Following the Fourth Industrial Revolution, the rapid activation of major megatrends such as 5G and artificial intelligence (AI) has significantly heightened the importance of memory devices, which are responsible for data storage in electronic system...
Following the Fourth Industrial Revolution, the rapid activation of major megatrends such as 5G and artificial intelligence (AI) has significantly heightened the importance of memory devices, which are responsible for data storage in electronic systems. As data generation and consumption continue to grow explosively, memory has evolved from a mere auxiliary component to a key element that determines the overall performance and efficiency of electronic systems. Accordingly, advances in memory technology have become essential for the continued growth of the semiconductor industry. In particular, reducing the bit cost (cost per bit) remains a critical issue. To address this, 3D NAND flash memory has been developed with increased layer stacks and optimized layout designs, such as placing peripheral circuits beneath the NAND cell array, maximizing storage density per unit area and lowering production costs.
Due to the three dimensional structure and the process requirements to realize it, polycrystalline silicon (poly-Si) has been adopted as the channel material. Poly Si consists of numerous grains and the grain boundaries (GBs) between them. GBs contain a high density of traps, which can be filled or emptied depending on the word line (WL) voltage. When traps are filled, potential barriers are formed within the GBs, which adversely affect electron mobility. Additionally, GBs have negative impacts on key device characteristics such as subthreshold slope and threshold voltage variability. As the number of layers in 3D NAND flash memory increases, the total string length also increases, resulting in a proportional rise in channel resistance. This, combined with the inherently low mobility of poly Si, leads to a significant decrease in current. As a result, securing a sufficiently large read current to ensure reliable operation has become increasingly difficult. To overcome this challenge, it is crucial to develop accurate models capable of predicting and analyzing the various electrical characteristics observed at poly Si GBs that strongly influence current behavior.
In this doctoral thesis, a model was developed to describe the potential barrier height at the GBs within the poly Si channel of 3D NAND flash memory. Starting from the energy band diagram of poly Si incorporating GBs, surface potential models for both the grains and the GBs were derived. Importantly, the derivation considered the cylindrical geometry of the macaroni-shaped cell structure specific to 3D NAND flash memory. From these surface potential models, a potential barrier model was derived and validated against results obtained through TCAD simulations. The model was further used to investigate the behavior of the potential barrier under varying WL voltages, temperatures, and GB trap distributions. In all cases, the model showed good agreement with TCAD results, confirming its accuracy. Since the barrier height has an exponential relationship with electron mobility in the channel, modeling this characteristic was a priority.
Subsequently, based on the derived potential barrier model, an effective mobility (μeff) model for the poly-Si channel was also developed. As previously mentioned, the barrier height affects mobility exponentially. Electrons traversing the channel must overcome this barrier, a process governed by the thermionic emission mechanism. In this process, electrons absorb phonon energy to overcome the barrier, which is closely related to the Meyer-Neldel Rule (MNR). Based on these mechanisms and the previously derived barrier model, an effective mobility model was formulated and validated through comparison with TCAD simulations. Various operating conditions—such as changes in WL voltage, temperature, and GB trap distributions—were explored to assess the behavior of μeff, and the model showed strong consistency with simulation results. As a result, this model enables accurate prediction and analysis of mobility in poly Si channels and provides insights into the associated current characteristics.
The physically based models presented in this doctoral thesis go beyond simple empirical fitting and offer deep physical insight into carrier transport in poly Si channels. These models are expected to play a critical role in evaluating the performance of future 3D NAND flash memory devices and guiding optimal device specifications. Furthermore, the modeling approaches introduced here can offer valuable insights for a wide range of poly Si-based device structures. Therefore, the proposed models and the compact model implemented in the SPICE simulator represent essential technological assets that can be effectively utilized in predicting device behavior, optimizing early-stage design, and ultimately contributing to shaping the future direction of the semiconductor industry.