Rare earth doped semiconductors have received much attention in the last decades because of potential applications in photonics and illumination. Because silicon is the most widely employed material in the semiconductor industry, much of this effort is devoted to obtain efficient light emitters involving silicon based alloys. The excitation cross section of rare earth ions is orders of magnitude higher in silicon-rich amorphous alloys than in ordinary oxides. Energy transfer is often very efficient in the presence of silicon nanocrystals or nanoclusters. This has led to energy transfer models involving quantum confinement and the electronic structure of silicon nanocrystals. Although these models are elegant and persuasive, they demand a highly homogeneous nanoparticle size distribution, absence of states associated with surface or strain in very small particles and the excitation of many rare earth ions by a single nanoparticle, all very unlikely in real systems.

We have studied the excitation of Tb³⁺ doped a-Si₃N₄:H. Our results show a correlation between the density of silicon dangling bonds and the luminescence efficiency under matrix excitation. These findings lead to a reinterpretation of published results in samples containing silicon nanoparticles. When nanoparticles are formed the density of silicon dangling bonds in fact increase as silicon-silicon, silicon-oxygen, silicon-nitrogen or silicon-hydrogen bonds are broken. Thus, the main effect of the formation of nanoparticles, crystalline or not, to the energy transfer to rare earth ions is through the increase of the density of silicon dangling bonds and consequently of midgap states. A very short range Auger process involving carrier recombination at silicon dangling bonds is most probably the main mechanism for excitation from a silicon rich host to rare earth ions. To optimize the luminescence efficiency, the density of dangling bonds must match the density of active rare earth ions.