Scaling Considerations of Inter-Poly Oxide-Nitride-Oxide Dielectric for High-Density Dram Applications

ZJ Ma,JY Choi,CD Lien
DOI: https://doi.org/10.1117/12.250866
1996-01-01
Abstract:ABSTRACT Highly reliable thin oxide-nitride-oxide (ONO) stacked dielectrics with oxide-equivalent thickness To;eq h1 the range of 34—68A were fabricated using conventional furnaces and an RTA machine. Rapid Thermal Nitridation(RTN) in NH3 ambient of smooth poly-Si prior to Si3N4 deposition is found to be critical for improving dielectric integrity and low-field leakage current of the ONO stacked dielectrics. This RTN process also significantly improves oxygen diffusion resistance of ultra-thin Si3N4 film. As a result, 22A LPCVD Si3N4 is adequate to sustainwet oxidation at 800°C without any oxygen diffusion through it, which results in a manufaturable and reliableToxeq45A ONO stacked dielectric. By using rugged poly-Si as a bottom electrode plus RTN process, ONO stackeddielectric can be scaled down to about To,ceq34A without any leakage or reliability problems.Keywords: Oxide/Nitride (ON), Oxide/Nitride/Oxide (ONO), Rapid Thermal Nitridation (RTN), LPCVD, DRAM. Th[TRODUCTION The decreased DRAM cell area requires a highly reliable ultra-thin dielectric to ensure a sufficient amount ofstored charge. For high-density DRAM applications, high storage capacitance, low leakage current, high breakdownvoltage, long TDDB lifetime and low infant mortality are key issues to be considered for the storage dielectric.ONO stacked dielectrics have been extensively studied for more than a decade and still remain the most practicalstorage dielectric6. As the ONO stacked dielectric is further scaled down, systematic study on how to manufacturea highly reliable and controllable ultra-thin ONO stacked dielectric continues to be necessary. In this study, 34—68AONO stacked dielectrics were fabricated by simply using the conventional LPCVD Si3N4 furnace, atmosphericoxide furnace as well as an RTA machine. Detailed characterization of the dielectrics includes I-V, C-V, TDDBmeasurements, etc.
What problem does this paper attempt to address?