Reliability assessment of a 0.35 μm silicided polysilicon fuse memory cell

Polysilicon fuses are utilised as one time programmable memory cells in submicron semiconductor circuits. Robust design and qualification strategies for the implementation of polysilicon fuse memory cells are necessary to ensure the long term reliability and stability of these memory cells embedded in high precision mixed signal semiconductor products. Analog Devices Inc, a world leader in the design and fabrication of mixed signal semiconductor products utilises a variety of one time programmable memory technologies across its product ranges. These memory cells are used for storing calibration trim factors in precision mixed signal circuits, therefore the long term reliability of the memory is critical to the continued reliable operation of the device in which they are incorporated.

In this work existing implementations of polysilicon fuse memory cells on a 0.35 μm BiCMOS semiconductor process are assessed for their long term reliability. The post programmed microstructure of the polysilicon fuse memory cells are analysed using transmission electron microscope techniques. From this analysis a theory for the physics of the programming process is derived. Implementations of the same polysilicon fuse memory cell fabricated on a different yet parametrically similar 0.35 μm BiCMOS processes are also analysed. Optimal programming conditions are determined by experimental means and the long term reliability is assessed using qualification testing strategies derived from a combination of a risk analysis of the potential fuse failure mechanisms and industry standards derived qualification strategies.

It is shown that the optimisation of the programming conditions contribute to the long term intrinsic reliability of the fuse memory cell. A complete understanding of the physics of the fuse programming process is necessary in order to properly optimise the programming conditions. The empirical data gathered through experimentation is validated using a mathematical model of the assumed physics of the programming mechanism. The mathematical model was generated in conjunction with the author and the Mathematics Applications Consortium for Science and Industry (MACSI) at the University of Limerick. The fuse programming mechanism in this thesis more accurately fits the observed post programmed fuse characteristics than that which is presented in the literature.

The long term reliability of the fuse structure is validated with 3 years of accumulated reliability testing data.

This work has contributed to the Analog Devices corporate quality and reliability qualification policy for devices containing polysilicon fuse memory cells.