Transition Metal-Catalyzed and MAO-Assisted Olefin Polymerization; Cyclic Isomers of Sinn’s Dimer Are Excellent Ligands in Iron Complexes and Great Methylating Reagents

Methylaluminoxane (MAO) is the most commonly used co-catalyst for transition metalcatalyzed olefin polymerization, but the structures of MAO species and their catalytic functions remain topics of intensive study. We are interested in MAO-assisted polymerization with catalysts L(R2)FeCl2 (L = tridentate pyridine-2,6-diyldimethanimine; imine-R = Me, Ph). It is our hypothesis that the MAO species is not merely enabling Fe–Me bond formation but functions as an integral part of the active catalyst, a MAO adduct of the Fe-precatalyst [L(R2)FeCl]. In this paper, we explored the possible structures of acyclic and cyclic MAO species and their complexation with pre-catalysts [L(R2)FeCl] using quantum chemical approaches (MP2 and DFT). We report absolute and relative oxophilicities associated with the Fe← O(MAO) adduct formation and provide compelling evidence that oxygen of an acyclic MAO species (i.e., O(AlMe2)2, 4) cannot compete with the O-donor in cyclic MAO species (i.e., (MeAlO)2, 7; MeAl(OAlMe2)2, cyclic 5). Significantly, our work demonstrates that intramolecular O→ Al dative bonding results in cyclic isomers of MAO species (i.e., cyclic 5) with high oxophilicities. The stabilities of the [L(R2)FeClax(MAO)eq] species demonstrate that 5 provides for the ligating benefits of the cyclic MAO species 4 without the thermodynamically costly elimination of TMA. Mechanistic implications are discussed for the involvement of such Fe–O–Al bridged catalyst in olefin polymerization.


Introduction
Ziegler-Natta polymerization [1,2] is the dominant method for the production of polyolefin [3,4]. Ziegler-Natta catalysts usually contain transition metal complexes (i.e., TiCl 4 ) and partially hydrolyzed aluminum alkyls as co-catalyst, among which methylaluminoxane [5][6][7] (MAO) is most commonly used today [8][9][10]. It was a major advance in the field when Brookhart discovered that Fe(II) complexes with di-nitrogen or tri-nitrogen ligands can serve as pre-catalyst with satisfying catalytic activity [11][12][13]. More recently, Sun and co-workers explored structurally similar bidentate bis(imino)pyridyl Fe(II) complexes [14], and tridentate 2,8-bis(imino)quinoline Fe(II) complexes [15], and related systems with nickel [16][17][18], cobalt [19,20], and titanium [21,22] also have been studied. However, these kinds of pre-catalysts require the presence of a very large excess of MAO as cocatalyst (>2000:1) and the structure and the function of MAO have not been clarified at all [23]. The experimental studies of aluminoxanes by Barron [24,25], Pasynkiewicz [26], and Sinn [27] suggested a plethora of chain, ladder, and cage structures. Theoretical studies by Ziegler et al. [28], by Hall and co-workers [29], and by Linnolahti et al. [30,31] explored possible species with ladder and cage structures derived from dimethylaluminum hydroxide (DMAH). In Scheme 2, three mechanisms of catalysis are outlined for a generic LFeCl2 pre-catalyst, where L signifies a tri-nitrogen ligand. All of them share the feature that at least one chloride is abstracted to generate an overall positively charged Fe(II) complex. The abstracted "first" chloride might aggregate with a MAO species and form a large "A − " anion which associated with the iron complex, and this scenario is shown in the top row. In this commonly considered mechanism [35], the next step consists in the replacement of the In Scheme 2, three mechanisms of catalysis are outlined for a generic LFeCl 2 precatalyst, where L signifies a tri-nitrogen ligand. All of them share the feature that at least one chloride is abstracted to generate an overall positively charged Fe(II) complex. The abstracted "first" chloride might aggregate with a MAO species and form a large "A − " anion which associated with the iron complex, and this scenario is shown in the top row. In this commonly considered mechanism [35], the next step consists in the replacement of the second chloride by a methyl anion (from MAO) forming an Fe-Me bond. After this alkylation step and on entry of an alkene into the vacant coordination site, a nucleophilic addition of the methyl anion to the coordinated alkene results in the replacement of the methyl anion ligand by a propyl anion ligand, and so on. Alternatively, one may consider mechanisms in which the MAO species does not alkylate the transition metal and instead coordinates to the transition metal (Scheme 2, center and bottom) and create a Fe-O-Al bridged catalyst. After their coordination to the transition metal, the MAO species would transfer a methyl group directly from Al to the olefin. An Al-based alkylation mechanism with Fe−O(MAO) bonding should discriminate strongly between acyclic and cyclic aluminoxanes (Scheme 3). The bridging oxygen in an acyclic aluminoxane may engage both of its p-lone pairs to engage in O→Al dative bonding. This is illustrated by resonance forms 4-IIa, 4-IIb, and 4-III in Scheme 3 for Me2Al−O−AlMe2 (R = Me). The strength of this O→Al dative bonding may be reduced by coordination of other donors to Al, but it will not be eliminated. Therefore, any O→Fe dative bonding in a complex of an acyclic species will always have to compete with two An Al-based alkylation mechanism with Fe-O(MAO) bonding should discriminate strongly between acyclic and cyclic aluminoxanes (Scheme 3). The bridging oxygen in an acyclic aluminoxane may engage both of its p-lone pairs to engage in O → Al dative bonding. This is illustrated by resonance forms 4-IIa, 4-IIb, and 4-III in Scheme 3 for Me 2 Al-O-AlMe 2 (R = Me). The strength of this O → Al dative bonding may be reduced by coordination of other donors to Al, but it will not be eliminated. Therefore, any O → Fe dative bonding in a complex of an acyclic species will always have to compete with two options for O → Al dative bonding. Cyclodialuminoxane 7 exemplifying cyclic MAO species in Scheme 2. Each oxygen in a cyclic aluminoxane can only engage its out-of-plane p-lone pair p oop for O → Al dative bonding and, thus, the one in-plane σ-lone pair σ ip will be available for O → Fe dative bonding. We will substantiate this argument below and it is for that reason that the mechanisms shown in Scheme 1 involve cyclic MAO species. The consideration of cyclodialuminoxane as the active co-catalyst warrants the study of two possible issues. First, the formation of cyclodialuminoxane 7 from the Sinn dimer 5a by TMA elimination might not be able to compete with oligomerization of the Sinn dimer 5a to the Sinn trimer 6. Second, the Fe-coordinated cyclodialuminoxane 7 is only a weak methyl donor.
R PEER REVIEW 4 of 16 options for O→Al dative bonding. Cyclodialuminoxane 7 exemplifying cyclic MAO species in Scheme 2. Each oxygen in a cyclic aluminoxane can only engage its out-of-plane plone pair poop for O→Al dative bonding and, thus, the one in-plane σ-lone pair σip will be available for O→Fe dative bonding. We will substantiate this argument below and it is for that reason that the mechanisms shown in Scheme 1 involve cyclic MAO species. The consideration of cyclodialuminoxane as the active co-catalyst warrants the study of two possible issues. First, the formation of cyclodialuminoxane 7 from the Sinn dimer 5a by TMA elimination might not be able to compete with oligomerization of the Sinn dimer 5a to the Sinn trimer 6. Second, the Fe-coordinated cyclodialuminoxane 7 is only a weak methyl donor. Here we report on the thermochemistry of the formation of the Sinn dimer 5 from the Sinn monomer 4. Detailed exploration of the potential energy surface of 5 reveals that the acyclic Sinn dimer structure 5a can rearrange easily into cyclic isomers 5b and 5c, which effectively are TMA-complexed cyclodialuminoxane, 7•TMA. We then discuss the ligand binding capacities of acyclic 4, cyclic 7, and cyclic 5 with

Results and Discussion
Isomer preference energies, activation, and reaction energies are discussed, and we report relative energies ΔE, enthalpies, ΔH0 = Δ(E + VZPE) and ΔH298 = Δ(E + TE), and free enthalpies ΔG298 = Δ(E + TE − 298.153·S). Results for the MAO species are collected in Table  1 and all relevant energy of complexation chemistry are collected in Table 2. Computated energies of MAO species and the iron pre-catalyst are shown in Tables S1 and S2. Structures of optimized MAO species are shown in Figures S1 to S3. Cartesian coordinates of all the stationary structures are shown in the supporting information. Here we report on the thermochemistry of the formation of the Sinn dimer 5 from the Sinn monomer 4. Detailed exploration of the potential energy surface of 5 reveals that the acyclic Sinn dimer structure 5a can rearrange easily into cyclic isomers 5b and 5c, which effectively are TMA-complexed cyclodialuminoxane, 7•TMA. We then discuss the ligand binding capacities of acyclic 4, cyclic 7, and cyclic 5 with Fe 2+ complexes of type [L(R 2 )FeCl] + . The organic ligand L is the tridentate pyridine-2,6-diyldimethanimine and we studied the N,N-diphenyl ligand L(Ph) 2 and the N,N-dimethyl ligand L(Me) 2 . The structures of the [L(R 2 )FeCl(MAO)] + species show that the 5 is a much better Me-donor than 4, and [L(R 2 )FeCl(5)] + illustrates well the mechanism shown in the bottom row of Scheme 2. The stabilities of the [L(R 2 )FeCl(MAO)] + species demonstrate that 5 provides for the ligating benefits of the cyclic MAO species 4 without the thermodynamically costly elimination of TMA.

Results and Discussion
Isomer preference energies, activation, and reaction energies are discussed, and we report relative energies ∆E, enthalpies, ∆H 0 = ∆(E + VZPE) and ∆H 298 = ∆(E + TE), and free enthalpies ∆G 298 = ∆(E + TE − 298.153·S). Results for the MAO species are collected in Table 1 and all relevant energy of complexation chemistry are collected in Table 2. Computated energies of MAO species and the iron pre-catalyst are shown in Tables S1 and S2. Structures of optimized MAO species are shown in Figures S1-S3. Cartesian coordinates of all the stationary structures are shown in the supporting information.

Formation of Sinn Dimer 5 and Its Structural Isomerization
The most likely path for the formation of trialuminoxane 5 is outlined in Scheme 1, stationary structures of 5 are shown in Figure 1, and relevant energies are collected in Table 1. Acyclic trialuminoxane 5a can be formed from 4 by adding DMAH and losing methane. The reaction energy ∆G 298 = −45.4 kcal/mol for this intermolecular CH 4 elimination between DMAH and 4 to form 5a is very similar to the respective value of ∆G 298 = −41.2 kcal/mol for the reaction of DMAH and TMA.
The acyclic structure 5a features two near-linear Al-O-Al moieties indicative of a strong O → Al π-dative bonding. Structure 5a can stabilize itself by intramolecular O → Al σ-dative bond formation and resulting in minimum 5b ( Figure 1). The cyclization energy of 5 is just about compensated by the loss in cyclization entropy, and 5b is only 0.6 kcal/mol more stable than 5a. The additional O-Al contact (1.91 Å) is only slightly longer than a normal O-Al bond (1.8 Å) and the exocyclic AlMe 2 moiety assumes a position to serve as the recipient of methyl-bridging by one CH 3 group of the endocyclic AlMe 2 group: d(Al endo -CH 3 Interesting is the fact that there exists a second cyclic minimum 5c that is essentially isoenergetic with 5b. Structure 5c is C s -symmetric, it contains two bridging methyl groups, the methyl bridges are asymmetric in the opposite direction compared to 5b, that is, d(Al endo -CH 3 ) = 2.256 Å > d(Al exo -CH 3 ) = 2.089 Å.
The stationary structures of 5a, 5b, and 5c are showed in the Figure 1, together with the transition state structure TSRF(5a,5b) for ring formation 5a 5b, TSMT(5b,5c) for isomerization 5b 5c, and ATS(5c) for automerization of 5c. All the isomerization processes are facile even at low temperature. Hence, we conclude that cyclic 5b is thermodynamically competitive and kinetically accessible.
Structures 5b and 5c can be viewed as the adducts of 7 and TMA. We previously discussed the formation of 7 via dimerization of 3 and the difficulty of generating 3 [32]. In this context, the formation of cyclic structures of 5 can be viewed as an "Umleitung" (German, detour) to a TMA adduct of 7 without going through 7.
The acyclic structure 5a features two near-linear Al−O−Al moieties indicative of a strong OAl π-dative bonding. Structure 5a can stabilize itself by intramolecular OAl σ-dative bond formation and resulting in minimum 5b (Figure 1). The cyclization energy of 5 is just about compensated by the loss in cyclization entropy, and 5b is only 0.6 kcal/mol more stable than 5a. The additional O−Al contact (1.91 Å) is only slightly longer than a normal O−Al bond (1.8 Å) and the exocyclic AlMe2 moiety assumes a position to serve as the recipient of methyl-bridging by one CH3 group of the endocyclic AlMe2 group: d(Alendo-CH3) = 2.091 Å, d(Alexo-CH3) = 2.223 Å. Interesting is the fact that there exists a second cyclic minimum 5c that is essentially isoenergetic with 5b. Structure 5c is Cs-symmetric, it contains two bridging methyl groups, the methyl bridges are asymmetric in the opposite direction compared to 5b, that is, The stationary structures of 5a, 5b, and 5c are showed in the Figure 1, together with the transition state structure TSRF(5a,5b) for ring formation 5a ⇆ 5b, TSMT(5b,5c) for isomerization 5b ⇆ 5c, and ATS(5c) for automerization of 5c. All the isomerization processes are facile even at low temperature. Hence, we conclude that cyclic 5b is thermodynamically competitive and kinetically accessible. Acyclic allene structure 5a, cyclic, monobridged structure 5b, and cyclic, dibridged structure 5c. Also shown are the computed structures of the transition state structures TSRF(5a,5b) for ring formation, of the transition state structure TSMT(5b,5c) for methyl transfer between 5b and 5c, and of the transition state structure ATS(5c) for automerization 5c 5c.

Iron-Complexes: Dichlorides and Monochlorides
Here we report on the complexation of iron(II) complexes derived from complexes of the type L(R 2 )FeCl 2 (Figure 2). The ligand L(R) 2 signifies a tridentate organic ligand and pyridine-2,6-diyldimethanimine (R = H) was considered as the parent organic tridentate ligand. Many of the organic ligands employed in Fe(II)/MAO catalyzed olefin polymerizations are N-aryl imines, and we studied the N,N-diphenyl ligand L(Ph) 2 and the N,N-dimethyl ligand L(Me) 2 .
Here we report on the complexation of iron(II) complexes derived from complexes of the type L(R2)FeCl2 (Figure 2). The ligand L(R)2 signifies a tridentate organic ligand and pyridine-2,6-diyldimethanimine (R = H) was considered as the parent organic tridentate ligand. Many of the organic ligands employed in Fe(II)/MAO catalyzed olefin polymerizations are N-aryl imines, and we studied the N,N-diphenyl ligand L(Ph)2 and the N,Ndimethyl ligand L(Me)2. The structures of dichloride complexes are shown in the first column of Figure 2. For L(Me2)FeCl2 there is one C2v symmetric minimum. Two minima exist for complex L(Ph2)FeCl2 (Figure 2) depending as to whether the phenyl twists occur in the same (conrot., C2) or in opposite (disrot., Cs) directions, and these complexes are essentially isoenergetic ( Table 2).
It is thought that the dichloride complex is a pre-catalyst and that the dissociation of one chloride generates the cationic complex [L(R2)FeCl] + . The cationic complexes [L(R2)FeCl] + (Figure 2) essentially retain the coordination geometry between iron and the organic ligand L(R2) as in the structure of the neutral dichloride complexes. There is one Cs-structure [L(Me2)FeCl] + and two perspectives are shown in columns two and three of The structures of dichloride complexes are shown in the first column of Figure 2. For L(Me 2 )FeCl 2 there is one C 2v symmetric minimum. Two minima exist for complex L(Ph 2 )FeCl 2 (Figure 2) depending as to whether the phenyl twists occur in the same (conrot., C 2 ) or in opposite (disrot., C s ) directions, and these complexes are essentially isoenergetic ( Table 2).
It is thought that the dichloride complex is a pre-catalyst and that the dissociation of one chloride generates the cationic complex [L(R 2 )FeCl] + . The cationic complexes [L(R 2 )FeCl] + (Figure 2) essentially retain the coordination geometry between iron and the organic ligand L(R 2 ) as in the structure of the neutral dichloride complexes. There is one C s -structure [L(Me 2 )FeCl] + and two perspectives are shown in columns two and three of Figure 2. Isomers are possible for [L(Ph 2 )FeCl] + depending on the conformation of the N-Ph bonds. In the monochloride, there are now two ways to obtain the minima with disrotatory N-Ph twists, I and II, in addition to the structure III resulting from conrotatory N-Ph twists. Structure I is slightly preferred over II and III (Table 2, top).
All structures of [L(R) 2 FeCl] + feature chloride in pseudo-axial position and there are no minima in which chlorine would be placed in the best plane of the organic ligand. We found this observation significant and wanted to explore the possible origin. In Figure 3 are displayed the electrostatic potential maps of [L(Me) 2 FeCl] + and the red and blue regions coincide with the location of the chlorine and iron, respectively. The iron center is the most electrophilic site by far (dark blue region in Figure 3) and the perfect attractor for olefin monomers. The placement of the chloride ligand in the pseudo-axial position concentrates the iron electrophilicity with the additional advantage of minimizing steric interaction with the incoming olefin. disrotatory N−Ph twists, I and II, in addition to the structure III resulting from conrotatory N−Ph twists. Structure I is slightly preferred over II and III (Table 2, top).
All structures of [L(R)2FeCl] + feature chloride in pseudo-axial position and there are no minima in which chlorine would be placed in the best plane of the organic ligand. We found this observation significant and wanted to explore the possible origin. In Figure 3 are displayed the electrostatic potential maps of [L(Me)2FeCl] + and the red and blue regions coincide with the location of the chlorine and iron, respectively. The iron center is the most electrophilic site by far (dark blue region in Figure 3) and the perfect attractor for olefin monomers. The placement of the chloride ligand in the pseudo-axial position concentrates the iron electrophilicity with the additional advantage of minimizing steric interaction with the incoming olefin.

Iron-Complexation by Sinn Monomer 4, Cyclic Aluminoxane 7, and Sinn Dimer 5
The iron dichloride complexes contain both chlorides in pseudo-axial positions, that is, the Fe-Cl directions are almost perpendicular to the LFe plane. Dissociation of one chloride leaves the other one in its pseudo-axial position in the [L(R) 2 FeCl] + complexes ( Figure 2). One might expect that the entry of a MAO ligand would occupy the pseudo-axial vacancy left by the removed chloride. Yet, as will be shown by the optimized structures below, this is not the case in any of [L(R) 2 FeCl(MAO)] + complexes. Instead, the MAO species will occupy a pseudo-equatorial position so that it may act as a bidentate ligand engaging in O → Fe bonding and also in Cl → Al bonding. This mode of bidentate MAO coordination is illustrated in Scheme 4.
It is a significant advantage of this MAO coordination mode that the second pseudoaxial position remains vacant and ready for the coordination of an alkene. For the olefin polymerization to proceed by the mechanism outlined in the bottom row of Scheme 2, it is essential that the MAO species is placed next to the olefin, and this requires complexes of the type shown in Scheme 4 with an axial chlorine and an equatorial MAO species.
The desirable [L(R) 2 FeCl ax (MAO) eq ] + complexes compete with the isomeric [L(R) 2 Fe-Cl eq (MAO) ax ] + complexes. These isomeric complexes also contain one O → Fe and one Cl → Al contact and one vacancy. However, the vacancy is next to the chloride rather than the MAO species.
We computed both the [L(R) 2 FeCl ax (MAO) eq ] + and [L(R) 2 FeCl eq (MAO) ax ] + complexes for the three MAO species 4, 7 and 5b for both the organic ligands L(Ph) 2 and L(Me) 2 . Molecular models of optimized structures of these twelve complexes are shown in Figures 4-6 for 4, 7 and 5b Table 2, and the energetically preferred isomers are shown with green frames in the figures. The MAO species 4 and 7 are symmetric and there is only one way for them to enter the coordination sphere of the [L(R) 2 FeCl ax ] + complex. On the other hand, the ligand 5 is distinctly asymmetric. The structure of free 5 suggested that 5b and 5c as the best structural isomers with one or two methyl groups in a bridging position (Figure 1). However, in the complexes, the ligand has the topology of 5e (Scheme 1), that is, the exocyclic AlMe 2 moiety is almost in the Al 2 O 2 plane with the possibility of O → AlMe 2 dative bonding. This topology occurs in all four complexes with 5 ( Figure 6). One of the in-ring Al atoms has only one methyl attached and coordinates to chloride (Al a ), while the other in-ring Al atom is bonded to two methyl groups (Al b ). For complex L(Ph) 2 FeCl ax (5)

Iron-Complexation by Sinn Monomer 4, Cyclic Aluminoxane 7, and Sinn Dimer 5
The iron dichloride complexes contain both chlorides in pseudo-axial positions, that is, the Fe−Cl directions are almost perpendicular to the LFe plane. Dissociation of one chloride leaves the other one in its pseudo-axial position in the [L(R)2FeCl] + complexes ( Figure 2). One might expect that the entry of a MAO ligand would occupy the pseudoaxial vacancy left by the removed chloride. Yet, as will be shown by the optimized structures below, this is not the case in any of [L(R)2FeCl(MAO)] + complexes. Instead, the MAO species will occupy a pseudo-equatorial position so that it may act as a bidentate ligand engaging in O→Fe bonding and also in Cl→Al bonding. This mode of bidentate MAO coordination is illustrated in Scheme 4. It is a significant advantage of this MAO coordination mode that the second pseudoaxial position remains vacant and ready for the coordination of an alkene. For the olefin polymerization to proceed by the mechanism outlined in the bottom row of Scheme 2, it is essential that the MAO species is placed next to the olefin, and this requires complexes of the type shown in Scheme 4 with an axial chlorine and an equatorial MAO species.
The desirable [L(R)2FeClax(MAO)eq] + complexes compete with the isomeric [L(R)2FeCleq(MAO)ax] + complexes. These isomeric complexes also contain one O→Fe and one Cl→Al contact and one vacancy. However, the vacancy is next to the chloride rather than the MAO species.
We computed both the [L(R)2FeClax(MAO)eq] + and [L(R)2FeCleq(MAO)ax] + complexes for the three MAO species 4, 7 and 5b for both the organic ligands L(Ph)2 and L(Me)2. Molecular models of optimized structures of these twelve complexes are shown in Figures  4-6 for 4, 7 and 5b, respectively. In each figure, the complexes with the L(Ph)2 ligand are shown in the top row and the complexes in the left column contain equatorial chloride. The energies of the [L(R)2FeCleq(MAO)ax] + complex relative to the [L(R)2FeClax(MAO)eq] + complex are included in Table 2, and the energetically preferred isomers are shown with green frames in the figures.    The MAO species 4 and 7 are symmetric and there is only one way for them to enter the coordination sphere of the [L(R)2FeClax] + complex. On the other hand, the ligand 5 is distinctly asymmetric. The structure of free 5 suggested that 5b and 5c as the best structural isomers with one or two methyl groups in a bridging position (Figure 1). However, in the complexes, the ligand has the topology of 5e (Scheme 1), that is, the exocyclic AlMe2 moiety is almost in the Al2O2 plane with the possibility of O→AlMe2 dative bonding. This topology occurs in all four complexes with 5 ( Figure 6). One of the in-ring Al atoms has

Kinetic vs. Thermodynamic Control of Complex Formation
We discussed the structure of the MAO complexed transition metal catalysts and our discussions emphasized the advantage of isomer L(R)2FeClax(5)eq for olefin polymerization. In Figures 3-5, the more stable isomers are highlighted by a green frame. As can be seen in Figure 4, there is a clear preference for the L(R)2FeClax(4)eq complexes with axial chloride irrespective of the nature of R. For the L(R)2FeCl(7) complexes in Figure 5, the complex with axial chloride is only preferred for the complex with the N,N-diphenyl ligand, L(Ph)2FeClax(4)eq. This result provides guidance for the improvement of the organic ligand. The structure suggests that a cyclic Al2O2 system will tend to align with one Narene group to benefit from parallel stacking. The interaction of the MAO species with the pre-catalyst will depend on the nature of the groups attached to the imine-Ns and the core of the ligand, which determines the distance between the imines. Inspection of Figure 6  The structure of L(Ph) 2 FeCl ax (5) eq exemplifies perfectly the mechanism postulated in the bottom row of Scheme 2, that is, the idea that the MAO species does not alkylate the transition metal and instead coordinates to the transition metal. The regiochemistry of this coordination places the AlMe 2 moiety right next to the coordination site of an H 2 C=CH 2 monomer and methyl transfer directly from the Al would result in the formation of an Al(Me)(CH 2 -CH 2 -Me) moiety. The catalysis would continue to cycle, converting Al(Me)([CH 2 -CH 2 ] n -Me) into Al(Me)([CH 2 -CH 2 ] n+1 -Me).

Kinetic vs. Thermodynamic Control of Complex Formation
We discussed the structure of the MAO complexed transition metal catalysts and our discussions emphasized the advantage of isomer L(R) 2 FeCl ax (5) eq for olefin polymerization. In Figures 3-5, the more stable isomers are highlighted by a green frame. As can be seen in Figure 4, there is a clear preference for the L(R) 2 FeCl ax (4) eq complexes with axial chloride irrespective of the nature of R. For the L(R) 2 FeCl(7) complexes in Figure 5, the complex with axial chloride is only preferred for the complex with the N,N-diphenyl ligand, L(Ph) 2 FeCl ax (4) eq . This result provides guidance for the improvement of the organic ligand. The structure suggests that a cyclic Al 2 O 2 system will tend to align with one N-arene group to benefit from parallel stacking. The interaction of the MAO species with the pre-catalyst will depend on the nature of the groups attached to the imine-Ns and the core of the ligand, which determines the distance between the imines. Inspection of Figure 6 and the isomer stabilities in Table 2 show a preference for the L(R) 2 FeCl eq (5) ax complexes, even for the one with N-phenyl groups.
It is important to distinguish between thermodynamic and kinetic control of isomer formation. In Table 2, we do not only provide thermodynamic isomer stabilities (Reactions (4)- (6) and (13)-(15)), but we also provide the reaction energies for the addition of the MAO species 4, 7, and 5 to the pre-catalyst leading to the isomeric complexes (Reactions (7)- (12) and (16)-(21)). The binding energies of the MAO species to iron are high and suggest irreversible MAO addition to the iron catalyst in fast reactions that will proceed without pronounced regioselectivity and not lead to thermodynamic equilibria. The positive consequence of this situation is that there will always be some active catalyst for the olefin polymerization. The art of catalyst design basically addresses the question as to how to provide a kinetic advantage for the formation of the desired L(R) 2 FeCl ax (MAO) eq isomer.

Absolute and Relative Oxophilicities and MAO Topology
The absolute oxophilicity Ω of the iron pre-catalyst [L(R) 2 FeCl] + is defined as the negative of the binding energies of the MAO species which are given in Table 2, and we discuss the ∆G data. We argued above that the oxygen in a cyclic aluminoxane would be a better donor in O → Fe dative bonding (Scheme 3). The results of the potential energy surface analysis allow to quantify this notion and we examined relative oxophilicities for MAO species in the equatorial position. The relative oxophilicity Ω r (8,10) of the iron pre-catalyst [L(Ph) 2 FeCl] + for cyclic 7 relative to acyclic 4 is given by the difference of the binding energies of reactions R8 and R10 in Table 2. In analogy, the relative oxophilicity Ω r (8,12) specifies the preferences for cyclic 5 relative to acyclic 4 based on reactions R8 and R12. The values Ω r (17,19) and Ω r (17,21) are defined in the same way for the N-methyl ligand. The computed oxophilicities Ω r are summarized in Table 3 and we discuss the ∆G values. The difference values ∆Ω r (10,12) and ∆Ω r (19,21) allow for a comparison of the two cyclic species 7 and 5, and ∆Ω r < 0 indicates that 5 is more strongly bonded ligand. The Ω r values all are greater than 10 kcal/mol, and in strong support of our hypothesis. Both 7 and 5 gave Ω r values of ∆G ≈ 16 kcal/mol for the phenyl-substituted ligands with a small preference ∆Ω r (10,12). The respective Ω r values for the methyl-substituted ligands are smaller and vary more; ∆G ≈ 10 kcal/mol for 7 and ∆G ≈ 15 kcal/mol for 5, respectively.

Mechanistic Hypothesis and Suggested Experimental Approach
The traditional mechanism for MAO-assisted transition metal olefin polymerization considers the MAO species solely as to purveyor of methyl anion to generate an initial Fe-Me species and the turnover of the catalytical cycle keeps adding olefin to generate Fealkyl species as shown in Equation (1). We are proposing an alternative mechanism which can be summarized by Equation (2). It is the essential feature of this mechanism that the polymerization occurs at aluminum. The transition metal is directly bonded to aluminum by way of an Fe-O-Al bridge and ensures that the iron-bonded olefin is positioned well for its aluminum-centered polymerization.
It is an essential feature of the proposed mechanism (Equation (2)) that the atoms of one Al-O bond of the MAO species coordinate both to the iron center and a remaining chloride ligand. This feature of the polymerization reaction described by Equation (2) is stressed in the bottom row of Scheme 2. The formation of O → Fe and Cl → Al dative bonds results in a four-membered ring, which should manifest itself in characteristic vibrational signatures. Therefore, we computed both the infrared and Raman intensities of the L(R) 2 FeCl ax (X) eq complexes with X = 4, 5b, and 7. In Table S3 we compiled the vibrational modes composed of any combination of Fe-O, Fe-Cl, Al-O, and Al-Cl stretches. The visualization of molecular vibrations is not trivial, but depictions of atom displacement vectors associated with normal modes are beneficial. In Figure S4, we show atom displacement vectors of the three most characteristic vibrational modes of complex L(R) 2 FeCl ax (5b) eq . We hope that these vibrational data encourage experimental studies.

Computational Methods
Computations were performed with Gaussian09 [36] on the Lewis3 cluster [37] of the high-performance computing center at the University of Missouri. Potential energy surface (PES) analyses of the MAO species were performed with second-order Møller-Plesset perturbation theory (MP2) [38] with all electrons included in the active space and in conjunction with the 6-31G* basis set [39,40]. Vibrational analyses were carried out analytically for each stationary structure at the level of optimization to compute vibrational frequencies and molecular thermal energies, enthalpies, and entropies.
Considering the size of the complexes, the complexes were studied with density functional theory, the popular B3LYP hybrid functional [41], and with the 6-31G* basis set. The six d-electrons of Fe 2+ are localized at the Fe center and their specific distribution has little effects on structures and complexation energies. We studied all systems as closed shell singlet using restricted Kohn-Sham orbitals.

Conclusions
We have presented the results of computational studies of the association of precatalysts [L(Ph) 2 FeCl] + with three MAO species. We have shown that the oxygen in cyclic MAO species is a much better ligand in iron complexes compared to the oxygen in acyclic iron species. Importantly, the cyclic MAO species does not have to be a cycloaluminoxane such as [(AlMe)O] 2 , 7 with all Al-O bonds being localized, but it can be an aluminoxane such as (MeAl)[O(AlMe 2 )] 2 , 5, which adopts a cyclic structure because of intramolecular O → Al dative bonding.
It is thought that the pre-catalyst L(R 2 )FeCl 2 goes through dissociation of one chloride and complexation of a MAO species to generate [L(R) 2 FeCl(MAO)] + complexes with O → Fe and Cl → Al dative bonding. To enable olefin polymerization, the remaining chloride must be in the pseudo-axial position relative to the Fe-pyridine plane. According to our calculations, the [L(R) 2 FeCl ax (MAO) eq ] + complex formations are highly exothermic and exergonic (∆G values of −19, −35, and −35 kcal/mol for MAO species 4, 7, and 5, respectively), and the regiochemistry of their formation most likely is under kinetic control.
Our study suggests that catalyst development should focus on the design of organic ligands (core and N-substituents) that will enhance the propensity for retaining one chloride in the axial position. Computed infrared and Raman spectra are reported and may guide experimental studies.