1973 2381A Study of Bonding in Some Organoaluminium Compounds by 27AINuclear Quadrupole Resonance SpectroscopyBy Michael J. S. Dewar," Dennis B. Patterson, and W. lrven Simpson, Department of Chemistry, TheUniversity of Texas at Austin, Austin, Texas 7871 2, U.S.A.27AI N.q.r. spectra have been measured for a wide range of monomeric and dimeric aluminium complexes of thetype Me,AIX and ( R1R2AIX),. Measurements are also reported for But,Al, (Me,AIOMe),, and (Me,AISMe), andfor 69Ga in trirnethylgallium. The results are interpreted in terms of the simple treatment of Townes and Daileywhich is shown to account well for all the observed trends. 35CI, 75As, 121Sb, and lZ3Sb n.q.r. spectra for theappropriate complexes are also reported. The potential of n.q.r. spectroscopy as a tool in structure determinationis stressed.NUCLEAR quadrupole resonance (n.q.r,) spectroscopy 1provides a uniquely useful procedure for studyingchemical bonding since the frequencies observed forcovalent compounds of non-transition elements dependentirely on the distribution of $-electrons in the valenceshell of the atom in question.Measurements of thiskind are of especial value in the study of chemicalbonding in molecules and much useful information hasbeen obtained in this way.lAluminium is an almost ideal subject for threereasons. First, naturally occurring aluminium consistsof a single isotope, 27Al. Secondly, this isotope has ahigh nuclear spin ($) allowing two independent transitions(j-4 f--). &- and &Q &$) to be observed; ffrom these one can calculate not only the nuclearquadrupole coupling constant (e2Qq/h) but also theasymmetry parameter (7).Thirdly, aluminium is avery versatile element, occurring in a variety of differentbonding situations. It forms salts and it also occurs ina variety of covalent compounds including some (e.g.,A1,Me6) that are ' electron deficient ' and contain three-centre bonds.*Preliminary studies 5 9 6 of a variety of organoaluminiumcompounds had established the potentialities of such anapproach. We were able to show5 that the originalformulation of hexamethyldialuminium by Longuet-Higgins4 is correct in contrast to an alternativerecently suggested by Nyburg et aL7 and the structuresof several mixed alkylaluininium halides were estab-lished in this way.Here we report a detailed study off In certain circumstances the transition -J-+ .f--t 5% canalso be observed. However the corresponding frequency isnecessarily the sum of those for the other two transitions.4 In view of an apparent misunderstanding * of our previouscommunication 6 we point out that the diagram reproduced therewas intended to imply only the type of bonding suggested byn'yburg et aZ.,' not the precise molecular geometry. Perhaps abetter representation would have been (I). This bears the samerelation to the Longuet-Higgins structure that edge-protonatedcyclopropane does to corner-protonated cyclopropane. RecentworkQ has further confirmed the correctness of the Lonauet-a number of organoaluminium compounds of variouskinds.EXPERIMENTALAll the compounds studied were highly sensitive both toair and moisture.They were handled under nitrogen in adry box (protected with sodium-potassium alloy), glovebags, Schlenk-type glassware, and standard vacuum tech-niques. Liquids and low-melting solids were finally puri-fied by degassing in a vacuum followed by vacuum transferTABLE 1Preparation of materialsLiterature ref.C. A. Smith and M. G. H. Wallbridge, J .N. Davidson and H. C. Brown, J . Amer. Chem.C. H. Hendrickson and D. P. Eyman, Inorg.S. Takeda and R. Tarao, J . Chem. SOC. Japan,J. L. Atwood and G. D. Stucky, J . Amer.H. Lehmkuhl, Ann. Chew., 1968, 719, 40.A. V. Grosse and J. M. Mavity, J . Org. Chem.,E.G. Hoffmann, Awn. Chem., 1960, 629, 104.T. Mole and J. R. Surtees, Austral. J . Chem.,W. C. Kaska, Ph.D. Thesis, University ofT. Mole and J. R. Surtees, Chem. and Ind.,T. Mole, Austral. J . Chem., 1966, 19, 373.D. G. Brauer and G. D. Stucky, J . Amev.Chem. SOC. ( A ) , 1970, 2675.SOL, 1942, 64, 316.Chem., 1967, 6, 1461.1965, 38, 1567.Chem. SOC., 1967, 89, 5362.1940, 5, 106.1964, 17, 310.Michigan, 1963.1963, 1727.Chew. SOC., 1969, 91, 5462.to ampoules which were sealed and used without openingin the n.q.r. studies. The remaining solids were sublimedor recrystallized as recommended in the literature (Table 1)and then sealed in ampoules.See E. A. C. Lucken, 'Nuclear Quadrupole CouplingConstants,' Academic Press, New York, 1969.M.H. Cohen, Phys. Rev., 1954, 96, 1278.R. Livingston and H. Zeldes, ' Table of Eigenvalues forPure Quadrupole Spectra, Spin +,' Oak Ridge National Labora-tory, No. 1913, 1955.H. C. Longuet-Higgins, J . Chew SOC., 1946, 139.M. J. S. Dewar and D. B. Patterson, Chem. Comm., 1970,544. * M. J. S. Dewar, D. B. Patterson, and W. I. Simpson, J . Amer.S. I<. Byram, J. K. Fawcett, S. C. Nyburg, and R. J. O'Brien,Chem. SOC., 1971, 93, 1030.Chem. Comm., 1970, 16.v * F. A. Cotton, Inorg. Chem., 1970, 12, 2804.a J. C. Huffman and W. E. Streib, Chem. Comm., 1971, 911. Higgins structureJ.C.S. DaltonThe simple complexes ( 1) of trimethylaluminium weremade by adding a slight excess of the ligand to trimethyl-aluminium (Alfa Inorganics) , mixing thoroughly, andremoving excess of ligand by vacuum distillation.Complex(la) does not seem to have been reported. The structurefollows from analogy and from its n.q.r. spectrum whichshowed it to contain both aluminium and antimony.The remaining compounds (4b-f), (5b), (5c), (6), (9b),and (9c) were donated by Texas Alkyls, Pasadena, Texas,in 25 mm 0.d. septum-sealed ampoules and were usedwithout purification except for (6) which was distilled onthe vacuum line.The n.q.r. frequencies were measured with a coherency-controlled super-regenerative spectrometer following thedesign of Peterson and Bridenbaugh lo but modified tooperate with maximum sensitivity over the range 3-50MHz. Switches allow the selection of six cathode chokesand five cathode resistors and a choice of feedbackcapacitors so that the lowest possible value can be used ineach frequency range.These modifications distorted theshape of the signal but greatly improved the sensitivity.For the lowest frequency range the time constant wasincreased to 100 s and the plate voltage raised. Fre-quencies were measured with a model 5245 L Hewlett-Packard Electronic Counter. Since a super-regenerativereceiver was used, the signals were accompanied by side-bands. The central peak was determined in the usual wayby varying the quench frequency. In the case of compound(loa), the 27Al signal was so weak that it is possible that thewrong peak may have been selected. The accuracy of theother 27Al frequencies was limited by the difficulty ofdetermining the centre of the peak because of the longtime constant that had to be used.Similar difficulties alsoarose in the case of Me,As and Me,Sb. The signals were inall cases reproducible within the limits of error of therecorder.A check on the reliability of our assignments was pro-vided in several cases by observation of the && f--f +*transitions. The frequencies for these were in each casethe sum of those already estimated for the &+ &$and &$ - j g transitions.Interpretation of N.q.r. Spectra.-Theoretical interpre-tations of n.q.r. spectra can be carried out at various levelsof sophistication 1 ranging from ab initio calculations to aninterpretation based on simple-minded MO theory.If weare concerned with the possible practical value of n.q.r.spectroscopy as an aid to the understanding of structureand reactivity in organic molecules, the latter is clearly themore valuable since i t ties in with the conventional re-presentation of molecules in terms of a simple MO approach.We will therefore discuss our results in terms of the treat-ment suggested by Townes and Dailey l311 in which the netelectric field gradient at the nucleus is attributed entirelyto unequal distribution of the $-electrons in the valenceshell, the populations of the corresponding p-AOs beingestimated by simple arguments. In the present connectionwe need the results for molecules of the type MX,Y, wherethe central atom (M) forms identical bonds to three groupsX and where the M y bond lies on a threefold axis ofsymmetry (e.g., CHCl,), and for molecules of the typeMX,Y, with tetrahedral geometries, the groups MX, andMY, lying in orthogonal planes (Figure).10 G.E. Peterson and P. M. Bridenbaugh, Rev. Sci. I.tzstv., 1964,35, 698; 1965, 36, 702.In the first case the asymmetry parameter is zero whilethe coupling constant (e2Qq) is given in terms of that(e2Qqo) for a single $-electron by equation (1) where a is theXMX bond angle, a is the population of an A 0 of M used toform an MX bond, and b the population of the A0 of NIused to form the MY bond.In the second case, i.e., tetrahedral molecules of the typeMX2Y2 (Figure), the usual assumptions concerning hybridiz-ation and orthogonality lead t o the relation (2) where 8 and(2)02 2cot2 - + cot2 + = 1$ are the XMX and YMY angles respectively (Figure).Itcan be shown that one of the three principal axes of the2y\ izII( b )(a) Principal axes for MX,Y,; (b) relation to a bimolecularaluminium complexfield gradient tensor lies along the intersection of the NIX,and MY, planes while the other two lie at right angles toit, in the MX, and M Y , planes respectively. If 8 < $, theaxis of maximum field gradient (2) lies in the MY, planeand that ( y ) of minimum field gradient in the MX, plane.The third axis, defined by the intersection of the planes, isthe axis ( x ) of intermediate field gradient [Figure, (a)].The coupling constant and asymmetry parameter (q) aregiven by equations (3) and (4) where c and d are the popu-(3)= -3COSO (4lations of the MX and MY AOs of M respectively.Notethat q = 1 when the central atom (M) is tetrahedral.In the case of bimolecular aluminium complexes of thetype (11) the principal axes are as indicated in the Figure, (b).Note that equations (l), (3), and (4) are derived on theassumption that hybrid orbitals follow ’ the correspondingbonds. In the case of bent (‘ banana ’) bonds, the anglesin equations (l), (3), and (4) refer to angles between thecorresponding AOs of M.In order to apply these relations, it is necessary to know(e2Qq0). The ground state of the aluminium atom has theconfiguration (1~)~(2~)~(2P)~(3~)2(3p) with a single 3p-electron. It therefore seems natural to equate the ob-11 See T.P. Das and E. L. Hahn, Nuclear QuadrupoleResonance Spectroscopy,’ ‘Solid State Physics,’ eds. F. Seitz andD. Turnbull, suppl. no. 1, Academic Press, New York, 19681973 2383served l2 quadrupole coupling constant for free aluminiumatoms to (e2Q40), as in equation ( 5 ) . This value rests of(ez&o)(e7Al) = 37.52 MHz (5)course on the assumption that the value for a free atom canalso be used for molecules (see below).RESULTS AND DISCUSSIONCoupling Constants for Monomeric Complexes.--Thefirst series of compounds to be considered contain asingle aluminium atom, being complexes (1) of trimethyl-aluminium with various donors. The results are inTable 2. Values are also included for tri-t-butyl-aluminium (2) and trimethylgallium (3), both of whichcases including some (e.g., X = AsMe,) where the ligandis symmetrical.The VaEue of e2Qqo.-The limiting member (2) of theseries is certainly planar.If hyperconjugation isneglected, a = 120" and b = 0 in equation (1). Sub-stituting the observed values for e2Qq [45.51 MHz;Table 21 and (e2Qq0) C37.52 MHz; equation (5)] we findequation (6). This result is not reasonable. It impliesthat the A1-C bonds are polarized in the sense Ala--Ca+.a = 1.21 (6)A similar anomaly is seen in trimethylgallium (3) wherethe observed coupling constant (162.10 MHz) is againmuch greater than the value for the free atom (125.04TABLE 2N.q.r. parameters (27Al) for trimethylaluminium complexes and tri-t-butylaluminium and (69Ga) for trimethylgalliumEstimatedpossibleObserved frequency/MHza error e2Qq AH?Compound (&$-a-+ &*) A+) (&MHz) (MHz) -4 kcal mol-'4.84 9.63 0.025 32.1 3 0.0624-29 8.525 0.01 28.45 0.0713.57 7-08 0.01 23.60 0.074 21.02 0.28"- 3.46 0.05 11.52 (0) d 29-96 f 0-19 C4.48 8.91 0,025 29.75 0.061 16.69 & 0.1884.56 9-03 0.01 30.12 0.093 16-95 f 0.1864.25 8.45 0.01 28.18 0.068 20.29 f 0-20*4-06 8.03 0.01 26-82 0.086 22.90 & 0.1964.63 9.25 0.01 30.83 0.026 -6.85 13.61 0.01 45-41 0-071 I(14(1b)(lc)(14(W( I f )(W(W( l i )(2)(3) 0.05 162.10 (0) * 81.05 -a At 77 K.Errors are estimated maximum errors. Heat of formation of the complex Me,Al-X from Me,Al f X in hexane.8 Ref. 16d. 1 Assumed since I = $?; "Ga signal c Ref.16c. d Assumed since second signal outside the range of spectrometer.observed a t correct ratio.are monomeric.13 Apart from minor differences betweenA1-CH, and Al-CMe, bonds, (2) can be regarded as thelimit of (1) in which the donor (X) vanishes.MHz 14), The situation is indeed even worse becausehyperconjugation must be important 15a9b in (Z), leadingto a transfer of electrons into the ' empty ' 3p-AO ofaluminium and so making b > 0. Thus either the MOtreatment used here is incorrect or the value for (e2Qqo)is not the same for a free atom as for an atom in aMe,AI f-- X Bu$Al Me,Ga(1) (2) (3)( a ) X = SbMe3( b ) X = AsMe3( c ) X = PMe3[ d l X = NMe3( e ) X = SMe2I f ) X S ( C H Z ) L( g ) X = O M e 2( h ) X = O ( C H ~ ) LAll these compounds conform to the conditionsassumed in deriving equation (1) since any differencesbetween the three A1-CH, bonds, due to asymmetry ofthe ligand, must be small.The measured asymmetryparameters do admittedly differ from zero; this, how-ever, can be attributed l to crystal-field effects, particu-larly since the values of y1 are much the same in alll2 H. Law and G. Wessel, Plays. Rev., 1953, 90, 1.l3 For evidence that compound (2) is monomeric, see H.l4 R. T. Daly, jun., and J. H. Holloway, Phys. Rev., 1954, 96,Lehmkuhl, A m . Chsm., 1968, 719, 40.539.molecule.There are in fact good reasons for believing that the'free atom' value for (e2Qqoo) must be too small. De-tailed ab ivtitio SCF calculations have shown that in anLCAO treatment of molecules one must use AOs thatare smaller than those appropriate to free atoms.15cSince the field gradient at the nucleus varies as theinverse cube of the linear dimensions of an A 0 and soas the cube of the effective nuclear charge, the value of(e2Qqo) to be used in equations (1) and (3) must be muchlarger than the ' free atom ' value [cf.equation @)I.If the orbital populations in compound (2) wereknown, (e2Qq0) could then be found from the measuredcoupling constant by using equation (1). Unfortunatelythere seem to be neither reliable theoretical estimates ofthe orbital populations nor values for the dipole momentof the C-A1 bond or the electronegativity of aluminium.l6 (a) K. A. Levison and P.G. Perkins, Discuss. Faraday Soc.,1969, 183; (b) K. A. Levison and P. G. Perkins, Theoret. Ckim.A&, 1970, 17, 15; (c) W. J. Hehre and J. A. Pople, J . Amer.Chern. Soc., 1970, 92, 2191J.C.S. DaltonHowever attempts to use equation (1) in a quantitativesense would probably in any case be a waste of timesince the treatment on which it is based is so approximateand since the sizes of AOs are further affected by theformal charge at the atom in question [so that (e2Qq0)varies correspondingly]. One can, therefore, useequation (1) only in a qualitative way to interpret trendsin observed coupling constants.Efect of the Ligand-On passing from compound (2)to (l), the 27Al coupling constant should decrease for tworeasons. First, the inter-bond angle cc decreases withformation of a bond between aluminium and the donor;secondly, formation of this bond raises the population[b in equation (l)] of the fourth aluminium AO.Botheffects should be greater, the stronger the bond betweenA1 and X in ( l ) , i.e., the better the group X is as a donor.In the formation of (la-d) the donor has only a singlepair of unshared valence electrons. The A1-X bonds inthese compounds are therefore a-bonds, analogous to theH-X bond in the conjugate acid HXi. The strength ofthe A1-X bond in compounds (la-d) should thereforebe greater, and the 27Al coupling constant correspond-ingly less, the greater the basicity of X. There is infact a large and progressive decrease in coupling constantin the series ( 2 ) > (la) > (lb) > (lc) > (Id), corre-sponding to the observed order of basicity in the seriesSbMe, > AsMe, > PMe, > NMe,.The increasingdegree of binding is also reflected by the thermodynamicstabilities of the complexes from AlMe, and X. This isknown l6 to increase in the series (lb) < (lc) < (Id) andthe antimony derivative (la) is unlikely to be out of step.The sulphur complexes (le and l f ) also seem to fit thesame pattern, judged by the correspondence betweencoupling constant and heat of complex formation shownby the data in Table 2. Thus the differences betweenthe coupling constants and heats of complex formationfor the trimethylamine (Id) and trimethylphosphine (lc)derivatives corresponds to a change in coupling constantof 1.37 MHz per kcal change in the heat of complexformation while the corresponding ratio for the trimethyl-amine (Id) and dimethyl sulphide (le) complexes is1.41 MHz kcal-l.Dimethyl sulphide is of course anextremely weak base.The A1-0 Bond.-The results for the ether complexes(lq-li) are out of step with those for the other ligands.Thus although ethers are far stronger bases than are thecorresponding sulphides, the coupling constants for theether complexes are only slightly less. Thus the heat offormation of (lq) is similar to that for the trimethyl-phosphine complex (Ic) but the coupling constant isalmost the same as for (lb).This discrepancy is presumably due to the samefactors that make A1-0 bonds in general anomalouslystrong.Our results indicate that the additionalstrength is not due to strong D-bonding. It seems likelyl6 N. Davidson and H. C. Brown, J . Amer. Chem. SOC., 1942,64, 316; C. A. Smith and M. G. H. Wallbridge, J . Chem. SOC.( A ) , 1970, 2675; C. H. Hendrickson and D. P. Eyman, Inorg.Chem., 1967, 6, 1461; C. H. Hendrickson, D. Duffy, and D. P.Eyman, Inovg. Chem., 1968, 7, 1047.that the additional pair of unshared oxygen electronsmay play a role, leading to p,,-d, bonding. The transferof charge from oxygen to aluminium by n-bonding wouldof course tend to reduce the corresponding transfer by0 bonding. Since only the latter contributes to changesin the 27Al coupling constant, n-bonding should tend toincrease both the heat of complex formation and thecoupling constant.This interpretation seems to be supported by thebehaviour of gallium where the bond to oxygen is notabnormally strong.Tong17 has found a fairly goodlinear correlation between the heats of complex form-ation and n.q.r. coupling constants for a series of com-plexes formed by gallium trichloride with nitrogen,oxygen, and sulphur bases. Further support also seemsto be provided by the crystal structure l8 of the dioxancomplex (li). Normally one might expect the oxygenatom in such a complex to have pyramidal geometrybut x-bonding to the acceptor should tend to make theoxygen atom planar. Although the oxygen atoms in(li) are not planar, the A1-O-C bond angles are large(122"). Indeed, the departure from planarity may wellbe due to the constraints imposed by the ring, the COCangles being only 108".A determination of the structureof the acyclic complex (lg) would be of interest in thisconnection.The High Basicity of Tetrahydrofitram-The tetra-hydrofuran complex (lh) is more stable than the di-methyl ether complex (lg) and the 27Al coupling constantis correspondingly less (by 1-3 MHz). These resultswould be expected since tetrahydrofuran is known tobe a stronger base than dimethyl ether. On the otherhand the coupling constants of the sulphur analogues(le) and (If) differ by an amount (0.3 MHz) comparablewith the ' noise ' due to crystal-field effects. The heatsof complex formation for (le) and (If) are also the samewithin the limits of experimental error.These resultssupport the current view that the high basicity of (111)is steric in origin, the increase in steric repulsion on saltformation being less for (lh) than for other ethers.Such steric effects should be less in the case of sulphidesboth because the sulphur atom is bigger than oxygen andbecause the CSC bond angles in sulphides are less thanthe COC bond angles in ethers.Note that the coupling constant for the dioxancomplex (li) is higher than that for (lg) by 2-65 MHz.Presumably (li), in which both oxygen atoms areattached to trimethylaluminium residues, is weakenedby the resulting dipole-dipole repulsion.Hybridization of A Zuminium and Ligand A&.-Wehave also measured the 75As, 121Sb, and lBSb n.q.r.spectra of compounds (la) and (lb), together with thoseof trime t hylarsine and trimet hyls tilbine for comparison.The results are shown in Table 3 together with thederived coupling constants.The observed frequenciescorrespond to values of the asymmetry parameter close17 D. A. Tong, Chern. Comm., 1969, 790.18 J. L. Atwood and G. D. Stucky, J . Awaer. Chem. SOL, 1967,89, 53621973 2385to zero (0.036 for SbRle,).two lowest signals could be detected, those being weak.In the case of (lb) only theTABLE 3‘;AS, lZ1Sb, and 123Sb parameters for AsMe,, SbMe,, andtheir A1Me3 complexes at 77 KCoupling Frequencies of transitions/MHz b constant.Compound &+-.+$ $+--++ Q-3 MHz- 193.4- 164.8SbMe, lZ1Sb 74.3 148.3 - 494-6lZ3Sb 45.2 89.9 135.1 630-3(lb) lZISb 66.4 (--) a - 442.0lZ3Sb 40.4 (-) a (-1 a 563.4a Signals not observed; see text. Estimated maximumAsMe, 96.7 -(la) 82.4 -possible error, 0.1 MHz.The n.q.r.frequency of monomeric trimethylalu-ininium is not of course known but it is unlikely that itcan differ much from that of tri(t-buty1)aluminium. OnE tE t(7)(10)(a) R = Me(b) R = PhSb or As n.q.r. frequencies. While no quantitativeconclusion can be drawn from these results, it seemsclear that the A0 used by aluminium in the A1-As orA1-Sb bonds must have much greater p-character thanthe As or Sb AOs. This in turn suggests that theMeAlMe bond angles in (la) and (lb) should be con-siderably larger than the tetrahedral value while theMeAsMe and MeSbMe angles should be correspondinglysmaller. It is interesting that the MeAlMe bond anglein (Id) is larger (11443°) than the tetrahedral va1~e.l~Coz@?ing Constants for Dimeric Complexes.-All buttwo of the remaining compounds studied were derivativesof dialane [6] including compounds with bridging chlorine(4), bromine (5), iodine (6), oxygen (7), nitrogen (S),alkyl (9), phenyl (lo), and phenylethynyl (1 1).We alsostudied the trimer (12) of dimethylaluminium methoxideand the polymer (13) of the corresponding sulphurcompound. The results for compounds (P-S), (la), and(13) are in Table 4 and those for (9)-(11) in Table 5.+(5)(a) R1 r; R2 = Me(c) R1 = R2 = Br(b) R1 = R2 = Et(9)(a) I< = Me(b) R = Et(c) R = Bu‘Me Me\ - /this basis one can see that the percentage change in the The symmetrical compounds in the series, with27Al coupling constant on formation of (la) or (lb) (37 formulae of the type (R,AlX), where X is the bridgingand 29% respectively) is very much greater than the group, conform to the conditions assumed in derivingcorresponding changes (15 and 11% respectively) in the equation (3).Since the XAlX angle is less than theChern. Scand., 1972, 26, 1947. the Figure (b). This assignment has been confirmed by 19 G. A. Anderson, F. R. Forgaard, and A. Haaland, Acts RAIR One> the principal be as indicated i2386 J.C.S. DaltonPeterson and Bridenbaugh 20 for the analogous dimer ofgallium chloride (Ga2C16) by a study of the n.q.r. Zeemaneffect in a single crystal.(13)Asymmetry Parameters : ' Bartana ' Bonding.-Theasymmetry parameters in Tables 4 and 5 are remarkablyangles in the four-membered rings which must of coursebe close to 90" (for which q = 0). Evidently theA1 AOs do not follow the bonds, the latter being bent' banana ' bonds like those in cyclopropane.2fThe asymmetry parameter of (8) changed slightlybetween 77 K and room temperature.This change isprobably a crystal-field effect, due to anisotropic thermalexpansion of the crystals.The external bond angles in compound (8) raise afurther point. The observed 22 value (115.9") is con-siderably greater than that (109.5') corresponding toTABLE 427Al N.q.r. parameters for polymeric aluminium complexesEstimatedpossiblee 2 0 qCompound T/I< ( & t Q - &+) (&Q -3~s) (AMHz) MHz - Observed frequency/lMHz error77 6.57 10.54 0.025 36.44 0.4577 6- 68 11-03 0-025 37-82 0.4277 6-59 10-79 0.025 37-18 0.4377 6.67 10.56 0.025 36-60 0.47196 6-57 10.44 0.025 36-16 0-4777 4.74 7.20 0.025 25.60 0.5177 4.83 7-70 0.05 26-66 0.46196' 4.66 7.00 0.05 24.48 0.53Room 4.66 6.94 0.05 24.32 0.5477 6.46 10.65 0.025 36-66 0.42196 6-42 10.58 0.025 36.41 0.4277 6-525 10.96 0.025 38.15 0.3777 3.034 3.83 0.01 13.86 c 0-7377 6-39 11.16 0.05 38-02 0.3477 6-01 6.01 6 0.01 22-71 1.0077 4.225 4-336 b 0.01 16.43 0.957Room 4.18 4.18 0.01 15.77 1.0077 4.25 5.89 0.025 20.91 0.624.56 6-76 0.025 30.81 0.73(4a)(4b)(44(4d)(44(4f)C4g)(54(5b)(54(6)(7)(8)(12)(13) 77 4.90 0-05 18-52 (1.00) dRoom 4.80 0.01 18.41 (1.00) d4.81 a 7.43 a 0.02511.35 0.025a Crystal-field splitting; mean value used in calculating e2Qq and 3.b (-Jt - ~ - - j 2%) transition also observed at expectedfrequency. c P. A. Cassbella, P. J. Bray, and R. G. Barnes, J. Chem. Phys., 1959, 30, 1393. d ,4ssumed; signal not resolved.TABLE 527Al N.q.r. parameters for alkyl-, phenyl-, and phenylethynyl-bridged dimersObserved frequency/MHzCompound TIK ( i z 4 t - t &%) (&%-&+) e2Qq r)77 5.38 6-51 b 23-71 0.78196 5.37 6.45 b 23-55 0.79196 5-62 6.2 23.23 0-8777 6.1 5 6.87 25-42 0.8777 6-78 24-27 (0-67) eRoom 4.93 6.54 23.41 0-67Room 4.53 5.39 19.71 0-79Room 6-74 6.74 25.49 1.00Q Estimated possible error, -+0.05 MHz, except for (1Oa) where the error could be as much as 0.1 MHz.(94(9b)(94( 104( 1 Ob)(11)b (-&+ A+)d Two unresolved signals of unequal intensity, in accordance with the crystal structure transition also observed.(ref.28a).e Assumed.large. Indeed compounds (7), (8), (ll), and (13) seemt o be the first non-ionic compounds for which q has beenfound to have the maximum possible value of unity.The corresponding interorbital angles, calculated fromequation (4), are much larger (109.5") than the XAlX20 G. E. Peterson and P. M. Bridenbaugh, J. Chem. Phys.,1969, 51, 235.the asymmetry parameter. If this discrepancy is takenseriously it would imply that the external bonds in (8)are also 'bent.' It seems more likely that, in orbital15y c.A. Coulson and W. E. Moffitt, J . Chew. Phys., 1947, 15,22 V. H. Hess, H. Hinderer, and S. Steinhauser, 2. anorg.Chem., 1970, 377, I.1973 2387terms, the AOs themselves are bent. The greater inter-electronic repulsions near the nucleus should tend toenforce a more nearly tetrahedral geometry in thatregion than in parts of the A0 more distant from thenucleus. The n.q.r. frequency is determined mainly bythe form of AOs near the nucleus while chemical bondingdepends on the outer parts of AOs. One should nottherefore expect any quantitative correspondence be-tween bond angles and interorbital angles deduced fromn.q.r., even in cases where strain is absent.Efleect of the Bridging Groups.-The dimers (R,AlX),fall into three categories. First, there are those [(4)-(S)] where X has an unshared pair of electrons so thatX and A1 are linked by normal covalent bonds.Secondly, there are compounds (9) with bridging alkylgroups which are linked to A1-X-A1 bonds. Thirdly,there are compounds [(lo) and (22)] with unsaturatedbridging groups where the x-electrons may participate inbridging.It is easily seen that in all cases the popu-lations of the aluminium AOs used in bridging should beless than those used in the terminal bonds. In com-pounds (4)-(8) the bridging atom carries a formalpositive charge and is therefore more electronegativethan the terminal atoms or groups while in (9) thebridging bonds involve sharing of pairs of electronsbetween three atoms instead of two.Moreover thebent ' banana ' bonds in the bridge must be weaker andso more polarizable than the terminal bonds; theelectronegative bridging atom should be Correspondinglymore successful in attracting electrons from aluminium.The compounds in Table 4 can be regarded as chelatedco-ordination complexes derived from the ion R,A1+which would be linear, corresponding to equations (3)and (4) with 8 = 90°, + = 180", d = 0. Co-ordinationwith the donor X introduces electrons into the previouslyempty aluminium AOs. The repulsion between theseelectrons should lead to an increase in 0 and a corre-sponding decrease in + [Figure and equation (2)]. Thenet effect [see equations (3) and (a)] in complexes of agiven type will be a decrease in the coupling constantand an increase in the asymmetry parameter, thesechanges being greater, the greater the donor activity ofX.The results in Table 4 are in agreement with thisprediction, the coupling constants decreasing in theorder (X =) : C1> RO > R2N. The asymmetry para-meter also increases, being in the range 0 . 4 4 - 5 formost of the compounds with halogen bridges but unity,or close to unity, for those with bridging oxygen ornitrogen. A value of unity for q implies sp3 hybridiz-ation of aluminium, 8 and (b having the tetrahedralvalue (109.5").Univalent and Bivalent Bridging Growps.--In com-pounds (4)-(8) the bridging atom has two AOs andfour electrons for bonding to aluminium. The A1-Xbonds are therefore normal two-centre covalent bonds.In the case of (9), however, the bridging group has oneA 0 only and is therefore linked to the aluminium atomsby a two-electron three-centre bond.Other things beingequal, one would then expect the electron densities in thealuminium AOs to be lower, and the coupling constantcorrespondingly greater, in the latter case. Thecoupling constants for (7) and (8) are indeed less thanthat for (9) but those for (4a), (5a), and (6) are muchgreater. Evidently the polarization of the Al-halogenbonds is so extreme that it outweighs the presence offour electrons rather than two. The Al-Cl bond iscertainly highly polar, as is shown by the very low 35Cln.q.r. frequencies (Table 6) in the correspondingcomplexes.TABLE 635Cl N.q.r. frequencies for alkylaluminium and phenyl-aluminium chloridesCompound (4b) (4c) (4d) (4e) (4f) (4g)Observed 9.89 9.97 10.18 10.46 10.82 10-48efrequency "/MHz 11.35b 11.72 11-47ca At 77 K unless otherwise stated.Estimated maximumpossible error, & 0.025 MHz. Crystal-field splitting. CI At196 K.11-505The results for compounds (7) and (9b) seem tocontradict the conclusions reached above concerning therelation between coupling constants and asymmetryparameters. Thus since the coupling constant for (7)is similar to that for (9b), one would expect the asym-metry parameters to be comparable. This is not thecase, the values of y for (7) and (9b) being 1-00 and 0.87,respectively. Yet even this small discrepancy can bereasonably explained in terms of the simple MO treat-ment.The interorbital angle near the nucleus dependson two factors; the angle between the orbitals as a whole,and the enhanced correlation effects near the nucleus.Bending a bond should therefore lead to correspondingchanges in the interorbital angle near the nucleus. It isimmediately obvious that the bending of the aluminiumorbitals in this way should be greater in (9b) with its' dimethylated double bond' than in (7) where thebridging atom contributes two AOs. This is seenclearly from the orbital diagrams in (14) and (15).E tIYE tIt is interesting that the external Me-Al-Me angle in(9a) (123") 23 is considerably larger than that (115.9") 2oin (S), as would be expected from the values for q (0.87and 1.00 respectively).Efect of the Terminal Grou9.-Replacement of terminalalkyl groups by halogen lowers the coupling constantand raises the asymmetry parameter [cf.(4b) with (4e),23 R. G. Vankra and E. L. -4rnrna, J . Amer. Chenz. SOG., 1967,89, 3121J.C.S. Dalton(4d) with (4f), and (5a) with (5c)l. This would also beexpected. An increase in the electronegativity of aterminal group increases the polarity of the bond linkingit to aluminium and so lowers the population of thecorresponding aluminium A0 [d in equation (3)]. Notonly will this lower the coupling constant but it shouldalso decrease the angle between the AOs used to formthe terminal bonds. This angle is greater (i.e., $ > 0)because the population of the terminal AOs is greaterthan that of the centre ones [c > d in equation (3)] andthe repulsion between the electrons is correspondinglygreater. A reduction in the terminal populations shouldtherefore allow 0 to increase at the expense of a decreasein #.This in turn should lead to an increase in theasymmetry parameter [see equation (a)].Choice between Bridging and Terminal Positions.-Ligands with two pairs of available electrons should bemuch more strongly bound to aluminium in bridgingpositions than in terminal ones. The difference shouldbe less for alkyl groups which have only one A 0 availablefor bonding. Alkyl bridging groups should thereforebe present only when no bivalent ligand is available,i.e., in hexa-alkyl-dialanes. Thus OMe and NMe,certainly occupy bridging positions in (7) and (8) andthe available evidence also indicates that halogen isalso in all cases preferred over alkyl in the bridgingpositions.It has been shown that the bridging groupsare chlorine by electron diffraction 24 in the case of(4a) and by X-ray crystallography25 in the case of(CH,AlCl,),. The n.q.r. data very strongly suggestthat the same is generally true. Thus the fact that(4b) has a much larger coupling constant than (9b), andalso than (4e) , can be explained only if replacement of abridging alkyl by chlorine raises the coupling constantwhereas analogous replacement of a terminal alkyllowers it. The data in Tables 4 and 5 indicate thatreplacement of a bridging alkyl by chlorine raises thecoupling constant by 6-7 MHz while replacement ofterminal alkyl lowers it by 5-6 MHz.As pointed outabove, these changes are those to be expected on thebasis of the Townes-Dailey theory and the changes inthe asymmetry parameter are also in the expecteddirection.Further confirmation is provided by the 35Cl n.q.r.data6 shown in Table 6. The fact that (4e) and (4f)gave two distinct * chlorine frequencies shows that thechlorine atoms in these occupy two chemically distinctlocations. These can only be the terminal and bridgingpositions. The fact that only two distinct resonanceswere observed, and the fact that the intensities of thetwo signals were similar, indicates that both bridgingpositions must be occupied by chlorine. Similar remarks* I.e., differing by more than the possible effects of crystaIfields.One of the resonances in compound (4e) showed an ad-ditional small splitting, probably due to this.24 L. 0. Brockway and N. R. Davidson, J . Amer. Chem. Soc.,1941, 63, 3287; see also J. Weidlein, J . Organometallic Chem.,213, 17, 1969, and refs. therein.25 G. Allegra, G. Perego, and I. Immirzi, Makromol. Chem.,1963, 61, 69.apply t o (4g), showing that chlorine also takes pre-cedence over phenyl in the bridging positions.In each case the lower of the two observed frequenciescan be assigned to the bridging chlorines. This wouldbe expected both theoretically and from the analogywith gallium trichloride where the assignment of fre-quencies has been established unambiguously bystudies 2o of the Zeeman effect.Next we have to consider the distinction betweenphenyl and alkyl.Since sp2-hybridized carbon is moreelectronegative than sp3-hybridized carbon 26 an A1-Phbond should be more polar in the sense AP+-CS- thanAl-Alkyl. The arguments given above indicate thatreplacement of terminal alkyl by phenyl in a dimericcomplex (&AlX), should lower the 27Al couplingconstant. This effect is seen in the comparison between(4e), (4f), and (4g). Replacement of alkyl by phenyllowers the coupling constant by 0-5-1 MHz per phenylgroup.The effect of replacing bridging alkyl by phenyl isharder to predict since the n-electrons of phenyl couldconceivably be used for back-~o-ordination.~~ Replace-ment of bridging alkyl by an equivalent but moreelectronegative group should raise the coupling constantby making the three-centre bonds more polar in thesense A1,8+-CS-.Back-co-ordination will, however, leadto transfer of charge in the opposite direction and so toa decrease in the coupling constant. It is impossible totell by qualitative arguments which of the two effectswill predominate. Fortunately (loa) has a highercoupling constant than (9a). Since replacement of aterminal methyl in (9a) by phenyl should have loweredthe coupling constant, this must imply that back-co-ordination is relatively unimportant. It has in factbeen established by X-ray crystallography 28a that thephenyl groups in (loa) occupy bridging positions.These conclusions are confirmed by the difference incoupling constant between (lOa) and (lob) (3.7 MHz)which is about four times the value deduced above forthe difference between terminal methyl and terminalphenyl.Sincesp-hybridized carbon is even more electronegative thansfi2-hybridized carbon, replacement of terminal methylgroups in (9a) by phenylethynyl should lead to a largedecrease in the coupling constant.Since the couplingconstant of (11) is in fact greater than that of (9a),the phenylethynyl groups must occupy the bridgingpositions. This has been shown to be so by lH n.m.r.spectroscopy.Bb Here again back-co-ordination by theunsaturated bridging group appears to be unimportant.Mosomeric aid Diineric CompLexes.-When R,AlSimilar arguments apply in the case of (11).26 See M.J. S. Dewar, ‘The Molecular Orbital Theory ofOrganic Chemistry,’ McGraw-Hill, New York, 1969, p. 147.27 M. J . S. Dewar, Bull. SOC. chim. France, 1962, 18, C71; seeM. J. S. Dewar and A. P. Marchand, Ann. Rev. Phys. Cheuvt., 1965,16, 321.28 ( a ) J. F. Malone and W. S. McDonald, Chenz. Comrvz., 1970,380; (b) E. A. Jeffery, T. Mole, and J. K. Saunders, Austral. J .Chem., 1963, 21, 1371973 2389combines with a donor X, the strength of the resultingA1X dative bond is limited by the charge transferproduced in its formation (R3A1--X+) . When, however,R,AlX dimerizes, the charges can be partly neutralizedby polarization of the pre-existing A1X bonds. Onewould therefore expect the AlX bonds in (R,AlX), to bestronger than those in R3A1--X+.The population ofthe corresponding aluminium AOs [d in equation (3)]should therefore be greater than that of the fourthaluminium A0 in R3A1X [b in equation (l)]. Since theRA1 populations should be similar in both cases [a 21 cin equations (1) and (3)], the factor la - b] in equation(1) should be greater than Ic - dl in equation (3). Ifthe aluminium atoms are in each case tetrahedral,cos 0: = -Q and I-, = 1 in equation (3). Both equationsthen reduce to the same form, i.e., (7). The coupling(7)constant for (R,AlX), should therefore be lower thanthat of R3A1X. This is true for the pair Me,Al*OMe,(28.18 MHz) and (Et,AlOEt), (22.71 MHz). Thedifference is far too great to be due to the substitutionof ethyl for methyl. Indeed, the comparisons (4a)-(4b)-(4c)-(4d) and (9a)-(9b) suggest that changing alkylgroups has little effect on the coupling constant.Steric Efects: Dimer-Polymer Equilibria.-Tervalentaluminium compounds R,AlX usually form cyclicdimers (R,AlX),.In certain cases, however, trimers[e.g., (12)] or polymers are formed. The bonding insuch compounds is essentially similar to that in thedimer. The only difference is that they contain nostrained rings. Conversion of the dimer into trimer orpolymer must therefore be exothermic. However italso leads to a decrease in entropy, owing to the replace-ment of a larger number of small molecules by a smallernumber of large ones. Since the entropy change willbe much the same in all cases, being due mainly tochanges in translational entropy, the equilibrium will bedetermined by the energy change on polymerization.This will depend on the strengths of the A1X bondssince the energy required to bend a bond in generalvaries with its strength.The bonds in turn are strongerfor bivalent ligands than for univalent ones and are alsostronger, the less electronegative the ligand. It istherefore easy to see why complexes with univalentbridging groups (e.g., alkyl) or with bridging halogenare invariably dimeric whereas compound (12) and(Me,AlNHMe), are trirneric and (13) is polymeric.The difference in bonding between dimers and higherpolymers is reflected in their n.q.r. spectra. As wehave seen, the deformation of the A1X bonds in thedimer should reduce the populations of the correspond-ing aluminium AOs. The populations of these AOs intrimeric or polymers should therefore be greater and the27Al coupling const ants correspondingly smaller.This29 K. Gosling, G. M. McLaughlin, G. A. Sims, and J. 0. Smith,Clzem. Comvn., 1970, 1617.effect is seen in the comparison (8). Unfortunately nodata are available for dimeric sulphur complexes.(7) 22-71 MHz (12) 20.86 MHz (8)However the huge difference between the couplingconstants for the linear polymer (13) (18.46 MHz) andthe monomeric complex (le) (29-75 MHz) shows thatthe value for an analogous dimeric complex would bemuch greater than that for (13).The fact that (7) is dimeric 22 whereas (5) is trimeric 29must be attributed to steric hindrance in the higherpolymers derived from (7), owing to the greater bulk ofthe ethyl groups.The same factor could account forthe fact that Me,AlNMe, forms the dimer (8) whereasMe,AlNHMe forms the trimer (Me,AlNHMe),. Thisinterpretation is supported by the fact that the couplingconstant for (8) is greater by nearly 5 MHz than thevalue for (la), although normally (see section on uni-and bi-valent bridging groups) the coupling constantsfor dimeric complexes are lower than those for analogousmonomeric ones. If steric effects are sufficient toprevent Me2A1NMe2 from forming higher polymers, theymay well also destabilize the dimer, thus weakening theA1-N bonds and so raising the coupling constant. Onewould then expect a very considerable difference incoupling constant between (8) and the (presumablystrain-free) trimer ( Me,A1NHMe),.29 This could havebeen responsible for our failure to observe signals forthe latter, the corresponding frequencies being belowthe lower limit of our spectrometer.The splittings of the signals in compounds (4e) and(5b) are probably due to the crystal field.The inten-sities of both components were in each case similar. Incompound (12) on the other hand the ratios of intensitieswere 2 : 1 for each pair of signals. Unless (12) has arather strange crystal structure, this must imply thatMe I ‘0-Me0- I I I de Me Me I he=‘Me= I \‘r-l--* \-MeMe-A 1OLA‘L \I ‘0-MeMe ‘ Me Me(17)the aluminium atoms are not chemically equivalent,one of them differing from the other two.If so, (12)must exist as the boat conformer (16) rather than as thJ.C.S. Daltonchair (17), presumably to avoid steric interactions ( 0 - *)between axial methyl groups. A study of the crystalstructure of (12) would clearly be of interest.A further curious feature is the low asymmetry para-meter of (12). One would certainly have expected thisto be unity like that of the analogous dimer (7).Conclusions.-Given the wide variety of compoundsdiscussed in this paper, it is gratifying to find that thesimple MO treatment of Townes and Dailey is so success-ful. Virtually all the observed trends can be explainedin these terms, including the ' non-classical ' bondingpresent in electron-deficient molecules such as hexa-methyldialane[6] (9a). These results not only help tojustify the use of simple MO arguments in connectionssuch as these but also indicate the potential of n.q.r.spectroscopy as a tool in the study of chemical bonding.Indeed, its success in this connection is such that it canalso serve as an aid in determining structures of com-pounds containing ' n.q.r.-active ' elements.In view of the growing importance of analogousorganometallic compounds as catalysts in organicchemistry, studies of this kind may prove of practicalimportance for two reasons. First, n.q.r. measurementscan be carried out much more easily and quickly thanother methods of structure determination. Secondly,a knowledge of the electronic structures of such catalystsis clearly prerequisite to any interpretation of theirreactivity.The discussion given here is of course based on a verynaive MO approach. Measurements of this kind mayprove still more significant in conjunction with moresophisticated theoretical treatments. We have recentlyshown30 that the 35Cl n.q.r. coupling constants of awide variety of aryl chlorides can be quantitativelyinterpreted in terms of semiempirical SCF MO calcu-lations. If similar success attends calculations for otherelements, this could not only provide even more detailedinformation concerning the electronic structure ofmolecules but could also prove of value in determiningthe parameters for elements (in particular metals) forwhose compounds thermochemical data are lacking.This work was supported by the Air Force Office ofScientific Research and by the Robert A. Welch Foundation.We thank Dr. G. E. Peterson, Bell Telephone Laboratories,for discussions and advice concerning the spectrometer,and Texas Alkyls Inc. for gifts of materials.[2/2598 Received, 16th November, 1972130 M. J. S. Dewar, D. H. Lo, D. B. Patterson, N. Trinajstic,and G. E. Peterson, Chem. Comm., 1970,238