On the synergy effect in moo3–fe2(moo4)3 catalysts

On the Synergy Effect in MoO3–Fe2(MoO4)3 Catalysts for Methanol Oxidation to Formaldehyde Methanol oxidation to formaldehyde was studied over a series of Fe–Mo–O catalysts with various MoFe atomic ratio and the end compositions Fe2O3 and MoO3. The activity data show that the specific activity passes through a maximum with increase of the Mo content and is the highest for Fe2(MoO4)3. The selectivity to formaldehyde, on the other hand, increases with the Mo content in the catalyst. A synergy effect is observed in that a catalyst with the MoFe ratio 2.2 is almost as active as Fe2(MoO4)3 and as selective as MoO3. Imaging of a MoO3 Fe2(MoO4)3 catalyst by SEM and TEM shows that the two phases form separate crystals, and HRTEM reveals the presence of an amorphous overlayer on the Fe2(MoO4)3 crystals. EDS linescan analysis in STEM mode demonstrates that the MoFe ratio in the amorphous layer is 2.1 in the fresh catalyst and 1.7 in the aged catalyst.

ORIGINAL PAPER On the Synergy Effect in MoO 3 –Fe 2 (MoO 4 ) 3 Catalysts for Methanol Oxidation to Formaldehyde Emma So ¨ derhjelm Æ Matthew P. House Æ Neil Cruise Æ Johan Holmberg Æ Michael Bowker Æ Jan-Olov Bovin Æ Arne Andersson Ó Springer Science+Business Media, LLC 2008 Abstract Methanol oxidation to formaldehyde was studied over a series of Fe–Mo–O catalysts with various Mo/Fe atomic ratio and the end compositions Fe 2 O 3 and MoO 3 . The activity data show that the specific activity passes through a maximum with increase of the Mo content and is the highest for Fe 2 (MoO 4 ) 3 . The selectivity to formaldehyde, on the other hand, increases with the Mo content in the catalyst. A synergy effect is observed in that a catalyst with the Mo/Fe ratio 2.2 is almost as active as Fe 2 (MoO 4 ) 3 and as selective as MoO 3 . Imaging of a MoO 3 / Fe 2 (MoO 4 ) 3 catalyst by SEM and TEM shows that the two phases form separate crystals, and HRTEM reveals the presence of an amorphous overlayer on the Fe 2 (MoO 4 ) 3 crystals. EDS line-scan analysis in STEM mode demon- strates that the Mo/Fe ratio in the amorphous layer is *2.1 in the fresh catalyst and *1.7 in the aged catalyst. The enrichment of Mo at the catalyst surface is confirmed by XPS data. Raman spectra give evidence for the Mo in the amorphous material being in octahedral coordination, which is in contrast to the crystalline Fe 2 (MoO 4 ) 3 bulk structure where Mo has tetrahedral coordination. X-ray diffraction (XRD) analysis gives no support for the for- mation of a defective molybdate bulk structure. The results presented give strong support for the Mo rich amorphous structure being observed on the Fe 2 (MoO 4 ) 3 crystal sur- faces being the active phase for methanol oxidation to formaldehyde. Keywords Methanol oxidation Á Formaldehyde Á Fe–Mo–O catalysts Á Iron molybdate Á XRD Á Electron microscopy Á SEM Á TEM Á HRTEM Á STEM–EDS Á XPS Á Raman spectroscopy 1 Introduction Formaldehyde is a reactive intermediate, which is used for the production of a large number of products [1, 2]. The largest amounts of formaldehyde are used to produce resins (con- densates) with urea, phenol and melamine, which are used for the production of adhesives and impregnating resins. Another market is the manufacture of molding compounds for surface coating. Formaldehyde is also used as an intermediate in the production of a variety of chemicals where the most important are polyacetals, MDI, 1,4-butanediol and polyols. The world consumption of formaldehyde in the form of its solution with water (37% HCHO) was about 28 million tons in the year 2005, and the present growth rate has been estimated to be around 3–4% per year [2]. Methanol and air are the raw materials for commercial production of formaldehyde in two competing technologies, E. So ¨ derhjelm Á N. Cruise Perstorp Specialty Chemicals AB, Process and Catalyst Development, 284 80 Perstorp, Sweden M. P. House Á M. Bowker School of Chemistry, Main Building, Cardiff University, Cardiff CF10 3AT, UK J. Holmberg Perstorp Specialty Chemicals AB, Perstorp Formox, 284 80 Perstorp, Sweden J O. Bovin Division of Polymer and Materials Chemistry, Department of Chemistry, Lund University, Chemical Center, P.O. Box 124, 221 00 Lund, Sweden A. Andersson (&) Department of Chemical Engineering, Lund University, Chemical Center, P.O. Box 124, 221 00 Lund, Sweden e-mail: arne.andersson@chemeng.lth.se 123 Top Catal DOI 10.1007/s11244-008-9112-1 which are referred as the silver process and the oxide pro- cess, respectively. The silver process operates at methanol- rich conditions with silver as catalyst, while the oxide pro- cess uses an iron molybdate catalyst under methanol-lean conditions. The first use of a silver catalyst was patented in 1910 [1]. In 1931, Adkins and Peterson [3]reportedthat methanol was selectively oxidised to give formaldehyde over an oxide catalyst with equimolar amounts of molyb- denum and iron. Although a catalyst with excess Mo gave lower activity, it was more selective to formaldehyde. However, it was not until the 1950s that that the molybdate catalyst got commercial importance. The basic chemical composition of the oxide catalysts has practically been the same over the years. It is well known that the fresh catalyst consists of two crystalline phases, namely MoO 3 and Fe 2 (MoO 4 ) 3 [4–9]. In spite of this fact, there have been several process improvements since the early 1960s as shown in Fig. 1 [2]. The intro- duction of gas recirculation from the absorber allowed the methanol inlet concentration to be increased from 6.5 to 7.5 vol.%. A few years later exhaust catalyst systems (ECS) became common, leading to increased steam production but no increase in productivity. Replacement of the gran- ular catalyst with ring-shaped catalysts allowed the gas flow rate to be increased and so the productivity. Dilution of the first part of the catalyst with inert rings (mixed layer) gave better temperature control and allowed increase of both the gas velocity and the methanol concentration up to 8.5 vol.%. A further development was pressurisation of the plants (0.3 bar g). More recently a refined catalyst activity profile (CAP) is used in new plants, allowing the heat of reaction to be properly distributed along the length of the reactor tube and making possible operation with a metha- nol concentration at the inlet of *10 vol.%. The durability of the catalyst depends on several factors including the methanol and oxygen concentrations, the temperature and the pellet diameter [10]. In practice it varies between 8 and 18 months depending on the oper- ating conditions and tolerances. By now, it is well established that the catalyst deactivates because it looses Mo during operation due to the formation of volatile spe- cies [9–13], causing lower activity and selectivity as well as increased pressure drop as molybdena needles condense in the lower part of the reactor. Another deactivation cause is sintering of the catalyst in the hot spot [11, 14, 15]. The deactivated catalyst consists mainly of Fe 2 (MoO 4 ) 3 , but contains as well the worse performing Fe 2 O 3 and FeMoO 4 phases [9, 11, 13, 16, 17]. It has been reported that the catalyst must contain both MoO 3 and Fe 2 (MoO 4 ) 3 for being active, selective and long- lasting [8, 9, 12, 13, 18, 19]. Different explanations have been given in the literature for the observation that both phases are required. One explanation is that Fe 2 (MoO 4 ) 3 is the active phase [6, 7, 20] and MoO 3 is required to secure that no iron rich phase is formed [12, 13, 21–23], which is also the opinion expressed in a comprehensive review to summarize the present knowledge about the active phase [4]. Other investigators have proposed that the active phase is Fe 2 (MoO 4 ) 3 with excess Mo in the structure. Fagherazzi and Pernicone [24] suggested from X-ray diffraction (XRD) data that the active ferric molybdate is Fe-defective due to Mo 6? substitutes for some of the Fe 3? ions in octahedral coordination and additional oxygen goes into interstitial positions. A similar conclusion was drawn by Leroy et al. [25] considering combined XRD and Raman results. According to Pernicone [8] the active structure for methanol oxidation has the composition Fe 2-3x (Mo 1?x O 4?1.5x ) 3 with x B 1/9 and usually below 0.05, consisting of nanoregions of MoO 6 octahedra included in the Fe 2 (MoO 4 ) 3 structure. Some researchers have concluded that the active phase is a solid interstitial solution of MoO 3 in the Fe 2 (MoO 4 ) 3 lattice [5, 26, 27], rather than being a substitution compound. Here it is also worth pointing out that Massarotti et al. [28] excluded any solubility of MoO 3 in Fe 2 (MoO 4 ) 3 , because they obtained very similar lattice-constant values for the stoichiometric molybdate and the molybdate in non-stoi- chiometric samples prepared by solid-state synthesis. Moreover, considering IR spectra, Sun-Kuo et al. [19] draw the conclusion that the active material consists of a dis- persed polymeric amorphous structure, and Aruanno and Wanke [14] have proposed that the activity is due to for- mation of a Mo rich surface on the Fe 2 (MoO 4 ) 3 crystals. Briand et al. [29] observed that bulk molybdates and supported molybdenum oxide catalysts possess similar activity and turnover frequency, indicating that the surfaces 0 5 10 15 20 25 Time Productivity (kg 37% HCHO/tube/day) Gas recirculation Ring-shaped catalyst Mixed layer Pressurization Refined activity profile Fig. 1 Gains in the productivity of formaldehyde per reactor tube and day since the early 1960s Top Catal 123 of bulk molybdates may be composed only by molybdenum oxide species in a two-dimensional overlayer. Besides being a dopant entering in small amount the interstial positions in the Fe 2 (MoO 4 ) 3 structure, Trifiro ` et al. [17] have pointed out that Mo in excess is necessary for improvement of the mechanical properties of the catalyst and in the process for securing efficient reoxidation of formed ferrous molybdate. In light of the disagreement in previous literature on the role of excess Mo for the performance of iron molybdate in methanol oxidation, the purpose of our work is to present new findings to clarify the origin of the observed synergy between the two catalyst constituents MoO 3 and Fe 2 (MoO 4 ) 3 . 2 Experimental 2.1 Catalyst Preparation A series of Fe–Mo–O catalysts was prepared with the atomic Mo/Fe ratios equal to 0.2, 0.5, 1.0, 1.5 and 2.2 starting from solutions of 100 ml ammonium heptamo- lybdate (BDH C 99%, 0.254 M for the Mo/Fe 1.5 catalyst) and 50 ml iron nitrate (BDH C 98%, 0.338 M). The molybdate solutions were acidified to *pH 2 using nitric acid (Fisher, Laboratory Grade), before drop wise addition of iron nitrate with stirring at 60 °C giving canary yellow precipitates, which were evaporated to near dryness at 90 °C. The resulting solids were then dried at 120 °C overnight before being calcined in air at 500 °C for 48 h. The single oxides of Fe 2 O 3 (Aldrich C 99%) and MoO 3 (BDH C 99.5%) used were commercially sourced. For comparative purposes the pure Fe 2 (MoO 4 ) 3 phase was prepared by solid-state reaction comprising mixing and grinding stoichiometric amounts of Fe 2 O 3 and MoO 3 , heating at 600 °C for 16 h, followed by a second grinding and heating for another 16 h, now at 650 °C. Some of the samples being characterized in this work have been supplied by Perstorp Formox. 2.2 Catalyst Characterisation The surface areas of the catalysts were measured either using a Micromeritics Gemini 2360 or a CE instruments QSurf M1. A five or six point BET method was used with adsorption of nitrogen at liquid nitrogen temperature and subsequent desorption at room temperature. All samples were degassed at 150 °C for 1 h. XRD was performed using either an Enraf Nonus FR590 fitted with a hemispherical analyser, or, a Seifert XRD 3000 TT diffractometer. In both cases Ni-filtered Cu Ka radiation was used. Fourier transform Raman (FT-Raman) spectra were recorded on a Bruker IFS66 FTIR spectrometer equipped with a Bruker FRA106 FT-Raman device, a Nd:YAG-laser and a germanium diode detector. The laser power was 100 mW, the resolution was 4 cm -1 and 400 scans were col- lected for each spectrum. XPS analysis was performed on a Kratos XSAM 800 spectrometer using Al Ka X-ray radiation (1486.6 eV). Quantifications were made using a linear background and instrumental sensitivity factors. Charging effects were corrected for by adjusting the main C 1s peak to a position of 285.0 eV. The anode was operated at an accelerating voltage of 13 kV and a current of 19 mA. The pass energy was 80 eV and the residual pressure in the spectrometer was 10 -8 torr, or, lower. SEM images were recorded on a JEOL 840A micro- scope using a tungsten filament and a voltage of 20 keV. EDS analyses were made using an Oxford instruments analyser with the INCA software. TEM imaging was performed using a JEOL 3000F microscope, operating at an acceleration voltage of 300 keV. The used EDS analyser from Oxford instruments was equipped with the INCA software. Linescanning was per- formed using EDS in STEM mode. Several dots in a line were analysed on each crystal, starting in the vacuum and then gradually moving onto the crystal. The typical dis- tance between the dots was between 20 and 30 nm. 2.3 Activity Measurements Activity and selectivity measurements on the laboratory prepared samples were made on a pulse flow micro reactor system [23]. The reactor basically consists of a U-tube mounted vertically within a Phillips PU 4500 GC oven with gas continuously flowing over the bed. A small amount of the outlet gas stream is monitored by a Hiden Analytical Hal 201 quadruple mass spectrometer, while the rest of the gas is vented via a Leybold Heraeus Trivac rotary pump. In this work, a heated gas with 10 vol.% O 2 in He (BOC) was flowed over 0.5 g catalyst at a rate of 30 ml/min (STP) with 1 ll methanol injections being made every 2 min, while the temperature in the furnace was ramped from *150 to *380 °C. Before each run, methanol injections were made over a bypass to account for the daily drift in the mass spectrometer. Comparative activity measurements on fresh and used commercial samples were made under adiabatic conditions in a stainless-steel micro reactor with a diameter of 21 mm. A flow with 1.75 vol.% methanol in dry 25 l/min (STP) air with an inlet temperature of 260 °C was passed over 13 g of 1 mm particles of crushed and sieved catalyst. The conversion of methanol was calculated comparing the signal from a Ber- nath Atomic 3006 FID-analyser before and after the reactor. Top Catal 123 3 Results 3.1 Influence of the Phase Composition on the Catalytic Performance The laboratory prepared Fe–Mo–O samples were tested in a pulse flow micro reactor system for their performance in methanol oxidation. In Table 1 are the metal compositions and the specific surface areas of the prepared samples lis- ted. XRD and Raman spectroscopy showed that the two single cation catalysts are single phase haematite Fe 2 O 3 (JCPDS file no. 33-664) and orthorhombic MoO 3 (JCPDS file no. 35-609), respectively [30]. The catalysts with the Mo/Fe ratios 0.2 and 0.5 consist of Fe 2 O 3 and Fe 2 (MoO 4 ) 3 (JCPDS file no. 31-642) [30]. In the samples with the Mo/ Fe ratios 1.0 and 1.5 only diffraction peaks from mono- clinic Fe 2 (MoO 4 ) 3 were observed. The stoichiometry of the former sample, however, dictates that there must be an additional iron phase, presumably Fe 2 O 3 , which is either amorphous or highly dispersed. The catalyst with Mo/Fe equal to 2.2 is a mixture of Fe 2 (MoO 4 ) 3 and MoO 3 . The catalysts below the stoichiometric level (Mo/Fe \ 1.5) presented a brown colour, the Mo/Fe 1.5 catalyst presented a yellow colour, while the Mo/Fe 2.2 catalyst was light green. For comparing the catalytic activity of the samples, methanol conversions at 180 °C are given in Table 1.Itis seen that the samples with 0.2 and 0.5 Mo/Fe ratios give the highest conversions, mainly due to their high surface areas. For a better comparison, the conversions should be normalised with respect to surface area. The normalised values in Table 1 indicate that the two samples with the Mo/Fe ratios 1.5 and 2.2 are the most active preparations per unit surface area. To better account for the differences in methanol conversion, first order surface area normalised rate constants were calculated. The values clearly show that the activity increases when the Mo/Fe ratio is increased from 0 (Fe 2 O 3 ) up to 1.5 (Fe 2 (MoO 4 ) 3 ). With further increase of the Mo/Fe ratio, the activity declines. Concerning the selectivity to formaldehyde at high meth- anol conversion the data in Table 1 for 90% conversion, show that the selectivity steadily increases with the Mo content in the catalyst. Of the pure phases, Fe 2 O 3 does not produce any formaldehyde, whereas the selectivity on Fe 2 (MoO 4 ) 3 and MoO 3 is 73% and 90%, respectively. Considering both activity and selectivity, the sample with Mo/Fe = 2.2 is outstanding, being almost as active as the pure Fe 2 (MoO 4 ) 3 and as selective as the pure MoO 3 . Also in industrial operation of Fe–Mo–O catalysts there is an optimal Mo/Fe ratio for the catalyst to perform well with good activity, selectivity and durability [9]. In Fig. 2 are the activities of used commercial catalysts compared with data for the corresponding freshly prepared catalysts. The used catalyst samples had been collected from the Table 1 Specific surface area, activity and selectivity of prepared Fe–Mo–O catalysts a Catalyst Mo/Fe ratio Specific surface area (m 2 /g) Conversion at 180 °C (%) Activity (conversion/m 2 surface area) First order rate constant k (cm 3 /min/m 2 ) Selectivity (%) to formaldehyde at 90% conversion b 0 (Fe 2 O 3 ) 2.1 *2 1.90 0.577 0 (322 °C) 0.2 55.4 55 1.99 0.865 18 (204 °C) 0.5 38.7 50 2.58 1.075 27 (210 °C) 1.0 16.3 38 4.66 1.760 47 (244 °C) 1.5 7.8 35 8.97 3.314 73 (249 °C) 2.2 6.7 29 8.66 3.067 90 (256 °C) ? (MoO 3 ) 1.0 *2 4.00 1.212 90 (377 °C) a Activity and selectivity measured in a pulse flow reactor injecting 1 ll liquid methanol every 2 min into a 30 ml/min gas flow with 10 vol.% O 2 in He, passing over 0.5 g catalyst while ramping the temperature from *150 to *380 °C (see Section ‘‘Experimental’’ ) b The temperature giving 90% methanol conversion on 0.5 g catalyst is given within brackets 0 5 10 15 20 25 30 35 40 45 Methanol conversion (%) Inlet layer Fresh catalyst Outlet layer Fresh catalyst Outlet layer Used catalyst Inlet layer Used catalyst Fig. 2 Conversion of methanol as measured in an adiabatic micro reactor (21 mm in diameter) over freshly prepared commercial catalysts and the corresponding used samples. The used samples had been collected from the inlet and outlet parts of an industrial multitube reactor after a full lifetime cycle. Reaction conditions: 1.75 vol.% methanol in dry 25 Nl/min air with an inlet temperature of 260 °C. The amount of catalyst in the micro reactor was 13 g of 1 mm particles of the crushed and sieved catalyst Top Catal 123 upper and lower part of the reactor, respectively, after the operation of a catalyst load in an industrial reactor had been terminated due to normal ageing. According to XRD the inlet fraction consisted mainly of Fe 2 (MoO 4 ) 3 , and elemental analysis indicated the presence of another 0.55 wt.% of Fe 2 O 3 in agreement with the catalyst surface being covered with a thin reddish brown layer. The fraction from the outlet consisted of the MoO 3 and Fe 2 (MoO 4 ) 3 phases, however, the Mo/Fe ratio was higher than in the unused catalyst and needle-like MoO 3 crystals were observed on the catalyst rings. As previously has been explained, Mo in the upper part of the catalyst forms volatile species with methanol [9, 10], which species decompose and condense as needles in the lower part of the bed [9, 11]. Thus, the activity data in Fig. 2 for industrial catalysts are in general agreement with the data in Table 1 for the laboratory prepared samples. The deactivation of the catalyst at inlet conditions is due to the surface being coated with a thin layer of Fe 2 O 3 , which according to Table 1 has lower activity and selectivity for methanol oxidation. A deacti- vation cause for the catalyst at the outlet of the reactor is that it contains more MoO 3 compared to the fresh catalyst. The data in Table 1 shows that the pure MoO 3 is less active than the composition Mo/Fe = 2.2 with both MoO 3 and Fe 2 (MoO 4 ) 3 . 3.2 Catalyst Characterisation with Electron Microscopy and XPS MoO 3 /Fe 2 (MoO 4 ) 3 catalysts were characterised with SEM, TEM, HRTEM and STEM–EDS before and after use in an industrial reactor. SEM shows the presence of large plate- like MoO 3 crystals being about 10 lm in size (Fig. 3). The crystal composition was verified by EDS analysis and, moreover, the observed crystal habit is typical of orthorhombic MoO 3 [31]. Generally it was observed that the MoO 3 crystals are more frequent in fresh catalyst than in used catalyst. Besides the plate-like crystals, the major part of the catalysts consists of smaller crystals, which appear to be randomly close packed as Fig. 3 shows. To gain information about the small crystals, samples were investigated by TEM and EDS. Most of the observed crystals are Fe 2 (MoO 4 ) 3 but also some are MoO 3 . The two types of crystals are in the same size range as Fig. 4 shows. Fe 2 (MoO 4 ) 3 presents crystals which mostly are rectangular/ elliptic, while the MoO 3 crystals are thin and plate-like. HRTEM imaging of the Fe 2 (MoO 4 ) 3 crystals revealed a 5–10 nm thick amorphous surface structure, extending around the crystals. The amorphous structure, which was observed both in fresh and aged samples, can be seen on the edges of the crystal in Fig. 5. Fig. 3 SEM image of a fresh MoO 3 /Fe 2 (MoO 4 ) 3 catalyst Fig. 4 TEM image of a fresh MoO 3 /Fe 2 (MoO 4 ) 3 catalyst Fig. 5 A HRTEM image of a typical Fe 2 (MoO 4 ) 3 crystal with an amorphous layer extending around the edge of the crystal Top Catal 123 The very surface of the Fe 2 (MoO 4 ) 3 crystals in a freshly prepared and an aged fraction of the same catalyst were analysed using EDS in STEM mode. Line scans were col- lected starting in the vacuum outside the edge of the crystal and then gradually moving in onto the bulkier parts of the material. By using this technique, it is expected to get information about any difference in elemental composition between the surface and the bulk of the crystal. For the wedge-like crystal terminations, the surface composition should be analysed at the edge of the crystal and with increasing distance from the edge, the analyses should show an increasing contribution from the bulk as the thickness of the amorphous layer being only about 5–10 nm. The ana- lysed Mo/Fe ratios are plotted in Fig. 6 against the distance from the edge of the analysed crystal. For the unused cat- alyst the Mo/Fe ratio is the highest at the edge, where the ratio is *2.1. With increase of the distance from the edge, the Mo/Fe ratio drops to a value of about 1.5. Compared with the fresh catalyst, the data for the corresponding aged sample shows a considerably lower Mo/Fe ratio on the edge (*1.7) and an identical value 1.5 further from the edge. The measured ratio 1.5 for the bulk is in perfect agreement with the stoichiometry of the Fe 2 (MoO 4 ) 3 crystal. As a complement to the EDS point analyses performed in STEM mode on separate Fe 2 (MoO 4 ) 3 crystals, XPS analy- ses were performed on a number of MoO 3 /Fe 2 (MoO 4 ) 3 preparations to give information about the average Mo/Fe ratio in the surface region. Figure 7 shows a comparison of Fresh Catalyst 1 1,2 1,4 1,6 1,8 2 2,2 2,4 0,0E+00 5,0E-08 1,0E-07 1,5E-07 2,0E-07 2,5E-07 Distance from crystal edge (m) Mo:Fe ratio crystal 1 crystal 2 crystal 3 crystal 4 crystal 5 crystal 6 crystal 7 crystal 8 Aged Catalyst 1,00 1,20 1,40 1,60 1,80 2,00 2,20 2,40 0,00E+00 5,00E-08 1,00E-07 1,50E-07 2,00E-07 2,50E-07 Distance from crystal edge (m) Mo:Fe ratio crystal 1 crystal 2 crystal 3 crystal 4 crystal 5 crystal 6 crystal 7 Fig. 6 The Mo/Fe ratios as determined by STEM–EDS line scan analysis on a Fe 2 (MoO 4 ) 3 crystal in (upper figure) a freshly prepared MoO 3 / Fe 2 (MoO 4 ) 3 catalyst and (lower figure) the corresponding catalyst after ageing in an industrial reactor Top Catal 123 the bulk ratio with the corresponding surface ratio as determined by XPS. In agreement with the EDS analyses, the XPS data clearly shows a general trend, namely that the catalyst surface is richer than the bulk in Mo. 3.3 XRD and Raman Characterisation of the Fe 2 (MoO 4 ) 3 phase Figure 8a shows an overlay of the XRD patterns of a MoO 3 /Fe 2 (MoO 4 ) 3 catalyst and a phase pure Fe 2 (MoO 4 ) 3 prepared by solid-state reaction from stoichiometric amounts of MoO 3 and Fe 2 O 3 . The difference pattern in Fig. 8b clearly exhibit peaks only from orthorhombic MoO 3 , indicating no difference in unit cell between the phase pure Fe 2 (MoO 4 ) 3 and the molybdate in the catalyst with excess MoO 3 . Included in the figure is a spectrum recorded for a pure MoO 3 sample, showing very intense (0 k 0) peaks due to the preferred orientation of the plate- like crystals in the sample holder. The most intense peak is (040), while in the difference pattern the (021) peak is the most intense for the MoO 3 in the catalyst matrix. According to data calculated from the unit cell (JPDS file no. 35-609) [30], the (021) peak should be the most intense peak for a non-oriented sample, which obviously is the case for the catalyst sample where the MoO 3 crystals have no preferential orientation in a matrix of Fe 2 (MoO 4 ) 3 . The Raman spectrum of a deactivated molybdate catalyst from a full-scale reactor is shown in Fig. 9 together with the spectrum of pure MoO 3 . According to XRD and atomic absorption the used catalyst consists of Fe 2 (MoO 4 ), and the recorded Raman spectrum is in perfect agreement with spectra reported in the literature for the pure Fe 2 (MoO 4 ) 3 phase [29, 32, 33]. The spectrum in Fig. 9 of the pure MoO 3 shows strong bands at 996, 818, 666 and 284 cm -1 together with a number of less intense bands in the region below 500 cm -1 . In the spectrum of the Fe 2 (MoO 4 ) 3 phase, bands are seen at 988, 967, 934, 818, 781 and 348 cm -1 . Thus, it is clearly seen that the molybdate spectrum does not include any contribution from crystalline MoO 3 because there are no bands at 666 and 284 cm -1 and, moreover, the bands at 988 and 818 cm -1 are of equal size, whereas in MoO 3 the band at 818 cm -1 is more than twice as large as the band at 996 cm -1 . It has been suggested for Fe 2 (MoO 4 ) 3 that the two bands at 988 and 818 cm -1 can be from some minor octahedrally coordinated Mo species [32] in contrast to the dominant Mo species, which is tetrahedral in a perfect Fe 2 (MoO 4 ) 3 lattice [24]. 4 Discussion The catalytic data in Table 1 reveal a pronounced synergy effect in that an atomic Mo/Fe ratio above the stoichiom- etric ratio for Fe 2 (MoO 4 ) 3 is needed for a catalyst to be 0 0.5 1 1.5 2 2.5 3 3.5 123456789101112 Catalyst # Mo/Fe ratio Bulk Surface Fig. 7 Comparison of the Mo/Fe bulk ratios with the corresponding ratios determined by XPS for a number of MoO 3 /Fe 2 (MoO 4 ) 3 preparations 0 20 40 60 80 100 120 0 5 10 15 20 25 30 Theta ( ˚ ) XRD intensity Pure iron molybdate Commercial fresh catalyst * * * * MoO 3 [0 k 0] peaks (a) -20 0 20 40 60 80 100 120 0 5 10 15 20 25 30 Theta ( ˚ ) Difference in XRD intensity Difference MoO3 (020) (110) (040) (021) (111) (060) (200) (061) (002) (081) (112) (211) (b) Fig. 8 A comparison of the XRD patterns recorded for a MoO 3 / Fe 2 (MoO 4 ) 3 catalyst and the pure Fe 2 (MoO 4 ) 3 phase prepared by solid state reaction (see Section ‘‘Experimental’’). The upper figure (a) shows an overlay of the two diffractograms. In the lower figure (b) is shown the resulting difference pattern obtained by subtracting the XRD for the pure phase from that of the catalyst. Here it is seen that the remaining peaks in the difference pattern correspond well with the XRD recorded for orthorhombic MoO 3 . The difference peaks have been indexed according to the JPDS file no. 35-609 [30] Top Catal 123 both active and selective to formaldehyde formation. This result is in general agreement with previous reports, although these in most cases have not considered both activity and the selectivity at high methanol conversion in the whole range of compositions from iron to molybdenum oxide. A surface area normalised activity maximum for methanol oxidation has been reported for the ratio Mo/ Fe = 1.7 [18, 19], which is not in conflict with the data in Table 1 showing that the catalysts with the Mo/Fe ratios 1.5 and 2.2 present similar activity. Moreover, in another report [22] is described that the highest specific activity was observed for the stoichiometric composition Fe 2 (MoO 4 ) 3 . It has also been reported that too high Mo content in the Fe–Mo–O catalyst results in lower activity [11], which is in agreement with the data in Fig. 2 for a catalyst collected from the outlet of an industrial reactor. This catalyst with excess molybdena needles on the surface is less active than the corresponding unused catalyst. The observation in Table 1 that the normalised activity (k value) of the pure MoO 3 is almost a factor three lower than that for the sample with the Mo/Fe ratio 2.2 is in perfect agreement with a previous comparison of a Har- shaw MoO 3 /Fe 2 (MoO 4 ) 3 catalyst with a MoO 3 catalyst [34]. Concerning the selectivity to formaldehyde, the val- ues in Table 1 show a steady increase with the Mo content up to the Mo/Fe ratio 2.2 with a selectivity of 90% at 90% methanol conversion, which is identical with the value for the pure MoO 3 . A similar trend has been reported previ- ously [18, 23]. For instance, Kolovertnov et al. [18] reported that the selectivity was low in Fe rich samples, whereas samples with Mo/Fe ratios above 1.5 and the pure MoO 3 are highly selective to formaldehyde at high meth- anol conversion. In our work, we have chosen first order rate constants as a means to account for the differences in surface area and conversion among the catalysts (see Table 1). The same method has been adopted by others [18]. According to Machiels and Sleight [35], the kinetics for a large number of molybdates and MoO 3 follows a power-law rate model with about 0.5 order in the methanol concentration. Therefore, we also calculated the half order rate constants for the catalysts in Table 1. The values, however, are not shown as they perfectly confirmed the trend indicated by the first order rate constants. In general, several possibilities have to be considered to explain an observed synergy between phases. One of the possibilities is dual phase catalysis, where some of the reaction steps occur on one phase while following steps take part on another phase. An example here is the ammoxidation of propane over the Mo–V–Nb–Te–O system with the two phases designated M1 and M2, where propane reacts to give propene on M1 and the formed propene then readsorbs on M2, which is more selective than M1 for the transformation of propene to acrylonitrile [36]. In the case of methanol oxidation on MoO 3 /Fe 2 (MoO 4 ) 3 , such type of mechanism is unlikely since the reaction pathway to form- aldehyde involves no intermediate gaseous product and, moreover, the methoxy intermediate does not migrate over the surface as it is relatively strongly bound [23]. Another alternative to be considered is that the catalysis may occur on the grain boundaries between the phases as has been suggested for 3-picoline ammoxidation on vanadia phases [37]. However, the electron microscopy results in Figs. 3 and 4 show separate crystals of the MoO 3 and Fe 2 (MoO 4 ) 3 phases. Although in contact with each other, no intergrowth is observed of the crystals from the two phases. In a study of the reduction behaviours of MoO 3 , Fe 2 (MoO 4 ) 3 and their mixtures by in situ electron micros- copy in a CH 3 OH/He atmosphere [38], it was noted that the crystals in the mixtures remain as distinct phases with the same reduction behaviour as the individual phases. In view of the above results, it seems that the observed synergy effect is not primarily related to the grain boundaries. In other cases an observed synergy between two phases has been explained by oriented growth of one phase on another, e.g. in the case of Sb–Sn–O mixed oxides for propene oxidation [39]. However, our microscopic investigation showed no oriented growth of MoO 3 on Fe 2 (MoO 4 ) 3 , see Figs. 3 and 4. In some catalyst systems an activation process occurs under influence of the catalytic reaction leading to the for- mation of new phases, which may be more active and selective than the original composition [40]. This occur- rence is not the case in methanol oxidation on MoO 3 / 1003005007009001100 Wavenumber (cm -1 ) Raman intensity (a. u.) Fe 2 (MoO 4 ) 3 MoO 3 Fig. 9 The Raman spectra recorded for MoO 3 and a used catalyst sample consisting of Fe 2 (MoO 4 ) 3 Top Catal 123 Fe 2 (MoO 4 ) 3 catalysts. On the contrary, in practice a steady deactivation of the catalyst is observed with time-on-stream as the data in Fig. 2 confirms, which also concurs with previous results [9, 11]. During the deactivation, Fe 2 O 3 and FeMoO 4 may form [6, 9, 11]. The iron oxide has low activity and is unselective (Table 1), and FeMoO 4 has been reported to be selective with an activity comparable to that of MoO 3 [35]. HRTEM imaging of the Fe 2 (MoO 4 ) 3 phase in the catalysts revealed an amorphous surface structure on the edges of the crystals as shown in Fig. 5. In fact the same type of structure was observed by Gai and Labun [38], although they just briefly mention it as a note in their article when referring to a small area in one of the images. The investigators focused in their work on the bulk structures and their reduction. The EDS data in Fig. 6 shows that the amorphous structure on the fresh catalyst is rich in Mo (Mo/Fe *2.1) while the amor- phous material on the aged catalyst has a lower content (Mo/Fe *1.7). Considering there being a link between the ageing and the composition and the performance of the catalyst (Fig. 2), we believe that the active material is the amorphous structure with excess Mo as compared to the bulk Fe 2 (MoO 4 ) 3 structure. The fact that the Mo/Fe ratio is higher on the molybdate surface than in the bulk, moreover, is confirmed by the XPS analyses in Fig. 7. A similar trend has been observed by other investigators using EDS and XPS [12, 13, 27]. The Raman spectrum of the Fe 2 (MoO 4 ) 3 phase in Fig. 9 confirms the existence of an additional structure on the molybdate. In agreement with published spectra of the pure phase [29, 32, 33], Fig. 9 shows two bands at 988 and 818 cm -1 , respectively, appearing as shoulders on the strong bands at 967 and 781 cm -1 from the tetrahedrally coordi- nated Mo in the bulk [33]. In agreement with a previous assignment [32], the two shoulder bands can be from Mo in octahedral environment considering that MoO 3 with octa- hedrally coordinated Mo gives two bands at similar wavenumbers i.e. 996 and 818 cm -1 (Fig. 9). Also in IR, Fe 2 (MoO 4 ) 3 gives a weak band at 990 cm -1 [27, 32], which has been assigned to Mo in octahedral coordination [32]. Thus, from these facts it can be proposed that the Mo in the amorphous layer may be in octahedral coordination sur- rounded by six oxygen atoms. The formation of an amorphous surface layer on Fe 2 (MoO 4 ) 3 can be understood considering its crystalline bulk structure, which in idealized form is illustrated in Fig. 10 as built up by regular tetrahedra and octahedra. Looking at the structure it is seen that it is very open. Therefore, the formation of an amorphous overlayer can be a means for stabilizing the surface. It is not clear whether excess Mo is needed in the synthesis only to give an amorphous layer with a ratio Mo/Fe[2 in the catalyst, or, if the MoO 3 crystalline phase in the finished catalyst has an additional role to sustain the desired Mo/Fe ratio in the active structure during operation in methanol oxidation. Fig. 10 The ferric molybdate structure with Mo in tetrahedral and Fe in octahedral coordination shown in idealized form as built up by regular polyhedra. The red and yellow polyhedra have Mo and Fe, respectively, in the center, and oxygen in the corner positions. Each oxygen is shared by two polyhedra and therefore 2-coordinated. As indicated in the three figures, the structure is viewed along the [100], [010] and [001] directions, respectively Top Catal 123 Previously, it has been proposed that surplus Mo in the catalyst is needed for avoiding the formation of Fe 2 O 3 in the reoxidation of formed ferrous molybdate FeMoO 4 to the desired ferric molybdate Fe 2 (MoO 4 ) 3 with a higher Mo/Fe ratio [17, 21]. In view of the fact that SEM and TEM imaging shows separate MoO 3 and Fe 2 (MoO 4 ) 3 crystals with only physical contact (Figs. 3 and 4), it seems doubtful whether MoO 3 should have such a role. It is true that iron oxide is not formed until the Mo/Fe ratio in the catalyst approaches the value 1.5 and the catalyst is free from any crystalline MoO 3 [9]. However, a kinetic study of the Mo loss from MoO 3 /Fe 2 (MoO 4 ) 3 preparations has shown that the loss from the MoO 3 crystals occurs considerably faster than from the remaining Fe 2 (MoO 4 ) 3 [41], which is sup- ported by results in our previous deactivation study [9]. In several cases, it has been concluded that Fe 2 (MoO 4 ) 3 is the active phase in MoO 3 /Fe 2 (MoO 4 ) 3 catalysts [4, 6, 7, 20–22]. According to our results this is correct though incomplete. A better description is that the molybdate is a support for the active structure, which is an amorphous surface layer with a higher Mo/Fe ratio compared to the bulk and with Mo in octahedral coordination. This finding is in partial agreement with some of the earlier proposals, which however were based on indirect evidence. In these works the active material has been described as a Mo rich molybdate surface [14], a dispersed amorphous structure [19] and molybdenum oxide species forming a monolayer [29]. Concerning the previous indications of a defective molybdate bulk structure being formed when prepared in excess of molybdenum, we have found no evidence for the formation of either a solid interstial solution of MoO 3 in the Fe 2 (MoO 4 ) 3 lattice [5]orMo 6? substituting for some Fe 3? and extra oxygen entering the interstial positions [24]. The XRD patterns in Fig. 8 of a MoO 3 /Fe 2 (MoO 4 ) 3 cata- lyst and the stoichiometric Fe 2 (MoO 4 ) 3 phase, indicate no difference between the crystalline bulk structure of the ferric molybdate in the catalyst and the corresponding phase prepared without excess molybdenum. This finding is in agreement with previous reports [6, 7, 21, 28]. 5 Conclusions The present study of methanol oxidation to formaldehyde on Fe–Mo–O catalysts has demonstrated that the best performing catalyst is a mixture of the crystalline phases MoO 3 and Fe 2 (MoO 4 ) 3 . A synergy effect is observed in that the catalyst is almost as active as the pure Fe 2 (MoO 4 ) 3 , which is less selective, and as selective as the pure MoO 3 , which is less active. HRTEM imaging of a fresh MoO 3 /Fe 2 (MoO 4 ) 3 catalyst and a corresponding aged catalyst discloses that the active structure may be an amorphous overlayer on the surface of the crystalline ferric molybdate. Line-scan EDS analysis reveals that the Mo/Fe ratio in the amorphous surface is higher in the fresh catalyst than in the used catalyst. The Mo/Fe surface ratio for the latter approaches the value 1.5, although the surface still is amorphous. The Raman spec- trum of a catalyst consisting of Fe 2 (MoO 4 ) 3 shows bands indicating that in the amorphous layer the Mo is in octa- hedral coordination. A comparison of the XRD pattern of a MoO 3 / Fe 2 (MoO 4 ) 3 catalyst with that for the pure Fe 2 (MoO 4 ) 3 phase evidences that the crystalline bulk structure of ferric molybdate is the same in both samples, ruling out previous proposals of the formation of a defective molybdate with excess Mo in the crystal lattice. Acknowledgements Perstorp Specialty Chemicals AB and the EPSRC in the UK are acknowledged for support of a studentship to M.P.H. References 1. Reuss G, Disteldorf W, Grundler O, Hilt A (1988) In: Ullmann’s encyclopedia of industrial chemistry, 5th edn, vol A11, VCH Publishers, Weinheim, pp 619–651 2. Chrichton B (2006) In: Informally speaking (newsletter from Perstorp Formox), spring/summer 2006, pp 2–8. http://www. perstorpformox.com 3. Adkins H, Peterson WR (1931) J Am Chem Soc 53:1512 4. Soares APV, Portela MF, Kiennemann A (2004) Catal Rev 47:125 5. Abaulina LI, Kustova GN, Klevtsova RF, Popov BI, Bibin VN, Melekhina VA, Kolomiichuk VN, Boreskov GK (1976) Kinet Katal 17:1307 6. Carbucicchio M, Trifiro ` F (1976) J Catal 45:77 7. Alessandrini G, Cairati L, Forzatti P, Villa PL, Trifiro ` F (1977) J Less-Comm Met 54:373 8. Pernicone N (2003) CATTECH 7:196 9. Andersson A, Hernelind M, Augustsson O (2006) Catal Today 112:40 10. Popov BI, Bibin VN, Boreskov GK (1976) Kinet Katal 17:371 11. Burriesci N, Garbassi F, Petrera M, Petrini G, Pernicone N (1980) In: Delmon B, Froment GF (eds) Catalyst deactivation, Stud Surf Sci Catal, vol 6, Elsevier, Amsterdam, pp 115–126 12. Soares APV, Portela MF, Kiennemann A (2001) Catal Commun 2:159 13. Soares APV, Portela MF, Kiennemann A, Hilaire L (2003) Chem Eng Sci 58:1315 14. Aruanno J, Wanke S (1975) Can J Chem Eng 53:301 15. Aruanno J, Wanke S (1977) Can J Chem Eng 55:93 16. Ma YH, Kmiotek SJ (1988) J Catal 109:132 17. Trifiro ` F, Carbucicchio M, Villa PL (1998) Hyperfine Interact 111:17 18. Kolovertnov GD, Boreskov GK, Dzisko VA, Popov BI, Tarasova DV, Belugina GG (1965) Kinet Katal 6:1052 19. Sun-Kou MR, Mendioroz S, Fierro JLG, Palacios JM, Guerro- Ruiz A (1995) J Mater Sci 30:496 20. Kim TH, Ramachandra B, Choi JS, Saidutta MB, Choo KY, Song S-D, Rhee Y-W (2004) Catal Lett 98:161 21. Boreskov GK, Kolovertnov GD, Kefeli LM, Plyasova LM, Karakchiev LG, Mastikhin VN, Popov VI, Dzisko VA, Tarasova DV (1966) Kinet Katal 7:144 Top Catal 123 [...]... JCPDS International Centre for Diffraction data (1991) Powder diffraction file, Swarthmore, PA 31 Hansen S, Andersson A (1988) J Solid State Chem 75:225 32 Forzatti P, Villa PL, Ferlazzo N, Jones D (1982) J Catal 76:188 33 Hill CG Jr, Wilson JHIII (1990) J Mol Catal 63:65 34 Cheng W-H (1996) J Catal 158:477 35 Machiels CJ, Sleight AW (1982) In: Barry HF, Mitchell PCH (eds) Proceedings of the 4th international... 4th international conference on the chemistry and uses of molybdenum, Golden, Colorado, Climax Molybdenum Co, Ann Arbor, pp 411–414 36 Holmberg J, Grasselli RK, Andersson A (2004) Appl Catal A 270:121 37 Andersson A, Bovin J-O, Walter P (1986) J Catal 98:204 38 Gai PL, Labun PA (1985) J Catal 94:79 39 Figueras F, Forissier M, Lacharme JP, Portefaix JL (1985) Appl Catal 19:21 ´ 40 Nilsson J, Landa-Canovas... Labun PA (1985) J Catal 94:79 39 Figueras F, Forissier M, Lacharme JP, Portefaix JL (1985) Appl Catal 19:21 ´ 40 Nilsson J, Landa-Canovas AR, Hansen S, Andersson A (1999) J Catal 186:442 41 Dazhuang L, Jianhong Z, Huimin T, Dongxia H (1997) React Kinet Catal Lett 62:347 123 ... Alouche A, Ent¨ wistle C, Torncrona A (2002) Catal Lett 83:165 24 Fagherazzi G, Pernicone N (1970) J Catal 16:321 25 Leroy J-M, Peirs S, Tridot G (1971) C R Acad Sci Paris C 272:218 26 Popov BI, Pashis AV, Shkuratova LN (1986) React Kinet Catal Lett 30:129 27 Soares APV, Portela MF, Kiennemann A, Hilaire L, Millet JMM (2001) Appl Catal A 206:221 28 Massarotti V, Flor G, Marini A (1981) J Appl Cryst 14:64 . although they just briefly mention it as a note in their article when referring to a small area in one of the images. The investigators focused in their work on the bulk structures and their reduction analysed at the edge of the crystal and with increasing distance from the edge, the analyses should show an increasing contribution from the bulk as the thickness of the amorphous layer being only about. the plate- like crystals in the sample holder. The most intense peak is (040), while in the difference pattern the (021) peak is the most intense for the MoO 3 in the catalyst matrix. According