Mechanistic Insight into the Ring-Opening Polymerization of ε-Caprolactone and L-Lactide Using Ketiminate-Ligated Aluminum Catalysts

The reactivity and the reaction conditions of the ring-opening polymerization of ε-caprolactone (ε-CL) and L-lactide (LA) initiated by aluminum ketiminate complexes have been shown differently. Herein, we account for the observation by studying the mechanisms on the basis of density functional theory (DFT) calculations. The calculations show that the ring-opening polymerization of ε-CL and LA are rate-determined by the benzoxide insertion and the C–O bond cleavage step, respectively. Theoretical computations suggest that the reaction temperature of L–LA polymerization should be higher than that of ε-CL one, in agreement with the experimental data. To provide a reasonable interpretation of the experimental results and to give an insight into the catalyst design, the influence of the electronic, steric, and thermal effects on the polymerization behaviors will be also discussed in this study.


Introduction
Utilization of the polyester such as poly(ε-caprolactone) (PCL) and poly(L-lactide) (PLA) as eco-friendly materials has been paid more and more attention [1][2][3][4][5][6]. Therefore, development of efficient synthetic strategies becomes of great interest to chemists. Among these strategies, ring-opening polymerization (ROP) catalyzed by the less toxic and inexpensive aluminum complexes is one of the useful and well-controlled methods . The ring-opening polymerization catalyzed by an organometallic complex usually proceeds through a coordination-insertion mechanism [31,32]. Recent works have revealed that coordination number [33][34][35], the size effect of the chelate rings [36], steric and electronic effects of the auxiliary ligands surrounding the aluminum complexes [22,25,34,[37][38][39][40][41][42][43][44] as well as the co-initiators would alter the reactivity and the properties of the products. Besides the reaction conditions would be very dependent on the ring size and the bulkiness of the monomers [45,46]. For example, L-lactide has been found to be bulkier than caprolactone since the Al-O bond enthalpy of the former are weaker than the later [47]. Several kinds of Al complexes as the initiators for the ring opening polymerization have been designed, and different factors that influence the reactivity have been discussed separately. Nevertheless, systematically studying the structures-activity relationship of the catalysts remains ongoing.
In 2015, Chen et al. reported a series of aluminum complexes bearing ketiminate ligands to process the ring-opening polymerization of both ε-caprolactone (ε-CL) and L-lactide (L-LA) ( Figure 1) [48,49]. By screening the pendant groups on the ketiminate skeleton of the aluminum complexes, the steric, electronic and chelating effects of the ring-opening polymerization of ε-caprolactone and L-lactide could be delineated by comparison of the experimental results. Previous experimental data has revealed that in the ketiminate-ligated Al systems, ring-opening polymerization of ε-caprolactone can proceed at room temperature. Nevertheless, if using L-lactide as a monomer, similar reactions cannot occur until raising up temperature to 333 K. Besides, the k obs values of the ring-opening polymerization are influenced by the steric and electronic effects on the pendant N donors of kind of Al complex. It is worth noting that the orders of k obs values in the polymerization of ε-caprolactone and L-LA have different tendencies. For examples, catalysts with the bulky isopropyl substituents facilitates the polymerization of ε-caprolactone, but formation of polylactide is not promoted by this kind Al complex ( Figure 2). To get more insight into the mechanism, we decided to employ density functional theory (DFT) calculation to compare the two ring-opening polymerization systems catalyzed by the Al complex bearing a ketiminate ligand. Although only few cases of tetradentate Al catalysts have been evaluated by DFT calculations [50][51][52], it has been proven to be a useful tool to model the reaction intermediates and comprehend the reaction pathways of the tris-and penta-coordinate Al complexes [29,42,44,[52][53][54][55][56][57][58][59][60][61][62]. Our aims in this theoretical study are: (a) to realize why polymerization of ε-caprolactone and L-lactide require different reaction temperatures; (b) to understand how the steric and electronic effects on the pendant group of the Al complexes would alter the catalytic abilities. To answer the above questions, we tried to compute the entire reaction pathways and estimate the activation barriers of the rate-determining steps. For the thermodynamic consideration, we also evaluated the bond enthalpy for the monomer coordination to the Al center. In addition, we chose L, L Cl , and L iPr (Figure 1) as pendant groups of the catalysts to evaluate the electronic and steric effects in both poly(ε-caprolactone) and poly(L-lactide) synthesis.

Materials and Methods
All the energies and the geometries for the intermediates and transition states were evaluated using the Gaussian 16 package (Gaussian Inc., Pittsburgh, PA, USA) [63]. The Becke 3-parameter Lee-Yang Parr (B3LYP) hybrid functional combined with the D3 empirical dispersion correction developed by Grimme and the 6-31g(d) basis set was chosen as the calculation level [64][65][66][67][68][69][70][71]. The polarizable continuum model (PCM) was employed to describe the solvent effect of toluene [72]. To identify all the stationary-point structures, all reaction species were subjected to perform the frequency computation. All the resting states show real values in all types of motional frequencies, while the species with only one imaginary frequency were confirmed as transition states. Finally, the intrinsic reaction coordinate (IRC) or Quasi-IRC calculations were employed to make sure the connection between intermediates and the transition state [73][74][75][76]. The optimized Cartesian coordinates for all structures are provided in the Supplementary Materials.

Results and Discussion
Ring-opening polymerization of ε-caprolactone and L-lactide catalyzed by Al complexes would follow the coordination-insertion mechanism. The process starts from monomer coordination to the Al center, followed by benzoxide insertion to the C=O group of the monomer. Then, the resultant intermediate adjusts to the proper conformation to facilitate the breakage of C-O bond, completing the first ring-opening process. In the ketiminate-ligated Al systems, the monomer is approaching the Al center to give a penta-coordinate species. Considering 13 possible isomers can be generated, all of them were submitted as initial geometries and optimized by DFT calculation. (Figure 3). Calculation results show that only isomer I, II, III, and IV can be obtained as the optimal geometries, so mechanisms starting from isomer V-XIII are not considered. In order to obtain a more comprehensive view of the ROP mechanism, we discussed the four possible pathways from four triangular bipyramidal (TBP) isomer I-IV as the starting species (Scheme 1). By scanning the reaction coordinates, we confirmed that for the species N donor resides in the axial site (isomer III and IV), the monomer tends to be driven away from the aluminum center when benzoxide is approaching. As a result, Route 3 and 4 in Scheme 1 can be ruled out. The mechanistic calculations were concentrated on the pathways starting from I and II, which are initiated from the coordination isomers with the axial O donor (Route 1 and 2: the starting species is the isomer that monomer respectively coordinates on the equatorial and axial site).

Ring-Opening Polymerization of ε-Caprolactone
Computing results showed that ε-caprolactone can occupy either equatorial or axial sites to give coordinated species in all three systems with different pendant groups. The ε-caprolactone equatorially and axially coordinated isomers are respectively denoted as C eq 1a-c and C ax 1a-c (a: L, b: L Cl , c: L ipr ; Figure 4). The Al-O2 (Al-O CL ) bond lengths in C eq 1a, C eq 1b, and C eq 1c are respectively 2.013, 2.012, and 2.024 Å, shorter than those in C ax 1a-c (C ax 1a and C ax 1b: 2.069; C ax 1c: 2.085 Å). However, the bond enthalpies of Al-caprolactone in C ax 1a-c (C ax 1a: −12.7; C ax 1b: −13.4; C ax 1c: −9.1 kcal/mol) are slightly larger than those in C eq 1a-c (C eq 1a: −9.4; C eq 1b: −10.4; C eq 1c: −7.5 kcal/mol), showing coordination of ε-caprolactone at the axial site is energetically favorable. Both the equatorial and axial caprolactone coordination orientation are affected by the steric effect. Bulky substituents of isopropyl groups would hinder the monomer coordination. As a result, the bond enthalpies in the L iPr systems are the smallest among the three systems. On the other hand, the fact the Al-O2 bond enthalpies in the L Cl systems are largest can be attributed to the chloro substituent, whose electron withdrawing property would make the Al center more positive and favorably binding to the ε-caprolactone. The Gibbs free energy profiles are shown in Figures 5-7. The calculations reveal that the benzoxide insertion is the rate determining step of the ε-caprolactone ring-opening polymerization. The relative Gibbs energies in Route 1 are computed as 14.4 kcal/mol for C eq TS1a, 14.9 for C eq TS1b, and 13.2 for C eq TS1c (The differences in barrier height between C eq 1a-c and C eq TS1a-c are 10.3, 10.9, and 9.1 kcal/mol). On the other hand, the activation Gibbs energy barriers of C ax TS1a, C ax TS1b and C ax TS1c are 8.2, 8.3, and 7.6 kcal/mol, respectively. The distances of C1-O4 in C eq TS1a, C eq TS1b and C eq TS1c are 1.923, 1.921, and 1.886 Å, while they are 1.768, 1.769, and 1.715 Å in C ax TS1a, C ax TS1b and C ax TS1c. The shorter distances of C1-O4 in C eq TS1c and C ax TS1c are probably due to the steric congestion imposed by the isopropyl groups. The benzoxide insertion steps which give the products of C eq 2a-c (Route 1) and C ax 2a-c (Route 2) were evaluated as endergonic reactions. Noteworthily, the L iPr systems of both routes require least input of energies to proceed the insertion step, probably because of the release of crowded space with the benzoxide group approaching to the ε-caprolactone [77]. The distances between Al and O4 of the inserted benzoxide in C eq 2a-c are similar (2.141, 2.148, and 2.147 Å for C eq 2a, C eq 2b and C eq 2c), while the aforementioned distance in C ax 2c (2.047 Å) is longer than those in C ax 2a (2.021 Å) and C ax 2b (2.015 Å). After the benzoxide insertion, the Al-O2 bond in C eq 2a-c/C ax 2a-c rotates to yield C eq 3a-c/C ax 3a-c, which hold the structures in proper orientations to facilitate the C1-O3 bond breaking. Although C eq 2a-c/C eq 3a-c and C ax 2a-c/C ax 3a-c can be viewed as two pairs of linkage isomers, the computing result showed that the free energy differences between C eq 3a-c/C ax 3a-c and C eq 2a-c/C ax 2a-c are very dependent on their configurations and substituents. For the less bulky L and L Cl ligand systems, the free energies of C eq 3a-b and C ax 3a-b are more favored than those of C eq 2a-b and C ax 2a-b with around 3 and 6.5 kcal/mol, respectively. For the L iPr ligand systems, 3.1 kcal/mol is required to drive C eq 2c to C eq 3c, while the pathway from C ax 2c to C ax 3c is exergonic by 3.7 kcal/mol. Al and O3 are separated with the distance of 2.087, 2.082, and 2.092 Å in C eq 3a, C eq 3b, and C eq 3c and 2.113, 2.123, and 2.123 Å in C ax 3a, C ax 3b, and C ax 3c. The distances reveal the Al . . . O3 interaction in C eq 3a-c/C ax 3a-c, driving the C1-O3 bond cleavage to complete the ring opening. C eq TS2a-c and C ax TS2a-c represent the possible transition-state structures for the ring-opening steps. The C1-O3 are elongated to 1.924, 1.932, and 1.898 Å in C eq TS2a, C eq TS2b, and C eq TS2c. The corresponding distances in C ax TS2a, C ax TS2b, and C ax TS2c are 1.932, 1.922, and 1.781 Å. The reactions then generate the final products C eq 4a-c and C ax 4a-c by conquering small energy barriers of~2 (Route 1) and~0.5-5 (Route 2) kcal/mol, respectively. In C eq 4a-c and C ax 4a-c, the C1-O3 of ε-caprolactone has been broken and reformed to a nine-membered ring with Al. Consequently, the coordination-insertion ring-opening reactions are slightly endergonic by less than 2 kcal/mol in both routes of L and L Cl system. For the cases of L iPr systems, Route 1 is predicted to be 3.6 kcal/mol endergonic, whereas Route 2 is 1.8 kcal/mol exergonic.   Comparing the reaction barriers, Route 2 requires much lower free energy than Route 1 to make the reaction take place, indicating that coordination of ε-caprolactone in the axial position is the better direction to undergo nucleophilic addition. The corresponding structures of the transition states in Route 2 display a C1 . . . O4 distance about 1.75 Å, shorter than those in Route 1 (about 1.92 Å). The shorter distance of C1 . . . O4 in Route 2 also supports the fact that the benzoxide fragment can attack the carbonyl group of ε-caprolactone more effectively. As a result, Route 2 is a preferential pathway for the ring-opening polymerization of ε-caprolactone. Moreover, the activation barriers in the L ipr system is about 1 kcal/mol smaller than those in the L and L Cl systems, in consistency with the experimental results. It can be attributed to the sterically encumbered groups obliging the benzoxide fragment to get closer to the ε-caprolactone thus speeding up the occurrence of the insertion step, which determines the rate of the ring-opening polymerization of ε-caprolactone.

Ring-Opening Polymerization of L-Lactide
The ring-opening polymerization mechanism of L-lactide shows different characteristics compared to ε-caprolactone polymerization. Although L-lactide can approach the Al center from the equatorial site, the results of the scan calculation show that benzoxide cannot attack the carbonyl group from this orientation. Hence, Route 2 is the only possible pathway for the ring-opening polymerization of L-lactide. The Al-O2 (Al-O LA ) bond lengths of L ax 1a, L ax 1b, and L ax 1c are 2.091, 2.072, and 2.099 Å, respectively ( Figure 8). The values are slightly longer than those in C ax 1a-c, probably because of the bulkier monomer of L-lactide. The bond enthalpies of Al-O2 in L ax 1a-c are calculated to be larger than those in C ax 1a−c (L ax 1a: −12.6 kcal/mol; L ax 1b: −13.9 kcal/mol; L ax 1c: -12.3 kcal/mol). Among these species, the electron withdrawing inductive effect in L ax 1b would lead to the more stable Al-O2 bond. Nevertheless, the calculation result in L ax 1c implies that the steric hindrance does not exhibit obvious influence on the Al-O2 bond enthalpy. DFT calculated Gibbs energy profiles of the proposed mechanism for the L ax 1a-c catalyzed ring-opening polymerization of L-lactide are shown in Figures 9-11. Different from the case of ε-caprolactone, the cleavage of the C-O bond is the rate-determining step in the ring-opening polymerization of L-lactide. The activation barriers generated from the nucleophilic addition of benzoxide to the carbonyl group on the L-lactide are about 7 kcal/mol, similar to the case of the ring-opening polymerization of ε-caprolactone. The corresponding transition states are determined to be a distorted square pyramidal structure, and the C1 . . . O4 distances are separated by 1.934 Å in L ax TS1a, 1.932 Å in L ax TS1b, and 1.869 Å in L ax TS1c. The bulky pendant groups in L ax TS1c would lead to confined space and therefore result in the shorter C1 . . . O4 distance. The intermediates produced from the insertion step are the species L ax 2a-c with Al . . . O4 distances of 2.149, 2.135, and 2.195 Å in L ax 2a, L ax 2b, and L ax 2c, respectively. L ax 2a-c then undergo a bond rotation and yield L ax 3a-c, the conformations more suitable for C-O cleavage. Compared to L ax 2a-c, the Al . . . O3 distances in L ax 3a-c are shortened while C1-O4 bonds are elongated. Noteworthily, the Al . . . O3 distance in L ax 3c (2.342 Å) is much longer than in L ax 3a (2.215 Å) and L ax 3b (2.174 Å). The breakage of the C1-O3 bond involves the transition states L ax TS2a-c which lie 12.1, 11.6, and 15.1 kcal/mol above L ax 3a, L ax 3b, and L ax 3c, respectively. The highest potential Gibbs energies of L ax TS2a-c reveal the C1-O3 breaking step determines the rate of the ring opening polymerization of L-lactide. The calculation results show that the C1-O3 distances are 2.002, 2.001, and 2.069 Å in L ax TS2a, L ax TS2b, and L ax TS2c, respectively. Following the intrinsic reaction coordinate, L ax TS2a-c give rise to the unstable intermediates L ax 4a-c which are better described as a distorted tetrahedral species. Finally, L ax 4a-c rearrange to the conformations of L ax 5a-c, which adopt a TBP structure with the bidentate coordination of the ring-open L-lactide moiety.   The theoretical prediction shows that the L iPr system has the largest activation barrier among the three catalysts. Therefore, it would initiate the ring-opening polymerization of L-lactide more slowly, in agreement with the experimental results. The reaction is rate-determined by the ring opening step, the produced fragment of which would occupy more space. As a consequence, the reaction will be retarded by the bulky substituents. On the other hand, the reaction barriers of the ring-opening polymerization of L-lactide are estimated to be about twice larger than those of ε-caprolactone. The computational results have elucidated the fact that the ring-opening polymerization of L-lactide need to proceed at higher temperature in comparison to the case of ε-caprolactone.

Conclusions
In this study, we have shown that the ring-opening polymerization of ε-caprolactone is rate-determined by the step of benzoxide insertion, while the bond cleavage step is the slowest step in the case of ring-opening polymerization of L-lactide. Besides, theoretical analysis revealed that the reaction barriers of the ring-opening polymerization of L-lactide are higher than those of ε-caprolactone, corresponding to the experimental results. Finally, the pendant groups on the ketiminate ligand of the aluminum complexes play an essential role in catalyzing the ring-opening polymerization. In the ε-caprolactone ring-opening polymerization system, the crucial transition state is the stage benzoxide attacking the caprolactone. When the ketimate ligand bears more bulky groups, the congestion can be more largely released. Consequently, the reaction barrier is lowered to facilitate the occurrence of insertion when the pendant groups are more bulky. On the contrary, when L-lactides are loaded as monomers, the aluminum complexes supported by bulky groups are not conducive to the ring-opening polymerization. Since the rate-determining step of bond cleavage would lead to the occupation of space, the bulky substituents would go against the reaction to prevent from losing space. The calculation results are in agreement with the experimental data.

Conflicts of Interest:
The authors declare no conflict of interest.