We report an interplay between magnetism and charge transport in the ferromagnetic semiconductor GdN, pointing to the formation of magnetic polarons centred on nitrogen vacancies. The scenario goes some way to resolving a long-standing disagreement between the measured and predicted Curie temperature in GdN. It further constitutes an extension of concepts that relate closely to the behaviour of ferromagnetic semiconductors generally, and EuO in particular.PACS numbers: 75.50. Pp, Intrinsic ferromagnetic semiconductors, which offer the freedom to dope into specific conducting channels without destroying their magnetic behaviour, are of obvious technological interest. From a fundamental point of view they are again interesting; it is a substantial challenge to understand the interplay between their magnetic and transport properties. Among the earliest discovered, and even now the most studied, is EuO, which shows a magnetoresistance as large as thirteen orders of magnitude across the metal-insulator transition at the Curie temperature (T C ) [1][2][3], the largest in any compound. The strong transport-magnetism interplay is further emphasised by the enhanced T C seen in electron-doped samples [4,5]. Debate continues about whether the ferromagnetic transition is homogeneous [2,6,7] or whether it involves magnetic polarons nucleating around magnetic impurities [3,[8][9][10][11][12].GdN, the prototypical rare-earth nitride compound, provides a rich set of comparisons with EuO. Divalent Eu and trivalent Gd share the same half-filled 4f shell, with S = 7/2, L = 0, and a net moment of 7 µ B . They share also the same rock salt structure, their reported T C are both near 70 K [13][14][15][16], and GdN also shows a strong magnetoresistance at T C [13,17]. Theoretical treatments of reproduce the measured electronic band features well, showing agreement as regards its semiconductor nature [13,14], the direct band gap [22][23][24], and features in the conduction-band density of states [13,25]. Even more important for device development is that GdN, unlike EuO, has a dispersive valence band of delocalised nitrogen states, which raises the possibility that it can support both n− and p−type conduction. Its potential in realistic spintronics has already been demonstrated by its use in a spin filter [26].In contrast the ferromagnetic exchange mechanism is still poorly understood. The atomic-like nature of the 4f electrons necessitates indirect exchange, but those levels lie too far below the Fermi level to call on the third-order perturbation theory applied to EuO [4,27,28]. The validity of a proposed exchange channel involving the excited Gd 4f8 level is uncertain [29]; in particular note these levels are similarly far from Fermi level [25]. Small energy differences are found among various spin orderings determined within the LDSA+U treatment that has successfully reproduced the band structure, but they lead to an estimated T C below 25 K [30], far below the experimental values. There is even a 1980 report [31], raised again re...

We report on the growth and properties of GaN films grown on Si(111) substrates by molecular beam epitaxy using ammonia. The properties of the layers show that our growth procedure is very efficient in order to overcome the difficulties encountered during the growth of nitrides on silicon substrates: first, no nitridation of the silicon substrate is observed at the interface between the AlN buffer layer and the silicon surface; second, there is no Si autodoping coming from the substrate and resistive undoped GaN layers are obtained; and, also, strain balance engineering allows one to grow thick GaN epilayers (up to 3 mm) without formation of cracks. The optical, structural and electrical properties of these films are studied. In order to evaluate the potentialities of III-V nitrides grown on silicon substrates, we have grown heterostructures to realize light emitting diodes (LEDs), photodetectors and high electron mobility transistors (HEMTs).Introduction GaN-based devices have demonstrated a large potential for applications in electronics and optoelectronics [1]. For all these applications, the III-V nitride layers are usually grown on sapphire or silicon carbide substrates. However the development of some of these applications on silicon substrates has obvious technological advantages. This is the reason why significant research efforts have been made recently in this direction as shown by the wide variety of techniques which have been assessed. Indeed, techniques including conventional growth by MBE [2][3][4] or MOCVD [5-9] using various buffer layers, combination of MBE and MOCVD [10], growth on patterned substrates [11,12], selective area epitaxy [13,14], conventional ELOG [15,16], mask-less ELOG [17], pendeo-epitaxy [18], wafer bonding and film lift-off [19,20] etc. have all been already used in order to find a successful process for the growth and/or the realization of GaN-based devices on silicon substrates.With a such successful process it is expected that high-quality single crystalline GaN layers on silicon will be obtained despite the large lattice mismatch between GaN and Si. However, to achieve this goal, some other difficulties encountered during the growth of nitride on silicon substrates have also to be solved. Up to now as demonstrated for the first time by Watanabe and co-workers [21], the use of high temperature AlN buffer layers seems to be the best candidate for the heteroepitaxial growth of GaN on silicon substrates. The formation of the AlN/Si interface is already a critical issue because aluminum can diffuse in the silicon substrate and form a p-type inversion layer [22,23] in the substrate. Moreover the presence of reactive nitrogen species at the

No abstract