SE2 reaction in noncarbon system: Metal-halide catalysis for dehydrogenation of ammonia borane.

An electrophilic substitution (SE) reaction of BN isosteres has been investigated for the dehydrogenation of ammonia borane (AB) by metal chlorides (MCl2) using various ab initio calculations. In contrast to the typical SE reaction occurring at the carbon atom, the nitrogen atom in AB serves as the reaction center for the SE reaction with the boron moiety as the leaving group when the MCl2 approaches the AB. The SE2 backside reaction is favored as a trigger step for the dehydrogenation of AB by the MCl2 The SE2 reaction is found for 3d-transition-metal chlorides (e.g., FeCl2, CoCl2, NiCl2, CuCl2, and ZnCl2), while PdCl2 leads to the dehydrogenation of AB by a direct B-H σ-bond activation, similar to most organometallic catalysts. Interestingly, the polymerization of AB promoted by MCl2 can be explained with the similar SE2 mechanism, and the dehydrogenation of the BN derivative 3-methyl-1,2-BN-cyclopentane (CBN) bearing a carbon backbone ring also follows the SE2 reaction. In particular, the experimental observation that the use of metal-chloride catalysis decreases the by-products obtained during the hydrogenation of AB can be explained by our mechanism involving the SE2 reaction. This work is helpful for the development of novel metal-halide catalysts for practical hydrogen storage materials, including the BN moiety.

Substitution reactions, a class of chemical reactions in which one functional group in a chemical compound is replaced by another functional group, are ubiquitous in organic chemistry (https://en.wikipedia.org/wiki/Substitution_reaction). Substitution reactions are classified either as nucleophilic (SN) or electrophilic (SE), depending on the underlying mechanism. Although SE reactions are less common than SN reactions, the SE reaction mechanism is more complicated, indicating that the variation among these processes is wider, which in turn means that there is a higher probability of unknown reactions (1–5). Therefore, whereas the SN reaction has been usually observed in carbon systems, the SE reaction would be able to occur in noncarbon systems.

Recently, ammonia borane (AB) has received significant attention as a promising onboard hydrogen storage material for use in transportation due to its high hydrogen storage capacity (19.6 wt % H2) (6–9). There has been considerable effort to lower the operating temperature for fuel cells to 80 °C; consequently, homogeneous catalysis has become a promising alternative means of lowering the activation energy barrier of the dehydrogenation reaction pathway (10–17). In particular, organometallic catalysts lower the dehydrogenation barrier of AB by the coordination of the hydrogen atoms bonded to boron or nitrogen to the metal centers of the catalyst (16), where the coordination activates the B–H or N–H bond, followed by hydrogen transfer to the metal center or ligands. Recently, the dehydrogenation mechanism of AB and amine borane by various complex metal hydrides has been thoroughly reviewed by Rossin and Peruzzini (18). As more cost-effective promoters than organometallic catalysts, metal halides have been considered as homogeneous catalysts for AB dehydrogenation (19–21). Homogeneous metal-halide catalysis has also been applied to improve the dehydrogenation of 3-methyl-1,2-BN-cyclopentane (CBN), which is a promising liquid-phase H2 storage material (22). For catalytic reactions with metal halides, it has been suggested that similar to the case of organometallic catalysis, the metal cation in metal halides initiates AB dehydrocoupling and retains a B–N unit during AB dehydrogenation (19, 20); however, the details of this proposed mechanism have not been investigated further.

While extending our search for an optimal metal-halide catalyst for the efficient dehydrogenation of AB, we have thoroughly explored the reaction pathway and kinetics of the dehydrogenation process using ab initio calculations. These efforts have led to the finding that an SE2 reaction occurs between the metal center and AB at the initial step of dehydrogenation and determines the performance of metal-halide catalysts, showing the possibility of an SE2 reaction in a noncarbon system. In contrast to previous SE reactions, the possibility of such a reaction on the B–N unit in an AB molecule can provide an opportunity for BN isostere chemistry, particularly with regard to the replacement of two carbon atoms (C–C) by the B–N moiety to produce various structures with interesting electronic structure and chemistry (23–27). In this work, the catalytic effects of various metal halides in the SE2 reaction for AB dehydrogenation are explored by several ab initio calculations, and their catalytic activity is discussed on the basis of quantum-mechanical energetics and molecular orbital (MO) analysis. Our calculations are also supported by a comparison with previous experimental results. Furthermore, it is confirmed that a similar SE2 mechanism is also applicable to catalysis for the CBN system, which has been recently reported as a potential hydrogen storage medium (22).

Results and Discussion

Direct Hydrogen Activation.

According to a previous study (28) of organometallic catalysts for AB dehydrogenation, the dehydrogenation reaction occurs through the activation of a B–H or N–H bond at a metal center. Thus, for the metal-halide catalysts considered in this work, we first investigated the dehydrogenation mechanism occurring through B–H activation, similar to the behavior observed in the case of organometallic catalysts. shows the AB dehydrogenation reaction pathways via B–H activation of metal chlorides, where FeCl2 and PdCl2 are used as representatives of 3d- and 4d-metal halides, respectively. For FeCl2 (), the iron atom binds with two σ-B–H bonds in the AB molecule with a binding energy of 29.6 kcal/mol. However, a careful investigation reveals that no further reactions occur. This indicates that the dehydrogenation via B–H activation is not favorable, in contrast to the experiments (21, 29) reporting that FeCl2 can promote the dehydrogenation of AB. We also investigated other 3d-metal halides, such as CoCl2, NiCl2, CuCl2, and ZnCl2, and found that similar to the FeCl2 case, these metal halides cannot activate the B–H bond.

On the other hand, PdCl2 can activate a B–H bond in AB (). PdCl2 interacts with one σ-B–H bond, that B–H bond is shifted toward a neighboring H atom, and then a H2 molecule is released in a step with an energy barrier of 12.9 kcal/mol. According to a previously reported experiment (21), the grinding of AB and PdCl2 does indeed lead to H2 generation. It is expected that the dehydrogenation reaction would occur via the mechanism shown in . This difference between FeCl2 and PdCl2 results from the fact that Fe2+ is a hard acid, while both Pd2+ and BH3 are soft acids (30). Thus, PdCl2 and BH3 can interact more readily than FeCl2 and BH3.

SE2 Reaction at a B–N Unit.

Although our calculations indicate that direct B–H activation in AB by nonprecious metal halides is not favorable, several experiments reported their catalytic activity (19–21, 29), implying the possibility of an alternative route for the dehydrogenation of AB via MCl2. The electropositive nature of the metal centers in the metal halides and the electron-rich nature and attachment to a potential leaving group, BH3, of the nitrogen atom in AB support the occurrence of an SE-type reaction in the AB and MCl2 system for AB dehydrogenation, especially for the occurrence of an SE2-type pathway over an SE1-type pathway.

Although the SN2 reaction occurs mostly through one route with rare exceptions (31), SE2 reactions can proceed through various routes (2). Since there is no severe steric hindrance at the NH3 or BH3 group in AB, two plausible pathways for the MCl2 + AB system can be considered: (i) the front-side open-retention pathway, SE2(FS), and (ii) the backside inversion pathway, SE2(BS). Fig. 1 shows the SE2(FS) and SE2(BS) pathways for FeCl2 + AB system as calculated at the density-functional theory (DFT) with a dispersion correction for two-body energy (D2) level of theory. We have confirmed that the DFT-D2 reaction energies are similar to the coupled-cluster singles and doubles plus perturbative triples [CCSD(T)] energies, as shown in . For the SE2(FS) pathway, the Fe atom in FeCl2 approaches the B–N bond, leading to an increase in the length of B–N bond from 1.63 to 2.80 Å, a reaction that is uphill by 15.9 kcal/mol and has an energy barrier of 16.6 kcal/mol. At the final state, full dissociation of the B–N bond did not occur because the H atom in BH3 is bound to the Fe atom by electrostatic interactions. A further reaction does not occur, indicating that the SE2(FS) reaction does not lead to the dehydrogenation of AB.

Fig. 1.

Two plausible SE2 reaction pathways for AB dehydrogenation with the help of FeCl2. Upper and lower routes are SE2(FS) and SE2(BS), respectively. Numerical values in each structure are relative energies (in kilocalories per mole) to separate FeCl2 and AB as a reference. In each structure, the B–N and M–N distances in angstroms are included.

On the other hand, for the SE2(BS) reaction, the Fe metal center in FeCl2 attacks the N atom in AB through the backside route, followed by the BH3 group leaving. The transition state (TS) for the reaction involves an inversion of the tetrahedral geometry at the N center, as observed in typical substitution reactions. The calculated energy barrier is 5.7 kcal/mol, much lower than that for the SE2(FS) route. Moreover, the reaction is slightly downhill. After the SE2 reaction, an additional process involving the recombination of the B–N bond can occur due to the absence of steric or rotational hindrance between the two dissociated groups. The second activation energy for the dehydrogenation is 20.8 kcal/mol, much lower than that for AB in the absence of a catalyst (34.2 kcal/mol). A long distance (2.87 Å) between the Fe–N moiety and the BH3 leaving group is the origin of the barrier-lowering effect, while in the noncatalyst case, a considerable contribution of the activation energy results from repulsion inducing the rotation and bending of the N and B atoms in AB (32, 33). Here, BH3 can be an intermediate product during the dehydrogenation reaction of MCl2 + AB. However, because BH3 itself is a very active species, it combines with another BH3 to dimerize into B2H6. According to a previous experiment (29), B2H6 was indeed observed as a by-product during the dehydrogenation reaction of AB with various MCl2 catalysts (e.g., FeCl2, CoCl2, NiCl2 CuCl2, and ZnCl2). This result clearly supports our SE2 mechanism.

Comparison of Various Metal-Halide Catalysts for the Dehydrogenation of AB.

At this point, it is plausible that the catalytic reaction pathway through the SE2(BS) route using FeCl2 lowers the energy barrier to 20.8 kcal/mol (Fig. 1), enabling the reaction to proceed near 80 °C. However, according to an experimental report (29), not all metal halides are effective for the dehydrogenation of AB, and the catalyst efficiency is correlated with the intrinsic properties of the metal center, such as electronegativity. Although a higher electronegativity generally leads to more efficient catalysis, several exceptions to this rule are found for the first-row transition metals. For example, the electronegativities of Co and Ni are higher than that of Fe; however, their catalytic efficiencies are lower than that of Fe (29). Therefore, we need to further explore the dehydrogenation process involving the SE2(BS) reaction pathway for various metal halides.

Using CCSD(T) calculations, we investigated reaction pathways involving the SE2(BS) reaction for AB dehydrogenation by several first-row transition-metal chlorides (FeCl2, CoCl2, NiCl2, CuCl2, and ZnCl2) and compared their reaction energies in Table 1. The information is also summarized in more detail in . Here, it is necessary to mention that the dehydrogenation process is a two-step reaction involving the SE2(BS) mechanism and the recombination of the B–N bond (RCBN). According to Table 1, the second step corresponding to the recombination reaction of the B–N bond is the rate-determining step for all first-row MCl2; however, irrespective of the type of MCl2, the relevant energy barrier is ∼20 kcal/mol. On the other hand, the first energy barrier concerning the SE2 reaction is more sensitively affected by the identity of MCl2. According to the reaction kinetics, the reaction rate for a two-step reaction (fast followed by slow) is proportional to kslow(kfast_front/kfast_back) (34), indicating that the fast reaction can be a critical factor for the determination of the reaction rate in the case of similar energy barriers for slow reactions. Therefore, the SE2 reaction can determine the AB dehydrogenation rate, even though the energy barrier for the SE2 reaction is lower than that of the B–N bond recombination reaction. Experimentally (29), H2 desorption peak temperatures from AB by the MCl2 catalysts show the following trend: ZnCl2 > CoCl2 ∼ NiCl2 > FeCl2 > CuCl2, indicating that the CuCl2 is the best candidate (Fig. 2). Indeed, this trend is well-correlated with the calculated SE2 energy barrier, where the barrier order is ZnCl2 > CoCl2 > NiCl2 > FeCl2 > CuCl2.

Table 1.

Comparison of the calculated energy barriers for AB dehydrogenation by various MCl2

MXn	SE2(BS)	RCBN	ΔE(2)N → B*	ΔEΠu*
FeCl2	6.9	22.1	19.6	6.9
CoCl2	10.9	19.9	29.7	8.2
NiCl2	8.3	20.5	30.6	7.5
CuCl2	4.0	20.7	32.4	5.6
ZnCl2	13.0	19.8	15.9	8.2

Mechanism is similar to that shown in Fig. 1 and . Dehydrogenation is a two-step reaction including the SE2(BS) and the recombination of B–N bond (RCBN). The energy barrier value is relative to the isolated AB and MCl2. The energy difference (ΔEΠu*) between the πu* MO level of single MCl2 and the HOMO level at the TS for the SE2 reaction is also compared. The energies are obtained at the calculation level of CCSD(T). All energies are in kilocalories per mole.

Fig. 2.

Correlation between the experimental (29) H2 desorption peak temperature (left y axis) and the SE2(BS) activation energy (right y axis) for AB dehydrogenation of various first-row transition-metal chlorides.

To confirm that the same mechanism for dehydrogenation of AB and metal halide also occurs in the solid state, periodic slab calculations were additionally performed, in which a nudged elastic band calculation () and a first-principles molecular dynamics simulation () were considered. The relevant computation details are described in . Indeed, both of the slab calculations clearly show that the dehydrogenation of AB involving the SE2 reaction can still occur in solids of AB and a metal halide.

To elucidate the effects of the MCl2 catalysts on the SE2 reaction barrier, we have also investigated the MCl2 electronic structures. Because this SE2 reaction accompanies a charge transfer similar to a general donor–acceptor interaction, the second-order perturbation stabilization energies from i to j* [ΔE(2) → ] were calculated using a natural bonding orbital (NBO) method, in which the interaction between occupied and vacant orbitals is represented to stabilize the superposition of each orbital. From the calculation, we find that the TS energy for the SE2 reaction is mainly determined by an intrinsic energy-lowering effect of the donor–acceptor interaction between a lone-pair orbital of nitrogen in AB and an unfilled lone-pair orbital of boron, as seen in . Calculated ΔE(2)N → B* values are summarized in the third column of Table 1. A comparison of the ΔE(2)N → B* values between MCl2 catalysts shows that the energy is higher for a metal with a larger electronegativity (Cu > Ni > Co > Fe > Zn). Moreover, the energy inversely correlates with the energy barrier of the SE2 reaction, with the exception of FeCl2. In other words, the ΔE(2)N → B* of FeCl2 is expected to fall between those of CuCl2 and NiCl2 because the order of the SE2 reaction barriers is CuCl2 < FeCl2 < NiCl2 < CoCl2 < ZnCl2; however, it is in fact lower than those of CoCl2 and NiCl2. The exceptional case of FeCl2 can be explained by another effect that is discussed below.

During the SE2 reaction, a linear symmetry of the MCl2 molecule is broken by interacting with the lone-pair orbital of the nitrogen atom (), which reveals that the SE2 energy barrier is also associated with an energy to retain this linear symmetry of the MCl2 molecule. The πu* near the highest-occupied MCl2 MO (HOMO) level has a suitable symmetry for the interaction with the AB MO, as seen in . It can be inferred from the above discussion that the energy barrier of the SE2 reaction would be associated with the energy-level difference (ΔEΠu*) between the πu* MO level of MCl2 and the HOMO level of MCl2 at the TS for the SE2 reaction (). The ΔEΠu* is also summarized in Table 1. Here, it is noticeable that the ΔEΠu* of FeCl2 is larger than that of CuCl2; however, it is lower than those of CoCl2 and NiCl2. In other words, although the energy-lowering effect of the donor–acceptor interaction in FeCl2 is less significant than that in CoCl2 or NiCl2, FeCl2 shows a stronger stabilization effect of the TS by the πu*-orbital interaction than that in CoCl2 or NiCl2, which can make the SE2 reaction barrier with FeCl2 lower than those with CoCl2 and NiCl2. Among the MCl2 compounds considered in this work, CuCl2 has the highest ΔE(2)N → B* and the lowest ΔEΠu*, which leads to the lowest SE2 reaction barrier.

Expansion of the SE2 Reaction to the Polymerization of AB and Other BN Derivatives.

Previous experiments reported that after the removal of a single H2 molecule from AB, NH2BH2 oligomers, including dimers, trimers, pentamers, and other intermediates with the empirical formula (NH2BH2), are generated (7, 35). Zimmerman et al. (36) reported that based on ab initio calculations, amine borane (NH2BH2) can serve as an autocatalyst for AB oligomerization by the addition of AB across the B=N double bond of amine borane with an activation energy of 29.5 kcal/mol. However, in the presence of the MCl2-type catalysts, the AB dimerization and even subsequent polymerization can be explained by the SE2 reaction framework in a similar manner to that for the monomer, as illustrated in Fig. 1 and . Fig. 3 shows a calculated reaction pathway of AB dimerization with an FeCl2 catalyst through the SE2 reaction. Following the removal of a single H2 molecule from AB (m-Opt2 in ), the boron atom can become a radical center with a positive partial charge, allowing it to be reused as an electrophile. The boron site in m-Opt2 attacks the nitrogen atom of another AB molecule (d-TS1 structure in Fig. 3), leading to the B–N bond elongation in the AB with an inversion (d-Opt2). The energy barrier for the SE2 reaction is 16.3 kcal/mol, which is higher than 5.7 kcal/mol in the case of the AB monomer. Here, for the final product (d-Opt3), the geometry of the dimer is not cyclic but rather a twisted zigzag type, which could be regarded as the reason why the quantities of side products are reduced by using MCl2 catalysts (29).

Fig. 3.

Calculated reaction pathway for AB dimerization with the help of FeCl2 catalysis, where the DFT-D2 method was used.

It is also known that the FeCl2 catalyst improves the dehydrogenation property of the CBN (22); therefore, we have explored the reaction pathway involving the SE2 reaction in Fig. 4. Overall, the reaction pathway is similar to the case of AB, shown in Fig. 1. The Fe atom in FeCl2 attacks the N center in CBN with simultaneous dissociation of the B–N bond. Despite the hindrance of the carbon backbone, the CBN ring can be opened through the B–N bond dissociation caused by the SE2(BS) reaction. This reaction is thermodynamically favorable by 14.5 kcal/mol. In particular, the energy barrier for this reaction is as small as 4.4 kcal/mol, which is smaller than that in the case of AB (5.7 kcal/mol) at the same calculation level (DFT-D2). The lower energy barrier in CBN results from the increased repulsive force induced by the eclipsed conformation of the hydrogen atoms between B and N in the TS. The RCBN step is thermodynamically exothermic by 14.1 kcal/mol, and the energy barrier is 24.6 kcal/mol. We denoted the solvation effects on the reaction pathway shown in Fig. 4 by red numbers; however, the effect is not significant.

Fig. 4.

Calculated reaction pathway for CBN dehydrogenation with the help of FeCl2 catalysis, where the calculation is performed at the DFT-D2 level. Red numbers in parentheses indicate the energies considering the solvent effect.

Conclusion

Thus far, we have elucidated the dehydrogenation mechanism of AB by the MCl2 catalysts using several ab initio calculation methods. A type of electrophilic substitution reaction in the B–N unit, in particular, an SE2 type reaction, was found during the catalyzed dehydrogenation reaction of AB catalyzed by MCl2. The SE2 reaction determines the catalytic activity of the metal chlorides for the dehydrogenation of AB, as supported by the reported experiments. Interestingly, a similar mechanism is also found in the AB polymerization reaction and even in the dehydrogenation reaction of a BN derivative with a carbon backbone ring (i.e., CBN). This work helps the development of metal-halide catalysts for practical hydrogen storage materials with the BN moiety. We also expect that this work could pave the way for the use of a new reaction in the field of BN chemistry.

Models and Methods

The catalytic effects of various metal chlorides (MCl2 type, M = 3d- or 4d-transition metals) on the dehydrogenation reactions of AB were investigated with several ab initio calculation methods. To appropriately describe the long-range correlation effect on the substitution reactions occurring during the dehydrogenation, all MCl2 adducts were fully optimized in the Becke, three-parameter, Lee–Yang–Parr (B3LYP) (37, 38) DFT framework, with the 6–311++G(d,p) basis set (39) plus Grimme’s empirical dispersion correction, called the DFT-D2 method (40). A TS for the dehydrogenation reaction was confirmed by the existence of one imaginary frequency, and the zero-point energy correction to the total energy was applied to the total energy.

When higher precision than that available from DFT was required, the CCSD(T) method (41) employing the correlation-consistent triple-ζ plus polarization basis set (42, 43) was applied to the structures fully optimized at the Møller–Plesset second-order perturbation theory (44, 45) level with the same basis set.

Moreover, to explore the relationship between their catalytic effect and electronic structure, we investigated the MO levels of the MCl2 compounds for the geometries that were fully optimized by CCSD(T). Here, the Stuttgart–Bonn set (SRSC) effective core potential (46) was used for the transition-metal elements. In fact, we also attempted the calculations with the Los Alamos National Laboratory for double zeta quality (LANL2DZ) core potential (47) for transition-metal elements, which has been widely used in the field of organometallic catalysts; however, we found that this approach did not provide a correct energy change for FeCl2 as a function of a multiplicity. The details are explained in .

In the case of CBN adducts, a solvation effect was taken into account using the polarized continuum model scheme (48, 49) with a dielectric constant of 2.38 corresponding to the toluene solvent used in a previous experiment (22). In this work, all atomic charges were calculated by the NBO 5.0 scheme (50, 51). All ab initio calculations were performed using the Q-Chem package (52).