Using the spectroscopies based upon x-ray absorption, we have studied the structural and magnetic properties of Zn1−xCoxO films (x = 0.1 and 0.25) produced by reactive magnetron sputtering. These films show ferromagnetism with a Curie temperature TC above room temperature in bulk magnetization measurements. Our results show that the Co atoms are in a divalent state and in tetrahedral coordination, thus substituting Zn in the wurtzite-type structure of ZnO. However, x-ray magnetic circular dichroism at the Co L2,3 edges reveals that the Co 3d sublattice is paramagnetic at all temperatures down to 2 K, both at the surface and in the bulk of the films. The Co 3d magnetic moment at room temperature is considerably smaller than that inferred from bulk magnetisation measurements, suggesting that the Co 3d electrons are not directly at the origin of the observed ferromagnetism.PACS numbers: 75.50. Dd, 75.30.Hx, 61.10.Ht Among the most investigated topics in the field of spin electronics, dilute magnetic semiconductors (DMSs) occupy a prominent position, because they would allow one to exploit efficiently the spin and the charge of the electrons in the same device. In fact electronic devices have been working for decades omitting the spin of the electron. In 1990 Datta and Das proposed a new magnetoelectronic device (a field effect transistor), 1 whose practical realisation has been hindered by the weak spin injection efficiency from a ferromagnet to a semiconductor. A ferromagnetic semiconductor would constitute therefore an alternative route towards the efficient spin injection into normal semiconductors. Up to very recently, however, the main concern was related to the low Curie temperature of the known DMSs, which is well below room temperature and precludes therefore potential applications.2 A significant breakthrough was achieved recently, since room temperature ferromagnetism was predicted 3 and observed 4,5,6,7,8,9 for semiconductors such as GaN and ZnO doped with Co, Mn or other transition metals. However, many reports remain controversial and the nature of the magnetic coupling has not been revealed yet. In fact, the original prediction of high T C ferromagnetism in these systems by Dietl et al.3 lies on the assumption that they can be properly doped with p-type carriers, which would mediate the magnetic interactions. However, in order to account for the numerous experimental observations of ferromagnetism in n-type ZnO, alternative models have been proposed, which are mainly relying on the presence of defects (like for example vacancies or interstitials).10,11 In most of these models, the presence of a magnetic impurity such as Co or Mn is a necessary ingredient for the appearance of ferromagnetism, but other models show that this might not be needed and that ferromagnetism can appear also in undoped oxides.12 To date, however, there has not been any clear experimental proof of the validity of any of these models and of the role of the magnetic dopants.In order to tackle these problems, we have performed e...