The possibility that the apparent room-temperature ferromagnetism, often measured in Co-doped ZnO, is due to uncompensated spins at the surface of wurtzite CoO nanoclusters is investigated by means of a combination of density-functional theory and Monte Carlo simulations. We find that the critical temperature extracted from the specific heat systematically drops as the cluster size is reduced, regardless of the particular cluster shape. Furthermore the presence of defects, in the form of missing magnetic sites, further reduces T C . This suggests that even a spinodal decomposed phase is unlikely to sustain room-temperature ferromagnetism in ZnO:Co. DOI: 10.1103/PhysRevB.81.054441 PACS number͑s͒: 75.50.Pp, 75.50.Tt, 75.75.Ϫc In recent years the search for ferromagnetism in insulating oxides doped with small quantities of transition metals has become a topic generating much debate in the literature. Taking ZnO:Co as the prototypical example for this class of materials, many experimental groups have reported roomtemperature ferromagnetism 1-3 whereas several other have failed to find any such evidence. 4,5 Notably, growth conditions, sample morphology, and spatial Co distribution play a crucial role in determining the magnetic properties. In particular, there is now an emerging view that samples with high structural quality and uniform Co distribution do not result in long-range ferromagnetism at high temperature and that the magnetism may be related to structural 6 or point defects. 7 Given the unsettled experimental landscape it should not be a surprise that a number of interesting and competing theoretical models have been proposed. In general explanations involving standard mechanisms for the magnetic interaction appear problematic. Schemes leading to short-range magnetic coupling such as superexchange need a Co concentration, ͓Co͔, exceeding the percolation threshold. This is around 20% for an interaction extending to the nearestneighbor sites of an fcc lattice and it is about a factor of five larger than the typical experimental concentrations. Similarly carrier-mediated mechanisms are not sustained by experimental evidence, which show both paramagnetism in presence of abundant-free carriers 6 and ferromagnetism deep in the insulating region of the phase diagram.2 More generally carrier concentration and mobility have an apparent little correlation to the magnetic properties.8 Furthermore carriermediated mechanisms are difficult to validate on a solid theoretical ground by using first-principles calculations since the empty Co d levels are usually erroneously predicted too shallow at the edge of the ZnO conduction band.
9Thus one has to look for more complex mechanisms for the magnetic interaction. Among these, the donor impurityband-exchange model ͑DIBE͒ ͑Ref. 10͒ has enjoyed considerable popularity in the experimental community. According to the DIBE the magnetic interaction among Co 2+ ions is mediated by donors, whose charge density is localized over large hydrogenic orbitals so that the relevant per...