopant: Au, Mo, Fe) on the catalytic activity for the production of CO, through the RWGS reaction, was also investigated. The latter approach is considered as an attempt to create a reference profile for the catalytic performance of the candidate electrodes, under the same experimental conditions as in co-electrolysis (including the presence of current collector). This is an important step, because it will provide detailed experimental feedback on the possible contribution of the RWGS reaction to the production of CO and the extent of this contribution to the electrochemical reduction of CO2, during the co-electrolysis mode.

In regards to the "homogenous" catalytic production of CO, the rate was found to be negligible, in the range of rco, homogenous = 0.05 μmol/s, for a fuel feed comprising 30 vol.% He − 24.5 vol.% H2O − 24.5 vol.% CO2 − 21 vol.% H2. Concerning the current collector, Ni-mesh shows some catalytic activity for the production of CO (rCO), which increases by increasing the operating temperature and the partial pressure of H2. However, comparative rCO measurements (see Figure S1 in Supplementary Material) of an Ni/GDC electrode with and without the presence of Ni mesh as the current collector, as well as of the bare Ni mesh, suggest that there is no direct catalytic correlation/contribution of the Ni mesh to the activity of the electrocatalysts. This is mainly explained by the fact that Ni/GDC is a porous electrocatalyst with SSABET and thus with more active sites, compared to the metallic nickel mesh, which does not possess similar properties. Consequently, from the point where they co-exist as the electrode and current collector, the catalytic activity is mainly attributed to the electrocatalyst.

Figure 4 presents the comparison of the produced rCO for the Ni/GDC and for modified-Ni/GDC electrodes, under three different H2O-CO2-H2 mixtures. The corresponding %CO2 conversions are depicted in Figure 5. The values are low enough to be considered in the differential region, apart from the case of 2Fe-Ni/GDC in the reaction mixture (A), which is relatively high. The % conversion of CO2 was calculated by the following formula:

$$\text{CO}\_2\text{ conversion } (\%) = \frac{\text{F}\_{\text{CO}}^{\text{out}}}{\text{F}\_{\text{CO}\_2}^{\text{in}}} \cdot 100 \tag{6}$$

where: Fout CO and Fin CO2 correspond to the rate (μmol/s) of the produced CO and of the introduced CO2 in the reactants feed, respectively.

**Figure 4.** CO production rates (μmol s−<sup>1</sup> g−1) on ESCs comprising: Ni/GDC, 3Au-Ni/GDC, 3Mo-Ni/GDC, 3Au–3Mo-Ni/GDC, and 2Fe-Ni/GDC, as fuel electrodes, in the range of 800–900 ◦C under three different mixtures: (**A**) 24.5% H2O − 24.5% CO2 − 21% H2 (PH2/PCO2 = 0.86), (**B**) 28% H2O − 28% CO2 − 14% H2 (PH2/PCO2 = 0.50), and (**C**) 31.5% H2O − 31.5% CO2 − 7% H2 (PH2/PCO2 = 0.22). Dilution of He: 30 vol.% and Ftotal = 140 cm3/min (STP conditions: 0 ◦C, 1 atm) in all cases. The dash line (– –) corresponds to the rCO values in thermodynamic equilibrium. All studied electrodes have similar loading in the range of 10–12 mg/cm2.

**Figure 5.** The corresponding %CO2 conversions of ESCs comprising: Ni/GDC, 3Au-Ni/GDC, 3Mo-Ni/GDC, 3Au–3Mo-Ni/GDC, and 2Fe-Ni/GDC, as fuel electrodes, in the range of 800–900 ◦C under three different mixtures: (**A**) 24.5% H2O − 24.5% CO2 − 21% H2 (PH2/PCO2 = 0.86), (**B**) 28% H2O − 28% CO2 − 14% H2 (PH2/PCO2 = 0.50), and (**C**) 31.5% H2O − 31.5% CO2 − 7% H2 (PH2/PCO2 = 0.22). Dilution of He: 30 vol.% and Ftotal = 140 cm3/min (at STP conditions: 0 ◦C, 1 atm) in all cases. All studied electrodes have similar loading in the range of 10–12 mg/cm2.

It is shown (Figures 4 and 5) that 2Fe-Ni/GDC and 3Mo-Ni/GDC have the highest catalytic activity of the examined electrodes for the RWGS reaction. In fact, 2Fe-Ni/GDC is the most active in terms of the produced CO. The above performance is observed for all applied fuel feeds (PH2/PCO2 = 0.86, 0.50 and 0.22), whereas it is enhanced by increasing (i) the operating temperature and (ii) the partial pressure of H2 in the fuel feed. The enhanced catalytic performance of the Fe-modified sample, can be primarily ascribed to the possible stronger adsorption and consequent catalytic dissociation of CO2 on the active sites of the catalyst. This first conclusion is going to be further examined with specific CO2 Temperature Programmed Desorption (TPD) measurements.

Another noteworthy remark is that the produced CO rates were compared to the thermodynamic equilibrium rates (dash line in Figure 4) for the three different H2O-CO2-H2 reaction mixtures. The equilibrium rates were calculated by using the equilibrium constant (Keq) formula that is reported in [57,58]. It was found (Figure 4) that Ni/GDC, the binary Au-, and the ternary Au–Mo- modified samples exhibit CO production rates, which are lower than the thermodynamic equilibrium for all the examined reaction conditions. The performances of 2Fe-Ni/GDC and 3Mo-Ni/GDC are closer to the equilibrium, but cannot be considered as thermodynamically limited.

The results from the measurements in Figure 4 can also be presented as Arrhenius plots, depicted in Figure 6, where the derived apparent activation energies (Ea, app) for the production of CO and consequently for the RWGS reaction are listed in Table 3.

**Figure 6.** Arrhenius plots of the CO production rates (μmol s−<sup>1</sup> g−1) on ESCs comprising: Ni/GDC, 3Au-Ni/GDC, 3Mo-Ni/GDC, 3Au–3Mo-Ni/GDC, and 2Fe-Ni/GDC, in the range of 800–900 ◦C, under three different mixtures: (**A**) 24.5% H2O − 24.5% CO2 − 21% H2 (PH2/PCO2 = 0.86), (**B**) 28% H2O − 28% CO2 − 14% H2 (PH2/PCO2 = 0.50), and (**C**) 31.5% H2O − 31.5% CO2 − 7% H2 (PH2/PCO2 = 0.22). Dilution of He: 30 vol.% and Ftotal = 140 cm3/min (at STP conditions: 0 ◦C, 1 atm) in all cases.

**Table 3.** Apparent activation energies (Ea,app, kJ/mol) for the RWGS reaction on ESCs for three different mixtures: (**A**) 24.5% H2O − 24.5% CO2 − 21% H2 (PH2/PCO2 = 0.86), (**B**) 28% H2O − 28% CO2 − 14% H2 (PH2/PCO2 = 0.50), and (**C**) 31.5% H2O − 31.5% CO2 − 7% H2 (PH2/PCO2 = 0.22).


*\* A:* The pre-exponential factor in the Arrhenius equation, r <sup>=</sup> <sup>A</sup>· exp(<sup>−</sup> *Ea*,*app <sup>R</sup>*·*<sup>T</sup>* ).

The Arrhenius plots and the calculated Ea, app show that 2Fe-Ni/GDC, 3Mo-Ni/GDC, and Ni/GDC have practically the same and the lowest Ea, app. However, the Fe- and Mo- modified samples exhibit higher pre-exponential factors, compared to Ni/GDC, which explains the higher production rates of CO. 3Au-Ni/GDC shows overall the highest Ea, app, which is an additional indication for its worst catalytic activity. Finally, the ternary 3Au–3Mo-Ni/GDC sample shows an apparent activation energy, which lies between that for 2Fe-Ni/GDC and 3Au-Ni/GDC. However, the catalytic activity of the ternary sample is the lowest, due to the lower pre-exponential factor compared to the binary Au-modified sample. According to the knowledge of the authors, there are no literature data available for experimentally measured Ea, app, for similar samples and reaction conditions. The so far available data come from theoretical investigations and there is a recent study from Cho et al. [41], who performed DFT calculations to evaluate the ability of various transition metals to increase the activity of Ni for the H2O/CO2 co-electrolysis. In particular, they computed the activation energies of specific elementary reaction steps and in the case of the RWGS on Ni(111), the Ea was found to be approximately 46 kJ/mole, which is very close to the values that were experimentally calculated in the present study.

The effect of H2 partial pressure on the catalytic rate of CO production is further verified in Figure 7 for all samples at 900 ◦C and 800 ◦C. The 2Fe-Ni/GDC sample is the most active and 3Au–3Mo-Ni/GDC the least one. The 3Au-Ni/GDC catalyst at 900 ◦C shows similar performance with that of Ni/GDC. In addition, by decreasing the temperature at 800 ◦C, the catalytic activity of 3Au-Ni/GDC exhibits the highest reduction. This is apparent from the significant decrease in the slope of the corresponding curve (Figure 7) from 800 ◦C to 900 ◦C and can be further explained from the calculated Ea, app (Table 3), which is the highest from all samples. Finally, the kinetic behavior of all samples suggests that the production rate of CO exhibits a positive order dependence on the partial pressure of H2.

**Figure 7.** Steady state effect of the H2 partial pressure (pH2, kPa) on the CO production rates (μmol s−<sup>1</sup> g<sup>−</sup>1) on ESCs comprising: Ni/GDC, 3Au-Ni/GDC, 3Mo-Ni/GDC, 3Au–3Mo-Ni/GDC, and 2Fe-Ni/GDC, for 900 ◦C and 800 ◦C. The pH2O/pCO2 ratios are fixed and their values are presented in Table 3.

The fact that in the majority of the samples there is no trend in the Ea, app, by varying the pH2/pCO2 ratios, primarily shows that all samples, apart from the binary Au-modified, have similar intrinsic catalytic activity for the RWGS reaction. This is also concluded from the remark that the effect of the H2 partial pressure on the CO production rate is almost linear. On the other hand, in the case of 3Au-Ni/GDC, there is an increase in Ea, app, by increasing the pH2/pCO2 ratio. This is corroborated from the results in Figure 7, where rCO seems to reach a plateau at high pH2. The latter remarks indicate that the majority of the samples have actives sites that follow a similar reaction mechanism, in regards to the dissociative adsorption of H2 and CO2 towards the RWGS. On the other hand, the reaction mechanism seems to be different in the case of 3Au-Ni/GDC. In particular, it is implied that the high coverage of adsorbed H2 may inhibit the CO2 dissociative adsorption and thus decreases the intrinsic catalytic activity of the specific sample. This suggestion is going to be further clarified through the currently performed H2O/CO2 co-electrolysis measurements.

Carbon formation, through the Boudouard reaction (6), was also investigated and the results are presented in Figure 8. The slight scattering of the experimentally measured rates in combination with the theoretical values, where r[CO2]inlet is equal to r[CO + CO2]outlet, suggest that there is no carbon deposition under the examined reaction conditions. This result is also in accordance with the fact that the Boudouard reaction is not thermodynamically favored above 750 ◦C [59].

**Figure 8.** Investigation of the Boudouard reaction at 900 ◦C and 800 ◦C under OCP conditions on ESCs comprising: Ni/GDC, 3Au-Ni/GDC, 3Mo-Ni/GDC, 3Au–3Mo-Ni/GDC, and 2Fe-Ni/GDC. Theoretical case where [CO2]inlet is equal to [CO+CO2]outlet is also depicted with a solid line.

Previous studies [1,59,60] reported that at realistic CO2/CO concentrations, during CO2 electrolysis in the range of 650–750 ◦C, the equilibrium of (6) is shifted towards CO production (reaction (7)) and therefore carbon will not be formed, catalytically, during CO2 electrolysis. In the case where the cell is operated at OCP conditions, as in our case, any coke deposited within the porous, modified or not, Ni/GDC electrode would be oxidized to CO according to the reverse Boudouard reaction (7) and thereby removed. Furthermore, the addition of steam in the feed is reported to remove carbon depositions according to reaction (8) [1,8]:

$$\text{Boundouard reaction:} \qquad \text{2CO} \rightarrow \text{CO}\_2 + \text{C} \tag{7}$$

Reverse Boudouard reaction: CO2 + C → 2CO (8)

$$\text{Reaction of coke with H}\_2\text{O:}\qquad \text{C} + \text{H}\_2\text{O} \rightarrow \text{CO} + \text{H}\_2\tag{9}$$

#### **3. Materials and Methods**

#### *3.1. Preparation of Electrocatalysts*

The modified cathode powders were prepared via the Deposition-Precipitation (D.P.) and Deposition-Co Precipitation (D.CP.) methods by using the commercial NiO/GDC cermet (65wt% NiO-35wt% GDC, Marion Technologies, Verniolle, France) as the support. The precursors for the 3wt% Au-NiO/GDC, 3wt% Mo-NiO/GDC, 3wt% Au−3wt% Mo-NiO/GDC, and 2wt% Fe-NiO/GDC samples were the HAuCl4 (Sigma-Aldrich) or/and (NH4)6Mo7O24 (Sigma-Aldrich, St. Louis, MO, USA) and Fe(NO3)3x9H2O (Sigma-Aldrich, St. Louis, MO, USA) solutions, respectively. Full details about the synthesis can be found elsewhere [42,43]. After filtering, the precipitate was dried at 110 ◦C for 24 h. All dried powders were calcined at 600 ◦C/90 min and a part of them at 1100 ◦C/75 min. The first batch was used for the paste preparation, which is described in the following paragraph. The batch at 1100 ◦C was used for the physicochemical characterization. In this way, the prepared catalysts were studied at similar thermal stress as the calcined electrode-electrolyte assemblies.

#### *3.2. Preparation of Electrolyte-Supported Half Cells*

The electrolyte-supported half cells consisted of circular shaped planar 8YSZ electrolyte (by Kerafol) with a 25 mm diameter and 300 μm thickness. The fuel electrode was deposited by using the screen printing technique as reported in previous studies [23,46]. In particular, a paste was prepared by using an appropriate amount of the electrocatalyst (modified NiO/GDC powder), terpineol (Sigma-Aldrich, St. Louis, MO, USA) as the dispersant, and PVB (polyvinylbutyral, Sigma-Aldrich, St. Louis, MO, USA) as binder. After the deposition of the paste, the cell was sintered at 1150 ◦C with a heating/cooling ramp rate of 2 ◦C/min. The last temperature is the lowest possible in order to obtain proper adherence of the electrode on the electrolyte, whereas it is equivalent with the calcination temperature of the characterized powders (1100 ◦C). The examined electrodes were approximately 20 μm thick and their loading varied in the region of 10–12 mg/cm2 with a 1.8 cm<sup>2</sup> geometric surface area (Figure 9). The prepared half cells were adjusted on a ceramic YSZ tube and sealed airtight with a glass sealing material manufactured by Kerafol.

(A)

**Figure 9.** Images of a (half) electrolyte supported cell (ESC) prepared by screen printing: (**A**) NiO/GDC electrode and (**B**) SEM cross section perpendicular to the NiO/GDC//YSZ (half) electrolytesupported cell.

#### *3.3. Physicochemical Characterization*

The samples, in the form of powders, were characterized with BET, H2-TPR, Air-TPO, and TGA re-oxidation measurements in the presence of H2O.

The BET specific surface area values of the samples were determined from the adsorption isotherms of nitrogen at −196 ◦C, recorded with a Micromeritics TriStar 3000 apparatus (Micromeritics, Norcross, GA, USA).

The re-oxidation measurements in the presence of H2O, as well as H2-TPR and Air-TPO measurements were carried out with a TA Q50 instrument. The H2O re-oxidation properties of the powders were studied with TGA measurements, at a constant temperature in the range of 650–800 ◦C. Before the measurement, the samples were pre-reduced in-situ with 80 vol.% H2/Ar at 800 ◦C for 100 min and then the feed was changed to 15.5 vol.% H2O/Ar. The total flow rate (at STP conditions: 0 ◦C, 1 atm) was 100 cm3/min and the loading of the measured samples was approximately 50 mg. Steam was introduced in the reactor by passing Ar through a saturator, which was maintained at a fixed temperature (65 ◦C).

#### *3.4. Catalytic-Kinetic Measurements*

The prepared half cells were catalytically investigated at Open Circuit Potential (OCP) conditions for the RWGS reaction, in the presence of Ni mesh. The catalytic experiments were accomplished at temperatures between 800–900 ◦C under various H2O/CO2/H2 mixtures by keeping in all cases the PH2O/PCO2 ratio constant. In regards to the experimental part of these measurements, H2O was added and handled in the feed in the form of steam, as in the SOEC measurements. Before its evaporation, liquid H2O was pressurized in a container and circulated in the system by means of a liquid water mass flow controller. Then, liquid H2O was evaporated through lines and valves, heated at 160 ◦C, to prevent water condensation. The flow rate was fixed at 140 cm3/min (at STP conditions: 0 ◦C, 1 atm), avoiding any mass transfer limitation effects in the reactor. Reactants and products were determined by using an on-line gas chromatograph (Varian CP-3800) with a thermal conductivity detector. Further details regarding the experimental parameters are indicated in the corresponding Figures.

#### **4. Conclusions**

The presented study deals with the kinetic investigation of Ni-based (modified or not) electrodes towards their performance for the RWGS reaction. The samples were examined in the form of electrolyte-supported (half) cells and the measured kinetic parameter was the production rate of CO. The main objective was to clarify the effect of the modification on the catalytic activity for the RWGS, which is considered as a key reaction for the CO production under H2O/CO2 co-electrolysis operation. The reaction conditions were similar to those that are applied under co-electrolysis mode.

Redox stability measurements in the presence of H2O showed that the ternary 3Au–3Mo-Ni/GDC electrode is the least active sample for the dissociation of H2O, having the highest Ea,app and consequently the lowest binding energy for the H2Oads. On the other hand, the binary 2Fe-Ni/GDC is the most active sample for the dissociation of H2O, thus having the potential to experience faster re-oxidation. Complementary characterization suggests that the interaction of Ni and Fe (FeOx species in the oxidized form of the sample) during the H2-reduction process increases the SSABET and affects the bulk properties of the binary Ni-Fe/GDC. The interaction can be realized through the possible formation of an Ni-Fe solid solution, which is currently under further clarification and may be responsible for enhancing the catalytic dissociation of H2Oads.

The kinetic study of the candidate electrocatalysts showed that Au modification inhibits the catalytic production of CO, through the RWGS reaction, while modification with Fe or Mo induces an enhancement of rCO. In fact, the 2wt% Fe-Ni/GDC sample is the most active both in terms of %CO2 conversion and of the produced CO. In addition, a negative synergy was observed for the ternary Au–Mo-Ni modified sample. Specifically, the 2wt% Fe-Ni/GDC and 3wt% Mo-Ni/GDC samples

showed similar apparent activation energy for the RWGS reaction as that of Ni/GDC, while the 3wt% Au-Ni/GDC and 3wt% Au–3wt% Mo-Ni/GDC samples showed lower Ea,app. The above performance is observed for all applied fuel feeds (PH2/PCO2 = 0.86, 0.50, and 0.22), whereas it is enhanced by increasing (i) the operating temperature and (ii) the partial pressure of H2.

Consequently, the kinetic behavior of all samples suggests that the production rate of CO exhibits a positive order dependence on the partial pressure of H2, whereas carbon formation was not detected. In the case of the most active 2Fe-Ni/GDC sample, it is proposed that Fe-modification may enhance the catalytic dissociative adsorption of CO2 towards the production of CO and this is further investigated. Finally, it is worth mentioning that the RWGS catalytic performance of both Fe- and Mo- modified samples is close to the equilibrium, but cannot be considered as thermodynamically limited.

Overall, the presented results correspond to a reference catalytic profile of the examined modified Ni/GDC samples for the RWGS reaction. These candidate electrocatalysts are currently being examined, as full electrolyte supported cells, in SOEC measurements for the H2O/CO2 co-electrolysis reaction to elucidate any additional effects by the applied current. Moreover, further investigation of the Fe-modification on NiO/GDC for the H2O/CO2 electrolysis processes will occur in the future.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/2/151/s1, **Figure S1.** Kinetic study of the R.W.G.S. reaction on the nickel mesh (), Ni/GDC with Ni mesh (●) and Ni/GDC without Ni mesh (), in the range of 800–900 ◦C, under two different mixtures: (**A**) 24.5% H2O − 24.5% CO2 − 21% H2 (PH2/PCO2 = 0.86) and (**B**) 28% H2O − 28% CO2 − 14% H2 (PH2/PCO2 = 0.50). Dilution of He: 30 vol.% and Ftotal = 140 cm3/min (at STP conditions: 0 ◦C, 1 atm) in all cases. The loading of Ni/GDC is 11 mg/cm2.

**Author Contributions:** Conceptualization, S.N. and D.N.; Data curation, E.I.; Formal analysis, E.I.; Funding acquisition, S.N. and D.N.; Investigation, E.I.; Methodology, S.N. and D.N.; Supervision, S.N. and D.N.; Validation, S.N. and D.N.; Writing—original draft, D.N.; Writing—review & editing, S.N., D.N. and E.I.

**Funding:** The research leading to these results has received funding from the Fuel Cells and Hydrogen 2 Joint Undertaking under the project SElySOs with Grant Agreement No: 671481. This Joint Undertaking receives support from the European Union's Horizon 2020 Research and Innovation Programme and Greece, Germany, Czech Republic, France, and Norway.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Catalysts* Editorial Office E-mail: catalysts@mdpi.com www.mdpi.com/journal/catalysts

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18