New reactions and reactive intermediates in the pyrolysis of cyclic phosphonium ylides

Pyrolysis, either neat or in diphenyl ether solution, results in the conversion of both 4-triphenylphosphoranylidenetetrahydrofuran-2,3,5-trione and 4-triphenylphosphoranylidenetetrahydrothio--phene-2,3,5-trione into 3,5-bis(triphenylphosphoranylidene)cyclopentane-1,2,4-trione. These reactions involve extrusion of CO 2 or COS to give 3-triphenylphosphoranylidenecyclopropane-1,2-dione which further loses CO to give triphenylphosphoranylideneketene. The precise way in which these two reactive phosphorus compounds combine to give the observed product has been examined by chemical and isotopic labelling studies. Cyclotrimerization of triphenylphosphoranylideneketene upon thermolysis in diphenyl ether has also been observed for the first time. The erroneous literature interpretation of the 13 C NMR spectrum for triphenylphosphoranylideneketene is corrected.


Introduction
We recently introduced the value of the two-bond NMR coupling constant 2 J P-CO as a diagnostic parameter for the reactivity of β-oxophosphonium ylides 1 towards thermal extrusion of Ph 3 PO to give alkynes (Scheme 1). 1 All known ylides that do undergo such extrusion to form an alkyne have 2 J P-CO < 11 Hz, while those with values above this do not.The latter group includes formyl and alkoxycarbonyl ylides (R 2 = H, O-Alkyl) where the failure of extrusion as well as the high J value is associated with the C=P and C=O functions being aligned anti rather than syn to one another.However the rule also applies to ylides where the functions are constrained syn such as the cyclic examples 2 and 3.With a J value of 11.3 Hz, 2 does not undergo extrusion under any circumstances whereas 3 with a value of 3.7 Hz readily eliminates Ph 3 PO upon flash vacuum pyrolysis (FVP) at 750 °C to give products derived from cyclohexyne. 1  A wide range of heterocyclic phosphonium ylides are known, 2,3 and among those with published 2 J P-CO values, our attention was drawn to ylide 4 and its thio-analogue 5.These are readily prepared from dichloromaleic (thio)anhydride and, with 2 J P-CO values for the ketone carbonyl of 8 Hz 4 and 7 Hz 5 respectively, seemed possible precursors for thermal generation of dehydromaleic anhydride and its thio analogue.Such heterocyclic alkynes are elusive and highly reactive compounds that have attracted considerable interest, 6 and we describe here for the first time the thermal decomposition of 4 and 5.

Results and Discussion
In fact, FVP of both 4 and 5 gave disappointing results, with the only products obtained in the cold trap being a trace of Ph 3 PO in the first case and a mixture of Ph 3 PO, Ph 3 PS, Ph 3 P and triphenylphosphoranylideneketene 11 in the second.However there was also substantial decomposition in the inlet tube and analysis of the residue from attempted FVP of 4 at 200 ˚C showed it, most surprisingly, to be the trioxo bis(ylide) 6 (10%) previously reported by Bestmann and co-workers. 7This must obviously be formed by an intermolecular process and, once this was clear, much better results could be obtained by heating 4 or 5 in a small volume of boiling diphenyl ether (bp 260 ˚C) for 24 h or in the case of 4 by heating the neat material at 250 ˚C for 3 h.In each case the major product was the trioxo bis(ylide) 6 which was readily identified by its characteristic NMR spectra [δ P +9.0; δ C 196.4 (t, J 10) and 187.6 (dd, J 20, 6)] and also by comparison with an authentic sample.
There are several mechanistic possibilities for the formation of this unexpected product (Scheme 2).It seems likely that either CO 2 or COS is first lost to give the triphenylphosphoranylidenecyclopropanedione 7 which may then dimerise probably by way of the dipolar ring-opened forms 9 or 10 to give 8 and this could then lose CO to give 6.Of course the diradical intermediates corresponding to 9 or 10 could also be considered but these seem less likely in view of the highly polar groups present.A second major possible route is for 7 to lose CO to give 11 which can then undergo direct cycloaddition with 9 or 10 to give 6.Evidence in favour of the latter pathway was provided by heating an equimolar mixture of 4 and 11 8,9 in boiling diphenyl ether for 2 h, which led to formation of 6 as the only phosphorus-containing product.Interestingly a control experiment of heating 11 alone for 30 min under these conditions led to formation of its cyclic trimer 12 (δ P +13.9) which was previously obtained by Bestmann and co-workers by a different route. 7We can thus conclude that whatever intermediate (9 or 10) is being formed by 4 in its decomposition to 6 reacts efficiently with added 11 to give the same product.Compound 6 was previously prepared by HCl catalysed trimerisation of 11 to give 12 which was then subjected to oxidation by 2-tosyl-3-phenyloxaziridine to afford the tetraoxo bis(ylide) 13.This underwent oxidative ring-contraction upon treatment with hydrogen peroxide and KOH to afford 6. However an additional lower-yielding route to 6 is the treatment of ethyl triphenylphosphoranylidenepyruvate 14 with sodium hexamethyldisilazide to give the product in 4% isolated yield (Scheme 3). 10 This was carried out in the expectation of isolating 7 and, although the author proposed that 6 was formed by a different route not involving 7, the new evidence presented here makes it likely that this route may also proceed by initial baseinduced elimination of ethanol from 14 to give 7 which thereafter reacts as shown in Scheme 2 to give 6.In order to differentiate between the possible modes of bond breaking in 7, we first attempted chemical labelling by preparing a tri-p-tolyl analogue.We were unable to obtain the tri-p-tolyl analogue of 4 but the tri-p-tolyl analogue of 11, compound 17, was prepared as shown in Scheme 4. Ethyl bromoacetate was reacted with tri-p-tolylphosphine 11 to give the salt 15 and treatment of this with sodium hydroxide gave the new stabilised ylide 16, which was fully characterised.

Scheme 4
When this was reacted with sodium amide in toluene a new product with characteristic NMR signals at δ P +5.3 and δ C 142.6 (d, J 39, P=C=C=O) attributed to 17 was formed.However heating this under a wide variety of conditions, either with or without added 4, did not lead to formation of a p-tolyl analogue of 6 and tri-p-tolylphosphine oxide was the only identifiable product derived from 17.
We then resorted to isotopic labelling and prepared 20% 13 C-labelled methoxycarbonyl ylide 19 by the literature method 12 starting from labelled methyl bromoacetate via the phosphonium salt 18 (Scheme 5).

Scheme 6
We cannot however exclude the possibility that, in the absence of added 11, formation of 6 is by dimerisation of an intermediate, probably 9, to give 8 which then loses CO.Whichever mechanism is operating, it should be possible to trap the intermediates with other added dipolarophiles leading to new types of stabilised ylides and this is currently being examined.
Finally, in the course of this study we have become aware of conflicting information regarding the 13 C NMR spectrum of triphenylphosphoranylideneketene 11.This compound was first reported by Birum and Matthews, 13,14 but was later obtained in a more convenient way by base-induced elimination of methanol from methoxycarbonylmethylenetriphenylphosphorane as shown in Scheme 5. 8,9 There are three literature reports of its 13 C NMR spectrum and there is widespread agreement on the rather remarkable chemical shifts of the ketene carbons [δ C -10.5 (d, J 189, P=C), 145.6 (d, J 43, P=C=C=O)].The situation for the P-phenyl signals is however confused as summarised in Table 1 with erroneous data shaded.Analysis of the 13 C NMR spectra of nearly 200 stabilised triphenylphosphonium ylides, 1 shows that the P-phenyl signals form a regular pattern with the ipso-C having a large coupling constant to phosphorus of between 90 and 100 Hz, the ortho-C coming around 132-133 ppm with a coupling constant of 9-11 Hz, the meta-C coming around 128-129 ppm with a larger coupling constant of 12-13 Hz, and the para-C around 132 ppm with a coupling constant of 0-2 Hz.The main problem in correctly interpreting the spectrum of 11 is the close proximity of the ortho and para signals and partial or complete overlap of one half of the ortho doublet with the para singlet giving two apparent singlets.The previous reports have all made this mistake, 9,15,16 with two also interchanging the ortho and meta signals. 9,16The corrected data as obtained in our work is shown and this agrees with the revised data newly acquired by one of the original authors. 17

Conclusions
The result of the isotopic labeling experiment indicates that formation of the trioxo bis(ylide) 6 from both 4 and 5 proceeds by initial loss of CO 2 or COS respectively to give ylide 7.This undergoes both loss of CO to afford the ketene ylide 11 and ring-opening to the dipolar species 9 which then combine in a cycloaddition reaction to form 6. Purely thermal cyclotrimerisation of the ketene ylide 11 to give 12 has been observed and correct interpretation of the 13 C NMR data for 11 shows this to be consistent with the usual pattern of chemical shifts and P-C coupling constants for triphenylphosphonium ylides.

Experimental Section
General.Melting points were determined using a Reichert hot-stage microscope and are uncorrected.NMR spectra were recorded using a Varian Gemini 2000 instrument at 300 MHz for 1 H, 75 MHz for 13 C, and 121 MHz for 31 P.All spectra were recorded on solutions in CDCl 3 with internal Me 4 Si as reference for 1 H and 13 C and external H 3 PO 4 as reference for 31 P. Chemical shifts (δ) are given in ppm to high frequency from the reference and coupling constants (J) are in Hz.Flash vacuum pyrolysis was performed using the set-up previously described. 18Compounds 4, 4 5, 5 and 11 8,9 were prepared by the published methods.
Flash Vacuum Pyrolysis of 4. FVP of 4 (36.5 mg) at 400 °C and 10 -2 Torr gave a brown solid at the furnace exit that was mainly Ph 3 PO (δ P +29.2) and gaseous products in the cold trap that condensed as a white solid.Addition of methanol to the cold trap followed by warming to RT did not result in trapping of any reactive products and no phosphorus compounds were present.Repeat FVP of 4 (320 mg) at 200 °C and 10 -2 Torr gave a white solid at the furnace exit that proved to be a 5:3 mixture of Ph 3 PO and Ph 3 P (δ P -5.0).In the inlet tube a pink solid was left which proved to be 3,5bis(triphenyphosphoranylidene)cyclopentane-1,2,4-trione 6 (22 mg, 9%).(Methoxycarbonylmethyl)tri-p-tolylphosphonium bromide (15).To a stirred solution of tri-p-tolylphosphine 11 (2.0 g, 6.6 mmol) in dry toluene (40 mL) a solution of methyl bromoacetate (1.0 g, 6.6.mmol) in dry toluene (20 mL) was added dropwise.After stirrring at RT for 18 h, the resulting white precipitate was filtered off and washed with diethyl ether to give the product (2.6 g, 87%) as a white powder, mp 187-189 °C.
1eat Pyrolysis of 4. A sample of 4 (20 mg) was heated in a kugelrohr distillation apparatus at 250 °C for 1.5 h.The resulting dark coloured solid was shown by1H and 31 P NMR to consist mainly of bis(ylide) 6. Neat Pyrolysis of 5. A sample of 5 (10 mg) was heated in a kugelrohr distillation apparatus at 250 °C for 3 h.The resulting dark coloured solid was shown by1H and 31 P NMR to consist mainly of unchanged 5. Solution Pyrolysis of 4 and 11 Together.A solution of 4 (40 mg, 0.11 mmol) and 11 (32 mg, 0.11 mmol) in diphenyl ether (1 mL) was heated under reflux for 2 h.The resulting dark coloured solution was shown by 31 P NMR to contain 6 (20%) in addition to Ph 3 PO and Ph 3 P. Solution Pyrolysis of 11.A solution of 11 (46 mg) in diphenyl ether (1.5 mL) was heated under reflux for 30 min.The dark solution was shown by 31 P NMR to contain Ph 3 PO, Ph 3 P and 2,4,6tris(triphenylphosphoranylidene)cyclohexane-1,3,5-trione 12 (δ P +13.9).Neat Pyrolysis of 11.