Nucleofugal behavior of a β-shielded α-cyanovinyl carbanion

Sterically well-shielded against unsolicited Michael addition and polymerization reactions, α-metalated α-(1,1,3,3-tetramethylindan-2-ylidene)acetonitriles added reversibly to three small aldehydes and two bulky ketones at room temperature. Experimental conditions were determined for transfer of the nucleofugal title carbanion unit between different carbonyl compounds. These readily occurring retro-additions via C–C(O) bond fission may also be used to generate different metal derivatives of the nucleofugal anions as equilibrium components. Fluoride-catalyzed, metal-free desilylation admitted carbonyl addition but blocked the retro-addition.


Introduction
A carbanionic fragment shows nucleofugal behavior on breaking its single-bond connection with a developing electrophilic center. Apart from deprotonation reactions or formation of a separated ion pair and other trivial examples, the (at least formally) heterolytic cleavage of C-C single bonds can provide cases of interest if it generates organometallic compounds under unusual conditions. The well-known cases of alkoxide fission [1][2][3] (top line of Scheme 1) may be viewed as a reversed formation of an alkoxide A 1 M 1 from an organometallic C-M 1 and a carbonyl compound (R 1 ) 2 C=O. Such a C-C fission (retro-addition reaction) can occur already near room temperature (rt) if the nucleofugal carbanion proves to be an electronically stabilized N≡C-CH 2 - [1] or allylic [2] species or a short-lived equilibrium component [3].
A cleavable alkoxide A 1 M 1 can be employed in at least three ways: (a) its equilibrium component C-M 1  ated alkoxide A 1 M 2 begins to release C-M 2 that would otherwise be protonated by any unconsumed portion of A 1 H. We had used this technique [3] to generate the short-lived, NMR-invisible pivaloylmetals (H 3 C) 3 C-COM from the overburdened alkoxides of tri-tert-butylacyloin that were readily cleaved with relief of internal strain as a driving force. We now report here on the above-mentioned options (a) and (c) with the focus on some nucleofugal traits of the carbanion unit of the stable [4], β-shielded α-metalated acrylonitriles.

Results and Discussion
The α-cyanoalkenyllithium 2Li (top line of Scheme 2) may be prepared [4] either by LDA-mediated deprotonation (LDA = iPr 2 NLi) of the β-shielded acrylonitrile derivative 1 or by an Sn/Li transmetalation reaction of n-butyllithium ("n-BuLi") with the α-stannyl derivative 3. Due to intermolecular LiN coordination, 2Li was deduced [4] to form a clustered ground state that showed no obvious rate anomaly with previously [4] studied electrophiles; this should also hold true here for pivalaldehyde t-BuCH=O (4, "t-Bu" = tert-butyl). The chiral adduct 7 of 4 exhibited four different NMR signals for the four methyl groups at the 1-and 3-positions. Analytically pure 7 was obtained only through distillation as a crystalline sample that decayed during an attempted recrystallization from methanol. A possible reason for this instability at a temperature far below the boiling point became clear on addition of some solid potassium tert-butoxide (KOt-Bu) to an NMR tube that contained purified 7 in DMSO at rt: After the immediate appearance of a yellow tint, the next 1 H NMR spectrum revealed that 7 was almost completely transformed into the acrylonitrile derivative 1 (bottom line of Scheme 2). This suggested that the nucleofugal carbanion unit of 2K had escaped from the potassium alkoxide 10 with formation of 2K (yellow tint) and t-BuCH=O, whereupon 2K was trapped by proton sources such as DMSO or unconsumed 7 to produce 1. As a consequence, the general procedure (GP) of adding carbonyl compounds to deprotonated 1 commends to use an acidic work-up and a strict exclusion of bases that are stronger than diluted aqueous NaHCO 3 .
Benzaldehyde (5) and cyclopropanecarboxaldehyde (6) added also readily to 2Li with production of 8 and 9, respectively. As already reported above for 7, adducts 8 and 9 showed also four CH 3 NMR signals on account of the stereogenic carbon centers C-OH. The NMR AB-type interproton 3 J splittings of the H-C-O-H moieties in 8 and 9 disclosed that intermolecular scrambling of the hydroxy protons was retarded through steric congestion.
The adduct 13 (Scheme 3) of adamantan-2-one (11) was formed (like the above aldehyde adducts) through addition at C-α (rather than at nitrogen), as shown by the δ values of the tetramethylindane part which were consistent with those of 7-9.
Complete NMR assignments for the 2-hydroxyadamantan-2-yl part of 13 were achieved with the two-dimensional NOESY and HSQC techniques at rt. Except for the OH signal, all other resonances were changed only insignificantly on cooling to −60 °C. Cleavage of 13 via the lithium alkoxide 12 (top line of Scheme 3) was fast on the laboratory time scale at rt in the solvents THF, Et 2 O, t-BuOMe, or toluene. As mentioned in the Introduction, the successful transfer of 2Li from 11 to 4 depends on the exclusion of proton sources (including the alcohol 13) that would protonate 2Li with formation of 1. Thus, the slow addition of 13 to a well-stirred solution of methyllithium (MeLi, 2 equiv) liberated gaseous CH 4 (1 equiv) so that 13 was completely consumed before the electrophile t-BuCH=O (4; 4 equiv) was introduced and furnished adduct 7 but no trace of 1. The worst case (with 1 as a preponderant product) may be encountered when the deprotonating base is added to the alcohol 13 either too slowly or in a less than stoichiometric amount. For instance (bottom line of Scheme 3), the heterogeneous, slow deprotonation of 13 in THF by the insoluble base potassium hydride (KH) afforded 11 and 1 as the only products via cleavage of the potassium alkoxide 14a and protonation of the emerging 2K by residual 13, so that the final introduction of pivalaldehyde (4) produced no adduct 7. However, the same result was also found when 13 was added to a homogeneous solution of PhCH 2 K in THF; this suggested that 13 had transferred its proton to the emerging, thermally stable [4] cleavage product 2K faster than to the residual PhCH 2 K (perhaps a problem of slow mixing). On the other hand, deprotonation of 13 by the Grignard reagent MeMgBr in THF was complete (quantitative CH 4 evolution) within 20 min at 0 °C; the subsequent cleavage reaction of the generated magnesium alkoxide 14b was very slow at rt, creating 2MgBr (Scheme 3) in the presence of t-BuCH=O (4) and therefrom 7 together with only a small amount of 1. This almost clean production of 7 agrees with the versatile [5] reactions of H 2 C=C(CN)-MgBr with many electrophiles. In most of the above trapping experiments, a slow disproportionation of t-BuCH=O was observed to generate t-BuCH 2 OH and t-BuCO 2 Li as side-products.
The adduct 18 (Scheme 4) of fluoren-9-one (15) had again a triangular ligand-binding system at C-α, as shown by the completely assigned NMR resonances. As the only unexpected contrast to 7-9 and 13, the 3-CH 3 protons were observed to absorb at a higher δ value (1.98 ppm) than that of the 1-CH 3 protons, which amounts to an increase of 0.34 ppm relative to 3-CH 3 of 13. We ascribed this to a ring-current effect by the 9´-hydroxyfluoren-9´-yl substituent that can be expected to prefer a perpendicular conformation relative to the indane rings with the 9´-OH group pointing toward 3-CH 3 . As a peculiar line-width effect, the NMR signals of 3-CH 3 ( 1 H and 13 C) and of 4-H were significantly broadened at rt but had the usual narrow line widths at −45 °C. Formation of the primary lithium alkoxide 16 (Scheme 4) was shown to be readily reversible as follows. Without protolysis and work-up of the green-colored solution, 16 was kept in THF at rt for 30 min and then treated with t-BuCH=O (4; 1.5 equiv); after a further period of 2 hours at rt, the GP work-up furnished the adduct 7 of t-BuCH=O along with the co-product fluoren-9-one (15), some fluoren-9-ol [6], and a small amount of 1, but no 18 (would derive from residual 16). Thus, 16 had released the nucleofugal carbanion unit of 2Li which was trapped by 4 to give the alkoxide 17 of 7.
In view of the apparently modest spatial demand of the α-cyano substituent in the adducts of 2Li, it seemed surprising that the following ketones generated no (or no kinetically stable) C=O adducts at rt: t-Bu 2 C=O, pivalophenone (t-Bu-C(O)-Ph), benzophenone (Ph 2 C=O) [7], bis(1-methylcyclopropyl)ketone [8], and dicyclopropyl ketone (19). Instead of an adduct of 19 to 2Li, the carboxyalcohol 21 (a known [9] product of 19 with LDA) in Scheme 5 and the acrylonitrile derivative 1 were obtained from 19; this run was conducted in the absence of LDA with a sample of 2Li that had been prepared in Et 2 O through an Sn/Li interchange reaction [4] of 3 with n-BuLi. Therefore, 2Li should have deprotonated 19 with formation of 20 that was trapped by 19 to give 21. But why was 21 not formed from 19 with the α-phenylalkenyllithium 22 (related to 2Li) that produced [10] the "normal" carbonyl adduct 23 of 19? This "normal" C=O addition reaction was obviously faster than the deprotonation of 19 and also apparently irreversible under the reaction conditions, whereas the unobserved C=O addition of 19 to 2Li might be possible yet quickly reversible with a terminating proton transfer from 19 to 2Li as shown in Scheme 5. Alternatively, carbonyl additions to 2Li might be generally retarded (relative to protonations) on account of the clustered [4] ground state structure of 2Li in solution.
Electrophilic cations (Li + or K + ) are not necessary (albeit perhaps helpful) for the addition of the carbanion unit of 2Li or 2K to carbonyl compounds: Generated through desilylation of 24 (Scheme 6) by tetrabutylammonium fluoride (Bu 4 N + F − ; ≤0.05 equiv), the ion pair 25 was trapped by pivalaldehyde (4) with formation of 7 along with a comparable amount of the alkene 1. However, the primary alkoxide product, as formed by 25 and 4, was supposedly blocked by FSiMe 3 and hence unsuitable for an analysis of the retro-addition reaction (see Supporting Information File 1).

Scheme 6:
Metal-free release of the carbanion unit in 25 and its seizure by t-BuCH=O (→ 7); Bu = n-butyl.

Conclusion
(i) The nucleofugal carbanion (α-deprotonated 1) can escape with surprising ease from alkoxides, perhaps with some assistance by a metal cation if present. Conversely, metal cation assistance was not necessary for the rapid carbanion release from the α-silyl compound 24 in the presence of Bu 4 N + F − in catalytic amounts.
(ii) Most of the above alkoxide fission reactions were conducted in the presence of an electrophile for trapping the released nucleofugal carbanion; this provided a first evidence for the retro-addition process with an alkenylmetal intermediate. These fissions were slow in case of the magnesium alkoxide but rapid for the lithium or potassium alkoxides at ambient temperatures.
(iii) The alternative trapping [3] of the carbonyl component by means of a nucleophile was not tried here but might serve to accumulate the organometallic equilibrium component for the purpose of spectroscopic characterization or X-ray diffraction analyses.
(iv) Fission of the lithium alkoxide of the adduct 13 of adamantan-2-one appeared to be comparably fast on the laboratory time scale in the solvents THF, Et 2 O, t-BuOMe, or toluene. A similarly weak solvent dependence had been observed [4] for the heterolytic cis/trans stereoinversion of 2Li.
(v) For more profound mechanistic investigations, one should be aware of the established [4] clustered ground state of 2Li and the possibility that clustered species might be the reactive components even in THF solutions, whereas the retro-addition processes may perhaps involve monomeric nucleofuges.

Experimental
General remarks. Stoichiometric formation of alkoxides through deprotonation of the alcohols was conveniently performed with methyllithium or methylmagnesium bromide and gas-volumetric control of the liberated methane (ca. 25 mL/mmol at 22 °C). All 1 H and 13 C NMR shifts δ were referenced with internal Me 4 Si. NMR abbreviations were as follows: d = doublet, m = multiplet, q = quartet, quat = quaternary, s = singlet, t = triplet.