J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2649-2652 EPR Data do not support the P=O Representation for Trialkyl Phosphates and Phosphine Oxides or Sulfides Uma S. Rai Department of Chemistry, Banaras Hindu University, Varanasi 221005,India Martyn C. R. Symons* Department of Chemistry and Biological Chemistry, University of Essex, Wivenhoe Park, Colchester, Essex, UK C043SQ Almost all representations of phosphates [(RO),P=O], trialkyl phosphine oxides (R,P=O) and related species include a double bond between phosphorus and oxygen. However, we strongly oppose the double-bond repre- sentation and maintain that it is misleading. EPR evidence for electron-loss centres are shown to be in accord with the structures (R),+P-0', with small, negative spin density on phosphorus.Also, electron addition shows that there are no low-lying vacant 3d orbitals available for the excess electrons, which occupy a largely non- bonding CT orbital with high spin-density in the 3s orbital on phosphorus. These results cannot be reconciled with any of the concepts which lead to the P=O formulation. It is almost universal practice to represent bonding between phosphorus and oxygen, sulfur, selenium or tellurium (X) as R,P=X. In a recent review, a variety of models for this bonding were described, and, despite much uncertainty as to the proper description of the 'real' nature of the bonding, it was firmly concluded that the double-bond formulation is correct: 'The PO bond is a double bond, formulated as P=O'.' Much weight is placed on the idea that these bonds are shorter and stronger than P-0 single bonds.NMR results are said to show that there is only a small positive charge on phosphorus, and IR results are also said to accord with expectation for the P=O formulation.' However, certain evidence was not cited in this review, nor elsewhere in this book on phosphorus compounds, namely, EPR data for related radicals. In our early studies, we argued that these results preclude significant double but these and later results have been ignored. The aim of the present study was to generate more radicals of the type R,P-X', (or R,P=X') having an unpaired elec- tron in a px orbital on ligand X. The results give further strong evidence for almost complete localisation on X, with no significant delocalisation, by any route, onto phosphorus. Results are also discussed for electron adducts, (R3PX)- which give no evidence for 3d-orbital participation, and which also disfavour the double-bond formulation.Experimental All compounds were of the highest grades available, and were not further purified. CFCl, was purified by passing it down an alumina column, followed by drying and deoxygenating it using oxygen-free nitrogen. Dilute solutions (ca. 1 : lo00 mole fraction) were irradiated at 77 K with doses in the region of lo3 Gy. EPR spectra were recorded on a Varian El09 X-band spectrom- eter using 100 kHz modulation. This was interfaced to an Archimedes computer. Samples were annealed in situ, with continuous monitoring of the EPR spectra, and recooled to 77 K whenever significant changes were observed.Results and Discussion The present results, together with EPR results from related studies, are given in Table 1. A typical EPR spectrum for Ph,PS'+ radical cations, is shown in Fig. 1. These results are -f This paper was presented at the 27th International ESR Con-ference at the University of Wales, Cardiff, 21st-25th March, 1994. unexceptional, and conform with our expectations for the species concerned. In this study, the Ph3PS'+ radical cations were formed by exposing dilute solutions of Ph,PS in CFC1, to y-rays at 77 K, since this is a standard procedure for pre- paring radical and no other product is expected. In our related studies of phosphine oxide radical cations, R,PO'+, hyperfine coupling to a single solvent chlorine nucleus was observed in addition to 31Pcoupling.In the present work, two centres were found for radical cations of Ph,PS, one showing a clear 31P doublet with a large Ag,, as expected for non-complexed Ph3PS'+, and the other, with a greatly reduced g-shift, showing complex features from hyper- fine coupling to 35Cl and 37Cl (Table 1). Thus, in this case, a borderline situation seems to apply. We have previously shown that, as the ionization potential decreases, the ten- dency to bond to chlorine is reduced.' So the sulfides should form weaker complexes than the oxides. The key result is that, for the uncomplexed radical cation (Table 1 and Fig.2), the ,'P hyperfine coupling is very small (ca. 22 G), and almost isotropic. This can be understood entirely in terms of a localised orbital on sulfur with spin polarisation of the P-S CT electrons (structure I). \ I p'-pl.. For irradiated (CH,),PO, well defined features for H2cP(0)(CH,), radicals were detected on annealing. The 'H and 31P hyperfine coupling constants were 22 and 33 G, Table 1 31P Hyperfine splittings for a range of R,P-X' centres ~~ 31P hyperfine coupling constants/G ~ ~~ Ph PS' + 22" ca. 20 ca. 20 ca. 20.7 +Ph3PS(CFC13)' ca. 25b ca. 25 25 ca. 25 (j5Cl) 56, (37)C1)46 +H$P(CH,)~SH 35 35 35 35 ('H) 27 18 18 21 18Sb 20.0 18.7 19.1 16 16 16 16 (H3C)zNPCN(CH3)2120C ('HI ---27.5 (I4N) 36 ca.0 ca. 0 ca. 12 g, = 2.13; g,, z g, z 2.002. * Single-crystal study (R. A. Serway, S. A. Mar-shall, J. A. Marshall and W. D. Ohlsen, J. Chem. Phys., 1969, 51, 4978). Irradiated hexamethyl phosphoramide (ref. 12). hl Fig. 1 First-derivative X-band EPR spectrum for a dilute solution of triphenylphosphine sulfide in CFC1, after exposure to ionizing radiation at 77 K, showing features assigned to (a)the parent radical cations, Ph,PS'+, (b) the solvent adducts thereof and (c) central fea- tures assigned to the phenyl-based cations, Ph'+ -P(Ph),S respectively, again establishing almost complete localisation of the singly occupied molecular orbital (SOMO) on carbon; this can be compared with the CH,cH, radical in which the CH, protons have hyperfine splitting constant, aiso= 22 G 1 3230 G 20 G 'H -1 0 +1 -1 0 +1 Fig.2 First-derivative X-band EPR spectrum for a dilute solution of trimethylphosphine sulfide after exposure to ionizing radiation and annealing, showing features assigned to H,c-P(CH,),SH+ radicals J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 whilst for 'CH,, a = 23 G, showing that delocalisation is indeed small. Evidence for Spin Localisation The importance of the EPR results given in Table 1 is that they all suggest strong localisation of the SOMO on the ligand (-0, -S, -cH2) rather than delocalisation onto phosphorus. In particular, for the -CH, unit, the 'H coup- ling constants suggest ca. 95% localisation on carbon, similar to the value for ethyl radicals (CH3-cH2).The difference between these results and those reported in ref. 1 is that hyperfine coupling data probe the wavefunction of the SOMO on phosphorus unambiguously. Since the SOMO for the radical cations must be comparable with the HOMO for the parent molecules, the results can safely be used to describe the HOMO of the R3PX molecules. In the case of oxygen and sulfur derivatives some distortion or solvent interaction is required to lift the orbital degeneracy of the pn or pnb orbitals. The large shifts in grr values, however, shows that these are still similar in energy. It could be argued that the filled orbital constitutes the double bond, whilst the half-filled orbital is indeed non-bonding. It is difficult to envisage this in terms of any theoretical models that have been recently described.8 Also for the H2e- PR, radicals there is only one potential orbital that can comprise a n bond, and that is the SOMO for the radical.The EPR results show that this orbital is not n-delocalised, and is no more delocalised than that in the ethyl radical. Note that for the carbonyl radical cations, R2C=O'+, there is a clear distinction between the 2pn orbital on oxygen and the in-plane 2p non-bonding orbital. This distinction is well defined in the EPR results for the radical cations, for which the SOMO is unambiguously the formally non-bonding 0rbita1.~ These orbitals are now well separated, as shown by the small g-shifts. Hyperconjugation The EPR spectra for ethyl and related radicals have been interpreted in terms of hyperconjugation, or Q--7t delocal-isation.' This involves partial electron donation from the o-frame to give bonding and anti-bonding orbitals, the latter being the SOMO.This type of delocalisation largely involves electron donation from the B(C-H) Q orbitals and hence is of major importance for carbocations, of significance for rad- icals, but is not important for carbanions. The alternative, sometimes called reverse hyperconjugation, involves donation from the n unit into the Q* orbitals. This is clearly what is required in the present case, since the formal bonding involves moving from R3P+-X- to R,P=X with partial electron transfer from X to the R3P unit. If, as is now under- stood from theoretical calculations,'.* p(n)-dn bonding is not important, then the alternative of reverse hyperconjugation needs to be involved, in order to justify the double-bond for- malism.For normal hyperconjugation there is a nodal surface close to the 'central' atom. Thus in radicals such as ethyl, the CH, carbon acquires very little spin density, whilst the CH, protons gain positive spin density. However, for reverse hyperconjugation, there is a nodal surface between the atoms of the Q bonds [e.g. carbon and phosphorus in (H,C),P-OJ so that in the radical [(H,C),P-O'+] positive spin will be acquired by both, and hence the ,'P contribution should be positive. Hence, for (CH3)3P-O'+ and related species, spin polarisation of the P-0 CJ electrons will put negative spin density onto phosphorus, whilst the pn-o* contribution will be positive.There is strong evidence that the net contribution is, in fact, negative.' ' J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 One possible orbital is that of e symmetry shown in struc- ture II.’ X /3 I 0 This shows clearly that both the ligands and phosphorus should gain spin-density, but that for phosphorus this is in the 3p, orbital. This should give rise to a significant aniso- tropic term in the hyperfine coupling. In fact, the coupling is almost isotropic in those cases where both parallel and per- pendicular splittings are well defined. Also, the maximum values lie along the z axis rather than along the x axis. The conclusion must be that delocalisation via this process is small.Another result that gives an interesting insight into struc- ture is that for the parent radical cation of dimethyl phos- phonate (structure 111).” H .+\ In H&O-P-O /H3CO If either hyperconjugation process were involved, the spin density on hydrogen should reflect this (see structure 11). We accept that if the p orbital on oxygen were to adopt an x axis position, which placed the P-H bond in the y-z plane, this argument would be invalid, but there seems to be no reason for such a specific preference. In fact, no ‘H hyperfine coup- ling was detected, which set an upper limit of ca. 5 G, corre-sponding to < 1% delocalisation. It is significant that this hydrogen readily migrated onto oxygen to give (MeO),POH + radicals,” which would be unlikely to occur if the hydrogen were strongly confined to the nodal plane. Yet another result that establishes ca.unit spin density on one ligand is the ‘H and 14N hyperfine coupling for (CH,),NP[N(CH,),],O radicals form from hexamethyl phosphoramide.” Data are shown in Table 1. The 14N and ’H data are almost identical with those for (CH,),NH+ rad- icals. For these radicals, delocalisation is not expected to extend to the N-H bond except via spin polarisation, which puts negative spin density on H. The inference that this also applies to the HMPA radical is very strong indeed. Electron Addition An important experimental result that rules out the extensive use of 3d orbitals on phosphorus comes from the EPR results for electron adducts of the species under consideration.’ The contrast with comparable transition-metal complexes is clear; e.g.electron addition to Cr0,’- is into a pure 3dn orbital, with some delocalisation onto oxygen, and the EPR results are in full agreement.15 In marked contrast, electron addition to P043-,16,17 or (RO),PO-lts in a distor- tion of structure such that one 0-P-0 bond angle increases towards 180”, thus reducing the antibonding effect of the extra electron. This is in a formally non-bonding s-p hybrid orbital on phosphorus, with about equal 3s and 3p character (estimated from the ‘P hyperfine coupling).’ The o* character is retained in the form of some localisation, but only onto the two ‘axial’ ligands.Alternatively, in some cases a c* radical is formed of the type R,P-X- in which one specific P-X bond has elon- gated and hence trapped the electron in a localised c* orbital. 20~2’ These bending and stretching distortions, which increase the electron affinities of the parent molecules, do 265 1 not occur in the structures of the parents or their electron- loss analogues. The contrast between electron gain and loss can be judged by comparing the isotropic hyperfine splitting for a typical electron adduct ca. 700 G with that of an electron-loss centre from the same molecule (ca. 25 G). Also, the extent of anisotropy for the former (2B z 100 G) is far greater than those found for the latter (0-4 G).Conclusions These results confirm our previous conclusions, that n-type delocalisation is not important for (R,P-X’)’ systems. Are we justified in extrapolating back to the closed-shell mol- ecules, R,P-X (or R,P=X)? In view of the symmetry of the parent molecules, this extrapolation seems to be fully justi- fied. The original double-bond formulation, almost universal- ly used by chemists when X is oxygen, etc., originated when pn-dn ‘back-bonding’ was very much in vogue. It is, of course, of great importance for transition-metal complexes, and the use of a double bond for (L,V=O) systems, for example, is quite reasonable. However, current theoretical calculations now seem to rule out the involvement of 3d orbitak8 Hence, one would have expected that the double- bond concept would also be abandoned.Since it is still widely in use, attempts have been made to justify this in terms of the concept of c-n bonding, as discussed above. However, this is a universal phenomenon, so, if the double bond is intended to represent this, its use should be extended. One example is that of the N-oxides, R,N-0, which should, on this theory, be written as R,N=O. Alternatively, if R3N-O is accepted, then R,P-O should also be accepted. We advocate this in preference to the more cumbersome R,P+-O-. We stress that if the compound (R,P-OH)’ has a single bond (as claimed) then there is no case for inventing a double bond on proton loss. Again, it should be used for both or neither.Since we are all happy with positive charges on phosphorus, e.g. R4P+ ions, then why not with R,P+-O-? Clearly the o-electron densities will re-adjust to neutralise the charges extensively and no-one would expect to find unit charges experimentally: hence if it is wished to indi- 6+ 6-cate formal charges perhaps R,P-0 is best. Does it matter?. We believe it does, especially at the teaching level. Students have learned all about the o + n representation of bonds in, for example, carbonyl compounds, so when they see P=O they expect to be able to apply the same criteria, espe- cially with respect to reactivity of the double bond. However, such comparisons are most unsatisfactory. Finally, we should mention the formation of solvent adducts referred to above.These centres are members of the very important class of adducts that we have labelled o* rad-icals. These are sometimes said to have a ‘three-electron bond’. Their formation in freon systems is related to the fact that the solute radical cations have high electron affinities, so if their structures permit the formation of localised bonds, the problem is reduced by electron donation from a chlorine ligand of a solvent molecule into the o bond, leaving the unpaired electron in the corresponding o* orbital, e.g. struc-ture IV.7 \Q This reduces the spin density on oxygen and hence the 31P hyperfine coupling, but such reductions (always <50%), make no difference to the conclusions drawn herein. 2652 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 We thank the Commission of the European Communities for 12 R. James and M. C. R. Symons, J. Chem. SOC.,Faraday Trans. 2, financial assistance. 1990,86,2173. 13 I. S. Ginns, S. P. Misha and M. C. R. Symons, J. Chem. SOC., Dalton Trans. 2, 1973,2509. References 14 P. J. Krusic, W. Mahler and J. K. Kochi, J. Am. Chem. SOC., 1 2 3 4 5 6 7 8 9 D. G. Gilheany, in The Chemistry of Organophosphorus Com- pounds, ed. F. R. Hartley, Wiley, New York, 1992, vol. I1 p. 1. A. R. Lyons, G. Neilson and M. C. R. Symons, J. Chem. SOC., Faraday Trans. 2, 1972,68,1063. S. Subramanian, M. C. R. Symons and W. H. Wardale, J. Chem. Soc. A, 1970, 1239. S. Subramanian and M. C. R. Symons, J. Chem. SOC. A, 1970, 2367.M. C. R. Symons, Chem. SOC. Rev., 1984,12,393. M. Shiotani, Magn. Reson. Rev., 1987, 12, 333. T. Clark, A. Hasegawa and M. C. R. Symons, Chem. Phys. Lett., 1985, 116,79. See, for example, E. Magnusson, J. Am. Chem. SOC., 1990, 112, 7940. P. J. Boon, M. C. R. Symons, K. Ushida and T. Shida, J. Chem. 15 16 17 18 19 20 21 1972,94,6033. N. Bailey and M. C. R. Symons, J. Chem. SOC., 1957,203. H. Lozykowski, R. G. Wilson and F. Holuj, J. Chem. Phys., 1969,51,2303. M. C. R. Symons, J. Chem. Phys., 1970,53,857. D. Nelson and M. C. R. Symons, J. Chem. SOC., Perkin Trans. 2, 1977,286. M. C. R. Symons, in Chemical and Biochemical Aspects of Elec-tron Spin Resonance Spectroscopy, Van Nostrand Reinhold, Wokingham, 1978, p. 30. A. R. Lyons and M. C. R. Symons, J. Chem. SOC., Faraday Trans. 2,1972,68,1589. A. Abu-Ragabah and M. C. R. Symons, J. Chem. SOC., Faraday Trans. 2, 1990,86,3293; M. C. R. Symons and R. L. Peterson, J. Chem. SOC.,Faraday Trans., 1,1979,75,216. Soc., Perkin Trans. 2, 1984, 1213. 10 M. C. R. Symons, J. Chem. SOC., 1959,277. 11 A. Begum, A. R. Lyons and M. C. R. Symons, J. Chem. SOC., Faraday Trans. 2, 1967,1770. Paper 4102250H; Received 14th April, 1994