Palladium-catalyzed regioselective and stereo-invertive ring-opening borylation of 2-arylaziridines with bis(pinacolato)diboron: experimental and computational studies

A palladium catalyzed regioselective borylative ring-opening reaction of 2-arylaziridines to give β-amino-β-arylethylborates was developed.


Introduction
Aziridines, three-membered strained azaheterocycles, serve as versatile building blocks in modern organic synthesis. 1 Regioselective ring opening of aziridines with nucleophiles, which is driven by release of its ring strain, is one of the most useful transformations of aziridines into ubiquitous b-amino-functionalized motifs. 2 Aziridines undergo oxidative addition to lowvalent late transition metal complexes at the terminal C-N bond in an S N 2 fashion, generating the corresponding oxidative adducts, azametallacyclobutanes. [3][4][5] In light of this reactivity, diverse transition metal-catalyzed regio-, stereo-, and/or chemoselective transformations of aziridines are feasible, [6][7][8][9] where the oxidative addition is smoothly coupled with subsequent elementary processes, such as transmetalation, migratory insertion, and reductive elimination; regioselective Rh-and Cocatalyzed carbonylative ring expansion of aziridines, pioneered by Alper and co-workers, represent the embodiment of this idea. 6 Over the past few years, numerous advances have been made in transition metal-catalyzed ring-opening cross coupling of aziridines, where aziridines can be utilized as a non-classical alkyl electrophilic partner [eqn (1)]. 10,11 In 2012, Doyle reported the rst Ni-catalyzed regioselective Negishi alkylation of 2-aryl-N-tosylaziridines that leads to branch-type products via the regioselective cleavage of a benzylic C-N bond. 10a Since this report, several groups including us 11b have developed Ni-10 and Pd-catalyzed 11 regioselective ring-opening cross coupling reactions of aziridines to form C(sp 3 )-C(sp 2 )/C(sp 3 ) bonds, with the regioselectivity seemingly governed by the aziridine substrate rather than the catalyst [eqn (1)]. Despite the above-mentioned successes in C-C cross couplings, other catalytic ring-opening C-E (E s C) bond forming reactions of aziridines based on a cross-coupling mechanism are also unexploited. 12 Our latest ndings 11b led us to seek new catalytic ringopening C-E coupling systems of aziridines. Moreover, we became intrigued with the C-B coupling of aziridines to give bamino-functionalized alkylboronates; alkylboronic acid derivatives serve not only as useful building blocks in organic synthesis, but are also biologically interesting motifs in medicinal chemistry. 13 Only a few precedents in transitionmetal-catalyzed borylative substitution of vinylaziridines via C(sp 3 )-N bond cleavage have been reported. Szabó and Pineschi disclosed a Pd(II) pincer complex 14 and a Ni(0)/BINAP 12 system as catalysts for the borylative ring opening of vinylaziridines with diboron reagents via the cleavage of an allylic C-N bond to give g-amino alkylboronic acid derivatives (formal S N 2 0 -type reaction), 15 respectively [eqn (2)]. Nevertheless, to the best of our knowledge, the catalytic direct displacement of the C(sp 3 )-N bond of aziridines with a C(sp 3 )-B bond on the same carbon (formal S N 2-type reaction) has never been described to date.
We herein report a regioselective ring-opening C-B crosscoupling reaction of 2-arylaziridines that is realized using a Pd/P(t-Bu) 2 Me/bpy catalytic system to give b-amino-alkylboronates [eqn (3)], which can serve as versatile building blocks to synthesize amino-functionalized compounds and biologically relevant b-amino acid surrogates. 16,17 Notably, our reaction features the opposite regioselectivity in C-N bond cleavage (at the 3-position) to those previously reported for C-C cross couplings using the same type of aziridine substrate (at the 2-position). 10a,c,11b Also, our C-B coupling system represents the rst example of a formal S N 2-type borylative C(sp 3 )-N bond cleavage of neutral compounds. 18 It should be noted that the present C-B coupling proceeds smoothly under neutral conditions as in the cases of borylative substitution of allyl carboxylates 19 and carbonates. 19a,20 In these cases, the leaving groupderived oxy anions (RCO 2 À and RO À ) serve as the internal bases, while the metal catalyzed borylative substitution of alkyl halides usually requires the addition of stoichiometric external base or activator to promote transmetalation. 21 In addition to experimental elaboration, we have performed theoretical calculations, applying density functional theory (DFT) and the multicomponent articial force-induced reaction (MC-AFIR) method to determine the mechanism of the reaction and explain the origin of the selectivity, and to clarify why external base is not required in this reaction system.

Results and discussion
Experimental part Development of Pd-catalyzed borylative ring-opening reaction of 2-arylaziridines. To identify the reaction conditions for the borylative ring-opening reaction, optimization studies applying racemic 2-phenyl-N-tosyl-aziridines (1a) as a model substrate with bis(pinacolato)diboron B 2 (pin) 2 (2) were performed (Tables S1-S13 in the ESI †). 22 The executive summary of the optimization study is shown in Table 1.
The highest yield (81%, isolated yield: 71%) of borylated product 3a was obtained when 1a was treated with 1.2 equiv. of B 2 (pin) 2 at 60 C in the presence of catalytic amounts of Cp(allyl) Pd (2 mol%), P(t-Bu) 2 Me 23 (1 mol%), and 2,2 0 -bipyridine (bpy, 20 mol%) in a mixed solvent of methyl-tert-butylether (MTBE) and H 2 O (entry 1, "standard conditions"). Most importantly, the regioselectivity of this ring opening and C-B coupling was opposite to those observed with the C-C coupling of the same substrate, 10a,c,11b implying that the oxidative addition occurred at the C-N bond on the terminal carbon (the 3-position of the aziridine). The regioisomer of 3a was not detected in the crude 1 H NMR spectra, indicating that the regioselectivity in the ringopening of 1a should be almost perfect. In fact, the regioselective oxidative addition dictated by the interactions between the substrate and the Pd(0) catalyst was supported by theoretical calculations (vide infra). The L : Pd ratio [L ¼ P(t-Bu) 2 Me] was found to have a signicant impact on the product distribution (entries 2 and 3). As the L : Pd ratio was increased from 0.5 (entry 1) to 2 to 3, the yields of 3a decreased to 74% and 45%, respectively (entries 2 and 3). This irregular ratio of L : Pd implies that there might be a complex equilibrium of PdL n species generated in situ, some of which are specically active in the catalytic cycle. In fact, this speculation was partly supported with 31 P NMR experiments and theoretical calculations (vide infra). Addition of electron decient unsaturated compounds, which can coordinate to metal complexes, is an effective strategy to suppress undesired b-hydride elimination or to accelerate reductive elimination, thereby oen leading to better results in alkyl cross coupling reactions. 24 Indeed, the effect of adding bpy was signicantly effective for suppressing the formation of byproduct 6 as well as 5, which was presumably derived from the hydrolysis of 6 with H 2 O (entries 4-7). The effect of the ligands was also signicant (entries 8-10). Nheterocyclic carbene (NHC) ligands, SIPr for instance, 11b promoted the consumption of 1a and produced undesired products 4 and 5 (entry 8). The results obtained from sterically demanding trialkylphosphine/Pd catalysts (entries 9 and 10) suggest a sluggish oxidative addition step, which is in good agreement with Wolfe's report. 9 Furthermore, as mentioned in the introduction section, the ring opening borylation disclosed herein does not require the addition of any external base, which is usually used to activate the transmetalation step for the borylative substitution of alkyl halides. In connection with this, addition of water is crucial. Moreover, in the absence of H 2 O, no conversion of 1a was observed (entry 11), while at least 1 equiv. of H 2 O allowed the borylation (Table S11 in the ESI †).
Theoretical calculations implied that H 2 O serves as a proton source (H + ) as well as a source of internal base (Pd-OH) to promote the four-membered transmetalation between B 2 (pin) 2 and Pd (vide infra). To investigate the fate of the other B(pin) moiety of B 2 (pin) 2 in the reaction, the 11 B NMR spectrum of the crude product was acquired in benzene-d 6 . The only peak other than the remaining B 2 (pin) 2 was detected at d ¼ 22.6 ppm as a slightly broad singlet. The comparison with the reported 11 B resonances of (pin)B-OH (d ¼ 22.5 ppm in benzene-d 6 ) 25a and (pin)B-O-B(pin) (d ¼ 21.6 ppm in benzene-d 6 ) 25b would suggest that one of the B(pin) moieties of the starting diboron reagent was converted to (pin)B-OH during the reaction.
Substrate scope and synthetic utility. Aer establishing the optimized conditions, we explored the substrate scope of the borylative substitution reaction ( Table 2). In terms of 2-arylaziridines, p-, m-, and o-tolyl-substituted aziridines 2b, 2c, and 2d smoothly underwent borylative ring opening in a regioselective manner to give the corresponding b-amide-b-tolylethylboronates 3b, 3c, and 3d in good yields. Aziridine bearing a p-uorophenyl substituent at the 2-position was efficiently borylated, giving the corresponding alkylboronate 3e in a high yield. It is important to note that the p-Cl functionality on the aryl ring of aziridine 2f survived the reaction conditions to provide the borylative product 3f in a high yield. Nevertheless, the reaction with p-bromophenyl-substituted aziridine did not proceed at all. An invaluable functional group in medicinal chemistry, a CF 3 -installed aziridine 2g gave the corresponding CF 3 -bearing alkylboronate 3g. Notably, this ring opening borylation was applicable to the arylaziridines bearing a strong electron withdrawing group (NO 2 , OAc, and CO 2 Me), and kept these functionalities intact. This highlights an advantage of using neutral conditions. Furthermore, the borylation was scalable, and 3a was prepared on a gram scale (2.89 g, 72% yield) from 10 mmol of 1a (pS17 of the ESI †). With respect to the limitation of substrate, 2pyridyl-(1k) and 2-alkylaziridines (1l, 1m, and 1n) were not consumed under the reaction conditions. Furthermore, this reaction is very sensitive to the steric hindrance around the 3positioned carbon, where oxidative addition to Pd(0) has occurred. For example, 2-methyl-2-phenylaziridine (1o) and trans-2-phenyl-3-methylaziridine (1p) were not consumed at all under the standard conditions. As for the scope of the N-functional groups of aziridines, p-MeOC 6 H 4 SO 2 and t-BuSO 2substituted 2-phenylaziridines smoothly underwent borylation in a regioselective manner (at the 3-position) to produce the corresponding b-amino-b-phenylethylboronates 3r (65%) and 3s (52%), respectively. On the other hand, N-p-nosyl-2-phenylaziridine did not give borylated product (Table S12 in the ESI †). The use of an N-acylated aziridine, N-Boc-2-phenylaziridine for instance, did not give borylated products at a noticeable level, highlighting the efficacy of the N-tosyl group for the borylation. Regarding the diboron reagents, the use of bis(neopentyl glycolato)diboron gave the corresponding aminoboronate product (3a 0 ) in 39% yield, while the borylation using bis(hexylene Table 1 Evaluation of reaction parameters glycolato)diboron produced the corresponding borylated products in a low yield as an inseparable diastereomeric mixture, presumably due to the existence of two chiral centers on the benzylic and glycolate carbons (Table S13 in the ESI †). Borylated product 3a was successfully transformed into bamino-b-aryl-substituted alkyltriuoroborate 7 in excellent yield, which can serve as a versatile building block to synthesize b-amino-b-arylethanes through Pd-catalyzed cross coupling reactions [eqn (4)]. 26 In order to understand the stereochemical outcome of the secondary stereogenic center of the borylated product and to demonstrate the utility of the alkylboronate, an enantiopure aziridine (R)-1a (99% ee from chiral HPLC analysis) was subjected to the standard reaction conditions (Scheme 1). HPLC analysis of the borylated product 3a revealed that the stereochemical information was completely retained (99% ee, S) through the reaction. This result excludes the possibility of the reaction pathway of oxidative addition/b-hydride elimination/ re-insertion of [Pd-H] species leading to 3a, although it cannot perfectly exclude such a pathway. 27 To demonstrate the synthetic utility of the enantiopure b-amino-alkylboronate (S)-3a, the asymmetric synthesis of 3-phenyl-1,2,3,4-tetrahydroisoquinoline derivative was conducted (Scheme 1). The tetrahydroisoquinoline skeleton is a ubiquitous motif in alkaloid natural products and constitutes a biologically important azaheterocycle. 28 Enantiopure alkylboronate (S)-3a was subjected to the Pd-catalyzed Suzuki-Miyaura cross coupling 29 with chlorobenzene to give the optically active 1,2-diphenylamine derivative (S)-8 in 69% yield, while keeping the enantiopurity intact (99% ee, chiral HPLC). The following Pictet-Spengler reaction 30 successfully gave the enantiopure tetrahydroisoquinoline product (S)-9 in excellent yield.
Mechanistic aspect. To obtain the stereochemical course in the oxidative addition step, a similar synthetic sequence using deuterated aziridine cis-1a-d 1 as a substrate was conducted (Scheme 2). The stereochemistry on the terminal carbon (the 3position) was completely inverted (for details, see the ESI †). This result agrees with the S N 2 nature of the oxidative addition process of aziridine toward Pd(0) complexes reported by our group 11b and others. 5,11a Moreover, this stereo-invertive process is supported with computational studies (vide infra).
To identify the Pd species generated by mixing P(t-Bu) 2 Me and Cp(allyl)Pd, we monitored the 31 P{ 1 H} NMR spectra of these mixtures in THF-d 8 at 60 C (Fig. 1). Upon addition of 2 equiv. of Cp(allyl)Pd to L (i.e., L : Pd ¼ 0.5), the sharp resonance corresponding to free L at d 15.3 ppm (Fig. 1a, lit. d 12.5 ppm at room temperature in toluene-d 8 Fig. 1c). These peaks were not changed when bpy (10 equiv.) was added (Fig. S1b in the ESI †). This result would indicate that bpy does not coordinate to these Pd species, which was also supported by our computational results (vide infra). Moreover, upon addition of another equivalent of L (i.e., L : Pd ¼ 1), the strongest sharp peak at d 37.0 ppm (dinuclear Pd complex) disappeared, and a broad peak around d 42.6 ppm appeared (Fig. 1d). Furthermore, addition of one more equivalent of L (L : Pd ¼ 2) gave a signicantly broadened peak (d $ 30 ppm, Fig. S1a in the ESI †), indicating signicant exchange among several PdL n species. Our observations could indicate that PdL 2 undergoes rapid exchange with PdL n in THFd 8 when the L : Pd ratio is greater than 0.5. Therefore, the same would be the case with the real reaction system.
Based on the NMR experiments, [Pd 2 L 2 (m-Cp)(m-allyl)] or PdL 2 would be the catalytically active Pd species (i.e., the starting point of the catalytic cycle). To determine which is the active species in the catalytic reaction, both complexes were generated separately as the sole components according to Baird's conditions (i.e., 1 : 1 and 2 : 1 mixtures of L and Cp(allyl)Pd were mixed in toluene-d 8 and stirred at 77 C for 1 h), 32 and the resulting Pd complexes were used as a catalyst in the borylation reaction in toluene (Schemes 3). As a result, while the dinuclear Pd(I) complex was found to be totally inert for the borylation reaction, the bisphosphine complex Pd(0)L 2 gave almost the same result as the real reaction system. Therefore, we conclude that Pd(0)L 2 would be the starting active species in the catalytic cycle, and this conclusion was strongly supported by theoretical calculations (vide infra).

Computational part
Computational methods. Geometry optimizations were performed with the B3LYP-D3 33-36 functional as implemented in the Gaussian09 program. 37 The polarizable continuum model (PCM) was used as the implicit solvent model with a dielectric constant (3) of 2.3714. 38 The SDD basis set and associated effective core potential were applied for Pd, 39,40 and 6-31G(d) basis sets were used for the other atoms (BS1). [41][42][43][44] All stationary points, local minima (LMs) or transition states (TSs), were optimized without any constraint. Vibrational frequency calculations were performed to conrm the nature of the stationary points. Pseudo-IRC calculations were performed to conrm the connectivity between the TSs and LMs. The nal potential energies of the optimized structures were calculated as singlepoint energies on the optimized structures, where a SDD basis set was used for Pd, and cc-pVTZ basis sets were used for the other atoms (BS2). [45][46][47] In the results section, we report both the relative Gibbs free energy (DG) at 298.15 K and 1 atm and the relative electronic energy with the zero point energy correction (DE).
The aziridine ring opening step of the mechanism takes place via many different TSs, and therefore a proper sampling is very important. An automatic exploration of all important reaction pathways was accomplished using the multi-component articial force induced reaction (MC-AFIR) method, as implemented in the global reaction route mapping (GRRM) strategy. [48][49][50] An articial force parameter (g) of 300 kJ mol À1 was applied, and this is suitable for nding TSs within 300 kJ mol À1 . The MC-AFIR calculations were terminated when no new AFIR LM was found for 10 consecutive AFIR paths (NFault ¼ 10). In AFIR calculations, the energy and derivatives were obtained using the ONIOM(B3LYP-D3:PM6-D3) method. [51][52][53][54][55][56][57] Partitioning of the molecular system is shown in Fig. 2a. A model catalyst was used for MC-AFIR calculations (Fig. 2). SDD basis sets and the associated effective core potentials were used for palladium, and 3-21G basis sets (BS3) were applied for the high-level region of ONIOM calculations. [58][59][60] All AFIR paths were inspected and approximate TSs were identied. Then, the real phosphine ligand was introduced to all approximate TSs, and the TS structures were fully optimized (without articial force) with B3LYP-D3/BS1 method. A Boltzmann distribution of the transition states was used to calculate the regioselectivity.  Energy decomposition analysis (EDA) 61,62 was performed for the key TSs leading to the desired and the undesired products. B3LYP-D3/BS2 level and PCM were used for the EDA. In this analysis, a TS structure was divided into A (catalyst) and B (substrate) (Fig. 3). INT AB is the interaction energy between A and B at their optimized TS structure. The deformation energy (DEF) is the energy of A and B at the optimized TS, relative to the optimized structures of isolated A and B (denoted as A 0 and B 0 ). The energy difference (DDE) between the optimized transition states, (AB) 1 and (AB) 2 , is the sum of the internal energy difference (DINT AB ) and the deformation energy difference (DDEF).
Thermodynamically stable complexes in solution. Our starting point was to explore thermodynamically stable complexes in solution. First, we checked the relative stability of PdL, PdL 2 , PdL 3 , and PdL 4 complexes [L ¼ P(t-Bu) 2 Me]. Among them, the PdL 2 complex (I) is the most stable complex, and this is in agreement with our NMR studies. The PdL 3 and PdL complexes are 3.4 and 26.0 kcal mol À1 higher in energy, respectively. Despite several attempts, no PdL 4 complex was found, as one of the four ligands (L) always dissociated due to steric repulsion. We used PdL 2 (I) as a candidate to generate other possible complexes in solution. Moreover, the two vacant sites of the PdL 2 complex, A and B, can be lled by the potential ligands in solution, specically H 2 O, solvent (denoted as Sol), and additive (denoted as bpy). Table 3 summarizes the possible complexes and energies relative to the PdL 2 complex (I).
Starting from the PdL 2 complex (I), coordination of one H 2 O, L, and bpy on I gives rise to PdL 2 (H 2 O) (Iw, À0.5 kcal mol À1 ), PdL 3 (3.4 kcal mol À1 ), and PdL 2 (bpy) (4.0 kcal mol À1 ) complexes, respectively. Among the three-coordinate complexes, Iw is the thermodynamically most stable complex, and is only 0.5 kcal mol À1 more stable than I. Therefore, both I and Iw can be formed in solution. In Iw, an H 2 O molecule coordinates to the metal with one of the two hydrogen atoms. It is important to note that the additive (bpy) can coordinate to the metal, and the resulting complex, PdL 2 (bpy), is, however, 4.5 kcal mol À1 higher than the most stable complex, Iw. All fourcoordinate complexes found in the calculations, PdL 2 (bpy)(H 2 -O) (2.8 kcal mol À1 ), PdL 2 (H 2 O) 2 (4.7 kcal mol À1 ), and PdL 3 (H 2 O) (7.9 kcal mol À1 ) are relatively higher in energy. Coordination of solvent (Sol) is not possible due to steric repulsion between the bulky groups of the solvent molecules (MTBE) and I. We concluded that the thermodynamically most stable complexes in solution are Iw and I. The general rules that lead to the formation of the thermodynamically stable complexes in solution are: (1) coordination of two ligands on PdL 2 , which gives rise to four-coordinate complexes, is not favorable; (2) in terms of making three-coordinate complexes, the binding preference of the third ligand follows the order H 2 O > L > bpy; (3) solvent MTBE would not coordinate to the PdL 2 complex due to steric repulsion; (4) coordination of the additive (bpy) is not favorable. According to Table 3, additive coordination on PdL 2 is not favorable as the subsequent complex PdL 2 (bpy) is 4.0 kcal mol À1 higher than PdL 2 , while the PdL 2 (bpy) 2 complex is not stable. At the same time, it is important to note that aziridine binding on PdL 2 is relatively easier (vide infra), and the corresponding adduct is only 2.6 kcal mol À1 higher than PdL 2 . Therefore, the additive would not coordinate to the PdL 2 complex before the aziridine substrate binding. In our present mechanistic study, we do not consider the role of the additive in Fig. 2 (a) ONIOM partitioning of the model complex into high-and low-levels, (b) the real complex, (c) the artificial force (g ¼ 300 kJ mol À1 ) was applied between fragm1 (catalyst) and fragm2 (substrate) without or with an explicit water molecule.  the mechanism because the reaction works even in the absence of additive (Table 1, entry 4). Aziridine ring-opening step. Our next task was the exploration of the aziridine ring-opening step. For this purpose, we used the most stable I and Iw complexes as the active intermediates for the reaction, and 2-phenyl-N-tosyl-aziridine (1) was used as the substrate. Therefore, two independent MC-AFIR calculations, with and without an explicit water molecule, were performed as shown in Fig. 2c. The fully optimized TS structures, 73 TSs from Iw and 33 TSs from I, were categorized into 15 groups: TSI-II A through TSI-II J from Iw, and TSI-II K through TSI-II O from I, based on structural similarities (Fig. 4). The lowest energy TS of each group, their relative energies, and their existence probability are depicted in Table 4 (see Table S14 in the ESI † for a full description of all the TSs).
Among the calculated TSs (Table 4), TS1-t-lw (12.6 kcal mol À1 ) which belongs to the TSI-II K group is the lowest energy TS of this step. In this TS, the aziridine ring opening takes place at the less hindered carbon in an S N 2 fashion, leading to the desired product forming. The same product can be obtained through TS2-t-lw (12.9 kcal mol À1 ), TS4-t-l (13.3 kcal mol À1 ), TS5-t-lw (13.4 kcal mol À1 ), TS6-t-l (13.5 kcal mol À1 ), TS7-t-l (13.6 kcal mol À1 ), TS8-t-l (13.7 kcal mol À1 ), and TS9-t-lw (13.9 kcal mol À1 ). On the other hand, at TS3-b-lw (13.2 kcal mol À1 ) of the group TSI-II L , aziridine ring opening occurs at the hindered carbon (the 2-position), and this is the lowest TS leading to the undesired product. Based on a Boltzmann distribution over the calculated TSs, the regioselectivity is calculated to be 89 : 11, which is qualitatively in agreement with the experimental results (99 : 1).
In the calculated TSs for TSI-II E (31.5 kcal mol À1 ), TSI-II F (30.9 kcal mol À1 ), and TSI-II M (24.7 kcal mol À1 ), aziridine ring opening occurs at the less hindered carbon (Fig. 4). However, these TSs are relatively higher in energy due to the fact that the N-p-toluenesulfonyl and 2-phenyl substituents of the substrate are much closer to the bulky groups of Iw or I. Similarly, TSs in the group of TSI-II G (31.3 kcal mol À1 ), TSI-II H (31.9 kcal mol À1 ), and TSI-II N (28.1 kcal mol À1 ) show high energy barriers, where aziridine ring opening occurs at the hindered carbon. TSs in the group of TSI-II I (31.8 kcal mol À1 ), TSI-II J (31.8 kcal mol À1 ), and TSI-II O (28.1 kcal mol À1 ) represent the aziridine ring opening through the carbon-carbon bond, which is, however, not possible due to very large reaction barriers. Furthermore, we checked the effect of H 2 O molecules on aziridine ring opening. Starting from the lowest energy TS for this step (TS1-t-lw, 12.6 kcal mol À1 ), up to four H 2 O molecules were introduced, and the corresponding TSs were calculated. The calculated DG of TS1-tlw with one H 2 O molecule (TS1-t-lw-w, 16.1 kcal mol À1 ), two H 2 O molecules (TS1-t-lw-w 2 , 14.3 kcal mol À1 ), three H 2 O molecules (TS1-t-lw-w 3 , 15.6 kcal mol À1 ), and four H 2 O molecules (TS1-t-lw-w 4 , 16.4 kcal mol À1 ) suggested that explicit H 2 O molecules in the system would not stabilize the TS1-t-lw.
TS1-t-lw is a major contributor to the aziridine ring opening at the less hindered carbon, while TS3-b-lw is the lowest TS for aziridine ring opening at the hindered carbon (Fig. 5). The calculated Gibbs free energy difference between TS1-t-lw and TS3-b-lw is 0.6 kcal mol À1 , while the potential energy difference is 3.2 kcal mol À1 . When we do not consider the zero point energy corrections, the potential energy difference is 3.5 kcal mol À1 , and we have used this value for the EDA. According to the EDA (Table 5), the origin of this difference comes from DINT AB (3.9 kcal mol À1 ), indicating better interactions at TS1-tlw. This is due to the fact that the aziridine substrate can approach closer to the metal in TS1-t-lw (Pd-C ¼ 2.47Å) (vs. Pd-C ¼ 2.61Å at TS3-b-lw) (Fig. 5) than in TS3-b-lw due to the lower steric repulsion.
The free energy prole for the early stages of the mechanism is shown in Fig. 6. The reaction starts from the thermodynamically most stable complex, Iw. Therefore, we report energies relative to Iw. Then, aziridine coordination on I leads to a complex II (2.6 kcal mol À1 ), and the subsequent aziridine ring opening occurs through TS1-t-lw (TSII-III, 12.6 kcal mol À1 ). Aziridine ring opening is also possible from Iw (not shown in Fig. 6) with an overall barrier of 13.3 kcal mol À1 (TS4-t-lw), which is, however, 0.7 kcal mol À1 higher than TS1-t-lw. Beyond TS1-t-lw, an intermediate, III (8.4 kcal mol À1 ) is formed.
Proton transfer to anionic amine. The next step of the mechanism would be a proton transfer from H 2 O molecules in solution to the anionic nitrogen atom in the water-coordinated intermediate III. This proton transfer process takes place smoothly from IIIw 3 (0.3 kcal mol À1 ), which is 8.1 kcal mol À1 more stable than III. Furthermore, three H 2 O molecules make a hydrogen bonding network starting from the anionic amine that is connected to the metal through a water oxygen coordination. The TS for this proton transfer process is 12.0 kcal mol À1 (TSIII-IVw 3 ). Proton transfer is also possible from the analogous intermediate with four H 2 O molecules, IIIw 4 (11.3 kcal mol À1 ), and the subsequent transition state is, however, 1.3 kcal mol À1 higher than TSIII-IVw 3 (not shown in Fig. 6). We were unable to locate TSs starting from the analogous intermediates with two H 2 O molecules (IIIw 2 , 1.1 kcal mol À1 ) and one H 2 O molecule (IIIw, 2.3 kcal mol À1 ), where the potential energy surface for the proton transfer processes was found to be repulsive ( Fig. S2 and S3 †). In IIIw 3 , three water molecules form a complete head-to-tail chain between the anionic amines and Pd centres. As a result, the synchronous three proton transfer via TSIII-IVw 3 takes place and gives a protonated N, two H 2 O molecules, and a hydroxyl group bound to the metal center. Based on our analysis, we conclude that three H 2 O molecules are required for the proton transfer process, leading to an intermediate IVw 3 (2.0 kcal mol À1 ). Once the proton is transferred, the chain of water molecules is no longer required, and the corresponding intermediate without H 2 O, IV, is 2.7 kcal mol À1 more stable than IVw 3 .
B-B bond coordination and cleavage. Now, the B(pin)-B(pin) species binds to IV and the subsequent intermediate V (0.5 kcal mol À1 ) is formed. Starting from V, inner-sphere B-B bond cleavage does not take place due to steric repulsion between the bulky alkyl groups in L and B(pin)-B(pin). One of the two phosphine ligands of the catalyst must dissociate, and the resulting intermediate, VI, is only 1.8 kcal mol À1 higher than V. Then, B-B bond cleavage takes place in an inner sphere fashion with a barrier of 10.8 kcal mol À1 (TSVI-VII), giving rise to VII (À2.1 kcal mol À1 ). We have checked the possibility of ligand exchange between the (pin)B-OH and a stronger phosphine ligand L on intermediate VII, and the resulting complex, VIII, is 24.5 kcal mol À1 more stable than VII.
cis-trans isomerization and reductive elimination. The energy proles for the nal stages of the reaction are shown in Fig. 7. In both the intermediates VII and VIII, B(pin) is trans to the alkyl group derived from the aziridine substrate, and therefore cis/trans isomerization must take place before the reductive elimination. The cis isomer VIII 0 is 2.5 kcal mol À1 more stable than the trans isomer VIII. This step may occur through ligand dissociation processes. Since we expect this isomerization to be a low energy process, we did not study this in depth. The subsequent reductive elimination from VIII 0 occurs through a barrier of 0.1 kcal mol À1 (TSVIII 0 -IX). In the resulting intermediate IX, the product (P) is still at the metal   coordination sphere, and P can be removed easily to recover the active form of the catalyst, PdL 2 (I), for the next catalytic cycle. Similarly, the formation of the cis isomer of VII, the less stable intermediate VII 0 (À19.4 kcal mol À1 ), and the subsequent reductive elimination can occur easily with a barrier of 5.7 kcal mol À1 (TSVII 0 -X). Then, a phosphine ligand (L) can coordinate to the resulting intermediate X, which ultimately yields the product, P.

Side product formation
Under the optimized reaction conditions, imine 6 is formed as a side product (Table 1). We have explored the mechanism for the side reaction as shown in Fig. S4 in the ESI. †

Catalytic cycle
Putting together the present results from the experimental and theoretical studies, the proposed catalytic cycle for the borylation is shown in Fig. 8 (black cycle). Pd(0)L 2 is generated through the sequential reduction of Cp(allyl)Pd. Then, Pd(0)L 2 attacks the less hindered carbon of the aziridine in an S N 2 fashion (a) to give a stereo-inverted oxidative adduct. A hydrogen bonded chain of H 2 O molecules plays two roles in the following steps: (i) as a proton source to quench TsN À (b) and (ii) as an internal base to form the [Pd(OH)] species. This intermediate activates the B-B bond of B 2 (pin) 2 (c). The dissociation of a phosphine ligand (d) facilitates the transmetalation (e). The free phosphine ligand participates in the catalytic cycle again to form the (alkyl)PdL 2 B(pin) complex (f). The trans-cis isomerization (g) followed by reductive elimination (h) leads to the C-B cross-coupled product and the completion of the catalytic cycle. A side reaction may occur aer the oxidation addition (brown cycle): the oxidative adduct of aziridine undergoes hydrogen transfer (i) followed by reductive elimination (j) to produce an imine byproduct. Our computational study suggests that the aziridine ring-opening step for substrate 1a has barriers of 12.6 kcal mol À1 (terminal position) and 13.2 kcal mol À1 (benzylic position) ( Table 4). For an alkyl aziridine, in particular 1l, the aziridine ring-opening step showed barriers of 15.4 kcal mol À1 (2-position) and 16.3 kcal mol À1 (3position). In the absence of the aryl group, the aziridine ring opening of 1l is 2.8 kcal mol À1 higher than 1, where the reaction is difficult with 1l. This is qualitatively in agreement with our experimental results, where the reaction does not proceed with 1l.

Conclusions
In conclusion, we have developed a Pd-catalyzed regioselective borylative ring opening reaction of 2-arylaziridines to give baminoalkylboronates that are otherwise difficult to synthesize using existing methodologies. Importantly, the regioselectivity of the ring opening is controlled by the interactions between the catalyst and the substrate. The S N 2 nature of the oxidative addition of aziridine was veried using deuterated aziridine and computational studies. Furthermore, the borylative reaction is applicable under neutral conditions that allow for high functional compatibility. The mechanism of the full catalytic cycle was proposed using DFT and MC-AFIR methods. The aziridine ring opening is initiated by the active species PdL 2 . TSs were systematically determined for the selectivity-determining aziridine ring opening step, and the calculations reasonably reproduced the experimental regioselectivity. The next step of the mechanism is a proton transfer that is facilitated by a H 2 O hydrogen bond chain. The resulting Pd-hydroxo species activates the transmetalation step, where inner sphere boronboron bond cleavage occurs, and leads to the nal reductive elimination step. These experimental and theoretical ndings open up an avenue to the further development of transition metal-catalyzed ring-opening C-E bond forming cross couplings of aziridines.