Fabrication and chracterization of single and multilayer tunnel dielectrics for advanced floating gate flash memory
The floating gate device has been the workhorse for the non-volatile memory technology since the beginning of flash memory era. However, as the device is scaled down towards the realms of nanometer dimension, floating gate flash faces a very steep scaling path. The tunnel oxide scaling has a practic...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Subjects: | |
| Online Access: | http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/44368/1/P.1-24.pdf http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/44368/2/Full%20Text.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The floating gate device has been the workhorse for the non-volatile memory technology since the beginning of flash memory era. However, as the device is scaled down towards the realms of nanometer dimension, floating gate flash faces a very steep scaling path. The tunnel oxide scaling has a practical limit of approximately 8 nm due to data retention requirement. Therefore, the purpose of this work is to characterize and to assess the performances of single and multi-layer tunnel oxide, which primary focus is to further scale it beyond 8 nm. This study was carried out in two steps. Firstly, device I-V characteristics were simulated using the MATLAB software, based on the most recent compact physical model. Programming speed and data retention were calculated based on the simulated I-V curves. Secondly, MOS capacitors were then fabricated and characterized to validate the simulation result. The performance of single layer tunnel oxide has been successfully demonstrated. Its performance has been mainly evaluated from two perspectives, namely the programming time τprog, and data retention τret. The τprog for 4 nm single layer oxide and oxynitride were calculated to be 110 μs and 130 μs respectively, not too far off from 100 μs technological requirement. However, their τret performance was well below 10-year requirement, with both dielectrics just been able to achieve 3.1 and 4.6 year respectively. In that sense, one can conclude that both 4 nm single layer oxide and oxynitride have failed to comply with the requirement of 18 nm technology node. However, it has been proved that nitrided oxide could improve the τret of single layer SiO2. Furthermore, it has also been demonstrated that the thickness of a single layer oxide and oxynitride of 8.25 and 6.4 nm respectively, would be required to achieve the 10-year data retention requirement. It has also been shown that nitrided oxide could serve as an effective way of suppressing trap generation which in turn would suppress low field device leakages, especially in the form of SILC. In the case of multi-layer dielectrics, it has been shown that the best configuration is the one with the thinnest bottom SiO2 / thickest Si3N4. Device simulation shows that for 2 and 3-layer dielectrics, the τprog was in the range of 18 to 41 μs for the EOT of 4 to 8 nm, while experimentally it’s in the range of 2 to 104 μs. Taking τret requirement into consideration however reveals that only configurations with the EOT of 6 nm for both 2 and 3-layer dielectrics and 8 nm of 3-layer dielectric have successfully met the requirement for 18 nm technology nodes. |
|---|