Analysis of Reliability Issues and Lifetime Estimation in NAND Flash Memory

Abstract

ABSTRACT

Analysis of Reliability Issues and Lifetime Estimation in NAND Flash Memory

Kyunghwan Lee School of Electrical Engineering and Computer Science College of Engineering Seoul National University

As NAND flash memory continues to be aggressively scaled down, it becomes more susceptible to reliability problems. As a result, the lifetime estimation of the device is now serious topic for mass production. However, the apparent activation energy (Eaa) in the conventional temperature-accelerated lifetime test method of NAND flash memory does not follow the Arrhenius model, since various failure mechanisms occur concurrently. Therefore, this conventional Arrhenius model has a huge error in the lifetime prediction. Generally well-known dominant failure mechanisms in NAND flash memory are detrapping, interface trap (Nit) recovery, and trap-assisted tunneling (TAT). In this thesis, we propose an advanced charge loss model and completely separate three dominant failure mechanisms in terms of the time-constant (τ) and the final ΔVth in various generations (A, B, C) of NAND flash test element group (TEG) cells and sub 20-nm multi-level cell (MLC) NAND flash memory main-chip. As a result, it is observed that each τ of the mechanisms follows the Arrhenius law well, which means that each has its own activation energy (Ea). In addition, we deeply investigate the retention characteristics of the dominant mechanisms in various conditions, such as the number of P/E cycling times, probability level of the Vth cumulative distribution, and states in sub 20-nm MLC NAND flash memory. We also extracte the contribution rate (CR) of each failure mechanisms at criterion of

ΔVth_Total

according to baking temperature. The results give the physical reason for abnormal retention behaviors such as Eaa roll-off at the PV3 and negative Eaa at the ERS. P/E cycling stress generates traps in the tunnel oxide layer. We extract the trap profile in the tunnel oxide in space and energy distributions using 3D TCAD simulation. For the first time, we reveal the origin of abnormal Eaa characteristics and derive a mathematical formula for Eaa as a function of each Ea(mechanism) in NAND flash memory. We propose two different accurate lifetime estimation models for sub 20-nm NAND flash memory. The first model is the Eaa integration method. Using the analytically modeled Eaa equation, the lifetime of NAND flash memory is accurately predicted. The second model is the advanced extrapolation method using the trends of extracted parameters. Using the proposed model, accurate lifetime is estimated in all states (PV3, PV2, PV1, and ERS). Since the proposed model takes into account the retention characteristics for various mechanisms, this model provides much accurate prediction on the lifetime of NAND flash memory. Also, the lifetime estimation for the next generation of NAND flash memory is analyzed using 3D TCAD simulation. As a result, the lifetime for the next generation is expected to decrease as much as 66 % of the lifetime for the current generation.

Keywords: MLC NAND flash memory, failure mechanisms, Arrhenius model, activation energy (Ea), P/E cycling stress, lifetime estimation

Student Number: 2011-20892