GENERAL DISCUSSIONProf. N, N. Greenwood (Newcastle upon Tyne) said: I would ask Perkins to whatextent is d orbital participation considered to be essential in stabilizing dimericAl,Me, on this model, and can the absence of such participation in BMe, be takenas an important factor in determining its inability to form a stable dimer?Dr. I. H. Hillier (University of Manchester) said : Could Perkins comment on thereliability of the CNDO method applied to the molecules considered in his paper, andon the 3d orbital exponent which should be used in such calculations ?Prof. P. G, Perkins (University of Strathclyde) said: In reply to Greenwood, wehave not yet analyzed the corresponding results for organoboron compounds butbelieve that the lack of participation of the 3d orbitals may well be the factor whichmilitates against formation of compounds such as B,Me,.In reply to Mrs.Green, one should indeed include further bases such as the 4s and4p atomic orbitals. However, the whole calculation is subject to a variational pro-cedure and it is the extent to which a set of new functions is likely to contribute whichdetermines their choice.In reply to Hillier and Urch, the extent to which the d orbitals are included isdetermined by the orbital exponent Z* and this should be determined during thecalculation by minimizing the energy with respect to it. However, such a procedurewill be very difficult in the CNDO framework and the present work therefore, as afirst approximation, takes two fixed cases. Each compound should be treated onits merits.In reply to Hill, the orbital exponent Z* which emerges from a variational calcula-tion need not have any precise physical meaning because it may relate to non-integralnumbers of electrons.This is true for simple systems (e.g., He atom) and so Z* isbest regarded as simply a variable parameter defining the 3d radial function.In reply to Hillier, in order to try and assess the reliability of the CNDO methodwe have made some comparison between it and an extended basis set ab initio calcula-tion on BH2 NH2. We find that the charge distribution is well reproduced by theCNDO method, and the eigenvalues are in satisfactory agreement. We believe atpresent that the latter method should be reasonably reliable for differential calculationswhere gross structural changes are involved.Prof.Theodore L. Brown (University of Illinois, Urbana) said: L. M. Ludwick inour laboratories has recently observed that the presumed 5-coordinate methylcobaltbis-dimethylglyoxime is in fact a dimer in methylene chloride. Dimerization probablyoccurs through the interaction of an oxygen from one molecule with the axial positionof the cobalt atom in a second species. Association in this manner is not uncommon.We have also been examining the kinetics of exchange reactions involvingreplacement of the axial ligand in the methylcobalt bisdimethylglyoxime system.Our evidence is not as yet extensive, but there is a suggestion in our results that theability of the axial ligand to act as a pi acceptor towards cobalt may be an importantfactor in coordination at this position.For example, triethylphosphite, whenbound in this axial position, does not exchange with excess ligand in methylenechloride, as evidence in the n.m.r. spectrum, at temperatures as high as +60°C.19200 GENERAL DISCUSSIONBy contrast, nitrogen and oxygen donors typically exchange much more rapidly,with coalescence temperatures of - 20 to - 50°C.Dr. H. A. 0. Hill (Oxford University) said: The suggestion made in the paper byDuncan et al. that the field metal d-orbitals interact directly with the 1s-orbitals ofthe methyl hydrogens may be most important. What effect would an increasingelectron density on a neighbouring metal atom acting directly on the methyl hydrogenshave on the JH--H coupling constants?Prof.Theodore I;. Brown (University of Illinois, Urbana) said: We have recentlyobserved the 35Cl quadrupole resonance spectrum of Zeise's salt. Absorptionswere observed, at 25"C, at 15.9, 20.10 and 20.37 MHz. The latter two absorptionsare ascribed to the cis chlorines, which are crystallographically distinct. The lowerfrequency absorption is assigned to the trans chlorine.The difference in quadrupole resonance frequencies of the two types of chlorineis substantial, and suggests that ethylene does indeed exert a trans influence insofaras this particular spectroscopic property is concerned. Chlorine quadrupoleresonance frequencies can generally be discussed satisfactorily in terms of p orbitalpopulations.The quadrupole resonance frequency is a measure of the differencein populations of the p,, orbital and the p x orbitals. In the present case, assumingthat the asymmetry parameter is small, the lower frequency for the trans chlorinecan be explained in terms of a more ionic Pt-Cl bond to the trans chlorine, as aresult of a strong c polarization by ethylene. There may, however, be an additionalfactor which would decrease the population difference in the p,, and pTc orbitals,namely, a n-donor action on the part of the trans chlorine toward platinum, resultingfrom the x-acceptor behaviour of ethylene. I suspect that this is a small effect, andthat most of the frequency difference must be ascribed to polarization in the bondsystem. The results are consistent with the observation of Denning regarding thePt-C1 stretching frequencies in Zeise's salt.Dr.M. G. Clark (University of Cambridge) said: An approach of the type usedby Green, Smith, and Tasker may be employed to study the stability of a 4-coordinatedsquare-planar system against the addition of two-electron donating ligands in eitheror both of the axial positions. The increase in coordination number is assumed tooccur without change in spin-state or equatorial bond length.Consider the effect of adding one or two axial ligands for each realistic orderingof the valence orbitals in a square-planar complex. We may work with the pointgroup C4u appropriate to the addition of one ligand, since the group D4h for theaddition of two is related to C4" by the addition of a centre of symmetry. Theaddition of either one or two ligands may be treated simultaneously, since in thelatter case only the centro-symmetric axial-ligand symmetry-orbital has correctsymmetry to interact with the orbitals of interest in the complex.The generalfeatures of the appropriate energy level diagrams are sketched in fig. 1, using CaUnotation. The diagrams may be considered as fragments of more complicatedmolecular-orbital diagrams, but since we require only qualitative features of theenergy levels they are adequate for our purpose, provided that it is borne in mind thatthey do not show all the interactions between levels explicitly. The only assumptionmade is that in the enlarged complex the antibonding a, orbital is more antibondingthan the bonding a, orbital is bonding.For a complex of a metal ion having configuration d" there are n+2 electronsto be assigned to the levels shown in the appropriate diagram. Table 1 shows foGENERAL DISCUSSION 201b*ale\\?\\ / - \ -FIG.1.TABLE 1ion n i c e e<ullow-spin d6 *high-spin d6 Ulow-spin d7 * *high-spin d7 Ulow-spin d8 S Shigh-spin d8 * .Ed9 s ss = stable against addition of ligands ; u = unstable against addition of ligands ; * = may or mayak**not be unstable depending on precise energetics.6<n<9 the results of considering whether or not the initially square-planar systemprofits energetically by taking on extra ligands under our assumptions. Not sur-prisingly, the results are indecisive in several cases ; but two definite points do emerge :(1) d9 and low-spin d8 ions are the obvious candidates for forming stable 4-coordinatedsquare-planar complexes, in agreement with observation ; (2) any truly 4coordinatedsquare-planar complexes of high-spin d6 and d7 ions should have al<e.Therarity of suitable examples makes the second prediction difficult to test; certainlythe mineral gillespite (BaFeSi,O, o), in which high-spin Fe2+ is square-planar co-ordinated by four oxygens, has a, <e.lThe results obtained above should be treated with considerable caution, since,apart from the simplicity of the arguments used, it is not obvious that other factorsR. G. Burns, M. G. Clark and A. J. Stone, Inorg.Chem., 1966,5,1268 ; M. G. Clark and R. G.Burns, J. Chem. SOC. A , 1967, 1034; M. G. Clark, G. M. Bancroft and A. J. Stone, J. Chem.Phys., 1967, 47, 4250202 GENERAL DISCUSSIONare negligible. For example, the iron in gillespite may have its coordination geometryto some extent forced upon it by the layer structure of the silicate framework.Dr. H. A. 0. Hill (Oxford University) said: The apparent reluctance of cobaltamben derivatives to give rise to alkyl cobalt(II1) derivatives may be caused by reasonsother than those mentioned in the paper by Greene et al. Except for ligand (V), onlyin the amben complexes will the nature of the coordinated nitrogen interfere with theplanarity of ligand when complexed. The steric requirements of the NH hydrogensmay cause a distortion of the complex such that in the cobalt(I1) derivative the struc-ture is intermediate between square-planar and tetrahedral.Even if formed, analkyl cobalt(l1l) derivative may be kinetically unstable. Though n-propylcobalaminis stable with respect to decomposition in solution, iso-propylcobalamin decomposesby homolytic fission of the cobalt-alkyl bond. This instability is presumably causedby steric interactions of the isopropyl ligand with the ring. Therefore it is possiblethat the alkyl cobalt amben complexes are unstable, or difficult to isolate, because ofsteric interaction of the alyl ligand with the non-planar amben ligand.Dr. Michael Green (University of York) said : In reply to Hill's remark about theplanarity of cobalt amben, conjugated '' N,-type " ligands are remarkable for theirtendency to form planar complexes even when geometric factors dictate otherwise.Weber has found that complexes of IX are planar not only when R is -(CH&--and -(CH2)3- but also when it is -(CH2),- in which case a tetrahedral configura-tion would be preferred stericalIy.We have found no significant variation either inthe visible/ultra-violet spectra or the magnetic properties of the cobalt ambenfamily (VIII), when R is -4CH2)2-, --(CH2)3 or --(CH&--. If a modification ingeometry occurred, the first complex would be expected to be planar and the lasttetrahedral. (The corresponding compounds from the cobalt salen family (VII) showdifferences in spectroscopic and magnetic properties compatible with such a modi-fication).Moreover, there is no notable difference between cobalt amben (VIII,R = -CH,CH,-), cobalt ambphen (VIII, R = 1,2-phenylene) and the cobaltomp pound,^ (X). It is difficult to see how these last two compounds could be dis-torted tetrahedrally.In view of this strong tendency for " N4 " systems to be planar and in view of theirgreater field strength compared with " Nz02 " ligands, any distortion in an alkylatedamben compound would be small and be most likely to involve the cobalt atom movingslightly out of the plane of the ligand which would remain flat as for CH,CoBAE whichHill mentions.Dr. H. A. 0. Hill (Oxford University) said: We have measured the e.p.r. spectraof the mono- andlor dipyridine complexes of the cobalt(I1) derivatives of ligandsJ.H. Weber, Inorg. Chem., 1967, 6,258.M. Hariharan and F. L. Urbach, Imrg. Chem., 1969,8, 556 and references therein.M. Green and P. A. Tasker, Chem. Cum., 1968,518GENERAL DISCUSSION 203I,l 11,2 131,l IV,3 VI,2 and VIII,2 in the paper by Greene et al. and in every case thespectra are interpretable in terms of a low-spin d7 complex in which the unpairedelectron is well-described by the d,Z orbital corresponding to position C and D in fig. 1.It would be of interest to examine the e.p.r. spectra of cobalt(I1) amben derivatives.Dr. Michael Green (University of York) (communicated) : Hill's data for (VII) areinteresting as they imply a reversal in the order of the dz2 and dx,, orbitals in going fromnickel salen to the cobalt salen/pyridine system.The presence of an axial ligandwill tend to cause changes A+B+C-+D, as we mention in our paper. As cobaltamben is reluctant to add pyridine as a ligand it will have a greater chance of being ofA-type than cobalt salen in pyridine. I agree the e.p.r. spectrum will be interesting.Dr. H. A. 0. Hill (Oxford University) said: A recent determination of the structureof CH3CoBAE, which has just been communicated to us prior to publication byProf. Rondaccio, University of Trieste, Italy,4 shows that the cobalt is indeed j u e -coordinate, with the BAE ligand planar and the cobalt lying 0.16 A above the plane ofthe BAE ligand. Thus, the suggestion made in our paper that in the five-coordintealkyl Co(II1) complexes, the cobalt atom lies above the plane of the ligands is confirmedin this case at least.Such a distribution may have important consequences forthe enzymatic role of the five-coordinate complexes. The importance may lie, notin the vacant coordination site trans- to the alkyl ligand but rather, that in the five-coordinate complex, the cobalt and its coordinated alkyl ligand lie out of the planeof the corrin ligand, making both the cobalt and the attached carbon atom moreaccessible for reaction with the substrate.Dr. J. F. Ogilvie and Dr. M. J. Newlands (Memorial University of Newfoundland,Canada) said : One of the most important applications of the theory of bonding inmetallo-organic compounds must be to elucidate the binding to metals of reactants,molecular intermediates and products in reactions which are catalyzed by materialscontaining metals, especially transition metals.A process of some current interestis the fixation of nitrogen in nature, in which molybdenum, perhaps cobalt, and parti-cularly iron are believed to be invol~ed.~ Inorganic nitrogen complexes of iron havealready been prepared.6* A commonly postulated intermediate in the ammonia-producing process is di-imide.5 However, di-imide, HN=NH, has been shown toinhibit fixation of nitrogen by, for instance, Clostridium pasteurianum.In connection with our consideration of multiple bonding in silicon and germaniumimine derivatives * we have been led to suggest that an imine intermediate, M=NHHN-M, rather than M . .. HN-NH . . . M, containing nitrogen atoms stronglybound to the appropriate metal atom M held rigidly in some location by some co-ordinating molecule such as a porphin type of chelating agent. An arylimino complexof rhenium has been prepared which has a relatively short lo strong bond betweennitrogen and rhenium atoms, indicating that such a strongly covalent linkage isindeed possible for transition metals.Cockle, Hill, Pratt and Williams, Biochern. Biophys. Acta, 1969, 177, 686.Hill, Morallee and Pellizer, to be published.Hill, MacFarlane and Williams to be published.S. Brucker, M. Calligaris, G. Nardin and L. Randaccio, to be published.R. Murray and D. C. Smith, Courd. Chem. Rev., 1968,3,429.A. Sacco and M. Aresta, Chem. Comm., 1968, 1223.' G. M. Bancroft, M. J. Mays and B. E. Prater, this discussion. * J. F. Ogilvie and M. J. Newlands, Trans. Farday SOC., in press.J. Chatt, J. D., Garforth, N. P. Johnson and G. A. Rowe, J. Chem. SOC., 1964, 1012.lo D. Bright and J. A. Ibers, Inorg. Chern. 1968,1,1099204 GENERAL DISCUSSIONProf. N. N. Greenwood (Newcastle upon Tyne) said: With regard to Bryce Smith’spaper, what is the origin of the antiferromagnetism in the dl0 compounds formulatedas Ag*Br and Ag*I? If there are, indeed, small domains of metal-metal clusterswould it not be expected that the compounds would be superparamagnetic ratherthan antiferromagnetic since the volume exchange interaction energy is known toapproach thermal energies for particles smaller than - 100 A and so magnetic orderingcannot occur ?Prof.D. Bryce-Smith (University of Reading) said. In reply to Greenwood, Ishould state fist of all that we have used the term “ antiferrornagnetism ” for con-venience because it presently seems to provide the closest description of the observedmagnetic behaviour; but we accept that the behaviour is not classically antiferro-magnetic. Since Ag‘ is isoelectronic with PdO, we agree that Agl clusters might atfirst sight exhibit paramagnetic properties similar to those of palladium metal. Thedifferent behaviour observed might be explained in several ways, one of which is thatwe could be dealing with sublattices of Ago and Ag”.Small domains of metal clusters would lead to superparamagnetism only if thesewere individually ferro- or ferrimagnetic.Antiferromagnetic domains would notresult in superparamagnetism. Tt is not certain whether metamagnetic domains,i.e., domains in which both ferro- and antiferromagnetic interactions occur, wouldresult in a form of superparamagnetism or what the bulk magnetic properties of asubstance containing such domains would be. Certainly, the forms of the curvesshown in fig. 5 and 6 are inconsistent wih normal superparamagnetic behaviour.Dr. M. G. Clark (University of Cambridge) said: I would ask Bryce-Smith if it ispossible that the new complexes may have polymeric or macromolecular structureswith the carbon suboxide, bromide, or iodide acting as bridging ligands. This wouldhelp to explain their solid-state magnetic properties and lack of crystalline form,but difficulties might be met with their solution properties and the new form of silver“Ag* ”-Prof. D. Bryce-Smith (University of Reading) said. In reply to Clark, the greatinsolubility of the silver complexes is certainly consistent with polymeric structures.The magnetic properties of the suboxide complexes, in so far as a comparison can bemade, appear to parallel those of the derived iodides, a fact which suggests that theseproperties arise from a common structural feature involving the silver atoms ratherthan the ligands. For this reason, we believe that super exchange does not contributein a major way to the magnetic properties, although we do not exclude the possibilitythat C302 units can chemically link cluster centres