Silicon Nanocrystal Charging Dynamics and Memory Device Applications

Citation

Feng, Tao (2006) Silicon Nanocrystal Charging Dynamics and Memory Device Applications. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/N8JK-ZQ70. https://resolver.caltech.edu/CaltechETD:etd-06052006-141803

Abstract

The application of Si nanocrystals as floating gate in the metal oxide semiconductor field-effect transistor (MOSFET) based memory, which brings many advantages due to separated charge storage, attracted much attention in recent years. In this work, Si nanocrystal memory with nanocrystals synthesized by ion implantation was characterized to provide a better understanding of the relationship between structure and performance -- especially charge retention characteristics.

In the structural characterization it was demonstrated that scanning tunneling microscopy (STM) and non-contact atomic force microscopy (nc-AFM) enable much more accurate measurements of the ensemble size distribution and array density for small Si nanocrystals in SiO₂, estimated to be around 2-3 nm and 4 x 10¹² -3 x 10¹³ cm⁻², respectively. The reflection high energy electron diffraction (RHEED) pattern further verified the existence of nanocrystals in SiO₂. Capacitance-voltage (C-V) measurements demonstrated the memory effects. The comparison between charge density and nanocrystal density suggests single charge storage on individual Si nanocrystals.

The electronic property of tunnel oxide layer is a key factor influencing charge retention, and was characterized by conductive atomic force microscopy (C-AFM). An overall high conductance observed between the nanocrystal floating gate and the substrate is believed to be responsible for the relatively short retention time for electrons. A narrowed denuded zone contaminated with nanocrystals is suggested to be the reason for the high conductance, which is further supported by switching events and fluctuations in local current-voltage (I-V) curves. From the results of C-AFM, a better control of nanocrystal distribution close to the channel is shown to be critical for non-volatile nanocrystal memory made via Si ion implantation.

Nanoscale charge retention characteristics of both electrons and holes were probed directly by ultrahigh vacuum (UHV) nc-AFM, in which a highly doped Si tip was applied to inject charges into the nanocrystal layer and monitor subsequent charge dissipation. The results reveal a much longer hole retention time (e.g., >1 day) than that for electrons (e.g., <1 hour), which is consistent with the charge retention characteristics from electrical characterization of nanocrystal floating gate MOS capacitors as well as time-resolved photoluminescence measurements. The large difference in charge retention times for electrons and holes is attributed to the difference in tunneling barrier heights: 3.1 eV and 4.7 eV for electrons and holes, respectively. Based on the charge injection and retention characteristics obtained from UHV nc-AFM and nanocrystal floating gate MOS devices, we suggest that hole programming in Si nanocrystal memory is an interesting choice in improving data retention or in further device scaling.

UHV nc-AFM guarantees high detection sensitivity and stability in charge imaging experiments due to a lack of air damping, so a three-dimensional (3D) electrostatic model can be developed to provide quantitative information regarding the distribution and evolution of the localized charges. For example, a transition from initial complementary error function distribution to Gaussian distribution was suggested in the simulation. In addition, charge detection sensitivity was found to increase with the scanning height, showing much room for further improvement of the sensitivity in UHV nc-AFM. The limitation of the electrostatic model is also discussed, and some knowledge regarding the charge distribution obtained from theoretical analysis and other experimental methods is suggested to be necessary supplements to the quantitative charge analysis by nc-AFM.

Finally, the approach used in the electrostatic simulation of nc-AFM was applied in 3D simulation of Si nanocrystal memory. The dependence of Coulomb charging energy on dielectric environment is analyzed. From the local variation of channel minority carrier density due to separated charge storage, the threshold number density of charged nanocrytals for 1D approximation to break down is shown to be 10¹² cm⁻² in the sample geometry investigated.

Thesis Files