1 Introduction

Hydrogen gas that is produced via steam-reforming of hydrocarbons or alcohols and subsequent water-gas-shift (WGS) still contains high levels of CO. For pure H2 streams, e.g., required for fuel cells, the CO must be removed, e.g., by preferential oxidation of carbon monoxide, CO-PROX [1,2,3]. PROX means that the oxidation of CO (CO + 0.5 O2 → CO2) is preferred over H2 oxidation (H2 + 0.5 O2 → H2O), avoiding the consumption of valuable hydrogen. As mentioned, this process is crucial for using the resulting H2 stream in polymer electrolyte membrane fuel cells (PEMFCs), because CO poisons the Pt electrodes, and significantly decreases the efficiency [3,4,5,6,7]. Accordingly, the CO concentration must be reduced to 10–50 ppm before feeding the H2 stream to the cells [8,9,10]. A variety of heterogeneous catalysts has shown promising performance in CO-PROX, for example, noble metals (Pd, Pt, or Rh) supported on alumina [3, 11, 12]. Nevertheless, the high-cost of precious metals inspired the search for low-cost alternatives.

Transition metal oxide catalysts containing Cu and Co have received attention, as their catalytic PROX performance was comparable to that of noble metal catalysts [3, 13,14,15,16]. Among them, CuO supported on ceria (CuO/CeO2) is a well-known low-cost low-temperature PROX catalyst with good activity and high selectivity for CO oxidation [3, 17,18,19,20]. Selective CO oxidation has been reported to occur in the temperature range of ~ 80–200 °C [3] and was attributed to sites at the interface of CuO and CeO2 [21,22,23,24]. Thus, a high dispersion of CuO clusters on ceria is clearly beneficial [25]. For PROX and low temperature CO oxidation, Co3O4 has also been frequently used [26,27,28,29].

Wang et al. prepared a catalyst of CuO particles on hollow CeO2 nanospheres, based on a metal-organic framework (MOF) precursor and using a template-free microwave method. In CO-PROX, 100% CO conversion was reached at about 80 °C [30]. In another work, using a hard template method [31] CuO nanoparticles were loaded on the outer surface of hollow CeO2 nanospheres in a layer-by-layer deposition technique. As reported in literature, the CuO surface plays a vital role as an active site for CO and H2 oxidation while, after the formation of CO2 and H2O, CeO2 provides oxygen to fill the vacancies [24, 32,33,34,35]. The oxygen transfer potential of the CeO2 support is also promising for other oxides [36, 37].

In the current work, CuO/CeO2 nanosphere catalysts were characterized by X-ray diffraction (XRD), high resolution transmission electron microscopy (HRTEM), and CO/H2- temperature programmed reduction (TPR), before being examined for CO-PROX in an atmospheric flow reactor. Combining experiment with density functional theory (DFT) modeling of surface structures and adsorbed CO and H2 on various metal and oxide sites enabled us to identify the active sites and the origin of high selectivity of CuO/CeO2.

2 Methods

2.1 Catalyst Synthesis and Characterization

The synthesis of CuO/CeO2 nanosphere catalysts was described in detail in Ref [31]. It is based on the hard template method, with CuO nanoparticles loaded on the outer surface of hollow CeO2 nanospheres via layer-by-layer deposition (Fig. 1a). Details of the preparation (including a final calcination at 500 °C) are presented in the Supporting Information. CO, H2, O2, and He were obtained from Messer Group GmbH, (purity 4.7 for CO and 5.0 for the others).

2.1.1 XRD

X-ray diffraction (XRD) patterns of oxidized and reduced catalysts were collected on a Philips XPERT-PRO diffractometer using Cu K-α radiation (1.5406 Å; 45 kV; 40 mA) operating in Bragg-Brentano reflection geometry with 2θ scanning from 20–90° (step size of 0.02°). Phase analyses and Rietveld refinements were performed with the HighScore Plus software (JCPDS data base) [38].

2.1.2 TEM

To study the size, morphology, and distribution of CuO/Cu0 particles on the ceria support, powder samples were drop-casted on carbon-coated gold grids and examined by transmission electron microscopy (TEM) in a TECNAI F20 microscope operated at 200 kV. Before TEM, the CuO/CeO2 catalyst was oxidized at 400 °C with 20% oxygen in He.

2.1.3 CO- and H2−TPR

CO- and H2-temperature programmed reduction (TPR) techniques were applied to characterize the reducibility of the catalysts at atmospheric pressure. Approximately 20 mg catalyst was loaded in a continuous-flow fixed-bed quartz reactor between two quartz wool plugs. Before each TPR run, contaminants were removed by pretreatment with 20% O2 in He (50 mL min−1) at 400 °C for 30 min (heating rate of 10 °C min−1). Samples were then cooled down to 30 °C in a flow of the same composition and purged with He for 30 min at the same temperature. The pretreated samples were then exposed at room temperature to either a mixture of 5 vol% CO in He (CO-TPR) or 5 vol% H2 in He (H2-TPR) at a flow rate of 50 mL min−1, before being subsequently heated to 400 °C (heating rate of 10 °C min−1). CO or H2 consumption and CO2 or H2O evolution, respectively, were analyzed by an online quadrupole mass spectrometer (QMS, Prisma Plus QMG220, Pfeiffer Vacuum) equipped with a secondary electron multiplier (SEM) detector.

2.2 Flow Reactor Studies (CO Oxidation and PROX)

Catalytic reactions of CO oxidation and preferential CO oxidation (PROX) were carried out in the same reactor as for TPR. Concentrations of reactants and products in the outlet stream were monitored by the MS and additionally by gas chromatography (GC, Agilent 6890) using a HP-PLOT Q column, a flame-ionization detector (FID) with a methanizer and a thermal conductivity detector (TCD).

Before CO oxidation and PROX experiments, each fresh catalyst (20 mg) was pretreated as described in the following. In all cases, the heating rate was 10 °C min−1 and the pretreatment time was 30 min each before cool-down to room temperature. For oxidation 20 vol% O2 in He was used, for reduction 5 vol% H2 in He, with a total flow rate of 50 mL min−1. For both a heating rate of 2 °C min−1 was applied.

Temperature-dependent CO oxidation was performed for three differently pretreated catalysts: (i) oxidation at 300 °C, (ii) oxidation at 300 °C followed by reduction at 300 °C, and (iii) oxidation at 500 °C followed by reduction at 500 °C. The feed composition was 5 vol% CO and 10 vol% O2 in He (heating rate of 2 °C min−1). Additionally, to mimic PROX without H2, 1 vol% CO and 1 vol% O2 in He was supplied for a sample oxidized at 400 °C.

For PROX, two different pretreatments were applied: (i) oxidation at 400 °C, (ii) reduction at 300 °C after pretreatment). The PROX reaction was then performed in 1 vol% CO, 1 vol% O2, and 50 vol% H2 in He.

The conversion of CO (\({X}_{CO}\)) and O2 (\({X}_{{O}_{2}}\)), and the CO2 selectivity (\({S}_{C{O}_{2}}\)) were calculated from the GC peak areas of CO and O2 in the inlet and outlet of the reactor using the following formulae:

$${X}_{CO}\left(\%\right)=\frac{C{O}^{in}-C{O}^{out}}{C{O}^{in}}\times 100$$
$${X}_{{O}_{2}}\left(\%\right)=\frac{{O}_{2}^{in}-{O}_{2}^{out}}{{O}_{2}^{in}}\times 100$$
$${S}_{C{O}_{2}}\left(\%\right)=\frac{1}{2}\frac{C{O}^{in}-C{O}^{out}}{{O}_{2}^{in}-{O}_{2}^{out}}\times 100$$

To determine the activation energy (\({E}_{a}\)) of PROX via the Arrhenius equation, the CO conversion (in the range below 30%) was measured for the oxidized and reduced catalyst at 75, 80, 85, and 90 °C.

Furthermore, we measured the reaction orders via varying the concentration of one reactant (at 70 °C and a total flow of 50 mL min−1). Accordingly, to determine the CO order, the CO concentration was varied between 0.5 and 2 vol%, while the O2 and H2 concentrations were kept constant at 1 vol% and 50 vol% (yielding about 8–9% CO conversion). To determine the O2 order, the O2 concentration was varied between 0.5 and 2 vol%, while the CO and H2 concentrations were kept at 1 vol% and 50 vol%, respectively.

2.3 Computational Methods

Density functional theory calculations were carried out in spin-polarized fashion using the Vienna Ab initio Simulation Package (VASP) [39] utilizing the projector augmented-wave method (PAW) [40, 41]. The generalized gradient approximation (GGA) was used in the parameterization according to Perdew, Burke, and Ernzerhof (PBE) [42]. The plane wave cutoff energy was set to 450 eV for all surface calculations. The following electron configurations were calculated explicitly, Cu(3d10 4s1), O(2s2 2p4), C(2s2 2p2), and H(1s1). The strong correlations of d states were treated by a Hubbard model using the Dudarev formalism (DFT + U) [43]. The electronic self-consistent loop was considered converged when the energy changes became smaller than 1 × 10−8 eV. Atomic positions were optimized until the Hellmann-Feynman forces acting on each atom dropped below 0.02 eV/Å. According to the role of CeO2 described in the Introduction, our DFT study focused on the CuO surface solely.

For these calculations, the most stable CuO(111) surface was chosen, as it has the lowest surface energy (see Tables S1, S2 in the Supplementary Information (SI)). The CuO(111) surface was cut from the calculated CuO unit cell, using 8 × 8 × 8 \(\varvec{k}\)-points and a cut-off of 600 eV. In this study we used a 2 × 1 periodic surface supercell p(2 × 1)-CuO(111) with a slab thickness of 5 layers and a vacuum gap of 15 Å keeping both “bottom” layers fixed at the bulk distance. A 2 × 3 × 1 Γ-centered \(\varvec{k}\)-point grid was sampled for surface calculations of CuO(111). van-der Waals interactions were included using the DFT-D3 method [44].

We chose an effective \(U\) value (\({U}_{eff}\)) of 7 eV as it tested quite reliable to describe the lattice constant and electronic properties, e.g., magnetic moment, derived from experimental work, see Fig. S1 and Table S4 of the SI.

The oxygen vacancy formation energy, \({E}_{vac}\), on a surface slab, was computed according to

$${E}_{vac}={E}_{defect/surf}+{\frac{1}{2}E}_{{O}_{2}}-{E}_{surf} \left(1\right)$$

where \({E}_{defect/surf}\) is the total energy of a surface slab with an oxygen vacancy, VO, and \({E}_{{O}_{2}}\) is the total energy of isolated triplet O2 in the gas phase. The adsorption energy, \(E{\left(X\right)}_{ads}\), of an adsorbed molecule X (i.e., CO or H2) was calculated in the following way

$${E}_{ads}\left(X\right)= {E}_{X/surf}-\left({E}_{surf}+{E}_{X}\right) \left(2\right)$$

where \({E}_{X/surf}\) is the total energy of the molecule adsorbed on the surface, \({E}_{surf}\) is the total energy of the bare surface, and \({E}_{X}\) is the total energy of an isolated molecule in the gas phase. A negative \({E}_{ads}\) indicates a binding interaction. The desorption energy of molecule X, i.e., CO2 and H2O, \({E}_{des}\left(X\right)\) is the negative value of \({E}_{ads}\left(X\right)\). A charge analysis according to the procedure suggested by Bader [45] was carried out to evaluate the change in charge \({\Delta }q\) as described in the SI. All simulation files used in this study can be found on the ioChem-BD platform [46].

3 Results

3.1 Catalyst Characterization

Figure 1a illustrates the synthesis of the CuO/CeO2 nanosphere catalysts. Figure 1b–d shows TEM images, displaying CeO2 hollow spheres covered outside by CuO nanoparticles (partly decorated by CeO2). The average size/diameter of the CuO/CeO2 nanospheres was about 180 nm. The maps of energy dispersive X-ray analysis (EDX) (Fig. 1e and Table S1) display that cerium, copper, and oxygen are very homogeneously distributed over the hollow spheres and indicate a loading of ~ 10.5 wt% Cu or ~ 13 wt% CuO.

Fig. 1
figure 1

Synthesis and electron microscopy of CuO/CeO2: a schematic illustration of the synthesis, b–d TEM images, and e EDX elemental mapping of the CuO/CeO2 catalyst (oxidized at 300 °C)

Full size image

To determine the catalyst reducibility, CO- and H2-TPR were performed (Fig. 2a, b). For CO-TPR, a two-step reduction (Cu2+ → Cu+ → Cu0) was indicated by peaks at 180 and 215 °C. In contrast, for H2-TPR, rather a one-step reduction (Cu2+ → Cu0) was suggested by a dominant peak at 205 °C. Overall, this behavior is quite similar to that of Co3O4 reported before [27]. As discussed below, these temperatures (representative of bulk reduction) are much higher than the typical reaction temperatures of CO oxidation and PROX. Nevertheless, surface reduction may still occur at lower temperatures.

Figure 2c shows XRD patterns of the pre-oxidized (20 vol% O2 in He, at 400 °C) CuO/CeO2 catalyst, as well as of Cu0/CeO2, obtained by the reduction at 300 °C in 5 vol% H2 (i.e., much higher than the 205 °C in Fig. 2b). The crystallite size was calculated using the Rietveld refinement and was 6 nm for CuO (PDF no; 00-006-2679, monoclinic) and 17 nm for Cu0 (PDF no; 04-009-2090, cubic). The diffractogram confirms a complete reduction of CuO at 300 °C, while CeO2 (PDF no; 04-003-1755, cubic) remained unaffected in both oxidized and reduced samples. The nanosphere shell is composed of CeO2 crystals of about 10 nm in size (9.6 and 10 nm for oxidized and reduced, respectively).

Fig. 2
figure 2

Temperature-programmed reduction and X-ray diffraction of catalysts: Mass spectrometry profiles obtained during a CO-TPR of CuO/CeO2, b H2-TPR of CuO/CeO2, and c XRD patterns of oxidized CuO/CeO2 and reduced Cu0/CeO2

Full size image

3.2 CO Oxidation

Figure 3a, b displays results of CO oxidation (1 vol % CO and 1 vol % O2) on nanosphere CuO/CeO2, analyzed simultaneously by MS and GC. An ignition behavior [47, 48] was observed at 95 °C, leading to more than 90% CO conversion. Under similar feed conditions, a maximum CO conversion was observed at ~ 150 °C for Co3O4 [27]. For a conventional CuO/CeO2 morphology, a temperature of > 100 °C was needed for 90% CO conversion [49, 50]. For pure CeO2, CO oxidation sets in above 325 °C [51, 52]. Interestingly, for the nanosphere catalysts, the reaction onset temperature was independent of the catalyst pretreatment, i.e., the pre-oxidized and reduced (both at 300 and 500 °C) catalysts behaved identically (Fig. 3c). It seems that the same reactive phase is obtained under reaction conditions. The similar activity despite the ~ 2.5 times lower dispersion of Cu0/CeO2 may point to interface sites as active centers, as their number is less affected by particle size.

Fig. 3
figure 3

Temperature programmed CO oxidation on CuO/CeO2 (oxidized at 300 °C, 1 vol% CO and 1 vol% O2 in He) a MS and b GC of CuO/CeO2. c The same for different pretreatments: oxidized at 300 °C (blue); oxidized and reduced at 300 °C (green); oxidized and reduced at 500 °C (orange); CO oxidation (5 vol% CO and 10 vol% O2 in He, total flow 50 mL min−1)

Full size image

3.3 PROX

Analogous experiments were carried out for the PROX reaction. Figure 4a, b compares MS results for CuO/CeO2 and Cu0/CeO2, with the reaction properties being basically identical: up to 100 °C, the catalyst shows 100% CO2 selectivity. Once more, the same reactive state seems to manifest under reactive conditions. GC analysis for CuO/CeO2 is shown in Fig. 4c, d, confirming the MS results (GC is less sensitive to small amounts of water than MS, though). Caputo et al. reported 100% CO2 selectivity below 144 °C on 4 wt% CuO/CeO2 catalyst (both oxidized and reduced) for preferential CO oxidation [53]. Above 125 °C, significant water formation sets in, leading to a drop in selectivity, although methane formation did not occur. Above 175 °C, the CO conversion \({X}_{CO}\) and CO2 selectivity \({S}_{C{O}_{2}}\) further decreased, likely due to the onset of a reverse Water-Gas shift reaction (CO2 + H2 ↔ CO + H2O), for which the Cu-based catalyst supported on CeO2 is known to be activated at temperatures around 200–300 °C [54, 55].

It is noteworthy that pure CeO2 shows PROX performance at high temperature with CO oxidation starting at around 270 °C (Fig. S2). This confirms that CuO on CeO2 strongly decreases the temperature of CO oxidation, enabling PROX performance at low temperature for this type of catalyst.

Fig. 4
figure 4

Temperature programmed PROX reaction on CuO/CeO2 for a reaction mixture of 1 vol% CO, 1 vol% O2, 50 vol% H2 and 48 vol% He: a oxidized at 300 °C and b after oxidation/reduction at 300 °C. PROX on oxidized CuO/CeO2 catalyst: c CO and O2 conversion, and d CO2 selectivity

Full size image

The Arrhenius plots in Fig. 5a, b display the activation energies (\({E}_{a}\)) for CuO/CeO2 (60.2 ± 1.7 kJ mol−1) and Cu0/CeO2 (58.1 ± 2.1 kJ mol−1); note that the fitted values for both catalysts exhibit only a small difference (2.1 kJ mol−1) which is within the measurement accuracy. This once more demonstrates that the initial state of the catalyst after pretreatment is not decisive for PROX. The \({E}_{a}\) values agree with those reported in the literature for selective CO oxidation over copper–ceria catalysts (\({E}_{a}\) range of 55–57.2 kJ mol−1) [56, 57]. The \({E}_{a}\) of CO oxidation for copper-ceria catalysts is ca. 20 kJ mol−1 lower than that of pure CuO catalysts, confirming that ceria aids in catalytic CO oxidation [58].

For PROX, the apparent reaction order in CO and O2 at 70 °C (Fig. 5c, d) were determined to be \({n}_{CO}\) = 1.00 and \({n}_{{O}_{2}}\)= 0.56, respectively, very close to the values expected from the reaction stoichiometry. For CO oxidation over CuO/CeO2 similar values of \({n}_{CO}\)= 0.72–0.97 and somewhat smaller values for \({n}_{{O}_{2}}\)= 0.00-0.25 were typically reported [32, 56,57,58,59,60,61].

Fig. 5
figure 5

PROX on CuO/CeO2. Activation energies of a oxidized CuO/CeO2 and b reduced Cu0/CeO2 catalysts. Reaction orders of c CO and d O2 on CuO/CeO2

Full size image

The active state of the current CuO/CeO2 nanosphere catalysts has not yet been determined by in situ/operando studies [62, 63]. Nevertheless, sites at the Cu-O-Ce interface were often invoked previously to rationalize the high activity and selectivity [21,22,23,24]. In order to differentiate the contributions of Cu2+ and Cu+ on the CuO nanoparticles for low-temperature oxidation, we will take up an alternative approach herein. We employ computational modeling to rationalize the experimental results, with a focus solely on the active surface of CuO(111) [24].

3.4 Computational Modeling

In order to obtain atomistic insights into the PROX reaction and the CO2 selectivity, density functional theory (DFT + U) calculations were carried out to study the formation of an oxygen vacancy, its re-oxidation by O2, adsorption of CO and H2 and reaction at various sites on a p(2 × 1)-CuO(111) model surface. Finally, PROX reaction pathways were evaluated, with a focus on the CO2 selectivity over CuO at low temperature.

3.4.1 Surface Structure and Oxygen Vacancy Formation

The CuO(111) surface model has antiferromagnetic coupled spins in which the magnetic moment per Cu remained at 0.658 \({\mu }_{B}\), alike the bulk structure. This is consistent with previous theoretical and experimental reports of 0.63 and 0.68 \({\mu }_{B}\), respectively [64, 65]. Four distinct surface sites of Cu and O atoms exist at the CuO (111) surface, as indicated in Fig. 6a, b. A three-fold coordinated O atom, O3c, represents the outermost surface atom while the four-fold coordinated O atom, O4c, represents the innermost surface atom.

Fig. 6
figure 6

DFT study of the p(2 × 1)-CuO(111) structure: a Side view and b top view. Blue and red spheres represent Cu (Cu2+) and O atoms, while ocean green represents reduced Cu (Cu+). Oxygen vacancy sites VO are shown in c at O3c (VO@O3c) and d O4c (VO@O4c). The re-oxidation of the VO@O3c site by O2 is shown in e. The solid black line indicates the unit cell used

Full size image

CO oxidation and PROX on reducible oxides are well-known to proceed via a Mars-van-Krevelen mechanism, which involves the creation of oxygen vacancies [3, 24, 66,67,68,69]. Creating such an oxygen vacancy, VO, leads to two Cu atoms in the CuO(111) surface being reduced. At first, we removed the three-fold, O3c, and the four-fold coordinated O atoms, O4c, which allows a VO density of \(1/8\). Surfaces with VO are shown in Fig. 6c, d. The oxygen vacancy formation energies \({E}_{vac}\) were calculated using Eq. (1) and are listed in Table 1. Forming an oxygen vacancy at O3c site (VO@O3c) is by 0.47 eV energetically more favorable than at the site O4c (VO@O4c), most likely because the latter is coordinatively more saturated by four nearest neighbors (4NN) [70, 71].

Table 1 Oxygen vacancy formation energies \({E}_{vac}\), of three-fold O3c (VO@O3c) and four-fold O4c (VO@O4c) coordinated O atoms, and adsorption energies \({E}_{ads}\left(X\right)\), of CO, H2, and O2. The desorption energies \({E}_{des}\left(X\right)\) of CO2 and H2O are included. Cu species refer to Cu3c. Energies are given in eV
Full size table

When one VO is created (\(\theta\)=1/8), two excess electrons localize at and reduce specific Cu atoms (Cu2+ + e− -> Cu+), as drawn in ocean green spheres in Fig. 6c, d. In the calculations, these reduced species are indicated by a change in the local magnetic moment and their modified atomic charge \({\Delta }q\), gaining 0.41 e−, Table S6. To create metallic Cu (Cu0) on the CuO (111) surface, a VO coverage of \(\theta \ge\) 5/8 is needed, resulting in a change in charge \({\Delta }\text{q}\) of 0.94 e− [64].

Furthermore, we modeled the re-oxidation of the energetically favorable vacancy site (VO@O3c) by an O2 molecule, in addition creating the new site Oad at the defective surface (Fig. 6e). Adsorption of O2 is strongly favorable with an energy of − 1.99 eV compared to the adsorbed CO and H2 species on bare surface (Sect. 3.4.2). This is in line with the calculated VO formation energy. Notably, the adsorbed O2 molecule shows an O–O bond length of 1.50 Å, which is elongated as compared to gas phase O2, 1.24 Å, suggesting that adsorbed O2 is activated and nearly dissociated. In more detail, upon O2 adsorption, electrons from reduced Cu (Cu+) sites transfer to the oxygen moiety and occupy 2π∗ orbitals, leading to an elongation of the O–O bond, so that O2 becomes a peroxo species, O22− (see also the SI) [72,73,74]. Such peroxo species have also been discussed in the context of tuning the surface reactivity [75]. As reported in the literature, VO thus plays a vital role in the adsorption of the O2 molecule on metal oxide surfaces [73, 76, 77].

In view of the energetically demanding VO formation energies, promotion by interaction with CO or H2 is required. Thus, molecular adsorption of CO and H2, as well as surface reactions, are discussed in the next section.

3.4.2 CO and H2 Adsorption on the CuO(111) Surface

The stability of surface adsorption sites was probed in the presence and absence of oxygen vacancies. First, we studied the interaction of CO and H2 with surface oxygen species, O3c, and O4c. CO adsorbs exothermically on the activated O3c site, Fig. 7a, whereas at the O4c site, CO moves away from the surface. The adsorption of CO on activated O3c is quite exothermic \({E}_{ads}\left(CO\right)=\) − 1.13 eV. The activated lattice O3c atom is pulled out of the surface, moving by 2.43 Å and reacting with CO, forming an adsorbed CO2 molecule. Consequently, CO2 formation produces an oxygen vacancy VO (VO@O3c), yielding two excess electrons which localize at Cu2+ to form Cu+ species. The adsorbed CO2 molecule is linear with an OCO angle of 179.2° and an average C–O distance of 1.18 Å close to the typical CO2 distance of 1.16 Å in the gas phase. The calculated energy \({E}_{des}\) for subsequent desorption of such a CO2 molecule is 0.28 eV. Keeping in mind that entropy effects are not included in the presented energetics, we can safely assume that the desorption of the CO2 molecule would be favorable in free energy. The adsorption of CO at Cu2O(100) surfaces has also been modeled, with a CO2 formation energy of ~ 1.7 eV [78, 79].

Alike molecular CO adsorption, the dissociative adsorption of H2 at O3c forms a H2O molecule, as shown in Fig. 7e. The adsorption energy is − 0.87 eV, i.e., 0.26 eV less favorable than for CO. Furthermore, the desorption energy of the formed water molecule, \({E}_{des}\left({H}_{2}O\right)\) = 0.77 eV, is substantially larger than that of CO2. This rationalizes the more facile conversion of CO to CO2 as the formed water would remain on the surface, blocking further H2 activation. The calculated binding and desorption energies agree with previously published values of − 0.83 and 0.78 eV, respectively [64]. The formed water molecule binds to a surface Cu atom at a distance of 2.05 Å and the measured bond angle and average H–O distance are 108.16° and 0.99 Å, respectively. These values are close to reported theoretical work with a bond angle and H-O distance of 107.5° and 1.0 Å, respectively [64].

Adsorption of CO and H2 on Cu sites of the CuO(111) surface were studied next, locating two possible sites: Cu3c (Cu2+) and reduced Cu3c (Cu+) as listed in Table 1. The calculated adsorption energies support a spontaneous adsorption process for both molecules. CO likely adsorbs at the Cu3c (Cu2+) site with an adsorption energy of − 0.62 eV, Fig. 7b. In contrast, CO does not adsorb at the Cu4c (Cu2+) site and moves away, likely due to bond competition. As described above, when a vacancy is introduced (e.g., by a previous reaction with CO), two Cu3c (Cu2+) atoms are reduced and transformed to Cu+. In contrast to Cu2+, these Cu+ ions adsorb CO twice as strong with \({E}_{ads}\left(CO\right)=\) − 1.29 eV, forming a CO–Cu+ carbonyl species, Fig. 7c. This is in line with experimental reports of Cu+ suggested as active centers for CO oxidation [10, 80,81,82,83,84] while Cu2+ was also reported [53]. The CO molecule binds to metallic Cu (Cu0) on the surface exothermically with an adsorption energy of − 0.92 eV (Fig. S4d) that is in between the other two oxidation states. However, in oxygen–containing feed (high \({\Delta }{\mu }_{O}\)), Cu0 is hardly stable as the formation of CuOx species occurs [85].

The interaction of H2 with Cu sites on CuO(111) was explored as well, (Fig. 7f, g and Fig. S4i of the SI), with H2 placed on Cu3c (Cu2+), Cu+, and Cu0 atoms. The H2 molecule binds exothermically to the Cu+ site with an adsorption energy of − 0.46 eV, 0.30 eV stronger than the Cu2+ site, and agrees with a previously reported value of \({E}_{ads}\) = − 0.47 eV [86]. In contrast, it shows very weak adsorption at Cu0, \({E}_{ads}\left({H}_{2}\right)=\) − 0.07 eV, which is in good agreement with a previous study of H2 adsorption on metallic Cu(111)\(, {E}_{ads}\left({H}_{2}\right)\) = − 0.07 eV [87]. Overall Cu+ adsorbs both H2 and CO better than the other oxidation states of Cu.

During the reaction the formed vacancies on the surface will be spontaneously replenished by gaseous O2 forming an active oxygen species at the surface, Oad. Thus, the adsorption of CO and H2 at Oad was studied accordingly. Oad reacts with CO and spontaneously forms CO2 with a strongly exothermic reaction energy of − 4.02 eV (Fig. 7d). In the same way, the adsorption of H2 also forms H2O with an adsorption energy of − 3.85 eV (Fig. 7h). As expected, the Oad species at the VO–CuO (111) surface is quite reactive in the oxidation process. A study of oxygen vacancy rich La0.8Sr0.2CoO3 also suggested that the presence of nearby VO enhances the CO oxidation/interaction with adsorbed Oad species [73]. Note that the possibility of hydroxyl formation [88] from H2 adsorption at O3c was also observed at a VO–CuO(111) surface (see Fig. S5). To complete the catalytic cycle, the formed species CO2 and H2O desorb from the surface with energies of 0.30 and 0.88 eV. This again indicates a strongly favored CO2 desorption, which rationalizes the experimentally observed high CO2 selectivity. A possible reaction mechanism is described in the following section.

Fig. 7
figure 7

Adsorption complexes at bare surfaces and those with oxygen vacancies: Top views of CO and H2 adsorption on CuO3c sites a, e, while b, c and f, g show adsorption on Cu sites. d and h represent CO2 and H2O formed on the extra oxygen species, Oad, generated from the re-oxidation of the VO@O3c site, respectively. Blue, red, and pink spheres represent Cu (Cu2+), O, and H atoms, respectively, while ocean green represents reduced Cu (Cu+)

Full size image

3.4.3 Suggested Reaction Pathways on the CuO(111) Surface

Following the results on the interaction of O2, CO, and H2 at CuO(111) we suggest two reaction pathways (Fig. 8) based on a Mars-van-Krevelen (MvK) mechanism [24, 89, 90]. Adsorbing CO and H2 molecules from the gas phase on Cu2+ (a1, b1) is slightly exothermic, but to form CO2/H2O and continue with the catalytic cycle the adsorption at an oxygen site is needed. Adsorption at O3c forms CO2 (a2)/H2O (b2) and one VO, with the CO2 formation being slightly preferred. The formed CO2 easily desorbs into the gas phase, while H2O desorption is substantially more challenging. These energies are compatible with the experimental results where CO2 is preferentially produced up to ~ 100 °C, whereas above H2O formation sets in and dominates > 175 °C (Fig. 4). Note that the transformation of CO to CO2 and H2 to H2O on activated O3c occurs without barrier as a downhill process except for the very first activation. In line with previous studies [91, 92], our nudged elastic band calculations [93, 94] showed these barriers to be smaller than ~ 0.5 eV for CO and smaller than ~ 0.7 eV for H2, around/below the “barrier” for the desorption process. The reaction rate will thus be controlled by the desorption process, for which we have determined a significant difference between H2O and CO2.

The formed vacancy VO has two Cu+ neighbors allowing further adsorption of CO (a4)/H2 (b4). Alternatively, O2 may repopulate the vacancy site (a5-b5), resulting in an added oxygen species, Oad. At this special oxygen another CO or H2 may adsorb and form CO2 (a6) or H2O (b6). Yet again, CO2 (a7) desorbs more favorably by 0.6 eV into the gas phase than H2O (b7), completing the reaction cycle.

Fig. 8
figure 8

Reaction pathways of a CO and b H2 oxidation. Energy baseline of CO2 and H2O products are marked in red and blue. Energies are reported relative to the bare surface and CO/H2 in the gas phase. Blue, red, and pink spheres represent Cu (Cu2+), O, and H atoms, respectively, while ocean green represents reduced Cu (Cu+)

Full size image

4 Summary

CuO/CeO2 nanosphere catalyst particles are active in CO oxidation and CO-PROX and were tested in the present work via experimental and computational approaches. CO oxidation ignited at 95 °C, while under PROX conditions, 100% selectivity to CO2 was maintained up to 100 °C, overall showing that the nanosphere catalyst works fairly well. Oxidative or reductive pretreatments, initially forming Cu2+ or Cu0 moieties, respectively, yielded no discernable difference in catalytic steady-state performance and activation energies. Hence, the same active catalyst phase is formed under reaction conditions.

Using DFT modeling, the energetics of likely reaction pathways on Cu oxide surfaces was further studied, indicating a conventional Mars-van-Krevelen (MvK) type mechanism. Upon CO adsorption and conversion to CO2, oxygen vacancies on Cu oxide surfaces were easily formed, rather independent of the oxidation state of neighboring Cu centers. For further adsorption, we determined that Cu+ is the preferred binding site, with CO adsorption being much stronger than H2 adsorption. Molecular O2 adsorption and subsequent activation proceeds at a vacancy, creating highly active O species. Upon interaction with CO the formed CO2 binds only weakly at the surface allowing for rapid desorption and freeing sites for further turnover, promoting CO2 selectivity as observed experimentally. The adsorption/activation of H2 is less feasible and the resulting water binds much stronger to the surface. This explains the undesired water formation only at high CO conversions and above 100 °C. Ceria is certainly beneficial due to its oxygen vacancies, but the support was not considered herein. Operando studies are planned for the future to elucidate active phases under reaction conditions.