| MOSFETs employing programmable floating gates have drawn much attention in the semiconductor memory market in the past decade and are found in many applications today such as computers, MP3 players, digital cameras, mobile phones or PDAs. Regardless of their enormous success, there has been little investigation of use of these nonvolatile devices outside of persistent storage memories. One reason for this is because they have been engineered to achieve long data retention time (>10 years) while operating voltage, switching speed and endurance are compromised in favor of retention time. Lower operating voltages and faster switching can be achieved by using band-gap engineered gate stacks, multiple metal floating gates, thinner oxides, and tunneling as the main programming mechanism. In principle, this also enables reduced stresses which results in increased endurance. Such devices can then be used as logic circuit elements in interesting applications with the goal to minimize power consumption, maximize performance, and provide high functionality.;This dissertation suggests that nanocrystal floating gate devices may be very useful for static applications such as a FPGA. Since FPGAs are typically reconfigured only a few times during their lifetime, fast switching programmable elements such as the SRAMs are not needed and only consume a large fraction of board area and power. They can be replaced by these nonvolatile devices to enable an area and power efficient FPGA board where a time-consuming configuration on every boot up is no longer required.;Furthermore, this dissertation presents the invention of a new unified memory device with merged dynamic and nonvolatile memory operations. It may also operate in both modes simultaneously as a 2-bit device. It is based on a dual floating gate structure that can potentially be built with high write cycle endurance in the dynamic memory domain by using direct tunneling for programming and/or metal nanocrystals in the floating gates. It is scalable to at least the 16 nm node and has potential for 3D stacking for ultra-high density. Applications of such a memory include instant-on computing, power proportional computing, and resilient computing. It could also be very useful in network on chips, synaptic processors or other applications that could benefit from a unified memory with multi-bit capability. |
