Mild reductive rearrangement of oximes and oxime ethers to secondary amines with hydrosilanes catalyzed by B(C 6 F 5 ) 3 †

The strong boron Lewis acid tris(penta ﬂ uorophenyl)borane, B(C 6 F 5 ) 3 , has been found to catalyze the reductive rearrangement of oximes and their ether derivatives at room temperature with hydrosilanes as the reducing agents. Cyclic substrates undergo ring enlargement, and the secondary amine products are generally formed in good yields. Control experiments combined with a DFT computational analysis of the reaction mechanism suggest that there are three energetically accessible reaction pathways (paths A – C), either or not involving hydroxylamine derivatives. Paths A and B proceed through the intermediacy of a common N , O -bissilylated hydroxylamine, and the ring-expanding rearrangement yields an iminium ion. With no intermediate at the hydroxylamine oxidation level (path C), the reaction mechanism resembles that of the Beckmann rearrangement where an O -silylated oxime converts into a nitrilium ion. The reduction – rearrangement sequence (paths A and B) is slightly preferred over the rearrangement – reduction order of events (path C), especially at ambient temperature.


Introduction
A few years ago, our laboratory disclosed the previously elusive hydrogenation of various oxime ethers to the corresponding O-protected hydroxylamines by making use of FLP-type dihydrogen activation with B(C 6 F 5 ) 3 as the catalyst 1 (Scheme 1, top). 2 The goal of that project had been to avoid the otherwise typical cleavage of the weak N-O bond to afford primary amines. 3 For this, the use of dihydrogen is critical because hydrosilanes as alternative reductants 4 engage in deoxygenation. 5 Treatment of those oxime ethers with hydrosilane/ B(C 6 F 5 ) 3 combinations results in exactly that, and these reactions were not clean but there was evidence for a reductive rearrangement. 2a Rearrangements of this sort that convert oximes and their ether derivatives into secondary amines can be achieved by the stoichiometric use of aluminum- 6 and also boron-based 7 reductants; the aforementioned reduction to primary amines is a common side reaction (Scheme 1, bottom). In the light of our earlier observations, we decided to investigate the B(C 6 F 5 ) 3 -catalyzed reductive rearrangement of oximes with hydrosilanes to turn it into a useful chemoselective methodology.

Results and discussion
We began our investigations with optimizing the reductive rearrangement of α-tetralone-derived oxime into the corresponding 2,3,4,5-tetrahydro-1H-benzo [b]azepine [(E)-1aa → 2a; Table 1]. Treatment of (E)-1aa with 4.0 equiv. of PhSiH 3 in the presence of 10 mol% of B(C 6 F 5 ) 3 in 1,2-C 6 H 4 F 2 at room temperature afforded 2a in 40% yield after 20 h (entry 1). This reaction was highly chemoselective, favoring migration of the aryl group over simple reduction; no formation of any primary Scheme 1 Reduction and reductive rearrangement of oximes and oxime ethers catalyzed by combinations of B(C 6 F 5 ) 3 and dihydrogen and hydrosilanes (n = 1-3), respectively. R 1  amine product 3a was detected (gray box). Among several solvents tested, toluene (toluene-d 8 used) brought about the best result with 95% yield and 88% isolated yield (entries 2-6). In toluene-d 8 , excellent and good yields were also obtained with MePhSiH 2 and Ph 2 SiH 2 , respectively (entries 7 and 8). Other hydrosilanes such as Et 2 SiH 2 and Me 2 PhSiH were however completely unreactive even at higher temperature (entries 9 and 10, see the ESI † for details). Reactions conducted with a shorter reaction time (entry 11), less PhSiH 3 (entry 12), or decreased catalyst loading (entry 13) resulted in lower yields. Incomplete conversion at reduced catalyst loadings could be compensated by longer reaction times.
The yield of the above reductive rearrangement was also high on a 7.0 mmol scale, and we continued exploring the reaction scope under this optimized reaction setup (Scheme 2; cf. Table 1, entry 6). Related ring enlargements worked equally well [(E)-1ba,ca → 2b,c]. The functional-group tolerance was assessed with acetophenone-derived oxime derivatives (E)-1dapa bearing a range of electron-donating and electron-withdrawing substituents on the aryl group. The four halo groups as in (E)-1ia-la were compatible with the reaction conditions but aryl iodide (E)-1la underwent partial hydrodeiodination to afford 2l in 57% yield along with 2d in 25% yield. A thioether as in (E)-1na and a free phenol as in (E)-1oa undergoing temporary silylation were tolerated as well. Exhaustive hydrodefluorination of the trifluoromethyl group in (E)-1ma occurred, furnishing 2h rather than 2m. 8 The reaction of (E)-1pa with a dimethylamino group in the para position only gave a trace amount of 2p, and the deamination product 4-ethyl-N,N-dimethylaniline was formed in 64% yield. 9 It is important to note that the configuration of the oxime does not affect the reaction outcome; 6g both (E)-1ka and (Z)-1ka yielded 2k in 98% and 93%, respectively.
We briefly investigated the reductive rearrangement of other alkyl-substituted benzylic oximes (Scheme 2). (E)-1ra and (E)-1sa with a 1°alkyl and a benzylic substituent converted selectively and in high yields into the corresponding secondary amines. However, substrates with 2°and 3°alkyl groups did lead to mixtures of rearranged products because the migrating ability of these groups (out)competes with that of the aryl group. This finding is different from previous reports. 6g The results obtained with isopropyl, cyclohexyl, and tert-butyl are presented in the ESI. † Symmetric ketoximes such as diarylsubstituted 1ta and 1ua as well as dialkyl-substituted 1va-xa reacted cleanly but for the latter a reaction temperature of 120°C was necessary to achieve high conversion. The rearrangement of unsymmetrically substituted (E/Z)-1ya furnished 2y selectively with migration of the phenethyl group. Additional substrates were subjected to the standard procedure (Scheme 3). Aldoxime (E)-1za rearranged to the aniline derivative 2z in a good yield. Moreover, diamine 5 was obtained from bisoxime (Z,Z)-4 by two-fold rearrangement in a moderate yield. We then turned our attention towards the reductive rearrangement of various oxime ethers (Scheme 4). After a slight reoptimization of the reaction conditions, we found that the reaction of O-silylated (E)-1db-de and O-alkylated ketoximes (E)-1df and (E)-1dg with 2.0 equiv. of PhSiH 3 in the presence of 2.5 mol% of B(C 6 F 5 ) 3 afforded N-ethylaniline (2d) in moderate to near-quantitative yields. The scope also includes the successful reductive rearrangement of hydroxylamine 6aa and O-silylated hydroxylamine 7de (Scheme 5). Both yielded the desired secondary amines 2a and 2d in 61% and 81%, respectively.
The hydroxylamine derivatives were not only chosen as an expansion of the method but also for their potential intermediacy in the reductive rearrangement of oximes. The above results indicate that O-silylated oximes, free hydroxylamines, and O-silylated hydroxylamines may be involved. Further evidence came from deuterium labeling when employing PhSiD 3 instead of PhSiH 3 (Scheme 6). Bisdeuteration of the α-methylene group was obtained in the rearrangement of the oxime [(E)-1aa → 2a-d 2 ] while monodeuteration was seen in the same position in the ring enlargement of the hydroxylamine (6aa → 2a-d 1 ). This suggests that oximes and their congeners are reduced to the hydroxylamine oxidation level during the course of the reaction. 2, 10 Cho, Tokuyama, and coworkers had reported the same for the related DIBAL-Hmediated rearrangement, and their computed mechanism proceeds through a three-centered transition state for the concerted aryl migration followed by hydride reduction of the resulting iminium ion 6g (silyliminium ion in the case of hydrosilane/B(C 6 F 5 ) 3 combinations).
We then conducted density functional calculations (DFT) to explore the mechanistic details of the reductive rearrangement of oximes to secondary amines. The calculations were performed at the M06-2X/cc-pVTZ//M06-2X/6-311G(d,p) level of theory 11 using the Gaussian 16 software, 12 with toluene solvation incorporated via a polarizable continuum model (PCM) at the same level of theory. 13 The reaction between oxime (E)-1aa and PhSiH 3 was chosen as a model system. Optimized geometries and free energy profiles (including the computational analysis on the relative stability of the B(C 6 F 5 ) 3 ·(E)-1aa Lewis pair and the B(C 6 F 5 ) 3 ·hydrosilane η 1 -adduct I1 14 ) involved in the catalytic cycle are provided in the ESI. † 15 As shown in Scheme 7, the reaction begins with the activation of the Si-H bond in I1 by the nitrogen atom or oxygen atom in oxime (E)-1aa 16,17 to afford N-silyliminium ion I2 ( path A) or silyloxonium ion I8 ( path B) with the borohydride [HB(C 6 F 5 ) 3 ] − as the counteranion. Our calculations indicate that the corresponding S N 2-Si transition states do not exist in the case of PhSiH 3 . 17,18 Instead, two stable trigonal bipyramidal transition complexes (rather than transition state) were located (I2′ or I8′, see Fig. S70 in the ESI †). 19 These results are different from that found in a computational study on B(C 6 F 5 ) 3 -catalyzed carbonyl hydrosilylation with a tertiary silane (Me 3 SiH). 17 Subsequent hydride transfer from the borohydride to the silyliminium ion or dehydrogenation gives the N-silylated hydroxylamine (I2 → I3) and O-silylated oxime Scheme 3 Scope II: B(C 6 F 5 ) 3 -catalyzed reductive rearrangement of aldehyde-and benzil-derived oximes. Isolated yields after purification by flash chromatography on silica gel.
Scheme 4 Scope III: B(C 6 F 5 ) 3 -catalyzed reductive rearrangement of ketone-derived oxime ethers. Yields determined by 1 H NMR spectroscopy using mesitylene as an internal standard.
Scheme 5 Scope IV: B(C 6 F 5 ) 3 -catalyzed reductive rearrangement of a hydroxylamine and an O-silylated hydroxylamine. Yields determined by ether (I8 → I9), respectively. B(C 6 F 5 ) 3 is regenerated in either of these steps. The formation of O-silylated oxime ether I9 is favored over N-silylated hydroxylamine I3 because the dehydrogenative silylation of oxime (E)-1aa is calculated to be kinetically more favorable than the CvN hydrosilylation reaction (ΔG ‡ = 11.3 versus 21.0 kcal mol −1 ), probably due to the release of dihydrogen gas. The N-silylated hydroxylamine I3 or O-silylated oxime I9 then undergoes similar dehydrogenative silylation or CvN hydrosilylation steps to form N,O-bissilylated hydroxylamine I5 through either I3 → I4 → I5 (path A) or I9 → I10 → I5 ( path B). The hydrosilylation of O-silylated oxime ether I9 requires an activation barrier of 24.1 kcal mol −1 , which is higher than that of the dehydrogenative silylation reaction of I3 (ΔG ‡ = 13.3 kcal mol −1 ). These computational results suggest that the formation of N,O-bissilylated hydroxylamine I5 proceeding through CvN hydrosilylation/Si-O dehydrogenative coupling ( path A) is the dominant reaction sequence in the current system. I5 could further react with the B(C 6 F 5 ) 3 ·hydrosilane adduct I1 to form ion pair I6, which undergoes the rearrangement involving aryl migration and N-O bond cleavage to generate the iminium ion I7 along with the disiloxane byproduct (I6 → I7, ΔG ‡ = 23.5 kcal mol −1 ). 20 Finally, I7 further reacts through a barrierless hydride transfer process from the borohydride to the iminium ion to form the reductive rearrangement product 2a-Si (gray box). The overall reaction is exergonic by 98.1 kcal mol −1 , and both path A and path B are likely related to the formation of the reductive rearrangement product because of the close activation barriers in the rate-determining steps involved ( path A, I6 → I7, ΔG ‡ = 23.5 kcal mol −1 ; path B, I10 → I5, ΔG ‡ = 24.1 kcal mol −1 ). These computational results can account for the experimental observation that oxime ethers and several hydroxylamine derivatives are also applicable to this method.
We also considered the possibility of the classical Beckmann rearrangement 21 ( path C, see Fig. S68 and S69 in the ESI † for details). This pathway proceeds through the reaction of I9 with the B(C 6 F 5 ) 3 ·hydrosilane adduct I1 to form the bissilyloxonium ion I11 with the oxygen atom as the nucleophilic center. Aryl migration and N-O bond cleavage of I11 further leads to the formation of ternary complex I12, in which both the borohydride and the disiloxane are loosely bound to the carbon atom of the nitrilium ion (see Fig. S69 † for the structural details of I12). Subsequent hydride reduction accompanied by disiloxane displacement yields the ringexpanded imine I13 along with B(C 6 F 5 ) 3 . A subsequent imine hydrosilylation 10 through N-silyliminium ion I14 furnishes the reductive rearrangement product 2a-Si (gray box). The ratedetermining step of this pathway is the aryl migration/N-O bond cleavage in intermediate I11 with an activation barrier of 25.7 kcal mol −1 . Although this process is kinetically less favorable than path A or B, we cannot exclude whether this pathway contributes to the formation of the related reductive rearrange-ment product, especially in cases where higher reaction temperatures are applied.
The direct hydride transfer from the borohydride to the nitrogen atom of intermediate I6 or I11 to eventually form the primary amine 3a is kinetically disfavoured (cf. gray box in Table 1). The corresponding barriers are estimated to be at least 65.9 and 47.6 kcal mol −1 , respectively (relative to separated reactants; see Fig. S64, S68, and S72 †). These results can account for the high selectivity in favor of the secondary amine 2a under the experimental conditions.

Conclusion
In summary, we have developed an effective catalytic reductive rearrangement of oximes and their congeners to selectively give secondary amines. Aryl-substituted secondary and primary hydrosilanes work as stoichiometric reductants, and a catalyst loading of 10 or 2.5 mol% of B(C 6 F 5 ) 3 is sufficient to secure synthetically useful yields. Quantum-chemical calculations revealed that this reductive rearrangement can follow two distinct orders of events (Scheme 7): reduction-rearrangement involving hydroxylamine derivatives ( paths A and B) or rearrangement-reduction ( path C). The ring expansion yields an iminium ion ( paths A and B) or a nitrilium ion ( path C). All pathways are energetically accessible but paths A and B are kinetically favored at room temperature; path C becomes competitive in those cases where high reaction temperatures are required. Hence, a Beckmann rearrangement ( path C) cannot be excluded but reduction followed by ring-expanding migration is more likely. The method indeed extends to hydroxylamine derivatives which is in agreement with the computed mechanism and control experiments. The new protocol avoids the previously used stoichiometric aluminum-and boron-based reductants.

Conflicts of interest
There are no conflicts to declare.