Con mun cadons On the Statistics Governing the Defect Density Determination in Si02 Films

the statistical method first employed by Fritzsche (2) and then by Chou and Eldridge (3) to determine the defect density in the SiO2 films on Si in their assess-ment of the oxide quality in terms of dielectric strengths. It was contended that the defects in SiO2 films ought to be regarded as indistinguishable, and that Bosse-Einstein rather than Boltzmann statistics should have been used in deriving the functional re-lationship among the defect density p [or ~ in Ref. (1)], the electrode area, F, and the fraction, P, of the MOS capacitors exhibiting primary breakdown when tested for dielectric strengths. A parallel was drawn between the fraction of pri-mary breakdown and the yield of integrated circuits (IC). The work of Price (4) was cited to show that different expressions for the IC yield could be ob-tained depending on the choice of statistics. For dis-tinguishable defects, for which Boltzmann statistics are appropriate, the IC yield, P, [or Y1 in Ref. (4)], is given by P = exp (--Fp) or i n P p = [1] F where F now denotes the active area of the IC devices. In the case of indistinguishable d fects, where Bose-Einstein statistics are applicable, the IC yield, P [or Y2 in (4)], is expressed b