Structural Requirements for Chemoselective Ammonolysis of Ethylene Glycol to Ethanolamine over Supported Cobalt Catalysts

Abstract

Ethylene glycol is regarded as a promising C₂ platform molecule due to the fast development of its production from sustainable biomass. This study inquired the structural requirements of Co-based catalysts for the liquid-phase ammonolysis of ethylene glycol to value-added ethanolamine. We showed that the rate and selectivity of ethylene glycol ammonolysis on γ-Al₂O₃-supported Co catalysts were strongly affected by the metal particle size within the range of 2–10 nm, among which Co nanoparticles of ~4 nm exhibited both the highest ethanolamine selectivity and the highest ammonolysis rate based on the total Co content. Doping of a moderate amount of Ag further promoted the catalytic activity without affecting the selectivity. Combined kinetic and infrared spectroscopic assessments unveiled that the addition of Ag significantly destabilized the adsorbed NH₃ on the Co surface, which would otherwise be strongly bound to the active sites and inhibit the rate-determining dehydrogenation step of ethylene glycol.

1. Introduction

Ethylene glycol is the simplest vicinal diol and has been applied as an essential raw chemical in various fields, such as solvent, antifreeze agent, surfactant, and plastics [1]. At present, the major industrial route of manufacturing ethylene glycol consists of hydrolysis of ethylene oxide, while the latter is obtained from partial oxidation of petroleum-derived ethylene. New routes are also explored to synthesize ethylene glycol from other carbon resources (e.g., syngas and renewable biomass feedstocks) [2,3]. Therefore, it is expected that such emerging non-petroleum routes will increase the availability and sustainability of the ethylene glycol supply and enable ethylene glycol to act as an alternative C₂ platform molecule to replace ethylene oxide, which is very reactive and difficult to handle and transport [4].

In this last context, the synthesis of ethanolamine is an illustrative example showing the potential replacement of ethylene oxide by ethylene glycol in the production of basic chemicals. Ethanolamine is widely used as an absorbent for CO₂ and SO₂ and building blocks for pharmaceuticals, agrochemicals, and detergents, having an annual demand of two million tons [5]. In the industry, ethanolamine is obtained, traditionally, by the ammonolysis of ethylene oxide carried out with excess of NH₃ [4]. An alternative, as an attractive industrial pathway, is the chemoselective ammonolysis of ethylene glycol when solely H₂O is the by-product (Scheme 1). However, the latter route generally requires a higher temperature (e.g., 453~523 K) to occur, under which condition ethanolamine can readily convert to secondary amination products (e.g., piperazine, ethylenediamine, and diethanolamine) [6,7,8,9]. Consequently, developing novel catalysts for efficient formation of ethanolamine from ethylene glycol under mild conditions is highly desired.

Metal catalysts have been reported to be efficient in the ammonolysis of alcohols containing at least a primary or secondary hydroxyl group (e.g., aliphatic primary monohydric alcohols and dihydric primary alcohols) via a hydrogen-borrowing methodology [6,7,10,11,12,13,14,15,16]. This hydrogen-borrowing pathway generally consists of three elementary steps:

(1) Dehydrogenation of one of the hydroxyl group in the alcohol to form a C=O bond;

(2) Transformation of the C=O bond to a C=N bond through a nucleophilic attack by NH₃ and a subsequent dehydration.

(3) Hydrogenation of the C=N bond to give a primary amine using the hydrogen generated in the first step.

It is worth nothing that the primary amine product has higher nucleophilicity than NH₃ and is thus more reactive than the latter for the second step, which often leads to undesired formation of secondary amines [7]. Compared with the case of simple monohydric alcohols, both of the hydroxyl groups in ethylene glycol can be replaced by -NH₂ to form diethanolamine and the primary ethanolamine product can undergo a further cyclization to form piperazine (Scheme 1) [6]. These issues make the achievement of a high selectivity of ethanolamine from the ammonolysis of ethylene glycol quite challenging.

Although studies on the catalytic ammonolysis of ethylene glycol are still limited [7], various metal catalysts (such as Co [17,18,19], Ni [20,21], Cu [22], Ru [23,24], Pd [25], Pt [26]) have been applied for the hydrogen-borrowing amination of monohydric alcohols or the corresponding carbonyl compounds. Among them, Co, Ni, and Ru have drawn attention the most, because of their high activity and chemoselectivity for primary amines. It has also been reported that the size of the metal nanoparticles [7] and the doping with a second metal—Cu [27], Ag [28], Ru [29], Bi [30], Re [31], or Pt [32]—can significantly improve the catalyst performance, providing a basis for further optimizing the amination processes. For example, the liquid-phase ammonolysis of 1-octanol showed a higher chemoselectivity to the primary amine product on smaller supported Ru nanoparticles than those on larger ones, because the latter favored the self-coupling of primary amines [33]. The doping of Ag with a moderate amount in γ-Al₂O₃-supported Co catalysts (i.e., Co/γ-Al₂O₃) exhibited a strong Co-Ag interaction, improving both the activity and chemoselectivity for the 1-octanol ammonolysis [34].

In this study, we attempt to systematically investigate the ammonolysis of ethylene glycol in liquid phase using Co/γ-Al₂O₃ as a model catalyst and inquire the structural requirements of Co-based catalysts for the chemoselective formation of ethanolamine from ethylene glycol. In doing so, the size of the Co nanoparticles was controlled by tuning the Co loading, and typical doping metals (i.e., Cu, Ag, and Ru) were introduced separately. The activity and selectivity of these catalysts for ethylene glycol ammonolysis were rigorously compared at similar conversions and correlated with the catalyst structure. For the optimized Co-Ag/γ-Al₂O₃ catalysts, kinetic and spectroscopic assessments were further conducted to unveil the nature of the Ag-doping effect at a molecular level.

2. Results and Discussion

2.1. Preparation and Caharacterization of Co/γ-Al₂O₃ Catalysts with Different Co Particle Sizes

Co/γ-Al₂O₃ catalysts with four different Co loadings of 2, 5, 10, and 15 wt% were first prepared by the incipient wetness impregnation and used in the ammonolysis of ethylene glycol. According to X-ray diffraction patterns (XRD, Figure S1 of the supporting information), the metallic Co nanoparticles in these catalysts were well dispersed when the Co loading was below 10 wt%, because only the 15 wt% Co/γ-Al₂O₃ sample exhibited apparent diffraction signals ascribable to the Co metal (e.g., 2θ = 44.2°, 51.5°, and 75.9°). These high Co dispersions imply a strong interaction between the Co₂O₃ species and the γ-Al₂O₃ support in the oxide precursors of the Co/γ-Al₂O₃ catalysts (i.e., Co₂O₃/γ-Al₂O₃), supported by the partially reducible nature of these γ-Al₂O₃-supported Co₂O₃ species below 1027 K (40.2–87.1% of the total Co amount, determined by the H₂-temperature-programmed reduction (H₂-TPR),

Table 1. Mass loading, reducibility, dispersion, and particle size of Co for the Co/γ-Al₂O₃ catalysts used in this study.

Catalyst	Co Loading a (wt%)	Co Reducibility b (%)	Co Dispersion c (%)	Average Co Particle Size c (nm)
2 wt% Co/γ-Al2O3	1.7	40.2	44.9	2.1
5 wt% Co/γ-Al2O3	4.9	80.1	23.1	4.2 (4.3 ± 0.2 d)
10 wt% Co/γ-Al2O3	9.8	85.3	14.9	6.4
15 wt% Co/γ-Al2O3	14.7	87.1	9.8	9.7

^a determined by the inductively coupled plasma optical emission spectrometry. ^b estimated from the reducible content of Co below 1073 K in the H₂-TPR process of the oxide precursors of the Co/γ-Al₂O₃ catalysts (Figure S2a). ^c determined by the N₂O-chemisorptoion method (Figure S2b). ^d determined by TEM (Figure S3).

and Figure S2) [35,36]. The dispersions of the metallic Co atoms were further measured by the chemisorption of N₂O at 323 K, which is able to selectively oxidize the surface of the Co nanoparticles at mild conditions [37,38]. As shown in Table 1, the measured Co dispersion decreased gradually from 44.9% to 9.8% by increasing the Co loading from 2 wt% to 15 wt%, corresponding to an increase of the Co nanoparticle size from 2.1 to 9.7 nm. The average Co particle size of 5 wt% Co/γ-Al₂O₃ was also determined by transmission electron microscopy (TEM, 4.3 ± 0.2 nm, Figure S3), consistent with that obtained by the N₂O-chemisorption method (4.2 nm, Table 1). These results confirm that the Co particle size of the Co/γ-Al₂O₃ catalysts were successfully tuned by changing the Co loading.

2.2. Effects of Co Particle Size for Co/γ-Al₂O₃ Catalysts in Ammonolysis of Ethylene Glycol

Ammonolysis of ethylene glycol on these Co/γ-Al₂O₃ catalysts were carried out at 453 K in tetrahydrofuran solvent with initial NH₃ and H₂ pressures of 0.6 and 3.0 MPa (0.13 mol/L ethylene glycol; ethylene glycol/NH₃/H₂ molar ratio = 1.0/8.5/42.5), respectively. For a rigorous comparison of the ammonolysis selectivity among these catalysts, the conversion of ethylene glycol for each catalyst was controlled at a similar level (~20%) by adjusting the catalyst amount loaded and the reaction time (2–4 h). On 2 wt% Co/γ-Al₂O₃ (2.1 nm Co), ethanolamine and glycolaldehyde were found as the main products with their carbon selectivity of 49.6% and 27.6%, respectively (Figure 1). According to the hydrogen-borrowing mechanism, glycolaldehyde, formed from dehydrogenation of ethylene glycol on the Co sites, served as an active intermediate for the ammonolysis of ethylene glycol ethanolamine (Scheme 2). It is worth noting that glycolaldehyde may exist in the dimeric form instead of the monomeric form in the reaction solution, because the former state is thermodynamically more stable. Minor products mainly included those formed from secondary amination of ethanolamine (ethylenediamine, piperazine) and dehydration of ethylene glycol (ethylene oxide, ethanol, diethylene glycol).

As the average Co particle size increased from 2.1 nm (2 wt% Co/γ-Al₂O₃) to 4.2 nm (5 wt% Co/γ-Al₂O₃), the carbon selectivity of ethanolamine significantly increased from 49.6% to 73.6%, concomitant with a decrease of the carbon selectivity of glycolaldehyde from 27.6% to 11.8% (Figure 1). This indicates that larger Co nanoparticles favor the C-N coupling between glycolaldehyde and NH₃. A further increase of the average Co particle size from 4.2 to 9.7 nm (i.e., 15 wt% Co/γ-Al₂O₃) led to the selectivity of ethanolamine decreased slightly from 73.6% to 69.3%, mainly because of an increase of the ethanol selectivity from 1.9% to 8.1% (Figure 1). These data show that the selectivity of ethylene glycol ammonolysis was sensitive to the size of the Co nanoparticles and those with an average size of ~4.2 nm appear to be most selective to ethanolamine.

The effect of Co particle size on the ammonolysis activity of the Co/γ-Al₂O₃ catalysts was also apparent. As shown in Figure 2a, the rate of ethylene glycol conversion normalized by the exposed metallic Co atoms (i.e., the turnover frequency, TOF) gradually increased from 5.5 to 14.5 mol (mol_Co-surface·h)⁻¹ when the average Co particle size was changed from 2.1 to 6.4 nm, but then declined to 11.1 mol (mol_Co-surface·h)⁻¹ as the Co particle size further increased to 9.7 nm. Accordingly, the Co nanoparticles of around 6.4 nm (i.e., 10 wt% Co/γ-Al₂O₃) possessed the highest intrinsic activity for ethylene glycol conversion among the examined catalysts. As elucidated in Section 2.5, the rate-determining step of ethylene glycol ammonolysis on Co/γ-Al₂O₃ was the dehydrogenation of ethylene glycol to glycolaldehyde, and the observed effect of the Co particle size reflected, in fact, the corresponding difference of the dehydrogenation ability of the Co catalysts. When the rates of the ethylene glycol reaction were compared based on the total Co amount of the catalyst in order to assess the utilization efficiency of the Co metal, a similar trend of the rate change with the Co particle size was obtained (Figure 2b), while the maximum shifted to 5 wt% Co/γ-Al₂O₃ (4.2 nm Co) as a result of its higher Co dispersion than that for 10 wt% Co/γ-Al₂O₃.

Taken together, the above results clearly show that ammonolysis of ethylene glycol on the Co/γ-Al₂O₃ catalysts was a typical structure-sensitive reaction, as observed widely for ammonolysis of monohydric alcohols [7]. Considering 5 wt% Co/γ-Al₂O₃ (4.2 nm Co) exhibited the highest overall catalytic activity for ethylene glycol conversion and also the highest chemoselectivity of ethanolamine, this catalyst was used next as a reference for investigating the effects of metal-doping on the catalytic performance of the Co/γ-Al₂O₃ catalysts.

2.3. Preparation and Caharacterization of Co/γ-Al₂O₃ Catalysts Doped by a Second Metal

Cu, Ag, and Ru metals were chosen here as a second component to modify the 5 wt% Co/γ-Al₂O₃ catalyst. These bimetallic catalysts were prepared via a co-impregnation method with the molar ratio of Co/M (M = Cu, Ag, Ru) fixed at 98.5/1.5 (denoted as Co_98.5M_1.5/γ-Al₂O₃ henceforth). Cu(NO₃)₂, AgNO₃, and RuCl₃ were chosen here as the dopant precursor, which were converted to the corresponding oxide and metallic states sequentially during the catalyst preparation (details in Section 3.1). As shown in Figure 3a, the introduction of Cu, Ag, or Ru metals into 5 wt% Co/γ-Al₂O₃ did not bring any observable change to the XRD pattern of the catalysts. In other words, no diffraction signals belonging to Co or the second metal appeared, indicating both of the metals dispersed well on the γ-Al₂O₃ support. The energy-dispersive X-ray spectroscopy (EDS) mapping of these bimetallic catalysts confirmed the homogeneous distribution of the two metal elements (Co_98.5Ag_1.5/γ-Al₂O₃ as an example shown in Figure 3b–d). Further TEM characterization revealed that the addition of the second metal slightly decreased the size of the Co nanoparticles (e.g., from 4.3 ± 0.2 to 4.0 ± 0.3 nm in the case of Co_98.5Ag_1.5/γ-Al₂O₃, Figures S3 and S4).

The interaction between Co and the second metal in the Co_98.5M_1.5/γ-Al₂O₃ catalysts was assessed by the H₂-TPR process of their oxide precursors (i.e., Co_98.5M_1.5O_x/γ-Al₂O₃), which were obtained after the thermal treatment (in air, 673 K, 4 h) involved in the catalyst preparation process (details in Section 3.1). The overlapped multiple reduction peaks within the wide temperature range of 473-1073 K for the monometallic Co/γ-Al₂O₃ sample reflect the stepwise nature of the Co₂O₃ specie during their reduction by H₂ (Figure 4a). When metallic Cu was introduced into the Co/γ-Al₂O₃ catalyst, the shape of the H₂-TPR profile was only slightly changed with the reduction peaks shifted to low temperature by ~6 K, and the reduction degrees of the Co₂O₃ specie for Co/γ-Al₂O₃ and Co_98.5Cu_1.5/γ-Al₂O₃ were similar (80% vs. 77%, Figure 4b), indicative of a weak interaction between Co and Cu in Co_98.5Cu_1.5/γ-Al₂O₃. In contrast, the addition of Ag rendered the reduction of the Co₂O₃ specie much easier, exhibiting one broad reduction peak centered at 616 K. The reduction degree of Co_98.5Ag_1.5/γ-Al₂O₃ was also higher than that for Co/γ-Al₂O₃ (86% vs. 80%, Figure 4b). These changes revealed that the presence of Ag promoted the reduction of the Co₂O₃ specie, suggesting a close contact between Ag and Co in Co_98.5Ag_1.5/γ-Al₂O₃. Similar with the case of Ag doping, Co_98.5Ru_1.5/γ-Al₂O₃ showed a main reduction peak centered at 635 K, while it also had a minor reduction peak appearing at 478 K that was likely relevant to the reduction of RuO_x or highly dispersed Co₂O₃ species. The reduction degree of Co_98.5Ru_1.5/γ-Al₂O₃ reached 1.0 (Figure 4b), reflecting a strong promoting effect of Ru doping on the reduction of the Co₂O₃ specie. The above distinct changes of the H₂-TPR process brough forth by the addition of Cu, Ag, and Ru metals suggested the catalytic performance of Co_98.5M_1.5/γ-Al₂O₃ would be influenced by these doped metals to different extents.

2.4. Effects of Doping a Second Metal for Co/γ-Al₂O₃ Catalysts in Ammonolysis of Ethylene Glycoy

As shown in Figure 5, the addition of Cu, Ag, or Ru metal improved the catalytic activity of Co/γ-Al₂O₃ in ammonolysis of ethylene glycol (453 K, 0.067 mol L⁻¹ ethylene glycol, 0.6 MPa NH₃, 3.0 MPa H₂), and the reaction rates based on the total Co amount increased with an order of Cu, Ag, and Ru. However, it is noticeable that only Co_98.5Ag_1.5/γ-Al₂O₃ maintained a high selectivity of ethanolamine (72.7%) similar to that of the monometallic Co/γ-Al₂O₃ catalyst (72.2%) at ~20% conversion of ethylene glycol. The selectivity of ethanolamine decreased significantly in the presence of Cu (16.9%) or Ru (50.5%), in concomitance with an obvious increase of the selectivity of ethanol (Figure S5), which was ascribable to the high activity of Cu and Ru for hydrogenolysis of C-O bonds in polyols [39]. Therefore, compared with Cu and Ru, the doped Ag exhibited a superior ability to promote the selective ammonolysis of ethylene glycol to ethanolamine on Co/γ-Al₂O₃.

In order to optimize the promoting effect of Ag doping, Co-Ag/γ-Al₂O₃ catalysts with the Ag/(Co+Ag) atomic ratio ranged from 0 to 10% were prepared. On one hand, the rate of ethylene glycol conversion based on the total Co amount increased from 1.43 to 2.53 mol (mol_Co-total·h)⁻¹ with increasing the Ag content from 0 to 5%, but then declined to 1.50 mol (mol_Co-total·h)⁻¹ as the Ag content further increased to 10% (453 K, 0.067 mol L⁻¹ ethylene glycol, 0.6 MPa NH₃, 3.0 MPa H₂; Figure 6). On the other hand, the selectivity of ethanolamine on the Co-Ag/γ-Al₂O₃ catalysts was nearly insensitive to the Ag content and kept around 73% at ~20% conversion of ethylene glycol for these bimetallic Co-Ag catalysts. Extra experiments confirmed that 5 wt% Ag/γ-Al₂O₃ was almost inert for catalyzing ammonolysis of ethylene glycol at the same reaction condition (Figure S6). Moreover, X-ray photoelectron spectroscopy (XPS) characterization showed that the Ag element was enriched on the Co-Ag/γ-Al₂O₃ catalysts. For instance, the measured surface Ag/(Co+Ag) atomic ratios for Co_98.5Ag_1.5/γ-Al₂O₃ and Co₉₅Ag₅/γ-Al₂O₃ were 3.2% and 6.4%, respectively (Figure S7), in line with the lower sublimation heat of Ag than that for Co (287 vs. 426 kJ mol⁻¹). These data implied that the doped Ag metal promoted the activity of the Co/γ-Al₂O₃ catalysts via surface modification, but too high Ag concentrations would inhibit the ammonolysis of ethylene glycol because of the dilution of the active Co sites.

2.5. Kinetic Insights into the Ag-Doping Effects on Co/γ-Al₂O₃ Catalysts

In an attempt to unveil the nature of the Ag-doping effects, kinetic assessments were carried out for the Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ samples with their Co loadings fixed at 5 wt%. These control experiments were done by varying the concentrations of ethylene glycol, NH₃, and H₂ individually at 453 K (Figure 7a–c), and the conversions of ethylene glycol were kept at ~20% as described in Section 2.2. The stirring speed of the autoclave reactor was set at 800 rpm, which was confirmed to be high enough to avoid any significant mass diffusion limitation at the examined reaction condition. It was observed that the ammonolysis of ethylene glycol to ethanolamine acted as the predominant reaction pathway in these control experiments, as reflected from the high chemoselectivities of ethanolamine (>70%) in most of the cases (Figures S8–S10). It is also worth noting that the product distribution of ethylene glycol conversion on the Co-based catalysts were not influenced strongly by the concentrations of ethylene glycol, NH₃, and H₂ (Figures S8–S10). Therefore, the following discussion was only focused on the effects of these reactants on the rates of ethylene glycol ammonolysis.

As the concentration of ethylene glycol, NH₃, or H₂ was changed, the rate of ethylene glycol conversion on Co_98.5Ag_1.5/γ-Al₂O₃ varied in a manner similar to that on Co/γ-Al₂O₃ (Figure 7), albeit the bimetallic catalyst showed higher rates. These similar trends of rate change were indicative for the reaction mechanism of ethylene glycol conversions on the Co-based catalysts being unaltered by the doping of Ag. Specifically, the rates for the Co-based catalysts increased gradually as the concentration of ethylene glycol changed from 0.067 to 0.33 mol L⁻¹ (Figure 7a), while the rate dependence on the concentration became weaker at higher concentrations, which is likely ascribable to an increase of the coverage of adsorbed ethylene glycol on the active Co sites. In contrast, the rates of ethylene glycol conversion were inhibited by increasing the partial pressure of NH₃ within the range of 0.3–0.7 MPa (Figure 7b), indicative of strong adsorption of NH₃ on the active Co sites under the reaction condition and thus a negative reaction order with respect to NH₃. Different from the effects of ethylene glycol and NH₃, the rates of ethylene glycol conversion were nearly insensitive to the partial pressure of H₂ within the range of 2.0–4.0 MPa (Figure 7c), reflecting a typical zero-order kinetics as also observed for amination of alcohols on Cu-based catalysts [40].

As illustrated in Section 2.2 (Scheme 2), the ammonolysis of ethylene glycol mainly consists of three steps, including (1) dehydrogenation of ethylene glycol to glycolaldehyde, (2) condensation between glycolaldehyde and NH₃ to imine, (3) hydrogenation of imine to ethanolamine. The measured rate dependance on the pressures of NH₃ and H₂ excluded the kinetic relevance of the second and third steps to the conversion of ethylene glycol, otherwise a positive reaction order with respect to NH₃ or H₂ would be observed. Consequently, dehydrogenation of ethylene glycol to glycolaldehyde appeared to be rate-determining, consistent with previous studies on the ammonolysis of ethylene glycol [41] and other alcohols [7,10,14,42,43,44]. In particular, the dehydrogenation of ethylene glycol on the metal surface involves an initial O-H cleavage to form an alkoxide-type intermediate (Step 2, Scheme 3) and a subsequent abstraction of the α-H atom in this intermediate to form glycolaldehyde (Step 3, Scheme 3). Kinetic analysis was applied next to discern the kinetic relevance of these two steps.

If the O-H cleavage step was rate-determining and irreversible on the Co sites (Step 2 in Scheme 3), taken together with the assumptions for the quasi-equilibrated nature of the adsorption/desorption processes on the catalyst surface and the adsorbed ethylene glycol and NH₃ as the predominant surface species, the rate equation for ethylene glycol conversion (r_EG) on the Co-based catalysts is given by:

$$r_{\mathrm{EG}}=\frac{k_{2+}{K}_1\left[\mathrm{EG}\right]}{{\left(1+K_1\left[\mathrm{EG}\right]+K_6\left[{\mathrm{NH}}_3\right]\right)}^2}$$

(1)

Here, [EG] and [NH₃] are the concentrations of ethylene glycol and NH₃, respectively, and K₁ and K₆ are the corresponding adsorption constants for these two reactants. The k₂₊ term is the forward kinetic constant for the O-H cleavage step, representing the intrinsic ability of the active Co sites to cleave the O-H bond. As shown in Figure 7d, an excellent fitness of Equation (1) with the rates of ethylene glycol ammonolysis was found for both of the Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ catalysts, supporting the hypothesis that the initial activation of O-H bond in ethylene glycol is the rate-determining step.

In constant, if the α-H abstraction of the alkoxide-type intermediate (Step 3 in Scheme 3) was the irreversible rate-determining step instead, r_EG is converted to the form as

$$r_{\mathrm{EG}}=\frac{k_{3+}{K}_1{K}_2{K}_5{}^{-1}\left[\mathrm{EG}\right]{\left[{\mathrm{H}}_2\right]}^{-1/2}}{{\left(1+K_1\left[\mathrm{EG}\right]+K_6\left[{\mathrm{NH}}_3\right]\right)}^2}$$

(2)

in which [H₂] and K₅ are the concentration and adsorption constant for H₂, respectively, k₃₊ is the forward kinetic constant for the α-H abstraction step, and K₂ is the equilibrium constant for the O-H cleavage step (equal to k₂₊/k_2- with k_2- as the backward kinetic constant for this step). Equation (2) indicates that r_EG is inhibited by H₂ because of the −1/2 order of H₂ appeared in the numerator, inconsistent with the measured zero-order rate dependance (Figure 7c). As expected, a poor fitness of Equation (2) with the rate data of ethylene glycol ammonolysis was obtained (Figure S11). Therefore, the kinetic relevance of the α-H abstraction step was ruled out.

Table 2. Regressed kinetic parameters for ammonolysis of ethylene glycol on Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ catalysts (Co 5 wt%) at 453 K.

Catalyst	k2+ (h−1)	K1 (L∙mol−1)	K6 (MPa−1)	k2 +K1 (L∙mol−1∙h−1)
Co/γ-Al2O3	1.4	4.9	8.0	6.9
Co98.5Ag1.5/γ-Al2O3	1.6	3.2	3.3	5.2

compared the regression-fitted parameters of Equation (1) for the Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ catalysts. It was shown that the doping of 1.5% Ag led to a slight increase of k₂₊ from 1.4 to 1.6 h⁻¹, while the values of K₁ and K₆ declined from 4.9 to 3.2 L mol⁻¹ and from 8.0 to 3.3 MPa⁻¹, respectively. As a consequence, the k₂₊K₁ term in the numerator of Equation (1) was slightly higher on the Co/γ-Al₂O₃ catalyst than that for Co_98.5Ag_1.5/γ-Al₂O₃ (6.9 vs. 5.2 L (mol·h)⁻¹), suggesting the presence of Ag, in fact, increased the apparent activation barrier of the O-H cleavage on the Co surface and made the Co catalyst intrinsically less active. On the other hand, the decrease of the values for K₁ and K₆ brought forth by the Ag-doping indicated that the Co active sites on Co_98.5Ag_1.5/γ-Al₂O₃ were less covered by ethylene glycol and NH₃ compared with the case of Co/γ-Al₂O_3. Under the reaction condition typically used in this study (453 K, 0.067 mol L⁻¹ ethylene glycol, 0.6 MPa NH₃, 3.0 MPa H₂), the coverages of ethylene glycol and NH₃, estimated by the K₁ and K₆ terms, were 5% and 78% for Co/γ-Al₂O₃, respectively, while these coverages for Co_98.5Ag_1.5/γ-Al₂O₃ were changed to 7% and 62%, respectively. It can be seen that the adsorbed NH₃ was the predominate surface species during ammonolysis of ethylene glycol on both of the catalysts. More importantly, the addition of Ag into the Co surface significantly weakened the adsorption of NH₃ and made the percentage of the unoccupied active Co sites nearly double (16% vs. 31%), which probably accounted for the promoting effect of Ag-doping on the rate of ethylene glycol ammonolysis.

According to the above kinetic analysis, the Ag dopant mainly acted as an inhibitor for the adsorptions of NH₃ and ethylene glycol on the Co metal surface (based on the changes of K₆ and K₁, respectively, Table 2) and did not significantly affect the intrinsic activity of the Co sites (based on the change of k₂₊, Table 2). This conclusion is consistent with the negligible activity of monometallic Ag/γ-Al₂O₃ catalysts in ammonolysis of ethylene glycol at the reaction condition of this study (Figure S6). A recent theoretical study also showed that the Ag dopant had a weak effect on the intrinsic activity of Co catalysts for ammonolysis of methanol [29]. However, it is worth emphasizing that the above kinetic treatment was based on an ideal assumption that the Co_98.5Ag_1.5/γ-Al₂O₃ catalyst possessed uniform active sites. The Ag dopant may lead to the formation of Co-Ag surface site pairs or even single-atom alloys that are highly active for dehydrogenation of ethylene glycol, considering that Ag-based catalysts have been widely used for alcohol dehydrogenation [45,46]. This latter possibility cannot be excluded because of the lack of atomic-level structural characterization at the current stage of our study.

2.6. Spectroscopic Evidence for the Ag-Doping Effect on the Stability of Adsorbed NH₃ Species

Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was further applied to validate the effect of Ag-doping on the stability of adsorbed NH₃ species. The pre-adsorption of NH₃ on Co/γ-Al₂O₃, Co_98.5Ag_1.5/γ-Al₂O₃ and the γ-Al₂O₃ support was conducted at the real condition of catalytic ammonolysis (453 K, 0.067 mol L⁻¹ ethylene glycol in tetrahydrofuran, 0.6 MPa NH₃, 3.0 MPa H₂), and the resultant samples were treated in vacuum to remove the physically adsorbed NH₃, ethylene glycol, and tetrahydrofuran before acquiring infrared spectra. Using the spectra collected at ambient temperature as the background, three reverse IR bands appeared at around 1651, 1486, and 1432 cm⁻¹ within the range of 1200–1800 cm⁻¹ for both of Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ at 343 K (Figure 8a), which reflected the desorption of chemically bound species from the catalyst surface. In contrast, the γ-Al₂O₃ support merely exhibited the IR band at ~1651 cm⁻¹, assignable to the NH₄⁺ ions formed from the adsorption of NH₃ on the Brønsted acid sites of γ-Al₂O₃ [47,48]. Therefore, the two bands at 1486 and 1432 cm^–1 appeared to be derived from those adsorbed on the Co surface. Specifically, the band at 1486 cm^–1 was attributable to the C-C stretching/C-H bending vibration (ν_C-C/δ_C-H) of ethylene glycol-derived species [49], while the band at 1432 cm^–1 was attributable to the deformation vibration of NH₃ coordinated to the metal site. As a consequence, the intensity of the latter band was used as a descriptor of the amount of NH₃ adsorbed on the Co surface.

As shown in Figure 8b, the desorption of NH₃ gradually occurred on both of Co/γ-Al₂O₃ and Co_98.5Ag_1.5/γ-Al₂O₃ as the temperature increased from 293 to 393 K (detailed spectra shown in Figure S12). It is noticeable that the bimetallic Co_98.5Ag_1.5/γ-Al₂O₃ catalyst completed the desorption at a temperature apparently lower than that for Co/γ-Al₂O₃ (348 vs. 373 K). This result indicated that the adsorption of NH₃ on Co_98.5Ag_1.5/γ-Al₂O₃ was weaker than on Co/γ-Al₂O₃, consistent with the above kinetic assessment. This spectroscopic evidence clearly unveiled the unique ability of the Ag dopant in destabilizing the NH₃ species bound to the Co surface, although the effect of Ag-doping on the electronic state of the Co sites was not observable from the deformation vibration bond of the adsorbed NH₃. More efforts are still required to understand the nature of the Ag-doping effects for ammonolysis of ethylene glycol on Co-based catalysts.

3. Experimental

3.1. Preparation of Supported Co Catalysts

Co/γ-Al₂O₃ catalysts (Co loading: 2, 5, 10, and 15 wt%) were prepared by the conventional incipient wetness impregnation method [34]. Using 5 wt% Co/γ-Al₂O₃ as an example, 0.617 g Co(NO₃)₂·6H₂O (AR, Sinopharm Chemical, Shanghai, China) was dissolved in 1.5 mL deionized water. This aqueous solution was then added dropwise to the γ-Al₂O₃ support (Sigma-Aldrich, Waltham, MA, USA, 154 m².g⁻¹) of 2.38 g, and the resultant powder was dried in an oven at 393 K overnight. Subsequently, the dried powder was treated in stagnant air at 673 K for 4 h (2 K min⁻¹) to decompose the Co(NO₃)₂ precursor to form Co₂O₃/γ-Al₂O₃, followed by a reduction process in flowing H₂ (40 mL min⁻¹, ≥99.999%, Linde Gases, Shanghai, China) at 773 K for 1 h (2 K min⁻¹) to obtain the Co/γ-Al₂O₃ catalysts. Bimetallic Co-M/γ-Al₂O₃ catalysts (M = Cu, Ag, Ru) with the Co loading fixed at 5 wt% were prepared through co-impregnation, in which Cu(NO₃)₂·3H₂O (AR, Sinopharm Chemical), AgNO₃ (AR, Sinopharm Chemical), and RuCl₃·3H₂O (98%, Energy Chemical, Shanghai, China) were chosen as the precursors of the second metal. The thermal treatments for the resultant bimetallic samples were identical to that in the preparation of the Co/γ-Al₂O₃ catalysts.

3.2. Catalyst Characterization

Powder X-ray diffraction (XRD) patterns of the Co-based catalysts were recorded in the 2θ range of 10–90° (scan speed 0.167°·s^–1) on a Panalytical X’Pert PRO diffractometer (Cu K_α radiation, λ = 0.15406 nm) at 40 kV and 30 mA. Transmission electron microscopy (TEM) images were collected on a Tecnai G2 F20 transmission electron microscope (FEI) operating at 200 kV. Before measurements, the samples were ultrasonically dispersed in ethanol solvent and then loaded onto carbon-coated Cu grids. The average size of metal particles was calculated on the basis of ca. 200 particles in different regions of the TEM images. Ex-situ X-ray photoelectron spectroscopy (XPS) spectra of the catalysts were recorded at 12 kV using a Thermo Scientific K-Alpha spectrometer equipped with a monochromatic Al K_α radiation source. The binding energy (BE) values were referred to the C 1s peak at 284.8 eV. The metal loading of the supported catalysts was determined by the inductively coupled plasma optical emission spectrometry (ICP, Thermo Fisher Scientific, ICAP7400).

The reducibility of the oxide form of the Co-based catalysts was assessed via H₂-temperature-programmed reduction (H₂-TPR) tests on a fully-automated chemisorption analyzer (DAS-7200, HUASI). Each sample (100 mg) was first treated in flowing N₂ (40 mL min⁻¹, >99.999%, Linde Gases) at 573 K for 1 h to remove physical adsorbed H₂O. After the temperature decreased to 313 K, the gas was switched to 5% H₂/Ar mixture (40 mL min⁻¹, Linde Gases), and the sample was heated to 1173 K at a ramp rate of 10 K min⁻¹. The H₂-TPR profiles were recorded and quantified using a thermal conductivity detector.

The dispersion of metallic Co for the Co-based catalysts was measured by the N₂O-chemisorption method [38,50] on the same chemisorption analyzer. Each pre-reduced catalyst (100 mg) was first exposed to a 5% N₂O/N₂ flow (60 mL min⁻¹, Linde Gases) at 323 K for 1 h to selectively oxidize the surface of the Co nanoparticles. The amount of chemically-adsorbed O atoms on the Co nanoparticles was determined from a subsequent H₂-TPR process using the protocol described above. The dispersion of Co was then calculated based on the hypothesis that metallic Co can be oxidized to Co³⁺ by N₂O under the chemisorption condition, and the average size of the Co nanoparticles (<d>, unit in nm) was estimated according to the following equation [51,52],

$$\mathrm{Dispersion}=\frac{96}{\langle d\rangle }\times 100\%$$

(3)

which assumes a spherical geometry of the Co nanoparticles.

Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed for monitoring the desorption of NH₃ on Co/γ-Al₂O₃, Co_98.5Ag_1.5/γ-Al₂O₃, and γ-Al₂O₃ as a function of temperature, which were carried out using a Nicolet iS50 FT-IR spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a MCT/A detector (4 cm⁻¹ resolution, 64 scans for each spectrum). The pre-adsorption of NH₃ for each sample was conducted in a stainless-steel autoclave under the real condition of ethylene glycol ammonolysis (453 K, 0.067 mol L⁻¹ ethylene glycol in tetrahydrofuran, 0.6 MPa NH₃, 3.0 MPa H₂) for 0.5 h. The resultant powder was separated without further washing by any solvent and then dried in a vacuum oven at 303 K for 12 h. Background spectra were collected at ambient temperature after the IR cell was purged by a He flow. The cell was heat stepwise in flowing H₂ from 293 to 393 K (10 K per step, 5 K min⁻¹ of the ramp rate), and the IR spectra of the samples were collected at each step after the temperature kept stable for 0.5 h.

3.3. Catalytic Ammonolysis of Ethylene Glycol

Ammonolysis of ethylene glycol was carried out in a 50 mL stainless-steel autoclave (HAC-1040, HUASI, Changsha, China). In a typical test, 0.124 g ethylene glycol (AR, Sinopharm Chemical) and 15 mL tetrahydrofuran solvent (AR, Sinopharm Chemical), together with 0.100 g Co-based catalyst (ground powder, <0.075 mm) pre-reduced at 773 K by H₂ as described in Section 3.1, were added into the autoclave. The reactor was then sealed and purged three times with pure NH₃ gas. After 0.6 MPa NH₃ and 3.0 MPa H₂ were added sequentially into the autoclave, the ammonolysis of ethylene glycol was conducted at 453 K under stirring (800 rpm) for 2–4 h. It was confirmed that the possibility of mass transport limitations was excluded at the conditions of our kinetic measurements based on the tests of the stirring speed and the catalyst particle size (Figures S13 and S14). After the reaction, the liquid phase and the catalyst were separated by centrifugation. The concentrations of ethylene glycol and reaction products were quantitively analyzed off-line by gas chromatography (GC7820, Shandong Huifen) with an OV-1701 capillary column connected to a flame ionization detector, and isopropanolamine (AR, Sinopharm Chemical) was selected as the internal standard. These products were identified using known standards and speciation by mass spectrometry after chromatographic separations (Agilent 7890B-5977BGC/MSD). The carbon balance was accurate to within 6% for each catalytic test. The conversion of ethylene glycol (C_EG) and the selectivity of product i (S_i) were calculated as reported elsewhere [34]:

$$C_{\mathrm{EG}} (\%)=\frac{n_{EG}^0-n_{EG}}{n_{EG}^0}$$

(4)

$$S_i (\%)=\frac{{\alpha }_i\times n_i}{2\times \left(n_{EG}^0-n_{EG}\right)}$$

(5)

Here, $n_{EG}^0$ and $n_{EG}$ refer to the initial and final molar amounts of ethylene glycerol in the reactor, respectively, $n_i$ is the molar amount of product i formed after the reaction, and ${\alpha }_i$ is the carbon number of product i. Rates of ethylene glycol conversion were calculated based on both of the surface Co atoms (i.e., turnover frequency) and the total Co amount. The regressed fitting of these rates with the proposed kinetic equations was conducted using the Bayesian estimation method, as implemented in the Athena Visual Studio software package.

4. Conclusions

For the liquid-phase ammonolysis of ethylene glycol on Co/γ-Al₂O₃ catalysts, a moderate size of the Co nanoparticles (~4 nm) is a requisite for obtaining both of a high amination rate and a high chemoselectivity to ethanolamine. Smaller Co nanoparticles (~2 nm) are not only intrinsically less active in the ammonolysis of ethylene glycol but also exhibit a higher selectivity to glycolaldehyde due to the lack of enough ability in catalyzing the condensation between glycolaldehyde and NH₃. The activity of the Co/γ-Al₂O₃ catalysts can be further improved by adding a moderate amount of Cu, Ag, or Ru, while the high chemoselectivity to ethanolamine is only maintained in the case of Ag-doping. Combined kinetic and infrared spectroscopic assessments reveal that the dehydrogenation of ethylene glycol to glycolaldehyde is the rate-limiting step for ethylene glycol ammonolysis on the Co-based catalysts, the surface of which is highly covered by NH₃ species at the reaction condition. The introduction of Ag dopant weakens the adsorption of NH₃ on the Co surface and renders more Co sites available for catalyzing ethylene glycol dehydrogenation, which accounts for the promoting effect of Ag-doping on the ammonolysis activity of the Co/γ-Al₂O₃ catalysts.