Icosahedral symmetry versus local icosahedral environments in Al–Mn alloys from NMR

Abstract

In a recent report¹, Pauling suggested that the icosahedral diffraction found in the aluminium–manganese system is due to a cubic crystalline phase with a large unit cell of more than 1,000 atoms, multiply twinned to mimic icosahedral symmetry, instead of a crystalline or quasicrystalline phase with fivefold symmetry. Pauling claims that most of the manganese atoms are surrounded by twelve aluminium atoms arrayed in a nearly perfect icosahedron, a cluster that occurs in the Al₁₂W structure. To account for the high Mn content in the icosahedral phase, he assumed that each aluminium atom was shared with a neighbouring icosahedron. Here, we report a comparison of the nuclear magnetic resonance (NMR) spectra obtained from aluminium–manganese alloys in the crystalline 'G' phase (in which each of the Mn atoms is surrounded by an icosahedron of Al atoms, exactly as in Pauling's proposal structure) with the spectra of the icosahedral phase. We find that the NMR spectrum in the 'G' phase is entirely different from the spectrum in the icosahedral phase. Twinning should not affect the NMR spectra except for a small contribution from atoms on or near the composition planes, so the Al–Mn icosahedral phase is not Pauling's twinned cubic structure.