|
11. |
Stochastic models of multi-species kinetics in radiation-induced spurs |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2673-2689
Peter Clifford,
Preview
|
PDF (1187KB)
|
|
摘要:
J . Chem. SOC., Faruday Truns. I, 1986, 82, 2673-2689 Stochastic Models of Multi-species Kinetics in Radiation-induced Spurs Peter Clifford Mathematical Institute, Oxford University, St. Giles, Oxford OX1 3LB Nicholas J. B. Green, Mark J. Oldfield, Michael J. Pilling and Simon M. Pimblott Physical Chemistry Laboratory, Oxford University, Oxford OX1 3QZ Diffusion-limited kinetics in multi-species radical clusters, typical of those resulting from the passage of high-energy radiation through a liquid, have been analysed using several theories. A Monte Carlo (MC) model, in which the trajectories of the diffusing reactants are modelled by time-discretised random flights and in which reaction occurs when particles undergo pairwise encounters, is proposed as a measure of reality. The random reaction times (RRT) model, an efficient stochastic simulation technique based on the approximation that interparticle distances evolve independently, is general- ised to multi-species spurs and validated by comparison with the full Monte Carlo model.Further comparisons with other analytical models are less favourable and demonstrate the importance of a stochastic rate equation in predicting product ratios. The RRT model is also extended to include reactive products in three different ways, each of which provides an acceptable approximation to the observed (MC) kinetics in all systems tested so far. We have recently developed techniques for the analysis of kinetics in non-homogeneous systems containing small numbers of reactive s~ecies.l-~ The techniques have been designed primarily to describe the processes following the passage of high-energy radiation through a liquid.The fast kinetics observed following the energy deposition are largely diffusion-controlled and are associated with reactions within, and diffusive spreading of, small clusters of reactive particles known as spurs.4 The clusters are microscopic in extent and are produced by the physicochemical processes that follow the inelastic interactions of the high-energy particle with the solvent molecules. As the spurs are microscopic and non-homogeneous, their behaviour cannot be described using conventional theories of homogeneous reactions. Spur evolution, however, is an important stage in the chemistry of the irradiated medium, since a significant fraction of the overall chemical reaction takes place during this phase.Furthermore, the most direct experimental information about the chemical consequences of the energy deposition events is provided by a study of spur kinetic^.^ The need to extract information from spur kinetics has led to the development of numerous theoretical m~dels.~-l~ It is clearly of prime importance that these models accurately represent what they purport to describe. In recent papers we have demonstrated the importance of considering stochastic effects, which result directly from the small number of particles involved in the spurs. In addition, the use in such models of rate constants pertaining to homogeneous kinetics cannot be justified, and the microscopic non-homogeneities should be described in terms of the evolution of interparticle distances.We proposed three models that fulfill these requirements. The first model undertakes a full Monte Carlo simulation of diffusion and reaction in a spur. It provides an accurate description of the kinetics within the 26732674 Mu1 t i-species Kinetics in Radia t ion -induced Spurs limitations of the diffusion equation. The other models are approximate and are based on the assumption that interparticle distances evolve independently. One model depends on a master-equation formulation containing rate parameters derived from the distri- bution function of the reaction times of the pairs treated as if they were in isolation. The other model, the random reaction times (RRT) model, is a Monte Carlo method which makes use of the independent-pairs approximation to achieve much faster realisations than are possible with the full Monte Carlo model.The RRT model has many advantages over the master-equation approach and it is the model that we feel is most able to deal, effectively and efficiently, with real systems. Following the proposal of the underlying approach, we have been concerned with the development of the models so that they can be used to analyse experimental data. Thus far we have only considered single species spurs, with or without the presence of a homogeneous scavenger, and the present paper examines the generalisation of the full Monte Carlo model and of the RRT model to deal with reaction schemes involving two or more species. We also describe the development of procedures which can deal with the production of reactive products, which are formed, for example, via the reaction O+OH + HO,.We compare our results with those of a standard deterministic model and show that to account correctly for the ratios of product yields it is necessary to use a stochastic formulation. As many observational parameters used in the modelling of radiation- induced spurs have been derived from such ratios, this calls into question the conclusions drawn. In previous papers, we have demonstrated the validity of the random reaction times model by comparison of the time dependence of the number of surviving radicals, obtained from simulation using the RRT model, with those obtained using a full Monte Carlo simulation, which we have used as our criterion of reality. The validity of the latter assumption depends on the ability of the Monte Carlo model to describe diffusion and reaction accurately; these properties were assessed by comparison of the results of the MC simulation with the analytic solution for reaction of an isolated pair.The present paper retains the same overall approach, i.e. comparison of RRT and MC models. In the following sections we consider first the modification of the MC model, to make it easier to deal with more than one type of reactive species, and then the formulation of the RRT model, describing the procedures necessary to obtain realistic results where more than one type of radical is present. Three different ways are developed for handling the problems presented when the products themselves can react.Throughout these latter sections, the validity of the techniques is assessed by comparison with the results of the Monte Carlo simulation. Monte Carlo Model The Monte Carlo model employed in the simulation of single species spurs has been described in detail.2 The steps followed in the modified version for multi-species spurs are shown in fig. 1. Initial particle positions are generated from a model distribution function. The separation of each potentially reactive pair is calculated, checked against the corresponding reaction distance and if the particles are sufficiently close to react, then the reaction is counted and the particles are removed from consideration. When every pair has been checked for ‘zero-time reaction’ the particles are allowed to diffuse and react.In our original Monte Carlo model, as in the earlier models of Botkr et al.,15-17 the diffusive motion of the particles is simulated by a random walk on a cubic lattice. TheP. Cliford et al. generate initial ‘I positions 2675 begin (7 input information react any particles closer than a 1 calculate time period for random flight react each pair with Brownian bridge bridge reaction probability for no - each particle takes random flight Fig. 1. Flow chart for MC simulation. a represents the reaction distance of the pair. position of a particle is allowed to change by & I in each Cartesian coordinate during each time step, so that the diffusion coefficient is defined by D = v12/2 where v is the jump frequency (the reciprocal of the time step).To extend this to more than one type of particle would require different values of v or 1 for each species, neither of which is wholly satisfactory. Consequently we adopt a random-flight description of diffusion in which each particle is allowed to undergo a Gaussian random flight for a time period dt. The random flight is simulated by letting each coordinate change independently by a Gaussian random variable of mean zero and standard deviation 2/(2Ddt). The time step, dt, is calculated in such a way that the probability of any reactive particles meeting within the time period is small. Thus, if all the particles are far apart, simulation of large time steps is possible, increasing the efficiency of the program. The magnitude of the time step, dt, must be chosen with care.It must not be made too small, otherwise no reactive encounters may occur during d t ; on the other hand, speed of computation demands as large a value of dt as is feasible. The time step was calculated by considering the time-dependent survival probability for each pair, m, conditioned on reaction ultimately occurring, and a time, z,, evaluated for which the conditioned probability of reaction by that pair reaches a small, preset value. It has been2676 Multi-species Kinetics in Radiation-induced Spurs shown that the conditioned process has the same survival probability as a one-dimensional Wiener process,l* so that if a preset reaction probability of 1% is used, z, = [(r, - a,)/erf~-l(O.O1)]~/4D~ where rm and a, are the interparticle and reaction distances, respectively, for the mth pair and DL the relative diffusion coefficient.If this prescription were used throughout the calculation, then the computation would be very slow - as a pair approached to small distances, the time steps would get progressively smaller to keep the reaction probability below the preset value. To obviate this problem, since the aim of the calculations is, after all, to model reaction, a small constant time, z, is added to z, to give a significant reaction probability if the separation r , is small: z; = z, + T O . Since no times shorter than z, are examined, no reaction is permitted within the model between t = 0 and t = z,. Typically, a value of z, = 1 ps was employed so as to minimise errors at very short times.Thus the time step dt, for all the particles, is set to the minimum value of 7;. [Z.e. the time step is such that the probability of a reaction occurring is less than the preset value for all reactive pairs except if certain pair(s) are separated by a short distance, when zo dominates the determination of zL.1 The random-flight model follows the trajectory of the particles by determining their positions at the end of each time step, St. Reaction is simulated by the separation, r,, of a pair of particles attaining a value less than or equal to the reaction distance a,. If r , is less than a, after the time step, dt, then a reaction is registered. It is possible, however, for the particles to diffuse to a separation rm < a, and then to separate again during the course of the time step.Thus reaction should be registered, but would not be within the confines of the simple random-flight model outlined above. This problem may be overcome using a Brownian bridge, which is detailed e1~ewhere.l~ The probability that encounter of the mth pair at separation a, occurred during the time step, at, with initial and final separations r k and r,, respectively (r;, r , > a,), is given by W = exp [ - (rh - a,) ( r , - a,)/D& st]. The Brownian bridge is based upon an assumption of constant radial drift throughout the time step, i.e. the 2 / r radial drift term in the diffusion equation is considered to be invariant. This implies that, while for small time steps, and consequently small changes in the interparticle separation, the model will be accurate, for large time steps it will begin to break down.The Brownian bridge must therefore only be applied when particles are close together and there is a significant probability of reaction within 6t. Its use with large 6t leads to an overestimate of reaction rate. A set of predefined conditions, based on either a maximum value of 6t or a range of maximum interparticle separations, is necessary to prevent the use of the Brownian bridge when its assumptions are not valid. Tests have shown that the optimum constraint is to employ the Brownian bridge only when r/m or r , is less than twice the reaction distance. the random-flight model was validated by simulating recombination of a pair of particles with initial positions drawn from a Gaussian distribution.The simulations were compared with analytic solutions for first passage from the same initial distribution. Over the time regime studied ( t < 100 ns) the agreement was good, even without constraints on the Brownian bridge, and was improved when the limits were incorporated. As with the original lattice diffusion A typical set of results for the geminate recombination A+B+AB (1) is shown in fig. 2. The particle positions were independently generated from a Gaussian distribution of standard deviation 1 nm and the particles were given a relative diffusionP . Cliford et al. 2677 I I I 1 I I I 1.0 2.0 3.0 4.0 log (tips) Fig. 2. Geminate recombination of a radical pair. 0, MC simulation with Brownian bridge interpolation.(-) Analytic solution. Initial distribution of each particle concentric sphere Gaussian, (T = 1 nm. coefficient of lops m2 s-l and a reaction distance of 1 nm. This plot demonstrates the early time deviation caused by the method of calculating 6t and the quick swamping of this error. Random Reaction Times Model In this section we discuss the RRT model for spurs with more than one type of reactive particle and for schemes where reactive products can be formed. The RRT model will form the basis of our proposed code for simulating reaction in realistic tracks and so the aim of this section is to develop approximations whose accuracy is checked with the Monte Carlo model. Multi-species Spurs A flow chart describing the steps followed in the RRT simulation of multi-species spurs is shown in fig. 3.This is the same basic approach as that employed in the single-species simulation.l* The reactant particles are positioned by sampling from a Gaussian distribution; all the distances between potentially reactive pairs are calculated, and then examined for ‘zero-time reaction’. This procedure is equivalent to the initial stages of the full Monte Carlo simulation. All pair distances between remaining particles are then used to generate reaction times from the distribution function of the reaction time for a single pair at a separation r : W(t) = ( a / r ) erfc [(r-a)/.\/(4D’t)]. This reaction probability is a strictly increasing function of time and easily inverted by numerical methods : thus, as reported previ~usly,~ the inverse-function method can be used to generate a random time from this distribution function using a single random number uniformly distributed on (0,l).(It is at this point where the ‘independent-pairs assumption’ is introduced;l the assumption is essentially equivalent to that made in the standard theory of diffusiori-controlled reactions in homogeneous reaction systems.20)2678 react corresponding pair of particles ( i &i) Multi-species Kinetics in Radiation -induced Spurs remove all RRTs i o r j - for pairs containing begin 7 input information +’ $q--G? generate initial positions I react any particles closer than a I generate random reaction times (RRT) I Fig. 3. Flow chart for RRT simulation. The minimum of the resulting ensemble of random reaction times is found and the corresponding pair of particles is deemed to have reacted and both the particles are removed from consideration. The smallest of the random reaction times for the remaining pairs is determined and is treated in the same manner.The procedure is repeated until all reactions are completed, or the cut-off time is reached. With the modification of Monte Carlo and RRT models to deal with more than one type of radical, it is necessary to consider the ordering of the particles. At ‘zero time’ in both models and during the early period of spur evolution in the MC model, where time is discretised, the local radical density is high, thus presenting the possibility of multi-body interactions. If the order in which particles were examined depended on the species and if there were three particles, all close enough to react, only one type of reaction would be counted.For instance, consider the case of two A particles and a B particle, all mutually overlapping, and a reaction scheme : A+A+AA A+B + AB. If the A-A distances were always checked before the A-B distances, then only A+ A reactions would be counted from such three-body overlaps. Randomising the order in which pairs of particles are checked is most easily achieved by randomising the order in which reactant particles are considered.P. Cliford et al. 2679 Simulation of Spurs Symmetric in Two Species A number of test systems was studied, each involving two reactive species, designated A and B. Both species have diffusion coefficients of 5 x m2 s-l and all reaction distances are assumed to be 1 nm.In each test performed there were initially two A particles and two B particles. Their initial positions were generated from Gaussian distributions, whose standard deviations were varied (1 nm and 2 nm were used). Increasing the width of the original distribution decreases the number of zero-time interactions and causes more radicals to be present at short times and their increased dispersion reduces the rate of loss of particles so that less reaction is seen overall. The reaction scheme : A+A+AA\ I B+B+BBJ involves two truly independent pairs so the RRT model should describe the kinetics exactly. It is found that the programs do indeed treat the pairs as being independent with the results of both the Monte Carlo simulation and the RRT model closely matching the analytic solution.Applying scheme (I) (i.e. only cross reactions) to the 2A/2B system provides the first test in the present investigation of the independent-pairs assumption of the RRT model. There are four potentially reactive distances which do not evolve independently. The presence of the approximation is demonstrated by a small deviation between the full Monte Carlo and the RRT results. This error, which is small and only just outside the statistical limits after 1 O5 realisations, increases with time. The more complex reaction scheme : (IV) 1 A+A+AA A+B+AB B+B-+BB deals with species that have the potential to react with more than one type of partner and so examines the effects of competition. The plots shown in fig.4 demonstrate a small but just noticeable deviation between the models. We can therefore conclude that the independent-pairs approximation, as implemented in the RRT model, provides an acceptable approximation to the true diffusion-controlled kinetics in spurs containing several species, generalising our earlier conclusions.1-3 Comparison with Analytic Models In this section we compare the RRT model results with Schwarz’s deterministic modelg and with our analytic stochastic model based on a master-equation formu1ation.l The Schwarz model is a deterministic model in that it is based on the macroscopic rate equations, which for scheme (IV) are written in the form: As we have discussed previously,l this formulation is accurate only for systems containing relatively large numbers of particles (> 50), owing to the implicit deterministic averaging of particle numbers, and for systems where concentrations are initially locally homogeneous, because of the use of rate constants valid for constant concentrations.Neither of these conditions is true, either in real spurs or in the idealised spurs we are considering here. It should be recalled that agreement between the RRT and the full Monte Carlo simulations is almost exact, so that any significant deviations apparent in2680 Multi-species Kinetics in Radiation-induced Spurs I I I I 1.0 2.0 3.0 4.0 log (tips) Fig. 4. Comparison of MC and RRT simulations for scheme (IV). 0, MC simulation. (-) RRT simulation. (Gaussian spurs: oA = oB = 1 nm, N , = NB = 2) D, = D , = 5 x lo-’ m2 S-l, aAA = aAR = aBB = 1 nm.1 I I 1 OA 1.0 2.0 3.0 4.0 log (tips) Fig. 5. Comparison of RRT simulation and Schwarz model for scheme (IV). (-) RRT simulation, (-.-.) Schwarz model. Details as for fig. 4. a comparison of RRT and deterministic simulations demonstrate inaccuracies in the latter. In scheme (I) only cross reactions are possible and the agreement between the Schwarz and the RRT models is good, although the final yield of AB is rather too low for the deterministic model. This case, however, is the most favourable of the comparisons. For scheme (111), as the two reactive pairs in the system are truly independent, the RRT model corresponds to the exact solution and the Schwarz model underestimates theP. Cliford et al. 268 1 time-dependent product yield by between 15 and 30 % .The results for scheme (IV), where all reactions are possible, are shown in fig. 5 . Under the conditions considered in scheme (IV), A and B particles are identical in all but name. Thus, for the simulation of spurs symmetric in the two reactants, as each pair reacts with the same time-dependent rate, the ratio of AA:AB:BB products can be calculated by simple probabilistic arguments. In a spur containing one of each reactant the expected product ratios are 0: 1 : 0; with two of each they are 1 : 4: 1 ; with three of each 1 : 3: 1 etc. Our simulations always show the correct ratio within statistical limits. Solution of the deterministic rate equation, on the other hand, always gives the limiting ratio for a large number of particles, 1 : 2: 1 , whatever the initial number of particles, demonstrating clearly the unreliability of the use of deterministic models to evaluate not only radical time dependences, but also product ratios.As product yields are a major source of data for the parametrisation of spurs, this fault is not insignificant when linking the theoretical and the experimental aspects of radiation chemistry. The master-equation model described previously1* can also be generalised to these reaction schemes, at the expense of introducing a second dimension into the state space of the master equation. If we label the probability that the spur occupies the state containing NA A particles and NB B particles by P(NA, NB), then the independent-pairs approximation gives rise to the master equation? where &(t) = d In (llij)/dt and llij is the pair-survival probability for each ij pair, given the initial relative distribution of i and j particles.The component equations for the extremal values of NA and NB are readily formulated by comparison with ref. (1). The master equation may be shown21 to imply the kinetic equations: where ( N i ) is the expectation of Ni, thus demonstrating correct stochastic averaging. The master equation was solved numerically by Adams’ method. Agreement between master equation formulation and RRT simulation for scheme (111) is near perfect because the two reactive pairs are truly independent. Small deviations are found with the other two reaction schemes, the master equation making a small overestimate of the reaction rate.This observation shows the surprisingly small effect of competition and particle localization, which is analogous to the results on scavenging we have recently rep~rted.~ [The decay/formation curves for scheme (IV) are shown in fig. 6.1 The essential difference between the RRT and ME methods is that the ME method extends the independent-pairs approximation to zero time, whereas the RRT model starts with a realisable initial configuration of particles. The error introduced by applying the independent-pairs approximation to the initial distribution is seen to be small for Gaussian spurs. 89 FAR 12682 Mult i-species Kinetics in Radia tion-induced Spurs I I 1 I 1-0 2.0 3.0 4-0 log (tips) Fig. 6. Comparison of RRT simulation and master equation model for scheme (IV).(-) RRT simulation, (-.-.) ME model. Details as for fig. 4. Simulation of Reactive Products Reactive products are spur species, created by reaction of the initially generated radicals, which are capable of further reaction. When formed they must be positioned at the point of interaction and their subsequent diffusion and reaction with other radicals in the system must then be investigated in the same manner as the initially generated particles. As the full Monte Carlo simulation of a spur follows the positions of all the particles within the region, its extension to include reactive products is both conceptually and practically simple. This is not the case with the random reaction times model. The random reaction times approach uses the initial positions of the particles to generate a reaction time from the first-passage time distribution function for each potentially reactive pair.(The only important quantity is the interparticle distance at time t , set as t = 0.) If a reactive product is formed at time t = 0, its exact position is known and its subsequent reaction times can be generated. If a reactive species is created at time t > 0 then the required positional information is not available since the RRT model, as presently formulated, does not follow the diffusive motion of the particles within the spur. The model must therefore be supplemented by including an approxi- mation which allows the calculation of either new interparticle distances at time t > 0, or reaction times without interparticle distances.We have formulated several alternative approximations, which incorporate varying amounts of positional information about the system. These approximations must stand or fall by comparison of the results they produce with the full Monte Carlo simulation, as they are difficult to test in any other way. Models for Reactive Products Position Approach The rationale behind this model is that the positions of the particles which form a reactive product are determined explicitly and the separations between the product and all the remaining particles evaluated. Consider the first interaction to produce a reactive product, which occurs at time t .P. Cligbrd et al. 2683 In the spirit of the independent-pairs approximation, all the particles except those involved in the reaction can be assumed to have diffused freely for the time period t.Thus, remembering that diffusion is Gaussian in nature, the positions of the particles can be generated by allowing them to undergo a random flight, i.e. Ri = R; +N3(0, of 1) (1) where R; and Ri are the position vectors of the particle i at times 0 and t, respectively, N3 is a normally distributed, three-dimensional random vector whose mean is zero and variance/covariance matrix is of 1 (1 is the identity matrix) and at = 2D6t. The positions of the two reacting particles (labelled 1 and 2) are constrained by the fact that at time t they must be at the reactive separation a. IRl-Rz( = a To generate the positions of the radicals at reaction we consider the vectors Sl and Sz defined by Sl Rl-Rz S, = Rl+bRz where b is defined such that the S vectors are statistically independent, i.e.they have zero covariance. As N3 is spherically symmetric, the three directions of the Cartesian system are independent. Thus if xi = x;+o, Bi where xi and x; are x components of Ri and R;, respectively, and Bi is an independent normally distributed random number of mean 0 and variance 1, then cov (XI - x,, X, + bx,) = ((01 B1- 02 B,) (01 B1+ b02 B,)) = ~ q - b ~ t . Thus for statistical independence of Sl and S,, b = o:/o~. Sz diffuses independently of the interparticle vector Sl, and therefore independently of reaction up to the first passage time. It is defined by: and can be generated by independent random sampling from a normal distribution with the appropriate variance.The length of the vector Sl is defined by the reaction distance a. It is therefore m l y necessary to calculate its direction in order to position the reactive product. To evaluate the angular distribution of Sl we make the assumption that the distribution is that which would pertain at the reaction distance in the absence of any reaction. This approximation is probably not severe as the angular distribution of Sl is not the most important constituent of the vectors R, and R,. Given this approximation, the argument proceeds as follows : the initial interparticle vector is designated r’. The vector diffuses freely till the reaction time t when r’ has become Sl. As we require the direction of S, and already know its length, we consider the angular density function at time t conditioned on the separation being a: The density function describing the evolution of Y has the form p(r, t I r’) = ( 4 ~ D ’ t ) - ~ / , exp [ - (r - ~’)~/4D’t].89-22684 Multi-species Kinetics in Radiation-induced Spurs If the coordinate system is chosen so that r’ lies along the z axis, then the marginal radial density function is given by Substituting in the conditional density function gives P@, t ) = l j P ( a , 894 ; 0 do d4 p(a, t ) = a exp [ - (rt2 - a2)/4D’t] sinh [ar’/2D’t]/r’2/(7tD’t). p(8,4; t I a, t ) = [ar’ sin 8 exp (ar’ cos 8/2D’t)]/[87tD’t sinh (ar’/2D’t)] which is the joint density function of the random angles, 0 and 4, at reaction, given the reaction time t . The system we are considering is cylindrically symmetric about r’, which implies that 4 must have a uniform density of 1/2n as is borne out in the observation that p(r, t ) is independent of 4.It is therefore possible to generate 4 by 0 = 2nu, where U, is a random number uniformly distributed between 0 and 1. density function for 0 is obtained: If the equation for the conditional density function is integrated with respect to 4, a p(8; t ) = [ar’ sin 8 exp (ar’ cos 8/2D’t)]/[4D’t sinh (ar’/2D’t)] which can be used to define the probability that the random angle, 0, is less than 8. P(6 < 8) = [ar’/4D’t sinh (ar’/2D’t)] sin 8 exp (ar’ cos 8/2D’t) d8 s = [exp (ar’/2D’t) - exp (ar’ cos 8/2D’t)]/[exp (ar’/2D’t) - exp ( - ar’/2D’t)]. Given the distribution function P(O < 8) we can generate a random angle 0 from the prescription 0 = arccos[l +(l/a) ln(1- U2[1 -exp(-2a)])] where a is ar’/2D’t and U2 is another (uncorrelated) random number.Based on the assumption of an angular distribution which is unaffected by reaction, we can therefore generate the direction of S, and we can completely characterise R, and R,, the positions of the reacting radicals, from S, and S2 whilst those of all the other particles, Ri(i # 1,2) may be generated from eqn (1). Assuming that a reactive product is created either midway between (1 product particle) or at one of (2 product particles) the final positions of its parent radicals a completely revised set of radical positions can be determined and a set of RRTs can be generated for the new particle. D ifusion Approach The diffusion approach is based upon the independent-pairs approximation.New positions are not generated for the particles in the system. Only the interparticle distances are considered and they are assumed to vary independently. The separation between two diffusing particles is a Bessel process, which is the radialP . Cliflord et al. 2685 component of a multi-dimensional Wiener process. If, at time to the interparticle vector is Y’, then at time t, the vector has evolved to Y = r’ + NJO, 2D’(t - to) 13. As we are only interested in separations, we may assume a coordinate system in which Y’ is the z axis, so that I Y I = [ni + n i + (r’ + n,)2]1/2 where n,, n, and n, are independent, normally distributed random numbers. In the diffusion approach a reactive product replaces one of its parent radicals.The separations of the chosen parent particle from all other radicals are assumed to evolve independently and can be generated according to the prescription above. From each distance a new random reaction time can be generated for each reactive product/initial radical pair and the simulation continues as before. Time Approach Like the diffusion approach, this model is based on the independent-pairs approximation and on the replacement, by a reactive product, of one of its parent radicals taken at random. Random reaction times for the new particle are generated from the RRTs of the parent particles, diffusion coefficients and reaction distances. Consider the particles i, j and k, with RRTs tij, ti, and tjk. Suppose that if i reacts with j it produces I which can react with k ; let I replace i.The separation lk follows the same ‘path’ as the distance ik, up to tij, thereafter the same trajectory is followed with a new relative diffusion coefficient. Thus the remainder of the time until the first passage at aik is just scaled by the ratio of the relative diffusion coefficients. t(ai,) = tij +(ti, - tij) Dik/DLl t(aik) only represents the reaction time of k and I if the reaction distances aik and akl are the same. If the reaction distances differ, then ?(a,,) is not equivalent to t,,. When akl < aik the reactive product will not have reached its encounter distance with k at time t(aik). It still has to pass through the shell at aik and diffuse until it reaches the encounter shell at akl. If t& is the random reaction time that would have been generated for the ik pair if their reaction distance had been akl, then the true encounter time of k and I can be approximated by When akl > aik it is possible that t& will be less than tij: To simplify the calculation we assume that whenever tij > t& the particles k and I are within their encounter distance and instantaneous reaction takes place.Thus random reaction times are calculated according to the formulation : t,l = tij+(t&- tij) Dik/Dtl t& > tij t,, = tij t;, < tij. In view of the possible errors caused by this last assumption, we have also developed an RRT model which applies the diffusion approach when t:, < t i j and the time approach in all other cases. Simulations of Systems involving Reactive Products We have examined three systems containing reactive products.At the beginning of each test, we generate the positions of two A and two B particles from a Gaussian distribution of mean zero and standard deviation 1 nm. Every species involved in a reaction scheme2686 is given a diffusion coefficient of 5 x 1 nm. Multi-species Kinetics in Radiation-induced Spurs m2 s-l and all reaction distances are taken as The first reaction scheme is I A+A+AA A+B+AB B+B+BB AB+A + AAB AB+B+ABB. As no pairs are initially unreactive, this is the simplest possible system for the generation of reaction times for pairs containing reactive products. Our results show only minor deviations between the various RRT approximations and the full Monte Carlo approach.The diffusion and the time approximations give almost identical particle decay and formation curves to those of the complete simulation. The position approximation appears to cause a slight overestimate of the rate OY reaction of the reactive product, AB. The magnitude of the error is, however, small ( < 5 %). The identical nature of the reactants and the similarity of the pathways to products yield superimposable decay curves for A and B, and formation curves for AA and BB and for AAB and ABB. A typical set of decay and formation curves are shown in fig. 7. The reaction scheme \ A+A+ AA I A+B+AB B+B+BB AA+B+AB+A is not symmetric. It involves the regeneration of one of the initial species from a reactive product. This removes the correspondence between decay and formation curves for reactants (A, B) and products (AA, BB).The results obtained show close agreement between the full Monte Carlo approach and all the RRT approximations. The deviations are never greater than 3% and for the diffusion approximation are negligibly small. The deficiencies of the time approach to RRT simulation are expected to be greatest when some of the initial species do not react with each other. Unreactive pairs have an effective reaction distance of zero, which results in a large number of cases where tij is potentially greater than t&. A reaction scheme demonstrating this is (VII) A+B+AB AB+A + AAB AB +B + ABB. Here a(A + A) and a(B + B) are zero. The time approximation generates potentially erroneous reaction times for A+AB and B+AB pairs that were originally A+A and B + B pairs, respectively. Our simulations show the predicted overestimate of the rate of AB reaction. At long times there is approximately an 18 % deficiency of AB particles.The A/B decay curves and the AAB/ABB formation curves also have smaller long time errors. The position and diffusion approaches to the RRT simulation of reaction scheme (VII) give results that correspond much more closely to those of the complete Monte Carlo model. Like the time approach, the position approach gives a long time underestimate to the number of AB particles (7%). This is again due to an increased rateP. Cliflord et al. 2687 Fig. 7(a-c). For legend see following page.2688 Multi-species Kinetics in Radiation-induced Spurs log (tips) Fig.7. Comparison of different approximations for incorporating reactive products in the RRT simulation for scheme (V). 0, MC simulation; (-.-.) position approach; (--) time approach; (-) diffusion approach. Details as for fig. 4 with DAB = 5 x 10-9m2 s-l and uAAB = aBAB = 1 nm. In (a) and (b) all RRT simulations are coincident. for the reaction AB+C (C being A or B). The diffusion approach gives a slight overestimate for the rates of the reactions A+B and AB+C. The deviation of the decay/formation curves from the Monte Carlo results is always <2% and can be considered as negligible. For reaction schemes (V) and (VI), the results of our hybrid time-diffusion model are indistinguishable from those of the time RRT model. The results do not correspond for scheme (VII).The incorporation of the diffusion approach for the problem AB + C pairs results in more favourable comparison with full Monte Carlo simulation. The A/B decay curves and the AB formation curve show small errors. The AAB/ABB formation curves appear exact. The correspondence is probably fortuitous, representing the cancellation of a small overestimate of the A+B reaction rate and a small underestimate of the AB + C reaction rate. Approach to Track Simulation The ultimate aim of our research is to establish a detailed method that can be used in the simulation of realistic radiation chemistry systems. The model used must be accurate and generally applicable, yet efficient in terms of C.P.U. time. The full Monte Carlo simulation of a high-energy electron track is clearly not feasible. The RRT models are much more favourable and require only one twentieth of the computer time. Within the RRT framework, C.P.U. time efficiency depends upon the approach employed. The accuracy and ease of application of the RRT models also depends on the approximation used. The time approach is of limited use (all initial pairs must be reactive), whilst the diffusion approach and the hybrid time-diffusion approach can be applied in a more general manner. It is likely that one of the latter methods will form the basis of our approach to the simulation of an electron track. References 1 P. Clifford, N. J. B. Green and M. J. Pilling, J. Phys. Chem., 1982, 86, 1318. 2 P. Clifford, N. J. B. Green and M. J. Pilling, J. Phys. Chem., 1982,86, 1322. 3 P. Clifford, N. J. B. Green and M. J. Pilling, J. Phys. Chem., 1985,89, 925. 4 E. Kara-Michailova and D. E. Lea, Proc. Cambridge Philos. SOC., 1940, 36, 101.P. Cliflord et al. 2689 5 J. W. Hunt, Ado. Radiat. Chem., 1976, 5, 185. 6 A. H. Samuel and J. L. Magee, J. Chem. Phys., 1953,21, 1080. 7 L. Monchick, J. L. Magee and A. H. Samuel, J. Chem. Phys., 1957, 26, 935. 8 A. Mozumder and J. L. Magee, Radiat. Res., 1966, 28, 215. 9 H. A. Schwarz, J. Phys. Chem., 1969,73, 1928. 10 W. G. Bums and A. R. Curtis, J. Phys. Chem., 1972,76, 3008. 11 A. Kuppermann, in Physical Mechanisms in Radiation Biology, USAEC 721001 (US. Atomic Energy 12 W. G. Burns, H. E. Sims and 3. A. B. Goodall, Radiat. Phys. Chem., 1984,23, 143. 13 C. N. Trumbore, D. R. Short, J. E. Fanning and J. H. Olson, J. Phys. Chem., 1977,82, 2762. 14 W. Naumann & W. Stiller, Int. J. Radiat. Phys. Chem., 1976, 8, 407. 15 L. Botar, T. Vidoczy and I. Nemes, Proc. 4th Symp. Radiat. Chem. (Hungarian 16 L. Bodr and T. Vidbczy, React. Kinet. Catal. Lett., 1979, 12,485. 17 L. Botir and T. Vidoczy, Chem. Phys. Lett., 1984, 104, 16. 18 N. J. B. Green, D. Phil. Thesis (Oxford, 1982). 19 P. Clifford and N. J. B. Green, Mol. Phys., 1986, 57, 123. 20 N. J. B. Green, Chem. Phys. Lett., 1984, 107, 485. 21 D. A. McQuarrie, J. Appl. Prob., 1967,4, 149. Commission, Washington D.C., 1975), p. 155. Academy of Sciences, Budapest, 1976), p. 1061. Paper 5/ 1789; Received 15th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202673
出版商:RSC
年代:1986
数据来源: RSC
|
12. |
A new equation for the retention of solutes in liquid–solid adsorption chromatography with mixed mobile phases |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2691-2705
Gerhard H. Findenegg,
Preview
|
PDF (1085KB)
|
|
摘要:
J. Chem. Soc., Faraday Trans. I, 1986, 82, 2691-2705 A New Equation for the Retention of Solutes in Liquid-Solid Adsorption Chromatography with Mixed Mobile Phases Gerhard H. Findenegg* and Franz Kostert Physikalische Chemie II, Ruhr- Universitat Bochum, 0-4630 Bochum 1, Federal Republic of Germany A new equation for the dependence of the capacity ratio k’ of solutes on the composition of a binary mobile phase in liquid-solid adsorption chromatography has been derived. Starting from the general definition of the chromatographic retention in terms of surface excess quantities, and then introducing a monolayer model for adsorption from three-component mixtures, a three-parameter relation for k’ in terms of the compositions of the surface layer and the mobile phase is obtained. This equation is tested against experimental data for the adsorption isotherms of two binary solvent systems (cyclohexane-benzene and cyclohexane-l,2,-dichloroethane) on silica gel (p-Porasil) and new measure- ments of the capacity ratio of non-polar and polar solutes (aromatic hydrocarbons and dinitrobenzenes) on the same chromatographic systems.The surface excess isotherms of the binary solvent systems are analysed on the basis of the parallel-layer model to determine the composition of the adsorbed layer. Using that information, the retention data for the solutes are then analysed in terms of the new equation. The resulting parameters (molecular size ratio, rs, interaction parameter, and surface phase fraction, p ) and their physical significance are discussed. In liquid-solid chromatography (LSC) the retention of solutes can vary over wide ranges when the composition of the mobile phase changes.It is now widely accepted that solute retention in LSC is based on a competitive adsorption mechanism of solvent and solute molecules distributed between the bulk liquid and the adsorbed 1ayer.l In recent years theoretical models of the retention in adsorption LSC with mixed mobile phases have been developed, taking into account the non-ideality of solvent-solvent and solvent-solute interactions, differences in size and shape of solvent and solute molecules, and the heterogeneity of adsorption sites on the surface of the sub~trate.~-~ On the other hand, it is now recognized that in LSC the measurable retention is related to the surface excess concentrations rather than the absolute surface concentrations of the solute and solvents, and that surface excess isotherms can be measured directly by chromatographic techniques.6-12 In the present work an attempt has been made to combine the most important factors influencing the retention of solutes in mixed mobile phases to obtain a physically acceptable model, yet to keep the number of parameters as small as possible. Starting from a general definition of retention in terms of surface excess quantities, the concentration profile of solvents and solute as a function of the distance from the solid surface is approximated by a sharp transition from the bulk liquid phase to a surface layer (‘adsorbed phase’), and the adsorption equilibrium is formulated in terms of exchange reactions of solvents and solute between these two phases. Activity coefficients t Present address : Spectra Physics GmbH, SiemensstraBe 20, D-6100 Darmstadt, Federal Republic of Germany.269 12692 L iqu id-Solid Chromatography in the bulk liquid and in the surface layer are defined on the basis of a Flory-Huggins-type lattice model (parallel-layer model).13 In the resulting relation [eqn (20); see later] the capacity ratio of the solute is expressed in terms of the concentrations of the solvents in the two phases (x& xi) with only three relevant parameters, viz. the size ratio, rs, of the solute and solvent molecules, a composite interaction parameter SZ1, and the fraction p of the total column hold-up amount of fluid that is accommodated in the surface layer.A self-consistent set of experimental data is used to test the new equation. Surface excess isotherms of the binary solvent systems on silica gel were determined by the chromatographic frontal-analysis technique as reported elsewhere. lo Measurements of the capacity ratio of several solutes as a function of the composition of these binary eluents are reported in the present work. The chosen solvent systems have one common component (cyclohexane) and differ in regard to deviations from Raoult's law, and the strength of preferential adsorption of the second component (benzene or 1,2- dichloroethane) onto the polar surface. Aromatic hydrocarbons and polar aromatic compounds are used as solutes. A related method of combining chromatographic data with liquid adsorption measurements for examining theoretical models for the capacity ratio in LSC was published by Jaroniec and Okik-Mendyk.14 The present work demonstrates clearly that the retention of the solutes is correlated with the surface concentration of the more strongly adsorbed component of the solvent system, and the new equation accounts for the observed dependence of the capacity ratio on the composition of the mobile phase in a reasonable way.Theoretical General Definition of the Capacity Ratio In analytical chromatography under isothermal and isocratic conditions the retention volume of a small sample of a solute s is given bys where v1 is the molar volume of the eluent, ns is the total amount of that solute in the column and xi is the mole fraction in the mobile phase.The vector xi defines the composition of the eluent before the solute is injected. In LSC the adsorption of the solute is defined in terms of the Gibbs formalism of surface excess quantities by15 yl4n) = nS-xi T/l(n)/vl. (2) Here represents the difference between the amount of solute in the actual column (n,) and in a reference system of uniform composition equal to that of the mobile phase, and containing the same total amount of fluid (ntot) as the column. With this convention for the reference system, the sum of the surface excess amounts nf(n) over all component i of the fluid phase is zero [nNa convention in the terminology of ref. (S)] and, by summation of the expressions for nf(n) over all components, Vn) = v1 n,,,.Hence, from eqn (1) and (2) Introducing the capacity ratio k: of the solute by the definition and using eqn (3) we obtain (4) ( 5 )G . H . Findenegg and F. Koster 2693 where the derivative represents the slope of the surface excess isotherm of the solute. In the limiting linear regime of the isotherm the capacity ratio becomes nu(n) (6) n d n ) k ' = s - S - xt ntot n, - n$n) ' This definition is more general than the conventional definition of k; as the ratio of the amounts of solute in the stationary and mobile phases. Whereas the (absolute) amount of solute in the liquid boundary layer is always positive, the surface excess amount nj5(n) can be positive or negative, depending on whether the solute is preferentially adsorbed or not. Thus, eqn (6) allows for negative values of k;, and such negative values have indeed been reported.ls Capacity Ratio in Elution with Binary Solvents Theories of the capacity ratio in adsorption LSC are commonly based on a two-phase model in which the concentration profile in the liquid boundary layer is approximated by a step function, separating the adsorbed surface phase (a) from the bulk liquid (1).The adsorption equilibrium of the solute s adsorbed from a pure solvent i is then characterized by a displacement reaction r,(i)a + (s)l = rS(i)l + (s)~ where the coefficient rs denotes the ratio of the areas occupied by solute and solvent molecules in the surface layer, a,/ai. The equilibrium condition for this exchange reaction is13 where the x are again mole fractions and the y are activity coefficients in the phases indicated by the superscripts, and ei is the thermodynamic equilibrium constant.If the surface can accommodate na = A,/ai mol of the solvent i (with A , the total surface area), then the surface excess amount is related to x; and x: by Under the conditions of analytical LSC (x, < 1) the second term in the denominator is negligibly small. Thus, from eqn (6) and (8) k: = p [ lim (x;/xl) - 11 xs+o (9) wherep = na/ntot is the surface phase fraction, i.e. the fraction of the total column hold-up amount of fluid that forms the adsorbed phase. From eqn (7) and (9) with the abbreviationf, = (y:/y~),,+o. If the solute is eluted by pure solvent i, as has been assumed so far, then xr/xt --$ 1 and y r / y ; --$ 1, and the capacity ratio is given by k&) = p ( q i f ; i ) - 1) (pure solvent i ) (1 1) where&(,, denotes the ratio of the activity coefficients of the solute s at infinite dilution in pure solvent i.For the capacity ratio of a solute eluted by a mixture of two solvents (i = 1, 2), of which2694 Liquid-Solid Chromatography component 2 is preferentially adsorbed onto the solid, the following relation is obtained by eliminating Ksi between eqn (10) and (1 1 ) for i = 2: As the solute has a very low concentration it will not perturb the adsorption equilibrium of the solvent mixture. Therefore, the ratio of the activities xt yt/x2; y i in eqn (12) can be taken from the adsorption equilibrium of the binary system (1 + 2), in the absence of a solute. The equilibrium condition for the adsorption of a solvent mixture on a homogeneous solid surface has the same form as eqn (7).In the simplest case, when the two types of molecules occupy equal areas (a, = a,), the adsorption isotherm becomes A generalization of eqn (1 3) for molecules of different sizes adsorbed onto a heterogeneous surface that is characterized by a Gaussian distribution of adsorption energies, with a width c, yields an isotherm equation of the form1' where r,, = a,/a, is the size ratio of the two molecules 1 and 2, g,, = y?/(yt)1'r21, and m = RT/c is the heterogeneity parameter. Eqn (14) is based on the so-called condensation approximation which is valid for highly heterogeneous surfaces (m < 1). The importance of surface heterogeneity effects in elution chromatography with mixed mobile phases has been stressed by Jaroniec et ~ 1 .~ 9 l4 and by Rudzinski et aL4 According to eqn (12), however, surface heterogeneity should influence the capacity ratio of solutes only indirectly, through the adsorption equilibrium of the solvent mixture. As the solute is present in very low concentrations these molecules will probe not all types of adsorption sites but only sites with energies E >, E*, for which the adsorption equilibrium constant is g2 [cf. eqn (7)Ie4 Activity Coefficients Eqn (12) is based on the concept of an adsorbed phase in equilibrium with bulk solution, without any specific assumptions about the nature of the adsorbed phase. To express the activity coefficients rP in terms of the concentrations in the two phases we now introduce the parallel-layer model in which the adsorbed phase is assumed to consist of a single monolayer of molecules arranged parallel to the surface.13 For a binary system of similar-sized molecules the activity coefficients of the two components in the liquid and adsorbed phases are expressed in the form (i = 1,2): (15) (16) where ai is the molecular size (a,/a, = r,,), 4; is the volume fraction of i in the mobile phase and I@ is the corresponding area fraction of i in the adsorbed monolayer; x12 is a Flory-type interaction parameter for 1-2 contacts.The above expression for yf results from the assumption that molecules in the adsorbed monolayer interact not only with their neighbours in the surface layer but also with molecules in the adjacent layer of the liquid phase (Ap and Av denote fractions of nearest neighbours to a given lattice site in the same layer and in each of the adjacent layers parallel to the surface). With this definition of the surface activity coefficients the function g,, of eqn (14) becomes In rt = (ai/al) ~12(1- 4; In I? = (ai/al) ~12[43(1 -4f)'++v(l -&I2] g12 = r:/(rt)1'r21 = exp { - x 1 2 Vp(1 - 24;) + AV(1 - 24311. (17)G .H. Findenegg and F. Koster 2695 Extension of this formalism to three-component systems consisting of two solvents (1 and 2) of equal-sized molecules (a, = a2 = a) and a solute s with a size parameter rs = a,/a yields for the activity coefficient of the solute in the mobile phase In Y: = r s k s 4: +x2s 4; - x 1 2 4: 4 3 (18 a) where xis is the interaction parameter for a solvent molecule i and one of the rs segments of a solute molecule.At infinite dilution of the solute in a binary solvent (&+& = 1) eqn (18a) reduces to with J2, = xZs -xis -x12, and at infinite dilution in one of the pure solvents (i = 1,2) ( W In Y;?) = YsXis. ( W In Y P = r s k 1 s +J21 xB +x12(m21 In the adsorbed layer, the activity coefficient of the solute is again made up from contributions of the interactions with neighbours within the adsorbed layer and with molecules in the adjacent layer of the liquid phase. For simplicity we assume that the Flory parameters for these two types of interactions are the same as the Flory parameter in the bulk solution. The surface phase activity coefficient of the solute at infinite dilution in the mixed solvent then becomes, by analogy with eqn (18 b), In Y Y = ~ s ~ , [ x l s + J 2 1 ~ ~ + X 1 2 ~ ~ ~ ~ 2 1 + ~ s Avkls+Jzl x;+x12(x;)21- In YE% = r d l p + Av 1 Xis- (194 (19b) At infinite dilution in one of the pure solvents (i = 1, 2) this expression reduces to Eqn (18 b), (1 8 c), (19 a) and (19 b) yield an explicit expression for the activity coefficient ratiofs(,?/fs in eqn (12).However, a more compact form of that equation is obtained by combining all activity coefficients on the right-hand side of eqn (12), using eqn (15), (1 6), (1 8) and (19) for the special case of equal-sized molecules (rZl = l), when in eqn (15) and (16) the volume fractions can be replaced by mole fractions. Finally, by specifying the lattice geometry in the usual way (Ap = 1/2, Av = 1/4), we obtain k; + P = K ( 2 ) +P) [(x;/x;) F(x2 )ITS (20) with where we use the abbreviation S,, = y 2 s - ~ l s + ~ 1 2 .Eqn (20) is a generalization of the treatments of Boehm and Martire2 and Bor~wko,~ taking into account differences in molecular size of solvent and solute molecules (parameter rs), and starting from the definition of the capacity ratio as a relative surface excess amount [eqn (6)]. In the form presented here, eqn (20) can be tested experimentally if the capacity ratio of solutes has been measured over a range of compositions of the binary eluent, starting from pure solvent 2 (but not necessarily including pure solvent I), and if the surface composition xg =Ax;) is known. Alternatively, eqn (20) can be written in a symmetrical form, containing the capacity ratio k&) in both pure solvents (i = 1,2); but that form is less suitable for experimental verification, as k;,,, (component 1 being the less effective eluent) may be too large to be measured accurately.F(x2) = exp [ - S2,( 1 + 2% - 3xi)/4] (21) Experiment a1 Method The chromatographic apparatus has been described previously.l0l l8 For the present elution measurements a Knauer variable-wavelength monitor has been used as a detector. Experimental flow rates (ca. 1 cm3 min-l) were measured with a precision of ca. 0.15% and the resolution in the retention times was ca. 0.05% ; hence the relative2396 Electroactivity of Polyaniline The second oxidation reaction occurs during the second half of peak I to give the dark green protonated form of emeraldine : + + f .( C 6 H 4 ) - N ( H ) - ( C 6 H 4 ) - N ( H ) ~ (C6H4)-N(H)=(C6H,)= N(H)+x + 2x c1- c1- I cl- f(C6H,)-N(H)-(C6H4)-N(H)~(c6H4)-~(H)=(c6H4)= N(H) f 2 ~ + 2x e-. (9) Cl- c1- The electrochemical reduction reactions are the reverse of those given above. These reactions occur during the time taken for a single cyclic voltammetric scan (ca. 24 s for the oxidation step). This study does not indicate whether the protonated units15 would spontaneously undergo deprotonation with the loss of HCl and be converted to + ~(c6H4)-N=(c6H4)=N(H)~ (2s’) or [-(C6H4)-N=(C6H,)=N-] (2A) c1- units if the polymer were permitted to stand in the electrolyte for an extended period in the absence of an applied potential. However, our previous investigationsR.l4 based on elemental analyses of material dried under dynamic vacuum for ca. 48 h show clearly that equilibration of emeraldine base with aqueous HCI for 55 h results in protonation of ca. 25% of the nitrogen atoms at a pH of 1 and ca. 50% of the nitrogen atoms at a pH of - 0.2 (1 .O mol dm-3). At pH values > 4 there is essentially no protonation. Hence spontaneous deprotonation cannot be extensive in the more acidic solutions, even on standing for an extended period in the absence of an applied potential. Note that the highly conducting form of polyaniline, i.e. the emeraldine salt, can be formed in two different ways as depicted below: leucoemeraldine base (insulator) oxidation (no protonation or deprotonation) [-42 e-+45 A-1 A- I f N(H)-(c6H4)-N(H>=(c6H4)=~(H)-(c6H4)-N(H)-(c6H4) 325 A- emeraldine salt (conductivity in metallic regime) protonation (no oxidation) In reaction (10) the emeraldine salt is formed by oxidation involving no change in the number of hydrogen atoms attached to nitrogen atoms.In reaction (1 1) it is formed by protonation with no accompanying change in the formal oxidation state of the polymer.G. H. Findenegg and F. Koster 2697 OO Fig. 1. Reduced adsorption of component 2 per unit mass of solid, n$%) in mmol g-l, us. mole fraction in the bulk liquid phase, xi, of the binary systems cyclohexane-benzene (C-B) and cyclohexane-l,2-dichloroethane (C-D) on silica gel at 25 "C. Experimental data : A, C-B; 0, C-D. The curves represent the monolayer isotherm equation (14) with different parameter sets as given in table 1 [lines (a)-(e)].'0 0.5 1 Fig. 2. Surface concentration $ us. bulk concentration x!, for the binary systems cyclohexane- benzene (C-B) and cyclohexane-1,2-dichloroethane (C-D) on silica gel. +, C-B; x , C-D: from surface excess data of fig. 1 by eqn (22), with na and rZ1 as given in table 1 [lines (a) and (41; curves from eqn (14) with the same parameter sets as in fig. 1 (see table 1). which yields na = a,/ai = 1.86 mmol g-l, and r,, = a2/al = 0.878 (system C-B) and 0.810 (system C-D). Fig. 2 shows a plot of the surface concentration Xa; as a function of the concentration of the bulk liquid. The points were derived from the experimental surface excess data by eqn (22), with na and r2, values as given in table 1 [lines (a) and ( 4 1 - The curves shown in fig.1 and 2 were obtained by fitting the surface excess data by eqn (14) in combination with eqn (22), and with the activity coefficients as given by eqn (1 5). The exchange energy parameter xlz was found by fitting eqn (1 5) to literature data2698 Liquid-Solid Chromatography Table 1. Parameters for adsorption isotherms [eqn (14) and (22)] of the binary solvent systems cyclohexane-benzene (C-B) and cyclohexane-l,2-dichloroethane (C-D) on silica gel at 25 "Ca system na r21 XlZ K21 m CJ C-B (a) 1.86 0.878 0.373 2.50 0.921 0.007 C-B (b) 1.98 1 0.373 2.36 0.910 0.004 C-B (a') 1.86 1.15 0.373 2.58 0.933 0.010 C-B (b') 1.73 1 0.373 2.71 0.923 0.005 C-D (c) 1.86 0.810 1.078 4.21 0.925 0.029 C-D (e) 2.06 1 1.078 4.15 1.037 0.036 C-D ( d ) 2.06 0.810 1.078 3.60 0.901 0.029 a Best-fit values of the equilibrium constant K,, and heterogeneity parameter rn for different choices of the monolayer capacity na/mmol g-l, size ratio r,, and exchange energy parameter x,, (see text). CJ is the standard deviation in the reduced areal adsorption n;(n)/mmol g-l. for the activity coefficients of the binary bulk systems [see ref.(lo)]. This bulk value of x12 was then used also in eqn (16) for the activity coefficients in the adsorbed layer (with A, = 1/2 and A, = 1/4). For given na, r21 and x,,, the remaining parameters of eqn (14), the adsorption constant K,, and the heterogeneity parameter m, were determined by a least-squares fitting procedure. The resulting sets of parameters with the respective standard deviations are summarized in table 1.For the system C-B, the monolayer isotherm equation (14) gives a good representation of the experimental surface excess data. The curves shown in fig. 1 and 2 are based on the following choices of the geometrical parameters na and rzl (see table 1): (a) na from the molar area a, and r,, = a,/a,; (b) na from the mean molar area (a, + a2)/2 and r21 = 1. These two sets of parameters fit the experimental surface excess data nearly equally well (fig. I), and the surface concentrations =Ax;) are almost indistinguishable (fig. 2). However, the above formula for ai may be unrealistic when applied to non-spherical molecules like benzene. Application to benzene gives 1.85 x lo5 m2 mol-l or 0.307 nm2 per molecule, which is significantly less than the value of 0.40 nm2 per molecule that is often used in interpreting the adsorption of benzene by solids.20 The two sets of parameters corresponding to this choice for benzene are denoted as (a') and (b') and are also given in table 1.The surface concentrations xi resulting from (a') and (b') are nearly indistinguishable, but a little higher than those for (a) and (b), yet the difference between (a') and (a) is less than between the isotherms (c) and ( d ) in the system C-D. The system C-D exhibits much larger deviations from Raoult's law than the system C-B. Furthermore, this system is less symmetrical in regard to the size ratio r21, and component 2 is more strongly preferentially adsorbed onto silica gel than in the system C-B. In fig. 1 and 2 the curves for the system C-D are based on the following choices for na and r2, (see table 1): (c) na from the molar area a, and rZ1 = a,/al; ( d ) na ca.10% greater than the estimated true monolayer capacity; (e) na as before and rzl = 1. The parameter sets (c) and (d), which are based on a physically realistic size ratio r21 but different na values, fit the experimental surface excess data equally well (cf. fig. l), but yield different surface concentrations xg, as shown in fig. 2. From studies of multilayer models of adsorption from binary mixtures2l9 22 there is evidence that adsorption in the second layer becomes significant for systems with large positive deviations from Raoult's law. It is to be assumed, therefore, that the surface excess amount of the system C-D contains a contribution due to adsorption in the second layer.On the basis of our monolayer model, such multilayer effects can be accounted for only indirectly byG. H. Findenegg and F. Koster 2699 adopting an na value greater than the estimated true monolayer capacity. Therefore, the parameter set ( d ) of table 1 is believed to yield more realistic surface concentrations than set (c). In contrast to the former system, the ' symmetrical' parameter set (e), with r2, = 1, does not fit the experimental surface excess isotherms as well as the parameter sets (c) and ( d ) , which account for the fact that the molar area of the strongly preferentially adsorbed component 2 of this system is ca. 20% smaller than the molar area of component 1. For both binary systems a heterogeneity parameter rn = 0.91 is found on the basis of eqn (14) if realistic values of na, r,, and x12 are chosen.However, the condensation approximation on which eqn (14) is based is valid only for rn < 1, which is clearly not the case here. Therefore, the interpretation of rn as a heterogeneity parameter is unjustified in our case. In fact, for rn = 1 (corresponding to a substantial heterogeneity) and r2, = 1, eqn (14) becomes identical with eqn (13) which applies to adsorption from regular solutions onto a homogeneous surface. Capacity Ratios In eqn (20) the dependence of the capacity ratio of solutes on the composition of a binary eluent is described essentially by the function ( ~ / X ; ) ~ S , while the factor F(x2) accounts for deviations from ideality, being either greater or less than unity depending on the sign of the interaction parameter S2,.The function xi/xl resulting from the isotherm equation (14) with the parameters for the present solvent systems (table 1) is shown in fig. 3. The function l/xi (corresponding to x i = 1 for all xt, i.e. infinitely strong preferential adsorption of component 2) is shown for comparison. Fig. 4 illustrates the influence of the size parameter r, and the interaction parameter S,, on the function [(xt/xi) F(x)]'s. For a quantitative analysis of the present k; data the parameterp = ?/n,,, (the surface phase fraction) was estimated from the dead-space volume of the column, VM, the mean molar volume, 8, and the corresponding mean molar area, ii, of the binary solvent system, and from the total surface area of the solid in the column (rn, a,).The resulting value, p = 0.08, is of similar magnitude as our k; data (0.017 < k; < 0.45), and therefore p cannot be neglected in eqn (20). Fig. 5 illustrates the influence of the function x;/xi on the calculation of the capacity ratio. The experimental data show k; for phenanthrene eluted by the solvent system C-D. The curves were obtained by eqn (20) and are based on the surface concentration isotherm x; =Ax;) given by eqn (14), with the parameters of table 1. For all curves, p = 0.08 and S,, = 1.1 has been chosen (see below), and the best-fit value of the size parameter r, was calculated by an iterative procedure. In this and in most other cases, the best fit of the experimental ki data is obtained when representing the surface concentration isotherm of the binary eluent by the physically most realistic parameters [set ( d ) of table 1, for the binary eluent C-D].However, the higher surface concentration isotherms (c) and (e) yield quite similar curves for k; as a function of xi (see fig. 5). The values of the size parameter obtained on the basis of these three isotherms are similar too (r, = 1.92 _+ 0.05). Much larger or much lower values of r, can be obtained, however, if these realistic surface concentration isotherms are replaced by some strongly deviating x;/xi function. For example, the surface concentration isotherm (a) (weaker preferential adsorption of component 2) or the isotherm 1 /xi (infinitely strong preferential adsorption of component 2) yield r, = 2.8 and 0.9, respectively, instead of a value near 2.0 based on the isotherm (d).Incidentally, if in eqn (20) the function x;/xi is approximated by l/xi [curve df) in fig. 31, and if S,, and p are both taken as zero, our equation reduces to the simple Snyder-Soczewinski (SS) equation, which gives a poor fit of the present experimental data, with rs close to unity for the aromatic hydrocarbon solutes of table 2 . kfi = k&,, ( l / ~ l ) ~ s (23)2700 4 3 .;I.: 2 Liquid-Solid Chromatography curve rs S,, 1 2 0 2 1.5 0 3 2.5 0 4 2 1 5 2 -1 Fig. 4. The function [ ( x t / x i ) F(x2)Irs us. mole fraction xi for typical combinations of the parameters rs and S21. The curves shown are based on the surface concentration isotherm =Ax;) for the system C-D [curve (d) in fig.1-31, Table 2 summarizes the results of an analysis of all experimental capacity ratio data in terms of eqn (20). The surface concentration isotherms of the two binary eluents are represented again by eqn (14), with the physically most realistic parameter sets (a) and ( d ) of table 1 . The solute-specific parameters rs and S,, have been obtained by an iterative parameter-fitting procedure, and are given in table 2 together with the respective least-squares standard deviation of the fit. Fig. 6 shows the experimental k: data of three solutes eluted by the binary solvent system C-B. While the aromatic hydrocarbons (naphthalene and phenanthrene) have small and nearly constant capacity ratios at high benzene concentrations, the capacity ratio of the polar dinitrobenzene is distinctly greater and increases more steeply as xi is decreased.Fig. 7 compares the capacity ratio ofG. H. Findenegg and F. Koster 270 1 b I Fig. 5. Capacity ratio k; of phenanthrene us. mole fraction x i of the binary eluent cyclohexane-l,2- dichloroethane : 0, experimental data. Curves represent eqn (20) with different functions g / x i (notation as in fig. 3); S,, = 1.1, p = 0.08, and best-fit values of rs (see text). Table 2. Experimental capacity ratios of solutes eluted by pure solvents, k;,,, and k;,,,, and best-fit values of parameters rs and S,, of eqn (20) for elution with binary eluents (silica gel, 25 "C) solute eluent : cyclohexane (1)-benzene (2)a 1,2-dinitrobenzene 1,3 -dini trobenzene 1,4-dinitrobenzene anthracene phenanthrene fluorene diphenyl naphthalene - 0.45 1 - 0.429 - 0.397 4.66 0.049 3.65 0.053 2.93 0.034 3.14 0.040 1.67 0.045 2.5 -1.3 2.5 -1.2 2.5 -1.0 2.5 0.7 2.4 0.7 2.4 0.6 2.3 0.7 1.9 0.7 eluent: cyclohexane (1)-1,2-dichloroethane (2)a 1,2-dini tro benzene 1,3-dini tro benzene 1,4-dinitrobenzene ant hracene phenant hrene fluorene diphenyl naphthalene - 0.123 0.178 - 0.172 4.66 0.017 3.65 0.017 2.93 0.017 3.14 0.017 1.67 0.017 - 2.3 -0.7 2.2 -0.7 2.2 -0.6 2.1 1.3 2.0 1.2 1.9 1.1 1.8 1.2 1.5 1.1 0.036 0.049 0.027 0.032 0.024 0.025 0.026 0.013 0.0 10 0.012 0.022 0.032 0.024 0.02 1 0.026 0.017 a g / x i from eqn (14) with parameter set a (system C-B) and d (system C-D) of table 1; p = 0.08, 0 is the standard deviation in kk.2702 Liquid-Sol id Chromatography Fig.6. Capacity ratio ki of (a) 1,2-dinitrobenzene, (b) phenanthrene and (c) naphthalene us. mole fraction xt of the binary eluent cyclohexane-benzene: 0, 0, a, experimental data. Curves show fits by eqn (20): (-) best-fit values for rS and S,, (table 2); (. . . . . .) S,, = 0 and p = 0, Y , = 4.3 (a), 2.7 (b), 2.3 (c). 4 3 k: 2 1 0.5 1 xi "0 Fig. 7. Capacity ratio ki of 1,4-dinitrobenzene us. mole fraction xi of the binary eluents (a) cyclohexane-benzene (C-B) and (b) cyclohexane-1 ,Zdichloroethane (C-D) : A, 0, experimental data. Curves show fits by eqn (20): (-) best-fit values for r, and S,, (table 2); (----) S,, = 0, p = 0.08, r, = 4.0 (C-B), 2.8 (C-D); (. . . .) S,, = 0 and p = 0, Y, = 4.3 (C-B), 3.1 (C-D). 1,4-dinitrobenzene in the two solvent systems, and fig.8 shows a similar plot for naphthalene. For both solutes, the k;,,, value in pure benzene is higher than in pure 1,2-dichloroethane, reflecting the fact that the latter solvent competes more effectively for adsorption sites, and is more strongly adsorbed from the mixed eluent than benzene (see fig. 2). In fig. 6-8, the full curves represent the best fit of the k: data by eqn (20), with the parameters given in table 2. The dashed and dotted curves show the influence of assuming S,, = 0 and p = 0, as explained in the figure captions. Some important results of the present analysis are summarized below.G. H . Findenegg and F. Koster 2703 0.6 iB Fig. 8. Capacity ratio kk of naphthalene m. mole fraction xi of the two binary eluents.Similar plots as in fig. 7, with best-fit values for rs and S,, (table 2). Size Ratio Parameter For the individual solutes of table 2, the values of the size ratio parameter rs resulting from the capacity ratio data in the two solvent systems agree within reasonable limits ( f 15 % of the mean of the two values). This consistency of the resulting parameters is significant because the two solvent systems exhibit distinctly different adsorption isotherms, i.e. different functions xt/xf (cf. fig. 3), and rs is the exponent to this function in eqn (20). Furthermore, the resulting rs values (e.g. 1.5-1.9 for naphthalene and 2.0-2.5 for anthracene, phenanthrene and dinitrobenzenes) are compatible with the estimated ratio of the areas occupied by adsorbed solute and solvent molecules.Interaction Parameter Positive values of the interaction parameter S,, are found for the aromatic hydrocarbon solutes, but negative values for the dinitrobenzenes, in both solvent systems. From the definition S,, depends on the sign and the magnitude of the binary interaction parameters solvent-solute klS and x,,) and solvent-solvent k12). For the present solvent systems xlz is positive (see table 1); the parameters xlS and xZs are expected to be positive too, and the quantity xZs -xis should be of similar magnitude for a class of chemically related solutes (like our aromatic hydrocarbons) in a given solvent system. Furthermore, as the present solutes are less soluble in cyclohexane (component 1) than in benzene or 1,2-dichloroethane (components 2 of the solvent systems), xlS should be greater than x , ~ , and we can expect the binary interaction parameters to follow the order S21 = x2s -x1s +x12 XlS ’ x2s = x12 > 0- Correspondingly, the negative S,, values found for the dinitrobenzenes imply that for these solutes xlS is significantly greater than xZs kls > xzs+xI2), whereas the positive S,, values for the aromatic solutes imply that xlS is not much greater thanx,, kls < xZs +x12).For all solutes the S,, value in the solvent system C-D is ca. 0.5 units greater than in the solvent system C-B (see table 2). This increment is of similar magnitude as the2704 Liquid-Solid Chromatography difference between the x12 values of the two systems (viz. 1.08-0.38 = 0.70) and can be rationalized by the conjecture that for the present solutes X , ~ - X ~ , has similar values in the two solvent systems.Although a systematic test of the new equation will require more experimental data on a variety of systems, the above qualitative discussion suggests that positive and negatives values of S,, are indeed to be expected, depending on the chemical nature of the solutes and solvents. On the other hand, if deviations from ideality are entirely neglected by assuming S,, = 0, then the experimental data can be fitted only by physically unrealistic values of the size parameter r,. Assuming S,, = 0 and p = 0 yields relatively good fits for those solutes for which a positive S,, had been found, but fails in those cases in which S,, had been negative (see fig. 6). Conclusion The present work has shown that the capacity ratio of solutes in adsorption LSC with binary mobile phases can be explained in terms of a monolayer model of adsorption from three-component systems.The capacity ratio of the solute depends on the adsorption equilibrium of the solvent system ( x t / x i ) . It is strongly affected by the ratio rs of the areas occupied by solute molecules and molecules of the preferentially adsorbed solvent, and by the interaction parameter S,,, which represents a linear combination of the three binary Flory parameters xii for solvent-solute and solvent-solvent interactions. Furthermore, a phase ratio parameter p appears in the theory, as a consequence of the fact that the capacity ratio is defined in terms of surface excess quantities.If the adsorption isotherm of the solvent system is accounted for with sufficient accuracy, the new equation gives a good representation of the experimental k; data, and the resulting parameters rs and S,, have physically reasonable values. Simplified versions of the theory (e.g. neglecting the non-ideality of the system) can still represent the experimental data reasonably well in favourable cases, but the values of the (remaining) parameters can no longer be interpreted in terms of a physical model. Financial assistance from the Minister fur Wissenschaft und Forschung des Landes NRW for this research is gratefully acknowledged. We also thank Dr U. Deiters for much help with the numerical analysis and useful discussions. References 1 L. R. Snyder and H. Poppe, J . Chromatogr., 1980,184, 363. 2 R. E. Boehm and D. E. Martire, J. Phys. Chem., 1980,84, 3620. 3 M. Borowko, J. Colloid Interface Sci., 1984, 102, 519. 4 W. Rudziriski, J. Narkiewicz-Michalek, Z. Suprynowicz and K. Pilorz, J. Chem. SOC., Faraday Trans. 5 M. Jaroniec, D. E. Martire and M. Bbrowko, Adv. Colloid Interface Sci., 1985, 22, 177. 6 J. F. K. Huber and R. G. Gerritse, J . Chromatogr., 1976, 58, 137. 7 H. L. Wang, J. L. Duda, C. J. Radke, J. Colloid Interface Sci., 1978, 66, 1277. 8 F. Riedo and E. Kovhts, J. Chromatogr., 1982, 239, 1. 9 N. Le Ha, J. Ungvaral and E. Kovats, Anal. Chem., 1982, 54, 2410. I, 1985, 81, 553. 10 F. Koster and G. H. Findenegg, Chromatographia, 1982, 15, 743. 11 W. Markowski, K. Czapinska and H. Poppe, Chromatographia, 1983, 17, 221. 12 J. Jacobson, J. Frenz and C. Horvath, J . Chromatogr., 1984, 316, 53. 13 D. H. Everett, Trans. Faraday Sac., 1965, 61, 2478; S. G. Ash, D. H. Everett and G. H. Findenegg, 14 M. Jaroniec and B. Oicik-Mendyk, J. Chem. Soc., Faraday Trans. I , 1981,77, 1277. 15 D. H. Everett, Pure Appl. Chem., 1972, 31, 579. 16 K. Groh and I. Halasz, J . Chromatogr., 1980, 199, 23. 17 W. Rudzinski, J. Zajac and C. S. Hsu, J. Colloid Interface Sci., 1985, 103, 528. 18 F. Koster, Doctoral Thesis (Ruhr-Universitat Bochum, 1984). 19 K. K. Unger, Porous Silica, J. Chromatogr. Library, vol. 16, (Elsevier, New York, 1979). Trans. Faraday SOC., 1968, 64, 2639.G. H. Findenegg and F. Koster 2705 20 D. H. Everett and R. T. Podoll, in Colloid Science, ed. D. H. Everett, (Specialist Periodical Report, The Chemical Society, London, 1979), vol. 3, chap. 2. We are grateful to one referee for this criticism. 21 S. G. Ash, D. H. Everett and G. H. Findenegg, Trans. Faraday Soc., 1968,64, 2645; 1970, 66, 708. 22 W. Rudzinski, J. Narkiewicz-Michalek, K. Pilorz and S. Partyka, J . Chem. SOC., Faraday Trans. 1,1985, 81, 999. Paper 511796; Received 8th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202691
出版商:RSC
年代:1986
数据来源: RSC
|
13. |
Surface acidity of some Re2O7-containing metathesis catalysts. Anin situFourier transform infrared study using pyridine adsorption |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2707-2718
Xu Xiaoding,
Preview
|
PDF (825KB)
|
|
摘要:
J . Chem. SOC., Faraday Trans. 1, 1986,82,2707-2718 Surface Acidity of some Re,O,-containing Metathesis Catalysts An in situ Fourier Transform Infrared Study using Pyridine Adsorption Xu Xiaoding,* Johamnes C. Mol and Cornelis Boelhouwer University of Amsterdam, Institute of Chemical Technology, Nieuwe Achtergracht 166, 1018 VW Amsterdam, The Netherlands Surface acidity of several metathesis catalysts, Re,O,/Al,O,, M, 0, - Re,O,/Al,O, (M, 0, = V20,, MOO, or WO,) and Re,O,/ SiO;Al,O,, have been studied in situ by Fourier transform infrared spectroscopy using pyridine adsorption. A distinct correlation has been found between the activity for metathesis and Bronsted acidity of the catalysts, while their Lewis acidity showed no correlation with the activity. Surface acidity is an interesting subject in heterogeneous catalysis since it is closely related to the activity of many acid-base catalysed reactions, e.g.isomerization, polymerization and catalytic cracking.l-* Although alkene metathesis is not commonly regarded as an acid-catalysed reaction, there are indications that in some cases surface acidity may have an important impact on the metathesis activity of some heterogeneous catalyst ~ystems.~-l* Treating alumina with strong acids can improve the overall metathesis activity of some catalyst^.^. Low-loading Re,O,/Al,O, catalysts prepared by impregnation with NH,ReO,/HCl show much higher activity in the metathesis of methyl oleate [(Z)-methyloctadec-9-enoate] than the corresponding Re,O,/Al,O, catalysts prepared without HCl.9 Ellison et ~ 1 .~ 1 ~ reported that 5 wt % Re,O, catalysts supported on various alumina carriers showed negligible activity except for one distinct catalyst that possessed some acidity at that rhenium loading. Several metathesis catalyst systems were reported, viz. Re,O,/SiO, * Al,O,1° and M, 0, * Re,O,/Al,O, (M, 0, = V,O,, MOO, or wo3),73 11-13 which show much higher activities for metathesis than the corresponding Re,O,/Al,O, catalyst systems. For these catalyst systems the high activity has been attributed to their high acidity, especially the Bronsted acidity.1° Up to now the acidity of Re,O,/Al,O, catalysts has not been studied thoroughly.l*? l5 Our knowledge of the acidity of Re,O,/SiO, - Al,O,1° and M, 0, - Re,O,/A1,O,ll catalysts is practically unexplored.The aim of the present research is to study the acidity of these metathesis catalysts in more detail. The results are discussed in the light of their activity in alkene metathesis. Various methods have been developed to measure surface acidity, e.g. amine titration using distinct indicators, and i.r. measurements with a base probe.l-* With the amine titration method using different indicators, information about the total acidity as well as about the distribution of acidities of various strengths can be obtained.l7 However, this method cannot distinguish Lewis acidity from Bronsted acidity. 1.r. measurements using a base probe molecule allow determination of both Lewis acid sites and Bronsted acid sites at the same time.16-21 We applied the latter method. Pyridine, a weak base (pK, M 9),16 was chosen as the base probe so that strong acid sites which may be involved in metathesis, could be measured.Since Parry's article was published,16 the i.r. study of surface acidity by pyridine adsorption has become a classic method to study surface acidity.'-"* Various surface 27072708 Re,O,-containing Metathesis Catalysts species, viz. hydrogen-bonded pyridine (HPy), physisorbed pyridine, coordinatively bonded pyridine (LPy) and pyridinium ions (BPy) can be determined from an analysis of i.r. bands in the ‘ring’ region (1400-1700 ~m-l).l-~9 l6 The i.r. bands of LPy sites appear at ca. 1620, 1580, 1490 and 1450 cm-l, while those of the BPy sites appear at ca. 1640, 1620, 1540 and 1490 cm-l.16 The band at ca. 1540 cm-1 is characteristic of Bronsted acid sites, while the one at ca.1450 cm-l is characteristic of Lewis acid sites, provided there is no interference from other surface The exact band positions, however, depend among other factors on the pretreatment, the conditions during the measurement and the nature of the samples. From the band heights or integrated areas of the i.r. bands, quantitative data of these species can be ~btained.l-~ Evacuation at elevated temperatures (> 423 K) can remove the H-bonded and physi- sorbed pyridine.l6> 1 7 9 2o By varying the temperature of evacuation after pyridine adsorp- tion and by varying the method of pretreatment of the samples, information about the strength of Lewis acidity can be obtained from the ease with which the adsorbed species can be pumped off and from the frequency shift of the band at ca.1450 cm-l.l-* Experiment a1 Fourier transform infrared (FTIR) spectra in the region 400-4000 cm-l were recorded on a Nicolet 7199 FTIR spectrophotometer with a liquid-nitrogen cooled HgCdTe detector. The resolution used was 2cm-l and the number of scans was 32 for each spectrum. The i.r. cell has been described el~ewhere.~ A pair of NaCl windows (35 mm) was used and silicon-rubber rings were fixed between the cell and the windows at both sides of the windows to ensure that the metal cell was vacuum tight under the conditions used. The self-supporting wafer was mounted on a metal sample holder which was located at the centre of the cell. A thermocouple was inserted inside the cell and positioned near the wafer.During the measurements water flowed through the cooling jacket of the cell to prevent the window part from overheating. The temperature of the cooling water was maintained by a thermostat at a temperature sufficiently high to prevent moisture from condensing on the windows (303-323 IS). The heating of the cell was monitored by a proportional temperature controller. The cell could be moved vertically in order to measure the background spectra and the sample spectra, separately. By ratioing the sample energy spectrum to the background energy spectrum, recorded under the same conditions, the proper i.r. spectrum was ~btained.,~ The i.r. cell was connected with a conventional gas and vacuum system, which provided the possibility of evacuation (up to ca.0.1333 Pa), the admittance of various gases (He, 0,) and substrates (e.g. pyridine) and the introduction of a liquid reactant via the septum of an injection port. The oxygen used was dried by 5A molecular sieves. The helium was dried and deoxygenated by activated molecular sieve 5A and a heated Cu/A1,0, catalyst. The pyridine used was distilled under reduced pressure after being dried by molecular sieves and only the middle-cut was used for the experiments. The catalysts used to prepare the self-supporting wafers were prepared by a wet incipient impregnation method, as described earlie~,~-ll and were ground to a particle size suitable for i.r. purposes. The i.r. wafers were pressed at 5 ton (on the ram) for a period of 1 min. The diameter of the wafers was 13 mm and their weights ca.10-20 mg. In pretreating the catalyst wafers, they were first heated at 773 K in a dry oxygen stream overnight, followed by heating at the same temperature for 2 h in a dry and deoxygenated helium stream. After cooling to 423 K and evacuation, a small amount of pyridine (up to a pressure of 4666Pa) was admitted into the system at that temperature. The sample was kept in the pyridine atmosphere at 423 K for 1 h and the system was then evacuated at 423 K for 1 h. The temperature was raised stepwise (steps of 50 or 100 K) during evacuation till 773 K was reached. Spectra were recorded afterX . Xiaoding, J . C. Mol and C . Boelhouwer 2709 stabilizing the conditions at each temperature for 0.5 h. Integration of the peak areas was conducted by using either the integration program of the FTIR, according to the usual tangent baseline method,17 or, in the case of overlapping peaks, by a cut-and-weigh method.The peak positions of the i.r. bands were picked up via a computer program of the FTIR spectrometer. In some cases distinct amounts of water were admitted to the system via the injection port after the sample was heated in vacuum to 773 K and cooled down to 423 K to see the effect of adding water. The activity for the metathesis of the catalysts were measured either in the liquid phase in glass batch reactors (for the metathesis of methyl oleate9-11), or in a fixed-bed microcatalytic flow reactor system at 323 K at a propene pressure of 0.05 MPa (for the metathesis of propeneg* 24) as described earlier.Results Fig. 1 shows the i.r. spectra of catalysts containing 0, 3, 6, 12 and 18 wt % Re,O, after activation in pyridine at 423 K for 1 h. The positions of the i.r. bands are given in table 1. It appears that the 12 and 18 wt % catalysts show a peak at 1539 cm-l and a shoulder peak at 1637 cm-l. The band at ca. 1540 cm-l remained on the surface even at 773 K in vacuum. Although from the spectra in fig. 1 no peak at 1539 cm-l can be observed for the 6 wt % Re,07/A1,0, catalyst, the subtracted spectrum (the spectrum after 1701 1611 1521 1431 1341 v/cm-' Fig. 1. 1.r. spectra of Re,O,/Al,O, catalysts after activation and pyridine adsorption at 423 K.2710 Re,O,-containing Metathesis Catalysts Table 1. Positions of the i.r. bands for various Re,07/A1,0, catalysts at 423 K after pyridine adsorption wt% Re207 band positions/cm-l 0 - 1623, 1614, 1596,a 1577, - 1493, 1451.3 - 1622, 1616, 1592,a 1577, - 1494, 1451. 6 - 1622, 1616, 1592,a 1576, 1539,b 1494, 1455. 12 18 1637, 1623, 1617, 1594,a 1575, 1539," 1494, 1455. 1636, 1622, 1616, 1593,a 1575, 1539,c 1494, 1455. a Only observed on spectra in pyridine. Obtained from the subtracted spectrum at 423 K in vacuum after and before pyridine adsorption. This i.r. band was observed at 773 K in vacuum at 1544 and 1545 cm-', for 12 and 18 wt "/, Re,O,/Al,O, catalysts, respectively . I I I I 1701 1633 1565 1497 1429 v/crn-' Fig. 2. 1.r. spectra of 3 wt % Re,O, catalysts on (a) HA-, (b) LAH-, (c) LAL-silica-alumina carriers and ( d ) on y-alumina. pyridine adsorption and evacuation minus the spectrum before pyridine adsorption in vacuum) did show a small peak at 1539 cm-l.The integrated areas of the ca. 1540 and ca. 1450 cm-1 bands were used to calculate Bronsted and Lewis acidity, respectively.20 A 6 wt % Re20,/A1203 catalyst prepared by HCl impregnation was also measured. A weak peak at ca. 1550 cm-l was observed at 423 K in vacuum. The addition of water ( 5 x cm3) to the sample increased the intensity of this peak.X . Xiaoding, J . C. Mol and C. Boelhouwer 271 1 Table 2. Band positions" of SA-supported Re207 catalysts SA LAH LAL HA HA HA HA HA wt% Re207 i.r. band positions/cm-l 1636 1622 - 1544 1491 1636 1622 - 1545 1491 1638 1622 - 1545 1491 1637 1622 - 1545b 1491 1637 1621 1545d 1490 1637 1622 1612c 1545e 1491 1640 1622c 1610 1545 1489 1455 1455 1455 1454 1454 1454 1454 a The spectra were recorded at 423 K after pyridine adsorp- tion and evacuation for 1 h at 423 K.1451 cm-l at 673 K. Shoulder. 1546 cm-l at 773 K. 1547 cm-l at 773 K. Table 3. Weight normalized peak areas of Lewis acid sites, Bronsted acid sites and the ca. 3740 cm-l band of SA-supported catalysts with different Re207 wt % at 423 K in vacuum after pyridine adsorptiona SA wt% Re207 LAH LAL HA HA HA HA HA Bronsted acidity A/cm-l g-l 46.3 29.0 65.6 99.8 100.2 142.3 158.0 Lewis acidity A/crn-l g-l OH (3740 cm-l) Alcm-l g-l 148.6 100.7 205 262 236 171 122 - 952 862 710 555 375 a All obtained by an FTIR integration program. Re20,/Si02.A120, (SA) catalysts of 0, 1, 3, 6 and 9 wt % Re20, supported on HA-SA (24.3 wt % Al,O,, 374 m2 g-l) and 3 wt % Re,O, on LAH- (13.0 wt % Al,O,, 460 m2 g-l) and LAL- (15.3 wt % A120,, 144 m2 g-l) silica-aluminal* were also measured as described above.Fig. 2 shows the i.r. spectra at 423 K in vacuum of 3 wt % Re207 catalysts supported on HA-SA, LAH-SA, LAL-SA and on y-alumina carriers. It appears that all the silica-alumina-supported catalysts contain a band at 1545 cm-l and a peak at ca. 1636 cm-l, while the alumina-supported one does not. The areas of the ca. 1545 cm-l band decrease in the order of HA-SA, LAH-SA, LAL-SA and alumina ( ~ 0 ) . The band positions of silica-alumina-supported Re,O, catalysts are given in table 2. Table 3 gives the integrated peak areas of i.r. bands at ca. 1540 and 1450 cm-l of some silica-alumina-supported Re207 catalysts, representing their respective Bronsted and Lewis acidities and the integrated peak area of the i.r.band at ca. 3740cm-l, representing the isolated OH groups on the catalyst surfaces. Fig. 3 shows the OH stretching region of a 1 wt % Re,O,/SA (HA) catalyst at various temperatures after pyridine adsorption in vacuum. It appears that the adsorption of pyridine decreases the band at 3741 cm-l and the evacuation at increasing higher temperatures gradually restores the OH band. cm3 of water was added to a 1 wt % Re,O,/SA (HA) catalyst after pyridine adsorption and evacuation at 423 K. The intensity of the peak at 1544 was greatly 5 x2712 Re@,-containing Metathesis Catalysts I I I I 3801 3711 3621 3531 3431 v/crn-' Fig. 3.1.r. spectra in the OH stretching region of a 1 wt % Re20,/Si02~A1203 (HA) catalyst.(a) Before pyridine adsorption in vacuum at 423 K, (b) in pyridine at 423 K, (c)-cf) after pyridine adsorption in vacuum: at 423 K (c), 573 K (d), 673 K (e) and 773 K (f). Table 4. FTIR results of some mixed-oxide catalysts (based on a 3 wt % Re20,/A1203 catalyst) and some related samples catalysta Bronsted acidity Lewis acidity A/cm-l g-l A/cm-l g-l Mo/Re 0 Mo/Re 2 Mo/Re 4 Mo/Re 6 Mo/Re 8 W/Re 6 V/Re 6 V/Re 2 6 wt % Re20,/A1203b 18 wt % W03/A1203 18 wt % Mo03/A1203 0 4.6 10.8 80.7 106.0 73.0 31.1 8.4 30.1 191.6 216.7 0 306.2 394.2 419.6 149.7 245.3 470.6 328.6 219.5 531.4 45 1.4 240.2 265.6 a The numbers after M/Re are the metal to rhenium atomic ratios of various mixed-oxide catalysts (Re : A1 = 1 : 154). Prepared by HC1 impregnation.X . Xiaoding, J .C. Mol and C. Boelhouwer 0 2 4 6 8 10 Mo/Re 2713 Fig. 4. Bronsted acidity (a) and conversion/wt % Re,O, (v) for methyl oleate as a function of Mo: Re atomic ratio for a series of MOO,. Re,O,/Al,O, (Re : A1 = 1 : 154) catalysts. increased, while the peak at 1453 cm-l decreased. The band at 3741 cm-l slightly changed, while a peak at 1648 cm-l appeared. Table 4 shows the FTIR results of various mixed-oxide catalysts. The conditions used for the measurements and the method to calculate the acidity were the same as for table 3. Fig. 4 shows the change of Bronsted acidity as a function of Mo: Re atomic ratios for a series of MOO,. Re,O,/Al,O, catalysts, based on 3 wt % Re,O,/Al,O,, together with their activity for the metathesis of methyl oleate using SnMe, as co-catalyst, expressed in conversion/wt % Re,O, at 90 min reaction time.ll Fig. 5 shows Bronsted and Lewis acidity as a function of wt % Re,O, of Re,O,/Al,O, catalysts together with their activity for the metathesis of propene, expressed in initial reaction rate (mol g-l s-l), reported by Kapteijn.,, Fig.6 shows Lewis and Bronsted acidity of HA-SA-supported catalysts as a function of rhenium loading, together with their activities for the metathesis of propene, expressed in initial reaction rate (mol g-l s - ~ ) , ~ calculated according to the formula of Kapteijn.,, Repeated i.r. experiments using the same samples deviated within less than ca. 10% of the reported values based on measurement of the integrated peak areas.Data for the peak areas calculated by the cut-and-weigh method or by the computer program are the same within f 2 % . 90 FAR 12714 Re20 -con tain ing Metathesis Catalysts B 80 - I OD d I 4 0 200 0 5 10 15 wt% Re20, Fig. 5. Bronsted (+) and Lewis (m) acidity and the initial reaction rate for propene metathesis (0) of various Re,O,/Al,O, catalysts. Discussion On solid surfaces, several species can be observed after pyridine adsorption, viz. hydrogen-bonded pyridine (HPy), physisorbed pyridine, coordinated pyridine (LPy) and pyridinium ions (BPy).lP4 Since evacuation at temperatures higher than 423 K results in desorption of the physisorbed and hydrogen-bonded pyridine,l6? 1 7 9 2o the integrated areas obtained after evacuation at 423 K are representative for the acidity of these surfaces.The spectra of Re,07/A120, show two peaks (or a peak with a shoulder) at ca. 1620 and two peaks at 1450 cm-l. The intensities of the peaks (or shoulders) with the lower wavenumbers decreased after evacuation. It appears that there are two types of Lewis acid sites on the Re207/A1203 catalysts under investigation. This phenomenon was also observed with pure aluminas.lq l6* l9 Moreover, the intensities of the peaks with higher wavenumbers seem to increase with the rhenium loading. This indicates that rhenium oxide is deposited preferentially on the weak Lewis sites and this deposition turns the weak Lewis sites into stronger ones. A similar phenomenon was also observed in the case of Mo03/A120, with the formation of protonic sites.20 The frequency of the ca.1450 cm-1 band gives an indication of the strength of Lewis acidity.l69 l8 From table 1 it can be seen that the frequency shifts to a higher value for rhenium-alumina catalysts with an increasing rhenium loading, indicating that stronger Lewis acid sites are formed2715 200 r( I en I H 1 T 100 0 X . Xiaoding, J . C. Mol and C. Boelhouwer / 0 2 4 6 8 wt% Re20, Fig. 6. Bronsted (m) and Lewis (+) acidity and the initial reaction rate for propene metathesis (a) of various Re,O,/SiO, .A1,0, (HA) catalysts. on Re,O,/AI,O,. That the ca. 1594 cm-l band was only observed in pyridine and not in the spectra after evacuation shows that pyridine at these sites is easy to desorb and we attribute this band to H-bonded or physisorbed pyridine.16 From fig.1 and 5, it appears that Bronsted acidity was developed on Re,O,/AI,O, catalysts at Re,O, loadings higher than ca. 6 wt % . The activity for metathesis increases with an increasing Bronsted acidity. Although Lewis acidity also increases with rhenium loading, the pattern of the change of activity agrees better with that of Bronsted acidity (fig. 5). That a 6 wt % Re,O,/Al,O, catalyst prepared by HCI impregnation shows a weak peak at ca. 1550 cm-l indicates that HCI impregnation increases Bronsted acidity.,~ Silica-alumina (SA) is well known for its strong a~idity.l-~ Acid sites with H, -= - 8.2, even H , < - 12.0" have been reported on SA.l* 3 9 25 The acidity is influenced by the alumina content, and the method of preparation and measured data of the acidity depend also on the method of measurement.l SA with 25 wt % A1,0, has been reported to contain maximum acidity.l? 26 Fig.2 shows that among the four 3 wt % Re,O, catalysts the one supported on HA-SA contains the highest Bronsted acidity; however, the LAH-SA-supported catalyst has a higher surface area. Alumina has no Bronsted acidity, in accordance with literature data.14 The frequency of the band at ca. 1450 cm-l of * Where H , is the Hammet index. 90-22716 Re,O,-containing Metathesis Catalysts (a) n . n Re20,/A1203 Re207/Si02 A1203 Fig. 7. Distribution of Bronsted acid sites (a) and rhenium species (0) on the surface of Re20,/A120, and of Re20,/Si02.A1,0, catalysts with (a) low, (b) medium and (c) high Re,O, loadings. silica-alumina (1455 cm-l) is higher than the one of alumina (145 1 cm-l), indicating a higher strength of the Lewis acidity of the silica-alumina-supported catalysts.Although the Lewis acidity of the alumina-supported catalyst is higher than that of all the silica-alumina-supported ones, its activity is the lowest in the metathesis of methyl oleatelo and of p r ~ p e n e . ~ The activity for metathesis decreases in the order of decreasing Bronsted a ~ i d i t y . ~ ~ lo It thus appears that there is a distinct correlation between Bronsted acidity and metathesis activity, but not between Lewis acidity and the activity. A similar correlation has been observed with the activity for metathesis of the series of Re,O,/SA (HA) catalysts with different rhenium loadings (fig. 6). In the spectra of silica-alumina and Re,O,/SiO, .Al,O, catalysts, a sharp band at ca. 3745 cm-l and a broad one at a lower frequency were observed in the OH region (fig. 3). The broad peak is attributed to bridged OH groups27 and the 3745 cm-1 band to isolated surface OH g r o ~ p s . ~ ~ - ~ ~ Although some authors suggest that there is only one type of isolated OH group on silica-alumina, due to its similarity to corresponding Si-OH groups on silica, others think that there are two types of isolated surface OH groups on SA with different acidities.l-*? 26-30 The integrated areas of the 3745 cm-l band of the series of HA-SA-supported Re,O, catalysts after pyridine adsorption and evacuation decrease with increasing rhenium loading. That this band does not totally disappear, even after a long time in pyridine, indicates that not all the OH groups on Re,O,/SiO, * Al,O, (HA) react with pyridine, which in turn indicates that the OH groups corresponding to the band at ca.3745 cm-l have different acidities. It appears that the deposition of rhenium oxide on silica-alumina initiates the acidity of distinct surface OH groups, hence decreasing the amount of non-acidic OH groups. Bronsted acidity increases with rhenium loading of the catalysts, Lewis acidity passes through a maximum at ca. 1 wt % Re,O,, while the reaction rate of the catalysts increases with an increasing rhenium oxide loading (fig. 6).9910 This again indicates that Lewis acidity does not correlate with metathesis activity, while Bronsted acidity does. It is known that the addition of certain metal oxides, e.g. MOO,, WO, or V,O,, to Re,07/A1,0, catalysts increases the metathesis a ~ t i v i t y .~ ~ 11-13 The deposition of these oxides on alumina also generates Bronsted acidity.l79 2o From fig. 4 and table 4 it is obvious that the addition of these metal oxides also increases the Bronsted acidity of catalysts with low rhenium loading and their activity indeed increases with increasingX . Xiaoding, J. C. Mol and C. Boelhouwer 2717 Bronsted acidity. This holds for Mo:Re ratios up to ca. 6; higher amounts of Mo apparently are an excess, covering active rhenium sites. BET data9 show that the catalyst with a Mo: Re ratio of 8 has a smaller surface area (192 m2 g-l) than the one with the ratio of 6 (220 m2 g-l), and this appears to reinforce the view mentioned above.Alumina-boria, a well known solid a ~ i d ~ l - ~ * with H, < - 8.2l~ 25 (its acidity being mainly of the Bronsted type1v34), is also a good carrier for rhenium oxide metathesis catalyst^.^^ 35 This again proves the vital role of Bronsted acidity in metathesis. Furthermore, the results of hexamethyldisilazane (HMDS) poisoning experiment^,^^ 3 5 7 36 in which poisoning of Bronsted acid sites by HMDS leads to a decrease and eventually to a total termination of metathesis activity of all the Re20,-containing catalysts studied, is a convincing proof of this correlation. Based on the acidity of the catalysts observed, a model is proposed (fig. 7 ) to illustrate the distribution of Bronsted acid sites and rhenium species on the surfaces of rhenium oxide-alumina and rhenium oxide-SA catalysts.The addition of a suitable third metal oxide to low rhenium loading Re20,/A1203 catalysts causes a transition from the situation in the left to the right in fig. 7(a). With this simple picture in mind and with the model presented in ref. (36) and (37), in which the formation of the metallacarbene intermediates includes the function of a reduced rhenium species and a neighbouring Bronsted acid site, we can explain the activity of the Re20,-containing catalysts studied so far. Nevertheless, this model is not necessarily applicable to other catalyst systems. For example, WO,/SiO, catalysts show completely different behavio~r,~~ indicating that other types of active sites might be operating. Conclusions Bronsted and Lewis acidity increase with increasing rhenium loading of Re20,/A1203 catalysts.The former starts to appear at ca. 6 wt % Re,O,. HCl impregnation leads to the formation of Bronsted acidity of catalysts with lower rhenium loading. The addition of a suitable third metal oxide to Re20,/A120, catalysts with low rhenium loading leads to the formation and increase of Bronsted acidity. Addition of rhenium oxide to silica-alumina increases its Bronsted acidity, while its Lewis acidity passes through a maximum. The metathesis activity of all the rhenium oxide-containing catalysts studied shows a distinct correlation with their Bronsted acidity. No clear correlation with Lewis acidity of these catalysts could be established. Very active rhenium oxide-containing metathesis catalysts are characterized by a pronounced Bronsted acidity.The key to the increase of the activity of low rhenium loading catalysts, therefore, lies in the increase of their Bronsted acidity by proper modifications. Xu Xiaoding (on leave of absence from the Department of Chemistry, Fudan University, Shanghai) is the recipient of a fellowship on the basis of an exchange programme between the Netherlands and the People’s Republic of China. We thank G. C. N. van den Aardweg for his technical help in building the equipment for the measurements. References 1 K. Tanabe, Solid Acids and Bases (Kodansha, Tokyo; and Academic Press, New York, 1970). 2 K. Tanabe, Catalysis, Science and Technology, ed. J . R. Anderson and M. Boudart (Springer-Verlag, 3 H. A. Benesi and B.H. C. Winquist, Adv. Catal., 1978, 27, 97. 4 H. Knozinger, Adv. Catal., 1976, 25, 184. 5 A. Ellison, A. K. Coverdale and P. F. Dearing, Appl. Catal., 1983, 8, 109. 6 A. Ellison, A. K. Coverdale and P. F. Dearing, J. Mol. Catal., 1985, 28, 1 . 7 R. Nakamura and E. Echigoya, Recl. Trav. Chim., Pays-Bas, 1977, 96, M31. Berlin, 1981), vol. 2, pp. 232-271.2718 Re@, -containing Me tat hesis Catalysts 8 N. Calderon, The Chemistry of Double-bonded Functional Groups, ed. S . Patai (John Wiley, London, 9 Xu Xiaoding, Ph.D. Thesis (University of Amsterdam, 1985). 1977), part 2, pp. 913-963. 10 Xu Xiaoding and J. C. Mol, J. Chem. Soc., Chem. Commun., 1985,631. 11 Xu Xiaoding, P. Imhoff, G. C. N. van den Aardweg and J. C. Mol., J. Chem. SOC., Chem. Commun., 12 E.I. Bogolepova. R. A. Fridman and A. N. Bashkiev, Izv. Akad. Nauk. SSSR, Ser. Khim., 1979, 7 , 13 R. Nakamura and E. Echigoya, Chem. Lett., 1977, 1227. 14 A. A. Olsthoorn and C. Boelhouwer, J. Catal., 1976, 44, 207. 15 P. F. Dearing, Ph.D. Thesis (Humberside College of Higher Education, Hull, 1984). 16 E. P. Parry, J. Catal., 1963, 2, 371. 17 F. E. Kiviat and L. Petrakis, J. Phys. Chem., 1973, 77, 1232. 18 M. R. Basila, T. R. Kantner and K. H. Rhee, J. Phys. Chem., 1964,68, 3197. 19 M. R. Basila and T. R. Kantner, J. Phys. Chew., 1966, 70, 1681. 20 Koh-Ichi Segawa and W. K. Hall, J. Catul., 1982, 76, 133. 21 T. R. Hughes, H. M. White and R. J. White, J. Catal., 1969, 13, 58. 22 E. A. Pankshtis, R. I. Soltanov and E. N. Yurchenko, React. Kinet. Catal. Lett., 1982, 19, 105. 23 G. Visser-Luirink, E. R. A. Matulewicz, J. Hart and J. C. Mol., J. Phys. Chem., 1983, 87, 1470. 24 F. Kapteijn, Ph.D. Thesis (University of Amsterdam, 1980). 25 H. A. Benesi, J. Am. Chem. SOC., 1956,78, 5490. 26 P. Andreu, G. Martin and H. Noller, J. Catal., 1971, 21, 255. 27 R. E. Sempels and P. G. Rouxhet, J. Colloid Interface Sci., 1976, 55, 263. 28 M. Takahashi, Y. Iwasawa and S. Ogasawara, J. Catal., 1976,45, 15. 29 H. Boehm and H. Knozinger, Catalysis, Science and Technology, ed. J. R. Anderson and M. Boudart (Springer-Verlag, Berlin, 1983), vol. 4, pp. 4&195. 30 J. M. Guil, J. E. Herrero and A. R. Paniego, J. Colloid Interface Sci., 1984, 102, 1 1 1. 31 A. Ozaki and K. Kimura, J. Catal., 1964, 3, 395. 32 M. Sato, T. Aonuma and T. Shiba, Proc. 3rd Int. Congr. Catul., ed. W. ha. H. Sachtler, G. C. A. Schuit and P. Zwietering (North-Holland, Amsterdam, 1964), vol. 1, pp. 396407. 33 V. A. Dzisko, Proc. 3rd Int. Congr. Catal., ed. W. M. H. Sachtler, G. C. A. Schuit and P. Zwietering (North-Holland, Amsterdam, 1964), vol. 1, pp. 422432. 34 Y. Izumi and T. Shiba, Bull. Chem. Soc. Jpn, 1964, 37, 1797. 35 Xu Xiaoding, C. Boelhouwer, J. I. Benecke, D. Vonk and J. C. Mol, J. Chem. Soc., Faraday Trans. 1, 36 Xu Xiaoding, C. Boelhouwer, D. Vonk, J. I. Benecke and J. C. Mol, J. Mol. Catal., in press. 37 D. T. Laverty, J. J. Rooney and A. Stewart, J. Catal., 1976, 45, 110. 38 A. J. Van Roosmalen and J. C. Mol, J. Cutal., 1982, 78, 17. 1985,273. 1623. 1986,82, 1945. Paper 5 / 18 13 ; Received 18th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202707
出版商:RSC
年代:1986
数据来源: RSC
|
14. |
Chemisorption and catalysis by metal clusters. Hydrogenation of ethene and hydrogenolysis of ethane catalysed by supported ruthenium clusters derived from Ru3(CO)12and from H4Ru4(CO)12 |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2719-2727
Richard B. Moyes,
Preview
|
PDF (679KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans. I, 1986,82, 2719-2727 Chemisorption and Catalysis by Metal Clusters Hydrogenation of Ethene and Hydrogenolysis of Ethane catalysed by Supported Ruthenium Clusters derived from RU,(CO)~, and from H,Ru,(CO),, Richard B. Moyes and Peter B. Wells* Department of Chemistry, The University, Hull HU6 7RX S . David Jackson" and Robin Whyman I.C.I., New Science Group, P.O. Box 11, Runcorn WA7 4QE Properties are described for catalysts containing active clusters derived from Ru,(CO),, and H,Ru,(CO),, and supported on silica, alumina and titania. The clusters, protected against sintering by retained ligand-CO, ligand-C, and a support-cluster interaction, are stable under the range of reaction conditions used (0.101 MPa, 273-590 K) and provide highly reproducible activity for the hydrogenation of ethene and the hydrogenolysis of ethane.Freshly prepared catalysts each exhibit an initial non-steady state during which hydrocarbon is progressively retained and activity rises, passes through a maximum and declines to a steady-state value. Catalysts in the steady state with respect to activity continue to accumulate hydrocarbon and such retained hydrocarbons mediate hydrogen-atom transfer to reacting adsorbates. The concentration of the retained hydrocarbon species, which have been determined by material balance, are compared with the known site concentrations associated with fresh cluster-derived catalysts. Catalysts in the steady state exhibited low activity for ethene hydrogenation. This behaviour is interpreted in terms of the strength of bonding between ethene and the cluster.In ethane hydrogenolysis, the site and bonding limitations of the cluster allow C-C bond breaking before extensive dehydrogenation takes place. Ruthenium cluster catalysts show lower specific activity than corresponding osmium catalysts. Supported ruthenium carbonyl clusters have been used as precursors for catalysts for a variety of rea~tionsl-~ and novel effects have been observed.,? However, often either the catalysts have not been characteri~ed~ or the preparation method is such that decomposition of the cluster and subsequent sintering of the residual species is assured., In a previous paper,5 catalysts, prepared from Ru,(CO),, and H,Ru,(CO),, supported on alumina, silica and titania were thermally activated and characterised using a variety of techniques (e.8.temperature-programmed decomposition, i.r. and u.v.-visible spectroscopies and selective chemisorption). These same samples have subsequently been used as catalysts for ethene, carbon monoxide and carbon dioxide hydrogenations and ethane hydrogenolysis and various novel effects have been observed. One such effect already reported is the insensitivity of these catalysts to poisoning by air at room temperature.6 In this paper attention will be focussed on two of these reactions, ethene hydrogenation and ethane hydrogenolysis. Both reactions have been studied on conventional supported ruthenium catalysts7. * and, where possible, results have been compared. Similar work has been carried out on supported osmium clusters*12 and again comparisons will be made.27192720 Catalysis by Supported Ru Clusters Table 1. Mean empirical composition of Ru,(CO),,C,, Z,, sample composition" CO adsorbedb derived from Ru3(CO),,/alumina Run(Co)2.8n '0.4, '0.4, 41.5 derived from Ru,(CO),,/silica Ru,(Co)O.O, '1.5, '0.6, 61 .O derived from Ru,(CO) / ti tania Run(Co)l.O, ' 0 . 6 , z0.8n 68.9 derived from Ru,H4(CO)12/alumina Ru~(Co)2.2n '0.2, '1.6, 154.0 derived from H4Ru4(CO)12/silica Ru,(C0)1.5?2 '0.5, '0.2n 15.5 derived from H4Ru4(CO),,/titania Run(Co)O.O?& c1.3?2 ' 0 . 6 , 60.0 a Two Z-entities constitute an adsorption site for CO and n has a value in the region of 10. (g catalyst)-l. pmol Experimental Catalyst Preparation and Nomenclature Supports used were alumina [Aluminium Oxide C (Degussa)], silica (Cab-0-Sil) and titania (Tioxide).The titania, which was prepared specially for this investigation from an organic titanate, was anatase; its purity was better than 99.9% TiO, and it had a silica content of less than 100 ppm. Each support was dried by heating to 773 K for 16 h in a stream of dry nitrogen. The surface areas of the dried materials, measured by the B.E.T. method of N, physisorption, were: silica, 110 k 5 m2 g-l; alumina, 97 k 5 m2 g-l; titania, 43 5 m2 g-l. Supports were impregnated with solutions of Ru,(CO),, or H,Ru,(CO),, in trichloromethane, the solutions being added to suspensions of support. Evaporation to dryness was carried out in a stream of dry nitrogen at 293 K; all of the cluster compound appeared to be impregnated on to the support. Weights of cluster compound and support were such as to give 2.0% by weight of ruthenium in all catalysts.Heating of the freshly impregnated materials in vacuum or in a helium stream to 523 K provides enhanced nuclearity clusters Ru,(CO),, Cy, Z,, (where n = 10 and Z represents adsorption sites). The species Ru,(CO),, Cy, Z,, can chemisorb H,, O,, CO and hydrocarbons and can catalyse ethene hydrogenation and ethane hydrogenolysis. The supported catalysts are referred to by the parent cluster and the support, e.g. Ru,(CO),,/silica, H,Ru,(CO),,/alumina etc. even though the active entities present may be enhanced clusters. The mean empirical compositions of Ru,(CO),, Cy, Z,, and the amount of carbon monoxide adsorbed, as given in ref.(9, is shown in table 1. Apparatus and Materials Reactions were carried out in a pulsed-flow microcatalytic reactor connected to a dual-column gas chromatograph for the analysis of products (fig. 1). The columns contained 13X molecular sieve and Porapak Q. The reactor consisted of a cylindrical glass vessel fitted with a coarse glass sinter on which the catalyst (0.30 g) rested. The depth of catalyst bed was typically 1 cm. Temperatures were measured by use of a thermocouple enclosed in a glass sheath located at the centre of the catalyst bed. Helium flowed through the reactor at a constant rate of 23 cm3(s.t.p.) min-l. The reactants stored in the vessels (R) were mixed in volume ( V ) before being metered (typically at 6.6-40.0 kPa) either into the standard volume (3.6 cm3) or the static reactor.The pulses in the standard volume were introduced into the carrier stream and hence to the catalyst and analytical system. The titania-supported samples were found to be unsuitable for use in the flow system owing to the powder being blown out of the reactor. This problem was not found with either the silica- or alumina-supported samples. Therefore the titania-supported catalysts were examined in a static reactor (50 cm3). Pre-mixedR. B. Moyes, P . B. Wells, S. D . Jackson and R. Whyman 272 1 I 1 +- I pJ-+g, 7-Lq-J E vacuum system 1 pulse-flow system Fig. 1. Schematic representation of the apparatus. The flow system is shown at the right-hand side; arrows indicate the direction of helium flow. A, katharometer. B, input flowmeter.C, standard volume. D, pulsed-flow reactor. E, trap. F1 and F2, g.1.c. columns. G, Geiger-Muller counter. H, scalar-ratemeter. I, output flow meter. P1 and P2, pressure transducers. (G and H were used in work to be presented in a future paper.) reactants were admitted to the reactor, and samples extracted for analysis. Catalyst weights were again 0.30 g. Compressed gases, H,, C,H,, C,H, (all >99.9%) were used as received. Activation energies quoted are uncertain to +4 kJ mol-l. The steady state referred to in this paper is defined as when the catalyst gave reproducible results for a given set of experimental conditions. Results The six cluster/support combinations, activated by heating tq 523 K as described in the Experimental section and in ref.(5), were immediately active for ethene hydrogenation and ethane hydrogenolysis. For both reactions the initial activity shown by each catalyst was separated from the steady-state activity that was finally achieved by a period of non-steady state behaviour. Non-steady State Behaviour Fig. 2 shows two examples of non-steady state behaviour; both show (i) a definite initial activity at pulse 1, (ii) a maximum in activity and (iii) a region of declining activity. Reproducible activity, at a level substantially lower than that achieved at the maximum, was finally achieved (see steady-state section). Fig. 2(a) shows this behaviour for ethane hydrogenolysis initially at 523 K (pulses 1-6), and then at various temperatures. The temperature was varied to examine whether the effect was temperature dependent.Fig. 2 (b) shows similar behaviour for ethene hydrogenation at various temperatures showing that the effect is at least partially temperature independent. Product analyses during the non-steady state period showed an imbalance in the sense that hydrocarbon was progressively retained by the catalyst. Table 2 shows the extent of hydrocarbon retention at the maximum of the curve of rate against pulse number and at the point where the steady state is reached. The extent of hydrocarbon loss is significantly less than the number of sites as determined by CO-chemisorption. Even by the onset of steady-state behaviour the extent of hydrocarbon retention is not equal to the chemisorption site density. Steady-state Behaviour After passage through the non-steady state, catalysts showed reproducible activity, during which kinetic information was obtained.For each catalyst conditions were2722 Catalysis by Supported Ru Clusters *1 (a' 523 0 ' 0 5 10 pulse number Fig. 2. Initial non-steady state behaviour. (a) Ethane hydrogenolysis (C,H,:H, = 1 : 1) at 523 K (pulses 1-6) then at various specified temperatures over catalyst derived from H4Ru,(CO),,/silica; (b) ethene hydrogenation (C2H4: H, = 1 : 1) at various temperatures over a catalyst derived from Ru,(CO),,/silica. Table 2. Hydrocarbon retention during non-steady state behaviour hydrocarbon re ten tionb siteby at end of density at activity non-steady (fresh catalyst reactiona maximum state catalyst) 61.0 24.4 Ru3(C0)12/Si02 C2H4 + H, -+ C& 5.5 C,H, + H, -+ CH, 9.8 H4Ru4(C0)12/A1203 C2H4 + H2 -+ C2H6 8.6 44.6 C,H, + H, -+ CH, 73.9 21.9 1 154.0 a C2H4: H, = 1 : 1, C,H6:H, = 1 : 1.Units pmol (g catalyst)-'. From CO adsorption measurements. selected so as to achieve conversions mostly in the range 0.5 to 10%. The reaction rate, r, obeyed the equations r = kPy PF and k = A exp (- Eapp/RT) where PI represents the hydrocarbon pressure, pZ the hydrogen pressure and Eapp the apparent activation energy. Thus the temperature dependence of the rate was described by the equation: log rT = - Eapp/2.303RT+ C where C is a constant.R. B. Moyes, P. B. Wells, S . D. Jackson and R. Whyrnan 2723 Table 3. Kinetic parameters for ethene hydrogenation" order in catalystb ethene 1% GooC Eapp/kJ mol-1 Ru,(CO),,/alumina 0.9 -0.98 32 (465-515 K) Ru,(CO),,/silica 0.0 0.47 64 (345-410 K) RU,(CO)~,/ titania 0.2 -0.25 37 (390-410 K) H,Ru,(CO),,/alumina d 0.15 21 (370-475 K) H,Ru,(CO),,/silica d - 0.4 18 (40M75 K) H,Ru,(CO),,/titania d 1.27 42 (273-340 K) a C,H,:H, = 1: 1.measured. Prepared by thermal activation. r has units pmol s-l (g catalyst)-'. Not Table 4. Kinetic parameters for ethane hydrogenolysis" catalystb log r476c Eapp/kJ mol-' Ru,(CO), , /alumina Ru,(CO),,/silica Ru,(CO),,/titania H, Ru,(CO), , / alumina H,Ru,(CO),,/silica H,Ru,(CO),,/titania Ru/silicad - 3.45 0.95 0.1 0.25 - 1.65 - 1.75 0.56 76 (505-580 K) 61 (400-455 K) 118 (370-435 K) 71 (520-590 K) 77 (410-490 K) 134 (450-483 K) 36 (445-5 15 K) a C,H6: H, = 1 : 1. r has units pmol s-' (g catalyst)-'. 5 % Ru by weight, prepared from RuCl, by reduction in flowing H, at 723 K for 2 h [taken from ref.@)I. Prepared by thermal activation. Ethene Hydrogenation Catalysts were active at 273 K and above. Values of the order of reaction, the apparent activation energies and logr at 400 K are given in table 3. Rates are stated for 400 K to show the relative activity pattern. The apparent activation energies show considerable variance, both from each other and from that reported for ethene hydrogenation catalysed by a conventional supported ruthenium,13 i.e. ca. 35 kJ mol-l. Ethane Hydrogenolysis Immediately following the investigation of ethene hydrogenation, the same catalysts were used for ethane hydrogenolysis. In each case a further cycle of non-steady state behaviour preceded steady-state activity. Table 4 gives apparent activation energy values together with values of logr at 476 K; also given are values obtained by Sinfelt and Yates8 for a conventional polycrystalline Ru/silica.All apparent activation energies obtained from the cluster catalysts are significantly lower than that obtained by Sinfelt and Yates.8 No pattern is observed either in relative/specific activity or in apparent activation energy, although four of the six catalysts give similar activation energies.2724 Catalysis by Supported Ru Clusters Discussion The Non-steady State The non-steady state was characterised by a progressive loss of hydrocarbon throughout, during which the activity rose, passed through a maximum and then declined to a steady state.We will interpret these findings for the situation involving ethene, but the model is applicable to other reactions. ' R R CHB R 1 1 1 H-transfer I I I I I I I CH, - H2CmCH2 CH, - 2 CH2 CH,_, -M-M- M-M- -M-M- + I I t L+H (from RU-H) ,I+H -I I------ 6 The initial activity observed with the first pulse over a freshly prepared catalyst represents the nearest approach that can be made to an activity measurement in the absence of retained hydrocarbon. Ethene molecules adsorb at vacant sites and react with hydrogen which adsorbs in competition. This represents the inherent activity of the ruthenium clusters. The first trend in the non-steady state was that activity increased; this occurred despite the fact that hydrocarbon was being progressively retained so that the concentration of sites for ethene adsorption and reaction decreased.We infer from this that the retained hydrocarbon acted as a hydrogen-transfer agent and that the effective concentration of adsorbed hydrogen was thereby increased. Thomson and Webb13 and Somorjai and coworkers1* have both claimed, with substantive evidence, that hydrocarbon retained at the conventional catalyst surface plays an active role in hydrogen transfer to adsorbed hydrocarbons, Therefore ethene hydrogenation is envisaged to occur as shown in scheme 1, where the retained hydrocarbon is represented as R-CH,-M or as R-CH,,_,,-M. The rise in activity indicates that hydrogen atom acquisition by the retained hydrocarbon species and its subsequent transfer to adsorbed -C,H, and -C,H, (steps 2, 3,4 and 5) is a faster process than direct acquisition from the ruthenium atom sites (steps 6 and 7).It is worthwhile to note at this point that the site, -M-M-, will adsorb only one CO or 0, rnolec~le,~ so that some correlation between the amount of deposited hydrocarbon and the site density, as measured by CO adsorption, may be expected. Indeed a relationship was observed with osmium cluster catalysts where the amount of hydrocarbon deposited was similar to the number of sites as measured by CO adsorption. However, with the ruthenium-cluster catalysts no relationship is observed. In fact it has been shown for most of the Group VIII metals that there is no simple relationship between adsorptive capacity and the fraction of adsorbed molecules active in hydr0genati0n.l~ The maximum in activity was soon reached and further hydrocarbon retention led to a decline in activity.This may occur because both sites of the pair, symbolised in scheme 1, may have become occupied by retained hydrocarbon or the retained species represented as R-CH,-M may have homologated to R'-CH,-M, which would result in increased steric hindrance toR. B. Moyes, P. B. Wells, S. D . Jackson and R. Whyrnan 2725 ethene adsorption at the second site of the pair. This model is in keeping with that proposed for ethene hydrogenation and ethane hydrogenolysis over osmium-cluster catalysts. The Steady State Kinetic information obtained on ethene hydrogenation in the steady state is detailed in table 2. Positive orders of reaction for ethene obtained with Ru,(CO),,/alumina and Ru,(CO),,/titania are unusual in that most Group VIII catalysts give a zero or negative order in ethene, e.g.Ru/silica -0.6, Ru/alumina -0.2. However, these results are in keeping with those from osmium-cluster catalysts, where the positive orders in ethene were interpreted as suggesting weak ethene adsorption on the cluster catalysts. This conclusion was supported by the fact that as the commitment of the metal to residual carbonyl ligands increased, the more positive were the reaction orders in ethene. A similar relationship is found with the ruthenium catalysts, i.e. number of carbonyl catalyst order in ethene ligands per Ru Ru3(C0)1 2 /A1203 0.9 0.2 Ru3(C0)12/Si02 0.0 Ru, (CO) 12 / Ti02 2.8 0.0 1 .o In ethane hydrogenolysis results were obtained which suggested novel behaviour.Sinfelt and Yatess calculated, using the mechanism of Boudart et a1.,16 that the species preceding bond-breaking on Ru/silica was adsorbed C, with no hydrogen bonded to carbon. However, this mechanism requires a large site, e.g. Martin1' has shown for a nickel catalyst that an ensemble of ca. 12 atoms is required for ethane hydrogenolysis; this site requirement is unlikely to be met by the cluster, where its total number of atoms is ca. 10. However, the apparent activation energies obtained with the cluster-derived catalysts are significantly lower than that obtained from a conventional ruthenium catalyst (table 4). These low activation energies may be indicative of less-extensive dehydrogenation of the adsorbed ethane, in which case hydrogenolysis could proceed on a smaller number of sites than required on a conventional catalyst.Therefore, it is unlikely that the mechanism proceeding on conventional ruthenium catalysts is occurring on the supported ruthenium cluster catalyst, although the same mechanism could occur in both systems if that accepted for conventional metal catalysts was incorrect. This has been suggested in part by Frennet et aZ.,ls but it is beyond the scope of this paper to discuss the matter further. Comparison of Ruthenium- and Osmium-cluster Catalysts Specific activities, and ratios of metal: (adsorbed CO), metal: (ligand CO) and metal: (ligand C) obtained for catalysts derived from H,Ru,(CO),,, Ru,(CO),,, H,Os,(CO),, and OS,(CO),, are compared in tables 5 and 6.Generally : (i) the ruthenium-cluster catalysts adsorb more CO per metal atom than do osmium-cluster catalysts, (ii) the amount of retained carbonyl ligands per metal atom is higher for osmium than ruthenium catalysts and (iii) the quantity of carbon retained per metal atom is greater for ruthenium than osmium catalysts (table 5). These three points suggest that ruthenium-derived catalysts have gone further towards complete ligand removal than osmium-derived catalysts. Table 6 gives the turnover frequencies for each of the catalysts for ethene hydrogenation and ethane hydrogenolysis. For ethene hydrogenation the osmium-cluster catalysts have higher turnover frequencies than their ruthenium counterparts, the only exception being2726 Catalysis by Supported Ru Clusters Table 5.Comparison of ruthenium and osmium catalyst characteristics no. of C ligands no. of CO ligands per metal atom adsorbed per per metal atom present in the metal atom after activation activated cluster no. of CO molecules catalysta Ru 0 s Ru 0 s Ru 0 s M,(CO),,/alumina 0.2 0.1 2.8 2.2 0.4 0.1 (5) M3(C0), ,/ ti tania 0.3 (6) 0.3 (9) 1 .o 2.0 0.6 0.8 M ,( CO), ,/silica 0.3 0.2 '0.0 2.8 1.5 0.3 H,M,(CO),,/alumina 0.8 0.1 2.3 2.4 0.1 (5) 0.2 H,M,( CO),,/silica 0.1 0.0 (3) 1.5 1.6 0.5 0.3 H,M,(CO),,/titania 0.3 (1) 0.2 (8) 0.0 2.2 1.3 0.3 a Prepared by thermal activation. Table 6. Comparison of specific activities of Ru and 0 s catalysts catalysta turnover frequency turnover frequency for ethene hydrogenation for ethane hydrogenolysis / 103 S-1 /lo3 S-1 - Ru 0 s Ru 0 s M 3(CO)1 ,/alumina 3 4 0 (0.01) 4 M,(CO),,/silica 49 260 150 100 M,(CO),,/titania 8 17 18 3 H,M,(CO),,/silica 26 59 1 100 H,M,(CO),,/titania 310 1600 0 (0.3) 8 H,M,(CO),,/alumina 9 4 12 65 c c M/silicab 24 330 a Prepared by thermal activation.temperature reduction in flowing H, [taken from ref. (8) and (19)]. 5% Ru, 1% 0 s w/w, both catalysts prepared by high- Data not available. H,Ru,(CO),,/alumina in comparison with H,Os,(CO),,/alumina. For ethane hydrogenolysis four of the six catalyst pairs have osmium catalysts with higher turnover frequencies than their ruthenium analogues; Ru,(CO),, and Os,(CO),, on silica gave similar values whereas that for Ru,(CO),,/titania is an order of magnitude greater than that for Os,(CO),,/titania. Generally for both ruthenium and osmium catalysts with both reactions the silica supported catalysts are the most active.However, great caution must be exercised in the interpretation of comparisons of turnover frequency. Turnover frequency is the number of molecules converted per site for CO-adsorption per second and we have concluded [see also ref. (15)] that, for ruthenium-cluster catalysts, the number of active sites for ethene hydrogenation and ethane hydrogenolysis is not well measured by the extent of CO-adsorption. Conclusion The catalytic behaviour of these ruthenium-cluster catalysts for ethene hydrogenation and ethane hydrogenolysis mirrors the behaviour already observed for osmium analogues.11t12 The change in metal did not bring about any significant change inR.B. Moyes, P . B. Wells, S. D . Jackson and R . Whyman 2727 behaviour, although the results reinforce the view that the initial unsteady-state period of a catalyst’s lifetime fundamentally influences the later stages. Furthermore, the mechanism of ethane hydrogenolysis catalysed by these clusters requires further mecha- nistic study, We thank Dr D. Urwin of Tioxide Limited for a gift of pure titania. This work was carried out at Hull University as part of a Joint Research Scheme with ICI plc. References 1 J. Robertson and G. Webb, Proc. R. SOC. London, Ser. A , 1974,341, 383. 2 L. Guczi, Z. Schay, K. Matusek, I. Bogyay and G. Steffler, Proc. VZIth Int. Congr. Catal., ed. T. Seiyama and K. Tanabe (Elsevier, Amsterdam, 1981), part A, p. 21 1. 3 A. F.Simpson and R. Whyman, J. Organomet. Chem., 1981,213, 157. 4 C. S. Kellner and A. T. Bell, J. Catal., 1982, 75, 251. 5 D. J. Hunt, S. D. Jackson, R. B. Moyes, P. B. Wells and R. Whyman, J. Chem. SOC., Faraday Trans. 6 D. J. Hunt, S. D. Jackson, R. B. Moyes, P. B. Wells and R. Whyman, J. Chem. SOC., Chem. Commun., 7 G. C. Bond, G. Webb and P. B. Wells, J. Chem. SOC., Faraday Trans. I , 1965,61,999; G. C. A. Schuit 8 J. H. Sinfelt and D. J. C. Yates, J. Catal., 1967, 8, 82. 9 G. Collier, D. J. Hunt, S. D. Jackson, R. B. Moyes, I. A. Pickering, P. B. Wells, A. F. Simpson and I, 1986, 82, 189. 1982, 85. and L. L. van Reijen, Adv. Catal., 1958, 10, 242. R. Whyman, J , Catal., 1983, 80, 154. 10 D. J. Hunt, S. D. Jackson, R. B. Moyes, P. B. Wells and R. Whyman, J. Catal., 1984,86, 333. 11 S. D. Jackson, R. B. Moyes, P. B. Wells and R. Whyman, J. Catal., 1984,86, 342. 12 D. J. Hunt, S. D. Jackson, R. B. Moyes, P. B. Wells and R. Whyman, Proc. VIZZth Znt. Congr. Catal. 13 S. J. Thomson and G. Webb, J. Chem. SOC., Chem. Commun., 1976, 526. 14 G. A. Somorjai, Proc. VZZZth Znt. Congr. Catal. (Verlag-Chemie, Basel, 1984), vol. 1, p. 113. 15 D. Cormack, S. J. Thomson and G. Webb, J. Catal., 1966, 5, 224; G. F. Taylor, S. J. Thomson and G. Webb, J. Catal., 1968, 12, 191; J. U. Reid, S. J. Thomson and G. Webb, J. Catal., 1973, 29, 421; S. J. Thomson and G. Webb, personal communication. (Verlag-Chemie, Basel, 1984), vol. v, p. 27. 16 M. Boudart, A. Cimino and H. S. Taylor, J. Phys. Chem., 1954,58, 796. 17 G. A. Martin, J. Catal., 1979, 60, 345. 18 A. Frennet, L. Degols, G. Lienard and A. Crucq, J. Catal., 1974, 35, 18. 19 J. H. Sinfelt, J. Catal., 1973, 29, 308. Paper 511827; Received 21st October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202719
出版商:RSC
年代:1986
数据来源: RSC
|
15. |
Electron spin resonance evidence for the structure of AgIIand AgOsolvates in acetonitrile |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2729-2733
Martyn C. R. Symons,
Preview
|
PDF (432KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans. 1 1986,82, 2729-2733 Electron Spin Resonance Evidence for the Structure of Ag" and Ago Solvates in Acetonitrile Martyn C. R. Symons,* David Russell and Andrew Stephens Department of Chemistry, The University, Leicester LEI 7RH George Eastland Saginaw Valley State College, University Center, MI, U.S.A. There are currently three quite disparate concepts concerning the structures of the solvates of AgII and Ago centres formed from solutions of AgI ions in acetonitrile. These are derived from e.s.r. and electron spin-echo modulation studies. We have now established that identical centres are formed from the crystalline derivative AgClO,(MeCN),, whose crystal structure is known. Analysis of the e.s.r. spectra suggests that the original tetrahedral configuration with a-bonded MeCN ligands is initially retained on electron addition, but that three of the four solvent molecules are lost on annealing or photolysis.Electron loss results in relaxation to give square planar coordination for the AgII centre, but the a coordination is retained. We find no evidence for (71) coordination. When solutions of silver perchlorate in acetonitrile (MeCN) are exposed to ionizing radiation at low temperatures AgII and Ago centres are formed. The latter has e.s.r. spectra which vary on exposure to light and on annealing. Three quite different conclusions have been drawn regarding the structures of these s~ecies.l-~ We attempt here to select one, on the basis of a study of the pure crystalline material, AgClO,(MeCN),. The Silver Atom Centre We have shown that when solutions of AgClO, in acetonitrile (CH,CN or CD,CN) are exposed to y radiation at 77 K a centre described as AgO(MeCN), is formed, characterized by a relatively low hyperfine coupling to lo9Ag (- 509 G) with isotropic coupling to four equivalent 14N nuclei (6 G).1)2 This was taken to mean that AgI ions are normally solvated by four MeCN molecules, probably arranged tetrahedrally about silver.End-on (a) coordination was assumed because of the isotropic nature of the 14N hyperfine coupling. On annealing, no new features for Ago centres were detected. Addition of water or methanol had almost no affect on the e.s.r. spectra up to ca. 0.5 mole fraction, showing, in our view, that AgI is strongly and preferentially solvated by MeCN.However, for those mixed solvent systems, a new species [A in ref. (2)] was formed in low concentration relative to the Ago(MeCN), centre, but the latter was converted to A on annealing above 77 K. Species A had a coupling to lo9Ag close to the atomic value (ca. - 720 G) and clear hyperfine coupling to only one 14N nucleus. We took this to mean that desolvation of the primary AgI solvates occurred more readily in these mixed solvents, species A being an intermediate between the tetrasolvated centre, characteristic of AgI solvation, and the unsolvated atom, which was also detected in some of these systems. These processes are indicated in scheme 1. Implicit in our discussion, as in that of our work on Ago centres in rigid aqueous systems,'4 is the idea that the primary Ago centres, when formed from AgI will initially 27292730 Structure of Ag solvates in Acetonitrile Ago (MeCN) + 3 MeCN I Ago + MeCN Scheme 1 .2 7 retain the Agl solvation, but that this will be lost on annealing to give essentially unsolvated silver atoms. We expect silver atoms to be very weakly solvated relative to AgI, in such good cation-solvating media as water or acetonitrile. These ideas are not shared by Kevan et al., who interpret changes that occur to Ago centres in terms of specific Ago s o l ~ a t i o n . ~ ~ ~ This view is maintained in their recent studies of silver perchlorate in acetonitrile. Their results at 77 K were the same as ours.l9 However, at 4 K they obtained the Ago (MeCN), centre in low yield with relatively poor resolution.On annealing, the yield was enhanced to the 77 K value, almost certainly as a result of reaction with (MeCN); centres. However, on photolysis (with a 500 W slide-projector lamp), a new centre was formed (labelled CNI) in very low concentration, having parameters similar to our species A. This species was lost on annealing to 77 K, but could be reformed on further photolysis at 4 K. Li and Kevan7 conclude that photolysis releases electrons from (MeCN); units and that these add to AgI to give species A (CN I). They go on to suggest that this centre is actually the primary electron-capture centre, representing the state of solvation of AgI ions in acetonitrile. This implies that AgI ions have a solvation number of one in MeCN. If this is correct, it is surprising that only the Ago(MeCN), (species CN 11) centre was detected after radiolysis, there being no trace of species A at this temperature.This was explained by the novel postulate that 'the most energetic secondary electrons' were responsible for Ago formation in the radiolyses and that this extra energy led to the conversion of species A (CN I) to the tetrasolvate (their species CN 11). The growth in solvation number from one to four was described as the growth of atom s~lvation,~ just as the changes observed with Ago in water were ascribed to atom s o l ~ a t i o n ~ ~ ~ (scheme 2). This is the complete inverse of our interpretation and expectation. 4 K Ag' (MeCN) + (e-)* + 3 MeCN - Ago (MaN)4 (CN 11) i h v t 7 7 K Scheme 2.' Very recently, Ichikawa, et aZ.8 have produced another interpretation of these results.It is now suggested that the parent Agl solvate is, after all, AgI(MeCN), rather than Ag(MeCN)r, but that this has a square-planar configuration of four n-coordinated solvate molecules, as in fig. 1 (a), rather than a Td (sigma) coordination as we originally proposed.lq2 It is now suggested that electron addition at 4 K gives the square-planar complex, Ago(MeCN), and that on annealing this changes to a tetrahedral complex, still retaining the n-coordinated MeCN groups, as in fig. 1 (b). These reactions are summarized in scheme 3 .M . C. R. Symons et al. 273 1 I Me Fig. 1. Structures postulated for Ag'I, AgI and Ago solvates in CD,CN:' (a) for AgII and AgI, (b) for Ago. Ag' (MeCN)., + e- - Ago (MeCW4 (planar, r) (planar, r) 1 h' Ago (MeCN)4 T d , Scheme 3.s The Ag" Centre All three groups agree that these must be square-planar complexes.However, in our original study,l9 we stated that the MeCN ligands were a-coordinated, whilst Ichikawa, et aL8 claim that they are n-coordinated, as in fig. 1 (a). The aim of this work was to attempt to draw a clear distinction between these three quite different structural and mechanistic schemes by studying the well defined crystalline compound, AgClO,( MeCN), . Experimental Single crystals of AgClO,(MeCN), were grown by cooling saturated solutions in MeCN. These were cooled to 77 K for radiolysis and e.s.r. studies. Since the crystals lose solvent rapidly at room temperature, they were studied immediately after preparation for the collection of X-ray data.Good results were obtained, but before our analyses were complete a paper by Nilsson and Oskarsson appeared which exactly reproduces our Their analyses are clearly accurate and there was therefore no point in our continuing with our studies. Suffice it to say that the two sets of data are in good agreement. For e.s.r. studies we concentrated on polycrystalline samples for comparison with previous work on frozen solutions. Some arbitrary spectra from crystals were obtained to confirm that the 14N hyperfine coupling for the Ago(MeCN), centre is indeed almost isotropic, as previously claimed. Samples were exposed to 6oCo y-rays at 77 K of doses of up to 1 MRad. E.s.r. spectra were studied at 77 K.2732 Structure of Ag solvates in Acetonitrile Results The most important result is that the e.s.r. spectra for the dQ AgII(MeCN), complex and the Ago(MeCN), complex in the crystalline compound were indistinguishable from those obtained previously using frozen solutions.l, We can therefore conclude that there were no special constraints imposed on the complexes by the rigid crystal structure and that arguments based on results for one medium must be transferable to the other.The X-ray results show conclusively that the AgI(MeCN), units are almost perfectly tetrahedral, with nearly linearly coordinated -NCMe units. Thus for AgI, a-coordination occurs rather than the n-coordination postulated by Ichikawa et a1.8 for AgI ions. This confirms our previous surmise1* and common expectation. This result suggests that a-coordination is likely to be retained for the AgII and Ago centres, but does not constitute proof.We stress that Ichikawa et al. were led to postulate n-coordination, as in fig. 1, in order to fit their electron spin-echo modulation results (e.s.e.).8 Certainly the curve-fitting procedure illustrated in this work shows that the n-models apparently fit the experimental features better than the particular a-models selected for comparison. Thus it could be argued that for some reason there is a structural switch from a- to n-coordination on electron gain and loss. However, in our view, the e.s.r. results preclude this interpretation. The Silver Atom Centre There is a reduction in the isotropic coupling for lo9Ag of ca. 25 % for this centre relative to atomic silver.This is almost certainly due to delocalization onto the ligands as evidenced by the 14N interactions. The results show that there are four equivalent nitrogen nuclei and that the hyperfine coupling is almost isotropic. The estimated total 2s-orbital population is ca. 5%. This leaves a balance of ca. 20% spin density which can be assigned to 2p orbital population on nitrogen, and 5p population on silver. Whilst we do not claim any high accuracy for this calculation there can be no doubt that the 2p:2s ratio must be less than ca. 4. The small anisotropy resulting from this 2p population would be difficult to detect for a tetrahedral complex. However, had the n-structure been correct, the 14N isotropic coupling would have arisen indirectly from spin polarization.This is normally ca. 4% of the total orbital population, which would therefore need to be greater than 100% on N in order to produce an isotropic coupling of 6 G. This is clearly impossible. We conclude, firmly, that the ligands remain a-bonded to the silver atom following electron capture. The Agn Centre Similar arguments apply in this case. Since the complex is now square planar, and since the parallel and perpendicular g-tensor features are well separated, it is possible to estimate the extent of the anisotropy of the 14N coupling quite accurately. The apparent values are ‘Al1’(l4N) = 21 G and ‘A1’(14N) = 25 G. The apparent ‘ A l l ’ value is for field perpendicular to the plane of the complex and hence corresponds to A , for the I4N tensor both for o- or n-coordinated MeCN.The apparent ‘Al’ value is, for the a-complex, expected to be close to the average of the true All and A , values. This gives, for the true parameters, All = 29 G and Al = 21 G. This corresponds to 2s and 2p orbital populations of ca. 4.3 and 16%, again in reasonable accord with expectation for the a-structure (corrections for orbital magnetic and dipolar effects would reduce this ratio slightly, but would not modify the conclusions). For the n-complex, the difference between All and A1(I4N) should be very large, since the ligand orbital is now pure 2p on nitrogen. Hence the spectra should be more complexM. C. R. Symons et al. 273 3 than observed, with turning points dictated by the hyperfine anisotropy. If All and Al are taken as positive, as is normally the case, and if Aiso arises only from spin polarization, we again require an impossibly high 2p orbital population, and a much larger value for All (14N) than that observed.Hence the n-structure can be eliminated. We conclude that our original suggestions were satisfactory1* and that, in the process of relaxing to the square-planar structure, the MeCN ligands retain their a-coordination. One important conclusion of this work is that the predictions based on the fitting of the e.s.e. data are apparently not correct. Unfortunately, we are not in a position to check this work or to attempt to discover better agreement than that reported when using the a-model. We stress that in a wide range of studies, Kevan and coworkers have obtained e.s.e.data which appear to fit with expectation very precisely and, so far as we know, this is the only example in which there has been some apparent error in interpretation. Aspects of Mechanism We are left with the contradictory aspects of the interpretation of the mechanism of electron addition given in schemes 1-3. It is now clear that the parent AgI(MeCN), complex has tetrahedral, a-coordination. Since the major species formed at 77 K discussed above also has tetrahedral a-coordination, it is clearly the primary product of electron addition. Our work with mixed solvents2 shows that species A(2) or CN I, which exhibits hyperfine coupling to one 14N nucleus, is a secondary product. We stress that this species was not formed directly at 4 K, but only after photolysis.It seems probable that s -+p excitation of an Ago(MeCN), centre could well lead to loss of coordinated solvent molecules, though we remain puzzled at our inability to detect intermediate tri- or di-solvates. Finally, we return to the usage of the terms ‘solvation’ and ‘desolvation’. We reiterate that, in our view, the spectral changes observed on annealing for all systems in which Ago centres are derived from solvated AgI ions are best viewed as ‘desolvation’ processes. This is essentially because AgI ions are always strongly solvated in media such as water, alcohols and MeCN, whereas the neutral atoms are not expected to interact strongly. This is borne out by the fact that, on annealing, the end product from ‘solvated’ Ago centres is a species with very narrow features and parameters similar to those for matrix-isolated atoms. Thus we are, in effect, measuring the rejection of solvent molecules which are initially held in a configuration dictated by the parent Ag* ions. This is surely most appropriately viewed as a desolvation, just as the growth of primary solvation when 0; ions are generated from unsolvated 0, molecules is viewed as a solvation process. We than the S.E.R.C. for a grant to A.S. References 1 D. R. Brown, G. Eastland and M. C. R. Symons, Chem. Phys. Lett., 1979, 61, 92. 2 C. K. Alesbury and M. C. R. Symons, J. Chem. SOC., Faraday Trans. 1, 1980, 76, 244. 3 D. R. Brown and M. C. R. Symons, J . Chem. SOC., Faraday Trans. 1, 1977, 73, 1490. 4 M. C. R. Symons, J . Chem. Phys., 1978, 69, 3443. 5 L. Kevan, J . Chem. Phys., 1978, 69, 3446. 6 L. Kevan, H. Hase and K. Kawabata, J . Chem. Phys., 1977,66, 3834. 7 A. S. N. Li and L. Kevan, J. Chem. Phys., 1981,85, 2557. 8 T. Ichikawa, H. Yoshida, A. S. W. Li and L. Kevan, J. Am. Chem. SOC., 1984,106,4324. 9 K. Nilsson and A. Oskarsson, Acta Chem. Scand., 1984, A38, 79. 10 M. C. R. Symons and J. M. Stephenson, J . Chem. SOC., Faraday Trans 1, 1981,71, 1579. Paper 511872; Received 25th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202729
出版商:RSC
年代:1986
数据来源: RSC
|
16. |
Direct measurement of temperature-dependent interactions between non-ionic surfactant layers |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2735-2746
Per M. Claesson,
Preview
|
PDF (1293KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans. I , 1986, 82, 2735-2746 Direct Measurement of Temperature-dependent Interactions between Non-ionic Surfactant Layers Per M. Claessont and Roland Kjellandert Department of Physical Chemistry, Royal Institute of Technology, S-I00 44 Stockholm, Sweden Per Stenius Institute for Surface Chemistry, Box 5607, S-I 1486 Stockholm, Sweden Hugo K. Christenson* Department of Applied Mathematics, Research School of Physical Sciences, The Australian National University, GPO Box 4 , Canberra, ACT 2601 , Australia The force between two surfaces coated with pentaoxyethylene dodecyl ether, CI2E5, and immersed in aqueous solution has been measured as a function of surface separation in the temperature range 15-37 "C. The surfaces were prepared by allowing the non-ionic surfactant to adsorb on to hydrophobised mica from the solution.At 15 "C the interaction is repulsive at all separations, but above 20 "C an attractive minimum appears at separations below ca. 4 nm. The attraction increases rapidly with temperature and is identified with the interaction that gives rise to the phase separation ('clouding') in micellar solutions of many non-ionic surfactants as well as in the poly- (ethylene oxidekwater system. The net interaction is composed of two large parts, one entropic that is attractive and one enthalpic that is repulsive. The two parts nearly cancel, making the net interaction repulsive at low temperatures but attractive at high. They apparently originate mainly from the hydration interactions between the oxyethylene head groups on one surface with those on the other, and are of the same order of magnitude as in pure poly(ethy1ene oxidekwater solutions.The thickness of the non-ionic surfactant layers increases with temperature, implying a decrease in surface area per head group. This indicates that the intralayer head-group interaction also becomes more attractive (or less repulsive) with temperature. 1. Introduction In recent years much effort has been expended on investigating the intermolecular interactions that determine the properties of surfactant and lipid aggregates like micelles and bilayers in water. In parallel, there has been an increased interest in the interactions between such aggregates, particularly for systems with ionic or zwitterionic groups. For instance, Rand, Parsegian and coworkers' and later Marra and Israelachvili2 have investigated the forces between lipid bilayers in water.The short-range repulsive force between lecithin bilayers in water was the first example of the kind of interaction that has become known as the hydration force. A repulsive hydration force has also been measured in other ~ystems,~ and the attractive hydrophobic interaction measured t Also at the Institute for Surface Chemistry, Box 5607, S-11486 Stockholm, Sweden. $ Also at the Department of Applied Mathematics, Research School of Physical Sciences, The Australian National University, GPO Box 4, Canberra ACT 2601, Australia. 27352736 Temperature-dependent Forces between Non-ionics between hydrophobised mica s ~ r f a c e s ~ - ~ may be considered as a kind of hydration force as well.Comparatively little attention has been devoted to interactions in non-ionic surfactant systems (from which class we exclude the zwitterionics). The non-ionic surfactants are interesting not only because they have found many applications and are commercially very important but also for theoretical reasons. Their behaviour is quite complex; minor differences in the head groups or hydrocarbon chains often lead to dramatic changes in the phase diag~am.~-~ In particular, there are large shifts of the phase equilibria with temperature. The behaviour is also very sensitive to additives like salt or oil.l0 The sensitivity to these changes indicates that the interactions in the system are delicately balanced, most likely that large, opposing forces are operative.This would apply to both the intra- and the inter-aggregate interactions, which are related to each other. For instance, forces between the head groups of different aggregates must also be operative between the head groups within an aggregate. A prominent feature of many non-ionic surfactant-water systems is a large, closed solubility gap in the micellar phasees The gap is sometimes intersected by other phases like the lamellar The existence of the gap (specifically its lower consolute boundary) indicates that the forces between the surfactant molecules become more attractive with increasing temperature. The lower critical consolute point, often referred to as the cloud point, shows a strong, systematic variation with the molecular structure of the surfactant.For example, C1&67 has a critical point at ca. 48 "C, C12E, at ca. 27 "C, C12E, at ca. 4 "C and CloE, at ca. 44 "C. [Ref. (7) gives an extensive list of lower critical points.] The location of the critical point depends in a rather intricate manner on the balance between different contributions to the force between the micelles and on the micellar size and shape. It is not simply related to the strength of the interaction alone. The shape of the micelles is largely determined by the intra-aggregate interactions, and may change with temperature since the interactions do so. The range of existence of the various liquid-crystalline phases, like that of the hexagonal and the lamellar phases, also depends strongly on the surfactant.Here the interaggregate interactions also play an important role alongside the intra-aggregate ones. In this paper we present direct measurements of the forces between two surfaces coated with non-ionic surfactant molecules. This system models the surfactant aggregates mentioned above, though some aspects of the highly dynamic interface of real aggregates may be lost. Also, our experimental system does not reproduce the highly curved surfaces of surfactant aggregates, like micelles. Nevertheless, we believe that our results provide an essentially correct picture of the interaggregate interactions. We have determined the dependence of the interaggregate force on surface separation as well as on temperature. The temperature dependence is analysed with respect to the enthalpic and entropic parts of the interaction, and various contributions to the force are discussed. The thickness of the surfactant layers is also measured as a function of temperature, which gives some information about the intra-aggregate interactions. The picture of the non-ionic surfactant systems emerging from this investigation is in line with that given by Kjellanderll in his analysis of the phase behaviour of these systems.In order to measure the forces between the oxyethylene headgroups of the non-ionic surfactants it is necessary to ensure that the surfactants are oriented with the hydrocarbon tails towards each surface. This is accomplished by first making the mica substrate hydrophobic by deposition of a compressed monomolecular film of dioctadecyldi- methylammonium bromide.This cationic surfactant adsorbs electrostatically to the mica surface, forming a densely packed, hydrophobic surface. Measurements of the strong, 'hydrophobic attraction' between two such surfaces in aqueous solution will be t C, Em stands for C, H,,+,(OCH,CH,), OH.P . M. Claesson, R. Kjellander, P . Stenius and H. K. Christenson 2737 Fig. 1. Sketch of the C,,E,-coated surfaces. The mica surfaces are first covered with a double-chain cationic surfactant (dioctadecyldimethylammonium bromide, [Me],[C,,],NBr) oriented with the head group towards the mica and the hydrocarbon chains outwards. The very hydrophobic surface thus obtained is coated with the non-ionic surfactant by letting it adsorb from an aqueous solution.The oxyethylene groups of C,,E, face the aqueous phase. The radius of curvature of the crossed cylindrical surfaces is ca. 1 cm, and is greatly exaggerated in the figure. published elsewhere;6 here we allow the non-ionic surfactant to adsorb from solution onto this hydrophobic surface and measure the force between the resulting ‘ lamellae’ of non-ionic surfactants. (Fig. 1 shows a sketch of the surfaces.) Thus the surfactant molecules can adsorb or desorb at will, so that the area per head group can attain its equilibrium value during the experiments. This is similar to the situation for real surfactant lamellae. 2. Materials and Methods 2.1. Chemicals The dimethyldioctadecylammonium bromide was obtained in recrystallised form from Eastman Kodak. The pentaoxyethylene dodecyl ether was purchased from Nikkol Co., Japan and used as supplied (>99% by gas chromatography according to the manufacturer). The water used in the Langmuir trough and the measuring chamber was purified as follows, without being exposed to laboratory air : decalcination and prefiltration, followed by reverse osmosis, treatment with two mixed-bed ion exchangers, activated charcoal, Organex and finally a second filtration.The purification units were all Millipore products except for the final filtration stage through a 50 nm Nuclepore filter. A subsequent distillation caused no detectable improvement in the water quality. The purified water was deaerated for several hours prior to filling the measuring chamber (this stage was omitted for water used in the Langmuir trough).2.2. Surface Preparation The mica sheets were glued onto supporting silica discs with an epoxy resin (Epon 1004 from Shell Chemical Co.). The mica mounted on the silica discs was then made hydrophobic by deposition of dimethyldioctadecylammonium bromide using a standard Langmuir-Blodgett technique.6t l2 The spreading solvent was a 5 % solution of ethanol in n-hexane. The surfactant was deposited at a headgroup area of 0.6 nm2, corresponding to a surface pressure of 30-35 mN m-? The advancing contact angle of water on the2738 Tempera ture-dependen t Forces bet ween Non -ion ics hydrophobised mica was 94+2" for freshly prepared surfaces and did not change after storage in water for several days. 2.3. Force Measurements The surface forces apparatus designed by I~raelachvilil~ was used to directly measure the force as a function of separation between two hydrophobised mica surfaces immersed in an aqueous solution of pentaoxyethylene dodecyl ether.Multiple-beam interferometry is employed to measure the separation (to 0.1-0.2 nm) between the surfaces mounted in a crossed-cylinder configuration and the force is determined to N by observing the deflection of a spring on which one of the surfaces is mounted. The apparatus and experimental techniques have been described in detail in ref. (1 3). 2.4. Experimental Procedure The hydrophobised mica surfaces were initially brought into contact in dry nitrogen to establish that the contact zone was flat and free of particles. The surfaces were then separated and the measuring chamber filled with a solution of non-ionic surfactant in water.The concentration of C12E, was 6 x mol dm-3, which is slightly above the c.m.c. (4 x moldm-3).14 On introduction of the pentaoxyethylene dodecyl ether solution into the measuring chamber the non-ionic surfactant starts to adsorb onto the hydrophobic surfaces. This adsorption is a slow process taking several hours and can be followed by observing a decrease in the attraction and a gradual outwards (towards larger separations) shift in the contact position. Once equilibrium was established the measured forces did not change over periods of up to one week. The adsorption is reversible and if the solution is diluted by a factor of 100 at the end of an experiment an equally slow desorption takes place until eventually a hydrophobic attraction is once again measured. The force measurements were performed at varying temperatures using either (i) external control of the temperature in the room (1 5 and 20 "C) or (ii) an external heating fan directed towards the measuring cell (30 and 37 "C).The temperature in the cell was measured with a thermistor during the experiment. 3. Results The force measured (at equilibrium) at 30 "C between Cl,E5-coated surfaces is shown in fig. 2. (Contact, d = 0, has been defined throughout as the separation to which the surfaces come under a load of 0.1 N m-l. A further tenfold increase in the load compressed the system by only a few iingstrom, and the C12E5 monolayer on each surface remained intact.) At large surface separations (d > 5 nm) there is a repulsive double-layer interaction, in good agreement with a theoretical curve calculated15 from the solution of the non-linear Poisson-Boltzmann equation, using a Debye length I C - ~ of 100 nm and a surface charge density CT,, of 0.4 mC m-2 (1 charge per 420 nm2).This corresponds to a 1 : 1 electrolyte concentration of 9 x mol dm-3 and a surface potential at infinite separation vo0 of 46 mV. The surfaces appear to interact under constant-charge conditions but it is unclear whether the slight deviation from the theoretical double-layer force found at separations of 5-10 nm is due to some degree of charge regulation (which would lower the repulsion) or a van der Waals attraction or a combination of both. There is a force maximum at 4.5 nm and a steep, short-range repulsion at separations below 2.1 nm.At 2.1 nm there is a weak adhesion, the measured adhesion force being - 0.16 mN m-l. The long-range double-layer force is independent of temperature, but small variationsP . M . Claesson, R. Kjdander, P . Stenius and H. K . Christenson h 0.1 0.10 ----___ - 0.05 - - ,$ \. -. 0'050 20 40 60 80 100 \.* ? A ----_ --. / 1 2739 - I E 1.0 0.5 0 N I E E h --. G I / , t 10 I / ' -0.05 0 5 10 15 20 0.5 dlnm Fig. 2. The measured force F,/R (where R is the mean radius of curvature of the surfaces) between two C,,E,-coated surfaces at 30 "C as a function of surface separation. The corresponding free energy change per unit area, AGf, for flat surfaces is given by the right-hand ordinate scale (see section 4.2).The dashed portion of the curve indicates the region that is inaccessible by the force measurement technique. The inset shows the long-range tail of the force on a logarithmic scale. The dashed-dotted line in both the inset and the main figure is a force calculated from solutions to the non-linear Poisson-Boltzmann equation with constant charge boundary conditions and the following parameters: K-' = 100 nm, o,, = 0.4 mC m-2, corresponding to a 1 : 1 electrolyte concentration of 9 x mol dm-3 and ya, = 46 mV. 1.5 1.0 - 0.5 0.2 0.1 r4 I E E h 0: - 1.0 0 1 2 3 4 5 6 7 dlnm Fig. 3. The force between the C,,E,-coated surfaces at various temperatures for separations in the region of the minimum (linear scale). Symbols: 0, 15; 0, 20; A, 30; A, 37 "C.are found from experiment to experiment, due to slightly different surface charge densities and electrolyte concentrations. The short-range component of the force shows a dramatic change with temperature as indicated in fig. 3. At 15 "C the repulsion continues to increase with decreasing separation and becomes much larger than the extrapolated double-layer interaction below CQ. 3.5 nm. On increasing the temperature to 20 "C a weak minimum, still in the2740 Temperat we-dependen t Forces bet ween Non -ionics 10 0.1 0 1 2 3 4 5 6 d/nm Fig. 4. The short-range repulsive part of the force between the C,,E,-coated surfaces at various temperatures (logarithmic scale). Symbols: 0, 15; 0, 20; A, 30; A, 37 "C. Table 1. Thickness of C,,E, monolayer as a function of temperature T/OC thickness/nm 15 1.4 & 0.4 20 1.9 & 0.4 30 2.2 f 0.4 37 2.6 Ifr 0.4 repulsive regime, appears while the short-range repulsion decreases in range.At 30 "C the minimum has moved in and is now attractive, while the repulsion is further reduced. With increasing temperature the minimum becomes more attractive and shifts towards smaller separations as the short-range repulsion decreases and at 37 "C appears as indicated in the figure. The short-range repulsion has been plotted on a logarithmic scale in fig. 4. As can be seen, there is a large decrease in the repulsion with temperature: the measurable range is reduced fourfold on going from 15 to 37 "C. The thickness of the adsorbed non-ionic surfactant layers at the various temperatures is given in table 1.These values have been computed by comparing the absolute distance value at d = 0 for each temperature with contact measured between the bare hydrophobic surfaces in dry nitrogen, using experimentally determined corrections for the thermal expansion of the hydrophobised mica. 4. Discussion 4.1. Various Force Contributions The experiments show that a rise in temperature gives a decreased repulsion or an increased attraction (depending on separation and temperature) between the C,,E,-coated surfaces. The temperature dependence of the long-range repulsive double-layer force isP . M . Claesson, R. Kjellander, P . Stenius and H. K. Christenson 274 1 negligible over the range studied, and in the following discussion dealing with the change in interaction with temperature we may disregard it.The presence of a double-layer force indicates that the surfaces carry a slight charge, which most likely arises from an imperfect neutralization of the charges on the mica surface by the adsorbed cationic surfactant. Results from measurements on the purely hydrophobic surfaces, without adsorbed non-ionic surfactants, show very similar double-layer forces.6 This indicates that the long-range repulsive forces are not due to the non-ionic surfactants. The very strong repulsion (F,/R x 10-100 mN m-l) that is present for all temperatures at small separations is most likely due to steric effects. A repulsive contribution also arises when the compression of the surfactant layers is large enough to affect the adsorption-desorption equilibrium of the non-ionic surfactant.It is reasonable to assume the existence of a van der Waals attraction between the surfaces across the aqueous solution, but it is difficult to give an accurate estimate of the effective Hamaker constant (A) in the absence of dielectric data for the non-ionic surfactants. We can nevertheless assess its variation with temperature sufficiently well for our purposes. The temperature dependence of the van der Waals force is due almost entirely to the zero-frequency term,16 which is nearly proportional to kT, thus giving at most a 7 % increase over the studied temperature range for this term only. The remaining part of the interaction has the same sign and can accordingly only decrease the relative magnitude of this increase.Thus it is clear that the temperature dependence of any van der Waals interaction is also negligible compared to that observed experimentally. Obviously, the DLVO theory cannot explain the experimental results and we must invoke some 'additional force' to account for them. In discussions concerning interactions between non-ionic surfactant aggregates, it is not uncommon in the literature to a s ~ u m e ~ 1 ~ ~ 9 that there are two main force contributions: the attractive van der Waals force and some kind of hydration repulsion. It is then assumed that the repulsion decreases with increasing temperature due to a 'dehydration' of the headgroups; the water in the hydration shell becoming more like bulk water (often described as a breakdown of the hydration-shell water structure).The repulsion is thought to be important at low temperatures but to become insignificant compared to the van der Waals force at high temperatures. Another suggestion has been brought forward by Kjellander.ll In a fairly detailed analysis of the phase diagram and the thermodynamic properties of non-ionic surfactant micelles, he reached the conclusion that essentially the same kind of forces are dominant between the micelles as between poly(ethy1ene oxide) molecules in aqueous solution. In both cases the major force is composed of a large repulsive, enthalpic and a large attractive, entropic part, both apparently originating from the same molecular mechan- ism. At low temperatures the enthalpic part is largest and the net interaction is repulsive, while at higher temperatures the interaction is attractive due to a dominance by the entropic part.In the poly(ethy1ene oxidekwater systemlg the turning point is at ca. 15 "C, and the temperature dependence of the two parts is such that the net interaction remains increasingly attractive up to at least 230 "C (when the system is studied under constant volume conditions above 100 "C). It was suggested1'$ l9 that this interaction is a hydration force of an essentially hydrophobic kind, modified by the ether groups. This hydrophilic influence shifts the interaction towards less attraction or, for low temperatures, gives rise to a repulsion. Thus, in this view, the hydration repulsion does not simply vanish at increased temperatures, but it turns into a hydration attraction. It was also pointed outll, l9 that the breakdown of the hydration-shell structure at increased temperatures (as reflected by the heat capacity) actually disfavours the decrease in repulsion (or the increase in attraction), and that the 'dehydration' of the head groups therefore can not explain the observed behaviour. Our direct measurements of the forces between the C,,E, layers allow us to separate2742 Temperature-dependent Forces between Non-ionics out the enthalpic and the entropic parts of the interaction free energy.This gives further evidence supporting the latter viewpoint, as will be shown below, but owing to experimental limitations it has not been possible to determine directly whether the hydration force contribution remains purely repulsive, or whether it turns attractive.However, we will present some strong arguments to show that there most likely exists a hydration attraction at higher temperatures. 4.2. The Enthalpy and Entropy of Interaction The measured force, F, at separation d is where G, is the total free energy of the whole system, i.e. the medium between the cylindrical surfaces and the liquid in the remainder of the box (index c stands for cylindrical). Since the cylinder radius R w 1 cm % d, the Derjaguin approximation20 is valid : F,(d) = 2nRGf(d) (2) where Gf(d) is the corresponding free energy for a system with immersed planar surfaces of unit areas (index f stands for flat). The force F, is practically zero beyond some large distance d , < R, for which the Derjaguin approximation is still valid. The zero point of the free energy Gf(d) can therefore be taken as Gf(d,), which of course for all practical purposes equals Gf(co).Thus we have for flat surfaces where AG,(d) = Gf(d)-Gf(co); the zero point of G now being arbitrary. The temperature dependence of F, is given by a&(d,T) 1 aAG,(d,T) ( aT ),==( aT )p = -ASf(d, T ) (4) where AS,(d, T ) = Sf(d, T ) -Sf(oo, T ) equals the change in entropy at constant tempera- ture for the whole system when two macroscopic, flat surfaces of unit area are taken from infinite separation to separation d. The corresponding change in enthalpy is and we have as always AGf(d, T ) = AHf(d, T ) - TASf(d, T ) . (6) Now, from the measurements of & at the different temperatures, we obtain from eqn (4) that ASf(d, T ) w 60 35 pJ m-2 K-l when 0.8 5 d/nm 5 2.1 and 15 5 T/"C 5 37, which means that the entropy contribution to AGf equals TASf(d, T ) w 18 10 mJ m-2.(7) The large error limits originate mainly from the uncertainty in separation of the surfaces; it is hard to obtain an accurate estimate of the difference in force at two different temperatures but at the same separation. An error in the separation of a few ingstrom can change the force by a large amount. Because of the errors involved, we were not able to determine the d- and the T-dependences of AS,, but only its approximate magnitude in the intervals indicated.P. M. Claesson, R . Kjellander, P. Stenius and H. K . Christenson 2743 The values of AG, are obtained from eqn (3) by using the measured Ii, in the corresponding d-interval for 20 5 T/"C 5 37 -0.1 5 AG,/mJ m-2 5 4 (8) (where the lower limit shows the value of the minimum at 37 "C and the higher shows the value at d = 0.8 nm and T = 20 "C).Comparing TAS, in eqn (7) with AG,, we see from eqn (6) that AHf has to be of about the same magnitude as TAS,, i.e. AH,(d, T) w 20k 12 mJ m-2. (9) Note, that the large error limits of TAS, and AH, only apply to their absolute values; their difference, AGf, is known with a much higher precision. Thus the force between the surfaces is composed of two large, opposing parts that nearly cancel : one entropic that is attractive and one enthalpic that is repulsive. In this situation, only a small change in temperature is necessary in order to shift the balance of the two terms and make the net force change between repulsion and attraction.To the first order in T, the contributions from the temperature dependence of AH, and ASf cancel in AG, (as always when the external pressure is constant), and hence they also cancel in the force. Provided A T is not too large we have A(AGf) w A(AHf) - TA(ASf) - (AS,) A T M -ASfAT (10) which implies that AG, in eqn (6) mainly changes through the factor T. Near the temperature where the force F, changes from repulsion to attraction (where AG, = 0), the balance between the enthalpy and entropy terms is shifted through a change in this factor T. Thus we see that the attraction originates from a dominant entropic part of the force at the higher temperatures. Of course, the large entropic part is present also at lower temperatures when the force is repulsive due to the preponderance of the large enthalpic term and vice versa. Thus it is not a question of a disappearance of a repulsive contribution (or an appearance of an attractive one); there is only a slight shift in balance between two large terms of each kind present at all temperatures in question. Since the AG, at the attractive minimum for the temperatures 30 < T/"C < 37 is rather small, it is of interest to assess the influence of the van der Waals attraction in this temperature range.In fact, by using a quite reasonable Hamaker constant ( A = 5 x J), we obtain a theoretical AG,,, that has about the same magnitude as the total AG, observed at the minimum. (For this value of A , the DLVO theory correctly describes the force curves from the force maximum at 4.5 nm and outwards.) However, the enthalpic and entropic contributions to AG,,, = AH,,, - TAS,,, are completely negligible compared to the observed, total AH, and TASf.As an example, consider the van der Waals interaction between two hydrocarbon phases across a water phase, in which case the entropic contribution is unusually large.16 The TASVdW component is attractive and it is ca. 45% of AGvdw. Since AHvd, is also attractive, the magnitudes of both TAs,dw and AH,,, are less than AGvdW M AG, (at the minimum), which is far smaller than the total AHf and TAS,. Accordingly, it is only when the dominant contributions to AH, and TAS, nearly cancel each other in AG,, i.e. in a temperature range around the switch-over between repulsion and attraction, that the van ders Waals force can play any significant role.Even so, it will not change the overall behaviour - (the shift from repulsion to attraction due to the large AH, and TAS,) but only affect the balance so that the switch-over will occur at a different temperature than it would in the absence of a van der Waals force. The experimental temperature range apparently lies close to the switch-over between repulsion and attraction, where the contribution from the hydration force is small. However, from the value of AS, follows that the attraction at a fixed separation d (at2744 Tempera t we-dependen t Forces be tween Non- ionics least within 0.8 5 d/nm 5 2.1) will increase by some 0.6 mJ m-2 for a temperature increase of only 10 "C.For comparison, the total van der Waals attraction AG,,, in this range lies between 0.2 and 0.03 mJ m-,. Therefore, one does not need to go to much higher temperatures before the hydration attraction dominates over the van der Waals force for a fixed d. This extrapolated behaviour is a direct consequence of the large value of AS,. The hydration force must, in fact, become attractive unless the entropy suddenly drops from its high value to practically zero at some temperature close to our experimental range. Such a decrease could only happen at a phase transition in the surface layer system. Furthermore, the transition temperature has to be dependent on the surface separation. Since there exists no evidence of a phase transition, we conclude that the hydration repulsion observed at low temperatures will not simply vanish at higher temperatures as ~ u g g e s t e d , ~ ~ ~ ~ ~ ~ ~ but that it will turn into a hydration attraction.Some further arguments supporting this conclusion are given in the next section. 4.3. Comparison with Related Systems From the results in the previous section it is clear that the interaction between the C12E5 coated surfaces qualitatively is very similar to that between non-ionic surfactant aggregatesll and between poly(ethy1ene oxide) segments in waterlg as described in section 4.1. Therefore, it is reasonable that the measured short-range interaction originates from the interactions between the oligo(ethy1ene oxide) residues on the two surfaces as suggested in ref.(1 1). To see that this is indeed very likely to be the case, let us compare the interactions between poly(ethy1ene oxide) molecules in aqueous solution with the AH, and TAS, values obtained in section 4.2. If we assume that a (-CH,-CH,-0-) unit (= E) occupies, say, 20 A2 on the C12E5 coated surfaces, we obtain AH, w TAS, w 2 kJ mol-l. Since the interaction between the surfaces surely involves more than one pair of E units when counted per 20 A2 surface area (there are several E units below the top ones on each surface), the 2 kJ mol-1 ought to be some small multiple of the H and TS parts of the average pair interaction (disregarding many-body effects). The corresponding quantities in the pure(ethy1ene oxidekwater system can be estimated from the enthalpic and entropic parts of the interaction parameter w [cf.ref. (19)], which is related to the second virial coefficient. These two parts are ca. 4 kJ mol-l each, as obtained from calorimetric and vapour pressure measurements compiled in ref. (19). This value is also some small multiple of the H and TS parts of the average pair interaction (in lattice theories the multiple equals half the coordination number). Therefore, we can conclude that the enthalpic and entropic parts of the interaction free energy are of the same order of magnitude in both cases. For the poly(ethy1ene oxidekwater system the entropic part of the interaction free energy remains positive at least up to 230 "C,lg which means that the interaction between the oxyethylene units remains increasingly attractive up to that temperature. This system has a lower consolute point ('cloud point') between 96 and 180 "C depending on the chain length.21 Thus the attraction increases more rapidly than kT up to 180 "C (otherwise there would be no phase separation). It is reasonable to assume that AS, and AHf for the C12E5 coated surfaces as well as the corresponding quantities for the non-ionic surfactant aggregates have a similar temperature dependence as for poly(ethy1ene oxidekwater.This implies that AS, remains positive and the interaction increasingly attractive up to very high temperatures. In fact, there exist C, En surfactant systems that have lower consolute points well above 90 O C a 9 Remember that the location of the cloud point in these systems is not simply given by the strength of the interaction, but also depends on several other factors like the chain length of the polymer or, in the surfactant systems, the micellar size and shape.In ourP . M . Claesson, R. Kjellander, P . Stenius and H . K. Christenson 2745 experiments we measure the force between lamellae both above and below the temperature where C,,E, has a cloud point in the micellar state (27 "C). The suggestionll9 l9 that the interaction between the oxyethylene segments is essentially of a hydrophobic kind is supported by the following facts. The hydrophobic interaction is dominated by a strongly attractive entropic contribution, the temperature derivative of which (aS/aT) is large and negative22 (cf. the anomalous heat capacity of mixing for typically aqueous solutions).The same is true for poly(ethy1ene oxide)-water.lg For small inert molecules in water the enthalpic contribution to the hydrophobic interaction is repulsive, while it becomes attractive for larger hydrophobic molecules (probably because the molecule cannot be contained in a small clathrate-like cage, implying that hydrogen bonds must be broken). For molecules with a single hydrophilic group like monoalcohols, the enthalpic contribution is more repulsive than for the corresponding hydrocarbon, but the thermodynamic behaviour is still dominated by the hydrophobic effect from the hydrocarbon groups. (The repulsive, hydrophilic contribution in com- bination with the ideal entropy of mixing makes the molecules water-soluble when the hydrocarbon chain is short despite the hydrophobic interaction.) Poly(ethy1ene oxide) behaves mainly like molecules in the latter group. The enthalpic repulsion is sufficiently strong in this case to outweigh the entropic attraction at low temperatures.This is probably due to the fact that it is possible to have a hydrophobic type of hydration around the methylene groups (cf. small hydrophobic molecules) and at the same time have hydrogen bonds to the ether oxygens.lS 5. Conclusions and Implications The hydration force between oxyethylene groups that we have measured is strongly temperature dependent because of an enthalpy-entropy compensation in the interaction mechanism. An important consequence of this situation is that the force can be repulsive or attractive depending on the temperature.It dominates the overall behaviour of the total interaction for large variations in temperature. It is interesting to note that the temperature of switchover from repulsion to attraction lies in the temperature range of most importance for aqueous systems at normal pressure, i.e. between 0-100 "C. This is most likely a major reason why poly(ethy1ene oxide) and the related surfactants have many applications. Owing to the self-annihilation of the hydration force at its switch-over temperature, the total force between, for example, surfactant aggregates will depend quite strongly on other contributions to the interactions near that temperature. Furthermore, small modifications of the very large enthalpic or entropic contributions to the hydration force will lead to relatively large changes in free energy. The transition temperatures for various phases will therefore be sensitive to changes in surfactant chain and to the concentration of various additiveslot 23 as observed.The hydration force may be described as a modified hydrophobic interaction. The hydrophobic methylene groups apparently have a major influence on the behaviour (giving a large, negative entropy and a fairly large, negative enthalpy of hydration) while the hydrophilic ether groups modifies it (mainly making the enthalpy even more negative). The oxyethylene groups become more and more hydrophobic with increased temperatures [up to at least 180 "C, as judged from the phase behaviour of poly(ethy1ene oxide)-water]. At low temperatures, on the other hand, the behaviour approaches that of more hydrophilic systems, and the hydration force turns repulsive.The increase in hydrophobicity of the head groups with temperature is, for instance, apparent in the ternary systems non-ionic surfactant-oil-water, where the oil solubility increases with temperature.l0 The location of the microemulsion phase is also affected in the same manner. Our measurements of the thickness of the non-ionic surfactant layers indicate that the 91 FAR 12746 Temperature-dependent Forces between Non-ionics intra-aggregate interactions between the headgroups also turn more attractive (or less repulsive) with increased temperature : the area per headgroup decreases and the layer thickness increases. Since a decrease in area affect the shape of surfactant aggregate^,^^ this implies that one would expect the sequence: spherical + rod-shaped + oblate aggregates + inverse structures when increasing the temperature.At least the first three members of this sequence apparently occur in the isotropic phases of some C , Em-water systems,llT 25 but this is subject to a considerable controversy in the literature. Note that one would expect a substantial variation in behaviour depending on the surfactant chain lengths, for similar reasons as in the case of the interaggregate interaction. This means that some surfactant systems may show a growth of the micelles when the temperature is increased, while others may only show spherical micelles all the way up to the cloud point, if qny. In addition, at high surfactant concentrations, the interaggregate interactions also influence the size and shape of the aggregates.Since both the critical temperature and the critical concentration are strongly dependent on the micellar shape (within certain limits), this fairly intricate interplay between the intra-aggregate forces (via the shape) and the inter-aggregate ones is a further reason for the great variability of the phase diagrams of the non-ionic surfactant-water systems. We are grateful to Prof. B. W. Ninham for his encouragement of this work and to our colleagues in the Department of Applied Mathematics for helpful comments on the manuscript. References 1 D. M. Le Neveu, R. P. Rand and V. A. Parsegian, Nature (London), 1976, 259, 601. V. A. Parsedan, 2 J.Marra and J. N. Israelachvili, Biochemistry, 1985, 24, 4608. 3 R. M. Pashley, J. Colloid InterJcace Sci., 1981, 80, 153. R. M. Pashley and J. N. Israelachvili, J. Colloid 4 J. N. Israelachvili and R. M. Pashley, J. Colloid Interface Sci., 1984, 98, 500. 5 R. M. Pashley, P. M. McGuiggan, B. W. Ninham and D. F. Evans, Science, 1985,229, 1088. 6 P. M. Claesson, P. Herder, C. Blom and B. W. Ninham, J. Colloid Interface Sci., in press. 7 J. S. Clunie, J. M. Corkill, J. F. Goodman, P. C. Symons and J. R. Tate, Trans. Faraday Soc., 1967,63, 2839. J. S. Clunie, J. F. Goodman and P. C. Symons, Trans. Faraday SOC., 1969,65, 287. 8 J. C. Lang and R. D. Morgan, J. Chem. Phys., 1980, 73, 5849. 9 D. J. Mitchell, G. J. T. Tiddy, L. Waring, T. Bostock and M. P. McDonald, J. Chem. Soc., Faraday Trans. 1 , 1983, 79, 975. 10 S. Friberg and I. Lapczynska, Progr. Colloid Polym. Sci., 1975, 56, 16. H. Saito and K. Shinoda, J. Colloid Interface Sci., 1967, 24, 10. K. Shinoda and H. Takada, J. Colloid Interface Sci., 1970,32, 642. 11 R. Kjellander, J. Chem. SOC., Faraday Trans. 2, 1982, 78, 2025. 12 J. Marra, J. Colloid Interface Sci., 1985, 107, 446; H. K. Christenson, J. Phys. Chem., 1986, 90, 4. 13 J. N. Israelachvili and G. E. Adams, J. Chem. Soc., Faraday Trans. I , 1978, 74, 975. 14 P. Mukerjee and K. J. Mysels, Critical Micelle Concentrations of Aqueous Surfactant Systems (U.S. 15 D. Y. C. Chan, R. M. Pashley and L. R. White, J. Colloid Interface Sci., 1980, 77, 283. 16 V. A. Parsegian and B. W. Ninham, Biophysics J., 1970, 10, 664. 17 E. J. Staples and G. J. T. Tiddy, J. Chem. Soc., Faraday Trans. I , 1978,74,2530; G. J. T. Tiddy, Phys. 18 L. Reatto and M. Tau, Chem. Phys. Lett., 1984, 108, 292. 19 R. Kjellander and E. Florin, J. Chem. Soc., Faraday Trans. 1, 1981,77, 2053. 20 B. V. Derjaguin, Kolloidn. Zh., 1934, 69, 155. 21 S. Saeki, N. Kuwahara, M. Nakata and M. Kaneko, Polymer, 1976, 17, 685. 22 C. Tanford, The Hydrophobic Eflect : Formation of Micelles and Biological Membranes (John Wiley, 23 P-G. Nilsson and B. Lindman, J. Phys. Chem., 1984,88, 5391. 24 J. N. Israelachvili, D. J. Mitchell and B. W. Ninham, J. Chem. Soc., Faraday Trans. 2, 1976,72, 1525. D. J. Mitchell and B. W. Ninham, J. Chem. Soc., Faraday Trans. 2, 1981, 77, 601. 25 P-G. Nilsson, H. Wennerstrom and B. Lindman, J. Phys. Chem., 1983,87,1377; W. Brown, R. Johnsen, P. Stilbs and B. Lindman, J. Phys. Chem., 1983, 87, 4548. P-G. Nilsson and B. Lindman, J. Phys. Chem., 1984,88,4764. Paper 511864; Receiued 26th October, 1985 N. Fuller and R. P. Rand, Proc. Natl Acad. Sci. USA, 1979, 76, 2750. Interface Sci., 1984, 101, 511. National Bureau of Standards, Washington, 1970). Rep., 1980, 57, 1. New York, 1980).
ISSN:0300-9599
DOI:10.1039/F19868202735
出版商:RSC
年代:1986
数据来源: RSC
|
17. |
Photocatalytic dehydrogenation of liquid propan-2-ol by TiO2. Part 2.—Mechanism |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2747-2754
Ian M. Fraser,
Preview
|
PDF (600KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans I, 1986,82, 2747-2754 Photocatalytic Dehydrogenation of Liquid Propan-2-01 by TiO, Part 2.-Mechanism Ian M. Fraser and James R. MacCallum Department of Chemistry, University of St Andrews, St Andrews KY16 9ST, Scotland The photodehydrogenation of propan-2-01 mediated by TiO, has been studied. Using basic additives it has been established that the sites responsible for the bulk of the reaction are acidic in nature. A fibre-optic monitoring system has been used to study changes in absorption of radiation by the TiO, in the presence of N,, 0, and H,O,. It is concluded that the active sites are Ti4+ ions. A mechanism involving 0; and H,O, is proposed. The photodehydrogenation of aliphatic alcohols, especially propan-2-01, mediated by TiO, has been studied in both gas and liquid Some doubt exists about the nature of the surface sites involved in the reaction.It has been ~uggested,~ for example, that the alcohol is chemisorbed at coordinatively unsaturated 0,- ions followed by photohole trapping to produce activated 0, which reacts to give ketone and H,O,. On the other hand, the 0; is proposed as the reactive intermediate,, with the site of reaction unspecified. In Part 1 it was shown4 that the kinetics of photodehydrogenation of propan-2-01 could be studied by following the absorption of 0,. It was established that the rate of reaction depended on the concentration of non-associated alcohol in the neat liquid or in solution in inert solvent. It is the object of this paper to present experimental details regarding the nature of the proposed active sites on which the alcohol conversion and oxygen consumption processes occur.The involvement of Ti3+ on the TiO, surface has been established by using a specially designed fibre-optic monitoring apparatus. Both 0, and H,O, react with this species. Blocking of surface sites with subsequent reduction in oxygen-uptake rates was observed upon adding a strong base, 1,4-diazabicyclo[2,2.2]octane (DABCO). In addition, the effect of varying the incident light intensity on oxygen-uptake rates was examined. The observed dependence of uptake on the square root of the intensity is in agreement with studies in which the rate of propanone formation was monitored,, thereby confirming the validity of the experimental technique used.Experiment a1 Experiments involving oxygen-uptake measurements were carried out using the apparatus described previ~usly.~ For the intensity dependence studies, neutral density filters (Barr and Stroud) were positioned normal to the incident light beam, between the lamp and uptake cell. Detection of H202 on the propan-2-01-anatase slurry was determined qualitatively by first centrifuging a slurry sample and then adding a few drops of the resulting clear liquid to an acidic solution of Ti"' ion. The characteristics orange-yellow colouration due to Ti Iv-H ,O, complex formation was observed. The interaction of H,O, and 0, at the anatase surface was studied by monitoring the 2747 91-22748 Photocatalysis of Propan-2-01 by TiO, transmitted visible light passing through the slurry (same composition as before) to the vortex generated by a magnetic stirrer.A special Pyrex cell was constructed of ca. 50 cm3 capacity which was placed in the sample compartment of a spectrofluorimeter (Perkin- Elmer MPF-2A). Excitation was provided by a medium-pressure mercury arc lamp with stabilised power supply (Hanovia). A tube was used to direct the beam onto the cell with the minimum amount of scatter of incident light. The cell itself was masked, such that light could only reach the vortex by passing through the slurry. One end of a fibre-optic bundle was slotted into the top of the cell, while the other was directed into the analysing monochromator of the spectrofluorimeter. By using this technique, the intensity of mercury lines passing through the vortex could be monitored and the influence of H202 on this distribution recorded.Surface blocking effects on anatase were studied by using 1,4-diazabicyclo[2.2.2]octane (DABCO) (Aldrich). The anatase and rutile used throughout the experiments were uncoated and supplied by Tioxide, (CLD 1729/F) and (CLD 1729/A), respectively. The propan-2-01 used was of Analytical grade (Fisons). Results The role of lamp intensity in influencing the rate of conversion of propan-2-01 to propanone has been studied by others,’ who found that conversion rates varied linearly with intensity at extremely low light intensities, while a I$ dependence applied for intensities varying over a higher range. Such observations are consistent with a process in which electron-hole recombination is the dominant path to exciton deactivation in a semiconductor.Under these circumstances, trapping of excitons by surface traps would be much less probable. Determination of the oxygen-uptake rate dependence on intensity was carried out by employing the procedure outlined in the Experimental section. Three suitable neutral filters were available. The relationship between uptake rate (R) and intensity ( I ) may be written as R K In (1) where n is the order of dependence. A plot of R against the square root of the normalised incident lamp intensity is shown in fig. 1. Linear-regression analysis of the plot indicates a high degree of linearity (correlation coefficient of 1 .OO). Clearly, therefore a square-root dependence of oxygen uptake rate on intensity applies for our experimental conditions. This observation establishes comparability between monitoring oxygen-uptake rates and measuring rates of propanone formation established previ~usly.~ The identification and characterisation of the nature of surface sites has received considerable attention in the literat~re.~? Defect sites on anatase (potential exciton traps) have been identified as being acidic.1° The role of such sites in the photodehydro- genation of propan-2-01 was tested by adding base to the reaction mixture.The most important factors governing selection of the base were that it should not absorb any radiation entering the uptake cell and that it be stable to oxidati.on. The base selected was 1,4-diazabicyclo[2.2.2]octane (DABCO). This material was added to the usual slurry composition (1 g anatase per 100 cm3 propan-2-01) to give different DABCO concentrations.On carrying out an oxygen-uptake run at 298 K, the slope of the linear region was reduced as compared with a run using identical conditions of lamp position and intensity but without the added base. This effect is shown in fig. 2, where the uptake isotherms have been corrected for the initial equilibration period. Table 1 shows the effect on the rate of reaction of varying the concentration of DABCO. The method for determining the extent of alcohol association described previ~usly,~ was used to confirm that DABCO has no detectable affect on the self-association properties of propan-2-01 in the concentrations used. Hence, the rate reduction cannot be explained by a perturbation of alcohol association by DABCO.Clearly therefore, these experimentalI. M . Fraser and J. R. MacCallum 2749 0.2 0.4 0.6 0.8 1.0 1 . 2 13 Fig. 1. Plot of the are of oxygen uptake against the square root of the relative lamp intensity. All experiments carried out at 302 K and for anatase-propan-2-01 suspensions. 100 80 mE -!?. 60 -3 44 n 40 20 0 10 20 30 40 50 60 t/min Fig. 2. Effect of the presence of a strong base (DABCO) on the oxygen uptake rate: 0, anatase-propan-2-01 only ; A, anatase-propan-2-01-1 x 1 O-, mol dm-3 DABCO. Both determina- tions carried out at 296.5 K. (Isotherms are normalised to take into account the initial equilibration period.) Table 1. Effect of added base to rate of 0, consumption relative [DABCO] rate /mol dm-3 1 0 0.3 1 10-2 0.15 10-1 0.46 10-32750 Photocatalysis of Propan-2-01 by TiO, 0 4 0 20 30 40 50 60 70 tlmin Fig.3. Oxygen uptake isotherm at 298 K for an anatase-propan-2-01 suspension containing ca. 2 cm3 of a 30 wt% solution of H202. observations can be regarded as absorption of the DABCO at acidic surface sites on anatase such that surface adsorption of alcohol is inhibited. Analyses of DABCO solutions containing the TiO, slurry gave no detectable change in the Concentration of base, indicating that the number of sites interacting with additive was very small. The participation of H,O, in the mechanism of photodehydrogenation has been suggested previously,2 although the exact nature of H,O, involvement has not been established. The presence of this species in air anatase-propan-2-01 slurry samples was detected by the method outlined in the Experimental section.This observation is in agreement with other workers, who also detected low concentration of H20, in similar compositions.l. Production of this species could only be detected after irradiation of the slurry, although no attempt was made to quantify the amount produced. The effect on oxygen-uptake rate of added H,O, was tested by adding ca. 2 cm3 of a 30 wt % solution of H,O, to a fresh anatase-propan-2-01 slurry. After equilibration in the dark at 298 K, the uptake cell was connected with the lamp source. After the normal thermal equilibration period, a fast rate of oxygen consumption was observed, lasting for ca. 32 min. This slope was ca.30% greater than that observed without addition of H,O,. After this time, the rate of uptake decreased to give a new linear slope, the gradient of which was close to the expected value at 298 K, in the absence of added H,O,. The oxygen-uptake isotherm is shown in fig. 3. Rudham et al., monitored the concentration of H,O, added to a photolysed Ti02-propan-2-ol slurry to which a fixed amount of peroxide had been added. They detected a fall in H,O, concentration until a low, steady value was reached. It is clear from the experimental observation outlined previously that H,O,, when added to the oxygen-uptake cell, does not decompose to give gaseous products. If this were the case, then an enhanced oxygen consumption rate would not have been observed. An n.m.r. measurement showed no modification of the association properties of propan-2-01 by H,O, in the concentration used in the above experiment.Furthermore, irradiation in the cell of H20, dissolved in propan-2-01 in the same proportions without added TiO, gave no consumption of 0,. The direct photodecomposition of the peroxide does not result in measurable reaction. Hence, it must be concluded that the H,O, interacts with the surface such that more sites are available to incoming reactant molecules.I. M . Fraser and J . R . MacCallum Table 2. Effect of different experimental conditions on the intensity of mercury lines transmitted through the vortex of a magnetically stirred anatase-propan-2-01 slurry intensity of 546 nm line intensity of 436 nm line experimental conditions ratio of (1) sample equilibrated with (2) 0, purged for 20 min (3) N, purged for 40 min (4) condition (3) immediately followed by addition of 2 drops of H,O, air 2.32 4.00 0.80 1.23 275 1 It was observed that if oxygen is removed from the slurry during irradiation, for example by purging with nitrogen, the normal intense white colouration of the TiO, became dull.This phenomenon applies to both anatase and rutile and as been attributed to excited electrons trapped at surface sites.ll These reduced surface sites have been identified by e.s.r. spectroscopy.12 In the presence of oxygen it is believed that oxidation of such sites can occur with subsequent formation of the superoxide radical l4 In order to obtain further information regarding the kinetic behaviour outlined previously, and in particular the role of H,O, in influencing these processes, a special cell was constructed.This utilised a fibre-optic bundle to monitor only that portion of a medium-pressure mercury lamp output which reached the vortex of a magnetically stirred anatase-propan-2-01 slurry. Direct connection of the fibre-optic system to the lamp revealed that it was capapable of transmission down to the near-u.v. region. Sampling of the light passing through the vortex was carried out while oxygen or nitrogen was passed through the slurry. As expected, the 366 mm line of the mercury lamp output was completely absent, being totally absorbed by the anatase. When rutile was substituted for anatase, the 405 nm line was also absent. These observations are excellent confirmation of the fact that the band gap of rutile is less than that of anatase.On exposure of the anatase-propan-2-01 slurry to air, the vortex spectrum showed the yellow-green (YG) lines (at 546 and 519 nm) to be more intense than the blue (B) (405 and 436 nm). This YG: B ratio increased gradually on purging with oxygen, reaching a maximum after ca. 20 min. However, on passing nitrogen through the slurry during irradiation, the trend gradually reversed, with the intensity of the blue line predominating, culminating in a minimum YG: B ratio after ca. 40 min. This trend of line intensity ratios is shown in table 2. By carefully ensuring that the optical alignment remained constant, it was found that all changes in YG:B ratios could be accounted for by alterations in the absolute intensities of the yellow-green (> 500 nm lines).Stronger absorption of these lines by the slurry was observed under nitrogen than in the presence of oxygen, with the lines at 405 and 436 nm remaining at constant intensity. Discussion Both 0, and H,O, reduce absorption of the yellow-green lines emitted by the mercury lamp. It is known that Ti3+ species absorb around 520 nm15 and thus it is reasonable to conclude that the variation in absorption above 500 nm is due to the presence of this ion. The conversion of TiO, from a dull grey to a white colour can be accounted for by the following processes : Ti74 + 0, + Tit; + O;(absorbed) Ti?; + H,O, -, Ti4+ + OH(absorbed) + OH-(absorbed).2752 Photocatalysis of Propan-2-01 by TiO, The latter process has been studied in solution,l6 where e.s.r.evidence has revealed that the radical species produced was stabilised by complexation with Ti4+ ions. It has also been shown that alkoxy and peroxy radicals can form stable ligands with transition metals. l7 The results reported in this and the previous paper have established that (a) the amount of free hydroxyl groups determines the rate, (b) addition of strong base inhibits the dehydrogenation and (c) the reaction proceeds by converting Ti3+ to Ti4+. Previously observed effects have been reproduced using the gas absorption to follow the kinetics of conversion, viz. (a) the production, and then consumption, of H,O, during the reaction and (b) the dependence of the rate of reaction on the square root of the intensity of radiation.Other workers have demonstrated the involvement of specific sites on the crystal surfaces, although the number of such sites was not e~timated.~ The following mechanism is proposed : Ti!& +HOCH(CH,), =Tit&[HOCH(CH,),](ads) (1) Tit&[HOCH(CH,),](ads)+e -+ Tif&[HOCH(CH,),](ads) (2) Tif&[HoCH(CH,),](ads)+O, -+ Ti$j[O;, HOCH(CH,),](ads) (3) Tii',?[O;, HOCH(CH,),](ads) -+ Ti?&[HO;, HOC(CH,),](ads) (4) Tit$[HO;, HOC(CH,),] -+ Ti3+[H,0,, (CH,),CO](ads) ( 5 ) Tif&[H,O,, (CH,),](ads) -+ Ti!& + H,O, + (CH,),CO. (6) Ti?$ + h -+ Ti4+ -+ step (1) Ti:& + HOCH(CH,), + Ti:&[HOCH(CH,),] --+ step (3) Ti?& + H,O, -+ Tii',S[oH, OH-](ads) Ti!$[oH, OH-] + HOCH(CH,), -+ Tit&[HOc(CH ),, OH-] Tif&[HOC(CH,),, OH-] + h --+ Tit$ + (CH3),C0 + H,O.The surface TifJ can undergo a number of reactions, thus (7) (8) (9) (10) (1 1) + H,O The overall effect of this mechanism is that two pathways interlink. The first, involving photoelectron trapping, produces 1 mol of H,O, for each mole of 0, and alcohol consumed. The second pathway produces 2 mol of water for each mol of H,O, and alcohol used up. The observed experimental behaviour can be rationalised in terms of this proposed mechanism. The alcohol is absorbed by coordination of the lone pair of the oxygen to the acidic surface site, resulting in an equilibrium as given in step (1). Blockage of this site by addition of a strong base, DABCO, results in a diminution in the rate of reaction. The increase in absorption of radiation around 500 nm on irradiation in N, is explained by step (2) and decrease on the addition of 0, and H,O, by steps (3) and (9). H,O, is produced by step (6) and removed by step (9); thus during a run its concentration rises, or if previously added in high concentration falls, to a steady value determined by the relative rates of the two processes.Steps (2) and (1 1) involve electrons or holes produced by photoexcitation of the TiO, crystals. The dependence of the rate on the square root of the intensity of radiation has been accounted for by the domination of h+e annihilatioq2 with a small fraction escaping to the surface sites and becoming involved in the photoinduced dehydrogenation. It is not possible to develop a kinetic expression for such a complex sequence of steps.Z.M . Fraser and J . R. MacCallum 2753 However, it an be concluded that step (2) plays a significant role in determining the rate of reaction since our observations give the following experimental relationship : rate = kZi[(CH,),CHOH]. The rate for step (2) is given by k,[e] [Tit;] [HOCH(CH,),], and since [Tit; HOCH(CH,),](ads) = K[Tii',t] [HOCH(CH,),] where K is the equilibrium constant for step (l), and [el = k , fi then rate = k, IiK[Tit4] [HOCH(CH,),]. All that can be deduced about the surface sites is that they are few in number. However, the regenerative cycle ensures their reactivity. It has been suggested that the constant that governs the concentration of electrons arriving at the surface is temperature dependent, and thus the observed activation energy will be a composite involving the enthalpy of adsorption and the activation energy required for electrons to populate the surface., The mechanism proposed by Cunningham and Hodnett, following their gas-phase studies is developed on coordinatively unsaturated 0; as the active sites. The results reported in this paper are not consistent with such a proposal. A strong base such as DABCO would not interact with 0; sites, and the variation in absorption of radiation cannot be explained on this basis.A major difference between our observations and other work2 on solution photo- dehydrogenation is the effect of 0, on the rate of reaction. We found that going from an atmosphere of air to pure 0, at laboratory pressure increased the rate by 5% for pure alcohol, whereas Rudham and coworkers report a larger increase.2 Note that the rates of conversion, and therefore the intensity of radiation, were much higher in Rudham's experiment.A significant feature in our mechanism which relates to the role of oxygen and which differs from Rudlam's mechanism is that radicals remain coordinated to the active sites. If, as has been proposed,2 the radical species (CH,),cOH were to be released into the bulk of the liquid then the addition of 0, to this radical must be considered as probable: 0-0' I (CH,),COH + 0, + (CH,),-C-OH followed by /0°' (CH3),C -+ (CH,),CO+'O,H. 'OH The rate constant for such addition is high, and this process is much more likely that any bulk biradical reactions. If high intensity resulted in significant desorption of radicals then oxygen consumption and ultimate formation of propanone via a homo- geneous radical reaction could result in a rate dependence on the concentration of 0,.References 1 R. B. Cundall, R. Rudham and M. S. Salim, J . Chem. Soc., Faraday Trans. I , 1976,72, 1642. 2 P. R. Harvey, R. Rudham and S. Ward, J. Chem. SOC., Faraday Trans I , 1983, 79, 1381. 3 J. Cunningham and B. K. Hodnett, J. Chem. Soc., Faraday Trans. I , 1981,77, 2777. 4 I. M. Fraser and J. R. MacCallum, J. Chem. Soc., Faraday Trans. I , 1986, 82, 607. 5 Vogel's Textbook of Macro and Semimacro Qualitative Inorganic Analysis, ed. G. Svehla (1979). 6 T. A. Egerton and C. J. King, J. Oil Colour Chem. ASSOC., 1979, 62, 386. 7 P. F. Cornaz, J. H. C. van Hooff, F. J. Pluijm and G. C. A. Schuit, Discuss. Faraday SOC., 1966, 41, 290.2754 Photocatalysis of Propan-2-ol by TiO, 8 R. D. Iyengar, M. Codell, J. S. Karra and J. Turkevich, J. Am. Chem. SOC., 1966,88, 5055. 9 J. Woning and R. A. van Santen, Chem. Phys. Lett., 1983, 101, 541. 10 H. P. Boehm, Discuss. Faraday SOC., 1971,52, 264. 11 M. A. Malati and N. J. Seager, J. Oil. Colour Chem. Assoc., 1981, 64, 231. 12 A. R. Gonzalez-Elipe, G. Munuera and J. Soria, J. Chem. SOC., Faraday Trans. I , 1979, 75, 748. 13 A. R. Gonzalez-Elipe, G. Munuera and J. Soria, React. Kinet. Catal. Lett., 1981, 18, 367. 14 A. R. Gonzalez-Elipe, G. Munuera and J. Soria, Chem. Phys. Lett., 1978, 57, 265. 15 F. A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry (Wiley, New York, 1972). 16 Y. S. Chiang, J. Craddock, D. Mickewich and J. Turkevich, J. Phys. Chem., 1966,70, 3509. 17 A. Tkac, Int. J. Radat. Phys. Chem., 1975, 7 , 457. Paper 511912; Received 30th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202747
出版商:RSC
年代:1986
数据来源: RSC
|
18. |
Conductivity of unsymmetrical and mixed electrolytes. Dilute aqueous cadmium chloride and barium chloride–hydrochloric acid mixtures at 298.15 K |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2755-2762
Korbratna Indaratna,
Preview
|
PDF (638KB)
|
|
摘要:
J. Chem. Soc., Faraday Trans. I, 1986,82, 2755-2762 Conductivity of Unsymmetrical and Mixed Electrolytes Dilute Aqueous Cadmium Chloride and Barium Chloride-Hydrochloric Acid Mixtures at 298.15 K Korbratna Indaratna, A. James McQuillan and R. Archibald Matheson" Chemistry Department, University of Otago, P.O. Box 56, Dunedin, New Zealand This paper presents conductivity measurements at 298.15 K for two aqueous systems containing three ionic species : barium chloride-hydrochloric acid mixtures and dilute cadmium chloride. The results have been analysed in terms of the Lee and Wheaton conductivity equation. Although it is an improvement on earlier approaches to the conductivity of electrolyte mixtures, the equation is found to be less satisfactory for mixtures than for pure electrolytes.The formation constants resulting from the analysis of cadmium chloride solutions are shown to be sensitive to the inclusion or otherwise of certain disputed terms in the Lee and Wheaton equation. Although the modern literature contains several studies of the equilibrium and transport properties of aqueous cadmium ~hloride,l-~ there are no accurate dilute solution conductivities, possibly because the refined conductivity theories of Fuoss and Onsager5 and others are restricted to symmetrical binary electrolytes. Hence these theories cannot be used for cadmium chloride, which is unsymmetrical and contains more than two ionic species. (For concentrations below 0.05 mol dm-3, the predominant species are Cd2+, CdCl+ and C1-.) For many years only theories of limiting-law accuracy, e.g.those of Onsager and Kim6 and of Pikal,7 were available for solutions containing more than two ions. However, Lee and Wheatonsq9 have now published a new conductivity equation which avoids the limiting-law approximations and which is applicable to solutions containing any number of ionic species of any valence type. We have measured the conductivities of aqueous cadmium chloride solutions over the concentration range 1.5 x lo-* to 0.1 mol dm-3 and have analysed the results of our more dilute solutions in terms of the Lee and Wheaton (L-W) conductivity theory. We have also analysed our data assuming that the equivalent conductivity of a cadmium chloride solution is an additive function of the equivalent conductivities of the hypothetical completely dissociated electrolytes (Cd2+, 2C1-) and (CdCl+, C1-) at the ionic strength of the cadmium chloride solution.In the past such an assumption has frequently been made when calculating ion-pair formation constants from the conduc- tivities of unsymmetrical electrolytes. It is, however, contrary to the L-W, Onsager-Kim, and Pikal treatments, which all imply that the equivalent conductivity of an ion in a mixture of two or more electrolytes will, in general, depend on the composition of the mixture as well as on the total ionic strength. This aspect of these three theories is best tested in systems where it is possible to vary the ionic proportions and total ionic strength independently. We have therefore measured the conductivities of a number of BaC1,-HCl mixtures and used the results to test the L-W and Pikal equations and also the simple additivity assumption.Apparatus Experimental Conventional conductivity cells with lightly platinised electrodes were used for the more concentrated solutions. A flask cell with bright platinum electrodes allowed in situ preparation of the most dilute solutions. The temperature was maintained constant 27552756 Conductivity of Aqueous CdCl, and BaC1,-HCl ( f. 0.002 "C) by means of a transformer oil thermostat and measured (k 0.006 "C) with a calibrated calorimeter thermometer. An a.c. bridge with Wagner earthing network was used to compare cell resistances against an 0.01% grade Sullivan and Griffiths non-reactive decade box. Location of the balance point to 0.0 1 % was achieved via an oscilloscope. Resistances (R) were measured at frequencies (f) of 1, 2, 3, 4 and 6 kHz and, if frequency dependent, extrapolated to infinite frequency by plotting R us.f-l. Cells were calibrated with potassium chloride solutions using literature data.loY l1 Various dilute solutions prepared in situ were used for the flask cell calibration and 0.1 and/or 0.01 demal solutions for the others. Reproducibility of cell constants was 0.01 % ( k 0.03 % for the flask cell). An intercomparison procedure showed that the various cell constants were mutually consistent within these limits. A conventional glass-electrode pH meter was used to determine the pH of the cadmium chloride solutions. Materials and Solution Preparation Conductivity water [IC = (4-5) x lo-' W1 cm-l , pH ca.6.01 was prepared by passing distilled water through an Elgastat deionising column. For some of the more concentrated solutions air-saturated conductivity water (IC = 1.1 x loys l2-l cm-l, pH ca. 5.7) was found more convenient. The pH results suggest that most of the conductivity of both types of water was due to an acid impurity, probably dissolved CO,. Stock solutions of potassium chloride were prepared by weight from twice recrystallised B.D.H. AnalaR salt dried at 400 "C. B.D.H. AnalaR CdCl, - 2.5 H,O which gave an 0.038 mol dm-3 solution with pH 4.9, and A(kCdC1,) = 121.19 a-1 cm2 mol-1 was recrystallised twice to give a product with pH 5.6, and A = 121.09 at the same concentration. The hydrolysis study of Gayer and Haas12 reports pH 5.8, for 0.037 mol dm-3 CdCl,.These data suggest that the original B.D.H. material contained sufficient excess acid (ca. 0.03%) to cause an error of ca. 0.1 % in A and that the recrystallisation procedure reduced this error to ca 0.01 % . Ca. 0.5 mol kg-l stock solutions were prepared from twice-recrystallised salt and analysed gravimetrically as AgCl (reproducibility of analysis better than 0.01 % ). Analysis of standard potassium chloride solutions showed the procedure to be reliable to 0.02%. A ca. 0.7 mol kg-l solution of barium chloride was made up by weight from BaC1, prepared by heating B.D.H. AnalaR BaC1;2H20 to constant weight at ca. 110 "C. The equivalent conductivities of ca. 0.05 mol kg-l solutions prepared from the stock by weight dilution were within ca.0.05% of literature data.13 Baker analysed concentrated hydrochloric acid was diluted to prepare a stock solution which was further diluted by weight to ca. 0.03 mol kg-l for conductimetric analysis using the data of Stokes.14 Solutions for conductivity measurements were prepared by weight dilution of the appropriate stocks using Mettler analytical balances. Buoyancy corrections were applied to all weights and densities required for calculating concentrations were measured pycnometrically for the BaC1,-HCl mixtures and obtained from the literature15 for all other solutions. Results and Discussion BaC1,-HCI Mixtures The results, which are based on the international ohm, are given in table 1, where I is the ionic strength, ci the molarity of electrolyte i in the mixture and Amix is the equivalent conductivity of the mixture [ I C / ( C ~ ~ ~ + 2cBaClZ)], where IC is the measured conductivity. The additivity rule requires CHCl A(HC1) + 2CBaC12 A(=!iBaC12) CHCl + 2CBaC12 &nix =K.Indaratna, A . J . McQuillan and R. A. Matheson 2757 Table 1. Equivalent conductivities of BaC1,-HCl mixtures at 298.15 K I = 0.005 mol dm-3 3.9989 0.3334 3.7487 0.4165 2.5006 0.8332 1.2500 1.2498 1 .OW4 1.3332 16.001 1.3329 15.000 1.6667 10.000 3.3331 5.000 4.9995 4.00 1 5.3326 I = 0.02 mol dm-3 I = 0.05 mol dm-3 40.037 3.336 37.522 4.168 25.000 8.332 12.500 12.500 9.996 13.330 373.76 362.48 299.92 223.75 206.66 365.26 353.96 29 1.04 214.92 197.98 356.99 345.76 282.78 207.14 190.11 0.005 0.36 0.70 0.27 0.21 0.20 0.02 0.71 2.5 0.24 0.26 0.13 0.05 0.74 5.6 0.28 0.31 0.23 where A(HC1) and A(iBaC1,) are the equivalent conductivities of the respective pure electrolytes at the ionic strength of the mixture.These quantities were obtained from literature datal3? l4 by appropriate graphical means. At each ionic strength there are deviations from additivity at least ten times the random experimental error, (Amix)calc being invariably greater than (Amix)exptl- These deviations are summarized as AADD where A denotes a mean value of Much greater deviations (Ap) resulted when Pikal’s simplified limiting law7 was used to calculate Amix from literature values of the limiting equivalent conductivities (A:) of the ions involved. The L-W equation for the equivalent conductivity of ion j in a solution containing s ion is s s where the various terms are as defined in the original paper.9 Since t = Icd, the equation involves a distance of closest approach d which is assumed the same for all pairs of oppositely charged ions. Lee and Wheaton in fact omitted the terms of order 1c3 [those involving Cg(t) and q2)(t)] in their analyses of data for unsymmetrical electrolytes92758 Conductivity of Aqueous CdCl, and BaC1,-HCl because Wheaton's studieP on the distribution functions for such electrolytes made subsequent to the derivation of the L-W equation showed these terms to be incomplete.? However, they are not negligible for electrolytes such as aqueous BaCl,, their combined contributions to the conductivity of this electrolyte being ca.0.4% at 5 x mol dm-3. In such cases Pethybridge and Tabal' find that an improved fit results if the q2)(t) terms are retained and the C$(t) terms omitted.(They could not obtain a satisfactory fit if both terms were included.) While it is clear that terms of order I C ~ are needed in the L-W equation, their correct formulation has yet to be achieved. Three different procedures were used to calculate Amix via the L-W equation. (a) Ion association was assumed to be absent, both I C ~ terms were omitted and the necessary parameters (A;, d ) derived by fitting1 the equation to literature data13* l4 for the separate pure electrolytes (omitting the ic3 terms in the case of BaC1,). The fit for HC1 was excellent [A = 0.007 for I < 0.02mol dmP3 with Ao (HCl) = 426.55 R-l cm2 mol-l and d = 0.5 nm], but that for BaCl, was rather poorer [A = 0.07 for I < 0.02 mol dm-3 with A"(iBaC1,) = 139.90 R-l cm2 mol-1 and d = 0.45 nm].A&+ and were obtained from A"(HC1) and A"(+BaCl,) assuming A&- = 76.35 l2-l cm2 moi-l and a mean value of d[(0.5+0.45)/2 nm] was used for all mixtures. Deviations between calculated and experimental AMix (summarized as A%,) are smaller than for the additivity rule but the fit for the mixtures is significantly worse than for either pure electrolyte. ( t ) terms were retained for both pure BaCl, and the mixtures and formation of BaCP and HCl ion pairs allowed. The formation constants K(BaCl+) and K(HC1) were obtained in the course of fitting the pure electrolyte conductivities to the L-W equation. The activity coefficients needed to calculate ionic concentrations from the formation constants were calculated from the Debye-Huckel equation - log yi = A@ I+/( 1 + BdB) with the d value used in the L-W equation.The fit of the pure electrolytes was improved [HCl: A = 0.005 for I < 0.08 mol dm-3 with AO(HC1) = 426.5 l2-l cm2 mol-l, d = 0.68 nm and K(HC1) = 0.23 dm3 mol-l; BaC1,: A = 0.02 for I < 0.06 mol dm-3 with AO(iBaC1,) = 140.1 R-l cm2 mol-l, AgaC1+ = 70 R-l cm2 mol-l,$ d = 0.45 nm and K(BaCl+) = 7.04 dm3 mol-l], but that for the mixtures was little changed (see AL-, values). (c) The procedure was as for (b), but allowing d to vary with mixture composition according to the formula (b) The foregoing procedure was modified as follows: The A slight improvement in fit resulted. (See ALw values.) No better results were gbtained when other powers of z (0, 1 or 3) were used in the above formula.The conductivities of the KCl-LiCl mixtures studied by Krieger and Kilpatrickl* were also calculated via the L-W equation using the procedure employed in (b), except that both ic3 terms were retained (because the system contained symmetrical electrolytes only). The fit, for I < 0.05 moldm-3 (A z 0.06) was very much better than for BaC1,-HCl. However, the deviations from additivity were also much smaller (A z 0.06). In summary, while the L-W equation is less successful for mixtures than for pure electrolytes, it is distinctly better than other approaches to the conductivity of electrolyte mixtures. t This difficulty does not arise in the case of symmetrical electrolytes for which the ic3 terms are invariably $ The appropriate 1" and d values were successively adjusted so as to minimise A.5 In fitting the BaC1, conductivities, I"(BaCl+) was held at this value while A"($BaCl,), d and K were optimised. Changing 1"(BaCl+) to 50 R-l cm2 mol-' altered the optimum K (BaCl+) but made little difference to the fit for either pure BaC1, or for the mixtures. retained.K . Indaratna, A . J . McQuillan and R. A . Matheson 2759 Table 2. Equivalent conductivities of aqueous cadmium chloride at 298.15 K A($CdCl,)/ A(BCdCl,)/ c/ 1 0-3 mol dm-3 c/ 1 0-3 mol dmP3 0-1 cm2 mol-l 0-l cm2 mol-I 0.151 00 0.214 55 0.245 81 0.294 80 0.443 45 0.545 70 0.654 50 0.775 43 0.791 42 0.795 45 0.804 80 0.846 75 0.971 88 125.74 124.39 123.71 122.73 120.54 119.20 117.89 1 16.53 116.31 116.39 116.28 115.77 114.53 1.334 3 1.730 2 4.450 2 6.610 5 8.406 0 10.156 11.844 22.691 36.698 59.380 72.285 101.46 11 1.42 108.56 95.64 89.39 85.43 82.33 79.73 68.91 61.14 53.63 50.63 45.57 CdCI, Solutions Our results, again given in terms of the international ohm, are listed in table 2.No hydrolysis corrections have been applied to these data and in the analyses which follow we assume that Cd2+, CdCl+ and C1- are the only conducting species present. At high dilution the most likely cause of failure in this assumption is the presence of H+ as a result of one or other of the hydrolysis reactions postulated in the literature:l2? 1 9 3 2o Cd2+ + H,O + CdOH+ + H+ CdCl+ + H,O s CdOHCl+ H+. The literature values of the hydrolysis constants are somewhat discordant and in any case refer mainly to concentrated perchlorate media.? It is therefore not possible to calculate accurately the extent of hydrolysis in our dilute cadmium chloride solutions.However, one may estimate the contribution of hydrolysis products to the conductivity of any CdC1, solution by comparing its pH with that of the parent conductivity water. In this way we estimate any such contribution to be less than 0.1% of A for all except the two most dilute solutions which were, therefore, excluded from the analyses which follow. Literature formation constants1 indicate that > 99.8% of total cadmium is present as Cd2+ and CdCl+ in cadmium chloride solutions more dilute than 2 x mol dm-3. Accordingly the 13 data points from 2.5 x mol dmP3 were analysed assuming Cd2+ + C1- e CdCl+ to be the only significant equilibrium.The L-W equation [including the c2)(t) terms] was used to calculate the various ionic conductivities. Activity coefficients were calculated as for the BaCl, solutions. With this approach to 1.7 x A = f[Aicd2+, AEdC1+, A&, d, K(CdCl+)]. 7 For the first reaction reports pK = 9.0 in 3 mol dmP3 NaClO,, while Dyrssen and Lumme20 find pK = 9.9 in the same medium. Biedermann and Ciavatta21 obtain pK = 10.0 in 3 mol dm-3 LiC10, and Chaberek et aLZ2 find pK = 1 1.6 in 0.1 mol dm-3 KCI. However, Gayer and Haas', were unable to obtain a constant K for this reaction from their studies of the pH of CdC1, solutions and concluded that the second reaction was the predominant hydrolysis reaction in their solutions.0.2 - n s b o * l - v , I I 0.2 n 0 .1 n s bo**I: 0 . 1 t 90 100 102 K(CdC1') *.* y n 0.3 0.5 0.7 d Fig. 1. Fit of cadmium chloride conductivities. A&- was set at the literature value23 of 76.35 0-l cm2 mol-1 and the remaining parameters adjusted in turn to minimize a(A), the standard deviation between calculated and experimental equivalent conductivities. The fit is quite sensitive to the values of l&de+ and K(CdCl+), but not very sensitive to the value &Cl+ and quite insensitive to the value of d (See fig. 1). The best fit [o(A) = 0.04,] was obtained with AiCd2+ = 53.95, A&cl+ = 33 0-l cm2 mol-l, d = 0.45 nm and the formation constant K(CdCl+) = 100.2 dm3 mol-l. (The fit is only slightly worse than for the alkaline-earth halides,17 which are less subject to hydrolysis than cadmium chloride.) The optimum values of $Cd2+ and K(CdCl+) differ somewhat from published data (liCd2+ = 53.5 derived by applying a Shedlovsky extrapolation to the conductivities of dilute cadmium perchlorate and K(CdCl+) = 85 from e.m.f.studies of dilute cadmium chloride solutions1), but the value of &C1+ agrees with the figure of 32.5 calculated from the mobility of this ion which Reilly and Stokes2 derived from the diffusion coefficients of cadmium chloride solutions. The difference betweeen the two values of AiCdz+ may be partly due to the absence of any hydrolysis correction in the present work and an over- correction for hydrolysis in the earlier work on cadmium perchlorate.More recent work20* 21 suggests that the hydrolysis constant used to make this correction19 may be too large by a factor of ca. 10. Both the goodness of fit and the values of the optimum parameters are somewhat changed if the q2) ( t ) terms are omitted [o(A) = 0.064, with &d2+ = 53.7, A",,,,+ = 26 Sz-l cm2 mot1, d = 0.4 nm and K(CdCl+) = 87 dm3 mol-l]. dvidently the I C ~ terms in the L-W equation play a significant role in the analysis. This implies that the uncertainty as to the correct form of these terms may lead to significant uncertainty in K(CdCl+).K. Indaratna, A . J . McQuillan and R. A . Matheson 2761 It is possible to fit the conductivities for a wider range of concentrations (2.5 x lo-* to 1.2 x lo-, mol dm-3) if one assumes the additional equilibrium Cd2+ + 2C1- + CdCl, with an equilibrium constant K(CdC1,) [q2)(t) terms included: a(A) = 0.07, with AiCde+ = 53.95, A&C1+ = 33 R-1 cm2 mo1-1, d = 0.45 nm, K(CdCl+) = 98.5 dm3 mo1-1, K(CdC1,) = 550 dms mok2. Without c2’(t) terms: a(A) = 0.07, with 1 i c d 2 + = 53.7, &C1+ = 26 0-l cm2 mol-l, d = 0.4 nm, K(CdCl+) = 86.2 dm3 mol-l, K(CdC1,) = 123 dms m01-~].The large differences between the two values of K(CdC1,) shows that this quantity is very sensitive to alterations in the conductivity equation. In the past, the additivity rule has often been used to calculate formation constants from the conductivities of associated unsymmetrical electrolytes. Here this takes the form A = CXA,:~+? Al:l where a is the degree of dissociation of CdCl+ and A2: and A1: are, respectively, the equivalent conductivities of the hypothetical completely dissociated electrolytes (Cd2+, 2Ck) and (CdCl+, C1-).These quantities were calculated from their limiting values by the equations 41 1 + B d d f A = Ao-(BlAa+B2) Over the range 2.5 x 10-4-1.7 x mol dm-3 this procedure gave almost as a good a fit as the much more complicated L-W analysis. [o(A) = 0.05, with &d2+ = 53.75, A&C1+ = 27.2 cm2 mol-l, d = 0.4 nm, K(CdCl+) = 90.9 dm3 mol-l.3 Again the range of fit could be extended to 1.2 x lo-, mol dm-3 if the additional complex CdCl, was assumed. [o(A) = 0.10, A&+ = 53.75, &(-I+ = 26 R-l cm2 mol-l, d = 0.4 nm, K(CdCl+) = 88.5 dm3 mol-l, K(CdC1,) = 750 dms m ~ l - ~ . ] With both the L-W and the additivity treatments one obtains a better fit for the cadmium chloride solutions than for the barium chloride-hydrochloric acid mixtures, possibly because the proportions of the various ions vary much more in these mixtures than in the cadmium chloride solutions. We thank Drs R.J. Wheaton and A. D. Pethybridge for helpful correspondence. K.I. acknowledges the award of a Colombo Plan postgraduate scholarship. Appendix Corrections to Conductivity Equations The following corrections were made to the Pikal and L-W equations before they were used in this work. Pikal Missing factors (zt [eqn (16)] and lz, zjI [eqn (17)]) were inserted into the denominators of the electromagnetic terms in eqn (16) and (17) of ref. (7). It should be noted that the omission of these factors does not affect the analyses carried out in Pikal’s original paper since he considered 1 : 1 electrolytes only.2762 Conductivity of Aqueous CdCl, and BaC1,-HCI L-W We are indebted to Dr R.J. Wheaton for pointing out that certain equations in his papers should be corrected to read as follows: ref. (8): eqn (9, /? = e2/DkT eqn (76), denominator of last term (1 +&p lcR+qp~,R2/2) eqn (99), last rerm on the third line of H,, p , ,(lcR), (1 + K R ) ~ ( 1 + q; KR) Tr(lcR)/q, + appendix to 8(a): pi in the equation for AXJo* @/X should read ui. ref. (9): eqn (52), replace ui by pi eqn (64), numerator of 1st term ef I C ~ X eqn (96), denominator of last term in first line, (1 +q; rcR+qplc2R2/2) eqn (104), extra factor (mi + mv) required in denominator of last term eqn (123), z = Fce/6nq References 1 P. J. Reilly and R. H. Stokes, Aust. J. Chem., 1970, 23, 1397. 2 P. J. Reilly and R. H. Stokes, Aust. J. Chem., 1971, 24, 1361. 3 A. J. McQuillan, J. Chem. Soc., Faraday Trans. I , 1974, 70, 1558. 4 E. Pitts, Proc. R. SOC. London, Ser. A, 1953, 217, 43. 5 R. M. Fuoss and L. Onsager, J. Phys. Chem., 1957,61, 668. 6 L. Onsager and S. K. Kim, J. Phys. Chem., 1957, 61, 215. 7 M. J. Pikal, J. Phys. Chem., 1971, 75, 3124. 8 W. H. Lee and R. J. Wheaton, J. Chem. Soc., Faraday Trans. 2, 1978,74, 743. 9 W. H. Lee and R. J. Wheaton, J. Chem. Soc., Faraday Trans. 2, 1978, 74, 1456. 10 R. A. Robinson and R. H. Stokes, Electrolyte Solutions (Butterworths, London, 2nd edn, 1959), 11 R. M. Fuoss, J. Am. Chem. Soc., 1959,81, 1559. 12 K. H. Gayer and R. M. Haas, J. Phys. Chem., 1960,64, 1764. 13 T. Shedlovsky and A. S. Brown, J. Am. Chem. Soc., 1934,56, 1066. 14 R. H. Stokes, J. Phys. Chem., 1961,65, 1242. 15 International Critical Tables (McGraw-Hill, New York, 1928), vol. 111. 16 R. J. Wheaton, personal communication. 17 A. D. Pethybridge and S. S. Taba, J. Chem. SOC., Faraday Trans. I , 1982, 78, 1340. 18 K. A. Krieger and M. Kilpatrick, J. Am. Chem. SOC., 1937,59, 1878. 19 Y. Marcus, Acta Chem. Scand., 1957, 11, 690. 20 D. Dyrssen and P. Lumme, Acta Chem. Scand., 1962, 16, 1785. 21 G. Biedermann and L. Ciavatta, Acta Chem. Scand., 1962, 16,2221. 22 S. Chaberek, R. C. Courteney and A. E. Martell, J. Am. Chem. SOC., 1952,74, 5057. 23 R. A. Robinson and R. H. Stokes, Electrolyte Solutions (Butterworths, London, 2nd edn, 1959), p. 463. 24 R. A. Matheson, J. Phys. Chem., 1962,66,439. p. 462. Paper 5/1914; Received 30th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202755
出版商:RSC
年代:1986
数据来源: RSC
|
19. |
Transport numbers of dilute aqueous cadmium chloride solutions at 298.15 K |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2763-2771
Korbratna Indaratna,
Preview
|
PDF (636KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans. I, 1986,82, 2763-2771 Transport Numbers of Dilute Aqueous Cadmium Chloride Solutions at 298.15 K Korbratna Indaratna, A. James McQuillan" and R. Archibald Matheson Chemistry Department, University of Otago, P.O. Box 56, Dunedin, New Zealand This paper describes an indirect moving boundary apparatus for determining transport numbers and presents transport numbers of dilute aqueous cadmium chloride solutions at 298.15 K. Below 0.02 mol kg-l, the variation of transport number with concentration can be reproduced using the Lee and Wheaton equation for the relevant ionic conductivities if the solutions are assumed to contain CdCl+ and CdC1, as well as Cd2+ and C1-. However, it is not possible to account for the concentration dependence of both transport number and conductivity by means of this equation unless different values of are used in the two analyses.The associated nature of aqueous cadmium chloride, cadmium bromide and cadmium iodide solutions has long been recognised. These solutions contain the species CdXiWn (n = 0-4) in varying proportions at different CdX, concentrations and this autocomplexation contributes to these salts' high solubilities in water. The concentrated aqueous solutions are dominated by the higher n complexes and at sufficiently high concentrations the measured transport numbers become negative. In spite of this unusual behaviour the transport numbers of these systems have not been well established over the experimentally accessible concentration range. This is more surprising for the dilute region, where these halides have often been used as following electrolytes in direct moving-boundary transport number measurements on other halides. In 1929 Lucassel reported cadmium chloride transport numbers obtained from e.m.f.measurements at 25 "C on concentration cells with transference. These measurements on 0.01 to 6.0 mol kg-l solutions showed that the transport number of the cadmium ion constituent2 becomes negative at high concentrations. More recently, McQuillan3 has reported transport numbers of aqueous cadmium chloride solutions at 25 "C from 0.1 to 6.0 mol kg-' by both Hittorf and concentration-cell measurements, with good agreement between the two methods. At the highest concentrations there is satisfactory agreement between the McQuillan and Lucasse data, but below ca.4 mol kg-l, where the transport number changes from negative to positive values, the two sets of data differ significantly. Furthermore, there is an unusual minimum in the Lucasse data at ca. 0.03 mol kg-l. The purpose of this work was to establish clearly the transport-number behaviour of aqueous cadmium chloride solutions below 0.1 mol kg-l. The direct moving-boundary method is the most precise method for this concentration range. However, the very reasons which have made cadmium chloride a popular choice as an indicator (following) electrolyte in direct moving boundary measurements, i.e. its low cation constituent electrical mobility and high density, virtually prohibit its use as a leading electrolyte. It is therefore surprising that cadmium halide solutions have not been studied by the indirect moving-boundary method,2 which gives the transport number of the indicator electrolyte.27632764 solution adjacent to the boundary adjusts to a value mB such that4 Transport Numbers of CdCl, Solutions In a moving-boundary experiment the molality of the indicator electrolyte in the where TA is the transport number of the non-common ion constituent A in the leading electrolyte AY, TB is the transport number of the non-common ion constituent B in the indicator electrolyte BY, mA is the molality of AY adjacent to the boundary, while 2, and 2, are the respective ionic charges. Thus, providing TA is known, TB can be calculated from the experimental value of the adjusted molality mB.While the precision of this indirect moving-boundary method may approach that of the direct method, it depends critically on the accuracy of mB, which is normally obtained by analysis of a small sample of solution. Considerable care must be given to both the analytical method and the sampling procedure. It is also essential that the apparatus is such as to give a stable boundary unaffected by any extraneous mixing processes. This paper reports the determination of transport numbers in dilute (0.004-0.1 mol kg-l) aqueous cadmium chloride solutions by the indirect moving- boundary procedure. The measurements were obtained using a cell design which minimizes solution irregularities and allows external conductometric analysis. Experimental Materials and Solutions The preparation and analysis of CdCl,, HC1 and KC1 solutions have been described el~ewhere.~ Potassium nitrate (Hopkin and Williams analytical grade) and silver nitrate (99.8 % pure Baker analysed) were dried at 100 "C and their solutions prepared directly by weight.Solutions were prepared with distilled water that had been passed through a deionising column (Elgastat type B102) to give a conductivity of ca. 1 x $2-l cm-l. The silver nitrate solutions were protected from light. Moving-boundary Apparatus In the indirect moving-boundary method it is important to obtain an authentic sample of the adjusted indicator solution and to avoid mixing of the solutions near the boundary with those around either electrode (where gross concentration changes may occur). Failure to meet these requirements may become evident in the following ways.(a) The adjusted molality mB may depend on the precise position from which the sample was taken or be generally irreproducible. (b) The value obtained for TB may depend on the boundary velocity and therefore on current. (c) The value obtained for TB may change when the leading electrolyte is changed. Considerable experimentation was needed to produce an apparatus free from these signs of anomalous behaviour. Satisfactory results could not be obtained when tubing of > 3 mm i.d. was used or when the boundary was formed by the autogenic method. It was found advantageous to separate the cathode compartment from the main apparatus by a glass frit and to have a substantial length of tubing between the anode compartment and the stopcock at which the boundary was initially formed.Tests on the apparatus finally adopted showed that: (a) when two samples of indicator solution were taken at different distances from the boundary (see Procedure) the two values of mB usually agreed within rfr 0.1 % . (b) The experimental value of Tg8C12 for a given solution was independent of current ( 0.1-0.2% ) for two- or three-fold increases in current and was unaltered ( If: 0.1 % ) when the leading electrolyte was changed from KCl to HCl. As a final check, the apparatus was used to determine the cation transport numberK. Indaratna, A . J. McQuillan and R. A . Matheson 2765 9 IJ I AglAgCl electrode IF Fig. 1. Indirect moving-boundary apparatus. of 0.1 mol kg-l aqueous AgNO, solution using KNO, as a leading electrolyte.With an initial AgNO, concentration 10% greater than the adjusted value and a 2 mA current, the two analysed sample concentrations agreed within 0.1 % and an average cation transport number of 0.4683 was obtained. This is in excellent agreement with the accepted value2 of 0.468. The Pyrex apparatus finally adopted for the indirect moving-boundary measurements is shown in fig. 1. A sheared-type boundary was initially formed at stopcock B and rose beyond the ball and cup joint A. Generally KC1 and occasionally HCl were used as leading electrolytes. The central part of the apparatus between joint J and the anode compartment was made of 3 mm i.d. tubing and, to minimize any boundary disturbances at A and B, both the bore of the ball and cup joint and the stopcock were also 3 mm. The ends of the hollow stopcock key were removed to facilitate thermal regulation. The 28 cm length of tubing between A and B had a volume of 2 cm3 and provided two samples for analysis.h The cathode unit was connected to the apparatus 10 cm above A by a 7/16 joint J.The cathode unit consisted of an H-shaped vessel, the two halves of which were separated by a glass frit. The silver-silver chloride cathode consisted of a 3 mm 0.d. helix of silver wire joined to a platinum wire which was sealed into the end of a glass tube attached to a 10/19 joint. An appropriate amount of silver chloride was formed by anodising at 1 mA in 1 mol kg-l hydrochloric acid. The cadmium anode consisted of a short length of 7 mm cadmium rod (> 99.98% pure) sealed into a 10/19 cone with epoxy resin glue.The cadmium electrode was cleaned prior to use by successive immersion in 2 mol kg-l nitric acid, 2 mol kg-l hydrochloric acid, followed by thorough rinsing in deionised water and air drying. The 23 cm length of tubing from the anode to stopcock B was formed in a loop to reduce the overall apparatus length. Procedure Using small amounts of Apeizon N grease the lower part of the apparatus containing the anode, stopcock B and the ball joint were assembled. With the aid of a length of 0.7 mm 0.d. polypropylene tubing and a syringe this section of the apparatus was filled with indicator solution of concentration ca. 10% above that expected in the adjusted solution. (It was established that the adjusted concentration was independent of the initial concentration of the indicator solution.) The stopcock was then closed and the excess solution removed under suction.The cell tubing above stopcock B was then rinsed with deionised water and the cathode section attached at joint J. The cell above stopcock B was then rinsed with oxygen-free leading solution and completely filled. The use of oxygen-free solution avoids complications arising from oxygen reduction at the cathode. The complete cell, held vertically by a long glass frame and rubber bands, was lowered into a glass-sided water bath maintained at 25.00 0.02 "C and clamped to a vibration- free support. The cell was left for 30 min to reach bath temperature, stopcock B was opened and a regulated current d.c.supply was switched in. At higher concentrations and currents the boundary was clearly visible and electrolysis was terminated when the boundary had passed the ball joint J. When the boundary was not visible the solution above J was tested with 0.5 mol kg-l NaOH and a white precipitate confirmed the presence of cadmium ions. At completion of electrolysis, the cell was carefully removed from the bath, stopcock B closed and the tube above joint A tilted through 30" to isolate the solutions above and below this joint. The solution above the joint was then removed under suction and the joint separated. Ca. 0.5 g samples of solution were then removed by syringe from the top and bottom halves of the adjusted indicator solution column and delivered into weighed pipettes for ca.100-fold dilution before conductometric analysis. Analysis The conductometric analysis made use of the cadmium chloride solution conductivity data previously obtained in this lab~ratory.~ Standard conductance measurement techniques were followed5 strictly during the conductometric analysis. Deionised water, stored overnight in a PVC container, was used for weight dilutions. Solvent corrections were ca. 1.0 x The following procedure was used to deduce the concentration of a diluted sample from its conductivity. After obtaining a first estimate from a plot of conductivity us. concentration, the exact concentration was obtained by successive approximations using a large scale plot of the slowly varying function A+ 5902/c us. c. The molality of the original sample was then calculated using the weight-dilution data and densities interpolated from literature values.6 Tests on the reliability of the method indicated that the results were accurate and reproducible to within f 0.1 % .R-l cm-l.Table 1. Indirect moving-boundary transport numbers of CdCl, boundary system initial molalities of leading electrolyte and following electrolyte /mol kg-l mW) a current CdClz /mA /mol kg-l 0.200 00 KCl/O. 100 CdCl, 0.133 79 KC1/0.064 CdCl, 0.100 26 KC1/0.047 CdCl, 0.170 59 HC1/0.047 CdCl, 0.072 39 KC1/0.034 CdC1, 0.054 45 KC1/0.026 CdCl, 0.038 29 KC1/0.018 CdC1, 0.019 98 KC1/0.010 CdCl, 0.033 74 HC1/0.010 CdCl, 0.009 022 KC1/0.0042 CdC1, 1 .o 2.0 1 .o 2.5 0.8 1 .o 2.5 1.8 0.6 1.6 0.5 1.2 0.4 1 .o 0.2 0.6 0.6 0.1 0.2 0.087 35 0.087 71 0.058 70 0.059 00 0.043 98 0.044 21 0.044 24 0.044 11 0.031 65 0.031 78 0.023 80 0.023 71 0.016 63 0.016 58 0.008 561 0.008 540 0.008 582 0.003 837 0.003 829 0.087 67 0.087 70 0.058 90 0.058 93 0.043 97 0.043 96 0.043 96 0.043 95 0.031 68 0.031 88 0.023 82 0.023 72 0.016 60 0.016 57 0.008 579 0.008 575 0.008 554 0.003 867 0.003 855 0.4305 0.4292 0.4296 0.43 18 0.4296 0.43 19 0.4322 0.4309 0.4283 0.4301 0.4283 0.4267 0.4256 0.4243 0.4200 0.4 190 0.42 12 0.4 172 0.4 166 0.4292 0.4295 0.4295 0.4295 0.4293 0.4288) 0.43 14 0.4269 0.4248 0.4242) 0.4208 0.4206) 0.4198 0.4205 0.4 190} 0.430, _+ 0.0010 0.430, & 0.0007 0.430, k 0.0010 0.429, k 0.001 1 0.427, f 0.0008 0.424, f 0.0005 0.420, & 0.0006 0.41 8, f 0.0014 fl ? a U and L label the data derived from samples which are drawn, respectively, from the upper and lower halves of the adjusted indicator solution.N2768 Transport Numbers of CdCl, Solutions 0.48 0.46 T:;C'? 0.44 0.42 1 .o fi 0.5 Cd C l i 0 I 6 - - - ---- c * 0 0.1 0.2 0.3 0.4 (m/mol kg-'); Fig. 2. (a) Relative abundance of species in aqueous cadmium chloride so1utions.h = rn,/m, where mi is the molality of species i and rn is the stoichiometric molality. (b) Cadmium ion constituent transport numbers. 0 , Lucasse data; 0 , this work; 0, McQuillan. Results The cadmium ion constituent transport numbers obtained for 0.004-0.1 mol kg-l aqueous CdCl, solutions are listed in table 1. The experiments when the leading electrolyte was HCl rather than KC1 have been included and demonstrate that the results are independent of the leading electrolyte.The mean transport numbers at each concentration were obtained by averaging over independent runs and the uncertainties in each case are average deviations from the means. The transport numbers were calculated using eqn (1) and existing transport-number data for aqueous KCl and HCl solution^.^ Solvent corrections8 were found to be less than 0.1% for these indirect moving-boundary measurements and were consequently ignored. Discussion The transport numbers reported here are compared with the earlier data of Lucassel and McQuillan3 in fig. 2(b). A smooth curve can be drawn through the present results and those of McQuillan, but there is a marked difference between both these sets of data and those of Lucasse.The smooth curve of transport number versus mi extending down to 0.004 mol kg-l suggests a simple extrapolation to obtain the limiting transport number. However, the experimental slope of this curve is positive at the lowest concentrations, whereas the limiting law requires a negative slope for any 2: 1 electrolyte with cation transport number less than 2/3. Hence there must be a minimum in the curve below the lowest experimental concentration, so that a simple extrapolation cannot be expected to give a reliable limiting transport number. This difficulty is a consequence of the association which takesK . Indaratna, A . J . McQuillan and R. A . Matheson 2769 Table 2. Cadmium ion constituent transport numbers in aqueous CdC1, at 298.15 K transport number calculateda molality /mol kg-l exptl A B C D 0.003 847 0.4183 0.4360 0.4182 0.4 130 0.4 178 0.008 565 0.4202 0.45 10 0.4221 0.4139 0.4220 0.016 595 0.4247 0.4644 0.4244 0.4128 0.4244 a A, Calculated via the L-W equation including the q2)(t), but not the CE(t) terms, using the optimum parameters from the corresponding conductivity analysis5 (A&p+ = 53.95, A&,,+ = 33 Q-' cm2 mol-l, K(CdC1,) = 550 dm6 mol-2); B, as for A, but with A&cl+ = 25.5 f2-l cm2 mol-l; C, calculated via the L-W equation excluding the q2)(t) and C,P(t) terms, using the optimum parameters from the corresponding conductivity analysis5 = 53.7, AEdCl+ = 26 a-1 cm2 mol-l, d = 0.4 nm, K(CdCl+) = 86.2 dm3 mol-l, K(CdC1,) =?23 dms mol-2); D, as for C, but with &C1+ = 28.2 R-l cm2 mo1-l.d = 0.45 nm, K(CdCl+) = 98.5 dm3 mol-l, place in cadmium chloride solutions even at high dilution [see fig. 2(a), which is based on the stability constants of Reilly and Stokesg]. Assuming Cd2+, C1- and CdCl+ to be the only ions present, the cadmium ion constituent transport number is given by where A, denotes an equivalent conductivity and c, a molar concentration. The transport number thus depends on the relative concentrations and on the conductivities of the various ions. In an associated electrolyte like cadmium chloride the concentration dependence of the transport number is largely determined by variations in the first factor, but the effect of changes in the ionic conductivities is not negligible. A simple modification of the procedure used to account for the concentration dependence of conductivity in dilute (< 0.012 mol dm-3) cadmium chloride solutions5 allows both factors to be considered and the transport number to be calculated via eqn (2).In this procedure we assumed the solutions to contain CdCl, and CdC1+ in equilibrium with free Cd2+ and C1- and calculated the various Ai by means of the Lee and Wheaton (L-W) equation : A; = A; 1 + zi i x; i t, x; [ ~ ; ( t ) (~lc) + B;(t) v K ) 2 + cg(t) (BK)~]} p-2 21-1 I -m { 1 + Vjl)(t) @c) + Vj2)(t) W I C ) ~ + llj5) t/6 i I + t where the various terms are as defined in the original paper.1° Because of uncertainty as to the correct form of the terms of order I C ~ in this e q ~ a t i o n , ~ ~ l1 two sets of calculations were made: one using the equation without the C$(t) and q 2 ) ( t ) terms and one with the q 2 ) ( t ) terms alone included.In both cases the parameters of the analysis were &d2+, A&l+, the distance of closest approach (d) and the two formation constants. (A&- was fixed at 76.35 C2-l cm2 mol-l.) The results for solutions of comparable concentration to those considered in the conductivity analysis are shown in table 2. The agreement with experiment is poor if the optimum parameters from the conductivity treatment5 are used in the calculations, especially when the V,j2)(t) terms are included in the L-W equation. The transport number is more sensitive toh 0 . 4 3 0.42 T:P 0.414- 0 . 4 1 \ \ \ \ \ 7 I I I 0.05 0.1 0 0.15 (rnlmol kg-')$ Fig. 3. Cadmium ion constituent transport numbers. 0, Experimental points.(-) Calculated from L-W equation. (a) With best fit parameters from the conductivity analysis as in table 2, column A, (b) with adjusted value of as in table 2, column B. (---) Onsager limiting law, (-) Concentration range of the L-W conductivity analysis. changes in than is the conductivity and substantial changes in this parameter were needed to obtain a fit (see table 2 and fig. 3). We find a similar situation arises when literature conductivities12 and transport numbers13 for aqueous calcium chloride are analysed as above, but assuming only one complex (CaCl+) to be formed. It is difficult to avoid the conclusion that there are faults in the L-W equation which affect the electrolyte conductivity and the transport number to different extents. We are indebted to a referee for drawing our attention to a recent paper by Justice et aZ.,14 confirming that the relaxation terms cancel in the calculation of transport numbers of symmetrical binary electrolytes.As far as we are aware, there has been no recent consideration of the situation for other types of electrolytes. While it is not clear that the relaxation terms in the L-W equation cancel completely in the calculation of transport numbers for unsymmetrical electrolytes, we find near cancellation for CdCI, and CaCl,. Failures of conductivity equations to account for the concentration dependence of transport numbers are not new and Spiro15 has raised the possibility of errors which cancel out in the sum of the ionic conductivities but not in their ratio, thus leading to errors in the transport number but not in the conductivity. We are grateful to Dr M. Spiro for helpful correspondence. K. I. acknowledges the award of a Colombo Plan Postgraduate Scholarship. References 1 W. W. Lucasse, J. Am. Chem. Soc., 1929,51, 2605. 2 M. Spiro, in Physical Methods of Chemistry, ed. A. Weissberger and B. W. Rossiter (Wiley, New York, 3 A. J. McQuillan, J. Chem. Soc., Faraday Trans. I , 1974, 70, 1558. 4 J. R. Gwyther and M. Spiro, J. Chem. SOC., Faraday Trans. I , 1976,72, 1410. 1971), part IIA.K . Indaratna, A . J. McQuillan and R. A . Matheson 277 1 5 K. Indaratna, A. J. McQuillan and R. A. Matheson, J. Chem. SOC., Faraday Trans. I , 1986,82, 2755. 6 International Critical Tables (McGraw Hill, New York, 1928), vol. 111. 7 D. G. Miller, J. Phys. Chem., 1966, 70, 2639. 8 J. R. Gwyther, M. Spiro, R. L. Kay and G. Marx, J. Chem. SOC., Faraday Trans. I , 1976,72, 1419. 9 P. J. Reilly and R. H. Stokes, Aust. J. Chem., 1970, 23, 1397. 10 W. H. Lee and R. J. Wheaton, J. Chem. SOC., Faraday Trans. 2, 1978, 74, 1456. 1 1 A. D. Pethybridge and S. S. Taba, J. Chem. SOC., Faraday Trans. I , 1982,78, 1340. 12 T. Shedlovsky and A. S. Brown, J. Am. Chem. SOC., 1932, 56, 1066. 13 L. G. Longsworth, J. Am. Chem. SOC., 1935, 57, 1185. 14 J. C. Justice, J. PCriC and M. P&e, J. Solution Chem., 1980, 8, 583. 15 M. Spiro, in Physical Chemistry of Organic Solvent Systems, ed. A. K. Covington and T. Dickinson (Plenum Press, New York, 1973), chap. 5, part 2. Paper 5/1915; Received 30th October, 1985
ISSN:0300-9599
DOI:10.1039/F19868202763
出版商:RSC
年代:1986
数据来源: RSC
|
20. |
Infrared study of water and pyridine adsorption on the surface of anhydrous vanadyl pyrophosphate |
|
Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases,
Volume 82,
Issue 9,
1986,
Page 2773-2779
Simon J. Puttock,
Preview
|
PDF (495KB)
|
|
摘要:
J. Chem. SOC., Faraday Trans. 1, 1986,82, 2773-2779 Infrared Study of Water and Pyridine Adsorption on the Surface of Anhydrous Vanadyl Pyrophosphate Simon J. Puttock and Colin H. Rochester* Chemistry Department, The University, Dundee DDI 4HN Infrared spectra of vanadyl pyrophosphate contained a broad maximum at 2550-3550 cm-l which exchange experiments with deuterium oxide showed could be primarily attributed to residual OH-containing species in the bulk crystal lattice. However, the presence of Brsnsted-acidic surface hydroxy groups was exhibited by the generation of adsorbed pyridinium ions when (VO),P,O, was exposed to pyridine vapour. Pyridine was also adsorbed on Lewis-acidic surface sites, some of which were converted to Brsnsted-acidic sites by the adsorption of water.Vanadyl pyrophosphate (VO),P20, has been shown to be an important constituent of vanadium-phosphorus mixed-oxide catalysts which are active for the oxidation of butanel or but- 1 -ene2p to maleic anhydride. Temperature-programmed desorption studies of the interactions between but- 1 -ene and vanadium-phosphorus mixed-oxide catalysts led to the conclusion that (VO),P,O, was responsible for the oxidation of butene to butadiene and crotonaldehyde, which were subsequently converted to maleic anhydride via catalysis involving p-VOP0,.3 The t.p.d. data suggested that five types of surface centre existed on active catalyst, each type of centre interacting with but-1-ene to form a particular surface complex which gave characteristic desorption products at elevated temperatures. A peak in the t.p.d.curve at 453 K corresponded to the desorption of but-1 -ene and but-2-enes. Active sites responsible for the isomerization reaction may be Brarnsted-acidic surface hydroxy groups, which interact with but- 1 -ene to form an adsorbed intermediate carbocation. An alternative would be that incompletely coordi- nated exposed vanadium cations (Lewis-acidic surface sites) provided centres for the adsorption of but-1-ene as a n-ally1 intermediate. Ai and Suzuki4 have shown that the selectivity of vanadium-phosphorus mixed-oxide catalysts is influenced by surface acidity. The present infrared study of the adsorption of water and pyridine on vanadyl pyrophosphate was undertaken to probe the existence of surface hydroxy groups and Bramsted- and Lewis-acidic sites5 on the surface of pure (VO),P20,.Experimental Vanadyl pyrophosphate was prepared by the reaction of vanadium pentoxide and orthophosphoric acid in strongly acidic solution and subsequent thermal activation of the reaction product in vacuum. The X.r.d. pattern and infrared spectrum of the product were consistent with data for (V0)2P20,.2 Samples of ca. 55 mg (surface area 11 m2 g-l) were compressed (3.5 ton) into self-supporting discs (diameter 25 mm) and mounted in a conventional infrared cell fitted with fluorite windows and an external furnace6 and glassblown to a vacuum apparatus capable of maintaining a dynamic vacuum of ca. lop4 N m-2. Discs were heated at 673 or 713 K for 17 h before spectroscopic study of the adsorption of water or pyridine.Spectra were recorded using a Perkin-Elmer 580A spectrophotometer. Water was ion-exchanged and triply distilled, once from alkaline potassium per- manganate and twice from itself, all under nitrogen. Water and deuterium oxide 27732774 Infrared Spectra of (VO),P,O, I I I 3500 3000 2500 17 wavenumber/cm -l 10 1600 Fig. 1. Spectra of (VO),P,O, after evacuation (17 h, 673 K) and (a)-(d) evacuation (ca. 300 K, 1 h) following exposure to water vapour (ca. 300 K) at pressures of (a) 0, (b) 57, (c) 114 and ( d ) 228 N m-,; (e)-(f) exposure to water vapour (228 N m-,, ca. 300 K) and evacuation at (e) 493 and df) 693 K. (Fluorochem 2 99.8%) stored in bulbs on the vacuum line were freed from permanent gases by series of freeze-thaw cycles. Pyridine (2 99.5%) was purified by double distillation from solid KOH.Results The infrared spectrum of (VO),P,O, exhibits a series of bands in the 700-1300 cm-l spectral region which may be assigned to vibrations of V02+ and P,Ot- ions in the bulk phase., Spectra of pressed discs contained a broad band envelope in the range 2500-3550 cm-l [fig. 1 (a)] due to OH-stretching vibrations of trace amounts of water or OH-containing species retained after the vacuum decomposition of the precursor, (VO),P,O, - 2H,O. Exposure of discs to water vapour enhanced the intensity of absorp- tion in the 2550-3700 cm-l region and led to the appearance of a maximum at 1620 cm-l due to the deformation vibrations of non-dissociatively adsorbed water molecules [fig. 1 (b)-(d)]. The adsorbed water was not desorbed by evacuation at the ambient temperature in the sample beam of the spectrophotometer [fig.l(d)] and some water was retained even after evacuation at 493 K [fig. 1 (e)]. Exchange experiments involving deuterium oxide were carried out to distinguish between bulk and surface OH-containing species. The broad band envelope at 2550-3550 cm-l [fig. 1 (a)] in spectra of (VO),P,O, before the adsorption of water is predominantly ascribed to bulk species because the envelope was retained in spectra after exposure of discs to excess deuterium oxide vapour [fig. 2(c)]. Adsorbed deuterium oxide gave a broad maximum centred at 2600cm-l, equivalent to the maximum at ca. 3500 cm-l for adsorbed water. A sequence of alternate treatments with water and deuterium oxide established the complete reversibility of the exchange process [fig.2 (c) and (41. A final treatment with water gave a closely similar spectrum [fig. 2(d)] to thatS. J. Puttock and C. H. Rochester 2775 I I I 3 500 3000 2500 wavenumber/ cm Fig. 2. Spectra of (VO),P,O, after evacuation (17 h, 673 K), exposure to excess water vapour (ca. 300 K) and evacuation at (a) ca. 300 and (b) 693 K, (c) after subsequent exposure to excess D20 vapour and evacuation, then excess H20 vapour and evacuation and finally excess D,O vapour and evacuation (all at ca. 300 K), ( d ) after subsequent re-exchange with H,O and evacuation at ca. 300 K. [fig. 2 (a)] initially recorded after exposure of (VO),P,O, to water vapour and evacuation at ca. 300 K. Spectra of (VO),P,O, after heat treatment in vacuum at 71 3 K and exposure to pyridine vapour at ambient temperature are shown in fig.3A. Infrared bands at 1610, 1575 and 145Ocm-l for low pressures of pyridine may be assigned to vibrations of pyridine molecules7 adsorbed through coordinative interaction with Lewis-acidic surface sites.8-11 Th concomitant appearance of a weak band at 1540 cm-l suggested the presence of some Brmsted-acidic sites on the (VO),P,O, surface. A band in spectra at 1490 cm-l must have contained contributions due to vibrations of both pyridinium ions and coordicatively adsorbed pyridine lo, l1 Spectra of discs in contact with high vapour pressures of pyridine [fig. 3 A ( d )-(f)] contained no additional features due to adsorbed pyridine, but exhibited additional bands at 1587 and 1440 cm-l, in particular, due to vibrations of pyridine molecules in the vapour phase [fig.3A(d)]. Removal of pyridine vapour by evacuation with the sample at room temperature gave a spectrum [fig. 3 B (c)] showing the retention of both adsorbed pyridine on Lewis-acidic sites and adsorbed pyridinium ions. Subsequent evacuation at 323 K had little effect, but the extents of desorption of both species progressively increased as the sample temperature was raised during evacuation of the infrared cell until at 673 K desorption was complete (fig. 3B). Water vapour was added to a (VO),P,O, disc on to which pyridine had been pre- adsorbed in order to test for the conversion of surface Lewis-acidic sites to Brarnsted- acidic sites in the presence of adsorbed water.Strong evidence for conversion was provided by the observed decrease in the intensity of the maximum at 1450 cm-l and the parallel increase in intensity of the band at 1540 cm-l as the surface concentration of2776 Infrared Spectra of (VO),P,O, . . I I 1600 1400 1600 1400 w avenumber/ cm Fig. 3. A, Spectra of (VO),P,O, after evacuation (17 h, 713 K) and exposure to pyridine vapour (ca. 300 K) at pressures of (a) 0, (b) 0.18, (c) 0.36, ( d ) 0.54, (e) 1.17 and d f ) 1.35 kN m-,. (g) Spectrum of pyridine vapour (1.35 kN rn-,). B, (a) as A(a), (b) as A(f), (c)-(i) after subsequent evacuation (c) at 293 K (17 h), followed by 1 h at (d) 323, (e) 393, d f ) 463, (g) 543, (h) 593 and (i) 673 K. adsorbed water was increased (fig. 4A). Adsorbed water gave a broad maximum at 1620 cm-l which obscured bands in the 1570-1670 cm-l region due to adsorbed pyridine.The reversal [fig. 4A(b) and (f)] of the relative intensities of the maxima at 1450 and 1490 cm-l confirmed that the latter contained contributions due to pyridine interacting with both Lewis- and Brarnsted-acidic sites. Subsequent thermal activation in vacuum led to the desorption of both water and pyridine (fig. 4B) at temperatures consistent with those observed in the separate experiments with water (fig. 1) and pyridine (fig. 3) alone. The addition of deuterium oxide to coadsorbed water and pyridine on (VO),P,O, decreased the intensity of the band at 1620 cm-l due to the bending vibrations of water molecules and hence revealed the narrower band at 1610 cm-l [fig.4C(d)-( f)] due to a vibration of pyridine adsorbed on Lewis-acidic surface sites. The exchange of adsorbed water by adsorbed deuterium oxide caused the disappearance of the weak band at 1540 cm-l ascribed to pyridinium ions (pyH+) because of the concomitant conversion of the latter to their deuterated analogue pyD+. The effects of the H/D-isotope exchange were reversed by the reintroduction of water to the infrared cell [fig. 4C(g)]. The absence of the band at 1540 cm-l for (VO),P,O, with coadsorbed pyridine and deuterium oxide was further confirmed by admission of deuterium oxide to a pyridine-covered surface, which had not also been exposed to water (fig. 5A). In accordance with the present conclusions, the spectrum of a solution of pyD+Cl- in deuterium oxide contained no band at 1540 cm-l [fig.5 A (g)], at which position there was a strong adsorption maximum in the spectrum of pyH+Cl- in water [fig. 5A(h)].12 The spectrum of pyD+Cl- contained a strong band at 1483 cm-l which was responsible for the growth in intensity and small shift to lower wavenumbers of the band atS. J. Puttock and C . H. Rochester 2777 B 1600 1400 1600 1400 1600 1400 w avenumber/ cm -' Fig. 4. A, Spectra of (VO),P,O, after (a) evacuation (1 7 h, 673 K), (b) exposure to pyridine vapour (0.54 kN m-,) and evacuation (ca. 300 K), and subsequent exposure to water vapour (ca. 300 K) at pressures of (c) 0.19, (d) 0.38, (e) 0.67 and (f) 1.25 kN m-2. B, (a) as A(a), (b) as A(f), (c)-(h) after subsequent evacuation (1 h) at (c) ca. 300, ( d ) 383, (e) 433, cf) 533, (g) 603 and (h) 673 K.C, (a) as A(a), (b) after subsequent evacuation (1 h, ca. 300 K) and exposure to pyridine followed by water [cf. spectrum A(f)], (c)-(f) after evacuation (1 h, ca. 300 K) following subsequent exposure to D,O vapour at pressures of (c) 96, (d) 192, (e) 384 and (f) 768 kN rn-, and (g) after re-exchange with H,O (125 kN mb2) followed by evacuation (1 h, ca. 300 K). 1490 cm-l [fig. 5A(b)] as pyD+ ions were generated on the (VO),P,O, surface by the adsorption of deuterium oxide [fig. 5 A (b)-(f)]. A concurrent weakening of the maximum at 1610 cm-l accompanied the conversion of pyridine molecules at Lewis-acidic sites to pyD+ ions. Desorption of pyridine from Lewis- and Brsnsted-acidic surface sites was complete after evacuation of (VO),P,O, at 713 K (fig.5 B). Discussion The existence of surface hydroxy groups on (VO),P,O, which had been heated at 7 13 K in vacuum was shown by the appearance of the band at 1540 cm-l in spectra of adsorbed pyridine. The absence of narrow infrared bands due to OH-stretching vibrations suggests that the hydroxy groups were involved in hydrogen-bonding interactions, probably with adjacent surface phosphate ions. Contributions to the infrared spectra due to the hydroxy groups were obscured by the broad maximum at 2550-3550 cm-l, primarily attributed to bulk species, and were too small and diffuse to be detectable by the deuterium-exchange experiments. Isolated surface P-OH groups, not involved in hydrogen bonding, would have been expected to give a narrow infrared band at ca.3700 cm-l.13 The surface of (VO),P,O, is Brernsted-acidic after heat treatment at temperatures typical of those used in the catalytic oxidation of butane or butenes to maleic anhydride. Brsnsted-acid sites could therefore be involved in the catalytic process, possibly at the 92 FAR 12778 Infrared Spectra I I - 1600 1400 1600 1400 w avenumberlcm-' Fig. 5. A, Spectra of (VO),P,O, after (a) evacuation (17 h, 673 K; 17 h, ca. 300 K), (b) exposure to pyridine vapour (0.54 kN m-2) and evacuation (ca. 300 K), (c)-cf) subsequent contact with D 2 0 vapour at pressures of (c) 96, (d) 192, (e) 288 and cf) 384 kN m-2. (g) Spectrum of a solution of pyridine (22 ~ 0 1 % ) in concentrated DC1 in D20, (h) spectrum of pyridine (22 ~ 0 1 % ) in concentrated aqueous HC1.B, (a) as A(a), (b)-(h) disc treated as for spectrum Acf) followed by evacuation (1 h) at (b) 293, (c) 323, ( d ) 443, (e) 553, (f) 603, (g) 673 and (h) 713 K. early stage of butene is~merization.~ An alternative source of Brarnsted acidity to surface hydroxy groups would be non-dissociatively adsorbed water molecules coordinatively bound to exposed vanadium cations. Intensities of the infrared band at 1620cm-l showed that water could be the source of Brarnsted acidity at least up to 493 K, but that heating (VO),P,O, at 693 K left an undetectable level of retained surface water. Analysis of the intensities of the bands at 1540, 1490 and 1450 cm-l in spectra of adsorbed pyridine have enabled the relative numbers of Lewis and Brarnsted sites on (VO),P,O, to be estimated.Using a formula containing the intensities of the bands at 1490 and 1450 cm-l [ref. (S)] gives a value of 2.1 times as many Lewis sites as Brarnsted sites on (VO),P,O, which had been heated at 673 K in vacuum. Subsequent saturation with water at room temperature resulted in the conversion of Lewis-acidic to Brarnsted-acidic sites and the surface concentration of the former fell to 0.7 times that of the latter. An alternative method of calculation using the intensities of the maxima at 1540 and 1450 cm-l lo* l1 gave a value of ([L]/[B]) of 2.0 for the initial surface and 0.5 after the adsorption of water. Lewis sites on (VO),P,O, heated in vacuum are probably exposed V4+ cations. We thank the S.E.R.C. for a Studentship (to S. J.P.). References 1 A. I. Pyatnitdkaya, G. A. Komashko, V. A. Zazhigalov, V. M. Belousov, 0. Y. Polotnyuk, S. P. 2 E. Bordes and P. Courtine, J. Catal., 1979, 57, 236. Chaikovskii and G. Ladwig, Kinet. Katal., 1979, 17, 94.S. J. Puttock and C. H. Rochester 2779 3 L. Morselli, F. Trifiro and L. Urban, J. Catal., 1982, 75, 1 12. 4 M. Ai and S. Suzuki, Bull. Chem. SOC. Jpn, 1974, 47, 3074. 5 E. P. Parry, J. Catal., 1963, 2, 371. 6 A. Buckland, J. Ramsbotham, C. H. Rochester and M. S. Scurrell, J. Phys. E, 1971, 4, 146. 7 C. H. Kline and J. Turkevich, J . Chem. Phys., 1944, 12, 300. 8 M. R. Basila and T. R. Kantner, J. Phys. Chem., 1966, 70, 1681. 9 E. P. Parry, J. Catal., 1963, 2, 371. 10 J. W. Ward, J. Catal., 1968, 11, 271. 11 P. Pichat, M-V. Mathieu and B. Imelik, Bull. SOC. Chim. Fr., 1969, 2611. 12 N. S. Gill, R. H. Nuttall, D. E. Scaife and D. W. A. Sharp, J . Inorg. Nuct. Chem., 1961, 18, 79. 13 J. B. Peri, Discuss. Faraday SOC., 1971, 52, 55. Paper 511922; Received 31st October, 1985 92-2
ISSN:0300-9599
DOI:10.1039/F19868202773
出版商:RSC
年代:1986
数据来源: RSC
|
|