Trap Parameter Extraction of Deep Defects in Semiconductors Using Noise Measurements.

In this dissertation, generation-recombination (G-R) noise is used as a tool for investigating the properties of deep defects. The first experiment we describe deals with low-field noise measurements, which were done on p ^+-p^--p ^+ silicon devices doped with a boron concentration of 10^{15} cm ^{-3} and an unknown gold concentration. The resistance and the G-R noise spectral density of the device were measured as a function of temperature and used to extract the trap energy, hole capture cross section, degeneracy factor, and trap density. The resistance and the noise data show the same activation energy of 0.336 eV, which stems from the generally accepted gold donor level, hereafter referred to as the gold first-donor level. The values of trap parameters so obtained are in good agreement with the published data in the literature, indicating that this method is accurate for investigating deep-defect properties. The second experiment described in this thesis was done on p^+-p^ --p^+ silicon devices doped with a boron concentration of 2.7 times 10^{15} cm^ {-3}. The resistance and low frequency voltage noise of the devices were measured as a function of temperature. The resistance data indicate that gold produces an activation energy of 0.267 eV, and the noise data indicate that gold produces an activation energy of 0.336 eV. This observation is explained in terms of double donor centers interacting with the valence band. Then the current-voltage (I-V) characteristics of the devices used in the first experiment were measured. They showed sublinear behavior at moderately high fields. With the use of low frequency voltage noise data measured as a function of electric field, we determined that the sublinear I-V behavior was caused by a decrease in the hole concentration with increasing electric field. This decrease is caused by a field-induced decrease in the hole emission coefficient, which is larger than the associated field-induced decrease in the hole capture coefficient. Device design schemes for investigating deep defects by noise measurements are presented. We analyzed the noise of several multi-level trap centers and derived the conditions for which the noise of a multi-level trap center reduces to the noise of a single-level trap center. We also devised a method by which G-R noise measurements can be used to discriminate between donor-like and acceptor-like trap behavior. In addition to the above described macroscopic G-R noise models, a microscopic model for carrier capture in deep traps is derived. Finally, theories for G-R noise generated in field-effect devices and polysilicon resistors are presented.