Department of Chemistry, Machine VP
Regioselectivity of Enolate Alkylations
The project involves ab initio molecular orbital calculations of the detailed geometric and energetic changes caused by systematic variations in transition state structures formed transiently during an important class of organic chemical reactions, the alkylation of ambident anions derived from ketones and aldehydes. The reactions investigated have two major alternate pathways and prediction/control of the selectivity is of prime significance in synthetic organic chemistry. The transient species (reaction transition states) cannot be observed directly, yet a knowledge of their structural features is needed to understand the factors controlling the reaction selectivity. Even with the full resources of a vectorised `supercomputer', the extent and large size of the chemical systems investigated limits the calculation level to low to moderate levels (within reasonable time allocation). The current results show that lower level calculations are ineffective here, moderate level ab initio molecular orbital calculations are required to give reasonably reliable reproduction of known qualitative features in these reaction systems. The mid-level calculations give quantitive information on the detailed geometric, electronic, and energetic features of these model reactions, which are matched to the chemical and electronic properties of the altered reactant structures and then compared to those changes predicted by current theories of chemical reactivity (based on extensive empirical observations). Such comparisons have provided good theoretical support for some aspects of current organic reactivity, whilst throwing doubt on other more recent proposals.
What are the basic questions addressed?
Alkylation of anions derived from ketones and aldehydes give mixtures of oxygen (O-) and carbon (C-) alkylated products, the ratios of which depend primarily on the precise nature of the substrate structures and solvent (see eqn below). A systematic investigation, using ab initio molecular orbital calculations, of the energetic and geometric variations in SN2 transition states for alkylation of acetaldehyde enolate anions by various substituted alkylating agents, R-CH2-X, was started to give insight on those factors controlling the regioselectivity (C- versus O-alkylation pattern) of such reactions.
CH2=CH-O- + R-CH2-X = CH2=CH-O-CH2-R + R-CH2-CH2-CHO + X-
O-alkylation T.S. (anion) C-alkylation T.S. (anion)
The starting point was to investigate the relative energetics of the three plausible, optimised approach geometries for O-alkylation and the (degenerate) approach geometry for C-alkylation (see figures above). This initial searching for global energy minima was by AM1 semi-empirical level calculations. Then for the preferred orientations, a more detailed study was made of the precise geometric and electronic variations induced in the optimised SN2 transition states by systematic modification of the electron donating and withdrawing properties of substituents (R) at the alkylation centre (R-CH2-X) and also by modifying the leaving group (X). The ordered range of structural variations were: alkylation centre substituent, R = CN, CHO, H, CH3, OH (increasing electron donation) and leaving group, X = F, OH, OCHO, OSO2CH3, OH2+ (increasing ease of cleavage). This involved reoptimisation of the preferred geometries for both O- and C-alkylation in each system with both low level RHF/3-21G and moderate level RHF/6-31+G* calculations using the extensive GAUSSIAN 92 package of programs.
What are the results to date and the future of the work?
Initial AM1 calculations revealed only small energy differences between syn- and anti- approach geometries for O-alkylation (perpendicular approach is, however, clearly disfavoured), with slightly larger energetic variations due to different torsional orientations about the O - CH2R - X axis. These AM1 calculations, however, incorrectly indicated preference of C- over O-alkylation in all systems, in contradiction to well established gas phase experimental results. Geometry reoptimisations of the transition state structures at the HF/3-21G level confirmed only small differences in the syn-/anti- O-alkylation energies, but now correctly indicated these as lower in energy than the C-alkylation (less preferred).
These 3-21G calculations failed to indicate any discernible trend in the C-/O- energetics ([[Delta]]E(C-O)) as the alkylation centre substituent (RCH2) was varied from electron withdrawing (R=CN, CHO; `tight' SN2 transition states) through mid (R=H) to electron donating (R=OH, CH3; `loose' SN2 transition states). Calculations at the 6-31+G* level showed a smaller general preference for O-alkylation, but now with clearly increasing O-preference as the alkyl substituent varied from electron withdrawing to donating (CN, CHO, H, CH3, OH). These energy trends were mirrored by related changes in the geometries and charge distributions within the transition states. Thus 3-21G results indicated only slight loosening of the reacting centre bonds, O -- C -- X, and irregular variation of charge distribution, whereas the 6-31+G* calculations showed distinct loosening of the bonds, with increasing positive charge at the alkylation centre and (about equal) negative charge on the two nucleophiles, as the electron donation increased (series CN, ..., OH) in the transition states. There is no correlation between the changes in activation energies and the overall reaction energetics (both level calculations), in keeping with the so-called `perpendicular' nature of these structural variations.
Calculations at both 3-21G and 6-31+G* levels for the ordered series of leaving groups, OH, F, OCHO, OSO2CH3 (R=H), confirm a shift from `later' to `earlier' transition states as leaving group ability improves, with increasing preference for O-alkylation in keeping with expected increasing dominance of charge control. These calculations also support the Bell-Evans-Polanyi analysis with activation energies proportional to the overall reaction energies. In contrast to recent proposals by Shaik and Pross, there is clear evidence that the transfer of charge from enolate to leaving group correlates directly with the late to early geometry shifts in these transition states with different leaving groups.
At this stage the major trends are obvious and there remains only some of the larger calculations to complete in the main series (some of the X=OSO2CH3 systems). Some less extensive parallel calculations, where both alkylation centre and leaving group are varied, are desirable to confirm that the observed leaving group trends hold also for the looser (R=OH) and tighter SN2 (R=CN) transition states; these remain to be attempted.
What computational techniques are used and why is a supercomputer required?
The computationally intense nature of the GAUSSIAN 92 program, requiring up to a gigabyte of temporary storage with vectorisation rates of 50-60% and computation times of up to 6 - 8 hr per iterative geometry optimisation for the larger systems (CH2CHO- + CH3OSO2CH3) at the moderate HF/6-31+G* level, means such calculations are only feasible with supercomputer facilities.