摘要:

J. Chem. SOC., Faraday Trans. I, 1988, 84(6), 2155-2171 Mass Transport in Channel Electrodes The Application of the Backwards Implicit Method to Electrode Reactions (EC, ECE and DISP) Involving Coupled Homogeneous Kinetics Richard G. Compton," Matthew B. G. Pilkington and Geoffrey M. Stearn Physical Chemistry Laboratory, Oxford University, South Parks Road, Oxford OX1 3QZ It is shown how the backwards implicit (BI) method may be applied to the description, under steady-state conditions, of complex electrode reaction mechanisms involving coupled homogeneous kinetics and several kinetic species, taking place at channel electrodes. Specifically the EC, ECE, DISPl and DISP2 mechanisms are examined, and it is shown how the transport- limited currents and current-voltage curves associated with these processes may be computed.The advantages of the method, particularly when compared with analytical approaches, are stressed. The channel electrode has been recently developed into a popular and uniquely powerful hydrodynamic electrode, both for analytical purposes and for the mechanistic investi- gation of those electrode reactions which involve coupled homogeneous kinetics.' * It consists of an electrode embedded in the wall of a rectangular duct through which electrolyte solution is flowed, as shown in fig. 1. In the absence of any kinetic complications, and in the presence of an adequate quantity of supporting electrolyte, the mass transport to the electrode is described by the following time-dependent convective- diffusion equation : where x and y are defined in fig.1, c is the concentration of the species of interest, D its diffusion coefficient and v, represents the solution velocity profile in the x-direction. The latter is parabolic, provided one is considering laminar flow and that a sufficiently long lead-in section is present for the flow to become fully de~eloped.~ Quantitatively, (2) where v, is the velocity at the centre of the channel, y' = h - y and h is the half-height of the channel. In writing eqn (2) for the channel we are assuming both that the width, w, of the electrode is sufficiently less than the channel width, d, for edge effects to be neglected and also that h 4 d. A further assumption implicit in eqn (1) is that diffusion in directions other than normal to the electrode surface may be neglected.This approximation has been thoroughly explored and has been shown to be entirely satisfactory for practical electrode^.^ Two approaches to the solution of the mathematical problem stated by eqn (1) and (2) have been adopted. In one case one proceeds by introducing the Lkvique approximation, which was first introduced in the context of the corresponding heat- transfer problem in pipes.5 This facilitates analytical methods and involves the approximation of linearising the parabolic velocity profile near the electrode : V , = v,( 1 -y'2/h2) V, z 221,(1 -y'/h). (3) 21552156 Mass Transport in Channel Electrodes ' 0' xe flow Fig. 1. The coordinate system for the channel electrode cell. A considerable variety of channel electrode problems have been solved with this method,4* although the solutions only apply to electrodes of certain geometries and to flow rates constrained within certain bounds such that concentration changes induced by the electrode are confined to be close to its surface and are thus located in a region of space where eqn (3) is a reasonable approximation.Alternatively, and with considerably more generality, Anderson has retained the full parabolic term in eqn (1) in his treatment of some steady-state problems.''. l2 This, however, dictates the adoption of numerical methods; in particular, the backwards implicit (BI) method was found to be optimal. l2 The method is economical in computer time, since, whereas alternative methods involve matrix calculations based on a two-dimensional grid representing the (x, y) space, the BI method only involves vector calculations.12 The vectors describe concentrations in the y-direction for different values of X.The calculation proceeds 'downstream', each vector enabling the calculation of the next, starting from the vector defining the boundary conditions specified for upstream of the electrode. An additional advantageous feature is that the computation is independent of the value of (O,O), i.e. the concentration at the extreme upstream edge of the electrode. This value would otherwise be problematic, since if, for example, one wished to calculate the transport- limited current, one would have to decide whether (0,O) was electrode or upstream wall : in other words, whether the concentration at (0,O) was either zero, corresponding to complete conversion of electroactive material arriving at the electrode, or else was equal to the upstream (bulk) concentration so that there was no flux through the wall.Alternative computational strategies to the BI method (such as the explicit finite difference method or the Crank-Nicholson method) may give answers dependent on this choice and so may be regarded as less satisfactory in that respect.'l It is apparent from the above that channel electrode problems are best tackled via the BI method, since this allows for the inclusion of a realistic solution flow profile and hence makes fewer limitations on the type of electrode to which the results are applicable. To date the BI method has been applied to the calculation of steady-state limiting currents at channel electrodes of various geometries.".l2 These calculations have so far thus involved kinetically stable species only and very simple boundary conditions at the electrode surface. The purpose of this paper is to show how the BI method can be extended to deal with complex electrode reaction mechanisms involving coupled homogeneous kinetics and several chemical species. The computation of both limiting currents and current-voltage curves (which require potential-dependent boundary conditions at the electrode surface) will be demonstrated. In respect of the former we will consider the ECE-DISP mechanism, and as regards the latter the EC mechanism will be examined. The nature of, and the background to, these mechanistic pathways will now be briefly defined.The ECE-DISP problem arises in a number of electrode reactions that behave as if two or more electrons are being transferred simultaneously. Thus, consider a reactionR. G. Compton, M. B. G. Pilkington and G. M. Stearn 2157 mechanism in which the first electron transfer is followed by a chemical reaction, the product of which is more readily reduced (oxidised) than the starting material, thus producing an overall two-electron process, viz : A+e- = B (9 B = C (ii) C+e- = D (iii) B+C+A+D. (iv) We use the notation k(n) for the forward rate constant of reaction (n), k(-n) for the rate constant of the reverse of reaction ( n ) and K(n, for the corresponding equilibrium constant. It is apparent that the second electron transfer in the above scheme may occur either at the electrode (ECE mechanism); via reaction (iii), or in the bulk solution, via disproportionation (DISP mechanism) through reaction (iv).Within the DISP scheme there are two further possibilities, DISP1 or DISP2, according to whether reaction (ii) or (iv) is rate-determining.13q1* The distinction is an important one because of the large number of reactions it includes and the kinetic and synthetic implications when other reactions (H-atom transfer, dimerisation etc.) compete with the two electron process. Whilst DISP2 is readily indentified by conventional electrochemical techniques (such as linear sweep voltammetry) ECE and DISP 1 frequently give identical responses. l5 Thus for example the variation of the transport-limited current at the rotating-disc electrode with rotation speed is so similar as to preclude the discrimination between ECE and DISPl, at least at the experimental accuracies normally employed," and spectro- electrochemical methods have been re~0mrnended.l~ However, we have recently pointed out' that because of the non-uniformity of the channel electrode and also through the wide range of flow rates that are readily available at this electrode, the flow- rate dependence of the transport limited current may represent a realistic method of tackling the ECE-DISPl problem.As the flow rate is decreased a transition from a one- to a two-electron process is observed corresponding to the intermediate B spending longer in the vicinity of the electrode and hence being more likely to decay into electroactive C.In this paper we generate 'working curves' using the BI method for ECE, DISPl and DISP2 processes at channel electrodes and examine the factors which cause a transition between these limiting cases. In the cases of ECE and DISPl analytical theory developed within the LCvtque approximation exists,"* and the conditions under which these apply are examined in the light of the more general theory. The EC mechanism is defined by the following kinetic scheme: A+e- = B (v) B -+ C. ( 4 In this case variation of the solution flow rate has no effect on the transport-limited current but, providing that the redox couple, A/B, is reversible, there is an effect on the half-wave potential in that the following kinetic process causes the voltammetric reduction wave to be shifted anodically and this effect is more pronounced at lower flow rates.An approximate analytical treatment of the problem at channel electrodes has been givenlg by invoking the Leveque approximation. A fuller theory is now developed. Theory In this section we show how the backwards implicit method may be extended to cover electrode reactions which include coupled homogeneous kinetics. In addition the theory is extended to cover the calculation of currents over the full potential range. The full theory is first summarised for the case of the calculation of the limiting current for a2158 Mass Transport in Channel Electrodes Y J J J - - 2 h J - j + i J - 0 k X I I wal I ~ channel K-2 - , K-electrode X , 1 2 k-I k k+l Fig:2. The finite difference grid.simple electron-transfer reaction and then the necessary developments detailed, first in general terms, and then for specific cases of interest. Consider a simple electron transfer, A+e- = B. The steady-state transport equation for A is where a = [A] and a,, is the bulk concentration of A. In order to apply the BI method we cover the xy plane with a two-dimensional finite-difference grid (fig. 2). Increments in the x-direction are Ax and in the y-direction, Ay. y j =jAy j = 0,l ... J where Ay = 2h/J ( 5 ) xk = kAx k = 0,l ... K where Ax = x,/K. (6) The derivatives in eqn (4) are approximated to - g t k + l - & k ax Ax where gA is the normalised concentration, gA = a/a,, and Thus combining eqn (4), (7) and (8) we obtain (7)R. G .Compton, M. B. G. Pilkington and G. M. Stearn 2159 where DAx(2h) 3d 6 Vf ~ ( A Y ) ~ (2h - jAy) A. = ’ and 6 is the average volume flow rate. The appropriate boundary conditions for the calculation of the transport-limited current are x=O a = a , g t , = l (1 1) y = O a = O g $ = O (12) y = 2h &Z/ay = 0 g;, k = gtFl, k. (13) Thus &,k = (2Al+ l)gf,k+l-Algt,k+l (14) g t k =-Aig~-1,k+l+(21”j+l)gjqk+l-AigjAfl,k+l j = 233 .-- J-2 (15) & l , k = -IZJ-1&-2,k+1+(’J-1+ ‘)&-l,k+l* (16) These (J-1) simultaneous equations may be expressed as a (J-l)x(J-1) matrix equation : i.e. where 0 . ai , 0 ‘,. Ci aJ-2 0 ‘J-2 ‘J-2 ‘J-1 ’J-1 di=g& j = 1,2 ... J-1 ui = gtk+l j = 1,2 ... J - 1 a, = -Ai j = 2,3 ... J-1 bi = (24+ 1) j = 1,2 ... J-2 bJ-l = AJ-l + 1 ci = --Ai j = 1,2 ...J-2. The matrix [TI being a tridiagonal form allows the use of the Thomas algorithm2’ (see Appendix), to give ( u } ~ from (djk. The boundary condition g?, = 1 supplies the vector { d } , from which {u}, is calculated. in the absence of homogeneous chemical complications, {d}k+l = (u},, so (u), is calculated from (d}, (= (u},) and so on Until (u}, is obtained. Thus all the values gtk ( j = 1,2 .. . J- 1, k = 1 ... K ) are evaluated. The expression for the current, I, at the electrode takes the following form: K gl, k-gO, k k-12160 Mass Transport in Channel Electrodes 13.61 13,O: I ’ ’ I I I I 1.0 1.5 2.0 2.5 3.0 Fig. 3. The convergence of the calculated value for the transport-limited current at a channel electrode as the grid size is increased.The electrode geometry was 2h = 0.04 cm, d = 0.6 cm, x, = 0.4 cm, w = 0.4 cm, D = cm2 s-I and = cm3 s-l. where F is Faraday’s constant. The boundary conditions give g;,k = 0, so that wFDa, Ax I = c g 1 , k - ‘Y k-1 Using the theory outlined above, the limiting current was computed (on a Norsk ND 500 computer) and convergence examined by varying J and K values. The convergence of typical limiting-current data is portrayed in fig. 3 with equal J and K. If values of J and K greater than 1000 were used it was necessary to work in double precision, as round- off errors became significant in single precision. Convergence to four significant figures was obtained with values J = 200, K = 500. Results for the flow rate and electrode geometry dependence of the transport-limited current were identical to those reported hw Anrlprcnn rrpf 1131 fin 31 We consider next, in general terms, the modifications in the above which arise when electrode reaction mechanisms involving more than one species are examined.Concentrations of each species, i, present along a column in the finite difference grid parallel to the y-axis are described by vectors {u}:. Adjacent vectors, (u>: and {u}:+~, are related by a set of ( J - 1) simultaneous equations. These are obtained for each species present from the appropriate convective-diffusion equations using the approximations in eqn (7) and (8). In all cases the ( J - 1) simultaneous equations may be expressed by a matrix equation (26) where [TIt is the tridiagonal matrix defined in eqn (17) above.In all reaction schemes considered, ui are equal to the normalised concentrations g;, k+l (= i/ao): ai and cj are both equal to -A, [see eqn (lo)] ; bj ( j = 2 , 3 . . . J - 2) is equal to 2Aj + 1 and bJ-l is equal to AJ-, + 1. The parameters dj and b, are dependent on the system being modelled and are listed below for specific mechanistic cases of interest. {d}i = [ TIi {uji The ECE-DISP Problem We consider first the generation of ‘working curves ’ for each of the ‘pure ’ mechanisms, ECE, DISPl and DISP2, respectively. The question of mixed kinetics is addressed subsequently.R. G. Compton, M. B. G. Pilkington and G. M . Stearn 2161 The appropriate steady-state transport equations for the ECE mechanism [see a2c ac D --ZI -+k(ii, b = 0. ay2 ax The boundary conditions relevant to the evaluation of the effective number of electrons transferred under transport-limited conditions are x = 0: a = a,, b = 0, c = 0 y = 0: a = 0, ab/ay = -aa/ay, c = 0 y = 2h: aa/ay = 0, ab/ay = 0, ac/ay = 0.(30) (31) (32) On applying the general theory outlined above three matrix equations result, one for each of i = A, B and C. The mechanism dependent elements in eqn (26) are given in tables 1-3. The matrix equation for B is dependent on g;k+l because of the boundary condition (31), so that the equations are solved, for each column of the grid, in the order of A, then C , and finally B. The current is the sum of that due to the reduction of both A and C: The generalisation of eqn (25) to eqn (33) for computational purposes is obvious.We consider next the pure DISP 1 mechanism. The pertinent steady-state transport equations are a2a aa D - - U , -+k(ii) b = 0 ay2 ax a2b ab D - - V , --2k(ii) b = 0. ay2 ax The boundary conditions are x = 0: a = a, b = 0 y = 0: a = 0, ab/c?y = -aa/ay y = 2h: = 0, ab/ay = 0. These expressions are again used to construct matrix equations fo both A a (34) (35) Table 1 shows the matrix elements that are unique to this mechanism. It is apparent the equation for B again requires g:,,,. For this reason we first calculate g t k + , by solving eqn (26) with the dj defined as in table 1, i.e. since we know g& We then solve for g;".+' by solving the matrix equation for B, since then g t k t l is known. The calculation again proceeds column by column, finding first A and then B.Notice that this strategy is possible since we know both gt, and g;,. The DISPl current is then evaluated using eqn (25). We turn now to DISP2. The required transport equations are (39)2162 Mass Transport in Channel Electrodes Table 1. Values for the matrix element dl mechanism species ECE ECE ECE DISP 1 DISP 1 DISP2 DISP2 mixed ECE/DISP mixed ECE/DISP mixed ECE/DISP 'rev 'rev EC EC Table 2. Values for the matrix elements dj ( j = 2,3 . . . J- 1) mechanism species d j ( j = 2 , 3 ... J-1) ECE ECE ECE DISPl DISP 1 DISP DISP2 mixed ECE/DISP mixed ECE/DISP mixed ECE/DISP b r e v E r e , EC EC The boundary conditions are x = 0 : a = a,, b = 0 y = 0 : a = 0, = - ab/ay y = 2h: aa/ay = 0, abpy = 0. (42) (43) The matrix equations for A and B utilise the elements shown in table 1.Again computation proceeds column by column ' down ' the electrode, with A being computed before B for the same reason as above. The DISP2 current is found using eqn (25). In order to investigate the factors governing the transition from ECE to DISPl orR. G. Compton, M . B. G. Pilkington and G. M . Stearn Table 3. Values for the matrix element b, 2163 mechanism species bl ECE ECE ECE DISP 1 DISP 1 DISP2 DISP2 mixed ECE/DISP mixed ECE/DISP mixed ECE/DISP Ere, Ere, EC EC DISP2, we investigated in addition the situation in which all the reactions (i)-(iv) are significant. This implies the following transport equations : a2a aa D - - V , -+kav) bc = 0 ay2 ax a2b ab D -- V, -- k(ii) b + k(-ii) C- kBv) bc = 0 a p ax a2c ac D - - ~ , - - + k ( ~ ~ , b - k ~ - ~ ~ ) c - k ( ~ , , b c = 0 a y ax (44) (45) and the following boundary conditions : x = 0 : a = a,, b = 0, c = 0 y = 0: a = 0, ab/ay = -aa/ay, c = 0 y = 2h: aa/ay = 0, ab/Q = 0, ac/Q = 0.(47) (48) (49) The matrix elements for the equations for A, B and C in this 'mixed' mechanism appear in table 1. Again these definitions require us to compute B after A and C . The current is deduced from eqn (33). The EC Mechanism As a preliminary the BI method was extended to calculate the current-potential curve for the following simple reversible electron transfer (Ere") : A+e- = B. The relevant transport equations are a2a aa D--v, -= 0 a y ax ayz ax a2b i3b D - - V , - = 0 71 (51) FAR I2164 Mass Transport in Channel Electrodes with boundary conditions x = 0 : a = a,, b = 0 y = 0 : a / b = exp (0), ab/@ = -aa/ay (53) y = 2h: aa/ty = 0 , a b p y = 0 (54) where 8 = (F/RT) (E- E O ) .(55) The matrix elements needed for this problem are presented in table 1. A difficulty arises since, owing to the y = 0 boundary condition, the two matrix equations are inter- dependent, i.e. we can only find g t k + l with a knowledge of g:k+l and vice versa. An iterative method was employed whereby, initially, gf, k+l was approximated by g:, in order to calculate g;k+l. This was used to obtain a better value for gfk+l, and the procedure was continued until both the gf,k+l and g:k+l values were unchanged on further iteration. Currents were calculated using a range of O values from - 10 to 10 so as to produce the full voltammetric wave.Currents at 0 = 0 were found to be half those as 0 + - co, and Tafel plots21 were linear, with a slope of 59 mV per decade. The validity of the general approach for potential-dependent boundary conditions was thus confirmed. Turning next to the EC mechanism, the boundary conditions of eqn (52), (53) and (54) are unchanged, as is the transport eqn (50). Eqn (51) is, however, replaced by The required matrix elements are shown in table 1. The same iterative method as used for the Ere, case was employed. Results and Discussion We consider the pure ECE and DISPl results first since, for these mechanisms, ana- lytical theory derived under the Lkvlique approximation is available for comparisonZ7* l8 with the numerically generated results. The former work describes17* l8 how the effective number of electrons transferred, neff, depends on the parameter Kl = k ( i i ) ( h 2 ~ 3 4 ~ i D)i.(57) Fig. 4 shows results generated for a flow cell of the following (typical) dimensions:22 2h =0.04 cm; d=0.6 cm; xe=0.4 cm; w = 0 . 4 cm; D = 3 x cm2 s-l and 6 = cm3 s-' (6 = 4v, hd/3). Also depicted are points calculated using the analytical theoryZ8 with these parameters. Excellent agreement is found, as is also seen to be the case for the DISPl mechanism (fig. 5). The parameters given above for the computation of fig. 4 and 5 correspond to the conditions for which it would be expected that the LkvCque approximation would work well. That is when the diffusion layer at the electrode is very thin compared to the height (2h) of the cell, so that concentration changes resulting from electrolysis are confined to where eqn (3) is a good approximation.Experimentally this corresponds to deep (large h) channels with fast electrolyte flow rates. The agreement with analytical theory evident from fig. 4 and 5 gives confidence in the general computational approach developed above and encourages the use of the BI method for reasons of generality, encompassing non-Levich geometries, and for those problems, such as DISP2, for which analytical approaches are not available. The effect of varying h and 6 outside of the Livique limit can be seen from fig. 6 and 7 which display the 'working curves' pertinent to the DISP2 and EC mechanisms. Considering the DISP2 case first, the application of the LevCque approximation to eqn (39) and (40) shows that under conditions where this is a valid procedure, the effectiveR. G .Compton, M. B. G. Pilkington and G. M. Stearn 2165 Fig. 4. The effective number of electrons transferred in a pure ECE process as a function of the normalised rate constant K, calculated via the BI method (-) or via approximate analytical theory ( x ). The cell parameters employed are given in the text. 2 .o 1 .a 1.6 k p: 1.4 1.2 1 .o log K1 Fig. 5. The effective number of electrons transferred in a pure DISPl process as a function of the normalised rate constant Kl calculated via the BI method (-) or via approximate analytical theory ( x ). The cell parameters employed are given in the text. 71-22166 Muss Transport in Channel Electrodes 1 I I I I I I 40 80 120 1.01 K2 Fig. 6.The effective number of electrons transferred in a pure DISP2 process as a function of the normalised rate constant K, for various values of 2h: A, 0.001; A, 0.0025; +, 0.005; 0, 0.01; V, 0.02 and x , 0.04 cm. The other cell parameters are defined in the text. 4 I I I I I - 0.5 0 0.5 1 0 K1 Fig. 7. The shift in half-wave potential, EL, with Kl for an EC reaction at a channel electrode with 5 = (a) 0.001, (b) 0.01 and (c) 0.1 cm3 s-l: The other parameters are as in fig. 3. Also shown is the ' Levich ' behaviour (- - - -). l9 number of electrons transferred would be expected to be a unique function of the following normalised rate constant K2 = (2k,,;, K(,,,) a,(h2x:/4ui D);. (58) Fig. 6, which was generated using the same parameters as for fig.4 except that nowR. G. Compton, M . B. G. Pilkington and G. M. Stearn 2167 Fig. 8. A three-dimensional plot showing neff as a function of k(iv) and K,. The cell parameters are as in fig. 3. x = log k(iv), y = K,, z = n,,,. 6 = cm2 s-l, shows how neff depends on K2 as 2h is increased from a very small value. For sufficiently deep channels the curves become essentially superimposable, and this corresponds to the Livcque limit. Notice that as h -+ 0 the cell operates essentially as a ' thin-layer ' cell23 in which there is effectively complete conversion of A into D, and a two-electron current is seen for all but the smallest rate constants. The corresponding effect of flow rate may be seen in fig. 7, which shows the shift in half-wave potential with Kl for an EC reaction for three different flow rates (AE;).The curve was generated by computing the full current-potential curves as described above and hence finding the half-wave variation. Also shown is the behaviour obtained by Coles and Comptonlg using the Livique approximation. At the highest flow rate shown the two curves are virtually identical, illustrating the fact that the LCvCque approximation works best at higher flow rates.' The above examples show the quantitative success of the BI method. The method provides theory for more general channel electrode geometries, so that the adoption of the Levich equation is not demanded. It also allows for the production of 'working curves' for cases (such as DISP2) where analytical methods, even with the Livcque approximation, prove intractable.The only limitation of the method lies in the amount of computer time that has to be invested. This is most severe for mechanisms involving fast homogeneous kinetics, since these may give rise to sharply changing concentration profiles, and thus the finite difference grid has to be made finer in the x-direction, i.e. the value of K increased. Nevertheless the data shown in fig. 4-7 were obtained for K equal to no more than 2000 or 3000. However, for significantly faster rate constants than depicted the computing time becomes prohibitive. In addition to the advantages cited above, it should be stressed that the method, in general, offers both simplicity and flexibility in programming, in that changes in mechanism are readily incorporated, thereby affording the experimentalist the opportunity to fully explore the various nuances of any proposed mechanism.The last point is well illustrated by consideration of the 'mixed' ECE/DISP reaction, where all the steps (i)-(iv) are considered. Specifically we will examine the transition from pure ECE to pure DISP as k,!,, is increased. For convenience we put k(+) and k(-iv, = 0. This constrains the DISP limit to be DISPl since reaction (ii) is thus 'forced' to be irreversible. Fig. 8 shows a three-dimensional plot in which neff is plotted vertically. cm3 s-l and D =2168 Mass Transport in Channel Electrodes 7.5'0 Fig. 9. A difference plot showing the transition between ECE and DISPl for an electrode of the geometry specified for fig.3 and a flow rate of lo-' cm3 s-l. x = logk,iv), y = K,, z = (nef, -neff. DISP1). Fig. 10. A difference plot showing the transition between ECE and DISPl for an electrode of the geometry specified for fig. 3 and a flow rate of cm3 s-l. x = logk,iv), y = K,, z = (neff - 'eft, nlSP1)' The two horizontal axes correspond to the normalised first-order rate constant, K,, and to second-order rate constant, k(iv,, respectively. This plot defines the mechanistic behaviour for all cases within the limits imposed and for the flow rates and cell geometry specified. The ECE to DISPl transition is best revealed by plots such as those in fig. 9 and 10, in which the 'working curve' for pure DISPl, shown in fig. 5, has been used to generate values of difference (n,,, - n,,,, DISPl) where n,,,, DrsPl represents the effective number of electrons transferred under pure DISPl conditions.These difference plots are shown for two separate flow rates ( 10-1 and lod3 cm3 s-l) for the same electrode geometry as employed for fig. 3. The changeover from ECE to DISPl as k(iv, is increased is clear. Moreover, the smaller Kl the lower the values of k(iv, needed to induce the transition. This is because for low Kl the intermediate B escapes further from the electrode surface before it is converted into C, and so there is more opportunity for C to react via reaction (iv) because it will then take longer for C to diffuse back to theR. G. Compton, M. B. G. Piikington and G. M. Stearn 2169 electrode surface to react via reaction (iii). Comparison of fig. 9 and 10 shows the effect of flow rate on the ECE-DISP1 transition.Superficially it appears that faster flow rates promote the ECE process. However, note that the y-axis of the three-dimensional difference plots is K l , and this depends on 6. When this is allowed for it is seen that for fixed k(ii) and k(,", and a given cell geometry an increase in flow rate promotes DISPI over ECE, since the enhanced convection removes B from the electrode vicinity and thus induces the homogeneous pathway. In conclusion we have shown how the BI method may be applied to embrace electrode reaction mechanisms involving coupled homogeneous kinetics. Both transport-limited current behaviour and current-potential curves may be generated, without losing the advantages of the BI method over other numerical strategies, i.e.economy of computing time and an insensitivity to the value of the concentration at the extreme upstream edge of the electrode (see above). The approach extends theory beyond electrode geometries for which analytical theories are applicable and hence extends the range of rate constants open to study for a given reaction mechanism type. Moreover, the ease with which modifications of mechanistic detail can be incorporated into the computational procedure enables the consideration of mechanisms of a complexity well beyond that conceivable within the confines of analytical Levich limit theory. The limitations of the method arise purely in terms of the computer time required when very fast rate constants are implicated in the mechanism under study.- 0 ' J - 2 ' J - 1 We thank the S.E.R.C. for a studentship for M. B. G.P. Appendix : The Thomas Algorithm The Thomas algorithm,20 a simplified form of Gaussian elimination, is an efficient method for the solution of tridiagonal matrix systems. The matrix equation to be solved is written as The Algorithm operates by factorising the tridiagonal matrix [TI into two bidiagonal matrices [TL] and [Tu] such that A solution is then found for the vector ( f ) in {dl = [TI{u). [TI = CTLl[TUl. and { f ) is used to give a final solution [T,l{ul = {fl. Since [TL]-'{d) = { f ) , [TJ {u) = [T,]-'{d} and thus ([TI] {T,]){u) = {d}. The matrix [TI of ( J - 1)2 elements can be [TJ = written as a2 62 c2 a J - 2 ' J - 2 0 ' J - 1 -2170 On factorisation [TI takes the form Mass Transport in Channel Electrodes - 0 0 - 1 P 1 0 P J - 2 0 1 - where aj and equation [TI = [TL] [T,], the following relations are obtained : are to be determined. By equating the left- and right-hand sides of the a, = b, 81 = c1/a, aj = bj-ajpj-l j = 2,3 ...J- 1 8, = c j / a j j = 2,3 ... J-2 (aj # 0). After obtaining aj and 8,, the equation for (f} is solved: [TLl{f> = (4 f l = dJa1 where elements f, of ( f} are given by S, = (dj-ajf,-l)/aj ( j = 2,3 ... J- 1). {f> is then used to determine the elements uj of {u) ' J - 1 = f J - 1 uj = f , - & ~ ~ + ~ By this procedure the matrix equation [TI (u] = { d ) is solved for (u). (j = 1,2 ... J-2). References 1 R. G. Compton and P. R. Unwin, J. Electroanal. Chem., 1986, 205, 1. 2 R. G. Compton and P. R. Unwin, Comprehensive Chemical Kinetics, 29, in press. 3 V. G. Levich, Physicochemical Hydrodynamics (Prentice-Hall, Englewood Cliffs, NJ, 1962). 4 J. B. Flanagan and L. Marcoux, J. Phys. Chem., 1974, 78, 718. 5 M. A. LkvZque, Ann. Mines. Mem. Ser. 1928, 12/13, 201. 6 W. J. Blaedel, C. L. Olson and L. R. Sharma, Anal. Chem., 1963, 35, 2100. 7 W. J. Blaedel and L. N. Klatt, Anal. Chem., 1966, 38, 879. 8 R. G. Compton and P. R. Unwin, J. Electroanal. Chem., 1986, 206, 57. 9 R. G. Compton and G. R. Sealy, J. Electroanal. Chem., 1983, 145, 35. 10 T. Singh and J. Dutt, J. Electroanal. Chem., 1985, 182, 259. 11 S. Moldoveanu and J. L. Anderson, J. Electroanal. Chem., 1984, 175, 67. 12 J. L. Anderson and S. Moldoveanu, J. Electroanal. Chem., 1984, 179, 107; 119. 13 C. Amatore and J. M. Saveant, J. Electroanal. Chem., 1977, 85, 27. 14 C. Amatore, M. Gareil and J. M. Saveant, J. Electroanal. Chem., 1983, 147, 1. 15 R. G. Compton and A. R. Hillman, Chem. Br., 1986, 22, 1088.R . G. Compton, M. B. G. Piikington and G. M. Stearn 2171 16 R. G. Compton, R. G. Harland, P. R. Unwin and A. M. Waller, J. Chem. Soc., Faraday Trans 1, 17 R. G. Compton, P. J. Daly, P. R. Unwin and A. M. Waller, J. Electroanal. Chem., 1985, 191, 15. 18 R. G. Compton, D. J. Page and G. R. Sealy, J. Electroanal. Chem., 1984, 161, 129. 19 B. A. Coles and R. G. Compton, J . Electroanal. Chem., 1981, 121, 37. 20 L. Lapidus and G. F. Pinder, Numerical Solution of Partial Dzflerential Equations in Science and 21 W. J. Albery, Electrode Kinetics (Oxford University Press, Oxford, 1975), p. 63. 22 B. A. Coles and R. G. Compton, J. Electroanal. Chem., 1983, 144, 87. 23 B. H. Vassos and G. W. Ewing, Electrotlnalytical Chemistry (Wiley, New York, 1983). 24 T. Singh and J. Dutt, J . Electroanal. C k m . , 1985, 190, 65. 1987, 83, 1261. Engineering (Wiley, New York, 1982). Paper 7/ 1613; Received 3rd September, 1987

ISSN:0300-9599    

DOI:10.1039/F19888402155

出版商:RSC    

年代:1988

数据来源: RSC

摘要:

J. Chem. Soc., Faraday Trans. I, 1988, 84(6), 2173-2180 Structure Dependence in the Hydrogenation of Diolefins over Ru Thin Films Jun Tamaki, Toshiaki Miyanaga and Toshinobu Imanaka* Department of Chemical Engineering, Faculty of Engineering Science, Osaka University, Toyonaka, Osaka 560, Japan Toshimi Yamane Department of Materials Science and Engineering, Faculty of Engineering, Osaka University, Suita, Osaka 565, Japan Ru thin films were prepared by a radiofrequency (r.f.) sputtering method. The structure of the Ru film changed with film growth. This influenced the activity and the selectivity in the hydrogenation of diolefins. The alterations in the activity and the selectivity were explained in terms of the columnar structure of the Ru film and the Ru (002) intensity in X-ray diffraction patterns.The partial hydrogenation of diolefins proceeded preferably over the (002) face of Ru. Elevating the substrate temperature, the more remarkably (002) orientated Ru film was prepared and showed the higher selectivity for the partial hydrogenation than that on the Ru film prepared at room temperature. It is well known that the structure of a metal film prepared by a vacuum evaporation changes from an island structure to a channel structure to a continuous film with film growth. Wilman and coworkers14 o b s e d d the preferred orientation of Ag, Au, Cu and A1 films condensed in a vacuum on a glass substrate. The crystal orientation of these films changed from the inital random orientation to the (1 11) orientation then to the (21 1) orientation with film growth.The most densely populated face, (1 11) for these f.c.c. metals, was initially parallel to the substrate surface. Boichot5+ found a similar preferred orientation of some compounds such as PbS, PbSe and PbTe condensed in a vacuum on an amorphous substrate. Thornton7 observed the tapered columnar structure of a continuous film prepared by a sputtering method. This columnar structure was found to be affected by the substrate temperature. The microstructures of the various metal films were classified with regard to T/T,, where T is the substrate temperature and T, is the coating metal melting point (K). At TIT, < 0.1 (zone l), tapered columnar crystallites develop. Intergrain boundaries are voids rather than true grain boundaries. At higher TIT, (0.1-0.3, zone T), films consist of a dense array of fibrous grains.Above T/T, (0.3-0.5, zone 2), columns extending through the entire coating thickness develop and recrystallization is possible. For TIT, in the range 0.5-0.75 (zone 3) columnar grains tend to be facetted. The facets were often replaced by smooth flat surfaces at higher temperature. In this investigation, Ru thin films were prepared by an r.f. sputtering method. Ru metal has the h.c.p. structure. Microstructures of the Ru films are expected to be varied by film growth and by substrate temperature. The change in the film structure would influence the catalysis by the film. Therefore, the structure dependence was examined to the activity and the selectivity in the hydrogenation of diolefins (buta-l,3-diene and cyclopen tadiene).21732174 Hydrogenation of DioleJins over Ru Experiment a1 The detailed procedure of film preparation was described p r e v i o ~ s l y . ~ ~ ~ An Ru ingot ( 6 ~ 2 0 ~ 3 . 5 mm3) was employed as a target. The ingot was made of Ru powder (99.99 YO) using a sintering technique. The sputtering gas was high-purity Ar (99.9995 %) at a pressure of 0.05 Torr (1 Torr = 133.3 Pa). Under these conditions, an Ru thin film was deposited at a rate of 3 nm min-' onto a Pyrex glass substrate which was placed opposite the target and was maintained at room temperature or 573 K. The thickness of the films was controlled by changing the discharge time. The hydrogenation of diolefins (buta- 1,3-diene and cyclopentadiene) was carried out over the Ru thin films of varying thickness using a closed-circulation system at 323 and 423 K.The ratio H,/hydrocarbon was 6 and the total pressure in the system was 140 Torr. The products of the hydrogenation were analysed by gas chromatography using a thermal conductivity detector. The columns used were a sebaconitrile 5 m column for the hydrogenation of buta-1,3-diene and a dimethylsulpholane 3 m column for cyclopentadiene. The structures and electronic states of the Ru films were investigated using X-ray diffraction analysis (X.r.d.) and X-ray photoelectron spectroscopy (X.P.S.). X.r.d. patterns and X.p. spectra were recorded with a Shimazu VD-1 X-ray diffractometer using Cu K, radiation (35 kV, 20 mA) and a Shimazu ESCA-750 X-ray photoelectron spectrometer using Mg K, radiation (8 kV, 30 mA).Results and Discussion Film Structure and Surface Electronic State Fig. 1 shows the X-ray diffraction patterns of the Ru thin films with varying thicknesses. Three peaks due to Miller indices (101), (002) and (100) of the Ru h.c.p. structure" appeared at 44.4, 42.4 and 38.8". The (101) peak intensity increased and the (002) intensity decreased with increasing film thickness. It was found that the microstructures of the Ru thin films were altered with the film thickness (film growth). The (002) face is the most densely populated face in the Ru h.c.p. structure. In the initial stage of film growth the (002) face is assumed to grow preferentially and to be parallel to the substrate surface. The preferred orientation of the Ru film then changes from the (002) orientation to the (101) orientation.These observations agreed with the results of Wilman and The surface electronic state of the Ru films was evaluated by X.P.S. All binding energies (EB) were referenced to the Ag 3d5/, level (368.2 eV) from a silver paste because the C 1s peak overlapped the Ru 3d3,, peak. The binding energy of Ru 3d5/, was 280.1 eV and did not change with the film thickness. Fig. 2 shows the X.p spectra of the Ru valence band (VB) for Ru films of varying thickness. The shape of the VB spectra was also unaltered by the film thickness. It was found that the electronic state of Ru for the Ru thin films was unchanged with the film thickness. in terms of the change in the film orientation. Hydrogenation of Diolefins The effect of the alterations in the microstructure of the Ru film on the activity and the selectivity in the hydrogenation of diolefins was examined.The activity was estimated on the basis of the geometrical surface area of the substrate. The selectivity was evaluated by extrapolation to 0 % conversion of the product distribution curve. Fig. 3 and 4 show the activities and selectivities for n-butane and but- 1 -ene in the hydrogenation of buta- 1,3-diene at 323 and 423 K, respectively, as a function of the relative intensity of the (002) face which was a proportion of the (002) peak in all the X.r.d. peaks. In both cases,J. Tamaki et al. 2175 45 40 35 28 I" Fig. 1. X-Ray diffraction patterns of the Ru films with various thickness: (a) 50, (b) 100, (c) 200 nm." I il: il: 4- f g . > E E t (II .C t .- A 1 I 1 I 15 10 5 0 binding energy/eV Fig. 2. X-Ray photoelectron spectra of valence band for the Ru films with various thickness: (a) 30, (b) 120, (c) 400 nm.2176 Hydrogenation of DioleJins over Ru E: 0 .r( 0 50 relative intensity of Ru (002) (%) Fig. 3. Initial product distribution and initial rate in the hydrogenation of buta-1,3-diene over the Ru films at 323 K as a function of the relative X.r.d. intensity of Ru (002): ., initial rate; 0, but-1-ene; A, n-butane. h 50 0 .- cm h I P 2 1 Y m V 0 50 relative intensity of Ru (002) (%) Fig. 4. Initial product distribution and initial rate in the hydrogenation of buta- 1,3-diene over the Ru films at 423 K as a function of the relative X.r.d. intensity of Ru (002): ., initial rate; 0, but-1-ene; A, n-butane.the hydrogenation activity decreased with increasing (002) intensity. In other words, the activity increased with increasing film thickness. If the structure of the Ru film is the columnar structure which is characteristic of sputtered films,' the surface area of the film increases with increasing film thickness, resulting in high hydrogenation activity. On the other hand, the selectivity for but-1-ene increased and that for n-butane decreased with increasing (002) intensity. The selectivity to trans- and cis-but-2-ene was independent of the (002) intensity. It was found that the (002) face of Ru was preferred for the formation of but- 1-ene, viz. the partial hydrogenation of buta- 1,3-diene. The adsorbed hydrogen concentration may be lower on the (002) face of Ru than on other faces, resulting in the high selectivity for the partial hydrogenation.It is well known that the strength of adsorption of a diene is stronger than that of a monoene on a Group VIII metal surface." The adsorption state of H, or CO was slightly different on the (OOl), (1 10) and (101) face of Ru with respect to temperature-programmed desorption (t.p.d.) spectra.12-15 Therefore, the strength of adsorption of diene and monoene may also differ on low-index faces of Ru, although few investigations on adsorption of organic molecules have been performed on these faces. When the reaction temperature wasJ. Tamaki et al. 2177 loo c 1 50 n E c .3 ." m E U .- --.- U & - .r( U .- E .3 relative intensity of Ru (002) (%) Fig.5. Initial product distribution and initial rate in the hydrogenation of cyclopentadiene over the Ru films at 323 K as a function of the relative X.r.d. intensity of Ru (002): m, inital rate; 0, cyclopentene. 100 1 130 20 2 P 0 relative intensity of Ru (002) (%) Fig. 6. Initial product distribution and initial rate in the hydrogenation of cyclopentadiene over the Ru films at 423 K as a function of the relative X.r.d. intensity of Ru (002): ., initial rate; 0, cyclopentene. raised from 323 to 423 K, the selectivity to but-1-ene was much enhanced. A similar effect of the film structure was observed in the hydrogenation of cyclopentadiene which, like buta-1,3-diene, is a conjugated diene. Fig. 5 and 6 show the activities and selectivities for cyclopentene in the hydrogenation of cyclopentadiene at 323 and 423 K, respectively, as a function of the (002) intensity.The hydrogenation activity decreased and the selectivity for cyclopentene increased with increasing (002) intensity. The (002) face of Ru was also found to be favourable for the partial hydrogenation of cyclopentadiene. Raising the reaction temperature greatly enhanced the selectivity to cyclopentene. It was reported that there were two types of hydrogen adsorptive site on the Ru (001) face,'' i.e. an h.c.p. site and an f.c.c. site. The h.c.p. sites are defined as three-fold hollow sites such that there is an Ru atom directly below in the second Ru layer. The f.c.c. sites2178 Hydrogenation of DioleJins over Ru R u (101) Ru (002) 45 40 35 201" Fig. 7.X-Ray diffraction patterns of the Ru films of 100 nm thickness prepared at different substrate temperatures : (a) room temperature, (b) 573 K. have no such Ru atom in the second underlying layer. High-resolution electron energy loss spectra for the hydrogen adsorption on the Ru (001) face showed two kinds of symmetric vibration for hydrogen at 88 and 105 meV.'' The t.p.d. spectra of hydrogen on the Ru (001) face showed two desorption peaks at ca. 320 and 430K.15 The adsorption energies on the h.c.p. and f.c.c. sites were calculated by ab initio SCF calculations. l7 According to these calculations, the adsorption energy was larger at the f.c.c. site than at the h.c.p. site. Therefore, the site due to the desorption peak at the higher temperature was found to be the f.c.c.site. At 423 K hydrogen is not assumed to be adsorbed on the h.c.p. site, which has the lower adsorption energy. The reduced surface hydrogen concentration brings about the increase in the selectivity for partial hydrogenation. The extent of this increase was pronounced in the low (002) intensity range. The surface structure of the Ru film is naturally presumed to be changed by raising the reaction temperature. The selectivity in the hydrogenation of diolefins would be affected not only by the adsorptive property of hydrogen but also by the surface structure of the Ru film. Effect of the Substrate Temperature The change in the film structure at high substrate temperature (573 K) was also examined. The melting point of Ru metal is 2523 K.At room temperature, according to the classification of Th~rnton,~ the substrate is zone 1. At 573 K, the substrate is classified as zone T. Films prepared at these two substrate temperatures have different microstructures. Fig. 7 shows X.r.d. patterns of the Ru films with the same thickness prepared at room temperature and 573 K. The (002) intensity obviously increased with increasing substrate temperature. At higher substrate temperatures, the diffusion of sputtered atoms becomes appreciable on the substrate surface. The (002) face, which was the most densely populated face of Ru, was more remarkably oriented. The Ru film prepared at the high substrate temperature was expected to show a higher selectivity for partial hydrogenation. The activities and selectivities in the hydrogenation of buta- 1,3-J.Tamaki et al. 2179 Table 1. Product distribution and initial rate in the hydrogenation of buta-1,3-diene over the Ru films" prepared at different substrate temperatures. distribution (%) substrate reaction trans- cis- initial activity temp. / K temp./K n-butane but-1-ene but-2-ene but-2-ene /lo-'* mol s-' cm-2 room temp. 323 52.0 27.0 12.0 9.0 31.5 573 35.6 36.4 16.7 11.3 3.80 room temp. 423 30.2 43.0 14.3 12.5 30.5 573 11.3 50.8 22.2 15.7 11.4 a Film thickness = 100 nm. diene over the Ru films prepared at different substrate temperatures are tabulated in table 1. The selectivity for but-1-ene is indeed higher and that for n-butane is lower over the Ru film prepared at 573 K than over the Ru film prepared at room temperature and the partial hydrogenation of diolefins proceeded preferably over the (002) face of Ru.On the other hand, the hydrogenation activity was decreased on the film prepared at high substrate temperature. The unevenness due to the columnar structure and the voids between the columnar grains are supposed to be reduced at high substrate temperature. The decrease in the activity is attributable to the reduction of the surface area of the film. Conclusion The microstructure of Ru films prepared by the r.f. sputtering method changes with film thickness (film growth). The alterations in the film structure affect the activity and selectivity in the hydrogenation of diolefins. It was found that the Ru(002) face is favourable for the partial hydrogenation of diolefins.Raising the reaction temperature, the selectivity was raised owing to the difference in the adsorption energy of hydrogen. The more (002) orientated Ru film was prepared at the high substrate temperature and exhibited higher selectivity for the partial hydrogenation. The surface electronic state of the Ru film was unchanged with film thickness. Therefore, the changes in the activity and the selectivity contributed to the microstructure of the Ru films. The r.f. sputtering method is a good method for the preparation of the well oriented crystallite films. New film catalysts will be developed by the effective use of these structure-controlled films. References 1 P. K. Dutta and H. Wilman, J. Phys. D, 1970, 3, 839. 2 P. K. Dutta and H. Wilman, J. Phys. D, 1971, 4, 1971. 3 K. K. Kakati and H. Wilman, J. Phys. D, 1973, 6, 1307. 4 N. K. Sandle and H. Wilman, J. Phys. D, 1973, 6, 1025. 5 S. J. Boichot, J. Phys. D, 1978, 11, 499. 6 S. J. Boichot, J. Phys. D, 1978, 11, 2553. 7 J. A. Thornton, J. Vuc. Sci. Technol., 1974, 11, 666. 8 T. Imanaka, J. Tamaki and S. Teranishi, Nippon Kuguku Kuishi, 1985, 1064. 9 J. Tamaki and T. Imanaka, Chem. Lett., 1986, 679. 10 ASTM Powder Diffraction Cards, 6-663. 11 G. C. Bond, Catalysis by Metals (Academic Press, New York, 1962), chap. 12. 12 T. E. Madey and D. Menzel, Jpn J. Appl. Phys., Suppl. 2, Part 2, 1974, 229.2180 Hydrogenation of DioleJins over Ru 13 P. D. Reed, C. M. Comrie and R. M. Lambert, Surf. Sci., 1976, 59, 33. 14 D. W. Goodman, T. E. Madey, M. Ono and J. T. Yates, J . Catal., 1977, 50, 279. 15 P. Feulner and D. Menzel, Surf. Sci., 1985, 154, 465. 16 M. A. Barteau, J. Q. Broughton and D. Menzel, Surf. Sci., 1983, 133, 443. 17 P. Cremaschi and J. L. Whitten, Surf. Sci., 1981, 112, 343. Paper 711628; Received 7th September, 1987

ISSN:0300-9599    

DOI:10.1039/F19888402173

出版商:RSC    

年代:1988

数据来源: RSC

摘要:

J . Chern. Soc., Faraday Trans. I, 1988, 84(6), 2181-2193 Solvation of Cyanoalkanes [CH,CN and (CH,),CCN] An Infrared and Nuclear Magnetic Resonance Study? Graham Eaton, Anthony S. Pena-Nuiiez and Martyn C. R. Symons* Department of Chemistry, The University, Leicester LEI 7RH The C E N stretching band (v,) has been studied for dilute solutions of CH,CN and (CH,),CCN in a range of aprotic and protic solvents. The former induce a low-frequency shift, whereas the latter induce a high- frequency shift relative to dilute solutions in hexane. In both cases there is a large increase in oscillator strength with increasing shift. This is the first example of a solvent that has an absorption band displaying such a dichotomy, the normal behaviour being a progressive low-frequency shift on going uia aprotic to protic media.In contrast such ‘normal’ behaviour is observed in the n.m.r. spectra for 14N shifts, but the I3C (CN) shifts are small and seem to be random. In contrast to our previous studies of ‘probe’ molecules the v 2 bands for solutions in water are almost identical to one of the bands in methanol ; however, the band in water is a single feature, whilst that in methanol is a doublet, the low-frequency feature being close to the unsolvated region. The interpretation is that MeCN in water is fully monosolvated (hydrogen-bonded), whilst in MeOH it is only ca. 50% monosolvated. However, the effects of temperature changes and studies of mixed water-aprotic solvent systems suggest that this may not be correct, and the possibility that MeCN forms two very weak hydrogen bonds in water is also considered.The methanol doublets are well defined at low temperatures ( - 50 “C) but resolution is lost on warming. At ca. 50 “C there is only one symmetrical band. For mixed water-aprotic solvent systems, the band remains a narrow singlet throughout the whole mole fraction range, there being no indication of twin bands for hydrogen- bonded and non-hydrogen-bonded units, in contrast with the results for methanol at low temperatures, and our normal experience with other probe molecules. One explanation is that there is rapid equilibrium between hydrogen- bonded and non-hydrogen-bonded units which is fast on the i.r. timescale. Results for other mixed-solvent systems are also reported. We have attempted to use changes in the first and second overtone 0-H stretch bands for HOD in D,O on adding MeCN to obtain a measure of the number of hydrogen bonds.This is at best only qualitative because of the proximity of the 0-H band for solvated MeCN and the band for water. However, for MeCN in Me,COH the and OH---(NCMe) bands are clearly resolved. The results support the concept that MeCN is dihydrated in water. A less extensive study has been made for 2-cyano-2- methylpropane (Me,CCN) for comparative purposes. The results are broadly similar, the major difference being a reduction in the extent of hydrogen bonding in methanolic solutions. We have recently shown that the combined use of i.r. and n.m.r. studies of dilute solutions of ‘probe’ molecules in pure and mixed solvents is a powerful method for studying the solvation of the probe, especially for estimating the number and strength ? Solvation Spectra Part 79.21812182 Solvat ion of Cyanoalkanes of the hydrogen bonds formed in mixed protic-aprotic solvents. The i.r. data give separate bands for each hydrogen-bonded solvate so that approximate compositions can be estimated. Generally there is a linear correlation between i.r. shifts and n.m.r. shifts for well defined (pure solvent) systems, and using this, the n.m.r. shifts for mixed solvent systems can be reproduced from the i.r. data. Details of these experiments are given in ref. (1)-(5) and the overall results have been summarised.6 Cyanomethane (MeCN) is an important solvent both for neutral compounds and for electrolytes.It is also widely used in its mixtures with water or methanol and has the advantage of being a good optical solvent. Whilst we knew of no systematic study of this solvent using spectroscopic techniques, there have been several i.r. and Raman studies of its complexation with metal ion^.^'^ In all cases there is a shift to high frequencies in the C r N stretching band (v2). It is assumed that bonding to the metal ions occurs via the N non-bonding (a) electrons (I), and in the particular case of the Ag(MeCN),f ion this linear coordination has been established by X-ray diffraction.' Studies of MeCN---H-X complexes in low-temperature matrices'' and in the gas- phase'l have also been unambiguously interpreted in terms of linear hydrogen-bonding via the Nn,b. orbital (11).H,C-C-N -+ M+ (I) H,C-C-N---H-X (11) Controversy has arisen regarding dilute solutions of MeCN in water.12*13 Using the OH overtone band (2vstretch) for HOD in D,O, Jolicoeur and his coworkers found an increase in intensity in the band often assigned to units (i.e. 0-H bonds which are not involved in hydrogen bonding). This they interpreted in terms of the ' structure- breaking' property of MeCN. In contrast, our view was that MeCN would not be soluble in water if it really caused the water structure to break, so we postulated that hydrogen bonding occurred, but, being very weak, it was difficult to distinguish between free OH units and such weakly bound ~ n i t s . ' ~ One aim of the present study was to attempt to substantiate this suggestion. Experimental 1.r.spectra were recorded on either a Perkin-Elmer 580 or 681 spectrophotometer coupled to a microcomputer for data analysis. Cells were demountable with calcium fluoride windows and pathlengths were determined by use of PTFE spacers (25 pm- 1 mm) as required. N.m.r. spectra were recorded on a Bruker WM-400 (courtesy of Warwick University) with TMS as reference for 13C spectra and DMF as reference for 14N spectra (externally locked and referenced in both cases). Solvents were purified by distillation from a suitable drying agent (usually calcium hydride) and were stored over molecular sieve or calcium hydride granules. Water was purified by a Milli-Q system. Samples were made up by volume using Eppendorf pipettes and Hamilton syringes. Results and Discussion I.R.Studies of Solutions of MeCN in Aprotic Solvents On going from inert solvents to dipolar aprotic solvents (including MeCN itself) there is a shift in v2(,,,) ( C r N ) to low frequencies, with a concomitant increase in band intensity (fig. 1). The shifts for the aprotic solvents correlate approximately with acceptor numbers (N,) for these solvents [fig. 2(a)], but there is no correlation for the protic solvents. This is in contrast with some of our other studies.6 There is an improved correlation between NA and intensity [fig. 2(b)], but two lines are still required. For mixed aprotic solvent studies, we obtained no evidence for the gain and loss of separate bands. Only smooth shifts occurred together with intensity gain as the moreG. Eaton, A . S.Pena-NuZez and M. C . R . Symons 2183 I l l 2270 I I I I I I I MeOH I 11 1 L CI LL I I- 2260 2250 Fig. 1. Changes in absorbance (q,,J and in v,(C=N) for MeCN in a range of pure protic and aprotic solvents. Since MeOH solutions exhibit two bands, the true absorbance for the high- frequency band would be greater if only this species were present. This is indicated by the dashed line. polar cosolvent was added. As in other studies, DMSO was the most perturbing solvent, probably because it has an exposed dipole. We envisage complexes of the type (111) as being involved, but stress that since no well defined bands were observed, this structure must be only a limit for a wide range of such species, with varying dimensions and shapes 6+ 6- H,CC G N in rapid equilibrium.There is evidence from X-ray scattering that such 'parallel' complexes are formed in pure MeCN at room temperat~re,'~ but since the shift for pure MeCN is less than that for DMSO solutions, this 'self'-interaction must be weaker. Clearly this interaction is one which involves the n-electrons of MeCN rather than the N(,n.b./o) electrons. It must result in an increase in the IC charge separation, and this must result in weaker n-bonding. However, the increased polarisation results in an in- crease in the intensity of the band. [We note that there is a weak band at ca. 2294 cm-l which has been assigned to the (v3 + v,) combination band. This gains in intensity as the v2 band shifts to high frequencies, probably as a result of this resonance. However, these changes are too small to account for the very large changes in v, observed in this study .]2184 r( 1 I 0 E 2260- 3 \ Solvation of Cyanoalkanes HEX 0 I CF3 1,CHOH 2270 t CF CH OH 3e MeOH H,.O 1 I 1 I 10 20 30 40 50 60 N , (solvent) CFJt / 5 6oot DMA NODMSO 0 I I I I I I 0 10 20 30 40 50 60 NA Fig.2. (a) v,(CrN) as a function of solvent acceptor numbers. The line a indicates an approximate correlation for solutions in aprotic solvents, but there is no unique curve for the protic solvents. (6) Absorbance for v 2 as a function of N , . I.R. Studies of Solutions of MeCN in Protic Solvents Solutions in methanol and ethanol exhibit two bands, but all other protic solvents used [H,O, CF,CH,OH and (CF,),CHOH] gave only one band. The doublet for methanol is well defined at low temperatures, but the bands merge into a broad singlet on heating (fig. 3).The low-frequency feature (2) is remarkably close to that for MeCN in inert solvents. This suggests that these MeCN molecules are surrounded by self-hydrogen- bonded methanol units which do not have any major influence on the -CN oscillator. The high-frequency feature (1) must then be assigned to singly hydrogen-bonded MeCN molecules. It is not easy to estimate the relative amounts of these two species since their absorbance varies drastically with frequency (fig. 1). Since the high-frequency band (1) occurs very close to the water band we can guess that they have about the same absorbance, in which case ca. 50 % is hydrogen-bonded. [On cooling, this is reduced and, given a considerable reduction in absorbance for band (2) relative to band (I), the 'free' form clearly dominates at low temperatures.] The single band for aqueous solutions shifts to low frequency on heating (fig.4). Curiously, the band narrows slightly in the 0-30 "C range, but broadens from then on.G. Eaton, A . S. Pena-NuEez and M . C. R. Symons 2185 2260 2260 2240 wavenumber/cm-' Fig. 3. Change in the v,(C-N) doublet for MeCN in MeOH with change in temperature ("C). Note the apparent formation of a single band at 50 "C. 2260 2259 3 I 5 1 az 2258 2257 0 20 40 60 80 TI" C Fig. 4. Effect of temperature on the band maximum (v,) for MeCN in water2186 Solvation of Cyanoalkanes I I 2 300 2250 2200 wavenum ber1cm-l ( b ) I 0 0.4 0.8 a mole fraction of DMSO Fig. 5. (a) 1.r.spectra for MeCN (v,) in MeOH-DMSO solutions, showing loss of the band for the MeOH---NCMe species and shift of the band by DMSO (25 OC): solution mole fraction DMSO 1 .oo 0.94 0.88 0.80 0.64 0.54 0.30 0.00 0.72 (b) Estimated changes in concentration for the two species detected in MeOH-DMSO solutions on the basis of equal concentrations of the two species in pure MeOH. Curve a gives [MeCN---HOMe], curve D gives [MeCN] not hydrogen-bonded.G. Eaton, A . S. Pena-NuiTez and M. C. R. Syrnons 2187 I I I I 2275 2250 222 5 wavenumber/cm-' Fig. 6. 1.r. spectra for MeCN (v,) in H,O-DMSO solutions. Note the lack of resolution into two bands at any stage (25 "C). solution mole fraction DMSO 1 0.00 2 0.03 3 0.15 4 0.20 5 0.37 6 0.50 7 0.70 8 1 .oo One possible explanation for the temperature insensitivity in this range is discussed below.As discussed above, mono-hydrogen bonding involves the formally non-bonding (n.b.) 'lone-pair' electrons on nitrogen rather than the '/I electrons. Again the charge separation is enhanced, hence there is a marked increase in intensity, but the high- frequency shift suggests that the C-N a-bonding is enhanced as a result of hydrogen bonding, just as in the case on complexation with metal ions. This increase in effective C=N bonding may reflect a decrease in antibonding character of the lone pair of electrons, or possibly a change in orbital hybridisation, resulting in more 2s character in the C-N a-bond with consequent increase in bond strength. The fact that the high-frequency band (1) for methanolic solutions is very close to the single band found for aqueous solutions is unexpected.In all our other studies bands for aqueous solutions are more strongly shifted than those for alcoholic solutions. Possible reasons for this difference have been discussed in our work on acetone.2 They all require that water forms more bonds to the 'probe' molecules than do the alcohols. Thus acetone forms two hydrogen bonds to water but only one to methanol. The obvious conclusion is that MeCN forms only one bond with water; however, as discussed below, this may not be correct. Both fluorinated alcohols [CF,CH,OH and (CF,),CHOH] induce large high- frequency shifts, which, relative to water, are far larger than expected from their N ,2188 i. 2260 2256 - I 5 1 a 2252 2 2 4 8 Sohat ion of Cyanoalkanes I \ Fig.7. Shift for v , , , ~ ~ for MeCN in H,O-DMSO solutions at 25 "C. values and in comparison with other probe [cf. fig. 2(a)]. Since these alcohols are stronger acids than water, enhanced shifts are expected, but we do not understand why they are so large. Cyanomethane is a relatively weak base, so there can be no question of actual proton transfer. One possible explanation stems from the opposed shifts observed for 71 and 0 interactions. We suggest that the degree of linearity makes an important contribution and that the stronger the hydrogen bond, the greater the precision in the linearity of the CEN---HA unit. For weaker bonds there may be a core of interactions around the -C-N axis, taking the hydrogen-bond away from the 0 orbitals and towards the n orbitals. This would reduce the high v shift from the 0 interaction and add a low v shift from the 71 interaction.[This concept is discussed below in connection with the mode of solvation by water.] Mixed Protic-Aprotic Solvents MeOH-D MSO There is a rapid loss of the band due to hydrogen-bonded MeCN as DMSO is added (fig. 5). This accords with the fact that DMSO is a far stronger base than MeCN and hence the reaction MeCN---HOMe + Me,SO % MeCN + Me,SO---HOMe (1) is strongly favoured in the forward direction. [The trends shown in fig. 5(6) are constructed on the assumption of equal concentration of free and bound MeCN in pure methanol at ca. 20 "C.] The band for non-hydrogen-bonded MeCN shifts steadily to low frequencies as [DMSO] increases, with no sign of any specific complex formation.The results show that Me,SO molecules are more effective than Me,SO---HOMe units at shifting the -C-N band.G. Eaton, A . S. Pena-Nuiiez and M. C . R. Symons 2189 0 0.4 0.8 mole fraction THF Fig. 8. ( a ) 1.r. spectra for MeCN (v,) in (CF,),CHOH-THF solutions at 25 "C. a: gives the value for MeCN---HOCH(CF,), units in CCl,. solution mole fraction THF 1 2 3 4 5 6 7 8 9 10 0.00 0.25 0.36 0.46 0.57 0.66 0.75 0.84 0.92 1 .oo (b) Changes in the relative concentrations of MeCN---HOCH(CF,), (a) and MeCN (p) (not hydrogen-bonded) in (CF,),CHOH-THF solvent systems.2190 Solvation of Cyanoalkanes H,O-DMSO The results for this system (fig. 6) differ markedly from those for the MeOH system (fig. 5). This system is unique in our experience since the band only broadens slightly as it shifts through the ‘zero’ (unsolvated) position, there being no indication of two bands as in the MeOH case.There is a small decrease in intensity as the band passes through the ‘free’ region (2256 cm-l), but the intensity at this position remains high compared with that in hexane. We suggest that this occurs because of a fast exchange between ‘free’ MeCN and MeCN bound to water. This accords with our explanation of the effect of an increase in temperature on the MeOH bands noted above (fig. 3). Since only a shift is observed, with no gain or loss of bands, we are unable to deduce the changes in concentrations of the species involved in the normal manner.1-6 Also, when the shift is plotted as a function of solvent mole fraction (ignoring the concentration of the probe), there is a steady shift to low frequency, with no region of insensitivity in the 1.0-0.8 mole fraction (H,O) region (fig.7), as might have been expected if two spectroscopically distinct species were involved. Fluor oalco ho 1 Sy s terns Because of the large shifts involved, mixed-solvent systems involving (CF,),CHOH as one component were studied in the hope of obtaining evidence for the number of hydrogen bonds formed to MeCN. The system (CF,),CHOH-CCl, gave two separate, well defined bands in the CC1,-rich range, the one at high frequencies (2267 cm-’) being assigned to the complex MeCN---HOR in a CCl, environment. The system (CF,),CHOH-tetrahydrofuran (THF) was more complicated, as can be seen from the spectra in fig.8. On going from pure THF the band due to non-hydrogen- bonded MeCN shifts towards a ‘free’ value as THF becomes hydrogen-bonded. In the 0.1-0.2 mole fraction alcohol region a new band centred on the CC1, value (ca. 2267 cm-l) grew in and was almost completely formed in the 0.5 mole fraction range. However, on going towards pure alcohol, the band continued to grow in intensity and to shift rapidly to the final value of 2276 cm-l for the pure alcohol. In this latter range the shift and intensity increases closely resemble the behaviour found for, say CC1,-DMSO systems, there being no sign of any gain or loss of bands. It therefore seems that this large change is not due to the formation of a disolvate, but is caused by secondary solvation as indicated in (IV).This type of bonding reinforces the bond to MeCN increasing its R I MeCN--- HO 0 \ \ \HO--- 1 R strength and causing the extra shift. We are surprised that it does not result in specific changes since only one bond is being reinforced. It is also surprising that such secondary solvation has only a minor effect for water and methanol, although it is surely also involved. Indeed, in our other studie~,l-~ quite large ‘ secondary solvation ’ shifts have been detected, although that for MeCN in (CF,),CHOH is exceptionally large. Because of the very large changes in intensity for the hydrogen-bonded species, it is not possible to monitor the concentration of this species very accurately as the mole fraction of THF is changed. However, this can be done for the ‘free’ band, and on this basis, using theG. Eaton, A .S. Pena-Nutiez and M. C. R. Symons 2191 14C h E, a v 130 - 0 *g 5 a" 120 s z 9 V \ 0 ( CFJ,CHOH I I I I I I 10 m 30 40 50 60 N A Fig. 9. Shift in the 14N resonance for CD,CN in a range of solvents as a function of N , of the solvents. assumption that only one equilibrium involving MeCN is involved, we have obtained an appproximate estimate of the overall changes in composition, as indicated in fig. 8 (b). 14N and I3C Resonance Shifts The 13C (13CN) and (13CH3) resonance shifts were small, and showed no well defined solvent dependence. In contrast, the 14N shifts were well defined, and gave a reasonable correlation with solvent NA (fig. 8). The two most notable aspects of fig. 8 are that the trend for aprotic solvents is in the same sense as that for protic solvents, in contrast with the i.r.shifts. Clearly the factors which govern these shifts are not the same, and there is no correlation between i.r. and n.m.r. shifts. This contrasts with the good n.m.r.-i.r. correlations noted for several other The other important result is that the 14N shift for water is greater than that for methanol. This arises because methanol is only partially hydrogen-bonded, as implied by the i.r. data, and the 14N shift is a fast time-average value. In the absence of a good correlation curve, we cannot use the results to obtain a precise estimate of the degree of solvation, but the shift is ca. half-way between the water shift and those for weakly interacting solvents, in qualitative agreement with our estimate of ca.50 % hydrogen bonding. Because we have not been able to obtain an i.r.-n.m.r. correlation plot, we have not undertaken mixed-solvent n.m.r. studies. Results for Me3CCN These were so similar to those for MeCN that we do not reproduce the majority of the data here. 1.r. data for pure solvents are included in table 1, and full details are given in2192 Solvation of Cyanoalkanes Table 1. 1.r. results [v,(max)] for MeCN and Me,CN in a range of solvents v(CrN) ~~ solvent MeCN Me,CCN cyclohexane tetrachloromethane cyanomethane tetrahydro fur an dimethyl sulphoxide methanol water trifluoroet hanol hexafluoroisopropanol 2256 2237 2255 2235.5 2254 - 2252 2234 2249 2233 2260/2254 2239/2245 2260 2238 2268 2252/2240 2276 225912240 ref.(16). For methanolic solutions the concentration of ‘free’ MeCN is enhanced, probably because there is some steric hindrance to hydrogen-bond formation, but, once formed, the bond is about as strong as that for MeCN, as judged from the shift from the hexane value. Solvation by Water Finally, we return to the question of the mode of solvation of MeCN by water. All but one of the present results point reasonably to mono-solvation [MeCN---HOH]. Only the effect of temperature on the bandwidth is difficult to understand on this basis, since heating is expected to result in a steady increase in width, caused by a general weakening of hydrogen bonds and a consequent increase in the spread of bond energies. What would be expected if MeCN, like Me,CO, were disolvated in water? Two limiting structures can be depicted [(V) and (VI)].The latter is expected by symmetry, 1 HOH / MeC N 0 \ \ \ HOH but its effect on the C-N stretch is likely to be similar to the former, which illustrates the factors involved more readily. In structure (V) one relatively strongly bound water molecule (a) will give rise to the normal high-frequency shift, but the other, weakly bound molecule, p, should induce a small low-frequency shift because of its effect on the 71 electrons. In the expected structure (VI), an average bond strength is expected and the overall shift should be the same. [Similar considerations are invoked above to explain the large shifts caused by the fluorinated alcohols.] Ideally, loss of one water molecule would cause a small shift to high frequencies.However, a careful study of the initial shift on adding DMSO to aqueous solutions revealed no high-frequency shift. It seems that if there are two water molecules, the two effects cancel each other.G. Eaton, A . S. Pena-Nuiiez and M . C. R. Symons 2193 A variety of other results have suggested that MeCN be dibasic in water. One, mentioned in the introduction, is that there is a strong growth of a band, indistinguishable from the 'OHfree' band for HOD in D20 in the 2v(OH) region on adding MeCN, which led Jolicoeur and coworkers to conclude that MeCN is a structure-breaker." The present results show unambiguously that MeCN is completely hydrogen-bonded in water, so this concept cannot be correct. Since MeOH shows no well defined 'OHfree' band in the 2v(OH) region17 we cannot readily compare water with methanol.However, t-butyl alcohol gives a well defined 'OHfree' band, and on the addition of MeCN, this band is reduced, with the concomitant growth of a well defined band shifted ca. 150 cm-I to low frequencies, which is clearly due to MeCN---HOCMe, units.16 These contrasting results suggest that the average hydrogen-bond strength for MeCN in water is less than that in Me,COH, which is most unexpected. When the v2 (CN) band is studied for MeCN in Me,COH, a broad band similar to that for MeOH solutions at 50 "C (fig. 1 ) was obtained. This implies the presence of both 'free' and hydrogen-bonded MeCN, as is the case for solutions in methanol. Thus, in this case, we can be reasonably sure that MeCN is only mono-hydrogen-bonded. One possible explanation is that water forms two hydrogen bonds, as in (VI), both being weaker than the single bond formed by the alcohols. The net effect on the C-N stretch would be similar to that for one bond.Other reasons for suspecting that water forms two bonds to MeCN include the fact that in some of our studies of other probes the spectral trends for mixed solvents on going from H,O to MeCN are similar in the water-rich region to those for other solvents such as Me2C0 or DMSO, which are known to form two hydrogen bonds.lP5 If MeCN is monobasic in water, a clear diflerence would be expected. We conclude that our present results, whilst in themselves suggesting that water forms only one hydrogen bond to MeCN, do not preclude the possibility that two very weak bonds are formed, at least at or below room temperature. References I M. C . R. Symons and G. Eaton, J . Chem. Soc., Faraday Trans. I , 1982, 78, 3033. 2 M. C. R. Symons and G. Eaton, J. Chem. Soc., Faraday Trans. I , 1985, 81, 1963 3 K. B. Patel, G. Eaton and M. C. R. Symons, J. Chem. Soc., Faraday Trans I, 1985, 81, 2775. 4 M. C . R. Symons and A. S. Pena-Nutiez, J . Chem. Soc., Faraday Trans. 1, 1985, 81, 2421. 5 G. Eaton, Ph.D. Thesis (Leicester University, 1983). 6 M. C. R. Symons, Pure Appl. Chem., 1986, 58, 1121. 7 K. F. Purcell and R. S. Drago, J . Am. Chem. Soc., 1966, 88, 919. 8 I . S. Perelygin, Opt. Spektrosk.. 1962, 13, 360. 9 K. Nilsson and A. Oskarsson, Acta Chem. Scand., Sect. A, 1984, A38, 79. 10 L. Andrews, J . Phys. Chem., 1984, 88, 2940. 1 1 B. A. Wofford, J. W. Bevan, W. B. Olson and W. J. Lafferty, J . Chem. Phys., 1985, 83, 6188. 12 P. Paguetter and C. Jolicoeur, J . Solution Chem., 1977, 6, 403. 13 M. C . R. Symons, J. M. Harvey and S. E. Jackson, J . Chem. Soc., Faraday Trans. I , 1980, 76, 256. 14 Von A. Kratochwill, J. W. Weidner and H. Zimmermann, Berichte. 1973, 77, 408. 15 Y. Koga, S. Kondo, S. Saeki and W. B. Person, J . Phys. Chem., 1984, 88, 3152. 16 A. S. Pena-Nutiez, Ph.D. Thesis (Leicester University, 1985). 17 H. L. Robinson and M. C . R. Symons, J . Chem. Soc., Faraday Trans. I , 1985, 81, 2131. 18 H. L. Robinson, Ph.D. Thesis (Leicester University, 1984). Paper 7/ 1663 ; Received 15th September, 1987

ISSN:0300-9599    

DOI:10.1039/F19888402181

出版商:RSC    

年代:1988

数据来源: RSC

摘要:

J . Chem. Soc., Faraday Trans. I, 1988, 84(6), 2195-2207 Spectroscopic Investigation of the Interaction of Co,(CO), with MgO and SO, Kanithi Mohana Rae,? Giuseppe Spoto, Eugenio Guglielminotti and Adriano Zecchina" Istituto di Chimica Fisica, Universita di Torino, Italy Co,(CO),, when absorbed in the gas phase, in vacuo, on fully dehydrated MgO, forms a variety of carbonyl clusters both neutral and ionic. These clusters disintegrate into monometallic species upon dosing with either CO or NH,, resulting in Co(C0); and Co2+(CO), (n = 2 or 3) with CO and Co(C0); and Co2+(NH,), with NH,, respectively. The negatively charged species are formed via several routes, including disproportionation and nucleophilic attack by 0,- on the carbonyl groups of different cobalt carbonyl clusters.These effects are attributed to the strongly basic nature of the highly dehydroxylated magnesia used in the present investigation. Diffuse reflectance and e x . spectral studies performed in similar conditions on magnesia confirm all these transformations. The results on a highly dehydrated SiO, surface are quite different : Co,(CO), is mostly adsorbed without appreciable chemical modification, giving two isomers containing linear and bridged CO. Only by removing CO by outgassing at the beam temperature, is the total transformation into Co,(CO),, and possibly Co,(CO) observed. Cluster chemistry is a rapidly developing field, and clusters have a valid contribution in providing plausible models of metal surfaces. Work on clusters is in its early stage, and Co,(CO), is one of the few metal carbonyl clusters used in industrial catalysis because it readily gives catalytic intermediates, i.e.coordinatively unsaturated cobalt carbonyl derivatives. lP3 The interaction of Co,(CO), in vucuo with silica (Cabosil), outgassed at 673 K, was studied by Schneider et al. :4 the immediate formation of CO,(CO),, by surface-catalysed decarbonylation of Co,(CO), was reported. The fully hydrated SiO, surface contains ca. 500H per 100 Dehydration at 673 K causes the elimination of only 2 OH per 100 A,. Consequently, as ca. 60% of the surface hydroxyls are still present on the surface outgassed at 673 K,5 the observed catalysed decarbonylation could be associated with the presence of silanols (as hypothesized by the same author^).^ Gopal and Watters' have studied the interaction of Co,(CO), with MgO: they concluded that the Co,(CO), undergoes a series of transformations with the formation of coordinatively saturated and unsaturated compounds.The magnesia which they used had been treated at 673 K, and therefore a non-negligible number of OH groups (20% of the monolayer capacity) was still present on the surface.' Iwasawa et a1.8 supported Co,(CO), on silica by a dry-mixing method and observed the formation of an abundant amount of [Co(CO),], surface species. The formation of these oxidized species could be due to the preparation procedure adopted (the dry mixing was apparently carried out in air). In order to separate the role of adsorbed water from that of the free surface in determining the surface transformations of adsorbed carbonyls, Guglielminotti et al.' made a preliminary investigation on the adsorption of Co,(CO), on totally dehydrated MgO.In the present study we report a more extensive investigation of the interaction t On leave from Indian Institute of Technology, Bombay, India. 72 2195 FAR I2196 Interaction of Co,(CO), with MgO and SiO, of CO,(CO),, in vacuu, with MgO outgassed at 1073 K and at Torr (1 Torr = 133 N m-,), i.e. characterized by a fully dehydroxylated surface (as monitored by the total absence of OH groups) and with a surface area of 180 m2 g-l.g The effect of CO and NH, over the resulting species was also investigated. Diffuse reflectance spectra (d.s.r.) and electron spin resonance (e.s.r.) studies were also performed along the similar lines to provide supporting evidence for the i.r.assignments. In addition, the adsorption of Co,(CO),, in vacuo, on silica which had been outgassed at 1100 K (the highly dehydroxylated form) and the effect of pumping over the species formed in this way are reported. Experimental The MgO was obtained by vacuum decomposition of Mg(OH), following a procedure described elsewhereg and the surface was then cleaned of adsorbed impurities at 1073 K and at Torr for a few hours. The specific surface area of the sample treated in this way was typically 180 m2 g-l. Silica (Aerosil, Degussa, specific surface area 370 m2 g-') was outgassed at 1100 K for soeveral hours in order to remove most of the hydroxyl groups (only ca. 1 OH per 100 A' was left on the ~urface).~ The CO and NH, used in this experiment were of high-purity grade (Matheson).Co,(CO), was obtained from ICN Labs. Infrared studies were carried out on a PE 580B IR instrument equipped with a data station, using a low-path (0.7 cm) i.r. cell and thin (0.01-0.02 cm) wafers of silica and magnesia, containing ca. 15-20 mg cm-, of powder. The diffuse reflectance spectra were recorded from 200 to 1900nm on a Varian Cary 2390 instrument, equipped with a diffuse reflectance attachment and a data station. E.s.r. studies were performed on a Varian El09 instrument operating in the X-band mode (9.4 GHz). Varian pitch ( g = 2.0029) was used as a reference. The adsorption of Co,(CO), from the gas phase on fully dehydrated MgO is a highly irreversible process with a high sticking probability.Consequently the Co,(CO), vapour is adsorbed immediately on the MgO pellet with the formation of reddish-brown penetration front which moves progressively from the external to the internal part of the pellet. The front divides the white cobalt-free part of the pellet from the coloured one, where the coverage is constant and near or equal to the maximum value. As already shown for the systems" where this phenomenon occurs, the i.r. spectra do not change qualitatively with the amount adsorbed, only the intensity being effected by the change in coverage. The same phenomenon does not occur (or it occurs to much lower extent) on silica. Results and Discussion Transformation of Co,(CO), on the SiO, Surface Fig. l(a) shows the i.r. spectra obtained when Co,(CO), is dosed on SiO, in a vacuum through gas-phase sublimation (the different curves correspond to increasing doses).The first small dose gives a doublet at 2080-2065 cm-l and a less intense but broader peak at 1870 cm-l. Successive doses cause the formation of several bands in the range 2100-2000 cm-l(2032,2045,2057 and 2072 cm-l). Besides these bands, a broad peak in the region of 1860 cm-' grows in parallel with a shoulder at 1970 cm-'. By comparing these bands with the data shown in schsme 1, we can conclude that the first dose (corresponding to ca. 0.1 mol per 100 A,; vide infra) is totally transformed into Co,(CO),, by surface decarbonylation (in fact the doublet is readily assigned to the two tetrameric species adsorbed on slightly different sites ; vide infra).The successive doses give both the bridged and linear Co,(CO), species (the last one monitored by the peakK. Mohana Rao, G. Spoto, E. Guglielminotti and A . Zecchina 2197 t wavenumberlcm-' wavenum ber/cm-' Fig. 1. (a) Infrared spectra (220&1700 cm-l) of increasing doses of Co,(CO), adsorbed through gas-phase sublimation on silica (Aerosil : specific surface area 370 m2 g-') previously outgassed under high vacuum Torr] at 1073 K. The most intense spectrum corresponds to 0.5 molecules of Co,(CO), per 100 A2. (b) Effect of 15 min outgassing, at the i.r. beam temperature and under high dynamic vacuum on the i.r. spectrum of adsorbed Co,(CO),. at 2032cm-l). When these species are subjected to evacuation at the infrared beam temperature, the spectrum is modified according to fig.l(6). The bands at 2032 and 2045 cm-' and the shoulder at 1970 cm-' disappear and new bands appear at 2 120,2 108, 2080,2065 and 1870 cm-l. [The peak at 1870 cm-l is slightly shifted towards the positive side with respect to that in fig. 1 (a) and is increased in intensity.] The presence of two doublets at 2120-2100 and 2080-2065 cm-l together with the broad peak at 1870 cm-l indicates that two forms of CO,(CO),, adsorbed on slightly different sites are becoming the predominant species present on the surface. From the known extinction coefficient of the bridged carbonylic species15 of Co,(CO),, it is possible to conclude that the Co,(CO),, coverage (broken curve) is still very low (ca. 0.5 molecule per 100 A2). The presence of a small amount of Co,(CO),, cannot be excluded : in fact, upon evacuation a weak shoulder at 1780 cm-l grows which can be assigned to Co, clusters.We can describe the effect of evacuation in the following way: Co,(CO),(ads) -+ CO,(CO),, -+ CO,(CO),,. (1) Schneider et aL4 reported that a fast and complete decarbonylation process occurs in vacuo when Co2(CO), i: dosed on SiO, outgassed at 673 K. In our case only the first dose (0.1 molecule per 100 A2) is quickly decarbonylated. Extensive decarbonylation occurs on subsequent doses only by prolonged outgassing at the i.r. beam temperature. We think that the smaller decarbonylating activity observed in our case is closely related to the lower concentration of residual OH groups which, as proposed by Schneider et aZ.,, are acting as catalytic centres.A detailed examination of the data illustrated in scheme 1 and fig. l ( a ) and (b) 12-22198 Interaction of co,(Co), with MgO and sio, 1 CO/Co-SiO, CO/Co-MgO l 3 coZic0)8 (linear) (bridged) " CO~(CO)~ I I , . . 2 m rn 1800 1600 wavenum ber/ cm Scheme 1. indicates that for clustered cobalt carbonylic compounds (both in solution or in the adsorbed state) and for dispersed cobalt particles, the Following conclusions can be drawn: (I) the linear carbonyls absorb in the 2120-2000 cm-' range; (2) the bridged carbonyls absorb in the 1900-1 800 cm-l range; (3) in the 2000-1900 cm-l interval, no substantial absorption occurs. Observations (1)-(3) are well known and fairly obvious : however, they must be borne in mind because they will be useful in discussing the results of the Co,(CO),-MgO system.Transformation of Co,(CO), on MgO In fig. 2 the i.r. spectra of increasing doses of Co,(CO), on MgO, in vacuum, i.e. through gas-phase sublimation, are illustrated. The spectrum is very complex : however, the relative intensity of the components does not change much with coverage, as expected for a highly irreversible adsorption process. Bands and shoulders are observed at: 21 12(vw), 2090(sh), 2070(sh), 2038(vs), 1960(sh), 1935(sh), 1870(s), 1820(m), 1760(w), 1675(m), 161 5(m), 1520(br, w), 1380(w), 1320(m), 1280(w), 1240(w), 1 180(vw), 980(vbr), 930(w) and 850(w) cm-l. With respect to the spectra of Co,(CO), adsorbed on SiO,, the most important differences are : (1) the presence of non-negligible absorptions in the region 2000-1900 cm-l [a region which is totally clear for the Si0,-Co,(CO), system]; (2) the presence of a strong and broad absorption tail at frequencies as low as 1700 cm-l (well documented in the strongest i.r. spectra of fig.2) which is totally absent on SiO,; (3) the presence of discrete bands at frequencies lower than 1700 cm-l (1675, 161 5, 1520, 1380, 1320, 1280, 1240, 1180, 980, 930 and 850 cm-l) which can be partially ascribed to carbonate-like species. Without entering into details of the assignments, we can briefly say that, as for the SiO,/Co,(CO), system, many of the bands observed in the 2120-1800 cm-l interval (particularly the strong and complex peak centred at 2038 cm-') can be explained on the basis of the stretching frequencies of a variety of neutral carbonyl clusters [Co,(CO),,K.Mohana Rao, G. Spoto, E. Guglielminotti and A . Zecchina 2199 Co,(CO),,, CO~(CO),~ etc.] and even of CO adsorbed on small metallic cobalt particles (see scheme 1). This fact suggests that an extensive release of CO is immediately occurring upon contact of Co,(CO), with the totally dehydrated MgO surface, with subsequent clustering, following reaction (1). This conclusion is proved by simultaneous measurements of the pressure, which show that a gaseous product is immediately released. CO abstraction in this case is favoured by the capacity of highly dehydrated MgO to adsorb CO strongly to give complex dimeric and polymeric structures described ear1ier.l' A weak shoulder at 2100cm-l and a peak at 1380cm-' coincide with the vibrational modes of the dimeric structures observed on the CO-MgO system." I I I co,(co):; l7 u co,(co);; l2 U PhCCo,(CO); A I.L '... ..,_,,p cO(c0); (distorted)l8' l9 . . , . . . . . . . . . CO(C0)'i ~ . . . (Td)I73 l9 I I 2203 ' xxx> 1803 1833 w avenumberlcm-' Scheme 2. However, the large spectral differences described in points (1)-(3) show that, unlike the Co,(CO),-SiO, system, negatively charged entities like those illustrated in scheme 2 must also be taken into consideration. In fact, negatively charged carbonyl compounds show strong bands in the 2000-1900 and 1700-1600 cm-' regions. In the inset of fig. 2, the u.v.-visible-near-i.r. reflectance spectrum of Co,(CO), (measured at maximum coverage) absorbed on MgO under the same conditions described before is reported.Several broad absorptions are observed at 37 300, 26 600, 18600 and 14800(sh) cm-', and when these bands are compared with the data shown in scheme 3 (where the spectra of several neutral and negatively charged carbonylic complexes are reported) the following conclusions can be deduced: (1) the complex adsorption at frequencies 225000 cm-' can be explained on the basis of a mixture of neutral carbonylic clusters [Co,(CO),, Co,(CO),, and co6(co)16], Con polymers and small cobalt particles; (2) the band at 19000 cm-' and the tail extending to lower frequencies find a closer analogy among the polynuclear carbonylic anionic species. It is evident that both the i.r. and reflectance experiments lead to the same (general) conclusion : the absorption of Co,(CO),, on MgO, in vacuo, occurs with loss of CO with the formation of both larger neutral clusters (and possibly metallic particles) and negatively charged carbonylic species.This conclusion totally differentiates the surface chemistry of highly dehydrated SiO, and MgO towards the adsorption of Co,(CO),. The formation of negatively charged cobalt carbonylic species from Co,(CO), and other neutral clusters is well known in homogeneous cherni~try.,~ It is useful to summarize briefly the main mechanisms leading to the formation of mononuclear and polynuclear negative species under homogeneous conditions.2200 Interaction of Co,(CO), with MgO and SiO, I \ wavenumberlcm - l Fig. 2. Infrared spectrum (2000-900 cm-') of increasing doses of Co,(CO), adsorbed through gas- phase sublimation on MgO (180 m2 g-l) fully dehydroxylated at 1073 K ui;lder high vacuum Torr).The most intense spectrum corresponds to cu. 60 x lo-, per 100 A2 cobalt atoms as determined via the spectroscopic method described in the text. Mon onuclear Negative Species [ Co (CO) 4 '1 (1) Reaction of cobalt carbonyls with a Lewis base B: 3Co2(CO), + 12B -+ 2(CoB6),+ + 4Co(CO), + 2CO 3c04(co),2 + 24B -+ 4(C0B6)2+ + 8cO(co), + 4CO (2 a) (2 b) (and equivalent reactions involving higher nuclearity clusters). These reactions occur with a release of CO and they involve disproportionation of zerovalent clusters with the simultaneous formation of both Co2+ and Co-. The stoichiometry of the positively charged complexes could be different with respect to what has been stated previously: in the presence of CO and an insufficient number of B molecules, Co2+ complexes with formulae Co(CO),B~ could also be formed.(2) Reaction with a Brernsted base: 8Co,(CO), + 8CO + 320H- + 16Co(CO), + 8CO;- + 16H,O. (3) According to this equation Co,(CO), is transformed into Co(C0)i and COi- (only in the presence of CO); in other words Coo is reduced to Co- while CO is oxidized to CO:-. (3) Reaction with a Brarnsted base and disproportion : 1 lCo,(CO), + 320H- -+ 2C02+ + 20Co(CO), + KO:- + 16H20. (4)K. Mohana Rao, G. Spoto, E. Guglielrninotti and A . Zecchina 220 1 - cO(c0); 21 - Fe(C0)i2 2 1 -- Con matrixz2 L -coZ-co, matrix 22 --. Co2(CO), [P(OMe3], 23 A -- COZ(C0)8 1 1 , 2 3 wavenumber/ 1 0-3 cm-’ Scheme 3.This reaction can be considered as a mixture of reactions (2) and (3) and is not accompanied by the release or consumption of CO. However, in the presence of a CO atmosphere, the Co2+ ions could form Co2+(CO), complexes. Polyn uclear Nega t iue Species Reaction between Co,(CO), and Co(C0)i : ~co,(co), + 2Co(CO); -+ Co,(CO);; + 9CO This reaction occurs with release of CO. Similar reactions can be hypothesized between Co(C0)i and higher cobalt carbonyl clusters formed via reaction (1). It is most noticeable that the direction of these relations can be reversed easily by increasing the CO pressure. ‘The fully dehydroxylated MgO surface contains a large variety of surface 02- species in different coordination states (five-fold, four-fold or three-fold coordinated).26 These 0,- species possess basic properties and can act as Lewis bases or behave like OH- anions. If we consider the heterogeneous analogues for the reactions (2)-(4), the possible equations are as follows: 3cO,(Co), + 20,- --+ 2Co2+O2- + 4co(cO), + 8CO 3Co,(CO),, + 402- -+ 4C02+02- + 8Co(CO); + 4CO (2 a’) (2 b’) (and equivalent reactions involving higher-nuclearity clusters). The reactivity of surface 0,- is in the order: five-fold < four-fold < three-fold.l6. 26* 27 The four-fold and three-fold coordinated 02- ions are formed by dehydration at temperatures > 773 K.l6, 26,27 Consequently, we think that the higher complexity of the i.r. spectra of Co,(CO), adsorbed on MgO dehydrated at 1073 K in respect to that dehydrated at 673 K6 can be mainly ascribed to the more abundant participation of low- coordinated oxygen ions in CO abstraction and disproportionation.In presence of CO2202 Interaction of Co,(CO), with MgO and SiO, gas (CO),Co2+02- complexes could be formed. However, the stoichiometry of the surface positive complexes, (CO),Co2+02-, cannot be predicted a priori because the number and nature of the ligands depend very much upon the CO pressure and upon the location of the Co2+ ions on the surface [e.g. on the surface each Co2+ ion is in contact with more than one 0,-, in contrast to the scheme shown in reaction (2a')l. The surface analogues of reactions (3) and (4) involving 02- are as follows: 8C02(CO), + 8CO + 1602- 16Co(CO), + 8CO;- (3') (4') 1 1 Co,(CO), + 1 60,- -+ 2C02+ + 2OCo(CO), + KO:-.[In the presence of CO gas Co2+(CO), complexes could be formed.] Both the reactions (3') and (49, if occurring at the surface, lead to the formation of C0:- adsorbed groups. As we have definite proof of the formation of a small amount of such species in our present experiment, we infer that the reactions (2') and/or (4') must also have occurred on the MgO surface. It is evident that if reactions (2')-(4') are operating simultaneously, the incoming Co,(CO), can react with preformed Co(CO),(ads) to give a variety of negatively charged carbonyl clusters. All the previous considerations suggest that the adsorption of Co,(CO), on fully dehydrated MgO leads to a variety of species both neutral and negatively charged, and all these species can be partially transformed into one other by changing the external conditions (e.g.the pressure of CO). An overall scheme is shown in scheme 4. Scheme 4. According to Indovina et a1.,28 the Co2+(CO), complexes should be e.s.r.-active. Furthermore, it is not excluded that as a result of the interaction of Co(C0); with neutral clusters, other paramagnetic structures of the type Co,(CO); can also be formed.', In conclusion, the adsorption of Co,(CO), in vacuo on MgO could lead to the formation of paramagnetic species with a subsequent appearance of e.s.r. activity. To observe these species we have performed e.s.r. experiments, keeping all the experimental conditions as described earlier. A very broad e.s.r. signal centred at g = 2.006 is detected. The signal breadth is probably associated with spin-spin interactions occurring in the reddish-brown adsorbed layer, which (for the reasons already discussed in the experimental section) is always highly concentrated.Unfortunately the signal breadth prevents a detailed interpretation of the spectrum. However, it must be stressed that both the CO~+(CO),~~ and Co,(CO);'' species or a mixture of the two are consistent with the observed spectrum. Scheme 4 suggests that the relative proportion of low- and high-nuclearity neutral andK. Mohana Rao, G. Spoto, E. Guglielminotti and A . Zecchina 2203 1 I rn 1800 1600 1400 1 2 0 Id00 wavenumberlcm-' Fig. 3. Infrared spectra (2000-900 cm-I) of increasing coverages of Co,(CO), adsorbed through gas-phase sublimation in presence of 40 Torr of CO, on MgO (180 m2 g-l) fully dehydroxylated at 1073 K under high vacuum Torr).charged species should be governed by the amount of adsorbed Co,(CO), and by the CO pressure. In particular, high CO pressures should favour low nuclearity complexes like Co,(CO),(ads) and Co(CO),, while in vacuo the clustered systems should be predominant (in agreement with the results illustrated in fig. 1). The reliability of the network of reactions illustrated in scheme 4 for describing the interaction of Co,(CO), with MgO is further reinforced by the results shown in fig. 3 and 4. In fig. 3 the i.r. spectra of Co,(CO), adsorbed at increasing coverages in the presence of 40 Torrf of CO are shown. These spectra are very different from those reported in fig. 2. The first spectrum is simply that of CO adsorbed initially on Mg0.l' The other spectra are mainly due to the superimposition of two families of carbonyls, Co2+(CO), [or less probably Co2+(CO),] and Co(C0); (see fig. 3 and 4).As demonstrated in previous papers,19 the spectra of Co(C0)i absorbed at the surface of MgO (and hence with a symmetry lower than G) consists of a quartet of bands in the 2100-1800 cm-' range. While the first peak [deriving from the i.r.-inactive mode of the free Co(C0)i molecule in G symmetry] is weak and fairly fixed in frequency (2030 cm-l), the remaining three modes [coming from the triply degenerate mode of the free Co(C0); ion by a splitting caused by the interaction with the surface] are strong and variable in the 2000-1 800 cm-l range. (Their separation depends upon the distortion introduced by the 7 1 Torr = 101 325/760 Pa.2204 Interaction of Co,(CO), with MgO and SiO, c .J' wavenumbed cm-' r . c0;- L 2000 1800 1600 1400 7200 I d 0 0 w avenumber/cm -' Fig. 4. Effect of dosing CO (40 Torr) (broken line) on Co,(CO), preadsorbed on MgO through gas-phase sublimation in vacuum (full line). interaction with the surface.) The spectra of Co(CO)i+ on an MgO-Coo solid consists of a doublet of peaks in the 2100-2040 cm-l range, with the low- frequency component (corresponding to the degenerate asymmetric stretching mode of the three carbonyls) being three times more intense than the high-frequency one. On this basis we conclude that the two peaks at 2097 and 2040cm-l are associated with Co(CO):+ species, while the peaks at 1975, 1862 and 1818 cm-l are due to the Co(C0); entity (as shown in fig.3 and 4). The i.r. spectra of these species obscure those of the residual neutral clusters. The shoulders at 1950 and 1780 cm-' could be due to the residual clustered anionic compounds. Also in this case the presence of the COi- species is well documented. In the inset of fig. 3 the reflectance spectra in the u.v.-visible-near-i.r. range corresponding to the first and last spectra are shown. From these spectra the following comments can be made: (1) the first spectrum is typical of CO adsorbed on MgO and does not require further comment because it has been discussed in detail in an earlier paper;27 (2) the difference between the spectrum shown in the inset of fig. 2 and the strongest spectrum of the inset of fig.3 (which corresponds to the most intense i.r. spectrum) can be explained in terms of different distributions of neutral and negatively charged carbonylic species. In presence of CO, lower-nuclearity species (in agreement with i.r. assignments) are predominant; consequently the absorbance in the lower- frequency range of the reflectance spectrum is definitely smaller. From all these experiments the following conclusions are drawn: in the presence of CO the formation of high-nuclearity clusters (both neutral and negatively charged) isK. Mohana Rao, G. Spoto, E. Guglielminotti and A . Zecchina 2205 0.6- - 0 5 2 04- 0, -s 0.2- wavenumb er/ cm -' Fig. 5. Effect of dosing NH, on Co,(CO),.preadsorbed on MgO through gas-phase sublimation. (---) Initial spectrum; (-----) after dosing with 2 Torr NH,; (---): after dosing with 6 Torr of NH,.depressed, and the primary disproportionation reaction leading to Co3+(CO), and Co(C0)i mononuclear products is dominating. The role of CO in the gas phase, to shift the equilibria towards the low-nuclearity compounds as described in scheme 4, is further proved by the experiment shown in fig. 4, where the effect of dosing CO on the i.r. and reflectance spectra (inset) of Co,(CO), absorbed in vacuo is illustrated. The i.r. spectra of the Co(C0)l species grow at the expense of the absorption centred at 2000-2100 cm-' [associated with terminal CO of adsorbed Co,(CO),, Co,(CO),, and other clusters]. Meanwhile, in the reflectance spectra the low-frequency absorption associated with high-nuclearity clusters (both neutral and negatively charged) is heavily eroded.In the inset the dotted broken curve shows the difference spectrum. Upon CO contact, two main bands, one centred at 25000 cm-l and the second at 15000 cm-', are eroded. On the basis of scheme 4, we believe that the disappearance of the two peaks corresponds to the destruction of zerovalent and negatively charged clusters, respectively. Transformation of Co,( CO), Adsorption Products in the Presence of Coadsorbed Base If, on vacuum adsorbed Co,(CO),, NH, is dosed at two different pressures (2 and 6 Torr), the spectrum in the range 2150-1700 cm-' changes abruptly as shown in fig. 5. (NH,)2206 Interaction of CO,(CO), with MgO and SiO, The absorption at 2100-2000 cm-l disappears and a new (very complex) one appears at lower frequency (2000-1 800 cm-l).NH, is a Lewis base. Therefore reaction (2) should readily occur between adsorbed neutral carbonyls [Co,(CO),(ads) etc.] and adsorbed NH,, leading to the extensive formation of Co(C0); and Co(NH,)i+. The main peak (which is generally associated with terminal CO in zerovalent clusters) totally disappears, while a quartet of new bands in the 2030-1 750 cm-l interval clearly associated with Co(C0)L is produced. When the NH, pressure is increased, the three low-frequency bands coalesce into a single absorption at 1900 cm-l typical of Co(C0); in a nearly tetrahedral ~ituati0n.l~ (This is because the Co2+ and Mg2+ ions are completely solvated by NH,, and consequently tight surface cation-anion pairs are not formed in these conditions.19) This experiment clearly shows that the same type of reaction, well known in homogeneous conditions, also occurs on the surface via the intervention of extra bases coming from the gas phase. symmetry in solvents with high solvation power is known,30 we can infer via a purely spectroscopic method that the number of Co(C0); groups (as%umed to be uniformly distributed on the whole MgO surface) is ca. 15 x lop2 per 100 A2. As a consequence we can also deduce that the total amovnt of Co atoms responsible for the full-line spectrum of fig. 5 is ca. 21 x per 100 A2. In fact the concentration in the sample is not uniform (for the reasons briefly summarized in the experimental section), and consequently in the external layers it is certainly larger.However, an accurate figure for the dependence of the concentration upon the distance of the pellet from the surface cannot be given. As the extinction coefficient of the Co(C0); anion with This research was carried out with financial support from the Minister0 Pubblica Istruzione, Progetti di rilevante interesse nazionale (Italy). M. R. thanks the ‘Third World Academy of Sciences’, ICTP, Trieste, Italy for awarding a grant for this research. References 1 (a) A. J. Clark and J. F. Harrod, Adv. Organomet. Chem., 1968,6, 119; (b) R. Whesman, J. Organomet. 2 K. Muruta, A. Mutasuda, T. Masuda, M. Ishino and M. Tamura, Bull. Chem. SOC. Jpn, 1987, 60, 3 T. Masuda, K. Murata, T. Kobayashi and K. Matsuda, Nippan Kagagu Kaishi, 1986, 171. 4 R.L. Schneider, R. F. Howe and K. L. Watters, J. Inorg. Chem., 1984, 23, 4593; 4600. 5 E. Borello, A. Zecchina, C. Morterra and G. Ghiotti, J. Phys. Chem., 1967, 71, 2945. 6 P. G. Gopal and K. L. Watters, 8th Int. Congr. Catal., Berlin (Verlag Chemie, Weinheim, 1984), 7 P. J. Anderson, R. F. Horlock and J. F. Oliver, Trans. Faraday SOC., 1965, 61, 2754. 8 Y. Iwasawa, M. Yamada, Y. Sato and H. Kuroda, J. Mol. Catal., 1984, 23, 95. 9 (a) E. Guglielminotti, A. Zecchina, F. Boccuzzi, E. Borello, in Growth and Properties of Metal Clusters, ed. J. Bourdon (Elsevier, Amsterdam, 1980), p. 165; (b) A. Zecchina and G. Spoto, J. Catal., 1985,%, 586. Chem., 1974, 81, 97. 438. vol. 5, p. 75. 10 E. Guglielminotti, G. Spoto and A. Zecchina, Surf. Sci., 1985, 161, 202. 11 (a) G.Bor, Spectrochim. Acta, 1963,19, 1209; (b) G. Longoni, S. Campanella, A. Ceriotti and P. Chini, J . Chem. SOC., Dalton Trans., 1980, 1816; (c) R. A. Friedel, 1. Wender, S. L. Shufler and H. W. Stenberg, J . Am. Chem. Soc., 1955, 77, 3951. 12 P. Chini, J. Chem. SOC., Chem. Commun., 1967, 440. 13 (a) N. Sheppard and T. T. Nguyen, in Advances in Infrared and Raman Spectroscopy, ed. R. J. H. Clark and R. E. Nester (Heyden, London, 1978), vol. 5, chap. 2, p. 67; (b) K. Mohana Rao and A. Zecchina, unpublished. 14 K. Sato, Y. Inove, I. Kojima and E. Miyajaki, J. Chem. SOC., Faraday Trans. 1, 1984, 80, 841. 15 R. M. Wing and D. C. Crocker, Inorg. Chem., 1967, 6, 289. 16 E. Guglielminotti, S. Coluccia, E. Garrone, L. Cerruti and A. Zecchina, J . Chem. SOC., Faraday 17 P. Chini and V. Albano, J. Organomet. Chem., 1968, 15, 433. Trans. 1, 1979, 75, 96.K , Mohana Rao, G. Spoto, E. Guglielminotti and A . Zecchina 2207 18 B. M. Peake, B. H. Robinson, J. Simson and D. J. Watson, Inorg. Chem., 1977, 16, 405. 19 (a) A. Zecchina, G. Spoto and E. Garrone, J . Phys. Chem., 1984, 88, 2587; (b) P. S. Braterman, Metal Carbonyl Spectra (Academic Press, London, 1975) and references therein; (c) W. F. Edgell and J. Lyford IV, J. Chem. Phys., 1970, 52, 4329; ( d ) W . F . Edgell, M. T. Yong and N. Koinzumi, J. Am. Chem. SOC., 1965, 87, 2563. 20 G . L. Geofferey and R. A. Epstein, Inorg. Chem., 1977, 16, 2795. 21 (a) H. Heiber and E. H. Schubert, Z. Anorg. Allg. Chem., 1965,338,32; (b) F . Calderazzo and R. Ercoli, Chzm. Ind. (Milan), 1962, 44, 990; (c) R. C. Dunbar and B. B. Hutchinson, J . Am. Chem. SOC., 1974, 96, 3816. 22 G. A. Ozin and A. J. Leehanlan, Inorg. Chem., 1979, 18, 1781. 23 H . B. Abrahamaen et al., Znorg. Chem., 1977, 16, 1554. 24 L. A. Hanlan, H. Huber, E. P. Kundig, B. R. Mcganvey and G. A. Ozin, J . Am. Chem. Soc., 1975,97, 25 I. Wender and P. Pino, Organic Synthesis via Metal Carbonyls (Wiley, New York, 1968), vol. I, 26 A. Zecchina, M. G . Lofthhouse and F. S . Stone, J. Chem. SOC., Faraday Trans. 1, 1975, 71, 1476. 27 A. Zecchina and F. S . Stone, J . Chem. Soc., Faraday Trans. 1, 1974, 74, 2278. 28 V. Indovina, D. Cordischi and M. Occhiuzzi, J . Chem. Soc., Faraday Trans. I , 1981, 77, 81 1. 29 A. Zecchina, G. Spoto, E. Borello and E. Giamello, J . Phys. Chem., 1984, 88, 2582. 30 S. F. A. Kettle and I. Paul, A h . Organomet. Chem., 1972, 10, 199. 7054. p. 63. Paper 7/1689; Received 21st September, 1987

ISSN:0300-9599    

DOI:10.1039/F19888402195

出版商:RSC    

年代:1988

数据来源: RSC

摘要:

J . Chem. SOC., Faraday Trans. 1, 1988, 84(6), 2209-2214 Pulse Radiolysis Study of Salt Effects on Reactions of Aromatic Radical Cations with Cl- Part 2.-Spectral Shifts and Decay Kinetics of Diphenylpolyene Radical Cations in the Presence of Tetrabutylammonium Hexafluorophosphate Yukio Y amamoto," Takahisa Aoyama and Koichiro Hayashi The Institute of ScientiJic and Industrial Research, Osaka University, 8-1 Mihogaoka, Ibaraki, Osaka 567, Japan Pulse radiolysis of all-trans a,o-diphenyl-substituted polyenes, such as 1,4-diphenylbuta- 1,3-diene, 1,6-diphenylhexa- I ,3,5-triene and I ,&diphenyl- octa- 1,3,5,7-tetraene, in I72-dichloroethane solution has been undertaken in the absence and presence of Bu,NPF,. The absorption maxima of the diphenylpolyene radical cations are shifted to shorter wavelengths in the presence of the salt.This is evidence for the formation of the contact ion pairs between the radical cations and PF;. The decays of the radical cations, which are due to neutralization reactions with C1-, are retarded by the addition of the salt. The rate constants for the neutralization reactions have been determined for the free-ion and ion-paired states. The pulse radiolysis results are compared with those for other aromatic compounds such as triphenylethylene, tetraphenylethylene and perylene. The appreciable spectral shift is characteristic of the diphenylpolyene radical cations. Extensive spectrophotometric studies of ion pairing have been undertaken on hydrocarbon radical anions and carbanions. 1-3 The absorption maxima of the anionic species are shifted to shorter wavelengths when they form contact ion pairs with alkali- metal cations.The effects of ion pairing in pulse radiolysis have also been studied on anionic species generated in tetrahydrofuran solutions containing alkali-metal salts and quaternary ammonium Recently, we reported that the lifetimes of aromatic radical cations generated in chlorohydrocarbon solutions are extended by the addition of quaternary ammonium salts having non-nucleophilic PF,, BF, and ClO, anions.''. l8 The result has been interpreted in terms of ion pairing of the radical cations with the anions from the salts. The radical cations investigated in the previous studies are those of biphenyl (BP), anthracene, pyrene and trans-stilbene. No apparent spectral shift due to ion pairing was observed for these radical cations.The present pulse radiolysis study was undertaken for the radical cations of all-trans a,w-diphenyl-substituted polyenes such as 1,4-diphenylbuta- 1,3-diene (DPB), 1,6-diphenylhexa- 1,3,5-triene (DPH) and 1,8-diphenylocta- 1,3,5,7-tetraene (DPO) generated in 1,2-dichloroethane solution. It was found that the absorption spectra of the diphenylpolyene radical cations are shifted to shorter wavelengths by ion pairing with PF; derived from added Bu,NPF,. The results for the radical cations of other aromatic compounds having comparable molecular weights, such as triphenylethylene (TrPE), tetraphenylethylene (TePE) and perylene (Pe), are presented for the sake of comparison. The influence of the structure of the radical cations on ion pairing is described.Experimental 1,2-Dichloroethane and Bu,NPF, were the same as those used in the previous study.ls DPB, DPH and DPO (all Aldrich) were recrystallized from hexane, hexane-benzene 22092210 Decay Kinetics of Diphenylpolyene Radical Cations Oa2 t o ' 5 k 5iO 520 5kO 5;; X/nm dm-3 Bu,NPF,. Fig. 1. Transient absorption spectra of DPB'+ in the absence (0) and presence (0) of 6 x mol mixtures and benzene, respectively. TrPE (Aldrich) and TePE (Wako) were recrystallized from methanol and benzene, respectively. Pe (Aldrich) was recrystallized from benzene. The experimental procedures and the techniques of pulse radiolysis have been described in the previous paper.13 The pulse radiolysis experiments were carried out by using 8 ns electron pulses (beam diameter, ca.4 mm) and at room temperature (ca. 22 "C). Results Pulse radiolysis of aromatic compounds in 1,2-dichloroethane solutions results in the formation of the radical cations via charge transfer from the solvent cation. Solute concentrations above ca. 4 x mol dmP3 are necessary for the complete capture of the solvent cation.18 The formation of the radical cations is followed by the second-order decay due to the neutralization reactions with C1-, which is a product of electron attachment to the solvent. mol dm-3 DPB solution irradiated in the absence and presence of 6 x mol dm-3 Bu,NPF,. The absorption peak, assigned to DPB'+,l9 is shifted from 546 to 542 nm in the presence of the salt. This is direct evidence for the formation of a contact ion pair between DPB'+ and PF,.The shapes and peak positions of the spectra were unchanged during the decay after the pulse irradiations. The decay rate of DPB'+ was decreased by a factor of ca. 5 in the presence of the salt. The decay rate was constant at salt concentrations (6-10) x mol dm-3 as well as the peak position of the spectrum. This means that DPB'+ is completely paired with PF, at the salt concentrations used. Fig. 2 and 3 show the spectral shifts of DPH'+ and DPO'+ caused by the addition of 6 x lop3 mol dmF3 Bu,NPF,. Because of the low solubilities of DPH and DPO, the pulse radiolysis experiments were carried out at a low solute concentration, 3 x mol dm-3 BP. The addition of an excess of BP is for the complete capture of the solvent cation, followed by the charge transfer to the solutes.For DPB'+ the absorption spectra and the decay rates observed with the 5 x mol dm-3 DPB solution were similar to those with the 3 x mol dm-3 DPB solution containing 6 x mol dm-3 BP. The decay rates of DPH'+ and DPO" were decreased by the addition of the salt similar to the case of DPB'+. Fig. 1 shows the transient absorption spectra for the 5 x mol dm-3, in the presence of 6 xY. Yamamoto, T. Aoyama and K. Hayashi 221 1 X/nm Fig. 2. Transient absorption spectra of DPH'+ in the absence (0) and presence (a) of 6 x mol dm-3 Bu,NPF,. I I I OS2 t h/nm Fig. 3. Transient absorption spectra of DPO'+ in the absence (0) and presence (@) of 6 x mol dm-3 Bu,NPF,. The second-order kinetic plots for the decays of the radical cations gave straight lines.The slope of the plot corresponds to k / d , where k is the second-order rate constant, E is the molar extinction coefficient and Z is the optical pathlength. The absolute rate constants for the neutralization reactions with Cl- were determined in the absence and presence of 6 x mol dm-3 Bu,NPF,. Irradiation cells of 1 and 2 mm optical pathlengths were used for the determination. Fig. 4 shows the second-order kinetic plots for the decays observed with the 1 mm cell. It was confirmed that the slope of the second- order kinetic plot determined with the 2 mm cell is equal to half the slope with the 1 mm cell. The values of E were obtained by comparison of the absorption intensities of the radical cations based on the literature value of E(BP'+), 1.66 x lo4 dm3 mol-1 cm-' at 680 nm." The data outlined in the present study are summarized in table 1.Similar experiments were carried out for TrPE'+, TePE'+ and Pea+. Because of the low solubilities of TePE and Pe, the data for TePE'+ and Pe'+ were obtained with the2212 Decay Kinetics of Diphenylpolyene Radical Cations '0 31 200 400 600 tlns - ' tins '0 200 400 600 Fig. 4. Second-order kinetic plots for the decays of (a) DPB'+, (b) DPH" and (c) DPO'+, monitored at the peak positions, in the absence (1) and presence (2) of 6 x mol dm-3 Bu,NPF,. The measurements were carried out with the 3 x mol dm-3 solutions containing 6 x mol dm-3 BP in the irradiation cells of the 1 mm optical pathlength. Table 1. Spectral shifts and rate constants for the neutralization reactions with C1- determined in the absence (k,) and presence (k,) of 6 x mol dm-3 Bu4NPF, rate constant/dm3 mo1-I s-' cation A/nm AA/nm &/dm3 mol-I cm-I k0 ks k0PS DPB'+ 546 4 1.2 x 105 (1.6k0.1) x lW1 (2.9kO.l) x l W o 5.5 DPH" 606 6 2.2 x 105 (1.4f0.1) x l W 1 (2.5k0.1) x l W o 5.6 DPO'+ 663 5 2.5 x 105 (8.6 +_ I .2) x 101O (1.8 +_ 0.2) x l W o 4.8 TePE" 490 0 2.9 x 104 (2.8 fO.l) x 109 (2.7k0.1) x 108 1.0 Pe'+ 545 0 6.4 x 104 (7.2k0.8) x 108 ( 6 .7 f 0 . 8 ) ~ 109 1.1 TrPE'+ 496 < 2 2.8 x 104 (1.1 kO.1) x 1W' (2.0k0.2) x l W O 5.5 ' - 'a s .ff 5, -2 4t 1 '0 u 200 400 600 8.0 I-, 4.2 5*0r-l 1.8 t i 1.0 0 400 800 1200 tins tlns tlns Fig. 5. Second-order kinetic plots for the decays of (a) TrPE'+, (b) TePE'+ and ( c ) Pe'+, monitored at the peak positions, in the absence (1) and presence (2) of 6 x mol dm-3 Bu,NPF,.The measurements were carried out with the 3 x mol dm-3 solutions containing 6 x mol dm-3 BP in the irradiation cells of 2 mm optical pathlength.Y. Yamamoto, T. Aoyama and K. Hajlashi 2213 3 x lop3 mol dm-3 solutions containing 6 x mol dmP3 BP. The lowering of the Pe concentration prevents the formation of the dimer radical cation.21 The results for TrPE'+ obtained with the 5 x mol dm-3 TrPE solution were similar to those with the 3 x l 0-3 mol dmP3 TrPE solution containing 6 x 1 0-2 mol dm-3 BP. The spectral shift of TrPE'+ caused by the addition of 6 x lop3 mol dm-3 Bu,NPF, was extremely small, Ad < 2 nm for the 496 nm band. On the other hand, the absorption spectra of TePE'+ and Pe'+ in the presence of the salt were completely superposed upon those in the absence of the salt.Fig. 5 shows the second-order kinetic plots for the decays of TrPE'+, TePE'+ and Pea+ observed with the 2 mm cell. The decay rate of TrPE'+ is decreased in the presence of the salt by a factor of ca. 5, whereas those of TePE'+ and Pee+ are not affected by the salt. The data for these radical cations are also presented in table 1. Discussion The rate constants, k, and k,, determined in the absence and presence of the salt, correspond to those for the reactions in the free-ion and ion-paired states, respectively. For DPB'+, DPH'+, DPO" and TrPE", the k, values are near the diffusion-controlled limit.ls The k, values are smaller than the k , values by a factor of ca.5. The k,/k, ratios are close to those previously determined for the trans-stilbene and anthracene radical cations (5.6 and 4.6, respectively''). This means that these radical cations are stabilized to the same extent by ion pairing with PF;. Among these radical cations, the appreciable spectral shifts by ion pairing are observed only for the diphenylpolyene radical cations. Preliminary experiments revealed that the absorption maximum of the retinal radical cation is shifted from 595 to 590 nm by the addition of 6 x low3 mol dmP3 Bu,NPF,. It can be said that the spectral shift by ion pairing is characteristic of a conjugated polyene radical cation. Of interest is the result that the conjugated polyene radical cations form contact ion pairs with PF,, even if the conjugated systems are long.This may be explained in terms of the charge localization induced by the counter-ion. On the other hand, the spectral shift by ion pairing is very small for TrPE". It should be noted that similar extents of the spectral shifts, AA < 2 nm, are observed for the 480 nm band of the trans-stilbene radical cation and for the 724 nm band of the anthracene radical cation in the presence of 6 x Neither peak positions nor decay rates of TePE'+ and Pea+ are affected by the addition of the salt. The k, values for these radical cations are more than one order of magnitude smaller than those for the others. That is to say, these radical cations have smaller reactivities or interactions with the anions, C1- or PF;.The result for TePE'+ can be attributed to the steric crowding due to the phenyl groups which surround the positively charged centre located on the olefinic double bond. The k, and k, values for Pe'+ are close to those previously determined for the pyrene radical cation ( 9 . 4 ~ lo9 and 8.1 x lo9 dm3 mol-1 s-l, respectively''). Therefore, the result for Pe'+ is explained in terms of the formation of a complex between Pe'+ and C1- similar to the case of the pyrene radical cation : Pe'+ + Cl-G Pe'+..-Cl- --+ neutral product. In this case the k, and k, values correspond to k,k,/k,. The charge delocalization of the large aromatic radical cations may be responsible for the results for the Pe and pyrene radical cation. mol dmP3 Bu,NPF,, as revealed by a careful reinvestigation.kc kn kd We are grateful to Kunihiko Tsumori, Norio Kimura, Tamotsu Yamamoto, Toshihiko Hori and Dr Seishi Takeda for help with the pulse radiolysis experiments.2214 Decay Kinetics of Diphenylpolyene Radical Cations References 1 J. Smid, Ions and Ion Pairs in Organic Reactions, ed. M. Szwarc (Wiley-Interscience, New York, 1972), 2 M. A. Fox, Chem. Rev., 1979, 79, 253. 3 H. M. Parkes and R. N. Young, J. Chem. SOC., Perkin Trans. 2, 1980, I137 and references therein. 4 J. H. Baxendale, D. Beaumond and M. A. J. Rodgers, Trans. Faraday Soc., 1970, 66, 1996. 5 B. Bockrath and L. M. Dorfman, J. Phys. Chem., 1973, 77, 1002. 6 B. Bockrath and L. M. Dorfman, J. Phys. Chem., 1973, 77, 2618. 7 B. Bockrath and L. M. Dorfman, J. Am. Chem. SOC., 1974, %, 5708. 8 G. A. Salmon and W. A. Seddon, Chem. Phys. Lett., 1974, 24, 366. 9 G. A. Salmon, W. A. Seddon and J. W. Fletcher, Can. J. Chem., 1974, 52, 3259. 10 J. R. Langan and G. A. Salmon, J. Chem. SOC., Faraday Trans. 1, 1982, 78, 3645. 11 J. R. Langan and G. A. Salmon, J. Chem. SOC., Faraday Trans. I , 1983, 79, 589. 12 B. Hickel, J. Phys. Chem., 1978, 82, 1005. 13 Y. Yamamoto, S. Nishida, K. Yabe, K. Hayashi, S. Takeda and K. Tsumori, J. Phys. Chem., 1984,88, 14 M. Ogasawara, N. Kajimoto, T. Izumida, K. Kotani and H. Yoshida, J. Phys. Chem., 1985, 89, 15 Y. Yamamoto, S. Nishida, X-H. Ma and K. Hayashi, J. Phys. Chem., 1986, 90, 1921. 16 For flash photolysis study see: M. Fisher, G. Ramme, S. Claesson and M. Szwark, Chem. Phys. Lett., 1971, 9, 306; 309; G. Ramme, M. Fisher, S. Claesson and M. Szwark, Proc. R. SOC. London, Ser. A, 1972, 327, 467. vol. 1, p. 85. 2368. 1403. 17 S. Mah, Y. Yamamoto and K. Hayashi, J. Phys. Chem., 1983, 87, 297. 18 Y. Yamamoto, S. Nishida and K. Hayashi, J. Chem. Soc., Faraday Trans. 1, 1987, 83, 1795. 19 T. Shida and W. H. Hamill, J. Chem. Phys., 1966,44,4372. 20 Y. Wang, J. J. Tria and L. M. Dorfman, J. Phys. Chem., 1979, 83, 1946. 21 A. Kira, S. Arai and M. Imamura, J. Phys. Chem., 1972, 76, 1119. Paper 711872; Received 19th October, 1987

ISSN:0300-9599    

DOI:10.1039/F19888402209

出版商:RSC    

年代:1988

数据来源: RSC