A series of 0.5-4.3 wt% Rh/Al2O3 catalysts were prepared by flame synthesis. STEM indicated relatively narrow particle size distributions for all catalysts and the mean particle size increased almost linearly with the Rh content in the range 0.96-1.65 nm. A DRIFTS study of CO adsorption on as prepared Rh/Al2O3 and after heat treatment in hydrogen at 400 degrees C revealed that there was no Rh oxide present at the catalyst surface after the high temperature reduction, which procedure is commonly used prior to enantioselective hydrogenation. In the hydrogenation of ethyl pyruvate and ethyl 3-methyl-2-oxobutyrate the cinchona-modified 4.3 wt% Rh/Al2O3 gave considerably higher ee than those achieved with the best known Rh catalyst. A decrease of the metal loading and thus the mean Rh particle size, led to a loss of ee to (R)-lactate by a factor of up to seven at I bar and up to two at 10-100 bar. Our interpretation is that the performance of Rh/Al2O3 is strongly distorted at atmospheric pressure by catalyst deactivation due to the Al2O3-catalyzed aldol condensation of the substrate. During the fast reactions at 100 bar the contribution of strongly adsorbed impurities is small and the variation of ee is mainly due to an intrinsic particle size effect. The structure sensitivity observed under optimal conditions, at high surface hydrogen concentration, is mainly due to steric effects: a small, ca. 1 nm Rh particle cannot accommodate the enantiodifferentiating diastereomeric substrate-modifier complex and the hydrogenation on its surface leads to racemic product. A practical conclusion is that there is no advantage of using small nanoparticles and low metal loading in the enantioselective hydrogenation of alpha-ketoesters. (C) 2008 Elsevier Inc. All rights reserved.