Currently, electrical interconnections based on metal lines represent the most important limitation on the performances of silicon-based microelectronic devices. The parasitic capacities generated at the metal/insulator/metal capacitors present in the complex multilevel metallization schemes actually used, the intrinsic resistivity of the metal lines, and the contact resistance at the metal/metal interfaces constitute the main contributions to the delay in the signal propagation. Recently, a reduction of the delay times was achieved by replacing the traditional metallization schemes based on Al and SiO 2 with new materials, such as copper-based alloys and low-dielectricconstant insulating layers, but as soon as the minimum feature size of the devices will be further reduced, the delay resulting from metal interconnections will again represent an unacceptable bottleneck for device performances.[1] A definitive solution to this problem could be the use of optical interconnections for the transfer of information inside a chip or for chip-to-chip communications. To develop this strategy, siliconcompatible materials and devices able to generate, guide, amplify, switch, modulate, and detect light are needed. Recent major breakthroughs in this field have been the observation of optical gain in Si nanocrystals, [2] the development of a Si Raman laser, [3] the realization of a high-speed Si electro-optic modulator, [4] and the observation of electroluminescence from ultrapure Si diodes [5] and Si nanocrystal field-effect transistors. [6] A primary requirement for the materials proposed for the above applications is compatibility with current Si technology. However, because Si is intrinsically unable to efficiently emit light, owing to its indirect bandgap, it is evident that the main limitation to the approach described above is the lack of an efficient silicon-based light source. Among the efforts of the scientific community to efficiently produce photons from silicon, the introduction of light-emitting impurities, such as erbium ions, has a leading role. A relevant advantage of this approach is that standard silicon technology can be used to introduce erbium as a dopant and to process the material. Furthermore, Er ions emit at 1.54 lm, which is a strategic wavelength for telecommunication because it corresponds to a minimum in the loss spectrum of the silica optical fibers. Incorporation of Er in crystalline silicon (c-Si) emerged as the first promising method to turn silicon into a luminescent material, [7] but doping concentration was limited (ca. 1 × 10 18 cm -3) by the low solid solubility of Er. A co-implantation of Er and O [8,9] allowed to limit Er segregation and precipitation, owing to the formation of Er-O complexes. However, at room temperature a relatively low luminescence efficiency was obtained as a result of the strong nonradiative processes competing with the radiative Er de-excitation in c-Si. More recently, it was shown that by using a SiO 2 matrix containing Er-doped Si nanoclusters, an intense room-t...