Control Experiments

To learn which architecture and functional groups in the polyfunctional systems are essential for their performance, a number of control catalyst systems CC was investigated. For high endo selectivity the imidazolium moiety in C4 is advantageous compared to a 1,2,3-triazolium moiety in the control catalyst CC1 (Table 6, entry 1). Azolium rings are beneficial for high enantioselectivity, as judged from CC2 missing this feature, which gives nearly racemic product (entry 2). Furthermore, the axially chiral betaine element provides activity advantages compared to the simpler imidazolium phenolate in CC3. While good results were attained with 5 mol% of CC3 (entry 3), no product was detected with 0.1 mol% (entry 4). This is in strong contrast to the use of C4 (see Table 1, entries 6–7). Like expected, imidazolium catalyst CC4 lacking the basic aryloxide showed a further decrease in activity (Table 6, entries 5 and 6).^[69-75]

Table 6. Study of control catalysts CC for the endo-selective Co(II) catalysis.

	X / (mol%)	CC	additive	conv.a) / (%)	yielda) / (%)	endo: exoa)	eeb) / (%)
1	5	CC1	−	99	89	84:16	92
2	5	CC2	−	87	77	89:11	14
3	5	CC3	−	90	90	97:3	80
4	0.1	CC3	−	<2	<2	−	−
5	5	CC4	−	63	57	95:5	84
6	0.1	CC4	−	<2	<2	−	−
7	5	CC5	−	99	54	85:15	1
8	5	CC5	Cs2CO3	>99	11	76:24	74
9	5	CC5 + CC6	−	>99	71	67:33	82

• ^a) Determined by ¹H-NMR in the presence of an internal standard.
• ^b) Enantiomeric excess of the endo isomer determined by HPLC.

The simple complex CC5 also displayed catalytic activity at high loadings, but gave nearly racemic product (entry 7). In the presence of Cs₂CO₃ as an external base the formation of large side product quantities was found, resulting in a poor yield, while enantioselectivity was largely improved (entry 8). Combining CC5 with betaine fragment CC6 in a binary catalyst system resulted in decreased endo selectivity and also considerable side product formation. These results indicate that an intramolecular interplay of the various functional groups is essential for the high performance of C4.

A similar program was performed for the exo-selective reaction (Table 7). Surprisingly, catalyst CC8 featuring an N-mesyl ligand donor provided mainly endo product (entry 1), revealing the impact of the N-triflyl donor in C7. We also investigated phenolate CC9 (entry 2), lacking the element of axial chirality. Compared to the standard system C7, it is less active and does not favor the formation of the exo isomer. The binaphthyl architecture thus seems to help to accomplish a suitable structure to overwrite the inherent endo preference. Also, the aryloxide metal distance seems to be crucial, because for biphenolate CC10, which is structurally more similar to C7, we noticed exo-selectivity (entry 3), yet still much lower than for C7.

Table 7. Study of control catalysts CC for the exo-selective Ni(II) catalysis.

	CC	Additive	Conv.a) / (%)	Yielda) / (%)	exo: endoa)	ee exob) / (%)
1	CC8	−	97	97	31:69	82
2	CC9	−	86	76	46:54	83
3	CC10	−	99	98	77:23	78
4	CC11	−	0	0	−	−
5	CC12	−	0	0	−	−
6	CC12	Cs2CO3	0	0	−	−
7	CC12 + CC13	−	36	11	22:78	11

• ^a) Determined by ¹H-NMR in the presence of an internal standard.
• ^b) Enantiomeric excesses determined by HPLC.

Interestingly, the catalysts CC11 and CC12, lacking an aryloxide, showed no catalytic activity (entries 4–6). In contrast, the binary catalyst system of CC12 and CC13 gave a small amount of product (entry 7) with poor stereoselectivity.

Table 7 thus suggests that for the exo-selective synthesis, the sophisticated intramolecular interplay between the Ni center and the triazolium aryloxide with a precisely elaborated chiral environment is crucial for high performance.

Spectroscopic Studies

The properties of C4 and C7 were studied by paramagnetic Evans-NMR, SQUID (superconducting quantum interference device) magnetometry, and EPR to gain insight into the electronic structure of the catalyst in a solid state as well as in solution. SQUID measurements of the molar magnetic susceptibility χ_m of both C4 and its precatalyst (featuring the neutral naphthol unit) were performed on powder samples. Both samples show a constant χ_mT of around 2.3 cm³ mol⁻¹ K at high temperatures and a strong decrease below 100 K (Figures S25 and S26).

Simulations of the susceptibility-temperature product χ_mT for both compounds using the spin Hamiltonian formalism reveals a high-spin state (S = 3/2) with zero-field splitting (ZFS) parameters of D = +38 cm⁻¹ and E/D = 0.30(3).^[76] The slight difference in χ_mT at room temperature between C4 and its precatalyst is due to different g_iso in C4 (g_iso = 2.22(1)) and its precatalyst (g_iso = 2.28(1)). Furthermore, Evans method NMR measurements on C4 and the precatalyst were performed in THF.^{[77, 78]} The paramagnetic shift confirms the high-spin state (S = 3/2), but g_iso is slightly lower. EPR experiments using frozen solutions confirm these trends and are shown in the Supporting Information (Figure S28).

SQUID measurements on C7 and its precatalyst (again featuring the neutral naphthol unit) show low χ_mT of 0.07 cm³ mol⁻¹ K across the whole temperature range with a slight increase above 270 K to 0.2 cm³ mol⁻¹ K at 300 K for C7. For a Ni d⁸-system in a high spin configuration (S = 1) χ_mT = 1 cm³ mol⁻¹ K (assuming g_iso = 2) and for low spin (S = 0) χ_mT = 0 cm³ mol⁻¹ K would be expected. These results indicate that C7 does not exist in a single magnetic conformation and a small part of the compound is paramagnetic. CASSCF calculated magnetic parameters based on PBEh-3c/def2-mSVP geometries support these results. The quantum chemical calculations strongly suggest this species to be the catalytically relevant one.

Concentration-dependent UV–Vis studies of catalysts C4 and C7 show a linear absorption to concentration relationship thus pointing to monomeric catalyst species (see Supporting Information, Figures S32 and S34). This assumption is further supported by DFT computations, according to which the dimerization of the catalysts is either endergonic (C7) or not possible due to steric hindrance (C4). Moreover, the linear dependencies of catalyst ee and product ee values for both systems are consistent with monomeric catalyst species (see Supporting Information, Figures S29 and S30).

Computational and Kinetic Studies

Based on the control experiments and the spectroscopic results in combination with previous experimental and computational studies on related Cu(II) catalysts reported by our groups,^{[54, 65, 69]} we used the simplified general catalytic cycle shown in Scheme 2 as working hypothesis for both the endo- and exo-selective methods.

According to this mechanistic model, iminoester 1 initially coordinates to catalyst C, thus forming I. This is followed by a proton shift from the iminoester to the catalyst's naphthoxide moiety, giving azomethine ylide complex II. As a next step, maleimide 2 binds to II via a hydrogen bond to the generated naphthol-OH, forming III. This allows for a quasi-intramolecular (concerted or stepwise) cycloaddition. After the formation of the cycloadduct complex IV, a conformational change of the catalyst to IV* is necessary to enable protonation of the cycloadduct-metal bond by the naphthol moiety to form the catalyst/product adduct V. Subsequent product dissociation closes the catalytic cycle.

Following that mechanistic model, comprehensive computational studies were conducted on the (B3LYP-D3(BJ)/def2-TZVP/ COSMO(THF)) level of theory on PBEh-3c/def2-mSVP geometries (for details see Supporting Information)^[79-92] to give—in combination with kinetic and spectroscopic investigations—detailed mechanistic pictures for both the endo- and exo-selective methods. In fact, according to our calculations, both reactions follow the general mechanism outlined above, but differ in various aspects that are important for the catalytic outcome.

Endo-selective Approach With Catalyst C4

As the magnetic measurements in solid state and solution (vide supra) show C4 to be a high-spin Co(II) complex, we evaluated the reaction pathway by DFT calculations on the respective quartet potential energy surface (PES). The predicted Gibbs free energy profile for the endo-configured main stereoisomer with a thermodynamic driving force of Δ_rG° = −50.0 kJ mol⁻¹ is shown in Figure 1. Our DFT calculations suggest that the initial iminoester coordination to C4 proceeds in a monodentate way via the ester moiety and is slightly endergonic. This step is followed by a slow and endergonic intramolecular deprotonation of the coordinated iminoester by the naphthoxide moiety to form the ylide-like II^Co, in which the Co(II) center adopts a trigonal-bipyramidal geometry with elongated axial bonds. The formally neutral donor atoms coordinate axially, while the negatively charged ones form the equatorial plane. The subsequent imide association via H-bond to the naphthol unit in III^Co is again endergonic.

In contrast, the actual cycloaddition step forming two new C,C-bonds is calculated to be strongly exergonic. Interestingly, it proceeds in a concerted manner for the endo pathway, while typically stepwise mechanisms have been suggested and calculated for other catalysts.^[93-95] In transition state TS(III^Co→IV^Co) (Figure 2, left), in addition to the hydrogen bond activation of the maleimide, the C–H acidic imidazolium forms a double hydrogen bond with the 1,3-dipole and the phenoxy moiety of the ligand. The latter can serve as a structural anchor, rigidifying and stabilizing the assembly. As a result of the simultaneous activation of both substrates adopting a suitable spatial orientation for a quasi-intramolecular step, a remarkably low barrier of 15.5 kJ mol⁻¹ (enthalpic part: 9.8 kJ mol⁻¹, entropic part: 5.7 kJ mol⁻¹) is found for this event thus demonstrating the power of the polyfunctional catalysis concept in this key step. The chiral metal sphere and the axially-chiral binaphthyl moiety form a well-defined chiral binding pocket that stabilizes the spatial orientation required for the concerted mechanisms and effectively reduces the activation entropy. Moreover, by the precisely defined alignment of both substrates within the catalyst's chiral environment, high enantioselectivity is achieved.

For the competing exo-pathway with C4, a stepwise cycloaddition mechanism was found requiring to surpass a significantly higher barrier (see Supporting Information, Figure S37).

The reaction steps following the C,C-bond formations are particularly characteristic for the catalytic mechanism with C4. As shown in Figure 2, the naphthol unit in IV^Co forms a hydrogen bond to the phenoxy unit of the ligand sphere. To enable the highly exergonic protonation of the cycloadduct, this bond must be broken, allowing for an isoenergetic conformational change of the imidazolium/naphthol unit. This change of IV^Co to its conformer IV^Co* restores the trigonal-bipyramidal coordination of the Co(II) center and permits the formation of a hydrogen bond between the naphthol unit and the pyrrolidinate nitrogen of the initial cycloadduct, thus facilitating a quasi-barrierless proton transfer to form V^Co. This Brønsted acid/base reaction step closely resembles enzymatically catalyzed proton transfer reactions.^{[96, 97]}

The product release from V^Co closes the catalytic cycle by regenerating C4. V^Co is energetically favored compared to the iminoester/ catalyst complex I^Co by −7.5 kJ mol⁻¹.

According to the energetic span model,^[98] the catalytic activity of a system, measured as TOF, is usually determined by only two states: the turnover-determining intermediate (TDI) and the turnover-determining transition state (TDTS). Whenever the TDI appears before the TDTS in the catalysis cycle, the energetic span is determined by the Gibbs free energy difference between those two states. Otherwise, the energetic span is the sum of that Gibbs free energy difference and the thermodynamic driving force. The energetic span model enables the direct comparison between Gibbs free energy profiles obtained by quantum chemistry and the experimental kinetic investigations. Following the DFT evaluation of our mechanistic model (Figure 1), the energetic span (+80.8 kJ mol⁻¹) of the investigated, endo-selective catalytic reaction is given by V^Co as TDI and the TS of the deprotonation step (I^Co → II^Co,a) as TDTS. Interestingly, the calculated entropic part of the activation barrier is almost zero (0.04 kJ mol⁻¹), which is another similarity to enzymatic catalysis and might be explained by the rigidified catalyst structure, resulting from structural anchoring by H-bond interactions.^{[99, 100]}

Using a microkinetic model, barriers can be assigned to individual chemical steps from experimentally measured concentrations. For this purpose, the ordinary differential equation system resulting from the mechanistic hypothesis is solved, and the barriers of the individual reaction steps are optimized in order to best replicate the experimentally measured concentration profiles. In order to better understand the endo-selective DCA, we decided to fit such a microkinetic model using methods (see Supporting Information) similar to those previously reported for an asymmetric hydroboration by our groups.^[102] According to our microkinetic model, the barrier of the iminoester deprotonation has the major influence on the kinetics, which is consistent with the identification of the corresponding TS as TDTS. Furthermore, the apparent barrier for the catalytic reaction 1a + 2A → 3aA from our microkinetic model is found to be +75.6 kJ mol⁻¹, which again agrees well with the calculated energetic span of + 80.8 kJ mol⁻¹ that can be deducted from the Gibbs free energy profile (Figure 1) and the apparent barrier from VTNA analysis (vide supra).

Exo-selective Approach With Catalyst C7

Since the magnetic measurements of the activated catalyst C7 in solid state and solution (vide supra) combined with ¹H-NMR spectra suggest the presence of a low-spin Ni(II) complex as largely dominating species which, however, also contains a paramagnetic high-spin Ni(II) fraction, it was unclear which species is catalytically active. Hence, we studied central reaction intermediates on both the singlet and triplet PES (see Supporting Information). Our calculations indicate that the azomethine ylide coordination as a bidentate ligand to the catalyst is strongly endergonic on the singlet PES, while that step is only slightly uphill on the triplet PES. Moreover, the calculated iminoester deprotonation barrier in the singlet case is +111.8 kJ mol⁻¹, which contrasts to the found triplet barrier of + 58.9 kJ mol⁻¹ and is too high for the experimental reaction conditions and outcome. Thus, it is very likely that the minor high-spin fraction of the Ni(II) complex is the catalytically active species.

The predicted Gibbs free energy reaction profiles for the main exo- and minor endo-product 3bA with thermodynamic driving forces of Δ_rG° = −62.1 and −54.0 kJ mol⁻¹, respectively, are shown in Figure 3. In case of C7, the iminoester coordination to the catalyst is exergonic. The subsequent proton shift to the aryloxide moiety of the catalyst is endergonic, but has a comparatively low barrier. In contrast to the analogous Co(II)-species derived from C4, the Ni(II) center in II^Ni (and III^Ni as well as IV^Ni) adopts a square pyramidal geometry by coordination of both the ligand and the ylide.

Again, imide binding to II^Ni is endergonic. From imide adduct III^Ni C,C-bond formations occur in either concerted or stepwise manner, depending on the stereoisomer formed. To further investigate this result, 2D PES scans were performed for both possible exo enantiomers, varying both C─C bond lengths systematically (Supporting Information). It was found that only in the case where stabilization of a potential Michael intermediate by hydrogen bonding between it and the naphthol-OH group is possible, a stepwise C,C-bond formation can be achieved. In all other cases, an asynchronous concerted mechanism is observed. The described stabilization is only possible for one of the two exo-product enantiomers, which is the main product using C7. In that case, in the TSs of both C,C-bond formation steps (Figure 4a,b), the C–H acidic triazolium moiety forms a hydrogen bond with the phenoxy moiety of the ligand in addition to the hydrogen bond activation of the maleimide. Similarly to the concerted C,C-bond formation with C4 (Figure 2a), this serves as a structural anchor to stabilize the spatial position of the azolium / naphthol moiety.

After C,C-bond formation, also IV^Ni (Figure 4c) needs to undergo a conformational change to form IV^Ni*, which is stabilized by a hydrogen bond between the naphthol OH group and the sulfonyl motif before protonation of the cycloadduct can take place. This change also results in a strong twisting of the azolium/naphthol moiety. In IV^Ni*, it is possible to bind the naphthol OH to the Ni(II) center in an octahedral coordination geometry, allowing protonation of the exo-configured cycloadduct (Figures 3 and 4d). Protonation of the minor endo-isomer follows the same mechanism. Regeneration C7 by product release closes the catalytic cycle. That dissociation is calculated to be endergonic for both exo and endo products.

Due to the low barrier of the C,C-bond formations, they are reversible for both the exo and the endo reaction paths. In addition, the barrier of the cycloadduct protonation from IV^Ni* is higher than the barrier of the back-reaction to III^Ni. This results in a Curtin–Hammett situation, which means that IV^Ni* (exo) and IV^Ni*(endo) are in equilibrium with each other. As the barrier to cycloadduct protonation is lower for the exo-product than for the endo-product, the exo-product is preferentially formed.

Based on our computational model, the exo:endo-ratio is determined by the Gibbs free energy difference between the protonation TSs, i.e., TS IV^Ni* → V^Ni. This value is found to be sensitive to the DFT functional used (see Supporting Information). B3LYP-D3(BJ)/def2-TZVP/COSMO(THF) yields a value of about 68:32. Using COSMO-RS^[103-105] instead of COSMO^[106-108] as an implicit solvent mode predicts an exo:endo-ratio of 88:12 (experimental: 92:8).