A robust alternative to η refinement for assessing the hand of chiral compounds

Assessment of the hand of chiral molecular compounds by least-squares refinement of a chirality parameter with all reflections is not statistically robust and can lead to incorrect answers due to a single, unnoticed, gross intensity-measurement error. An alternative is described in which the most easily observable Bijvoet differences are derived from the solved structure and remeasured at slow speed and possibly longer wavelength. Errors due to absorption, multiple reflections and anisotropic extinction can be cancelled out in the measurement for the sense of Bijvoet pairs with the proper experimental technique. An estimate of the probability of having derived the wrong hand, based exclusively on the number of remeasurements and the number of contradictions, and therefore insensitive to individual measurement errors, is developed. With this estimate, determination of the hand of compounds with oxygen as the heaviest scatterer is statistically unambiguous with Cu Kα intensity data, and feasible even with Mo Kα data. Consequently, the method can be applied to Cu Kα data for molecules with carbon as the heaviest scatterer. This approach has been automated as the utility BIVOET in the NRCVAX system of programs.