Catalytic performance and mechanism of KF-loaded catalysts for biodiesel synthesis

Abstract

A number of KF-loaded heterogeneous base catalysts were prepared by doping KF on mixed oxide or single oxide supports containing Mg, Cu, Zn, Co, Al, Cr, Ni or Fe. The catalysts were characterized and tested for the transesterification of vegetable oil with methanol to produce biodiesel, in order to elucidate the active phase(s). The experimental results showed that KF doping increased activity irrespective of the nature of the support (i.e. whether it was mixed oxide or single oxide). For all catalysts KOH, formed during the treatment of each support with KF, was demonstrated to be the active phase. The activity of KF doped catalysts is promoted by the surface F-species. On this basis, it is possible to rationalize the general effect that KF has in promoting base catalyzed activity and select suitable supports in the design of highly active KF-loaded catalysts.

---

1. Introduction

Heterogeneous catalysts are economically and ecologically important because they are non-corrosive, environmentally benign and present fewer disposal problems. They are also much easier to separate from liquid products than their homogeneous counterparts and they can be designed to give high activity, selectivity and long catalyst lifetimes.^1–3 As a consequence of these advantages, there is increasing interest in the application of heterogeneous base catalysts for processes of importance to the fine chemical industry.^4–7 Many different types of bases, such as alkaline earth metal oxides, anion exchange resins and alkali metal compounds supported on alumina or zeolite, can be applied as catalysts to base catalyzed reactions such as isomerization, aldol condensation, Knoevenagel condensation, Michael condensation, oxidation and transesterification.⁸

KF/Al₂O₃ catalysts were originally reported in 1979 by Yamawaki and Ando,⁹ and since then interest in KF-loaded systems has grown considerably, with hundreds of papers being published detailing their application as heterogeneous catalysts.^10–12 Due to their importance as heterogeneous base catalysts, a significant degree of attention has been focused upon the elucidation of the active phase and mechanism of reaction and currently no consensus has been reached. KF/Al₂O₃ is an illustrative example in this respect. Ando et al.¹³ concluded that there were origins of the basicity of KF/Al₂O₃ of importance for catalytic activity: (1) the presence of active fluoride, (2) the presence of the [Al–O⁻] ion, which generates OH⁻ when water is added, and (3) co-operation of F⁻ and [Al–OH]. However, Kabashima et al.¹⁴ reported that the KF/Al₂O₃ surface species relevant to catalytic activity (particularly for the double bond isomerization and Tishchenko reactions), was a F⁻ containing species, which gave a peak at −150 ppm in ¹⁹F MAS NMR. In contrast, Weinstock et al.¹⁵ have proposed that the basicity of KF/Al₂O₃ is induced by the formation of KOH by reaction of KF with the alumina support during preparation. Recently, Verziu et al.¹⁶ investigated the nature of the active site for mesoporous alumina supported alkaline fluorides applied to the transesterification of vegetable oils. They concluded that the active sites of alkaline fluorides were generated by co-operation between the fluorine and oxygen species. In addition to Al₂O₃, many other supports have been applied to KF based catalysts. These include ZnO,¹⁷ zeolites,¹⁸ fluorapatite,¹⁹ Zn(Al)O,²⁰ and Mg–Al hydrotalcite.²¹ The application of different supports complicates the study of the catalytic mechanism and this results in significant difficulty in the identification of the most suitable support for a given application.^9–21

The number and strength of base sites are highly pertinent parameters for the activity of heterogeneous base catalysts.⁸ In KF-loaded catalysts, it is known that the basicity is generated as a consequence of the KF loading process.^9–21 The identification of the basicity is crucial for the elucidation of the catalytic mechanism and the identification of the active site. Two explanations can be advanced to explain why the source of basicity (i.e. the active sites and catalytic mechanism) is still not understood despite 30 years of extensive study on KF-loaded catalysts. One reason is the complexity of the catalyst surface. The other reason is that most of the studies performed to date have concentrated upon systems employing a single support.^9–21 However, since each system is complex, it may be possible to gain a greater degree of understanding by taking an overview of the common features of supported KF catalysts employing a very wide range of supports. This general approach has been adopted in the present study as detailed below.

In recent years, hydrotalcite-like compounds (hereafter denoted HTlcs) have been applied to numerous processes, and have found use as catalyst precursors or supports, ion-exchangers, stabilizers and adsorbents.^4,22 HTlcs are layered double hydroxides belonging to the class of anionic clays. Their general formula can be represented as: [M(II)_1−xM(III)_x(OH)₂·(Aⁿ⁻_x/n)·mH₂O], where M(II) (e.g. Mg, Zn, Cu, Mn, Co or Fe) and M(III) (e.g. Al, Cr, Fe, Ni or La) are divalent and trivalent metals, respectively; the value of x is equal to the molar ratio of M(II)/(M(II) + M(III)) and is generally in the range 0.2–0.33; Aⁿ⁻ is the anion that compensates for the positive charge of the hydroxide layers. Thermal decomposition of HTlcs leads to mixed oxides. The structure of HTlcs can accommodate a wide variation of different M(II) ions, M(III) ions, mixtures of M(III) and M(II) ions, M(II) : M(III) atomic ratios (represented by x = M(III)/M(II) + M(III) ranging from 0.1 to 0.33), and different types of inter-layered anion with charges ranging from −1 to −3.^4,7,23,24 As a result of this, it is possible to tune the structure and properties of mixed oxide catalyst supports by employing HTlcs as precursors. Accordingly, the structure and properties of KF supported upon mixed oxides can be modified. In this way, it is possible to generate a wide range of supported KF systems, which can be investigated to yield information on the nature of the reaction system and/or the active sites.

In this work, we have applied different HTlcs and single hydroxides as support precursors, from which we have prepared a series of KF-loaded catalysts. These KF-loaded catalysts have been characterized and their catalytic activity for the transesterification of vegetable oil with methanol to produce biodiesel has been determined. The aim of this study was to determine the catalytic mechanism and the structural and catalytic influence exerted by the support. Our experimental results demonstrate that KOH is the source of the basicity and is the active phase for the supported KF systems. This work is of interest in relation to the on-going discussion of the active phase and catalytic mechanism of KF-loaded catalysts, and it is useful for selecting suitable supports with the aim of designing highly active KF-loaded catalysts.

2. Experimental

2.1. Preparation of catalysts

2.1.1. Preparation of support. HTlcs with M²⁺ : M³⁺atomic ratios of 3 : 1 were prepared using a standard aqueous co-precipitation method at constant pH and temperature.²⁵ An aqueous solution (166 mL) of the metal nitrates in the desired M²⁺ : M³⁺ molar ratio with a total concentration of 1.5 M was mixed slowly with an alkaline solution of NaCO₃/NaOH with continuous stirring. The molar quantity of NaCO₃ employed was twice that of M³⁺. The pH value of the mixture was kept constant, typically at values between 9 and 10, by adjusting the rate of addition of the alkaline solution. The temperature was maintained at 60 °C. Following this, which resulted in the formation of heavy slurry, the mixture was aged at 60 °C for 18 h with stirring, to enhance the selective formation of the precipitated hydrotalcite phase. The slurry was then cooled to 25 °C, filtered, and washed with water until the pH value of the filtrate was near 7. The precipitate was dried at 90 °C for 16 h. The resulting materials were HTlcs, which were then calcined at 500 °C for 3 h in static air to generate mixed oxides (hereafter denoted as M²⁺(M³⁺)O). In this study, the M²⁺(M³⁺)O materials prepared were Mg(Al)O, Cu(Al)O, Co(Al)O, Ni(Al)O, Mg(Fe)O, Cu(Fe)O, Ni(Fe)O, Co(Cr)O and Zn(Cr)O. Table 1 summarizes the systems prepared.

Table 1 Nitrogen physisorption data of the M²⁺(M³⁺)O catalyst supports

Entry	Catalyst	Surface area (m^2 g^−1)	Pore Volume (cm^3 g^−1)	Average pore width (Å)
1	Mg(Al)O	190	0.60	114
2	Ni(Al)O	176	0.56	103
3	Co(Al)O	106	0.50	155
4	Cu(Al)O	54	0.20	120
5	Mg(Fe)O	141	0.88	212
6	Ni(Fe)O	97	0.64	225
7	Cu(Fe)O	9	0.05	280
8	Zn(Cr)O	52	0.53	359
9	Co(Cr)O	100	0.37	119

---

In order to compare the activities of the catalysts containing mixed oxides with their single oxide component counterparts, single oxides (hereafter denoted as MO) were prepared by calcining single hydroxides (hereafter indicated as M(OH)_x) in air at 500 °C for 5 h. The M(OH)_x precursors were obtained by precipitation using aqueous NaOH from the corresponding aqueous nitrate salt solutions. The MO supports prepared are listed in Table 2. The γ-Al₂O₃ used was a commercial sample (Alfa Aesar, S_BET = 220 m² g⁻¹).

Table 2 Nitrogen physisorption data of MO catalyst supports

Entry	Catalyst	Surface area (m^2 g^−1)	Pore volume (cm^3 g^−1)	Average pore width (Å)
1	MgO	151	0.85	148
2	NiO	20	0.1	132
3	CuO	8	0.02	449
4	Fe2O3	27	0.26	303
5	Co3O4	16	0.09	253
6	Cr2O3	29	0.2	275

---

2.1.2. Preparation of KF-loaded catalyst. An 80 wt% (KF to M²⁺(M³⁺)O weight ratio) KF on M²⁺(M³⁺)O support was prepared by an impregnation method using an aqueous solution of KF synthesized according to literature procedure.^9,20 A solid M²⁺(M³⁺)O support (3 g) was mixed with KF (2.4 g) in 15 mL of water, and the resulting solid was dried in an oven at 90 °C for 12 h and then crushed. Hereafter, the dried solid is referred to as KF/M²⁺(M³⁺)O-dy. The KF/M²⁺(M³⁺)O-dy materials were further calcined in air at 500 °C for 3 h to yield KF/M²⁺(M³⁺)O-cn materials. An 80 wt% KF on MO was also prepared in a similar way to KF/M²⁺(M³⁺)O-cn, and was denoted KF/MO-dy (before calcination) and KF/MO-cn (after calcination).

2.2. Catalyst characterization

X-Ray diffraction patterns were recorded using a D/Max-3C X-ray powder diffractometer (Rigaku Co., Japan), using a Cu-Kα source fitted with an Inel CPS 120 hemispherical detector.

Infrared transmission spectra were recorded in the range 400–4000 cm⁻¹ using an FTIR spectrometer (Bruker Tensor 27, Germany) and applying the KBr pellet technique. The IR spectra of the samples are shown in supplementary data (ESI, Fig. S1 and S2†).

The surface area and pore characteristics of the catalysts were determined using a Micromeritics ASAP 2020 instrument. The sample was degassed at 250 °C for 4 h in N₂ prior to surface area measurement. Nitrogen adsorption and desorption isotherms were measured at −196 °C, and the specific surface areas of the catalysts were determined by applying the BET (Brunauer-Emmett-Teller) method to nitrogen adsorption data obtained in the relative pressure range from 0.06 to 0.30. Total pore volumes were estimated from the amount of nitrogen adsorbed at a relative pressure of 0.995. Pore volume and pore-size distribution curves were obtained from analysis of the desorption branches of the nitrogen isotherms using the BJH (Barrett-Joyner-Halenda) method (Table 1 and 2; ESI, Table S1 and Fig. S3†).

Hammett indicator experiments were conducted to determine the H₋ value range of the basic sites for white solids only, since the sensitivity of this method was affected for colored solids. The conditions of pre-treatment of the solid for the determination of basic strength using Hammett indicators are given in Table 3. The Hammett indicators used were methyl yellow (pK_a = 3.3), neutral red (pK_a = 6.8), bromthymol blue (pK_a = 7.2), phenolphthalein (pK_a = 9.3), alizarin yellow R (pK_a = 11.0), indigo carmine (pK_a = 12.2), 2,4-dinitroaniline (pK_a = 15), 4-nitroaniline (pK_a = 18.4), 4-chloroaniline (pK_a = 26.5), and diphenylmethane (pK_a = 35). Typically, 25 mg of the catalyst was mixed with 1 mL of a solution of Hammett indicator diluted with cyclohexane and allowed to equilibrate for at least 1 h, after which the color of the catalyst was noted. The basic strength of the catalyst was taken to be higher than the weakest indicator that underwent a color change and lower than the strongest indicator that underwent no color change.

Table 3 Results of Hammett indicator measurements on catalysts^a

Sample	Pre-treatment T (°C)	Basic Strength
Mg–Al HTlcs	100	7.2 < H− < 9.3
Mg(Al)O	500	12.2 < H− < 15.0
KF/Mg(Al)O-cn	500	18.4 < H− < 26.5
MgO	500	11.0 < H− < 12.2
KF/MgO-cn	500	18.4 < H− < 26.5
γ-Al2O3	500	7.2 < H− < 9.3
KF/γ-Al2O3-cn	500	18.4 < H− < 26.5

---

CO₂ temperature programmed desorption (TPD) experiments were performed using a microreator (AutoChem II 2920, Micromeritics Instrument Corporation, USA) loaded with 70 mg of sample. The sample was pre-treated under He (30 mL min⁻¹) ramping at 15 °C min⁻¹ up to 500 °C where it was held for 30 min, and then cooled to 40 °C prior to the adsorption of CO₂. After the adsorption of CO₂ (30 mL min⁻¹) for 60 min, the catalyst was flushed with He (30 mL min⁻¹) for 60 min at 40 °C to remove the physisorbed gas from the surface of the catalyst, and the desorption profile was recorded employing a heating ramp of 10 °C min⁻¹ between 40 to 500 °C where it was held for 30 min. TPD profiles of CO₂ adsorbed on M²⁺(M³⁺)O and KF/M²⁺(M³⁺)O-cn materials are shown in the ESI (ESI, Fig. S4†).

Thermogravimetric analysis (TGA) experiments were carried out using a Q600 SDT thermal analysis machine (TA Instruments, USA) under a flow of nitrogen. The sample weight used was about 20 milligrams, and the temperature ranged from 38 to 1000 °C with a ramping rate of 20 °C/min.

2.3. Reaction procedure

The catalytic transesterification reactions were carried out under vigorous stirring in a round-bottomed flask equipped with a reflux condenser. Typical reactions were performed with 19.6 mL of vegetable oil (100% soybean oil; Xi'an Jiali Grease Industrial Co., Ltd., Xi'an, China) and 4.6 mL of methanol (methanol to vegetable oil molar ratio 6 : 1) using 3 wt% (catalyst to oil weight ratio) of catalyst at methanol reflux temperature (65 °C) for the specified reaction time. The reaction products were analyzed using the following procedure: the samples were separated from the catalyst and glycerol by centrifugation, with the separation of glycerol being achieved because it was insoluble in the esters and had a much higher density. The methanol was removed under vacuum and the product was added to chloroform-d for ¹H-NMR spectroscopy. The methyl ester could be readily determined quantitatively from the NMR spectra using the method described by Gelbard et al.²⁶

3. Results and discussion

3.1. BET surface area and pore size

The surface area of the mixed oxides was a function of their composition. Mg(Al)O had the highest surface area (190 m² g⁻¹). Cu(Fe)O had the lowest surface area (9 m² g⁻¹). Most of the prepared mixed oxides had high surface area (>54 m² g⁻¹). The surface areas of single oxides were also determined and are shown in Table 2. Among them, MgO had the highest surface area (151 m² g⁻¹) and the surface areas of the remaining single oxides were lower than 30 m² g⁻¹. Hence, compared to the mixed oxides, most of the single oxides had lower surface area.

3.2. XRD analysis

Fig. 1 presents the XRD patterns of the HTlcs, M²⁺(M³⁺)O, KF/M²⁺(M³⁺)O-dy and KF/M²⁺(M³⁺)O-cn materials. The X-ray diffractograms of the HTlcs display characteristic diffraction peaks at 11.4°, 23.0°, 34.9° (Fig. 1 (a)).^4,27 The thermal pre-treatment of the HTlcs results in changed XRD patterns, caused by the structural changes associated with the loss of CO₂ and H₂O from the starting material (Fig. 1 (b)).

Fig. 1 XRD patterns showing (a) HTlcs, (b) M²⁺(M³⁺)O, (c) KF/M²⁺(M³⁺)O-dy, and (d) KF/M²⁺(M³⁺)O-cn materials.

Mg(Al)O displayed reflections characteristic of MgO (Fig. 1A (b)).²⁷ MgO was rehydrated to form Mg(OH)₂ in the process of preparing KF/Mg(Al)O-dy. The XRD pattern of KF/Mg(Al)O-dy showed, in addition to Mg(OH)₂, a series of new diffraction peaks (Fig. 1A (c)). Among them, the reflections at 29.8° may correspond to K₃AlF₆ and the reflections at 31.8°, 39.2° and 45.6° could be associated with KMgF₃. The XRD patterns of KF/Mg(Al)O-cn showed diffraction peaks of MgO, K₃AlF₆ and KMgF₃.^10,16,21,27 The MgO resulted from the removal of H₂O from the Mg(OH)₂ of KF/Mg(Al)O-dy. The formation of the new crystalline phases (KMgF₃ and K₃AlF₆) can be ascribed to a chemical reaction between KF and the Mg(Al)O support (eqn (1) and (2)) during the preparation of KF/Mg(Al)O-dy.^10,16,21,27 Stoichiometric considerations indicate that KOH will also be produced when KF is loaded on the Mg(Al)O support, despite the fact that no XRD reflections characteristic of KOH were found in the XRD pattern of the KF/Mg(Al)O catalysts, possibly as a consequence of high dispersion (Fig. 1A (d)).

3KF + MgO + H₂O → KMgF₃ (K₂MgF₄) + 2KOH (1)

12KF + Al₂O₃ + 3H₂O → 2 K₃AlF₆ + 6KOH (2)

12KF + Fe₂O₃ + 3H₂O → 2 K₃FeF₆ + 6KOH (3)

3KF + ZnO + H₂O → KZnF₃ (K₂ZnF₄, ZnF₂) + 6KOH (4)

12KF + Cr₂O₃ + 3H₂O → 2 K₃CrF₆ + 6KOH (5)

Co(Al)O displayed the reflections characteristic of Co₃O₄ (Fig. 1B (b)).²⁸ The XRD patterns of KF/Co(Al)O-dy and KF/Co(Al)O-cn showed, in addition to Co₃O₄, the presence of K₃AlF₆ (Fig. 1B (c) and (d)), the formation of which can be ascribed to the chemical reaction between KF and the Co(Al)O support (eqn (2)) during preparation.

The XRD patterns of Cu(Al)O and its supported materials were similar to that of Co(Al)O. Cu(Al)O displayed reflections characteristic of CuO (Fig. 1C (b)).²⁹ The XRD patterns of KF/Cu(Al)O-dy and KF/Cu(Al)O-cn showed, in addition to CuO, K₃AlF₆ (Fig. 1C (c) and (d)). The formation of the new phase (K₃AlF₆) can be ascribed to the chemical reaction between KF and the Cu(Al)O support (eqn (2)) during preparation.

Mg(Fe)O exhibited diffraction reflections characteristic of MgO (Fig. 1D (b)) and KF/Mg(Fe)O-dy showed diffraction peaks characteristic of Mg(OH)₂ and KMgF₃ (Fig. 1D (c)).³⁰ The XRD patterns of KF/Mg(Fe)O-cn indicated the presence of KMgF₃ and MgO (Fig. 1D (d)).

Ni(Fe)O showed diffraction reflections characteristic of NiO (Fig. 1E (b)) and the XRD patterns of KF/Ni(Fe)O-dy showed, in addition to NiO, a series of new diffraction peaks characteristic of K₃FeF₆ (Fig. 1E (c)).³¹ The presence of K₃FeF₆ may be ascribed to chemical reaction between KF and the Ni(Fe)O support (eqn (3)) upon preparation of KF/Ni(Fe)O-dy. The diffraction peaks characteristic of K₃FeF₆ became weak after calcination (Fig. 1E (d)).

The Cu(Fe)O displayed reflections characteristic of CuO (Fig. 1F (b)).²⁹ The XRD patterns of KF/Cu(Fe)O-dy (Fig. 1F (c)) and KF/Cu(Fe)O-cn (Fig. 1F (d)) were similar to that of the Cu(Fe)O support, suggesting that KF doping did not apparently change the crystalline structure of the support.

Zn(Cr)O exhibited diffraction reflections characteristic of ZnO and ZnCr₂O₄ (Fig. 1G (b)).³² The XRD patterns of KF/Zn(Cr)O-dy showed, in addition to ZnO and ZnCr₂O₄, a series of new diffraction peaks ascribed to KZnF₃, KF and K₂ZnF₄ (Fig. 1G (c)).^10,33 KF arises from residues from the preparation and the new crystalline phases (KZnF₃ and K₂ZnF₄) may be ascribed to the chemical reaction between KF and the Zn(Cr)O support (eqn (4)). Calcination enhanced the crystallinity of KF/Zn(Cr)O-cn (Fig. 1G (d)). Compared with KF/Zn(Cr)O-dy catalysts, the diffraction peak intensity of ZnO, KZnF₃ and K₂ZnF₄ was higher on KF/Zn(Cr)O-cn. Furthermore, the KF/Zn(Cr)O-cn displayed, in addition to ZnO, KZnF₃ and K₂ZnF₄, reflections characteristic of ZnF₂ and K₃CrF₆.^34,35 The ZnF₂ and K₃CrF₆ may have resulted from chemical reaction between KF and the Zn(Cr)O support (eqn (4) and (5)) during the calcination process.

Co(Cr)O displayed reflections characteristic of Co₂CrO₄ (Fig. 1H (b)). The XRD patterns of KF/Co(Cr)O-dy were similar to that of Co(Cr)O (Fig. 1H (c)). Calcination made KF/Co(Cr)O-cn more crystalline (Fig. 1H (d)). Furthermore, KF/Co(Cr)O-cn showed, in addition to Co₂CrO₄, a series of new diffraction peaks characteristic of K₃CrF₆ (Fig. 1H (d)).The new crystalline phase (K₃CrF₆) may be ascribed to the chemical reaction between KF and the Co(Cr)O support (eqn (5)) during calcination.

Fig. 2 presents the XRD patterns of the prepared M(OH)_x, MO and KF/MO catalysts. The M(OH)_x sample showed X-ray diffractograms typical of M(OH)_x (Fig. 2A–F (a)). The thermal pre-treatment of the M(OH)_x sample resulted in a change in the XRD pattern, caused by the removal of H₂O from the starting material (Fig. 2A–F (b)).The XRD pattern was further changed by the KF doping on the MO support (Fig. 2A–F (c)). The XRD pattern of KF/MgO-cn showed, in addition to MgO, a series of new diffraction peaks ascribed to K₂MgF₄ (Fig. 2A (c)).

Fig. 2 XRD patterns of (a) M(OH)_x, (b) MO, and (c) KF/MO catalysts.

KF/Cr₂O₃-dy displayed only diffraction peaks characteristic of Cr₂O₃ (Fig. 2F (c)). However, the XRD pattern of KF/Cr₂O₃-cn showed, in addition to Cr₂O₃, a series of new diffraction peaks ascribed to K₂Cr₂O₇ (Fig. 2F (d)). The K₂Cr₂O₇ compound was further validated by qualitative analysis. The newly formed K₂Cr₂O₇ compound was first separated from the KF/Cr₂O₃-cn material, and then treated with a solution of AgNO₃ or BaCl₂. A red or yellow precipitate was observed for Ag₂CrO₄ or BaCrO₄, respectively, demonstrating that the new compound was K₂Cr₂O₇. The production of K₂Cr₂O₇ during calcination is given by (eqn (6)):

4KOH + 2Cr₂O₃ + 3O₂ → 2 K₂Cr₂O₇ + 2H₂O (6)

where KOH was a component of KF/Cr₂O₃-dy, produced during the impregnation process of doping KF on the Cr₂O₃ support and O₂ originates from the air atmosphere, in which the calcination was performed. The reaction consumed KOH, implying that the basicity of KF/Cr₂O₃-cn will be lower than that of KF/Cr₂O₃-dy. The participation of O₂ in the chemical reaction indicated that the weight of the catalyst may be changed after calcination, which is supported by the results of TGA experiments on the KF/Cr₂O₃-dy sample (Fig. 3(b)).

Fig. 3 DSC/TG curves of (a) KF/Mg(Al)O-dy under a flow of nitrogen, (b) KF/Cr₂O₃-dy under a flow of oxygen, and (c) KF/Zn(Cr)O-dy under a flow of oxygen.

The XRD patterns of the other four MO supported KF catalysts (KF/NiO, KF/CuO, KF/Co₃O₄, KF/Fe₂O₃) showed, in addition to their corresponding supports, diffraction peaks ascribed to residual KF (Fig. 2B–E (c)), suggesting that KF was not well dispersed and crystallized on the MO supports, which may be a consequence of their low surface area. The results also indicated that no other crystalline phases were formed in the process of preparing these MO supported catalysts.

Analysis of the XRD patterns demonstrates that the formation of KOH and KM_xF_y occurs for some supports when they are treated with KF. These supports are Mg(Al)O, Co(Al)O, Cu(Al)O, Mg(Fe)O, Ni(Fe)O, Zn(Cr)O, Co(Cr)O and MgO. However, these new compounds were not observed in the XRD patterns of KF supported on Cu(Fe)O, NiO, CuO, Co₃O₄ and Fe₂O₃, suggesting that any KOH and KM_xF_y formed may be in an amorphous state since KF can react with all the supports during preparation.

3.3. Thermogravimetric analysis

The structure characterization by XRD analysis showed that calcination affected the structure of KF-loaded catalysts. In order to explain the effect of calcination, we have investigated the calcination process of KF/Mg(Al)O-dy, KF/Cr₂O₃-dy and KF/Zn(Cr)O-dy by thermogravimetric analysis. The DSC/TG measurements of KF/Mg(Al)O-dy (Fig. 3a) showed the temperatures at which the precursors decomposed when heated in a nitrogen atmosphere. The weight of KF/Mg(Al)O-dy decreased with temperature until 1000 °C. The weight loss that occurred below 211 °C was ascribed to the removal of physisorbed water, and the weight loss above 211 °C was attributed to dehydroxylation and loss of CO₂.

The discussion of KF/Cr₂O₃-cn in the previous section showed that O₂ participated in the process of calcining the KF/Cr₂O₃-dy samples. The proposed reaction pathway was illustrated by eqn (6), which indicated that sample mass increased after reaction. In order to verify the proposed reaction mechanism, thermal analysis of KF/Cr₂O₃-dy was undertaken in an O₂ atmosphere. The DSC/TG result (Fig. 3b) showed that its weight decreased with temperature until 400 °C, after which the weight began to increase reaching a maximum at about 626 °C. Above 626 °C, the mass started to drop. The weight loss below 400 °C may be ascribed to the removal of physisorbed or chemisorbed water and CO₂. The apparent increase of weight between 400 and 626 °C may be attributed to the chemical reaction between O₂ and the sample, as indicated in eqn (6). Above 626 °C, the mass loss may possibly result from the decomposition of the sample and the removal of volatile gas. The DSC/TG result of KF/Zn(Cr)O-dy (Fig. 3c) was similar to that of KF/Cr₂O₃-dy. Its weight also increased between 400 and 626 °C, and decreased below 400 °C as well as above 626 °C, indicating that O₂ also reacts with KF/Zn(Cr)O-dy as illustrated in eqn (6).

3.4. The basic strengths and basicity of the catalysts

The strength of the basic sites in the white samples has been analyzed qualitatively using Hammett indicators. As shown in Table 3, the Mg–Al HTlcs, Mg(Al)O, MgO and γ-Al₂O₃ samples were found to be weakly basic, possessing H₋ values in the range 7.2–15.0. Loading KF onto the support material promoted the basic strengths of the catalysts. KF/Mg(Al)O, KF/γ-Al₂O₃ and KF/MgO were found to possess H₋ values in the range of 18.4–26.5.

3.5. Catalytic transesterification

3.5.1. Effect of calcination. Table 4 shows the effect of calcination upon the catalytic activities of KF/M²⁺(M³⁺)O. The activity of KF/Mg(Al)O-cn, KF/Mg(Fe)O-cn, KF/Cu(Al)O-cn and KF/Ni(Fe)O-cn were similar to their corresponding precursors KF/Mg(Al)O-dy, KF/Mg(Fe)O-dy, KF/Cu(Al)O-dy and KF/Ni(Fe)O-dy. The activity of KF/Ni(Al)O-cn, KF/Co(Al)O-cn, and KF/Cu(Fe)O-cn was slightly higher than their corresponding uncalcined precursors (KF/Ni(Al)O-dy, KF/Co(Al)O-dy, and KF/Cu(Fe)O-dy). A contrary phenomenon was observed for both Zn(Cr)O and Co(Cr)O supported catalysts. The yield of biodiesel was 82% in the presence of KF/Co(Cr)O-dy but decreased to 70% over KF/Co(Cr)O-cn; the yield of biodiesel was 87% in the presence of KF/Zn(Cr)O-dy and dramatically decreased to 4% in KF/Zn(Cr)O-cn. The result suggested that calcination was detrimental for the activity of the KF/Zn(Cr)O-cn and KF/Co(Cr)O-cn catalysts.

Table 4 Biodiesel yields over KF/M²⁺(M³⁺)O-dy (before calcination) and KF/M²⁺(M³⁺)O-cn (after calcination) catalysts

Catalyst	Biodiesel yield (%)
	-dy	-cn
KF/Mg(Al)O	100	100
KF/Mg(Fe)O	93	96
KF/Cu(Al)O	92	90
KF/Ni(Fe)O	83	86
KF/Ni(Al)O	94	100
KF/Co(Al)O	81	95
KF/Cu(Fe)O	78	87
KF/Co(Cr)O	82	70
KF/Zn(Cr)O	87	4
KF/γ-Al2O3	82	86

---

The activity of KF/Cr₂O₃ was also affected by calcination (Table 5, entry 12). The biodiesel yield was 70% with KF/Cr₂O₃-dy as the catalyst; it decreased to 0 with KF/Cr₂O₃-cn as the catalyst. It seems that calcination leads to a decrease of activity for KF-loaded catalysts with supports containing Cr. The deactivation of KF/Zn(Cr)O-cn, KF/Co(Cr)O-cn and KF/Cr₂O₃-cn may be due to the decrease of KOH, resulting from reaction between Cr₂O₃ and KOH (eqn (6)) during calcination. The results suggest that the major active component of the prepared KF-loaded catalysts may be KOH.

Table 5 Biodiesel yields over MO and KF/MO catalysts

Entry	Catalysts	Biodiesel yield (%)
a Biodiesel yield in the presence of KF/Cr2O3-dy.
1	MgO	0
2	KF/MgO-cn	98
3	NiO	0
4	KF/NiO-cn	73
5	CuO	0
6	KF/CuO-cn	92
7	Fe2O3	0
8	KF/Fe2O3-cn	83
9	Co3O4	0
10	KF/Co3O4-cn	77
11	Cr2O3	0
12	KF/Cr2O3-cn	0 (70)^a

---

3.5.2. Effect of catalyst support. Table 4 also displays the effect of mixed oxide catalyst support. KF/Mg(Al)O-cn, KF/Ni(Al)O-cn, KF/Co(Al)O-cn and KF/Mg(Fe)O-cn exhibited the highest activity, with a biodiesel yield of >95%; KF/Cu(Al)O-cn, KF/Cu(Fe)O-cn, KF/Ni(Fe)O-cn, KF/Co(Cr)O-dy and KF/Zn(Cr)O-dy show similar activity to KF/γ-Al₂O₃, which gives a biodiesel yield in the range of 80–90%. On this basis, it appears that the promotional effect of Mg(Al)O, Co(Al)O, Ni(Al)O and Mg(Fe)O supports was higher than γ-Al₂O₃ support, whereas, the promotional effect of the Cu(Al)O, Cu(Fe)O, Ni(Fe)O, Co(Cr)O and Zn(Cr)O supports was similar to the γ-Al₂O₃ support. Therefore all the KF/M²⁺(M³⁺)O could replace KF/γ-Al₂O₃ as active heterogeneous basic catalysts.

The single oxide supports did not show any activity for biodiesel production under the reaction conditions studied (Table 5). However KF loading was observed to increase activity. The KF/MgO catalyst showed the highest activity, with a biodiesel yield of 98%; followed by the KF/CuO catalyst (biodiesel yield 92%). Other KF/MO systems had lower activity, with biodiesel yields in the range of 70–80%. Comparing the activity between the mixed oxide and single oxide systems, it was found that the activity of the mixed oxides was similar to, or a little higher, than their corresponding single oxide counterparts.

The results also showed that KF doping increases catalyst activity irrespective of whether the support was mixed oxide or singe oxide. It is apparent that the yield of the methyl esters does not directly correlate with the surface area of the catalyst support. The highest surface area was observed for the Mg(Al)O (190 m² g⁻¹) and the lowest surface area was for CuO (8 m² g⁻¹), yet the difference in their biodiesel yield was only 6%. The increased activity of the KF-loaded catalyst could be mainly ascribed to the active sites on the surface. The KF/M²⁺(M³⁺)O catalysts contained KF, KOH, KM_xF_y and M²⁺(M³⁺)O on their surfaces, while KF/MO catalysts contained KF, KOH, KM_xF_y and MO compounds on their surfaces (Table 6, entry 6 and 7). Of the surface compounds, KF was the dopant and M²⁺(M³⁺)O or MO was the support, which all showed low activity (Table 6, entry 1, 3 and 4). Since KOH and KM_xF_y were formed during the reatment of supports with KF, the high activity of the KF-loaded catalysts may result from the formation of either or both of the two new compounds. As the true active component should be common to all the prepared KF-loaded catalysts, KM_xF_y can be ruled out. In addition, since KM_xF_y are salts with no apparent basic properties it is not possible for them to act as basic catalysts for biodiesel synthesis. Furthermore, the commercially available K₃AlF₆, which has been tested in the transesterification of vegetable oil with methanol to biodiesel (Table 6, entry 5) showed no activity, as was expected. KMgF₃ has also been proved to be ineffective for biodiesel synthesis as described in section 3.6. On this basis it can be deduced that KOH should be the most active component because KM_xF_y, KF and the supports were inactive for biodiesel synthesis (Table 6, entry 6 and 7). This conclusion was also supported by the structure analysis and activity of the KF-loaded catalysts whose supports contained Cr. It suggests that any oxide could be the active support of KF-loaded catalysts, if the support reacts with KF to produce KOH. On this basis a further question can be posed as to whether materials other than oxides can be active supports if they react with KF to produce KOH.

Table 6 Structure analysis and study of catalytic mechanism

Entry	Catalysts	Comprised compounds on the surface	Active components^b	Biodiesel yield (%)^c
a Characterized by structure analysis. b Deduced from the correlation between structure and activity of catalyst. c Experimental results.
1	KF	KF	None	<1
2	HTlcs	HTlcs	None	<3
3	M^2+(M^3+)O	M^2+(M^3+)O	None	<8
4	MO	MO	None	<8
5	K3AlF6	K3AlF6	None	0
6	KF/M^2+(M^3+)O	KF	KOH	80–98
		M^2+(M^3+)O
		KMxFy^a
		KOH^a
7	KF/MO	KF	KOH	70–98
		MO
		KMxFy (K2Cr2O7)^a
		KOH^a
8	KF/Mg–Al HTlcs	KF	KOH	95
		Mg–Al HTlcs
		KMxFy^a
		KOH^a

---

In order to answer this question, a KF/Mg–Al HTlcs catalyst was prepared by doping KF on the Mg–Al HTlcs support. Powder X-ray diffraction showed that the KF/Mg–Al HTlcs catalyst contained KMgF₃ (ESI, Fig. S5†). It is evident that Mg–Al HTlcs can react with KF to produce KOH. The activity of the KF/Mg–Al HTlcs catalyst was tested under the same conditions as the KF/M²⁺(M³⁺)O and KF/MO catalysts. A high yield (>95%) was obtained (Table 6, entry 8), indicating that Mg–Al HTlcs could be an active support of KF-loaded catalysts. Therefore it can be concluded that materials other than oxides are effective supports if they can also react with KF to produce KOH.

3.6. Comparison of KF doped catalyst and an equivalent KOH doped catalyst

The role of KF was investigated by comparing the activity of KF doped catalysts with that of an equivalent KOH doped catalyst (Table 7). The amount of KOH liberated from the KF-support reaction was determined by titrimetric analysis according to literature procedures.¹³ Aqueous washing of the KF doped reagents gives an alkaline solution, titration of which gives a guide to the amount of soluble base formed during the reagent preparation. Table 7 gives the measured basicities for the similarly prepared KF doped reagents (KF/Mg(Al)O, KF/Al₂O₃, KF/MgO, and KF/Fe₂O₃). The amount of soluble base for KF/Mg(Al)O was the highest (0.0493 g). Next was that of KF/Al₂O₃ (0.0474 g). The amount of soluble base for KF/Fe₂O₃ was far less than that of others (0.0007 g), being about 1.4% of the amount for the KF/Mg(Al)O and KF/Al₂O₃ catalysts. Table 7 also gives the activity for pure KOH and KOH doped catalysts. The activity of 0.0493 g of pure KOH (equivalent to the amount of base liberated by KF/Mg(Al)O) was similar to that of KF/Mg(Al)O (Table 4, biodiesel yield 100%), but higher than that of KOH/Mg(Al)O (Table 7, entry 2; biodiesel yield 81%). A similar phenomenon was observed for the KF/Al₂O₃ catalyst (Table 7, entry 3 and 4; Table 4). The activity of KF/MgO (Table 5, entry 2; biodiesel yield 98%) was higher than that of both pure KOH (Table 7, entry 5; biodiesel yield 87%) and KOH/MgO (Table 7, entry 6; biodiesel yield 60%). The activity of the pure KOH (Table 7, entry 7; biodiesel yield 1%) and KOH/Fe₂O₃ (Table 7, entry 8; biodiesel yield 1%) was far less than that of KF/Fe₂O₃ (Table 5, entry 8; biodiesel yield 83%). The results demonstrate that the activity of the KF doped catalyst is higher than that of an equivalent KOH doped support. Furthermore, except the case about KF/Fe₂O₃, even the activity of pure KOH is higher than that of an equivalent KOH doped support. This indicates that the KF doped catalysts can not be prepared directly by adding KOH to the supports. The different activity of the KF doped catalyst and KOH doped catalyst may be related to the difference of their structure. The main surface species of the KOH doped catalysts is KOH, while KOH, KF and KM_xF_y are the main surface species of the KF doped catalysts. Therefore, the higher activity of the KF doped catalysts may be ascribed to the synergistic relationship between the KOH (active sites) and the other surface species (KF or KM_xF_y). The KOH pure is an active homogeneous catalyst for biodiesel synthesis, while both KF pure and KM_xF_y pure show no activity. Hereby, the surface F-species (KF or KM_xF_y) may work as cocatalysts for KF doped catalysts.

Table 7 The measured basicities for the similarly prepared KF doped reagents and the activity of pure KOH and KOH doped catalysts

Entry	Catalyst	Amount (g)	Biodiesel yield (%)
			-dy	-cn
a The amount of KOH liberated from 0.5 g of KF/Mg(Al)O-cn. b The amount of KOH liberated from 0.5 g of KF/γ-Al2O3-cn. c The amount of KOH liberated from 0.5 g of KF/MgO-cn. d The amount of KOH liberated from 0.5 g of KF/Fe2O3-cn.
1	KOH	0.0493^a	97
2	KOH/Mg(Al)O	KOH (0.0493)/Mg(Al)O (0.278)	81	81
3	KOH	0.0474^b	87
4	KOH/γ-Al2O3	KOH (0.0474)/γ-Al2O3 (0.278)	69	50
5	KOH	0.0295^c	87
6	KOH/MgO	KOH (0.0295)/MgO (0.278)	63	60
7	KOH	0.0007^d	6
8	KOH/Fe2O3	KOH (0.0007)/Fe2O3 (0.278)	10	10

---

3.7. Recycling experiments for catalysts

Three catalysts KF/Mg(Al)O, KF/Cu(Fe)O and KF/Fe₂O₃ were reused (Table 8). The biodiesel yield of the three fresh catalysts was in the range of 84–99% (Table 8, entry 1). Upon second use (Table 8, entry 2) a decrease in the activity of KF/Mg(Al)O and KF/Fe₂O₃ of about 30% was evident, with a greater decrease for KF/Cu(Fe)O being apparent, for which the biodiesel yield sharply dropped to 16%. Upon third use, activity was further reduced, and the catalysts were completely deactivated after being used more than 4 times. Although there were differences in the stability, the activity of the three tested catalysts declined with each reuse, which is a similar observation to other reported KF-loaded catalysts.^16,20 The XRD results of KF/Mg(Al)O-cn before use and after being used 5 times are shown in Fig. 4, in which similar patterns are evident. Reflections characteristic of KMgF₃ were observed in the XRD patterns of KF/Mg(Al)O-cn before use and after being used 5 times, further indicating that KMgF₃ was not an active phase for the biodiesel synthesis since the catalyst was almost totally deactivated after being used 5 times. As discussed previously, analysis of the structure–activity relationship revealed that KOH was the active phase. Since KOH is soluble in the reactant methanol, the deactivation of reused catalysts may be due to the leaching of the active component KOH. Based on the experimental results, it is suggested that future work should be concentrated on the development of methods to prevent the leaching of the active component, KOH, and to improve the stability of KF-loaded catalysts.

Table 8 Recycling experiments for the catalysts^a

Entry	Used times	Biodiesel yield (%)
		KF/Mg(Al)O-cn^b	KF/Cu(Fe)O-cn^b	KF/Fe2O3-cn^c
a The used catalyst was separated from the reaction solution by centrifugation. It was taken as the catalyst for the repeated reactions. The repeated reactions were run under the same reaction conditions to the last reaction. b Reaction conditions were supplied in experimental section. c Catalyst amount was 5 wt%, other reaction conditions were supplied in experimental section.
1	First use	99	84	97
2	Second use	59	16	72
3	Third use	46	<1	10
4	Fourth use	16	—	<1
5	Fifth use	8	—	—

---

Fig. 4 XRD patterns of KF/Mg(Al)O-cn (a) before use and (b) after being used 5 times.

4. Conclusions

In this work, a series of KF-loaded heterogeneous base catalysts have been prepared, characterized and tested for the transesterification of vegetable oil with methanol to yield biodiesel. The deposition of KF on supports generates KM_xF_y and KOH species. The results showed that KF loading could increase the activity of each prepared catalyst irrespective of whether the support was a mixed oxide or singe oxide. The supports showed no activity (biodiesel yield <8%), but when loaded with KF high activities were apparent with biodiesel yields in the range of 73–98% being observed. The active phase of the KF-loaded catalysts was studied by systematically investigating the relationship between structure and activity. It was demonstrated that the active site of the KF-loaded catalysts is KOH, which is formed during the treatment of supports with KF. The activity of the KF doped catalysts is promoted by the other surface species (KF or KM_xF_y).

Based on the results above, it is proposed that materials other than oxides can be active supports if they react with the KF dopant to produce KOH. This supposition is further supported by the experimental data concerning the direct application of hydrotalcite as a support for KF. On this basis, it is possible to identify potential novel highly active supports and these will not solely be limited to γ-Al₂O₃ and other reported supports. Furthermore, this study also shows that future work on the KF-loaded catalysts should be directed towards the development of methods to prevent leaching of the active KOH phase, hence improving the stability of the KF doped catalysts. Therefore, this work will be useful for selecting suitable supports in the design of highly active KF-loaded catalysts.

Acknowledgements

This work was supported by the Scientific Research Foundation for the Returned Overseas Chinese Scholars and the Project funded by Natural Science Basic Research Plan in Shaanxi Province of China (Program No. 2010JM2003). We thank Dr J. S. J. Hargreaves (Department of Chemistry, University of Glasgow, UK) for language help.

References

1. X. Xie, Y. Li, Z. Q. Liu, M. Haruta and W. Shen, Nature, 2009, 458, 746.
2. R. Juárez, P. Concepción, A. Corma, V. Fornés and H. García, Angew. Chem., Int. Ed., 2010, 49, 1286.
3. A. Thomas and M. Driess, Angew. Chem., Int. Ed., 2009, 48, 1890.
4. D. P. Debecker, E. M. Gaigneaux and G. Busca, Chem.–Eur. J., 2009, 15, 3920.
5. S. Shylesh, A. Wagener, A. Seifert, S. Ernst and W. R. Thiel, Angew. Chem., Int. Ed., 2010, 49, 184.
6. M. Toda, A. Takagaki, M. Okamura, J. N. Kondo, S. Hayashi, K. Domen and M. Hara, Nature, 2005, 438, 178.
7. C. M. Jinesh, C. A. Antonyraj and S. Kannan, Catal. Today, 2009, 141, 176.
8. G. Busca, Chem. Rev., 2010, 110, 2217.
9. J. Yamawaki and T. Ando, Chem. Lett., 1979, 8, 755.
10. J. M. Clacens, D. Genuit, L. Delmotte, A. Garcia-Ruiz, G. Bergeret, R. Montiel, J. Lopez and F. Figueras, J. Catal., 2004, 221, 483.
11. B. E. Blass, Tetrahedron, 2002, 58, 9301.
12. N. Boz, N. Degirmenbasi and D. M. Kalyon, Appl. Catal., B, 2009, 89, 590.
13. T. Ando, J. H. Clark, D. G. Cork, T. Hanafusa, J. Ichihara and T. Kinura, Tetrahedron Lett., 1987, 28, 1421.
14. H. Kabashima, H. Tsuji, S. Nakatab, Y. Tanaka and H. Hattori, Appl. Catal. A, 2000, 194–195, 227.
15. L. M. Weinstock, J. M. Stevenson, S. A. Tomellin, S. H. Pan, T. Utne, R. B. Jobson and D. F. Reinhold, Tetrahedron Lett., 1986, 27, 3845.
16. M. Verziu, M. Florea, S. Simon, V. Simon, P. Filip, V. I. Parvulescu and C. Hardacre, J. Catal., 2009, 263, 56.
17. W. Xie and X. Huang, Catal. Lett., 2006, 107, 53.
18. J. Zhu, Y. Chun, Y. Qin and Q. Xu, Microporous Mesoporous Mater., 1998, 24, 19.
19. A. Smahi, A. Solhy, H. E. Badaoui, A. Amoukal, A. Tikad, M. Maizi and S. Sebti, Appl. Catal., A, 2003, 250, 151.
20. C. Xu, J. Sun, B. Zhao and Q. Liu, Appl. Catal., B, 2010, 99, 111.
21. L. Gao, B. Xu, G. Xiao and J. Lv, Energy Fuels, 2008, 22, 3531.
22. S. Abelló, F. Medina, D. Tichit, J. Pérez-Ramírez, J. C. Groen, J. E. Sueiras, P. Salagre and Y. Cesteros, Chem.–Eur. J., 2005, 11, 728.
23. G. Centi and S. Perathoner, Microporous Mesoporous Mater., 2008, 107, 3.
24. F. Cavani, F. Trifiro and A. Vaccari, Catal. Today, 1991, 11, 173.
25. W. T. Reichle, J. Catal., 1985, 94, 547.
26. G. Gelbard, O. Bres, R. Vargas, F. Vielfaure and U. Schuchardt, J. Am. Oil Chem. Soc., 1995, 72, 1239.
27. A. H. Padmasri, A. Venugopal, V. Durga Kumari, K. S. Rama Rao and P. Kanta Rao, J. Mol. Catal. A: Chem., 2002, 188, 255.
28. J. Pérez-Ramíraz, G. Mul and J. A. Moulijn, Vib. Spectrosc., 2001, 27, 75.
29. A. Alejandre, F. Medina, P. Salagre, X. Correig and J. E. Sueiras, Chem. Mater., 1999, 11, 939.
30. P. S. Kumbhar, J. Sanchez-Valente and F. Figueras, Tetrahedron Lett., 1998, 39, 2573.
31. M. Del Arco, P. Malet, R. Trujillano and V. Rives, Chem. Mater., 1999, 11, 624.
32. L. Zou, X. Xiang, J. Fan, F. Li and D. G. Evans, J. Phys. Chem. Solids, 2008, 69, 1075.
33. R. Hua, Z. Jia, D. Xie and C. Shi, Mater. Res. Bull., 2002, 37, 1189–1195.
34. D. S. Jacob, L. Bitton, J. Grinblat, I. Felner, Y. Koltypin and A. Gedanken, Chem. Mater., 2006, 18, 3162.
35. F. Heras, J. Dufour, A. López-Delgado, C. Negro and F. López-Mateos, Environ. Eng. Sci., 2004, 21, 583.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c1cy00022e

---