Quantitative failure analysis for MEMS materials with multiple active flaw populations

Tungsten-coated and uncoated polycrystalline silicon microscale specimens of various dimensions were used to investigate the applicability the use of Weibull statistics to predict component failure when multiple flaw populations are present. The addition of W coating was employed to generate the second flaw population because of W precipitation in polysilicon. For uncoated polysilicon specimens the tensile strength scaled with the specimen length as the major detrimental flaws were located at the specimen sidewalls. For coated specimens the scaling of specimen strength was not as evident. The strength of W-coated polysilicon was considerably lower than that of uncoated polysilicon with a greater scatter in values. A three-parameter Weibull model provided accurate predictions for the probability of failure for W-coated specimens with strength data sorted according to the specimen length. However, the scaling of strength with specimen dimensions was mixed: strength data were found to scale with the surface area for probabilities of failure greater than 60% and with the specimen length for probabilities of failure less than 40%.