**Manganese and Cobalt Doped Hierarchical Mesoporous Halloysite-Based Catalysts for Selective Oxidation of** *p***-Xylene to Terephthalic Acid**

**Eduard Karakhanov 1, Anton Maximov 1,2, Anna Zolotukhina 1,2, Vladimir Vinokurov 3, Evgenii Ivanov <sup>3</sup> and Aleksandr Glotov 3,\***


Received: 18 November 2019; Accepted: 14 December 2019; Published: 18 December 2019 -

**Abstract:** Bimetallic MnCo catalyst, supported on the mesoporous hierarchical MCM-41/halloysite nanotube composite, was synthesized for the first time and proved its efficacy in the selective oxidation of *p*-xylene to terephthalic acid under conditions of the AMOCO process. Quantitative yields of terephthalic acid were achieved within 3 h at 200–250 ◦C, 20 atm. of O2 and at a substrate to the Mn <sup>+</sup> Co ratio of 4–4.5 times higher than for traditional homogeneous system. The influence of temperature, oxygen, pressure and KBr addition on the catalyst activity was investigated, and the mechanism for the oxidation of *p*-toluic acid to terephthalic acid, excluding undesirable 4-carboxybenzaldehyde, was proposed.

**Keywords:** halloysite; hierarchical materials; *p*-xylene oxidation; terephthalic acid

#### **1. Introduction**

Terephthalic acid (TPA) and its dimethyl ester are important monomers in thermoresistant and mechanically stable polymer production, such as polyethylene terephthalate (PET) and poly-paraphenylene terephthalamide (Kevlar) [1,2]. The world production of terephthalic acid has exceeded 5 Mt/yr. is still growing [1,2]. Currently up to 70% of terephthalic acid is produced by direct oxidation of *p*-xylene through the AMOCO/Mid-Century process being developed in the 1960s [1–6].

*p*-xylene in the AMOCO process is subjected to oxidation by molecular oxygen or air in the presence of homogeneous Mn and Co catalysts with KBr as a promotor and free-radical source in acetic acid medium under the severe conditions (170–220 ◦C, 15–30 atm of O2 or 50–70 atm of air), giving terephthalic acid yield of 90–97% within 4–12 h (Scheme 1) [7–10].

**Scheme 1.** Oxidation of *p*-xylene to terephthalic acid in the AMOCO process [1,2,7,8].

In spite of the obvious advantage of the AMOCO process, such as quantitative yield of high purity terephthalic acid, acetic acid medium with bromides cause corrosion that makes it necessary to use expensive titanium reactors [11–13]. It challenges the development of new environmentally friendly, safer and less corrosive reaction media, promoters and additives [7]. Catalyst heterogenization is considered as one of the possible ways for improving the AMOCO process.

Thus, it was earlier demonstrated, that the use of CoO and Co3O4 as catalysts together in the presence of MnO, NiO or CeO co-catalysts and *p*-toluic acid as a promotor allowed for oxidation without KBr [14]. Nonetheless, this system appeared to be much less effective in comparison with the conventional homogeneous system, including Mn(OAc)2, Co(OAc)2 and KBr, and the yield of terephthalic acid did not exceed 65–70% within 8 h [14].

In the presence of MCM-41 doped with Fe and Cu, terephthalic acid underwent further oxidation to 2,5-dihydroxy-1,4-terephthalic and p-benzoquinone-2,5-dicarboxylic acids even under much milder conditions, than in the AMOCO process (H2O2 as an oxidant, 80 ◦C, AcOH, CH3CN, 5 h) [15]. The highest selectivity to terephthalic acid did not exceed 45% at a *p*-xylene conversion of 10% in the neat acetonitrile. It was found out, that iron additive increased the selectivity to terephthalic acid, whereas copper addition, vice versa, favored the further oxidation of TPA to 2,5-dihydroxy-1,4-terephthalic and p-benzoquinone-2,5-dicarboxylic acids [15].

High yields of terephthalic acid (99% within 2 h) were obtained in the presence of polynuclear μ3-oxo-bound complexes of Co and Mn, encapsulated in the cavities of Y zeolite, under the conditions similar to AMOCO process (KBr, 200 ◦C, 610 atm of the air) [16]. The said catalyst appeared as highly resistant to the metal leaching due to close sizes of polyoxo metal clusters and zeolite cavities [8,16].

In this connection, halloysite-based materials appeared to be promising as heterogeneous catalysts for *p*-xylene oxidation under the typical conditions. Halloysite is a natural clay with the rolled tubular structure, appearing as a multiwall nanotube (halloysite nanotube, HNT) with a length of 0.5–1.5 mm, an outer diameter of 50–60 nm and an inner cavity diameter of 10–15 nm (Figure 1) [17,18]. Halloysite clays were successfully applied as carriers for the tubular Ru nanocatalysts, revealing high activity in the hydrogenations of aromatic compounds and phenols [19–23].

**Figure 1.** Schematic visualization (**left**) and TEM image (**right**) of halloysite clay.

Grafting of the ordered mesoporous materials, such as MCM-41 or SBA-15, onto halloysite template allows us to obtain new hierarchical systems with stronger mechanical properties and surface area up to 650 m2/g (Figure 2) [21,24,25]. La-doped MCM-41/HNT composite revealed high efficacy as a sulfur-reducing additive for FCC (fluid catalytic cracking) catalyst, resulting in decrease of sulfur content by 25% and in the yield of gasoline fraction of about 45% [24,26,27]. Modified with CaO and MgO, MCM-41/HNT and SBA-15/HNT composites demonstrated high activity in the cracking of sulfones, formed after the oxidative desulfurization of diesel fraction, decreasing sulfur content from 450 up to 100 ppm [28]. The catalysts said were recycled several times without significant loss of activity, and with high resistance to metal leaching and structure maintenance under the reaction conditions [28].

**Figure 2.** Schematic visualization (**left**) and TEM image (**right**) of MCM-41/HNT composite.

In this work, we present for the first-time synthesis of new heterogeneous bimetallic MnCo catalyst, based on mesoporous hierarchical MCM-41/HNT composite that can be successfully applied for quantitative oxidation of *p*-xylene to terephthalic acid in the AMOCO process, proving its efficiency, that was 4–4.5 times higher, as compared with a traditional homogeneous system.

#### **2. Results and Discussion**

#### *2.1. The Synthesis and Characterization of Hierarchical Mesoporous MCM-41*/*HNT Composite, Doped with Mn and Co*

Bimetallic MnCo catalyst, based on MCM-41/HNT composite was synthesized by the wetness impregnation method. Herein Mn2<sup>+</sup> and Co2<sup>+</sup> were deposited from the water solution of Mn(OAc)2 and Co(OAc)2 in molar ratio of 1:10 (Scheme 2) [16]. Mn(OAc)2 and Co(OAc)2 tetrahydrates were chosen as sources for Mn2<sup>+</sup> and Co2<sup>+</sup> respectively to avoid the influence of other counter-anions on the adsorption and oxidation processes, and because of acetic acid, used as a solvent in the conventional oxidation process [2,8].

**Scheme 2.** Mn(OAc)2 and Co(OAc)2 deposition onto MCM-41/HNT composite.

The material obtained was characterized by atomic emission spectroscopy with inductively coupled plasma (ICP-AES), transmission electron microscopy (TEM) and X-ray photoelectron spectroscopy (XPS). The physical and chemical properties are listed in Table 1.


**Table 1.** Physical and chemical properties of the synthesized MnII1CoII10@MCM-41/HNT catalyst.

As seen from Table 1, weight content of Mn and Co in the sample reached 0.15% and 1.29% respectively, which was approximately three times less than corresponding theoretical values. Herein Co/Mn ratio appeared as 8.6:1, which was in accordance with the literature data [29].

According to XPS data (Table 1, Figure 3), both Mn and Co were presented in the forms of simple oxides MnO [30–32] and CoO [33–35] and bound complexes [MnO4] [36,37] and [CoO6] [32–34,38–41], arising from initial Mn(OAc)2 and Co(OAc)2 tetrahydrates as well as from the aluminosilicate. Herein MnO form strongly predominated over [MnO4] bound form for Mn, whereas for Co [CoO6] form, vice versa, predominated over CoO, that might be due to the much stronger oxygen affinity for Co in comparison with Mn [42,43]. Taking in account the presence of nitrogen in the XPS spectra (Table 1), carbon in the sample, found out in –CH2CH2–, –CH2CH2N– and H3CC(=O)– forms [44,45], may be related not only to adsorbed acetate anions, but also to cetyl trimethyl ammonium bromide template, partly remained in the MCM-41/HNT composite after calcination.

**Figure 3.** XPS spectra of Mn (**left**) and Co (**right**) in MnII1CoII10@MCM-41/HNT composite at 2p line.

As TEM analysis showed (Figure 4), deposition of manganese and cobalt acetates did not destruct the MCM-41/HNT matrix. The MCM-41 framework observed consisted of regular hexagonal pores of 2.9 nm in diameter (Figure 4B) and inner located halloysite nanotubes with the inner diameter of 12–15 nm and outer diameter of 31–35 nm, and interplanar space of 3.2 nm (Figure 4A).

**Figure 4.** TEM images of MCM-41/HNT composite (A—HNT templated MCM-41, B—MCM-41) after deposition of Mn(OAc)2 and Co(OAc)2.

#### *2.2. Oxidation of p-Xylene in the Presence of Hierarchical Mesoporous MCM-41*/*HNT Composite Doped with Mn and Co*

The catalyst synthesized was tested in the liquid-phase oxidation of *p*-xylene and compared with traditional homogeneous system Mn(OAc)2/Co(OAc)2/KBr. It should be noted, that *p*-xylene oxidation is a multi-stage radical process (Scheme 3) [2,7,8,46,47]. Its kinetics is strongly affected by the reaction conditions, such as temperature, oxygen pressure, substrate concentration, substrate: Co:Mn ratios, KBr loading, etc. [2,7,8]. Herein KBr is required as a free radical source, and Mn(OAc)2 acts as promotor at the first stage of reaction—activation of the methyl group in the *p*-xylene molecule [2,10]. As a rule, *p*-xylene conversion to *p*-toluic acid proceeds fast, whereas oxidation thereof to terephthalic acid is a slow highly activated process [2,8]. The use of acetic acid as a solvent promotes dissolution of both KBr and *p*-xylene, the ion-radical oxidation process, and deposition of terephthalic acid due to its poor solubility. The TPA product is easily isolated from the reaction medium as white needle-like glittering crystals [2].


**Scheme 3.** Radical involved mechanism of *p*-xylene oxidation in the presence of Mn<sup>2</sup>+/Co<sup>2</sup>+/KBr/AcOH system [2,7,8].

The best conditions for *p*-xylene oxidation, found out in academic studies and widely used in industry (AMOCO process) are as follows: temperature of 180–225 ◦C, oxygen pressure of 15–30 atm, *p*-xylene to acetic acid molar ratio of 0.08–0.16, and molar ratios Mn/Co and Co:Br being 1:8–10 and 3–4 respectively [2,7,8,48–50]. We have used a temperature of 200 ◦C, oxygen pressure of 20 atm and molar ratios *p*-xylene/acetic acid and Mn/Co of 0.08 and 1:10 respectively, both in homogeneous and heterogeneous processes. Herein the reaction turnover frequency (TOF) was calculated as the amount of substrate reacted (νsubstr) per mole of metal (ν*(*Mn <sup>+</sup> Co)) per unit of time with account of each product yield and the number of oxygen atoms added, according to formula:

$$\text{TOF}(\mathcal{O}\_2) = \frac{\nu\_{\text{substr}} \ast (\omega\_{p\text{TALd}} \ast 1 + \omega\_{\text{pTAc}} \ast 2 + \omega\_{4-\text{CBA}} \ast 3 + \omega\_{\text{TPA}} \ast 4 + \omega\_{\text{N}\_\text{I}\text{I}} \ast 3}{\nu\_{(\text{Mn}+\text{Co})} \ast t},$$

where ω is the yield of the certain product, expressed in the unit fractions, N/I is a not identified product (presumably 4-hydroxymethylbenzoic acid) and *t* is the minimal reaction time, for which the reaction progress is measured. The results obtained are listed in Table 2.


**Table 2.** Comparison of homogeneous system Mn(OAc)2/Co(OAc)2 and heterogeneous MnII1CoII10@MCM-41/HNT catalyst in the liquid-phase oxidation of *p*-xylene 1.

<sup>1</sup> Reaction conditions are: 0.5 mL (4.055 mmol) of *p*-xylene, 5 mL of AcOH, 4.5 mg (0.038 mmol) of KBr, 200 ◦C, 20 atm of O2, 3 h–for both homogeneous and heterogeneous processes; 32 mg (0.127 mmol) of Co(OAc)2 and 3.1 mg (0.13 mmol) of Mn(OAc)2 for homogeneous system; 150 mg of bimetallic catalyst for the heterogeneous system.

As one can see, both the homogeneous Mn(OAc)2/Co(OAc)2 system and heterogeneous MnII1CoII10@MCM-41/Hall catalyst give a quantitative conversion and similar product distribution with a TPA yield of 94–95% within 3 h (Table 2). Similar results were obtained earlier for μ-oxo-bridged complexes of Mn and Co, encapsulated into the zeolite Y pores [16]. It should be noted, that near to the quantitative yield of terephthalic acid in the presence of heterogeneous hierarchical catalyst MnII1CoII10@MCM-41/HNT was achieved at much higher substrate to the Mn + Co ratio, giving rise to the higher TOF value (Table 2). It proves the efficacy of the new heterogeneous catalyst developed, comprising of the MCM-41/HNT composite, for the AMOCO process. On the other hand, due to the same loading of KBr for both homogeneous and heterogeneous systems, Co:Br in the latter appeared as high as 1:1, resulting in the increased concentration of Br· radicals and, as a consequence, the easier activation of methyl groups in *p*-xylene and *p*-toluic acid molecules.

Temperature, oxygen, pressure and the presence of KBr were found to be crucial factors for the effective *p*-xylene oxidation process, which was in accordance with the literature data [29]. As seen from Table 3, decrease in oxygen pressure down to 5 atm resulted in corresponding decrease in conversion to 88–89%, and *p*-toluic acid appeared as the major reaction product with the yield of 50–63% for both homogeneous and heterogeneous catalysts. Moreover, lower metal loading in heterogeneous system gave rise to a lower rate of further oxidation of *p*-toluic acid, characterized by the higher activation energy [8], resulting in the extremely low yield of terephthalic acid, about 0.5% (Table 3, Entry 2).

**Table 3.** The influence of the oxygen pressure and KBr presence on the effectiveness of the *p*-xylene oxidation 1.


<sup>1</sup> Reaction conditions are: 0.5 mL (4.055 mmol) of *p*-xylene, 5 mL of AcOH, 4.5 mg (0.038 mmol) of KBr, 200 ◦C, 3 h—for both homogeneous and heterogeneous processes; 150 mg of bimetallic catalyst for heterogeneous system; 32 mg (0.127 mmol) of Co(OAc)2 and 3.1 mg (0.13 mmol) of Mn(OAc)2 for homogeneous system. <sup>2</sup> TPAld is terephthalic aldehyde.

The removal of KBr from the reaction medium led to abrupt downfall in conversion for heterogeneous MnII1CoII10@MCM-41/HNT catalyst even at high oxygen pressure, when all other conditions are equal (Table 3, Entry 3). *p*-Toluic aldehyde was found to be the major reaction product with the yield of 1% only. The results obtained seemed to be connected not only with the absence of the source of Br· radicals, initiating the oxidation stepwise process, but also with the extremely low concentration of Mn3<sup>+</sup> ions in the system, being also responsible for methyl group activation in the substrate. For the successful oxidation of *p*-xylene, Pd-containing systems were suggested [51–53] or compounds, such as *N*-hydroxyphthalimide, being able to generate stable free radicals [8,54,55]. In these cases, however, a very long time (12–48 h) is needed to obtain the yield of terephthalic acid, exceeding 70%.

The influence of temperature on the rate of the *p*-xylene oxidation and product distribution in the presence of MnII1CoII10@MCM-41/HNT heterogeneous catalyst is presented in Table 4. As one can see, the quantitative yield of terephthalic acid may be obtained in 5 h at 200 ◦C and in 3 h at 250 ◦C, and the TOF values being 401.7 and 424.8 h−<sup>1</sup> respectively (Table 4, Entries 6 and 8–9). Hence, an increase in temperature from 200 to 250 ◦C results in a slight rise of the catalyst activity and, therefore, the oxidation rate only.


**Table 4.** The influence of the temperature on the effectiveness of the *p*-xylene oxidation in the presence of MnII1CoII10@MCM-41/HNT catalyst 1.

<sup>1</sup> Reaction conditions are: 150 mg of bimetallic MnCo heterogeneous MCM-41/HNT-based catalyst, 0.5 mL (4.055 mmol) of *p*-xylene, 5 mL of AcOH, 4.5 mg (0.038 mmol) of KBr and 20 atm. of O2.

Vice versa, decrease in temperature down to 150 ◦C led to an abrupt downfall in activity, in accordance with literature data [29,50]. Herein the *p*-xylene conversion did not exceed 40% even after 5 h, with *p*-toluic aldehyde being obtained as the major reaction product (Table 4, Entry 3).

As seen from Tables 2–4, the yield of 4-carboxybenzaldehyde in the presence of heterogeneous MnII1CoII10@MCM-41/Hall catalyst (under 20 atm of O2) did not exceed 0.5%, and 4-hydroxymethylbenzoic acid being formed as the main by-product with the yield up to 5%. At low oxygen pressures the yield of 4-carboxybenzaldehyde increased and reached 12% for homogeneous Mn(OAc)2/Co(OAc)2 system and 4% for heterogeneous MnII1CoII10@MCM-41/HNT catalyst (Table 3, Entries 1–2). It should be noted, that 4-carboxybenzaldehyde is the undesirable product of *p*-xylene oxidation: it is slowly oxidized to terephthalic acid and cocrystallizes with it at separation due to structural similarity [2,7,8].

One may assume, that, when MnII1CoII10@MCM-41/Hall is used as the catalyst, the oxidation passed through the intermediate alcohol formation (Scheme 3, right). Due to the low metal content and/or specific microenvironment of active sites in the MnII1CoII10@MCM-41/HNT catalyst, caused by the interaction of Mn (II) and Co (II) with halloysite and/or MCM-41 matrix, ArCH2OO· radical apparently interacted not with Co2<sup>+</sup>, resulting in aldehyde formation, but with each possible reactant molecule, resulting in the alcohol formation. Noticeably, the oxidation of *p*-xylene to *p*-toluic acid mainly proceeded through the formation of *p*-toluic aldehyde, and only traces of *p*-toluic alcohol were observed at low conversions of *p*-xylene. Hence, the "alcohol pathway" was presumably the characteristic of the oxidation *p*-toluic acid. Thus obtained 4-hydroxymethylbenzoic acid underwent further Co (II) catalyzed oxidation to terephthalic acid, initiated by O2 or Br· as radicals.

We think the explanation above is consistent with the influence of oxygen pressure and KBr observed, as well as with a high reaction rate even at very low metal loading, for the selective oxidation of *p*-xylene to terephthalic acid in the presence of heterogeneous MnII1CoII10@MCM-41/HNT catalyst. Therefore, MCM-41/HNT composite, doped with Mn and Co, may be considered as a prospective catalyst for the AMOCO process, far superior to the original homogeneous Mn(OAc)2/Co(OAc)2 system.

Unfortunately, metal leaching took place during the reaction. The quantity of leached metals depended on the reaction temperature and time: a higher temperature and longer reaction time resulted in less metal maintained in the catalyst (Table 5). Herein Co leached the first followed by Mn. When carrying out the reaction for the longest time (5 h), the concentrations of both Mn and Co dropped approximately by 13–15 times (Table 5, Entry 5). XPS analysis revealed no Mn or Co on the surface of the samples recycled.


**Table 5.** Comparative Mn and Co weight content in the recycled samples of MnII1CoII10@MCM-41/HNT catalyst depending on reaction conditions.

We suppose, the metal leaching observed occurs due to dissolution of the adsorbed Mn and Co species by acetic and hydrobromic acids. The last one is resulted from the interaction of Br· radicals with substrate or intermediate compounds, bearing H-atoms at side carbons (Scheme 3). Nonetheless it should also be noted, that catalyst, based on MCM-41/HNT composite possesses very low packed density and, simultaneously, high apparent adsorption capacity, and, therefore, occupy all of the reaction volume (5 mL of AcOH vs. 150 mg of catalyst). To extract reaction products for HPLC analysis, up to 40 mL of DMSO (the best solvent for terephthalic and *p*-toluic acids) was required, that strictly reduced the possibility for the hot filtration and recycling tests. Mn and Co content in the catalyst recycled was measured just after filtration from DMSO solution. Therefore, the metal leaching could take place not only during the reaction, but also at washing by DMSO.

Hence one may conclude, that pores and cavities of MCM-41 and halloysite moieties act as microreactors for the oxidation of *p*-xylene under the reaction conditions. However these pores appear too wide and not able to effectively retain Mn and Co species inside the carrier at filtration and washing as compared with polynuclear μ3-oxo-bound complexes of Co and Mn, encapsulated in the cavities of Y zeolite [16]. This challenges for the further development and modification of MCM-41/HNT composites to provide more effective metal retention and leaching resistance under the severe reaction conditions, to make possible the repeated use of MnII1CoII10@MCM-41/HNT catalyst in *p*-xylene oxidation.

#### **3. Experimental**

#### *3.1. Chemicals*

The following substances were used as substrates and reference compounds: *p*-xylene 4-H3CC6H4CH3 (pX; Reachim, Purum, Moscow, Russia); *p*-toluic aldehyde 4-H3CC6H4C(=O)H (pTAld) (Aldrich, 97%; Steinheim, Germany); *p*-toluic acid 4-H3CC6H4C(=O)OH (pTAc; Aldrich, 98%; Steinheim, Germany); 4-carboxybenzaldehyde 4-HO(O=)CC6H4C(=O)H (4-CBA; Aldrich, 97%; Steinheim, Germany) and terephthalic acid *p*-HO(O=)CC6H4C(=O)OH (TPA; Acros Organics, 99+%; Geel, Belgium).

For the synthesis of Mn/Co oxidation catalyst based on MCM-41/HNT composite the following reagents were used: manganese (II) acetate tetrahydrate (CH3COO)2Mn × 4H2O (Aldrich, ≥99%; Steinheim, Germany), cobaltous (II) acetate tetrahydrate CH3COO)2Co × 4H2O (Aldrich, ≥98%; Steinheim, Germany) and MCM-41/HNT composite, earlier prepared according to literature procedure [24].

Acetic acid CH3COOH (Chimmed, Chemical Grade; Moscow, Russia), dimethyl sulfoxide (DMSO) (H3C)2SO (Ruschim, Imp; Moscow, Russia) and potassium bromide KBr (Chimmed, Chemical Grade; Moscow, Russia) were used as solvent and additive respectively in the procedure of catalytic experiments.

Double-distilled water, methanol CH3OH (J.T. Baker, HPLC grade; Gliwice, Poland) and acetonitrile CH3CN (J.T. Baker, HPLC grade; Gliwice, Poland) were used as solvents and phosphoric acid H3PO4 (Component-Reactive, Chemical Grade; Moscow, Russia) was used as a pH regulating additive while conducting HPLC analysis.

#### *3.2. Analyses and Instrumentations*

Analysis by transmission electron microscopy (TEM) was carried out using LEO912 AB OMEGA (Carl Zeiss, Jena, Germany) and JEM-2100 JEOL microscopes (Jeol Ltd., Tokyo, Japan) with an electron tube voltage of 100 kV.

The isotherms of nitrogen adsorption/desorption were measured at 77 K on Micromeritics Gemini VII 2390 t instrument (Micromeritics Instrument Corp., Norcross, GA, USA). Before the measurements, the samples were degassed at 350 ◦C for 4 h. The specific surface area was calculated with the Brunauer–Emmett–Teller (BET) and Langmuir methods applied to the range of relative pressures P/P0 = 0.05–0.30. The pore volume and pore size distributions were determined from the adsorption branches of the isotherms based on the Barrett–Joyner–Halenda (BJH) model.

Weight content of manganese and cobalt in the catalyst was determined by means of atomic emission spectroscopy with inductively coupled plasma (ICP-AES) and X-ray fluorescent spectrometry (XFS). ICP-AES analysis was performed on the IRIS Intrepid II XPL instrument (Thermo Electron Corp., Waltham, MA, USA) in the radial observation modes at wavelengths of 257.6 and 259.4 nm for Mn and 228.6 and 237.9 nm for Co. XFS analysis was conducted on the ARL PERFORM'X spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with Rh anode and tube diameter of 3 mm in the etalon-free mode with the relative error of 5%.

X-ray photoelectron studies (XPS) were carried out on a Versa Probe II instrument (ULVAC-PHI Inc., Hagisono, Chigasaki, Kanagawa, Japan), equipped with a photo-electronic hemispherical analyzer with retarding potential OPX-150. X-ray radiation of the aluminum anode (Al Kα = 1486.6 eV) with a tube voltage of 12 kV, emission current of 20 mA and power of 50 W was used to excite photoelectrons. Photoelectron peaks were calibrated with respect to the carbon C 1s line with a binding energy of 284.8 eV.

Qualitative and quantitative analysis for the products of the *p*-xylene oxidation was carried out by means of high-performance liquid chromatography (HPLC) similar to the procedure, earlier described in literature [56] on the Agilent 1100 Series instrument (Agilent Technologies, Santa Clara, CA, USA) with a Zorbax SB-C18 column (5 μm, 2.1 mm × 150 mm) and UV detector. The eluent consisted of methanol (25%), acetonitrile (25%) and water (50%). For better peak resolution, phosphoric acid was added with concentration of 1 mL per 1 L of the eluent. The flow rate was 0.1 mL/min and the injection volume was 0.1 μL. Chromatograms were recorded at 210, 230, 254 and 280 nm simultaneously and analyzed on a computer using Agilent 1100 Software (Agilent Technologies, 2008, Santa Clara, CA, USA). Conversion of *p*-xylene and the yield of each oxidation product were calculated using calibrating curves with exponential approximation.

#### *3.3. The Synthesis of Mn*/*Co-Containing Oxidation Catalyst Based on the MCM-41*/*HNT Composite*

Deposition of Mn and Co on the MCM-41/HNT composite was performed according to the following procedure [16]. Of MCM-41/HNT composite 1000 mg and 20 mL of distilled water were placed into a 100 mL one-neck round-bottom flask, equipped with a reflux condenser and a magnetic stirrer. Then 20 mg of Mn(OAc)2 × 4H2O (0.08 mmol) and 200 mg of Co(OAc)2 × 4H2O (0.8 mmol) were placed into a chemical beaker and dissolved in 10 mL of distilled water at room temperature while stirring. Then the resulting solution of the pink-purple color was added dropwise to the suspension

of MCM-41/HNT composite in water at room temperature while stirring, and additional 10 mL of distilled water was passed through the dropping funnel to wash the residues of Mn(OAc)2 and Co(OAc)2 solution. The reaction was carried out at 60 ◦C for 12 h. After that the material obtained was centrifuged, washed twice with ethanol for the better removal of water and then dried in the air. The catalyst obtained was isolated as a pale-pink powder, weighing 1065 mg (92%).

ωMn (XFS): 0.15%. ωCo (XFS): 1.29%.

XPS (eV): 74.5 (Al2O3, Al 2p, 2.5%); 103.5 (SiO2, Si 2p, 25.9%); 119.5 (Al2O3, Al 2s); 154.5 (SiO2, Si 2s); 284.9 (–CH2CH2–, C 1s, 1.27%), 285.9 (–CH2CH2N, H3CC(=O)–, C 1s, 3.15%), (H3CC(=O)–, –CH2N<sup>+</sup>, C 1s, 0.82%), 290.1 (H3CC(=O)–, C 1s sat, 0.55%); 400.5 (–CH2CH2N, [NR4] <sup>+</sup>, N 1s, 0.6%); 530.6 (MnOx, CoOx, O 1s, 1.7%), 532.1 (CH3C(=O)O–, (Al2O3)x\*(SiO2)y, O 1s, 14.8%), 533.1 (SiO2, O 1s, 43.0%), 534.6 (O ... H2O, O 1s, 4.6%); 642.5 (MnO, Mn 2p3/2, 0.08%), 646.9 ([MnO4] bound, Mn 2p3/2, 0.02%), 654.0 (MnO, Mn 2p1/2), ([MnO4] bound, Mn 2p1/2); 686.5 (O–Si–F, F 1s, 0.2%); 780.0 (CoO, Co 2p3/2, 0.05%), 782.5 ([CoO6] bound, Co 2p3/2, 0.28%), 785.2 (CoO, [CoO6] bound, Co 2p3/<sup>2</sup> sat, 0.14%), 788.2 (CoO, [CoO6] bound, Co 2p3/<sup>2</sup> sat, 0.17%), 791.6 (Co 2p3/<sup>2</sup> sat, 0.06%), 794.6 (CoO, Co 2p1/2), 798.4 ([CoO6] bound, Co 2p1/2), 802.4 (CoO, [CoO6] bound, Co 2p1/<sup>2</sup> sat), 805.1 (CoO, [CoO6] bound, Co 2p1/<sup>2</sup> sat), 808.9 (Co 2p1/<sup>2</sup> sat, 0.06%).

#### *3.4. Protocol for the Catalytic Experiments*

Oxidation of *p*-xylene in the presence of Mn/Co MCM-41/HNT catalyst was carried out according to the literature procedures [16,29]. Here 150 mg of the catalyst, 4.5 mg of KBr, 0.5 mL of *p*-xylene and 5 mL of acetic acid were placed in a titanium autoclave, equipped with a magnetic stirrer. The autoclave was sealed, filled with oxygen up to a pressure of 2 MPa and placed in an oven with a thermostat control. Then the oxidation reaction was conducted at 150, 200 or 250 ◦C for 1, 3 or 5 h. After reaction, the autoclave was cooled down to room temperature and depressurized. The reaction products were additionally diluted by DMSO as the best solvent for terephthalic acid and, after the catalyst sedimentation and filtration, analyzed by the HPLC method.

The catalyst activity (TOF = turnover frequency) was calculated as the amount of reacted substrate (νsubstr) per mole of metal (ν*(*Mn <sup>+</sup> Co)) per unit of time with account of the yield of each product and number the oxygen atoms added, according to the formula:

$$\text{TOF}(\mathcal{O}\_2) = \frac{\nu\_{\text{substr}} \ast (\omega\_{p\text{TALd}} \ast 1 + \omega\_{\text{pTAc}} \ast 2 + \omega\_{4-\text{CBA}} \ast 3 + \omega\_{\text{TPA}} \ast 4 + \omega\_{\text{N}\_\text{I}\text{I}} \ast 3}{\nu\_{(\text{Mn}+\text{Co})} \ast t},$$

where ω is the yield of the certain product, expressed in the unit fractions, N/I is a not identified product (presumably 4-hydroxymethylbenzoic acid) and *t* is the minimal reaction time, for which the reaction progress is measured.

Each experiment at the same conditions was carried out two or three times, with the results differing by no more than 5% from the corresponding average value. These average values were presented in Tables 2–4. The measurement error did not exceed 5%.

#### **4. Conclusions**

Heterogeneous bimetallic Mn/Co-containing catalyst, based on MCM-41/halloysite composite, have been synthesized for the first time and tested in *p*-xylene oxidation under the conditions of the AMOCO process. It was demonstrated, that in the presence of bimetallic heterogeneous catalyst MnII1CoII10@MCM-41/HNT and KBr as a free radical source, the quantitative yield of terephthalic acid can be obtained in 3 h at temperature of above 200 ◦C, and oxygen pressure of 20 atm. The substrate to the Mn + Co ratio was 3.5–4 times higher than that for the traditional homogeneous

Mn(OAc)2/Co(OAc)2 system, hence proving a high efficacy and superiority of the heterogeneous catalyst, based on the hierarchical material MCM-41/HNT.

The influence of the oxygen pressure, temperature and KBr presence on the catalyst activity and product distribution was investigated. It was established that a decrease in oxygen pressure to 5 atm. resulted in the corresponding decrease of *p*-xylene conversion to 88–89%, with *p*-toluic acid obtained as the major reaction product with the yield up to 63% within 3 h. Decrease in temperature from 200 to 150 ◦C led to the abrupt downfall in the reaction rate. Herein the conversion did not exceed 40% after 5 h and *p*-toluic aldehyde was the major reaction product. Vice versa, rise in temperature from 200 to 250 ◦C did not result in the significant increase of the reaction turnover frequency. The presence of KBr was found to be crucial for the effective process of the oxidation of *p*-xylene, whose conversion did not exceed 2%, when KBr was removed from the reaction medium.

It was found that in the presence of MnII1CoII10@MCM-41/HNT the further oxidation of *p*-toluic acid to terephthalic acid mostly proceeded through the formation of 4-hydroxymethylbenzoic acid, thus eliminating the stage of undesirable 4-carboxybenzaldehyde. This pathway supposes the diminished role of the Co (II) in the oxidation of *p*-toluic acid and, therefore, allows the elevated substrate to catalyst ratios, but requires high oxygen pressures and Br· radicals as initiators.

Halloysite clay is a cheap and available in thousands tons aluminosilicate, which makes it a prospective nanomaterial for catalysts support. Despite the low resistant to metal leaching under the reaction and separation conditions the heterogeneous catalyst showed a phenomenally high activity in the oxidation of *p*-xylene to terephthalic acid under the conditions of AMOCO industrial process. This MnII1CoII10@MCM-41/HNT catalyst based on new hierarchical support with halloysite aluminosilicate nanotubes could be easily scaled up after stability improvement.

**Author Contributions:** Conceptualization, A.G., V.V. and A.M.; Methodology, A.G. and A.Z.; Software, A.Z. and E.I.; Validation, E.K. and A.M.; Formal analysis, E.I.; Investigation, A.Z. and A.G.; Resources, E.I.; Data curation, A.Z.; Writing—Original draft preparation, A.Z. and A.G.; Writing—Review and editing, A.M. and E.K.; Visualization, A.G. and A.Z.; Supervision, V.V.; Project administration, V.V. and A.M.; Funding acquisition, E.I. All authors have read and agreed to the published version of the manuscript.

**Funding:** The study was financially supported by the Ministry of Science and Higher Education of the Russian Federation, the unique project identifier is RFMEFI57717X0239.

**Acknowledgments:** We also thank National University of Science and Technology 'MISIS' (Moscow, Russia) for XPS facilities and Valentine Stytsenko (Gubkin University) and Yusuf Darrat (Louisiana Tech University, USA) for language editing.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Influence of Preparation Method on the Physicochemical Characteristics and Catalytic Activity of Co**/**TiO2 Catalysts**

#### **John Vakros**

Department of Chemistry, University of Patras, Rion, GR26504 Patras, Greece; vakros@chemistry.upatras.gr

Received: 23 December 2019; Accepted: 6 January 2020; Published: 7 January 2020

**Abstract:** Two Co/TiO2 catalysts with 7% CoO/g loading were prepared using equilibrium deposition filtration and the dry impregnation method. The two catalysts were characterized with various physicochemical techniques and tested for the degradation of sulfamethaxazole (SMX) using sodium persulfate (SPS) as the oxidant. It was found that the two catalysts exhibit different physicochemical characteristics. The equilibrium deposition filtration (EDF) catalyst had a higher dispersion of cobalt phase, more easily reduced Co(III) species, and a higher ratio of Co(III)/Co(II) species. The interactions between Co-deposited species and the titania surface were monitored with diffuse reflectance spectroscopy in all the preparation steps, and it was found that they increased during drying and calcination, while EDF favored the formation of surface species with strong interactions with the support. Finally, the EDF catalyst was more active for the degradation of sulfamethaxazole due to its better physicochemical characteristics.

**Keywords:** cobalt catalyst; titania; diffuse reflectance spectroscopy; sulfamethaxazole; persulfates; point of zero charge

#### **1. Introduction**

Cobalt-supported catalysts are quite promising for many reactions of environmental and industrial interest such as volatile organic compounds oxidation [1], CO2 hydrogenation [2], the Fischer–Tropsch process [3,4], and electrocatalysis [5].

In the last years, many studies dealt with the cobalt catalysts supported on titania [6–9]. In most of them, extensive physicochemical characterization was performed in order to combine the catalytic activity with the physicochemical characteristics. At this point, it should be stated that, generally and for Co catalysts specifically, the physicochemical characteristics and consequently the catalytic activity are strongly dependent on the preparation method and the precursor-supported species [10–14].

Thus, there are many studies dealing with the preparation of Co/TiO2 catalysts with different methods, although the most common method used for the preparation of supported Co catalysts is the dry impregnation (DI) method. This is because of its simplicity and low cost, as the other methods used are quite complicated and costly. Using DI, the speciation of Co-supported species is usually determined by the loading of the Co phase and the TiO2 properties such as the specific surface area, surface groups, and point-of-zero charge. The deposition mechanism is mainly the bulk precipitation during the drying step, while only a small portion of Co species adsorbs onto the titania surface. It is difficult to adjust the impregnation parameters during dry impregnation and consequently to control the mechanism of deposition of active species and the interactions of the supported phase.

One very attractive method for the preparation of catalysts is the equilibrium deposition filtration method (EDF) [15–17]. This method, compared to the other non-conventional preparation methods used, is simple and it does not need expensive reagents and infrastructure. Using this method, a quite large volume of dilute solution of the precursor compound is used. An amount of support is added to the solution. The suspension is allowed to reach equilibrium under well-defined and controlled impregnation parameters. During equilibrium, a quantity of the precursor ion is deposited onto the support surface. The ion deposition is finished when equilibrium is reached; then, filtration is applied for the collection of the solid and the removal of the non-deposited precursor species. The deposition of the active species can be controlled using proper conditions such as the pH, temperature, initial concentration, etc. There is a variety of different deposition modes such as adsorption, surface reaction, interfacial deposition, etc., depending on the impregnation parameters [11].

Compared to DI, EDF gives us the ability to adjust the impregnation parameters and the interactions of the supported phase with the support surface, and therefore to control the final speciation of the deposited species and the catalytic activity, as it was shown in many studies [1,11,15,16,18,19].

An extensive study about the preparation of Co/TiO2 catalysts using EDF was performed by Lycourghiotis et al. [20], where the interfacial chemistry of the impregnation step was fully monitored. In that study, it was found that the deposition mechanism involves the formation of mononuclear/oligonuclear inner sphere complexes upon deposition of the [Co(H2O)6] <sup>2</sup><sup>+</sup> ions at the "titania/electrolytic solution" interface. Depending on the Co(II) surface concentration, mononuclear complexes are formed, almost exclusively, at low and medium Co(II) surface concentrations. Binuclear and three-nuclear inner sphere complexes are formed, in addition to the aforementioned mononuclear ones, at relatively high Co(II) surface concentrations, while the deposition followed by H<sup>+</sup> released from the solid to bulk solution lowered the pH solution. Finally, it was found that impregnation pH was the most important parameter for the deposition of Co(II) species. Higher pH resulted in a higher amount of Co(II) deposited onto the titania surface; however, pH higher than 8 should be avoided, as Co(II) precipitates either as Co(OH)2 or Co(OH)NO3 in bulk solution. That study focused on the impregnation step and deposition mechanism of Co(II) onto the titania surface. Co/TiO2 samples using EDF have not been prepared and applied to catalytic tests, which is due to the low Co content.

The limited loading of Co onto the TiO2 surface is the only disadvantage of the EDF method. The Co loading is low and less than 3% CoO/g of the catalyst, even under the most favorable conditions. This is probably due to the limited solubility of Co2<sup>+</sup> ions in pH >8. In addition, the deposition mechanism for the Co(II) species onto the titania surface was exclusively through the formation of inner sphere complexes between monomeric or oligomeric Co species and surface–OH groups, and it did not involve surface precipitation. In the case of alumina, surface precipitation can be achieved in the interfacial region in lower pH values than the values required for bulk precipitation, which usually leads to higher deposited amounts of Co and more active catalysts [17,21].

The aim of the present work is the preparation of Co/TiO2 catalysts with higher loading with two different methods, EDF and DI, and the determination of how the preparation method alters the physicochemical characteristics and the catalytic activity for the degradation of sulfamethaxazole (SMX) with sodium persulfate (SPS) as the oxidant of the prepared catalysts.

In the present work, the EDF method was modified in order to increase the Co loading. Specifically, in the impregnation solution, the pH increased with the addition of a base to a value of 8.5, where extensive hydrolysis and polymerization between Co(II) ions and finally, the precipitation of Co(II) species occurred. The pH remained about 8.5 while Co(II) species precipitated. The amount of base added is about half of the amount required stoichiometrically for the complete precipitation of the Co(II) ions. Under these conditions, only a part of Co(II) precipitates and the rest of the Co(II) ions are dissolved in the suspension, with a higher degree of hydrolysis and polymerization. The pH is still 8.5, which is higher than the one used for the deposition of Co(II) ions in the previous study. This higher pH will be beneficial for the deposited loading of Co(II) species, probably resulting in higher deposited amounts of Co onto titania. In addition, under these conditions, the deposited Co(II) surface complexes onto the surface of titania are expected to have more than 3 Co atoms in contrast with the binuclear or three-nuclear inner sphere complexes formed at lower Co surface coverage.

#### **2. Results**

#### *2.1. Preparation of the Catalyst*

The solution of Co2<sup>+</sup> with a concentration of 0.03 M was placed in a double-walled beaker at a constant temperature of 25 ◦C. The pH was increased to about 8.5 using 2 M of NH3 solution. The amount of NH3 was half of the stoichiometric required for the complete precipitation of Co ions. As a result of the limited solubility of Co ions in high pH, the hydroxide (Co(OH)NO3 or Co(OH)2) was precipitated according to equations:

$$\rm{^2Co^{2+}(aq) + NO\_3^-(aq) + OH^-(aq) \leftrightharpoons Co(OH)NO\_3(s)}\tag{1}$$

$$2\text{ Co}^{2+}(\text{aq}) + 2\text{ OH}^-(\text{aq}) \leftrightharpoons \text{Co}(\text{OH})\_2(\text{s}).\tag{2}$$

The precipitation was followed by color changes from pink to blue–green. At that point, 1.0 g of TiO2 was added into the suspension. The surface of TiO2 was negatively charged in that pH and started to adsorb Co2<sup>+</sup>(aq) ions. The Co2<sup>+</sup> ions concentration decreased due to the adsorption process, and Equations (1) and (2) shifted to the left. The positive charge in the interfacial region and the surface of TiO2 increased due to the accumulation of Co2<sup>+</sup> ions. The surface reduced the increasingly positive charge by releasing H<sup>+</sup> to the bulk solution. This process resulted in a decrease in pH, and this also shifted Equations (1) and (2) to the left. As a result, the precipitated Co phase acted as a reservoir of Co2<sup>+</sup> ions to be deposited onto the TiO2 surface. This Co(II) deposition is usually pH dependable, and generally, it is higher in high pH values. After 5 days of impregnation, the pH is lower than 7.5, and the precipitate Co phase was dissolved. Then, the suspension was filtrated, and the Co/TiO2 was collected. The supernatant was pink, proving the dissolution of the precipitate Co phase, while the solid Co/TiO2 had a brown color. The adsorbed Co was measured spectophotometrically, and it was found to be 7% CoO/g of catalyst. On the other hand, the maximum Co content obtained using the standard EDF method was not exceeding 3% under optimal conditions. That was the result of the limited pH used (lower than 7). The preparation route can be seen in Figure 1.

**Figure 1.** (**a**) Impregnation solution of Co(H2O)6 <sup>2</sup>+, (**b**) after base addition and pH = 8.5, (**c**) after the addition of TiO2, (**d**) after deposition for 5 days, (**e**) image of equilibrium deposition filtration method (EDF) catalyst after filtration, and (**f**) image of dry impregnation (DI) catalyst after impregnation.

One other catalyst with the same Co loading was prepared using the DI method. Usually, DI is the most common method to prepare Co-supported catalysts, and the prepared catalysts exhibit high stability and activity.

After the preparation of the catalyst with dry impregnation, the two samples had significant differences, starting from the color of each sample. The DI sample was pink due to the octahedral Co(H2O)6 <sup>2</sup><sup>+</sup> complex, indicating that the Co phase is mainly the precipitate Co(H2O)6(NO3)2 phase with low interactions between Co and the surface, while the EDF sample had a brown color suggesting that the local environment of Co(II) was different than the Co(H2O)6 <sup>2</sup>+. The color change and the differences in the local environment of Co(II) species could be attributed to the partial substitution of water molecules ligands with either O–Ti or O–Co groups. Thus, strong interactions were present between Co ions and surface groups or among Co ions.

#### *2.2. Physicochemical Characterization with Di*ff*use Reflectance Spectroscopy (DRS) before Calcination Step*

For the elucidation of the interactions between the Co phase and TiO2 surface, the diffuse reflectance spectra of the two catalysts were collected after impregnation (wet samples, Figure 2a) and after drying (dry samples, Figure 2b). An inspection of Figure 2a clearly shows the significant differences between the two samples. This was expected, since the two samples had different colors after the impregnation step. Generally, there are three broad peaks in the spectrum. The first peak was centered at about 350 nm, the second was centered at about 500 nm, and a wide broad peak was centered at about 650 nm. In addition, the F(R), which was an analogue to the absorbance of the sample, was higher in the case of the EDF sample, especially in the first and third peaks. Taking into account that the two samples had the same Co loading, the differences in F(R) can be related to the dispersion of the Co phase. It provides evidence that the deposition of the Co phase resulted in higher dispersion in the case of the EDF sample. The first peak, which is located at about 350 nm, is due to the charge transfer between Ti–O–Co, and it can serve as a measurement of the interactions between the surface groups and the Co phase. This peak was absent in the DR spectrum of the Co(H2O)6 2+ (Co(H2O)6(NO3)2), pointing out that this was generated from the interactions between the surface –OH groups and Co ions.

**Figure 2.** Diffuse reflectance spectra of wet (**a**) and dried (**b**) samples.

Generally, this peak has been observed in many cases where a transition metal ion (either as oxyanion, MOx z–, or hydrated cation, M(H2O)x <sup>z</sup><sup>+</sup>) deposited on the TiO2 surface and was attributed to the charge transfer between Ti–O–M bonds [18–20,22,23]. The intensity of the peaks denotes that the interactions were higher in the case of the EDF sample, while the shift observed in lower energy supports this finding. The peak centered at about 500 nm was due to the Co–(OH2) bonds, and it can be seen in the Co(H2O)6(NO3)2 spectrum. It is characteristic of the octahedral symmetry of Co aqua complexes. The third peak was very interesting, because it features the partial substitution of the H2O molecules in the [Co–(OH2)6] <sup>2</sup><sup>+</sup> complex. The peak can be deconvoluted in two other peaks (inset of Figure 2a), which are centered at about 625 and 680 nm. The first peak is assigned to Co–O–Ti, while the second peak is due to Co–O–Co polymerization in accordance with previous works for Co deposition on an Al2O3 surface [21,24].

The same three broad peaks can be seen after drying in Figure 2b. The main difference was the intensity of these peaks. Due to the heat provided during drying, the peaks were more intense, indicating that the interactions were increased as a result of the release of water molecules. A more pronounced increment is evident in the EDF sample. This can be attributed to the better dispersion of the Co phase in the EDF sample. In the EDF case, the main mechanism of the Co phase deposition is the adsorption/surface reaction of Co ions with the –OH surface groups. During drying and water evaporation, the Co phase and the surface groups can interact further, and the corresponding peak is more intense. On the other hand, in the DI case, the main mechanism is the bulk precipitation of the Co phase onto the surface of TiO2 resulting in agglomerates with higher dimensions. This means that a significant amount of Co phase precipitates onto the previously deposited Co phase, and it is not in contact with the surface –OH groups. The majority of this part was not connected with the surface groups, and during drying, it could not interact with the TiO2 surface. As a result, the increment of interactions and consequently the intensity of the peak at 400 nm was not significant.

#### *2.3. Physicochemical Characterization of the Calcined Samples*

#### 2.3.1. Measurement of Specific Surface Area (SSA)

The deposition of 7% CoO on TiO2 did not significantly alter the SSA of the two samples. The SSA of TiO2 was found to be equal to 53 m2/g, while the EDF sample had an SSA equal to 50 m2/g, and the DI sample had an SSA equal to 45 m2/g.

The pore diameter distribution (Figure 3) showed small changes for the two samples. DI created a new group of mesoporous with a mean diameter of 28 nm. The EDF impregnation altered the porosity of TiO2. Specifically, the TiO2 pores shifted to a lower mean diameter, from 87 nm to 68 nm. This was probably a partial pore blocking due to the deposition of the Co phase onto the surface of TiO2, which also exhibited a significant increase of porosity, which was probably due to the generation of new pores after Co phase deposition.

**Figure 3.** Pore volume distribution of the calcined samples and pure TiO2.

#### 2.3.2. Diffuse Reflectance Spectra

The samples were calcined at 400 ◦C. The supported Co phase was transformed to Co3O4, as it can be seen in Figure 4. According to the literature, the Co in the Co3O4 oxide can be found either in octahedral symmetry (peaks at 425 nm and 710 nm) or in tetrahedral symmetry (the triplet centered at about 600 nm) [1]. As it can be seen in Figure 4, the spectra of the two samples were similar to that of Co3O4 oxide. The higher intensity found for the DI sample suggested an increased formation of Co3O4 oxides for the DI sample compared to the EDF sample. In addition, the ratio of the peaks at 710 nm and 450 nm was different between the two samples. The DI sample had a very close value to unsupported Co3O4, while the value for the EDF sample was higher. These were the results of the different deposition mechanisms and the different speciation of the Co phase before the calcination step. Taking into account that the peak at 710 nm is due to Co(III) ions, it can be claimed that the EDF samples exhibited higher amounts of the Co(III) oxide phase.

**Figure 4.** Diffuse reflectance (DR) spectra of the calcined samples.

#### 2.3.3. X-ray Diffraction, XRD

The XRD patterns of the two samples (Figure 5) exhibit only the characteristic XRD peaks of TiO2 P25. The peaks of the Co3O4 supported phase can hardly be observed. Only one corresponding to (311) can be seen at 37◦. This indicates that the Co3O4-supported particles are either highly dispersed on TiO2 or the particles are too small to be detected with XRD. Similar results can be found in the literature for catalysts with 10% loading [25–28].

**Figure 5.** XRD patterns of EDF and DI catalysts and TiO2 support.

#### 2.3.4. X-ray Photoelectron Spectroscopy, XPS

The X-ray photoelectron spectra for Ti 2p, O 1s, and Co 2p are presented in Figure 6. The two spectra had significant differences as a result of the different deposition mechanism. The region of Ti2p for the DI catalyst showed two symmetrical peaks, which were centered at 459.4 and 465.0 eV. The shape, the position, and the peak separation (5.6 eV) were very close to the Ti(IV) species in TiO2 nanoparticles. The spectrum of the EDF catalysts was more complicated. The two peaks were broader. In addition, each peak had a shoulder in lower binding energy (BE). The deconvolution of those peaks clearly showed the contribution of other Ti species in the peaks. These Ti species were located in lower values of BE: 457.0 eV and 462.3 eV respectively. These lower values in BE were probably due to the interactions with Co–O species adsorbed onto the TiO2 surface groups and the electron charge transfer to Ti atoms in accordance with the DRS results.

**Figure 6.** XPS spectra of Ti 2p (**a**), O 1s (**b**), and Co 2p (**c**) photoelectrons for the samples prepared.

The O 1s photoelectron spectrum was also different for the two samples. The DI sample showed one well-formed symmetrical peak centered at 531.4 eV and a shoulder at 533.4 eV. The first peak was characteristic of O in TiO2, while the second peak was assigned to the surface –OH groups. The O 1s spectrum for the EDF catalyst exhibited one additional new peak centered at 529.1 eV. The lower BE value indicated the increasing of electron cloud density of the lattice oxygen because of the deposition of Co species.

Figure 6c shows the Co 2p spectra of the two samples. There were two main peaks centered at 781.3 and 797.5 eV for DI and 781.5 and 796.4 for the EDF sample and two satellite peaks centered at 786.4 and 801.4 for DI and 788.4 and 807.7 for EDF. The first peak assigned to Co 2p3/<sup>2</sup> and the second peak was assigned to Co 2p1/2. It should be pointed out that it is rather difficult to determine the oxidation state of Co only by the peak position. Other parameters such as the shape of the satellites and the energy gap between the satellites and the main lines are used to discriminate between the different oxidation states of Co. Generally, the satellite peaks for Co(II) are located at BEs lower than 10 eV with respect to the main peak, while the satellite peaks for Co(III) are detected at BEs higher than 10 eV. For the two samples studied, the satellites peaks were located in higher BEs for the EDF sample, suggesting

that higher amounts of Co(III) ions were present on the surface of the EDF sample. Additionally, peaks exhibiting a shoulder are characteristic of Co(II) species in a high spin state, while the diamagnetic low-spin Co(III) ion does not show shake-up structures [25]. In this study, the shoulders were relatively lower in the case of the EDF catalyst, meaning that the Co(III) surface concentration was higher.

On the other hand, higher values of BE for the EDF sample denoted that the Co species interact better with the TiO2 surface. This suggests that the Co dispersion was higher in the EDF catalyst. Indeed, the atomic ratio Co/Ti for the two samples was found to be 0.11 and 0.20 for DI and EDF respectively, which was almost twice that of the EDF catalyst, proving the better dispersion of the Co phase for the EDF catalyst.

#### 2.3.5. Temperature Programmed Reduction (TPR) with H2

The TPR profiles of the DI and EDF catalysts are presented in Figure 7. The unsupported Co3O4 (inset of Figure 7) usually exhibits one intense peak centered at 454 ◦C and a shoulder at 369 ◦C, which is characteristic of the Co(III) species. These peaks were also present in the TPR profile of the DI sample, although the main peak of the DI sample was located at much higher temperature, at about 607 ◦C. This peak was due to Co(II) species interacting with the TiO2 surface. The TPR profile for the EDF sample presented three reduction peaks. The first one was located at 373 ◦C and denotes the Co(III) species of the EDF catalysts. The peak was well formed and of high intensity, suggesting that this catalyst had a significant higher amount of Co(III) species in accordance with the DRS and XPS results. The other two peaks were located at 450 ◦C and 553 ◦C, which were both very close to the temperatures of the peaks of the DI samples.

**Figure 7.** Temperature-programmed reduction (TPR) profiles for the samples studied, inset: the TPR profile for Co3O4.

#### 2.3.6. Acid Base Behavior of the Prepared Catalysts

The titration curves of TiO2 and two catalysts, EDF and DI, are presented in Figure 8. In this figure, there is also the titration curve of the blank solution. The section point of the curve of the blank solution with each of the suspension curves denotes the point of zero charge (pzc) of each solid according to the potentiometric mass titrations (PMT) method.

**Figure 8.** Potentiometric mass titrations for the prepared catalysts, TiO2 and the blank solution. Inset: the section points of titration curves of each suspension with blank solution, corresponding to the point of zero charge (pzc) of each solid.

As it can be seen from the inset of Figure 8, the pzc for TiO2 was equal to 5.9. The pzc values for the two catalysts were higher: 6.4 for the DI sample and 7.6 for the EDF sample. As it can be seen from the EDF curve, there was a significant consumption of H<sup>+</sup> at pH values lower than 6. This is evidence that the deposition of Co(II) species occurs through inner sphere complexes and there are significant interactions between the TiO2 surface and Co-supported phase. The Co(II) deposition process was not pH-independent, but the surface groups were involved. Thus, the acid–base behavior of the catalyst was changed, at least in an aqueous suspension.

The above can be seen more clearly in Figure 9, where the H<sup>+</sup> consumptions as a function of pH suspension were presented for the two catalysts and TiO2 and Figure 10, where the differential curves of the H<sup>+</sup> consumptions were presented, respectively.

**Figure 9.** H<sup>+</sup> consumption curves for the three samples as a function of pH suspension.

**Figure 10.** Differential curves of the H<sup>+</sup> consumption for each sample studied.

Indeed, for the EDF sample, a significant amount of H<sup>+</sup> was consumed at a pH lower than 6. This was not expected for the TiO2, as it is known from the literature that the surface –OH groups do not appear in this pH [27]. Under realistic conditions, where the effect of a double layer is significant, there are two peaks for the TiO2 centered at about 10 and lower than 3.6, respectively [28]. The peak at a pH of about 10 can be seen in the differential curve for TiO2 in Figure 10. For the EDF sample, there is a new peak at about 5.5, which can be attributed either to the cobalt phase or to the shift in higher pH of the lower peak of the TiO2 surface groups. This peak was also present in the DI sample, although it was less intense.

#### *2.4. Catalytic Activity*

The prepared catalysts were tested for the degradation of SMX using SPS as the oxidant. This process is part of the advanced oxidation processes (AOPs), which are based on the oxidative power of specific chemicals. In this case, the catalyst activates the SPS to form SO4 <sup>−</sup> radicals. These radicals exhibit high redox potential; they are more selective for oxidation by electron transfer reaction and exhibit higher activity in carbonates solution. The SO4 <sup>−</sup> radicals could be generated by initiating SPS with a transition metal ion. The results showed that Co ions are the most efficient metal ions for the activation of KHSO5, and Co3O4 has been used for the degradation of various phenols and dyes [29].

The degradation of SMX with time is presented in Figure 11. Two different initial concentrations of SMX, 150 and 250 μg/L, were used. At the beginning of the process, the catalyst was immersed into the SMX solution. After a period of 15 min, 1000 mg/L SPS was added into the solution, and the monitoring of degradation was started (time = 0). The concentration of SPS was chosen to be 1000 mg/L, because higher concentrations had no significant positive effect on the degradation of SMX, while lower concentrations of SPS resulted in lower degradation. Although increased oxidant concentrations will expectedly generate more radicals, these radicals in excess may suffer from partial scavenging and be converted to less reactive species [30]. In addition, higher concentrations of SPS will increase the sulfate ions on the water matrix, which is not desirable.

**Figure 11.** Degradation of sulfamethaxazole (SMX) using 1000 mg/L sodium persulfate (SPS) as oxidant and 100 mg/L catalyst with time. The concentration of SMX is presented.

As it can be seen from Figure 11, the degradation was higher for the EDF catalyst. The influence of the initial concentration was almost negligible. This is evidence that the production of radicals and the amount of adsorbed SMX were not the limiting steps under these conditions. If the rate-determining step was the production of free radicals, then the degradation would have been higher for the low concentration of SMX.

In addition, if the adsorption step was the one that determines the rate, then higher concentrations will be in favor for the degradation. In this study, the adsorption of SMX onto the catalyst surface is low, and the process is not under mass transfer phenomena. The adsorption of SMX was higher in the case of EDF, although it was still low, and almost zero for the DI catalyst. The low adsorption of SMX on the catalyst surface is rather expectable. The pzc of the catalysts were found to be 6.4 for DI and 7.6 for EDF, while the degradation process occurred at ambient pH (around 6). Generally, SMX exists as a cation only in very low pH (<1.4), while at pH >6, SMX is negatively charged, so SMX can be considered as neutral at 1.4 < pH < 5.8 [31]. Under the pH of the degradation process, SMX and DI are almost neutral, while EDF has a slightly positive charge due to the surface sites with pH close to 5.5, making the adsorption of SMX easier; this also explains the small difference in the adsorption between the two catalysts.

The degradation data can be represented as a first-order reaction. The general kinetic model for a first order reaction rate is

$$\text{Rate} = -\text{dC}/\text{dt} = \text{k C} \tag{3}$$

where C is the concentration of SMX, and k is the rate constant.

The value of k can be calculated from the slope of the integrated form of Equation (3):

$$
\ln \mathbf{C} = \ln \mathbf{C} \mathbf{o} - \mathbf{k} \text{ t.}\tag{4}
$$

The k value for the EDF samples was found to be equal to (4.0 ± 0.1) × 10−<sup>3</sup> min−<sup>1</sup> and for the DI catalyst (3.0 ± 0.1) × 10−<sup>3</sup> min<sup>−</sup>1, indicating the higher activity of the EDF catalyst, and the linear regression of the data is presented in Figure 12.

**Figure 12.** Linear dependence of the logarithm of SMX concentration ratio for the catalysts studied. The concentration of SMX is presented.

Higher activity of the EDF catalyst was not expected, since the EDF catalyst exhibited higher amounts of Co(III) species. According to the proposed mechanism for the activation of persulfates with Co3O4, the Co(II) species are the most active [32,33]. They can activate S2O8 <sup>2</sup><sup>−</sup> ions to SO4 <sup>−</sup> radicals by the oxidation of Co(II) to Co(III), although after the reaction cycle, Co(II) regenerates. It has been shown that the surface speciation of Co3O4 after five reaction cycles for the Orange G degradation does not alter significantly [33]. The surface –OH are very important in the catalytic performance. According to the literature, the surface –OH groups are crucial for the catalytic performance [32]. They can help the transformation of Co(III) to Co(II). The enhanced activity of the EDF catalysts can be explained from the almost doubled surface concentration of Co species, as it was determined by XPS, and by the higher reducibility of the Co(III) species, as it was found by TPR experiments. Finally, the EDF catalyst exhibits higher interactions between the titania surface –OH groups and Co species.

#### **3. Materials and Methods**

#### *3.1. Modification of the Typical EDF Procedure for The Preparation of Co(II)*/*TiO2 Catalysts*

The titania used in this work was the P25 aeroxide from Degussa (Rellinghauser Strasse 1-11, Essen 45128, Germany), and all the other chemicals used in this study were analytical grade from Merck (Merck KGaA, Headquarters of the Merck Group Frankfurter Strasse 250, Darmstadt 64293, Germany) The modified EDF procedure followed for the deposition of Co2<sup>+</sup> ions on the titania involved the preparation of an aqueous solution of 0.03 M Co(NO3)2, the addition of 2 M NH4OH solution for regulating the suspension pH of about 8.5, the impregnation by this solution of 1.0 g of TiO2 at 25 ◦C, and the equilibrium of the suspension for 5 days, under nitrogen atmosphere. This helped to complete the deposition of Co(II) species and the pH to reach 7.5. The deposition of Co(II) was followed by filtration, drying at 110 ◦C for 2 h, and calcination in air at 400 ◦C for 3 h. The Co(II) concentration before and after deposition was determined in the impregnation solution using the Nitroso R-salz procedure [34]. From the difference in the Co(II) concentration before and after deposition, the Co loading was determined. The Co loading was also confirmed with desorption of the Co from the uncalcined sample and with atomic adsorption spectroscopy for a dissolved part of the final catalyst.

#### *3.2. Physicochemical Characterization*

The diffuse reflectance spectra (DRS) of the samples studied were recorded in the range 200–800 nm at room temperature, using a UV-vis spectrophotometer (Varian Cary 3, Agilent 5301 Stevens Creek Blvd, Santa Clara, CA 95051, USA) equipped with an integration sphere. Titania was used as a reference in all the cases. The powder samples were mounted in a quartz cell, which provided a sample thickness greater than 3 mm and thus guaranteed "infinite" sample thickness. Nitrogen adsorption isotherms at liquid N2 temperature (Tristar 3000 porosimeter Micromeritics Instrument Corp. 4356 Communications Drive, Norcross, GA 30093-2901, USA) were used for the determination of specific surface area (SSA). X-ray diffraction (XRD) patterns were recorded in a Bruker D8 (Billerica, MA, USA) Advance diffractometer equipped with a nickel-filtered CuKa (1.5418 Å) radiation source. The XPS analysis of the oxidic specimens was performed at room temperature in a UHV chamber (base pressure 8 × 10−<sup>10</sup> mbar), which consists of fast specimen entry assembly, preparation, and an analysis chamber with residual pressure below 10−<sup>8</sup> mbar, equipped with a hemispherical electron energy analyzer (SPECS, LH10) and a twin-anode X-ray gun for XPS. The unmonochromatized Mg *K*α line at 1253.6 eV and a constant pass energy mode for the analyzer were used in the experiments. The temperature-programmed reduction (TPR) experiments were performed in laboratory-constructed equipment. An amount of sample, 0.1 g, was placed in a quartz reactor, and the reducing gas mixture (H2/Ar: 5/<sup>95</sup> *<sup>v</sup>*/*v*) was passed through it for 2 h with a flow rate of 40 mL·min−<sup>1</sup> at room temperature. Then, the temperature was increased to 1000 ◦C with a constant rate of 10 ◦C·min<sup>−</sup>1, and the reducing gas mixture was monitoring with a thermal conductivity detector (TCD). The reducing gas mixture was dried in a cold trap (−95 ◦C) before reaching the TCD. To study the acid–base behavior of the Co/TiO2-prepared catalysts, potentiometric mass titrations were applied [35]. According to PMT, the point of zero charge value of the catalyst is the common intersection point of the titration curves of suspensions with different amounts of solid and the titration curve of a blank solution. The latter is a solution that contains exactly the same amounts of inert electrolyte and base solution without solid. Applying the mass balance equation for the H<sup>+</sup> ions for each titration curve [36], the H<sup>+</sup> consumption on the biochar surface was determined. Using the differential curves of the H<sup>+</sup> consumption curves, valuable conclusions about the surface sites and the interfacial region can be made according to the literature [28].

#### *3.3. Catalytic Activity*

The prepared catalysts were tested for the degradation activity of sulfamethaxazole (SMX), which is an antibiotic using sodium persulfate (SPS) as the oxidant. Experiments were conducted in a cylindrical glass reaction vessel of 200 mL capacity, which was open to the atmosphere. An SMX solution was prepared (150 or 250 μg/L) in ultrapure water (UPW), and the appropriate amounts of SPS (1000 mg/L) and the Co/TiO2 catalyst (100 mg/L) were added to start the reaction under magnetic stirring and ambient temperature. Samples were periodically withdrawn from the vessel and analyzed by high-performance liquid chromatography (HPLC) using an Alliance HPLC system equipped with a photodiode array detector (Waters 2996), More details about the catalytic experiments and SMX analysis can be found in [37].

#### **4. Conclusions**

Two Co/TiO2 catalysts, with 7% CoO, were prepared using EDF and DI methods, which were characterized with various physicochemical techniques and tested in the SMX degradation. The EDF method was modified to increase the Co amount and achieved more than double Co loading. It was found that the Co loading was not the determinant factor for the activity of the catalysts. Furthermore, the production of radicals and the amount of adsorbed SMX were not the limiting steps of the reaction. The SMX degradation was found to be surface structure-sensitive, increasing with the Co surface concentration and the reducibility of the Co phase. The interactions between the Co phase and titania surface groups are important for the catalytic performance. During EDF, the Co species interact strongly with the titania surface groups, resulting in higher dispersion, a higher Co(III)/Co(II) ratio, more easily reducible Co species, and a different speciation of O surface groups, and therefore different acid–base behavior, pointing out the significance of the preparation method.

**Acknowledgments:** E. Siokou is gratefully acknowledged for the XPS spectra. P. Nanou is gratefully acknowledged for helping with the catalytic activity measurements.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Determining the Location of Co2**<sup>+</sup> **in Zeolites by UV-Vis Di**ff**use Reflection Spectroscopy: A Critical View**

#### **Andrea Bellmann 1,2, Christine Rautenberg 1, Ursula Bentrup 1,\* and Angelika Brückner 1,\***


Received: 10 December 2019; Accepted: 10 January 2020; Published: 15 January 2020

**Abstract:** UV–Vis spectroscopy as well as in situ FTIR spectroscopy of pyridine and CO adsorption were applied to determine the nature of Co species in microporous, mesoporous, and mixed oxide materials like Co–ZSM-5, Co/Na–ZSM-5, Co/Al–SBA-15, and Co/Al2O3–SiO2. Because all sample types show comparable UV–Vis spectra with a characteristic band triplet, the former described UV–Vis band deconvolution method for determination and quantification of individual cationic sites in the zeolite appears doubtful. This is also confirmed by results of pyridine and CO adsorption revealing that all Co–zeolite samples contain two types of Co2<sup>+</sup> species located at exchange positions as well as in oxide-like clusters independent of the Co content, while in Co/Al–SBA-15 and Co/Al2O3–SiO2 only Co2<sup>+</sup> species in oxide-like clusters occur. Consequently, the measured UV–Vis spectra represent not exclusively isolated Co2<sup>+</sup> species, and the characteristic triplet band is not only related to γ-, β-, and α-type Co2<sup>+</sup> sites in the zeolite but also to those dispersed on the surface of different oxide supports. The study demonstrates that for proper characterization of the formed Co species, the use of complementary methods is required.

**Keywords:** Co–ZSM-5; UV–Vis diffuse reflection spectroscopy; FTIR spectroscopy; pyridine adsorption; CO adsorption

#### **1. Introduction**

Cobalt zeolites are important catalysts for the selective catalytic reduction (SCR) of nitrogen oxides by hydrocarbons, especially by methane [1–14]. It is widely accepted that depending on the used zeolite, its Si/Al ratio, Co content, and preparation method, different Co species are formed comprising: (i) Co2<sup>+</sup> cations located in the zeolite channels, (ii) CoOx microaggregates, and (iii) CoO, Co3O4, as well as Co silicate mainly in the case of over-exchanged zeolites. The specific role of these species in the SCR reaction is contrarily discussed.

As Co ions are most stable in the form of Co2<sup>+</sup>, two [AlO4] <sup>−</sup> sites are needed to stabilize the charge. Thus, only Al–O–(Si–O)2–Al units occurring in close proximity in the zeolite rings are able to charge balance divalent cations, while single Al atoms can only balance monovalent ions [15]. Therefore, the concentration and location of incorporated Co species are controlled by the distribution of Al atoms in the zeolite framework. Reversely, the zeolite exchange capacity for Co2<sup>+</sup> can be used to probe the Al distribution in the zeolite framework [16]. Assuming a fully exchanged zeolite, the quantification of the rings containing two Al atoms in the same channel is possible. The preparation conditions to guarantee a complete exchange were described in this reference. Thus, in the group of Wichterlová [17–20], a procedure was developed enabling the discrimination between those sites using

UV–Vis diffuse reflection spectroscopy (UV–Vis–DRS). It was assumed that the d–d transition bands of bare Co2<sup>+</sup> ions in the visible region of the respective dehydrated zeolites were characteristic for the individual cationic sites. Thus, the observed bands around 21,000 cm−<sup>1</sup> (475 nm), 17,250 cm−<sup>1</sup> (580 nm), and 16,000 cm−<sup>1</sup> (625 nm) were assigned to the γ-, β-, and α-type sites, respectively. By analyzing the deconvoluted UV–Vis–DR spectra, the percentages of the respective sites occupied by Co2<sup>+</sup> were determined. This approach was applied for Co–mordenite [17], Co–ferrierite [18], Co–ZSM-5 [19], and Co–beta [20]. The fundamental requirement for using this method is that all Co, represented by these UV–Vis bands, is present in form of isolated Co2<sup>+</sup> and that these sites are created by wet ion exchange at a pH of 5.5 according a selected procedure [15,16].

The work of the Wichterlová group is highly recognized in the scientific community, and the developed approach to probe the Al distribution in the zeolite framework was also applied by other groups [21,22]. However, there are some doubts related to the nature of the Co sites created by ionic exchange and the assignment of the observed band triplet in the UV–Vis–DR spectra. Thus, Campa et al. [10] and Gil et al. [23,24] demonstrated that the formation of Co oxidic clusters is possible even if the formal Co/Al ratio is smaller than 0.5. This is due to the hydrolysis of the [Co(H2O)6] <sup>2</sup><sup>+</sup> and [Co(OH)]<sup>+</sup> complexes, which proceeds easily. Hence, intermediates like [Co–OH–Co]3<sup>+</sup> (by olation) and [Co–O–Co]2<sup>+</sup> (by oxolation) can be formed which are subsequently converted into oxidic clusters during the calcination process.

On the other hand, in papers of the Busca group, [11,25,26] besides comprehensive spectroscopic characterization of Co-exchanged zeolites, UV–Vis–DR spectra of Co/SiO2–Al2O3 samples, having an open pore structure, are also reported for comparison. These spectra were found to be very similar to those measured on Co–MFI and Co–Fer zeolites showing a triplet of band in the region 22,000–14,000 cm−<sup>1</sup> too. Furthermore, as deduced from additional infrared spectroscopic studies, the presence of Co species located at the external surface of the zeolite has to be considered besides those located at exchange positions within the zeolite framework. As a consequence, both internal as well as external Co2<sup>+</sup> species contribute to the UV–Vis spectrum and cannot be distinguished from each other. Therefore, it was stated that the assignment of the observed bands, being very similar for Co-exchanged MOR, FER, and MFI zeolites, to three different types of sites (α, β, γ) located in the internal cavities only is quite unreliable [26].

Inspired by these controversial statements concerning the use of UV–Vis–DRS for determining nature and distribution of Co2<sup>+</sup> species in Co-exchanged zeolites and based on results of own investigations of Co–ZSM-5 catalysts used in the CH4–SCR of NOx, we present here a comprehensive characterization study enabling a deeper insight concerning the nature of Co species created by wet exchange procedure. Therefore, different samples were selected comprising Co–ZSM-5, Co/Na–ZSM-5, and Co/Al–SBA-15 prepared by ion exchange, as well as Co/Al2O3, SiO2 prepared by incipient wetness impregnation. All samples were characterized by UV–Vis–DRS and FTIR (CO and pyridine adsorption). The aim was to demonstrate the requirement of applying complementary methods for proper characterization the formed Co species. The exchange degree reflected by the availability of acid sites was examined by pyridine adsorption, while the nature of the formed Co species (isolated and/or agglomerated) was characterized by CO adsorption at low temperature and UV–Vis–DR spectroscopy.

#### **2. Results and Discussion**

#### *2.1. Catalysts*

The catalysts together with their metal contents and the preparation methods applied in this work are listed in Table 1. Except for the 1.96 Co sample, which was prepared according to the procedure described by Dˇedeˇcek et al. [19], all zeolite-based catalysts were prepared by an own method adapting the procedure from Beznis et al. [27] in which the used Co salts, the applied Co concentrations in the solutions, and temperatures for ion exchange were varied, while NH4–ZSM-5 and Na–ZSM-5 were used as starting materials for ion exchange. Because the Si/Al ratio in the starting zeolite material was constant, the exchange degree is mainly influenced by the applied Co salt concentration. When using Na–ZSM-5 as starting material, the Na ions were not completely exchanged. For comparison, two catalysts were also prepared utilizing materials that have no microporous zeolite structure. On the one hand a mesoporous Al–SBA-15 was used for ion exchange, and on the other hand a mixed oxide SiO2/Al2O3 with open pore structure was loaded with Co by incipient wetness impregnation.


**Table 1.** Overview of catalysts studied, applied preparation methods, and metal contents.

<sup>1</sup> Preparation method according Dˇedeˇcek et al. [19].

#### *2.2. UV–Vis DRS Studies*

The UV–Vis–DR spectra of the catalysts measured at room temperature after dehydration in Heat 400 ◦C are depicted in Figure 1. For all samples, similar spectra were obtained showing a characteristic triplet of bands with maxima around 495 nm, 585 nm, and 645 nm as earlier described in literature for Co–ZSM-5 samples [19]. The intensity ratios of these band maxima vary, as can be seen in detail by inspecting the deconvoluted spectra (Figure S1), but a clear trend in terms of Co concentration and/or used materials is not visible. Interestingly, comparable spectra with a characteristic band triplet were also obtained from the samples 1.79 Co–SiAl and 0.8 Co–AlSBA, which have no zeolite structures. This suggests, in agreement with conclusion made by the Busca group, [11,26] that the assignment of the observed band triplet to three different types of sites (α, β, γ) located in the internal cavities of pentasil-containing zeolites [17–19] might not be correct, particularly since such a characteristic band triplet is not restricted to zeolite materials and was also described, e.g., for cobalt-based blue pigments, [28] cobalt spinels, [29,30] cobalt oxide-apatite materials, [31] Co/Al2O3, [32], and Co-doped ZnO [33].

**Figure 1.** UV–Vis spectra measured in diffuse reflection mode at room temperature after heating the samples at 400 ◦C in He for 30 min.

Octahedral and tetrahedral Co2<sup>+</sup> (d7) complexes, particularly in zeolite materials, have been widely studied by optical spectroscopy in the past [34–39]. It is known that Co2<sup>+</sup> changes from high-spin octahedral state in hydrated samples to tetrahedral coordination in the respective dehydrated ones, visible by the color change from pink to blue. Tetrahedral Co2<sup>+</sup> (d7) exhibits three symmetry and spin allowed transitions from the ground state: ν<sup>1</sup> = 4A2(F) → 4T2(F), ν<sup>2</sup> = 4A2(F) → 4T1(F), ν<sup>3</sup> = 4A2(F) → 4T1(P) [36,39]. While ν<sup>1</sup> is observed in the infrared region between 2500 and 6000 cm<sup>−</sup>1, the <sup>ν</sup><sup>2</sup> and <sup>ν</sup><sup>3</sup> transitions are observable in the near infrared and visible region. For the <sup>ν</sup><sup>2</sup> and ν<sup>3</sup> transitions, a band splitting is observed, which may be due to (i) a low symmetry perturbation, which lifts the degeneracy of the two 4T1 excited levels and which induces an asymmetrical band splitting; (ii) a dynamic Jahn–Teller effect, which splits the two bands into three symmetrically spaced peaks; (iii) spin-orbit coupling [38,39]. Taking these facts into account, it is not surprising that the spectra of all dehydrated Co2<sup>+</sup>-containing samples, not only those of zeolites, exhibit a band triplet in the visible region between 400–750 nm.

The observed different shapes of the band triplets might by related to the specific coordination geometry of Co2<sup>+</sup> influenced by the individual surrounding of the Co2<sup>+</sup> ions in the respective solids. Thus, it can be expected that the splitting pattern in the case of Co-containing zeolites changes depending on the occupied positions (γ-, β-, or α-type sites), e.g., in pentasil-containing zeolites [17–20]. Therefore, one can consider that the splitting pattern provides information concerning the successive occupation of respective sites with increasing Co content as described by Dˇedeˇcek et al. [19], e.g., for Co–ZSM-5 samples. Indications for that can also be seen by comparing the spectra of 0.32 Co and 2.44 Co in Figure 1. However, when Co2<sup>+</sup> ions are evenly distributed on γ-, β-, and α-type sites, then each type of Co species gives rise to an own triplet band, consequently leading to three superimposed triplet bands, which contribute to the observed spectrum. Therefore, it is practically impossible to discriminate between Co species at different sites and to quantify the percentage of occupied sites by deconvolution of the observed spectra. Nevertheless, based on spectral deconvolution, Dˇedeˇcek et al. [19] described different components which were ascribed to three types of Co ions located at specific cationic sites in Co–ZSM-5: A single band at 15,000 cm−<sup>1</sup> for α-sites; four bands at 16,000, 17,150, 18,600, 21,200 cm−<sup>1</sup> for β-sites; and a doublet at 20,100, 22,000 cm−<sup>1</sup> for γ-sites. This might be mathematically correct, but not from the physical point of view. Moreover, as we have shown, the triplet bands can also be properly deconvoluted on the basis of only three components (cf. Figure S1).

It has to be mentioned here that for determining the location of the Co2<sup>+</sup> ions, it was presupposed [15,19] that all Co2<sup>+</sup> ions are isolated due to the exchange of protons (Brønsted sites), which are created by the incorporation of Al in the zeolite matrix. Hence, only the Co/Al ratio was taken into account. We demonstrate in the next paragraph that this supposition cannot be made without inclusion of additional characterization methods.

#### *2.3. FTIR Studies: Pyridine and CO Adsorption*

Brønsted and Lewis acidic sites can be well characterized by FTIR spectroscopic investigation of pyridine adsorption [40,41]. The typical bands resulting from pyridine adsorbed on Brønsted acid sites (PyH+) are observed around 1540 cm<sup>−</sup>1, while the bands from pyridine interacting with Lewis acid sites (L–Py) are detected at around 1450 cm−<sup>1</sup> and in the region of 1625–1595 cm<sup>−</sup>1. From the band positions in the latter region, the strength of Lewis acid sites can be evaluated [41]. The pyridine adsorbate spectra of the studied catalysts measured at 150 ◦C are depicted in Figure 2. As expected, the Co-free H–ZSM-5 sample shows an intense PyH<sup>+</sup> band at 1545 cm<sup>−</sup>1, the intensity of which decreases with Co loading. Upon comparing the band intensities of the PyH<sup>+</sup> band, it is obvious that even the highly loaded sample 3.19 Co possesses a quite high number of Brønsted acid sites, although the Co/Al ratio is very close to 0.5. This means that a part of the Co2<sup>+</sup> ions is obviously not located on exchange positions, which points to an exchange process with a more complex stoichiometry as also observed by other authors [10].

**Figure 2.** Pyridine adsorbate spectra measured at 150 ◦C.

By comparing the intensities of the L–Py band of H–ZSM-5 with that of the Co-containing samples, it is seen that additional Lewis acid sites are created with increasing Co content as indicated by the bands at 1450/1610 cm−<sup>1</sup> (Co2<sup>+</sup>) and 1443/1596 cm−<sup>1</sup> (Na+). Furthermore, a splitting of the L-Py band in the region above 1600 cm−<sup>1</sup> into two components at 1610 and 1616 cm−<sup>1</sup> is observable, revealing two Co2<sup>+</sup> Lewis sites of different strength, possibly located at different sites in the zeolite lattice.

A suitable method for characterizing the nature of Co species is adsorption of CO at low temperatures [12,13,42–44]. Thus, one can distinguish Co2<sup>+</sup> located at exchange positions from Co2<sup>+</sup> being part of oxide-like clusters because the respective νCo2+−CO bands appear at different positions, namely at 2204 cm−<sup>1</sup> for Co2<sup>+</sup> in exchange positions and at 2194 cm−<sup>1</sup> for Co2<sup>+</sup> in oxide-like clusters. The CO adsorbate spectra of the studied samples are compared in Figure 3. Here, the spectra measured at −60 ◦C are shown because the band stemming from the interaction of CO with OH groups at 2174 cm−<sup>1</sup> [45] has negligible intensity at this temperature. This band is intense at the adsorption

temperature of −120 ◦C but vanishes at higher temperatures because the interaction of CO with hydroxyl groups is not as strong as the interaction with Co2<sup>+</sup> ions [14]. It is obvious that all Co–zeolite samples (Figure 3a) contain Co2<sup>+</sup> species at exchange positions (2204 cm<sup>−</sup>1) as well as in oxide-like clusters (2195 cm<sup>−</sup>1) independent from the Co content. This is in accordance with the results of pyridine adsorption, where two Co2<sup>+</sup> Lewis sites of different strength have been identified. The bands at 2177 and 2112 cm−<sup>1</sup> in the 0.35 Co/1.98 Na sample are related to νNa+−CO and νNa+−OC vibrations, respectively [46].

**Figure 3.** CO adsorbate spectra of Co-ZSM-5 samples (**a**) and 1.79 Co-SiAl as well as 0.8 Co-AlSBA (**b**) measured at −60 ◦C.

In the CO adsorbate spectra of 1.79 Co–SiAl and 0.8 Co–AlSBA (Figure 3b), only Co2<sup>+</sup> in oxide-like clusters is observed as reflected by the band position. With respect to the UV–Vis spectra, which are similar for both types of samples (cf. Figure 1), the results of CO adsorption lead to the conclusion that the triplet bands observed in the spectra are related to tetrahedral Co2<sup>+</sup> sitting at exchange positions as well as in oxide-like clusters, which can be located inside as well as outside the zeolite channels, depending on the cluster size. The latter conclusion is based on CO adsorption experiments in which o-toluonitrile was preadsorbed on Co–H–MFI [12,13]. In this way, the discrimination between cationic sites inside and outside the channels was possible because the bulky molecule is not able to penetrate the channels of zeolites like ZSM-5 or MOR and therefore block respective sites. Thus, it could be demonstrated for Co–H–MFI that Co2<sup>+</sup> ions are distributed at both the external and internal surfaces. This has to be assumed too for the Co–ZSM-5 catalysts studied in this work.

To be sure that the findings of pyridine and CO adsorption are not influenced by the used preparation methods, being different from that described by Dˇedeˇcek et al. [19] we have also prepared a sample according to this protocol. As a result, the sample 1.96 Co was obtained (cf. Table 1), the characterization results of which in comparison with those of 2.44 Co are shown in Figure 4.

**Figure 4.** Comparison of UV–Vis–DR spectra measured at room temperature after heating the samples at 400 ◦C in He for 30 min (**a**), pyridine adsorbate spectra measured at 150 ◦C (**b**), and CO adsorbate spectra measured at −50 ◦C (**c**) of the samples 1.96 Co and 2.44 Co.

It is seen that the UV–Vis–DR as well as the pyridine and CO adsorbate spectra are very similar. While the UV–Vis spectra are nearly identical (Figure 4a), the Brønsted acidity of the 1.96 Co sample is slightly higher compared to that of 2.44 Co (Figure 4b), which might be due to the lower Co content and therefore lower exchange level. This is also true for the intensities of the observed CO adsorbate bands, whereas the intensity ratio of the bands at 2204/2195 cm−<sup>1</sup> is nearly the same. In summary, it can be stated that in both catalysts, independent of the applied exchange procedure, Co2<sup>+</sup> species are located both at exchange positions and in oxide-like clusters. Hence, as already mentioned above, the respective UV–Vis spectra reflect not only Co2<sup>+</sup> ions located at α, β, γ sites in the internal cavities of the zeolite but also those in oxide-like clusters inside and/or outside the zeolite channels.

The existence of oxide-like clusters implies the possible presence of Co3<sup>+</sup> besides Co2<sup>+</sup> ions. Considering the fact that the samples were calcined in synthetic air at 500 ◦C, the presence of Co3<sup>+</sup> has to be taken into account, which was also proven by NO adsorption in a former study [14]. For investigating the possible influence of calcination conditions on the nature of formed Co species, we analyzed sample 1.96 Co sample by respective FTIR experiments with CO and exemplarily the 2.44 Co catalyst by UV–Vis–DRS.

In Figure 5a, the normalized CO adsorbate spectra of 1.96 Co are shown measured at −60 ◦C after pretreatment of the fresh, non-calcined sample in situ in the FTIR reaction cell by heating in vacuum for 1 h at 500 ◦C (vac.) and heating in synthetic air at 500 ◦C for 45 min (air). For comparison, the spectrum of the calcined (calc.) sample is also shown. The spectrum of the latter is similar to that of the vacuum-treated sample exhibiting the typical bands of CO adsorbed on Co2<sup>+</sup> at exchange positions as well as Co2<sup>+</sup> in oxide-like clusters (vide supra). In the spectrum of the air-treated sample, only Co2<sup>+</sup> in oxide-like clusters is detectable. The high intensity of the νOH−CO band around 2172 cm−<sup>1</sup> in this sample in relation to the respective ν Co2<sup>+</sup>−CO bands suggests an agglomeration of the Co2<sup>+</sup> species that recreates Brønsted sites, i.e., not exchanged protons.

**Figure 5.** CO adsorbate normalized spectra measured at −60 ◦C (**a**) of the samples 1.96 Co and (**b**) 2.44 Co.

Figure 5b shows the UV–Vis spectra of the sample 2.44 Co measured at room temperature after heating at 400 ◦C for 60 min followed by heating in 2 vol% O2/He at 400 ◦C for 45 min. After oxidation a new band around 340 nm appears, which is attributed to a ligand to metal charge transfer transition from framework oxide ions to Co3<sup>+</sup> ions [37,39,47]. In parallel, the intensity of the triplet band of Co2<sup>+</sup> decreases accordingly. This means that a part of Co2<sup>+</sup> species, most probably those occurring in oxide-like clusters, has been oxidized to Co3<sup>+</sup>.

These experiments reveal that the calcination/dehydration conditions influence the nature and location of formed Co species, but even treatment in vacuum at high temperature does not inevitably lead to Co2<sup>+</sup> located only at exchange positions. The fact that part of the Co2<sup>+</sup> species can be easily oxidized leads to the conclusion that these Co2<sup>+</sup> ions occur mainly in oxide-like agglomerates located inside and/or outside the zeolite channels. This means in terms of the UV–Vis spectra of such samples that the observed spectra represent not exclusively isolated Co2<sup>+</sup> species but also those located in agglomerated, oxide-like species inside and/or outside the zeolite channels. The same conclusion was drawn from the results of pyridine and CO adsorption as described above.

These findings are in accordance with those from former studies [12,25,26] revealing that the distribution of Co species is by far more complex, as deduced from previous papers in which UV–Vis, EXAFS, and XRD data have been discussed [17–19,48]. Insofar, a detailed analysis and quantification of Co positions in the zeolite matrix on the basis of deconvoluted UV–Vis spectra is practically not possible because the spectrum reflects Co species occupying positions inside the zeolite channels as well as positions on internal and external surfaces. This conclusion is also supported by the fact that for Co sites in zeolite cavities as well as for those dispersed on the surface of different oxide supports, similar UV–Vis spectra with the characteristic band triplet are obtained. The characteristic band splitting is due to symmetry perturbation, dynamic Jahn–Teller effect, and spin-orbit coupling [38,39] and is observed in general for Co2<sup>+</sup> ions independently from the surrounding matrix. For determining nature and distribution of Co2<sup>+</sup> species in Co-exchanged zeolites, the use of only UV–Vis–DRS cannot reflect the real situation. For this purpose, the combination of several, complementary methods have to be applied.

From the presented results the following conclusions can be derived. The observed triplet bands in the UV–Vis spectra of Co2+-containing samples reflect the sum of spectra of all existing Co2<sup>+</sup> species (isolated and oxide-like) located at different positions inside the zeolite channels as well as on internal and external surfaces. The characteristic band splitting is due to symmetry perturbation, dynamic Jahn–Teller effect, and spin-orbit coupling and is generally observed for Co2<sup>+</sup> ions in zeolite

as well as oxide materials. Because comparable UV–Vis spectra with a characteristic band triplet is observed for Co2<sup>+</sup> ions independently from the surrounding matrix, the former described UV–Vis band deconvolution method for determination and quantification of individual cationic sites in zeolites is not applicable.

#### **3. Materials and Methods**

The Co–ZSM-5 catalysts were prepared starting from NH4–ZSM-5 (Zeolyst, CBV 2314, SiO2/Al2O3 = 23). Co was introduced by wet ion-exchange (IE) using solutions of Co(NO3)2–6H2O (Sigma-Aldrich, St. Louis, MO, USA) or Co(CH3COO)2–4H2O (Sigma-Aldrich, St. Louis, MO, USA), adapting the method described by Beznis et al. [27]. The used conditions are summarized in Table 1. The zeolite was suspended in the respective Co salt aqueous solution and preheated to 60 ◦C. The pH was adjusted to 6.9, and the solution was stirred for 24 h. After filtration, the material was washed with 400 mL desalinated water/g solid. After drying over night at 90 ◦C, the solids were calcined in synthetic air at 500 ◦C for 6 h. For the preparation of the Na–ZSM-5, a solution of NaNO3 (Sigma-Aldrich, St. Louis, MO, USA) was used. The exchange procedure and subsequent washing was repeated two times. Finally, the solid was dried at 100 ◦C for 24 h and calcined in synthetic air at 500 ◦C for 6 h. For comparison, the respective H–ZSM-5 catalyst was prepared applying the same IE method but using only water. The sample 1.96 Co was prepared by the method described by Dˇedeˇcek et al. [19] with adjusted pH of 5.5.

The Co–SiAl sample was prepared by incipient wetness impregnation (IWI) of a Siralox sample from Sasol (49.2% SiO2, 50.8% Al2O3), which was dried overnight at 120 ◦C. After dropwise adding of a respective volume of a solution of Co(CH3COO)2·4H2O (Carl Roth, Karlsruhe, Germany) under stirring and further stirring for 2 h, the suspension was dried overnight and calcined at 500 ◦C for 6 h in synthetic air.

For the preparation of Co–AlSBA, the starting Al–SBA-15 material was prepared according the method described by Wu et al. [49]. The obtained solid was dried for 12 h at 70 ◦C and then calcined in synthetic air at 650 ◦C for 16 h. This Al–SBA-15 (Si/Al = 20) was used for the IE procedure described for Co–ZSM-5 above.

The Si/Al ratio of H–ZSM-5 and the contents of Co, Al, Na were determined by ICP-OES analysis.

The UV–Vis–DR spectra were recorded in diffuse reflection mode between 200 and 800 nm on a Cary 5000 spectrophotometer (Agilent, Santa Clara, CA, USA) equipped with a diffuse reflection accessory (Praying mantis) and a Harrick reaction cell. After treatment, the samples were heated at 400 ◦C for 30 min, and the spectra were measured at room temperature.

In situ FTIR measurements were carried out in transmission mode on a Bruker Tensor 27 spectrometer or Nicolet iS10 spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with custom-made reaction cells with CaF2 windows for adsorption studies of pyridine and CO, respectively. The sample powders were pressed into self-supporting wafers with a diameter of 20 mm and a weight of 50 mg. For each experiment, the sample was pretreated in synthetic air at 400 ◦C for 1 h. After cooling down to the respective adsorption temperature and evacuation of the cell a background, spectrum of the sample was recorded. Pyridine was adsorbed at 150 ◦C temperature until saturation. Then, the reaction cell was evacuated to remove physisorbed pyridine, and an adsorbate spectrum was recorded. For CO adsorption, a 5% CO/He mixture was used, dosed in pulses at −120 ◦C until saturation. To remove the physisorbed CO, the cell was evacuated, and the CO adsorbate spectrum was recorded. Subsequently, the CO desorption under vacuum was followed by continuously heating the sample and measuring a spectrum every 10 ◦C. Generally, the spectrum of the catalyst measured after pretreatment at adsorption temperature was subtracted from the respective adsorbate spectra of pyridine and CO.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/10/1/123/s1, Figure S1: Deconvoluted UV–Vis–DR spectra of selected samples measured at room temperature after heating the samples at 400 ◦C in He for 30 min.

**Author Contributions:** Conceptualization, U.B. and A.B.; investigation, A.B. and C.R.; data curation, A.B. and C.R.; writing—original draft preparation, U.B.; writing—review and editing, A.B.; supervision, A.B.; funding acquisition, U.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Leibniz-Gemeinschaft, grant number SAW-2013-LIKAT-2.

**Acknowledgments:** The authors thank Anja Simmula for the ICP-OES analysis and Dominik Seeburg for the preparation of Al–SBA-15.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Catalysts* Editorial Office E-mail: catalysts@mdpi.com www.mdpi.com/journal/catalysts

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18