The incorporation of Si into vapor−liquid−solid GaAs nanowires often leads to p-type doping, whereas it is routinely used as an n-dopant of planar layers. This property limits the applications of GaAs nanowires in electronic and optoelectronic devices. The strong amphoteric behavior of Si in nanowires is not yet fully understood. Here, we present the first attempt to quantify this behavior as a function of the droplet composition and temperature. It is shown that the doping type critically depends on the As/Ga ratio in the droplet. In sharp contrast to vapor−solid growth, the droplet contains very few As atoms, which enhance their reverse transfer from solid to liquid. As a result, Si atoms preferentially replace As in GaAs, leading to p-type doping in nanowires. Hydride vapor phase epitaxy provides the highest As concentrations in the catalyst droplets during their vapor−liquid−solid growth, resulting in n-type dopant behavior of Si. We present experimental data on n-doped Sidoped GaAs nanowires grown by this method and explain the doping within our model. These results give a clear route for obtaining n-type or p-type Si doping in GaAs nanowires and may be extended to other III−V nanowires.

Despite the unavoidable presence of silicon atoms in the catalyst alloy droplets during the vapor−liquid−solid growth of III−V nanowires on silicon substrates, it remains unknown how the nucleation of nanowires is affected by these foreign atoms. In this work, we present the first attempt to quantify the nanowire nucleation rate versus the silicon concentration in the droplet. We calculate the chemical potential difference per Ga−As pair in the quaternary Au−Ga−As−Si liquid alloy droplet and in solid state, and compare it to the ternary Au−Ga−As droplet without silicon. This allows us to compute the nucleation rates of GaAs nanowires versus the silicon concentration under different conditions. We find that the presence of silicon in the droplet decreases the nucleation probability of GaAs nanowires for gallium-rich droplets (with the gallium contents c GA greater than 0.6) and increases it for goldrich droplets (c GA < 0.6). The model is used to explain our experimental data for hydride vapor phase epitaxy of gold-catalyzed GaAs nanowires, which easily nucleate on Si(111) covered with different SiO 2 layers but do not grow on the bare Si(111). In the latter case, more silicon is etched from the substrate and enters the gallium-rich droplets, which suppresses the nanowire nucleation. We discuss other relevant data, including the known difficulties in obtaining self-assisted GaAs NWs on silicon by chemical epitaxy techniques. These results may be useful for the fine-tuning of III−V nanowire properties and integrating them with silicon electronics.