Ammoxidation of Ethane to Acetonitrile and Ethylene: Reaction Transient Analysis for the Co/HZSM-5 Catalyst.

Ethane ammoxidation to acetonitrile and ethylene over the Co/HZSM-5 catalysts was revisited based on both transient and steady-state performance evaluation to elucidate the structure/reactivity relationships. We suggested that the exchanged Co2+ cation encapsulated in the zeolite favors the formation of acetonitrile and ethylene, whereas nanosized cobalt oxide particles without close proximity with the HZSM-5 only favor CO2 formation. Excess Brønsted acid sites of the zeolites may act as a reservoir for NH3, which inhibits the CO2 formation through the NH3-mediated oxidative dehydrogenation mechanism. According to the transient kinetic analysis, the time constants τ from the back-transient decay for NH3 and CO2 are both 7.7 min, which decreased to 2.7 min for acetonitrile and further decreased to 3-4 s for ethane, ethylene, and O2. Assuming first-order reaction kinetics, the rate constants for the formation of acetonitrile and CO2 are 0.37 and 0.13 min-1, respectively, from their corresponding reactive intermediates.

Introduction

The current shale gas revolution has revitalized research interests on the catalytic conversion of light C1–C3 alkanes. While significant attention has been devoted to methane and propane conversion, a catalytic process for ethane conversion has been relatively less concerned probably due to the fact that it is difficult to compete with the well-established steam cracking process (although it is high carbon and energy-intensive). The catalytic processes for ethane conversion include oxidative/nonoxidative dehydrogenation (ODH),[1−10] aromatization,[11−14] ammoxidation, partial oxidation, reforming, and so forth.[15−21] The advantages and disadvantages of these catalytic processes are summarized in Table . In the present paper, we revisited the Co/HZSM-5 catalyzed ethane ammoxidation because such a catalytic system produces acetonitrile and ethylene simultaneously with inhibited CO2 formation (compared to the ODH) because of the presence of NH3.[15]

Table 1

Advantages and Disadvantages of Various Catalytic Processes for Ethane Conversion

processes	products	catalysts	advantages	disadvantages
aromatization[11−14]	ethylene + aromatics	metal-modified HZSM-5	high conversion, high total selectivity	low stability, high reaction temperature,
ODH[1,2,7]	ethylene	mixed metal oxides	lower reaction temperature, higher stability	low selectivity at high ethane conversion
dehydrogenation[3−6,8−10]	ethylene	Pt-based	high selectivity	low stability, high reaction temperature, conversion is equilibrium-limited
ammoxidation[15]	ethylene + acetonitrile	metal-modified zeolites or mixed metal oxide	lower reaction temperature, higher stability, lower CO2 formation than ODH	low selectivity at a high ethane conversion; NH3 can also be oxidized to NOx.
partial oxidation[16−18]	oxygenates	metal-modified HZSM-5 or metal oxides	lower reaction temperature, higher stability	low conversion, high operation pressure, or use H2O2
reforming[19−21]	H2 + CO	transition metal-based	industrial application is feasible	products need to be converted through FT or methanol synthesis and so forth.

The ammoxidation reaction was initially invented in 1959 by Standard Oil of Ohio (the SOHIO Process) for the production of acrylonitrile from propylene, NH3, and O2.[22] The ethane ammoxidation was then investigated about 20 years later by Aliev and Sokolovskii with a Cr–Sc–Mo–O catalyst.[23] However, the selectivity to acetonitrile (<30%) was limited because of the formation of CO2 and HCN over such a mixed metal oxide catalyst. It is quite surprising that the formation of ethylene was not mentioned in the early study.[23] Other mixed metal oxide catalysts investigated for ethane ammoxidation include Nb–Sb–O/Al2O3,[24] V–Mo–Nb–O,[25] Ni–Nb–O,[26,27] and Mo–V–Te–Nb–O.[28] However, the selectivity to acetonitrile over these mixed oxide catalysts was low (<30%) and the typical propylene and propane ammoxidation catalyst V–Sb–Al–O seems to be not effective in the ethane ammoxidation.[24]

The most effective catalyst for ethane ammoxidation was the cobalt-modified zeolites (prepared through an aqueous solution or solid-state ion-exchange) initially developed by Li and Armor.[29−35] The rate of acetonitrile formation over the Co-β catalyst at 475 °C was 1–2 orders of magnitude higher than the mixed metal oxide catalysts at 500 °C.[29] Li and co-workers then investigated systematically the effect of zeolite topology, types of metal cations, and the effect of reaction conditions (temperature and reactants partial pressure) on the catalytic activity and selectivity, as well as the reaction pathways and mechanism.[30−32] Among various zeolite topologies investigated, the Co-modified MFI (ZSM-5) and BEA (β) catalysts seem to be more effective: ethane conversion up to 35% and acetonitrile selectivity around 50% was obtained at 450 °C. Although Co–Y has the highest acetonitrile selectivity (60%), the conversion of ethane was only 8.4%, which increased significantly after dealumination treatment.[33] Besides the zeolites, Co/silica-alumina and CoO/Al2O3 were also investigated for ethane ammoxidation, which showed significantly lower activity and selectivity than the Co-ZSM-5 and Co-β catalysts, indicating the importance of the zeolites framework structure. Later studies on the effect of acidity further indicated that the negative charge of the zeolite framework and the accessibility of the reactants to the exchanged cobalt cations to be the decisive factors for controlling the activity.[36] With respect to the effect of metal cations, Co2+ has been found to be the most effective one.[32] The catalytic activity for other metal cations, such as Ni2+, Fe3+, and Mn2+ was significantly lower than that of Co2+, whereas the Cu-, Pd-, Ag-, Rh-, and Pt-modified ZSM-5 catalysts were found to be totally inactive in acetonitrile formation (produce mainly ethylene and CO2). In terms of the catalytically active sites, Li and Armor suggested that acetonitrile, ethylene, and CO2 can all be produced over the exchanged Co2+ cation; however, the formation of acetonitrile and ethylene was favored.[31]

Although Li and co-workers have carried out extensive studies on the Co/zeolites catalysts, such a catalytic system has been frequently revisited recently for both ethane and ethylene ammoxidation by others for the purpose of further improving the catalytic performance or understanding the catalytic mechanism.[15,37−43] For example, Essid et al.[15] recently reported that the activity and selectivity of ethane ammoxidation both improved if [Co(NH3)6]2+ was impregnated into the BEA zeolite instead of [Co(OH2)6]2+. The improved catalytic performance was attributed to the formation of Co4N with the presence of NH3 during the activation or impregnation.[15]

In this work, a series of Co modified HZSM-5 catalysts were prepared for ethane ammoxidation through incipient wetness impregnation, ion-exchange, and physical mixing. The effect of Co loading, SiO2/Al2O3 ratio, and various reaction conditions on the catalytic performance has been investigated. Specifically, the kinetics and mechanism of such catalytic systems have been discussed based on the early-stage reaction transient analysis as well as the back-transient kinetics. The physicochemical properties of the catalysts were extensively characterized by N2 physisorption, transmission electron microscopy (TEM), scanning TEM–energy-dispersive X-ray spectroscopy (STEM–EDX), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), ultraviolet–visible diffuse reflectance spectroscopy (UV–vis-DRS), and NH3- and n-propylamine temperature-programmed desorption (TPD). The effect of catalyst preparation, Co loading, and the Si/Al2 ratio of the zeolite host on the physicochemical properties and their relationships to the catalytic performance was discussed. We anticipate that a balance between the function of Co species and the Brønsted acid density of the HZSM-5 is crucial in the ethane ammoxidation.

Results and Discussion

Catalytic Results

Effect of the Catalyst Preparation

We first did the blank test of the reactor system, which only produces a negligible amount of CO2 (∼0.2%). Other products were not observed during the blank test. While pure HZSM-5 seems to be active in ethane ammoxidation, the rates of acetonitrile, ethylene, and CO2 were negligible in comparison to the Co-modified HZSM-5 catalyst (see Figure S1). The catalytic performance of ethane ammoxidation over the cobalt-modified HZSM-5 (Si/Al2 = 30) catalysts prepared through ion-exchange, impregnation, and physical mixing is shown in Figure a,b. As we can see, all of the three catalysts are quite active in ethane ammoxidation for acetonitrile and ethylene. The cobalt-exchanged HZSM-5 (Co-IE) prepared according to Li and Armor[31] shows slightly higher activity (in terms of ethane conversion) and relatively lower selectivity to acetonitrile than the catalysts prepared through impregnation (Co-IM-2 wt %). However, the catalyst prepared through physical mixing (Co-PM-2 wt %) shows relatively higher CO2 selectivity and lower activity than the Co-IE and Co-PM-2 wt %. The rates for the formation of acetonitrile, ethylene, and CO2 are shown in Figure b. The Co-PM-2 wt % catalyst clearly shows a lower rate in terms of acetonitrile and ethylene. We anticipate that over the Co-PM-2 wt %, the catalytically active sites (exchanged Co2+ cation) were formed during the activation and ongoing reaction process through the solid-state ion-exchange (see STEM–EDX characterization for evidence). Hence, all three catalysts finally show similar selectivity after reaching the steady state. However, the Co-PM-2 wt % catalyst takes a longer induction period before reaching the steady state (see Figure S2), which also provides evidence for the occurrence of solid-state ion-exchange during the reaction process. It must be noted that for all of the catalysts, the NH3 conversion is typically 50%, which is significantly higher than the ethane conversion (10–20%). Such different reaction rates regarding NH3 and ethane conversion suggested that large amounts of NH3 were nonselectively oxidized into NO instead of forming acetonitrile [The mass spectrometer signal of NO (m/z = 30) was clearly identified, but quantification of NO is quite difficult]. The acetonitrile selectivity based on NH3 is 25–30%, which is similar to the results reported by Li and Armor.[33,34]

Figure 1

Influence of catalyst preparation (a,b), Si/Al2 ratio (c,d), and Co loading (e,f) on the catalytic performance of ethane ammoxidation. (a,c,e): Selectivity (ethylene, acetonitrile, and CO2) and conversion (C2H6, NH3, and O2); (b,d,f): formation rate of ethylene, acetonitrile, and CO2. Reaction conditions: 10% C2H6, 10% NH3, and 6.4% O2, balanced with Ar, total flow rate 80 mL/min; T = 475 °C, W = 0.2 g.

Effect of the Si/Al2 Ratio

Using the impregnation method for catalysts preparation, we then investigated the effect of the Si/Al2 ratio and Co loading on the catalytic performance. As shown in Figure c, both ethane conversion and acetonitrile selectivity decrease with the increasing Si/Al2 ratio. Accordingly, the selectivity to CO2 increases significantly and ethylene selectivity increases slightly. Specifically, the ethane conversion was decreased from 18% to almost zero with increasing Si/Al2 ratio from 30 to 80. However, the conversion of NH3 and O2 decreased with increasing Si/Al2 from 30 to 50 and then increased with further increasing Si/Al2 from 50 to 80. In terms of the rates for the formation of acetonitrile, ethylene, and CO2 (see Figure d), they all decreased almost linearly with the increasing Si/Al2 ratio from 30 to 80 or decreasing the Brønsted acidities of the zeolite host. Therefore, the acid properties of the zeolite host are important for such a catalytic system. According to our recent study,[44] the Brønsted acidities of the HZSM-5 with Si/Al2 ratio of 30, 50, and 80 are 531, 387, and 224 μmol/g, respectively. Because the exchanged Co2+ cation has been considered as the catalytic active sites for ethane ammoxidation[31] and for the zeolite host with a higher Brønsted acid density, the concentration of the Co2+ active sites should also be higher. Without sufficient Brønsted acid sites (for high Si/Al2 ratio zeolite), excessive cobalt species will most likely form cobalt oxide particles without close proximity with the zeolite (see STEM–EDX characterization), which will result in the formation of CO2 and NO. Consequently, NH3 and O2 conversion further increased with the increasing Si/Al2 ratio from 50 to 80.

Effect of Co Loading

Because it has been generally accepted that the exchanged Co2+ cation is responsible for the ethane ammoxidation,[29−35] the effect of Co loading on the HZSM-5 (Si/Al2 = 30) host was investigated. The Co loading was varied from 0.5 to 10 wt %, which represents the excess amount of either H+ or Co2+ [assuming bridge-type-exchanged Co2+ cations (...Al–O–Co–O–Al...)]. As shown in Figure e, the acetonitrile selectivity decreases, CO2 selectivity increases, and ethylene selectivity remain almost unchanged with increasing Co loading from 0.5 to 10 wt %. The conversion of all three reactants increased almost linearly with increasing Co loading from 0.5 to 2 wt %. With further increase of the Co loading from 2 to 10 wt %, the conversion of ethane and NH3 remain unchanged, but the conversion of O2 further increased. From the involved reaction stoichiometry, such further increased O2 conversion is likely related to the increased CO2 selectivity because the conversion of ethane and NH3 conversion remain unchanged. In terms of the formation rate of different products, as shown in Figure f, the maximum rate for acetonitrile was obtained at 2 wt % Co loading. When the Co loading was further increased from 2 to 10 wt %, the rate of acetonitrile decreased; meanwhile, the rate of CO2 further increased and the rate of ethylene remains almost unchanged. As already mentioned, excessive cobalt species will form cobalt oxide particles without close proximity with the zeolite (see STEM–EDX characterization), which is undesired for ethane ammoxidation. Therefore, we suggested that 2 wt % be the optimal Co loading for the ethane ammoxidation. Although for the 2 wt % loading sample Co is deficient compared to the stoichiometric Co/Al ratio, the excessive Brønsted acid sites could act as a reservoir for NH3. Actually, if we normalize the rate by the Co loading instead of the mass of the catalyst, the rates for all three products decreased continuously with increased Co loading (see Figure S3), which indicated that the excessive Brønsted acid sites could have a beneficial effect to the ammoxidation.

Catalyst Characterization

The characterization results of the selected Co-modified HZSM-5 catalysts were measured by means of N2-physisorption, XRD, XPS, STEM–EDX, NH3-TPD, and n-propylamine-TPDec. As shown in Table , the physical properties of the catalysts are clearly influenced by the method of catalyst preparation and the Co loading. The Co-IE and Co-PM samples show lower total and microsurface area than the Co-IM samples because of the partially blocked micropores. Additionally, with increasing Co loading for the Co-IM samples, the total and microsurface area and pore volume are also clearly decreased. The XRD patterns (see Figure a) of the Co-modified HZSM-5 catalysts with different Co loadings show almost identical diffraction patterns for the MFI type of zeolites. The diffraction of Co species was completely absent for Co loading ≤2 wt %, indicating very high dispersion of the Co species in the HZSM-5 zeolite. The Co species underwent significant solid-state ion-exchange during the high-temperature pretreatment to form the exchanged Co2+ cation.[34] When the Co loading ≥5 wt %, the diffraction of Co3O4 (311) was identified at 2θ of 36.8° (JCPDS no. 42-1467), indicating that nanosize Co3O4 particles are formed when the Co loading is higher than the nominal amount (assuming Co/Al = 0.5, the Co loading for the ion-exchanged sample is 3.1 wt %). Such oxidic Co particles usually do not have close proximity with the zeolites (see Figure i) and could be responsible for the nonselective ODH and NH3 oxidation to NO on the high Co-loading catalyst.

Table 2

Physical Properties for Co-IM, Co-IE, and Co-PM Samples

			surface area (m2 g–1)		pore volume (cm3 g–1)
sample	pore size (nm)	external surface area (m2 g–1)	total	micro	total	micro
Co-IE	4.00	70.1	239.7	169.7	0.24	0.08
Co-PM	4.00	72.0	241.1	169.1	0.24	0.08
Co-IM-0.5 wt %	2.61	91.1	337.1	246.1	0.22	0.11
Co-IM-1 wt %	2.63	86.0	321.4	235.4	0.21	0.11
Co-IM-2 wt %	2.62	85.6	324.6	239.0	0.22	0.11
Co-IM-5 wt %	2.69	73.2	295.0	221.8	0.20	0.10
Co-IM-10 wt %	2.63	67.3	275.2	208.0	0.18	0.10

Figure 2

Characterizations of the Co-modified HZSM-5 catalyst. (a) XRD patterns; (b) Co 2p XPS spectra; (c) NH3-TPD profiles; and (d) n-propylamine-TPD, rate of propylene desorption.

Figure 3

UV–vis-DRS spectra of Co-IM-0.5 wt %, Co-IM-1 wt %, Co-IM-2 wt %, and Co-IE samples. Peaks at 12 000–24 000 and 31 000 cm–1 are assigned to Co(II) ions and CoO species, respectively.

The bulk XRD characterization was further complemented by the results from XPS surface analyses. The Co 2p spectra of the samples are displayed in Figure b, while the calculated Co and Al ratios from the spectra are shown in Table . There are no apparent peaks present at the Co 2p spectra for both Co-IE and Co-IM-0.5 wt %, indicating that the concentration of Co species on the surface of the zeolite is too low to be detected by the XPS. The peaks start to emerge for other samples and increases with increasing Co loading. The Co 2p spectrum of Co-IM-2 wt % contains a peak at ∼782 eV which is usually ascribed to a divalent Co2+. The peak shifted to ∼780 eV for the Co-IM-5 wt %, indicating the presence of Co3+,[45] which formed the spinel-type Co3O4 particles as identified by the XRD. Compared to the fresh catalyst, the Co 2p intensity of the Co-IM-2 wt % after ethane ammoxidation further attenuates. During the reaction process, the migration of the Co2+ into the inner pores of the zeolites is likely happening which prevents the detection of the Co species by the XPS measurement.

Table 3

Effect of Co Loading on the Chemical Properties of the Catalysts

	Co/Al bulk atomic ratioa	Co (at %) surfaceb	Al (at %) surfaceb	total acid (μmol/g)	Brønsted acid (μmol/g)
HZSM-5				1225.0	531.0
Co-IE	0.50	0.3	1.5
Co-IM-0.5 wt %	0.08	0.5	1.6	741.1	381.7
Co-IM-1 wt %	0.16			555.1	369.8
Co-IM-2 wt %	0.32	0.4	1.4	542.7	321.3
Co-IM-5 wt %	0.80	1.1	1.6	555.9	262.4
Co-IM-10 wt %	1.61			381.1	221.6

Nominal Co/Al ratio based on Co loading.

Surface concentration measured by XPS.

While the Co species was largely absent from the XPS spectra for the samples with low Co loading, the presence of Co was clearly observed from both UV–vis-DRS spectra and the STEM–EDX mapping. As shown in Figure , the UV–vis-DRS spectra clearly show Co(II) ions at wavenumbers of 12 000–24 000 cm–1 for the Co-IE and Co-IM (with Co loading ≤ 2 wt %) samples. Additionally, the peak at 31 000 cm–1, which can be assigned to the CoO species, was almost absent from the UV–vis-DRS spectra, indicating that most of the cobalt species are coordinated with framework aluminum located in the pores of zeolites.[51] The TEM image and STEM–EDX chemical mappings of the Co-IM-2 wt % (on HZSM-5 with Si/Al2 = 30) catalyst are shown in Figure a–c. A comparison of the Co Kα and Al Kα mappings of the catalyst both before and after reaction demonstrated the homogeneous distribution of the cobalt species on the zeolite. Aggregated CoO nanoparticles were not observed from the TEM, which again confirmed the formation of exchanged Co2+ cations. Different from the catalyst prepared through impregnation, the Co-PM-2 wt % catalyst after the reaction showed nanosized particles supported on the HZSM-5 (Figure d). Additionally, the occurrence of solid-state ion-exchange during the activation and reaction process was proved by the STEM–EDX characterization. As shown in Figure f, homogeneously distributed Co species were clearly identified in the selected region (see Figure e), where nanosized particles are absent. While Co and zeolite in close proximity was found for both Co-IM and Co-PM with 2 wt % loading on the HZSM-5 with Si/Al = 30, aggregated Co oxide particles without such proximity was observed when either the Co loading or the Si/Al2 (of the zeolite host) is too high because of the off-ratio of the Co/H+. As shown in Figure g,h, aggregated Co species without close proximity with the zeolite were clearly identified for the catalyst with Si/Al2 = 280 (Co loading 2 wt %). The same aggregated Co species were found for the Co-IM-10 wt % (Si/Al2 = 30) catalyst (see Figure i).

Figure 4

TEM and STEM–EDX of different catalysts. (a,b) Fresh catalyst after activation, (c–i) catalyst after reaction at 475 °C. (a) TEM image of Co-IM-2 wt % on Si/Al2 = 30, (b,c) STEM–EDX chemical mapping of Co-IM-2 wt % on Si/Al2 = 30 before and after the reaction, (d,e) TEM images of Co-PM 2 wt %, (f) STEM–EDX chemical mapping of image (e), (g,h) TEM image of Co-IM-2 wt % on Si/Al2 = 280 and the corresponding STEM–EDX mapping, (i) TEM image of Co-IM-10 wt %.

The effect of Co loading on the acidity of the catalysts was characterized by NH3- and n-propylamine-TPD. As shown in Figure c, two distinct NH3 desorption peaks at 280 and 490 °C are assigned to NH3 desorbed from weak acid sites and strong acid sites, respectively. The total acid sites quantified by NH3-TPD are varied from 381 to 741 μmol/g for different catalysts. With increasing Co loading, the intensity of the high-temperature peak decreased, which suggested that the Brønsted acid sites (H+) of the host HZSM-5 were largely replaced by the exchanged Co2+ cation. Evidence for the decreased Brønsted acid density can also be found from the n-propylamine-TPD results (Figure -d), which provide quantitative results of the Brønsted acid concentration. In such methods, “C3H7NH3+···ZSM-5–” was formed between propylamine and the Brønsted acid site of the zeolite, which was decomposed to NH3 and propylene according to the Hofman-elimination type of mechanism at ∼400 °C.[46−49] In our previous paper,[44] the desorption of NH3 was employed to quantify the Brønsted acid sites because of the secondary reaction of the propylene. However, over the present Co-modified HZSM-5 catalyst (not a typical catalyst for aromatization), the formation of secondary products from propylene was insignificant. Therefore, propylene desorption during the propylamine-TPD was employed for quantification as suggested by the literature.[46] The Brønsted acid density of the catalysts was quantified to be 381.7 μmol/gcat for Co-IM-0.5 wt % and decreased to 221.6 μmol/gcat for the Co-IM-10 wt % catalyst. The Brønsted acid sites only slightly decreased when the Co/Al ratio exceeds significantly the stoichiometric ratio for ion-exchange, indicating that part of the Brønsted acid sites (most probably the sinusoidal channels) is inaccessible for solid-state ion-exchange.

Transient Kinetics Analysis and Mechanism

Relaxation-type transient kinetic analysis provides important information about how the steady-state reaction was reached upon introducing perturbation, which was realized by fast partial pressure step changes for the reactor influent gases in the present study. During the early-stage transient, the reactor flux was changed abruptly from Ar to the reactant mixture. Information about how a clean surface catalyst reaches the steady-state ammoxidation was obtained. The results over the optimal Co-IM-2 wt % catalyst are shown in Figure a. The normalized (with respect to the steady-state outlet molar flow rate) outlet flow of reactants and products take about 2 min to reach 1, which means the ethane ammoxidation over such catalysts takes 2 min to reach the steady state. The reactants ethane and O2 appear almost immediately in the outflow gas phase and reach the maximum at 0.3 min after switching from Ar to the reactant mixture. Ammonia, however, appears in the gas phase with clear delay (0.25 min) and it takes about 2 min to reach the maximum. Significant delay of NH3 with respect to ethane and O2, indicating the strong chemisorption of NH3 on the (Lewis and Brønsted) acid sites of the catalyst. Additionally, ethane and O2 reach a maximum that exceeds their steady-state level at 0.2–0.5 min, which is most probably because of the increase of partial pressure caused by the consumption of NH3 through chemisorption. The O2 then decreased to below steady-state when acetonitrile was largely produced. Such early-stage features of delay were totally absent during the blank test (see Figure S4a,b), in which the influence of the reactor system on the delay can be excluded. Additionally, delay of NH3 over the HZSM-5 sample (without Co function) is similar to the Co/HZSM-5 catalysts, indicating that the initial delay of NH3 (∼0.25 min) mainly originated from the chemisorption on the acid sites of the zeolite.

Figure 5

Transient kinetics analysis of ethane ammoxidation over Co-IM-2 wt % at 475 °C. (a) Normalized outlet molar flows of reactants and products during the early-stage run-in period; (b) extended back-transient showing the decay of the reactants and products; (c) near steady-state back-transient where the single Ar indicates the reactor response; (d) extended back-transient of acetonitrile and CO2 and a linear relation between ln(nFi) and t indicating the first-order kinetics ln(Fi) = ln(Fi0) – kt. The inset numbers are the rate constant k.

In terms of product formation, ethylene and CO2 were produced simultaneously when ethane appeared in the gas phase. Both reached a maximum at 5 s and decreased immediately to a minimum after 20–30 s. They both take about 1 min to reach the steady-state level. The peaks of ethylene and CO2 showed up before the full appearance of NH3 in the gas phase, which suggested that the clean catalyst surface without NH3 is active for ethane ODH to ethylene (mainly produces CO2 rather than ethylene). Such a nonselective ODH mechanism was replaced by NH3-mediated ODH (ammoxidation) when the catalytical active sites were occupied by NH3. Different from the production of ethylene and CO2, ammoxidation involves two different mechanisms during the early-stage run-in period: the formation of acetonitrile dependent exclusively upon the chemisorption of NH3 on the catalytic active sites. Ethylene and CO2 appear immediately in the gas based on the nonselective ODH mechanism, whereas the formation of acetonitrile was delayed (the same as NH3) because it requires building a catalytically active surface through NH3 pre-chemisorption. The production of acetonitrile reaches the steady-state at 1.5 min. Quite similar features of delay on acetonitrile and NH3 were also observed for the Co-IE and Co-PM-2 wt % catalysts (see Figure S5b,d). However, the delay of O2, as well as the quantitative information of such early-stage features for different reactants and products, seem to be affected by the catalyst preparation.

The back-transient kinetics was initiated by changing the reactor flux from the reactant mixture back to the inert gas (Ar). Kinetics information about the reactivity of the chemisorbed intermediates (on the catalyst surface) can be obtained from the time constants of the product decay. Assuming first-order reaction kinetics on single-type catalytically active sites, the rate for the formation of final products during the extended back-transient can be expressed as ln(ratetransient) = ln(ratesteady-state) – t/τ (or ln(ratetransient) = ln(ratesteady-state) – kt). Details about the derivation of such a linear equation can be found in our previous paper.[50] The entire back-transient behavior of the normalized outlet flow for both reactants and products is shown in Figure b. The time constant τ for the decay of NH3 and CO2 is 7.7 min, which decreased to 2.7 min for acetonitrile and further decreased to only 3–4 s for ethane, ethylene, and O2. The time constant(3–4 s) is close to the reactor response of the inert gas (2 s) and the response of the blank test (see Figure S4a). During the entire back-transient process, the first few minutes after removal of reactants were considered as the near steady-state behavior (see Figure c). It takes only about a few seconds for the complete decay of O2, while the decay of ethane and ethylene take slightly longer than that for O2. They all disappeared from the gas phase within about 10 s, which is similar to the time required for the complete appearance of Ar in the gas phase. Therefore, we suggested that the chemisorbed surface intermediates do not lead to the formation of ethylene in the absence of gas-phase ethane and NH3. Additionally, ethane and O2 either chemisorbed weakly (completes poorly with NH3) on the catalyst or the chemisorbed ethane and O2 converted immediately to another type of intermediates (presumably the most abundant reactive intermediates or MARI). The decay of acetonitrile, CO2, and NH3 requires a longer time than that for ethane and ethylene. The rate constant for the formation of acetonitrile and CO2 is 0.37 and 0.13 min–1, respectively, (see Figure d) from their corresponding reactive intermediates. A strict linear relation between ln(ratetransient) and time during the extended back-transient indicates that such reactions follow the first-order kinetics. The near steady-state back-transient also shows a sudden decay of acetonitrile, NH3, and CO2 within the initial 30 s, which then level-off for various time periods before the final first-order decay. Such unique behavior is most probably because of the formation of two different types of catalytically active sites, presumably Co oxide particles located on the external surface of the HZSM-5 zeolite and exchanged Co2+ cations inside the pores of the zeolite. Additionally, the time constant for the decay of NH3, CO2, and acetonitrile over the Co-IE and Co-PM-2 wt % catalysts (see Figure S5a,c) was quite similar to that over the optimal Co-IM-2 wt % catalyst, indicating that the same type of reactive intermediates and catalytically active sites are involved. It must be noted that the decay of CO2 over the HZSM-5 sample (Figure S4c) is significantly faster than that over the Co-modified HZSM-5 catalysts (note that the activity of the HZSM-5 sample is very low in ethane ammoxidation, see Figure S1), indicating the formation of CO2-related species (probably formates or carbonates) on the Co sites.

In terms of the reaction pathways and mechanism, we suggested that ammonia mediated the initial step ethane ODH. The C–N bond was formed immediately after the initial C–H bond activation, which acts as an intermediate for acetonitrile. Meanwhile, such an intermediate can also be decomposed into ethylene and NH, which might be partially oxidized into CO2 and NO, respectively. The overall kinetic expression for the rate of acetonitrile has been proposed by Li and Armor.[31] They suggested that the formation of the C–N bond between ethylene and chemisorbed NH3 on the Co–OH sites could produce the ethylamine-type intermediate, whereas the formation of ethylene from ethane ODH was proposed on a clean Co–OH site in the absence of NH3, which is less likely under the ammoxidation conditions because NH3 terminates such chemisorption on Lewis acid Co sites. Li and Armor also proposed the formation of N2 from the ethylamine-type intermediate with an additional chemisorbed NH3. However, the formation of NO, which was clearly observed from our study, was not considered. With respect to the catalytic active sites, Li and Armor suggested that the exchanged Co2+ cations to be responsible for the formation of all products from ethane ammoxidation. We anticipate that CO2 may originate from the ethylene oxidation, and such a reaction was particularly favored because of the presence of the Co nanoparticles without close proximity with the zeolites. The efficient formation of acetonitrile may also be related to the remaining Brønsted acid sites of the zeolite host, which could provide protonated NH4+ in the first step, followed by the formation of “C2H5NH3+···ZSM-5–” (as the reactive intermediate) through a mechanism similar to the reverse Hofman-elimination. Therefore, supported by our catalytic results, we suggested that a proper balance between the exchanged Co2+ cations and the Brønsted acidity density would adjust and optimize the activity/selectivity to acetonitrile in this catalytic system.

Conclusions

While the cobalt-modified HZSM-5 catalyst has been extensively studied for the ethane ammoxidation since the later 1990s, more detailed discussion about the structure/performance relationships and the kinetics of the reaction mechanism has been less concerned. We present here our understanding of these questions based on the performance of the catalyst with tuned metal/acid functions. Such steady-state catalytic performance was complemented with the early-stage transient analysis and the back-transient kinetics of the product decay study. We identified that the catalyst prepared through simple impregnation shows similar catalytic performance with that prepared through the ion-exchange method. The same type of the active site, namely, the exchanged Co2+ cation was formed by two preparation methods after activation. The catalyst prepared through physical mixing also leads to the formation of exchanged Co2+ sites (with the co-presence of nanosized Co oxide particles), showing similar product selectivity. The effect of the Si/Al2 ratio and Co loading on the activity and selectivity suggested that excess Co amounts or deficient Brønsted acid sites density resulted in the formation of CO2 rather than acetonitrile and ethylene, which most probably due to the formation of aggregated cobalt oxide particles as identified by the STEM–EDX, XRD, XPS, and UV–vis-DRS. We suggested that the excess Brønsted acid sites of the zeolites might act as a reservoir for NH3, which promotes the formation of acetonitrile while inhibiting the CO2 formation. The early-stage catalytic behavior during the transient kinetic analysis indicated that NH3 mediated the initial ODH of ethane. Ammonia and acetonitrile appear in the product stream simultaneously before building the NH3 reservoir, whereas CO2 was mainly produced. According to the back-transient kinetic analysis, the decay of NH3 and CO2 shows the same large time constant τ at 7.7 min, indicating that the presence of NH3 strongly reduced the rate of CO2 formation from its intermediate precursor. The time constant for acetonitrile was decreased to 2.7 min and further decreased to only 3–4 s for ethane, ethylene, and O2. Finally, the kinetics from the back-transient analysis suggested the first-order reactions for the formation of acetonitrile and CO2 from their corresponding reactive intermediates. Their rate constants are 0.37 and 0.13 min–1, respectively.

Experimental Section

Catalyst Preparation

Co/HZSM-5 with different Co loadings was prepared through incipient wetness impregnation. The NH4-ZSM-5 zeolites were purchased from VWR International (SBET ≈ 400 m2/g). Before impregnation, the NH4-ZSM-5 zeolites were converted into proton-type through calcination at 550 °C under air for 6 h. For a typical impregnation process, 6 mL of an aqueous solution of cobalt nitrate (Sigma-Aldrich) was added dropwise to the calcined HZSM-5 (3 g). The obtained sample was kept in air for 12 h at 60 °C. Finally, the sample was calcined under flow of air in a muffle furnace at 560 °C for 4 h. The obtained samples were named as Co-IM # wt %, where the # represents the nominal metal-based Co loading. The obtained powder samples were pressed into pellets and sieved to obtain a particle size of 20–45 mesh for characterization and catalytic performance tests.

For the purpose of comparison, Co-modified HZSM-5 catalysts were also prepared through ion-exchange (named as Co-IE) and physical mixing (named as Co-PM). For ion-exchange, 3 g of NH4-ZSM-5 zeolite was exchanged with 0.05 M cobalt nitrate aqueous solution at 80 °C for 6 h and repeated three times. The obtained slurry was centrifuged and washed with distilled water three times. For physical mixing, 0.534 g of cobalt nitrate was mixed with 3 g NH4-ZSM-5 using a Fritsch Pulverisette ball mill for 5 min. The obtained sample was finally dried and calcined according to the same procedure described above.

Catalyst Characterization

Nitrogen adsorption/desorption isotherms were measured at −195.8 °C on a TriStar II 3020 gas adsorption analyzer of Micromeritics. Prior to the measurement, the sample was degassed under high vacuum at 300 °C for 8 h. The total surface area was calculated from the adsorption branch in the range of relative pressure from 0.05 to 0.25 by the Brunauer–Emmett–Teller (BET) method, whereas the total pore volume was estimated at a nitrogen relative pressure of 0.99. The micropore volume and external surface area were calculated from the isotherms by the t-plot method; the micropore surface area was obtained from the difference between the total surface area and external surface area.

Ammonia-TPD and n-propylamine-TPD experiments were performed in a quartz reactor with a volume of 2 mL (ID, Φ = 1/2 in). Prior to NH3 and n-propylamine adsorption, the catalysts (0.1 g for NH3-TPD and 0.06 g for C3H7NH2-TPD) were pretreated in Ar at 650 °C for 20 min. Adsorption of NH3 or n-propylamine was carried out at 120 °C under pure NH3 for 30 min or through pulsing n-propylamine injection until saturation, respectively. For both experiments, the samples after pre-adsorption were flushed with Ar (at 120 °C for NH3 and 200 °C for n-propylamine) for 2 h to remove the physically adsorbed NH3 and n-propylamine. Finally, the temperature of the sample was increased to 620 °C at a ramp of 10 °C/min under the flow of Ar at 20 mL/min. Desorption of NH3 (m/z = 17) was measured during the NH3-TPD by an online mass spectrometer (Agilent 5973). In the C3H7NH2-TPDec experiment, the desorption signals of NH3 (m/z = 17), C3H6 (m/z = 41), benzene (m/z = 78), toluene (m/z = 92), and xylene (m/z = 106) were measured with the same mass spectrometer. More detailed experimental procedures for NH3- and n-propylamine-TPD have been described in our previous work.[44]

The solid-phase composition prepared by different methods and various cobalt loading percentages was identified by X-ray powder diffraction with a scanning rate of 4° min–1 in the range of 2θ from 4 to 50°. The purity and crystallinity of zeolites of the sample obtained by pressing the powder into schistose were measured by the X-ray diffractometer on the Rigaku Ultima III XRD system with Cu Kα radiation (154.06 pm, 40 kV, and 44 mA).

TEM/STEM–EDX characterization of the sample was performed on a JEOL 2100TEM (accelerating voltage 200 kV) equipped with a Gatan camera.

The surface chemistry and composition of the catalysts were measured using a PHI Quantum 2000 XPS. The samples were crushed and pressed into indium metal foils, which were then mounted on the XPS sample holders. A monochromated Al Kα radiation (1486.6 eV) at an operating power of 18 kV was used for the analysis. The core-level spectra were monitored by employing a pass energy of 23 eV for the high-resolution scans (except for trace elements) and 188 eV for the survey scans. The C 1s line at 284.8 eV of adventitious carbon was taken as an energy reference to compensate for surface-charging effects. The spectra obtained were processed and analyzed using the CasaXPS software.

The ultraviolet–visible diffuse reflectance spectra (UV–vis-DRS) were collected on a Cary 5000 UV–vis–NIR spectrophotometer (Agilent) equipped with a diffuse reflectance attachment with a polytetrafluoroethylene integrating sphere. Before each measurement, the hydrated Co-ZSM-5 samples were transformed to their dehydrated form under 10–1 Pa and 400 °C for 7 h, which were then transferred into a sample cell in a glovebox (O2 < 0.1 ppm; H2O < 0.1 ppm). Operated at a scan speed of 10 nm s–1, a step length of 1 nm, and a slim width of 5 nm, the UV–vis-DRS were collected in a differential mode referenced to their parent H-ZSM-5 zeolites.

Catalytic Testing

The ammoxidation of ethane was conducted in a home-made U-shape quartz reactor equipped with an on-line Agilent 5973 mass spectrometer. The manifold reactor setup has two parallel gas lines, switched by a four-way valve, connecting to the reactor. The reactor influent gas can be abruptly switched from inert (Ar) to the reactant mixture (C2H6/NH3/O2) and vice versa, which provides the dynamic information for both the early-stage transient period and the back-transient reaction fade away.[50] Typically, 0.2 g of the catalyst was loaded in the reactor and pretreated in Ar (20 mL/min) at 650 °C (ramp 10 °C/min) for 20 min to remove the impurities adsorbed on the catalyst. The temperature of the reactor was then cooled to 475 °C for the reaction. Prior to each experiment, bypass feed spectra were recorded for mass spectrometer calibration and used as a reference for activity calculation. Finally, the reaction was initiated abruptly by switching the influent gas from Ar to the reactant mixture (10% C2H6, 10% NH3, and 6.4% O2, balance with Ar) at a total flow of 80 mL/min. The corresponding gas hourly space velocity is 32 000 h–1. All of the connection gas lines between the reactor and mass spectrometry were heated to 150 °C to avoid the condensation of acetonitrile and water.