Inspection of CPK models of ligand-metal complexes (in which the ligand was a derviative of TPA), suggested that derivatization of the a carbon would control the direction of the twist. However, CPK models suffer from inability to model metal ions with the appropriate complexation geometry. So computational modeling was used to refine our designs.
To estimate the energy differences between the conformers, suitable structures were optimized using the MNDO semiempirical method (as implemented in Spartan 3.0.1 ). The degree of dynamic control of the asymmetry in the complexes was modeled using the mixed mode dynamics MC/SD [2-4] algorithm as implemented in MacroModel 4.5 , with a modified AMBER* forcefield to take into account the Zn(II)-ligand interactions. Clustering analysis to obtain population distributions was performed using XCluster 1.0  on the output of sampled structures from the MC/SD runs.
First we looked at the interconversion dynamics in [Zn(TPA)Cl] obtaining a trajectory plot by monitoring the values of the Nam-Zn(II)-Npy-Cpy (Fig. 1) dihedrals, which shows a quasi-concerted "flipping" of the rings:
Analyzing one of the dihedral angles we can obtain an approximately even distribution of the lambda and d conformers (see Figure 3 for definitions of the twists):
When the same dynamic studies were done using TPA derivatives substituted in the a carbon, we found that one of the conformers was preferred, and that this preference was dictated by the absolute configuration of the chiral a carbon. Thus for a carbon center of "R" configuration, the metal-ligand complex prefers a conformer with lambda twist and with the substituent in the anti orientation (see diagrams below).
Performing similar simulations to obtain conformer distributions for Zn(II)-ligand complexes, in which the ligand presents a particular substitution in the a carbon, produces plots with an asymmetric distribution of populations, e.g. Figure 5 below.
The above distribution plot shows that only two conformers predominate. Using the relative distributions of the preferred conformers, an estimate of the energy differences can be made (Table 1), which compares well with values calculated using a semi-empirical (MNDO) method.
|Ligand (a)||XCluster (b)||Histogram