Further Studies on the Pyrolytic Domino Cyclization of Stabilized Phosphonium Ylides Bearing an Ortho-Aminophenyl Group

Four new, stabilized phosphonium ylides containing a 2-(benzyl(methyl)amino)phenyl group have been prepared and characterized and are found, upon pyrolysis under gas-phase flow conditions, to lose Ph3PO and benzyl radicals to afford new heterocyclic products resulting from domino cyclization of both C- and N-centered radicals. Most products arise from processes of the former type and have quinoline, phenanthridine, or ring-fused phenanthridine structures, while in one case, a process of the latter type leads to a benzocarbazole product. The X-ray structure of a 2-(methyl(tosyl)amino)phenyl ylide is also reported.


Introduction
The synthetic use of flash vacuum pyrolysis (FVP) is now well-established and provides the method of choice to obtain certain, otherwise inaccessible products [1]. Particularly in the area of heterocyclic chemistry, the method has been applied to a wide variety of reaction types and product classes [2]. The inherent advantages of a clean, solvent-, reagent-, and catalyst-free method are further enhanced if several reaction steps can be achieved in sequence in a single gas-phase flow pyrolysis process, which is a good example of the "domino" reaction approach [3][4][5]. In previous studies we have developed a series of methods for the synthesis of fused-ring heterocyclic compounds by FVP of suitably designed carbonyl-stabilized phosphonium ylides. The principle is demonstrated by the case of ylide 1, which, at 850 • C, eliminates both triphenylphosphine oxide and a methyl radical (Scheme 1). The resulting alkynylphenoxyl radical 2 undergoes spontaneous 5-endo-dig cyclization to give the benzofuryl radical 3, bearing a suitably-placed styryl group to undergo (after E/Z isomerization) an intramolecular S H Ar process, leading to the fused ring product naphtho[2,1-b]benzofuran 4 in 44% isolated yield [6,7]. Interestingly, in this case a second isomeric product 5 could also be identified in 14% yield resulting from a rearrangement of 3 prior to cyclization. Scheme 1. Example of the domino cyclization approach to fused-ring heterocycles.
The method was later extended to a wide range of examples, including an eight-stage cascade process leading from allyloxy ylide 6 to the 7-(2-benzothienyl)benzofuran 7 (Scheme 2) [8], formation of thieno [2,3-b]pyridine products including the benzofurothienopyridine 9 from ylide 8 [9], formation of a iminobenzopyranone 11 from ylide 10 [10], and synthesis of a wide range of products with structure 13, including 24 different fused-ring systems from the general starting ylide structure 12 [11]. All the preceding examples rely on cyclization of phenoxyl or thiophenoxyl radicals, and extension of the method to ylides with a nitrogen-based cyclizing radical introduces an unexpected complication. Preparation of the ylides requires a carbon-based protecting group on nitrogen as well as the leaving group, and transfer of the reactive site from N to C may lead to different products. Thus, while simpler ylides, such as the N-methyl-N-tosyl compound 14, give the 3-substituted quinoline 15 resulting from cyclization of -NH-CH2 (Scheme 3), the more extended cinnamoyl-type analogues 16-18 do undergo the expected cyclization of -N(Me) to give ring-fused carbazole products, 19-21 [12]. Scheme 1. Example of the domino cyclization approach to fused-ring heterocycles.
The method was later extended to a wide range of examples, including an eight-stage cascade process leading from allyloxy ylide 6 to the 7-(2-benzothienyl)benzofuran 7 (Scheme 2) [8], formation of thieno [2,3-b]pyridine products including the benzofurothienopyridine 9 from ylide 8 [9], formation of a iminobenzopyranone 11 from ylide 10 [10], and synthesis of a wide range of products with structure 13, including 24 different fused-ring systems from the general starting ylide structure 12 [11]. The method was later extended to a wide range of examples, including an eight-stage cascade process leading from allyloxy ylide 6 to the 7-(2-benzothienyl)benzofuran 7 (Scheme 2) [8], formation of thieno [2,3-b]pyridine products including the benzofurothienopyridine 9 from ylide 8 [9], formation of a iminobenzopyranone 11 from ylide 10 [10], and synthesis of a wide range of products with structure 13, including 24 different fused-ring systems from the general starting ylide structure 12 [11]. All the preceding examples rely on cyclization of phenoxyl or thiophenoxyl radicals, and extension of the method to ylides with a nitrogen-based cyclizing radical introduces an unexpected complication. Preparation of the ylides requires a carbon-based protecting group on nitrogen as well as the leaving group, and transfer of the reactive site from N to C may lead to different products. Thus, while simpler ylides, such as the N-methyl-N-tosyl compound 14, give the 3-substituted quinoline 15 resulting from cyclization of -NH-CH2 (Scheme 3), the more extended cinnamoyl-type analogues 16-18 do undergo the expected cyclization of -N(Me) to give ring-fused carbazole products, 19-21 [12]. All the preceding examples rely on cyclization of phenoxyl or thiophenoxyl radicals, and extension of the method to ylides with a nitrogen-based cyclizing radical introduces an unexpected complication. Preparation of the ylides requires a carbon-based protecting group on nitrogen as well as the leaving group, and transfer of the reactive site from N to C may lead to different products. Thus, while simpler ylides, such as the N-methyl-N-tosyl compound 14, give the 3-substituted quinoline 15 resulting from cyclization of -NH-CH 2 (Scheme 3), the more extended cinnamoyl-type analogues 16-18 do undergo the expected cyclization of -N(Me) to give ring-fused carbazole products, 19-21 [12]. However, when N,N-dibenzyl ylides 22-25 were examined, a more complex picture emerged [13] with all four examples giving mainly the styrylquinoline products 27 (Scheme 4) as mixtures of (Z)-and (mainly) (E)-isomers, together with the ring-fused carbazoles 26 for ylides 22 and 25 only, and the product 28 resulting from further cyclization of 27 from 25 only. The loss of a phenyl group in formation of products 27 and 28 is notable, and only for compound 24 were the products 29 and 30 retaining the N-benzyl-derived phenyl obtained. It is clear from these results that it is not yet possible to predict the pyrolysis outcome for the amino-functionalized ylides with any degree of certainty and further work is required to fully understand the processes involved. In this paper, we describe the synthesis and characterization of four new amino-functionalized ylides and the results of their pyrolysis, which leads to the formation of seven new fused-ring heterocyclic products.

X-ray Structure Determination of Ylide 14
Since many of the previously studied nitrogen-containing ylides were formed in disappointing yield and showed severe restricted rotation leading to very broad signals in their room temperature NMR spectra [12,13], it was clear that steric hindrance may be an important factor in their reactivity. To examine this further, we have determined the structure of the N-methyl-N-tosyl ylide 14 [12] by X-ray diffraction. The resulting structure ( Figure 1) confirms that this is indeed a sterically congested molecule with the two very bulky ortho substituents on the central benzene ring scarcely able to fit. Within the crystal there are no significant intermolecular interactions. The following bond lengths within the keto ylide functionality: P(7)-C(7) 1.746(2), C(7)-C(8) 1.408(2), C(8)-O(8) 1.263(2) A are indicative of a good degree of delocalization and a high contribution from the charge-separated structure with P + -C=C-O − rather than P=C-C=O. The torsion angle P(7)-C(7)-C(8)-O(8) is −9.0(2), which is within the range normally expected for keto-stabilized ylides [14,15], and indicates a high probability of successful thermal extrusion of Ph3PO, as does the low value of 6 Hz [12] for 2 JP-CO. However, when N,N-dibenzyl ylides 22-25 were examined, a more complex picture emerged [13] with all four examples giving mainly the styrylquinoline products 27 (Scheme 4) as mixtures of (Z)and (mainly) (E)-isomers, together with the ring-fused carbazoles 26 for ylides 22 and 25 only, and the product 28 resulting from further cyclization of 27 from 25 only. The loss of a phenyl group in formation of products 27 and 28 is notable, and only for compound 24 were the products 29 and 30 retaining the N-benzyl-derived phenyl obtained.  It is clear from these results that it is not yet possible to predict the pyrolysis outcome for the amino-functionalized ylides with any degree of certainty and further work is required to fully understand the processes involved. In this paper, we describe the synthesis and characterization of four new amino-functionalized ylides and the results of their pyrolysis, which leads to the formation of seven new fused-ring heterocyclic products.

X-ray Structure Determination of Ylide 14
Since many of the previously studied nitrogen-containing ylides were formed in disappointing yield and showed severe restricted rotation leading to very broad signals in their room temperature NMR spectra [12,13], it was clear that steric hindrance may be an important factor in their reactivity. To examine this further, we have determined the structure of the N-methyl-N-tosyl ylide 14 [12] by X-ray diffraction. The resulting structure ( Figure 1) confirms that this is indeed a sterically congested molecule with the two very bulky ortho substituents on the central benzene ring scarcely able to fit. Within the crystal there are no significant intermolecular interactions. The following bond lengths within the keto ylide functionality: P(7)-C(7) 1.746(2), C(7)-C(8) 1.408(2), C(8)-O(8) 1.263(2) A are indicative of a good degree of delocalization and a high contribution from the charge-separated structure with P + -C=C-O − rather than P=C-C=O. The torsion angle P(7)-C(7)-C(8)-O(8) is −9.0(2), which is within the range normally expected for keto-stabilized ylides [14,15], and indicates a high probability of successful thermal extrusion of Ph3PO, as does the low value of 6 Hz [12] for 2 JP-CO. It is clear from these results that it is not yet possible to predict the pyrolysis outcome for the amino-functionalized ylides with any degree of certainty and further work is required to fully understand the processes involved. In this paper, we describe the synthesis and characterization of four new amino-functionalized ylides and the results of their pyrolysis, which leads to the formation of seven new fused-ring heterocyclic products.

X-ray Structure Determination of Ylide 14
Since many of the previously studied nitrogen-containing ylides were formed in disappointing yield and showed severe restricted rotation leading to very broad signals in their room temperature NMR spectra [12,13], it was clear that steric hindrance may be an important factor in their reactivity. To examine this further, we have determined the structure of the N-methyl-N-tosyl ylide 14 [12] by X-ray diffraction. The resulting structure ( Figure 1) confirms that this is indeed a sterically congested molecule with the two very bulky ortho substituents on the central benzene ring scarcely able to fit. Within the crystal there are no significant intermolecular interactions. The following bond lengths within the keto ylide functionality: P(7)-C(7) 1.746 (2), C(7)-C(8) 1.408 (2) which is within the range normally expected for keto-stabilized ylides [14,15], and indicates a high probability of successful thermal extrusion of Ph 3 PO, as does the low value of 6 Hz [12] for 2 J P-CO .

Preparation and Characterization of Ylides 35-38
The four new ylides chosen for study have structures 35-38 (Scheme 5). Compounds 35 and 36 are obviously isomeric with ylides 17 and 18, which were well-behaved upon pyrolysis and gave the ring-fused carbazoles [12], but whose dibenzyl analogues 24 and 25 showed more varied and unpredictable behaviour [13]. The 2-methylcinnamoyl ylide 37 is similarly the analogue of dibenzyl ylide 23, and both it and its thiophene analogue 38 are of additional interest, since we previously observed cyclization involving the methyl group to give vinylnaphthalenes and further cyclized products when ylides analogous to 37 and 38 but only lacking the benzylmethylamino group were subjected to FVP [16,17]. As we described previously, it was advantageous to prepare the ylides using N-acylbenzotriazoles as the acylating agents [18] rather than acid chlorides, and the four required benzotriazole derivatives 31-34 (see Supplementary Material S1-S7) were prepared by treatment of the appropriate carboxylic acids with thionyl chloride (1 equiv.) and benzotriazole (4 equiv.).

Scheme 5. Preparation of ylides 35-38.
The four ylides were prepared as shown in Scheme 5 by treatment of phosphonium salt 39 [12] with butyllithium to generate the ylide 40, followed by addition of the appropriate Nacylbenzotriazole, and were obtained as stable crystalline solids with high melting points. The signals due to the P-phenyl groups in the 1 H NMR spectra were very broad, presumably due to similar steric hindrance and restricted rotation to that revealed by the X-ray structure of 14. The situation was significantly improved by running the spectra at 55 °C, as illustrated in the case of 36 (see

Preparation and Characterization of Ylides 35-38
The four new ylides chosen for study have structures 35-38 (Scheme 5). Compounds 35 and 36 are obviously isomeric with ylides 17 and 18, which were well-behaved upon pyrolysis and gave the ring-fused carbazoles [12], but whose dibenzyl analogues 24 and 25 showed more varied and unpredictable behaviour [13]. The 2-methylcinnamoyl ylide 37 is similarly the analogue of dibenzyl ylide 23, and both it and its thiophene analogue 38 are of additional interest, since we previously observed cyclization involving the methyl group to give vinylnaphthalenes and further cyclized products when ylides analogous to 37 and 38 but only lacking the benzylmethylamino group were subjected to FVP [16,17]. As we described previously, it was advantageous to prepare the ylides using N-acylbenzotriazoles as the acylating agents [18] rather than acid chlorides, and the four required benzotriazole derivatives 31-34 (see Supplementary Material S1-S7) were prepared by treatment of the appropriate carboxylic acids with thionyl chloride (1 equiv.) and benzotriazole (4 equiv.).

Preparation and Characterization of Ylides 35-38
The four new ylides chosen for study have structures 35-38 (Scheme 5). Compounds 35 and 36 are obviously isomeric with ylides 17 and 18, which were well-behaved upon pyrolysis and gave the ring-fused carbazoles [12], but whose dibenzyl analogues 24 and 25 showed more varied and unpredictable behaviour [13]. The 2-methylcinnamoyl ylide 37 is similarly the analogue of dibenzyl ylide 23, and both it and its thiophene analogue 38 are of additional interest, since we previously observed cyclization involving the methyl group to give vinylnaphthalenes and further cyclized products when ylides analogous to 37 and 38 but only lacking the benzylmethylamino group were subjected to FVP [16,17]. As we described previously, it was advantageous to prepare the ylides using N-acylbenzotriazoles as the acylating agents [18] rather than acid chlorides, and the four required benzotriazole derivatives 31-34 (see Supplementary Material S1-S7) were prepared by treatment of the appropriate carboxylic acids with thionyl chloride (1 equiv.) and benzotriazole (4 equiv.).

Scheme 5. Preparation of ylides 35-38.
The four ylides were prepared as shown in Scheme 5 by treatment of phosphonium salt 39 [12] with butyllithium to generate the ylide 40, followed by addition of the appropriate Nacylbenzotriazole, and were obtained as stable crystalline solids with high melting points. The signals due to the P-phenyl groups in the 1 H NMR spectra were very broad, presumably due to similar steric hindrance and restricted rotation to that revealed by the X-ray structure of 14. The situation was significantly improved by running the spectra at 55 °C, as illustrated in the case of 36 (see The four ylides were prepared as shown in Scheme 5 by treatment of phosphonium salt 39 [12] with butyllithium to generate the ylide 40, followed by addition of the appropriate N-acylbenzotriazole, and were obtained as stable crystalline solids with high melting points. The signals due to the P-phenyl groups in the 1 H NMR spectra were very broad, presumably due to similar steric hindrance and restricted rotation to that revealed by the X-ray structure of 14. The situation was significantly improved by running the spectra at 55 • C, as illustrated in the case of 36 (see Supplementary Material), and the 13 C NMR spectra for all four new ylides were run at this temperature. The value of the two-bond coupling between C=O and P was 5-7 Hz in each case, indicating a good prospect for successful thermal extrusion of Ph 3 PO [14].

Flash Vacuum Pyrolysis to Give Heterocyclic Products
The first ylide to be pyrolyzed, the 3-furyl compound 35, gave a surprising result. There was complete reaction at a furnace temperature of 700 • C with formation of Ph 3 PO and bibenzyl-as expected-but the sole heterocyclic product identified in low yield did not contain a furan ring at all and was identified as the 9-methylphenanthridine 41, by comparison of its 1 H NMR data with literature values. A suggested mechanism for formation of this product (Scheme 6) involves initial extrusion of Ph 3 PO and benzyl radical followed by transfer of the reactive site from N to C, by means of a hydrogen atom transfer. Cyclization of the NH-CH 2 onto the triple bond, aromatization, and further cyclization gives an intermediate that could very easily lose a hydrogen atom to form the expected furo[3,2-k]phenanthridine. However, this is apparently not favorable, and instead, ring-opening of the dihydrofuran ring, followed by rearrangement and loss of carbon monoxide leads to the observed stable methylphenanthridine product. Supplementary Material), and the 13 C NMR spectra for all four new ylides were run at this temperature. The value of the two-bond coupling between C=O and P was 5-7 Hz in each case, indicating a good prospect for successful thermal extrusion of Ph3PO [14].

Flash Vacuum Pyrolysis to Give Heterocyclic Products
The first ylide to be pyrolyzed, the 3-furyl compound 35, gave a surprising result. There was complete reaction at a furnace temperature of 700 °C with formation of Ph3PO and bibenzyl-as expected-but the sole heterocyclic product identified in low yield did not contain a furan ring at all and was identified as the 9-methylphenanthridine 41, by comparison of its 1 H NMR data with literature values. A suggested mechanism for formation of this product (Scheme 6) involves initial extrusion of Ph3PO and benzyl radical followed by transfer of the reactive site from N to C, by means of a hydrogen atom transfer. Cyclization of the NH-CH2 onto the triple bond, aromatization, and further cyclization gives an intermediate that could very easily lose a hydrogen atom to form the expected furo[3,2-k]phenanthridine. However, this is apparently not favorable, and instead, ringopening of the dihydrofuran ring, followed by rearrangement and loss of carbon monoxide leads to the observed stable methylphenanthridine product. By way of contrast, the 3-thienyl ylide 36 behaved analogously to the N,N-dibenzyl 2-thienyl ylide 25 [13] with mainly quinoline products 42, accompanied by a small amount of the tetracyclic thieno [3,2-k]phenanthridine 43 formed by a further cyclization event (Scheme 7). The latter product showed a distinctive 1 H NMR spectrum (Supplementary Material S22) which, though complex, was successfully analyzed to derive all coupling constants using simulation. Both chemical shifts and coupling constants were in good agreement with those already reported for the isomeric system 28 [13], thus supporting the structural assignment. Clearly, in this case, the process initially involves extrusion of Ph3PO and benzyl radical, hydrogen atom transfer, and cyclization, as shown in Scheme 6-but the greater thermodynamic stability of the thiophene ring means that it survives to appear in the products. The contrast between the behaviour shown here for 35 and 36 and that for the corresponding 2-furyl and 2-thienyl isomers 17 and 18, which give exclusively fused-ring carbazole products 20 and 21 [12], is most surprising. The balance between the N-and C-centered radical as the cyclizing species is obviously delicate. By way of contrast, the 3-thienyl ylide 36 behaved analogously to the N,N-dibenzyl 2-thienyl ylide 25 [13] with mainly quinoline products 42, accompanied by a small amount of the tetracyclic thieno [3,2-k]phenanthridine 43 formed by a further cyclization event (Scheme 7). The latter product showed a distinctive 1 H NMR spectrum (Supplementary Material S22) which, though complex, was successfully analyzed to derive all coupling constants using simulation. Both chemical shifts and coupling constants were in good agreement with those already reported for the isomeric system 28 [13], thus supporting the structural assignment. Clearly, in this case, the process initially involves extrusion of Ph 3 PO and benzyl radical, hydrogen atom transfer, and cyclization, as shown in Scheme 6-but the greater thermodynamic stability of the thiophene ring means that it survives to appear in the products. The contrast between the behaviour shown here for 35 and 36 and that for the corresponding 2-furyl and 2-thienyl isomers 17 and 18, which give exclusively fused-ring carbazole products 20 and 21 [12], is most surprising. The balance between the Nand C-centered radical as the cyclizing species is obviously delicate. extrusion of Ph3PO and benzyl radical, hydrogen atom transfer, and cyclization, as shown in Scheme 6-but the greater thermodynamic stability of the thiophene ring means that it survives to appear in the products. The contrast between the behaviour shown here for 35 and 36 and that for the corresponding 2-furyl and 2-thienyl isomers 17 and 18, which give exclusively fused-ring carbazole products 20 and 21 [12], is most surprising. The balance between the N-and C-centered radical as the cyclizing species is obviously delicate. This fact was emphasized by the result in the case of the 2-methylstyryl ylide 37 (Scheme 8). This is the methyl analogue of 16 which gave the benzocarbazole product 19 as the only major product in 66% yield [12]. Although products derived from cyclization of a C-centered radical, the quinolines 44 and the benzophenanthridine 45, amounted together to 33%, the largest single product was now the dimethylbenzocarbazole 46, resulting from direct cyclization of an N-centered radical, formed in 22% yield. This fact was emphasized by the result in the case of the 2-methylstyryl ylide 37 (Scheme 8). This is the methyl analogue of 16 which gave the benzocarbazole product 19 as the only major product in 66% yield [12]. Although products derived from cyclization of a C-centered radical, the quinolines 44 and the benzophenanthridine 45, amounted together to 33%, the largest single product was now the dimethylbenzocarbazole 46, resulting from direct cyclization of an N-centered radical, formed in 22% yield. As mentioned in the introduction, one reason for trying ylide 37 was that in earlier work we observed the involvement of an ortho-methyl group in such ylide pyrolyses to give naphthalenes [16,17]. This was not observed for 37 nor for the second methyl-containing ylide 38, which simply afforded the (E)-and (Z)-thienylvinylquinolines 47, corresponding to 42 in low yield (Scheme 9). However, even in this case, the contrast with the behavior of the ylide 18 caused by introduction of the single methyl group remote from the reactive site for initial cyclization is remarkable.

General Experimental Details
NMR spectra were recorded on solutions in CDCl3 using Bruker instruments, and chemical shifts are given in ppm to high frequency from Me4Si (H, C) or 85% H3PO4 (P) with coupling constants J in Hz. Both low-and high-resolution mass spectra were recorded using electrospray ionization. All chromatographic separation was carried out on silica gel.
Flash vacuum pyrolysis (FVP) was carried out in a conventional flow system by subliming the starting material through a horizontal quartz tube (30 × 2.5 cm), externally heated by a tube furnace to 700 °C and maintained at a pressure of 2-3 × 10 −2 torr by a rotary vacuum pump. The apparatus used is illustrated, and a detailed experimental procedure is given in a recent publication [19]. Products were collected in a liquid N2-cooled, U-shaped trap, and purified as noted.
The phosphonium salt 39 was prepared as previously described [12]. The α,β-unsaturated carboxylic acids were prepared from the corresponding aldehydes by a standard Doebner reaction with malonic acid in pyridine with catalytic piperidine. As mentioned in the introduction, one reason for trying ylide 37 was that in earlier work we observed the involvement of an ortho-methyl group in such ylide pyrolyses to give naphthalenes [16,17]. This was not observed for 37 nor for the second methyl-containing ylide 38, which simply afforded the (E)-and (Z)-thienylvinylquinolines 47, corresponding to 42 in low yield (Scheme 9). However, even in this case, the contrast with the behavior of the ylide 18 caused by introduction of the single methyl group remote from the reactive site for initial cyclization is remarkable. This fact was emphasized by the result in the case of the 2-methylstyryl ylide 37 (Scheme 8). This is the methyl analogue of 16 which gave the benzocarbazole product 19 as the only major product in 66% yield [12]. Although products derived from cyclization of a C-centered radical, the quinolines 44 and the benzophenanthridine 45, amounted together to 33%, the largest single product was now the dimethylbenzocarbazole 46, resulting from direct cyclization of an N-centered radical, formed in 22% yield. As mentioned in the introduction, one reason for trying ylide 37 was that in earlier work we observed the involvement of an ortho-methyl group in such ylide pyrolyses to give naphthalenes [16,17]. This was not observed for 37 nor for the second methyl-containing ylide 38, which simply afforded the (E)-and (Z)-thienylvinylquinolines 47, corresponding to 42 in low yield (Scheme 9). However, even in this case, the contrast with the behavior of the ylide 18 caused by introduction of the single methyl group remote from the reactive site for initial cyclization is remarkable.

General Experimental Details
NMR spectra were recorded on solutions in CDCl3 using Bruker instruments, and chemical shifts are given in ppm to high frequency from Me4Si (H, C) or 85% H3PO4 (P) with coupling constants J in Hz. Both low-and high-resolution mass spectra were recorded using electrospray ionization. All chromatographic separation was carried out on silica gel.
Flash vacuum pyrolysis (FVP) was carried out in a conventional flow system by subliming the starting material through a horizontal quartz tube (30 × 2.5 cm), externally heated by a tube furnace to 700 °C and maintained at a pressure of 2-3 × 10 −2 torr by a rotary vacuum pump. The apparatus used is illustrated, and a detailed experimental procedure is given in a recent publication [19]. Products were collected in a liquid N2-cooled, U-shaped trap, and purified as noted.
The phosphonium salt 39 was prepared as previously described [12]. The α,β-unsaturated carboxylic acids were prepared from the corresponding aldehydes by a standard Doebner reaction with malonic acid in pyridine with catalytic piperidine.

General Experimental Details
NMR spectra were recorded on solutions in CDCl 3 using Bruker instruments, and chemical shifts are given in ppm to high frequency from Me 4 Si (H, C) or 85% H 3 PO 4 (P) with coupling constants J in Hz. Both low-and high-resolution mass spectra were recorded using electrospray ionization. All chromatographic separation was carried out on silica gel.
Flash vacuum pyrolysis (FVP) was carried out in a conventional flow system by subliming the starting material through a horizontal quartz tube (30 × 2.5 cm), externally heated by a tube furnace to 700 • C and maintained at a pressure of 2-3 × 10 −2 torr by a rotary vacuum pump. The apparatus used is illustrated, and a detailed experimental procedure is given in a recent publication [19]. Products were collected in a liquid N 2 -cooled, U-shaped trap, and purified as noted. The phosphonium salt 39 was prepared as previously described [12]. The α,β-unsaturated carboxylic acids were prepared from the corresponding aldehydes by a standard Doebner reaction with malonic acid in pyridine with catalytic piperidine.

X-ray Structure Determination for 14
The compound was prepared as previously reported [12] and recrystallized from ethyl acetate/diethyl ether (1:1) to give crystals suitable for X-ray diffraction.
Crystal data for C 40

Conclusions
The results from FVP of these four ylides further emphasize the sensitivity of the outcome in these reactions on the precise combination of substituents present. As compared to the previously studied N-benzyl-N-methylamino ylides, the present compounds show an almost complete change from carbazole products derived from cyclization of N(Me) · to quinoline and phenanthridine products derived from hydrogen atom transfer and cyclization of the resulting NHCH 2 · . The results are more similar to those obtained from certain N,N-dibenzyl ylides, but even then, significant differences in product distribution are caused by the slightest change in ylide structure. The difference in pyrolysis outcome between the isomeric compounds 35 and 36 and 17 and 18 is striking, as is the effect of adding a single methyl substituent remote from the reactive site, as seen going from 16 to 37 or from 18 to 38. We conclude that, as also found in our recent synthesis of Eustifoline D using this method [13], model studies are of little value in this area, and which pyrolytic routes will predominate in a given case can only be determined by experimentation.