LiAlH4‐catalyzed Imine Hydrogenation with Dihydrogen: New DFT Mechanistic Insights

Recently, it was found that the widely used stoichiometric reducing reagent LiAlH4 can also be used as a useful catalyst for imine hydrogenation with H2 under relatively mild conditions. In this work, extensive state‐of‐the‐art DFT calculations are conducted to explore the detailed catalytic mechanism. In sharp contrast to the recent proposal involving heterolytic H2‐activation over Al−N amide bonds after multiple imine insertion, our new mechanism highlights the dual role of the Lewis‐acidic lithium cation in frustrated‐Lewis‐pair‐like activation of both imine substrate and H2 when combined with the aluminum‐bound hydride and amide ligands, respectively. Ionic Li−N amide bonds are about 13 kcal/mol more reactive than Al−N amide bonds for H2 activation, thus providing a useful structural feature for rational design of active hydrogenation catalysts.

too high to account for the observed mild reactions conditions, thus putting some doubt on the proposed catalytic mechanism.
To gain more general mechanistic insight into the LiAlH 4catalyzed imine hydrogenation with H 2 , extensive state-of-theart dispersion-corrected DFT calculations at the PW6B95-D3 + COSMO-RS//TPSS-D3 + COSMO level in implicit THF solution (see below for computational details) are conducted using the typical imine substrate PhCH = NtBu (Im). The inclusion of COSMO solvation in geometry optimization is essential to identify potential ionic reactions (such as the formation of separated ion pairs) that are highly disfavored in the gas-phase. In this work, multiple imine insertion and H 2 addition reaction steps are examined in order to identify the true catalytic species in solution. In contrast to recent dispersion-uncorrected DFT calculations based on gas-phase structures, [9] a rather different catalytic mechanism is found in our better DFT calculations including well-defined and chemically reasonable intermediates and novel H 2 activation steps (Scheme 1, bottom).
Our DFT calculations clearly show that a doubly-bridged Li/ Al complex is preferred in solution when the Lewis-acidic Li center is coordinated by imine PhCH = NtBu, but separated ions including Li(THF) 4 + are more stable by about 4 kcal/mol due to a higher affinity of THF to the lithium cation. Even more importantly, bridging LiÀ N amide bonds are shown to be much more reactive than AlÀ N amide bonds for H 2 activation, with decreasing reactivity with increasing steric hindrance due to multiple imine insertion. Such mechanistic insights can be useful for rational design of efficient hydrogenation catalysts.
As shown in Figure 1, possible solution structures of LiAlH 4 coordinated by bulky imine PhCH = NtBu (Im) and THF molecules are examined at first. Without coordinating THF molecules, LiAlH 4 should exist as imine-coordinated, double-hydridebridged bimetal complexes of Im 2 LiH 2 AlH 2 and ImLiH 2 AlH 2 that are nearly degenerate in free energy at room temperature, which should be favored in neat imine solution and under moderate heating, respectively. The separated ion pair of Im 2 Li + and AlH 4 À is 5.0 kcal/mol higher in free energy; further imine coordination to yield the cation Im 3 Li + is 3.3 kcal/mol endergonic and thus thermodynamically unfavorable even in neat imine solution. Triple-hydride-bridged structures found in gasphase DFT optimization [9] are unstable in solution and spontaneously optimized into double-hydride-bridged structures in our DFT calculations. With more coordinating THF molecules in solution, exergonic Im/THF ligand exchange and additional THF coordination at Lewis-acidic Li + may occur easily via various neutral and/or ionic channels, eventually leading to the stable separated ion pair of Li(THF) 4 + and AlH 4 À in solution. Since bimetal LiHAl hydride and Li[N]Al amide bridges are crucial for the respective imine insertion and H 2 addition reaction steps (as shown below), stronger stabilization of Lewis-acidic Li + with THF seems to be responsible for the relatively lower catalytic reactivity of LiAlH 4 observed in THF than in imine solutions. [9] As shown in Figure 2, starting from the double-iminecoordinated complex Im 2 LiH 2 AlH 2 , intramolecular hydride to imine carbon transfer may occur via the six-membered-ring transition structure TS1 over a free energy barrier of 24.9 kcal/ mol, which is À 14.8 kcal/mol exergonic to form the hydride/ amide doubly-bridged complex ImLi[N]HAlH 2 . Further H 2 addition to the LiÀ N amide bond of ImLi[N]HAlH 2 is 4.7 kcal/mol endergonic over a barrier of 24.1 kcal/mol (via TS2) to form the desired amine product PhCH 2 NHtBu (H[N]) along with the initial ImLiH 2 AlH 2 complex, formally completing the catalytic cycle of imine hydrogenation. Indeed, as the reverse reactions of metal amide hydrogenation, primary and secondary amines may react smoothly with LiAlH 4 to give aluminum amide products and H 2 . [12] In contrast, a high free energy barrier of 35.4 kcal/mol (via TS2a) is found for the H 2 addition to the terminal AlÀ N bond of ImLiH 2 AlH[N] that is easily accessible from ImLi[N]HAlH 2 but 2.2 kcal/mol higher in free energy; a comparable high free barrier of~37 kcal/mol was recently found for similar H 2 addition to AlÀ N bonds. [9] The FLP-like H 2 -activation over reactive ionic LiÀ N bonds found in this work is also consistent with our recent findings of reactive alkali metal species in the activation of small molecules such as CO and H 2 . [13] The new catalytic cycle involving ImLiH 2 AlH 2 and ImLi[N]HAlH 2 as the respective imine and H 2 activator is À 10.1 kcal/mol exergonic with an overall barrier of 24.9 kcal/mol, which should be efficient under moderate heating.
The actual situation is further complicated by potential imine insertion reactions of meta-stable ImLi[N]HAlH 2 . As shown in Figure 2 [9] the second imine insertion step is indeed kinetically 1.0 kcal/mol more favorable than the first one via TS1. Note that the imine insertion of ImLi[N]HAlH 2 (via TS3) is kinetically comparable with the competing H 2 addition (via TS2), but thermodynamically much more favorable. Since the final Gibbs free energies are computed for all species at 298 K and 1 mol/L reference concentration while under the applied experimental conditions (1-6 bar H 2 pressure in neat imine PhCH = NtBu solution), [8][9] the actual concentrations of H 2 and imine are lower and higher than 1 mol/L, respectively, leading to additional driven force of about 2 kcal/mol for imine insertion rather than H 2 addition. Even though, our DFT calculations suggest that LiAlH 4 may act as the actual imine hydrogenation catalyst under high H 2 pressure and relatively low reaction temperature conditions. As shown in Figure 2 [8][9] to realize the slow third imine insertion that is unlikely to be important in an efficient catalytic imine hydrogenation. According to our DFT calculations, the LiÀ N amide bonds of ImLi [N] 2 AlH 2 can also be directly involved in H 2 activation, which is however kinetically 1.0 kcal/mol less efficient than that via less stable ImLi[N]HAlH[N] (see ESI, TS3a). After the third imine insertion, further H 2 addition over the LiÀ N amide bond of ImLi [N]HAl[N] 2 is however prevented by an even higher barrier of 33.3 kcal/mol, likely due to steric hindrance (see ESI, TS6).
Interestingly, the cyclic complex ImLi[N] 2 Alc (formally resulted from ImLi[N] 2 AlH 2 via dehydrogenative ortho-phenyl metallation with two endocyclic AlÀ C bonds) was found to be unreactive with H 2 . [9] Our DFT calculation show that the H 2 addition to a LiÀ N bond of ImLi[N] 2 Alc is 15.4 kcal/mol endergonic over a moderate barrier of 22.9 kcal/mol (see ESI, TS7), which may cause facile H-isotope exchange at phenyl orthosites due to kinetically more favorable H 2 -release to regenerate the reactive LiÀ N bond. In contrast, the endocyclic AlÀ C bonds of ImLi[N] 2 Alc are shown to be even more stable than AlÀ N bonds towards H 2 -addition (see ESI, TS8 vs. TS2a). In this sense, the double-amide-bridged complex ImLi[N] 2 AlH 2 acts actually as a kinetic trap during the catalysis, with its slow hydrogenation with H 2 being crucial to recover the hydride/amide bridged bimetal species ImLi[N]HAlH 2 as an actual catalyst.
In conclusion, extensive state-of-the-art DFT calculations for LiAlH 4 -catalyzed imine PhCH = NtBu (Im) hydrogenation in solution clearly show that hydride/amide bridged Li/Al species serve as active bimetal catalysts for imine hydrogenation with H 2 , with the combinations of Lewis-acidic Li + and Lewis-basic hydride and amide groups acting as FLP-like active sites for imine insertion and heterolytic H 2 activation, respectively. The Li/Al bimetal complexes resulted from the first and the second imine insertions of LiAlH 4 are the actual catalyst and the kinetic trap, respectively, with relatively slow catalyst recovery via H 2 addition. In contrast to recent proposal with AlÀ N amide bonds for H 2 activation, ionic LiÀ N amide bonds are much more reactive thus may represent a useful structural feature for active hydrogenation catalysts.

Computational Methods
All DFT calculations are performed with the TURBOMOLE 7.4 suite of programs. [14] The structures are fully optimized at the TPSS-D3/ def2-TZVP + COSMO (THF) level, which combines the TPSS meta-GGA density functional [15] with the BJ-damped DFT-D3 dispersion correction [16] and the def2-TZVP basis set, [17] using the Conductorlike Screening Model (COSMO) [18] for THF solvent (dielectric constant ɛ = 7.58 and diameter R solv = 3.18 Å). The density-fitting RI-J approach [19] is used to accelerate the calculations. The optimized structures are characterized by frequency analysis (no imaginary frequency for true minima and only one imaginary frequency for transition states) to provide thermal free-energy corrections (at 298.15 K and 1 atm) according to the modified ideal gas-rigid rotorharmonic oscillator model. [20] More accurate solvation free energies in THF are computed with the COSMO-RS model [21] (parameter file: BP_TZVP_C30_1601.ctd) using the COSMOtherm package [22] based on the TPSS-D3 optimized structures, corrected by + 1.89 kcal/mol to account for the 1 mol/L reference concentration in solution. To check the effects of the chosen DFT functional on the reaction energies and barriers, single-point calculations at both TPSS-D3 [15] and hybrid-meta-GGA PW6B95-D3 [23] levels are performed using the larger def2-QZVP [17] basis set. The final free energies are determined from the electronic single-point energies plus TPSS-D3 thermal corrections and COS-MO-RS solvation free energies. As also noted previously, [24] the overall results from both DFT methods are in good mutual agreement (À 0.2 � 1.5 kcal/mol, mean � standard deviation for all relative energies) though as expected 2.1 � 0.7 kcal/mol higher reaction barriers are found at the PW6B95-D3 level. In our discussion, the more reliable PW6B95-D3 + COSMO-RS free energies (in kcal/mol, at 298.15 K and 1 mol/L concentration) are used unless specified otherwise. The applied DFT methods in combination with the large AO basis set provide usually accurate electronic energies leading to errors for reaction energies (including barriers) on the order of typically 1-2 kcal/mol and even better for relative barrier differences between similar reactions due to error cancellation. This has been tested thoroughly for the huge data base GMTKN55 [25] which is the common standard in the field of DFT benchmarking.