摘要:

J. Chem. SOC., Faraday Trans. I, 1985, 81, 939-949 Chemisorption and Catalysis on LaMO, Oxides BY GOJKO KREMENIC, JOSE M. L. NIETO, JUAN M. D. TASC~N AND LUIS G. TEJUCA* Instituto de Catalisis y Petroleoquimica, C.S.I.C., Serrano 119, Madrid 6, Spain Received 30th May, 1984 The adsorption of oxygen and isobutene and the catalytic activity for propene and isobutene oxidation have been studied on a series of LaMO, (M = Cr, Mn, Fe, Co and Ni) perovskite oxides. Coadsorption results point to the non-competitive adsorption of oxygen and isobutene; i.e. these molecules adsorb on different centres. Oxygen adsorption underwent a remarkable increase after isobutene had been preadsorbed on these oxides (enhanced adsorption). Activation energies for complete oxidation ranged between 16 kcal mol-l (LaMnO,, LaCoO, and LaNiO,) and 31 kcal mol-1 (LaFeO,).LaCrO, showed some activity for methacrolein formation. Adsorption and catalytic-activity profiles showed maxima for LaMnO, and LaCoO,. These results are discussed within the framework of the ideas of Dowden and Wells on the local symmetry of surface cations and its influence on chemisorption and catalysis and show the importance of localized interactions in the processes studied. Some LaMO, oxides (where M is a first-row transition metal) have a high catalytic activity for the total oxidation of CO and hydrocarbons. In addition, their activities are not significantly affected by poisons such as Pb and S present in automotive exhaust gases, thus making these materials promising as anticontamination catalysts.On the other hand, perovskite-type oxides ABO, having both bismuth atoms and cation vacancies in the A position have been shown to be catalysts for the partial oxidation of hydrocarbons.2 In a previous study3 two maxima in the catalytic activity for CO oxidation were found for LaMnO, and LaCoO, in the series of oxides LaMO, (M = V, Cr, Mn, Fe, Co and Ni). These results were discussed in terms of the ideas of Dowden and Wells on the local symmetry of surface cations and their influence on chemisorption and catalysis.* The purpose of the present work was to see if similar considerations could be applied to the chemisorption of oxygen and also to the total oxidation of propene and isobutene on these oxides. In the perovskite structure the M3+ ion is in an octahedral environment of oxygen ions which remains the same for each oxide.Differences in the surface structure of these compounds are expected to be less pronounced than, for example, in a series of simple oxides MO, where the lattice structure may change; thus changes in catalytic activity or adsorption due to changes in geometry should be smaller in the former case. In the reaction products, CO and CO, as well as some oxygenated organic compounds were analysed to determine whether the less active oxides for total oxidation exhibit activity for the partial oxidation of these hydrocarbons. EXPERIMENTAL SAMPLE PREPARATION AND GASES LaMO, oxides (M = Cr, Mn, Fe, Co and Ni; LaVO, is too unstable for catalytic studies and Ti3+ is unstable in solution as it is oxidized to Ti4+, which precipitates as TiO,) were 939940 CHEMISORPTION AND CATALYSIS ON LaMO, OXIDES Table.1. LaMO, oxides final heating oxide temperature/K S,, ET/m2 g-la LaCrO, I023 5.7 LaMnO, 973 12.1 LaCoO, 973 8.5 LaNiO, 1023 4.7 LaFeO, 923 6.4 a Cross-sectional area of N, molecule, 0.162 nm2. prepared by the decomposition of amorphous precursors. Metal nitrates and citric acid (Merck, reagent grade) were dissolved in water in the desired proportions. The water was evaporated in a rotary evaporator at 343 K and 10 mmHg (1 mmHg z 133.3 N mP2) until the precipitate acquired the consistency of a viscous syrup. The precursors so obtained were kept in a vacuum stove at 373 K for 5 h and were then heated for 4 h at the appropriate temperature for obtaining a single perovskite phase (table 1).In the X-ray diffraction spectra, peaks of simple oxides were absent. Further details of the method of preparation have been given el~ewhere.~ B.E.T. specific surface areas are given in table 1. Propene and isobutene (HC) (both 99.0% pure), oxygen (99.980,; pure) and helium (99.995% pure) from Sociedad Espaiiola del Oxigeno were used. ADSORPTION EXPERIMENTS AND INFRARED SPECTRA The adsorption experiments were carried out in a volumetric high-vacuum system in which a dynamic vacuum of lop6 mmHg could be maintained. Pressure measurements were made by means of an MKS capacitance manometer with a sensitivity of lop2 mmHg. The samples were first outgassed in a high vacuum for 15 h at 773 K; the temperature was then lowered to 298 K and the isotherm of total adsorption of oxygen or isobutene (on a clean surface) recorded.After pumping in a high vacuum for 1 h at the adsorption temperature a second isotherm (reversible adsorption) was run. Experiments on the adsorption of 0, at 298 K on a surface with preadsorbed isobutene were carried out after pumping off for 1 h the reversibly adsorbed isobutene at the adsorption temperature and running, as indicated above, a double isotherm of 0,. The same procedure was followed for isobutene isotherms. Infrared spectra were obtained with a Perkin-Elmer 225 spectrophotometer. The high-vacuum system and method of sample pretreatment used were the same as in the adsorption experiments. CATALYTIC ACTIVITY Catalytic-activity experiments were carried out in a flow system with a Pyrex glass tubular reactor of 0.9 cm i.d., the catalytic load (0.25 g, sieved between 0.42 and 0.59 mm) being mixed with Sic (of the same grain size) in a volume ratio of LaMO,: Sic = 1 : 10.Experiments were carried out between 500 and 650 K, the reactants, helium and water, being in molar ratios of HC:O,:He:H,O = u:b:c:40 (where u:b = 10:40 or 40:20 and c is balance to atmospheric pressure) and the contact time W/F = 15 and 4 g (catalyst) h mol g (HC)-l (for a: b = 10: 40 and 40: 20, respectively). In all cases the total flow was kept at 2.6 x lo3 cm3 h-l. Analyses of reactants and products were carried out with a Hewlett-Packard 5830 A chromatograph. Columns of zeolite 13X (for 0, and CO) and Porapak Q (for other gases and liquids) were used.Conversion into product i [Xi (?{)I, total conversion [XT ("/,)I and selectivity to product i [Si (%)I were defined as follows: rnol g of i x no. of C atoms of i x. = x 100 a mol g of HC x no. of C atoms of HC x, = cxi; si = X i / X T x 100.G . KREMEN~C, J. M. L. NIETO, J. M. D. TASCON AND L. G. TEJUCA 94 1 0 100 200 P/mmHg Fig. 1. Total (open symbols) and reversible (filled symbols) adsorption of 0, on a clean LaMnO, surface at 298 K (circlesj and on a surface with preadsorbed isobutene (triangles). Isothermal experiments (at 523 K) were also carried out either keeping the isobutene partial pressure (<sobutene) constant (and equal to 0.09 atm) and varying the 0, partial pressure (Po,) (between 0.09 and 0.40 atm) or keeping Po, constant (and equal to 0.18 atm) and varying j’i)is,,bu&ne (between 0.04 and 0.20 atm).H,O (at 0.40 atm) and He (the balance to atmospheric pressure) were added to the reacting mixture. The total conversion was kept below 3%. RESULTS AND DISCUSSION ADSORPTION OF OXYGEN As an illustration, isotherms of total and reversible adsorption of 0, at 298 K on a clean surface of LaMnO, and on a surface with preadsorbed isobutene are given in fig. 1. They are of type I (total adsorption) or type 111 (reversible adsorption) in Brunauer’s classification. The total adsorption underwent a remarkable increase after isobutene had been preadsorbed on the oxide surface. This phenomenon of enhanced adsorption has been observed, for example, in 0, adsorption on simple oxides after the preadsorption of p r ~ p e n e ~ - ~ and is responsible for the negative character of the adsorption constants found in some instances in the analysis of kinetic data when Langmuir-Hinshelwood type equations are applied.I0 The profiles of 0, adsorption (at 150 mmHg) on the series of perovskites studied are given in fig.2. Sharp maxima were observed for LaMnO, and LaCoO, (on a clean surface). Enhanced adsorption is greater for the oxides which adsorb larger amounts of hydrocarbon (LaFeO, and LaCoO,, cf. fig. 2 and 4). This seems to indicate that the increase in 0, adsorption is directly related to the amount of hydrocarbon preadsorbed. This effect causes the maxima for LaMnO, and LaCoO, on a clean surface to become attenuated by isobutene preadsorption. Reversible adsorption represents a small fraction of the total adsorption; its near942 CHEMISORPTION AND CATALYSIS ON LaMO, OXIDES Fig.2. Total (open symbols) and reversible (filled symbols) adsorption profiles of 0, on LaMO, oxides on a clean surface (circles) and on a surface with preadsorbed isobutene (triangles). P = 150 mmHg; T = 298 K. Table 2. Coverages of 0, on LaMO, oxides at 298 K and 150 mmHg" oxide @total %otal @re" LaCrO, 0.10 0.30 0.02 LaMnO, 0.57 0.77 0.02 LaFeO, 0.04 0.47 0.02 LaCoO, 0.52 0.82 0.03 LaNiO, 0.26 0.33 0.03 a OtOtal and t9iotal, coverages for total adsorption on a clean surface and on a surface with preadsorbed isobutene ; Or,,, coverage for reversible adsorption. Cross-sectional area of 0, molecule, 0.141 nm2. constancy for all the oxides and its independence of the state of the surface (clean surface or surface with preadsorbed isobutene) suggests physisorption.Coverages (taking the cross-sectional area of the 0, molecule to be 0.141 nm2) on LaMnO, and LaCoO, (clean surface) amount to about half a monolayer (table 2). Coverages for other oxides are much lower. These values are about equal (LaCrO, and LaFeO,) or higher by a factor of 2 (LaMnO,) to 5 (LaNiO,) than those found from t.p.d. by Seiyama et ~1Z.l~ for 0, adsorption on the simple oxides Cr,O,, MnO,, Fe,O,, Co,O, and NiO. ADSORPTION OF ISOBUTENE In fig. 3 isotherms of the total and reversible adsorption of isobutene on LaMnO, at 298 K are given. Unlike that observed for 0, (fig. l), all of them are of type I. The profiles of isobutene adsorption at 150 mmHg on LaMO, oxides at 298 K are represented in fig.4. Both the total and reversible adsorption on a surface withG. KREMEN~C, J. M. L. NIETO, J. M. D. TASCON AND L. G. TEJUCA 943 0 1 I I I I I 100 20 0 P/mmHg Fig. 3. Total (open symbols) and reversible (filled symbols) adsorption of isobutene on a clean surface of LaMnO, at 298 K (circles) and on a surface with preadsorbed 0, (triangles). ' C++ Mnh i e b cih iij+ Fig. 4. Total (open symbols) and reversible (filled symbols) adsorption profiles of isobutene on LaMO, oxides on a clean surface (circles) and on a surface with preadsorbed 0, (triangles). P = 150 mmHg; T = 298 K. 1 I preadsorbed 0, are equal or lower than the corresponding adsorption on a clean surface. These decreases should be due to steric hindrance to isobutene adsorption caused by preadsorbed 0,, this effect being more marked on the oxides which adsorb larger amounts of hydrocarbon (LaFeO, and LaCoO,).These results indicate that the adsorptions of isobutene and 0, are not competitive, i.e. these molecules adsorb on different centres. The maximum adsorption was observed on LaFeO,, an oxide which exhibited a minimum in its 0, adsorption; at present we have no explanation for this result. The coverages attained are remarkably higher than those measured for 0, (e.g. on LaFeO, and LaCoO, they are above one monolayer).944 CHEMISORPTION AND CATALYSIS ON LaMO, OXIDES 2 h E $ 1 % 0 500 600 TI K 2 h E 8 1 % 0 500 600 TI K Fig. 5. Conversion of propene (a) and isobutene (b) to CO, as a function of temperature. ., LaCoO,; 0, LaMnO,; a, LaNiO,; A, LaCrO,; 0, LaFeO,.Molar ratio HC:O, = 0.25: 1. INFRARED SPECTRA In the infrared spectra obtained after propene and isobutene adsorption on LaCrO,, of low activity, and LaMnO,, of high activity (see below), between room temperature (r.t.) and 673 K (after pumping for 0.5 h at r.t.), a wide band centred at 2320-2340 cm-l corresponding to linearly adsorbed CO, was observed. This band was partially or totally removed by pumping at r.t. when adsorption temperatures were in the range r.L-473 K. However, CO, produced after HC adsorption at 573-673 K was not removed at r.t. The CO, band yielded by the system HC + LaMnO, was more intense than that of HC+ LaCrO, because of the higher catalytic activity for total oxidation of the former oxide. In some cases (e.g.propene and LaMnO, at 673 K) wide bands centred at 1700 and 1520 cm-l of carbonate structures produced by interaction of CO, with the perovskite surface were found. CATALYTIC ACTIVITY FOR TOTAL OXIDATION The conversions of propene ( a ) and isobutene (b) (for a molar ratio HC:O, = 0.25: 1) referred to the main reaction product, CO,, as a function of temperature are plotted in fig. 5. They increase in the sequence LaFeO, < LaCrO, < LaNiO, -c LaMnO, < LaCoO,. As in CO oxidation,, remarkable differences amongst these oxides were observed. For example, 10% conversion of propene on LaMnO, and on LaFeO, was attained at 565 and 645 K, respectively. Catalytic activities for total oxidation of propene and isobutene at 573 K for molar ratios HC:O, = 0.25: 1 and 2: 1 are given in fig.6. Sharp maxima (more pronounced for isobutene) were found for LaMnO, and LaCoO,. Conversions are higher for isobutene, as would be expected from its higher reactivity. From Arrhenius plots of lnr, (the areal reaction rate) against 1/T (fig. 7), activation energies (E,) for isobutene oxidation (molar ratio HC: 0, = 2) were calculated. For LaMnO,, LaCoO, and LaNiO,, the value of E, ( 1 6 & 2.6 kcal mol-1 ; 1 kcal mol-l z 4.18 kJ mol-l) was remarkably lower than those for LaCrO, andG. KREMEN~C, J. M. L. NIETO, J. M. D. TASC~N AND L. G. TEJUCA 945 8 t Fig. 6. Catalytic activity profiles of LaMO, oxides at 573 K in propene (open symbols) and isobutene (filled symbols) oxidation. Molar ratio HC : 0, = 0.25 : 1 (circles) or 2: 1 (triangles). lo-& 1 .6 1.7 1 . 8 1.9 103 KIT Fig. 7. Reaction rates (ra) for isobutene oxidation as a function of temperature: ., LaCoO,; a, LaNiO,; 0, LaMnO,; A, LaCrO,; 0, LaFeO,. Molar ratio HC:O, = 2: 1.946 CHEMISORPTION AND CATALYSIS ON LaMO, OXIDES Table 3. Reaction rate constants and reaction orders for total oxidation of isobutene on LaMO, oxides at 523 K" ~ catalyst k m n LaCrO, 0.52 x lop2 0.50 0.62 LaMnO, 7.54 x 10-2 0.56 0.49 LaFeO, 0.57 x 10-3 0.00 0.67 LaCoO, 2.20 x 10-1 1.05 0.42 LaNiO, 2.08 x 0.58 0.56 a k , m and n were obtained from Y = kPgobuteneP&, as defined in the text. LaFeO, (30f0.2 and 31 k0.l kcal mot1, respectively) and was very near to that found by Yao12 for ethylene oxidation on LaMnO, (1 7 kcal mol-l).Reaction rate constants ( k ) and reaction orders (m,n) for the total oxidation of isobutene on LaMO, oxides at 523 K are given in table 3. They were calculated by fitting kinetic data of r against Pisobutene (or Poz) to a potential equation r .= kPgobutene P& [where r is the reaction rate in mol g (CO,) formed per g (catalyst) per hour]. Like the catalytic-activity pattern in fig. 6, k values showed maxima for LaMnO, and LaCoO,. Note that the highest values of m and n were found for LaCoO, and LaFeO,. Therefore r is most affected by increasing the pressure of isobutene or 0, for the most active or the least active catalyst, respectively. PARTIAL OXIDATION Following propene and isobutene oxidation, analyses for acrolein, methacrolein, acetone, acetic acid, CO and CO, were made.Representative results for isobutene oxidation on LaCoO,, LaMnO, and LaFeO, are given in table 4. With LaCoO, and LaNiO, no partial oxidation products were detected for any of the reacting mixtures used. In propene oxidation, no formation of oxygenated organic com- pounds was observed, except for traces of acetone in some cases. The formation of partial oxidation products was favoured by increasing the ratio HC:O, and thus with increasing coverage of hydrocarbon. This is in agreement with the results of Gerei et a1.' The activity for methacrolein formation increased in the series LaCoO, = LaNiO, c LaMnO, < LaFeO, c LaCrO,. Acetone was detected only when LaFeO, was used as catalyst. The trend in activity observed for CO formation depended strongly on HC: 0, molar ratio.We conclude that the catalytic activity of these oxides for partial oxidation follows the reverse sequence to that found for total oxidation. LOCALIZED INTERACTIONS As it can be seen in fig. 2 and 6, maxima were found for LaMnO, and LaCoO, in both oxygen adsorption and in catalytic activity for propene and isobutene oxidation. A similar pattern was found in the catalytic activity for CO oxidation on LaMO, oxides., Twin-peak patterns were also found by Dowden et al.,13 Dowden and Wells4 and Dixon et al.14 in reactions involving hydrogen catalysed by first-row transition-metal oxides and by Boreskov15 in oxidation reactions catalysed by a series of spinel-type oxides. In addition, Seiyama et a1.l1 found maxima for oxygen adsorption on MnO and Co,04 within a series of oxides of the first-row transition metals.In perovskite-type oxides, B ions (M in LaMO,) are situated in the centre of anG . KREMEN~C, J. M. L. NIETO, J. M. D. TASCON AND L. G. TEJUCA Table 4. Selectivity to product i [S, (%)I in isobutene oxidation at a total conversion (X,) of 3% oxide HC: 0, methacrolein acetone 0.25 LaFeO, LaCrO, 2.00 0.25 LaMnO, 0 1 3 4.4 7.8 8.2 0 1.2 1.6 0 1.5 4 0 0 0 0 0 a 947 a Traces of acetic acid. octahedron whose vertices are occupied by oxygen ions. Assuming that the most frequently exposed planes in the polycrystalline material are those with the lowest Miller index, on chemisorption of oxygen the M3+ ion should recover the coordination of the bulk, changing from square pyramidal to octahedral [(loo) plane], from tetrahedral to octahedral [( 1 10) plane] and from trigonal to octahedral [( 1 1 1) plane].The change in crystal-field stabilization energy (A&) due to the change in coordination of the M3+ ion usually exhibits two maxima when passing form d o to d10 ions.16 The position of the experimental peaks in the adsorption and catalytic processes studied does not coincide with the position of the AEc peaks, as was also observed by Dowden et all3 and by Boreskov.15 The surface chemistry of perovskite oxides has been little studied. However, it is known that the extent of reduction with hydrogen at a given temperature is different for each oxide (see below). This indicates that the M-0 bond strength in the perovskite structure differs for each member of the series.Therefore, under conditions of adsorption and catalysis, surface defects different in concentration and nature may be produced in each oxide and this could account for the observed peak displacement. In any case, the similarity between the AEc profile and the experimental profiles of adsorption and catalysis found in this work supports the ideas of Dowden and Wells4 on the relationship between the local symmetry of surface cations and chemisorption and catalysis and therefore shows the importance of localized interactions (i.e. those where free electrons or holes are not implied) in these surface processes. On the other hand, the parallelism observed between catalytic activity for total oxidation and oxygen adsorption (the best adsorbents being oxides exhibiting the highest catalytic activity) shows that adsorbed oxygen plays an important role in these catalytic reactions.This is in agreement with results indicating that total oxidation of hydrocarbons occurs through a suprafacial catalysis mechanism in which the catalyst provides orbitals of the appropriate energy and symmetry for bond formation with reactants and intermediates.ll In this mechanism (as opposed to an intrafacial catalysis mechanism, e.g. that occurring in partial-oxidation processes where oxygen from the oxide catalyst is consumed and regenerated continuo~sly~~ 9 la) adsorbed oxygen (not lattice oxygen) is the species that participates in the catalytic reaction.948 CHEMISORPTION AND CATALYSIS ON LaMO, OXIDES CONCLUDING REMARKS Marshneva and Boreskovlg found, from thermal-desorption measurements, that bond energies of oxygen weakly adsorbed on Cr,O,, Mn,O, and Fe,O, have a mini- mum in manganese oxide.On the other hand, the activation energy for the isotopic exchange of oxygen in simple oxides with molecular 0, (which is a measurement of the strength of the M-0 bond and is proportional to the total selectivity20) increases in the sequence Co,O, < MnO, < NiO < Fe,O, < Cr,O, < TiO,.,l Assuming that similar trends also hold for perovskite-type oxides, these findings would be consistent with the high activity for total oxidation of LaMnO, and LaCoO, and also with the higher selectivity for partial-oxidation products exhibited by LaCrO, and LaFeO,. Recent work carried out in our laboratory has indicated that in some instances the trend observed for 0, adsorption and catalytic activity also holds for reducibility of these oxides.For example, LaCo0,22 and LaNiO,,, (good adsorbents and catalysts) were both reduced (in H,) to 3e- per molecule ( i t . Co3+ or Ni3+ were reduced to Coo or Ni") at 770 K. LaCr0,24 (both a poor adsorbent and poor catalyst) was reduced to only lo-, e- per molecule at 1170 K. In the former oxides the initial outgassing treatment (15 h at 773 K) may produce a high concentration of surface oxygen vacancies25 and this may cause the 0, adsorption to be higher. However, for LaMnO,, which exhibited a maximum in both 0, adsorption and catalytic activity, temperatures of ca. 1070 K were needed for a reduction to le- per The results reported suggest that A& may be a significant factor in the energetics of chemisorption and catalysis and show the importance of surface metal ions M3+ as active centres in the processes studied.We thank CAICYT for sponsorship of this work. Thanks are also due to the Ministerio de Educacion' y Ciencia for a grant awarded to J. M. D. T. We are much indebted to the referees for their helpful suggestions. S. Katz, J. J. Croat and J. V. Laukonis, Ind. Eng. Chem., Prod. Res. Dev., 1975,14,274; H. Katman, L. Pandolfi, L. A. Pedersen and W. F. Libby, Chemtech, 1976 (June), 369. W. C. Conner Jr, S. Soled and A. Signorelli, in Proc. 7th Int. Congr. Catal., ed. T. Seiyama and K. Tanabe (Elsevier, Amsterdam, 1981), part B, p. 1224. J. M.D. Tascon and L. Gonzalez Tejuca, React. Kinet. Catal. Lett., 1980, 15, 185. D. A. Dowden and D. Wells, Actes 2eme Congr. Int. Catalyse (Technip, Pans, 1961), vol. 2, p. 1499. J. M. D. Tascon, S. Mendioroz and L. Gonzalez Tejuca, Z. Phys. Chem. (Neue Folge), 1981,124, 109. Y-F. Y. Yao and J. T. Kummer, J . Cutul., 1973, 28, 124. S. W. Gerei, E. F. Rozhkova and Ya. B. Gorokhovatsky, J . Catal., 1973, 28, 341. T. Ono, T. Tomino and Y. Kubokawa, J. Catal., 1973, 31, 167. K. Hata, S. Kawasaki, Y. Kubokawa and H. Miyata, Proc. 6th Int. Congr. Catal., ed. G. C. Bond, P. B. Wells and F. C. Tompkins (The Chemical Society, London, 1977), vol. 2, p. 1102. lo S. W. Weller, Adt. Chem. Ser., 1975, 148, 26. l 1 M. Iwamoto, Y. Yoda, N. Yamazoe and T. Seiyama, J. Phys. Chem., 1978, 82, 2564. l 2 Y-F. Y. Yao, J. Catal., 1975, 36, 266. l3 D. A. Dowden, N. Mackenzie and B. M. W. Trapnell, Proc. R. Soc. London, Ser. A, 1956, 237,245. l4 G. M. Dixon, D. Nicholls and H. Steiner, Proc. 3rd Int. Congr. Catal., ed. W. M. H. Sachtler, l 5 G. K. Boreskov, Proc. 5th Znt. Congr. Catal., ed. J. W. Hightower (North Holland, Amsterdam, 1973), Ifi A. Clark, The Theory of Adsorption and Catalysis (Academic Press, New York, 1970), p. 360. l8 G. W. Keulks, L. D. Krenzke and T. M. Notermann, Adv. Catal., 1978, 27, 183. lS V. I. Marshneva and G. K. Boreskov, React. Kinet. Catal. Lett., 1974, 1, 15. 2o T. Seiyama, N. Yamazoe and M. Egashira, Proc. 5th Int. Congr. Catal., ed. J. W. Hightower (North G. C. A. Schuit and P. Zwietering (North Holland, Amsterdam, 1965), vol. 2, p. 815. vol. 2, p. 981. R. J. H. Voorhoeve, D. W. Johnson Jr, J. P. Remeika and P. K. Gallagher, Science, 1977, 195, 827. Holland, Amsterdam, 1973), vol. 2, p. 997.G. KREMEN~C, J. M. L. NIETO, J. M. D. TASCON AND L. G. TEJUCA 949 41 A. I. Gelbshtein, S. S. Stroeva, Y. M. Bakshi and Y. A. Mischenko, Proc. 4th Znt. Congr. Catal. (Akademiai Kiado, Budapest, 1971), vol. I, p. 297. 22 M. Crespin and W. K. Hall, J . Catal., 1981, 69, 359. 23 J. L. G. Fierro, J. M. D. Tascon and L. Gonzalez Tejuca, J . Catal., 1984, 89, 209; 1985, in press. 24 J. L. G. Fierro and L. Gonzalez Tejuca, J . Catal., 1984, 87, 126. 25 L. Gonzalez Tejuca, C . H. Rochester, J. L. G. Fierro and J. M. D. Tascon, J . Chem. SOC., Faraday Trans. 1, 1984, 80, 1089. (PAPER 4/887)

ISSN:0300-9599    

DOI:10.1039/F19858100939

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. 1, 1985,81, 951-959 The Catalytic Cracking of Cumene Studied by Reversed-flow Gas Chromatography BY NICHOLAS A. KATSANOS* AND MICHAEL KOTINOPOULOS Physical Chemistry Laboratory, University of Patras, Patras, Greece Received 3 1st May, 1984 The new technique of reversed-flow gas chromatography has been applied to the catalytic crachng of cumene over LaY and HY zeolites. The delta mechanism of Campbell and Wojciechowski (J. Catal., 1971, 20, 217) was chosen as the basis for the derivation of the rate equation. Two limiting cases of this equation arise when one or two steps in the proposed mechanism are rate-controlling. With the aid of the equations derived, rate constants and activation parameters for the dealkylation reaction of cumene have been determined.For the LaY catalyst the formation of propene is governed by one rate-controlling step, whilst in the formation of benzene two such steps are involved, the second of which is the slow desorption of the product. For the HY catalyst the desorption of benzene appears to be the only rate-controlling step. The catalytic cracking of cumene over zeolite catalysts has been the subject of various investigations concerning the active sites on the catalyst surface, the identifi- cation and yields of the primary and secondary products and a determination of the reaction mechanism. The last has been based on kinetic studies performed with three common reactor types : pulse microcatalytic, differential and integral. Best and Wojciechowski' have tabulated the results obtained with these three reactor types by various authors.Prater and Lago,2 using a differential reactor to study the kinetics, were among the first to propose a mechanism for the dealkylation of cumene to benzene and propene. This mechanism was expanded by Campbell and Wojciechowski3 to the so-called delta mechanism shown in fig. 1 . Based on this mechanism, rate expressions, in terms of fractional conversions and various model parameters, were derived for the cases that the rate-controlling step is (i) the bond-breaking step,3 (ii) the desorption of the reaction products4 or (iii) the adsorption of cumene.* These expressions were extensively used by Wojciechowski and coworkers in various kinetic studies of dealkylation and other reactions occurring in cumene cracking over LaY and HY zeolites.lY 4-12 The rate equations mentioned above, however, are complicated expressions involving many parameters, defined in terms of other parameters (mainly adsorption equilibrium constants).The fit of these equations to the experimental data, leading to the determination of model parameters, activation energies etc. is a difficult task, solved only with the aid of a powerful computer. By contrast, the technique of reversed-flow gas chromatography (r.f.g.c.) is a much simpler differential method of studying the kinetics of heterogeneous catalysis. Its equations are simple, containing only the rate constants of the various steps as parameters. Fitting them to the experimental data is straightforward, leading to a determination of the true rate constants and activation energies rather than apparent values involving heats of adsorption.R.f.g.c. has been used to measure not only rate constants of chemical reactions but also the rate coefficients of various physical processes, such as gas diffusion coefficients 95 1952 CATALYTIC CRACKING OF CUMENE z s S - c s 2 - Y S k2 Fig. 1. The delta mechanism of Campbell and Wojciech~wski.~ C, Y and Z represent cumene, propene and benzene, respectively, S the active sites of the catalyst and CS, YS and ZS the corresponding chemisorbed species. The various k are rate constants for the chemical reaction and the adsorption-desorption processes shown. in binary mixtures13 and their temperature variation,l* rate coefficients for evaporation of liquids,15 diffusion coefficients in multicomponent gas mixtures,lG adsorption equilibrium constants'' and rates of the drying of catalysts.18 It was in the domain of surface-catalysed reactions, however, that the method was introduced, first in a preliminary f01-d~ followed by a detailed theoretical analysis20 and applications to the dehydration of alcohols21 and the deamination of primary amines.22 The method was then extended to more complicated reactions with two gaseous reactants, namely the oxidation of carbon monoxide with oxygen23 and the hydrogenation of ~ r o p e n e .~ ~ A review of the method is soon to appear.25 In the present paper the r.f.g.c. is used to study the kinetics of another class of organic reaction, namely catalytic cracking. The dealkylation of cumene over LaY and HY zeolites was chosen, giving propene and benzene as the main products.EXPERIMENTAL MATERIALS The LaY catalyst was prepared from Linde molecular-sieve base LZ-Y52 (Ventron, 10&150 mesh, SiO,:A1,0, ratio 4.74, surface area 900 m2 g-l). 10 g of this was mixed with LaCl, in aqueous solution (50 cm3 containing 6 g LaCl;xH,O, Merck, pro analpi) and left at room temperature for 16 h. It was then thoroughly washed with distilled water until free of chloride and dried at 393 K for 24 h. From this, pellets were prepared which were placed in a porcelain dish and heated in a water-vapour atmosphere at 2 atm* pressure for 11 h. They were then dried in an oven at 410 K for 2 h and crushed to 100-150 mesh particles. Using classical analytical methods it was found that the degree of ion exchange was 64%.The HY catalyst was prepared by heating as described below using Linde molecular-sieve base LZ-Y82 (Ventron, 10&150 mesh, SiO, : A1,0, ratio 5.38, surface area 770 m2 g-l). This is an ammonium zeolite. Cumene (Fluka, purity 2 99%), benzene (Merck, purity 2 99.7%) and propene (Matheson Gas Products, purity 2 99.7%) were used as such. The carrier gas was helium (Linde, Athens, 99.99% purity). APPARATUS AND PROCEDURE The experimental set-up and the procedure followed in the application of the r.f.g.c. method have already been reported.20 The catalyst was contained in two gas-chromatographic columns * 1 atm = 101 325 Pa.N. A. KATSANOS AND M. KOTINOPOULOS 953 I' and I connected in series with an injector between them.In the present work the lengths I' + I were 14.3 cm (i.d. 4 mm)+ 119.8 cm (i.d. 1.6 mm), containing 1.41 + 1.74 g LaY catalyst, and 2 + 43.6 cm (both of i.d. 4 mm), containing 0.15 + 3.77 g base LZ-Y82. Conditioning of the columns was carried out in situ under carrier-gas flow (0.33 and 0.65 cm3 s-l for the first and second columns, respectively). The starting temperature was 373 K and this was increased at a rate of 1 1 K min-l up to 683 K, where the columns were kept for 20 h. During this period the base LZ-Y82 lost ammonia and was transformed into the HY zeolite. Following each activation the columns were cooled to the working temperature and, after some preliminary injections of reactant, 2-5 mm3 of liquid cumene was introduced with a microsyringe through the injector between the two columns. The carrier gas flowed, with the same volumetric rates mentioned above, in the direction F, i.e.entering the short column I' and leaving the long column 1 towards the detector. The carrier-gas inlet and the detector were connected to the columns through a six-port valve, by means of which the flow of the carrier gas was reversed to the opposite direction R, then back to F and so on. The time interval between any two successive reversals of the gas flow was always greater than the sum, tk+ t,, of the retention times of the products on the column sections I'+1. Each flow reversal creates extra peaks (sample peaks) on the chromatographic elution curve,2o and the heights or areas under these peaks are used to calculate the rate constant of the reaction being studied.Plots and calculations were made on a Hewlett-Packard 9825 A desk-top computer connected to a 9872 B plotter. Identification of the products was made by injecting reference substances and also with the help of a time-of-flight mass spectrometer CVC MA 3A. THEORETICAL ANALYSIS DERIVATION OF THE GENERAL EQUATIONS This study belongs to the case where the reactant is strongly adsorbed on the catalytically active surface, and the catalyst also functions as a chromatographic material separating the products. Since the reactant cumene (C) is introduced as a pulse in the middle of the column I'+Z, and since the chromatographic process on C is repeatedly changing direction (forwards and backwards), the same approximation as previously20 is adopted, namely that the distribution with the distance x along the column of the reaction rate r(x, to) is described by a delta function d(x - 2').Thus we can write for the production rate of a substance r(to) 6(x - l'), where r(to) is only a function of the time, z,, from the injection of C and is expressed in mol cmP2 s-l, i.e. per unit cross-section a of the void space in the column. The general equation giving the area under the sample peaks of a product in the r.f.g.c. method has been derived.20 For the R-peaks, i.e. peaks obtained with the carrier gas entering the long column 1 and leaving the short column I', this equation, in the form of the to Laplace transform, is where f represents the area under the peaks, R(po) is the transformed rate of production r(to) of a substance, po is the transform parameter with respect to to and t , is the retention time of the product on column 1. For the F-peaks, obtained with the carrier gas flowing in the opposite direction, the same equation applies with the retention time tk on column I' substituted for tR.To apply eqn (1) to the dealkylation of cumene over LaY and HY zeolites we must (a) assume that the products propene and benzene are subject to linear chromatography on these materials, based on reversible physical adsorption, and (b) determine the rate of production of these products assuming a certain mechanistic model for the reaction. 32 FAR 1954 CATALYTIC CRACKING OF CUMENE By substituting the Laplace-transformed rate R(po) in eqn (1) and taking the inverse transform of po, the area f of the sample peaks will be found as an analytic function of to.This function can then be fitted to the experimentally determined peak areas to calculate the rate constants of the slow rate-controlling step(s). If we adopt the delta mechanism of Campbell and Wojciechowski (fig. 1) as the working mechanistic model and neglect the back reaction (the k-, branch), since the method is a differential one, we derive the following equations. (2) - k1 CS CC - k-1 cCS Rate of disappearance of C: -- dCC - dt0 (3) dccs - = k , cs cc - k-, ccs - k , ccs dt0 rate of change of CS : (4) rate of change of YS: - dcys = k, ccs + k-, cs cy - k, cys d to and rate of production of Y: dc, = k3 cys - k-, cs cy ( 5 ) dt0 where the various c are molar concentrations of the species shown as subscripts.Since the amount of catalyst is relatively large and the amount of the reactant very small, cs can be regarded as a constant and combined with k , in eqn (2) and ( 3 ) or with k-, in eqn (4) and ( 5 ) to give a new constant: k', = k , cs and k l , = k-, cs. (6) The system of differential equations (2)-(5) can be transformed into a system of algebraic equations by Laplace transformation with respect to to, using the initial conditions Solving this new system of equations for C, (the Laplace-transformed function of cy) we obtain (8) cC(0) = co and ccs(0) = c,,(O) = cy(0) = 0. (7) k , co - AB~O+kWO+k,) c - where and (9) The rate of formation of Y is simply r = V(dc,/dt,)/a, where V is the volume of the gas phase in the column.The Laplace transform of the rate, under the initial condition (7), is R = Vpo C y / a . This, substituted into eqn (l), gives Yto f Y = VCY [ 1 - exp ( -Po fR)1. (1 1) The inverse Laplace transform of this expression, with eqn (8) substituted for C,, gives the areafY under the R-peaks (or under the F-peaks with tk in place of tR) as an analytic function of the time to, for any step of fig. 1 being rate-controlling or even for all steps influencing the rate of appearance of Y .N. A. KATSANOS AND M. KOTINOPOULOS 955 Eqn (1 1) also gives the area under the R- and F-peaks if these consist of the other product Z. In this case C, in eqn (1 1) must be substituted by Cz, and this is given by an expression analogous to eqn ( 8 ) with k, and k l , (= k-,c,) substituted for k, and k:,, respectively.LIMITING CASES OF EQN (8) AND EQN (1 1) If only one of the steps of the mechanism in fig. 1 is slow and rate-controlling, while the others are fast, eqn (8) is greatly simplified by omittingp, compared with the rate constants of the fast steps. This simplified equation is then substituted for Cy in eqn (1 l), giving an expression whose inverse Laplace transform is easily found. Let us take, for example, the k, step (the dealkylation reaction) as rate-controlling, assuming the adsorption of C (the k, step) and the desorption of YS (the k, step) to be fast. Omitting po in comparison with ki and k,, eqn ( 8 ) becomes k2 co cy = ~ A Bk; k, with A and B now given by A = (Po+ k,)/k; and B = po/k3. This, substituted for C, in eqn (1 l), gives and the inverse transform of this expression (for to > tR) is Had adsorption of C (the k, step) been taken as rate-controlling, po would have been omitted in comparison with k, and k, in eqn (8), and the final result would again be eqn ( 14), where in place of k2 an apparent rate constant k, would appear: k; 1 + k+/k, * k, = Finally, if desorption of Y were rate-controlling (the k, step) the following equation would be obtained instead of eqn (14) : where k, = k, + kL,.Eqn (14) and (1 6) show that, if one step of fig. 1 is rate-controlling, a linear plot of In fu against to is predicted. From the slope of this plot a rate constant can be determined but it is impossible to ascertain whether this is k,, k, or k,, since the final equation givingf, as a function of to has exactly the same form in all three cases.This was also the case with the rate expressions derived by Best and Wojciechowski4 using the same mechanism. A second limiting case of eqn (8) and (1 1) is that the overall reaction rate is controlled by two steps in the mechanism of fig. 1, i.e. only one of the three steps (k,, k2 or k,) is fast compared with the other two. This possibility has never been explored before owing obviously to the very complicated rate equations which would result. This is not the case, however, with the r.f.g.c. technique employed here. Let us assume, 32-2956 CATALYTIC CRACKING OF CUMENE for example, that only adsorption of C (the k, step) is fast, while both other steps (k2 and k,) are slow.Then, omittingp, in comparison with k; only, eqn (8) becomes This is substituted for C , in eqn (1 1) as before, and the inverse Laplace transformation is found for to > tR: where ks = k3 + k:,. The factors before exp (- k, to) and exp (- k, to) inside the large parentheses are both approximately equal to tR if k, and k, are sufficiently small. The two other possibilities of having two slow steps occur when only the k2 step or only the k3 step is fast, the remaining two steps in each case being slow. Both these cases lead to the same integrated rate equation as eqn (18), with different rate constants. Therefore, the mere fitting of these equations to the experimental data does not distinguish between the three possibilities above. However, it does distinguish a mechanism with two slow steps from another with one such step, since the behaviour of eqn (18) is different from that of eqn (14) or (16).A product whose production rate follows eqn (1 8) exhibits an induction period, its amount increasing initially with time, passing through a maximum and then decreasing almost exponentially with a rate coefficient equal to the smaller rate constant. RESULTS AND DISCUSSION CUMENE DEALKYLATION OVER Lay With this catalyst both products of the dealkylation reaction Y and Z (propene and benzene, respectively) appear on the reversed-flow chromatogram forming separate sample peaks with different retention times. The peaks of propene decrease with time from the outset, according to a simple exponential law, as fig. 2 shows. Obviously, the formation of propene follows eqn (14) or (16), and therefore a limiting case of the delta mechanism with only one step being rate-controlling. From the slopes of plots such as that of fig.2, the rate constants of table 1 were calculated by standard least-squares procedures. The sample peaks of benzene have significant height only in the R-direction. As shown previously,20 this is due to the large t , value of benzene on column I as compared with the very small tk value on column l’. The areaf, under these peaks initially increases with time until a maximum value is reached and then decreases. This behaviour is predicted by eqn (18) but not by eqn (14). Thus the formation of benzene involves two slow steps in the delta mechanism. One of these must be the slow desorption of benzene with a rate constant k, = k,+kl_,, since adsorption of cumene and the bond-breaking act are steps common to both products, propene and benzene.Had both of these steps been slow, the formation of propene would also follow eqn (1 8) and not eqn (14). Unfortunately, the exact values of the rate constants for benzene formation were not determined, owing to some overlap of the propene and benzene peaks in the R-direction. For the same reason, the k values of propene formation were determined only from the F-peaks. As was pointed out in the Theoretical section, it is not possible from the rate equation to tell which of the three rate constants (k,, k, or k,) the k values of table 1N. A. KATSANOS AND M. KOTINOPOULOS 957 1 3 5 7 9 11 t , / 1 0 3 s Fig.2. Plot of eqn (14) or (16) for the cracking of cumene over LaY catalyst at a carrier-gas flow rate of 0.33 cm3 s-l. The fy values are the areas under the propene. 539 K with F-peaks of Table 1. Kinetic parameters for propene formation during dealkylation of cumene over LaY catalysta T/K k/lOP4 s-' EJkJ mol-1 ln(A/s-l) 525 0.52f0.01 135+4 2 1 f l 539 1.30 & 0.02 546 1.75 IfI 0.01 551 2.37k0.04 561 3.8f0.1 a The & values in this and the following table are standard errors analysis. found from regression represent. There is some indication, however, that this is k, because the activation energy of 135 kJ mol-l, calculated from the Arrhenius plot of fig. 3, is rather high to be connected with k, or k,, which represent adsorption-desorption processes. The plots in fig.3 indicate that both LaY and HY used in this work are diffusion-free catalysts, but the activation energy found by Corma and WojciechowskilO on a similar catalyst was 98 f 8 kJ mol-l, i.e. 37 kJ mol-1 lower than our value. This could be due to the different base material used by us, having a %/A1 ratio of 4.74 as compared with 2.25 of the above authors; or it could be due to the fact that a true rate constant is determined by the present method, pertaining strictly to the surface reaction and not combined with adsorption equilibrium constants. Another implication of k, being a true rate constant is that the value of the frequency factor found is due to the entropy of activation, which is calculated as -84 J K-' mol-I. This negative value indicates that the transition state CS# on the catalyst surface is more localized or generally has less freedom on the surface than the adsorbed species CS.Conventional rate equations like those of Wojciechowski et ~ 1 . ~ 9 ~ lead to the determination of k2cS, and Arrhenius plots based on this combination cannot give the entropy of activation.958 CATALYTIC CRACKING OF CUMENE -7.50 -8.00 & -8.50 -9.00 h - E 3 -9.50 -10 .oo lo3 KIT 1.60 1.65 1.70 1.75 1.80 I I I 7 -7.00 -7.20 h - - 7 - 4 0 > 5. K - -7.60 -7.80 1.75 1.80 1.85 1.90 lo3 KIT Fig. 3. Arrhenius plots for the cracking of cumene: 0, rate constants for formation of propene over LaY catalyst (lower abscissa, left ordinate); x , rate constants for formation of benzene over HY catalyst (upper abscissa, right ordinate).Table 2. Kinetic parameters for benzene formation during dealkylation of cumene over HY catalyst T/K from F-peaks from R-peaks Aka(%) 573 3.60 f0.07 3.8 & 0.2 5.6 582 4.4L0.7 4.8 f 0.1 9.1 595 5.3k0.1 5.6 f 0.1 5.7 603 6.7k0.2 6.4 f 0.6 4.5 611 7.7f0.1 7.2 f 0.6 6.5 EJkJ mol-l: 53 L 3, In (A/s-l): 3.1 k0.5 a Defined as 100 1 k,-k, I / k F . CUMENE DEALKYLATION OVER HY The reversed-flow chromatograms with this catalyst show only the peaks of benzene, while those of propene are delayed because of the formation of carbonium ions on the catalytic ~urface.~ These ions are probably very slowly decomposed, because of the high concentration of Bronsted sites on this catalyst. From the F- and R-peaks of benzene, which follow eqn (14) or (1 6), rate constants in both directions were determined and are collected in table 2.Note that the percentage difference between the two k at the same temperature is not at all significant, in spite of the fact that the F- and the R-peaks are due to such different column lengths I' and I (2 and 43.6 cm, respectively) containing such different amounts of catalyst (0.15 and 3.77 g, respectively). This has also been noted earlier2'> 22 for other reactions. Using the mean k value at each temperature, the Arrhenius plot for HY wasN. A. KATSANOS AND M. KOTINOPOULOS 959 constructed and is given in fig. 3. The activation energy and the frequency factor, calculated from the plot, are given in the lower part of table 2. The relatively low value for E, indicates that the rate-controlling step in benzene production is probably the desorption of thp, product and the k values of table 2 represent k, = k4+k14.This conclusion is consistent with that drawn earlier for the LaY catalyst, that one of the two slow steps there must be thz desorption of benzene. A second slow step is not observed in HY, and this is an indication that the dealkylation step is very fast here. CONCLUSION The r.f.g.c. method, applied to the catalytic cracking of cumene and based on the Campbell-Wojciechowski mechanism, leads to an easy determination of rate constants, activation energies and entropies of activation. The method uses simple integrated rate equations easily fitted to the experimental data, by means of which a limiting case of only one rate-controlling step can be distinguished from a case with two such steps.We thank Mr A. Niotis and Mrs Margaret Barkoula for their kind assistance. D. A. Best and B. W. Wojciechowski, J. Catal., 1977, 47, 343. Komarewski and P. B. Weisz (Academic Press, New York, 1956), vol. 8, p. 293. D. R. Campbell and B. W. Wojciechowski, J. Catal., 1971, 20, 217. D. A. Best and B. W. Wojciechowski, J. Catal., 1973, 31, 74. D. A. Best and B. W. Wojciechowski, J. Catal., 1977, 47, 1 1 . D. A. Best and B. W. Wojciechowski, Preprints Can. Symp. Catal., 1977, 5, 400. ' D. A. Best and B. W. Wojciechowski, J. Catal., 1978, 53, 243. D. A. Best and B. W. Wojciechowski, Can. J. Chem. Eng., 1978, 56, 588. A. Corma and B. W. Wojciechowski, Preprints Can. Symp. Catal., 1979, 6, 149. 2 C. D. Prater and R. M. Lago, in Advances in Catalysis, ed D. D. Eley, W. G. Frankenburg, V. I. lo A. Corma and B. W. Wojciechowski, J. Catal., 1979, 60, 77. 'I1 A. Corma and B. W. Wojciechowski, Can. J. Chem. Eng., 1980, 58, 620. l2 A. Corma, H. Farag and B. W. Wojciechowski, Int. J. Chem. Kinet., 1981, 13, 883. l 3 N. A. Katsanos and G. Karaiskakis, J. Chromatogr., 1982, 237, 1 . l4 N. A. Katsanos and G. Karaiskakis, J , Chromatogr., 1983, 254, 15. l5 G. Karaiskakis and N. A. Katsanos, J. Phys. Chem., 1984, 88, 3674. l6 G. Karaiskakis, N. A. Katsanos and A. Niotis, Chromatographia, 1983, 17, 310. G. Karaiskakis, N. A. Katsanos and A. Niotis, J. Chromatogr., 1982, 245, 21. G. Karaiskakis, A. Lycourghiotis and N. A. Katsanos, Chromatographia, 1982, 15, 351. l 9 N. A. Katsanos and I. Georgiadou, J. Chem. SOC., Chem. Commun., 1980, 242. N. A. Katsanos, J. Chem. SOC., Faraday Trans. 1, 1982, 78, 1051. *l G. Karaiskakis, N. A. Katsanos, I. Georgiadou and A. Lycourghiotis, J. Chem. Soc., Faraday Trans. I , 1982, 78, 2017. 22 M. Kotinopoulos, G. Karaiskakis and N. A. Katsanos, J. Chem. SOC., Faraday Trans. I , 1982, 78, 3379. 23 G. Karaiskakis, N. A. Katsanos and A. Lycourghiotis, Can. J. Chem., 1983, 61, 1853. 24 N. A. Katsanos, G. Karaiskakis and A. Niotis, Proc. 8th Int. Congr. Catal., West Berlin, 1984, vol. 25 N. A. Katsanos and G. Karaiskakis, Adv. Chromatogr., 1984, 24, 125. 111, p. 143. (PAPER 4/895)

ISSN:0300-9599    

DOI:10.1039/F19858100951

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. 1, 1985,81,961-972 Ul traviole t-Visi ble Spec tropho t ome tric Determination of Ion-association Constants for Alkylpyridinium Iodides N-Ethyl-4-cyanopyridinium Iodide in Mixed Solvents Containing Ethanol BY MOHAN PAL AND SANJIB BAGCHI* Department of Chemistry, University of Burdwan, Burdwan 71 3 104, India Received 5th June, 1984 A new method of analysis of spectrophotometric data has been developed to deal with the problem of ion-pair formation represented by the scheme contact ion pair + solvent-shared/solvent-separated ion pair + free ions for alkylpyridinium iodides. Variation of the composition of a mixed solvent enables the position of each equilibrium to be measured. It has been observed that solvent-shared ion pairs are distinct entities in ethanol solution (ca.80% of the associated species) and the extent of solvent-shared ion pairing decreases as the polarity of the solvent decreases. The primitive model can explain successfully the formation of an ion pair from the free ions in mixed solvents as long as the percentage of polar component is high, but effects caused by solute-solvent interactions appear at the non-polar end of the mixed-solvent system. Finally the association constants have been correlated with a suitable solvent-polarity parameter. It is now well established that in a solution of low dielectric constant an electrolyte exists as the following distinctly different species:' viz. (1) the contact ion pair (c.i.p.), where two oppositely charged ions are in physical contact, (2) the solvent-shared/ solvent-separated ion pair (s.s.i.p.), the two ions forming an ion pair remain separated by one/more solvent molecules, and (3) the free ions.Interionic attractions coupled with thermal motion lead to the following equilibria: KS c.i.p. & s.s.i.p. free ions K h J/' The distribution between the different species in a solution is determined by the equilibrium constants K, and KA, and a complete solution of this problem is to determine the nature of s.s.i.p. and to measure the position of each equilibrium. Although much work has been done on the thermodynamics of ion association, very little has been done on determining the equilibrium constant (K,) for the interconversion of the ion-pair sub-species. Recently Symons and coworkers2 have performed a spectrophotometric investigation of the nature of s.s.i.p.and Rao3 has made theoretical computations on the energitics of ion-pair solvation. FUOSS~ has developed a method for analysis of the conductance data leading to evaluation of K,. On the other hand, it has not become possible to distinguish between the s.s.i.p. and the c.i.p. by an ultraviolet-visible absorption spectrophotometric study except for some special cases where the s.s.i.p. shows characteristic ab~orption.~ 96 1962 U.V.-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS Table 1. Properties of different mixed solvent systems at 25 "C. W,, weight percentage; XI, mole fraction of ethanol; D12, dielectric constant of mixed solvent; dlz, density of mixed solvent. solvent w, dl 2 Dl, ethanol + dioxane ethanol +carbon tetrachloride ethanol + ethylacetate ethanol + dichloroethane pure ethanol 0.00 7.8290 16.0463 33.76 12 43.3278 75.3582 87.3 110 0.00 11.09 20.83 30.73 40.19 62.97 82.4 1 0.00 8.8852 17.9933 36.9 128 46.7419 67.190 88.7626 0.00 7.9568 14.4005 22.384 1 30.9686 40.2244 6 1.091 8 100.00 0.00 0.14 0.2677 0.4937 0.5940 0.8540 0.9293 0.00 0.2934 0.4772 0.5876 0.6920 0.8512 0.9406 0.00 0.1572 0.2957 0.5282 0.6267 0.7966 0.9379 0.00 0.1386 0.2686 0.3830 0.49 13 0.591 5 0.77 16 1 .oo 1.0269 1.0027 0.9785 0.9302 0.8940 0.8337 0.8093 1.5839 1.4228 1.3076 1.2092 1.1266 0.9675 0.8632 0.89455 0.8836 0.8727 0.8508 0.8398 0.8 180 0.7960 1.1667 1.1285 1.0904 1.0522 1.0140 0.9759 0.8995 0.785 1 2.209 4.33 6.55 10.9479 13.171 1 9.8446 22.072 2.25 4.45 8.12 11.76 14.60 19.67 22.69 6.02 7.8342 9.6579 14.3873 15.14 18.80 22.4665 10.00 1 1.4935 12.00 14.43 15.90 17.13 20.34 24.30 Alkylpyridinium iodides provide a suitable system for studying ion-association phenomena spectrophotometrically.The solutions of these compounds show a characteristic absorption band in the u.v.-vis. region, originatingfrom acharge-transfer (c.t.) process within the contact ion pairs. It has been observed by several workers6. that for this system the molar absorbance of the absorbing species in solution varies with the nature of the solvent; this has been attributed to the existence of the s.s.i.p. However, the values of K, and E ~ , the molar absorbance of the absorbing species (the c.i.p.), have not been determined. The existence of the s.s.i.p.finds support from ultrasonic relaxation studies by Hemmes et aZ.,* but there is no direct spectrophoto- metric support for its existence. In the present work we have determined the thermodynamic parameters and the molar absorbance of N-ethyl-4-cyanopyridinium iodide for a series of mixed solvents of varying composition and from these we have been able to find the values of K, and E,. For the binary solvent system we have used mixtures made up with ethanol (D = 24.3) and different aprotic cosolvents. As cosolvents we have used dichloroethane (D = 10.0), ethylacetate (D = 6.02), carbon tetrachloride (D = 2.25) and 1,4-dioxane (D = 2.21), i.e. four aprotic solvents with different polarities.M. PAL AND S . BAGCHI 963 EXPERIMENTAL MATERIALS 4-Cyanopyridine (Koch-Light) was quaternised by methods described previously.6 The solvents were purified by standard proceduresg. lo and distilled immediately before use.Mixed solvents were prepared by carefully mixing the components so as to minimise contamination by moisture. Physical properties of the binary solvents used in the spectrophotometric measurements are shown in the table 1. Densities were measured in a 20 cm3 pyknometer at 25 "C. Dielectric constants were obtained from capacity measurements on a standard bridge. The value of D = 24.30 for the dielectric constant of ethanol at 25 "C was used as a calibrating datum. We have also calculated the dielectric constants of binary solvents by assuming that the polarisation of a liquid solution is the sum of additive terms for the individual components.ll Thus where P12 is the molar polarisation of the mixture and (Di - 1) (20, + 1) M, p.= (i = 1,2) 9D, d, (3) the suffixes 1 and 2 are used for the solvent components and 12 is used to denote the mixture and X , M and d are the mole fraction, molecular weight and density, respectively. The values obtained are found to be in good agreement with the experimentally determined values. SPECTROPHOTOMETRIC MEASUREMENTS Spectrophotometric measurements were carried out in a Beckman model 26 instrument using stoppered 1 cm cells placed in a thermostatted cell holder. The solutions were prepared immediately before use and precautions were taken during transfer of the solution to minimise contamination by moisture and/or air. The concentrations of the salt were ca.mol dm-3. I; may be formed in some solvents and it was detectable by its characteristic absorption bands. Only those solutions which did not show absorption from tri-iodide were used for the experiments. ANALYSIS OF SPECTROPHOTOMETRIC DATA The association of free ions to form a contact ion pair may be described by the equilibrium where a are the activities, square brackets denote molar concentrations andf, is the mean ionic activity coefficient. The conversion between contact and solvent-shared/solvent-separated ion pairs is determined by short-range ion-ion and ion-solvent interactions. In the absence of detailed information about the short-range interaction it is customary to treat this transformation as the reaction (1),4 where the solvent has been assumed to be a continuous medium and as such the solvent molecules do not appear explicitly.However, a primitive model like this seems to be inadequate in the case of a mixed solvent where the nature of the ion-solvent interaction may differ significantly for the component solvents. We may, however, modify the scheme by introducing the solvent molecules explicitly into the reaction. We assume that for the binary solvent mixtures used here, the component with the higher dielectric constant (ethanol) will preferentially form an s.s.i.p. The other component, being essentially non-polar, does not take part in s.s.i.p. formation to a significant extent. Thus the formation of an s.s.i.p. from a c.i.p. may be represented as Pt, I-+nS,P,+ * - - nS I- or (Pi )I I-) (6)964 U.V.-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS A 3 .O 2 . 4 m I -0 E d 1.8 m 2 > ---. h 7 1.2 +I 5 01 I I I I I 0 . 4 0.5 0.6 0.7 0.8 0.9 f % / A Fig. 1. Representative plot of eqn (10). where (Pi 11 I-) represents an s.s.i.p. and n is the number of solvent molecules engaged in the formation of the s.s.i.p. n = 1 corresponds to a solvent-shared ion pair and n > 1 corresponds to solvent-separated ion pair. It will be assumed that the above equilibrium conforms to the law of mass action and the activity coefficients of both forms of ion pairs may be taken as unity. Thus Of the different species present in the solution only the contact ion pair has a characteristic c.t. absorption band.6* In the wavelength range under investigation (around the c.t.band maxima) where the contact ion pair is the only absorbing species, the absorbance of a solution of unit path length at a particular v is given by A’ = [Pi, I-] t$ (8) where e: is the molar absorbance of the contact ion pair at V. First we assume that only one form of ion pair, i.e. the contact form, is present. Then for a stoichiometric concentration, Co, of the salt one can derive the equation7 from which KA and &: can be calculated graphically. If the presence of an s.s.i.p. is also considered we have in place of eqn (9). Thus for a particular solvent composition a plot off+ C,,/dAV against f+ d/AV gives a straight line and from the slope and the intercept one can obtain the associationM. PAL AND S. BAGCHI 965 Table 2.EZ; as a function of solvent composition solvents &max a m c. t. band maximum XI /cm-l l2 /dm3 mol-l cm-l ethanol + dioxane ethanol +carbon tetrachloride ethanol + ethylacetate ethanol + dichloroethane pure ethanol 0.14 0.2677 0.4937 0.5940 0.8540 0.9293 0.2934 0.4722 0.5876 0.6920 0.8512 0.9406 0.1572 0.2957 0.5282 0.6267 0.7966 0.9379 0.1386 0.2658 0.3830 0.49 13 0.5915 0.7716 1 .oo 22200 23500 25300 25600 25950 26650 24700 25100 25300 25800 26300 26500 23500 24280 25 100 25500 26 100 26600 23250 24050 24280 24700 24900 25650 26800 1800 1180 870 767 566 548 900 780 706 650 555 540 1361 1110 780 690 580 538 1015 965 800 700 617 520 500 a &Fax = 2400 dm3 mol-1 cm-l; K, for ethanol = 3.76 at 25 "C; K& for dichloroethane = 0.62 at 25 "C. constant and the molar absorption coefficient.Note that when the presence of only one type of ion pair is assumed one interprets the constants as KA and E:, but if both types of ion pair are present one determines from such plots the values of and &Epp = &E( 1 + K, a,n)-l (12) instead of KA and E:, respectively. &ipp may be looked up on as the number-averaged molar absorbance of the different ion-pair species. The values offk were calculated using the Debye-Hiickel equation with a = 5 A, calculation off,, requiring knowledge of the concentration of the free ions and the dielectric constants of the binary solvents of varying composition. In a given set of mixed solvents the concentration of the free ions is taken to be Co-Av/~Lpp. For &ipp of the ion pair in a given solvent mixture an iterative procedure was adopted.The initial value of &ipp was obtained from the plot of C o / l / A y against 1/Av ( i . e . f , values were assumed to be unity) and the results then plotted as required by eqn (10) to yield an output value of elpp.. The whole operation was then repeated until the output value of &Lpp was within 1 % of the input value. A typical plot of eqn (10) is shown in fig. 1.966 U. V.-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS 2 . 2 - 1.8 - 1 . 4 1.0 0 . 2 1 I i I I I I 0 0.2 0.4 0.6 0.8 1.0 XI Fig. 2. l/es: as a function of X I , the mole fraction of ethanol for mixtures of ethanol with: A, carbon tetrachloride; 0, ethylacetate; 0, dioxane and a, dichloroethane. RESULTS AND DISCUSSION EFFECT OF SOLVENT COMPOSITION ON THE INTENSITY OF THE CHARGE-TRANSFER BAND The experimentally determined molar absorptivity at the c.t.band maxima, e g ; , of N-ethyl-4-cyanopyridinium iodide obtained from the plot off+ C,,/dAmaX against f+ 4 A m a x for different binary solvent systems are shown in table 2. The values are f o n d to increase with increasing mole fraction of the less polar component (i.e. of lower dielectric constant value) of the binary mixtures. A plot of l/eg; against X,, the mole fraction of ethanol in different binary solvent mixtures, is shown in fig. 2. It can be seen from the fig. 2 that 1/~g; is a linear function of the mole fraction of ethanol. However, the ethanol +carbon tetrachloride system shows deviation from linearity at the non-polar end ( X , < 0.2).Moreover for the aprotic cosolvents carbon tetrachloride, ethylacetate and dioxane the points fall on one straight line, particularly when the percentage of ethanol is high. For the ethanol + dichloroethane system a different straight line, almost parallel to that obtained for other systems, with a higher intercept value is obtained. The c.t. band maximum, being sensitive to a change in the environment around the contact ion pair, changes with solvent composition (table 2). The thermodynamic association constants for a fixed solvent composition derived from the plot of eqn (10) are independent of the wavelength of measurement. Moreover, the c.t. band in solution appears to be nearly Gaussian in nature and the band shape and width are independent of the sol- vent composition.Thus the molar absorption coefficient &%; determined at the band maximum is proportional to the oscillator strength (f) of the c.t. transition. Thus a comparison of eg; for various solvent compositions should give an idea of theM. PAL AND S. BAGCHI 967 Table 3. log KA as a function of solvent composition at 25 "C solvents - ethanol + dioxane ethanol +carbon tetrachloride ethanol + ethylacetate ethanol + dichloroethane pure ethanol Et /kcal mol-1 0.14 0.2677 0.4937 0.5940 0.8540 0.9293 0.2934 0.4772 0.5876 0.6920 0.8512 0.9406 0.1572 0.2957 0.5282 0.6267 0.7966 0.9379 0.1386 0.2658 0.3830 0.49 13 0.5915 0.77 16 1 .oooo 5.1631 4.3334 3.8910 3.6055 2.9326 2.3926 - 6.0653 4.6447 3.7857 2.9674 2.4381 4.3722 4.1691 3.7385 3.4646 2.8073 2.3076 4.3886 3.7235 3.442 1 3.2746 3.1 173 2.7688 2.1802 23.094 15.2671 8.3700 7.5924 5.0391 4.5306 22.47 12.31 8.50 6.85 5.08 4.40 12.76 10.35 6.95 6.66 5.32 4.45 8.70 8.33 6.93 6.29 5.83 4.92 4.1 152 63.53 67.27 7 1.475 72.50 74.24 76.24 70.59 71.83 72.37 73.68 75.23 75.83 67.20 69.39 71.83 72.93 74.60 76.03 66.48 68.72 69.73 70.59 71.11 73.30 76.648 variation offfor various dielectric media.However, such a wide variation of ~a"p"p" with solvent compositions as has been observed experimentally (table 2) cannot be rationalised solely in terms of the variation o f f of the c.t. transition in different dielectric media. According to our model the experimentally determined molar absorbance is &max = app &Fax/( 1 + Ks a?). For different solvent compositions a, varies, thus explaining the large variation of EZ;.The above equation may be rearranged to give We may replace the quantity a,, the activity of ethanol, by X,, the mole fraction of ethanol. Under the assumption that only solvent-shared ion pairs are formed (n = l), eqn (13) becomes This explains the linear plot of 1 /EZ; against XI and also supports the solvent-shared model for an s.s.i.p. in the present case. Hemmes et al. have made similar observationss with N-methylpyridinium iodide using ultrasonic relaxation studies in acetone +water968 U.V.-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS i I I I I I 2 4 6 8 10 1 OOlD 12 Fig. 3. Plot of log KA against 1 /D12 for N-ethyl-4-cyanopyridinium iodide in various mixed solvents. Symbols as fig. 2. mixtures.It also appears from eqn (14) that the slope and intercept of the 1/&g; against X , plot would not depend on the essentially non-polar component in the binary solvent mixture so long as the amount of the more polar component (ethanol) is fixed. It can be seen from fig. 2 that this is true within the limit of experimental error when the percentage of the more polar component is high. The values of &Fax and Ks can be evaluated from the graph. We have obtained the values of Ks = 3.76 and &Fax = 2400 dm3 mol-l cm-1 for N-ethyl-4-cyanopyridinium iodide in ethanol. Note that Ks is of the same order of magnitude as determined by ultrasonic relaxation techniques8 and picosecond dynamics.13 The observed deviation from linearity for the CCl, + ethanol system in the non-polar region may be due to the formation of higher ionic aggregates.It has already been pointed out by Hyne and Levy1, that the ion-pair complex may form more complex ionic aggregates to exclude CCl, molecules. If these complex ionic aggregates do not take part in the charge-transfer process in the region of our experiment, they would decrease the value of EZ$ and thus increase 1 /EF$ in the region where the percentage of CCl, is high. Alternatively, the ion pairs may be solvated selectively by the more polar component (ethanol), leading to an increase in the local concentration of ethanol near the ion pair. To attribute some significance to the deviation noted for theethanol + dichloroethane system, we observe that the assumption made during the derivation of eqn (lo), that only the more polar (ethanol) of the binary solvents takes part in the formation of an s.s.i.p., is justified only when the other part is almost non-polar (or D , - D , is large) and there is no specific solute-solvent interaction.However, the appreciable polarity of dichloroethane, which may be further enhanced owing to a modification of the equilibrium between the gauche form and trans form of the molecule in the local electric field of the ions,I5 may lead to the formation of dichloroethane shared ion pair, (Pi 11 I-)'. EZ; will then be given by a modification of eqn (12) asM. PAL AND S. BAGCHI 969 I I 33 I I I I 0 4 8 12 16 20 24 A,E,/kcal mol-' Fig. 4. Plot of A,AG(ions) against At& for EtfNI- from methanol to various solvents.(1) ethylacetate, (2) ButOH, (3) PriOH, (4) BunOH, (5) Pr*OH, (6) EtOH, (7) CH,ClCH,Cl, (8) CH,Cl,, (9) ethyl methyl ketone, (10) acetone, (1 1) dimethylformamide, (12) MeCN and (13) dimethyl sulphoxide. where the activities of the solvents have been replaced by their mole fractions, and where KL represents the equilibrium constant for the formation of (Pi 11 I-)' from (Pi, I-) and Xz is the mole fraction of dichloroethane, which in a dilute salt solution in a binary mixed solvent may be replaced by (1 - X I ) . Thus we obtain the following equation in place of eqn (14) : According to eqn (16) a plot of 1 /E$; against X, would give a straight line with an intercept value higher than that for the case described &ove. One may also calculate the value of K; using the value &Fax = 2400 dm3 mol-l cm-l.Thus for the system under investigation K; has been evaluated as 0.62. The results are in agreement with the fact that dichloroethane, being less polar than ethanol, will form solvent-separated species to a lesser extent. THERMODYNAMIC PARAMETERS The spectrophotometric method gives the association constant [eqn (1 1)) total concentration of ion pair K = KA(l +Ksa$) = [PY'I [I-] f"* This is also true when both solvents from an s.s.i.p. Thus the spectrophotometric method, like the conductometric method, cannot distinguish between different forms of ion pairs in a solution, and as such the two methods give the same association ~onstant.~? One can determine the value of the true association constant KA only when K$ are known. Above we described a method for the determination of K,.Using those values and replacing activities by mole fractions we have calculated the values of KA for various solvent compositions; the results are summarised in the table 3. It appears that the phenomenon of ion association increases as one goes from the polar to the non-polar end of the solvent mixture. Below we discuss the correlation of KA with solvent polarity.970 U.V.-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS 7 6 5 4 - 3 G? OD 2 1 0 I I I I I I 60 64 68 72 7 6 80 ET/kcal mol-' Fig. 5. Plot of log KA against Et for N-ethyl-4-cyanopyridinium iodide in various mixed solvents. Symbols as in fig. 2; x , pure ethanol. In the primitive model, which describes the solvent as a continuum, the bulk dielectric constant D serves as a measure of solvent polarity.FUOSS~~ and Eigenl' have derived an equation for ion-pair formation in 1-1 electrolytes: 47cNa3 exp (e2/DakT) 3000 KA = w i i c i c I V IS nvugauiu s iiuiiiutx, u IS LIIG c.~i~~it=~u-c;t;~iirt; U I ~ L ~ I ~ W ociwecn i ~ i e ivris and e is the electronic change. Eqn (1 7) demands a linear relationship between log KA and 1 ID. Fig. 3 shows a plot of log KA against the reciprocal of the dielectric constant (lOO/D). It can be seen that for D 2 10 a linear relation holds, irrespective of the nature of the mixed solvents. The linear variation for mixed solvent systems may be interpreted as meaning that the microscopic dielectric constants do not differ significantly from the bulk dielectric constant.The slope of the line allows an estimation of the a parameter. For the ion pair under study we obtain a = 4.5 A. Although the primitive model is a good physical approximation in mixed binary solvent systems where D 2 10, it is nevertheless apparent that each particular system usually exhibits some specificity at the non-polar end of the binary solvent mixture. T h i c nerhQnc e v n l Q i n c the A e w i Q t i n n frnm 1;necsAtw in n l n t c nf l n m K c s n c s i n c t 1 /n logK, AS A FUNCTION OF ET The characteristic c. t. absorption band in alkylpyridinium iodides depends largely on the nature of the solvent. In a previous communication1* we discussed how the transition energy (ET) in a particular solvent is a measure of the free energy ofM. PAL AND S.BAGCHI 97 1 solvation (AG) of an ion pair when the entropy change during the c.t. absorption process is neglected. The change in the transition energy, At ET, as one goes from one solvent to another will then be given by where the subscripts g and e denote the ground and excited states, respectively, and A, denotes the change due to transfer from one solvent to the other. Since the excited state is non-polar, AG, will be relatively insensitive towards solvent polarity changes. Thus Now the free energy of transfer from a pure solvent to a mixed solvent of an ion pair is given byI9 (20) where PA and KA are the association constants in pure ethanol and mixed solvent of ethanol and a non-polar aprotic cosolvent, respectively. Using eqn (19) and (20) we may write At ET = At AG,- At AG, (18) A, ET = A, AGg.A, AG(ion pair) = At AG(ions) + RT In ( K i / K A ) (19) RT In KA = RT In PA + A, AG(ions) -At ET = RT In PA + A, AG(ions) + E& - ET (21) where E$ and ET are transition energies in pure ethanol and mixed solvent, respectively. Eqn (21) gives a relationship between In KA and ET. For the At AG(ions) terms we proceed as follows. Abraham19 has determined the free energies of transfer from methanol to any solvent of the ions formed by the dissociation of the ion pair Et,N+I-. We have seen that the values vary linearly with At ET (fig. 4), at least for a series of related solvents, although complications may arise from specific solute- solvent interactions. Alternatively, an estimate of At AG(ions) may be obtained using the extended Debye-Huckel equation.It has been observed that the values estimated by this method also vary linearly with At ET. Hence it is expected that in the case where specific solute-solvent interactions are absent or constant, log KA would wry linearly with ET. In the present system, solute-solvent interactions may be assumed to be constant and we observe a linear plot (fig. 5), particularly when the percentage of non-polar component is not high. CONCLUSIONS The association constants for the ion-pairing process in alkylpyridinium iodides can be determined in a polar solvent (ethanol) by studying the variation of intensity of the c.t. band of the c.i.p. in a binary mixture of the solvent with a non-polar cosolvent. It has been found that interpretation of the data is possible if one assumes that (1) only the polar solvent forms a solvent-shared ion pair, particularly when the percentage of it is high, or (2) the binary solvent mixture behaves ideally so the activity of the components may be replaced by their mole fractions and (3) the activity coefficients of the ion-pair species in a mixed solvent are unity.Further work with other mixed solvents using various alkylpyridinium iodides is in progress. M.P. thanks the University Grants Commission, New Delhi, India for a scholarship.972 U.V .-VIS. DETERMINATION OF ION-ASSOCIATION CONSTANTS T. R. Griffiths and M. C. R. Symons, Mol. Phys., 1960, 3, 90. (a) I. M. Strauss and M. C. R. Symons, J . Chem. SOC., Farada-y Trans. I , 1978,74,2146; 2518; 1977, 73, 1796; (6) M. C. R. Symons, T. A. Shippey and P. Rastogi, J . Chem. SOC., Faraday Trans. I , 1980, 76, 2251. A. Gupta and C. N. R. Rao, J . Phys. Chem., 1973,77, 2888. R. M. Fuoss, J. Phys. Chem., 1978,82, 2427. T . F. Hogen-Esch and J. Smid, J . Am. Chem. SOC., 1966,88, 307. R. A. Mackey and E. J. Poziomek, J . Am. Chem. SOC., 1970, 92, 2432. S. Bagchi and M. Chowdhury, J. Phys. Chem., 1976, 80, 21 1 1 . P. Hemmes, J. N. Costanzo and F. Jordan, J. Phys. Chem., 1978, 82, 387. A. Weissberger, Techniques of Organic Chemistry (Interscience, New York, 1955), vol. 7. lo Solute-Solvent Interactions, ed. J . F . Coetzee and C. D. Ritcher (Marcel1 Dekker, New York, 1969). l1 H. Sadek and R. M . Fuoss, J. Am. Chem. Soc., 1954,76, 5897. l 2 M. Pal and S . Bagchi, Proc. Ind. Acad. Sci., Chem. Sci., in press. l 3 J. D. Simon and K. S. Peters, J . Am. Chem. SOC., 1982, 104, 6142. l4 J. B. Hyne and R. M. Levy, Can. J. Chem., 1962,40, 692. l5 Y. H. Inami, H. K. Bodeneseh and J. B. Ramsay, J. Am. Chem. SOC., 1961,83,4745. l6 R. M. Fuoss, J . Am. Chem. SOC., 1958, 80, 5059. l7 M. Eigen, 2. Phys. Chem., 1954, 1, 176. M. Pal and S. Bagchi, Ind. J . Chem., in press. l9 M. H. Abraham, J. Chem. SOC., Perkin Trans. 2, 1972, 1343. (PAPER 4/922)

ISSN:0300-9599    

DOI:10.1039/F19858100961

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985,81, 973-981 Spreading Pressures for Fatty-acid Crystals at the Air/Water Interface BY MAKIO IWAHASHI, NOBUKO MAEHARA,~ YOSHIHIDE KANEKO~ AND TSUTOMU SEIMIYA Department of Chemistry, Faculty of Science, Tokyo Metropolitan University, Setagaya-Ku, Tokyo 158, Japan AND STEVEN R. MIDDLETON,§ NORMAN R. PALLAST AND BRIAN A. PETHICA*(/ Clarkson University, Potsdam, New York 13676, U.S.A. Received 15th June, 1984 Published values for the spreading pressures of lipids at the air/water interface are inconsistent. As part of a collaborative programme to establish criteria for surface manometry a detailed study was made of the spreading pressures for tetradecanoic, pentadecanoic and hexadecanoic acid crystals prepared from melts and solutions. The spreading pressures vary considerably for a given acid.The removal of solvents from recrystallised acids and the solubilities of chosen crystal samples in hexane were measured. The spreading pressures were then correlated for the various crystal preparations. Optical and scanning electron microscopy were used to characterise crystal geometries, and the crystallographic forms were determined by X-ray diffraction. It was established that solvent retention in the crystals can lower the spreading pressure, and that crystals with microscopic points and edges show high spreading pressures. The results can be interpreted simply by thermodynamic arguments. Data on the effect of temperature on the spreading pressures and solubilities are also presented, and the enthalpies of spreading and solution are calculated.Many lipids spread spontaneously when placed in solid or liquid form at the air/water (A/W) interface to give monolayers in apparent equilibrium with the excess bulk lipid phase. The corresponding monolayer surface pressure (n) is generally known as the equilibrium spreading pressure (e.s.p.). The e.s.p. is, in principle, useful for obtaining thermodynamic information on the crystal-monolayer transition. Unfortu- nately the reported values for a range of lipids are diverse beyond reasonable experimental error, as shown in table 1 for several fatty-acid homologues. We have recently examined in some depth the conditions for reproducible surface manometry using several manometric techniques, identifying experimental artefacts and showing the large effects given by small impurity le~els.~-ll Using the techniques developed it was found that the spreading pressure of a given lipid varies significantly from pure sample to pure sample, whereas the pressures for the liquid-expanded-liquid-con- densed and the liquid-vapour phase transitions for the same samples did not.These various spreading pressures are often the metastable values for non-equilibrium solid t Present address: Toyo Ink Co. Ltd, Aoto, Katsushika-Ku, Tokyo 125, Japan. 3 Present address: Shinetsu Polymer Co. Ltd, Ohmiya, Saitama-Ken 330, Japan. 9 Present address: Pulp and Paper Research Institute of Canada, Point Claire, Montreal, Quebec, T Present address: Standard Oil of Ohio, Warrensville Research Center, 4440 Warrensville Center Road, 11 Present address: Electro-Biology Inc., 300 Fairfield Road, Fairfield, New Jersey 07006, U.S.A.Canada. Cleveland, Ohio 44128, U.S.A. 973974 SPREADING PRESSURES FOR FATTY-ACID CRYSTALS Table 1. Published spreading pressures (in mN m-l) for n-fatty acids on 0.01 mol dmP3 HCl at 25 "C ref. '14 '15 Cl6 Cl, Cl8 c19 c 2 0 c2 2 - - - 14.3 18.2 11.8 4.5 - 18.2 20.5 - - 17.4 22.0 12.5 - - 10.0 7.3 - 3.9 3.4 0.5 - 11.0 - - - 10.7 9.8 15.2 7.6 12.9 6.1 - 16.4 20.4 - 14.6 - 0.4 < 0.1 - - - - - - - - - - - - - - - - - - - < 0.1 __ - forms, and we may identify the true e.s.p. as the lowest spreading pressure observed for large well formed crystals of ultra-pure acids. In this paper we examine the reasons for these variations of the spreading pressures of a given fatty acid by correlating experiments on fatty-acid solubilities in hexane, observations on the microscopic geometries of crystals prepared from melts and solutions and measurements of solvent retention in crystals prepared from solution. Monolayer studies on pentadecanoic and hexadecanoic acids were made on three surface balances at Clarkson.Repeat experiments were made on a fourth balance in Tokyo and the data then extended to include tetradecanoic acid. EXPERIMENTAL MATERIALS Samples of hexadecanoic (C,,), pentadecanoic (C,,) and tetradecanoic (C,,) acids were obtained from Applied Science and NuCheck Prep. Two samples of C,, and one of C,, have been described previously in some detail.g One of these C,, samples, a third sample of C,, and one of C,, from Applied Science were used in the experiments in Tokyo.These later samples were analysed as received on an Iatroscan analyser to give purities of 99.5% (C,,) and 99.7% (C,,), i.e. less pure than the C,, and C,, acids obtained ear lie^.^ All samples as received were in the form of small crystals, presumably recrystallised from a solvent. Samples of all three acids were purified by repeated slow recrystallisation overnight from hexane to give purities better than 99.9% .g Some samples were also recrystallised from acetone. Crystals formed from solution were either allowed to lose solvent by exposure to the air or were pumped out for various times on a vacuum line. Large crystals were made by slow recrystallisation from solution or by cooling slowly from a melt.9 Some large crystals were reduced to a fine powder by gentle grinding in an agate mortar.Spreading pressures were measured with samples as received or after purification. The purified samples were examined as large crystals or as small crystals obtained by ordinarily quick cooling of solutions in hexane, pouring an acetone solution into water or by grinding large crystals. Crystals were examined by optical microscopy and by scanning electron microscopy on a JEOL instrument, model JSM-T3000. The crystallographic form was determined for the C,, acid with a Rigaku Denki (model 2001) X-ray diffraction instrument in Tokyo, and for the C,, acid by flat-film and Guinier camera methods by Dr J. D. Oliver of Procter and Gamble in Cincinnati. Purification of the water (surface tension 72.00 mN mP1 at 25 "C) and of the hydrochloric acid has been described ear lie^,^ l1 as have the techniques to clean the silica dishes by roasting.Glass equipment (including a Pyrex surface-balance dish in some later experiments) was cleaned by chromic acid followed by washing and steaming with pure water. All monolayers were studied at the surface of 0.01 mol dmP3 HC1 solutions. The three surface balances used at Clarkson have all been described earlie~,~-ll emphasizingM. IWAHASHI et al. 975 the importance of temperature and humidity controls. A fourth balance constructed in Tokyo has been reported separately.12 This balance used strain gauges for measuring the surface pressures. Note that the isotherms given in ref. (12) for pentadecanoic acid to illustrate the use of the Tokyo balance show a degenerate liquid-expanded-liquid-condensed phase transition, as distinct from the simple first-order transitions found by critical monolayer method^.^^ lo The curves of ref.(12) were obtained by continuous compression, thus illustrating the artefacts which this method can entail, as pointed out ear lie^.^^ lo Spreading pressures were measured with glass or paper Wilhelmy plates, or by the horizontal- force method using a float or the single-thread configuration with a central sensor rod?, lo Crystals (or a single crystal) were placed on the clean surface and the surface pressure recorded until constant to the sensitivity of the balances (0.01 mN m-l), requiring many hours for a single crystal and correspondingly less time when a number of small crystals were in the interface.The film was then either compressed or expanded and the return of II to the previous steady value checked. Alternatively, more solid was added or some crystals removed by suction and the constancy of Il re-examined. With purified samples ll did not vary with the number of crystals. With the as-received materials there was a measurable increase in ll on adding a large excess of solid, presumably owing to the small amounts of residual impurities. The solubilities of recrystallised smallcrystal fatty-acid samples in hexane were determined at temperatures (constant to kO.02 "C) over the range 20-25 "C by the asymptotic method,13 following changes in concentration for up to three days with solutions initially either under- or super-saturated.Samples were withdrawn through pre-extracted filter paper and aliquots evaporated and weighed.l49 l5 The solubility of large crystals was obtained by contacting a crystal with a small volume of a solution saturated with a sample of small crystals. Up to seven days were required to establish equilibrium. RESULTS AND DISCUSSION For a given sample, essentially identical results were obtained for the spreading pressures on each of the balances used, with agreement for paper and glass Wilhelmy plates and for floats or single threads in the horizontal configuration. The only exceptions to this broadly based identity of results for a given sample were that the spreading pressures for the as-received samples were a few percent higher if a large excess of crystals was placed on the surface, and that the horizontal balance intermittently gave values ca.0.1 mN m-l lower than those from the Wilhelmy method. The first exception we ascribe to the effect of the low level of impurities in the as-received samples. The second we ascribe to the difficulty of completely removing fatty-acid contamination from the 'clean' side of the float or thread. Very small amounts of these acids are required to bring the 'clean' side to the pressure of the liquid-vapour monolayer transition. In this state, expansion or compression of the 'clean' side will show no pressure change. Between samples, however, the spreading pressures vary considerably. The results fall into three groups: (i) small crystals free of solvent, (ii) large crystals from the melt or from solution but free of solvent and (iii) crystals containing residual solvent.The class of small crystals includes the as-received samples and purified specimens prepared in small-crystal form by rapid precipitation from hexane solution or by grinding large crystals. Small crystals lose residual solvent rapidly, and changes in spreading pressure after pumping were rarely observed. Large crystals prepared from solution lose solvent more slowly, and a series of spreading pressures can be observed as these crystals lose solvent on pumping or being allowed to dry in air. Since the spreading pressures are the same within each of the three groups irrespective of whether the samples were as-received or further purified, it appears that impurities at the low levels present do not influence our results to the accuracy claimed, with the exception of the as-received samples used in a large excess, as noted above.The data for pentadecanoic acid are the most extensive, and illustrate the essential976 SPREADING PRESSURES FOR FATTY -ACID CRYSTALS features of the varying spreading pressures. The results for small crystals free of solvent, obtained by recrystallisation or by grinding, all cluster into the range 19.6-19.8 mN m-l at 25 O C , with the exception of a very fine sample prepared by pouring an acetone solution into water, which gave a result at 19.9 mN m-l. The small-crystal result for pentadecanoic acid at 19.7 mN m-l is thus a useful practical pressure ~tandard.~ Large crystals all show lower spreading pressures than the small-crystal group.The large crystals are flat plates up to a few mm in size. Those prepared from the melt or from solution with subsequent pumping for several days all gave a spreading pressure of 18.6-18.7 mN m-l, which we take as the equilibrium spreading pressure for pentadecanoic acid. Large crystals freshly prepared from solution gave spreading pressures as low as 17.0 mN m-l, the values increasing steadily on protracted pumping. One fresh pentadecanoic acid crystal was weighed after brief air drying, and its spreading pressure was determined to be 17.0 mN m-l. The crystal was recovered and pumped on the vacuum line for five days, losing 2.8 % of its initial mass. The spreading pressure was redetermined and shown to be 18.7 mN m-l.The effect of retained solvent on the spreading pressure of large crystals of pentadecanoic acid can be estimated by assuming ideal mixing in the solid state and applying the Gibbs adsorption isotherm. Denoting the chemical potential of pentadecanoic acid in the monolayer and crystal as pm and pc, the spreading pressure in the coexisting monolayer as II and the surface density of pentadecanoic acid in the monolayer as r we have: dII = Tdu, (1) and dp, = RTd lnx, (2) where R is the gas constant, T the temperature and x, the mole fraction of acid in the crystal. We take r from recently published i s o t h e r m ~ ~ ' ~ ~ to be 7.91 x mol cm-2 at 18.7 mN m-l and 7.83 x 10-lo mol cm-2 at 17.0 mN m-l. Noting that in the fresh crystal the mole fraction of the fatty acid was 0.926, eqn (1) is readily integrated to give the difference in the spreading pressure between the fresh and fully dry crystals at 1.42 mN m-l, as compared with the observed difference of 1.7 mN m-l.The calculation assumes that the solvent is not retained in the monolayer and ignores solvent concentration gradients in the crystal and related crystal imperfections. These imperfections would also contribute to a raised spreading pressure and probably anneal out with time to give a final spreading pressure in agreement with that found for large crystals from the melt. It is therefore likely that the solvent retention in the crystals is greater than that implied by ideal mixing. The difference in spreading pressure between the large and small crystal groups is ca.1 mN m-l. Fig. 1 shows the results for pentadecanoic acid and tetradecanoic acids for a range of temperatures. Both large and small crystals were shown to be in the stable A' crystallographic form as expected,l6? l7 ruling out an explanation based on polymorphism. The gross dimensions of the small crystals (1&300 pm) are too large to expect the pressure changes to be due to the Kelvin effect. Mansfieldl* and BrooksL9 reported a spreading pressure for finely powdered octadecanol at 25 "C of 41.2 mN m-l. Roylance and Jones20 gave 39.9 mN m-l for large octadecanol crystals grown from acetone solutions. Mansfield suggested that the ' spreading pressure' of a crystal edge is less than that of a randomly oriented powder, noting that small random crystals could contact water at various faces.If Mansfield's suggestion were correct the solubility of small random crystals in a solvent would be the same as for a large crystal, since the solvent contacts the crystals at all faces and the gross crystalM. IWAHASHI et al. 977 Ti K Fig. 1. TIK Fig. 2. Fig. 1. Spreading pressures for pentadecanoic acid (A) and tetradecanoic acid (0) crystals at various temperatures. Open symbols represent crystals from solution, filled symbols represent large crystals. Fig. 2. Solubilities for small (open symbols) and large (filled symbols) crystals of tetradecanoic acid in hexane at various temperatures. The diamond is a value reported in ref. (24). The dashed line shows the calculated value for the solubilities of the large crystals.sizes are too big for the Kelvin effect to give a significant change. However, fig. 2 and 3 show that the solubility of small crystals of both pentadecanoic and tetradecanoic acids is greater than for large crystals. Both spreading pressure and solubility are larger for small crystals, inviting a common explanation. The correspondance between solubility and spreading pressure for the small and large crystals is readily established. Variations in chemical potential of a fatty acid in hexane solution may be represented by dp, = RTd lnx, (3) where x, is the mole fraction of fatty acid in the solution. Neglecting the mutual solubilities of water and the fatty acids, the terms dp in eqn (1) and (3) may be equated and the change in solubility predicted from the change in spreading pressure if the surface densities are known.The relevant density values for pentadecanoic acid are those quoted above. The values for tetradecanoic acid may be taken from the approximate FI against area ( A ) isotherms given in fig. 4 for several temperatures. These isotherms were obtained by continuous compression and extend well above the equilibrium spreading pressures for tetradecanoic acid. They are therefore strictly non-equilibrium isotherms. Note that the so-called liquid-expanded-liquid-con- densed phase transitions all occur at pressures above the spreading pressures at a given temperature. With these reservations we use the isotherms up to the known spreading pressures to integrate eqn (1) and calculate the solubilities of large tetra-978 SPREADING PRESSURES FOR FATTY-ACID CRYSTALS p / I I 1 I I I 293 294 295 296 297 298 T/ K Fig.3. Solubilities for small (open symbols) and large (filled symbols) crystals of pentadecanoic acid in hexane at various temperatures. The diamond symbol is a value reported in ref. (24). The dashed line shows the calculated values for solubilities of the large crystals. molecular area/mm2 molecule-' Fig. 4. ll against A isotherms of tetradecanoic acid on 0.01 mol dmP3 HC1 at various temperatures. II, and IIL denote the spreading pressures for small and large crystals at 25 "C, respectively. The isotherms were obtained by continuous compression. The liquid- expanded-liquid-condensed transitions occur above the spreading pressures and are non- equilibrium artefacts.The degenerate appearance of this transition is an artefact of the continuous-compression m e t h ~ d . ~J. Chem. SOC., Faraday Trans. 1, Vol. 81, part 4 Plate I Plate 1. (a) Scanning electron micrograph of the end of a small crystal of tetradecanoic acid from solution ( x 500). (6) Scanning electron micrograph of the end of a rod-like portion of a small crystal of tetradecanoic acid from solution ( x 5000). M. IWAHASHI et al. (Facing p . 978)J. Chem. Soc., Faraday Trans. I , Vol. 81, part 4 Plates 2 and 3 Plate 2. Scanning electron micrograph of the corner of a large crystal of tetradecanoic acid ( x 700). Plate 3. Scanning electron micrograph of a large crystal of tetradecanoic acid gently ground in an agate mortar ( x 700).M. IWAHASHI ef al.M. IWAHASHI et a[. 979 Table 2. Spreading pressures (in mN m-l) for n-fatty acids on 0.01 mol dm-3 HC1 tetradecanoic pent adecanoic T/"C large small large small large hexadecanoic - 20.0 11.1 11.8 16.2 16.9 22.5 12.1 13.3 - 18.5 - - 17.6 18.7 22.8 25.0 13.8 14.7 18.7 19.7 9.3 27.5 15.5 16.4 20.1 21 .o 30.0 - - - - - - - 11.5 decanoic crystals from the data on small crystals. The results of these calculations for both pentadecanoic and tetradecanoic acids are shown in fig. 2 and 3. The agreement with the experimental results with the large crystals is striking for both acids. It remains to account for the greater values of both the spreading pressure and the solubility in hexane for small fatty-acid crystals. Plate 1 (a) shows a scanning electron micrograph of a corner of a small crystal from hexane solution. There are numerous needle-like rods with sharp ends.By contrast, plate 2 shows that the edge of a large crystal does not exhibit these fine-pointed structures. When a large crystal was ground, the powder particles showed numerous irregular sharp edges rather than needle-like structures (plate 3). In addition, the optical clarity of the large crystals is lost on grinding, suggesting dislocations in the resulting powdered solids. An electron micrograph at higher magnification of a needle on a small crystal from solution is given in plate l ( b ) , showing a tip radius of ca. 0.2pm. This dimension is in a range to give a significant Kelvin effect on both solubility and spreading pressure.Kelvin effects on the solubility of inorganic salts are known for submicron particles.21$ 22 In mixed-crystal powders the evidence is that the finest particles determine the solubility, at least temp~rarily,~~ and it is reasonable to associate the fine structures shown in plate 1 with the raised solubility and spreading pressures. Assuming that the relation for the Kelvin effect for small spheres may be applied approximately to the needle tips of the small crystals, and taking 302.5 cm3 mol-l as the molar volume for pentadecanoic and using the data of fig. 3 at 25 "C, the calculated interfacial tension of the acid against hexane is 29 mN m-l. Using eqn (l), the corresponding interfacial tension of the fatty acid against water is calculated as 39 mN m-l.Both results are reasonable, the fatty-acid/hexane interface having the lower energy as expected. The agreement in spreading pressures for small crystals from solution and from grinding large crystals suggests that the microscopic dimensions in both cases are similar. This may be coincidental or the result of there being a fairly narrow size range over which the edge or point energies rise rapidly. The spreading pressure of hexadecanoic acid was studied less extensively. Large crystals prepared by slow crystallisation from hexane were examined and gave equilibrium spreading pressures of 1 1.4 and 9.3 mN m-l at 30 and 25 "C, respectively. Our accumulated results for the spreading pressures of the three fatty acids are shown in table 2. The values at 25 "C are based on more numerous experiments and are considered accurate to fO.l mN m-l.The results at the other temperatures are less extensively supported and are probably accurate to better thank 0.2 mN m-l. The solubility results of fig. 2 and 3 were interpreted to give the enthalpies of980 SPREADING PRESSURES FOR FATTY-ACID CRYSTALS solution using the Clausius-Clapeyron equation. The solution enthalpies for the pentadecanoic and tetradecanoic acids could be distinguished at 95 and 88 kJ mol-l, respectively, to an accuracy of ca. +2 kJ mol-l from the small-crystal data, the large-crystal results being less accurate. These values may be compared with the 113 and 96 kJ mol-1 obtained from the data of Hoerr and H a r w ~ o d , ~ ~ which give rather curved logarithmic plots against reciprocal temperature and show the C,, and C,, solubilities crossing over.The variation of spreading pressure with temperature was used to obtain the enthalpies of spreading from the crystal, using the Clausius equation in the form2, where N is Avogadro's number and the subscript e indicates that the A and l-I values refer to the e.s.p. for large crystals at 25 "C. On this basis AH becomes 32.7, 17.2 and 14.5 kJ mol-1 for the C14, C,, and c16 acids, respectively, with an accuracy of ca. 5% for the C,, and C,, acids and an uncertain error for the C,, acid since the calculation is based on results at two temperatures only. In interpreting these heats of spreading, note that the large AH for the C14 acid reflects the large A at the spreading pressure (32.7 A2 molecule-') as against the smaller values of 21 and 20 A2 molecule-l taken for the C,, and c16 acids.The main conclusions of this study are that impurities, inadequate technique and differences in microscopic crystal geometries are the main causes of the diversity of the published data on spreading pressures. If solvent retention in the solid phase is avoided, the spreading pressures of the fatty acids are reproducible, but with different values for small and large crystals. These differences relate to the presence of sub-micrometre edges or needle-like structures found with crystals prepared by rapid precipitation and grinding. The spreading pressures for large smooth crystals may be taken as the equilibrium spreading pressures. We thank Drs R. G. Laughlin and J.Oliver of the Procter & Gamble Co. for X-ray diffraction analysis of the fatty-acid samples and Mrs M. Muto and Y. Ono of JEOL Ltd for scanning electron microscope measurements. This work was partly supported by grants from the U.S. National Science Foundation (CHE782 7566) and the U.S. National Institute of Health (ROIGM240 68) and partly by Clarkson University, and also by a Scientific Research Grant from the Japanese Ministry of Education. ' A. Cary and E. K. Rideal, Proc. R . SOC. London, Ser. A, 1925, 109, 318. W. D. Harkins and G. C. Nutting, J. Am. Chem. Soc., 1939. 61, 1702. G. E. Boyd, J. Phys. Chem., 1958, 62, 536. R. E. Heikkila, C. Kwong and D. G. Cornwell, J. Lipid Res., 1970, 11, 190. N. L. Gershfeld and R. E. Pagano, J. Phys. Chem., 1972, 76, 1224. B. Sims and G. Zografi, J. Colloid Interface Sci., 1972, 41, 35. K. Motomura, S. Yoshino, K. Fujii and R. Matuura, J . Colloid Interface Sci., 1977, 60, 87. S. R. Middleton, M. Iwahashi, N. R. Pallas and B. A. Pethica, Proc. R . Soc. London, Ser. A, in press. N. R. Pallas and B. A. Pethica, Colloids S u r - , 1983, 6, 221. " A. K. Rakshit, G. Zografi, I . M. Jalal and F. D. Gunstone, J. Colloid Interface Sci., 1981, 80, 466. I " S. R. Middleton and B. A. Pethica, Furaday Symp. Chem. Soc., 1981, 16, 109. l 2 M. Iwahashi, S. R. Middleton, T. Seimiya and B. A. Pethica, Bull. Chem. Soc. Jpn, 1983, 56, 2525. l 3 C. J. McGinn, J. Phys. Chem., 1961, 65, 1896. l 4 M. Muramatsu, M. Iwahashi and K . Masumoto, J. Chem. Eng. Data, 1975, 20, 6. l5 M. Iwahashi, Y. Watanabe, T. Watanabe and M. Muramatsu, Bull Chem. SOC. Jpn, 1984,57, in press. l6 E. Van Sydow, Acta Chem. Scand., 1955, 9, 11 19. E. Van Sydow, Acta Chem. Scand., 1955, 9, 1685.M. IWAHASHI et al. 98 1 L. W. Mansfield, Aust. J . Chem., 1983, 16, 76. J. M. Brooks, Retardation of Evaporation by Monolayers, ed. V. K. LaMer (Academic Press, New York, 1962), p. 259. 2o A. Roylance and T. G. Jones, J. Appl. Chem., 1961, 11, 329. M. L. Dundon and E. Mack, J. Am. Chem. Soc., 1923,45, 2479. 22 M. L. Dundon, J. Am. Chem. Soc., 1923, 45, 2658. 23 B. V. Enustun and J. Turkevich, J. Am. Chem. SOC., 1960, 82, 4502. 24 C. W. Hoerr and H. J. Harwood, J. Org. Chem., 1951, 16, 779. 25 A. E. Alexander and F. C. Goodrich, J . Colloid Sci., 1964, 19, 468. (PAPER 4/ 10 1 8)

ISSN:0300-9599    

DOI:10.1039/F19858100973

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. 1, 1985,81, 983-990 Potentiometric Studies of Some Lanthanum Carboxylates at Constant Ionic Strength BY PAUZI B. ABDULLAH AND CECIL B. MONK* Edward Davies Chemical Laboratories, University College of Wales, Aberystwyth, Dyfed SY23 1NE Received 15th June, 1984 Overall stability constants Ppqr [where p is the number of lanthanum(rI1) atoms, q is the number of hydrogen ions and r is the number of ligands] have been determined at 25 "C for lanthanum glycollate, malate, maleate and tartrate at an ionic strength ( I ) of 2.0 mol dmd3 and for lanthanum malonate at I = 1.0, together with those of the corresponding acids. The technique consisted of e.m.f. measurements to & 0.01 mV of a cell containing glass and silver, silver chloride electrodes and calibration by reference to the acid stability constants.The lanthanum D values were computed by two methods and the answers are discussed in terms of the number of complexes formed with each ligand and the factors which influence the calculations. A previous paper1 reported potentiometric determinations of the overall stability constants Plor of some lanthamide glycollates at I = 0.5 and 25 "C. The data were sufficient only for evaluation of Plor values with Y = 1-3, whereas analyses by Sonesson2 of his own potentiometric work at Z = 2.0 and 20 "C extend to Y = 5 . Others3 have placed the limit at r = 3. For the present work we increased the ligand concentration range beyond that used previously for lanthanum glycollatel to see if r(max) could be decided.In addition we report a limited study of some lanthanum dicarboxylates. Most previous studies on these3 indicate the formation of LaL and LaL, (L = dicarboxylate) but4 for lanthanum malate and thiomalate LaL, may also be formed. Also, Dellien and Grenthe,5 from analyses of their potentiometric data for all the lanthanide malonates, derived 8103 values for the series dysprosium to lutetium but found Plo2 to be the upper limit for the lanthanum to terbium group. The analysis can be complicated by the formation of other species such as LnHL, LnHL, and NaL (in NaClO, media). This is illustrated by the work of Dellien and Grenthe5 and by that of Dunsmore and Midgley6 on the lanthanum tartrates. Our selection of dicarboxylates was made on the basis of the above observations.EXPERIMENTAL The preparation of sodium glycollate, sodium perchlorate, lanthanum perchlorate and hydrochloric acid stock solutions has been described previously.' Stock solutions of dicarboxylic acid and of sodium carbonate were made up by weight from the dried salts. E.m.f. measurements were made to kO.01 mV at 25 "C with the digital-voltmeter system described e l ~ e w h e r e . ~ ~ ~ Each silver, silver chloride electrode was made from 5 cm of 1.63 mm diameter silver wire soldered to copper wire and sealed by paraffin wax into a glass tube. It was etched in dilute nitric acid and a silver chloride coat formed by electrolysis for ca. 1 min in 0.1 mol dm-3 hydrochloric acid. This coat was renewed for each run. Details of the cell and 983984 POTENTIOMETRIC STUDIES OF LANTHANUM CARBOXYLATES techniques for additions of solutions were as described previo~sly.~ After the initial equilibration of the first solution for cell calibration, which took several hours, ensuing additions of stock buffers gave steady readings in ca.2 min. RESULTS AND DISCUSSION LANTHANUM GLYCOLLATE The cell and equations used for determining poll of the acid were as follows: glass electrode I NaL( C,), H[ClO,](C,), HCl(C3), Na[ClO,] to I = 2.0 mol dm-3 I AgCl, Ag electrode E/mV = E" - 59.158 log,, [(C2 + C3)C3] E/mV = E" - 59.158 log,, ([H+]C3) B o l l = (C, + c3 - [H+I)/{[H+I(C,- c, - c3 + [H+I)) (A) (1) ( 2 ) (3) where Ci are the concentrations of the various species in mol dmW3. IT, the standard cell potential, was calculated from eqn (1) and the e.m.f.when the cell contained only HCl + Na[ClO,]. poll values were calculated from eqn (2) and (3) and the e.m.f. values obtained by adding portions of stock buffer solution which contained sufficient Na[ClO,] to keep I = 2.0. The average of 2 runs is poll = 4130 & 30 mol-1 dm3. Sonesson,2 who used gold, quinhydrone electrodes found systematic trends in poll with buffer concentrations. Such trends were not found with cell (A). PIOr values of lanthanum glycollate were derived from e.m.f. values of the cell glass electrode I NaL(C,), H[ClO,](C,), NaCl(C,), La(ClO,),(C,), Na[ClO,] to I = 2.0 I AgCl, Ag electrode (B) and E/mV = E" - 59.158 log,, ([H+]C,) E" was calculated from the e.m.f. of an initial addition of stock buffer [no La(ClO,),] and eqn (4)-(6).One addition of stock La(ClO,), was then made followed by a series of additions of stock buffer. Values of A were calculated from the resulting e.m.f. values and eqn (4)-(9). plor values were calculated by Rydberg's methodLo (RYDBERG program) and by that of Alcock and Rogers1l7 l2 (DALSFEK program). The latter does not involve ii data and is based on a damped non-linear least-squares method which closely follows Marquardt's algorithm13 for obtaining the optimum parameter corrections for a set of iterations. DALSFEK required inputs of E = 59.158 log,, ([H+]T/[H+]A) (10) where [H+], is the total available [H+], i.e. C, of cell (B), and [H+IA is the true value of [H+], i.e. that given by eqn (4). There is an omission in ref. (12); the first E valueP.B. ABDULLAH AND C. B. MONK Table 1. Example run for lanthanum glycollate at 25 "C and I = 2.0 mol dm-3 (concentrations in mmol drn-,, e.m.f. values in mV)a 985 14.22 18.41 27.38 34.55 44.33 54.08 66.84 79.32 91.59 103.84 I 15.25 126.4 140.4 154.3 167.5 180.2 192.9 205.3 2 19.9 234.0 260.9 273.8 290.9 309.1 326.7 343.5 7.09 9.17 13.64 17.21 22.09 26.94 33.30 39.51 45.63 51.73 57.41 62.96 69.94 76.85 83.46 89.75 96.09 102.3 109.6 116.6 130.0 136.4 144.9 154.0 136.4 144.9 46.84 46.66 46.34 46.04 45.66 45.28 44.81 44.3 1 43.84 43.37 42.93 42.51 41.95 41.42 40.92 40.53 39.93 39.45 38.90 38.34 37.3 1 36.8 1 36.16 35.46 34.78 34.13 15.93 15.87 15.76 15.66 15.53 15.40 15.24 15.07 14.9 1 14.75 14.60 14.46 14.27 14.09 13.92 13.75 13.58 13.42 13.23 13.04 12.69 12.52 12.30 12.06 11.83 11.61 114.62 1 15.03 114.86 114.27 1 13.23 112.06 110.53 109.06 107.67 106.38 105.22 104.18 102.96 101.81 100.78 99.87 98.99 98.17 97.24 96.41 94.89 94.20 93.31 92.42 91.55 90.75 a For E" determination, C, = 7.919 x C, = 3.940 x C, = 5.320 x lo-,, E = -92.01, E" = -384.24. E(corrected) = E(obs)+3.5C1.should follow instruction (1 1). Also if more than 32 points are handled, each of 3 reactant concentrations, the dimensions of the arrays P, NPAR and NCODE must be increased accordingly from the original values of 100. Inputs of approximate PlO7 values and of the concentrations of all species present in the first mixture are also needed. Such data were taken from or estimated from the RYDBERG output. With RYDBERG it was found, in line with one of Sonesson's observations,2 that making small corrections to the e.m.f.values obtained minimised the percentage standard deviations of the calculated PIOr values. These corrections can be attributed to factors such as small changes in I as CL increases due to increasing concentrations of complexes. Such corrections were made with E(corrected)/mV = E(obs) +fCl (1 1) where C, is defined for cell (B) andfis a small adjustable empirical parameter. The corrected e.m.f. values found with RYDBERG were also used when running DALSFEK. Our A(max) values were ca. 3.2. These are lower than n(max) x 4 attained by Sonesson,2 but excessive ligand concentrations are needed to exceed A = 3. An example run is shown in table 1 and the resulting PIOr values obtained on combining 2 runs in table 2.33 FAR 1986 POTENTIOMETRIC STUDIES OF LANTHANUM CARBOXYLATES Table 2. Summary of calculated overall stability constants Ppqr at 25 "C glycollate ( I = 2.0 mol dm-3, two combined runs, 56 points, A = 0.15-3.18) method PI01 10-3 plO2 P103 P104 RYDBERG 133f3 6.55k0.17 7.4f0.9 8.3k 1.0 DALSFEK 137f1.3 6.14f0.09 8.0f0.2 8.3f0.25 Sonesson (20 "C) 155+8 5.7 & 0.4 6.5 f 0.6 1.2 k 0.3 malate ( I = 2.0, two combined runs, 33 points, A = 0.42-1.87) ~~~ ~ method PlOl PIo2 PI03 RYDBERG 3490k30 9.4kO.11 9.1f1.1 DALSFEK 3400 f 60 9.5 f 0.14 7.1 f 0.4 maleate ( I = 2.0, two combined runs, 37 points, A = 0.18-1.28) method PlOl Dl02 RYDBERG 364 & 8 7.8 f 0.3 DALSFEK 363 k 9 7.9 f 0.7 tartrate ( I = 2.0, one run, 16 points, A = 0.13-0.92) method D l O l P l 0 2 Plll RYDBERG 900 f 30 29.5 & 0.9 0 DALSFEK 896 k 7 29.9 & 1.5 0 ref.(6), I = 0.4 1280f 10 36.2k0.1 6.6k0.12 ref. (6) with ref. (6) with RYDBERG 1050 _+ 60 43.2 & 1.2 4.0" DALSFEK 1090+50 39.1 k0.5 4.5f0.9b malonate ( I = 1 .O, one run, 21 points, A = 0.1 1-1.11) method DlOl plOz All RYDBERG RY DBERCi DALSFEK DALSFEK ref. (9, runs 3 and 4 ref. (9, runs 3 and 4 ref. (3, runs 3 and 4 ref. (5), runs 3 and 4 RYDBERG RYDBERG DALSFJiK DALSFEK 740f 10 780f 10 720 f 50 770 k 20 2.66 + 0.15 3.27 f 0.15 2.69 f 0.50 3.27f0.19 0 5.0" 0 5.0" 990+ 10 1.05 k 0.07 0 1150k 10 1.14+ 0.08 2oa 990& 10 1.05 f 0.02 0 1150f 10 1.17 & 0.08 20" a Treated as a fixed parameter. Treated as a variable parameter.P. B. ABDULLAH AND C. B. MONK 987 LANTHANUM MALATE Boll and pOzl values for malic acid were estimated from e.m.f.measurements of the cell glass electrode I H,L( Cl), Na,[CO,]( Cz), HCl( C3), Na[ClO,] to I = 2.0 I AgCl, Ag electrode E/mV = E" - 59.158 log,, ([H+]C3) (C) (12) Po11 = [HL-l/([H+I[L2-I) (13) Do21 = [H2L1/([H+12[L2-1)* (14) and the equations The first e.m.f., using HCl +Na[ClO,] alone, gave E" [eqn (12)]. It was found better to calculate these p values by varying pOzl until poll showed no trend with increasing buffer additions rather than using a least-squares method. In this way one can monitor the range of buffer concentrations and find reasonable values of [HL-]/[L2-] and [H,L]/[L2-] for the following calculations. For each buffer addition, the e.m.f. and eqn (12) give [H+]. Then assuming [H,L] = 0 for the first cycle one has [HL-] = 2(C1 - C,) + C3 - 2[H2L] - [H+] [H,LI = [H+1{2(C1- C,) + c3 - [HL-l}/V;!A + 2CH+1) [L2-] = Cl - [H2L] - [HL-I.(15) (16) (17) An estimated first value of pOzl was used in eqn (1 6) and calculations continued until [HL-] was constant to within 1 x mol dm-3. The process was repeated until the selected value of /Iozl produced the minimum average deviation in the obtained value of Doll. From two runs we get poll = 3.37f0.13 x lo4 mol-l dm3 /3021 = 5.77k0.42 x lo7 moF2 dm6. /? values of lanthanum malate were calculated from e.m.f. values and concentrations of the cell glass electrode I H,L( C1),Na,[C03]( C,), HCl( C3), La(CIO,),(C,), NaCl(C3), Na[ClO,] to I = 2.0 I AgC1, Ag electrode. (D) The initial solution contained a low concentration of buffer but no La(ClO,),, and E" was calculated from the resulting e.m.f., starting with a first estimate of E" to obtain a first value of [H+] from eqn (12).Eqn (15) was used to obtain a first estimate of [HL-] assuming [H,L] = 0. [H,L] was obtained from [H&] = (2C2 - Cl - C3 + [H+])/@G,~[H+]~ - 1) (18) and the process continued until [H2L] was constant to within 1 x lop7. Then using eqn (17) to obtain [L2-], Doll was calculated. The estimated E' was varied until pol,(calc.) agreed with the above determined value to within 50 mol-1 dm3. From the e.m.f. values given by one addition of La(ClO,), and a series of stock buffer, together with Rl = D02l[H+I/P0ll (19) (20) [HL-1 = (C1- 2C2 + C3 - [H+])/( 1 + 2R1) 33-2988 POTENTIOMETRIC STUDIES OF LANTHANUM CARBOXYLATES CL = Cl - [HL-] - [H,L] n values are obtained from eqn (9).The fi(max) values approached 2, which suggests that LaL, formation is possible. This was supported by finding that the standard- deviation values of polo and /Ilo2 were reduced on inclusion of pi03 and by the work of Cefola4 at I = 0.1. Results from RYDBERG and DALSFEK are shown in table 2. LANTHANUM MALEATE The /? values of the acid are sufficiently different for separate determinations. Cell (C) was used for poll and for pOzl with C, = 0. Our results are poll = 4.15 0.01 x lo5 mol-1 dm3 pOz1 = I .74 & 0.09 x lo7 mo1k2 dm6. and Determinations of polo and PlO2 for lanthanum maleate are shown in table 2. LANTHANUM TARTRATE Dunsmore and Midgley6 have reported studies of lanthanum tartrate at 25 "C, at both varying and constant ionic strengths.They used tetramethyl ammonium chloride to control I since they found14 that association between the ions Na+ and tartrate2- occurs. Their cell comprised glass and silver, silver chloride electrodes in separate compartments linked by a salt bridge. Their analyses were based on the formation of LaL, LaL, and LaHL (where L = tartrate2-), and after preliminary estimates of the corresponding /? values these were refined simultaneously by the GAUSS program.15 The published data6 of the 3 runs at I = 0.4 (apart from the first point of the first run) have been combined and recalculated by the present methods. The RYDBERG program was modified to run with selected fixed values of Pill since the concentrations of LaHL have a pervading influence. Eqn (1 7) now becomes [L2-] = C , - [H,L] - [HL-] - [LaHL] and for eqn (9) C , is replaced by C,-[LaHL].The standard deviation in plo2 was minimised at pill = 4 x lo5 [plOl remained constant with pill = (3-5) x lo5]. The DALSFEK program was modified so that Dill could be treated and calculated as a variable parameter. This was successful and the results of the above calculations are shown in table 2. For the present work at I = 2.0, poll and pOzl of tartaric acid were determined in the manner described for malic acid. Our results are poll = 4.66f0.05 and pOzl = (3.49f0.69) x lo4. Na[ClO,] media were used on the assumption that errors due to association between Na+ and tartrate2- cancel out. Also, since an extrapolation of the results of Dunsmore and Midgley6 at I = 0.1, 0.2 and 0.4 against I indicates that pill is only ca.1000 at I = 2.0, this parameter was not considered for the results shown in table 2. LANTHANUM MALONATE Dellien and Grenthe5 determined stability constants of the lanthanide malonates at I = 1 .O (Na[ClO,]) and 25 "C. Their cell consisted of glass and silver, silver chloride electrodes in separate compartments linked by a salt bridge. /3 values were calculated by a least-squares program of the seriesL6 LETAGROPVRID assuming the complexes to be LnL, LnL,, LnL,, LnHL and LnHL, [where Ln = lanthanide(m)]. Experimental difficulties precluded determinations of 8103 for LnL, for the lanthanum to terbiumP. B. ABDULLAH AND C. B. MONK 989 series. Since it proved difficult to find suitable conditions for work at I = 2.0 (owing to hydrolysis and limited solubility) and also worthwhile to compare the present procedures with those of the above authors, we also used Na[ClO,] media at I = 1.0.The p values of the acid were determined as described for malic acid. The results are and poll = 1.15f0.04 x lo5 pOzl = 5.35f0.06 x lo7. The first agrees with (1.155 0.010) x lo5 given by Dellien and Grenthe5 but the second is much higher than their result of (4.49k0.08) x lo7. They used LETAGROPVRID, but since they gave no experimental data the difference may be due to the different methods of calculation and our previous comment concerning malic acid. They gave about one-half of their experimental data on lanthanum malonate (4 runs). We combined runs 3 and 4, since the reactant concentrations are of the same order although A(max) is only ca.0.49. With RYDBERG and various fixed values of pill the standard deviations of BlOl and plo2 scarcely change, but the actual value of plol increases as pill increases while that of ploz changes little. Similar effects were found on running DALSFEK. Examples are given in table 2. Some conclusions which can be drawn from the present work are as follows. (a) At the maximum values of CL used for lanthanum glycollate, the slow increase in n with CL confines the maximum value of plor that can be calculated to plOs. Sonesson2 reached much the same conclusion but also gave a rough estimate of p105. La111 may have 6 coordination sites, but, as Sonesson pointed out, glycollate may act as a chelating ligand and this obscures the issue. Allowing for temperature difference and methods of calculation, the present results and those of Sonesson (who used a graphical procedure) show no major differences except for QlO4 (see table 2).Our figures for this are much the larger. A possible reason for this is that at the highest CL values the e.m.f. values upon which Sonesson's calculations are based were small and subject to marked percentage corrections. (6) The present work with lanthanum maleate confirms the interpretation of Cefola4 that up to 3 bis-chelate ligands can coordinate with La1'*, i.e. a coordination number of 6 is attained. Maleate associates strongly3 with many metal ions so that relatively low C,> concentrations are needed to raise A to between 2 and 3.This may explain in part why the maximum PIOr values found with the other dicarboxylates used in the present work are plO2. These ligands associate less strongly with LaIII than does maleate, so a situation is reached similar to that found with glycollate, namely that beyond a certain value of C,, increases in A are small. ( c ) The studies of Dunsmore and Midgley6 and of Dellien and Grenthe5 show that due account should be taken of species such as LaHL and NaL. The calculation of p,,, may not be straightforward if pill is small or only small concentrations of LaHL are generated. In such cases it may be helpful to test the programs by treating pill as a fixed parameter and to examine the outcome as described earlier on for tartrate and malonate.(d) The Blor values and the standard deviations calculated from our own measure- ments by RYDBERG and by DALSFEK are in fairly good agreement. The most apparent differences concern the standard deviations. There is no systematic pattern in these differences. They depend both on the methods of calculation and the values of the assigned experimental uncertainties. With RYDBERG we used the proposed standard deviationlo of 0.7% in [L]. With DALSFEK the options are selected by the operator. We used error values of 0.01 mV and 0.03% for the 3 analytical concentrations H+, total ligand and total La111. Rossotti et al.17 state that standard deviations of the constants990 POTENTIOMETRIC STUDIES OF LANTHANUM CARBOXYLATES may be obtained from the leading diagonal of the inverted product of the transposed and the untransposed coordinate matrix, but these deviations are only correct when a properly weighted treatment is applied.These procedures have been incorporated12 into DALSFEK. Rossotti et al.17 have also reviewed and discussed numerous published computer programs for stability-constant determinations and the associated error limits. They emphasise that the ‘best’ set of constants, however obtained, is not necessarily the ‘right’ one. The answers can never rise above the status of values ‘compatible with’ the available experimental data. One of our aims has been to meet this criterion by devising experimental conditions that generate accurate data from glass-electrode types of e.m.f. cells with the help of currently available eq~ipment.~? We thank the National University of Malaysia for a financial support award to P. B. A. P. Carpenter, C. B. Monk and R. J. Whewell, J . Chem. SOC., Faraday Trans. I, 1977, 73, 553. A. Sonesson, Acta Chem. Scand., 1959, 13, 998. Stability Constants, ed. A. E. Martell and L. G. Sillen (Chem. SOC. Spec. Publ. no. 17, The Chemical Society, London, 1984). M. Cefola, A. V. Celiano and P. S. Gentile, Znorg. Chem., 1962, 1, 290. 1. Dellien and I. Grenthe, Acta Chem. Scand., 1971, 25, 1387. H. S. Dunsmore and D. Midgley, J. Chem. SOC., Dalton Trans., 1972, 1138. G . L. Cumming, J. S. Rollett, F. H. C. Rossotti and R. J. Whewell, J . Chem. SOC., Dalfon Trans., 1972,2652. F. J. C. Rossotti and R. J. Whewell, J . Chem. SOC., Dalton Trans., 1977, 1223. C. B. Monk and M. F. Amira, J. Chem. SOC., Faraday Trans. I, 1980, 76, 1773. lo J. C. Sullivan, J. Rydberg and W. Miller, Acta Chem. Scand., 1959, 13, 2023. l1 R. M. Alcock, F. R. Hartley, D. E. Rogers and J. L. Wagner, J . Chem. SOC., Dalton Trans., 1975, l2 Solution Equilibria, ed. F. R. Hartley, C. Burgers and R. M. Alcock (Ellis Horwood, Chichester, l3 D. W. Marquardt, J . SOC. Znd. Appl. Math., 1963, 11, 431. l4 H. S. Dunsmore and D. Midgley, J. Chem. SOC. (A), 1971,3238; J . Chem. SOC., Dalton Trans., 1972, l5 R. S. Tobias and M. Yasuda, Inorg. Chem., 1963, 2, 1307. l6 N. Ingri and L. G. Sillen, Ark. Kemi, 1965, 23, 97. l7 F. T. C. Rossotti, H. S. Rossotti and R. J. Whewell, J . Znorg. Nucl. Chem., 1971, 33, 2051. 2189; 2194. 1980). 64. (PAPER 4/1019)

ISSN:0300-9599    

DOI:10.1039/F19858100983

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. 1, 1985, 81, 991-1001 Solvent-induced Changes in the Visible Transmission Spectrum of Illinois no. 6 Coal Enhancement of Light Transmission through a Microporous Solid by Solvent-induced Index Matching of the Pores BY JON S . GETHNER Corporate Research Science Laboratories, Exxon Research and Engineering Company, Clinton Township, Route 22 East, Annandale, New Jersey 08801, U.S.A. Received 20th June, 1984 Coal is a high-surface-area microporous solid with an optical density considerably greater than that expected from an organic matrix containing only simple aliphatic and aromatic hydrocarbons. Mechanisms proposed to explain the anomalously high optical density of the dry carbonaceous material include the presence of large polycyclic aromatic species, charge- transfer complexes and light scattering from inhomogeneities. Initial experiments testing these mechanisms have been performed by examining the changes induced in the optical transmission of Illinois no.6 coal when the dry coal is saturated with solvents of different swelling and optical properties. Over the spectral region from 4000 to 7000 A, transmission through thin sections of coal increases substantially upon addition of solvent. Initial analysis of the data shows that the transmission increase and wavelength dependence of the resulting transmission spectra correlate with a model in which transmission losses are dominated by light scattering from the pores. The principal effect of the solvent is to index match partially the pores to the organic matrix, thereby reducing the scattering cross-section and increasing the transmission through the coal.Coal is a structurally complex, chemically heterogeneous solid consisting of pre- dominately carbonaceous material. All coals are microporous solids containing a significant void volume (from a few percent to as much as 50%) and a large surface area (as much as several hundred square metres per cubic centimetre of solid). Viewed in white light, most coals appear black or dark brown. Transmitted light does not pass through pieces of coal which are thicker than a few micrometres. As a result, many spectroscopically based techniques applicable to non-carbonaceous microporous solids cannot be applied to composition or reaction studies of coal. A variety of studies have focussed on the examination of the optical properties of solid Nevertheless, the basis for the high optical density of coal is not known.Observed optical densities are of the order of 0.5 per micrometre of thickness in the mid-visible region of the Therefore, the absorptivity in the mid-visible is ca. 2000 (assuming that the average absorber is a three ring aromatic species and that the aromatic-to-aliphatic ratio is 1). This is approximately two orders of magnitude greater than would be observed if a simple absorption mechanism were responsible for the observed absorptivity. Numerous mechanisms have been proposed to explain this discrepancy, including the presence of large polycyclic hydrocarbons, charge-transfer complexes, exiton interaction between nearby molecular species and light scattering from inhomogeneities caused by macerals, pores and mineral matter.2 If the structural and chemical basis for the optical density in concentrated organic matrices such as coal were understood, experimental methods which are not currently practical would be 99 1992 TRANSMISSION SPECTRUM OF ILLINOIS NO.6 COAL available for coal studies. For example, changes in the absorptivity of coal undergoing a chemical reaction could provide both kinetic and mechanistic data about the reaction. Transmission infrared difference spectroscopy studies of the binding of nitrogen- containing bases to the pore surface of coal indicated that treatment of coal by a swelling solvent (e.g. pyridine) might be causing a change in the amount of light scattered by the coal sample and arriving at the dete~tor.~ This suggested that light scattering from pores might be an important factor in the high optical density of coal.Solvent-induced changes to the light scattering from the pore structure in coal can be measured in the visible region of the spectrum where the absorption coefficient for most of the organic species in coal is small and the strong wavelength dependence of light scattering5 results in easily detected wavelength dependences. Changes in either the optical density or colour of coal may accompany changes and variations of thickness.6$7 To insure that optical-density changes do not solely result from thickness changes, it is essential that the wavelength measurements be made over a wide spectral range rather than at a single wavelength.This paper is an initial examination of the visible spectrum of coal (4000-7000 A) to test if light scattering might be a major contributor to the optical density of coal. Illinois no. 6 coal immersed in a variety of liquid solvents was used. The samples were optically thin sections specially mounted so that a starting coal spectrum could be measured in a microspectrophotometer, the sample immersed in liquid and the resulting spectrum obtained from the same area of the sample. Complex changes occur in the spectrum of the solvent-saturated coal. There is a substantial increase in the intensity of transmitted light over the entire spectral region measured. The changes apparently result from both structural changes, which depend upon the degree of swelling, and light scattering by voids, which become filled with solvent.Neither model is adequate to explain the data, although a model involving multiple scattering from the voids and partial index matching of the voids to the coal matrix seems to account best for the spectroscopic changes observed. The initial experiments indicate that the theory of light scattering from dense particulate systems may provide a basis for explaining the optical properties of coal. These observations, from one particular coal, should apply to microporous organic matrices in general. Some considerations for experiments to clarify further the basis for the high optical density in these materials are discussed. EXPERIMENTAL SAMPLE PREPARATION Thin-section samples of Illinois no.6 coal were prepared following the procedures described previously for preparing similar samples used in infrared studies.8 The nominal sample thickness used was either 0.4 or 1 .O pm. The samples were cut so that the knife travel was approximately parallel to the bedding plane of the block of coal. Several thin sections, each ca. 1 mm x 1 mm, were transferred onto a rectangular glass cover glass (thickness ca. 0.14 mm). The cover glass was used as the bottom of a liquid microcell which was prepared as follows. Two small strips of nickel micromesh acted as spacers between the sample-containing cover glass and a square cover glass (thickness ca. 0.14 mm), which was used to cover the samples. The top cover glass was sealed to the bottom cover glass by a thin seal of dental wax.A small opening was left in the wax on one side so that solvent could be placed in the interior of the cell. Samples were stored under nitrogen until they were used. SOLVENT TREATMENT Coal samples to be examined were immersed in tetrahydrofuran (THF), nitrobenzene, pyridine, chlorobenzene, acetonitrile or dimethyl sulphoxide (DMSO) using the procedures described below. The solvents were of standard reagent grade except that the pyridine was storedJ. S . GETHNER 993 Table 1. Indices of refractioqa swelling ratios,b absorbance ratiosC and values of nd for solvent-swollen coal solvent index of swelling absorbance refraction ratio ratio n acetoni trile 1.34 1.1 1.6 3.6 chlorobenzene 1.524 1.23 2.2 3.6 ni trobenzene 1.55 1.56 2.1 4.1 THF 1.407 1.90 2.2 3.4 DMSO 1.48 2.28 1.7 3.5 p yridine 1 SO9 2.61 2.8 3.5 a Ref.(9). A = 5144 8, log ( I s ) = log (BK+c//ln) as described in Experimental. Ref. (10). Absorbance ratio = (raw coal)/(solvent-swollen coal) measured at Value of n obtained from a fit of the data for the solvent-swollen coal and to in contact with 5 A molecular sieve to reduce contamination with water. The coal samples were immersed in the solvent using the capillary action provided by the cell to draw solvent from a pipette. At least 50% of the volume of the cell could easily be filled with solvent. The remaining air bubble was moved (if necessary) so that the coal sample was immersed in the solvent. Swelling of the coal was visually verified using the microscope.The sample cell was sealed by melting a small amount of dental wax over the opening through which the solvent had been introduced. Samples sealed in this manner were usually stable for long periods of time. We observed no significant discoloration of the solvent or loss of general sample shape over a period of 1 week. The indices of refraction of the solvents are assumed to be those measured for solvents of comparable grade by Aldrich;g they are shown in table I with the swelling ratios of Green and Larsen'O for volumetric swelling of unextracted Illinois no. 6 coal. Since sample-to-sample variations are expected, these swelling ratios are probably accurate only to ca. & 10% for our samples. SPECTRAL AND DEPOLARIZATION MEASUREMENTS Absorption measurements were performed using a single-beam microspectrophotometer constructed with a Leitz Ortholux microscope as the basic optical system.* A diagram of the spectrometer is shown in fig.1. The sample was imaged using a matched pair of 32 x quartz objectives (Zeiss Ultrafluor). The light transmitted through the sample was detected using a broad-band photodiode (EG&G UHV-4000) positioned in the aperture plane of the objective. Illumination was from a 150 W high-pressure xenon arc lamp. The light was monochromatized through a McPherson monochrometer (model 218). A bandpass of ca. 20 8, was used. Additional orders of light from the monochrometer were re'ected usingedge filters on the output 03SWPO19). The useable bandpass extended from slightly less than 4000 to ca. 70008,. Fluctuations in lamp intensity were compensated for by taking the ratio of the transmitted signal to a reference signal generated from a photodiode positioned in front of the sample (see fig.1). A 0.25 s RC time constant was used on the output from both the reference and the signal photodiodes. The entire system was interfaced to a minicomputer which controlled the monochrometer scanning and digitized the compensated signal (output from a PARC model 188 ratiometer). Single-beam spectra were obtained for both the sample and a blank area of the slide. The absorbance was then computed by taking the log (sample intensity/blank intensity). The area observed (ca. 0.0073 mm2) was fixed by an imaging diaphram positioned in front of the detector and calibrated by observing the nickel micromesh spacer.beam to reject light of wavelength < 4000 and > 6500 R (Melles Griot no. 03FCG055 and * Construction details available from the author994 TRANSMISSION SPECTRUM OF ILLINOIS NO. 6 COAL LS = LENS SYSTEMS C = CONDENSING LENS M = MIRRORS 0 = OBJECTIVE LENS MIJMD = MEASURING LENS 8 MEASURING OM = OBSERVATION MIRROR FYFD = FIELD LENS 8 FIELD DIAPHRAGM SS = SAMPLE STAGE DIAPHRAGM (SWING OUT) B S = BEAM SPLITTER PMT-1 = SAMPLE PHOTOMULTIPLIER TUBE PMT-2 = REFERENCE PHOTOMULTIPLIER TUBE Fig. 1. Schematic diagram of the microspectrophotometer. At least three areas of two or more different samples were measured for each solvent. Samples were cut from 3 separate blocks of coal. Spectral data were analysed by averaging several measurements for different samples.The variation in measurements of different areas of the same sample show a spread which is slightly less than the variation observed between different samples. Thus, experimentally significant sample-to-sample variability was not found. However, the greater spread in the measurements from different samples may indicate that a systematic variation between samples exists which is comparable to our reproducibility. This is likely to be caused by variations in cutting thickness, which are larger between samples than within a single sample. Depolarization measurements were performed using either standard Leitz plastic polaroids supplied for the microscope or a calcite air-spaced prism polarizer to polarize the incoming light in front of the condenser.An instrumental depolarization ratio of 1000 to 3000 was obtained over the useable spectrometer bandpass. The accuracy of high-depolarization-ratio measure- ments is limited by the gain of the photodiode used. TREATMENT OF DATA For light traversing a non-absorbing, optically thin medium, the total scattering, integrated over all angles, is related to the transmission by I, = 1,-I where I, is the integrated scattering, 1, is the intensity of the incident light and I is the intensity of the transmitted light. The transmission (T = I / I o ) is measured with the single-beam microspectropho tometer. Thus If T is close to unity, then - (I - 1 / T ) "N - log T = A , where A is the absorbance. For this case, I = I,, and the equation for the scattering becomes I, = -Z(1- 1 / q .I, = I o A .J. S . GETHNER o.O 3 995 0.0 4000 4500 5000 5500 6000 6500 7000 wave1engthlA Fig. 2. Transmission spectrum, plotted as absorbance against wavelength, of Illinois no. 6 coal. This approximation is valid only for values of the transmittance close to unity; it is approximately valid over the spectral range observed in these experiments. Thus, intensity data are displayed as absorbance. Light scattering for a two-phase system of pores in a uniform matrix can be parametrized according to an inverse A relati~nship.~ This has been done using a linear least-squares fit of the data to the equation carried out over the restricted range 4750-6500 A. The constant, BK, is empirically determined by the fitting routine. Convergence was based upon maximizing the correlation coefficient of the fit.The baseline correction is needed in order to account for the observation of small areas of the slide not containing any sample (and therefore resulting in the determination of an I,, value appropriate to a sample area larger than actually observed). log [I,(A)] = log (BK+ c / P ) RESULTS The transmission spectrum of Illinois no. 6 coal was measured between 4000 and 7000 A. A typical spectrum plotted as absorbance is shown in fig. 2. The large residual absorbance at 6500 A is indicative of the strong attenuation of the transmitted light beam even in the near-infrared region of the spectrum. The even larger absorbance at 4500 A (0.58) produces nearly complete extinction in samples thicker than a few micrometres.The strong wavelength dependence of the transmitted light results in a strong thickness dependence of the total transmitted light and a substantial variation in the colour with transmitted light from a white-light source when viewing samples of varying thickness. Fig. 3 shows spectra obtained when samples similar to that shown in fig. 2 are immersed in tetrahydrofuran (THF), pyridine, nitrobenzene and a nitrobenzene + pyridine mixture. It can be clearly seen that the overall shapes of the spectra are similar both to each other and to that of the untreated sample. A substantial decrease in absorbance occurs over the entire wavelength region for the solvent-swollen samples. This is shown in table 1 for the data at 5144 A for representative coal samples with each of the solvents used.The decrease is greatest at shorter wavelengths. In order to eliminate the possibility that the absorbance change is due to sample996 TRANSMISSION SPECTRUM OF ILLINOIS NO. 6 COAL 0.6 5 0.4 c .+ Y ._ M .4 e, * + 2 0.2 0.0 I nitrobenzene tetrahydrofuran I-\\ 4000 4500 5000 5500 6000 6500 7000 w avelengthla Fig. 3. Transmission spectra, plotted as absorbance against wavelength, of Illinois no. 6 coal immersed in tetrahydrofuran, pyridine, nitrobenzene and a 1 : 1 nitrobenzene + pyridine mixture. 0.55 [ I 0.50 0 .+ 0.45 c1 x Y .+ s Y c 0.40 .- 0.30 4000 4500 5000 5500 6000 6500 7000 wavelength/A Fig. 4. Difference spectrum for tetrahydrofuran-swollen Illinois no. 6 coal. loss, the ratio of the spectrum for the swollen coal to the initial coal spectrum was computed.An example of the resulting difference spectrum for THF-swollen coal is shown in fig. 4. The difference spectrum is not constant. Since sample loss of an absorbing sample would result in a wavelength-independent change in the absorbance, the change between initial and swollen state is due to a mechanism other than sample loss or expansion. Similar difference spectra are obtained for all of the solvent-treated coals. It is important to try to distinguish between the effect of swelling and the effect ofJ. S. GETHNER 0 4 - 0 3 x u - .- E 0 2 - c u - .- 0 1 997 - A / A / . . A A ' $ 1 . 1- . - A ;--- - - - A / A - 0.0 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 index of refraction 0.1 1 A 0.0 / I l l " l l l ' l l ' ' ' l l / l 1 .o swelling ratio Fig.5. Plot of the intensity of absorbance at 5144 8, against (A) index of refraction and (B) swelling ratio. index matching. However, treatment of the coal with solvents cannot be done under conditions of constant swelling. In order to test that the transmission changes observed were not simply caused by perturbations induced by swelling or the lack of pore filling, data were obtained over a wide range of solvent swelling ratios and indices of refraction. When possible, (swelling ratio, index of refraction) pairs were used which would vary one parameter while approximately fixing the other. The absorption at 5144 A is plotted for all the samples as a function of both index of refraction and swelling ratio in fig. 5. Qualitatively, the same curves are obtained if data at other wavelengths are plotted.The absorbance data shown appear to be weakly correlated with the index of refraction, exhibiting a broad minimum in the absorbance. The same data plotted as a function of swelling ratio do not show a simple monotonic change correlated to the degree of swelling. Since the swelling varies from ca. 10% volume increase to almost a 300% increase, this relatively weak sensitivity of the transmitted light to the degree of swelling is surprising if sample swelling is the principle cause998 TRANSMISSION SPECTRUM OF ILLINOIS NO. 6 COAL 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 index of refraction A- t 3.0 1 swelling ratio Fig. 6. Plot of --n against (A) index of refraction and (B) swelling ratio. of the transmission increase.Therefore, these results suggest that index matching of the pores to the surrounding medium plays an important role in the transmission of light through coal. Since the transmission may be dominated by the index mismatch of pores to the surrounding coal matrix, the data were parametrized in terms of a light-scattering model appropriate for a simple two-phase system of voids embedded in a uniform medium (see Experimental). The spectra of the initial samples and the swollen samples were fitted to the function absorbance = BK+ C/An in order to determine the wavelength dependence of the spectra. Representative results as shown in table 1 and the full set of data plotted as a function of index of refraction and swelling in fig. 6. The wavelength dependence of the fit of the raw coal gives 1.9-2.4 for the group of samples examined and is in agreement with our previously reported measurements made on similar sample^.^ The wavelength depen- dence for the solvent-treated coals as compared with the starting sample correlatesJ.S . GETHNER 999 Table 2. Depolarization ratios of coal and coal immersed in solvent ratioa sample 4700 8, 6300 8, coal > 2.7 > 2.4 + pyridine 1 .o 0.95 +THF 1.1 1.24 a Ratio = (I_JIl,)sample/(I_JII,)blank . Fractional depolarization ratio relative to the blank measurement taken either through the slide alone or the slide plus solvent. Ratio = 1 .O will result if the depolarization ratio of the sample is the same as the blank. A ratio > 1.0 indicates that the sample transmits more light than the blank.In all cases, I_JIIl 4 1.0. with the increase of transmission through the samples and is responsible for the wave- length dependence seen in the difference spectrum (fig. 4). The spectral measurements are not expected to be sensitive to any changes in optical anisotropy of the samples since unpolarized light is incident on the samples and all polarizations are detected equally by the microspectrophotometer. Depolarization measurements are, however, useful for assessing the validity of scattering models for dielectric media. Depolarization measurements were performed at two fixed wave- lengths on several of the samples used for the spectral measurements. The ratios of the depolarization (IJIll) for the samples relative to the blank are shown in table 2.While the precision of these measurements is not very high, several important trends are seen. The coal alone clearly allows considerably more light leakage (birefringence) than either of the two solvent-swollen coals. The pyridine-swollen coal is experimentally indistinguishable from the blank, while the THF-swollen sample still shows some depolarization of the light. Similar results were reported by Green et aL1l Since pyridine treatment results in greater light transmission through the coal than does THF, this trend is expected. The data indicate that the depolarization ratios exhibit a wavelength dependence. However, these data are not sufficiently accurate to quantify this trend. DISCUSSION The trend exhibited by the transmission data is towards increasing transmission at all wavelengths when solvent is added to the coal.This is not confined only to the visible portion of the spectrum but has been observed in infrared difference ~pectra.~ In fig. 5 the dependence of the transmission on the index of refraction of the solvent and the swelling ratio of the coal is shown. The very strong tranpmission changes induced by solvent treatment will tend to obscure changes in molecular absorptions which might occur. Relatively weak dependence of the transmission on the particular solvent used is seen. Therefore, molecular absorption changes, accompanying chemical changes, might be observable if the coal is saturated with an inert solvent before inducing the chemical reaction. The data shown in fig. 5 indicate that transmission through the sample depends upon the index of refraction of the solvent in which the coal is immersed. The absorbance goes through a minimum for a solvent having an index between 1.41 and 1.511000 TRANSMISSION SPECTRUM OF ILLINOIS NO.6 COAL (fig. 5A). The absorbance does not scale monotonically with the degree of swelling (fig. 5B), thus ruling out physical changes caused by swelling as the sole cause of the absorbance change. The dependence of absorbance on the index of refraction of the solvent, though weak, indicates that the optical density of the coal may be caused by transmission losses from scattering of the incident light out of the beam. The absorbance scale is then proportional to the intensity of scattering and is therefore proportional to the index of refraction mismatch between the coal and the solvent.Thus, the curve shown in fig. 5A should be symmetric about the minimum and the index-matching criterion appears to be at ca. 1.48. The scattering for a simple two-phase system would be close to zero at such a matching condition. Since the scattering is significantly greater than zero, a more complex dependence of the scattering on the index of refraction, wavelength and degree of swelling is therefore indicated. This implies that the index-matching criterion may depend upon pore size and/or shape and/or pore-wall composition. The deviation of the scattering as a function of index of refraction from a symmetric curve may be caused by variations in the composition of the matrix surrounding the pore.Organic constituents of coal can be expected to have a range of indices of refraction from > 1 to ca. 1.6. The index of refraction of mineral matter is greater than that of organic matter and can be as high as ca. 2.3. The sharp increase of scattering at high index of refraction seen in fig. 5A is a consequence of the fact that pores in the organic matrix will have wall material with an index of refraction < ca. 1.6. The lack of a sharp minimum is probably a consequence of variations in pore surface compositions resulting in a distribution of minima. The very high optical densities of the unswollen samples (> 0.4) indicate that considerable multiple scattering probably takes place. Compression fracturing is present in these samples. The fractures are observed to have a much larger spacing than the shortest wavelength of light used (0.4,um).The light scattering from such a fractured matrix would be expected to result in a wavelength dependence (n) of ca. 1 and thus cannot account for the strong wavelength dependence observed. The mean size of the fractures varies inversely with thickness. Experiments in which the degree of fracturing was varied by cutting samples having a variety of thicknesses did not show any dependence of n on the extent of fra~turing.~ The wavelength dependence of the light scattered from pores having diameters less than the wavelength of light would normally be expected to vary as A-n, where 3 < n < 4.5 The observed power-law exponent (n = 1.9-2.4 for the unswollen samples) indicates that a simple two-phase model of independent scatterers cannot explain the data from the unswollen samples.The power of the wavelength, however, increases when the transmission increases upon subsequent swelling. This is consistent with the notion that the solvent causes index matching of the pores to the coal matrix thereby producing a sample having optical properties closer to those predictable by a simple scattering theory. Multiple scattering from the pores would result in n < 3 and would also produce substantial depolarization of light by an otherwise isotropic m e d i ~ m . ~ The observed decrease of the depolarization ratio which accompanies swelling (table 2) indicates that the intensity of light transmitted between crossed polaroids decreases when the coal is immersed in solvent.This is consistent with a decrease in the amount of multiple scattering resulting when the pores are partially index-matched to the matrix. However, the data indicate that the actual situation is more complex since such a simple explanation cannot account for all of the data.J . S. GETHNER 1001 Q DEPENDENCE OF THE SCATTERING Light scattering from a microporous system will show a strong angularly dependent scattering inten~ity.~ The use of a microspectrophotometer in these measurements does not allow accurate determination of the angular dependence of the scattering and is convenient only for measuring the integrated function. The choice of instrumentation was dictated by the desire to obtain the full wavelength dependence rather than the full angular dependence.The angular dependence is extremely valuable in that it can be used to confirm the hypothesis that light scattering is a major source of transmission losses. If light scattering is found to be significant, the angular dependence can be used to obtain the pore size and shape distribution and the total pore volume. Angularly resolved light-scattering spectrometers operating at a limited number of wavelengths are readily available. They can easily be adapted to perform such angular-dependent studies over a continuous wavelength region. CONCLUSIONS The wavelength dependence of the transmission spectrum of visible light through solvent-treated Illinois no. 6 coal has been studied. The results suggest that the multiple scattering of light by the pore structure contributes significantly to the overall optical characteristics of coal.The application of the theory of light scattering from dense particulate systems may provide a basis for explaining the optical properties of coal and permit the optical properties to be related to details of the molecular structure of the coal matrix and the physical structure of the pore system. Angular- dependent light-scattering studies should be able to provide a definitive test of this model. The preparation of the thin sections of coal by G. W. Nalavany is gratefully acknowledged as is his assistance with some of the measurements. The communication of solvent-swelling data (table 1) by J. W. Larsen contributed to the final experimental design and is gratefully acknowledged. (a) D. W. van Krevelen, Coal (Elsevier, Amsterdam, 1961); (b) Coal Structure, ed. M. L. Gorbaty and K. Ouchi, Adr. Chem. Ser. no. 192 (American Chemical Society, Washington, D.C., 1981). A comprehensive list of references to studies of optical properties can be found in the review by H. Tschamler and E. De Ruiter, in Chemistry of Coal Utilization, ed. H. H. Lowry (Wiley, New York, (1963), suppl. vol. J. S. Gethner, Am. Chem. Soc., Symp. : Div. Org. Coat. Plast. Chem., Prepr., 1980, 43, 41 3. M. Kerker, The Scattering of Light and Other Electromagnetic Radiation (Academic Press, New York, 1969). C . G. Cannon and W. H. George, in Proc. Con$ Ultrafine Structure of Coals and Cokes (BCURA, London, 1944), pp. 29&3 15. .I J. S. Gethner, unpublished data. ’ M. T. Mackowsky, Brennst. Chem., 1958, 39, 540. * J. S. Gethner, Fuel, 1982, 61, 1273. CataloglHandbook of Fine Chemicals (Aldrich Chemical Company, 1982). T. Green and J. W. Larsen, personal communication. Press, New York, 1982). l 1 T. Green, J. Kovac, D. Brenner and J. W. Larsen, in Coal Structure, ed. R. A. Meyers (Academic (PAPER 4/1055)

ISSN:0300-9599    

DOI:10.1039/F19858100991

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985,81, 1003-1014 Surface Characterization of a Grafted Vanadium-Titanium Dioxide Catalyst BY GUIDO BUSCA* AND LEONARDO MARCHETTI Istituto Chimico, Facolta di Ingegneria, Universita di Bologna, Viale Risorgimento 2, 40136 Bologna, Italy AND GABRIELE CENTI AND FERRUCCIO TRIFIRO Istituto Tecnologie Chimiche Speciali, Facolta di Chimica Industriale, Viale Risorgimento 4, 40136 Bologna, Italy Received 28th June, 1984 Surface properties of a low-content vanadium catalyst, prepared by grafting a high-surface- area TiO, (anatase) support, have been studied by means of diffuse reflectance, e x . and F.t.i.r. spectroscopy, t.p.d. and its catalytic activity by butadiene oxidation. It has been found that the grafting reaction of vanadium occurs on all surface OH groups of TiO,, leaving exposed Ti4+ cations.Vanadium sites are isolated on the surface and have labile oxidative and coordinative states. Moreover, they interact cooperatively with titanium sites. Evidence for other vanadium species (surface, interstitial or reticular) has not been found. Titania-supported vanadium oxideslq are widely used as selective catalysts for the oxidation of o-xylene,’? b ~ t e n e , ~ b~tadiene,~ toluenes and the ammoxidation of picoline.’ Different preparation procedures result in different V: Ti ratios, as well as in different activities and selectivities. In general, the optimum vanadium content is reported to be between 2 and 10 mol % V,O,. Recently,8T a grafting technique has allowed the preparation of active and selective catalysts of very low vanadium content (< 1 % V,O,).This technique also allows better control of the preparation and of the surface structure. However, EXAFS-XANES results on a grafted V-Ti-0 catalyst have shown evidence of ‘intrinsic disorder of bond length’, considered as ‘an impor- tant factor in determining the catalytic performance’.lO Other authors have obtained high-performance low-vanadium-content catalysts by another preparation procedurell- l2 and have ascribed the resulting high activity to the special structure of vanadia on the TiO, (anatase) support. However, neither the mechanism of monolayer formation nor the nature of the resulting surface structure has yet been clearly identified. This paper reports an investigation, using several techniques of surface analysis, of a grafted V-Ti-O catalyst prepared following the method of Bond and Bruckmann.8 A high-surface-area sulphate containing anatase support (from Tioxide Ltd) was used in order to obtain a useful catalyst for fluidized-bed reactors.EXPERIMENTAL CATALYST PREPARATION* 1 5 g of TiO, (Tioxide Ltd CLDD 1587/1; B.E.T. surface area 173 m2 g-l; X-ray diffraction : anatase 100% ; impurities: SO, 5.75% ; P,O, < 0.06% ; ZrO, 350 ppm; Nb,O, 120 ppm; CaO 110 ppm; K,O 70 ppm) was moistened with doubly distilled water and dried at 403 K for 20 h. 10031004 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST The powder was then added to a solution of 0.40 cm3 of 99% VOCl, (Aldrich) in anhydrous benzene (1 50 cm3), heated under reflux for 2 h, filtered and dried for 1 h at 373 K.Subsequently hydrolysis was effected by adding a few drops of doubly distilled water to the hot solid, followed by drying at 383 K for 1 h and calcination at 723 K for 1 3 h. The amount of VOC1, used slightly exceeded that needed for complete reaction of the estimated amount of surface hydroxy groups,8 and would correspond, if completely retained, to V,O, content of 1.54 mol % . EXPERIMENTAL TECHNIQUES A Perkin-Elmer model 124 visible spectrometer, a Varian E 4 X-band first-derivative e.s.r. spectrometer and a Nicolet MX 1 Fourier-transform i.r. spectrometer were used, all connected with conventional gas manipulation and evacuation ramps and measurement cells.13. l4 Desorption spectra were obtained using conventional t.p.d.apparatus15 (t.c. detector; heating rate 3 0 K min-I), using constant-flow dry and deoxygenated He as the carrier (flow rate 60 cm3 min-l). The usual pretreatment of the samples was evacuation at 723 K for spectroscopic measure- ments and exposure to He flow at the same temperature for t.p.d. Adsorbate compounds were hyperpure products from Carlo Erba and SIO (Milano, Italy). Catalytic tests were made in a fixed-bed flow reactor, using a mixture of 0.6% butadiene and 11 % oxygen in nitrogen. Analysis of the products were performed using a gas chromatograph on-line to the reactor; a Porapak QS column (f.i. detector) and a dimethylsulpholane column (t.c. detector) were used for separation of organic products and carbon oxides, respectively. RESULTS DIFFUSE REFLECTANCE SPECTRA Fig.1 shows the diffuse reflectance spectra of the vanadium ‘monolayer’ catalyst (VT) after calcination and after a short evacuation, as well as the spectrum of the TiO, support (TS) for comparison. A high-intensity charge-transfer band at ca. 24000 cm-l and a weak band at ca. 16000 cm-l are observed. An absorption band near 24000 cm-l is characteristic of V5+ ions in octahedral coordination;l4? l6 the band at 16000cm-l corresponds to the 2B2 -+ 2Bl transition of V4+ ions in a distorted octahedron such as in VOSO, - 5H,0.17 A second band is expected at ca. 13000 cm-l for octahedrally coordinated V4+, corresponding to the ,B2 + ,E(I) transition; however, it has been suggested that in a more symmetrical environment the band at 16000 cm-l is the most intense.” After a short evacuation at 723 K, no strong modification occurs, but a new band appears at ca.20000 cm-l. A similar band has been reported by King and Goodl8 for V3+ ions; this suggests that even a short evacuation causes partial reduction of the surface. To confirm this result, reported in fig. 1 (c) is the spectrum of the catalyst following butadiene oxidation (T = 553 K, 0.6% butadiene + air mixture). A strong increase in the intensity of the band at ca. 20000cm-l occurs, confirming the attribution of this band to reduced vanadium centres. E.S.R. SPECTRA VT samples, after evacuation at 673 K, show [fig. 2 (a)] an intense e.s.r. signal where three anisotropic components are resolved, all having an eightfold hyperfine splitting, indicating the presence of isolated paramagnetic VIV species (3d1, S = 7/2) of orthorhombic symmetry.Adsorption of water causes a change in the e.s.r. signal, which then shows two anisotropic components and eightfold hyperfine splitting [fig. 2(6)]. Evacuation at 573-673 K restores the previous spectrum; this indicates that VIV species are accessible to water and reversibly change their coordination upon adsorption-desorp tion cycles.G. BUSCA, L. MARCHETTI, G. CENTI AND F. TRIFIRO 1005 wavenum berlcm-' 12500 15000 20000 25000 I I I 1 1 800 700 600 500 400 wavelength/nm Fig. 1. Diffuse reflectance spectra of (a) TS, (b) VT, (-) after calcination and (----) after calcination and short evacuation, and ( c ) VT after interaction at 533 K with a mixture of N,, 0.6 vol % butadiene and 11 vol % 0,.The e.s.r. parameters are typical of vanadyl speciesl9 and fit well with those of the dehydrated and hydrated forms, respectively, of V02+ ions in the cavities of zeolites X and Y.20 The e.s.r. parameters are also similar to those observed for other low-vanadium-content V-Ti-0 powders produced by impregnation2' but differ from those of V4+ ions in TiO, lattices,lg* 22 serving as a confirmation that such paramagnetic centres are located on the surface. When activated samples are contacted with dry oxygen at 150 K and later briefly evacuated, the e.s.r. spectrum of VIV is still observed, not perturbed, but superimposed by another sharp signal formed by three anisotropic components (fig. 3) (8, = 2.021, g, = 2.009 and g, = 2.004). This signal is typical of the superoxide 0; ion in orthorhombic ~yrnrnetry.~, The absence of superhyperfine splitting indicates that his species does not interact with vanadium nuclei.On TS an almost identical signal is observed under the same conditions but it is less intense. This indicates that 0; interacts with Ti cations still present on the surface after reaction with VOCl,, and this result confirms that reported by Serwicka21 for a low-vanadium-content V-Ti-0 sample produced by impregnation.1006 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST I I I I I I I I g3 = 1.910 (bl T.P.D. SPECTRA Fig. 4(a) shows the temperature-programmed desorption spectra of water from TS and VT samples. Two maxima are observed at ca. 650 and 710 K in the thermal desorption from TS, whereas only one (T,,, = 573 K) asymmetric maximum towards higher temperatures is present in the t.p.d.spectrum of VT. The temperature range of the peaks at 650 K on TS and at 570 K on VT are connected with desorption of molecular water coordinated on Lewis-acid sites, according to Munuera and Stone24 and Egashira et ~ 1 . ~ ~ The lowering of Tmsx when passing from TS to VT may be taken as an indication of a lowering of the Lewis-acid strength of the surface. This is confirmed by the t.p.d. spectra of ammonia [fig. 4(b)], where two peaks on TS correspond to a single peak on VT at a lower temperature (683 K). However,G. BUSCA, L. MARCHETTI, G. CENT1 AND F. TRIFIRb 1007 g , = 2.021 t g3 = 2.004 Fig. 3. E.s.r. spectrum (150 K) of activated VT after contact with dry oxygen (50 Torr) and evacuation at r.t.7 5 0 650 550 -<** c---. ', NH3 ( b ) ___/-- I 7 50 650 1 , --- , 1 800 600 4 5 0 3 50 T/K Fig. 4. Temperature-programmed desorption of (a) water, (6) ammonia and (c and d ) benzene from (- - -) TS and (-) VT samples. Adsorption at room temperature. comparison of these spectra with those reported by Topsoe et aL2'j and Hidalgo et aL2' for ammonia desorption from various zeolites suggests that strong acid sites are present on both TS and VT surfaces. Note that in contrast to previous authors, we have found only high-temperature peaks, assigned to desorption of strongly bonded chemisorbed molecules, for the low initial surface coverage we adopted in our experiments. 28 Fig. 4(c) shows the thermal desorption of reversibly adsorbed benzene.On TS two desorptions of benzene are observed (T,,, = 350 and 380 K) corresponding to two adsorption sites, whereas on VT only the lower-temperature peak is detected. In accordance with the t.p.d. spectra of water and ammonia, this confirms a general weakening of the adsorptive interactions on VT with respect to TS and the1008 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST / f +*-: ,' 0 T 2 z E U U L L 3800 3600 34 00 3200 3000 wavenum berlcm -I Fig. 5. F.t.i.r. spectra of (--) VT after contact with water (10 Torr) and evacuation at r.t., 523 and 673 K and (- - -) TS after evacuation at 673 K. l i s 0 " W 0 hr 2 500 2300 2100 wavenurnber/cm-' 1: 30 Fig. 6. F.t.i.r. spectra of (-) VT in contact with CO at 0, 20 and 100 Torr, (---) activated TS and (----) TS in contact with 100 Torr of CO.Insert: expanded spectra of VT activated (-) and reduced (- - . .) in a benzene atmosphere (10 Torr) at 673 K.G. BUSCA, L. MARCHETTI, G. CENTI AND F. TRIFIRO 1009 1800 1600 1400 1200 wavenumber/cm-' Fig. 7. F.t.i.r. spectra of ammonia adsorbed on VT: (a) activated VT, (b) after contact with 20 Torr ammonia for 15 min and evacuated at r.t. and ( c ) after evacuation at 473 K. disappearance of some adsorption sites. A small fraction of irreversibly adsorbed benzene desorbing at temperatures > 500 K is also present in both surfaces [fig. 4(d)]. F. T. I. R. SPECTRA Fig. 5 shows the i.r. spectra in the v(0H) region of VT for different evacuation conditions; for comparison, the spectrum of TS evacuated at 673 K is also reported. As discussed elsewhere,13 four v(0H) bands are observed on TS at 3710, 3690, 3660 and 3640 cm-l (shoulder).On VT only one band, relatively broader and more intense, is observed near 3650 cm-l and it is assigned to a free surface hydroxy group. This seems to indicate that all surface hydroxy groups of the TS support have reacted with VOC1, and a new OH has been formed, probably bonded to vanadium ions. We also observe a very broad band centred near 3250 cm-l that progressively disappears under evacuation between 523 and 673 K; this band corresponds to the water desorption peak observed at 560 K in fig. 4(a). Under these conditions we also observe the change in the e.s.r. spectrum discussed above. We may conclude that desorption of molecular water bonded to Ti and V centres or dissociated into hydrogen-bonded hydroxy groups occurs under similar conditions and that the two peaks cannot be resolved in t.p.d.experiments. Under these conditions the v(0H) band at 3650 cm-l slowly decreases in intensity.1010 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST 360 0 2800 '7 2000 1800 . 1600 . 1400 ' 1200 wavenumber/cm-' Fig. 8. F.t.i.r. spectra of activated VT and during contact with increasing pressures of benzene vapour (up to 5 Torr). Fig. 6 shows the i.r. spectra of VT and TS in the 2400-2000 cm-l region, both before and after adsorption of CO. In this region VT shows two absorption bands that are not observed on TS. A very strong band at 2345 cm-l, which only decrease slowly with evacuation at 670 K, is related to trapped CO, molecules produced by oxidation of the organic solvent during calcination.A weaker band shows a maximum at 2060 cm-l and a shoulder at 2015 cm-l. However, if the sample is reduced in contact with benzene at 673 K, the component at 2015 cm-l becomes predominant. This suggests assignment of the 2060 cm-l band to the first overtone of a V5+=0 stretching band expected near 1030 cm-l. The band at 201 5 cm-l could then be assigned to the corresponding overtone of a vanadyl species where vanadium takes a lower oxidation state, in agreement with the previous diffuse reflectance measurements. A value of 1010 cm-l for v(V=O) of this species seems to be reasonable. On TS13 four bands due to adsorbed CO are observed when in contact with 100 Torr of the gas.The more intense, whose maximum shifts from 2207 to 2195 cm-l with increasing CO coverage, has been assigned to CO linearly coordinated on Ti4+ Lewis-acid sites, while the two weaker components at 2238 and 21 5 1 cm-l have been tentatively attributed to two CO molecules interacting with a single Ti4+ cation havingG. BUSCA, L. MARCHETTI, G . CENTI AND F. TRIFIR~ 101 1 100 80 n E h Y .4 5 60 V a, I $ 40 .r( h a, d s 20 0 22 0 2 60 300 340 reaction temperature/'C Fig. 9. Conversion (0) and selectivity to maleic anhydride (A) and furan (m) formation in butadiene oxidation on the VT sample. Experimental conditions: 0.6 vol % butadiene, 11 vol % 0,; g.h.s.v. 5461 h-l; 0.55 g of catalyst. two coordinate unsaturation~.~~ The lower-frequency band is clearly due to CO bonded to a reduced Ti3+ centre. On VT, only the stronger component of similar intensity is observed, the maximum shifting from 2203 to 2190 cm-l with increasing coverage and always some cm-l lower than on TS.This result, in accordance with those obtained for t.p.d. and e.s.r., indicates that Lewis-acid sites of TS are still present on VT, even if they are weakened; however, some particularly exposed sites, such as those responsible for the formation of the bands at 2238 and 2 15 1 cm-l, are destroyed or affected by inductive effects. However, i.r. spectra of adsorbed CO indicate that there are no Ti3+ centres present on the VT surface, even after evacuation at 723 K. Fig. 7 shows the spectrum of ammonia adsorbed on VT. After r.t. evacuation, bands at 1605 and near 1200 cm-l are indicative of ammonia coordinatively bonded (aas and d,, respectively), while the broader, weakly split band centred near 1455 cm-l and the shoulder near 1680cm-l are due to ammonium cations.This indicates that the Bronsted acidity of the surface OH of VT is sufficient to protonate ammonia. However, evacuation at 420-470 K causes the complete disappearance of ammonium cations, while chemisorbed ammoniais still detected up to 623 K. This confirms the perturbation of the surface acidity induced by supported vanadium. In fact on TS, where sulphate ions are present and induce strong Bronsted acidity, both chemisorbed ammonia and ammonium cations are stable up to ca. 670 K. The band present at 1370 cm-l for the activated sample and shifted during NH, adsorption is due to surface ~u1phates.l~ Fig.8 shows the spectrum of VT during contact with increasing pressures of benzene vapour. After contact with very small pressures (lo-l Torr), adsorbed benzene is detected. This species is characterized by a high perturbation of the two out-of-plane combination vibrations observed at 1980 and 1845 cm-l. These are the spectroscopic features most sensitive to the strength of absorbate bonding.29 This species is also observed on TS and is assigned to molecules n-bonded to Ti surface cations. At higher1012 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST pressures (up to 5 Torr), new bands whose frequencies are similar to those of the liquid appear and are related to the perturbation of the v(0H) band from 3650 to 3470 cm-l.This indicates that such a species is hydrogen bonded to the surface hydroxy groups. The shift (Av = 180 cm-l) agrees with the medium Bronsted acidity of these groups.3o This shift does not correspond to those observed for the OH of the TS support.13 The behaviour of the OH group of VT towards both benzene and ammonia supports its assignment to a V-OH group, in accordance with the similarity of the v(0H) frequency with that observed for a grafted V-Si-0 powder, which is certainly related to V-OH groups.31 CATALYTIC TESTS The activity of VT for oxidation reactions was tested using the butadiene oxidation as an example. The relevant activity was observed at very low temperatures (activity range 523-573 K); however, the selectivity for furan and maleic anhydride formation was high only at low conversion, suggesting that the catalyst is able to form maleic anhydride, but at high temperatures and conversion, successive oxidation to carbon oxides occurs.The catalyst did not show a decrease (after ca. 40 h working time) in catalytic activity and selectivity. DISCUSSION FORMATION OF THE VANADIUM MONOLAYER The i.r. spectra of VT show that reaction of VOC1, results in the disappearance of all free surface OH of the support and the appearance of a new OH group (identified as a V-OH group) whose chemical behaviour is different. Moreover, both the e.s.r. spectra of adsorbed 0, and the i.r. spectra of adsorbed CO indicate that Ti cations are still present on the VT surface, in amounts comparable to those on TS.These results indicate that reaction with VOCl, involves the free surface OH of titania. However, the result of this reaction is not complete coverage of the surface TiO, with a vanadium oxide monolayer. Ti4+ cations are still exposed and accessible to adsorbates and can, in principle, cooperate with vanadium sites in the mechanism of catalytic transformations. In particular, both i.r. and t.p.d. data indicate that the adsorption sites for benzene are Ti4+ cations. NATURE OF THE VANADIUM SURFACE STRUCTURES Visible diffuse reflectance spectra of the fresh catalyst show the presence of relevant amounts of V4+, together with V5+. Brief evacuation at 723 K causes the appearance of large amounts of V3+ under the reaction conditions (butadiene oxidation). The e.s.r.spectra confirm the presence of V4+ and indicate that such ions are isolated and accessible for the adsorption of water, which changes their coordination state. The catalyst is selective at low conversion, where visible diffuse reflectance spectra [fig. 1 (c)] show that the vanadium at the surface of the catalyst is in a reduced oxidation state. At higher temperatures and conversions, the absorption at 20000 cm-l in the diffuse reflectance spectra decreases. According to our results on the formation of maleic anhydride on V-P o ~ i d e s , ~ , ~ ~ ~ increasing the rate of oxidation of vanadium increases the rate of the consecutive oxidation of maleic anhydride to carbon oxides. The results obtained on VT thus indicate that the catalyst is active and selective in maleic anhydride formation but at the same time the activity in maleic anhydride decomposition is high because of the higher oxidation state of the vanadium.The stable activity of the catalyst suggests that no inclusion of vanadium in the anataseG . BUSCA, L. MARCHETTI, G . CENTI AND F. TRIFIRO 1013 structure occurs, at least during the 40 h of reaction used in this study. Furthermore we did not observe any evidence for reticular or interstitial vanadium and under all our experimental conditions X-ray diffraction excluded the formation of crystalline phases other than anatase. We can conclude that the vanadium in our catalyst is all exposed on the surface, and both its coordination and oxidation state are labile. This indicates that our catalyst is different from that studied by Rusiecka et a1.,12 since the vanadium is highly dispersed.The observation of different oxidation and coordination states of vanadium may be interpreted as supporting the observation of 'intrinsic disorder of vanadium species' reported by Kozlowski et al.l0 in a sample prepared using the same procedure. The F.t.i.r. spectra of adsorbed CO show the absence of Ti3+ ions on the surface, usually observed for anatase samples under similar ~0nditions.l~ However, e.s.r. spectra of adsorbed oxygen indicate that superoxide ions, formed by the removal of electrons from oxidizable centres, are located on Ti cations. This may be interpreted as an indication that vanadium stabilizes Ti4+ cations on the surface, but participates in the adsorption of oxygen by freeing an electron. This may also be a further indication that vanadium and titanium centres may interact cooperatively with adsorbates, possibly through electron exchange. EFFECT OF VANADIUM ON THE SURFACE ACIDITY The t.p.d.and F.t.i.r. data for the adsorption-desorption of ammonia, water and carbon monoxide agree, indicating a decrease in the Lewis acidity induced by vanadium. However, this effect cannot be ascribed to the poisoning of sites but to inductive effects. The nature of Lewis-acid sites on the monolayer catalyst seems to be substantially unchanged with respect to the support. The F.t.i.r. spectra of adsorbed ammonia and benzene indicate that on VT medium Bronsted acidity is present, as is usual for vanadium-containing oxides. In fact, the presence of vanadium causes the disappearance of the strong Bronsted acidity induced on TS by the presence of sulphates.However, the acidity of the V-OH groups is higher than that of unperturbed Ti-OH groups.13 This work was supported by ' Minister0 Pubblica Istruzione', Rome, Italy. We thank Prof. Vincenzo Lorenzelli (University of Genova, Italy) for use of the F.t.i.r. spectrometer and for helpful discussions. We also thank Tioxide Ltd for the non-commercial TiO, sample used in the fluidized-bed reactor. M. S. Wainwright and N. R. Foster, Catal. Rev., 1979, 19, 21 1. D. J. Hucknall, Selective Oxidation of Hydrocarbons (Academic Press, New York, 1974). G. C. Bond and P. Konig, J. Catal., 1982, 77, 309. G. C . Bond, A. J Sarkany and G. D. Parfitt, J. Catal., 1979, 57, 476.A. J. Van Hengstum, J. C. Omnen, H. Bosch and P. J. Gellings, Appl. Catal., 1983, 8, 369. G. C. Bond and K. Bruckman, Faraday Discuss. Chem. SOC., 1982,72, 235. .I W. E. Slinkard and P. B. Degroot, J . Catal., 1981, 68, 423. ' S. Lars and J. Anderson, J . Chem. SOC., Faraday Trans. I , 1979, 75, 1356. '# P. Konig, U S . Patent 4228038, 1980. I" R. Kozlowski, R. F. Pettifer and J. M. Thomas, J. Phys. Chem., 1983, 87, 5176. l2 M. Rusiecke, B. Grzybowska and M. Gasior, Appl. Catal., 1984, 10, 101. l3 G. Busca, H. Saussey, 0. Saur, J. C. Lavalley and V. Lorenzelli, Appl. Catal., in press. lil G. Martini, F. Trifiro and A. Vaccari, J . Phys. Chem., 1982, 86, 1573. l6 J. L. Falconer and J. A. Schwarz, Catal. Rev. Sci. Eng., 1983, 25, 141. l6 W. Hanke, R. Bienert and H. G. Jerschkewitz, Z . Anorg. Allg. Chem., 1975, 414, 109. M. Gasior, I. Gasior and B. Grybowska, Appl. Catal., 1984, 10, 87. C. J. Ballhauses and H. B. Gray, Inorg. Chem., 1962, 1, 113.1014 SURFACE STUDIES OF VANADIUM-TITANIUM OXIDE CATALYST E. F. King and M. L. Good, Spectrochim. Acta, Part A , 1973, 29, 707. l9 P. Meriaudeau and J. C. Vedrine, Nouv. J. Chim., 1978, 2, 133. 2o G. Martini, M. F. Ottaviani and G. L. Seravalli, J. Phys. Chem., 1975, 79, 716. 21 E. Serwicka and R. N. Schindler, 2. Naturforsch., Teil A, 1982, 36, 992. 22 J. L. Marill and D. Cornet, J. Chim. Phys., 1973, 70, 336. 23 M. Che and A. J. Tench, Adu. Catal., 1983, 32, 1. 24 G. Munera and F. S. Stone, Discuss. Faraday Soc., 1971, 52, 205. 25 M. Egashira, S. Kawasami, S. Kagawe and T. Seizame, Bull. Chem. Soc. Jpn, 1978, 51, 3144. 26 N. Topsoe, K. Petersen and E. G. Derouane, J. Catal., 1981, 70, 41. 27 C. V. Hidalgo, H. Itoh, T. Hattori, M. Niwa and Y. Murakami, J. Catal., 1984, 85, 362. 28 R. J. Cvetanovic and Y. Amenomiya, Adv. Catal., 1967, 17, 103. G. Busca, L. Marchetti, T. Zerlia, A. Girelli, M. Sorlino and V. Lorenzelli, Proc. 8th Znt. Congr. Catal. (Verlag Chemie, Berlin, 1984), vol. 3, p. 299. 30 H. P. Boehm and H. Knozinger, in Catalysis: Science and Technology, ed. J. R. Anderson and M. Boudart (Springer-Verlag, Berlin, 1983), vol. 4, chap. 2. 31 W. Hanke K. Heise, H. G. Jerschkewitz, G. Liscke, G. Ohlmann and B. Paulitz, Z. Anorg. Allg. Chem., 1978,438, 176. 32 G. Centi, G. Fornasari and F. Trifiro, J. Catal., 1984,89,44. 33 F. Cavani, G. Centi, F. Trifiro and A. Vaccari, Proc. Symp. Adsorption and Catalysis on Oxide Surfaces, ed. M. Che (Elsevier, Amsterdam, 1984). (PAPER 4/ 1 107)

ISSN:0300-9599    

DOI:10.1039/F19858101003

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985,85, 1015-1024 Enthalpic McMillan-Mayer Coefficients from Literature Data on Excess Enthalpies Application to Solutions of Alkanes in Alkan- l-ols BY MICHAEL BLOEMENDAL, KEES BOOIJ AND Gus SOMSEN* Department of Chemistry, Free University, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands Received 2nd July, 1984 A method is described to calculate enthalpic McMillan-Mayer pair-interaction coefficients from literature data on excess enthalpies of binary mixtures. The calculation also yields the enthalpies of solution at infinite dilution. A test of the method of mixtures, for which directly measured enthalpic interaction coefficients are known, shows that the method gives results of reasonable accuracy. However, the method is able to obtain a large number of interaction coefficients in a relatively simple manner. This makes interpretation and correlation analyses of the results possible, and this is illustrated by the application of the method to alkane + alkan- 1-01 mixtures.Results for alkanes dissolved in alkan-1-01s are discussed in terms of the Savage-Wood group-additivity treatment using one parameter for each solvent. In addition we present an empirical relation which makes it possible to correlate 49 enthalpic interaction coefficients of 6 different alkanes dissolved in 10 different alkan-1-01s with only one adjustable parameter. Over the last few years, an increasing amount of work has been done on the determination of interaction coefficients in dilute solutions using the McMillan-Mayer the~ry.l-~ The pair-interaction coefficients provide information about the interaction between two solute particles in a solvent,6y 9 9 lo, l1 Gibbs-energetic interaction co- efficients have been determined by vapour-pressure and freezing-point measurements and enthalpic interaction coefficients have been obtained from microcalorimetrically determined heats of dilution.Work has been undertaken to obtain theoretical or predictive descriptions of the interaction coefficients,', 2, 9 7 12-16 but these attempts were hindered by a lack of data, especially for non-aqueous solvents, because of the fact that the experimental determination of interaction coefficients is laborious and time consuming.83 l5 On the other hand, many data have been published for excess enthalpies of binary mixtures [for a recent compilation see ref.(1 7)]. In principle, enthalpic interaction coefficients can be extracted from these data, provided that the measurements were performed with accuracy down to low mole fractions: as far as we know, such calculations have not yet been performed. Therefore we present in this paper a method of calculating enthalpic interaction coefficients from the excess enthalpies of mixtures; this calculation also yields enthalpies of solution at infinite dilution. The method is tested for several systems, where both enthalpies of dilution and excess enthalpies are known, and it is applied to the determination of the enthalpic pair-interaction coefficients of several n-alkanes in a series of alkan- l-ols as solvents.10151016 MCMILLAN-MAYER COEFFICIENTS FOR ALKANES IN ALKANOLS METHOD OF CALCULATION The McMillan-Mayer theory,ls applied to the enthalpy H of a dilute solution, results in a series expansion in the molality m.’ For a binary system containing nA moles of solvent (A) and nB moles of solute (B) this gives H = nA hz +n,(hg 4-B: rn+Bf: m2i-Bf: m3+ ...) where hz is the molar enthalpy of the pure solvent, hg is the partial molar enthalpy of B at infinite dilution and Bk,Bf: and Bf: are the enthalpic pair, triplet and quadruplet interaction coefficients, respectively. These enthalpic interaction coefficients are related to the cluster integrals in the McMillan-Mayer theory and to the McMillan- Mayer coefficients [see ref. (9)]. For a mixture of nA moles of A with nB moles of B, the excess enthalpy per mole mixture, HE, is given by where hg is the molar enthalpy of pure B and x A and X , are the mole fractions of A and B, respectively.Substitution of eqn (1) yields H E = xB(hg - hg+ B!m+ Bf:m2+ Btm3+ ...) H E / x , = Asol hW i- B! m i- Bf: m2+ Bf m3+. . , (3) (4) where Asol h” is the molar enthalpy of solution of B at infinite dilution in A. Hence in principle the enthalpic interaction coefficients can be determined by an analysis of the values of HE/xB as a function of m. However, the McMillan-Mayer approach, and hence eqn (4), holds only at concentrations for which the solvent activity approximates the activity of the pure ~olvent.’9~* Since most sets of data on excess enthalpies contain only few values at low concentrations of each of the components, and these values are often not very accurate, it may be expected that only Bk can be calculated with reasonable reliability and that even this coefficient should be treated with caution.The above-mentioned restriction means that the concentration range to which eqn (4) can be applied is limited. On the other hand, the statistical analysis requires as many values of H E / x , as possible. We have analysed literature data using two equations based on eqn (4). In one analysis we truncated the right-hand side of eqn (4) after the B: term (quadratic fit) and in the other after the Bt term (cubic fit). Starting the analyses with the minimum number of data points necessary (for the quadratic fit 4 and for the cubic fit 5), we performed a series of subsequent fits by adding one value of H E / x , at the next higher concentration.Using a computer program we obtained two sets of ASol hW and Bg values, together with corresponding values for the standard deviation of the fit, ofit. In general the variation of ofit with the number of values used for the fit is similar for both sets. For a few data points ofit is usually low and because of overfitting it is without statistical significance. Under these circumstances the addition of one value results in a strong variation in the adjustable parameters. Then, sometimes after a maximum in ofit, the standard deviation decreases because of better statistics and the different parameters become more or less constant. However, when the molality increases further, higher terms of eqn (4) should be taken into account and a strong increase in ofit results.We have selected the AsolhW and Bi values corresponding to the minimum in ofit in the statistically reliable area for both sets. In fig. 1 an example is given for the set of data belonging to the cubic fit for the solute ethylene glycol in the solvent water.” orM. BLOEMENDAL, K. BOOIJ AND G. SOMSEN N 16- L 2 12. 5 a - M 2 4 . h --- 4. 1017 J - 320 2 8 0 h E 240 2 M f - 2 0 0 160 B! =fit 20.360 Fig. 1. Plots of ofit and B: against the amount of input data (nd) for the cubic fit of ethylene glycol in water; 2 denotes the selected value. Table 1. Enthalpies of solution and enthalpic interaction coefficients in H,O. Comparison between calculated and directly measured values ASol h"(ca1c.) Asol hm(exptl) B:(calc.) B: (exp t 1) compound /kJ mo1-1 /kJ mol-l /J kg moF2 /J kg mol-2 formic acid acetic acid 1,4-dioxane tetrahydrofuran ethylene glycol methanol methanol methanol - 0.70a - 1.59a - 10.3d - 15.0' - 6.3J' - 7.3l - 7.3" - 7.0P -0.678b -1.176b - 10.8e -14.2h, - -7.1h - 7.3m - 7.3m - 7.3m 65a 33tja 999d 16.6e 1416' 305J' 31 l1 260" 218P 86c 303" 1183f 1 1822 362k 247n 247" 247n a Ref.(20); ref. (21); ref. (22); ref. (23); ref. (24); f ref. (25); 9 ref. (26); ref. (27); ref. (4); j ref. (19); ref. (13); ref. (28); ref. (29); " ref. (30); * ref. (31); P ref. (32). The values of Asol ha and Bt from the quadratic and cubic fits must be comparable. We considered only systems for which this condition was fulfilled and adopted the mean of the values from both fits.In some cases the values of HE at the lowest concentrations are very inaccurate. This results in large deviations in H E / x B from the H E / x B against m curve obtained at higher concentrations and in extremely high values of ofit. In such cases these data were omitted in the analyses. TEST OF THE METHOD We have tested the approach for several aqueous systems where enthalpic interaction coefficients are known from directly measured enthalpies of dilution. In addition, the values of Asol ha obtained can be compared with calorimetrically measured enthalpies of solution. Some results are collected in table 1 and show that values of both ASoI ha 34 FAR 11018 MCMILLAN-MAYER COEFFICIENTS FOR ALKANES IN ALKANOLS Table 2. Enthalpies of solution and enthalpic interaction coefficients for ethanol and propan- l-ol in water As,,, h"(ca1c.) A,,, hm(exptl) Bk(ca1c.) BfLl(expt1) compound /kJ mol-l /kJ mol-1 /J kg mo1-2 /J kg mo1F2 propan- l-ol - 9.4a - 10.lb 142Y 561" propan-1 -01 - 10.4d - 10.lb 1 368d 561" ethanol - 10.7e - 10.2h 737e 243c ethanol - 9.9f - 10.2b 577' 243c ethanol - 10.29 - 10.2b 3059 243c Ref.(33); ref. (29); ref. (34); ref. (35); ref. (36); f ref. (28); ref. (26). m/mol kg-' Fig. 2. Plot of H E / x , against rn for ethanol in water: A, ref. (36); x ref. (28); 0, ref. (26). and l3; as calculated from data for excess enthalpies compare reasonably well with values from direct measurements. The maximum deviation for a calculated value of B; with respect to the measured one is 30 % .Although this might appear to be large, note that for B; coefficients obtained from enthalpies of dilution substantial discrepancies between different authors are found. For Asol h" the situation is better, although in this case directly measured enthalpies of solution may also differ. The application of the approach is restricted to cases where accurate measurements have been performed down to low concentrations and where no strong curvature in the H E / x , against rn plot occurs at low concentration of B. Table 2 presents results for propan-1-01 and ethanol as solutes in water. Although the values for propan-1-01 are in agreement with each other, they are much higher than the value from direct measurements. For ethanol a comparable deviation is found for the first two values, whereas the third shows much better agreement with the directly measured value.The reason becomes obvious in fig. 2, where we present H E / x , as a function of the molality for three different data sets for ethanol in water. Only the accurate data of MarshTable 3. Bk and Asol h" for alkanes as solute in alkan-1-01s as solvent at 298.15 Ka solute E ethaneb propaneC butaned pentanee nonanef decaneg k ?5 solvent Bk Asolhm B! Asolhm Bi Asolh" Bi Asolhm Bk Asolhm BF Asolhm 3 methanol ethanol propan- 1-01 butan- l-ol pentan-1 -01 hexan- l-ol heptan- l-ol octan- 1-01 nonan- l-ol decan- l-ol - 48 - 12 + 17 + 87 + 26 + 32 - - 25 - 37 0.36 0.41 - 0.80 - 1.03 - 0.92 - 0.92 - - 0.66 - - 0.47 -116 1.95 -63h 0.85h +10 0.26 +75 0.04 +26 0.06 +21 0.09 - - -7 0.27 - - - 3OOh - 102 +3 + 3 + 12 + 27 + 10 - 17 - 30 - 63 3.13h 1.53 0.79 0.67 0.56 0.49 0.58 0.69 0.76 0.9 1 -618 - 139 - 23 + 4 + 60 + 23 -2 +3 + 37 + 23 4.0 1 2.0 1 1.16 0.92 0.64 0.7 1 0.80 0.78 0.73 0.84 -631 - 272h -118 - 56 - 34 - 32 + 19 4.22 -761 4.61 ?c B 2 .w -357 3.17 2.20 -144 2.50 1.86 -82 2.16 ' 1.65 -32 1.83 5 U 1.41 -6 1.56 c, z 1.54 -31 1.66 8 2 - - - - w - - W t4 e a Units: B! in J kg molW2, Asol h" in kJ mo1-I; from ref. (40) at 170 kPa; f from ref. (41) at 170 kPa; 9 from ref. (42) at 170 kPa; from ref. (37) at 6900 kPa; from ref. (38) at 2985 kPa; based on the cubic fit only. from ref. (39) at 2985 kPa; L E) \D1020 MCMILLAN-MAYER COEFFICIENTS FOR ALKANES IN ALKANOLS 0- - 2 0 0 - N -. 2 M 2d h r" -400- rp -600- - r- -800- -200 -400 -600- 0 L - .C -800- 6 1C nC, alkane 100-j oi<d\ -100 -50 "E -50 50- - 50- ' O 0 V 01 ./ -1ooy----- 6 I( Fig. 3. Plot of BF for alkanes dissolved in several alkan-1-01s against the number of carbon atoms in the alkane molecule, nc, alkane. (a) Methanol, (6) ethanol, (c) propan-1 -01, ( d ) butan- 1-01, (e) pentan-1 -01, (f)hexan-1-01, ( g ) heptan-1-01, (h) octan- 1-01 and (i) decan-1-01. and coworkers,26 who have measured a large number of excess enthalpies down to low concentration, yield a reliable value for B:. Thus although the method may fail for some systems and the resulting Bk coefficients are less accurate than those determined experimentally, the method is useful as it allows us to obtain a considerable number of interaction coefficients in a simple manner, and for a correlation analysis of B: values the possible lack in accuracy in the values is outweighed by the large amount of data that can be obtained.APPLICATION TO ALKANE+ ALKAN-1-OL SYSTEMS Accurate measurements on enthalpies of mixing of several alkane + alkan- 1-01 mixtures down to low mole fractions have been performed by Christensen and c o ~ o r k e r s . ~ ~ - ~ ~ We have applied our method to their data and found it applicable to alkanes as solutes in alkan-1-01s as solvents. Because of strong curvature in the plots relating HE/x, to rn at low alkanol concentrations, it is not possible to calculate values of B? for alkanol as solute and alkane as solvent. Some of the results of Christensen and coworkers have been obtained at high pressure (up to 6900 kPa).From the H E values of iso-octane dissolved in propan-2-01, determined at different pressures by H e i n t ~ , ~ ~ it can be calculated that Bt of iso-octane changes from - 118 J kg mok2 at 100 kPa to - 122 J kg mo1k2 at 4200 kPa ( T = 298 K) and thatM. BLOEMENDAL, K. BOOIJ AND G . SOMSEN 1021 substantial changes in BfLl occur only at pressures > 10 MPa. From the results of Heintz on heptane in propan-2-01 it follows that BfLl is reduced from - 205 J kg molP2 at 100 kPa to - 144 J kg m o t 2 at 19.7 MPa. Values of Asol ha are even less sensitive to pressure changes. Hence it may be expected that the changes in BfLl and Asolha in the pressure range used by Christensen and coworkers are small and can be ignored, in view of the uncertainty of the method.Consequently we have analysed the data of Christensen and coworkers without corrections for pressure differences. The results of our calculations are shown in table 3. We will now focus our attention on the B: values only. For most of our systems it appears that B? decreases with the size of the alkane (solute) molecule and that the decrease becomes smaller when the solvent molecules become larger. The decrease with solute chain length seems to become an increase for solvents with the largest molecules (octan- l-ol, nonan- l-ol and decan- 1 -ol), but since we have only few data available, the evidence is not conclusive. In fig. 3 we give a graphical representation of the BfLl values in the different solvents as a function of the number of C atoms in the solute molecules, nC,,lkane. In the lower alkanols the shape of the curves resembles that of substituted amides dissolved in N,N- dimethylformamide, which we have studied previo~sly.~ In that study the decrease in BfLl upon enlargxnent of the alkylic part of the solute was attributed to polarophobic interaction, which was defined as the tendency of apolar particles to attract each other in polar media.The results for the systems described here may be interpreted in a similar way. Large alkane molecules, dissolved in an alkanol with a short alkyl chain, disturb the hydrogen bonding in the solvent. Adherence of the solute molecules to each other counteracts the decrease in hydrogen bonding and corresponds to a negative interaction enthalpy and a negative value of BfLl.The effect will become larger when the chain length of the solute molecules increases. On the other hand, in alkanols with large molecules an alkane which is much smaller may fit into the solvent structure without substantial disruption of hydrogen bonding, so the variation of B! with n,,,,,,,, will be reduced. This is exactly what has been found. Often an analysis of pair-interaction coefficients is given in terms of the Savage and Wood additivity approach.I3 The basic assumption of this approach is to split the two interacting solute molecules A and B into functional groups, where each group of molecule A interacts independently with each group of molecule B. The total pairwise interaction is the sum of all group interactions. This leads to B i B = nf n? hij ( 5 ) i, j where BkB is the enthalpic pair-interaction coefficient between the molecules A and B in the solvent under consideration, nf and n? are the numbers of groups of type i in molecule A and of type j in molecule B, respectively, and h, is the contribution to BiB of the interaction of one group of type i with one group of type j i n that solvent. In order to reduce the number of variables, one CH, group is counted as 1.5 CH, groups.Though approximative, the group-additivity concept often leads to reasonable results in water [for a review of applications see ref. (44)]. When this approach is applied to our B! values for alkanes dissolved in different alkan-1-ols, only one group interaction coefficient per solvent, hCH2, CHz, is needed and eqn (5) reduces to This means that this approach fails to describe a change in the sign of B!.Consequently, as can be seen from fig. 3, it can be applied only to the lower alkan-1-01s.1022 MCMILLAN-MAYER COEFFICIENTS FOR ALKANES IN ALKANOLS Table 4. Results of Savage and Wood approach for some alkan-1-01s as solvent hCHz, CHza *fit solvent /J kg mo1P /J kg mol-2b methanol -14 (2) 105 ethanol - 5.9 (0.3) 50 propan- 1-01 -2.6 (0.3) 56 a The value in parentheses is the standard deviation in hCH2, CHz. ufit is the standard deviation of the fit. N + 0 E 00 24 h . 12 10 8 6 4 2 0 1 2 4 6 8 10 nC, alkanol Fig. 4. Plot of hTHz, against the number of carbon atoms in the solvent, ncs alkanol. The results of a least-squares analysis in terms of eqn (6) are presented in table 4 and show that the magnitude of hCH2,CHz decreases from methanol to propan-1-01 and that the standard deviations of the fit are rather high. We now discuss whether the use of the molality as a measure for concentration is appropriate for a comparison of Bt and hCHz, CHz values in different solvents. It could be argued that the enthalpy of interaction depends on the number density of solute molecules.This means that the molarity scale should be used, and since the densities of the pure alkan-1-01s do not differ substantia!:y, the molality scale should also be a reasonable measure of concentration. However, in our approach we have interpreted the Bi values in relation to hydrogen bonding in the solvent, which means that the number of solvent hydrogen bonds between two solute molecules is important for their interaction.Therefore a concentration measure which reflects the number of solute molecules per number of solvent particles may be more suitable. The aquamolality, m,, is such a measure of concentration. It is defined as the number of moles of solute dissolved in 55.51 moles of solvent, where M , and M , are the molar masses of solvent and water, respectively. m, is related to m by m, = (M,/M,)m. (7)50 0 -50 -100 - -150 CI E on Y 2 -200 CCI rg -250 -300 -350 - 400( M. BLOEMENDAL, K. BOOIJ AND G. SOMSEN 2 3 5 I I I t t I I I I I 1 0.3 05 0,7 0.9 2.0 3.0 4.0 6.0 r 1023 Fig. 5. Plot of B;? against r = nC,alkane/nC!alkanol. The numbers denote the number of carbon atoms in the alkane. Enthalpic pair-interaction coefficients on the aquamolality scale, B!* w, are obtained from (8) Bh,W = 2 ( ~ W I M d B ! .Of course trends in the value of Bkv in relation to the size of the solute in the different solvents do not differ from those of B;, but since Mw/Ms c 1 the variations are less pronounced, especially for the higher alkan- 1-01s. We have also determined the different Savage-Wood CH,, CH, group-interaction coefficients on the aquamolality scale. Values of hrHp, CH2 as a function of the number of C atoms of the solvent molecules, nc, alkanol, are presented in fig. 4. It appears that I hFHzl CH2 I decreases rapidly with increasing nc, alkanol and becomes almost zero for the higher alkan-1-ols, in accordance with what has been mentioned above about disturbance of solvent hydrogen bonds by the solute.A possible hydrogen-bond disruption in the solvent by dissolved alkane molecules suggests a relationship between B$ and the ratio of the size of the solute and the solvent, or roughly with nC,alkane/nC,alkanol F r . In fig. 5 we have plotted B;jw as function of r. All data fit well onto a single curve, which can be represented by the empirical relation with A = - 15.3 1 0 . 3 J kgmol-2 and a standard deviation of the fit of 10 J kgmol-,. From eqn (8) and (9) it follows that the ‘normal’ enthalpic pair-interaction coefficients depend on both Ms and r through B!? = Ar(r- 1) (9) B! = (M,/M,) Ar(r- 1). (10)1024 McMILLAN-MAYER COEFFICIENTS FOR ALKANES IN ALKANOLS Analysis of values of Bt in terms of eqn (10) gives A as - 14.6 f0.4 J kgmoF2, whereas the standard deviation of the fit becomes 40 J kg mo1-2.In this way it is possible to describe 49 enthalpic pair-interaction coefficients of alkanes in 10 different but related solvents with only one adjustable parameter. This work was carried out under auspices of the Netherlands Foundation for Chemical Research (SON) and with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO). We thank R. P. Tito and R. Bleeker for their assistance. P. J. Rossky and H. L. Friedman, J. Phys. Chem., 1980,84, 587. L. R. Pratt and D. Chandler, J. Solution Chem., 1980, 9, 1. W. J. M. Heuvelsland, C. de Visser and G. Somsen, J. Chem SOC., Faraday Trans. 1 , 1981, 77, 1191. F. Franks, M. Pedley and D. S. Reid, J. Chem SOC., Faraday Trans.1, 1981, 77, 1341. G. M. Blackburn, T. H. Lilley and E. Walmsley, J. Chem SOC., Faraday Trans. I , 1982,78, 1641. G. Barone, V. Elia and E. Rizzo, J. Solution Chem., 1982, 11, 687. ' I. R. Tasker and R. H. Wood, J. Phys. Chem., 1982,86,4040. a E. Matteoli and L. Lepori, J. Phys. Chem., 1982, 86, 2994. lo H. L. Friedman and C. V. Krishnan, J. Solution Chem., 1973, 2, 119. l1 J. E. Desnoyers, G. Perron, L. Avedikian and J-P. Morel, J. Solution Chem., 1976, 5, 631. I2 J. J. Kozak, W. S. Knight and W. Kauzmann, J. Chem. Phys., 1968,48, 675. l3 J. J. Savage and R. H. Wood, J. Solution Chem., 1976, 5, 733. l4 B. Y. Okamoto, R. H. Wood and P. T. Thompson, J. Chem SOC., Faraday Trans. 1, 1978,74, 1990. l5 A. Ben Naim, in Hydrophobic Interaction (Plenum Press, New York, 1980), chap.3. I6 G. Barone, P. Cacace, G. Castronuovo and V. Elia, J. Chem SOC., Faraday Trans. I , 1981, 77, 1569. l' J. J. Christensen, R. W. Hanks and R. M. Izatt, Handbook of Heats of Mixing (Wiley, New York, M. Bloemendal and G. Somsen, J. Solution Chem., 1983, 12, 83. 1982). W. G. McMillan Jr and J. E. Mayer, J. Chem. Phys., 1945, 13, 276. l9 Y. Matsumoto, H. Touhara, K. Nakanishi and W. Watanabe, J. Chem. Thermodyn., 1979,9, 801. 2o R. Vilcu and E. Lucinescu, Rev. Roum. Chim., 1974, 19, 791. 21 J. Konicek and I. Wadso, Acta Chem. Scand., 1971, 25, 1541. 22 A. L. Harris, P. T. Thompson and R. H. Wood, J. Solution Chem., 1980, 9, 305. 23 J. R. Goates and R. J. Sullivan, J. Phys. Chem., 1958, 62, 188. 24 H. Nakayama, Bull. Chem. SOC. Jpn, 1970, 43, 1683.25 B. Y. Okamoto, R. H. Wood, J. E. Desnoyers, G. Perron and L. Delorme, J. Solution Chem., 1981, 26 M. J. Costigan, L. J. Hodges, K. N. Marsh, R. H. Stokes and C. W. Tuxford, Aust. J. Chem., 1980, *' D. D. Macdonald, M. E. Estep, M. D. Smith and J. B. Hyne, J. Solution Chem., 1974, 3, 713. 2a R. F. Lama and B. C-Y. Lu, J. Chem. Eng. Data, 1965, 10, 216. 29 A. C. Rouw and G. Somsen, J. Chem. Thermodyn., 1981, 13, 67. 30 E. Lange and H. G. Markgraf, Z. Electrochem., 1950, 54, 73. 31 M. K. Duttachoudhury and H. B. Mathur, J. Chem. Eng. Data, 1974, 19, 145. 32 S. Murakami, R. Tanaka and R. Fujishiro, J. Solution Chem., 1974, 3, 71. 33 S. R. Goodwin and D. M. T. Newsham, J. Chem. Thermodyn., 1971,3, 325. 34 F. Franks, M. Pedley and D. S. Reid, J. Chem. SOC., Faraday Trans. 1, 1976,72, 359. 35 V. P. Belousov, Vestn. Leningr. Univ., Fiz. Khim., 1961, 16, 144 [cited in ref. (17)]. 36 J. A. Boyne and G. A. Williamson, J. Chem. Eng. Data, 1967, 12, 318. 37 T. A. McFall, M. E. Post, J. J. Christensen and R. M. Izatt, J. Chem. Thermodyn., 1981, 13, 441. 38 M. E. Post, T. A. McFall, J. J. Christensen and R. M. Izatt, J. Chem. Thermodyn., 1981, 13, 77. 39 T. A. McFall, M. E. Post, S. G. Collins, J. J. Christensen and R. M. Izatt, J . Chem. Thermodyn., 40 S. G. Collins, J. J. Christensen, R. M. Izatt and R. W. Hanks, J. Chem. Thermodyn., 1980, 12, 609. 41 J. J. Christensen, R M. Izatt, B. D. Stitt, R. W. Hanks and K. D. Williamson, J. Chem. Thermodyn., 42 J. J. Christensen, R. M. Izatt, B. D. Stitt and R. W. Hanks, J. Chem. Thermodyn., 1979, 11, 261. 43 A. Heintz, Ber. Bunsenges. Phys. Chem., 1981, 85, 632. 44 I. R. Tasker and R. H. Wood, J. Solution Chem., 1982, 11, 749. 10, 139. 33, 2103. 1981, 13, 41. 1979, 11, 1029. (PAPER 4/ 1 134)

ISSN:0300-9599    

DOI:10.1039/F19858101015

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J. Chem. SOC., Faraday Trans. I, 1985, 81, 1025-1030 Photoionization of N,N,N',N'-Tetramethylbenzidine in Dodecyltrimethylammonium Chloride Cationic Micelles under 337 nm Laser Irradiation BY RAFAEL ARCE~ AND LARRY KEVAN* Department of Chemistry, University of Houston, Houston, Texas 77004, U.S.A. Received 4th July, 1984 The 337 nm steady-state laser irradiation of N,N,N',N'-tetramethylbenzidine (TMB) in dodecyltrimethylammonium chloride (DTAC) cationic micelle produces monophotonic ioniz- ation of the molecule. Rapid back reaction of the photoejected electron with the cation due to electron attraction by the positively charged micellar surface reduces the TMB+ radical cation yield to 0.002, one-fifth of the yield found in anionic micelles under similar irradiation conditions. The TMB+ visible absorption band shows a maximum at 465 nm, but broadens and loses vibronic structure in the cationic micelle in comparison with the anionic micelle.In the presence of anionic electron scavengers the radical cation yield increases to levels at which its electron spin resonance spectrum can be detected in room-temperature cationic micellar solutions. In room-temperature micellar solutions, laser flash photolysisl and electron spin resonance spectroscopy (e.s.r.)2 have demonstrated that the net photoionization of N,N,N',N'-tetramethylbenzidine (TMB) is less in cationic micelles than in anionic micelles. This was attributed to a fast photoejected-electron back reaction with the radical cation due to hydrated-electron attraction back to the cationic micellar surface and to an increase in the intersystem-crossing probability of the excited singlet to the triplet compared with photoionization.1 t In frozen solutions the net photoionization probability depends on electron escape from the micelle rather than on electron return to the micelle and the intrinsic photoionizaton yield is two-fold greater in cationic micelles compared with anionic mi~elles.~ The TMB+ radical cation absorption spectrum in photoirradiated cationic micelles at room temperature has been observed by fast optical detection methods1 but not by C.W. e . ~ . r . ~ The transient resonance Raman spectrum5 of TMB+ in cationic hexadecyltrimethylammonium bromide micelles evolves with time, whereas in anionic sodium dodecylsulphate micelles it does not. It has been speculated that the time evolution in cationic micelles is due to movement of TMB+ from the positive headgroup region farther out into the micellar Stern layer and that this driving force is lacking in anionic micelles.For comparison with studies on the photoionization mechanism of TMB in the anionic sodium dodecylsulphate micellar system,6 we have studied the photoionization mechanism of TMB in dodecyltrimethylammonium chloride (DTAC) cationic micelles under steady-state irradiation conditions. It is demonstrated that the photoionization process is monophotonic in cationic micelles and that the TMB+ optical absorption spectrum shows less vibronic structure in cationic as against anionic micelles. Added anionic electron scavengers are shown to increase g?eatly the yield of TMB+ in cationic micellar systems and to permit the first e.s.r. observation of TMB+ in cationic micellar systems at room temperature.t On sabbatical leave from the University of Puerto k c o , k o Piedras, Puerto Rico. 10251026 PHOTOIONIZATION OF TMB EXPERIMENTAL DTAC and TMB were obtained from Eastman Kodak Co.; other chemicals were reagent grade. TMB was used as received while DTAC was recrystallized twice from a mixture of 10% ethanol in acetone. Aqueous solutions of the surfactants were prepared with triply distilled and deoxygenated water. TMB was solubilized (0.1-0.3 mmol dm-3) in deoxygenated 0.1 mol dm-, DTAC by stirring at 60 "C for 3 h. These stock solutions were quantitatively diluted with deoxygenated 0.1 mol dm-, DTAC solutions to give an absorbance in the range 0.9-1.2 at the wavelength of maximum absorbance in a 1 cm path length Suprasil quartz optical cell.Aliquots of 2.2 cm3 of the solution were transferred to 1 cm Suprasil optical cells, stoppered with a cork and sealed with Parafilm under an atmosphere of dry nitrogen. Absorbance measurements were made with a Cary 14 spectrophotometer before and after irradiation. The e.s.r. samples for room-temperature measurements were contained in 75 A pipettes and the spectra recorded in a Varian E-4 e.s.r. spectrometer. Solutions containing solute scavengers (NaNO,, NaNO, and NaBrO,) were prepared by weighing the proper amount of scavenger salt and diluting in a 5 or 10 cm3 volumetric flask with the stock solution of TMB in 0.1 rnol dmP3 DTAC.Irradiations were done with the optical cell in a cell holder at a distance of 16 cm from a Lumonics model 861-T excimer laser operated at 337 nm with a nitrogen+ helium mixture. The absorption spectrum of the photolysed sample was recorded 130-140 s after irradiation. RESULTS AND DISCUSSION A 3.5 x loF5 mol dm-3 TMB solution in 0.1 mol dm-3 DTAC shows an absorption spectrum similar to that observed in 0.1 mol dmP3 SDS, but with a 2 nm red shift and maximum absorption at 307 nm (fig. 1). Continuous laser irradiation at 337 nm produces a faint yellow colour in the solution. The absorption spectrum of the irradiated solution shows a decrease in the absorbance of the TMB band and the appearance of a broad band in the visible region with an onset at 530 nm and extending up to 390 nm (fig.1). The absorbance of the visible band levels off with increasing irradiation time or the amount of light absorbed, and a much lower yield of photoproduct is observed in comparison with that in SDS micelles. Note that in fig. 1 the absorption in the SDS system is shown after partial decay. The visible band shows a broad maximum with a wavelength of maximum absorbance at 465 nm. This represents a blue shift of 8 nm compared with the TMB+ radical absorption maximum in SDS micelles. No well defined vibronic structure is observed on this band in DTAC micelles in contrast to SDS micelles (fig. 1). Although the observed absorption band in the visible region in photolysed TMB in 0.1 mol dmP3 DTAC micelles appears in the same wavelength region as that of the TMB+ visible absorption band in SDS micelles, some features of both spectra are not identical. If the broad band observed in 0.1 mol dm-3 DTAC is due to TMB+, the e.s.r. spectrum of the photolysed solution should correspond to the reported spectrum of TMB+.So far, in photolysed aqueous solutions of TMB in 0.1 mol dm-, DTAC at room temperature no e.s.r. signal has been observed for TMB+.4 Attempts to generate a higher steady-state concentration of TMB+ radicals by irradiating at higher laser power or by in situ irradiation in the e.s.r. cavity with a high-pressure mercury arc lamp were unsuccessful. In order to increase the TMB+ yield to make it detectable by e.s.r. techniques, negatively charged electron scavengers (NaNO,, NaNO, and NaBrO,) were separately added to the cationic DTAC micellar solution.In the presence of 8 mmol dm-3 NO;, irradiation at 337 nm of a 0.38 mmol dmP3 TMB solution in 0.1 mol dm-3 DTAC for 120 s produces an intense yellow colour in the sample and upon examination by e.s.r. a multiplet spectrum is observed (fig. 2). A corresponding increase in absorbance of the visible absorption band is also observed for samples containing the anionic electron scavengers. The broad band still lacksR. ARCE AND L. KEVAN 1027 240 280 320 360 400 440 480 wavelength/nm Fig. 1. U.v.-visible absorption spectra of 3.5 x mol dmP3 TMB in 0.1 mol dm-3 DTAC compared with irradiated TMB in 0.1 mol dmP3 SDS : (A) TMB in DTAC before 337 nm laser irradiation, (B) TMB in DTAC after 337 nm irradiation and (C) TMB in SDS after 337 nm laser irradiation. The SDS system has been allowed to decay after irradiation until the TMB+ intensity is comparable to the TMB+ intensity initially observed in the DTAC system.This was done better to compare the TMB+ lineshapes between the two systems. 5 G - H Fig. 2. E.s.r. spectrum of 0.38 mmol dm-3 TMB laser irradiated at 337 nm for 120 s at room temperature in 0.1 mol dm-3 DTAC containing 8 mmol dm-, NaNO,. The e.s.r. conditions were: modulation amplitude 0.5 G, gain 2 x lo4, microwave power 20 db, scan range 3400+ 100 G and laser power 1.7 x W. vibronic structure, as observed in the absence of the electron scavengers. Analysis of the e.s.r. spectrum’ and comparison with the e.s.r. spectrum of TMB+ in SDS micelles4 confirms the presence of TMB+ in the DTAC micelles.Kalyanasundaram and Thomas* observed perturbation of the vibronic band intensities in pyrene monomer fluorescence in trimethylalkylammonium micelles, and1028 PHOTOIONIZATION OF TMB 3 E / 0 0.5 1 .o 1.5 2.0 laser power/ 1 O2 W Fig. 3. Dependence of TMB+ yield measured by e.s.r. in 0.1 mol dm-3 DTAC at 77 K with respect to incident laser power. interpreted this as being due to interactions between pyrene and the quaternary ammonium headgroups. Similar interactions might account for the loss of vibronic structure and broadening of the visible absorption band of the TMB+ radical cation in cationic DTAC micelles as compared with anionic SDS micelles. This interaction does not seem to affect the unpaired spin distribution since the e.s.r.spectrum of the radical cation does not show any real differences from the e.s.r. spectrum of TMB+ in SDS micelles. In irradiated solutions of TMB in 0.1 mol dm-3 DTAC, the visible band decays with time and after irradiation one-half of the initially photodestroyed TMB is regenerated, as determined from changes in the absorbance in the TMB band. No bands are left in the visible region after TMB+ has completely decayed. Similar decay behaviour is observed in SDS micelles. The transient laser Raman spectrum of TMB+ in hexadecyltrimethylammonium bromide (HTAB) micelles has been observed and found to evolve with time on a nanosecond time~cale.~ Shifts in the frequency and intensity of some of the Raman lines are observed in HTAB but not in SDS micelles.It is possible that these changes in the transient Raman spectra in HTAB but not in SDS micelles are related to the differences observed by us on a long timescale in the visible absorption band of TMB+ in DTAC compared with SDS micelles. Assuming that the molar absorption coefficient of the TMB+ in 0.1 mol dm-3 DTAC micelles is the same as in 0.1 mol dm-3 SDS micelles (4 x lo4 dm3 mol-l cm-l),l an average TMB+ quantum yield of 0.002 was calculated in comparision with a quantum yield of 0.010 in SDS at a similar absorbed light intensity. A photoionizationR. ARCE AND L. KEVAN 1029 0 1 2 3 4 5 [NaN03 ]/mmol dm-3 Fig. 4. Variation of TMB+ quantum yield in 0.1 mol dm-, DTAC as a function of NaNO, concentration. quantum yield of 0.08 and triplet yield of 0.46 have been determined for the laser flash photolysis of TMB in 0.1 mol dm-, HTAB.2 The smaller photoionization yield in cationic micelles relative to anionic SDS micelles192 is atrributed to a faster back reaction due to hydrated electron attraction back to the positively charged micellar surface.In the absence of an electron scavenger the maximum attainable absorbance of TMB+ is of the order of 0.15. With such low absorbance values it was difficult to measure absorbance changes by decreasing the incident-light intensity or laser power by an order of magnitude. Thus, several sets of experiments were done in which the incident-light intensity was decreased by two to three times less than the initial intensity. These results indicate that the TMB+ yield as measured from the absorbance at 465 nm varies with the first power of the incident-light intensity supporting a monophotonic ionization process.Using e.s.r. methods, the dependence of the TMB+ yield on incident-light intensity in the presence of added nitrate ion was determined by measuring the e.s.r. signal height of the TMB+ radical cation in 0.1 mol dm-, DTAC at 77 K as a function of laser power. The results are given in fig. 3 and again support a monophotonic mechanism over a biphotonic one. Experimental results on the effect of NO; concentration on the TMB+ yield indicate that it takes only a concentration of 0.2 mmol dm-, NaNO, to double the yield of TMB+ in 0.1 mol dm-, DTAC. At concentrations of 5.0 mmol dm-, the yield approaches the observed yield in SDS micelles at similar absorbed light intensities6 (fig.4). Higher NO; concentrations were not used because of overlap of the nitrate ion absorption band with the TMB absorption band. The TMB+ yield is significantly increased by added nitrite and bromate as well as by added nitrate. The simplest interpretation is that all these anions act as electron scavengersg to increase the net yield of TMB+. They may also oxidatively quench the TMB triplet, TMBT, to generate TMB+. Nitrite ion is an efficient triplet quencher,l0? l1 although it is not known whether this leads to formation of TMB+. However, a recent study has shown that nitrate ion can react with TMB or1030 PHOTOIONIZATION OF TMB N-methylphenanthiozine triplets in water + alcohol to generate the corresponding cations.12 The following sequence has been suggested : TMBT+NO; -, TMB++NO:- NO:-+ H,O -, NO, + 20H- NO,+TMB -+ TMB++NO;.Note that two molecules of TMB+ are produced for each NO; by this mechanism. In positively charged micellar systems an anion like nitrate is adsorbed on the micellar surface to generate a high local scavenger concentration even in dilute bulk In the presence of 2.7 mmol dm-3 NaBrO, the TMB+ quantum yield increases to 0.35, approximately ten times higher than at a similar NaNO, concentration. The rate constant for electron scavenging by BrO; is 2.1 x lo9 dm3 mol-1 s - ~ , ~ one-fourth that for NO;. Tetravalent bromine, Br0:- or BrO,, is formed as an intermediate when bromate ions are reduced by electrons,14 and these intermediates were found to act as oxidizing agents. The difference in TMB+ yield in the presence of similar concentrations of different anionic scavengers could be explained in terms of differences in the degree of adsorption of the scavenger at the positively charged micelle surface; however, such differences should be small among mononegative anions.It is perhaps more probable that pathways other than simple electron scavenging occur. We believe that oxidative quenching of TMB triplets may occur for nitrate; it might also occur for nitrite, and in the case of BrO;, further oxidation of TMB by the intermediate species resulting from the electron reduction of BrO; may occur. These scavenging results in cationic micelles imply that the initial photoionization yields in both cationic and anionic micellar solutions are similar at room temperature. This research was supported by the Department of Energy under contract no. DE-AS05-80ER10745. S. A. Alkaitis and M. Gratzel, J. Am. Chem. SOC., 1976, 98, 3549. S. Hashimoto and J. K. Thomas, J. Phys. Chem., 1984, 88, 4044. P. A. Narayana, A. S. W. Li and L. Kevan, J. Am. Chem. SOC., 1982, 104, 6502. P. A. Narayana, A. W. S. Li and L. Kevan, J . Am. Chem. SOC., 1981, 103, 3603. S. M. Beck and L. E. Brus, J . Am. Chem. SOC., 1983, 105, 1106. R. Arce and L. Kevan, to be published. J. M. Fritsch and R. N. Adams, J. Chem. Phys., 1965,43, 1887. K. Kalyanasundaram and J. K. Thomas, J. Am. Chem. SOC., 1977,99, 2039. E. J. Hart and M. Anbar, The Hydrated Electron (Wiley, New York, 1970), p. 237. lo M. Almgren, F. Grieser and J. K. Thomas, J . Am. Chem. SOC., 1979, 101, 279. l1 A. Treinin and E. Hayon, J . Am. Chem. SOC., 1976,98, 3884. I2 A. J. Frank and M. Gratzel, Inorg. Chem., 1982, 21, 3834. l3 M. Gratzel and J. K. Thomas, J . Phys. Chem., 1974, 78, 2248. l4 G. V. Buxton and F. S. Dainton, Proc. R. SOC. London, Ser. A, 1968, 304,427. (PAPER 4/ 1 155)

ISSN:0300-9599    

DOI:10.1039/F19858101025

出版商:RSC    

年代:1985

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. 1, 1985,81, 1031-1035 Isotopic Effects on the Tracer Diffusion of Water, Methanol and Ethanol Dissolved in Carbon Tetrachloride at 25 OC BY HERMANN WEINGARTNER Institut fur Physikalische Chemie und Elektrochemie der Universitat Karlsruhe, Kaiserstrasse 12, D-7500 Karlsruhe, West Germany Received 5th July, 1984 An experimental search has been made for isotopic mass effects on the tracer diffusion of water, methanol and ethanol dissolved in carbon tetrachloride. No effect has been observed within the experimental precision of ca. 0.5%. The tracer-diffusion coefficients of water and methanol are compared with data on the rotational motion of these molecules at high dilution in CCl,, as reported in the recent literature. Upon dilution in CCl, the barriers which impede rotational motion are removed, but the translational motion is not accelerated.Current theories on diffusion in atomic or molecular liquids predict, in general, that the diffusion rate of an isotopic species present in trace amounts in a solvent should depend on the mass of this species, although the mass effects may be As isotopically labelled species are frequently used in diffusion studies, it is important to know whether such effects do contribute to the tracer-diffusion coefficients. Recent critical reviews of the pertinent experimental data2?, show that mass effects may be small or even undetectable. This paper reports an investigation of isotopic effects in the tracer diffusion of labelled species of water, methanol and ethanol dissolved in carbon tetrachloride.In all these cases the solute is small and light in comparison with the solvent molecules. Interest in these systems has been stimulated by recent work in this laboratory on the dynamica14y and structural6 properties of CH,OH + CC1, mixtures. In the course of these investigations the mutual- and self- (or intra-) diffusion coefficients in this system have been measured. In neat methanol at 25 "C and atmospheric pressure Hurle and WoolP have determined the ratio of the self-diffusion coefficients of CH,OH and CH,OD to be D C H 3 ? H / D C H 3 0 D = 1.09. We were able to show that this isotopic effect vanishes upon dilution in CC1,.8 It seemed to be worthwhile, therefore, to undertake a study of the tracer diffusion of labelled methanol in CCl,. Additionally, the tracer diffusion of various isotopic species of water and ethanol in CCl, was investigated.In particular, we were concerned with the question of what may be inferred about the properties of water in a state where the hydrogen-bonded network is broken up. Isotopic effects on diffusion in neat water have been well in~estigated.~? lo EXPERIMENTAL MATERIALS CC1, was of analytical grade and was purchased from Merck, Darmstadt. It was stored over molecular sieves and distilled before use. The radioactive tracers were supplied by the Radiochemical Centre, Amersham. CD,OT was prepared by dissolving a small amount of 10311032 ISOTOPIC EFFECTS ON TRACER DIFFUSION highly active CH30T in CD30H and it was assumed that the amount of remaining CH,OT has a negligible effect.DTO and C2D,0T were prepared by a similar procedure. DIFFUSION MEASUREMENTS The measurements were performed using a magnetically stirred diaphragm cell and the techniques described by Mills and WoolP1 [see also ref. (3)]. Experiments with methanol and ethanol tracers followed the well established procedure^.^, l1 Additional precautions were incorporated in the experiments on the water + CC1, system in order to avoid phase separation and the formation of microscopic droplets in suspension. In H,O + CCl, the presence of water droplets can be examined by recording high-resolution proton n.m.r. spectra, as the resonance signal of the protons in neat water is well separated from that of dissolved water,', and such measurements were performed at various stages of the experiments.CCl, and H 2 0 were mixed together and the saturated CC1, + H 2 0 phase was separated. This saturated phase was then further diluted in CCl, in order to avoid phase separation upon changes in temperature. The procedure chosen may have caused small differences in the actual water concentration of the various samples that were used, but no attempts were made to determine the concentration quantitatively. The diaphragm cell was filled homogeneously with the H,O + CCI, solution and was immersed in a water thermostat. A solution of HTO in CCl, was prepared by the same procedures and was also thermostatted. To start the diffusion run the unlabelled solution in the top compartment of the cell was replaced by the labelled one.Similar procedures were applied to determine the tracer-diffusion coefficient of DTO in CCl,. LIQUID SCINTILLATION COUNTING The home-built scintillation counter was essentially the same as that described by Mills and Woolf.ll All samples were prepared by accurate weight dilution. The more active top- compartment solutions were diluted by weight with carrier solution to yield approximately the same counting rates as the corresponding bottom-compartment solutions. Counting in the presence of CCl, leads to problems because of quenching effects. Counting of 14C-labelled tracers was performed in a scintillator solution consisting of 6 g dm-3 butyl- PBDt in toluene, using a weight ratio of 1 part active liquid in 100 parts of scintillator solution. A different procedure was used for tritium counting as the application of the usual procedures did not yield satisfactory results.5 g of the original sample was mixed with 40 g H,O and stirred under slight heating. Using this procedure all the tritium should exchange into the aqueous phase because of the fast proton exchange in water. After settling down, various samples were extracted from the aqueous phase and were counted in a scintillator solution consisting of 9 g butyl-PBD, 750 cm3 toluene and 300 cm3 ethanol, as recommended by Mills.g RESULTS The tracer-diffusion coefficients D* (the asterisk denotes ' tracer concentration') at 25 "C of the various labelled species of water, methanol and ethanol in CCl, are given in table 1. The level of accuracy is estimated to be +0.5% for 14C-labelled systems and In table 1 are also included the limiting mutual-diffusion coefficients DYz of CH,OH and C,H,OH in CCl, reported by Longsworth.13 Theory requires DY2 to be identical to the tracer-diffusion coefficients D* of this species, and this equivalence has been borne out by e~periment.~ For this reason Longsworth's data are directly comparable with our data.The extrapolation values given in table 1 should possess an accuracy of a few tenths of a per cent. For diffusion of H,O in CCl,, Longsworth has quoted two experimental values which differ by ca. 1 % . Both values are listed in table 1. 0.5-1 % for tritium-labelled systems. t 2-(4-t-butylphenyl)-5-(biphenyl)- 1, 2, 3-oxadiazole.H. WEINGARTNER 1033 Table 1. Tracer-diffusion coefficients for water, methanol and ethanol diffusing in carbon tetrachloride at 25 "C mass ratio tracer mass no./solute m2 s-' runs solvent D/ 1 0-9 no. of H2O HTO DTO CH,OH l4CH3OH CH,OT CD,OT C2H,0H 14C,H50H C2H50T C,D50T 18 20 21 32 34 34 37 46 48 48 53 8.3 4.07, 4.12a - 4.1 15 - 4.085 4.7 2.61b - 2.635 - 2.625 - 2.617 3.3 1.95b - 1.948 - 1.945 - 1.947 6 3 3 2 2 2 2 2 - - Experimental values quoted in ref. (13). Extrapolation values from ref. (13). DISCUSSION Before discussing the data in detail some comments should be made as to the state of the solute, because in these experiments the solute concentration cannot be made extremely small. In the methanol and ethanol systems the solute concentration corresponds to ca. 0.01-0.05 mol % . At these concentrations the observed behaviour can be attributed to the diffusion of monomeric species.This follows from the association constants for the various possible complex equilibria, the magnitudes of which are well established.l49 l5 Moreover, the concentration dependence of 4, at low alcohol concentrations leads to the same conclusion, as discussed in detail by Longsworth.13 Thus, there is little doubt that the concentrations chosen are small enough to fulfill, in practice, the requirement of infinite dilution. For water dissolved in CC1, and in similar solvents the corresponding information is more scarce. The magnitude of the lH chemical shift suggests that water may primarily exist as monomer,13 and various other data reported in the literature16-19 give support to such a conclusion.In the course of our experiments we have noted that there is no detectable dependence of DgTO on the variations in the concentration which are inevitably connected with the method of sample preparation. Therefore, in the subsequent treatment we will consider water to be virtually in a state of infinite dilution, being aware that this presumption is not totally unambiguous. From the data in table 1 we conclude that there does not exist a measurable isotopic effect in the diffusion rate of methanol and ethanol in CCl,. For ethanol all values agree to within 0.4% and their scatter does not reflect any correlation with tracer mass. For methanol our data centre around a value of 2.625 x m2 s-l and also do not show a correlation with tracer mass. Note, however, that our values are consistently higher than those of Longsworth: the difference between DY2 and the average over D* is ca.0.5 % . If mass effects were dominant CH30H should diffuse the most rapidly. There are examples in the literature2O9 21 where the heavier tracer has been observed1034 ISOTOPIC EFFECTS ON TRACER DIFFUSION to diffuse faster than the lighter one. We do not believe that the observed small discrepancy has any physical significance. The interpretation of the data observed for water is more difficult. According to our results HTO diffuses faster than DTO: the isotopic effect is 0.7%, which is larger than the experimental scatter obtained in a series of 6 diffusion runs for the H,O + CCI, system. If the observed isotopic effect has physical significance then H,O is expected to diffuse the most rapidly, as the mass difference of the HTO/H,O pair is twice that of the HTO/DTO pair.However, at a first glance such behaviour is not obvious from the two experimental values quoted by L o n g s ~ o r t h . ~ ~ In fact, the two mutual-diffusion coefficients have been measured by Raleigh interferometry, which in favourable cases may yield an accuracy of better than 0.1 % . 3 The very low water concentration in CC1, requires measurements with extremely small path differences, as is obvious from the number of fringes reported for these measurements. In this case the experimental scatter may well be of the order of 1 %. Although no statement on the accuracy was made by Longsworth, this explanation of the discrepancy of the two mutual-diffusion values seems to be the most probable one.Assuming this, there is no systematic correlation between Dcz0, DgTO and DE,,, and so the observed difference between DLTo and DgTO is expected to be an experimental artefact. However, we cannot totally exclude the possibility that the difference from Longsworth's data reflects a dependence of D,, on H,O concentration. In this case D,, would extrapolate to a higher value, so that DGZ0 > DgTO > DgTO. Even in this case the isotopic effect in the H,O/HTO pair should not exceed 1 %. It serves no purpose to discuss the reliability of all the theoretical models which provide predictions on the mass effects in tracer Claims have been madez2 that tracer diffusion obeys an inverse-square-root-of-mass law, as is predicted from both activation or free-volume theories and dilute-gas t h e ~ r i e s .~ In this case water would constitute a very favourable solute to test such a relationship because of its low molecular mass. For the H,O/HTO pair we would expect D*,pO/D&TO = 1.054, but such a ratio has not been observed for self- and tracer diffusion in neat H,0,99 lo where the ratio D&zo/DcTO = 1.03 is well establishedg and has virtually been reproduced.,, If water is dissolved in CCI, the corresponding isotropic effect is even smaller, if present at all. In other systems a square-root-of-mass dependence has not been borne out by experiments.,? Mills and Harris, have pointed out that there seems to be a small isotopic effect in the tracer diffusion of benzene in organic hydrocarbons which decreases on going from the C , to the C,, hydrocarbon.For benzene diffusing in octamethylcyclotetra- siloxane (OMCTS) no isotopic effect was observed at all,23 and similar results have been obtained for other systems consisting of a light and small solute in a heavier and larger The molecular mass of CCl, is approximately five times that of methanol and eight times that of water, so the mass ratio solvent/solute is even larger than it was in the systems considered in ref. (23) and (24). Thus, our results give support to the findings in earlier studies on the diffusion of a light solute in a solvent of larger and heavier molecules. Note that this result is at variance with the results from molecular-dynamics calculations, as Alder et have observed D* to become larger with decreasing mass and size of the solute.Finally, we make some comments concerning the magnitude of the tracer-diffusion coefficients. The self-diffusion coefficients of pure water (2.3 x m2 s-'),~> lo methanol (2.41 x m2 s - ' ) ~ are lower than those of non-associated liquids of similar mass and size, and reflect the effect of hydrogen- bond association on diffusion. Upon dilution in CC1, one may expect the translational motions to be markedly accelerated. In fact, it has recently been shown5 that the m2 s-l), and ethanol (1.09 xH. WEINGARTNER 1035 reorientational motion of methanol molecules reflects the rupture of hydrogen bonds upon dilution in CCl,. The limiting value of the rotational correlation time z2 = 0.2 x s as determined from 2H magnetic relaxation time measurements is only one-twentieth of the value observed in neat methanol and is very close to the estimated value for methanol rotation in the free gas, z2 = 0.15 x 10-l2 s .~ The small amount of work done on 2H and 170 magnetic resonance of water dissolved in CCl, l8, 26 does not yield an unambiguous value for the rotational correlation time, but the experimental data show that the rotational motion is accelerated by a factor of 5-10 if the associates present in pure water are broken up in CCl,. On the other hand, the data in table 1 show that the translational motions of water and methanol are not accelerated. Obviously, the water and methanol molecules can only move distances of their diameter if the larger CCl, molecules get out of their way.Thus, upon dilution in CCl, the barriers which impede rotational motions are removed, but the molecule is still captured with respect to translational motion by the surrounding cage of CCl, molecules. Prof. H. G. Hertz and Prof. R. Mills are thanked for helpful discussions. H. L. Friedman, in Molecular Motions in Liquids, ed. J. Lascombe (D. Reidel, Dordrecht, 1974), p. 87. R. Mills and K. R. Harris, Chem. SOC. Rev., 1976, 5, 215. H. J. V. Tyrell and K. R. Harris, Diflusion in Liquids (Butterworths, London, 1984), chap. 7. H. Weingartner and S. Prabhakar, Z. Phys. Chem., Neue Folge, 1983, 137, 1. H. G. Hertz and M. Holz, 2. Phys. Chem., Neue Folge, 1983, 136, 81. W. Koch, H. Leiter and S. Mal, 2. Phys. Chem., Neue Folge, 1983, 136, 89. ’ R. L. Hurle and L. A. Woolf, Aust. J. Chem., 1980, 33, 1947. * H. Weingartner and M. Holz, to be published. R. Mills, J. Phys. Chem., 1973, 77, 685. lo H. Weingartner, Z. Phys. Chem., Neue Folge, 1982, 132, 129. l1 R. Mills and L. A. Woolf, The Diaphragm Cell (A.N.U. Press, Canberra, 1968). l2 H. G. Hertz and W:Spalthoff, Ber. Bunsenges. Phys. Chem., 1959, 63, 1096. l3 L. G. Longsworth, J. Colloid Interface Sci., 1966, 22, 3. l5 W. C. Coburn Jr and E. Grunwald, J. Am. Chem. SOC., 1958, 80, 1318. l6 J. R. Johnson, S. D. Christian and H. E. Affsprung, J. Chem. SOC., 1966, 77. W. L. Masterson and M. C. Gendrano, J. Phys. Chem., 1966, 70, 2895. l8 J. C. Hindman, A. Svirmickas and M. Wood, J. Phys. Chem.. 1968,72,4188. l9 E. Greinacher, W. Luttke and R. Meche, Z. Elektrochem., 1955, 59, 23. 2o R. Mills, J. Phys. Chem., 1976, 80, 888. 21 R. Freer and J. N. Sherwood, J. Phys. Chem., 1981, 85, 102. 22 L. B. Eppstein, J. Phys. Chem., 1969, 73, 269. 23 K. R. Harris and R. Mills, J. Phys. Chem., 1977, 81, 2191. 24 R. Freer and J. N. Sherwood, J. Phys. Chem., 1981, 85, 932. 25 B. J. Alder, W. E. Alley and J. B. Dymond, J. Chem. Phys., 1974,61, 141 5. 26 G. D. Mateescu and G. M. Benedikt, J. Am. Chem. SOC., 1979, 101, 3959. H. Wolff and J. E. Hoppel, Ber. Bunsenges. Phys. Chem., 1968,72, 1173 and references cited therein. (PAPER 4/ 1 163)

ISSN:0300-9599    

DOI:10.1039/F19858101031

出版商:RSC    

年代:1985

数据来源: RSC