| There are many different types of semiconductor memories. They can be divided into volatile and nonvolatile. Nonvolatile memories (NVM) are distinct from volatile memories in that the information stored is maintained after the power supply is removed. The basic principle of floating gate-based NVMs is shifting the threshold voltage of a metal-oxide-semiconductor field transistor (MOSFET) by charging or discharging a floating gate. Although these NVMs are more flexible and scalable than other nonvolatile memories, inherent “disturb” problems hinder the devices from expanding their market. The main cause for “disturb” originates from memory cell architecture designed for high packing density. Since total cell isolation from other cells is very impractical, it becomes important to solve and minimize the “disturb” issues for NVMs. Any unintentional change of threshold voltage can be considered as a “disturb” which causes erratic logic decisions. In terms of “disturb” origins, the current “disturb” and charge “disturb” are discussed in this dissertation. The pn junction leakage current brings about high charge injection into the floating gate during “writing” and “erasing” operations. Also under low electric fields, small charge compensation can be made by low gate leakage current through the floating gate. This current increases as the gate oxides are stressed after “writing” and “erasing” as well as “reading” operation. These stressed oxides tend to have traps generated by high electric fields, which shift the threshold voltage without compensating charges in the floating gate.; As the devices are scaled down, “disturb” problems are aggravated due to the increased leakage current, high electric fields and fabrication issues. In this dissertation, this author also discusses NVM fabrication issues, which can induce stress and gate oxide degradations. |
