Abstract
Small‐signal conductance peaks in metal‐oxide‐semiconductor (MOS) devices have been found experimentally to be much wider than predicted by the surface‐state continuum model of Lehovec. This increased width has been attributed by Nicollian and Goetzberger to surface potential fluctuations generated by inhomogeneities in the interface charge density. However, no theoretical estimate of the amount of additional width to be expected from such inhomogeneities nor of the dependence of this additional width upon device parameters has been presented. In this paper these theoretical estimates are provided for the case when the surface potential departs slightly from its value as estimated in the usual one‐dimensional charge sheet model of the MOS device. To do this we derive and approximately evaluate a theoretical relation between the variance of the surface potential and the autocovariance function of the interface charge distribution. For a random interface charge distribution we obtain a logarithmic dependence of the variance upon the sum of the oxide, space‐charge, and surface‐state capacitances. In the limiting case in which only long‐wavelength charge fluctuations contribute to the variance of the surface potential, this logarithmic dependence is replaced by the dependence appropriate to a quasiuniform model. The results indicate that the magnitude and bias dependence of the effect of a random distribution is in fair agreement with experiment, but that more data are needed at various oxide thicknesses on samples of low or uniform surface‐state density with known flat‐band shifts before it can be concluded that surface potential fluctuations are indeed responsible for the anomalous width of the conductance peaks. Should this prove to be true, the dependence of the conductance‐peak width upon device parameters would serve to identify both the presence and the nature of interface charge inhomogeneities.