Profiles ofZn in n-type InP(100) wafers after ampoule diffusion were measured by secondaryion mass spectrometry, Auger electron spectrometry, differential Hall-effect measurements, capacitance measurements, and scanning electron microscopy. The results can be explained by an interstitial-substitutional mechanism, in which Zn diffuses as a singly ionized interstitial and is incorporated in the In sublattice as an electrically active substitutional acceptor or as an electrically inactive complex. At Zn concentrations lower than the background donor concentration the profile is cut off', as interstitial diffusion breaks down. The acceptor solubility increases with background donor concentration. Activation energies for diffusion and solubility were found to be 1.40 and 1.0 eV, respectively.

Closed-ampoule Zn diffusion in InP results in a net acceptor concentration that is much smaller than the Zn concentration. After subsequent annealing of InP in an atmosphere without Zn, the Zn and net acceptor concentrations have become almost identical, due to a decreased Zn concentration and an increased net acceptor concentration. The annealing treatment gives rise to a decreased contact resistivity. The diffusion depth has not changed. Annealing with a SiN cap on the InP surface does not have this effect on the concentrations. These annealing effects also take place in InGaAsP on InP layers. The results can be explained quantitatively by assuming that Zn is incorporated as both substitutional acceptors and interstitial donors and that only the interstitial Zn is driven out by the annealing, owing to its large diffusion coefficient. Profiles calculated with this interstitial-substitutional model can be fitted to experimental profiles assuming Zn to diffuse as singly ionized interstitial donors. This model can also describe earlier reported results on Zn diffusion in n-type InP for which a profile cutoff is found at a depth where the acceptor concentration equals the background donor concentration and in which the acceptor solubility is higher than in undoped InP.

No abstract