Direct assessment of recombination noise in semiconductors using electron beam-induced conductivity.

Abstract

The level of internal noise of the transistors, diodes, and other semiconductor components limits the successful design of any low noise electronic system. All types of noise, namely, Johnson, 1/f, and so forth, are generated due to activity of crystalline defects such as vacancies, dislocations, and others. The intensity of the electron scattering and recombination processes, inflicted by defects (traps), controls the level of noise. Dependent on the dynamic operation condition of semiconductor devices, such as external biases and level of current injection, the traps will generate certain type and level of noise. Material growth or device processing technologies could introduce all kind of defects. Therefore, characterization of the semiconductor wafer in the early stages of processing (at least before packaging) could help to predict the level of noise due to the type and density of defects present on the wafer. Sorting out bad semiconductor chips could save money and effort in the radio frequency design of low-noise circuits. This current study focuses on 1/f noise modeling, which involves most powerful generators of noise and linear defects, named dislocations. The study also examines the possibility of assessing this noise by quantitative electron beam-induced conductivity (EBIC) measurements. These defects could be found in the bulk as well as at the epitaxial interfaces of a semiconductor device. The nanoscale size of these defects makes the scanning electron beam an instrument of choice for the proposed study. Conventional EBIC produces images of the defects, where contrast is proportional to the recombination rate at the site of a defect. Since contrast is measured as a fraction of one percent, the relative nature of contract value precludes quantitative measurements of the recombination rate, thus making quantitative assessment of 1/f noise impossible. In our model, using the Boltzman continuity equation, the recombination-generation processes per unit of length of a dislocation was defined for two operational conditions of EBIC, namely, for low and high intensity of an electron beam. The experimental technique of the quantitative measurement of carrier recombination (Mil'shtein 2001) consists of taking two EBIC scans along the selected defect at two different beam intensities, digitally subtracting the first scan from the second one and normalizing the result to the size of the electron range. The value of the recombination rate, extracted from the model, could then be used to predict the level of 1/f noise in the tested semiconductor sample.