Introduction

Ni-based catalysts are widely used in many important industrial reactions, such as hydrogenation^[1], oxidation^[2], and reforming of alcohols^[3]. However, carbon deposition on Ni at high temperature during operation often leads to deactivation of the catalyst^{[4, 5]}. Previous studies show that Ni/CeO₂ is an efficient and low-cost catalyst for many important catalytic reactions including water-gas shift reaction^[6], ethanol steam reforming^[7], dry reforming of methane^[8], which can also be very stable at reaction conditions. It is shown that Ni ions can be embedded in the ceria fluorite lattice, i.e., the Ce ions can be replaced by the Ni ions in the mixed oxide and the structure maintain the fluorite type^[6]. Such special structural configuration is considered as the key to the good catalytic properties of Ce_1-xNi_xO_2-y solid solution. X-ray diffraction (XRD) data, which is based on the long-range order of the materials, shows that the highest concentration of Ni ions in such solid solution is within the range of 10%~12%. However, the value of surface Ni species during catalytic reactions occurrence, which is important for catalytic performance, may be different from that extracted from XRD measurements.

Solid-state nuclear magnetic resonance (NMR) spectroscopy, a powerful characterization method providing rich short-range order information for solids, is complementary to the information obtained from XRD^{[9, 10]}. Recently, ¹⁷O solid-state NMR spectroscopy, combined with density functional theory (DFT) calculations, was demonstrated as a new approach to extract surface structural information of oxide nanomaterials^[11-14]. In particular, oxygen ions at 1^st, 2^nd and 3^rd layers of ceria nanoparticles^[15] can be differentiated based on their distinct NMR shifts. Since Ni ion is likely in a +2 oxidation state^[6], large electronic dipole interactions can lead to broad resonance peaks or decreases in intensities of sharp peaks. Such changes give an opportunity to analyze the concentrations of surface Ni ions. Here we demonstrate the analysis of Ni ions' concentration in the surface of Ni/CeO₂ based on ¹⁷O magic-angle spinning (MAS) NMR data.

1 Experimental section 1.1 Material preparation

Ceria nanorods exposing {111} facets were prepared by a hydrothermal method according to the procedures introduced by Mai et al.^[16]. The obtained sample was calcined at 700 ℃ and denoted as CeO₂-700. Ni/CeO₂ samples with different concentrations of Ni ion {n(Ni)/[n(Ni)+n(Ce)]=10% and 20% in the starting materials} were prepared by an impregnation method using CeO₂-700 followed by calcination of the Ni-containing CeO₂ at different temperatures (300 ℃ and 500 ℃). These samples were denoted 10 % NiCe-500, 20% NiCe-300 and 20% NiCe-500 according to their Ni concentrations and calcination temperatures. The materials still expose {111} facets after impregnation and calcination (Fig. S1 and S2, available in the online version).

1.2 ¹⁷O-enrichment procedures

Samples were first heated at 300 ℃ for 3 h under vacuum before cooling down to the room temperature. After that, a same amount of H₂¹⁷O (90 % in ¹⁷O, Isotec Inc.) were introduced to the samples via a vacuum system and the mixture were treated at 100 ℃ or the room temperature.

1.3 NMR experiments

¹⁷O MAS NMR experiments were performed on a 9.4 T Bruker Avance Ⅲ spectrometer with a 4.0 mm MAS probe operating at 54.2 MHz at ambient temperature. All the samples were packed inside the rotors in a N₂ dry glove box and spun at a MAS rate of 14 kHz. A single pulse sequence with an excitation pulse of 1.3 μs corresponding to a hard π/6 pulse was used in all experiments. The recycle delays were set to 0.5 s and 0.1 s, for CeO₂-700 and Ni-containing samples, respectively, both of which were optimized to ensure quantitative measurements. All the chemical shifts were externally referenced to water at δ 0.0.

2 Results and discussion 2.1 The influence of Ni concentration and calcination temperatures on ¹⁷O MAS NMR spectra

Fig. 1 shows the ¹⁷O MAS NMR spectra of different ¹⁷O-enriched samples treated at room temperature. For all the samples, a strong and sharp peak at δ 890, along with shoulders at δ 874 and δ 857 can be observed. According to our previous results^[15], the peaks at δ 857 and δ 874 can be assigned as oxygen ions in the 3^rd layer and the bulk part of the nanorods (4^th layer or deeper), respectively. The major resonance at δ 890 has been observed using dynamic nuclear polarization (DNP) NMR^[17], which was assigned to the oxygen ions at 2^nd layer. However, such strong signal arising from the oxygen species of the 2^nd layer seems to be contradicted to the very small peak at approximately δ 1 020 due to oxygen species in the 1^st layer. This discrepancy may be related to the presence of a relatively large amount of water adsorbed on the surface for the samples studied here and the one with DNP NMR, which was removed by heating the materials under vacuum at 100 ℃ in our previous studies. Therefore, the signal at δ 890 may be tentatively assigned to the surface oxygen species (1^st and 2^nd layer) with water molecules nearby. Nonetheless, a decrease in the intensity can be clearly observed for all the signals at lower frequencies (δ 890, 874 and 857). This can be attributed to the incorporation of Ni²⁺ in the lattice. The Ni²⁺ can broaden the resonance peak. The resonance signals of oxygen directly bound to Ni²⁺ are usually as broad as several thousand ppm due to large electron-nuclear spin interactions, while oxygen ions further away exhibit relatively narrow linewidths^{[18, 19]}. In our case, since the linewidths of all the resonance peaks at lower frequencies are only around δ 20~30, the signals due to the oxygen ions bound to Ni²⁺ are practically eliminated. Clearly, with the increase of Ni in the starting material, the signal intensity decreases, indicating more Ni²⁺ are incorporated in the lattice. At the same time, the increase in the thermal treatment temperature also leads to the decrease in the spectral intensity, suggesting that higher temperature can cause more Ce ions to be replaced by Ni ions. The relative intensities arising from the oxygen ions in the 3^rd layer in the Ni-containing samples compared to the parent CeO₂-700 sample, are extracted by using NUTS NMR data processing software (Acorn NMR Inc.) (Fig. S3, available in the online version) and shown in Table 1.

¹⁷O MAS NMR experiments were also performed on the ¹⁷O-enriched samples at 100 ℃ and the corresponding spectra are shown in Fig. 2. The resonance peaks at δ 1 033, 925 and 878 can be assigned to the oxygen ions in the 1^st, 2^nd layer and bulk part of ceria nanorods, respectively, although the resolution does not allow the oxygen ions at the 3^rd layer to be resolved. Similarly, the intensities of the peaks at 2^nd layer and the bulk part of the sample decrease with the incorporation of Ni. Unexpectedly, the intensity of the major resonance at δ 925 of 20% NiCe-500 arising from the oxygen ions at 2^nd layer is stronger than 10% NiCe-500. XRD analysis (Fig. S4, available in the online version) shows that NiO phase is present in 20% NiCe-500. Thus, further quantitative analysis was not performed on this sample. The intensity of the signal centered at δ 1 033 due to the oxygen ions on the surface does not seem to vary much with the amount of Ni, which may be associated with its much broader linewidths. The fractions of the intensities of the signals owing to the oxygen ions in the 2^nd layer in Ni/CeO₂ samples, as compared to CeO₂-700, are obtained and shown in Table 2.

2.2 Concentration of surface Ni ions calculation

The ceria nanorods prepared mostly expose {111} facets and oxygen ions are 3-(surface) or 4-coordinated (subsurface or other). If one of the Ce⁴⁺ bound with oxygen ion is replaced by Ni²⁺, such oxygen ion will be invisible in the ¹⁷O NMR spectrum due to the large electron-nuclear spin interactions. Therefore, the concentration of surface Ni²⁺ should correlate to the decrease of ¹⁷O signals in Ni/CeO₂, if another assumption is accepted that the ¹⁷O NMR resonance of oxygen ions in the 3^rd coordination shell is not affected (i.e., ¹⁷O-Ce-O-Ni). However, it is possible that one oxygen ion directly connects to more than one Ni²⁺ ions, and some further constraints on the distribution of Ni²⁺ needs to be added in order to extract the concentrations of surface Ni. For the convenience of quantitative analysis, the coordinated conditions of different oxygen ions and Ni ions are listed below, according to the structure of CeO₂ {111} facets (Fig. 3):

● For oxygen ions in the 2^nd layer: each Ni ion in the 1^st layer influences m oxygen ions in the 2^nd layer, while each Ni ion in the 2^nd layer influences 1 oxygen ion in the 2^nd layer;

● For oxygen ions in the 3^rd layer: each Ni ion in the 1^st layer influences 1 oxygen ion in the 3^rd layer, while each Ni ion in the 2^nd layer influences n oxygen ions in the 3^rd layer.

Parameters m and n reflect the distributions of Ni ions in the 1^st and 2^nd layer of metal ions (Ce and Ni). As shown in Fig. 4, the ranges of parameters m and n have two boundary conditions: dispersed and closed packing. The parameter m or n equals 3 in the condition of dispersed packing, in which no Ni-O-Ni environment is present [Fig. 4(a)]. If all the Ni²⁺ ions are connected and Ni-O-Ce environments are only found on the boundaries, parameter m or n is infinitely close to 1. Therefore, the range for the possible values of the parameter m or n is (1, 3].

Based on the above assumptions and constraints, the fractions of surface Ni ions in Ni/CeO₂ can be extracted. For a given sample, we can define:

x₁ = fraction of Ni ions in the 1^st metal ion layer calculated by Ni/(Ni+Ce);

x₂ = fraction of Ni ions in the 2^nd metal ion layer calculated by Ni/(Ni+Ce);

A = decrease in intensity of oxygen ions in the 2^nd layer compared to CeO₂-700;

B = decrease in intensity of oxygen ions in the 3^rd layer compared to CeO₂-700.

The decrease of the ¹⁷O NMR intensity arising from the oxygen ions in the 2^nd layer is caused by Ni ions at the 1^st layer and 2^nd layer. After removing double counting, the first equation is obtained:

$ m{x_1} + {x_2}(1 - m{x_1}) = A $

(1)

Similarly, there is a second equation for the decrease of the ¹⁷O NMR intensity arising from the oxygen ions in the 3^rd layer:

$ n{x_2} + {x_1}(1 - n{x_2}) = B $

(2)

A and B of different samples (Table 3) can be obtained from Table 2 and 1, respectively.

2.2.1 Concentration of Ni in the top two layers of 10% NiCe-500

Fig. 5(a) and 5(b) show the variations of x₁, x₂ with parameters m and n. The values of x₁ and x₂ vary a lot as the parameter m or n changes. The average fractions of Ni ions in the top two layers, which are most important in catalytic applications, however, vary in a much smaller range [Fig. 5(c)]. For example, Fig. 6 shows the variations of the values of x₁, x₂ and (x₁+x₂)/2 with the parameter n when m=1.1 in 10% NiCe-500. The value of (x₁+x₂)/2 is close to 0.16 whatever the value of n is although different values of n mean different Ni distribution in the 2^nd layer.

Similar results are found when the parameter m changes. Fig. 7 shows the variations of (x₁+x₂)/2 with the parameter n when the parameter m equals 1.1, 1.7, 2.3 and 3.0, the values of (x₁+x₂)/2 are in the range of 0.16~0.17, 0.12~0.16, 0.10~0.16 and 0.09~0.16, respectively. That is, the values of (x₁+x₂)/2 vary from 0.09 to 0.17 in all the possible values of parameters m and n. Therefore, the average surface Ni concentration in the 1^st and 2^nd metal ion layers of 10% NiCe-500 is in the range of 9%~17%.

It is clear that the higher the parameter m or n, the lower the values of (x₁+x₂)/2. When both the parameter m and n reach 3, meaning that Ni ions are distributed evenly in the top two layers of ceria, x₁=0.09 and x₂=0.10. This is consistent with the previous research of Ce_1-xNi_xO_2-y solid solution that the highest fraction of Ni is about 10%~12%^[6]. However, if Ni ions form some kinds of clusters, our data suggest a higher concentration of Ni in the top two layers would be likely.

2.2.2 Concentration of Ni in the top two layers of 20% NiCe-300

Similarly, the values of x₁ and x₂ change a lot with the variation of parameters m and n for 20% NiCe-300 [Fig. 8(a) and 8(b)], while the values of (x₁+x₂)/2 stay in a much smaller range [Fig. 8(c)]. Fig. 9 shows the variations of (x₁+x₂)/2 with the parameter n when the parameter m equals 1.1, 1.7, 2.3 and 3.0, the values of (x₁+x₂)/2 are within the range of 0.13~0.15, 0.10~0.14, 0.09~0.14 and 0.08~0.13, respectively. Therefore, the average surface Ni amount in the 1^st and 2^nd metal ion layers of 20% NiCe-300 is within the range of 8%~15%.

Again, the higher values of parameters m and n lead to the smaller values of (x₁+x₂)/2. The smaller fraction of surface Ni in 20% NiCe-300 as compared to 10% NiCe-500 can be predicted by observing smaller values for A and B. This suggests that a thermal treatment temperature of 300 ℃ is not as efficient to incorporate Ni in the ceria lattice as 500 ℃.

3 Conclusion

In this study, ¹⁷O MAS NMR experiments have been performed on ¹⁷O-labeled CeO₂ and Ni/CeO₂ samples. Considering the coordination of Ni and O ions, as well as the distribution of Ni in the 1^st and 2^nd layers of ceria nanorods, we have established quantitative relationships between the Ni fractions in top two metal ion layers and the relative peak intensities of oxygen ions in the 2^nd and 3^rd layers. Analysis shows that the Ni concentrations in top two metal ion layers of 10% NiCe-500 and 20% NiCe-300 are within the ranges of 9%~17% and 8%~15%, respectively. This method may be extended to other complicated oxides to extract the relative concentrations of the doped paramagnetic ions.