Back to search

National Technical University of Athens (NTUA)

Νανοηλεκτρονικές διατάξεις μνήμης

Abstract

dc:description

The ever-increasing market for memories due to the widespread of portable electronics, (smartphones, tablets, etc.), has led to an enormous expansion of non-volatile memories, namely memories which maintain the stored information even without voltage supply. The main leader is represented by Flash technology, which has dominated the market in the last decades and has now reached the 16 nm scaling node. However, Flash memory is reaching its scaling limits due to issues related with big leakage currents in thin film oxides, which render the write/erase states indistinguishable. In addition, there is a urgent need to explore new memory concepts capable of filling the gap which actually exists between the high-performance, low area density and expensive SRAM and DRAM, and the low cost, high density, but low-performance HDD.For this reason, in the last years there has been a huge effort in both industrial and academic research groups to overcome this scaling issue by realizing new and revolutionary non-volatile memories with novel materials, new physical concepts and increased scaling capability. The main purpose is to obtain a novel device with the high-performance and endurance of SRAM and DRAM, combined with the high density integration and robustness of conventional HDD. This new concept has been defined as Storage Class Memory. Although it seems not easy to find a new technology able to fit all the proposed requirements, some novel devices are promising for future non-volatile applications.In addition to the need of novel non-volatile memory concepts in order to satisfy the insatiable requirements of memory technology for enhanced capacity, new devices are also object of active research for novel logic and neuromorphic computation. Some operations like real-world image learning, pattern recognition and decision are extremely expensive for conventional Boolean CMOS processors, while for human brain, they represent quite easy processes. In this framework, the new devices could help in developing complex, high density and low power neuromorphic networks able to efficiently perform artificial learning and recognition tasks.One of the most promising technologies for both non-volatile memory and neuromorphic computation is represented by metal-oxide resistive switching memory, or RRAM. These devices have become one of the top competitors for new generation memory due to its extreme ease of fabrication in the back-end of a CMOS process, fast switching, relatively high endurance and low power operation. All these properties make RRAM extremely interesting for future Storage Class Memory. In addition, its intrinsic variability of the switching characteristics, which represents the major concern for industrial memory production, could become the winning feature for neuromorphic network synapses.This doctoral dissertation focuses on the study of resistive switching effect, on both theoretical and experimental level. For this reason, a variety of materials were used in order to investigate the influence of the complex fabrication parameters on the electrical performance of the memory cells, focusing mainly on titanium oxide (TiOx), hafnium oxide (HfOx) and tantalum oxide (TaOx). In the literature, exceptional properties have been presented for the above materials, whereas the last two have been successfully integrated in the CMOS process line. Thus, it was considered of outmost importance to study their response as non-volatile memories, focusing on the whole spectrum of their electrical characteristics. In addition, by employing theoretical calculations based on the Density Functional Theory (DFT), the basic conditions, which must apply on atomic scale in order for the resistive switching effect to take place, were examined.According to the theory of finite elements, the solution of three partial differential equations was achieved in 3D geometry, gaining significant insights in terms of local density of oxygen vacancies, temperature and electric potential profiles. The influence of the diffusion barrier and ion hopping distance on the current-voltage characteristics was also studied, elucidating on the origins of the sharp/gradual resistance change. At the same time, in order to understand the variability of the electrical characteristics of RRAM devices, a Monte-Carlo based algorithm was developed and a systematic study on the influence of various physical parameters (such as recombination energy, generation energy, migration energy) on the cumulative distribution functions of RRAM cells was carried out.The incorporation of metallic nanoparticles within TiOx dielectric matrix was proved a simple and promising technique, towards the enhancement of the statistical dispersion of RRAM devices, memory window and external operating voltage. The existence of a large memory window is of great importance for multilevel applications, induced either with external pulses or with compliance current limit control. The main reason for the improved electrical performance is the local increase of the electric field due to the presence of nanoparticles and the confinement of the potential percolation paths where resistive switching effect will take place.The integration of two, three or multiple layers as active material within the RRAM memory cell was also examined, in order to reduce the operating currents. The latter is achieved by depositing dielectric layers under low oxygen content conditions, acting thus as series resistances, resulting in bigger voltage drop in the layer with the biggest oxygen content. Consequently, the resistive switching effect is limited between a small region of the active layer, improving thus the statistical variability. Measurements towards neuromorphic applications were also carried out, in order to assess the potential applicability of the memory cells for artificial learning and pattern recognition. In addition, the simple structure of the RRAM cells, permits their integration into crossbar architectures, with aim to enhance the area density of the chip.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2017

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Bousoulas, Panagiotis
  • Μπούσουλας, Παναγιώτης

Subjects

dc:subject × 14

Rights

dc:rights
Statement dc:rights
  • BY
Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/42493

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Bousoulas, Panagiotis; Μπούσουλας, Παναγιώτης. Νανοηλεκτρονικές διατάξεις μνήμης. National Technical University of Athens (NTUA), 2017. http://hdl.handle.net/10442/hedi/42493