Part I

BY L. E. SUTTON

In this Volume there are only two Chapters about electron diffraction studies of molecular structure, but the second of them is unusually long.

The first is the customary, comprehensive survey of recent work: the period covered is August 1975 to August 1976. Our thanks are due to Dr D. W. H. Rankin for undertaking this task in three successive years. In introducing his detailed report he has made some interesting and pertinent general remarks. He draws attention to the increasing complexity of the molecules now being studied and, what in part arises from this, to the increasing tendency of investigators to use a variety of ancillary information in the analyses. He stresses the need for very clear statements of what such information is introduced and how it is used. This point has already been made by the Commission on Electron Diffraction of the International Union of Crystallography (see Acta Crystallogruphica, 1976, A32, 1013) but it is of increasing importance and can properly be emphasized more strongly. The analyses reported include some very high temperature work, a fair amount of looking for unusual bond lengths or bond angles, and a large number of conformational studies.

Because of the current interest in conformation it is appropriate that we have a Chapter, by Professor Robert K. Bohn, surveying the results obtained in recent years. It includes a series of massive tables of data which in themselves represent a very substantial work of scholarship and which should be of great value. The quantity of work came as a surprise to us all and, indeed, proved quite disconcerting. Professor Bohn is mainly concerned with systematizing the known facts and with the broad picture, so he does not give much space to detailed discussions of causes of conformation; but he draws attention to one important generalization which works for the great majority of cases. This is that a double bond, such as the carbonyl bond (C=O), can be regarded as two equal but opposite bent bonds. The description of the bond given by photo-electron spectroscopy is, however, that it is a σ-bond plus a π-bond, i.e. two different component elements. This further illustrates the dichotomy which is frequently found between the geometrical (or geographical) and the energetic descriptions of molecules. Neither one is universally right or wrong. Each is appropriate for one aspect of structure. Perhaps we need to understand better how to predict which is required for a particular purpose, and why.

My remaining task is the pleasant one of thanking the two contributors to this Part for their perseverence and cheerful co-operation which, as in previous years, have made the lot of the Senior Reporter a relatively happy one.

1

Electron Diffraction Determinations of Gas-phase Molecular Structures

1 Introduction

In this chapter the results are reported of nearly 100 structure determinations of molecules in the gas phase by electron diffraction, published in 81 papers, between August 1975 and August 1976. There are about 20 gas diffraction instruments active in the world, so the average output is only five published structures per unit. As there are ten or more structures reported for each of three instruments (EG100 in Moscow, KD.G2 in Oslo, and the 'Oslo apparatus') it is clear that there is considerable unused capacity. Why is this so? Is it that equipment is often out of commission for technical reasons? Is there a shortage of people interested in studying gas-phase structures? Or is there a lack of molecules suitable for study by this method? From my own experience as a preparative chemist with no diffraction apparatus of my own, it is not simply any one of these. The problem seems to be one of bringing together men, molecules, and machines.

There is evidence from papers reviewed in this chapter that it is becoming increasingly difficult to find simple molecules for study for electron diffraction. The Figure shows the distribution of compounds investigated in terms of number of atoms per molecule, compared with the corresponding data for two years previously. It is clear that the peak of the smoothed distribution has moved from about eight atoms per molecule to around 14 and that there has been a dramatic drop in the number of very small molecules studied.

This tendency to use more complex molecules may be demonstrated in two other ways. In the first place, larger and heavier species tend to have lower vapour pressures, and so to an increasing extent high nozzle temperatures are being used. This year, temperatures above room temperature were used for more than 65% of all the compounds investigated, with over 25% requiring more than 100 °C. Secondly, the diffraction experiments very often cannot give sufficient information to enable a full structure determination for a complex molecule to be carried out. Thus in many cases additional experimental data, such as rotational constants, are used, while in others heavy reliance is placed on vibrational amplitudes calculated from spectroscopic data, or on structures derived by theoretical means of varying sophistication including molecular mechanics and CNDO/2 and ab initio molecular orbital calculations. One consequence of this is that it is often difficult to determine which of the quoted results are experimental, and which are merely the opinions of a computer, faithfully reflecting its master's preconceptions! Indeed, a parameter may be dependent on both experimental and theoretical data, and without having the author's programs it may be impossible to assess fully the validity of the results. This state of affairs has developed gradually, and it seems appropriate now to urge authors to make clearer statements about the sources of their derived parameters. Otherwise there is a danger not merely of lack of clarity but of the development of circular arguments.

Because of such occasional doubts about their origin, the results quoted in this chapter are in the same form as given in the original papers. It may be assumed that distances are ra values and that errors (in parentheses) are estimated standard deviations, expressed in terms of the least significant digit given, unless specifically stated otherwise. Comments and comparisons with previous work are in general taken from the reports being considered; references to such earlier work will not be given here.

2 Main Group Inorganic Compounds

Group III. — As part of a series of studies of gases at very high temperatures, the structure of rubidium metaborate at 1240 K has been determined. The RbOBO chain has an angle of 112(5)° at oxygen, but is linear at boron, as had been found earlier for lithium metaborate and for boric oxide, which has an OBOBO chain. Bond lengths reported are r(Rb — 0) 2.57(6) and r(B — 0) 1.255(10) Å.

Further work has been reported on the intriguing problem of beryllium borohydride. A sample of this compound was divided into two portions; crystals were obtained from one, while the other was rapidly cooled in liquid nitrogen, to give an amorphous material. Diffraction patterns of gas samples obtained by warming these solids differed considerably and corresponded well with the two types of curve obtained in earlier attempts to solve the structure. The authors conclude, without attempting any refinements of structures, that there are at least two gaseous species obtainable from beryllium borohydride, although it is impossible to be certain that both have the formula BeB2H8.

In a new study of decaborane(l4) the mean B & — B distance is found to be 1.78(1) Å, and the B — H distances 1.18(2) and 1.34(2) Å for terminal and bridging hydrogen atoms respectively. The parameters are in good agreement with those found for the solid phase by neutron diffraction. Much older X-ray and electron diffraction results are shown to be in error.

1,l0-Dicarba-closo-decaborane(l0)( 1) has D4d symmetry, with B — B bond lengths of 1.87(1) and 1.80(1) Å, very similar to the corresponding distances in [B10H10]2-. The C — B distance, 1.60(1) Å, is much the same as in 1,6-dicarbadecaborane derivatives, but is substantially less than the 1.710 Å in 1,12-dicarbadodecaborane(l2). The structure of the 1,12-di-iodo-derivative of this last compound (2) has also been determined, using sets of data obtained with two sets of diffraction apparatus. The interatomic distances of the boronxarbon cage are very close to those in the parent carbaborane, the average values for the two determinations being C — B 1.706(9); B — B (within five-membered ring) 1.779(8); B — B (between rings) 1.778(12) Å. The two C — I distances obtained are 2.082(14) and 2.107(18) Å, shorter than in methyl iodide, longer than in iodoacetylene, but much the same as in tetraiodoethylene. It is suggested that the carbon atoms in 1,12-dicarbadodecaboranes behave as if sp2, hybridized.

Several studies of boron-Group VI compounds have been undertaken, particular emphasis being placed on evidence for π-bonding in boron-oxygen and -sulphur bonds. Trimethyl borate has a planar skeleton of C3h symmetry, with r(B — 0) 1.367(4) and r(C — 0) 1.424 Å, and [angle] BOC 121.4(5)°. The planarity of the skeleton, the shortness of the B — O bonds compared with a predicted single bond length of 1.43 — 1.47 Å, and the width of the angle at oxygen are taken as evidence for n-bonding, the extent of which is greater than in boron-sulphur or -selenium compounds.

In dimethyl boric anhydride, Me2BOBMe2, the skeleton is not planar, but there is a twist of 38.2(36)° about each B — O bond, with C, symmetry being preserved, and a dihedral angle of 72° between the two CBC planes. The short B — O distance, 1.359(4) Å, indicates the importance of π-bonding: r(C — B) is 1.573(4) Å. The two OBC angles at each boron atom are different, being 117.1(13) and 120.4(14)°, the larger one involving the carbon atom cis to the other Bme2 group. A wide BOB angle, 144.4(27)°, results from the short B — O bonds, by repulsion both between the boron atoms and between the electrons in the B — O bonds, increased in number by π-bonding.

Bis(methy1thio)methylborane (3) was found to be present probably only in the syn-anti form, although up to 20% of the syn-syn form could also be present. Important parameters are r(C — B) 1.567(10), r(B — S) (mean) 1.796(7), r(S — C) 1.818(6) Å; [angle] S(2)BC 116.4(4), [angle] S(4)BC 124.4(4), [angle] BS(2)C 104.5(10), [angle] BS(4)C 106.2(14)°. The S-methyl groups were found to be twisted about 17 [+ or -] 6° out of the plane of the other heavy atoms, but this could be a vibrational effect with the average skeletal structure being planar.

In contrast to this, bis(dimethylbory1)disulphane (4) has planar C2BSS units, but these are twisted so that the BSSB dihedral angle is 120(6)°. The S — S bond is unusually long, 2.078(4) Å, compared with 1.88 Å, in S2F2 and 2.05 Å in Me2S2, and the SBC angles are of interest, the ones trans to the S — S bond being 114.0(6)° but the cis ones being 123.4(4)°. Other parameters [r(B — S) 1.805(5), r(C — B) 1.573(5) Å; [angle] BSS 105.3(4)°] are as expected.

Assuming that 1,3-dimethyl-2-chlorodiazaboracyclohexane (5) has a planar boron atom and a plane of symmetry, all the heavy atoms except the carbon furthest from boron were found to be coplanar within experimental error. Most parameters are close to those in the analogous cyclopentane derivative, with r(B — Cl) 1.782(5), r(B — N) 1.417(5), and r(C — N) 1.454(4) Å. The angles in the ring are 120.8(5) at boron, 121.6(5)° at nitrogen, and 110.3(6) and 113.4° at C(4) and C(5). The methyl groups are bent away from the boron atom, with [angle] BNC 124.1(5) compared with [angle] CNC 112.4°. The various parameters are considered to indicate π-bonding in the B — N bonds, as expected from ab initio calculations and other investigations.

Full details of the structure of the adduct of gallium trichloride and ammonia have now been published. The wide ClGaCl angle, 116.4(3)°, is similar to that in the aluminium analogue, and probably indicates that free GaCl3 is planar. The Ga —N bond is fairly long [2.057(11) Å and this suggests a weaker interaction than in the aluminium compound. The Ga — C1 bond length is 2.142(5) Å.

Group 1V. — In a new determination of the structure of perfluorodisilane, the rα0 parameters are very close to those reported in an earlier study: r(Si-Si) 2.317(6), r(Si-F) 1.564(2) Å; [angle]FSiF 108.6(3)°. Whereas a single fixed conformation was used in the previous work, the new data have been analysed in terms of restricted rotation, the barrier height being determined as 0.5 — 0.7 kcal mol-l, depending on the assumed gas temperature. CND0/2 calculations indicate the barrier to be 0.5 kcal mol-1, compared with 3.9 kcal mo1-1 for perfluoroethane.

Hexamethylcyclotrisilazaneh as r(Si — N) 1.728(4) and r(Si — C) 1.871(4) Å, and the ring angles are SiNSi 126.8(8) and NSiN 108.4(10)°, with [angle] CSiC 108.9(23)°. The angles are therefore similar to those in non-ring compounds, and this seems to be generally true, so that it is possible to predict the degree of puckering (slight in this case) in six-membered rings with alternating atom types. Chair (C3υ) and boat (Cs) conformations fit the data equally well: a twist-boat (C2) conformation is not so satisfactory.

A series of structures of silicon-oxygen compounds has been published. Hexamethyldisiloxanel6 has Si — C and Si — 0 bond lengths of 1.865(4) and 1.631(3) Å respectively, and the SiOSi angle refines to 148(3)°. The molecule probably adopts a staggered (C2υ) conformation, but a twist angle of 30° for the trimethylsilyl groups is also possible. This molecule has large-amplitude torsional and bending (at oxygen) vibrations, and the effects of these on the apparent SiOSi angle and conformation are uncertain. The SiOSi angle may be very much wider than 148° in the average structure.

Tetramethoxysilanels and methyltrimethxysilane have r(Si — O) 1.613(1) and 1.632(4) Å, r(C — O) 1.414(2) and 1.425(4) Å, and [angle] SiOC 122.3(3) and 123.6(5)°. The tetramethoxy-compound is present as a single conformer, of S4 symmetry, slightly flattened along the axis, so that two OSiO angles are 115.5(10)°, and the others are 106°. In this respect, and in the COSiO dihedral angles (64°), this molecule is very like tetramethoxymethane. In methyltrimethoxysilane all the angles at silicon are close to the tetrahedral angle, and again there is probably one important conformer, with C3 symmetry and dihedral angles COSiC of 25 — 80°, with considerable torsional motion. Another conformer with dihedral angles of about 150° may also be present to a small extent.

A mixture of dichlorogermylene, GeCl2, and germanium tetrachloride resulted from heating polymeric GeCl2. Using the known GeCl4, structure, it was determined that 49(3) % of the gaseous material was GeCl2, and that the ClGeCl angle in this molecule was 107(5)°. This should be compared with 97° in GeF2, 101° in SiF2, 105° in CF2, and 105° in SiCl2. The Ge — Cl bond length of GeCl2 could be shorter than that of GeCl4, but it was impossible to be certain about this.

As part of a systematic investigation of methylhalogenogermanes structures have been determined for tetramethylgermane, difluorodimethyl- and trifluoro(methy1)germane, and dibromodimethyl- and tribromo(methyl)germane. The important parameters are given in Table 1. The general shortening of bond lengths with increasing halogen substitution is correlated in each case with changes in estimated bond polarities. The wide CGeC angles in the two dihalogenodimethylgermanes are accounted for in terms of a concentration of electron density in germanium orbitals to carbon (VSEPR model) or an excess of s character in the Ge — C bonds (hybrid atomic orbital model).

Germyl isocyanate and digermylcarbodi-imide both have non-linear heavy atom skeletons, with GeNC angles, uncorrected for shrinkage, of 141.3(3) and 138.0(5)° respectively. Important bond lengths in the isocyanate are r(Ge — N) 1.831(4), r (N — C) 1.19O(7), and r C — O) 1.182(7) Å, while in the carbodi-imide r (Ge — N) is 1.813(5) Å, and r(N — C) is 1.184(9) Å. in each case the bonds to carbon are collinear within experimental error. The dihedral angle between the Ge — N bonds in digermylcarbodiimide refines to 75°. Both structures are discussed in terms of valence bond models, (p -> d) π-bonding, and non-bonded contacts between germanium and carbon.

Group V. — Most of the work on Group V compounds published this year is concerned with four-co-ordinate phosphorus(v) derivatives. A careful study of thiophosphoryl fluoride showed rg(P — F) to be 1.538(3) and r(P — S) 1.866(5) Å; the FPF angle is 99.6(3)°. Experimental amplitudes of vibration agree well with those calculated from spectroscopic data, and the derived rotational constants are consistent with those determined directly. A survey of bond lengths in various PX3, and Y=PX3 molecules has been made, and an empirical relationship is given, expressing bond lengths in terms of covalent radii and electronegativities of X and Y. Another empirical equation relates angles in Y=MX3 compounds, where M is a Group V element, to the radii of M and X and the electronegativity of Y.