State dependence and temporal evolution of resistance in projected phase change memory

Benedikt Kersting, Vladimir Ovuka, Vara Prasad Jonnalagadda, Marilyne Sousa, Valeria Bragaglia, Syed Ghazi Sarwat, Manuel Le Gallo, Martin Salinga, Abu SebastianScientific 2020 · DOI

TL;DR

This paper presents a new device model for projected phase-change memory (PCM) that accurately captures resistance drift and state dependence. This model, validated experimentally, reveals the critical role of interface resistance and provides crucial guidelines for engineering better PCM devices for in-memory computing and neuromorphic applications.

What Problem Does This Paper Address?

Phase-change memory (PCM) is a promising technology for in-memory computing and neuromorphic systems due to its ability to store a continuum of resistance values, enabling analog operations like matrix-vector multiplications. However, a major hurdle is the inherent variability in PCM device resistance. This variability stems from factors like structural relaxation (resistance drift), 1/f noise, and temperature fluctuations. While the "projected PCM" concept has shown promise in mitigating these issues by diverting read current through a separate "projection layer," a deep understanding of its behavior, particularly the interplay between resistance drift and state dependence, has been lacking. This paper aims to fill this gap by developing and validating a comprehensive device model.

Methods

The authors developed a device model for a projected PCM device with a lateral line-cell geometry. This model represents the device as an equivalent electrical circuit composed of resistors corresponding to the crystalline and amorphous phases of the PCM material, as well as the projection layer. A key innovation is the explicit inclusion of an "interface resistance" between the PCM and projection layers, which is hypothesized to significantly influence device behavior.

The model was developed by considering two extreme scenarios: zero and infinite interface resistance. To validate the model, the researchers fabricated and characterized projected and unprojected PCM line-cells based on antimony. They performed electrical characterization at 200K to minimize recrystallization effects. The model's predictions for resistance drift and state dependence were then compared against experimental data. The authors also used the model to explore material selection guidelines by analyzing the impact of various material properties on device performance.

Datasets

Not discussed in the paper. The study relies on experimental data obtained from fabricated PCM devices.

Key Insights and Ideas

  • Interface Resistance is King: The study unequivocally demonstrates that the interface resistance between the phase-change material and the projection layer is a critical parameter. It dictates how effectively the projection layer bypasses the amorphous region, significantly impacting both resistance drift and state dependence.
  • Projected PCM Alters Drift Behavior: The temporal evolution of resistance (drift) in projected PCM devices deviates from the standard power law observed in unprojected devices. This deviation is attributed to the constant current through the projection layer, leading to a more complex drift suppression mechanism.
  • State Dependence Becomes Non-Linear: In projected PCM, the device resistance no longer scales linearly with the amorphous fraction. This non-linear behavior, influenced by interface resistance, presents challenges for precise state programming, especially at low resistance states.
  • Trade-offs in Device Engineering: While a high interface resistance can improve drift suppression, it can also lead to a non-linear resistance-amorphous fraction relationship. This highlights a trade-off that needs to be carefully managed during device design.
  • Material Selection Guidelines: The developed model provides a framework for selecting materials. It suggests that materials with a high amorphous-to-crystalline resistance ratio (e.g., doped-SbTe3, Ge2Te2Sb5) are most suitable for projected PCM devices, but careful optimization of projection material and interface resistance is still necessary.

Key Results

  • The developed device model accurately captures the resistance drift and state dependence observed in experimental projected PCM devices.
  • Experimental validation on antimony-based projected PCM line-cells confirmed the model's predictions, showing a clear curvature in the logarithmic resistance evolution and a separation between resistance states that decreases with time, indicative of non-zero interface resistance.
  • The effective drift coefficient in projected PCM devices can be suppressed by a factor of up to 10 compared to unprojected devices, depending on the amorphous volume.
  • The study identified material combinations and interface resistance ranges that can achieve a drift coefficient below 0.01 and maintain a near-linear scaling of resistance with amorphous length, crucial for reliable in-memory computing.

Limitations and Future Work

  • Ohmic Regime Assumption: The model's validity is primarily for the ohmic regime of device operation, where read voltages are kept low (e.g., <0.15V for antimony). The field dependence of amorphous phase transport is neglected.
  • Localized Interface Resistor: The model uses a localized interface resistor, which is an approximation. While supported by FEM simulations, the real interface resistance might be distributed.
  • Future Work: The authors suggest extending the model to include the field dependence of transport and exploring the impact of distributed interface resistance. Further research into optimizing material selection and device engineering based on the provided guidelines is also implied. The work lays the groundwork for designing large arrays of projected PCM devices for advanced computing applications.

Summary updated

Abstract

Phase change memory (PCM) is being actively explored for in-memory computing and neuromorphic systems. The ability of a PCM device to store a continuum of resistance values can be exploited to realize arithmetic operations such as matrix-vector multiplications or to realize the synaptic efficacy in neural networks. However, the resistance variations arising from structural relaxation, 1/f noise, and changes in ambient temperature pose a key challenge. The recently proposed projected PCM concept helps to mitigate these resistance variations by decoupling the physical mechanism of resistance storage from the information-retrieval process. Even though the device concept has been proven successfully, a comprehensive understanding of the device behavior is still lacking. Here, we develop a device model that captures two key attributes, namely, resistance drift and the state dependence of resistance. The former refers to the temporal evolution of resistance, while the latter refers to the dependence of the device resistance on the phase configuration of the phase change material. The study provides significant insights into the role of interfacial resistance in these devices. The model is experimentally validated on projected PCM devices based on antimony and a metal nitride fabricated in a lateral device geometry and is also used to provide guidelines for material selection and device engineering.
BibTeX
@inproceedings{kersting2020state,
  author = {Benedikt Kersting and Vladimir Ovuka and Vara Prasad Jonnalagadda and Marilyne Sousa and Valeria Bragaglia and Syed Ghazi Sarwat and Manuel Le Gallo and Martin Salinga and Abu Sebastian},
  title = {{State dependence and temporal evolution of resistance in projected phase change memory}},
  booktitle = {Scientific},
  year = {2020},
  doi = {10.1038/s41598-020-64826-3},
}
Figure