摘要:

J . Chem. SOC., Furuday Trans. I , 1985,81, 1685-1692 Calorimetric Measurement of Heats of Vapour Adsorption on Graphitized Thermal Carbon Black BY A. DERKAUI, THE LATE A. V. KISELEV~ AND B. V. KUZNETSOV* Laboratory of Adsorption and Chromatography, Chemistry Department, M. V. Lomonosov Moscow State University, 119899 GSP, Moscow, U.S.S.R. Received 18th September, 1984 Heats of adsorption of various vapours have been measured on Sterling FT-FF graphitized carbon black in a calorimeter at 293-383 K. Curves representing the dependence of the heat of adsorption on surface concentration have a maximum at temperatures of 352 K or higher. After treatment of the carbon black surface with hydrogen at 1423 K this maximum remains in the case of triethylamine and iso-octane but it is absent in the case of cyclohexane.Corresponding isotherms of adsorption have steps at the same surface concentration as the maximum. This phenomenon seems to be related to the penetration of molecules into defects of molecular size in the graphite basal planes. Such penetration can take place if there are increases of temperature and adsorbate concentration. Adsorption isotherms at low coverage are similar to Henry’s isotherms, and values of the heat of adsorption decrease when the temperature increases. These values are compared with the results of other authors. Recent extensive experimental information concerning the adsorption of various classes of organic substances on the homogeneous surface of graphitized thermal carbon black (GTCB) has shown that this adsorbent is highly selective to the separation of structural and spatial isomers by the method of gas adsorption chromatography.1-6 The combination of molecular-statistical methods, involving the atom-atom approximation in the theory of intermolecular 3 9 7-9 with experimental chromatographic determination of Henry’s constants for adsorption on a homogeneous monoatomic surface of GTCB has yielded a new chromatoscopic method of determining various structural parameters of molecule^.^^ 8-11 However, the higher accuracy and reliability afforded by the chromatoscopic determination of molecular parameters can only be obtained with homogeneous crystal surfaces of known structure, presumably monoatomic. The gas-chromatographic method, widely applied in studying the adsorption of complex molecules at low surface coverages, is dynamic, and the adsorbed molecules are in contact with the adsorbent for a relatively short time.Consequently this method should be complemented by a static method of analysis of equilibrium adsorption isotherms and calorimetric measure- ments of the heat of adsorption over a wide range of temperatures, approaching those used in gas chromatography. This can then be used to determine (in detail) the existence of particularly active adsorption centres and their distribution according to adsorption energies over a wide range of adsorbent surface coverages and to investigate the possible effect of surface inhomogeneity on the gas-chromatographic determination of thermodynamic characteristics of adsorption.The lowering of the heat of adsorption with increasing temperature, as shown by calorimetric measurements,12 results in a prejudiced attitude to experimental data based on the t Deceased 17th July, 1984. 16851686 HEATS OF ADSORPTION ON CARBON BLACK r A n - A - " o = - o 0 - - " (01 I I I 0.5 1.0 1.5 2.0 a/pmol ni-2 Fig. 1. Dependences of the differential heats of adsorption of benzene on GTCB on its surface concentration measured at various temperatures: (a) 297, (b) 332, (c) 352, ( d ) 363, (e) 373 and cf) 383 K. isosteric method of determining heats of adsorption using the isotherms obtained by gas-chromatographic methods. In this determination one usually assumes that the heat of adsorption is independent of temperature. Hence the direct measurement of heats of adsorption at different temperatures using a calorimeter is important in practical applications of adsorption and adsorption chromatography, as well as in furthering the corresponding molecular-s ta tis tical theory .EXPERIMENTAL The heats and isotherms of adsorption of vapours were measured using a DAK-1-1 calorimeter (U.S.S.R.) redesigned in such a way that there is no temperature gradient between the calorimeter cover and an ampoule filled with ad~orbent.'~ A vacuum system of adsorbate dosing using a membrane rni~romanometer~~ and a capillary microburette was employed which did not contain mercury or lubricant [for further details see ref. (12)]. Sterling FT-FF (Cabot Corporation) GTCB was used as adsorbent. The GTCB was evacuated at 473 K until a residual pressure of Q Torrt was attained. The specific surface area of this GTCB (s = 11.4 m2 g-l) was determined by the B.E.T.method using the adsorption of benzene vapour to find the area corresponding to one benzene molecule in a dense monolayer, a, = 40.1 A2.15 To remove surface oxides the GTCB was evacuated at 1423 K up to Torr and exposed to hydrogen at 1423 K for 30 h with three intermediate evacuation sessions and the subsequent introduction of pure hydrogen at the same temperature (this hydrogen-treated sample will be denoted as HTGTCB). Triethylamine, iso-octane, cyclohexane and benzene were used as adsorbates after being dessicated and distilled. t 1 Torr = 101 325/760 Pa.A. DERKAUI, A. V. KISELEV AND B. V. KUZNETSOV 1687 0.3 PI 'E 0.2 5 4 0 ---.0.1 2 . 0 CI '1.5 & - 0 5 1.0 . 0.5 5 10 15 PlTOrr Fig. 2. Isotherms of benzene adsorption on GTCB at different temperatures. (A) Isotherms up to p = 0.6 Torr and (B) isotherms up to p = 17 Torr. Legend as for fig. 1 . RESULTS AND DISCUSSION Fig. 1 shows differential heats of adsorption of benzene on GTCB, ijv, measured for different values of surface concentration a and calorimeter temperature from 297 to 383 K. The corresponding adsorption isotherms are shown in fig. 2. Heats of adsorption and adsorption isotherms were measured twice for 332 and 383 K and three times for 363 K. As seen from fig. 1, for a calorimeter temperature of 297 K and a surface concentration rising to 2 pmol m-2 (which corresponds to a surface coverage 8 x 0.5)15 the differential heat of adsorption remains practically constant (qv = 40 kJ mol-l).At 332 K this constancy ends and a further increase in the calorimeter temperature generates a peak starting from 352 K in the a region from 0.2 to 0.4 pmol me2. As the temperature rises the heat of adsorption in the region of the peak increases up to 50 kJ mol-1 with two narrower peaks appearing which shift towards lower a. Adsorption isotherms of benzene have steps which correspond t s the above peaks and at higher temperatures have sections of approximately constant values of a [fig. 2(B)]. A comparison of the heat of adsorption of benzene at 297 K with calorimetric data published previo~sly~~ leads to the conclusion that the Sterling FT-FF GTCB sample used is sufficiently homogeneous.2 However, increasing the adsorption temperature removes this homogeneity with additional centres of adsorption becoming available;1688 HEATS OF ADSORPTION ON CARBON BLACK 0 1 2 3 0 1 2 3 Fig.3. Isotherms of adsorption of cyclohexane (a) and benzene (b) on GTCB for (a) 0, 296 and 0, 365 K and (6) 0 , 2 9 7 and 0 , 393 K. p/Torr r 0 0.5 a/pmol m-' 1 .o Fig. 4. Dependences of differential heats of adsorption of 0, benzene and 0, cyclohexane on GTCB for 296297 K. these are revealed by peaks in the heats of adsorption with corresponding steps on the adsorption isotherms which shift towards lower a. Unlike benzene adsorption isotherms, no steps at either 296 or 365 K were observed on the isotherms of cyclohexane adsorption on the same GTCB (fig. 3). This implies that under these conditions the saturated cyclohexane molecule cannot reach additional centres of adsorption, while the less curved isotherms of cyclohexane adsorption indicate lower adsorption energy compared with that of benzene.15 Fig.4 shows that from 296 to 297 K, as already mentioned, the differential heat of adsorption, q,, of the flat benzene molecules (fig. 4) in the range of surface concentrations a < 1 pmol m-2 is practically constant, whereas the value of 4, for cyclohexane is less and increases gradually with increasing a, owing to a greater adsorbate-adsorbate interaction.16 Note that for the adsorption on HTGTCB of the non-rigid iso-octane molecule, which has internal rotation and is incapable of specific intermolecular interaction, a curve showing the dependence of the differential heat of adsorption on coverage exhibits a peak at 366 K for a ~ 0 .6 pmol mP2, althoughA. DERKAUI, A. V. KISELEV AND B. V. KUZNETSOV 3 c, 401 Y \ > : 1689 ;5o&PtP&zg= c - n = I 1 I 1 0 0 . 5 1.0 1.5 2.0 2.5 3.0 Fig. 5. Dependence of the differential heat of adsorption of iso-octane on HTGTCB (a) and the corresponding adsorption isotherms (b) for 0, 296 and 0, 366 K. 60 50 40 30 ::[30 (0) I 1 1 I 1 0.5 1.0 1.5 2 .O 0 Fig. 6. Dependence of the differential heat of adsorption of triethylamine on GTCB (opened points) and on HTGTCB (filled points) measured at various temperatures 0, 300; W, 362; 0, 383; 0, 296; 0, 363 and +, 383 K. broader than that for benzene [fig. 5(a)], and the adsorption isotherm has a corre- sponding step at this temperature [fig.5(b)]. To check whether the peaks in the heats of adsorption and the corresponding steps in the adsorption isotherms are a result of a stronger interaction between the adsorbate molecules and the surface polar oxygen-containing compounds, the GTCB surface was treated with hydrogen and both the original sample and that treated with1690 HEATS OF ADSORPTION ON CARBON BLACK 2 . 5 il 1 2 3 4 p/Torr Fig. 7. Isotherms of adsorption of triethylamine on GTCB and HTGTCB at various temper- atures. The legend is the same as in fig. 6. hydrogen, HTGTCB, were used to study the adsorption of a strong organic base, triethylamine. In this case dependences of q, on a and of a on p measured at different temperatures were similar to those for benzene, and the heat of adsorption in the region of the peak reached 55-60 kJ mol-l. After treatment with hydrogen the peaks were still observed on the curves showing the variation of qv with a for triethylamine at 363 and 383 K, and the corresponding curves measured at 296 and 300 K practically coincide (fig.6). Isotherms of triethylamine adsorption on GTCB and HTGTCB measured at 362, 363 and 383 K, respectively, practically coincide, the adsorption isotherms measured in the vicinity of the ambient temperature seem to differ owing to slight differences in the temperatures at which adsorption was measured (fig. 7). The above experimental data lead to the conclusion that adsorption centres causing peaks in the dependence of differential heats of adsorption on surface concentration and the corresponding steps in adsorption isotherms appear (1) at adsorption temperatures above 352 K, (2) only for low surface concentrations of the adsorbate (from 0.2 to 0.5 pmol m-2) but (3) for high pressures of the adsorbate vapour p in the gas phase (ca.1 Torr). In addition, the interaction of these centres with molecules of the adsorbate is not specific. It seems that the high temperature used in conventional calorimetric analyses of heats of adsorption, combined with low surface ccncentrations of the adsorbates studied, are the reasons for overlooking the phenomenon in earlier studies of heats and isotherms of benzene adsorption.l6? l7 One possible cause of peaks in the heats of adsorption and steps in the corresponding adsorption isotherms would be a gap in the exterior graphite layers and dislocations in the crystal lattice of the GTCB which, when heated, become more easily accessible to the adsorbing molecules.Dislocations of ca. 7 A were detected by high-resolutionA. DERKAUI, A. V. KISELEV AND B. V. KUZNETSOV 1691 3 00 350 4 00 T/K Fig. 8. Dependence of differential heat of adsorption of benzene, q,, extrapolated to the zero surface coverage, on temperature. Calorimetric meaurements: 0, 0, this work; ., taken from heat-capacity measurements in ref. (20); +, taken from heat-capacity measurements in ref. (1 5). A, 273-323 K; [ref. (17)] x , 297-383 K (present work). Chromatographic determinations: 0, 303-343 K [ref. (21)]; 0, 343-373 K [ref. (22)]; v, 333-418 K [ref. (23)]. The dashed line is L - RT where L is the heat of benzene condensation.electron microscopylS for various temperatures (including an extreme of ca. 3273 K) of graphitizing the thermal carbon black. In the points of these dislocations cavities of molecular size can be formed. At adsorption temperatures of 352 K and over, with sufficiently high pressure in the gas phase, these cavities are accessible to the flat and rigid molecules of benzene and branched but not rigid molecules (i.e. capable of internal rotation) of triethylamine and iso-octane, whereas molecules of cyclohexane, having a chair conformation, cannot penetrate them. In these cavities the adsorption energy increases owing to additional intermolecular interactions of the adsorbing molecules both with carbon atoms of graphite as entering the cavities or gaps in the lattice and with each other.Daccy and Evansl9 related the presence of 6-7 A wide slot-like pores to a 1.2% increase in volume of a microporous Saran-type carbon sample detected by the dilatometric method when the carbon black was saturated with benzene vapour. With GTCB, however, such points of raised energy of benzene adsorption are few. The greatest area occupied by peaks in fig. 1 is only ca. 5% of the total area expressing the integral heat of adsorption of a dense monolayer of benzene molecule^.^^ As the temperature grows the peaks on the curves of qv against a shift towards lower values of a. In fig. 2 these values of a correspond to high pressures of the adsorbate vapour in the gas phase (1-3 Torr). These pressures and elevated temperatures generate the excess of concentration needed for the adsorbate molecules to penetrate inaccessible portions of the adsorbent and to overcome the energy of activation.Further research will require new measurements on various samples of graphitized thermal carbon black, including a sample of the more homogeneous Sterling MT 3100 GTCB. Particular attention should be paid not only to reproducibility of the results but also to obtaining desorption points so that one may be certain that equilibrium has been reached. Peaks on the curves showing the dependence of the differential heat of adsorption on the surface concentration noticeably reduce the accuracy of extrapolation of qv to zero coverage. These extrapolations for adsorbed benzene using the data of fig.11692 HEATS OF ADSORPTION ON CARBON BLACK were done in two ways: extrapolation using the initial points (up to a = 0.3 pmol m-2) or over all points (except for those referring to peaks) (up to a = 2.0 pmol m+). As seen from fig. 8, the values of ijv,l (the index 1 corresponds to the value extrapolated to zero-coverage fit) thereby obtained are much reduced as the measurement temperature increases, uiz. 40 kJ mol-l at 297 K to 28 kJ mol-l at 383 K. This difference of 12 kJ mol-1 is far beyond the limits of measurement and extrapolation errors. Good agreement with our results is shown by data for the calorimetric determination of heats of benzene adsorption at 293 K from ref. (15) and the heat of benzene adsorption at 307 K calculated using thermal-capacity measurements.20 Isosteric heats of adsorption calculated by the least-squares method from the initial portions of the benzene adsorption isotherms determined herein [fig.2(a)J as well as isosteric heats obtained in ref. (1 7) by the gravimetric method and extrapolated to zero coverage fit the plot of the calorimetric measurements (fig. 1) well, provided that they are referred to the middle of the temperature ranges in which the corresponding adsorption isotherms were measured and that qv, = qst, - RT. Using chromatographic experiments a good agreement for qv, is obtained from the data of ref. (21) if the value of ijv, obtained there is referred to the middle of the temperature range in which the chromatograms were measured. A. V. Kiselev and Ya.I. Yashin, Gas-adsorption Chromatography (Nauka, Moscow, 1967), p. 267; Gas Ahorption and Liquid Chromatography (Chimiya, Moscow, 1979), p. 255. N. N. Avgul, A. V. Kiselev and D. P. Poshkus, Adsorption of Gases and Fluih on Homogeneous Surfaces (Chimiya, Moscow, 1975), p. 383. A. V. Kiselev, in Physical Chemistry: The Modern Problems, ed. Ya. M. Kolotyrkin (Chimiya, Moscow, 1982), pp. 180-213. 7%. Welsch, W. Engewald and J. Porschman, J. Chromatogr., 1978, 148, 143; J. Prakt. Chem., 1978, 320, 493. E. V. Kalashnikova, A. V. Kiselev and K. D. Shcherbakova, Chromatographia, 1983, 17, 52 1. A. V. Kiselev, V. I. Nazarova, K. D. Shcherbakova, E. Smolkova-Keulemansova and L. Feltl, Chromatographia, 1983, 17, 533. A. V. Kiselev, Chromatographia, 1978, 11, 691. A. V. Kiselev and D. P. Poshkus, Faraday Discuss. Chem. Soc., 1980, 15, 13. 'I D. P. Poshkus, Discuss. Faraday SOC., 1965, 40, 195. lo A. J. Grumadas, A. V. Kiselev and D. P. Poshkus, J. Chem. SOC., Faraday Trans. 2, 1982,78, 2013. l 1 A. V. Kiselev and R. S. Petrova, Dokl. Akad. Nauk SSSR, 1983,272, 1415. l 2 A. Derkaui, A. V. Kiselev and B. V. Kuznetsov, Zh. Fiz. Chim., 1985, 59, 159. l 3 B. V. Kuznetsov and A. Derkaui, Zh. Fiz. Chim., 1983, 57, 1326. l4 A. Derkaui and B. V. Kuznetsov, Zh. Fiz. Chim., 1982, 56, 1840. l5 A. A. Isirikyan and A. V. Kiselev, J. Phys. Chem., 1961, 65, 601 ; Zh. Fiz. Chim., 1962, 36, 1 164. l6 N. N. Avgul, G. I. Berezin, A. V. Kiselev and A. Ya. Korolev, Kolloid. Zh., 1958, 20, 298. l7 R. A. Pioretti and R. E. Smallwood, J. Colloid Interface Sci., 1966, 22, 459. l* L. L. Ban, in Surface and Defect Properties of Solids (The Chemical Society, London, 1972), vol. 1, l9 J. R. Daccy and M. J. B. Evans, Carbon, 1979,9, 579. 2o G. I. Berezin, A. V. Kiselev and V. A. Sinitsyn. Zh. Fiz. Chim., 1970, 44, 734. 21 L. D. Belyakova, A. V. Kiselev and N. V. Kovaleva, Bull. SOC. Chim. Fr., 1967, 285. 22 F. Dondi, M. F. Gonnord and G. Guiochon, J. Colloid Interface Sci., 1977, 62, 303. 23 E. V. Kalashnikova, A. V. Kiselev, R. S. Petrova, K. D. Shcherbakova and D. P. Poshkus, Chromato- p. 54. graphia, 1979, 12, 799. (PAPER 4/ 1 6 19)

ISSN:0300-9599    

DOI:10.1039/F19858101685

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. 1, 1985,81, 1693-1704 Phase Cooperation in Oxidation Catalysis Structural Studies of the Iron Antimonate-Antimony Oxide System BY RAYMOND G. TELLER, JAMES F. BRAZDIL AND ROBERT K. GRASSELLI* The Sohio Research Center, 4440 Warrensville Center Road, Cleveland, Ohio 44128, U.S.A. AND WILLIAM YELON University of Missouri Research Reactor, Columbia, Missouri 6521 1, U.S.A. Received 24th September, 1984 The compositions FeSb20, and FeSb,O,, of the two-phase FeSb0,-a-Sb,O, system, an active and selective catalyst for the oxidation and ammoxidation of propylene, have been structurally characterized by Rietveld analysis of powder neutron-diffraction data. Results of the analysis indicate that the presence of Sb20, has no effect on the bulk structural parameters of FeSbO,.Specifically (a) the unit cell of FeSbO, does not depend upon the presence of Sb,O, or calcination temperature, (b) antimony atoms are not found in the intersticies of the coexisting iron antimonate and ( c ) the apparent Sb/Fe ratio is 1 in iron antimonate. Additionally, the Sb/Fe occupancy in the rutile FeSbO, structure is random as no supercell reflections were observed. Results of scanning electron microscope and X-ray photoelectron spectroscopy experiments have been interpreted to show that Sb enrichment occurs on coprepared samples of the two-phase mixture. Based on this evidence and the lack of alteration of the bulk structures of both phases it is suggested that surface alteration in this two-phase system is the key to enhanced selective catalytic oxidation activity.Among the ternary oxide systems that catalyse olefin oxidation reactions, Bi,O, - nMoO,, SnO, - Sb,O,, USb, 0, and FeSb0,-Sb,O,, the latter is perhaps the least well understood. For the first three systems listed, descrete phases have been identified as being the active catalytic agent.l The iron antimonate-antimony oxide system is unusual in that a coexistence of both phases is needed to realize maximum The individual phases of the system, FeSbO, and a-Sb,O,, have been structurally characterized, but the synergistic effect of the two-phase system in olefin oxidation is not well understood. Sb,O, is, by itself, a selective but not active catalyst, while FeSbO, is active but not as selective., Most workers attribute activity to the FeSbO, phase, while numerous experiments have documented that good catalytic oxidative activity and selectivity is only realized when an excess of Sb204 is In order to correlate these findings, numerous proposals that deal with the coexistence of these phase have been forwarded.Some of these are listed here: ( I ) an excess of Sb20, is needed to promote formation of FeSb,O,, either as a discrete phase3 or on the surface of FeSb0,,2q (2) dissolution of Sb into FeSbO, confers improved catalytic properties on the phase,6$ (3) ‘phase cooperation’ between Sb,O, and FeSbO, effects catalytic behaviour in olefin oxidation2 and (4) an excess of Sb20, is needed to insure complete reaction of all Fe,03,5 a well known deep-oxidation (waste-forming) catalyst. As two of these postulates [( 1) and (2)] require bulk structural alteration of FeSbO,, it was concluded that a careful investigation of an FeSb0,-Sb,O, mixture was in order.16931694 PHASE COOPERATION IN OXIDATION CATALYSTS Although the two phases have been separately characterized, a simultaneous structure refinement of both phases on a coprepared mixture has never been reported. The application of Rietveld analysis (least-squares fits of models to powder diffraction profiles) to neutron-diffraction data results in a powerful tool for metal oxide structure determination. The ability to model accurately the line shape of a neutron-diffraction Bragg reflection and near equal scattering lengths of most of the elements makes powder neutron diffraction an extremely powerful and precise structural probe.Additionally, the inherent ability of Rietveld analysis to refine simultaneously multiple phases in a given sample allows precise structure investigations of multiphase powder samples, providing overlap is not severe. Consequently we have collected neutron- diffraction data on two powdered samples: FeSb206 (FeSbO, -iSb,O,) and FeSb,O,, (FeSb0,-2Sb20,). Hypotheses based on models from data of the former sample are applicable to issues in oxidation catalysis, as it is the Sb/Fe ratio (Sb/Fe = 2) that exhibits excellent catalytic activity and is often the composition of 4-7 for testing. From a structure-determining point of view this material is not ideally suited for a neutron-diffraction experiment because Bragg peaks due to Sb204 are an order of magnitude smaller than those due to FeSbO,, and this makes the resulting model of Sb20, imprecise. Neutron-diffraction data for the FeSb,O,, system, however, yield models of comparable precision for both phases.The results of the multiphase refinement for Sb,O, from this system (Sb/Fe = 5 ) can therefore be used as a fixed contribution to the refinements of the FeSb,06 system. Hence, by judicious combination of the results from both compositions one can make significant arguments regarding both the structural and catalytic features of FeSb, 0, materals. EXPERIMENTAL SAMPLE PREPARATION Samples of FeSb,O, were prepared as follows. An aqueous solution of Fe(NO,), (Baker reagent grade) was added to an Sb sol (12% Sb,O, by weight, Nalco Co.) and the mixture was heated and stirred until dry. After drying overnight at 120 "C the mixture was fired at 425 "C for 3 h.Heat treatment of FeSb,O,, consisted of firing the mixture at 800 "C for 6 h twice with grinding in between. X-ray diffractograms taken after each heat treatment indicated that no increase in crystallite size had occurred during the second calcination. Finally, the mixture was fired at 900 "C for 16 h. For FeSb,O, calcination was carried out at 800 "C for 3 h. A portion of the FeSb,O,, mixture was examined on a Cambridge Instruments scanning electron microscope equipped with a Kevex EDX analyser. The mixture was dusted on a silver-painted aluminium pin and then was sputtered with carbon. Examination in the miscroscope revealed a broad particle-size distribution, ranging from 10 to 200 pm.Elemental analysis of numerous particles also gave varied results; the Sb/Fe ratio ranged from 4.4. to 10.3. Significantly, no particle was observed with a Sb/Fe ratio of 1 or 2, which would correspond to an FeSbO, or FeSb,O, phase, respectively. Almost all particles observed appeared not to be isolated single crystals, rather each appeared to be coated with small crystallites. In view of the diffraction-experiment results (vide infra) it would appear that larger crystals of FeSbO, are coated with smaller crystals of Sb,O,. A micrograph of the FeSb,O,, material is presented in plate l(a) and an elemental scan (for Fe) in plate l(6). Comparison of the micrographs indicates that the large particles contain iron. In order to confirm the hypothesis that the surface is Sb-enriched, X-ray photoelectron spectra were taken of the FeSb,O,, sample' on a Kratos XSAM800 X-ray photoelectron spectrometer instrument at 15 keV, 20 mA power using Mg Ka X-rays.The relative atomic abundances were calculated with the following formula : where Q is the cross-section* and C is the area counts per sweep. The result of this calculation was 1 / 17.8, indicating a surface enrichment of Sb above that of the known 1 / 5 stoichiometry. NFe/NSb = cFe (TFe/CSb OSb (1)J. Chem. SOC., Faraday Trans. 1, Vol. 81, part 7 Plate 1 Plate 1. (a) Scanning electron micrograph of FeSbO, * 2Sb,O,. (b) An Fe map of the same area as (a). TELLER, BRAZDIL, GRASSELLI AND YELON (Facing p . 1694)R. G. TELLER, J. F. BRAZDIL, R .K. GRASSELLI AND W. YELON 1695 NEUTRON-DIFFRACTION DATA FOR FeSb,O,, Room-temperature powder neutron-diffraction data were collected at atmospheric pressure at the University of Missouri Research Reactor with the new position-sensitive detector diffractometer. Data were collected in (4) 25" spans of the detector from 12.5 to 112.5" (28) with neutrons of 1.287 8, wavelength. The sample was contained in a in. diameter thin-wall vanadium can to eliminate parasitic Bragg scattering from the container. An oscillating radial collimator in front of the detector eliminated neutrons scattered from the beam by air and minimized the experimental background. Data were collected for 12 h. The data were corrected for the efficiency variation of the detector (due to the parallax effect of the linear detector) and rebinned into 0.1" intervals. No correction was made for the variation in peak width with detector setting (another parallax effect), but previous experience has shown this to have only a small effect on R,, and almost no effect on RHragg or on the refined crystal parameters.Rietveld profile analysis was carried out with a locally modified version of DBW 3.2, obtained from Dr R. Young,g University of Georgia. Data from 15 to 110" (28) were used in the analysis. A contribution of 0.001 or greater of the maximum of any Bragg peak was included in the calculation of the profile. Starting parameters for FeSbO, were taken from an X-ray powder determinationlo and for Sb,O, from a powder neutron-profile analysis.ll Initially a four-term background function and a three-term Gaussian profile function were varied in the least-squaring process.As the refinement continued more variables were included; in the final cycles of least squares a total of 42 parameters were varied. Global parameters varied were four background and a 26 zero. Various parameters were varied for the individual phases. For FeSbO, [space group P4,/mnm, Fe/Sb located on special position (0, 0, 0), one 0 atom at special position (x, x, O)], these were a three-term Gaussian profile (half-width) function, scale factor, two cell parameters, one temperature factor for Fe/Sb, Sb and Fe occupancies, one position term and a temperature factor for 0. For a-Sb,O, (space group Pna2,) a three-term Gaussian profile function, scale factor, three cell parameters, a temperature factor for each unique Sb atom, one temperature factor for all 0 atoms and 17 position parameters [the z position parameter for Sb(1) was fixed] were varied.Results of the refinement are reported in table 1. Comparison of the Gaussian profile functions from the two phases indicates that the average iron antimonate crystal size is larger than that of the a-Sb,O,. Additional refinements were carried out making use of an extrapolated background. With the exception of the temperature factors, the results of the two kinds of refinements were virtually identical. Because the profile agreement factors for the refined background refinements are lower, only results from that least squaring are reported herein. Calculated, observed and difference diffractograms are presented in fig.1. All observed Bragg reflections are accounted for by the presence of the two phases FeSbO, and a-Sb,O,. Specifically, there are no extra Bragg peaks due to a trirutile phase (FeSb,O,) or P-Sb,O,. The occupancies of the Fe and Sb atoms in the FeSbO, phase were independently refined to establish the oxide stoichiometry (this is equivalent to refining the scattering length of the site). The final values for Fe and Sb are 0.12(1) and O.lO(l), respectively [results for the extrapolated background least squares being 0.12( 1) and 0.12( l)]. Additionally, refinements were attempted where Sb was placed in an octahedral hole (0, t, $) and a tetrahedral hole (0, i, a). In each case the temperature factor of this interstitial was constrained to be equal to that of the Fe/Sb at (0, 0,O).In both cases the occupancies could be refined to values not significantly different from zero [ -0.02( 1) and O.OO( l ) , respectively]. These results indicate that there is little bulk solubility of Sb in FeSbO, and that the Fe/Sb ratio is 1. NEUTRON-DIFFRACTION DATA FOR FeSb,O, Room-temperature time-of-flight (TOF) data were collected using the special-environment powder diffractometer (SEPD) at the intense pulsed-neutron source (IPNS) at Argonne National Laboratory at atmospheric pressure. A detailed description of the SEPD and IPNS has been published.', The sample was contained in a thin-walled seamless vanadium tube ca. 4 in. in diameter and 4 in. long capped with A1 plugs. Data from ca.2 x 10, pulses, collected over ca. 18 h, were used to analyse FeSb,O,. Data from the backscattering banks (26 = 150")1696 PHASE COOPERATION IN OXIDATION CATALYSTS Table 1. Final parameters for the multiphase refinements of FeSbO, and a-Sb20, parameter a-Sb,O, FeSbO, space group cell parameters (A) a b V C half-wid th parame tena U V W final agreement factors RWP R P Rexp RBragg cell parameters (A) a b C V final agreement factors RWP R P Rexp ~ (A) constant-wavelength data Pna2,, Z = 4 P4,/mnm, Z = 2 5.441 (2) 4.6365 ( 6 ) 4.800 (1) 4.6365 ( 6 ) 11.766 (3) 3.0742 (6) 307.72 66.09 2.22 0.95 0.61 0.49 - 1.13 -0.34 5.42% - 7.05% - 1.77% - 2.33 % (349 reflections) 2.49% (49 reflections) (B) time-of-flight datab 5.440 (2) 4.809 (1) 11.755 (3) 307.53 (8) 8.41 % 5.80% 2.99% 4.640 (1) 4.640 (1) 3.077 (2) 66.24 (4) a FWHM**2 = U tan2@+ V tan@+ W.Time of flight (in ps) = 7569.58 d-2.99 d2-9.92; maximum 19 ms, minimum 6 150 ps. with a minimum time of flight of 61 50 p s (dmin x 0.8 1 A) and a maximum time of flight of 19 ms were used in the least-squares refinement. A Rietveld-type analysis adopted for time-of-flight data with multiphase capabilities was used.13 A contribution of 0.00 1 or greater of the maximum of any Bragg peak was included in the calculation of the profile. Starting parameters for both compounds were taken from the data-analysis procedure described above for FeSb,O,,. During all the least-squares refinements described below, positional parameters for the a-Sb,O, were held fixed. This was dictated by the small size of the Bragg peaks of Sb20, as compared with those of FeSbO, (see fig.2). Parameters refined were as follows: three background parameters (YBK = BKl + BK2 * EXP [ - BK3 * d ] ) , one adsorbtion parameter ( A = 1 /[ 1 + ABS * d]), two peak-shape parameters for each phase, a scale factor for each phase, cell constants (three for orthorhombic antimony oxide, two for tetragonal iron antimonate), two isotropic temperature factors for each of the two phases (one for each of the two kinds of atoms and/or positions) and the x parameter for the oxygen at (x, x, 0) in FeSbO,. Additionally, in two separate refinements an Sb atom was placed in a tetrahedral and octahedral hole in FeSbO,, the isotropic temperature factor was fixed to be identical to that of the Fe/Sb at (0, 0, 0) and the occupancy was allowed to vary.The final values for the occupancies were O.OO(1) and - O . O l ( l ) , respectively. This establishes that there is little solubility of Sb in theR. G. TELLER, J. F. BRAZDIL, R. K. GRASSELLI AND W. YELON 0 - 1 1697 I I 1 I I I 11 I I11118111IIlII I1 I0 111111118ll11 I I I I 1 I I I 8 I l l l l l I I 8 I 8 I U I Y I I 8 I I I 1 I 8 Y I I I I l Y l Y I I I U I 8 ~ I l l 8 I I I I l I Y ' c - ------.CAM "A-.-/>/)b.--wiPt 1, A</.,./, ~ . ~ ~ . ~ * - ~ - ~ ~ ~ ~ , 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ~ 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 t 1 1 L68 9830 -- 0191 -- I 6 5 5 3 -- I 8 3276 1638 intersticies in iron antimonate for this composition. The experimental data, calculated diffractogram and difference plot are presented in fig.2. As for the data for FeSb,O,,, all observed Bragg peaks can be accounted for by the presence of two phases, FeSbO, and a-Sb,O,. COMPARISONS OF THE NEUTRON REFINEMENTS Fig. 1 and 2 display the raw data and calculated profiles for the two systems refined. Comparisons of the model parameters from the two refinements for FeSbO, are excellent. The thermal parameters [0.65(10) and 0.69(10) for Fe/Sb and 0, respectively, for FeSb,O,,, as against 0.4(5) and 0.4(6) for FeSb,O,] compare favourably, and values of the only structural parameter, x (x, x, 0), for the 0 atom are also identical [0.3055(6) and 0.3053(16) for FeSb,O, and FeSb,O,, respectively]. Final fractional coordinates for a-Sb,O, (from refinements of the constant-wavelength data for FeSb0,/2Sb20,) are presented in table 2 and a list of distances and angles in table 3.Scattering lengths for Fe, Sb and 0 were 0.95, 0.56 and 0.575 A, respe~tively.~ Refined cell constants for FeSbO, are a = 4.6365(6), c = 3.0742(6) A and V = 66.09( 1) A3 (constant- wavelength data) and a = 4.640( 1) & c = 3.077(2) 8, and V = 66.24(4) A3 (time-of-flight data). Because all Bragg peaks are included in the unit-cell determinations (349 reflections for a-Sb,O, and 49 for FeSbO,) the resultant lattice parameters are very precisely determined. Consequently the excellent agreement between these results and those of Amador and RasineslO [4.6388(2) A, 3.0773(2) A and 66.22 A3, respectively, based on X-ray data] are not surprising. The slight differ- ences are probably attributable to a small error in neutron wavelength for the reactor (constant- wavelength) data.For a-Sb,O, results here are a = 5.441(2) A, b = 4.800(2) and c = 11.766(3) A (reactor data on FeSb,O,,), a = 5.441(1) A, h = 4.808(1) A and c = 11.755(3) 8, (time-of-flight data on FeSb,O,). Cell constants from an earlier reported neutron powder-profile analysis are markedly different, a = 5.456(1) A, b = 4.814(1) 8, and c = 11.787(2) but are closer to results on ASTM card no. 11-694, a = 5.436 A, 56 F A R 11698 PHASE COOPERATION IN OXIDATION CATALYSTS 1 11468 0 - Iv t I 1 I 1 I I 61 50 a198 1 0 2 ~ 6 1 2 2 9 ~ 1 4 3 ~ 2 16 390 1 8 ~ 3 8 time of flight/ps Fig. 2. Experimental time-of-flight neutron data (points) and calculated profile (line) from Argonne National Laboratory for FeSb,O,.A difference plot is given below. Tick marks indicate the position for Bragg peaks of FeSbO, only. Smaller peaks without tick marks are due entirely to the a-Sb,O, phase. Table 2. Final atomic parameters (fractional coordinates) for a-Sb,O, atom lo3 x 103y 103z Biso Sb(1) -31 (3) 31 (2) -2" 0.12(10)b Sb(2) 377 (3) 17 (4) 247 (3) - 0(1) 322 (2) 159 (4) 93 (2) 0.47 (12) O(2) 152(2) 717(2) 192(3) - O(3) 82 (3) 192 (2) 307(2) - O(4) 361 (2) 824(3) 405 (2) - " The z coordinate of Sb( 1) was fixed. Temperature factors of like atoms were constrained to be identical. b = 4.810 8, and c = 11,76 A. The good match between our results for the two data sets discussed here and those of others indicates that the calibration of the instruments and the refinements are accurate.We consider the results here to be of the highest precision. The raw time-of-flight neutron-diffraction data for FeSbO,/OS Sb,O, and the constant- wavelength raw neutron-diffraction data for FeSb0,/2Sb2O, are available as Supplementary publication no. SUP 56190. See Notice to Authors, J . Chem. SOC., Faraday Trans. I , 1985,81(1).R. G. TELLER, J. F. BRAZDIL, R. K. GRASSELLI AND W. YELON 1699 Table 3. Selected distances (in A) and angles (in ") in FeSbO, and Sb,O,, and comparisons with earlier work Sb@, this work ref. (1 1) Sb( 1 )-O( 1)" Sb( 1WWb Sb( 1 W ( 3 ) Sb( 1 W(4)" Sb( 1)-0(4)b Sb(2tO( 1) Sb(2F0(2)" Sb(2w(2)b Sb(2)-0(3)" Sb(2)-0( 3)b Sb(2)-0(4) LO( I)-%( 1) - O( 1) LO( l)a-Sb( 1)-0(4)" L O( l)b-Sb( 1 )-0(4)b LO( l)a-Sb( 1)-0(4)" LO( l)b-Sb( 1 )-0(4)" L 0(2)a-Sb(2)-0(3)b L 0(2)b-Sb(2)-0(3)' L 0(4)-Sb(2W( 1) 2.30 (2) 2.03 (2) 2.51 (3) 2.01 (2) 2.22 (2) 1.94 (2) 1.98 (2) 2.00 (2) 1.92 (3) 1.94 (2) 2.08 (4) 2.37 1.86 2.62 2.09 2.17 2.01 1.98 2.05 1.92 1.92 2.04 82.5 145.6 89.2 66.8 78.6 170.3 175.6 176.6 FeSbO," this work ref.(10) M-Ob 1.996 (2) 1.960 M-Ob 2.005 (3) 2.066 weighted average 1.999 1.995 Alternate crystallographic positions of chemically identified oxygen atoms. M = Fe or Sb. RESULTS Structures of the individual phases have already been investigated, FeSbO, with powder X-ray diffractiong and a-Sb,O, with powder neutron diffraction,'l so only a brief discussion will be given here. a-Sb,O, is a layered compound in which corrigated sheets of edge-sharing Sb5+ octahedra are separated by Sb3+ layers with asymmetric oxygen coordination.The structure is represented in fig. 3. The octahedral Sb5+ coordination is quite symmetric and the Sb-0 distances range from 1.92(2) to 2.08(2) A, averaging 1.98(2) A. The oxygen coordination about the Sb3+ atoms is very asymmetric, presumably owing to the lone-pair effect. These lone pairs line up in channels perpendicular to the [loo] face. Sb-0 distances for these Sb3+ ions range from 2.00(2) to 2.51(3) A. As noted above in lattice-parameter comparisons, there are marked discrepancies between our model and that proposed earlier. Not surprisingly, comparison between the Sb-0 distances listed here and those of othersll are also at odds. These discrepancies are recorded in table 2. The excellent match between lattice parameters from two kinds of neutron sources (spallation source, time-of-flight data and nuclear reactor constant-wavelength data) and from two different samples reported here offers strong support for the precision of the model presented here.56-21700 PHASE COOPERATION IN OXIDATION CATALYSTS Fig. 3. Stereographic ORTEP diagram displaying the [IOO] face of a-Sb,O,. Open spheres represent 0 atoms and filled spheres Sb atoms. For FeSbO, the Fe and Sb atoms both occupy the special position (0, 0,O); the lack of supercell reflections attests to the random nature of the Fe/Sb occupancy. The only structural parameter to be varied is the x position parameter of the lone unique oxygen atom [located at (x, x, O)]. The results of our refinement yield a value of x different from that of the X-ray refinement [0.3055(6) and 0.3053( 16) as against 0.318, respectively].In the resultant model the MO, octahedran is more symmetric as compared with the X-ray model. Because of the near equality of neutron-scattering lengths of 0, Fe, and Sb as compared with X-ray scattering form factors, the ability to model accurately a neutron-diffraction peak profile and the good agreement between our results on two systems from two neutron sources we believe the precisions of the neutron refinements are significantly higher. Analysis of the diffraction data indicates that the only phases present are FeSbO, and a-Sb,O,. All diffraction peaks are accounted for by these phases. Specifically, there are no extra lines suggestive of a trirutile phase (with c' = 3c).Therefore, reports of the formation of a trirutile phase under these synthesis conditions are to be regarded with suspicion. Additionally, for each refinement that included nuclear density in the octahedral or tetrahedral holes in the FeSbO, structure, the occupancy of this intersticial was refined to a value not significantly different from zero. Based on this there is no evidence for the incorporation of an excess of Sb (or Fe) into the FeSbO, phase. The neutron-diffraction results also demonstrate that the Fe/Sb ratio is essentially 1 in FeSbO,, even in the presence of an excess of Sb,O,. Therefore, the refinements have demonstrated that the presence of Sb,O, has no effect on the bulk structure of iron antimonate. Furthermore, the cell volumes of FeSbO, in FeSb0,-2Sb20, calcined at 900 "C and FeSbO, in FeSbO4-O.5Sb,0, calcined at 800 "C are also virtually identical to each other and to that of pure FeSbO,.This is further evidence that the presence of a-Sb,O, (and variation in calcination temperature) have no effect on the bulk structure of iron antimonate. Further support for this argument is found in the excellent structural match between the FeSb,O,, and FeSb,O, systems.R. G. TELLER, J. F. BRAZDIL, R. K. GRASSELLI AND W. YELON 1701 Despite the structural evidence presented above, we are faced with a preponderance of data that establish the need to have a coexistence of both phases (prepared simultaneously, not mixed after preparation) to achieve maximum catalytic perf~rmance.l-~ Because there is no bulk structural modification, one is left with the possibility of surface modification of FeSbO, to correlate these data.These possible modifications fall into two general categories. Surface structure modification can occur, where an excess of Sb is incorporated into the FeSbO, crystal, perhaps forming islands with FeSb,O, stoichiometry,2 and/or Fe incorporation into the surface of a-Sb,O,. The second possibility leaves both structures unchanged but orients the Sb,O, crystals on iron antimonate, allowing the phases to produce synergistically the desired oxidation product. As direct observation of either of these possibilities may not be possible we are left to draw inferences from existing data. One such inference, the large difference in crystallite size as measured by the diffraction experiment and observed in the electron microscope, and the apparent coating of FeSbO, crystals with small Sb,O, crystallites, has lead us to investigate the latter hypothesis.To this end we have employed two programs, MATCH^ and MATCH^,', to investigate possible epitaxy between FeSbO, and Sb,O,. MATCH 1 generates two-dimensional crystallographic networks of the two compounds and searches for matches based on cell-dimension fits. MATCH^ takes the cell-matching information generated by MATCH^ and compares atomic positions in planes (motifs) for the two phases. Use of this software allows one to find planes of atoms in the two crystals that are suitably matched for epitaxial growth, looking for phase boundaries that minimize dislocation density and atom movement.Applying this procedure to FeSbO, and a-Sb,O,, MATCH^ found many excellent lattice fits of crystallographic faces with low indices and low dislocation densities (< 1Olo interfacial dislocations per cm2). Analysis of these networks with MATCH2 Sb M a 1 Sb a Scheme 1. Sb M ,M ~ b 2 + c 2 ) Scheme 2.1702 PHASE COOPERATION IN OXIDATION CATALYSTS oriented Sbz 0, c r y s t a l l i t e s ,-~..I , 111 01 Scheme 3. indicated that atoms on the [ 1101 face of FeSbO, matched those of the [OOl] face of Sb,O,. The [OOl] direction is perpendicular to the layers of Sb atoms described above. An outline of the SbV octahedral coordination layer is given in scheme 1 (metal atoms only). All distances are in ingstroms and a unit-cell outline is shown.A similar outline of the [110] face of FeSbO, is given in scheme 2. The mismatch in distances is small (in the range 2.2-3.6%) and it is easy to envisage a coherent phase boundary between these faces. Based on this a schematic diagram of a possible active catalyst is given in scheme 3. For a three-step mechanism of propylene oxidation there are two possibilities for active sites: /O\ Sb3+ Sb3+ CHJ-CH- CH2- CH2-CH-CH2 (allyl formation) /O\ Sb5+ Sb'+ CH2-CH-CH2-CH2=CH-CH0 (two step, 4e- oxidation) (3) (4) (catalyst reoxidation) I ( 5 ) Sb3+ + 2Fe3+ d 2Fe2+ + Sb5+ 2Fe2++ 02(g)-2Fe3++ 20(1) There is a great deal of evidence that elements with lone pairs of electrons are needed to activate propylene and form the allylic intermediate. Bi,O, is known to dimerize propylene selectively.l5 Sb,03 also displays a high selectivity to hexadiene formation from propylene.ls These facts implicate metalloids with free pairs of electrons not involved in M-0 bonding as being crucial to allyl formation, an intermediate widely held to be important in propylene oxidation.'' Metals in their highest oxidation states are well suited for 0 insertion. The resulting reduction of the metal (Mo6+ to Mo4+ in molybdates and Sb5+ to Sb3+ in atimonates) is chemically reasonable. Evidence in support of this is more plentiful for molybdenum oxides [e.g. MOO, oxidizes methanol to formaldehyde (with Fe)18 and allyl radicals to a ~ r o l i e n ~ ~ ] than antimonates as there are, strictly speaking, no 'pure' Sb5+ oxides. However, we ascribe the role of oxidation (or 0 insertion) to Sb5+.One function of the FeSbO, crystal may be inR. G. TELLER, J . F. BRAZDIL, R . K . GRASSELLI AND W. YELON 1703 active-site reconstruction (catalyst reoxidation). The facile redox couple (Fe3++e- -+ Fe2+) would facilitate 0, chemisorption with the FeSbO, acting as the oxidant for the reduced Sb204. This would require some reduction of Fe3+ to Fe2+ on a working catalyst (as has been observed20). Note also that for the scheme above, the [OOl] face of the Sb,O, crystal would not be suitable for an active site, as only Sb5+ or Sb3+ (but not both) are exposed on this face. Faces that present both Sb oxidation states of Sb are the [loo] and [OlO] faces (indicated by A and B in scheme 3). These crystal faces would then most likely contain the sites of catalytic activity.Oriented crystals of Sb,O, have been implicated in propylene oxidation catalysis in other systems.' The second possibility for an active-site model is at the interface between the two phases. At a coherent interface between [ 1 101 of FeSbO, and [OOl] of Sb,O,, the local stoichiometry is Sb/Fe = 3, with Sb5+, Sb3+, Fe3+ and Fe2+ present. In short, these coherent phase boundaries would contain all the necessary elements for olefin oxidation. Alternate possibilities for active sites include shear planes and intergrowth regions in either of the individual phases. The latter possibility appears to be ruled out based on the well defined crystallography of the system. The existence of shear planes (or zones) cannot be excluded on the basis of the crystal data if they are non-periodic and/or have a very large (ca.40 A) periodicity. If the periodicity is very large, however, the impact of such line defects upon catalytic activity (the number of active sites) would be small. Numerous non-periodic shear planes would give rise to diffuse scattering which may not be obvious in powder studies, hence the possibility of their existence cannot be excluded. High-resolution electron-microscopy studies are needed to clarify this hypothesis. CONCLUSIONS We have performed structure refinements of the two-phase (FeSb0,-Sb,O, ) system with powder neutron-diffraction data. Based on the results of these structure refinements we conclude the following. Under the synthesis conditions employed no trirutile phase (FeSb,O, or Fe,Sb,O,) is observed and the Fe/Sb occupancy in FeSbO, is entirely random and uniform.The effect of an excess of Sb,O, and calcination temperature on bulk FeSbO, is negligible. Specifically, there is no lattice expansion, contraction or incorporation of an excess of Sb into iron antimonate. Assuming oxidation states of Fe3+ and Sb5+ there is no oxygen or metal deficiency in bulk FeSbO,. Based on the conclusions arrived at above, the enhancement of selective oxidative catalytic activity of the two phases over that of the individual components is necessarily due to surface modification (possibly mutual) of the two phases and/or a phase cooperation effect. We (and others) have proposed a model whereby a-Sb,O, crystallites are oriented on the surface of FeSbO, and thereby create catalytically active sites not found in either of the isolated phases.Whether the active sites are at the phase interface or nearby on chemically modified single phases is impossible to determine given the experimental data presently available. That Sb can be found in excess in a surface layer of FeSbO, has already been suggested.,? Recent observations that small amounts of vanadium21 or molybdenum22 can exist in solid solution with Sb,O, (triggering the a to p transition) raises the intriguing possibility that surface modification of Sb,O, (Fe dissolution) can confer increased oxidation catalytic activity on this phase. At any rate it may be a moot point to ascribe oxidation activity to a single phase in this system, as it is clear that a coexistence of both a-Sb,O, and FeSbO, is necessary for optimum catalytic activity.1704 PHASE COOPERATION IN OXIDATION CATALYSTS We thank L.C. Glaeser (Sohio) for sample preparation, Dr J. Jorgensen (Argonne) for assistance in the time-of-flight data collection and the Standard Oil Company (Sohio) for permission to publish this work. We also thank the U.S. Department of Energy for supporting the Intense Pulsed Neutron Source at Argonne National Laboratory as a national users’ facility. For the bismuth molybdates see R. K. Grasselli, J. D. Burrington and J. F. Brazdil, Faraday Discuss. Chem. SOC., 198 1,72,203 and references therein. For the SnSbO, system see J. C. Volta, G. Coudurier, I. Mutin and J. C. Vedrine, J . Chem.SOC., Chem. Commun., 1982, 1044; J. C. Volta, B. Benachouba, I. Mutin and J. C. Vedrine, Appl. Catal., 1983,8, 215. For the USb,O system see R. K. Grasselli and D. D. Suresh, J . Catal., 1972, 25, 273. V. Fattore, A. Zygmunt, G. Mnara and B. Notari, J. Catal., 1975, 37, 223. (a) R. K. Grasselli, unpublished lecture, Milan; ( b ) F. Sala and F. Trifiro, J . Catal., 1976, 41, 1. * (a) I. Aso, S. Furukawa, N. Yamazoe and T. Sieyama, J . Catal., 1980, 64, 29; (b) S . F. Saito, Y. Y. Susaki, T. I. Nakamura, K. K. Moriya, Y. K. Nakamura and H. Y. Utsumi, Nitto Chemical Company Ltd, U S . Patent No. 4,083,804, 1978. C. K. Boreskov,S. A. Ven’yaminov,V. A. Dzis’ko,D. V. Tarasova,V. M. Dindoin,N. M. Sanobova, I. P. Olen’kova and L. M. Kefeil, Kinet. Katal., 1969, 10, 1530. T. V. Adamiya, Y. A. Michenko, D. A. Dulin and A. I. Gel’bshtein, Kinet. Katal., 1970, 11, 1168. ’ F. Trifiro, unpublished lecture, Standard Oil (Sohio) Research Center, 1984; M. Carbuciccho, G. Centi and F. Trifiro, J . Catal., 1985, 91, 85. J. H. Scofield, Lawrence Livermore Laboratory Report UCRL-51326 1973. D. B. Wiles and R. A. Young, J . Appl. Crystallogr., 1981, 14, 149. ln J. Amador and I. Rashes, J. Appl. Crystallogr., 1981, 14, 348. l 1 G. Thornton, Acta. Crystallogr., Sect. B, 1977, 33, 1271. l 2 R. B. Von Dreele, J. D. Jorgensen and C. G. Windsor, J . Appl. Crystallogr., 1982, 15, 581. l 3 J. D. Jorgensen and J. Faber, ICANS-I/, Proc. VIth Int. Collab. Adv. Neu. Sources, Argonne National Lab. June 28-July 2, 1982 (ANL-82-80, 1983); J. R. Hauman, R. T. Daley, T. G. Worlton and R. K. Crawford, IEEE Trans NucI. Sci., 1982, NS-29, 62. B. Dickens and L. W. Schroeder J . Res. Natl Bur. Stand., 1980, 85, 347. l5 B. Grzybowska, J. Haber and J. Janas, J . Catal., 1977, 49, 150. Y. N. Usov, I. M. Bolotov, N. I. Kuvshinova and V. I. Kitaev, Neftekhimiya, 1975, 15, 242. l7 (a) D. J. Hucknall, Selectiue Oxidation of Hydrocarbons (Academic Press, New York, 1977), p. 24 and references therein; (b) J. L. Callahan, R. K. Grasselli, E. C. Milberger and H. A. Strecker, Ind. Eng. Chem. Prod. Res. Deo., 1970, 9, 134. l8 (a) H. Atkins and W. R. Peterson, J. Am. Chem. SOC., 193 1,3, 15 12; (b) N. Pernicone, J . Less-Common Met., 1974, 36, 289; (c) G. Fagheruzzi and N. Pernicone, J . Catal., 1970, 16, 32. l 9 J. D. Burrington and R. K. Grasselli, J. Catal., 1979, 59, 79. 2n N. Burriesci, F. Garbassi, M. Petrera and G. Petrini, J . Chem. SOC., Faraday Trans. I , 1982, 78, 817. 21 F. J. Berry and M. E. Brett, J . Chem. SOC., Dalton Trans., 1984, 985. 22 R. G . Teller, M. R. Antonio, J. F. Brazdil, M. Mehicic and R. K. Grasselli, to be published. (PAPER 4/ 1647)

ISSN:0300-9599    

DOI:10.1039/F19858101693

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985,81, 1705-1714 Activity of Supported Tungsten Oxide Catalysts for the Metathesis of Propene BY AMEDEO ANDREINI* AND JOHANNES C. MOL University of Amsterdam, Institute for Chemical Technology, Plantage Muidergracht 30, 1018 TV Amsterdam, The Netherlands Received 24th September, 1984 The activity of several supported tungsten oxide catalysts for the metathesis of propene has been studied under temperature-programmed conditions. WO,/TiO, was found to be active at moderate temperatures; its activity passes through a maximum when the reaction temperature is decreased from 670 to 450 K. WO,/ZnO is less active and its activity as a function of the reaction temperature goes through a maximum under certain conditions. In contrast to these catalysts and to the earlier reported behaviour of the well known WO,/A1,0, catalyst, the activity of the WO,/SiO, catalyst steadily increases for reaction temperatures up to 860 K.In this respect the WO,/SiO,-Al,O, catalyst behaves like the WO,/SiO, catalyst and not like the WO,/Al,O, catalyst. The acidity of the supports has been determined by means of a temperature-programmed desorption technique, using t-butylamine as the adsorbed species. The results of this work do not suggest a relationship between the acidity of the supports and the activity for metathesis. A maximum in the activity for the metathesis of propene: 2CH,CH=CH, =t CH,=CH, + CH,CH=CHCH, (1) as a function of the reaction temperature was first found by Banks and Bailey' for a Coo-MoO,/Al,O, catalyst.This feature was later investigated in more detail by Moffat and Clark., Moreover, we have found that the metathesis catalyst WO,/Al,O, shows a point of maximum as well as of minimum activity for the metathesis of propene as a function of the reaction temperature., Banks4 has reported that there seems to be a relationship between the temperature range in which the metathesis catalysts MoO,/Al,O,, WO,/SiO, and WO,/SiO,-Al,O, attain maximum metathesis activity (or pass through a maximum) and the range in which the rate of deuterium exchange for the corresponding supports passes through a maximum. This would suggest a functional relationship between the rate of metathesis and the rate of the H-D exchange, the latter being a measure of the acidic properties of the support.In this paper we present the results of a study of tungsten oxide catalysts on different supports. As well as the well known supports SiO,, Al,O, and Si0,-A1,03, we tested TiO, and ZnO. The activity of the WO,/TiO, and WO,/ZnO catalysts for the metathesis of propene was studied as a function of the reaction temperature. The same was done for the WO,/SiO, catalyst, but at a higher temperature range than previously reported., The aim of the present study was to find out whether the occurrence of a maximum in the activity for metathesis as a function of the reaction temperature is a general one and can be considered a characteristic of the active sites involved in the reaction. 17051706 TUNGSTEN OXIDE METATHESIS CATALYSTS Moreover, the activity of the WO,/Al,O, and WO,/SiO,-Al,O, catalysts for the metathesis of propene was studied as a function of the activation temperature in order to test the hypothesis of a possible relationship between the catalytic activity and the acidity of the supports.All the supports used and the WO,/TiO, catalyst were also tested for their acidity at various activation temperatures. EXPERIMENTAL The experiments were carried out in a conventional fixed-bed microcatalytic flow reactor. The reactor itself was a section of a stainless-steel tubing (8 x m internal diameter), placed in a vertical oven, wherein a constant temperature could be provided and maintained. The gases were first passed through a separate purification section placed on all feed lines leading to the reactor. Alumina (Alcoa H 15 1) and 3A molecular sieves were used for drying the feed and a Cu/A1,03 catalyst was used to eliminate any oxygen impurity to < 1 ppm.The product gases were analysed on line by means of a gas chromatograph at room tem- perature with a 12 x 3.175 x mcolumn packed with 30% 2,5,8,11,14-pentaoxapentadecane on Chromosorb P, 149-180 pm, using flame-ionization detection and nitrogen as the carrier gas. The gas-chromatographic signal was real-time processed by a PDP 10/ 1 1 computer system, provided with a g.c. signal-integrating program. The properties of the supports used in this work are given in table 1. From all these supports catalysts containing 6 wt % WO, were prepared by wet impregnation. Screened fractions of the supports (1 80-212 pm) were impregnated with an aqueous solution of ammonium metatungstate (Koch-Light Laboratories, 99.9%) and then dried at 365 K under vacuum.Before testing, the catalysts were calcined in a stream of dry air at 773 K for 6 h unless otherwise stated. The weighted catalyst samples taken for each run were given a second heat treatment (activation) within the reactor in a nitrogen stream for one night. The acidicity measurements were carried out in a temperature-programmed desorption set up, using t-butylamine (TBA) (Merck, 98%) as the adsorbed species5 and thermal conductivity/flame-ionization detectors. RESULTS ACTIVITY TESTS A catalyst, different from those cormally used in the metathesis reaction (viz. WO,/SiO, and W03/A120,), was prepared with TiO, as the support. The 6 wt % WO,/TiO, catalyst was calcined at 675 K, activated and tested under temperature- programming conditions by increasing or decreasing the temperature stepwise by 40 K every 30 min (fig.1). In one experiment the catalyst, activated at 770 K, was studied coming from the low-temperature side. Substantial conversions were only attained when the temperature was raised above 550 K. Clearly, the catalyst went through a break-in for temperatures in the range 550-620 K, as within a 30 min period the conversion of the second sample was always higher than that of the first sample. In three more experiments the catalyst was studied coming from the high-temperature side, for three different activation temperatures, after it had been allowed to go through a break-in at 675 K.A point of maximum activity appears located in the temperature range 490-520 K. The maximum conversion values were found to increase when the activation temperature was decreased from 870 to 770 K. Fig. 2 shows repeated heating-cooling cycles obtained with the WO,/TiO, catalyst after activation at 770 K. The conversions measured during the cooling leg of the second cycle were much higher than those recorded through the first cooling leg. Isothermal runs at the lowest temperature of 470K indicated that the high conversions measured at that temperature could be kept for several hours. AfterwardsA. ANDREINI AND J. C. MOL 1707 Table 1. Propertiesa of the supports used support TiO, SiO, Al,O, Si0,-Al,O, ZnO type Harshaw T i 4 102-T 3.2 lo-, m composition 86% TiO, 14% A1,0, specific surface 78 pore volume/cm3 g-l 0.33 area/m2 g-l chemical composition A1 Ca c1 c u Fe Mg Mn Na S Ti Zn Zr Davison Ketjen Ketjen Harshaw Grace 62 CK-300 LAC 25 99.9% 99.9% 85% SiO 100% ZnO SiO, A1,0, 15% A1,0, 344 187 61 8 3 1 0.49 0.7 0.26 trace elements (ppm) - 1-10 - 5-500 - 1-10 - 1-10 100 300 10-100 - 1-10 - - 10 300 - 150 270 50-500 - 10-100 - 10-100 - - 100 - - - - - - - - a From manufacturers' specifications; blank spaces indicate data not available.0.1 2 0.0 8 X 0.0 4 0 400 600 T/ K Fig. 1. Plot of catalyst activity, expressed as conversion of propene, X, as a function of the reaction temperature for a 6 wt % WO,/TiO, catalyst. Heating leg after activation at 820 K (A), cooling legs after activation at 770 (+), 820 (0) and 870 (W) K.Reaction conditions: contact time, W/F= 1.89 kg s mol-l, pressure = 0.2 MPa.1708 TUNGSTEN OXIDE METATHESIS CATALYSTS 0.2 X 0.1 0.0 I I 400 500 600 700 T/K Fig. 2. Plot of catalyst activity, expressed as conversion of propene, X, as a function of the reaction temperature for the 6 wt % WOJTiO, catalyst activated at 770 K and after attaining steady state at 675 K. 0, First cooling leg; m, first heating leg and 0, second cooling leg. Reaction conditions: contact time, W/F = 1.89 kg s mol-l, pressure = 0.2 MPa. the catalyst went through a phase of apparently reversible deactivation. The measured deactivation rates for this catalyst were 3.4% h-l of the highest conversion measured after break-in at 675 K and 6% h-' of the highest conversion measured at 470 K.A 6 wt % WO,/ZnO catalyst was also tested in the way described above. The results, obtained by increasing (decreasing) the temperature stepwise by 50 K every 30 min, showed that the catalyst is slightly active at temperatures > 600 K. A maximum in activity appeared for an activation temperature of 870 K. A number of experiments was done to find out whether a maximum appears in the activity of the WO,/SiO, catalyst as a function of the reaction temperature, using 3, 6 and 12 wt % catalysts. Fig. 3 shows a representative plot for the 12 wt % catalyst. For reaction temperatures in the range 675-875 K in no case was a maximum found. In another test the catalyst was found to undergo irreversible deactivation and loose its catalytic activity at temperatures slightly in excess of 875 K.The supports used in the preparation of all these catalysts were found to be inactive for metathesis. The rate-temperature relationship for the 6 wt % WO,/Al,O, catalyst has been reported previ~usly.~ Fig. 4 shows the effect of the activation temperature on the conversion at 675 K. The catalytic activity increases when the activation temperature is raised. Note that the selectivity for the reaction to primary metathesis products is slightly lower than that of the WO,/SiO, catalyst and decreases when the activation temperature is increased from 770 to 970 K, but from the available data it was estimated to be 95% or better. (The selectivity for the WO,/SiO, catalyst in this study was 99% or better.) The slightly lower selectivity of the WO,/A1,0, catalyst shouldA.ANDREINI AND J . C. MOL 0.5 0.4 X 0.3 1709 - - - i 01 I 1 I TI K 600 800 Fig. 3. Plot of catalyst activity, expressed as conversion of propene, X, as a function of the reaction temperature for a 12 wt % WO,/SiO, catalyst after activation at 820 K. Reaction conditions: contact time, W/F = 3.78 kg s mol-l, pressure = 0.2 MPa. r 0 1 2 Fig. 4. Plot of the conversion of propene, X, against process time for a 7.4 wt % WO,/AI,O, catalyst activated at 770 (A), 870 (+ and m) and 970 (0) K. Reaction conditions: contact time, W/F = 1.89 kg s mol-l, temperature = 675 K, pressure = 0.2 MPa. t l h be partly attributed to the action of the support. This support was tested in blank runs at 675 K. The main products of the reaction with propene were ethene and 2-methylpropene7 cis- and trans-but-2-ene in minor quantities and high boiling products.Data from blank runs were used to correct some of the results from the actual runs. The rate-temperature relationship for the WO,/SiO,-Al,O, catalyst was studied in a number of experiments. No maxima were found on either the heating or cooling leg of the cycle over the temperature range from 600 to 800 K, at which temperature1710 TUNGSTEN OXIDE METATHESIS CATALYSTS 0.4 X 0.2 0 1 2 tlh Fig. 5. Plot of the conversion of propene, X, against process time for a 6 wt % WO,/SiO,-Al,O, catalyst activated at 770 (m) and 870 (0) K. Reaction conditions: contact time, W/F = 1.89 kg s mol-I, temperature = 675 K, pressure = 0.2 MPa. Table 2. Catalytic activity for catalysts containing 6 wt % catalystsa activation, reaction, N r0 rto catalyst T/K T/K /s-l mol g-' s-l mol s-' m-2 at point of maximum W03/Ti02 770 515 0.240 WO,/ZnO 870 710 0.074 wo3 / 820 630 0.166 WO,/SiO, 870 675 3.15 820 675 0.961 W03/A1203 870 675 0.425 770 675 0.103 W03/Si02 - A120, 870 675 0.848 770 675 0.414 at 675 K 1.62 0.265 0.737 23.6 4.30 2.77 0.50 1 7.83 1.98 2.08 8.83 0.394 6.86 1.25 1.48 0.268 1.27 0.320 a Pressure = 0.2 MPa.the active sites appeared to undergo irreversible deactivation. This catalyst is thus comparable to the WO,/SiO, ~atalyst.~ The effect of the activation temperature on its activity is shown in fig. 5 , showing that this catalyst can be more active than the alumina-based one and has no break-in period, as shown by the W0,/Si02 catalyst.The selectivity is comparable. In order to make the comparison of the various catalysts more complete, conversion values recorded in these and previous tests were used to calculate turnover frequencies and specific rates of the metathesis of propene (table 2). The initial reaction rates given in the last two columns have been calculated on the basis of the rate equation for the carbene model as proposed by Kapteijn et aLs From table 2 we can see that the W0,/Ti02 catalyst can be much more active than the WO,/Al,O, catalyst.A. ANDREINI AND J . C. MOL 171 1 Table 3. Acidity measurements material number of acid sites/g of support or catalyst /K Bronsted Lewis activation temperature Si02a Al,O," TiO, 6 wt % W0,/Ti02 ZnO 775 825 875 775 825 875 775 825 825 775 775 875 775 875 2.7 x 1019 2.8 x 1019 2.6 x 1019 1.6 x 1020 1.3 x lozo 1.2 x 102O 1 .1 x 1020 6.3 x 1019 7.2 x 1019 2.3 x 1020 2.1 x 1020 2.5 x 1020 1 . 1 x 1020 1.0 x 1020 0.9 x 1020 5.2 x lozo 4.2 x 1020 3.6 x 1020 2.3 x 1020 6.1 x 1019 6.2 x 1019 1.5 x 1Ol8 1.2 x 10'8 a The values for these catalysts are in general agreement with results reported by B. Scheffer, P. Grimberg and J. A. Moulijn, to be published. At a reaction temperature of 675 K the WO,/SiO, catalyst appears to be more active than those based on alumina or silica-alumina. The activity of this catalyst appears to be much more dependent on the activation temperature, whereby we estimate that its activity for an activation temperature of 770 K would be comparable to that of the alumina- or silica-alumina-based catalysts.On the basis of the initial reaction rates it appears that the alumina-based catalyst is more active than the silica-alumina-based catalyst. However, the two catalysts appear to have approximately the same activity when this is expressed on a unit-surface-area basis. ACIDITY MEASUREMENTS The results of the acidity measurements are shown in table 3. On the basis of these results the supports can be classified in order of decreasing Bronsted acidity at 775 K SO,-Al,O, > TiO, > Al,O, as follows: while the classification for the Lewis acidity at 775 K is Si0,-Al,O, > TiO, z SiO, > Al,O, > ZnO. Van Roosmalen et al.' found silica to contain little Lewis acidity, and this mainly associated with the aluminium impurity. The high value for SiO, in table 3 is due to a large amount of physisorbed t-butylamine.Further, SiO, was found to contain no Bronsted-acid sites. ZnO was also found to contain no Bronsted acidity. Its Lewis acidity is lower than that of A1,0, by a factor 100, but its specific surface area is also lower by as much. The peak corresponding to the Lewis-acid sites appears as a convolution of two peaks: a large one in the temperature range 375-425 K and a smaller one in the temperature range 515-535 K.1712 TUNGSTEN OXIDE METATHESIS CATALYSTS Al,O, also shows a convolution of two peaks: a large one in the temperature range 355-375 K and a smaller one in the temperature range 475-485 K. The location of these peaks does not shift when the activation temperature is increased.The trend of total acidity of Al,O, in this work as a function of the activation temperature is dissimilar from that reported by Tanabe.8 Si0,-Al,O,, like alumina, is found to contain both Lewis- and Bronsted-acid sites. The peak corresponding to the Lewis sites is not a convolution of two smaller ones and the peak corresponding to the Bronsted sites is located at a lower temperature, an indication that silica-alumina contains stronger acidity. A change in the activation temperature seems to affect only the Lewis acidity of alumina, but both the Bronsted and Lewis acidity of silica-alumina. The Lewis acidity of the WO,/TiO, catalyst appears to be lower than that of the original support by a factor of four, while the Bronsted acidity is lower by a factor of two.The change in activation temperature seems to affect only the Bronsted acidity of this catalyst. At the higher activation temperature, the peak corresponding to the Bronsted acid is located at a lower temperature, possibly an indication of stronger Bronsted-acid sites. DISCUSSION Because of the complexity it is impossible to come to a general theory for the relationship between catalytic activity for metathesis and the structural properties. Apart from the parameters defining the catalytic activity in general, the following should be considered as critically important for the metathesis reaction: (a) the reducibility of the catalyst, (b) the stability of the reduced species, (c) the acidity of the support and/or of the corresponding catalyst, (d) the precursor surface structures and (e) the type and concentration of the surface hydroxyl groups.In this article we restrict ourselves to the case of the tungsten oxide catalysts for the metathesis of propene. For the chemistry of the break-in or activation process, reference should be made to two previous articles.,. The beneficial effect of the increased activation temperature on the catalytic activity of the WO,/Al,O, and WO,/SiO,-AI,O, catalysts should be attributed to the loss of catalytic amounts of 0 and OH surface species from one or more tungsten oxide surface compounds, as was also observed in the case of WO,/SiO, catalysts., CATALYTIC ACTIVITY AS A FUNCTION OF THE REACTION TEMPERATURE The appearance of a maximum in the catalytic activity as a function of the reaction temperature appears to be a frequent occurrence in the metathesis of propene over a solid catalyst, as has been shown with various supported WO, catalysts.The WO,/TiO, catalyst, having a maximum in the temperature range 490-520 K, can be compared with the WO,/Al,O, catalyst, which has a maximum at ca. 630 K., At 500 K the WO,/TiO, catalyst is more active than the WO,/Al,O, catalyst at 630 K. Explanations for the appearance of a maximum can be found in a special relationship between the activation energy for the metathesis and the heat of adsorption of propene, or in a reversible change in the number of active sites taking place when poisons or fragments needed for the reaction are adsorbed or desorbed., In contradiction to the behaviour of the WO,/Al,O, (fig. 4) and W0,/Si0,3 metathesis catalysts, however, the activity of the WO,/TiO, catalyst increases when the activation temperature is decreased.The results of fig. 2 indicate a relationship between the observed maxima and the catalytic activity. As the maximum becomes more pronounced in the second cycle, showing higher catalytic activity, and theA. ANDREINI AND J. C. MOL 1713 deactivation rate is moderate and higher at the point of maximum, we conclude that deactivation does not vitiate such results. This catalytic system is active in a temperature range where the WO,/SiO, catalyst shows negligible activity, and therefore it is promising for practical applications. The W03/Zn0 catalyst showed a maximum in activity as a function of the reaction temperature for an activation temperature of 870 K.The point of maximum is located at a higher temperature than in the previous case. The presence of a maximum in the activity as a function of the reaction temperature occurs so generally in the case of the metathesis catalysts3 as to prompt the idea that it is a characteristic property of the active sites involved in this reaction. There are, however, exceptions preventing us reaching this conclusion. One exception is provided by the WO,/SiO, catalyst. For this catalyst the activity for the metathesis reaction does not pass through a maximum even when the reaction temperature is raised as high as 860 K. At that temperature the catalyst can work efficiently and selectively without any sign of a break-down. Another such exception is provided by the WO, / Si0,-Al,O, catalyst .RELATIONSHIP BETWEEN CATALYTIC ACTIVITY AND THE ACIDITY OF THE SUPPORTS The Bronsted acidity of the TiO,, Al,O, and SiO,-Al,O, supports does not seem to play a role in the metathesis reaction. This conclusion is based on the following considerations. Among these supports, TiO, yields the most active catalyst and yet it is not the most acid. The number of Bronsted-acid sites remains constant (Al,O,) or decreases (SO,-Al,O,) when the activation temperature is increased, while the catalytic activity of the corresponding catalyst increases in both cases. It is surprising that the Bronsted acidity of the WO,/TiO, catalyst is lower than that of the original support and increases with the activation temperature. The fact that at the higher activation temperature the catalyst is less active suggests that the Bronsted acidity present in this catalyst plays no role in the metathesis reaction.Therefore we cannot accept, in our case, the suggestion made by Laverty et a1.l0 that strong Bronsted-acid sites play an important role in metathesis. When a proton is needed for the formation of the initial metal carbene (the suggested intermediate in metathesis), a more likely source seems to be a Lewis site located on the transition-metal ion, carrying an alkene molecule, as proposed by Van Roosmalen and M o ~ , ~ or a product molecule originating from the reduction of the alkene. Furthermore, the classification of the various supports on the basis of decreasing Lewis acidity does not correspond to their classification on the basis of decreasing activity. Therefore we conclude that the Lewis acidity of the carriers is not directly involved in the metathesis reaction. This conclusion is further strengthened by the fact that the Lewis acidity is seen to decrease with increasing activation temperature, while the catalytic activity of the corresponding catalyst increases. The W03/Ti0, catalyst does not fit in this pattern, but its behaviour does not contradict the conclusion. We thank Mr B. Scheffer for his assistance with the acidity measurements. R. L. Banks and G. C. Bailey, Ind, Eng. Chem., Prod. Res. Det.., 1964, 3, 170. A. Andreini and J. C . Mol, J . Colloid Interface Sci., 1981, 84, 57. R. L. Banks, Prepr. Am. Chem. Soc., Div. Petr. Chem., 1979, 24, 399. H. C . Nelson, R. J. Lussier and M. E. Still, Appl. Cutal., 1983, 7, 113. F. Kapteijn, H. L. G. Bredt, E. Homburg and J. C. Mol, Ind. Eng. Chem., Prod. Res. Detl., 1981, 20, 457. * A. J. Moffat and A. Clark, J . Catal., 1970, 17, 264.1714 TUNGSTEN OXIDE METATHESIS CATALYSTS ’ A. J. Van Roosmalen, M. C. G. Hartman and J. C. Mol, J. Catal., 1980,66, 1 12. K. Tanabe, Solid Acids and Bases: Their Catalytic Properties (Academic Press, New York, 1970), p. 45. A. J. Van Roosmalen and J. C. Mol, J. Catal., 1982, 78, 17. lo D. T. Laverty, J. J. Rooney and A. Stewart, J. Cafal., 1976, 45, 110. (PAPER 4/ 1648)

ISSN:0300-9599    

DOI:10.1039/F19858101705

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985, 81, 1715-1724 Electrical-conductance Responses of Catalysts Exposed to Pulses of H, and 0, BY ROBERT B. BJORKLUND,* DENNIS MDERBERG AND INGEMAR LUNDSTROM Laboratory of Applied Physics, Linkoping University, S-58 1 83 Linkoping, Sweden Received 28th September, 1984 A computer-controlled gas-mixing system has been used to contact alternating hydrogen and oxygen pulses in an argon carrier with various catalysts at 400 "C. A simple pulsing sequence was used to obtain an oxidation-reduction pattern for each catalyst so that composition-related differences in conductance response could be observed. Catalyst groups studied were : alkali- promoted Fe,O,, multicomponent Fe-Sb-Ti-Mo oxides and Co supported on A1,0,. Results are discussed with respect to how the pulsing technique can be used for investigating the effect of promoters in oxide matrices, for comparison of granularities and in quality-control applications.We have previously reported results for in situ conductivity measurements of catalysts to illustrate how conductance changes can be used to monitor the composition of gaseous reactants in contact with the catalysts and how the catalysts are altered by the reactants.' In this paper we report the conductance changes observed for catalysts exposed to alternating pulses of hydrogen and oxygen in an argon carrier as a method of studying the oxidation-reduction properties of the catalysts. Wagner and Hauffe reported in 1939 a study of the reaction of gaseous oxygen and hydrogen adsorbed on Pd by measuring the electrical-conductivity change as a function of oxygen The kinetics of many other reactions have been studied using the pulse-microreactor technique since its introduction by Kokes et aL3 Pulse techniques have proved valuable in elucidating the kinetics and mechanisms of reactions occurring on catalyst surfaces and many examples are described in ref.(4). We have applied a pulse-conductivity technique to the study of catalysts using oxidation and reduction as the characterizing reaction and computer control to produce reproducible pulse sequences. Our results show that this technique can be used to evaluate many catalysts with respect to: (a) quality control in manufacture, (b) effect of promoters in multicomponent oxides and (c) inter-granular contacts in tableted catalysts.EXPERIMENTAL Conductivity measurements were performed in a stainless-steel cell (0.5 mm3 volume) constructed of two Balzers hgh-vacuum flanges (1.33 in?) that has been described elsewhere.' The catalyst was tightly held against the bottom flange by a stiff, gold-plated tungsten wire connected to an insulated current feed. Electrical contacts for all samples were tested by slowly sweeping between +2 and -2 V and observing the linearity and symmetry of the current response at 400 "C. Directcurrent measurements were performed by applying 1 V across the catalyst in series with a decade resistor and measuring the voltage drop across the resistor. The resistance was chosen so that the measured voltage drop was < 50 mV. For convenience we report all results 1715 7 1 in = 2 .5 4 ~ lo-* m.1716 CONDUCTIVITY OF CATALYSTS in units of microsiemens (1 pS = lopg W1 ). Notations such as x 5 or x 50 in the figures indicate that the sensitivity was increased to record certain curves. All measurements were performed at 1.5 bar? pressure, 240 cm3 min-l total flow and 400 "C. For convenience the temperature was measured using a chromel-alumel thermocouple connected to the external cell wall. The cell was placed in a resistively heated furnace during measurements. Pulses of 0, or H, in argon carrier were passed over the catalysts in rapid succession using a Commodore Business Machine CBM 3008 computer to control three Brooks thermal mass- flow controllers for Ar, 0, and H, flows. We used a simple pulsing pattern for the work reported here.For example, 0, could be first set at 0.3 vol% and then a series of seven H, pulses from 0.009 up to 0.3 vol% (each H, concentration 1.82 times larger than the preceding one) would be passed by repeating the sequence: 15 s H, on, 0, off-15 s 0, on, H, off seven times. After the seventh pulse the 0, concentration could be changed to 0.2 vol% and after a 4 min storage time the same H, pulse series could be done. This change in the carrier-gas composition after each pulse sequence is seen as spikes in the data curves when valves were opened and closed and resulted in some curves not appearing to return to the original base line. Details of this computerized gas-mixing system have been given el~ewhere.~ Gases used in the experiments were purchased from Alfax AB and used without further purification.H, was supplied to the flow controller as 1 % or 10% in Ar and 0, as 1 % in Ar or as pure 0,. In this way a wide partial pressure range of the gases could be contacted with the catalysts. The following commercial catalysts were studied : alkali-promoted Fe,O,, manufactured by Shell and UCI, and supported Co, manufactured by UCI. Several mixed-oxide catalysts containing Fe, Sb, Ti and Mo were also investigated. These were prepared by mixing the appropriate water-soluble salts of Fe, Sb and Ti following a description in the patent literature.s After calcination at 600 "C for 4 h, X.r.d. powder patterns indicated that FeSbO, and Sb,O, phases were present. Samples containing Mo were prepared by impregnating an aqueous solution of ammonium paramolybdate into the dried Fe-SbTi matrix.Powders of 100-200 mesh were pressed at 10 ton$ to obtain suitable samples for conductance measurements. RESULTS AND DISCUSSION ALKALI-PROMOTED Fe,O, Several types of promoted Fe,O, are commercially available for dehydrogenation reactions. In fig. 1 we compare Fe,O, styrene catalysts containing different amounts of promoter. The results for Shell 105 (91 % Fe,O,, 7% K,O, 2% Cr,O,) were quite straightforward and normal behaviour for n-type semiconductor Fe,O, was observed for all H, partial pressures, with the conductivity increasing in H, and decreasing in 0,. G-64C from UCI (60% Fe,O,, 20% K,CO,, 2.5% MOO,, 5% Ce0,)alsoexhibited normal behaviour at least for preparation batch B.However, batch A behaved quite differently. At low H, partial pressures it became less conducting in H, and more conducting in 0, and it was only at higher H, pressures (marked by an asterisk in fig. 1) that batch A samples exhibited a normal response to reduction. For these oxide catalysts the pulsing technique has clearly shown the effect of promoters on conductance changes upon oxidation and reduction. It is not surprising that G-64C batch B exhibited smaller conductance changes upon contact with H, than Shell 105 (fig. l), since the former contains much more K,CO, (K,O), which we assume is electrically inert. It is not as clear why patterns for G-64C batches A and B were so different from each other. Lee has previously reported that different promoters can cause Fe,O, to behave as either a p- or n-type semicond~ctor.~ However, batches A and B are of nominally the same composition so it is possible that other factors in the manufacturing t 1 bar = lo5 Pa.$ 10 ton/1.32 cm2 (area of the die) FZ 7500 kg cm-*.R. B. BJORKLUND, D. SODERBERG AND I. LUNDSTROM 70 50 1717 - - I 1 I A I I Fig. 1. Conductance response of alkali-promoted Fe,O, at 400 "C. 0, concentration was 0.2 vol% for the first 6 pulses in each sequence and 0.1 vol% for the seventh pulse. H, concentration was varied as shown along the axis (described in fig. 4). Each sequence consisted of 15 s H, then 15 s 0,, seven times, with the H, concentration increasing with each pulse. (a) G-64C batch A, (6) G-64C batch B and (c) Shell 105. process or prior history of the catalysts (e.g.prolonged exposure to moisture) have caused the difference. It has been reported that the electrical properties of ceramic Fe20, gas sensors are strongly influenced by the anion of the iron salt used as the starting material.8 After calcination at 600 "C it was found that sensors made from FeCl, - 6H20 exhibited p-type behaviour with respect to oxidation and reduction at 400 "C, while Fe20, prepared from nitrates and sulphates exhibited n-type behaviour. Thus anions present in the Fe precipitate can have an effect on the final Fe203 even if they are removed during calcination. We have also made Seebeck measurements in air on the catalysts and the results confirmed that Shell 105 and G-64C batch B were in fact n-type Fe,O, and that batch1718 CONDUCTIVITY OF CATALYSTS I I A I I 0.009 0.33 0.009 0.33 H, (VOI %) Fig.2. Conductance response of Fe-Sb-Ti-Mo oxides at 400 "C. For the left-hand sequence 0, concentration was 0.3 vol% and for the right-hand sequence 0.2 ~01%. H, concentration was varied as shown along the axis (described in fig. 4). Each sequence consisted of 15 s H, then 15 s 0,, seven times. 0, concentration was constant for the first six pulses, but was changed for the seventh pulse. (a) FeSb,., Tio.* Moo,, 0, and (b) Fe Sb,., Tioe8 0,. A was p-type Fe,O,. We assume that in an in situ cell9 the transition of batch A from p- to n-type behaviour with increasing H, pressure could also be observed by e.m.f. measurements as such a phenomenon has been reported for a so-called two-carrier cobalt ferrite.I0 MIXED-OXIDE OLEFIN-OXIDATION CATALYSTS In fig.2 we compare mixed metal oxides of Fe, Sb and Ti both with and without Mo. Note that the oxidation-reduction patterns shown are taken from a long series of pulse groups and have been chosen to illustrate certain points and thus are not positioned relative to each other: e.g. the data for 0.2 vol% 0, should be shifted to higher conductance (+ 1 pS) relative to the corresponding 0.3 vol% data. We investigated oxides containing Tio.8, Tila2 and Til.6, which either contained Moo., or no Mo at all (total of six different compositions). Because of variations caused by differences in the tablets formed from the powders and in electrical contacting when the tablets were positioned in the measurement cell, it was difficult to observe how the initial conductances of the samples in 0.3 vol% 0, in Ar before pulsing were dependent on composition even after several repetitions of the measurements were made.This problem was solved by collecting the conductance data in digital form andR. B. BJORKLUND, D. SODERBERG AND I. LUNDSTROM 2 1719 I 1 Ti 0.8 T i l . 2 Ti1.6 titanium fraction Fig. 3. Background conductances at 400 "C in 0.3 vol% 0, as a function of Ti content of catalysts with and without Mo: (a) FeSb,., Ti, Moo., 0, and (b) FeSb,., Ti, 0,. 0 .L 0.2 I 0.0 0 60 120 180 t l s Fig. 4. Comparison of the conductance responses at 400 "C of the six catalysts described in fig. 3 to alternating pulses of H, and 0,. 0, concentration was 0.2 vol% for each re-oxidation pulse. Bar graph depicts the concentration of each hydrogen pulse (~01% in Ar).1720 3.29 3.26 v) x ;s- 3.23 0 CONDUCTIVITY OF CATALYSTS /------- I I I 5 10 fl S 15 Fig.5. Comparison of the conductance responses at 400 "C of the six catalysts described in fig. 3 (a) and (b) during the re-oxidation pulses shown as A in fig. 4: 0.2 vol% 0, for 15 s. then using a computer to perform some simple analysis in order to compensate for non-linearities introduced into the data by experimental uncertainties. The conductance data for the six compositions over the measurement time (partially shown for the Tio.8 sample in fig. 2) were found to consist of two parts, one independent of the pulse sequence and one dependent on the pulse sequence. These two parts were separated using a least-squares fitting method in a similar procedure to that used previously to analyse long-term behaviour in gas l1 Fig.3 shows the trends for the independent part of the conductance data, which we define as the samples' background conductance at 400 "C in 0.3 vol% 0, in Ar, as a function of Ti composition. For samples with no Mo this background conductance decreased with increasing Ti content. For samples containing Moo., the Ti dependence of the conductance was diminished. More significant for our investigation of how pulse sequences can be used to study catalysts is the analysis of the gas-dependent part of the conductance data. It was convenient to make a comparison of how the different samples responded to the pulse sequences by scaling the data relative to one of the samples (arbitrarily chosen).This is shown in fig. 4 for the H, pulse sequence with a re-oxidation pulse at 0.2 vol% 0, (right-hand side of fig. 2). The sample containing Til.2 Moo., was used as the reference data set and the other five sets were compared point by point with the reference and then each was assigned a single scaling factor. The scaling factors were between 0.5 and 2 and were used to multiply all the points in each data set so that the five sets as nearly as possible matched the reference set. As seen in fig. 4 the scaled conductances for the six samples mostly coincided as increasing hydrogen pressures were contacted with the catalysts, but the curves for the re-oxidations after H, pulses 4, 5 and 6 show some differences.One can also observe an Mo-dependent difference in these re-oxidations in the primary data shown in fig. 2 for samples containing Tioe8. The re-oxidations markedR. B. BJORKLUND, D. SODERBERG AND I. LUNDSTROM 1721 by an asterisk in fig. 2 seem to show an Mo-induced diminishing of the conductance change upon re-oxidation for the 0.2% 0, concentration. This comparison can be made for all samples by expanding fig. 4A as we have done in fig. 5. After reduction with pulse 6 of 0.18 vol% H,, the gas in contact with the samples was changed to 0.2 vol% 0, at time 0. It is quite clear that after 5 s of re-oxidation a clear difference in slope develops between the samples which contain Mo and those without. A similar comparison when the re-oxidation pulses were 0.3 vol% 0, (left-hand side of fig.2) did not show such a clear Mo dependence as in fig. 5. Mixed-metal-oxide catalysts used for selective oxidation of hydrocarbons have been the subject of many investigations since the first description of a bismuth molybdate catalyst for propene (amrn)oxidation.l2 Many approaches have been developed for studying the role of promoter ions in the oxides, e.g. X.r.d. to detect the changes in solid-state structure brought about by incorporation of foreign ions in the original oxide matrix. 13-15 Conductivity measurements have also been used to characterize mixed oxides such as bismuth molybdate,16 tin-antimony oxide,” promoted TiO,l* and Pt-doped Ti0,.19 Because of experimental uncertainties, the results of such studies are generally of only qualitative value,,O but since the conductivity and selective- oxidation,’ properties of the catalysts are both dependent upon the degree of oxidation and reduction, conductivity measurements are a useful in situ monitor of redox changes in such oxide materials.The significant result obtained using the pulse-conductivity technique described here is that qualitative differences between catalysts of different compositions were observed only for certain 0, and H, compositions. This is illustrated in fig. 2, where the redox pattern for 0.3 vol% 0, is qualitatively the same for catalysts which contained Mo or had no Mo, but at 0.2 vol% 0, the same H, pulse sequence indicated a difference between the two catalyst groups with respect to re-oxidation (fig. 5).We attribute the observed hindered re-oxidation of the catalysts containing Mo to the fact that the Mo ions (shown by ESCA to be mostly on the catalyst surface) consume 0, upon re-oxidation but do not contribute much, if anything, to the conductivity change of the catalyst. This model of the role of the Mo ion on the catalyst surface is supported by the computer analysis of the data. Background conductances were found to decrease with increasing Ti content, but catalysts which also contained Mo exhibited less variation in background conductance (fig. 3). Thus even at constant 0, concen- tration (no pulsing) the surface Mo partially blocks the effect that Ti has on catalyst conductance. Further analysis of the pulsing data (fig. 4 and 5) showed that the presence of Mo affected re-oxidation for some of the H, and 0, concentrations. AI,O,-SUPPORTED C O We have previously reported that certain commercial supported catalysts are reversibly ca.lo4 times more conducting when oxidized than when reduced.l To investigate this phenomenon further we contacted two forms of a Co catalyst (UCI, G62RS, 32 wt% Co/Al,O,) with an oxygen pulse sequence (15 s 0,, then 15 s H,, increasing 0, concentrations). In fig. 6 we compare results for as-received tablets and tablets which were first pulverized and then re-tableted at 10 ton pressure. The strip-chart data on the right-hand side show that the re-tableted pellets exhibited a 10 times greater conductance response to the 0, pulses than the as-received tablets and that the latter had relatively greater response to the larger 0, pressures.Plotting conductance against vol% 0, (left-hand side of fig. 6) showed a linear relationship for the as-received tablets (the 0, pulse sequence is in fact a logarithmic progression) and a non-linear response for the re-tableted pellets. There have been only a few conductivity studies on catalysts containing SiO, or1722 CONDUCTIVITY OF CATALYSTS 1 .E m a ;s- 0.5 0.2 0.4 0 0 2 (vol %I Fig. 6. Conductance response at 400 "C of UCI G62RS Co/Al,O, catalysts. H, concentration was 0.3 vol% and 0, was varied as shown along the axis. Each sequence consisted of 15 s 0, then 15 s H,, seven times, with the 0, concentration increasing with each pulse. (a) Re-tableted pellets and (6) as-received pellets. Al,O, reported in the literature.l69 22* 23 A possible explanation as to why we have observed such large conductance changes for G62RS is that the metal crystallites become larger in 0, and smaller in H,, and so an electrically conducting pathway through the catalyst (which is 90 vol% Al,O,) is more extensive in 0,.This hypothesis is in agreement with electron-microscope data for supported Pd,24 and Ni.26 This is apparently not an unusual property, as we have recently observed similar behaviour by four additional commercial catalysts. These were (from Harshaw): Ni-3250T, 50 wt % Ni/proprietary support; Ni-3266E, 58 wt % Ni/Al,O,; Ni-0148T and Ni-O149T, 60 wt% Ni/kiselguhr. The catalysts are sold in a reduced and stabilized form. All four catalysts were electrically insulating at room temperature but exhibited the same conductivity behaviour with respect to H, and 0, exposure as previously reported for G62RS.Although we have observed this phenomenon for a wide range of catalyst compositions, it is not possible to make a definite statement on how such electrical behaviour is related to composition. The phenomenon seems to be related to the percolation theory of electrical conductivity in composite materials made up of aR. B. BJORKLUND, D. SODERBERG AND I . LUNDSTROM 1723 conducting phase dispersed in an insulating medium. Depending on the s t r u c t ~ r e , ~ ~ ~ 28 ca. 10-15 vol% conducting phase is necessary before any increase in the insulator's conductivity is seen. What this means for the supported catalyst is that > ca. 30 wt% metal must be present in order for a 'potential' electrical pathway to be present, i.e.one that can be established by metal-crystallite expansion in 0, and broken by metal-crystallite contraction in H, at elevated temperatures. Of course, many catalysts with > 30 wt% metal are conducting at room temperature and remain so in both H, and 0, at elevated temperature (although we have shown that for one such catalyst, alternate 0, and H, treatments at 425 "C changed it to G62RS-like behaviour,' probably because of a redistribution of the metal crystallite sizes2s). However, our present understanding is that if a catalyst contains > 30 wt% metal on an insulating support (SiO, or A1203) and is insulating at room temperature after reduction and stabilization, there is a good chance that it will exhibit G62RS-like behaviour with respect to conductivity changes in 0, and H, when heated to elevated temperatures.A connecting pathway of expanding and contracting crystallites through the catalyst must also 'hop' from grain to grain so it is not only the sizes of the metal crystallites which determine the conductivity response to oxidation and reduction but also the intergranular resistivity.20$ 29 We believe that this aspect of conductance response can be studied by the pulse technique. In fig. 6 we compare two forms of G62RS with respect to oxidation and reduction at 400 "C. Since we are not at liberty to disclose characterization data concerning any of the commercial catalysts we have used in this work, we shall propose a model tci explain the data in fig.6 based solely on our electrical measurements. A catalyst which had been powdered and then re-tableted exhibited ca. 10 times greater background conductance in 0.3 vol% H, and greater responses to 0, pulses than the original as-received material. We attribute this to the fact that the re-tableting procedure pressed the catalyst grains closer together, thus improving the electrical pathway. In addition, plots of conductance against 0, concentration indicated a saturation effect for the re-tableted catalyst. Since the intergranular contacts are better in the re-tableted catalyst they make a smaller contribution to the measured conductance. The change in conductance upon oxygen exposure is therefore more dependent on the adsorption of oxygen on the grain surfaces and the saturation observed in the conductance change may be due to saturation of the oxygen adsorption on the catalyst grains.In the as-received catalyst the grain boundaries have a larger influence on the conductivity and thus the conductance changes upon exposure to oxygen are not as large and exhibit no saturation effect. Further speculation on how such factors as grain geometries would affect the data in fig. 6 would require detailed electron-microscope studies similar to those described in ref. (20) - CONCLUSIONS We have presented a method for investigating catalytic materials which involves contacting the catalysts with alternating H, and 0, pulses at an elevated temperature while recording the electrical conductance.The catalysts were studied under transient conditions with the gas flow regulated by a computer-controlled gas-mixing system. The main advantage of the technique is that several catalysts can be simultaneously analysed over a wide H,/O, concentration range. Our results indicate that it is often necessary to cover a large concentration range in order to detect regions where catalysts of related compositions behave differently. The pulse-conductivity technique can be applied to areas of practical interest in catalysis, such as quality control in manufacturing (e.g. determination if promoted Fe,03 is of p- or n-type character or1724 CONDUCTIVITY OF CATALYSTS monitoring of the tableting process for certain supported metal catalysts) or as a complementary in situ technique for reactor studies of mixed metal oxides.The large numbers of catalysts involved in both areas, manufacturing and testing, makes the pulse-conductivity method a promising technique for further development. The work reported here was supported by a grant from the National Swedish Board for Technical Development. We thank R. A. Innes of Gulf Research for comments on the manuscript. R. B. Bjorklund and I. Lundstrom, J. Catal., 1983, 79, 314. C. Wagner and K. Hauffe, 2. Electrochem., 1939, 45, 409. R. J. Kokes, H. Tobin and P. H. Emmett, J. Am. Chem. SOC., 1955, 77, 5860. H. Kobayashi and M. Kobayashi, Catal. Rev. Sci. Eng., 1974, 10, 139. D. Soderberg, Computer Assisted Evaluation of Gas Sensors, Linkoping Studies in Science and Technology, Dissertation no. 89, 1982. R. A. Innes, personal communication. E. M. Lee, J. Catal., 1966, 6, 137. Y. Nakatani and M. Matsuoka, Jpn J. Appl. Phys., 1982, 21, L758. S. T. Hwang and G. Parravano, J. Electrochem. Soc., 1967, 114,482. lo G. H. Jonker, J. Phys. Chem. Solids, 1959, 9, 165. l 1 I. Lundstrom and D. Soderberg, Sensors and Actuators, 1981, 2, 105. l 2 J. D. Idol, U.S. Patent 2094580. l 3 R. D. Srivastava, A. B. Stiles and G. A. Jones, J. Catal., 1982, 77, 192. l4 P. Forzatti, P. L. Villa, N. Ferlazzo and D. Jones, J. Catal., 1982, 76, 188. l5 J. F. Brazdil and R. K. Grasselli, J. Catal., 1983, 79, 104. l6 J. M. Peacock, A. J. Parker, P. G. Ashmore and J. A. Hockey, J. Catal., 1969, 15, 381. *' J-M. Hermann, J-L. Portefaix, M. Forissier, F. Figueras and P. Pichat, J. Chem. Soc., Faraday l8 S. R. Morrison, J. Catal., 1974, 34, 462. l9 J.-M. Hermann and P. Pichat, J. Catal., 1982, 78, 425. 2o B. M. Arghisopoulos and S. J. Teichner, J. Catal., 1964, 3, 477. 21 A. Bielanski and J. Haber, Catal. Rev. Sci. Eng., 1979, 19, 1. 22 S. J. Thomson and G. A. Harvey, J. Catal., 1971, 22, 359. 23 F. J. Jansen and R. A. Schoonheydt, J. Chem. Soc., Faraday Trans. 1, 1973,69, 1338. 24 E. Ruckenstein and J. J. Chen, J. Colloid Interface Sci., 1982, 86, 1. 25 T. Wang, A. Vazquez, A. Kato and L. D. Schmidt, J. Catal., 1982, 78, 306. 26 T. Nakayama, M. Arai and Y. Nishiyama, J. Catal., 1983, 79, 497. 27 E. K. Sichel, J. I. Gittleman and P. Sheng, J. Electron. Muter., 1982, 11, 699. 28 V. B. Tare and J. B. Wagner, J. Appl. Phys., 1983, 54, 252. 29 J. R. Stetter, J. Colloid Interface Sci., 1978, 65, 432. Trans. I , 1979, 75, 1346. (PAPER 4/ 1673)

ISSN:0300-9599    

DOI:10.1039/F19858101715

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985, 81, 1725-1730 Mutual Diffusion and Self-diffusion in the Frictional Formalism of Non-equilibrium Thermodynamics BY HANS VINK Institute of Physical Chemistry, University of Uppsala, P.O. Box 532, S-751 21 Uppsala 1, Sweden Received 2nd October, 1984 Conceptual differences between mutual diffusion and self-diffusion are analysed using the frictional formalism of non-equilibrium thermodynamics. Expressions for the two diffusion coefficients are derived and it is shown that the friction coefficient operative in self-diffusion is different from the corresponding friction coefficient in mutual diffusion. A common error in the recent literature concerning the expression for the mutual-diffusion coefficient is corrected. The importance of diffusion measurements in the study of solution dynamics has increased considerably in recent years as a result of the development of new measuring techniques. Thus, in addition to the classical concentration-gradient method, quasi- elastic light scattering is routinely used to determine the mutual-diffusion coefficient, whereas the radioactive tracer and the pulsed-field-gradient nuclear magnetic resonance techniques1-* are used to determine self-diffusion coefficients. Whereas earlier diffusion experiments were mainly concerned with dilute solutions, the present trend is to extend the measurements into the region of semi-dilute and concentrated solutions, where the interpretation of experimental data is naturally much more complicated.The problem has recently been considered by many ~ o r k e r s , ~ - ~ ~ although the results are inconclusive and in some instances conflicting.However, an essentially correct phenomenological treatment, based on the frictional formalism of non-equilibrium thermodynamics, has been given by Laity.13 This approach is further elaborated in the present treatment and clearly demonstrates the difference between the two diffusion processes. Accordingly, we use the general friction equations for multi-component diffusion xhjcj(ui-uj) = -grad pi ( i , j = 1,. . . n ) (1) I where f i r is the friction coefficient between the components specified by the indexes (in N m* s rnoF2), ua is the average velocity, ci is the concentration (in mol m-3) and pi is the chemical potential. Eqn (1) is independent of the choice of the reference frame since only velocity differences occur in the equations. Here a laboratory-fixed coordinate system is used as the diffusion coefficient is always measured in this reference frame.Eqn (1) is supplemented by the Onsager reciprocal relations14 f i j =hi the Gibbs-Duhem equation X ci grad pi = 0 i 17251726 DIFFUSION AND NON-EQUILIBRIUM THERMODYNAMICS and the equation for the conservation of mass c. ciui = 1 (4) i where vi is the partial molar volume (in m3 mol-l), which in the present treatment is assumed to be independent of concentration. By defining the fluxes we obtain from eqn ( 1 ) of the components Ji = ci ui Cfij (3 Ji - J j ) = -grad pi- 3 ci It is also important in this connection to consider the dissipation function, which may be written in the f0d4 where 4 is the dissipation function and o is the entropy source strength.The positive definite character of this function demonstrates that the friction coefficients fij are positive quantities. MUTUAL-DIFFUSION COEFFICIENT Considering first mutual diffusion in a binary solution (component 1 solvent, component 2 solute), eqn (6) reduces to a single independent equation f12(:Jl-J2) = -gradpl. 1 For volume-conserving systems the volume flux vanishes and we obtain the additional equation J1 v1 + J, U , = 0. Thus, eliminating J1 and using eqn (4) we obtain (9) (10) C l V l J, = - gradpl. f 1 2 The concentration dependence of the solvent’s chemical potential is often expressed in terms of the virial expansion of the osmotic pressure: pl-p; = -v1n = - R T u ~ ( c , + A , c ~ + .. .) ( 1 1 ) where A , is the second virial coefficient in molar concentration units. Inserting this expression into eqn (10) and using eqn (4) we obtain RT( 1 - C, v,), J, = - ( 1 + 2 A , c , + . . .) gradc,. f 1 2 C l Since the measurable diffusion coefficient is given by Fick’s law in a laboratory-fixed coordinate system we have RT( 1 - c, u2)2 f ( 1 + 2 A , c , + . . .) D , =H. VINK 1727 where f = fi2 c, is the molar hydrodynamic friction coefficient of the solute. (Note that f,, is defined as the friction coefficient between one mole of component 1 and one mole of component 2.) Eqn (1 3) corrects a common error in the recent literat~rel~-~' where in the corre- sponding equation the linear factor (1 - c, u,) is used (or omitted altogether). The dis- crepancy cannot be removed by formally defining alternative friction coefficients as this would introduce extraneous factors into the expression for the dissipation function.SELF-DIFFUSION COEFFICIENT The usual definition of self-diffusion originates from the Einstein-Smoluchowksi equation,22*23 which in one dimension has the form (A") = 2Dst (14) where (A2x) is the mean-square displacement of a molecule during the time interval t and D, is the self-diffusion coefficient. This definition is not flawless as it presumes that it is possible in some way to follow the trajectory of a diffusing particle. However, this is not the case and we can only specify the probability distribution of finding the particle at a given position along the x axis at a given time. Thus, for a given particle which at zero time is at the origin of the x axis, we may define the probability frequency function f (x, t) of finding the particle at the position x-x+dx at the time t.The mean-square displacement in eqn (14) then becomes equal to the variance of the probability distribution and we obtain (A") = ~2f(x, t) dx = 20, t. s': We now consider n moles of labelled molecules (in a tube with unit cross-section along the x axis) at the origin for t = 0 and study their concentration distribution as time proceeds. Denoting the concentration distribution by c* (x, t) we have the equation for conservation of mass : +co n = c*(x, t)dx. (16) -co Thus, the concentration distribution may be normalised : Note that the normalised concentration distribution f* (x, t) is identical with the probability frequency function of finding an arbitrarily chosen labelled molecule at the position x-x+dx at the time t.Diffusion of the labelled molecules obeys the diffusion equation which can be solved for very general boundary conditions by the method of moments.24' 25 Thus, multiplying both sides of eqn (18) by x2 and integrating, we obtain for the left-hand side1728 DIFFUSION AND NON-EQUILIBRIUM THERMODYNAMICS where p: is the variance of the normalised concentration distribution of labelled molecules. The right-hand side of eqn (1 8) yields (after two partial integrations and because the integrated parts vanish at the boundaries) +a -a 2 ~ f * + 2 j + ~ -a f*dx) = 2D*. ( 2 0 ) af* (I+a -a ax ax2 a 2 f * xaD*-dx = D* Thus and after integration &(t) = 2D*t+pL,*(O) ( 2 2 ) where p t (0) = 0 if the initial concentration distribution is sharp (6 function).Eqn ( 2 2 ) is identical to eqn (15) and from this we may draw the important conclusion that the probability distribution in self-diffusion of unlabelled molecules is equivalent to the normalised concentration distribution of labelled molecules, both satisfying the macroscopic diffusion equation. Thus, provided the molecular interaction parameters remain unchanged in the labelling process, the diffusion of labelled molecules is identical to self-diffusion of unlabelled molecules and D* = D,. In close conformity with this requirement the labelling in tracer diffusion is accomplished by the use of radioactive isotopes and in pulsed-field-gradient nuclear magnetic resonance by the orientation of nuclear spin.To apply the frictional formalism to self-diffusion in a binary solution we consider a three-component solution consisting of solvent (I), unlabelled solute (2) and labelled solute (3). Eqn (6) then yields The conditions for self-diffusion involve the following relations : c, + c3 = c = constant ( 2 6 ) f 1 2 = f 2 l = f 1 3 = f 3 1 f 2 3 = f 3 2 = f 2 2 u2 = v,. Eqn (26) implies (because molecules 2 and 3 have identical interaction parameters) gradpl = 0 ( 3 0 ) ( 3 1 ) C C and eqn ( 2 3 ) yields J , = L ( J 2 + J,). The condition that the volume flux vanishes has the form J 1 u l + ( J 2 + J 3 ) u 2 = 0 J1 = 0 and J2+ J 3 = 0. and eqn (31) and ( 3 2 ) yield ( 3 2 ) ( 3 3 )H.VINK 1729 From eqn (3), (30), (33) and (24)-(26) we obtain the flux of the labelled molecules i t remains to consider the thermodynamics of the system of labelled solute molecules. The chemical potentials are of the form (35) (36) where we have put pi = pg and the activity coefficient f has the same value as in a solution of pure component 2 at the concentration c = c, + c,. Eqn (35) and (36) may also be written in the form p2 = pi + RT In dfc,) p, = pi + RT In Cfc,) p2 = pi + RT In f c + RT In (c,/c) p, = pi + RT In f c + RT In (c3/c). (37) (38) For the mean chemical potential ~ 2 3 of the labelled system we obtain C c2p2+c3p3 = pi+RTln(fc)+ RT(2 In ( c p / c ) + s ln(c3/c)). (39) C C p23 = As c2/c and c,/c are the mole fractions of unlabelled and labelled solute molecules we find that the labelled system forms an ideal solution between its components, the last term in eqn (39) representing the entropy of mixing of the system.From eqn (36) we obtain RT grad c, c3 gradp3 = and eqn (34) yields J3 = - grad c, (41) f 1 2 C l + f 2 2 c RT and thus Considering the meaning of the even friction coefficient f i 2 we observe that this coefficient is not defined in eqn (1) because the velocities in the friction equation are pre-averaged26 and therefore all molecules of the same kind have the same (average) velocity. By the labelling process velocity differences between molecules of the same kind become perceptible and the even friction coefficients acquire physical significance. DISCUSSION In eqn (13) and (42) the difference between mutual diffusion and self-diffusion is clearly displayed.Although the magnitudes of the two diffusion coefficients are not directly comparable, because of the thermodynamic factor present in the mutual- diffusion coefficient, we find that the friction coefficient in self-diffusion differs from the frictioncoefficient in mutual diffusion, because the former contains the solute-solute friction term. The physical significance of this term becomes apparent in the condition for the vanishing of the volume flux. In mutual diffusion the solute flux is compensated by the back-flow of solvent, and the frictional process may be characterised as solvent flow past a lattice formed by solute molecules. In concentrated polymer solutions the 57 FAR 11730 DIFFUSION AND NON-EQUILIBRIUM THERMODYNAMICS process actually resembles the swelling-deswelling of a gel (hence the alternative term ‘cooperative diffusion’ for this process).On the other hand in self-diffusion the solvent is stationary and the flux of labelled solute molecules is compensated by the back-flow of unlabelled solute. The process is characterised by the interchange of position between neighbouring solute molecules, resulting in solute-solute frictional inter- actions. In concentrated polymer solutions, where entanglement effects are important, these interactions may become very large, and it follows from scaling arguments2’ that the friction coefficient in self-diffusion increases much more rapidly with concentration than the friction coefficient in mutual diffusion.Although the form of the friction coefficients seems to indicate that the friction coefficient in self-diffusion is larger than the corresponding coefficient in mutual diffusion, this may not always be true, since the coefficientf,, is not necessarily the same for the two diffusion processes as they represent different modes of molecular motion. E. 0. Stejskal and J. E. Tanner, J. Chem. Phys., 1965,42, 288. T. L. James and G. G. McDonald, J. Magn. Reson., 1973,11, 58. P. T. Callaghan, C. M. Trotter and K. W. Jolley, J. Magn. Reson., 1980, 37, 247. P. Stilbs and M. E. Mosely, Chem. Scr., 1980, 15, 176. G. D. J. Phillies, J. Chem. Phys., 1974, 60, 976; 983. J. L. Anderson and C. C. Reed, J. Chem. Phys., 1976, 64, 3240.G. D. J. Phillies, J. Chem. Phys., 1977, 67, 4690. R. S. Hall, Y. S. Oh and C. S . Johnson, J. Phys. Chem., 1980,84, 756. J. A. Marqusee and J. M. Deutch, J. Chem. Phys., 1980,73, 5396. lo M. M. Kops-Werkhoven, A. Vrij and H. N. W. Lekkerkerker, J. Chem. Phys., 1983,78,2760. l1 J. M. Shurr, Chem. Phys., 1982,65, 217. l2 W. Brown, P. Stilbs and R. M. Johnsen, J. Polym. Sci., Polym. Phys. Ed., 1983, 21, 1029. l3 R. W. Laity, J. Phys. Chem., 1959, 63, 80. l4 H. Vink, Acta Chem. Scand., Ser. A, 1984, 38, 335. l6 H. Yamakawa, Modern Theory of Polymer Solutions (Harper & Row, New York, 1971), p. 262. l6 R. G. Kitchen, B. N. Preston and J. D. Wells, J. Polym. Sci., Polym. Symp., 1976, 55, 39. l7 G. D. J. Phillies, G. B. Benedek and N. A. Mazer, J. Chem. Phys., 1976,65, 1883. l8 R. Bergman and L. 0. Sundelof, Eur. Polym. J., 1977, 13, 881. l9 J. Roots, B. Nystrom, L. 0. Sundelof and B. Porsch, Polymer, 1979, 20, 337. 2o B. N. Preston, W. D. Comper, A. E. Hughes, I. Snook and W. van Megen, J. Chem. Soc., Faraday 21 W. Brown, P. Stilbs and R. M. Johnsen, J. Polym. Sci., Polym. Phys. Ed., 1982, 20, 1771. 22 A. Einstein, Ann. Phys., 1905, 17, 549. 23 M. v. Smoluchowski, Ann. Phys., 1906, 21, 756. 24 H. Vink, Nature (London), 1965,205, 73. 26 H. Vink, J. Chromatogr., 1977, 135, 1. 26 H. Vink, J. Chem. Soc., Faraday Trans. I , 1983,79, 2355. 27 P. T. Callaghan and D. N. Pinder, Macromolecules, 1981, 14, 1334. Trans. I , 1982, 78, 1209. (PAPER 4/ 1709)

ISSN:0300-9599    

DOI:10.1039/F19858101725

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. I, 1985,81, 1731-1744 Ion Exchange in Zeolites The Exchange of Cadmium and Calcium in Sodium X Using Different Anionic Backgrounds BY PHILIP FLETCHER~ AND RODNEY P. TOWNSEND* Department of Chemistry, The City University, Northampton Square, London ECl V OHB Received 18th October, 1984 Ion-exchange characteristics have been determined for the binary exchanges of Cd2+ and Ca2+ in sodium X zeolite using a variety of anionic backgrounds. A thermodynamic model, previously presented for the evaluation of the activity coefficients of salts in mixed electrolyte solutions, is applied to calculate the thermodynamic data for these exchanges. The application of this model shows that changes in cadmium selectivity as a function of changing anionic background can be rationalised entirely in terms of changes in the magnitude of solution-phase activity coefficients.The reversibility of the systems has been tested carefully in order to obtain information on the degree to which the zeolite is affected by hydronium exchange and/or acid damage of the structure. It is shown that the ability of divalent ions to undergo reversible and stoichiometric exchange in zeolite X is consistent with trends observed in their acid association constants. In addition, and contrary to some exchanges in clay systems, it is shown that the highly negatively charged environment of zeolite X efficiently excludes associated ion pairs. In recent years there have been many detailed studies of the ion-exchange characteristics of zeolites. Most if not all of these have been determined with an electrolyte background containing only one coanion.This use of a monoanionic background is adequate for thermodynamic characterisation of exchange phenomena. However, in applications of ion exchange in industry and also when exchange reactions occur in nature, the exchange process is likely to be in the presence of mixed background anions. Thus, the objectives of this paper are two-fold. First, a recently derived thermodynamic procedure, which allows the calculation of single-ion activity coefficient ratios in multicomponent electrolyte solutions, is presented and tested. Secondly, ion-exchange isotherms for the binary exchanges between cadmium(1r) and sodium and also calcium(rr) and sodium, in zeolite X, in a range of anionic media are presented.Consequently, some of the effects that can occur through changes in the anionic background are investigated and discussed. EXPERIMENTAL MATERIALS Synthetic sodium zeolites were supplied by Union Carbide. Chemicals used for both analysis and exchange purposes were of analytical-reagent grade. PREPARATION OF HOMOIONIC ZEOLITES Initially, 200 g of sodium X were treated with 500 cm3 of 0.5 mol dm-3 sodium chloride solution. After centrifuging and minimal washing, the zeolite was dried at 40 "C and stored 7 Present address: Department of Agricultural Science, Oxford University, Parks Road, Oxford OX1 3PF. 1731 57-21732 ION EXCHANGE IN ZEOLITES Table 1. Zeolite analysis zeolite unit cell formulae Na-X Ca-X Na84.6(A102)84.7(si02)107.3 .25 *7 H2O ca42.4(A102)84.9(si02)107.1 ' z6l *8 H,O Cd-X (prepared from NO;) Cd-X (prepared from Cl-) Cd43~l(A102)84~4(Si02)lo,,6 - 264.3 H,O Cd43,4(A102)84.5(Si02)lo7,5~ 264.9 H,O Table 2. Thermodynamic data for Cd + Na exchange in NO; 0.901 0.808 0.71 1 0.616 0.482 0.374 0.265 0.184 0.177 0.108 0.099 0.0591 0.0 185 5.63 x 10-3 1.25 x 10-3 4.8 1 x 10-4 2.01 x 10-4 0.953 - 0.924 0.899 0.875 0.850 0.83 1 0.819 0.794 0.779 0.752 0.738 0.698 0.608 0.534 0.339 0.260 0.191 .0.063 0.378 0.728 0.986 1.436 1.808 2.322 2.605 2.717 2.892 2.869 3.132 3.719 4.283 4.819 5.285 5.618 1.659 1.645 1.631 1.617 1.597 1.579 1.562 1.549 1.548 1.536 1.535 1.528 1.521 1.519 1.518 1.518 1.517 0.433 0.876 1.217 1.467 1.905 2.265 2.768 3.043 3.154 3.321 3.298 3.356 4.138 4.700 5.236 5.702 6.088 0.93 1 0.873 0.819 0.764 0.707 0.665 0.639 0.588 0.597 0.509 0.485 0.424 0.318 0.258 0.145 0.105 0.0709 0.203 0.262 0.3 17 0.374 0.435 0.482 0.51 I 0.571 0.560 0.666 0.695 0.769 0.887 0.939 0.985 0.997 1.009 in a desiccator over saturated sodium chloride solution for three weeks prior to analysis and exchange.Samples of this equilibrated sodium X (Na-X) were converted into the homoionic cadmium and calcium forms by exhaustively exchanging 3 g samples of Na-X with solutions containing 0.1 mol dmP3 of the salts Cd(NO,),, CdCl, or Ca(NO,),, respectively. The zeolites were then dried at 40 "C and water-vapour equilibrated over saturated sodium chloride solution in dessicators. The pH values of the cadmium solutions before adding zeolite were 4.34 [Cd(NO,),] and 4.79 (CdC1,).After equilibration with the zeolite, the pH values were 5.82 and 6.08, respectively. ANALYSIS OF HOMOIONIC ZEOLITES Samples were analysed for SiO, content by treatment with HF. Total aluminium was determined gravimetrically using 8-hydroxyquinoline. Cadmium was found not to interfere with this analysis as long as the precipitation was performed at pH < 5. Sodium and cadmium were determined by dissolution of the zeolite in 1 : 1 nitric acid and subsequent analysis using flame photometry and atomic absorption spectroscopy (a.a.s.), respectively. Water was determined thermogravimetrically. The combined analytical data are expressed as unit-cell compositions in table 1 . CONSTRUCTION OF ION-EXCHANGE ISOTHERMS The ion-exchange characteristics of the Cd2+/Na+ exchange in Na-X were determined using three different anionic backgrounds: (a) 0.1 mol dm-3 NO,, (b) 0.1 mol dmP3 C1- and (c) solutions containing equimolar quantities of C1- and NO;, the total concentration of anionP.FLETCHER AND R. P. TOWNSEND 1733 Table 3. Thermodynamic data for Cd Na exchange in C1- ECd ‘Cd In K , r In K , gCd 0.966 0.929 0.859 0.761 0.676 0.576 0.459 0.359 0.2 19 0.152 0.0938 0.0502 0.0 183 9.66 x 10-3 1.83 x 10-3 6.16 x 10-4 2.24 x 10-4 0.97 1 0.946 0.9 18 0.880 0.859 0.836 0.8 15 0.793 0.750 0.709 0.654 0.604 0.550 0.47 1 0.288 0.205 0.1 19 - 1.286 - 1.044 - 0.459 - 0.0866 0.293 0.663 1.111 1.444 1.900 0.070 2.285 2.628 3.354 3.531 4.129 4.656 4.919 4.530 4.53 1 4.534 4.538 4.543 4.550 4.562 4.575 4.601 4.618 4.636 4.65 1 4.662 4.665 4.669 4.669 4.669 0.225 0.467 1.053 1.426 1.808 2.I77 2.629 2.964 3.426 3.599 3.792 4. I65 4.893 5.073 5.660 6.196 6.459 0.963 0.922 0.867 0.786 0.740 0.690 0.645 0.599 0.517 0.447 0.368 0.3 10 0.258 0.200 0.109 0.078 1 0.0496 0.184 0.227 0.28 1 0.360 0.407 0.459 0.507 0.556 0.651 0.773 0.829 0.898 0.954 1.006 1.045 1.045 1.036 Table 4. Thermodynamic data for Cd Na exchange in CI-/N03 ECd ‘Cd In K , r In K , gCd 0.93 1 0.859 0.773 0.675 0.562 0.475 0.450 0.359 0.225 0.141 0.0783 0.05 1 I 0.02 10 8.31 x 10-3 1.33 x 10-3 0.959 - 0.93 1 0.899 0.871 0.848 0.832 0.828 0.810 0.772 0.740 0.678 0.624 0.570 0.503 0.312 .0.418 0.0238 0.288 0.623 1.052 1.367 1.463 1.776 2.266 2.586 2.804 2.895 3.487 4.02 1 4.746 2.747 2.740 2.73 1 2.723 2.7 13 2.705 2.703 2.695 2.683 2.674 2.664 2.666 2.669 2.663 2.662 0.593 1.032 1.293 1.625 2.050 2.362 2.457 2.768 3.253 3.570 3.784 3.876 4.469 5.000 5.725 0.945 0.895 0.830 0.771 0.721 0.688 0.679 0.642 0.577 0.51 1 0.418 0.347 0.291 0.234 0.1 13 0.202 0.250 0.31 1 0.366 0.413 0.446 0.454 0.490 0.552 0.622 0.72 1 0.789 0.841 0.889 0.970 again being 0.1 mol dmP3.The Ca2+/Na+ exchange was measured with two different back- grounds, viz. (a) 0.1 mol dm-3 NO; and (b) 0.1 mol dmP3 CIO;. In this case no mixtures of anions were examined. All equilibria were constructed at a temperature of 298 f 0.5 K. Each forward equilibrium point was determined by equilibrating a 0.2 g sample of Na-X with ca. 50 cm3 of a solution containing a binary mixture of sodium and the appropriate divalent cation and using the chosen anionic background.Equilibrium solutions varied in composition between 0 and 100% Na but the normality was always kept constant. In order to obtain points at the isotherm extrema, the ratio of zeolite to solution was varied suitably. It was assumed that equilibrium was obtained after 5 days. Subsequent separation of phases was by centrifugation. The resulting solution phase was analysed for its cation composition by flame photometry and a.a.s. The solid1734 ION EXCHANGE IN ZEOLITES Table 5. Thermodynamic data for Ca Na exchange in NO3 ECd ‘Cd In K , r In KG g C d 0.971 0.921 0.871 0.848 0.768 0.682 0.583 0.469 0.985 0.955 0.921 0.894 0.852 0.818 0.785 0.743 - 0.277 -0.448 -0.573 - 0.836 -0.607 -0.31 15 0.013 0.302 0.390 0.712 0.494 0.258 0.667 0.943 0.167 0.629 1.334 0.106 0.584 1.627 0.0455 0.5 15 2.171 0.0373 0.48 1 2.183 0.01 55 0.390 2.573 1.740 1.732 1.725 1.721 1.709 1.696 1.678 1.661 0.276 0.102 -0.0279 - 0.293 -0.0708 0.217 0.532 0.810 0.985 0.955 0.9 18 0.887 0.833 0.785 0.734 0.665 0.523 0.526 0.537 0.550 0.578 0.606 0.638 0.685 .648 0.993 0.6 12 0.723 .626 1.429 0.534 0.78 1 .611 1.81 1 0.470 0.833 .599 2.097 0.398 0.894 .589 2.634 0.301 0.183 .588 2.645 0.260 1.022 334 3.033 0.174 1.102 6.18 x 10-3 2.284 2.874 1.582 3.333 0.1 14 1.132 3.67 x 10-3 2.024 3.106 1.582 3.564 0.0989 1.127 1.743 x lod3 0.164 3.289 1.58 1 3.748 0.0847 1.097 1.08 x 10-3 0.144 3.287 1.581 3.748 0.0825 1.068 3.75 x 10-4 0.0521 3.43 1 1.581 3.888 0.0896 1.029 Table 6.Thermodynamic data for C a e N a exchange in ClO; 0.951 0.902 0.772 0.545 0.378 0.23 1 0.09 1 0.0 16 8.88 x 10-3 3.04 x 10-3 4.65 x 10-4 0.971 -0.540 0.943 -0.418 0.861 -0.511 0.774 0.141 0.714 0.580 0.66 1 1.080 0.59 1 1.859 0.443 2.850 0.334 2.812 0.229 3.226 0.07 15 3.573 1.606 1.599 1.579 1.542 1.514 1.489 1.463 1.450 1.448 1.447 1.447 -0.656 -0.01 18 - 0.0540 0.574 0.995 1.478 2.239 3.22 1 3.183 3.596 3.943 0.970 0.940 0.836 0.702 0.599 0.509 0.397 0.2 15 0.137 0.0959 0.0826 0.48 1 0.494 0.555 0.646 0.722 0.794 0.890 1.061 1.123 1.121 1.041 phase was washed (briefly) twice with 40 cm3 of distilled water and dissolved in ca.50 cm3 of 1: 1 nitric acid at room temperature prior to analysis by flame photometry and a.a.s. The resulting data are shown in tables 2-6. REVERSIBILITY Reverse points are constructed by exchanging 0.2 g samples of Na-X with 100 cm3 aliquots of solutions containing 0.05 mol dm-3 Cd2+ or Ca2+.This was followed by separation and re-equilibration with 50 cm3 of a suitable isotherm solution containing both sodium and the divalent cation. Total analysis of all cations in all phases was as before. Several reverse points for each isotherm were constructed in this manner. It has been reported that transition-metal ions may bind into zeolite X irreversibly. 1-3P. FLETCHER AND R. P. TOWNSEND 1735 Consequently, exhaustive re-exchanges of samples of homoionic Cd and Ca zeolites were performed using 0.1 mol dm-3 sodium nitrate solution. The resulting fractions of irreversibly bound Ca and Cd along with similar data for the exchange of Zn2+, Mn2+, Co2+, Ni2+ and Cu2+ in synthetic X were found to be (in percentage terms) Mn2+ 0.3 % , Zn2+ 0.5 % , Co2+ 4.8 % , Ni2+ lO.l%, Cu2+ 21.3%, Cd2+ 0.8% and Ca2+ 0.0%.THEORY Any binary ion-exchange reaction may be written as Z , A(S) -I- Z A B(C) ZB A(C) + Z A B(S) (1) where 2, and 2, refer to the valencies of the ions A'A and B Z ~ , respectively, and (c) and (s) refer to crystal and solution phases. The thermodynamic equilibrium constant is expressed as Ka = (EzB mgA/EgA mzB) r (&B/&) (2) where EA and EB are the equivalent fractions of ion A and B in the crystal, mA and mB are solution molalities of A and B, respectively, gA and gB represent crystal-phase activity coefficients and is defined as the ratio of solution-phase single-ion activity coefficients (yi) raised to the respective charges, i.e.In order to evaluate both the equilibrium constant and the crystal-phase activity coefficients, Gaines and Thomas4 derived the following: The crystal-phase activity coefficients g A and g B (at any crystal composition EA, EB) are given by P l and where KG is a measurable quantity defined as and the symbol * denotes that KG is the value at a composition (EA, E,). This particular thermodynamic formulation is valid under conditions where imbibition of neutral electrolyte is negligible, which for zeolites is at solution con- centrations < 0.5 mol dm-3.5 A second condition is that the overall contributions made by water-activity changes in the zeolite are negligible. This point has been discussed by Barrer and Klinowski,6 and their conclusions, in addition to other experimental evidence,'? justifies the assumption that this condition holds normally for zeolites.Consequently, a plot of In KG against EA for any binary reaction in zeolites should be independent of both total solution normality and the nature of the coanionic background. Thus, the requirements for a precise determination of K, are accurate experimental data over the complete isotherm range and correct evaluation of r.1736 ION EXCHANGE IN ZEOLITES EVALUATION OF r In order to calculate r it is necessary to consider appropriate theories of mixtures The mean molal stoichiometric activity coefficient (7 +) and the single-ion activity of electrolytes in aqueous solution. coefficients for any electrolyte AiXj are related by the general equation For a case involving two cations A and B and two anions X and Y there are four stoichiometric activity coefficients to be considered, viz.Eqn (9) and (10) share the term In yx. Thus, by multiplying eqn (9) by Z , (2, + Zx)/Z, and eqn (10) by ZA(ZB+Zx)/Zx the ratio ~ Z ~ A / Y ~ B can be evaluated by elimination of terms in yx, i.e. 1 In =-[zA(zB+zX) In Y + S X - z B ( z A + z X ) In ?_+AX]. (13) Z X Similarly, the ratio r can be derived by elimination of terms in yy from eqn (1 1) and (12), i.e. (14) 1 In =-[zA(z13+zY) In Y+BY-ZB(ZA+ZY) In Y+_AYI- Z Y The two r values in eqn (1 3) and (14) are by definition equal. This is a consequence of the fact that activity coefficients in mixtures are not independent quantities but linked by the cross-differentiation rule9 and the Gibbs-Duhem equation.lo This important point was emphasised long ago by Guggenheim.” Thus the general relationship for the ratio in any mixed electrolyte is In order to evaluate r from eqn (13), (14) or (15) the value of y + for the necessary salts must be correctly evaluated in the mixed electrolyte solution. For this purpose Fletcher and Townsend12 presented a model which allows calculation of the stoichio- metric activity coefficient of any component in a mixture in terms of the activity coefficients of the solution components in pure solutions at the same ionic strength as the mixture. This model is an extension of a treatment developed by Guggenheim13 and is consistent with similar approaches by Scatchard14 and Pitzer.15 In principle, the model is a regular solution theory16 based upon the elements of statistical mechanics,” which conforms to Bronsted’s theory of specific ion interaction.’* TheP.FLETCHER AND R. P. TOWNSEND 1737 fundamental principles behind these theories have been reviewed in depth by Harned and Robinson.19 The general equation of Fletcher and Townsend12 is I log y(,A;i;.$m* x 2 . ' . x n ) = log y*,A, XI + 4r(2A I + zX ,) [ii ( mAi [ail log 7: A, XI + log r r A i Xi + ai3 A( + / 'dl)-'l where is the mean molal stoichiometric activity coefficient for salt AIXl in the presence of (m- 1) and (n- 1) other cations and anions, respectively, I is the ionic strength of the mixture, A is the Debye-Huckel constant,1° which has the value 0.51 15 mol-4 dmi at 25 "C, the terms mpi etc.represent molalities (mol kg-l) of the cations and anions and the symbol * indicates a pure-solution measurement. For the simple case in question (i.e. cations A and B in the presence of anions X and Y) eqn (16) for the salt AX simplifies to where Similar equations for log y AY, log y+ BX and log y+ can be derived from eqn (1 6). Thus r can be evaluated using either eqn (13) -br (14) and the correct pair of stoichiometric activity coefficients from eqn (1 6). As explained previously, one condition of thermodynamic consistency is the uniqueness of the r value for a given solution composition and ionic strength. The symmetry of eqn (16) ensures that this is the case.1738 ION EXCHANGE IN ZEOLITES RESULTS AND DISCUSSION REVERSIBILITY ACID HYDROLYSIS All reverse points measured for the five isotherms were congruent with the forward isotherms within the limits of experimental uncertainty.* This reversibility is consistent with data for the Ca Na exchange in zeolite X presented by Sherry20 and also the results of Wolf et Stoichiometric, reversible exchange is a characteristic feature of alkaline-earth exchange in faujasites.20v 22 An isotherm for Cd T- Na exchange has been determined by Gal and R a d o ~ a n o v , ~ ~ but reversibility was not reported. However, isotherm irreversibility and non-stoichiometric exchange has been reported in studies on many transition metals in synthetic faujasites.2v3 In parallel studies to this one Ni2+ was suspected of precipitating basic salts within zeolite X,2 and Cu2+ showed signs of dealuminating this zeolite by acid hydr~lysis,~ thus reducing the exchange capacity, and binding copper into the zeolite framework.Other ions (Zn2+, Mn2+ and Co2+) showed some signs of irreversible binding293 but to a much smaller extent than Cu2+. In addition, Sherry24 has reported that cadmium exchanged to a level > 100% of the exchange capacity (as estimated from the aluminium content) of sodium A. In view of these previous r e p o r t ~ l - ~ ? ~ ~ further tests on the reversibility of both calcium and cadmium exchange were performed by exhaustive re-exchange (see Experimental) in an attempt to elute all the calcium and cadmium out of the zeolite. These data show that < 1% of cadmium in the zeolite was irreversibly bound, compared with > 20% copper, 10% nickel, 5% cobalt and smaller quantities of zinc and Calcium showed no sign of irreversible binding.The analytical data in table 1 confirm stoichiometric exchange of both calcium and cadmium, although cadmium exchange from both chloride and nitrate background does appear to go to ca. 102% of the exchange capacity as calculated from the zeolite aluminium content. This may either be a reflection of experimental uncertainty or it could be due to some precipitation and/or exchange of a small quantity of Cd(OH)+ species (see below). The observed ‘hydrolysis’ of zeolite X by transition metals may be partially rationalised by consideration of the association constants for acid hydrolysis of these ions in aqueous solution. Using data taken from various source^,^^-^^ values of the decadic logarithm of the first association constant for the formation of MOH+ from M2+ and OH- may be compared. Although there is some controversy over the precise values of these constants, the trends are nevertheless quite clear.Calcium and magnesium (log values of 1.4 and 2.2, respectively) have a much lower acidic reaction with water than do Co2+, Mn2+, Zn2+, Ni2+ and Cd2+ ions, all of which have log values of ca. 4. In contrast, Cu2+ and Fe2+ (log values of 6.0 and 5.7, respectively) have association constants some two orders of magnitude greater than this, and both of these ions are known to encourage hydronium exchange in zeolites, or even to dealuminate them.3 The intracrystalline environment in zeolites is clearly different from a purely aqueous environment, and this analogy with the aqueous acidic reaction must not be held too rigorously.Nevertheless it does appear that a first association constant of > lo4 is one criterion indicating the borderline of stability for divalent cations in zeolite X. Clearly cadmium is in this border region. If this simple analogy holds then it would be predicted that Pb2+ and Hg2+, having first association constants of ca. lo8 and ca. 10l2, respectively, would show significant irreversibility in zeolite X. This has been shown for lead; reversible exchange in zeolite X, or even zeolite Y , seems impossible.28P. FLETCHER AND R. P. TOWNSEND 1739 Table 7. Percentage of cadmium species in CdCl, isotherm solutions Cd2+ CdC1+ CdCli CdC1; CdC1;- ECd 0.966 0.929 0.859 0.761 0.676 0.576 0.459 0.359 0.219 0.152 0.0938 0.0502 0.0 183 9.66 x 10-3 1.83 x 10-3 6.16 x 10-4 2 .2 4 ~ 10-4 26.98 26.94 26.46 25.92 25.69 25.14 24.23 23.74 23.22 22.3 1 20.71 19.86 19.63 19.51 19.54 19.43 19.38 63.06 63.09 63.10 63.10 63.15 63.4 64.1 64.3 64.7 65.1 66.7 67.5 67.7 67.8 67.7 67.8 67.8 9.55 9.55 10.00 10.47 10.63 10.91 1 1 . 1 1 11.36 11.47 11.71 11.93 11.97 11.99 12.01 12.07 12.07 12.11 0.363 0.371 0.398 0.447 0.463 0.48 1 0.50 0.52 0.54 0.563 0.574 0.581 0.589 0.593 0.599 0.603 0.607 0.043 0.044 0.05 1 0.059 0.059 0.063 0.067 0.07 1 0.074 0.078 0.083 0.084 0.087 0.089 0.09 1 0.090 0.092 EXCHANGE OF COMPLEXED CADMIUM SPECIES Recently the question of non-stoichiometric exchange involving divalent ions in soils and clay minerals has been considered in detail by Sposito et al.29930 In these studies, the hypothesis was forwarded that such phenomena could be explained in terms of exchange involving not only the divalent ions but also ion pairs.It is known that almost all anions and cations form associated species to some extent, although (as emphasised above) there is some doubt about the precise values of association constant^.^^-^' Sposito et al.29 re-interpreted the results of Maes et al.31 (which showed non-stoichiometric exchange of transition metals in sodium montmorillonite) in terms of metakhloride ion pairs MCl+, and they also produced data for copper(r~),~~ calcium and magnesium exchanges in montmorillonite30 which supported this hypo thesis. Sherry2* suggested initially that cadmium acetate and chloride complexes were responsible for the degree of over-exchange they observed with cadmium in sodium A, although they concluded finally that sample irreproducibility was a more likely cause of the observed behaviour.In view of this, it seems appropriate to consider the effect that cadmium nitrate and cadmium chloride complexes may have upon the exchange of Cd2+ in zeolite X. Clearly, the stoichiometric and reversible exchange of cadmium leaves no doubt that cadmium complexes do not exchange to any large extent in zeolite X when the total solution normality is low (in contrast to the case with montmorillonite where 17 % over-exchange However, it is instructive to consider the extent to which nitrate and chloride complexes may form in the cadmium isotherm solutions. Chemical speciation calculations were therefore performed on these solutions using association constants taken from Martel and SmithZ5 and The numerical procedure was one of simple iteration, based upon techniques outlined by Sposito and M a t t i g ~ d .~ ~ The levels of complex formation calculated for each of the three sets of cadmium isotherm solutions (tables 2-4) are shown in tables 7-9. The concentration of each cadmium species is expressed as a percentage of the total cadmium present in solution. Clearly, cadmium forms only1740 ION EXCHANGE IN ZEOLITES Table 8. Percentage of cadmium species in CdC1,/Cd(N03), isotherm solutions ECd Cg+ CdCI+ CdCl; CdC1; CdC1;- CdNOl 0.93 1 0.859 0.773 0.675 0.562 0.475 0.450 0.459 0.225 0.141 0.0783 0.05 1 1 8.31 x 10-3 1.33 x 10-3 48.5 47.2 45.9 44.49 42.88 41.45 40.14 38.51 36.68 36.23 34.57 34.58 34.57 34.34 46.77 47.86 48.97 50.13 5 1.29 52.45 53.70 54.95 56.23 56.30 57.54 57.54 57.54 57.54 2.82 3.02 3.24 3.55 3.98 4.27 4.37 4.79 5.37 5.75 6.17 6.23 6.37 6.46 0.046 0.053 0.058 0.069 0.083 0.096 0.100 0.1 15 0.141 0.159 0.182 0.187 0.195 0.20 2.4 x 10-3 3.0 x 10-3 3.5 x 10-3 4.4 x 10-3 5.6x 10-3 7.3 x 10-3 9.1 x 10-3 6.8 x lop3 1.2 x 10-2 1.5 x 1.7 x 1.8 x 10-2 1.9 x lop2 2.0 x 10-2 1.86 1.86 1.82 1.78 1.76 1.72 1.68 1.63 1.57 1.54 1 S O 1.46 1.44 1.42 Table 9.Percentage of cadmium species in Cd(NO,), isotherm solutions ECd CdZi CdNO,+ 0.901 0.808 0.71 1 0.6 16 0.482 0.374 0.265 0.184 0.177 0.108 0.099 0.059 1 0.0 185 5.53 x 10-3 1.25 x 10-3 4.81 x 10-4 2.01 x 10-4 93.6 93.5 93.3 93.2 93.0 92.8 92.7 92.5 92.5 92.3 92.7 92.8 92.6 92.6 92.6 92.55 92.55 6.4 6.5 6.7 6.8 7.0 7.2 7.3 7.5 7.5 7.7 7.3 7.2 7.4 7.4 7.4 7.45 7.45 weak complexes with nitrate.However, the solutions containingchloride are dominated by the species CdCl+, and ClCli is present in substantial proportions (tables 7 and 8). In fact, the data in table 7 show that the solution concentration of hydrated Cd2+ in the chloride isotherm is only ca. 20% of the total cadmium concentration. Since these species are present in solution, and, in contrast to m~ntmorillonite,~~ little over-exchange is observed with Na-X, we must therefore conclude that the intra- crystalline environment of zeolite X is sufficiently negatively charged to preclude substantial penetration of these loosely bound associated species even when ionP.FLETCHER AND R. P. TOWNSEND 1741 Table 10. Equilibrium constants and free energies and polynomials for In K G as a function of E, In KG = 7.554- 10.409Ec,+ 17.6O1Pcd- 15.205E& In KG = 2.982 - 8.99 1 E,, - 26.638PCa + 14.486PC, ~~ reaction anion Ka A G e Cd(s) + 2Na(c) e Cd(c) + 2Na(s) C1- 3 1.10 -4.26 & 0.17 NO, 30.67 -4.24k0.21 CI-/NO; 37.33 -4.48 & 0.2 Ca(s)+2Na(c)*Ca(c)+2Na(s) NO; 3.25 - 1.46f0.10 c10, 3.45 - 1.53 f 0.10 association in the electrolyte solution is extensive. This is a Donnan exclusion effect, and therefore one must not exclude the possibility that at high solution concentrations these complexes may be imbibed.5 It is known that when the total solution normality is high, salt imbibition may O C C U ~ .~ SELECTIVITY TRENDS AND THE THERMODYNAMIC DATA Certain characteristic features are exhibited by the data presented in tables 2-6. Both cadmium and calcium exhibit a high affinity for zeolite X compared with sodium. The standard free energy for Cd f Na exchange (table lo), being -c - 4 kJ equiv, reflects the high preference for Cd2+ over Na+ ion. The values (table 10) are consistent with the value of -4.2 kJ equiv measured previously by Gal and R a d o ~ a n o v . ~ ~ Similarly, the Ca Na isotherms and their respective Kielland plots (fig. 1) are consistent with other reported measurements on the calcium exchange in Na-X.20-22 However, fig. 1 shows a change in selectivity for calcium at high calcium loadings, and since exchanges of Na+ ions in the hexagonal prism and sodalite cage sites are probably involved at this exchange level, this selectivity change may well be associated with this phenomenon.Since the difficulties of measuring accurately data at isotherm extrema are well known,34 we choose not to place overmuch significance on this selectivity change. Values of equilibrium constants and standard free energies for all the exchange reactions are summarised in table 11, together with the best fitting polynomial equations, which describe In KG as a power series in EA. EFFECT OF CHANGING THE COANION Interesting differences between the cadmium and calcium exchanges are exhibited when the exchanges occur in different anionic media; the two calcium isotherms are indistinguishable, whereas the cadmium isotherm shows a distinctly different shape for each of the three anionic backgrounds.8 The data of tables 2-4, plotted in fig.2 as mass-action quotients, where K , = ( E ~ B miA/E;A mzB) (18) confirm the fact that the experimental preference for cadmium increases in the order NO; > NO;/Cl- > C1-. In contrast, once the solution-phase activity correction has been applied using eqn (16) and (17), the corrected selectivity quotients (KG) determined for each system become indistinguishable (fig. 1). These data serve to establish two points. First, the equations of Fletcher and Townsend12 for activity1742 ION EXCHANGE IN ZEOLITES 5t -' t 0 .o 0.5 1 .o E M Fig. 1. - 6 5 L 3 kE - c 2 1 0 -1 0 0 0 0 vo 0 0 0 a.- x 0.0 0.5 1 .o EM - Fig. 2. Fig. 1. Logarithms of the corrected selectivity quotient KG [eqn (711 plotted against EM, where M = Cd (0, V, 0) or Ca (a, V) as appropriate. Note that the effect of applying the solution phase r correction has been to make all the cadmium data coincident, irrespective of the nature of the accompanying coanion (see text). 0 and 0, NO,; D, NO;/Cl-; 0, C1-; V, ClO;. Fig. 2. Logarithms of the mass action quotients K, [eqn (19)] for the data given in tables 2-6 plotted against the equivalent fraction EM, where M = Cd or Ca as appropriate. Symbols as in fig. 1. coefficients in mixed solutions are adequate at relatively low concentrations, even in solutions where significant ion association occurs. The changes in cadmium selectivity are simply due to changes in the solution-phase activity coefficients.Secondly, the correct evaluation of r is most critical for a precise thermodynamic description of some ion-exchange reactions. The magnitude of the excess free energy due to solution non-ideality (G,3 is Values of G,E for the five equilibria were calculated from data in tables 2-6 and are given in table 11. These data show quite clearly that in all cases the value of G,E is large enough to make a substantial contribution to the overall free energy of reaction. Indeed, the magnitude of G,E for the Cd/Na(Cl)-X isotherm is ca. 2 kJ equiv-l, which amounts to almost 50% of the value of the free energy of exchange (table 10). In order to emphasise the importance of eqn (1 6), fig. 3 shows the values of r evaluated forP. FLETCHER AND R.P. TOWNSEND 1743 Table 11. Excess free energies as defined in eqn (19) and their contributions to the standard free energies reaction anion -GF/J equiv-l % contribution to AG* ~~ Ca s Na NO; 603 Ca Na c10, 555 Cd e Na NO; 593 Cd + Na c1- 2003 Cd e Na Cl-/NO; 1366 41 36 14 47 30.5 I . . ' . I " " 0.0 0.5 - ECd 0 Fig. 3. Values of the solution-phase correction for the three cadmium systems. The NaCl/CdCl, mixing correction is seen to be very extensive in magnitude, whilst the NaNO,/CdNO, mixing correction is only small. (a) Cl-, uncorrected, (b) C1-, ( c ) NO;/CI-, (6) NO; and (e) NO;, uncorrected. the three cadmium isotherms both with and without the corrections evaluated using eqn (16). For cadmium nitrate, the mixing correction is small.This is not so for the chloride solutions where, in some cases, the uncorrected r is more than twice as large as the corrected value. The mixed chloride/nitrate system is of interest, since the application of eqn (16) leads to a unique value for r. If the correction for the mixed solution is omitted in this case, one is forced to make a choice between the use of r calculated from chloride data or r from nitrate data, and this choice cannot be made because of the absence of any determining criteria. In either case, failing to correct for the mixed electrolyte solution would lead to a serious error in AG.1744 ION EXCHANGE IN ZEOLITES CONCLUSIONS It appears that both calcium and cadmium exchange reversibily in Na-X. However, the acidic reaction between Cd2+ and water suggests that cadmium is on the borderline of stability to an extent that concentrated solutions may be found to hydrolyse this zeolite.In contrast to clays and soils, the zeolitic framework charge appears normally to prevent the penetration of associated ion pairs into the intracrystalline voids. In a recent publication8 an iterative procedure was outlined which demonstrated the accurate prediction of ion-exchange behaviour from a limited amount of thermodynamic data such as those given in tables 2-6 and I 1. This present paper emphasises the critical importance of evaluating correctly the solution-phase non-ideality term r; this is an essential prerequisite to accurate prediction of exchange data. We thank the S.E.R.C. for a Research Grant (to P.F.).P. P. Lai and L. V. C. Rees, J . Chem. Soc., Faraday Trans. I , 1976,72, 2650. P. Fletcher and R. P. Townsend, J . Chromatogr., 1982, 238, 59. P. Fletcher and R. P. Townsend, in Proc. 5th Int. Conf. Zeolites, Naples, 1980, ed. L. V. C. Rees (Heyden & Sons, London, 1980), p. 3 1 1. G. L. Gaines and H. C. Thomas, J . Chem. Phys., 1953, 21, 714. R. M. Barrer and A. J. Walker, Trans. Faraday Soc., 1964, 60, 171. R. M. Barrer and J. Klinowski, J. Chem. Soc., Faraday Trans. I , 1974, 70, 2080. ’ S. A. I. Barri and L. V. C. Rees, J . Chromatogr., 1980, 201, 21. * R. P. Townsend, P. Fletcher and M. Loizidou, Proc. 6th Int. Con$ Zeolites, Reno, 1983, ed. D. Olson and A. Bisio (Butterworths, Sevenoaks, 1984), p. 110. E. A. Guggenheim, Thermodynamics (North-Holland, Amsterdam, 1950), pp. 175 and 200. lo G. Sposito, The Thermodynamics of Soil Solutions (Oxford University Press, New York, 1981), p. 60. l 1 E. A. Guggenheim, Thermodynamics (North-Holland, Amsterdam, 1950), p. 302. l 2 P. Fletcher and R. P. Townsend, J . Chem. Soc., Faraday Trans. 2, 1981, 77, 2077. l 3 E. A. Guggenheim, Philos. Mag., 1932, 19, 588. l4 G. Scatchard, Chem. Retl., 1936, 19, 309. l5 K. S. Pitzer, J . Phys. Chem., 1973, 77, 268. l6 E. A. Guggenheim, Thermodynamics (North-Holland, Amsterdam, 1950), p. 208. l7 E. Fowler, Statistical Mechanics, (Cambridge University Press, Cambridge, I929), p. 186. l H Bronsted, J . Am. Chem. Soc., 1922, 45, 877. l9 H. A. Harned and R. A. Robinson, Multicomponent Electrolyte Solutions (Pergamon Press, Oxford, 2o H. S. Sherry, J . Phys. Chem., 1968, 72, 4080. 21 F. Wolf, D. Ceacardnu and K. Pilchowski, Z . Phys. Chem. (Leipzig), 1973, 252, 50. 22 R. M. Barrer, L. V. C. Rees and M. Shamsuzzoha, J . Inorg. Nucl. Chem., 1966, 28, 629. 23 J. J. Gal and P. Radovanov, J . Chem. Soc., Faraday Trans. I , 1975, 71, 1671. 24 H. S. Sherry, J . Phys. Chem., 1968, 72, 4087. 25 R. M. Smith and A. E. Martell, in Inorganic Complexes (Plenum Press, New York, 1976), vol. 4. 26 H. Helgeson, Am. J . Sci., 1969, 267, 729. 27 S. V. Mattigod and G. Sposito, Soil. Sci. SOC. Am. J . , 1977, 41, 1092. 29 G. Sposito, K. M. Holtzclaw, C. T. Johnston and C. S. Lavesque-Madore, SoilSci. Soc. Am. J . , 1981, 30 G. Sposito, K. M. Holtzclaw, L. Charlet, C. Jouany and A. L. Page, Soil Sci. Soc. Am. J . , 1985, 47, 31 A. Maes, P. Peigneur and A. Cremers, Proc. Int. Clay. Conf., p. 319. 32 K. S. Johnson, Marine Chem., 1981, 10, 195. 33 G. Sposito and S. V. Mattigod, GEOCHEM, A Computer Program for the Calculation of Chemical 34 P. Fletcher and R. P. Townsend, J. Chromatogr., 1980, 201, 93. 1968). J. F. O’Connor and R. P. Townsend, Zeolites, in press. 45, 1079. 51. Equilibria in Soil Solutions (U. C. Riverside, 1980). (PAPER 4/ 1776)

ISSN:0300-9599    

DOI:10.1039/F19858101731

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. 1, 1985,81, 1745-1754 Interaction Parameters and the Equilibrium Anionic Polymerization of a-Methylstyrene in Tetrahydrofuran BY VAN TAM BUI* Departement des Sciences Fondamentales, Universite du Quebec a Chicoutimi, Chicoutimi, Quebec, Canada G7H 2B1 AND JACQUES LEONARD Departement de Chimie, Universite Laval, Quebec, Quebec, Canada GI K 7P4 Received 18th October. 1984 Flory-Huggins interaction parameters, x , ~ , have been determined for a-methylstyrene + tetrahydrofuran and tetrahydrofuran + poly(a-methylstyrene) mixtures from static vapour- pressure measurements. For the first system, experiments were performed at five temperatures varying from 20.0 to 40.0 "C and yield values of x between -0.30 and -0.40. Experimental data for the second system were obtained at 25.0 and 35.0 "C, leading tox values slightly greater than 0.40.These results are applied to the data obtained for the equilibrium anionic polymerization of a-methylstyrene in tetrahydrofuran and used for the computation of the free energy of polymerization, AG,,. Values of AGlc obtained in this fashion are found to be in good agreement with previously obtained results. It has been shown1 that in the case of anionic polymerization of a-methylstyrene (aMS) in tetrahydrofuran (THF) at a given temperature, the equilibrium monomer concentration is not unique but varies with the concentration of polymer present under equilibrium conditions. The effect of solvent on the equilibrium may be verified by comparing results obtained for the polymerization of the same monomer in two solvents whose characteristics are different.l. * Taking into consideration the effect of solvent, Ivin and Leonard113 derived an expression for the free energy of polymerization.From the Flory-Huggins expression4 a for the partial molar free energy of monomer and polymer in a ternary system, they obtained AGlc = RT(lndm+ 1+Irms-Xsp(Vrn/V,)Ids+Xmp(dp-dm)-("dp+')/n~ (1) where AGlc is the free energy change upon conversion of 1 mole of pure liquid monomer to 1 base mole of pure liquid amorphous polymer, R is the gas constant, T is the absolute temperature, Vm and V, are the molar volumes of monomer and solvent, respectively, n is the average number of segments in a chain, 4 is the volume fraction of any component in the system under equilibrium conditions, x is the free- energy interaction parameter between any two components and the subscripts m, s and p refer to monomer, solvent and polymer, respectively.If n is large and &,, the volume fraction of polymer, is not too small, eqn (1) becomes AGIc = RT{ln d m + 1 +kms-Xsp(Vrn/ &)I ds+Xmp (dp-dm)}. (2) Upon examination of eqn (1) and (2) it is found that the effect of solvent on the equilibrium of the polymerization system can be summed up by the term 17451746 INTERACTION PARAMETERS FOR a-METHYLSTYRENE IN TH F The effect of monomer-polymer interaction on the equilibrium, represented by the term xmp (4, - &), is much smaller than the previous one because of the usual small values of (4p-4m). In order to determine the exact magnitude of the solvent effect on the equilibrium polymerization of a-methylstyrene in tetrahydrofuran, the x parameters for the a-methylstyrene + THF and poly(a-methylstyrene) + THF systems have been deter- mined and their values used for the computation of AG,,.The values of the x para- meters were obtained using static vapour-pressure measurements for various compositions and temperatures. EXPERIMENTAL MATERIALS Tetrahydrofuran and a-methylstyrene were the best available reagents. THF was kept over CaH, and degassed over a period of several days on the vacuum manifold. It was then distilled onto a first sodium mirror and kept over a second one. aMS was kept over CaH, and degassed on the vacuum line for several days. It was then distilled under vacuum, the head and the tail fractions being discarded.Regarding the poly(a-methylstyrene) (PaMS) samples used, four samples of different average molecular weights (1.7 x lo5, 4.5 x lo4, 2.65 x lo4 and 5300) had previously been prepared by anionic polymerization of a M S in THF. They were characterized by gel-permeation chromato- graphy, which showed a polydispersity index of < 1.10 for all polymer samples except for the 1.7 x lo5 sample, whose index was 1.7. APPARATUS The vapour-pressure assembly used for the measurement of static total vapour pressure of mixture was similar to that described by Singh and Ben~on.~ It was made of an MKS Baratron pressure gauge, a vapour-pressure cell and a vacuum manifold. The pressure gauge consisted of an MKS Baratron TM type 77M electronic pressure meter and a type 77H series pressure head.The pressure head was fitted with two ports of entry; differential pressures of up to 300 mmHg (40.0 kPa) could be measured and variations of 0.003 mmHg could be detected. The vapour-pressure cell was connected to the first entry port and to the vacuum manifold through two bellows stopcocks. The vacuum manifold was also connected to the second entry port of the pressure head through a three-way stopcock which could be opened to the atmosphere. The volume of the cell (including connecting tubes) was 65 f 3 cm3. The cell was immersed in a constant-temperature bath. The maximum temperature fluctuation was f 0.01 "C at the extreme temperature of 40 "C. Because of that fluctuation, the average pressure readings for the most volatile component (THF) was estimated at f0.03 mmHg (f0.004 kPa) at 40.0 "C and & 0.0 1 mmHg at 20.0 "C.The tubes emerging from the bath and the pressure head were kept at a temperature which was ca. 5 "C above the bath temperature in order to prevent any condensation. The flasks and ampoules containing the pure components were attached to the vacuum manifold. MEASUREMENT PROCEDURE The vacuum manifold was constantly heated at ca. 75 "C and the whole vapour-pressure assembly was thoroughly degassed prior to any series of measurements. Two ampoules were attached to the vacuum manifold and degassed. With the stopcocks connecting the vacuum manifold to the pressure head and the cell closed, ca. 25 g of each pure component was distilled into each ampoule, which was equipped with a vacuum stopcock. Then both ampoules were removed from the vacuum manifold, weighed and connected again to the vacuum manifold.The components were successively distilled into the cell. The amount of the components distilled into the cell was determined by weighing the ampoules before and after the distillation. In order to ensure that no distillate remained in the vacuum manifold, the cell was kept in liquid nitrogen for several minutes before closing the stopcock connecting the vacuum manifold with the cell. Then the mixture was allowed to reach the lowest desiredV. T. BUI AND J. LEONARD 1747 Table 1. Vapour pressure of pure tetrahydrofuran at various temperatures T/OC P"/kPa Pa/kPa 10 10.739 10.746 15 13.665 13.688 20 17.258 17.292 25 21.578 21.621 30 26.724 26.81 1 35 32.877 32.970 a From ref.(6). temperature (20 "C). The equilibrium was attained rapidly with the help of a stirring magnetic bar, coupled with an external rotor driven by compressed air. Subsequently the temperature of the mixture was raised by increments of 5 "C to higher temperatures. All manipulations were performed according to high-vacuum techniques. With regard to the experiments carried out with THF+PaMS mixtures, a second cell equipped with a special entry port was used. Ca. 2.0 g of previously dessicated PaMS was inserted into the cell through that port of entry; the cell was then kept in liquid nitrogen and degassed slowly via the bellows stopcock. This manipulation was carried out carefully in order to prevent any loss of polymer powder.Finally, a desired amount of THF was distilled into the cell and allowed to reach 25.0 "C. It was observed that the PaMS had been well solubilized in THF and the resulting solutions could be stirred easily. RESULTS TETRAHYDROFURAN + Q-METHYLSTYRENE SYSTEM In order to check the reliability of our results, P", the vapour pressure of pure THF in the temperature range 10.0-35.0 O C , was measured and compared with published data.6 As can be seen from table 1, our values are slightly lower than the reported values. The observed difference may arise from various causes, for example the use of either a dynamic or static method. However, the agreement is satisfactory as differential pressures rather than absolute pressures are required for the computation of intermolecular interaction parameters.Vapour pressures P of binary mixtures of aMS and THF for various mole fractions Xm are listed in table 2 for five different temperatures from 20.0 to 40.0 "C. The values of x, have been corrected for the loss of material occurring through evaporation from the cell. The amount of component i, AK, lost by the liquid phase through (4) evaporation is AK = Mi PYX: V/RT where Mi is the molecular weight, Pip is the vapour pressure of the pure component, V is the volume of the gas phase (ca. 55+3 cm3) and x i is the mole fraction as calculated from the amount (in g) distilled into the cell. Eqn (4) is used assuming that the vapour is made of two ideal gases and the activity coefficient of each component in the liquid phase is unity.The errors introduced by these two approximations to the computation of x, are extremely small. Table 2 and fig. 1 show the 'excess vapour pressure'PE for various mole fractions x, and various temperatures, PE being defined here as P E = P-P",,-P& where subscripts m and s refer to aMS and THF, respectively. The curves shown in1748 INTERACTION PARAMETERS FOR Q-METHYLSTYRENE IN TH F Table 2. Vapour pressure and ‘excess vapour pressure’ of mixtures of a-methylstyrene and tetrahydrofuran at various temperatures and mole fractions T / “C xrn P/kPa P”/kPa Xm P/kPa PE/kPa 20 25 30 35 40 0 0.1637 0.2239 0.3046 0.3737 0.3985 0.4733 0.4948 0 0.1640 0.2243 0.3050 0.3741 0.3991 0.4742 0.4954 0 0.1643 0.2247 0.3054 0.3745 0.4000 0.4753 0.4962 0 0.1646 0.2252 0.3058 0.3751 0.4006 0.4764 0 0.1653 0.3064 0.3757 0.40 1 5 0.4777 17.258 14.319 13.205 11.719 10.426 9.946 8.579 8.186 21.578 17.878 16.518 14.639 13.032 12.459 10.739 10.266 26.724 22.104 20.41 1 18.1 12 16.152 15.438 13.319 12.745 32.877 27.131 24.984 22.258 19.885 19.005 16.378 40.1 16 33.063 27.024 24.27 1 23.178 19.945 - -0.156 -0.245 -0.359 - 0.477 - 0.535 - 0.629 - 0.657 - -0.215 - 0.293 - 0.459 -0.597 -0.640 -0.764 - 0.787 - - 0.304 -0.41 1 -0.591 - 0.735 -0.779 - 0.920 - 0.944 - -0.435 - 0.627 - 0.752 - 0.889 - 0.947 - 1.127 - -0.552 - 1.044 - 1.072 -1.151 - 1.387 0.6098 0.6545 0.7 104 0.771 7 0.8092 0.9 182 1 .ooo 0.6108 0.6553 0.7106 0.7723 0.8097 0.9 186 1 .ooo 0.61 13 0.6564 0.71 1 1 0.773 1 0.8 104 0.9193 1 .ooo 0.4800 0.6 124 0.71 15 0.7740 0.81 12 0.9 197 1 .ooo 0.6 136 0.71 19 0.7752 0.8 120 0.9200 1 .ooo - - - -- 6.166 5.426 4.486 3.533 2.960 1.380 0.251 7.739 6.819 5.673 4.446 3.740 1.746 0.331 9.619 8.479 7.086 5.559 4.666 2.200 0.456 16.241 11.839 8.772 6.873 5.786 2.750 0.61 1 14.465 10.746 8.406 7.113 3.406 0.793 - - - -0.721 -0.701 - 0.689 - 0.600 - 0.536 - 0.263 - __ -0.861 -0.833 - 0.807 - 0.723 -0.635 -0.313 - 1 - I .048 - 1.003 - 0.959 -0.857 -0.771 -0.376 - - - 1.148 - 1.279 - 1.147 - 1.031 -0.916 - 0.45 I - - 1.523 - 1.377 - I .227 - 1.073 -0.533 - fig.1 have been calculated by the least-squares method according to the Redlich-Kister equation’ with two parameters: PE = x m X, [a, + a, (xm - x,)]. Combining eqn ( 5 ) and (6), the experimental vapour pressure, P, is given byV.T. BUI AND J . LEONARD 1749 Fig. 1. Excess vapour pressure PE plotted against the mole fraction x, for a- methylstyrene+ tetrahydrofuran mixtures, at various temperatures. 0, 20; @, 25; 0, 30; m, 35 and A, 40 "C. The interaction parameter zms is defined as4 where q& is the volume fraction of THF in the binary mixture and AGE is the excess free energy of mixing, which is given by where y is the activity coefficient in the liquid phase. The activity coefficients are calculated with the help of an iterative method described elsewhere*~Q using the following set of equations: lny, = - l n ( x m + ~ m s x ~ ) + x s Ams - ) (10) In ys = - In (x, + As, x,) - x, x m + Ams x s x s + Asm x m1750 INTERACTION PARAMETERS FOR a-METHYLSTYRENE IN THF Table 3.zms parameter for a-methylstyrene + tetrahydrofuran mixtures at various temperatures T/"C 4 s 20 25 30 35 40 0.849 0.715 0.593 0.484 0.385 0.294 0.212 0.135 - 0.305 -0.320 -0.330 -0.340 -0.355 -0.365 - 0.375 - 0.390 -0.295 -0.305 -0.315 -0.320 - 0.330 -0.340 -0.350 -0.360 -0.310 -0.315 -0.320 - 0.330 -0.335 - 0.340 - 0.345 -0.350 -0.290 - 0.295 - 0.300 - 0.300 - 0.300 - 0.305 - 0.305 -0.310 - 0.280 -0.285 -0.285 -0.290 - 0.295 -0.295 - 0.300 - 0.300 0 0.2 0.4 0.6 0.8 1.0 9s Fig. 2. Variation of zms with the volume fraction & at various temperatures for a- methylstyrene+ tetrahydrofuran mixtures: 0, 20; @, 25; m, 35 and A, 40 "C. where Am, and Asm are adjustable semi-empirical parameters of the Wilson equationlo and am and a, are correction factors for the non-ideality of the gas phase.l19l2 Table 3 and fig.2 show values of xms at five temperatures computed using eqn (7H 12). THE POLYMER SYSTEMS The solvent-polymer interaction parameter, xsp, was obtained directly from experimental data of vapour pressure P of poly(a-methylstyrene) (p) + tetrahydrofuran (s) mixtures using4 lna, = l n ( l - d ~ ) + ( l - l / n ) d p + x s ~ # ~ (13) where dP is the volume fraction of polymer in this mixture, n is the average number of segments in a chain and a, is the activity of solvent, calculated using13 P a, = exp [B(P-E)/RT] (14) zV. T. BUI AND J. LEONARD 1751 0.5 2 X 04- 0 3; Table 4. xsp parameter for tetrahydrofuran + poly(a-methylstyrene) mixtures at 25.0 "C - A> / 0 0 ---x== 0 0 I 1 I 1 sample (M,) P/kPa 1.7 x 105 0.257 0.234 0.140 4.5 x 104 0.337 0.21 1 2.65 x 104 0.253 0.202 0.174 0.142 5.3 x 103 0.368 0.195 0.167 21.380 21.36 21.53 21.07 21.37 21.30 21.39 21.45 21.49 20.71 21.31 21.39 0.46 0.46 0.42 0.45 0.42 0.42 0.4 1 0.4 1 0.40 0.43 0.38 0.37 Fig.3. Variation of xsp with the polymer volume fraction 4p for poly(a-methylstyrene) + tetrahydrofuran mixtures at 25°C for polymers with average molecular weight of: 0, 1.7 x lo5; A, 4.5 x lo4; A, 2.65 x lo4 and IJ, 5300. where B, the second virial coefficient, may be estimated from the Berthelot equation of state:14 where Pc and are the critical pressure and temperature, respectively, of the pure solvent and T is the working temperature. Experimental data of P at 25.0 "C and calculated values of xsp are given in table 4 and fig.3. Measurements carried out at 35.0 "C have revealed a slight decrease of xsp with increasing temperature. As for the monomer-polymer interaction parameter, xmp, measurements could not be made for this mixture because the vapour pressure of aMS was too low at working temperatures. Moreover, the head-pressure gauge did not allow measurements to be taken above 45.0 "C because of its maximum heating rate. Since the contribution of zmp to eqn ( I ) and (2) is small, the reported value of 0.4 for xmp at 70.0 "CI5 has been used.1752 INTERACTION PARAMETERS FOR a-METHYLSTYRENE IN TH F Table 5. and AG,,/RT for the equilibrium anionic polymerization of a-methylstyrene in tetrahydrofuran with xmp = 0.40 - 15 -0.38 0.46 - 1.36 - 10 -0.37 0.46 -1.15 -5 -0.36 0.45 -1.17 0 -0.35 0.45 - 1.01 5 -0.34 0.44 -0.81 10 -0.34 0.44 -0.92 15 -0.33 0.43 -0.80 20 -0.32 0.43 -0.79 AS,, = 137.8 J rnol-' K-l, AHlc! = -42.5 kJ mol-l, ~~ ~ - 3.45 - 1.12 - 3.00 - 1.10 - 2.70 - 1.08 - 2.24 - 1.07 - 1.82 - 1.04 - 1.58 - 1.04 - 1.22 - 1.02 - 0.95 - 1.01 AS,," = 121.7 J mol-l K-l AH,," = -38.4 kJ mol-l - 3.24 - 2.96 - 2.63 -2.31 - 2.02 - 1.68 - 1.39 - 1.12 " From ref.( I ) . 3 4 3 5 3 6 3 7 3 0 lo3 KIT Fig. 4. Plot of AG,JRT against 1/T for the polymerization of a-methylstyrene: 0, this work and A, ref. (1). COMPUTATION OF p AND AGlc With the foregoing results for xms and xsp, using an average value of 0.40 for xmp, and with the reported data1 for the equilibrium polymerization of aMS in THF, values of AGlc and p are obtained from eqn (2) and (3) and the results are shown in table 5 .Values of xms and xsp for below 20 "C listed in table 5 are extrapolated from data obtained above 20 "C. When x is found to vary with 4, values 4: and 4; for the binary mixtures are selected. These volume fractions are defined with respect to the volume fractions of the components in the polymerization system as 4; = d s / ( d s + dm) and 4; = 4p/ ( 4 p + 4s). (16) The monomer mole fraction used in eqn (8) is xk, the value corresponding to 4;. From the experimental results for equilibrium polymerization one can verify that at a givenV. T. BUI AND J. LEONARD 1753 polymerization 4; (and also xk) is remarkably constant3 so that only one value of zms is required at that temperature. The computation of /3 and AG,, was performed for temperatures corresponding to the polymerization data of ref.(1) with an average value of 1.60 for Vm/& Values of AG,,/RT and computed through eqn (2) and (3) are listed in table 5 together with results obtained previously.' Fig. 4 shows the variation of AG,,/RT with I/T for both sets of results. AHlc, the enthalpy change, and AS,,, the entropy change, were obtained by means of the least-squares method and their values are reported in table 5. DISCUSSION The results of xms parameter reveal two types of departure from the Flory-Huggins theory, which is based on the lattice model: (i) zms calculated from the excess free energy of mixing, AGE, is much smaller (in absolute values) than that obtained from the enthalpy of mixing16 (xH z - 1 .O) and (ii) xms is not constant and increases slightly with concentration of THF at 40.0 "C, with the change becoming more pronounced at lower temperatures.The former observation denotes a large excess entropic contribution to zms, while the latter observation indicates important contributions from the long-range interaction in addition to short-range interactions. In order to take into account the contributions of long-range interactions, a higher-order interaction parameter of the type zmss might be introduced so that" which allows xms to vary with concentration. However, in the present case it is found that 4: [eqn (16)] used for the computation of zms is virtually constant. Regarding the xsp parameter, in addition to a slight dependence on the molecular weight, it is found to increase with polymer concentration.As has been discussed above, the variation of the xsp parameter with concentration is regarded as the major deficiency of the lattice-model theory.18 The quantitative evaluation of the solvent effect through xms and xsp on the considered equilibrium polymerization is made directly compared with the aggregate approach proposed previously.l9 Taking into account the experimental uncertainties in .xms and xsp, it is found that agreement between both sets of results is fairly satisfactory. CONCLUSIONS The thermodynamic behaviour of an equilibrium polymerization in solution may be adequately described by only three first-order binary parameters of the Flory- Huggins theory under the following conditions : first, monomer-solvent interactions ought to be of the van der Waals type and associated species are not formed, secondly, polymerization mixture should not be too concentrated in polymer, and thirdly, the variation of thex parameters with the composition of the polymerization system ought to be taken into account when using eqn ( I ) and (2).We thank the National Sciences and Engineering Research Council of Canada for financial support and for the award of a fellowship (to V.T. B.).1754 INTERACTION PARAMETERS FOR Q-METHYLSTYRENE IN THF K. J. Ivin and J. LConard, Eur. Polym. J., 1970,6, 331. J. Leonard, Macromolecules, 1969, 2, 661. P. J. Flory, Principles of Polymer Chemistry (Cornell University Press, Ithaca, 1953), (a) chap. XI11 and (b) chap. XII. J. Singh and G. C. Benson, Can. J. Chem., 1968,46, 1249. T. Boublik, V. Fried and E. Hala, The Vupour Pressure of Pure Substances(Elsevier, Amsterdam, 1973). 0. Redlich, A. T. Kister and C. E. Turnquist, Chem. Eng. Progr. Symp. Ser., 1952, 2, 48. J. A. Barker, Aust. J. Chem., 1953, 6, 207. B. T. Trinh, R. S. Ramalho and S. Kaliaguine, Can. J . Chem. Eng., 1972, 50, 771. H. W. Prengle Jr and M. A. Pike Jr, J. Chem. Eng. Data, 1961, 6, 24. 2 J. LConard and S. L. Malhotra, J. Pofym. Sci., Part A , 1971, 9, 1983. lo G. M. Wilson, J. Am. Chem. SOC., 1964, 86, 127. l2 0. Redlich and J. N. S. Kwong, Chem. Rev., 1949,44, 233. l3 C. Booth and C. J. Devoy, Polymer, 1971, 12, 309. l4 J. 0. Hirschfelder, C. F. Curtis and R. B. Bird, Molecular Theory of Gases and Liquids (Wiley, New l5 S. G. Canagaratna, D. Margerison and J. P. Newport, Trans. Faraday SOC., 1966, 62, 3058. l6 B. J. Cottam, J. M. G. Cowle and S. Bywater, Makromof. Chem., 1965,86, 116. l7 G. Scatchard, Annu. Rev. Phys. Chem., 1952, 3, 259. l8 B. E. Eichinger and P. J. Flory, Trans. Faraday SOC., 1968, 64, 2035. York, 1964). (PAPER 4/ 1782)

ISSN:0300-9599    

DOI:10.1039/F19858101745

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faruduy Trans. 1, 1985,81, 1755-1759 Formation of the rn-Dinitrobenzene Radical-anion Dimer in the Triplet Ground State on Magnesium Oxide BY TOKIO IIZUKA Department of Chemistry, Faculty of Science, Hokkaido University, Sapporo 060, Japan Received 29th October, 1984 The adsorbed state of rn-dinitrobenzene on the surface of magnesium oxide has been studied by means of visible absorption and electron spin resonance spectroscopies. In the visible spectrum the adsorbed dinitrobenzene had two absorption maxima at 560 and 690nm, attributed to the nitro radical anion and radical dimer, respectively. The adsorbed species showed an e.s.r. signal of Ams = & 2 transitions characteristic of a triplet-state two-spin system in the g = 4 region, in addition to a monomer signal at g = 2.The number of radical dimers formed was sensitive to the conditions of preheating the oxide in air. The formation of negative radical ions upon adsorption of various acceptor molecules such as aromatic nitro c~mpounds,~-~ tetra~yanoethylene,~ phena~ine,~ pyridine,61 dimethylanthracenea etc. on the surface of alkaline-earth-metal oxides is a well known phenomenon and the mechanism of electron transfer involved has attracted much attention. Che et al. proposed that the donor centres were surface 02- ions in low-coordination sites.4 On the other hand, Garrone et al. have recently proposed a new mechanism for the formation of radical anions which does not require electron transfer from the solid.g The present work reports an observation of the triplet state of the rn-dinitrobenzene radical-anion dimer adsorbed on MgO by means of visible absorption and electron spin resonance spec t r oscopies.EXPERIMENTAL The magnesium oxide used in this investigation was prepared from Mg(OH), (Kanto Chemical Co., guaranteed reagent) by evacuation at various temperatures for 1-2 h. rn- Dinitrobenzene (Wako Pure Chemicals, guaranteed reagent) was adsorbed on MgO from benzene solution through a breakable seal. A Jeol X-band e.s.r. spectrometer operating with 100 kHz field modulation was employed to obtain the e.s.r. spectra. The visible absorption spectra were recorded at room temperature with a Shimadzu MPS-SOL spectrometer by using an in situ quartz cell in which a pressed wafer of Mg(OH), powder was fixed. RESULTS AND DISCUSSION Visible absorption spectra of rn-dinitrobenzene adsorbed on MgO pretreated at various temperatures are shown in fig.1. The observed spectrum has two absorption maxima at 560 and 690 nm. The peak at 560 nm corresponds to the adsorption band of a nitroaromatic radical anion.l0 When dinitrobenzene was adsorbed on MgO which I7551756 TRIPLET DIMER OF DINITROBENZENE RADICAL ANION ON MgO 400 500 600 700 800 wavelength/ n m Fig. 1. Visible absorption spectra of rn-dinitrobenzene adsorbed on MgO prepared by the decomposition of Mg(OH), in vacuum at 900 "C and further sintered in air at 900 "C for various periods: (a) background, (b) 0, (c) 0.25, ( d ) 0.5, (e) 0.75, (f) 1 and (g) 3 h. had been evacuated at 500-900 "C for 1 h the 560 nm band appeared as a shoulder of the 690 nm band.If the evacuated MgO was further heated in air at 900 "C and dinitrobenzene solution was adsorbed after cooling to room temperature in a stream of dry nitrogen, the 690 nm band decreased and the 560 nm band reappeared distinctly [fig. 1 ( c ) - ( f ) ] . In the case of MgO heated in air at 900 "C for 3 h or more the 690 nm band disappeared completely and only the 560 nm band was observed, as shown in On MgO evacuated at 900 "C for 1 h the same measurements were made in solutions where the concentration of dinitrobenzene ranged from 2.8 x lop4 to 5 x mol dm-3. These spectra are shown in fig. 2. At low concentrations only the 560 nm band was observed, as seen in fig. 2(b). However, the 690 nm band emerged clearly as the concentration of dinitrobenzene was increased. This indicates that the 690 nm band may be ascribed to the radical anion dimer of dinitrobenzene adsorbed on the surface.On the other hand, m-dinitrobenzene adsorbed on MgO gives an intense e.s.r. signal at g = 2.003 which has been attributed to the nitro radical ani0n.l In addition to this signal, a sharp signal at half-field strength of the g = 2 region was observed as shown in fig. 3. This signal in the g = 4 region was observed in correspondence with the 690 nm visible absorption band. Thus the signal at g z 4 can be ascribed to Ams = 2 transitions characteristic of a two-spin triplet electronic state. From these results it is apparent that the pair of nearest-neighbouring radical anions adsorbed on the surface of MgO tends to dimerize to form a biradical in the triplet ground state.In glassy-matrix fig. 1 (g).T. IIZUKA 1757 400 500 600 700 800 wavelength/ nm Fig. 2. Change in the visible adsorption spectra on MgO evacuated at 500 "C for 2 h with the concentration of rn-dinitrobenzene in benzene: (a) background, (b) 2.8 x ( d ) 8.3 x (e) 1.67 x and (f) 5.0 x mol dm-3. (c) 5.6 x I 1 160 200 240 280 i 320 HImT Fig. 3. E.s.r. spectra of m-dinitrobenzene adsorbed on MgO. (a) Monomer radical-anion signal at g = 2, (6) liquid-nitrogen-temperature spectrum on MgO prepared by the decomposition of Mg(OH), at 900 "C in a vacuum, ( c ) room-temperature spectrum of the same sample and ( d ) room-temperature spectrum on MgO after calcination at 900 "C in air for 3 h.1758 TRIPLET DIMER OF DINITROBENZENE RADICAL ANION ON MgO solutions some organic-radical ion pairs show the occurrence of spin-spin interactions at low temperatures (ca.77 K).ll The e.s.r. spectrum obtained in this work on the surface of MgO was almost the same as that of the nitrobenzene radical-anion dimer in glassy solution, except for the signal at Am, = f 1." The band of the Am, = f 1 transition was obscure on the surface, probably owing to the large anisotropy and to overlapping with the monomer signal. For the formation of the biradical, the radical pairs might exist at a distance of ca. 5 A or less.ll Since a large number of dimers are formed on MgO, as seen in fig. 1 and 2, the sites for the formation of the radical ion must aggregate on the surface. The simplest way to visualize the formation of radical anions on the oxide surface is to consider a direct electron-transfer process from a surface site towards a given acceptor molecule.Che et aL4 have suggested surface 02- ions in low coordination as sites for such a process. On the other hand, Garrone et al. have proposed a different mechanism to explain the formation of radical anions on the surface of alkaline-earth-metal oxide~.~ According to these authors the radical anions can be formed without electron transfer from the solid. In fact, an XH molecule can be heterolytically chemisorbed on a surface (M2+02-) site leading to the species M2+X- and OH;. Subsequently the carbanion can transfer an electron to a second XH molecule, leading to the radical anion XH'-. This mechanism might explain satisfactorily the formation of the dipyridyl radical anion from pyridine on MgO, CaO and SrO, a process which needs both the abstraction of hydrogen from the pyridine molecule and an electron-transfer If this mechanism is applied to the case of dinitrobenzene, the reaction is as follows: DNBH+M2+02- --* M2+DNB-+OH; M2+DNB-+ DNBH 4 M2+DNBH*-+DNB*. In this case the surface reaction is complex and the formation of either a biradical or a radical dimer will be possible.However, this possibility may be ruled out for the case of dinitrobenzene for the following reasons. The amount of dimer formed gradually decreased and the monomer radical became dominant in the calcination of MgO in air. This is due to the decomposition of 02- in low-coordination sites by the sintering of the MgO surface.From this it is clear that aggregated surface sites are necessary for the formation of dinitrobenzene radical-anion dimers. This suggests that the radical anion formed shows a one-by-one correspondence with the electron-donor site on the surface. Thus direct electron transfer to dinitrobenzene from the surface would be plausible in this case. Direct electron transfer probably occurs on molecules which possess an extremely high electron affinity, such as nitroaromatic compounds or tetracyanoethylene. When the dipyridyl radical anion was exposed to oxygen gas the radical anion disappeared instantly and the superoxide anion (0;) was formed.6y Electron transfer from the adsorbed anion to the oxygen molecule has been observed for the adsorption of benzene, ethylene, ammonia, toluene, hydrogen, propene, butene and acetylene on Mg0.12 However, methanol and water did not show electron transfer in the adsorbed state, although they were adsorbed on MgO dissociatively.12 Stone and Garrone concluded that the anions of those molecules are too stable to donate an electron to molecular oxygen.The adsorbed radical anion of dinitrobenzene is stable to exposure to oxygen gas, in contrast to the case of the dipyridyl anion radical. This difference might also be due to the strong stabilizing effect of the nitro group on the anion in the case of dinitrobenzene.T. IIZUKA 1759 A. H. Tench and R. L. Nelson, Trans. Faraday SOC., 1967, 63, 2254. K. J. Klabunde, R. A. Kaba and R. Moms, Inorg. Chem., 1978, 17,2684. M. Che, C. Naccache and B. Imerik, J. Cat& 1972, 24, 328. K. S. Seshadri and L. Petrakis, J. Phys. Chem., 1970, 74, 1317. T. Iizuka and K. Tanabe, Bull. Chem. SOC. Jpn, 1975,48, 2527. M. Che, S. Coluccia and A. Zecchina, J. Chem. SOC., Faraday Trans. 1, 1978, 74, 1324. V. Indovina and D. Coldischi, J. Chem. SOC., Faraday Trans. I , 1982, 78, 1705. E. Garrone, A. Zecchina and F. S. Stone, J. Catal., 1980, 62, 396. lo A. Ishitani, K. Kuwata, H. Tsubomura and S. Nagakura, Bull. Chem. SOC. Jpn, 1963, 36, 1357. l 1 C. A. McDowell and F. Nakano, J. Phys. Chem., 1971, 75, 1205. l 2 E. Garrone and F. S. Stone, Proc. 8th Int. Congr. Catal., 1984, vol. 111, p. 441. * T. Iizuka, H. Hattori, Y. Ohno, J. Souma and K. Tanabe, J. Catal., 1971,22, 130. (PAPER 4/ 1844)

ISSN:0300-9599    

DOI:10.1039/F19858101755

出版商:RSC    

年代:1985

数据来源: RSC