Supported metal catalysts are the most widely used type of heterogeneous catalysts and play an important role in the chemical industry [1]. The performance of a supported metal catalyst depends on the properties of the metal, support, and pretreatment [2]. In addition, some studies have found that the interaction between the metal and support can also influence the performance through the carrier effect, which can be used to improve the stability, enhance the activity, and study the underlying mechanism [3-8].

The impregnation method is a simple way to prepare the supported metal catalyst. With this type of method, the support is immersed in a solution of metal precursor, and capillary forces in the pores enable the solvent to be distributed throughout the full support. The resulting catalyst is then set aside for several hours to ensure the equilibrium of the precursor distribution. After drying and calcination, the catalyst is created [9, 10]. However, during conventional evaporation drying, the solvent will move within the absorbed layer, disturbing the distribution of the metal precursor. Moreover, in a porous support, the capillary pressure will transport the solvent to the external surface. Consequently, a large amount of metal dissolved in a solution will be transported to the external surface, forming significant aggregations [11]. Therefore, it is difficult to obtain sufficient metal dispersion and metal-support interaction over the catalyst. Some researchers have added some viscous materials into the impregnating solution to suppress the liquid movement, providing a more uniform distribution [12]. Vergunst et al. applied microwave drying in preparing a nickel-based monolithic catalyst to avoid metal accumulation [13]. However, research on a general method to increase the metal-support interaction in an impregnation method remains insufficient.

In our study, we used freeze drying instead of traditional evaporation drying to solve the above problem. Freeze drying is used to remove the solvent from a frozen sample through sublimation under vacuum conditions. During sublimation, the mobility of the precursor is greatly restricted [14, 15]. In the field of chemistry, freeze drying is typically used to synthesize nanometal powders [16]. Nevertheless, this method is rarely applied in impregnation to enhance the metal-support interaction of a catalyst. Herein, Fe and ZSM-5 were selected as the representative metal and support, respectively. FeZSM-F (1.0 wt% ⁵⁷Fe) was prepared through freeze-drying impregnation and compared with the evaporation-drying impregnation catalyst (FeZSM-E, 1.0 wt% ⁵⁷Fe). The structure of the Fe species in the prepared catalysts and their performance during N₂O decomposition are discussed. In addition, the formation mechanisms are proposed.

FeZSM-5 catalysts were prepared through impregnation. A certain amount of HZSM-5 was added into the ⁵⁷Fe(NO₃)₃ solution and stirred for 2 h. In addition, ⁵⁷Fe(NO₃)₃ was created by dissolving ⁵⁷Fe metal powder with a dilute HNO₃ solution at 90 ℃. After filtering, the well-mixed samples were separated into two parts. One part was dried in oven at 120 ℃ for 10 h, whereas the other was immediately frozen using liquid nitrogen and put into a vacuum freeze dryer at –10 ℃ and 1 Pa for 10 h. The samples were then heated at 550 ℃ in air for 8 h. Finally, the samples were calcined in helium flow at 900 ℃ for 1 h. The obtained samples were labeled FeZSM-E and FeZSM-F, where E and F indicate evaporation drying and freeze drying, respectively.

The magnetic properties were determined at room temperature using a superconducting quantum interference device magnetometer (SQUID, MPMS-XL-5, Quantum Design Company of USA). In FeZSM-5 catalysts, ZSM-5 is a major component with diamagnetism, which exerts a significant influence over the measurement of iron magnetization. Thus, the spectra of ZSM-5 are subtracted from the catalysts to obtain the real spectra of the iron species. The H₂ temperature-programmed reduction (H₂-TPR) experiment was conducted using Micromeritics ASAP 2920 to analyze the reducibility of the catalysts. Prior to the reduction, the catalyst (50 mg) was treated in flowing Ar (50 mL/min) at 500 ℃ for 1 h and cooled to 50 ℃. In addition, 10% H₂-Ar was passed through the sample. TCD signals were recorded from 50 to 900 ℃ at 10 ℃/min. The ⁵⁷Fe Mössbauer spectra were determined using a Wissel spectrometer within a velocity range of -11.2 to +11.2 mm/s at room temperature using a ⁵⁷Co (Pd) source. The isomer shift was given with respect to α-Fe at room temperature. The solid-state ²⁷Al NMR experiments were conducted on a Varian InfinityPlus 300 at room temperature with a magnetic field of 7.1 T. In addition, the resonance frequency was 78.13 MHz, and the spinning rate of the sample was 8.0 kHz. The spectra were referenced to 1 mol/L Al(NO₃)₃ at 0 ppm.

0.1 g catalyst (20–40 mesh size) was loaded into a tubular quartz fixed-bed reactor (I.D. = 8 mm) at atmospheric pressure. In addition, 1.0% N₂O with a He balance (150 mL/min) was introduced into the reactor. The reactants and products were analyzed using Porapak Q and molecular sieve 5A columns.

The magnetic property of the catalysts was investigated using SQUID. As shown in Fig. 1, two catalysts show different shapes, indicating that they have a different Fe magnetic species composition. Hysteresis cycles with a positive slope are observed in both catalysts, indicating at least a mixed property of ferromagnetism and paramagnetism. In iron-supported ZSM catalysts, γ-Fe₂O₃, α-Fe₂O₃, and Fe²⁺ bonded with the framework may appear as a ferromagnetic phase [17]. However, γ-Fe₂O₃ will be transformed into α-Fe₂O₃ at above 400 ℃ [18]. Because the calcination temperature is 900 ℃ during preparation, γ-Fe₂O₃ should not exist in the catalysts. In addition, the paramagnetic behavior may be ascribed to Fe³⁺ bonded with the framework [17].

The reducibility of the catalysts was investigated through H₂-TPR experiment (Fig. 2). The peaks at 350–600 ℃ can be classified as the reduction of Fe³⁺ to Fe²⁺. The Fe³⁺ species may include isolated Fe³⁺, Fe_xO_y clusters, or α-Fe₂O₃ nanoparticles [19-21]. Within this range, there are two peaks for FeZSM-E, indicating the diversity of Fe³⁺ species in FeZSM-E, whereas FeZSM-F only has a single peak. In addition, FeZSM-E has a larger amount of Fe³⁺ species than FeZSM-F. The peaks at a higher temperature (600–800 ℃) represent reduction of Fe²⁺ to Fe⁰ [19-21]. According to the integral results, the consumption of H₂ is higher for FeZSM-F (0.26 versus 0.24 mmol/g). Considering that both catalysts have the same iron content and fewer Fe³⁺ are reduced to Fe²⁺ in FeZSM-F, FeZSM-F must possess a greater amount of Fe²⁺ species.

⁵⁷Fe Mössbauer spectroscopy is a powerful tool for analyzing the Fe species [22]. In the Mössbauer spectra (Fig. 3), the valence state and coordination of the Fe species are reflected through the isomer shift (IS) and quadrupole splitting (QS), respectively. Table 1 shows the parameters obtained through a spectral devolution. According to the literature [23-25], 0.1–0.6 mm/s is assigned to Fe³⁺, whereas 0.7–1.4 mm/s is ascribed to Fe²⁺. For Fe³⁺, the IS of lower than 0.3 mm/s is within the T_d coordination, whereas the IS higher than 0.3 mm/s is in the O_h coordination. Doublet(Ⅰ) (0.1 mm/s ≤ IS ≤ 0.3 mm/s, 0.87 mm/s ≤ QS ≤ 1.24 mm/s) is assigned to Fe³⁺-Al-Si of the T_d coordination in the extraframework [19, 26, 27]. Doublet(Ⅱ) (0.30 mm/s ≤ IS ≤ 0.37 mm/s, 0.65 mm/s ≤ QS ≤ 1.06 mm/s) is attributed to small Fe_xO_y nanoparticles in the O_h coordination with superparamagnetism [23]. Some researchers have suggested that small Fe_xO_y nanoparticles were formed inside the channel [28-30]. Sextet(Ⅰ), Sextet(Ⅱ), and Sextet(Ⅲ) (0.36 mm/s ≤ IS ≤ 0.50 mm/s, 39.2 T ≤ B_hf ≤ 51.4 T) belong to large bulk magnetic iron oxide outside ZSM-5, which is larger than 10 nm [31]. To further confirm the state, we measured the spectra of 30 nm α-Fe₂O₃ (Fig. S1, Table S1) and found a similar IS value but with a different B_hf (hyperfine magnetic field). Thus, we ascribe Sextet(Ⅰ), Sextet(Ⅱ), and Sextet(Ⅲ) to large α-Fe₂O₃ particles (> 10 nm) with different sizes outside ZSM-5, which is also in line with a former magnetic characterization. For Fe²⁺, a higher QS is ascribed to higher coordination numbers. Here, Doublet(A) (QS > 1 mm/s) is assigned to Fe²⁺ in the O_h coordination, and Doublet(B) is assigned to mononuclear Fe²⁺ in the planer coordination [29, 31, 32].

Table 1 lists the quantitative results. For Fe³⁺-Al-Si, FeZSM-F has a larger amount (10.4% versus 3.0%). For Fe²⁺, FeZSM-E contains two types, namely, Fe²⁺ of the O_h coordination (1.7%) and planer coordination (7.2%), whereas FeZSM-F only contains Fe²⁺ in the O_h coordination (29.6%), which is much higher than the total Fe²⁺ amount (8.9%) in FeZSM-E. Because the formation of Fe²⁺ is connected with Al (as illustrated later), and the ²⁷Al MAS NMR spectra (Fig. S2) show that aluminum in ZSM-5 remains predominately in the framework (50 ppm), Fe³⁺-Al-Si and two Fe²⁺ types are formed through interactions between the iron and support. It is clear that FeZSM-F has more of these species (40.0% versus 11.9%). Thus, freeze drying contributes more to the enhanced metal-support interactions.

For small Fe_xO_y nanoparticles of the O_h coordination in the channel, FeZSM-F contains more than twice the amount of FeZSM-E (27.4% versus 12.2%). For large α-Fe₂O₃ particles, FeZSM-E has three types with a total amount of 75.9% and more than one type with 32.6% in FeZSM-F. As most of the framework aluminum exists in the channel [33], Fe³⁺-Al-Si and Fe²⁺ are mainly formed in the channel, similar to small Fe_xO_y nanoparticles, whereas large α-Fe₂O₃ particles are formed outside the ZSM-5. Thus, we can divide the Fe species into two categories: inside and outside the ZSM-5. In FeZSM-F, most of the iron species (~67.4%) are inside the ZSM-5. However, in FeZSM-E, most of the Fe species (~75.9%) are outside the ZSM-5. Clearly, freeze drying has a confining effect on the iron species.

Owing to the confinement of the narrow channel (diameter of less than 1 nm), the sizes of the iron species inside the ZSM-5 are much smaller than the iron species outside the ZSM-5 (larger than 10 nm). Thus, the overall size of the iron species in FeZSM-F is smaller than that in FeZSM-E. This conclusion can be confirmed from another perspective [20]. In the Mössbauer spectra, the particle size of the iron species has a significant impact on the peak type. When the size is below a certain value, the peak type will be the doublet; however, above the value, the peak will turn into the sextet [20]. In FeZSM-F, there are 67.4% iron species in the doublet, and 32.6% in the sextet. However, in FeZSM-E, there are only 24.1% iron species in doublet, and 75.9% in the sextet. All of the above factors indicate that freeze drying favors the formation of a smaller metal species.

For different valence states, FeZSM-F has more Fe²⁺ species than FeZSM-E (29.6% versus 8.9%), whereas FeZSM-E has more Fe³⁺ species than FeZSM-F (91.1% versus 70.4%), which is also consistent with the trend found in the H₂-TPR results. Furthermore, FeZSM-E has more types of iron species than FeZSM-F (7 versus 4), including more Fe²⁺ (2 versus 1) and Fe³⁺ (5 versus 3). Overall, freeze drying contributes more to the formation of Fe²⁺ and the diversity of the iron species.

Based on the results above, formation mechanisms are displayed in Fig. 4. First, both samples are immersed in a precursor, and the iron species are well dispersed owing to the capillary forces. During the preparation of FeZSM-F, the sample is then frozen, and water is removed through sublimation. In the absence of the capillary force, Fe(NO₃)₃ remains well distributed in the ZSM-5 after freeze drying (step (A)). With a large specific surface area and abundant micropores in the ZSM-5, most of the Fe(NO₃)₃ is confined in the channel, and there are more opportunities for Fe(NO₃)₃ to interact with the framework Al during calcination, resulting in more Fe³⁺-Al-Si and Fe²⁺. Because not all Fe(NO₃)₃ is able to contact the limited framework Al, the remaining Fe(NO₃)₃ in the channel will form small Fe_xO_y nanoparticles. Owing to the confinement of the channel, the sizes of the iron species are rather small. Only a few Fe(NO₃)₃ outside the ZSM-5 will form large α-Fe₂O₃ particles. Because the concentration of Fe(NO₃)₃ is uniform, large α-Fe₂O₃ particles of only one size are formed (step (B)). However, in the preparation of FeZSM-E, during evaporation drying, most of the Fe(NO₃)₃ is transported along the channel by the capillary force to the external layer, and accumulates on the surface of the ZSM-5 (step a). Without the restriction of narrow channels, many large α-Fe₂O₃ particles (> 10 nm) are formed after calcination. In addition, only a few Fe(NO₃)₃ particles are confined in the channel, producing fewer Fe³⁺-Al-Si, Fe²⁺, and small Fe_xO_y nanoparticles. Because the concentration differs in various parts, two Fe²⁺ types and three different sizes of large α-Fe₂O₃ particles are formed (step (b)).

Herein, we suppose that the two Fe²⁺ species contain different iron nucleus, which depends on the distance between the iron atoms. For FeZSM-E, the distribution of iron is non-uniform inside the ZSM-5. In the low iron concentration part, the distance between two iron atoms is relatively large, and it may be easier to form Doublet(B) (mononuclear Fe²⁺); however, in the high concentration part, the distance is relatively small, and it may be easier to form a multinuclear species. Doublet(A) (Fe²⁺ in O_h coordination) is speculated to contain two or more iron atoms. For FeZSM-F, the iron is uniformly distributed, and the concentration is insufficiently low to create the mononuclear species, and thus only Doublet(A) is formed.

To verify the difference in composition, the catalysts are tested through N₂O decomposition reaction. The active centers in the N₂O decomposition are a series of Fe species. After high-temperature treatment in inert gas, some Fe³⁺ will be reduced to Fe²⁺, which are the active centers [29, 31, 33-36]. Although the structure of the active Fe²⁺ remains under debate, there is a consensus that Fe²⁺ should be connected with Al. Some authors have ascribed the active centers to bare Fe²⁺ or a Fe²⁺-O-Fe²⁺ complex stabilized by Al³⁺ pairs, whereas others have considered square planer mononuclear Fe(Ⅱ) in the β-6MRs, which contains two Al T-sites, as the active center [32, 37]. Small Fe_xO_y nanoparticles are also responsible for such activity [30]. However, large Fe₂O₃ particles are inactive or have low activity, which is also shown through our experiment (Table S2). Thus, Doublet(A), Doublet(B), and Doublet(Ⅱ) are ascribed to active species. Because there are more active species over FeZSM-F than FeZSM-E (57.0% versus 21.1%), FeZSM-F should exhibit a higher activity. N₂O decomposition activities at 420 and 440 ℃ are illustrated in Table 2. With the same Fe content in both catalysts, different TOF_N2O (turnover frequencies, per Fe atoms per second) indicate the distinction of the Fe species. As expected, FeZSM-F has a higher TOF_N2O. At 420 ℃, TOF_N2O for FeZSM-F is 2.2-times that for FeZSM-E, whereas at 440 ℃, both TOF_N2O increase significantly, and TOF_N2O for FeZSM-F is 3.1-times that for FeZSM-E. Fig. 5 shows Arrhenius plots of the N₂O decomposition (430–510 ℃). The value of E_a (activation energy) is lower in FeZSM-F (123.2 kJ/mol) than in FeZSM-E (153.3 kJ/mol).

To summarize, freeze-drying impregnation is applied in the preparation of ⁵⁷FeZSM-5. Compared with the traditional evaporation-drying catalyst, more iron species are confined inside ZSM-5 in a freeze-drying catalyst, creating enhanced metal-support interactions, smaller sized species, and less diversity. The difference in the composition is also proved based on the superior performance of the freeze-drying catalyst during N₂O decomposition. For preparation of the supported metal catalysts, in most cases, the metal-support interaction and the size of the metal species are vital for the reaction, and our study provides a general way to tune the metal-support interactions and the size of the metal species in impregnation method, which can be extended to other metals and supports.