Ugo Russo

Self-Accelerated Thermal Dissolution Model for Reset Programming in Unipolar Resistive-Switching Memory (RRAM) Devices

This paper addresses the numerical modeling of reset programming in NiO-based resistive-switching memory. In our model, we simulate electrical conduction and heating in the conductive filament (CF), which controls the resistance of the low resistive (or set) state, accounting for CF thermal-activated dissolution. Employing CF electrical and thermal parameters, which were previously characterized on our NiO-based samples, our calculations are shown to match experimental reset and retention characteristics. Simulations show that reset transition is self-accelerated as a consequence of a positive feedback between the thermal dissolution of the CF and local Joule heating in the CF bottleneck, which can account for the abrupt resistance transition in experimental data. Finally, the model is used to investigate the reduction of the reset current, which is needed for device application.

Study of Multilevel Programming in Programmable Metallization Cell (PMC) Memory

Programmable metallization cell (PMC) memory, also known as conductive bridging RAM (CBRAM), is a resistive-switching memory based on non-volatile formation and dissolution of a conductive filament (CF) in a solid electrolyte. Although ease of fabrication, promising performance and multilevel (ML) capability make the PMC a possible candidate for post-flash non-volatile memories, further physical understanding is required to better assess its true potential. In this work, we investigate the kinetics involved in the programming operation (i.e., transition from the high resistance to the low resistance state), which occurs by voltage-driven ion migration and electrochemical deposition, and results in CF formation and growth. The main kinetic parameters controlling the programming operation are extracted from our electrical data. Also, CF growth and corresponding resistance decrease is shown to be controllable with reasonable accuracy in pulse mode by employing a variable load resistance which can dynamically control the programming kinetics. A semi-analytical physical model is shown to account for experimental data and allows for the engineering of fast and reliable ML programming in one transistor-one resistor (1T-1R) devices.

Filament Conduction and Reset Mechanism in NiO-Based Resistive-Switching Memory (RRAM) Devices

The physical understanding of the programming and reliability mechanisms in resistive-switching memory devices requires a detailed characterization of the electrical and thermal conduction properties in the low-resistance state of the memory cell. The aim of this paper is the characterization of the conductive filament (CF), which controls the localized current flow in the low resistive state of the cell. Based on a new technique for evaluating the CF temperature during operation, we perform a statistical characterization of the critical filament temperature for the reset operation, i.e., the transition to the high-resistance state by the thermal dissolution of the CF. The thermal resistance of the CF and the activation energy for the dissolution mechanism are then evaluated, allowing for a physics-based numerical modeling of the reset operation based on CF thermal breakup.

Conductive-filament switching analysis and self-accelerated thermal dissolution model for reset in NiO-based RRAM

This work presents detailed characterization and modeling of the reset operation in resistive-switching memories based on metal oxides. Our experimental results confirm previous observations that reset is controlled by Joule heating, providing an insight on the electrical and thermal parameters of the conductive filament (CF) in the low resistance state. The characterization of such parameters allows to model the CF rupture responsible for reset switching. Our model explains the switching by self-accelerated dissolution of the CF, and can quantitatively account for reset and data-retention experiments. The scaling of programming current is finally investigated by means of reduction of CF cross-section.

Modeling of Programming and Read Performance in Phase-Change Memories—Part I: Cell Optimization and Scaling

One of the major concerns for the feasibility of phase-change memories is the reduction of the programming current. To this aim, several efforts have been dedicated both on cell geometry and on material engineering. This paper addresses programming-current minimization by the optimization of the cell geometry and materials, programming-current scaling, and the tradeoff between programming and readout performances of the cell. A general procedure to find the optimum-cell geometry is proposed and applied to a prototype vertical cell. Then, the evolution of program and read performances through technology nodes is analyzed by numerical simulations with the aid of an analytical model, for both the isotropic- and nonisotropic-scaling approaches. The two scaling approaches are discussed and compared in terms of program and read cell performances. Finally, material optimization is considered for further program-read improvement.

Intrinsic Data Retention in Nanoscaled Phase-Change Memories—Part I: Monte Carlo Model for Crystallization and Percolation

The amorphous phase of chalcogenide material in phase-change memories (PCMs) is subjected to spontaneous and thermal-activated crystallization. This represents a critical reliability issue and has to be carefully investigated and modeled for physically based projection of retention failure up to ten years. A new three-dimensional percolation model describing the statistical crystallization behavior in an intrinsic PCM cell for the amorphous state is developed. With this physical model, the authors were able to calculate the resistance evolution with time in the cell and the statistical distribution of retention failure times in a cell array. From the impact of geometrical parameters on the cell retention performance, PCM design guidelines to minimize data-loss effects can be obtained. The model allows the evaluation of nucleation and growth parameters and statistical extrapolations of intrinsic retention failure, which is shown in part 2

Intrinsic Data Retention in Nanoscaled Phase-Change Memories—Part II: Statistical Analysis and Prediction of Failure Time

The statistical spread of intrinsic data retention times in phase-change memory (PCM) cells is studied. Based on the crystallization and percolation model described in part 1, the crystalline grain size in the amorphous volume after data loss is extracted. From the temperature dependence of grain size, the authors calculate the statistical shape factor for the distribution of failure times, allowing a statistical prediction of data retention in PCM large arrays. The scaling and optimization issues with respect to failure time statistical spread are finally addressed

Voltage-Driven On–Off Transition and Tradeoff With Program and Erase Current in Programmable Metallization Cell (PMC) Memory

The transition from the on (low-resistance) to the off (high-resistance) state is studied for programmable metallization cell nonvolatile memories. The stability of the on state under stress voltage and the erase operation were characterized as a function of the initial resistance in the timescale from 100 mus to 100 s. The data suggest that the on-off transition is limited by voltage-driven ion hopping and that filaments with larger size, and hence lower resistance, are more stable. Finally, results are discussed with the aid of an analytical model for erase, which is also used to address the tradeoff between on-state stability and program/erase currents.

Analytical Modeling of Chalcogenide Crystallization for PCM Data-Retention Extrapolation

Phase-change memory (PCM) reliability is affected by the crystallization of the amorphous chalcogenide material. To reduce measurement time, crystallization of the active material, usually a chalcogenide alloy, is generally studied at high temperature (T > 160degC), while data retention is to be predicted for lower temperatures (T > 120degC). Therefore, a physically based procedure to extrapolate crystallization dynamics from high to operation temperatures is required. This paper shows a simple analytical model for predicting the maximum PCM-operation temperature compatible with a ten-year retention lifetime of the device. Experiments are first analyzed to extract the average retention lifetime and average size of crystalline particles at crystallization in the PCM cell; this allowed the extraction of nucleation rate and growth velocity in the Ge2Sb2Te5 phase-change material. The classical theory for crystallization based on nucleation and growth (N/G) is then used to extrapolate lifetime data to relatively low temperatures for reliability assessment. Our study shows that the temperature dependence of retention lifetime may not obey to the Arrhenius law, as a result of non-Arrhenius nucleation. The dependence of reliability on N/G parameters is finally discussed with reference to different crystallization modes in phase-change materials.

Modeling of Programming and Read Performance in Phase-Change Memories—Part II: Program Disturb and Mixed-Scaling Approach

The scaling analysis of phase-change memory (PCM) cells is an essential step toward validation as a competitive technology in terms of array density and current consumption. While the current scaling has been addressed in a companion paper, we focus here on the thermal crosstalk, namely, the temperature increase in 1 bit in the array while an adjacent cell is being programmed by a high-current reset pulse. This parasitic heating may lead to partial crystallization in the amorphous phase and to a consequent resistance decrease after cycling. Our analysis shows that the thermal crosstalk strongly depends on the scaling approach used, e.g., isotropic or nonisotropic scaling. A new mixed-scaling option for PCM cells is proposed, which provides the maximum decrease of programming current compatible with the reliability requirements deriving from the thermal crosstalk. The effects of this new scaling approach on the programmed volume and data retention are finally addressed.