Unsaturated and Benzannulated N-Heterocyclic Carbene Complexes of Titanium and Hafnium: Impact on Catalysts Structure and Performance in Copolymerization of Cyclohexene Oxide with CO2

Tridentate, bis-phenolate N-heterocyclic carbenes (NHCs) are among the ligands giving the most selective and active group 4-based catalysts for the copolymerization of cyclohexene oxide (CHO) with CO2. In particular, ligands based on imidazolidin-2-ylidene (saturated NHC) moieties have given catalysts which exclusively form polycarbonate in moderate-to-high yields even under low CO2 pressure and at low copolymerization temperatures. Here, to evaluate the influence of the NHC moiety on the molecular structure of the catalyst and its performance in copolymerization, we extend this chemistry by synthesizing and characterizing titanium complexes bearing tridentate bis-phenolate imidazol-2-ylidene (unsaturated NHC) and benzimidazol-2-ylidene (benzannulated NHC) ligands. The electronic properties of the ligands and the nature of their bonds to titanium are studied using density functional theory (DFT) and natural bond orbital (NBO) analysis. The metal–NHC bond distances and bond strengths are governed by ligand-to-metal σ- and π-donation, whereas back-donation directly from the metal to the NHC ligand seems to be less important. The NHC π-acceptor orbitals are still involved in bonding, as they interact with THF and isopropoxide oxygen lone-pair donor orbitals. The new complexes are, when combined with [PPN]Cl co-catalyst, selective in polycarbonate formation. The highest activity, albeit lower than that of the previously reported Ti catalysts based on saturated NHC, was obtained with the benzannulated NHC-Ti catalyst. Attempts to synthesize unsaturated and benzannulated NHC analogues based on Hf invariably led, as in earlier work with Zr, to a mixture of products that include zwitterionic and homoleptic complexes. However, the benzannulated NHC-Hf complexes were obtained as the major products, allowing for isolation. Although these complexes selectively form polycarbonate, their catalytic performance is inferior to that of analogues based on saturated NHC.

We have continued our efforts to synthesize catalysts based on unsaturated NHC ligands, and present here, for the first time, bis-phenolate unsaturated-and benzannulated-NHC complexes of titanium and hafnium. The performance of these complexes in CO2/CHO copolymerization is compared to those of saturated NHC ligands. Finally, with the help of density functional theory (DFT) and natural bond orbital (NBO) analysis, the structural and electronic property differences of these complexes are presented, and the potential impact of these differences on CO2/CHO copolymerization is discussed. Scheme 1. Previously obtained complexes containing (i) saturated NHC with group 4, and (ii) unsaturated and benzannulated NHCs ligands (a and b, respectively) with zirconium.
We have continued our efforts to synthesize catalysts based on unsaturated NHC ligands, and present here, for the first time, bis-phenolate unsaturated-and benzannulated-NHC complexes of titanium and hafnium. The performance of these complexes in CO 2 /CHO copolymerization is compared to those of saturated NHC ligands. Finally, with the help of density functional theory (DFT) and natural bond orbital (NBO) analysis, the structural and electronic property differences of these complexes are presented, and the potential impact of these differences on CO 2 /CHO copolymerization is discussed.

Synthesis of Bis-phenolate NHC Complexes of Titanium and Hafnium
The N,N'-di (2-hydroxy-3,5-di-tert-butylphenyl) imidazolium chloride (a) and N,N'-di(2hydroxy-3,5-di-tert-butylphenyl) benzoimidazolium chloride (b) proligands were prepared according to previously reported procedures [26,67,68]. The alcohol elimination route involving the direct and slow addition of proligands a and b to a solution of Ti(OiPr)4 in THF at −30 °C was found to be the most appropriate protocol for the synthesis of both ([κ 3 -O,C,O]-I NHC)TiCl(OiPr)(THF) 1a and ([κ 3 -O,C,O]-Bz NHC)TiCl(OiPr) 1b complexes, respectively, in good yields and without the formation of side compounds, such as the homoleptic and zwitterionic complexes (Scheme 2). This protocol slightly diverges from the previously reported one in which N,N'-di(2-hydroxy-3,5-di-tertbutylphenyl) imidazolidinium chloride salt (c) was used as proligand and the addition was carried out at room temperature, leading quantitatively to ([κ 3 -O,C,O]-Is NHC)TiCl(OiPr)(THF) 1c [38]. The 1 H and 13 C-NMR spectra show that the proligands are fully deprotonated with the concomitant disappearance of both OH and Himidazolium protons and all chemical resonances are shifted downfield in agreement with the bonding of bis-phenolate NHC ligands to the titanium metal center for both 1a and 1b compounds (Supplementary Figures S1-S4). The only immediately observed difference between these two compounds is that the THF molecule in complex 1b is very labile and can easily be removed under prolonged vacuum. The 13 C-NMR spectra of both compounds 1a and 1b encompass typical imidazol-2-ylidene and benzimidazolin-2-ylidene NHC-Ccarbene resonances at δ 184.0 and 195.2 ppm (Supplementary Figures S2 and S4) [14,69], respectively, which are shifted upfield compared to that of the structurally analogous bis-phenolate saturated NHC complex of titanium ([κ 3 -O,C,O]-Is NHC)TiCl(OiPr)(THF) 1c (δ 198.6 ppm) [38].
As indicated above, the order of addition of the reagents is crucial here, contrasting the case of the Zr(OiPr)4(HOiPr) precursor with either a or b proligands in which the homoleptic and zwitterionic compounds are observed independently of the addition order [44]. For instance, the addition of Ti(OiPr)4 to proligand b in THF at room temperature leads to a mixture of compounds containing at least complex 1b (unambiguously deduced from by 1 H-NMR analysis, cf. Figure S5) along with the zwitterionic ([κ 2 -O,O]-Bz NHC-H)TiCl2(OiPr) 1b' and homoleptic ([κ 3 -O,C,O]-Bz NHC)2Ti 1b'' as minor products (9% and 5%, respectively). Consistent with the observations made earlier for the Bz NHC-Zr analogue [44], formation of zwitterionic 1b' was further confirmed by singlecrystal X-ray diffraction (SCXRD) analysis of a crystal sampled from the reaction mixture in toluene at -30 °C (Supplementary Figure S6 and Table S1). The 1 H and 13 C-NMR spectra show that the proligands are fully deprotonated with the concomitant disappearance of both OH and H imidazolium protons and all chemical resonances are shifted downfield in agreement with the bonding of bis-phenolate NHC ligands to the titanium metal center for both 1a and 1b compounds (Supplementary Figures S1-S4). The only immediately observed difference between these two compounds is that the THF molecule in complex 1b is very labile and can easily be removed under prolonged vacuum. The 13 C-NMR spectra of both compounds 1a and 1b encompass typical imidazol-2-ylidene and benzimidazolin-2-ylidene NHC-C carbene resonances at δ 184.0 and 195.2 ppm (Supplementary Figures S2 and S4) [14,69], respectively, which are shifted upfield compared to that of the structurally analogous bis-phenolate saturated NHC complex of titanium ([κ 3 -O,C,O]-Is NHC)TiCl(OiPr)(THF) 1c (δ 198.6 ppm) [38].
As indicated above, the order of addition of the reagents is crucial here, contrasting the case of the Zr(OiPr) 4 (HOiPr) precursor with either a or b proligands in which the homoleptic and zwitterionic compounds are observed independently of the addition order [44]. For instance, the addition of Ti(OiPr) 4 to proligand b in THF at room temperature leads to a mixture of compounds containing at least complex 1b (unambiguously deduced from by 1 H-NMR analysis, cf. Figure S5) 2 Ti 1b" as minor products (9% and 5%, respectively). Consistent with the observations made earlier for the Bz NHC-Zr analogue [44], formation of zwitterionic 1b' was further confirmed by single-crystal X-ray diffraction (SCXRD) analysis of a crystal sampled from the reaction mixture in toluene at −30 • C (Supplementary Figure S6 and Table S1).
Both 1a and 1b-THF show a slightly distorted octahedral geometry around the Ti(IV) center as a result of the mer-coordination of the tridentate NHC ligand, with ∠O Ar -Ti-O Ar bite angles of 159.14(4) • and 159.18 (8) • , respectively. This ligand coordination is similar to that observed earlier for other bis-phenolate NHC-Ti complexes [20,29,38,40,42,70]. The principal structural features of both 1a and 1b-THF include (i) a mer-NHC chelate deviating from planarity, with torsion angles ∠O Ar -C Ar -N-C carbene of 5.86/6.89 • and 27.04/-25.40 • , respectively, (ii) trans-dispositioning of the carbene moiety and the Cl atom, and (iii) trans-dispositioning of the OiPr and the THF ligand. Compared to the corresponding titanium complex bearing a saturated NHC, ([κ 3 -O,C,O]-Is NHC)TiCl(OiPr)(THF) 1c, the torsion angles ∠O Ar -C Ar -N-C carbene are similar for 1a but far more distorted from planarity for 1b-THF (Table 1) [38]. The Ti-C carbene bond length (2.221(2) Å) in complex 1b-THF is longer than that of 1c (2.166(3) Å), reflected in a shorter Ti-Cl bond distance (2.3459(8) Å) trans to the NHC. This is consistent with a weaker trans influence from the presumably less electron-donating benzimidazolin-2-ylidene moiety [2,[71][72][73]. Less electron donation from the latter ligand and its deviation from planarity might help explain the relatively short Ti-OiPr bond and the tendency toward sp 2 hybridization, suggested by the relatively wide ∠Ti-O-C iPr angle (160.5(3) • , 165.57 (19) • for isomer B), for this oxygen atom. The more sp 2 -like hybridization may bring about increased π-donation from the OiPr moiety and thus explain the apparent greater trans influence and the more weakly bound THF molecule in 1b-THF (Ti-THF avg ≈ 2.28 Å) and 1c than in 1a ( Table 1).
The latter complex has the shortest Ti-C carbene bond distance (2.1310(13) Å) of all reported NHC-Ti complexes of functionalized NHC ligands (Ti-C carbene = 2.14-2.33 Å) [12,14,29]. As expected, the short Ti-C carbene bond is, due to trans influence, reflected in a Ti-Cl trans bond that is longer than in 1b-THF and only slightly shorter than in 1c. Whereas ligands based on the imidazolidin-2-ylidene moiety are often reported to be more electron donating than those of the imidazol-2-ylidene moiety [2,[71][72][73], the short Ti-C carbene bond distance of 1a seems to suggest otherwise. The components of the Ti-C carbene bonds of the three ligands have thus been studied and compared using DFT and NBO analysis (vide infra). Furthermore, the relatively sharp ∠Ti-O-C iPr angle (139.95(5) • ) of 1a seems to suggest more sp 3 -like hybridization and less π-donation of the OiPr ligand. The sharp ∠Ti-O-C iPr angle appears not to be caused by steric repulsion between the imidazol-2-ylidene and OiPr moieties, since the NHC in 1a is only slightly less planar than that in 1c. Thus, the presumed weaker π-donation from OiPr in 1a is consistent with the short Ti-THF bond (2.2573(11) Å) which, in turn, is consistent with the trans influence of OiPr being weaker in 1a than in 1b-THF and 1c.
To further investigate the structural differences of the complexes and their relation to the electronic properties of the NHC ligands, we studied the ligands and the complexes using DFT and NBO analyses. First, the DFT calculations predict the experimentally obtained bond distances accurately, to within 0.01-0.02 Å (Table 1). More importantly, the trend in calculated Ti-C carbene bond distances between the complexes faithfully reproduces that obtained in X-ray crystallographic analysis. The large variation in Ti-C carbene bond distances (up to 9 pm when comparing the X-ray structures) are thus not the result of crystal-packing effects but must instead originate from the carbenes themselves. The experimentally and computationally obtained Ti-C carbene distances thus suggest that the strength of the interaction between the metal and the carbene diminishes in the order a > c > b for the three NHC ligands (Table 1). Valence and torsional angles are also well reproduced, except for the ∠Ti-O-C iPr angle. However, this angle varies by more than 20 • between the three complexes, presumably reflecting a very shallow bending potential.
Regarding the Ti-ligand bond energies and interactions, the bond "snapping" energies (Table 2), i.e., the bond energies calculated by dissociating the tridentate ligands heterolytically to frozen-geometry [TiCl(OiPr)(THF)] 2+ fragments M1a-c and dianionic NHC ligands Ma-c (Supplementary Scheme S1), are consistent with the trend in Ti-C carbene bond distances. Orbital interactions between these pairs of fragments might thus reveal the origin of the trends in both Ti-C carbene bond distances and bond energies. The calculated ligand-to-metal net electron donations (Table 2) are essentially identical for the three complexes, showing that further resolution is necessary for uncovering the factors determining the differences in bond distances and energies.
To uncover these factors, we performed NBO [74] analyses of the individual fragments M1a-c and Ma-c as well as of the metal-ligand orbital interactions in the three complexes. The most important fragment and complex orbitals obtained in these analyses are shown in Figures 2 and 3.  Table S2) by dissociating the tridentate ligands heterolytically to frozen-geometry fragments M1a-c and Ma-c. b The NHC fragment charge is the sum of all the natural atomic charges of the tridentate NHC ligand. c The NHC→Ti net donation is estimated as the number of electrons needed to reach neutrality for a NHC fragment in which the atomic charges of the two oxygen atoms have been subtracted (Table S3).
Molecules 2020, 25, x FOR PEER REVIEW 6 of 22  Table S2) by dissociating the tridentate ligands heterolytically to frozengeometry fragments M1a-c and Ma-c. b The NHC fragment charge is the sum of all the natural atomic charges of the tridentate NHC ligand. c The NHC→Ti net donation is estimated as the number of electrons needed to reach neutrality for a NHC fragment in which the atomic charges of the two oxygen atoms have been subtracted (Table S3).
To uncover these factors, we performed NBO [74] analyses of the individual fragments M1a-c and Ma-c as well as of the metal-ligand orbital interactions in the three complexes. The most important fragment and complex orbitals obtained in these analyses are shown in Figures 2 and 3.  Insight is offered, for example, by the calculated energies of the C σ orbitals of the three OHcontaining free-carbene ligands Ma-cH2 (Table 3 and Supplementary Scheme S1), which suggest that the σ-donating capacity should be greatest for c, followed by a and b. Whereas this ranking is consistent with the relative Ti-Ccarbene bond distances of 1a and 1b-THF, additional factors must explain why this distance is shorter in 1a than in 1c. An explanation might be offered by the interaction between the π-orbital of the ligand (C-N π) and a metal d-orbital of the same symmetry (Ti dπ). Insight is offered, for example, by the calculated energies of the C σ orbitals of the three OH-containing free-carbene ligands Ma-cH 2 (Table 3 and Supplementary Scheme S1), which suggest that the σ-donating capacity should be greatest for c, followed by a and b. Whereas this ranking is consistent with the relative Ti-C carbene bond distances of 1a and 1b-THF, additional factors must explain why this distance is shorter in 1a than in 1c. An explanation might be offered by the interaction between the π-orbital of the ligand (C-N π) and a metal d-orbital of the same symmetry (Ti d π ).  The calculated second-order perturbative estimate of the donor-acceptor interaction between these two orbitals is largest for 1a, followed by 1c, and 1b-THF (Table 4 and Supplementary Table  S4), consistent with the trend in bond distances and bond energies. In other words, π-donation from  The calculated second-order perturbative estimate of the donor-acceptor interaction between these two orbitals is largest for 1a, followed by 1c, and 1b-THF (Table 4 and Supplementary Table  S4  The calculated second-order perturbative estimate of the donor-acceptor interaction between these two orbitals is largest for 1a, followed by 1c, and 1b-THF (Table 4 and Supplementary Table  S4), consistent with the trend in bond distances and bond energies. In other words, π-donation from the ligand to the metal is suggested to be stronger for a and to modify the trend offered by the ligand  The calculated second-order perturbative estimate of the donor-acceptor interaction between these two orbitals is largest for 1a, followed by 1c, and 1b-THF (Table 4 and Supplementary Table S4), consistent with the trend in bond distances and bond energies. In other words, π-donation from the ligand to the metal is suggested to be stronger for a and to modify the trend offered by the ligand σ-donating capacity suggested by the C σ energies in Table 3. The importance of ligand-to-metal π-donation has already been noted for NHC complexes of early transition metals [75][76][77][78][79]. Table 4. Second-order perturbative estimates of donor-acceptor interactions in the NBO basis of 1a, 1b-THF and 1c.

Complex
Donor Orbital a Acceptor Orbital E2 (kcal mol −1 ) a LP refers to lone pair.
Whereas the above-described donation from largely filled ligand π-orbitals to largely empty d π -orbitals of the metal is estimated to contribute significantly to the Ti-NHC bonding, the low occupations of metal d-orbitals of early transition-metal complexes (see, e.g., Figures 2 and 3) suggest that π-back donation from titanium to the NHC is much less important for the present complexes than for complexes of mid-to-late transition metals [71,75]. The metal d-orbitals are considered to be "lone vacant" orbitals in the NBO analysis (Figures 2 and 3), and direct back-donation from the metal to the C-N π* orbitals does not appear in the analysis and is likely to be small.
In contrast, the C-N π* orbitals are reported to sometimes accept electrons from lone pairs of anionic ligands of early transition metals [76][77][78][79]. Weak contributions of this kind, between isopropoxide oxygen lone pairs and the C-N π* orbitals, are seen also in the present three complexes (Table 4). In addition, the second-order perturbation analysis also identifies analogous interactions between the THF oxygen lone pairs and C-N π*. The strongest of these interactions is in 1b-THF, where it is likely to be one of the driving forces behind the tilting of the NHC ligand toward the THF.
In conclusion, the calculations show that the strength of the interactions between the metal and the NHC follows the trend portrayed by the calculated and experimental Ti-NHC bond distances (a > c > b). Although the ligand-to-metal σ-donation is predicted to be stronger for ligand c (followed by  a, and b), the π-donation from a is stronger and contributes to giving the overall trend in metal-ligand interaction strength and bond distances. Whereas back-donation from the metal to the NHC seems to be unimportant, weak donor-acceptor interactions from THF and OiPr lone pairs to the C-N π* orbitals contribute and are probably involved in the tilting of the NHC seen in 1b-THF.
Due to their potential application in polymerization of CHO with CO 2 [29,[41][42][43], the bis-isopropoxide I NHC-and Bz NHC-titanium complexes 2a and 2b were also synthesized and were found to be readily accessible, in quantitative yields (Scheme 2), via salt metathesis of LiOiPr with complex 1a and 1b, respectively, similarly to the saturated NHC-titanium analogue [38].
The NMR spectra of 2a and 2b contain resonances typical of five-coordinate ([κ 3 -O,C,O]-NHC)TiX 2 complexes including a doublet resonance originating from the Me groups of the two OiPr moieties (Supplementary Figures S7-S10), which are consistent with C 2v -symmetric structures in solution for both complexes [38,42,43]. The corresponding 13 Figures S8 and S10) [14,69], respectively, shifted upfield compared to the saturated ([κ 3 -O,C,O]-Is NHC)Ti(OiPr) 2 complex 2c [38]. Furthermore, the complete molecular structures of 2a and 2b were confirmed by SCXRD analysis, showing that these complexes are five-coordinate and adopt a distorted square-pyramidal geometry according to the Addison and Reedijk geometric parameter (τ 5 = 0.49 for 2a and 0.27 for 2b), with one of the OiPr moieties in apical position (Figure 1; see Supplementary Table S6 for crystallographic data) [80]. Both geometries differ from that of saturated ([κ 3 -O,C,O]-Is NHC)Ti(OiPr) 2 complex 2c in which the five-coordinate Ti metal center adopt a trigonal-bipyramidal geometry (τ 5 = 0.51) [38]. The overall structural data for 2a and 2b resemble those previously observed for 1a and 1b-THF, with the following main particularities: (i) an even more pronounced deviation from planarity for the mer-NHC chelate, with torsion angles ∠O Ar -C Ar -N-C carbene of −12.88/19.19 • for 2a and  Table S6) [38]. The sharper ∠O Ar -Ti-O Ar angle (138.19(8) • ) observed for 2b compared to 1b-THF is most likely a result of steric interactions between the tBu and OiPr moieties (Supplementary Table S6).
Aiming to further explore NHC-hafnium compounds as precursors for the copolymerization of epoxide with CO 2 , attempts to synthesize ([κ 3 -O,C,O]-I NHC)HfCl(OiPr)(THF) 3a complex via addition of proligand a to Hf(OiPr) 4 (HOiPr) under the same reaction conditions as for titanium, invariably gave a mixture of unidentifiable compounds. Only when the addition of a to Hf(OiPr) 4 (HOiPr) was performed overnight at room temperature and extended reaction time did the 1 H NMR spectrum of the reaction mixture showed three distinct sets of signals attributable to three different compounds (in ratio ≈ 2:1:0.8), which unfortunately could not be further separated or isolated. The most intense signal set was tentatively attributed to ([κ 3 Figure S11) and 13 C-NMR spectra (with a typical Hf-Ccarbene at δ 201.8 ppm) [14,69]. The minor side-products presumably are the zwitterionic 3b' and the homoleptic 3b'' (Scheme 3). As previously observed for the reactivity of proligands a and b with the Zr-alkoxide precursor, the formation of the homoleptic complex cannot be completely avoided, most likely due to the reaction of a second proligand with the large metal ions such as Hf 4+ . In contrast, the smaller Ti 4+ leads to release of HCl, which, in turn, cleaves off the M-Ccarbene bond in the ([κ 3 -O,C,O]-NHC)MCl(OiPr)(THF) complex and thus to the formation zwitterionic species [44]. Even if 3b could not be further purified, the molecular structure was established by the recovery of single crystals of 3b suitable for SCXRD analysis from a solution of unpurified 3b in pentane at −30 °C (Figure 4). A crystallographic summary for 3b is included with the selected bond lengths, angles, and torsion angles in the electronic supplementary information  Figure S11) and 13 C-NMR spectra (with a typical Hf-C carbene at δ 201.8 ppm) [14,69]. The minor side-products presumably are the zwitterionic 3b' and the homoleptic 3b" (Scheme 3). As previously observed for the reactivity of proligands a and b with the Zr-alkoxide precursor, the formation of the homoleptic complex cannot be completely avoided, most likely due to the reaction of a second proligand with the large metal ions such as Hf 4+ . In contrast, the smaller Ti 4+ leads to release of HCl, which, in turn, cleaves off the M-C carbene bond in the ([κ 3 -O,C,O]-NHC)MCl(OiPr)(THF) complex and thus to the formation zwitterionic species [44]. Even if 3b could not be further purified, the molecular structure was established by the recovery of single crystals of 3b suitable for SCXRD analysis from a solution of unpurified 3b in pentane at −30 • C (Figure 4). A crystallographic summary for 3b is included with the selected bond lengths, angles, and torsion angles in the electronic supplementary information (Tables S5 and S7). As expected, complex 3b exhibits structural features closely related to those of 1b-THF. Similar observations can be made when 3b is structurally compared to its saturated analogue ([κ 3 -O,C,O]-Is NHC)HfCl(OiPr)(THF) 3c than between NHC-Ti complexes of 1b-THF and 1c [45]. The only exception is the angle ∠Hf-O-C iPr , which is sharper in the case of 3b than in 3c (162.4(3) • vs. 171.1(3) • ), indicating slightly diminished π-donation from the OiPr moiety. mixture of compounds, among which ([κ -O,C,O]-NHC)Hf(OiPr)(THF) 3b is identified to be the major product according to 1 H NMR (estimated yield 86%, Supplementary Figure S11) and 13 C-NMR spectra (with a typical Hf-Ccarbene at δ 201.8 ppm) [14,69]. The minor side-products presumably are the zwitterionic 3b' and the homoleptic 3b'' (Scheme 3). As previously observed for the reactivity of proligands a and b with the Zr-alkoxide precursor, the formation of the homoleptic complex cannot be completely avoided, most likely due to the reaction of a second proligand with the large metal ions such as Hf 4+ . In contrast, the smaller Ti 4+ leads to release of HCl, which, in turn, cleaves off the [44]. Even if 3b could not be further purified, the molecular structure was established by the recovery of single crystals of 3b suitable for SCXRD analysis from a solution of unpurified 3b in pentane at −30 °C (Figure 4). A crystallographic summary for 3b is included with the selected bond lengths, angles, and torsion angles in the electronic supplementary information (Tables S5 and S7). As expected, complex 3b exhibits structural features closely related to those of 1b-THF. Similar observations can be made when 3b is structurally compared to its saturated analogue ([κ 3 -O,C,O]-Is NHC)HfCl(OiPr)(THF) 3c than between NHC-Ti complexes of 1b-THF and 1c [45]. The only exception is the angle ∠Hf-O-CiPr, which is sharper in the case of 3b than in 3c (162.4(3)° vs. 171.1(3)°), indicating slightly diminished π-donation from the OiPr moiety. As previously reported for the bis-isopropoxide Bz NHC-titanium complex 2b, the Hf analogue 4b was also synthesized via salt metathesis from the reaction of LiOiPr with complex 3b (Scheme 3). As previously reported for the bis-isopropoxide Bz NHC-titanium complex 2b, the Hf analogue 4b was also synthesized via salt metathesis from the reaction of LiOiPr with complex 3b (Scheme 3). Although many attempts to isolate the 4b in its pure form were unsuccessful, the NMR data unambiguously allowed identification of 4b as the major product (Supplementary Figure S12).

Copolymerization of CHO with CO 2
The copolymerization of CHO and CO 2 was investigated by using unsaturated and benzannulated NHC-titanium and hafnium complexes in combination with 1 equiv. of bis(triphenylphosphine)iminium chloride ([PPN]Cl) as ionic co-catalyst in neat CHO (CHO:M = 1250:1) under mild conditions (P CO2 = 2 bar, at 65 • C) for 24 h ( Table 5).
The results were compared with the benchmark saturated-NHC complexes  (Table 5) [41][42][43]45]. As for the benchmark binary catalyst systems, all NHC-Ti and NHC-Hf catalysts gave completely alternating PCHC selectively (99% in carbonate linkage) without concomitant formation of CHC or PCHO. Another characteristic feature of the new catalysts was that they all produced PCHCs of molecular weights below 4.5 kg mol −1 , with bimodal distributions and relatively narrow polydispersities (Ð < 1.6), indicating a controlled polymerization (entries 1−5 and 7−8, Table 5). A noticeable trend among the NHC-Ti catalysts is that the unsaturated NHC-Ti 1−2a/[PPN]Cl systems are less active and productive than the benzannulated NHC-Ti 1−2b/[PPN]Cl systems (entries 1−3 and 5, Table 5). To allow for a direct comparison with the benchmark saturated NHC-Ti 1c catalyst, the reaction time was shortened to 5 h for avoiding recurrent mass transfer issues about half conversion in neat CHO [41]. It was found that    (Table 5) [41][42][43]45]. As for the benchmark binary catalyst systems, all NHC-Ti and NHC-Hf catalysts gave completely alternating PCHC selectively (99% in carbonate linkage) without concomitant formation of CHC or PCHO. Another characteristic feature of the new catalysts was that they all produced PCHCs of molecular weights below 4.5 kg mol −1 , with bimodal distributions and relatively narrow polydispersities (Ɖ < 1.6), indicating a controlled polymerization (entries 1−5 and 7−8, Table 5). A noticeable trend among the NHC-Ti catalysts is that the unsaturated NHC-Ti 1−2a/[PPN]Cl systems are less active and productive than the benzannulated NHC-Ti 1−2b/[PPN]Cl systems (entries 1−3 and 5, Table 5). To allow for a direct comparison with the benchmark saturated NHC-Ti 1c catalyst, the reaction time was shortened to 5 h for avoiding recurrent mass transfer issues about half conversion in neat CHO [41]. It was found that saturated catalyst system 5/  [62]. The more active catalysts 1c and 1b-THF have, according to SCXRD and DFT, longer Ti-THF bonds than 1a, which indicate a weakly bonded THF ligand and a higher rate of formation of the active species and/or, by analogy to other catalytic systems [46][47][48][49][50]52,81], a faster dissociation of the growing polymer chains during the copolymerization. The length of the Ti-THF bond, in turn, does not correlate in a straightforward fashion with the net electron donation from the NHC moiety to the metal center (which are very similar; see Table 2) or with the length or strength of the Ti-NHC bond. The latter bond appears to influence the lability of the THF ligand more indirectly, via the Ti-OiPr The trend in catalytic activity (1c > 1b-THF > 1a) might originate from the inherent stability of the complexes and might also, at least in part, originate from the lability of the THF molecule and the ease with which this ligand is displaced to form the putative  [62]. The more active catalysts 1c and 1b-THF have, according to SCXRD and DFT, longer Ti-THF bonds than 1a, which indicate a weakly bonded THF ligand and a higher rate of formation of the active species and/or, by analogy to other catalytic systems [46][47][48][49][50]52,81], a faster dissociation of the growing polymer chains during the copolymerization. The length of the Ti-THF bond, in turn, does not correlate in a straightforward fashion with the net electron donation from the NHC moiety to the metal center (which are very similar; see Table 2) or with the length or strength of the Ti-NHC bond. The latter bond appears to influence the lability of the THF ligand more indirectly, via the Ti-OiPr bond trans to the THF. For example, the long Ti-C carbene bond in 1b-THF results in a short Ti-OiPr bond and, due to trans influence, a long and presumably weak Ti-THF bond.
The NMR spectra (Bruker, Billerica, MA, USA) of air and moisture sensitive compounds were recorded by using J. Young valve NMR tubes at 25 • C on a Bruker-BIOSPIN-AV500 ultrashield 500 plus (5 mm BBO with z-gradient BTO, 1 H: 500.13 MHz; 13 C: 125.75 MHz), and a Bruker Ascend AV850 III HD (5 mm triple resonance CryoProbe, 1 H: 850.13 MHz; 13 C: 213.77 MHz). 1 H and 13 C shifts are referenced to internal solvent resonances and reported in parts per million relative to TMS. DRIFT spectra (Thermo Nicolet, Madison, WI, USA) were recorded by using a Nicolet protégé 460 ESP FTIR spectrometer and a DRIFT cell (KBr window) under argon atmosphere. The spectra were averaged over 64 scans; the resolution was ± 4 cm −1 . Elemental analysis of C, H and N elements was performed on an Elementar Vario EL III. GPC-SEC (Viscotek-Malvern, Worcestershire, UK) was measured, to determine M n and M w of the PCHC polymers obtained from the catalytic testing, from Viscotek. Narrow polystyrene PS-99K (M w = 99.284 kg mol −1 , M n = 97.381 kg mol −1 and IV = 0.477 dL g −1 ) and all calibration standards were obtained from Malvern PolyCAL. Approx. 30 mg of each polymer, obtained from the catalytic testing, were dissolved in exactly 10 mL THF (containing 250 ppm BHT inhibitor). The sample solutions (≈ 3.0 mg mL −1 in THF) were filtered through syringe filter Whatmann (0.45 µm pore size) prior to injection. Chromatographic separation was performed at a column temperature of 30 • C with a flow rate of 1 mL min −1 . SEC was performed with a pump supplied by Viscotek (GPCmax), employing two ViscoGel columns. Signals were detected by means of a triple detection array (TDA 302) and calibrated against polystyrene standards (Ð < 1.2, from 0.12-940 kg mol −1 ).

Synthesis of ([κ 3 -O,C,O]-Bz NHC)Hf(OiPr) 2 (THF) 4b
In a glovebox, to a solution of 3b (37.2 mg, 0.043 mmol) in 10 mL THF was added dropwise 1.1 equiv. LiOiPr (3.1 mg, 0.047 mmol) dissolved in 5 mL THF. The solution immediately turned yellow and then completely colorless after stirring at room temperature for 24 h. The reaction mixture was dried under vacuum and extracted with hexane. The colorless solution mixture was centrifuged, filtered, washed with pentane, and then dried under vacuum affording a white powder corresponding to compound 4b as major compound in quantitative yield along with minor unidentified side compounds. This compound could never be isolated in pure form even after repeated washings with hydrocarbon solvents. Major compound 4b: 1 H-NMR (500.13 MHz, C 6

Copolymerization of CHO and CO 2
A detailed copolymerization procedure is described as a typical example ( Table 5, Entry 1). In a glovebox, an reaction tube for low-pressure reactions equipped with a magnetic stirring bar, a solution of the [PPN]Cl cocatalyst (8 µmol) in dichloromethane (ca. 1 mL) was added under vigorous stirring to a solution of complex 1a (8 µmol) in dichloromethane (ca. 1 mL). The mixture was stirred at ambient temperature for 15 min and the solvent was removed under vacuum (30 min). The resulting solid was then dissolved in a precooled (−30 • C) solution of CHO (10 mmol). Then to the reaction tube was added 2 bar of CO 2 and the reaction mixture was heated to 65 • C. After 24 h, the reaction was cooled down and the pressure was released. An aliquot of the solution was taken for characterization of crude material by 1 H-NMR spectroscopy in CDCl 3 . Then, the reaction mixture was quenched with 1 mL of acidic methanol, precipitated with methanol, and dried for several hours at 80 • C until constant weight. The yield was determined gravimetrically.

X-ray Crystallographic Details
Suitable crystals for diffraction experiments were selected in a glovebox and mounted in a minimum of Parabar 10,312 oil (Hampton Research) in a nylon loop and then mounted under a nitrogen cold stream from an Oxford Cryosystems 700 series open-flow cryostat. Data collection was done on a Bruker AXS TXS rotating anode system with an APEXII Pt 135 CCD detector (Bruker-AXS, Madison, WI, USA) using graphite-monochromated Mo Kα radiation (λ = 0.710 73 Å). Data collection and data processing were done using APEX2 [85], SAINT [86], and SADABS [87] version 2012/1, whereas structure solution and final model refinement were done using SHELXS [88]

Computational Methods
All density functional theory (DFT) calculations were performed with the Gaussian 16 suite of programs [91].

Geometry Optimization
Geometry optimization was performed using the Gaussian 16 implementation of the generalized-gradient approximation (GGA) functional of Perdew, Burke and Ernzerhof (PBE) [92] including Grimme's D3 empirical dispersion term [93] with revised Becke-Johnson damping parameters [94] (overall labelled PBE-D3M(BJ) for brevity). All atoms except titanium were described by Dunning's correlation-consistent valence triple-ζ plus polarization basis sets (termed cc-pVTZ) [95,96], as retrieved from the EMSL basis set exchange database [97,98]. Titanium was described by the Stuttgart 10-electron relativistic effective core potential (termed ECP10MDF) in conjunction with its accompanying primitive basis set (8s7p6d2f1g) contracted to [6s5p3d2f1g]. Both the effective core potential and the accompanying basis set were retrieved from the Stuttgart/Cologne group website [99]. Numerical integrations were performed using the Gaussian 16 (99,590) "ultrafine" grid (keyword int = ultrafine), a pruned grid consisting of 99 radial shells and 590 angular points per shell, except when solving the coupled-perturbed Hartree-Fock equations (part of the analytical second-derivatives calculations) for which the pruned (75,302 grid) "fine" grid was used (keyword CPHF=(Grid=Fine)). Geometries were optimized using tight convergence criteria (max. force 1.5 × 10 −5 a.u., RMS force 1.0 × 10 −5 a.u., max. force 6.0 × 10 −5 a.u., RMS force 4.0 × 10 −5 a.u.), without symmetry constraints, using convergence criteria for the self-consistent field (SCF) optimization procedure that were tightened tenfold compared to the default settings. The tightened criteria were RMS change in density matrix < 1.0 × 10 −9 and max. change in density matrix < 1.0 × 10 −7 . All stationary points were characterized by the eigenvalues of the analytically calculated Hessian matrix and confirmed to be minima.

Single-Point Energy Calculations
All single-point energy calculations were performed with the same PBE-D3M(BJ) functional as described above for geometry optimization. For titanium, carbon, and hydrogen atoms, the basis sets were also the same as those used in the geometry optimizations. All other atoms (N, O and Cl) were described by correlation-consistent valence triple-ζ plus polarization basis sets augmented by diffuse functions (EMSL: aug-cc-pVTZ) [95,97,98,100]. Numerical integrations were performed with the "ultrafine" grid of Gaussian 16, and the SCF density-based convergence criterion was set to 10 −5 (RMS change in density matrix < 1.0 × 10 −5 , max. change in density matrix = 1.0 × 10 −3 ).

Natural Bond Orbital Calculations
All natural bond orbital calculations were performed with the NBO7 program [74], using the wavefunction produced by the Gaussian 16 single-point energy calculation as input. Keywords used in the NBO7 job include "bndidx", which requests the print-out of the NAO-Wiberg Bond Index array, "NBO" which requests the calculation and printing of NBO's, and "DMNAO" which requests the natural atomic orbital density matrix. To get a comparable set of orbitals among the complexes, the Lewis structures were explicitly restricted to those shown in Supplementary Scheme S2 via the $CHOOSE input section.

Conclusions
A series of titanium and hafnium complexes bearing unsaturated and benzannulated tridentate, bis-phenolate NHC ligands have been synthesized and characterized. The Ti-C carbene distances with which these ligands bind to the metal vary considerably (by 9 pm), and these differences manifest themselves, via trans influence and other "ripple effects", in significant variations also in the other metal-ligand bond distances. These structural differences and their relation to the metal-NHC bonds and the electronic properties of the ligands have been studied for titanium complexes 1a, 1b-THF, and 1c using DFT and NBO analyses. The shorter Ti-C carbene distance in 1a than in the other two complexes seems to originate from stronger ligand-to-metal π-donation, whereas the corresponding σ-donation is weaker than in 1c. In contrast, back-donation directly from the metal to the NHC ligand seems to be unimportant in these d 0 complexes. Still, the C-N π* NHC orbitals are involved in bonding as they interact with THF and isopropoxide oxygen lone-pair donor orbitals, an interaction that appears to contribute to the tilting of the NHC ligand toward the THF in 1b-THF.
The new complexes catalyze the copolymerization of CHO with CO 2 under mild reaction conditions (P CO2 = 2 bar and 65 • C) to exclusively give poly(cyclohexene carbonate) product, albeit with low-to-moderate yields. Among the new complexes, the benzannulated-NHC-coordinated titanium complex (1b-THF) gives the most active catalyst upon activation with [PPN]Cl. Including previously reported complexes, the order among the NHC ligands in terms of catalytic activity is as follows: imidazolidin-2-ylidene (saturated) > benzimidazolin-2-ylidene (benzannulated) > imidazolin-2-ylidene (unsaturated). Although further mechanistic studies are needed to uncover the factors governing this order, it might be influenced by the inherent stability of the complexes and possibly also the lability of the THF ligand, as suggested by the variation in Ti-THF distance among the complexes.