Field-sensitive photoconductive sampling and probes.

Abstract

This dissertation has investigated thoroughly the sensitivity of photoconductive sampling, including the limitations of picosecond photoconductive material, Schottky metal/semiconductor contacts, miniaturized MSM interdigitated switches, and the parasitic effects of external readout circuits. The dominant noise source is found to be the photovoltaic noise induced by the 1/f gating laser amplitude fluctuation. By using an integrated JFET source follower with an extremely high input impedance and low parasitic capacitance, the sampling method is converted to a voltage measurement. The sensitivity of photoconductive sampling with a conductive contact is improved to a few tens of nV/Hz$\sp{1/2}$, a factor of 100 over the conventional method. The invasiveness to the device under-test and the required laser power are both greatly reduced. In addition, this voltage measurement method can avoid the intrinsic high electric field effect of photoconductive material. Overall, this sensitivity has already reached the limitation of the most sensitive readout electronics available. The dynamics of photoconductive sampling circuits using an equivalent-time method are modeled and verified using a new charge injection-dissipation model, which including the intrinsic photoconductive charge injection rate and the leakage current due to the external readout circuit. By measuring the modulation bandwidth of sampling switches at various gating laser power for various sizes of MSM interdigitated switches, this model is verified to be consistent with the experimental results. This high impedance source follower has also been incorporated with low-temperature-grown MBE GaAs based, epitaxial lift-off layer probes and measures picosecond waveforms through insulating layers. A sensitivity of 2.5 $\mu$V/Hz$\sp{1/2}$ and a spatial resolution of 5 $\mu$m, consistent with the charge injection-dissipation model is verified. A conductive contact is no longer required and the invasiveness in term of current drainage from device-under-test is eliminated. The dominant noise source is reduced to a few nV/Hz$\sp{1/2}$ of the readout circuit, instead of the 1/f photovoltaic noise since there has no complete current loop. The possibility of a photoconductive scanning probe with a submicron spatial resolution and a voltage sensitivity of a few $\mu$V/Hz$\sp{1/2}$ is discussed by the charge injection-dissipation model.