As discussed in the introduction, silylammonium borohydride was detected in the stoichiometric experiment and proposed to undergo C–N cleavage to form deamination product. Because it is unknown which one among mono-, bi-, and tri-silylammonium borohydrides (i.e., int2, int6 or int10 shown in Figure 1) is the active intermediate that leads to deamination product, we calculated three possible pathways that involve different silylammonium borohydrides. As shown in Figure 2, the reaction is initiated by the association of substrate 1a with one PhSiH₃. This step needs to overcome a Gibbs energy barrier of 14.0 kcal/mol (TS1), generating a thermodynamically unstable amine-silane complex int1. Then, the Lewis acid B(C₆F₅)₃ abstracts a hydride from int1, which is accompanied by the N–Si bond formation to afford the monosilylammonium borohydride species (int2). The activation energy barrier for the silylation process is 27.3 kcal/mol (TS2). The monosilylammonium borohydride intermediate then undergoes C–N bond dissociation (pathway 1), where borohydride (C₆F₅)₃BH⁻ acts as a nucleophile to attack the benzylic carbon of silylammonium moiety via an S _N 2-type transition state (TS3) to afford the desired deamination product A and monosilazane B. The rate-determining step of pathway one is the silylation step (TS2) and overall reaction barrier is 27.3 kcal/mol.

Alternatively, the monosilylammonium borohydride intermediate may occur dehydrogenative reaction (pathway 2, blue color), in which (C₆F₅)₃BH⁻ accepts a proton from the amine group (TS4), which releases a H₂ molecule, B(C₆F₅)₃ and mono-silylated amine int5. The Gibbs energy barrier of TS4 is 1.7 kcal/mol higher than that of TS3, indicating that the monosilylammonium borohydride intermediate prefers C–N dissociation than dehydrogenation. Follow the dehydrogenative step, int5 further reacts with another PhSiH₃ molecule and B(C₆F₅)₃ catalyst to give the disilylammonium borohydride intermediate, int6 (Figure 3). Like the monosilylammonium borohydride, the disilylammonium borohydride intermediate can undergo an S _N 2-type C–N cleavage (TS6, ΔG^‡ = 26.3 kcal/mol) to yield the deamination species A and disilazane C, completing pathway 2. Pathways one and two bifurcate at the monosilylammonium borohydride intermediate and pathway two is less favorable since the monosilylammonium borohydride intermediate favors C–N dissociation.

In the alternative pathway 3 (in red color), the dehydrogenation of disilylammonium borohydride intermediate via TS7 (ΔG^‡ = 35.4 kcal/mol) releases one H₂ molecule and a disilyated amine (int9). int9 then reacts with another molecule of PhSiH₃ and B(C₆F₅)₃ for further silylation. The silylation step proceeds via a very high energy barrier transition state, TS8, (ΔG^‡ = 51.6 kcal/mol) and yields a highly unstable trisilylammonium borohydride species int10. Finally, the C–N dissociation of the trisilylammonium borohydride via TS9 releases product A and trisilazane D. In summary, the computational results suggest the direct C–N dissociation of the monosilylammonium borohydride to generate monosilazane and deamination product (pathway 1) is the most favorable pathway for this reductive deamination reaction. The first silylation step via TS2 is the rate-limiting step and the overall energy barrier is 27.5 kcal/mol. Moreover, the TSs of dehydrogenation of both mono- and di-silylammonium borohydride are higher in energy than their corresponding TSs of C–N dissociation energy (TS3 vs. TS4 and TS6 vs. TS7), suggesting the deamination is more favorable than dehydrogenation reaction which is consistent with experimental results. The lower activation barrier of deamination can be attributed to the weaker bond strength of C–N bond compared with the N–H bond. The calculated bond dissociation energy of C–N bond is lower than that of N–H bond by 8.5 and 13.2 kcal/mol for mono- and di-silylammonium borohydride, respectively (Supplementary Table S1). Thus, the C–N bond is easier to be cleavage than the N–H bond. In addition, well-ordered π-π stacking interactions between the naphthyl ring of substrate, the phenyl ring of silane and the pentafluorophenyl group of the catalyst were identified in TS3 and TS6 (Figure 4), which help to stabilize the deamination TSs. The NCI analysis (Humphrey et al., 1996; Lu and Chen, 2012) supports the existence of attractive π-π stacking interactions in TS3 and TS6.

Based on the theoretical calculation, the most favorable pathway (i.e., pathway 1) only consumes one equivalent of PhSiH₃ to afford the deamination product A, which contradicts with the experimental observation that four equivalents of PhSiH₃ are required to achieve good yields. Because both PhSiH₃ and B(C₆F₅)₃ are Lewis acids, we envision that PhSiH₃ and B(C₆F₅)₃ may compete to bind with the amine substrate to form amine-silane ([N-Si]) and amine-boron ([N-B]) Lewis adducts, respectively. As shown in Figure 5, the binding of 1a with B(C₆F₅)₃ has a Gibbs energy barrier lower by 2.0 kcal/mol than with PhSiH₃ and leads to a very stable [N-B] Lewis adduct int11 (ΔG ^o = −11.1 kcal/mol). Thus, computational results suggest that the formation of [N–B] Lewis adduct (int11) is kinetically and thermodynamically more favorable than [N–Si] Lewis adduct (int1). The favorable formation of [N–B] Lewis adduct can be attributed to the stronger Lewis acidity of B(C₆F₅)₃. Frontier molecular orbital analysis supports that B(C₆F₅)₃ is a much stronger Lewis acid than PhSiH₃ and thus easier to accept a lone pair electron from 1a (Supplementary Figure S2).

The computational results demonstrate that when B(C₆F₅)₃ and PhSiH₃ were added with a ratio of 1:1, the amine substrate prefers to bind with B(C₆F₅)₃ and form a very stable Lewis adduct (int11) which is the resting state of catalyst and substrate. Taking the resting state int11 as the start point for the deamination reaction, the overall reaction barrier is as high as 38.6 kcal/mol (int11 to TS2), which is difficult to overcome. This result is in line with the experimental observation that int11 rather than the deamination product was obtained as main product in the stoichiometric experiments with stepwise addition of B(C₆F₅)₃ and PhSiH₃.

In the catalytic reaction with 20% mol of B(C₆F₅)₃, the use of four equivalents of PhSiH₃ afforded deamination product as main product. We speculate this is related to the concentration effect which influences the competition between the formation of int1 and int11. Based on the reaction rate equation, the reaction rate (r ₁ and r ₂) for the formation of int1 and int11 can be calculated by (Eqs 1, 2), respectively. Thus, their ratio (r ₁/r ₂) is determined by Eq. 3 which is affected by the ratio of rate constants (k ₁/k ₂) and the concentration ([PhSiH₃]/[BCF]). k ₁/k ₂ is calculated based on the Erying equation (Eq. 4) r 1 = k 1 [ 1 a ] [ P h S i H 3 ] (1) r 2 = k 2 [ 1 a ] [ B C F ] (2) r 1 r 2 = k 1 k 2   [ P h S i H 3 ] [ B C F ] (3) k 1 k 2 = k B T h e − ∆ G 1 ≠ R T k B T h e − ∆ G 2 ≠ R T = e − ∆ G 1 ≠ − ∆ G 2 ≠ R T (4)

As the activation energy difference between TS1 and TS10 is 2.0 kcal/mol (Figure 5), k ₁/k ₂ is calculated to be 0.077 at 120°C. Therefore, r ₁/r ₂ for the reaction with equivalent amount of PhSiH₃ and B(C₆F₅)₃ is as small as 0.077, suggesting the formation of int11 is dominant. However, under the catalytic reaction condition with 20% mol of B(C₆F₅)₃ and 4 equivalents of PhSiH₃, the concentration ratio [PhSiH₃]/[ BCF] increases to 20. As a result, r ₁/r ₂ is increased by 20 folds compared to that of the stoichiometric reaction with equivalent of PhSiH₃ and B(C₆F₅)₃, and thus the formation of int1 is 1.54 times faster than int11. This result is in good agreement with the experimental observation that increasing the equivalent of PhSiH₃ gradually increases the yields of silylammonium borohydride intermediate and deamination product in the stoichiometric reaction with 1 equivalent of B(C₆F₅)₃. Therefore, the excess PhSiH₃ plays a crucial role to maintain a high [PhSiH₃]/[BCF] ratio so that PhSiH₃ can compete with B(C₆F₅)₃ for binding with the amine substrate, avoiding the deactivation of catalyst and substrate.

In the original experimental work, control experiments were performed to assess the relative reactivity of a variety of benzylamines. As shown in Scheme 2, the competition reaction with one equivalent of an equimolar benzylamines mixture (1b, 1c, 1d) demonstrates that the relative reactivity of benzylamines follows the order: 1b < 1c < 1d. To understand the relative reactivity of benzyl amines with different degrees of substitution at the α-carbon atom, we calculated the reaction pathway one for all substrates. Pathway one involves three main steps, i.e., the amine-silane binding, amine silylation, and deamination (Figure 6). It is worth noting that, for substrate 1d, the silylation step (TS2d, ΔG ^‡ = 26.2 kcal/mol) remains as the rate-limiting step, like the reaction with 1a. However, for substrate 1b and 1c, the C–N bond cleavage of monosilylammonium borohydride (i.e., deamination) via TS3b (ΔG ^‡ = 28.7 kcal/mol) and TS3c (ΔG ^‡ = 27.2 kcal/mol) becomes the rate-determining step. Thus, the overall reaction energy barriers for the deamination of 1b/1c/1d are calculated to be 28.7, 27.2 and 26.2 kcal/mol, respectively, which is consistent with the experimental observed reactivity order. Figure 6 clearly shows that the C–N bond cleavage (TS3) of 1d is more favorable than 1b and 1c, which is because the reacting benzylic carbon in TS3d is stabilized by more methyl substituents, leading to stronger stabilization effect on the transition state.

In the end, we turn our attention to the reactivity of hydrosilanes, another important reactant for the deamination reaction. To study the substituent effect of PhSiH₃, we calculated the reaction of a series of PhSiH₃ derivatives that carry electron-withdrawing groups (EWGs) or electron-donating groups (EDGs) at the phenyl ring. Figure 7 summarizes the Gibbs energies of all TSs of deamination reaction with different silanes (PhSiH₃, C₆F₅SiH₃, 1,3,5-C₆H₃F₂SiH₃, 1,3,5-C₆H₃Cl₂SiH₃, 1,3,5-C₆H₃Br₂SiH₃, and 1,3,5-C₆H₃Me₂SiH₃). In all reactions, the silylation step (TS2) is the rate-determining step. Compared to unsubstituted PhSiH₃, hydrosilanes with EWGs, such as F, Cl, and Br substituents, lower the barrier of TS2 by 0.4–2.0 kcal/mol. Moreover, the formation of amine-silane complex (TS1) becomes more favorable than the generation of amine-boron Lewis product as TS1 for the EWG-substituted hydrosilanes are lower in energy than TS10 by 0.7–1.2 kcal/mol. This indicates that the reaction with these hydrosilanes may not need excess amount of silane reagent. The increased reactivity and binding affinity of hydrosilanes caused by the EWGs may because the EWGs increases the acidity of hydrosilanes which makes them more reactive toward amine substrates. On the contrary, EDGs will decrease the acidity of hydrosilane and thus lower the reactivity. Indeed, the potential energy surface of 1,3,5-C₆H₃Me₂SiH₃ lies above the energy surface of PhSiH₃.