Base-promoted isomerization of CF₃-containing allylic alcohols to the corresponding saturated ketones under metal-free conditions

Introduction

We have previously reported [1-3] an extremely efficient isomerization process of γ-trifluoromethylated propargylic alcohols 1F to the corresponding α,β-unsaturated ketones (E)-5 by the action of a weak base like Et₃N under THF reflux conditions (Scheme 1). From the mechanistic point of view, deprotonation of the propargylic proton H^a was considered to be the initial step of this reaction and this step would be in competition with the proton abstraction from the OH^b group in 1F. Thus, it was our first understanding that during the equilibration between 3F-O and 3F-C, the irreversible reprotonation by way of the latter intermediate would be occurred at the CF₃-attached carbon atom to produce the corresponding allenyl type intermediate. Furthermore, its keto–enol interconversion would result in construction of the isomerized α,β-unsaturated ketones (E)-5 [4-6]. For a better comprehension of this interesting and efficient protocol, computation was performed [7] for obtaining the rough indication of the acidity of both protons H^a and H^b in 1F. Moreover, the corresponding allylic alcohols 2F as well as their non-fluorinated counterparts 1H and 2H were also employed for comparison whose results were summarized in Table 1.

For this purpose, evaluation was performed on the basis of the energetic difference between neutral compounds and the corresponding carbanion or alkoxide species: thus, in the former instance, ΔE_C as E_3X-C − E_1X (or E_4X-C − E_2X) or the latter ΔE_O as E_3X-O − E_1X (or E_4X-O − E_2X) in the case of propargylic (or allylic) alcohols, respectively, as the simple measure of destabilization. In the case of the non-fluorinated 1H and 2H series, as expected, a clear alkoxide preference was noticed irrespective of the carbon–carbon unsaturation pattern. For the corresponding fluorinated counterparts, in spite of the failure to locate the energy minimum for 4F-Ca, it was found out in other cases that ΔE_C values were unanimously smaller than ΔE_O which led to the definitive conclusion that protons at the propargylic and allylic positions should be more acidic than the ones at the corresponding OH groups. The strongly electron-withdrawing property of a CF₃ group is considered to play a key role in stabilization of both 3F-C and 4F-C, and the phenyl group as well as the carbon–carbon multiple bonds would also provide the additional preferable effect by their efficient resonance. In the case of the propargylic alcohol 1Fa with a methyl moiety as R, its electron-donating ability unambiguously destabilized the carbanionic species 3F-Ca while the electron-withdrawal of a CF₃ group nicely compensated this disadvantage, resulting in a better stability of 3F-Ca by 1 kcal/mol with respect to the alkoxide 3F-Oa. In the case of 1Fb, the phenyl group worked nicely for increase of the energetic preference of 3F-Cb to 3F-Ob of about 7.9 kcal/mol. For the allylic alcohol 2Fb (R = Ph), although the ΔΔE value was small, the carbanionic species 4F-Cb was calculated to be more (or at least almost equally) preferable to 4F-Ob. This result allowed us to similarly anticipate the successful isomerization of CF₃-containing allylic alcohols 2F to the corresponding saturated ketones 7 when the substrates possessed appropriate aromatic substituents as R.

With reference to these computational results, we decided to employ the CF₃-containing allylic alcohols (E)-6 at the γ-position as substrates for the amine-promoted isomerization, instead of the corresponding propargylic alcohols 1F. Although a similar Ru- [8-12] or Fe-catalyzed processes [13] have been previously demonstrated, our present work is considered to draw a clear line with these instances because of the apparently convenient metal-free process. Moreover, the same type of proton shift has also reported by two groups. For example, during the reaction of (E)-2-(trifluoromethyl)vinylsilane and benzaldehyde in the presence of an excess amount of CsF, the resultant product (E)-2Fb was further converted in situ to the corresponding saturated ketone (E)-5b (R = Ph). Confirmation of this process was performed by the action of DBU, resulting in 95% conversion of the starting allylic alcohol (E)-2Fb [14]. Quite recently [15], the same proton shift was published by using the bicyclic guanidine-type base, 1,5,7-triazabicyclo[4.4.0]dec-5-ene (TBD), as a catalyst. In spite of these two preceding studies, we have also succeeded in attaining a similar level of chemical yields as well as stereoselectivity to the latter process by using far-less expensive DBU (1,8-diazabicyclo[5.4.0]undec-7-ene), whose results are reported in this article in detail.

Results and Discussion

Preparation of substrates (E)-6 was carried out by way of our recently reported one-pot procedure [16]: thus, after reaction of an appropriate Grignard reagent and ethyl trifluoroacetate at −80 °C, a Horner–Wadsworth–Emmons reagent [17] activated by LiBr and Et₃N [18-20] was introduced to this solution at room temperature, leading to the formation of (E)-6 after the NaBH₄ reduction of the resultant products (see Supporting Information File 1 for the detailed procedure).

As shown in Table 2, we at first checked a type of bases suitable for this isomerization in THF under reflux for 3 h. Although Et₃N was appropriate in the case of isomerization of propargylic alcohols 1F [1], this was not the case for the allylic alcohol (E)-6a, forming the desired 7a only in a trace amount with recovery of (E)-6a in a large amount (Table 2, entry 1). Because Et₃N was proved to be inadequate, utilization of DBU with the higher basicity was tried at the next stage although this base gave detrimental results for the isomerization of 1F. Actually, DBU was found to work nicely in the present instance to afford the desired compound 7a in 67% isolated yield (Table 2, entry 2). The result that the stronger tertiary amine DBU worked more effectively than Et₃N was a clear support of our computation at least from a qualitative point of view which estimated the lower acidity of the allylic proton rather than the one in the propargylic position (ΔΔE of −0.321 and −7.869 kcal/mol for the allylic and propargylic series, respectively). Due to the unacceptable efficiency of K₂CO₃ and NaOH (Table 2, entries 3–5), the solvent effect was surveyed at the next stage with fixing the base to DBU. Polarity seemed to affect the reaction significantly, and aprotic MeCN, DMF, and DMSO (Table 2, entries 9, 10, and 12, respectively) showed a similar potency to THF with respect to chemical yields of 7a. Further investigation clarified that toluene was the best solvent from the standpoint of the material balance, and with taking efficiency into consideration, we eventually decided that, as described in Table 2, entry 18, 3 h reflux in toluene with 0.5 equiv of DBU were the conditions of choice.

Because we have successfully determined the appropriate reaction conditions for the present intriguing isomerization, clarification of its scope and limitation was carried out whose results were summarized in Table 3. Entry 1 depicts the result already shown in entry 18 of Table 2 where the substrate (E)-6a was transformed into the saturated ketone 7a in 91% yield. A slower reaction was suspected for (E)-6b in Table 3, entry 2 because of the possible destabilizing effect of the partial anionic charge at the 3 position by the electron-donating 4-MeOC₆H₄ group as R². However, the effect was only quite limited and isomerization of (E)-6b was occurred in the same 3 h period as entry 1 with recording 89% isolated yield. Moreover, as expected, the electron-withdrawing 4-fluorophenyl moiety worked nicely to quantitatively furnish the ketone 7c in shorter time (Table 3, entry 3). However, as shown in Table 3, entries 4 and 5, clear retardation of this process was noticed for substrates with alkyl groups as R². Different from the case of R², an electronic effect of the substituent on the benzene ring of R¹ affected this transformation quite significantly, and the substrate (E)-6f with 4-MeOC₆H₄ and Ph(CH₂)₂- groups as R¹ and R², respectively, experienced “double retardation” to require 48 h for attainment of an adequate level of conversion. Because of such substituent sensitivity of R¹, alkyl groups at this position completely inhibited the reaction (Table 3, entries 8 and 9). Entry 10 described the result for the non-fluorinated substrate (E)-6j possessing a CH₃ group instead of a CF₃ moiety, and its simple recovery was observed. Exchange of the strongly electron-withdrawing CF₃ group to CH₃ led to complete loss of the good stabilizing factor of the developing carbanionic species. As a result, the alkoxide was more energetically preferred and thus this proton shift became difficult.

For the purpose of obtaining mechanistic information on the present reaction, three representative optically active substrates (R,E)-6 were selected and submitted to the identical conditions as above (Table 4). Furthermore, in the case of (R,E)-6a, conditions in entries 6 (6 M NaOH aq in MeOH) and 19 (DABCO in toluene) in Table 2 were also tried for comparison (see entries 2 and 3, respectively). Like the case of the corresponding racemic compounds, the DBU-mediated proton shift was realized in a similar fashion to afford the ketones 7 in excellent isolated yields and, moreover, accomplished a quite high degree of chirality transmission (CT) on the basis of the chiral HPLC analysis (Table 4, entries 1, 4, and 5). Unanimous formation of (R)-stereoisomers at the 3 position of 7 from (R,E)-6 led to confirmation that the proton attached to C¹ was migrated to C³ from its si face, thus from the same back side if (R,E)-6 possessed its conformation as shown in Table 4. Although DABCO attained the same level of CT (Table 4, entry 3) albeit a slower reaction rate, this is not the case for the conditions of 6 M NaOH aq in MeOH and only 21% CT was observed. The latter result would be because of the competing occurrence of intermolecular reprotonation by the solvent.

The present interesting proton shift reaction was also computationally simulated [7] by employing (R,E)-6h with the substitution pattern of R¹ = Ph and R² = Me as the model substrate. For simplicity, 1,4-diazabicyclo[2.2.2]octane (DABCO) was employed as the representative base rather than DBU. Successful location of the transition state TS-8h at the B3LYP/6-311++G** level of theory led to clear analysis that TS-8h was 27.10 kcal/mol higher in energy than the combination of the substrate (R,E)-6h and the reactant DABCO under vacuum. This barrier became lowered to 24.71 kcal/mol after the single point calculation by consideration of the solvent effect of toluene by the conductor-like polarizable continuum model (CPCM). This energy barrier is considered to qualitatively correspond to the requirement of toluene reflux temperature (110–111 °C) for promotion of the desired proton shift. The cleaving C¹···H and forming C³···H bond lengths in TS-8h were calculated to be 220.0 and 238.1 pm which were significantly elongated from the ones of the corresponding substrate (R,E)-6h and product (R)-7h of 109.5 and 109.3 pm, respectively (Figure 1). The partial charge at C³ became more negative (–0.338, –0.110, and –0.299 for TS-8h, (R,E)-6h, and (R)-7h, respectively) and H more positive (0.528, 0.172 and 0.244). The DABCO molecule with activating the proton was found to be situated right behind the C² atom with the C²···H distance of 201.6 pm. The N···H distance of 104.4 pm was found to be only 2.4 pm longer than the same bond in protonated DABCO which would be one of the major reasons why the weaker base Et₃N did not work for this reaction.

Conclusion

As shown above, our original isomerization of CF₃-containing propargylic alcohols 1F to the corresponding α,β-unsaturated ketones (E)-5 was successfully extended to the transformation of allylic alcohols (E)-6 to saturated ketones 7. Different from the previously reported transition metal-catalyzed protocols by Cahard [8,9,13] and Liu [10-12], our transformation nicely proceeded with such a convenient and tractable tertiary amine as DBU, thus under metal-free conditions like the cases of the Ando [14] and Martín-Matute groups [15]. Moreover, this intriguing proton shift was clarified to be applied for optically active allylic alcohols whose chirality was transmitted almost in a perfect fashion. These results should stem from the acidic nature of the allylic proton which was successfully estimated from our independent computation. Further utilization of this reaction is being studied in this laboratory and the results obtained will be reported in due course.