GENERAL DISCUSSIONProf. J. N. Murrell (University of Sussex) said: The paper by McWeeny, Masonand Tow1 presents a theory which predicts that ligand geometries should be relatedto the geometries of their excited triplet states. I think this is based on doubtfulassumptions. In the first place, if one argues from the standpoint of separated-statecontributions to the ground state wave functions, then the molecular orbital expressionis a bad starting point. Thus, the expansion in expression (3) of their paper containsno terms corresponding to excited singlet states of the ligand, A(laLQMg), whereasthese must arise to represent the physical situation of the ligand being polarized bythe field of the metal. Both this state, and ionic states like AM(-L+), will have lowerenergies than the double triplet kf(3@D,@M) which they consider important, and thematrix elements between these and the ground state A(@,,QMg) will probably belarger (the matrix element between the ground state and the double triplet is secondorder in overlap).From a separated state theory, one is more likely to deduce that the ligand geo-metry is close to the geometry of the ligand excited singlet state or the ligand negativeion than that it is close to the triplet state geometry.As excited singlet and tripletstates that are related will have similar geometries it is difficult to use the experimentaldata to differentiate these theories.Prof. R. Mason (Shefleld University) said : In reply to Murrell, it is true that a mol-ecular orbital description would be a bad starting point in describing weakly interact-ing systems, for which a separated-state wave function might be more appropriate.Inthe present situation, however, overlap is considerable and we are particularly con-cerned with the overlap-dependent terms characteristic of covalency effects : the M.O.function has been used as a starting point simply to examine the types of separated-state contributions which arise when it is re-expressed in separated-state language.We think it significant that the function, mentioned by him is overlapindependent and therefore more appropriate in describing polarization effects at longrange.Without making a full variation caldation on a particular complex, we cannot becertain about the relative weights of the various ionic and locally excited contributionsreferred to in our expansion, but our view is that the ionic functions will predominatein complexes where the metal is in a high formal oxidation state, while the neutralexcited state functions will be important in complexes where the metal has a lowoxidation state (0 or + 1, say).As for the relative importance of singlet and tripletexcited states, it is true that the geometry of the excited ligand would be very similar :we think the triplet is most important for two reasons : (i) the expansion in powers ofoverlap gives the triplet term three times the weight of the singlet, and more signifi-cantly, (ii) when fragments in triplet states are coupled to a singlet, overlap is accom-panide by a stabilization effect (which would enhance their weights in a variationalcalculation), whereas fragments in singlet states are mutually repulsive as overlapincreases (indicating a diminishing variational importance).The structural results on several butadiene and oxygen complexes support ourbelief that the ionic functions are relatively unimportant.Thus, for butadiene,560 GENERAL DISCUSSIONCooper and McWeeny give the following n-bond orders (complete CI (M.O. andV.B.)) for the lowest triplet state and anion respectively.2 0.727 3 2 0-580 30.380 0.638 0.638 /7 I + - - n 0=148--- 4 q--0 . 3 8 004091--* 4Bu- 3 B ~ *There is no doubt that a number of butadiene complexes have geometries whichapproximate more nearly that of 3 B ~ * than of Bu- or Bu2- ; this extends even to the" squeezing " together of atoms 1 and 4 in the complexes (note the positive long-range bond order) compared with their separation in cis-butadiene.Again the bondlengths of the 0- and 02- ions are well characterized as 1.31 and 1.48 A respectively ;these are distances which are reproduced in a variety of bridged binuclear cobalt(II1)complexes but the 0-0 bond length of 1.65 A in the [(Ph2PCH2CH2PPh2)2Lr02]cation must imply, in our view, that the electron distribution in the ligand is muchmore nearly that of the 3C; neutral excited state than of the anion. In M.O. termsthese conclusions suggest that a population analysis would show the coordinate Iigandto be roughly n37r*3 rather than n47P3 or 7 t 4 ~ * 4 .Mr.A. F. Orchard (Oxford University) (partly communicated): I believe that theapproximate molecular orbital method employed by Dr. Hillier and Mrs. Canadine,whilst undoubtedly a considerable advance on the Wolfsberg-Helmholz approach,is unsatisfactory in its present form. I am particularly concerned that the quantitycfu (in eqn. (3) of the paper) is approximated simply as an atomic IP (or VSIE). Ifthe model is to be related to the LCAO-MO SCF theory in a reasonably well definedmanner then E ~ , must have the general formwhere = (iu I 7'- Vlore I iu), and nku is the effective occupancy of the atomicorbital Xku. Expressions of the type (1) may be obtained by applying either zerodifferential overlap (ZDO) or Mulliken-style approximations to the Roothaanequations.Such procedures are highly approximate, but unless they are pursuedconsistently the simple m.0. treatment will suffer all the ambiguity inherent in Huckelmet hods.According to eqn. (l), Eiu is a one-electron energy (for the a.0. xi, in the valencestate of the atom u) which includes the self-repulsion term, +nfuJlu,iu. In other words,for a closed-shell molecule the n,, electrons occupying xfu in the valence state mustbe treated as +niu of each spin type, a and p, and these repel each other. Now thispoint is a familiar one [e.g., ref. (6)], but deserves additional emphasis here, particularlyin connection with the reported ferrocene calculation.Suppose Xi, belongs to the atomic I-subshell a, which is symmetrically occupied inthe valence state.With modification of notation (and dropping the u subscript),Cooper and R. McWeeny, J. Chem. Phys., 1968,49,3223.Churchill and Mason, Adv. Orgunometallic Chem., 1967, 5, 93.e.g., Gerloch and R. Mason, Proc. Roy. SOC. A , 1964, 279, 170. Churchill and R. Mason,Proc. Roy. SOC. A , 1966, 292, 61.4 J. A. Pople, D. P. Santry and G. A. Segal, J. Chem. Phys., 1965,43, S129.5 J. W. Richardson and R. E. Rundle, A Theoretical Study of the Electronic Structure of TransitionMetal Complexes, U.S.A.E. Report ISC-830, Ames Laboratory, Iowa State College, Ames,Iowa, 1956).6 R. E. Watson and A. J. Freeman, Phys. Rev. A , 1964,134, 1526GENERAL DISCUSSION 61the average one-electron energy iswhere o, is the degeneracy of the subshell, and where Oab = (.Iik- Kik) is a fullyaverged two-electron interaction.Now a space and spin randomized IP (cf. V0IP)lis a different quantity from If we neglect orbital rescaling effects (as in Koop-mans' approximation, for example) then the fully averaged atomic IP from thesubshell a isIy[na,nb.. .] = - ,ye- (na - lIeixa - nbeab* (3)b#aAt this level of approximation, we therefore haveE , = - 1; + [ 1 - (na/2ua)]Oaa, (4)[cf. ref. (2)) Thus, E , may be identified with -lav only when the subshell a is fullyoccupied.In the general case, the second term in eqn. (4) is by no means trivial. Its magni-tude will often be comparable with that of the penetration integrals that Hillier andMrs. Canadine are much concerned with (eqn. (3) of the paper).There will besimilar difficulty with the off-diagonal elements of the simplified Fock matrix.The role of the self-repulsion terms in the valence state is in fact more complexthan my eqn. (4) indicates, because the degeneracies of the atomic 2-subshells arelifted in sufficiently low molecular symmetries. The concomitant self-consistenteffects must then be carefully taken into account. Let us consider the ferrocenecalculation reported here. Arguing directly from eqn. (l),Ed(eZg)-Ed(alg) = '&nal,(FO- 16F2-39F4)-+ne2@(Fo - 36F2 61F4) - 5neIn(F2 - 5F4), ( 5 )where the Fk are the usual Slater-Condon parameter^.^ If we now use the numericalresults of Hillier and Mrs. Canadine, with the same choice of metal 3d wave-functions,we obtain &d(e2g)-&d(~1g)N3'8 eV.The ad hoc adjustment of the reported m.0.eigenvalues then leads to an e2,-a,, energy separation of about +Om5 eV, and thee2g molecular orbital energy becomes about -7.2 eV. These figures are in moresatisfactory agreement with the photoelectron spectrum of ferrocene determined byTurner.4 A similar consideration of the chromium dibenzene results also suggestsa reversal of the relative positions of the mainly-metal ez, and a,, levels. We hopeshortly to measure the photoelectron spectrum of chromium dibenzene in order toexamine this point.The above arguments are incomplete, and the numerical features insubstantial :and, the excellent agreement with the ferrocene photoelectron data is fortuitous.However, it does seem likely that the self-repulsion terms represent an importantfactor determining the relative energies of the eag and a,, molecular orbitals.Itwould be interesting to see what effect these additional terms produce in a repeat ofthe full self-consistent calculation. Certainly my own experience with calculationsH. Basch, A. Viste and H. B. Gray, Theor. chirn. Acta, 1965,3,458.Press, 1963).P. Day (Wiley, 1968).2 L. C. Cusachs and J. W. Reynolds, J. Chem. Phys., 1965,43, S160.3 E. U. Condon and G. H. Shortley, The Theory of Atomic Spectra, (Cambridge University4D. W. Turner in Physical Methods in Advanced Inorganic Chemistry, ed. H. A. 0. Hill an62 GENERAL DISCUSSIONon simple transition-metal complexes is that the self-repulsion energies are a crucialaspect.For example, without their inclusion, one cannot reproduce the subtleeffects manifest in are cent, “ exact ” LCAO-MO SCF treatment of hypotheticalsquare-planar NiFz- to which Hillier has referred.A much improved methodology is needed if an approximate molecular orbitaltheory of the extended Huckel type is to contribute seriously in areas such as metallo-organic chemistry. Apart from the problems discussed above, some attention mustbe given to neutral penetration integrals, which are particularly significant in highlycovalent molecules. These are ignored in the treatment advocated by Hillier andMrs. Canadine. And there clearly remain difficulties with regard to the satisfactoryhandling of diffuse valence a.0. such as transition metal (n + l)s,p or d orbitals.Similar problems are apparent in the use of 3d valence orbitals for atoms belongingto the second short period.Dr.I. H. Hillier and Mrs. R. M. Canadine (University of Manchester) (communi-cated) : Orchard has suggested that our computational scheme may be improved bythe introduction of self-repulsion terms into the Etu of eqn (3). Any semi-empiricalscheme that is to retain a large measure of computational simplicity and thus besuitable for the treatment of large inorganic systems, must contain a number of grossapproximations and may be criticized for their introduction. However, the usefulnessof such schemes should be judged by comparison with more accurate calculations andwith experimental data for a wide range of compounds. For a series of transitionmetal halides, and carbonyl complexes we find our method gives encouraging resultswhere such comparisons are possible, although it is naturally not expected to repro-duce the subtle effects of ab initio calculations.Such comparisons with ab initiocalculations are limited by the small number which have been performed. For thisreason we have underway in our laboratory such calculations on a number of inorganiccomplexes, which will hopefully suggest ways of improving semi-empirical schemes.Such improvements may be attempted by reducing the approximations involved, butat the expense of additional computation which must be weighed against the improve-ment obtained in the wavefunction. It would thus certainly be interesting to see ifOrchard’s suggestion does lead to a significant improvement for a range of molecules.The other problems referred to by him, such as the treatment of the penetration terms,and the difficulty of handling diffuse valence atomic orbitals are well-known and donot require further comment, except to suggest that ab initio calculations referred topreviously should throw light on their correct inclusion.Prof.J. N. Murrell (University of Sussex) said: The C=O stretching bands ofcomplex carbonyls can be fitted adequately by assuming a harmonic force field withdiagonal C=O force constants and off-diagonal C=O, C=O interaction constantsk’. However, if these empirical force constants are to be interpreted in terms ofvalence calculations, then a more precise definition of k’ is required.In the system, O * L M - k b , the effective k’ between the carbonyl vibrationsdepends on all the interaction constants k14, k12, k34, k13, k24 and k23, where wedefine k14 = (d2V/dRlaR4) etc.The precise form in which these contribute tok’ depends on the G-matrix, i.e., on the masses of the atoms and the geometry. Tointerpret the experimentally determined k’ as k14, as has been done by Anderson andBrown, is an oversimplification.‘H. Basch, C. Hollister and J. W. Moskowitz, Application of MotecuIar Orbital Theory toProb[ems in Inorganic Chemistry, 154th Meeting, Amer. Chem. SOC., (Chicago, Illinois, 1967).A. F, Orchard, unpublished resultsGENERAL DISCUSSION 63Dr. D. S . Urch (Queen Mary College) said: Unfortunately reference to the workof Hayter was omitted from our paper.The relevant data are :v1 v2 v3 v4(Fe(C0)3P(CH3)2)2 2050 210 1977 19622043 2004 1973 1961(Fe(C0)3BrP(CH3)2} 2 2080 2038 2010 - {Fe(C0)3AS(CH3)2)2( Fe (CO) 3 €3 r As (CH 3) 2 1 2 2072 2034" 2008{Fe(C0)31As(CH3)2) 2 2066 2028 2006* given as2054 in solution but 2038 in mull. It seems reasonable to suppose that 2054 is a misprintfor 2034.The four frequencies reported for the first compounds are in accord with the modelpresented. In the last three compounds the Fe-Fe bond has been broken byhalogenation and the only remaining route for Fe(CO)3-Fe(C0)3 interaction isacross the chalcogen bridge. Under these circumstances only one high frequencyline is observed,- confirming that theformation of two. high-frequency linesis due to interactions associated with adirect Fe-Fe bond.The differencebetween the average of v3 and v4 andthe single high frequency line is stillca 50-55 cm-l. This is the magnitude ofsplitting expected in an isolated Fe(C0)3group, and is also strong evidence thatthe low-frequency splitting is due only tothe inductive effect of the bridge groups.Dr. J. R. MilIer (Essex University) said :The infra-red spectrum of [Fe(CO),SEt],in fig. 1 in cyclo-hexane solution wasmeasured by K. Edgar at ManchesterUniversity. Unlike the spectra reportedby Urch, it shows five strong bands anda number of weak ones. We thoughtinitially that the strong bands would bethe C-0 stretching fundamentalallowedby the Czu selection rules and that theweak bands could be attributed tonaturally occurring 3C0 species.However, in view of the commentsmade by Bor, it seems at least as likelythat this spectrum is of a mixture ofsyn-anti isomers.One of the maincm-l2100 2 000 1900I 1 IFIG. 1 .-Infra-red spectrum of [Fe(CO),SEt], inthe 2000 cm-' Region (cyclohexane solution).features of interest in these compounds is the interaction between C-0 oscillators atopposite ends of the molecule; Urch has suggested that this interaction may be thesame between any pair of C-0 groups on different iron atoms.R. G. Hayter, Inoi-g. Chem., 1964, 3, 71464 GENERAL DISCUSSIONOur spectra, if they be of one pure isomer, show that there is a detectable splittingof the low frequency bands into three components ; if they be of isomeric mixtures,then they show that the isomers do have different frequencies.Both of these possi-bilities, we would suggest, lead to the conclusion that the CO-CO interaction isdependent upon the path across the molecule. In order to clear up the interpretationof these spectra it is desirable that measurements and proper assignments should bemade on the naturally occurring 3C0 bands undoubtedly present, and the appropriatecalculations made.Dr. D. S. Urch (QueeB Mary College) said: I find it hard to understand how thespectrum presented by Miller could be that of a pure compound, and would concurwith his opinion that it must be of a mixture. A mixture of syn- and anti-isomers(at least?) seems reasonable, bands A , C, E, G (and possibly P) could comprise onespectrum and BD together with other bands in the EFG region could be the spectrumof another isomer.The latter would then be related to the former by a shift to lowerfrequencies of about 3-5 cm-I. As suggested in the reply to Greenwood there is noreason why a change in conformation should not manifest itself as a small changein the inductive effects of the XC6Y5 groups.It is not clear whether P must be regarded as due to an impurity or associatedwith the BD spectrum or whether ACEFG is a true five line spectrum. If the latteris found to be so then this may be easily understood as due to some deviation fromlocal trigonal symmetry by the carbonyl groups in X(CO),.Thus, neither thepossibility of slight shifts due to conformational isomers nor the observation of afive line spectrum is reason for doubting the basic validity of the simple theorypresented the original paper. It is interesting to observe that ACEG fits quite wellwith this theory.King has reported the infra-red spectra of two isomers of the related compound,(Fe(C0)3SCH3)2 :(i) 2085 2050 2000(ii) 2075 2040 2000 1995.A slight conformational shift can be observed as well as a slight change in the inductiveeffect due to substituted chalcogens, as suggested above the explanation of Miller’sresults.Dr. G. Bor (University College, London) said : The spectrum of [Fe(CO),SEt],given in Miller’s remark, and Urch’s proposed interpretation to it, prompts somefurther comments.Based on the published results on the i.-r. spectra ofpure ant- andsyn isomers of the same compound and on Cl80 enrichment studies with the ana-logous methyl comp~und,~ the following straightforward assignment of the publishedspectrum may be made :A, C , E, G : fundamental C-0 stretches of the anti-isomer,B, C, F, G : fundamental C-0 stretches of the syn-isomer,D, I, J (wavenumbers for the CH,-compound : 2 028.2, 1 956.4, and 1 943.3 cm-I,respectively 2, : 13C-0 bands of the anti-isomer.R. B. King, J. Amer. Chem. SOC., 1962,84,2460.G. Bor, J. Organometal. Chem., 1968, 11, 195.P. W. Robinson, private communicationGENERAL DISCUSSION 65The C frequencies of both isomers coincide and the G bands are separated onlyby 2 cm-1 resulting in the broader shape of this band if the spectrum is recorded on the“ natural ” mixture of the isomers.The 13CO-satellite band of A is overlapped inthis spectrum by the highest C-0 fundamental B of the syn-isomer. The muchlower concentration of the syn-isomer does not enable one to detect the isotopebands of this isomer from the published spectrum.Dr. J. R. Miller (Essex University) said: I would suggest that the term “forceconstant ” be reserved for quadratic potential coefficients of the general type[C3’V/drtC3rjl0, and that this term should not be used for the type of experimentalparameter under discussion. A convenient naming for these latter quantities wouldbe CK parameters (after Cotton and Kraihanzel).Concerning the relationship between these experimental parameters and the trueforce constants,05 4if one assumes that the vibrational force-field of a carboayl molecule is harmonic(which is open to criticism), then the true force constants for stretching co-ordinatesin a symmetric dicarbonyl molecule, as defined in the figure, may be related to theCK parameters in the following way :~ C I C = kz + (2*3/7)ki - @/7)k12,k& = ki, 4- (2’3/7)k23 4- (8/7)k13,where kCK and k& are the C-0 stretching and interaction parameters respectively.These equations are derived with the knowledge that so far as C-0 stretchingvibrations are concerned, there is negligible mechanical coupling across the metalatom (this is not true of M-C stretching vibrations).This result can be extendedto any system containing one set of symmetry-equivalent carbonyl groups and theinteresting feature is that, for a given interaction parameter, the component forceconstants are all of the off-diagonal type and are those which follow the same paththrough the molecule as the interaction parameter concerned. Thus in the octahedralcase,k&,(trans) = k,,(trans) + (2.3/7)kz3(trans) + (8/7)k13(trans),khK(cis) = k,,(cis) + (2*3/7)k,,(cis) - (8/7)k13(cis).Dr. G. Bor (University College, London) said: I would comment on the paper ofReddy and Urch. Our observations are based on the spectra of the same type ofcompounds as well as the cobalt complexes Co,(CO),(RC = CR’)2 having strictlyanalogous structures to the RS- (RSe- or RTe-) bridged compound^.^ Weconfirm that the organothio-bridged Fe2(CO),(SR), compounds have in fact onlyfour C-0 stretching bands, even if R = alkyl (having usually narrower bands andthus giving better resolution), or if the syn and anti forms are studied in pure formG.Bor, J. Organometal. Chem., 1968, 11, 195.G. Bor, Chem. Ber., 1963,96,2644.W. G. Sly, J. Amer. Chem. SOC., 1959,81, 18.L. F. Dahl and C. H. Wei, Inorg. Chem., 1963, 2, 328.G. Bor, Proc. Symp. Co-ord. Chem., Tihany (Hungary), Sept. 1964, p. 361.66 GENERAL DISCUSSION(cf. fig. 1 in ref. (1)). On the other hand, the acetylenic dicobalt hexacarbonylshave$ve C-0 stretching bands if the spectra are taken with good resolution (cf.fig. 1 and table 1 of ref. (2), and fig.3 of ref. (5)). The possibility that the lowest-frequency band of the Co,(CO),(acetylene) type compounds, being always weak,may be a 13C0 isotope band has been excluded by 13C0 enrichment studies. Weneed an explanation to account for all of these observations.If we construct the F matrix elements (in a factored off C-0 stretching model)without any a priori assumptions or constraints between the interaction constantsthen we have two direct and four indirect interaction (“ perturbation ”) constantsthe latter ones acting between the two halves of the molecule (ijk acting betweenthejth and kth CO ligand, according to the numbering scheme of the authors).Then if i36 # i34 there is a “splitting” between the a, and b, modes. Andfrom similar studies with other binuclear carbonyls these indirect interactions betweenCO ligands bound to different metal atoms are mainly distance-dependent and acis or cisoid type of interaction is usually 0.08 to 0.20 mdyn/A higher that the transinteraction.This difference results thus in a separation b2-a2 = 5-12 cm-’. Butsince the a, species is i.-r.-forbidden this does not appear in the spectrum of eitherof these types of compounds.On the other hand, the authors have assigned the entire splitting due to “ kd ”to the higher roots of the two second-order species a, and b, which generally is trueonly if iI2 = i14-again this very unlikely. But since i34 and i12, on the one hand,and i14 and i36, on the other, may be expected to have nearly equal values, thev2(al)-vv,(b,) separation may be very near to the value of the v,(b,)-v,(a,) splitting.Thus the alternative model is the following : the local symmetry of the M(CO)3groups is not C3, but C, which should give rise to a three-band spectrum, the magni-tude of the lower a’(v,) -v3(a”) splitting reflecting the deviation from the threefoldsymmetry.The secondary splitting due to the ( 0 C ) M . . . M’(C0)‘ coupling islarge (32-40 cm-l) for the v , ( L z , ) - v ~ ( ~ ~ ) separation, and smaller (7-18 cm-’) for thetwo lo w-frequency separations.Our view is that the number of five bands (being normal for point group C2, andobserved experimentally for the alkyne-bridged cobalt complexes) is reduced to fourfor the RS-bridged compounds by the overlapping of the weak band v,(bl) throughthe strong v6(bZ).If the (hypothetical) v,(a’) -v3(a”) splitting of the monomericM(CO)3 unit is small as compared with the “ dimerization splitting ”, this overlappingdoes not occur. Clearly the conditions for this case are the better given the nearerthe local symmetry of the M(CO)3 groups is to C3,. This may be the case for thecobalt compounds. If, on the other hand, the two types of splittings are nearlyequal, v5 should be overlapped by vg, or by the v 2 f v 6 doublet, which seems to bethe case with the iron compounds.Dr. D. S. Urch (Queen Mary College) said: I must disagree with the rather complexinterpretation of the infra-red spectra proposed by Bor. Let us consider the possi-bility that different carbonyl-carbonyl perturbation force constants may exist due todifferent interaction routes across the “ bridge ” ; these will then be, assuming C,,symmetry, and numbering the carbonyl groups as in fig.1 :3-4 = 5-6 = k,,4-5 = 3-6 = kp,1-2 = k,1-4 = 1-6 = 2-3 = 2-5 = kgGENERAL DISCUSSION 67Then the normal coordinate functions in table 2 have the corresponding forceconstants :a1 + a2 6-1 *(2ky +4kx+ 12k, + 4k, + 4ks + 8ka + 2ky),a1 - 6-' (2ky + 4kx + 12kc -4ka-4ka - 8ka-2ky)ya3 + 12-1 (8ky + 4kx- 12kC-4ka - 4ka + 16ka - 8ky),as + a6 12-1 (8ky +4kx- 12k,+ 4ka +4ka - 16ka + 8ky),a3 - a4a5 -a64-1 (4kx - 4kc - 4k, -E 4kp),4-1 (4k,- 4k,+4kU-4kp).Furthermore there will be some interaction between the two coordinates of symmetryal and the two coordinates of symmetry 6,.This has been ignored in the above tablebut the magnitude of the cross-term may not be negligible :for a,, 6-1 (4ky-4k,-4k,-4ka+4k,+4kd),and for bl, 6-' (4ky - 4kx + 4k, + 4ks - 4k, - 4kd).If all the " bridge " perturbation force constants are different and vary both withchalcogen and with the group attached to the chalcogen and with the relative orienta-tions of the groups attached to the chalcogens, it is difficult to see how any regularpattern could emerge from the infra-red spectra of a wide variety of compounds.And yet remarkably regular and constant features are observed. In all the four linespectra discussed above it was observed that, no matter what the compound, thedifference between the average of the two high frequency lines and the average of thetwo low frequency lines was a constant (-55 cm-').This is clearly associated withsome invariant feature of the molecules, e.g., an X(CO)3 group. It is of interest tonote that the splitting in carbonyl stretching frequencies associated with three carbonylgroups cis to each other is about 55 cm-l.An examination of the force constants for al +a2 and al -a2 in both the " simple "and " complex " theories show that the splitting here is associated with " bridge "interactions. However, in its " complex " form the theory suggests that the low-frequency lines will also be shifted, and as can be seen from the force constants, in amost complicated way. It is difficult in this case to see why the spectra should showa total of four and not five lines and why any simple relationship about averages offrequencies should ever hold.It therefore seems necessary to suppose that the low-frequency lines are not shifted by interactions across the " bridge " and this in turncan be achieved by postulating that ka = kp = ky = ka. I am surprised that thisis so. It is interesting to note that if all the " bridge " perturbation force constantsare equal then the coupling term between the two al and the two bl coordinatesreduces to (2/3) (ky-k,.). This is very small indeed (table 1) and explains why thiscoupling was safely ignored in the original paper.It should be remembered that the reason why the low frequency lines are notshifted by " bridge " interactions in the simple theory is because of the coefficientschosen for the carbonyl normal coordinate modes at each X(CO)3 site.Thesecoefficients depend on local C3" symmetry. Thus any deviation from this localsymmetry would result in slight shifts in the low frequency region and in particularmight well result in a weak line due to (as-a6) being observed (coincident witha,+&, in the simple theory). If kx>ky, then a g - c t g might still not be observedsince it would be displaced to lower frequencies in the region of strong absorptiondue a3+a,. On the other hand, when k,<ky, then a5-a6 would be the lowes68 GENERAL DISCUSSIONfrequency of all and might be most easily discerned as a shoulder or as an independentband.Thus the simple model presented above can easily be adapted to explain thepossible observation of five bands in some compounds but this does not affect thetype of bonding proposed in the bridge region but merely reflects the degree ofdistortion from local CSv symmetry at each X(CO)3 site.Dr.G. Bor (University ColZege, London) said : The spectra of [C1*0] enrichedsamples of the A isomer of di-p-methylthiodi-iron hexacarbonyl, Fe,(CO),(SME),,have been studied recently.2 The isotopic frequencies observed cannot be explainedon the basis of the model suggested by Urch and Reddy; moreover, they led to thedetermination of the two unobserved frequencies (species A2 i.-r. inactive ; speciesB1 obscured by the lowest observed strong band) which do not coincide with theCO stretching frequencies observed directly.This offers additional evidence forthe interpretation given in the discussion.Prof. N. N. Greenwood (Newcastle upon Tyne) : said With regard to Urch’s paper,in addition to the diamagnetism there exists strong evidence from Mossbauer spectro-scopy that compounds of the type shown in fig. 1 have a bent metal-metal bond, thusmaking the co-ordination symmetry around each iron atom octahedral. Thisevidence comes from the quadrupole splitting of such compounds which is in therange 0.65-1 a05 mm sec-l typical of &coordinate low-spin iron complexes ratherthan in the range 2.0-2.6 mm sec-l typical of 5-coordinate low-spin iron.There is also a question concerning the configuration of the bridging groupsand its effect on the symmetry arguments used in the paper. Geometric models ofsuch compounds show that the pendant groups on the bridging chalcogens can adopteither the cis or trans configurations and this can have a pronounced effect on severalphysical properties, e.g., colour, thermal stability etc.We have found that theMossbauer spectrum of cis-(OC),Fe(SMe),Fe(CO), has 6 = 0-2S5 and A = 0.895mm sec-l, whereas the trans form has 6 = 0.279 and A = 1434 mm sec-I. Otherexamples could be quoted. Were the compounds studied by Urch cis or transisomers, or were they mixtures of the two forms?Dr. D. S . Urch (Queen Mary CoZZege) said: May I thank Greenwood for pointingout that Mossbauer data also provides excellent confirmation of the structure proposedin fig. 1. Typical values of A, the quadrupole splitting constant, in Fe2(CO),*s&&)(c6Y5), are : Y = Y’ = H, 1.01 ; Y = Y’ = F, 1.33 ; Y = H, Y‘ = F, 1.16,suggestive of octahedral coordination about the iron (ref.(l), our paper).In reply to Greenwood’s question about the conformation of the groups attachedto the bridging chalcogen atoms the answer is that we do not know. In the compound[Fe(C0),S(C6F,)I2 there is some evidence (n.m.r.-ref. (l), our paper) for the exist-ance of syn- and anti- isomers in the ratio 1 : 6, but for the most part extensive work,lin an attempt to resolve syn- and anti-isomers, was without result. Thin layerchromatography of reaction products typically gave two or three spots, only one ofwhich corresponded to the required compound.The model discussed in our paper describes only one facet of bonding in themolecules [Fe(C0)3XC6Y5]2 so that it may well be there are differences in manyR.B. King, J, Amer. Chem. SOC., 1962, 84,2460.P. W. Robinson, Dept of Chemistry, University College, London, private communication.T. C . Gibb, R. Greatrex, N. N. Greenwood and D. T. Thompson, J. Chern. SOC. A , 1967,1663GENERAL DISCUSSION 69physical properties of the syn- and anti- isomers. The contention is that part of theelectronic structure which is relevant to the transmission of carbonyl group stretchingeffects involves mainly the interaction of iron 3d orbitals and is not greatly perturbedby the conformation of the groups attached to the chalcogens. This does notpreclude the possibility of slight conformational effects manifesting themselves via apurely inductive effect (see also later my reply to Miller).Dr. A.J. Rest (Cambridge) (communicated) : We have obtained the infra-redspectra of a number of transition metal carbonyl compounds in argon matrices(dilution 1 in 200) at 15°K. We find that the matrix isolated spectra, unlike thesolid state spectra described by Kettle, are analogous to those obtained in the gasphase. The absorptions are sharp with half-widths at half-height in the range 3-8cm-l.Dr. D. S. Urch (Queen Mary College) said: Kettle has raised the interesting questionof the orientation of the planar part of an olefinic molecule relative to the metal atomin complex formation. In Dewar’s model for the silver-olefinic complex the silveratom is situated above the double bond.It is interesting to speculate if this is theonly possible configuration.Gas chromatographic columns in which the active substrate is silver nitratedissolved in polymethylene glycol are effective at separating cis and trans isomersof olefins; cis isomers are retained about four times as long as the correspondingtrans isomer. Presumably this is because the cis isomer can form a more stablecomplex with Ag+ than the trans isomer. Could it be that the structure of the ciscomplex is different from the trans? Indeed, might not it be possible that the cis-isomer forms a complex with the silver ion in the plane of the olefinic atoms (sidecomplex) rather than above this plane (top complex)? (see fig. 1). From the symmetrypoint of view the side complex is just as viable as the top complex. (In the followingzC- HHI Y’top’ complex ‘side‘ complexFra. 1.argument I shall ignore the hybridization of orbitals which merely makes potentialoverlap more intuitively obvious.) Ag+ has the configuration 4dI0, 5s0, 5p0. Ifthe olefinic-Agf line is taken as the z-axis and the x-axis is parallel to the olefinicband then in the side complex the 5s orbital can receive electron by donation from theolefinic o-bond and the 5p, by donation from the n-bond. Effective back donationcan be established between 44, and the antibonding n*-orbital. All four lobes of4dx, and n* can interact. By contrast the overlap potential in the top complex ismore restricted. 5s and 5p, are now available to receive electrons from the 0 and nbonds but only one half of the x bond can be used. Similarly, 4dx, can back-bondto only half of the n* orbital. Thus, at a very naive level, it might be arguable tha70 GENERAL DISCUSSIONa side complex should be more stable than a top complex, provided the silver ion isnear enough to the olefin bond. Clearly steric requirements will be of great import-ance. The replacement of just one hydrogen by a methyl group may mean that thesilver ion cannot come as close as is necessary for the side complex to be more stablethan the top complex: such steric arguments would not apply to top complexes,(which is why some modification to the original Dewar proposal would seemnecessary to explain the differences in stability between cis and trans complexes).Thus, it is proposed that ethylene, propylene, 1-enes and cis-olefins could form themore stable side complex whilst other olefins would form the less stable top complex