Solid State Ionics 177 (2006) 2241 – 2244 www.elsevier.com/locate/ssi
BIMEVOX as dense membrane in catalytic reactor (ME = Co, Cu, Ta) C. Pirovano b,⁎, A. Löfberg a , H. Bodet a,b , E. Bordes-Richard a , M.C. Steil b , R.N. Vannier b a
b
Laboratoire de Catalyse de Lille, UMR CNRS 8010, USTL-ENSCL, Cité Scientifique, 59655 Villeneuve d'Ascq Cédex, France Laboratoire de Cristallochimie et Physicochimie du Solide, UMR CNRS 8012, USTL-ENSCL, BP 90108, 59652 Villeneuve d'Ascq Cédex, France Received 13 July 2005; received in revised form 10 January 2006; accepted 11 January 2006
Abstract A catalytic dense membrane reactor allows to physically separate the oxygen feed from the reactant (hydrocarbon) feed with a catalytic membrane chosen among oxide ion conducting materials. The membrane plays a double role, it provides the oxygen needed for selective oxidation and acts as a catalyst. The catalytic properties of BIMEVOX (ME = Co, Cu, Ta) membranes were examined in the mild oxidation of propene and of propane. During the complex transient state observed when the surface is rough, the nature and distribution of products are different from those obtained with mirror-polished membranes in a former work. In particular, syngas is formed with propene as well as with propane, and it precedes the production of hydrogen and coke. The complex behaviour differs according to ME and seems to be related to the different nature of electron semi-conduction induced by each dopant. © 2006 Elsevier B.V. All rights reserved. Keywords: BIMEVOX; Catalytic dense membrane reactor; Partial oxidation; Syngas; Oxide ion conduction
1. Introduction The upgrading of C1–C3 alkanes is aimed at producing oxygenated intermediates like methanol, acetic acid, acrolein, etc. (mild oxidation), alkenes (oxidative dehydrogenation, oxidative coupling), or syngas (partial oxidation) [1]. The selective oxidation of hydrocarbons proceeds by a redox mechanism according to which the surface lattice oxygen (O2− ) of the catalyst is directly introduced in the adsorbed intermediate to give the oxidised products. This step is followed by the reoxidation of the reduced catalyst by dioxygen, usually cofed with the reactant in usual (fixed or fluidised bed) reactors. Because of the presence of gaseous O2 that may directly react with the hydrocarbon in the gas phase, and/or produce adsorbed oxygen species which are too electrophilic [2], carbon oxides are produced and the selectivity to products is not optimum. To avoid this, one mean is to separate the two steps of the redox mechanism. Such a “redox decoupling” [3] may be performed in time (pulse or periodic feed reactors), or in space (circulating ⁎ Corresponding author. E-mail addresses:
[email protected] (C. Pirovano),
[email protected] (E. Bordes-Richard). 0167-2738/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ssi.2006.01.025
fluid bed or membrane reactors). The driving force in a catalytic dense membrane reactor (CDMR) is the gradient of oxygen partial pressure between two compartments separated by the membrane [4–8]. At the high oxygen partial pressure side, O2 dissociates to O2− ions which diffuse through the membrane to the low oxygen partial pressure side in which the hydrocarbon is introduced. BIMEVOX oxides, well known for their high oxide ion conduction at moderate temperature (300–700 °C) [9], are good candidates for selective oxidation catalysis because of the presence of both Bi and V cations [1,3,7]. They derive from the parent compound Bi4V2O11 and are obtained by partial substitution of vanadium with a metal [9]. BICOVOX.10 for instance corresponds to Bi2V1−xCoxO5.5−3x/2 with x = 0.10. Although these materials are highly oxide ion conducting, their oxygen semipermeability to oxygen gas is poor, because of low rate of oxygen exchange at the surface (in contrast to the high O2− diffusion in the bulk) and low electronic conduction [10]. In a former study of propene oxidation using mirrorpolished membranes of BIMEVOX as well as of Bi1−xErxO1.5/ Ag cermet in a CDMR [7], we have demonstrated that a high oxygen flux is not a key property in mild oxidation. Products were those found in selective oxidation reactions (hexadiene, acrolein) although CO was also present in large amounts. The
2242
C. Pirovano et al. / Solid State Ionics 177 (2006) 2241–2244
conversion of propene was low (up to 2 mol%) because of the limited number of active sites. Here we report on the catalytic properties of some BIMEVOX (ME = Co, Cu, Ta) membranes with rough surface to make it more active. 2. Experiments Bi2V0.9Co0.1O5.35, Bi2V0.9Cu0.1O5.35 and Bi2V0.8Ta0.2O5.5 (BICOVOX, BICUVOX and BITAVOX, respectively) powders were synthesized from pure oxides as described elsewhere [9]. Membranes of BIMEVOX (∅ = 15 mm, thickness 1.7 mm) were sintered at 750 °C (BICOVOX, BICUVOX) or 875 °C (BITAVOX). Their relative density was higher than 95% after sintering. After polishing to obtain a good parallelism of the surfaces, the catalytically active side was polished using a 220 grit SiC paper [7]. Details on the reactor assembly and procedure can be found in Ref. [7]. The gas phase composition was monitored by mass spectrometry for the catalytic reaction and an oxygen gauge was used for permeation measurements. Air was flowed in the HOP (retentate) side, and the LOP (permeate) side was fed with He (permeation mode), or 1% hydrocarbon in He (both F = 50 cm3 mn− 1) (reaction mode) at 550–750 °C and P = 1 atm. The contact time was 2.3 s. The surface polarization at the LOP side was in situ determined by measuring the electric potential difference between two gold electrodes, a grid (counter electrode) and a tip (working electrode) on the HOP and LOP sides, respectively [7]. The purity of BIMEVOX powders was checked by X-ray diffraction (XRD). Both sides of the membranes were analysed before and after experiments by XRD, Scanning Electron Microscopy (SEM-EDX), Laser Raman Spectroscopy (LRS) and X-ray photoelectron spectroscopy (XPS).
ena were observed instead of steady state, and the mild oxidation products were replaced mostly by H2 and CO, and by CO2, H2O and methane in lower amounts. In the case of Cu and Co membranes (Fig. 1), the conversion amounts to 10 mol% in the first minutes, then it increases strongly up to 60 mol% (Co) or passes through a maximum (Cmax = 53 mol%) for Cu (Fig. 1a). The conversion remains high on BITAVOX (70–52 mol%). These phenomena are fastened as temperature increases. When a steady state is finally reached, hydrogen is quite the only product with carbon (coke). Plotting the H2 / CO ratio vs. time (Fig. 1b) shows that it increases strongly for Co and Ta while it reaches a steady lower value with Cu. After burning of the coke and regeneration of the membrane by feeding air in LOP at the same temperature, the same figures were recovered, which indicates that the process is reproducible. Analyses performed on both sides of the membranes after experiments showed that the structural integrity of the membranes was maintained. No metallic bismuth (often detected in powder catalysts after use in fixed bed reactor [11]), nor other phase that could account for a bulk decomposition, could be identified by XRD. Scanning electron microscopy showed only that the size of grains had slightly increased. XPS analysis of the surface revealed that the Bi / V atomic ratio and the oxygen stoichiometry remained quite the same, except in the case of BICOVOX where an increase of Co was noticed. As in the case of mirror-polished membranes [7], a low difference of potential between LOP and HOP membrane sides was measured due to the polarisation at the LOP side. Its variation was exactly the reverse of propene conversion vs. time. Therefore, and although the active sites are more numerous than on mirror-polished membranes, the polarisation remains high, so that enough oxygen species are available for the reaction to proceed without reduction of the bulk.
3. Results 3.2. Propane oxidation 3.1. Propene oxidation As expected when compared to mirror-polished membranes [7], the initial conversion of propene was considerably increased when the surface was depolished. Transient phenomTa
60
(a)
Co
40 Cu
20
20
0 Co
15
Ta Cu
10
(b) 0
500
1000
1500
2000
2500
3000
3500
4000
H2/CO
Conversion (%)
80
When propene was replaced by propane in LOP side, a transient behaviour of BIMEVOX was observed whatever the dopant, but the phenomena were delayed because of the lower reactivity of propane. Fig. 2 shows that this activity peak occurs
5 0 4500
Time (s) Fig. 1. Comparison of a) conversion of propene, and b) H2 / CO ratio for BICOVOX, BICUVOX and BITAVOX membranes at 700 °C vs. time.
C. Pirovano et al. / Solid State Ionics 177 (2006) 2241–2244
2243
Conversion (%)
60
(a) Ta
40
Co
20 Cu
0 8
Ta
6 Cu
4
(b) 0
500
1000
1500
2000
2500
3000
3500
4000
2
H2/CO ratio
10
Co
0 4500
Time (s) Fig. 2. Comparison of a) conversion of propane, and b) H2 / CO ratio for BICOVOX, BICUVOX and BITAVOX membranes at 700 °C vs. time.
later and later in the Ta b Cu b Co series and that the conversion, before and after it, is remarkably the same (10–12 mol%) for the three membranes. The duration of the activity peak is larger for Ta and two close maxima at 250 s (4.1 min) and 285 s (4.75 min) are detected (Fig. 2a). Three activity peaks (maxima at 23.6, 26.6 and 30.2 min) occur for BICOVOX after a constant activity for ca. 20 min. In contrast, only one peak is observed for Cu. At the very first seconds, a small amount of propane was oxidatively dehydrogenated to propene with coproduction of CO2 and quickly replaced by syngas for a time depending on ME. After the activity peak(s) carbon was detected on the LOP side of the membrane. The variation of H2 / CO ratio is related to time and duration of the activity peaks. As in the case of propene oxidation (H2 / CO = 1.7 and ∼6.0), H2 / CO amounts to 1.8 for a short (Ta: 1.3 min) to longer (Cu: 5.2 min; Co: 21 min) period. Another constant value which is held for few seconds is close to 3.0–3.2 and is found at 4 min (Ta) b 30 min (Co) but it does not change in the case of BICUVOX after 7.5 min. None of these ratios correspond to the stoichiometric formation of syngas (H2 / CO = 1.0 and 1.33 for propene and propane, respectively), which indicates that the reaction mechanism is
more complex. Fig. 3 shows the distribution of products in the case of BICOVOX at 700 °C. Apart from the first seconds when little propene and more CO2 appears, there is a steady production of H2 N CO, CH4 N H2ON CO2N C3H6 molecules during 21 min. H2 and CO are the main products during the activity peak but little methane remains produced all the time. Although mainly coke and H2 are formed at the final steady state where 10 mol% of propane are converted, the production of CO cannot be accounted for by the oxygen that could permeate through the membrane [7]. For example, at 3000 s, 300 ppm CO are produced (pCO = 30 Pa) which means that 150 ppm O2 (or 300 ppm O2−) have been consumed, while only 5 ppm O2 would permeate in the absence of hydrocarbon. Therefore, even in these strongly reducing conditions the polarisation is high enough and the membrane is still active. 4. Discussion The reasons why these phenomena occur and can be restored if the membrane is cleaned and regenerated are not yet understood. The first phenomenon to understand is the activity
1200
60 50
H2
800
40
600 400
30
CO2 CH4
Conversion
H2O
20
CO
200
10
0 0
500
1000
1500
2000
Conversion (%)
Pressure (Pa)
1000
2500
3000
3500
4000
4500
0 5000
Time (s) Fig. 3. Conversion of propane and distribution of products vs. time for BICOVOX at 700 °C.
2244
C. Pirovano et al. / Solid State Ionics 177 (2006) 2241–2244
peak(s) during which coke is formed as shown by the variation of H2 / CO ratio which becomes higher and higher as CO is replaced by coke. This reactivity decreases along Ta N Cu N Co in propane (or propene) oxidation. Conversely, the steady production of syngas is the highest on BICOVOX. The H2 / CO ratio increases step by step before it reaches a constant value corresponding to a steady conversion with production of coke. Chemical engineering parameters may be involved because the design of the reactor is far from being optimised. A hot spot as those commonly encountered with fixed bed of particles because of poor local heat transfer is unlikely and no particular increase of temperature was noticed during the activity peak. However in the dead volume of the reactor, gas-phase reactions certainly play a role by modifying the product distribution since H2 / CO does not fit the stoichiometry of the syngas formation from propene or propane. We assume that most phenomena are related to the nature and reactivity of the membranes, which mainly depend on (i) their surface texture, and (ii) their ionic and electronic conductivity, the latter properties being related to the ME cation [12]. First, the surface texture could be an important factor, for example if the grain boundaries are more or less numerous according to the sample. SEM showed that initially the size of domains was quite the same for Cu and Co samples, but that it was smaller for BITAVOX. Moreover, the polishing step performed to get a rougher and then more active surface is not a standardised process. However, if we assume in first approximation that the surface state of the three membranes is the same, it means that the same amount of hydrocarbon molecules in LOP and the same amount of O2 molecules in HOP are converted. This seems to be true during the oxidation of propane (conversion ≈10 mol%). Another common feature is the H2 / CO ratio which reaches similar values at the two first plateaus (∼1.8 and ∼3.0 respectively) on all membranes before it increases more (Co, Ta) or less (Cu) strongly during formation of coke. These two steps are probably related to definite restructurations of the surface yielding given amounts of CO and H2, the active sites displayed on LOP surface of the three membranes being the same for a while. However, the influence of the nature and the amount of ME are seen in the fact that the duration of the steady periods, the onset, duration and intensity of the peak(s) activity are different: the larger (and sooner) the activity peak, the higher H2 / CO and the coke produced. It may be assumed that during the sudden increase of conversion, the reservoir of active oxygen decreases. We may therefore consider that, contrary to mirror polished membranes, the supply of O2− at the LOP surface is not sufficient to restore the active phase. Although it is low as compared to ionic conductivity, the electronic conductivity of BICOVOX is p-type, that of BICUVOX is n-type, while it is quite null in the case of BITAVOX. The presence of several activity peaks (two for Ta and three for Co) is not yet understood, these different electronic properties could be responsible for the different dynamic behaviours.
5. Conclusion When the surface of membranes is rough, the conversion of the reactant is enhanced but transient phenomena occur during the oxidation of propene and of propane. The high conversion (as compared to that observed with mirror-polished surfaces) favours the formation of syngas instead of products of mild oxidation, as if the oxygen supply was not rapid enough with respect to the increase in activity. After these activity peak(s), the overall production of oxygen is decreased since hydrogen and coke are the main products. Meanwhile, there are several surface restructurations which give rise to temporary but definite arrangements and activity of atoms since the catalytic performance (H2 / CO, conversion) is similar whatever the membrane. The main difference of behaviour seems related to the ME cation which is responsible for the electronic type of conductivity and therefore for the O2 dissociation step in HOP and O2− transfer. As the gaseous mixture (HxCy + H2 + CO) is strongly reducing, the LOP surface may endeavour partial reduction, the extent of which may depend on ME cation. However, the membrane retains its integrity. Explanations are scarce for the moment and more work is needed. In situ analyses are planned in order to characterize the surface reactivity of membranes. Acknowledgements The Centre National de la Recherche Scientifique and the Région Nord-Pas de Calais are acknowledged for providing a grant to one of the authors (H.B.). The authors are also grateful to Françoise Ratajczak for her help in powder preparation. References [1] [2] [3] [4] [5] [6]
[7] [8] [9] [10]
[11] [12]
S. Albonetti, F. Cavani, F. Trifirò, Catal. Rev.-Chem. Eng. 38 (1996) 413. A. Bielanski, J. Haber, Catal. Rev.-Chem. Eng. 19 (1979) 1. E. Bordes, C.R. Acad. Sci., Paris, Série IIc, Chimie, 3 (2000) 725. R. DiCosimo, J.D. Burrigton, R.K. Grasselli, J. Catal. 102 (1986) 234. S. Azgui, F. Guillaume, B. Taouk, E. Bordes, Catal. Today 25 (1995) 291. C. Burello, B. Taouk, E. Bordes, Réc. Prog. Génie Procédés 55 (1997) 125 ibid.; C. Burello, B. Taouk, E. Bordes, Catal. Lett. 66 (2000) 129. A. Löfberg, S. Boujmiai, E. Capoen, C. Steil, C. Pirovano, R.N. Vannier, G. Mairesse, E. Bordes-Richard, Catal.Today 91–92 (2004) 81. A. Löfberg, H. Bodet, C. Pirovano, C. Steil, R.N. Vannier, E. BordesRichard, Catal. Today, in press. F. Abraham, J.C. Boivin, G. Mairesse, G. Nowogrocki, Solid State Ionics 40 (1990) 934. R.N. Vannier, S.J. Skinner, R.J. Chater, J.A. Kilner, G. Mairesse, Solid State Ionics 160 (2003) 85 ibid.; R.N. Vannier, S.J. Skinner, R.J. Chater, J.A. Kilner, G. Mairesse, Solid State Ionics 160 (2003) 327. A. Chetouani, B. Taouk, E. Bordes-Richard, E. Abi-Aad, A. Aboukaïs, Appl. Catal., A Gen. 252 (2003) 269. M. Guillodo, J. Fouletier, L. Dessemond, P. DelGallo, Electrochim. Acta 47 (2002) 2809.
